Homooligomeric and Heterooligomeric Associations between K+-Cl– Cotransporter Isoforms and between K+-Cl– and Na+-K+-Cl– Cotransporters*

Little is known regarding the quaternary structure of cation-Cl– cotransporters (CCCs) except that the Na+-dependent CCCs can exist as homooligomeric units. Given that each of the CCCs exhibits unique functional properties and that several of these carriers coexist in various cell types, it would be of interest to determine whether the four K+-Cl– cotransporter (KCC) isoforms and their splice variants can also assemble into such units and, more importantly, whether they can form heterooligomers by interacting with each other or with the secretory Na+-K+-Cl– cotransporter (NKCC1). In the present work, we have addressed these questions by conducting two groups of analyses: 1) yeast two-hybrid and pull-down assays in which CCC-derived protein segments were used as both bait and prey and 2) coimmunoprecipitation and functional studies of intact CCCs coexpressed in Xenopus laevis oocytes. Through a combination of such analyses, we have found that KCC2 and KCC4 could adopt various oligomeric states (in the form of KCC2-KCC2, KCC4-KCC4, KCC2-KCC4, and even KCC4-NKCC1 complexes), that their carboxyl termini were probably involved in carrier assembly, and that the KCC4-NKCC1 oligomers, more specifically, could deploy unique functional features. Through additional coimmunoprecipitation studies, we have also found that KCC1 and KCC3 had the potential of assembling into various types of CCC-CCC oligomers as well, although the interactions uncovered were not characterized as extensively, and the protein segments involved were not identified in yeast two-hybrid assays. Taken together, these findings could change our views on how CCCs operate or are regulated in animal cells by suggesting, in particular, that cation-Cl– cotransport achieves higher levels of functional diversity than foreseen.

In mammals, NKCC1 as well as KCC1, -3, and -4 have been shown to exhibit wide tissue distributions, whereas KCC2 is apparently confined to the nervous system (4 -11, 16 -21). They have also been shown to coexist in certain cell types, such as erythrocytes or lens cells, where a number of isoforms/variants (KCC1, KCC3, KCC4) have been identified (10,11). In certain tissues, localization studies have suggested a more differential distribution (9, 16 -21).
Although very homologous to each other, the KCCs display variant affinities for each of the transported ions and for the drug furosemide. In addition, their transport capacity and response to various stimuli are not the same under controlled conditions. In Xenopus laevis oocytes, for example, heterologously expressed KCC2 displays higher K m values for Cl Ϫ but lower K m values for Rb ϩ compared with KCC1 and KCC3 (16,(22)(23)(24)(25). Along the same line, KCC4 is less active than KCC2 at low levels of intracellular Cl Ϫ but more sensitive to phorbol ester-triggered events (26). Not surprisingly, differences between KCCs and NKCCs are even more pronounced (4,16,27).
Several lines of evidence suggest that the NKCCs exist as homooligomers in cells. They are as follows. 1) The size of NKCC1 and NKCC2 has been found to increase by a ϳ2-fold factor when membranes expressing either protein were treated with cross-linking agents (12,28). 2) Through GST pull-down assays and yeast two-hybrid mapping analyses, the cytosolic carboxyl terminus (Ct) of the NKCCs was found to harbor two domains that are endowed with self-interacting properties, * This work was supported by grants from the Kidney Foundation of Canada and from the Canadian Institute of Health and Research (CIHR) through MOP-68949 and MOP-15405. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Table S1 and Figs. S1 and S2. 1 A CIHR Scholar. 2  implying that these carriers could assemble into pairs of monomers (14,15). 3) In heterologous expression systems, NKCC2 has been shown to coimmunoprecipitate with NKCC2 and NCC with NCC (29,30). The structural unit of the Na ϩ -independent CCCs is currently unknown. Based on the oligomeric configuration of the NKCCs and a recent study by Blaesse et al. (31), 4 one might predict that each of the KCCs can also exist as homooligomeric units assembled through the distal portion of single monomers. At the same time, it is unclear that heterooligomeric associations between isoforms would be supported by the Ct given that the KCCs, although quite homologous to one another, have still diverged to some degree from a common ancestor.
In this work, various studies were conducted to define the high order structure of the KCCs. In particular, we found that several of these carriers can assemble not only into homooligomers but also into functionally relevant heterooligomers enclosing different CCCs. Such findings bear great implications as to how CCCs actually operate in native tissues or as to how their inactivation in various disorders leads to specific phenotypes. They are also crucial from the perspective of drug development for individual CCC members.

EXPERIMENTAL PROCEDURES
Supplies-Chemicals, reagents, or kits were from different companies and included monoclonal antibodies (Roche Applied Science and Amersham Biosciences), oligonucleotides (Sigma), enzymes/buffers used for cDNA constructions (New England Biolabs, Fermentas, and Invitrogen), the TNT rabbit reticulocyte lysate system (Promega), the GST SpinTrap Purification Module (Amersham Biosciences), and the Yeast Matchmaker Two-hybrid System (Clontech). All cDNAs were propagated in XL1 blue cells (Stratagene).
Constructs-All of the cDNA sequences and oligonucleotides used are listed as supplemental material in Table 1S. Several of these sequences (rabbit KCC1, rat KCC2, human KCC3A and KCC3B, mouse KCC4, and human NKCC1) were already cloned in the appropriate vector from previous work (13-15, 25, 26).
CCCs in pGilda or pB42AD-Several CCC-encoding cDNAs were used for the yeast studies. They were designed with the premise that their localization is cytosolic based on previous hydropathy analyses (4 -8, 14, 15, 25, 26, 36) and that the KCCs might also form oligomeric units by interacting through their Ct as shown for NKCC1 (14,15). Regions minimally required for the NKCC1-NKCC1 interaction to occur are shown in Fig.  1S (provided in this work as supplemental material), where they are aligned as scaled horizontal bars with the various CCCderived protein segments used.
Based on these alignments, corresponding regions in the KCCs were tested in yeast to determine whether they were also endowed with interacting properties. After we realized that they were not (see below), additional protein segments from the cytosolic Ct and amino terminus (Nt) were tested. In particular, the Ct-derived protein segments (some of which are also illustrated in Fig. 1S as scaled horizontal bars) were all longer than the deduced interacting regions except for one, and all belonged to a large domain that extends from transmembrane domain 12 to the end of the Ct. pGilda and pB42AD are the yeast vectors in which inserts were cloned; one has coding sequences for the LexA DNAbinding domain and a His transformation marker (TM), and the other has sequences for the activating domain of a transcription factor, the hemagglutinin (HA) tag, and a Trp TM . The CCC inserts were obtained from several constructs by PCR through appropriate primers to which restriction sites were added. Final constructs were generated by cloning inserts in empty or insert-enclosing vectors through one or several ligation step(s).
KCC4-(921-1057)/pGEX4T3-To generate a bait protein for GST pull-down assays, the insert of KCC4/Pol1 was cut with restriction enzymes (SacII at bp 1965 and XhoI at bp 3175), and one of the fragments (length 1206 bp) was transferred to the vector pGEX4T3 (between restriction sites SmaI and XhoI after blunting SacII with Klenow and dNTPs). The resulting construct was recut at two BsmI sites (to exclude a 2758-bp fragment encoding most of KCC4), and the vector-containing fragment, which encodes the distal Ct of KCC4 between residues 921 and 1057, was recircularized to generate KCC4-(921-1057)/pGEX4T3 (see bottom bar of Fig. 1S).
CCCs in pCDNA3-To generate prey proteins for GST pulldown assays and to express KCC4 in HEK-293 cells for protein studies, two additional constructs were used. One of these constructs, termed NKCC1-(759 -1212)/pCDNA3 and available from previous work (14), encloses a fragment that encodes residue stretch 759 -1212 in the carrier's Ct. The other construct, termed KCC4/pCDNA3, was produced by transferring the insert of KCC4/Pol1 (previously tagged NH 2 -terminally with c-myc; see Ref. 26) as a 3312-bp HindIII-SmaI fragment between the EcoRV sites of pCDNA3, blunting HindIII with Klenow and dNTPs before the ligation.
CCCs in Oocyte Vectors-Several constructs were used for protein analyses and functional studies in X. laevis oocytes. They were available from previous work (25,26) or engineered de novo for the current work. The inserts correspond to CCCencoding sequences (cloned in Pol1 or pGEM-HE) and to the human Na ϩ -glucose transporter type 1 (SGLT1)-encoding sequence (transferred for this work from PBS to Pol1 as an EcoRI fragment). The vectors, on the other hand, correspond to modified pGEMs from which stable cDNA-derived cRNAs can be produced in X. laevis oocytes. They contain a T7 bacterial promoter, a cloning site flanked by the X. laevis ␤-globin untranslated regions, a poly(A) tract, and an NheI linearizing site.
A c-Myc epitope tag (MEQKLISEEDL) or HA epitope tag (MYPYDVPDYA) was added in front of several coding sequences through the following sequential steps: 1) by cutting these sequences (in their vector) at a restriction site "x" just before the first ATG and at a nearby downstream restriction site "y"; 2) by ligating the resulting fragments to a pair of complementary oligonucleotides designed to encode the tag in frame and to possess x-and y-compatible single-stranded cDNA ends once hybridized. The 5Ј-3Ј restriction sites used for rabbit KCC1, rat KCC2, human KCC3A, and human SGLT1 were, respectively, XbaI-HincII, XhoI-NarI, XbaI-NsiI, and XmaI-EcoRI. Note that tagged mouse KCC4/Pol1 and tagged human NKCC1/Pol1 were already available from previous studies (26).
Yeast Two-hybrid Mapping Studies-Cotransformants were generated and analyzed as previously described (14,15). In brief, the following steps were pursued. 1) EGY48 yeast (Ys) were transformed with the p8op-lacZ Ura TM -enclosing vector (which can encode Leu TM and lacZ through built in LexA-dependent operators) and were seeded on ϪUra (ϪU) plates to generate Y OUL s. 2) Y OUL s were transformed with a CCC/pGilda and were seeded on ϪUraϪHis (ϪUH) plates to generate Y OULH s. 3) Y OULH s were transformed with a CCC/pB42AD to generate Y OULHT s, after which they were tested for their ability to grow on ϪUraϪHisϪLeuϪTrp (ϪUHLT) plates or express ␤-galactosidase activity on ϪUHT plus X-galactosidase (ϪUHT ϩ X) plates.
Two procedures were conducted to establish whether the behavior of Y OULHT s seeded on relevant plates could be ascribed to the occurrence of CCC-CCC interactions. 1) Y OULH s were tested for expression of the hybrid protein and for autonomous reporter gene activation on ϪUHL ϩ X plates. 2) Y OUL s were transformed with a CCC/pB42AD and tested for expression of a hybrid protein after selection of colonies on ϪUT plates as well as for autonomous gene activation on ϪUTL ϩ X plates.
Expression of CCCs in X. laevis Oocytes or HEK-293 Cells-Mature oocytes were injected with H 2 O (controls) or cDNA-derived cRNAs (ϳ15 ng for the KCCs and ϳ1 ng for NKCC1 unless mentioned otherwise) and maintained for 3 days at 18°C in Barth medium plus 125 M furosemide (15, 25, 26, 32, or 33). Native HEK-293 cells were subjected first to a pulse-chase experiment (34) in 100-mm dishes to label newly formed proteins through sequential incubations in 1) Met-free DMEM low (30 min), 2) Met-free DMEM low plus 100 Ci/ml [ 35 S]Met (120 min), and 3) phosphate-buffered saline wash solutions. Subsequently, they were incubated in 12 ml of DMEM low plus 2.5 mM Met, transfected through standard protocols (13,35,36) with KCC4 or pCDNA3 (25 g of cDNA plus 60 l of Lipofectamine TM 2000 and 3 ml of DMEM low), and maintained in the same (but cDNA-free) medium for 48 h.
Functional Studies-Pretreated oocytes incubated in 86 Rb ϩcontaining solutions were transferred to 96-well plates prefilled with 2% SDS and scintillation fluid. After a brief stabilization period, 86 Rb ϩ was detected with the TopCountNXT counter (Packard). For each condition tested, counts were averaged and converted into flux rates as explained in the legend to Fig. 8.
Immunolocalization Studies-Pretreated X. laevis oocytes expressing one or two different CCCs heterologously were examined by immunofluorescence as previously described (15,26,36). Anti-c-Myc or HA was used as primary antibody and Alexa Fluor-conjugated IgGs were used as secondary antibodies. Note that H 2 O-injected oocytes were used as negative controls.
Protein Analyses-Proteins were extracted from yeast, oocytes, or HEK-293 cells in lysis buffers supplemented with protease inhibitors (Table 1). In some cases, cell lysates (from nonpretreated oocytes) were subjected (at 4°C in the same lysis buffer) to 1-h incubations with anti-c-Myc or anti-HA (1:100) followed by another 30-min incubation with 2% protein-Sepharose A (Amersham Biosciences) and centrifugation-wash cycles. In one series of assays, cell lysates (from oocytes exposed to Sulfo-NHS) were incubated at 4°C with 4% agarose-coupled streptavidin before the centrifugation-wash cycles.
In all studies, pellet-bound or solubilized proteins from whole cell extract were incubated for 2 min at 100°C in protein sample buffer, migrated on polyacrylamide SDS-Tricine gels, and revealed by Coomassie Blue staining, direct autoradiography, or chemiluminescence (Amersham Biosciences). For the latter detection procedure, proteins were first transferred to Immobilon-P nylon blots (Millipore), where they were incubated sequentially with a primary and a secondary antibody. When appropriate, whole-cell extract proteins were quantified with the DC protein assay (Bio-Rad), and protein bands separated by gel electrophoresis were quantified numerically.
GST Pull-down Assays-The proteins GST-KCC4-(921-1057) (used as bait) and GST (used as control) were generated in BL21 E. coli through isopropyl 1-thio-␤-D-galactopyranoside 14 by incubating ϳ4 g of bead-immobilized proteins with 10 l of labeled prey for ϳ6 h at 4°C in a binding buffer (200 mM NaCl, 1 mM EDTA, 2.0% Triton X-100, and 20 mM Tris-HCl) followed by several washes in the same buffer supplemented with 800 mM NaCl.
Reverse Transcription-PCR Analyses-To determine which KCCs are expressed in HEK-293 cells, gene-specific primers (see Table 1S in the supplementary material) mixed with cDNA templates (prepared from HEK-293 cells or present in a Clontech human kidney library) were diluted in a PCR buffer (50 l total) and subjected to 35 rounds of thermal cycling (annealing temperatures 45-55°C). After amplification, one-fifth of the PCR products were migrated on ethidium bromide-stained agarose gels.
Sequence Analyses and Statistics-DNA characterizations were performed by restriction analyses and automated sequencing using vector-or CCC-specific primers. For DNA or protein sequence analyses, we used a combination of programs, including DNAStar (Lasergene) and online bioinformatic work tools. When pertinent, all data (band densities, flux rates, etc.) were averaged among replicate experiments and expressed as mean values Ϯ S.E. In some instances, differences between groups of variables were also analyzed by Student's two-tailed t tests, rejecting the null hypothesis for p Ͼ 0.05.

RESULTS
Yeast Two-hybrid Studies-Forty-eight protein segments were tested for their ability to support interactions in yeast. Various residue stretches belonging to the putative cytosolic Ct of various CCCs (NKCC1, KCC1, KCC2, KCC3, and KCC4) were included in 38 of these protein segments, 19 of which were fused to LexA and 19 to HA. Those used to illustrate the behavior of yeast are depicted through horizontal bars in Fig. 1S and are termed according the accompanying figure legend). The remaining protein segments (not illustrated in this work) included the putative cytosolic Nts of KCC1, KCC2, KCC3A, KCC3B, and KCC4.
As a prerequisite to the testing of bait-prey interactions in this system, Western analyses (conducted as described in the legend to Fig. 1S) confirmed that yeast transformed with any of the 48 CCC-encoding constructs and selected on appropriate plates were able to express hybrid proteins even in the absence of reporter gene activation (results not shown). As another prerequisite to the testing of interactions in this system, various analyses also confirmed that none of the LexA-or HA-coupled Ct-derived protein segments formed in yeast were able to autoinduce the reporter genes (results also not shown). Unfortunately, this was not the case for the KCC-derived Nt protein segments, which were all shown to activate the reporter genes autonomously and thus could not be used in further analyses.
The behavior of Y OULHT s (yeast transformed with the construct p8op-lacZ and different combinations of protein seg-ments as described under "Experimental Procedures") is shown through representative examples in Fig. 2S (provided as supplemental material along with Fig. 1S to illustrate the protein segments used). In each horizontal line, a unique transformant coexpressing a pGilda-and a pB42AD-derived protein segment (columns i and ii, respectively) was seeded on ϪUHTL plates and ϪUHT ϩ X plates (columns iii and iv, respectively), where the occurrence of an interaction is suggested by the ability of yeast to develop into colonies or to generate strong ␤-galactosidase activity.
When the coexpressed protein segments were from KCC1 or KCC3, we observed that the reporter genes Leu TM /lacZ failed to become activated whether the expressed residue stretch included a truncated Ct, a full-length Ct, or an extended Ct and whether they were from the same isoform or from both of these isoforms. Relevant examples are provided in Fig. 2S (1) along with Fig. 1S to illustrate the protein segments used. The other combinations tested (over 50) were carried out with additional protein segments that varied in length but included at least one putative oligomerization domain based on alignments with NKCC1. Among other possibilities, these findings suggest that KCC1 and KCC3 are devoid of self-interacting properties, that their assemblies into quaternary structures are not mediated by the Ct, unlike NKCC1 (14), or that factors required for the association of certain CCC domains are lacking in yeast.
When, on the other hand, the protein segments coexpressed in Y OULHT s were from KCC2 or KCC4, we found that they were able to induce Leu TM /lacZ as long as both corresponded to full-length Cts. Even if these findings do suggest that certain KCCs interact with each other through their Ct, it is noteworthy that the truncated Cts all failed to activate Leu TM /lacZ regardless of the partner coexpressed. As for KCC1 and KCC3, thus, the yeast environment could have prevented certain residues in KCC2 or KCC4 from forming associations that normally occur in mammalian cells.
The last group of Y OULHT s that were analyzed in these studies expressed the full-length Ct of a KCC along with the fulllength Ct of NKCC1. Results, which are illustrated in Fig. 2S (3), show that none of the Y OULHT s tested behave as if bait-prey interactions have occurred, in contrast to a positive control that consisted of Y OUL s expressing the proximal and distal Ct of NKCC1 (14). These results suggest once more that the localization or binding requirements of oligomerization domains vary among the carriers or that KCC1 and KCC3 do not behave as self-interacting structures.
Homooligomeric Assemblies between KCCs-To confirm that certain KCCs can form homooligomeric units in cells, as suggested in part by the yeast data, coimmunoprecipitation studies were carried out with oocyte lysates expressing each isoform as a c-Myc-and/or HA-tagged protein. Results are shown in Fig. 1, A-D, but only for KCC1 and KCC3A, since they were similar for all of the isoforms tested (including KCC3B). In each panel, lanes 2 served as controls to establish that the signals observed in lanes 1 were not due to the antibody used for immunopre-cipitation (Ab I ), and lanes 3 served as controls to test for the specificity of Ab I as well as contamination of the preparation by nonprotein-Sepharose A-bound antigens.
Lanes 1 show that anti-c-Myc-immunoprecipitated c-Myctagged isoforms can be detected with anti-c-Myc (A and C) and that anti-HA-immunoprecipitated HA-tagged isoforms can be detected with anti-HA (B and D). The same bands are seen in lanes 2 (A-D) even if Ab I s were omitted before the Western blot, indicating that the bands in lanes 1 do not correspond to IgGs. As for lanes 3 (A-D), in which irrelevant Ab I s were used, the absence of bands also indicates that the Ab I s used for lanes 1 were specific and that the preparations were not contaminated. Last, in lanes 4, the presence of appropriately sized bands when HA is used as Ab I and anti-c-Myc is used for detection (A and C) or when anti-c-Myc is used as Ab I and HA is used for detection (B and D) implies that the differentially tagged isoforms can be coimmunoprecipitated with each other. Taken together, these findings suggest, in contrast to the findings obtained in yeast, that oligomerization domains are present in all of the KCCs (as will be discussed below).
In the course of these studies, we have excluded the possibility of a direct c-Myc-HA interaction by conducting similar experiments in which c-Myc-tagged KCC1 or KCC3 was coexpressed with an irrelevant protein (i.e. a HA-tagged huSGLT1). In Fig. 2A, for example, no bands are seen when anti-HA-immunoprecipitated HA-SGLT1 is detected with anti-c-Myc (lanes 2 and 4) and when anti-c-Myc-immunoprecipitated FIGURE 1. Coimmunoprecipitation studies using X. laevis oocytes coexpressing differentially tagged but otherwise identical KCC isoforms. Oocytes were microinjected with cRNAs derived from epitope-tagged cDNAs and incubated 3 days in Barth medium before cell lysis. A-D, proteins were from total cell lysates (lanes 2) or corresponded to protein-Sepharose A-bound antigens (lanes 1, 3, and 4). Isoforms expressed were c-Myc-tagged and/or HA-tagged KCC1 (A and B) and c-Myc-tagged and/or HA-tagged KCC3A (C and D). Studies carried out with oocytes expressing c-Myc-tagged KCC2 and HA-tagged KCC2 or expressing c-Myc-tagged KCC4 and HA-tagged KCC4 revealed identical results (not shown). For each of the panels shown, lysates were obtained by solubilizing oocytes in a protease inhibitor-supplemented solution (Table 1), and immunoprecipitated antigens were obtained by incubating these lysates with a primary antibody (mouse anti-c-Myc or anti-HA) and protein-Sepharose A. The quantity of solubilized protein was ϳ100 g/lane, and the quantity of immunoprecipitated proteins was from ϳ1 mg of total lysates. Detection was carried out by chemiluminescence assays (exposure times of ϳ1 min) after incubating the blots sequentially with mouse anti-c-Myc (1:400) 1 in B or C) suggest that coimmunoprecipitation efficiency was unexpectedly high. At the same time, less pronounced differences in other experiments (Ͻ5-fold) imply that certain factors (quantity of samples loaded per lane, protein degradation, etc.) might have accounted for some of the findings. Importantly, however, the results of longer exposure times always failed to reveal signals for the negative controls. Note also that in most lanes, positive signals appear as broad doublet-like bands, as would be expected for a CCC family member (1,12,17,20,29,36). Note also that in D, the signal of lane 1 appears as a narrower band with altered electrophoretic mobility compared with lane 2, suggesting that a specific form of post-translationally modified carrier might have been selected for through immunoprecipitation. On the other hand, it should be noted that longer exposure times revealed a broad doublet-like band in this lane as well, indicating that the carrier immunoprecipitated had not been completely confined to intracellular domains. M, c-Myc; H, HA; IP, antibody used for immunoprecipitation. FIGURE 2. Coimmunoprecipitation studies using X. laevis oocytes expressing only one KCC isoform. Oocytes were microinjected with cRNAs derived from epitope-tagged cDNAs and incubated 3 days in Barth medium before cell lysis. In all lanes, proteins were from protein-Sepharose A-bound antigens. A, carriers coexpressed in each oocyte were differentially tagged, but one corresponded to a c-Myc-tagged KCC, and the other corresponded to an HA-tagged huSGLT1. B, carriers were extracted from a mixture of lysates derived from c-Myc-KCC1-expressing oocytes and HA-KCC1-expressing oocytes. Antibodies used for immunoprecipitation or detection were as indicated below each panel. The number of experiments from which representative blot images are shown, the results of longer exposure times, and the overall conditions used were otherwise as described for Fig. 1. Although differences between the intensities of certain bands on a given blot tended to vary among experiments, the positive and negative controls revealed that these bands, as well as the antibodies used, were specific. M, ϭ c-Myc; H, HA; IP, antibody used for immunoprecipitation. *, molecular weight mark at which HA-tagged huSGLT1 should migrate. Heterooligomeric Assemblies between Different KCC Isoforms-To determine whether Na ϩ -independent CCCs can interact with each other to form heterooligomers, coimmunoprecipitation studies were conducted once more as above except that each oocyte was microinjected with different combinations of KCC isoforms (one c-myc-tagged and the other HA-tagged). Results, which are shown in Fig. 3, A and B, reveal specific bands in all lanes regardless of which antibody was used as Ab I or for detection. Various controls, such as those used to test for KCC homooligomerization, also confirmed that the bands observed and antibodies used were specific.

Structural Organization of the CCCs
Heterooligomeric Assemblies between the KCCs and NKCC1-Here, an additional series of coimmunoprecipitation studies were carried out to verify whether the KCCs (using KCC1 and KCC4 as models) can interact with NKCC1. Accordingly, oocytes were microinjected with c-myc-NKCC1 alone, HA-KCC1 alone, c-myc-NKCC1 and HA-KCC1, or c-myc-NKCC1 and HA-KCC4. Results, which were identical for both KCCs, are presented in Fig. 4 for KCC1 only. Quite interestingly, it is seen through this example that KCC1 and NKCC1 can be coimmunoprecipitated with either anti-c-Myc-or anti-HA (lanes 4), that the observed bands are carrier-specific (lanes 1 versus 2), and that the Ab I s used were specific as well (lanes 1 versus 3).
Localization of CCCs in Oocytes-Immunofluorescence studies and Western analyses of cell surface-biotinylated CCCs were performed to determine whether the cellular localization of a CCC in amphibian cells was affected substantially by the coexpression of another CCC and, thus, to exclude the possibility that the interactions identified only took place in subcellular structures. Fig. 5 is used to illustrate the data, but only for KCC1 and NKCC1 in A (similar results were obtained for other combinations, including KCC4 and NKCC1) and only for KCC4 and NKCC1 in B.
The immunofluorescence studies, which are shown in Fig.  5A, reveal that the carriers tested can reach the cell surface or a compartment just beneath it, whether they are expressed alone (Fig. 5A, 3 and 6) or in combination (Fig. 5A, 7 and 8), and that the signals observed are of similar intensity among these subpanels. Regarding the Western analyses, which are shown in Fig. 5B with the data of three separate experiments, they reveal that the quantity of cell surface c-Myc-NKCC1 (corresponding to the upper bands of the middle subpanel) only decreases modestly in the presence of KCC4 (compare lanes 2 versus lanes 3 or lanes 4). In conjunction with the coimmunoprecipitation studies, which show high levels of oligomeric formation in oocytes (see Figs. 1, 3, and 4), the localization studies suggest that the cell surface of oocytes do express CCC-enclosing oligomers. . Coimmunoprecipitation studies using X. laevis oocytes coexpressing different KCC isoforms. Oocytes were microinjected with cRNAs derived from epitope-tagged cDNAs and incubated for 3 days in Barth medium before cell lysis. In all lanes, proteins were from protein-Sepharose A-bound antigens. Carriers coexpressed in each oocyte were differentially tagged, and they also corresponded to different isoforms. A, antibodies used were anti-HA for immunoprecipitation and c-Myc for detection. B, antibodies used were c-Myc for immunoprecipitation and anti-HA for detection. As in Fig. 1, differences in the electrophoretic mobility of a given KCC are seen between certain lanes (e.g. compare lanes 2 with lanes 1 in B), and some of the signals do not appear as broad doublet-like bands (see lanes 5 and 6 in A and 4 and 6 in B). Here, however, these discrepancies could be an indication that the post-translational processing of certain isoforms varies according to which other isoforms have been coexpressed or which other isoforms are included in an association. At the same time, longer exposure times also revealed much broader bands in all lanes, implying that at least a fraction of the associations uncovered probably included highly, if not fully, processed carriers. The number of experiments from which representative blot images are shown and the overall conditions used were otherwise as described for  . Coimmunoprecipitation studies using X. laevis oocytes coexpressing different CCC isoforms. Oocytes were microinjected with cRNAs derived from epitope-tagged cDNAs and were incubated for 3 days in Barth medium before cell lysis. Proteins were from total cell lysates (lanes 2) or protein-Sepharose A-bound antigens (lanes 1, 3, and 4). Isoforms expressed were c-Myc-tagged NKCC1 with or without HA-tagged KCC1 (A), or they were HA-tagged KCC1 with or without c-Myc-tagged NKCC1 (B). The number of experiments from which illustrative blot images are shown, the results of longer exposure times, and the overall conditions employed were also as described in the legend to Fig. 1. Although differences between the intensities of certain bands on a given blot tended to vary among experiments (as in Figs. 1 and 3), the positive and negative controls revealed that the signals observed, as well as the antibodies used, were all specific here as well. M, c-Myc; H, HA; IP, antibody used for immunoprecipitation.

Analyses of CCC-CCC Interactions Using Other Approaches-
Because the yeast studies could have led to false positive or negative results, as they can do at times (14,37,38), and because the interactions described thus far were identified primarily in the yeast or oocyte environment, we exploited other experimental systems to confirm our main findings and determine whether other interactions could have been missed. Results for these additional studies are described below and shown through representative experiments in Figs. 6 and 7.
One of the assays exploited was to pulse-label native CCCs in HEK-293 cells (KCC1, KCC4, and NKCC1 but not KCC2 and KCC3, according to Fig. 6A), to incubate the cells in a chase medium, to transfect them with c-myc-KCC4, and to carry out coimmunoprecipitation studies with anti-c-Myc and T 4 , a NKCC-specific antibody (39). Based on Fig. 6, we observe that c-Myc-immunoprecipitated c-Myc-KCC4 can be revealed with anti-c-Myc (Fig. 6B), confirming heterologous protein synthesis and that it can associate with endogenous, 125-175-kDa radioactive proteins (Fig. 6C) but that it cannot be coimmunoprecipitated with detectable quantities of NKCC1 (Fig. 6D). At the same time, however, Fig. 6 shows that T 4 -immunoprecipitated NKCC1 can be detected with anti-c-Myc (Fig. 6E), suggesting that the radioactive proteins in Fig. 6C probably included endogenous NKCC1 in addition to endogenous KCC4. It should be mentioned that T 4 has often worked poorly in our hands when used as a detecting antibody in coimmunoprecipitation studies (36).
The other assay exploited was to determine whether radioactive NKCC1-(759 -1212) 5 or KCC4-(1-1083) could interact with GST-KCC4-(921-1057) more than it could with GST. The results of these studies, which were identical among three experiments and are presented in Fig. 7 (through Coomassiestained and autoradiographed protein gels that were loaded with washed protein-coupled Sepharose mixtures), showed that the beads were highly enriched in GST-KCC4-(921-1057) or GST (Fig. 7A) but retained prey only when enriched in GST-KCC4-(921-1057) (Fig. 7B). Taken together, these results confirm the KCC4-KCC4 and KCC4-NKCC1 interactions and suggest that the yeast studies did lead to false negatives.
Functional Studies in Oocytes-To determine whether some of the oligomers identified could behave as functionally relevant structures, we performed pilot experiments in which CCCdependent 86 Rb ϩ transport rates were measured in oocytes expressing NKCC1, KCC4, or both of these transporters under conditions that were intended to stimulate transport rates for NKCC1 but not for KCC4. The results of these studies, which are summarized in Fig. 8, show that both NKCC1 and KCC4 behave as predicted under such conditions but, interestingly, that transport rates by oocytes expressing both NKCC1 and KCC4 increase 2-fold relative to those by 5 As explained in a previous study by our group (14), translation of NKCC1-(759 -1212) in rabbit reticulocyte lysates produces two major bands. One band migrates at ϳ50 kDa and probably corresponds to residue stretch ϳ759 -1212 initiated at a non-AUG (CUG, ACG, and GUG) triplet (as can be frequently encountered in this system) (13,41). The other band migrates at ϳ46 kDa and probably corresponds to residue stretch 794 -1212 initiated at the first AUG.  H), or HA-tagged-KCC4 and c-myc-tagged NKCC1 (not shown). After 3 days of incubation in Barth medium, they were pretreated in medium LH ϩ for 1 h (to activate NKCC1 but maintain KCC4 inactivated) and exposed to a fixating agent at the end of the procedure ( Table 1) 3 and 4); the difference between lanes 3 and 4 is that the quantity of cRNA used was ϳ15 and ϳ3 ng of cRNA, respectively. After 3 days of incubation in Barth medium, all oocytes were pretreated in medium LH ϩ for 1 h (to activate NKCC1 but maintain KCC4 inactivated) and lysed in a protease inhibitor-supplemented solution at the end of the procedure ( Table 1). The quantity of streptavidin-precipitated proteins was from ϳ1 mg of total lysates. Detection was carried out by chemiluminescence assays (exposure times of ϳ1 min) after incubating the blots sequentially with mouse anti-c-Myc (1:50) and a horseradish peroxidase-coupled anti-mouse IgG ( oocytes expressing NKCC1 only. Given that this increase was presumably not due to a change in NKCC1 expression (as shown in Fig. 5B), these results suggest that a relevant cooperative interaction has formed between NKCC1 and KCC4 in oocytes (see "Discussion").

DISCUSSION
In this work, coimmunoprecipitation studies have shown that several of the CCCs (NKCC1 and KCC1, KCC2, KCC3, and KCC4) have the potential to form homooligomeric units when heterologously expressed in X. laevis oocytes and that they have the potential to form heterooligomeric units as well by interacting with each other in this cell type; in particular, we demonstrated that certain of these heterooligomers could be composed of the secretory NKCC1 linked physically to a KCC. We have also found that the some of the associations identified in amphibian cells could occur in HEK-293 cells as suggested by an experiment in which endogenous NKCC1 or KCC4 was shown to coimmunoprecipitate with heterologously expressed KCC4.
Various studies aimed at determining whether CCC-enclosing oligomers occur in native tissues or untransfected cell lines were not conducted to complement this work, since antibodies are not available for most of the isoforms. Still, it is unlikely that such studies would have allowed resolution of this issue with certainty based on the following rationale: 1) tissue lysates from most sources contain several cell types that cannot be sorted out easily; 2) colocalization of two proteins in a cellular domain through imaging analyses is not a proof of interaction; and 3) antibodies often lack the sensitivity to detect a protein that is produced at low levels, especially if purified by immunoprecipitation beforehand (as found in this study, for instance). The latter rational also implies that immunolocalization studies in which variant tissue distributions were observed for certain CCCs (9, 16 -21) should be interpreted cautiously.
Yeast two-hybrid studies and GST pull-down assays carried out to determine which domains underlie the formation of CCC-enclosing oligomers in oocytes and HEK-293 cells sug-  (Table 1S), and templates used (per reactions) were from 0.2 g of total RNA (HEK-293 cells) or corresponded to ϳ2 g of cDNA library (human kidney). The image shown is from an ethidium bromide-stained agarose gel that was used to separate the various reverse transcription-PCR products. Bands are expected to migrate at 408 bp for KCC1, 707 bp for KCC3A, 844 bp for KCC3B, and 589 bp for KCC4. For KCC4, therefore, the band observed in the KID lane is unexplainably higher than expected. B-E, coimmunoprecipitation studies using HEK-293 cell lysates. For these experiments, cells were first pulse-labeled with [ 35 S]Met, after which they were incubated in a chase medium and transiently transfected with c-myc-KCC4/pCDNA3 or pCDNA3 as control. For each of the panels shown, lysates obtained by solubilizing cells in a protease inhibitor-supplemented solution (Table 1)   . GST pull-down assays. A, Coomassie Blue stain of an SDS-polyacrylamide Tricine gel illustrating the quantity of GST-coupled Sepharose beads loaded relative to that of GST-KCC4-(921-1057)-coupled Sepharose beads (n-fold difference between band intensities is ϳ1.5). B, autoradiogram of another Coomassie-stained gel illustrating the quantity of radiolabeled prey 5 (full-length KCC4 or protein segment NKCC1-(759 -1212)) retained by the GST-coupled Sepharose beads relative to that retained by the GST-KCC4-(921-1057)-coupled Sepharose beads (n-fold differences between band intensities are Ͻ50 for each of the prey tested). Both the gel image and autoradiogram shown were chosen to illustrate the results of three experiments.
gest that the cytosolic Ct probably plays an important role. As will be explained below, however, these studies did not lead to the identification of interacting domains for all of the CCCs (i.e. the Ct domains of KCC1 and KCC3 were found to be devoid of self-interacting properties based on the yeast analyses). Left alone, these results would imply that the Cts of various CCCs can support the formation of NKCC1-NKCC1, KCC2-KCC2, KCC4-KCC4, or KCC2-KCC4 oligomers but not that of other CCC-CCC combinations.
Despite the yeast data, other observations obtained from previous investigations and from this work as well suggest that several, if not all, of the CCCs do have the potential of behaving as self-interacting partners. These observations are as follows. 1) All of the CCCs studied thus far (NKCC1, NKCC2, NCC, and KCC2) appear to be organized as oligomers (12, 28 -31). 4 2) We have identified residue sequences that behave as self-interacting domains for many CCCs, including NKCC1, NKCC2, KCC2, and KCC4 (14,15). 3) The CCCs are all highly conserved through evolution, implying that important functional domains should be found at the same locations among the isoforms. 4) In pull-down assays, used as an alternative strategy to identify oli-gomerization domains in the CCCs, NKCC1-(759 -1212) was found to interact with KCC4-(921-1057), although it was unable to do so in yeast. 5) Our own coimmunoprecipitation studies yielded clear cut results.
One reason that might have accounted for the inability of certain Ct-derived proteins segments to interact in yeast is that the bait or prey tested were missing key residues. In this study, such a hypothesis was verified for KCC1 and KCC3 by using bait or prey that extended from transmembrane domain 12 to the end of the Ct, but interactions still failed to occur. Attempts were also made to determine whether the missing residues could belong to the proximal extremity of these carriers, but this possibility could not be tested given that the Nts of all KCCs were found to activate the yeast reporter genes. In this regard, however, a study by Moore-Hoon and Turner (12) showed that a mutant NKCC1 lacking all of its Nt was still able to assemble into homooligomers in a cell line. Accordingly, and based on the observations that have already been detailed above, it would be surprising if the Nt played an important role in carrier assembly.
An alternative reason that could explain the yeast data for KCC1 and KCC3 is that improper folding of certain protein segments in this expression system precluded associations that normally occur in vivo when carriers are intact. This conformational constraint, often incriminated as a cause of false negative results in yeast two-hybrid studies (14,37,38), would imply that the Ct of various CCCs might have to adopt isoform-specific structural configurations for oligomerization to occur and that the stringency of this requirement is variable among the carriers. It would perhaps also explain why the proximal and distal Ct of various KCCs failed to interact with each other in this system. Additional studies aimed at identifying oligomerization domains for all of the CCCs and at confirming that KCC1-CCCx or KCC3-CCCx oligomers do occur in various cell types will be necessary to confirm these hypotheses.
As stated earlier, various studies have shown that some cells do express more than one CCC subtype; human erythrocytes, for example, were shown to produce at least five CCC isoforms/ variants (10,11), and HEK-293 cells (through this work) were shown to produce at least three different CCCs. In view of these observations, one might predict that CCC coexpression occurs in other cell types as well, a possibility that is supported further by at least three different observations: 1) Some CCCs (NKCC1 and KCC1) are widely distributed (1,2,4,16,39); 2) gene redundancy resulting from the coexpression of functionally homologous proteins is a common feature among several cell types; and 3) inactivation of any given CCC in mammals is generally nonlethal even for those carriers that exhibit wide distributions (40,(42)(43)(44).
Beyond the probable occurrence of CCC coexpression in certain cell types as an indirect argument for in vivo heterooligomerization, three other line of evidence are consistent with this possibility: 1) in humans or mice, NKCC1, KCC2, or KCC3 inactivation leads to a number of similar traits, including deafness, abnormal blood pressure, and neuronal hyperexcitability (40,(42)(43)(44); 2) some of the CCC isoforms have been found to exhibit cell-specific functional behaviors (35); and 3) previous studies have suggested that a variety of cation channels can FIGURE 8. Functional studies of X. laevis oocytes expressing KCC4 and/or NKCC1. Oocytes were injected with cRNAs derived from various cDNAs, including c-myc-tagged NKCC1 (bar 1), c-myc-tagged NKCC1 and HA-tagged KCC4 (bars 2 and 3), or HA-tagged KCC4 (bar 4); for bars 2 and 4, the quantity of cRNA used was ϳ15 ng, and for bar 3, it was ϳ3 ng. After 3 days of incubation in Barth medium, all oocytes were pretreated in medium LH ϩ for 1 h (to activate NKCC1 but maintain KCC4 inactivated) and assayed for 86 Rb ϩ flux rate measurements (Table 1). In these studies, radioactive counts were transposed into flux rates as follows: flux rates ϭ counts accumulated in an oocyte/ incubation time in flux medium ϫ specific activity, assuming that membrane surface areas among stage V-VI oocytes are the same. All of the values shown are expressed as mean background-subtracted flux rates Ϯ S.E. of 9 -12 oocytes among 3-7 different experiments, and each of the signs used (* or §) indicates that the corresponding value is significantly different statistically (p Ͻ 0.01) relative to the other three values. Note that in the biotinylation studies of Fig. 5B, cell surface expression of heterologous c-Myc-tagged NKCC1 was not increased by the coexpression of HA-tagged KCC4. M, c-Myc; H, HA. assemble with one another in various cell types to form heterooligomeric complexes (45).
For the CCCs, the potentially important implication of forming such complexes in native cells stems from the differential characteristics that these carriers exhibit in regard to substrate specificities and kinetic behaviors. In contrast to NKCC1, for instance, the KCCs behave as Na ϩ -independent carriers that are modestly sensitive to bumetanide (4,16,25,26,32,33), and compared with KCC1, KCC2 exhibits higher affinity for Cl Ϫ (16,(22)(23)(24)(25). Heterooligomers that are composed of NKCC1 and a KCC or of different KCCs could thus display characteristics that differ substantially from those of NKCC1 or KCC homooligomers, leading to diversity in the physiological and pharmacologic properties of cation-Cl Ϫ cotransport among various tissues.
Another implication to the formation of these complexes is that the residues forming each of the ion-or inhibitor-binding sites in an oligomer could be contributed by different carriers. As such, and presuming that all of the residues included in a binding site operate as a coordinated ensemble, oligomers generated through KCC-KCC or NKCC-KCC associations would be expected to interact with K ϩ , Cl Ϫ , and furosemide regardless of the isoform present in the complex but not with Na ϩ . An alternative scenario would be that all of the CCCs can interact with both Na ϩ and K ϩ but that they exhibit a preference for K ϩ when the subunits at work only include KCC isoforms. This scenario would explain why the KCCs behave as Na ϩ -independent systems when expressed heterologously (4 -8, 16, 22-26) and why NKCC1 can mediate some levels of K ϩ -Cl Ϫ cotransport in the absence of extracellular Na ϩ (13,32,33,46).
In the current study, we have begun to address the interesting question of whether oligomeric assemblies between certain CCCs could lead to the formation of functionally unique structures by conducting ion transport studies in oocytes injected with NKCC1 with or without KCC4. To our surprise, we found that KCC4 coexpression led to a 2-fold increase in 86 Rb ϩ flux rates even if the assays were designed to stimulate transport activity by NKCC1 but prevent that by KCC4. In conjunction with the biotinylation studies of Fig. 5B, the observed increase in 86 Rb ϩ transport suggests that NKCC1 and KCC4 behaved in oocytes as cooperating subunits rather than independent carriers, leading to changes in 86 Rb ϩ transport by simple summation. They are thus consistent with the idea that certain types of CCC-enclosing heterooligomers play a relevant role in vivo.
Several studies have shown that the KCCs not only differ in kinetic behaviors but also differ in their responses to changes in cell volume, intracellular [Cl Ϫ ], and cell surface kinase/phosphatase activities (16,26). As mentioned previously, the behavior of NKCC1 in this regard differs from that of KCCs to a considerable extent (4,16,27). If, in such a context, heterooligomeric associations between various KCCs or between NKCC1 and KCCs did take place in certain cell types, the molecular mechanisms of CCC regulation would probably be more complex than foreseen initially. For example, would each subunit within an oligomer be regulated independently from one another? And if so, would the regulation of one unit affect that of other units or of the ensemble?
Another question raised by the in vivo occurrence of differ-ent types of CCC-CCC complexes is whether certain factors could play a role in determining the ratio of homo-to heterooligomer formation. Conceivably, such factors would vary from cell to cell based on the abundance of any given isoform and type of auxiliary subunits present. Differences in the affinities at which given CCCs combine with each other could also correspond to an important factor. Although very challenging in essence, future studies are warranted to determine how much native cells are able to exploit this type of potential structural diversity by controlling the formation rate or composition of various assemblies. Several questions that have been raised thus far concerning the functional relevance of certain associations between various CCCs will remain unanswered until additional experiments are carried out to complement this study. In particular, the ion dependences, inhibitor sensitivities, and saturabilities of CCC activity in cell types that express different combinations of isoforms and are subjected to relevant environmental stimuli will have to be determined. The data obtained through these measurements will also have to be interpreted based on the quantity of homooligomers versus heterooligomers that can form in untransfected cells or expression systems and based on modelbased kinetic predictions to determine whether coexpressed carriers promote ion movement through cooperative associations or by summation.
In conclusion, our finding that several CCCs can assemble with each other into both homooligomeric and heterooligomeric units represents a key finding in the investigative field of (Na ϩ )-K ϩ -Cl Ϫ cotransport. Indeed, this new information will have to be integrated to that obtained from other studies in which molecular mechanisms for CCC family members have been or are being explored. Our findings are also highly relevant from the perspective of drug development, given that the CCCs are probably underexploited targets for the treatment of diseases, such as sickle cell anemia, systemic hypertension, and epilepsy.