Cell surface expression of 5-hydroxytryptamine type 3 receptors is controlled by an endoplasmic reticulum retention signal.

Two subunits of the 5-hydroxytryptamine type 3 (5-HT3) have been identified (5-HT3A and 5-HT3B) that assemble into homomeric (5-HT3A) and heteromeric (5-HT3A+5-HT3B) complexes. Unassembled 5-HT3B subunits are efficiently retained within the cell. In this study, we address the mechanism controlling the release of 5-HT3B from the endoplasmic reticulum (ER). An analysis of chimeric 5-HT3A receptor(R).5-HT3BR constructs suggests the presence of elements downstream of the first transmembrane domain of 5-HT3B subunits that inhibit cell surface expression. To investigate this possibility, truncated 5-HT3B subunits were constructed and assessed for their ability to access the cell surface in COS-7 and ts201 cells. Using this approach, we have identified the presence of an ER retention signal located within the first cytoplasmic loop between transmembrane domains I and II of 5-HT3B. Transplantation of this signal (CRAR) into the homologous region of 5-HT3A results in the ER retention of this subunit until rescued by co-assembly with wild-type 5-HT3A. The mutation of this ER retention signal in 5-HT3B (5-HT3BSGER) does not lead to cell surface expression, suggesting the presence of other signals or mechanisms to control the surface expression of 5-HT3BRs. The generation of truncated 5-HT3BSGER constructs confirmed that the CRAR signal does play an important role in the ER retention of 5-HT3B.

The electrophysiological actions of serotonin (5-hydroxytrptamine; 5-HT) 1 include the mediation of fast excitatory responses in neurons of the central and peripheral nervous systems. These responses are induced by the activation of ligand-gated ion channels (5-HT 3 Rs). Peripheral receptors are thought to modulate pain and intestinal and cardiovascular functions (1). In the central nervous system, 5-HT 3 Rs are important targets for the control of emesis induced post-operatively by chemotherapy and radiotherapy (2) and in the palliative care of patients with multiple sclerosis (3).
The first 5-HT 3 R subunit cloned (5-HT 3A ) is able to generate functional homomeric receptors in heterologous expression sys-tems (4,5). However, the conductance of these recombinant receptors is too small to be resolved directly (sub-picosiemens), and the receptors do not resemble many native neuronal 5-HT 3 Rs (9 -17 pS) (6,7). The second 5-HT 3 R subunit to be cloned (5-HT 3B ) (5,8) is not able to generate functional homomeric receptors because of retention in the ER (9). However, upon the co-expression of 5-HT 3A and 5-HT 3B , 5-HT 3B is rescued and expressed on the cell surface as a heteromeric complex with a channel conductance similar to that of most native receptors (9). Electrophysiological studies suggest that both homomeric and heteromeric 5-HT 3 Rs may co-exist within the same neuron (6,8,10,11).
The 5-HT 3 Rs belong to the superfamily of transmitter-gated ion channels that includes the nicotinic acetylcholine, GABA A , and glycine receptors. The structural relationship of the 5-HT 3 Rs to the other members of this group suggests that their assembly may involve similar post-translational events (12). The assembly of GABA A receptors is directed by specific assembly signals within the amino-terminal extracellular domain (13)(14)(15)(16)(17)(18)(19)(20). The export of receptors from the ER represents a critical checkpoint for surface expression, with quality control within the lumen of the ER performed by the chaperone proteins BiP, calnexin, and protein disulfide isomerase (9,21,22). Interaction with such proteins can cause the intracellular retention of immature proteins by virtue of ER retention signals within the chaperone molecules. In addition, cytoplasmically exposed ER retention signals within the cargo proteins themselves have been identified as elements that control protein export from the ER (23)(24)(25)(26)(27). A unifying mechanism for the regulation of ER export by these cytoplasmically localized signals is not known. However, a role for COPI and COPII recruitment of proteins into ER transport vesicles has been postulated (Refs. 28 -31, but see Ref. 24).
To investigate the mechanisms involved in the ER retention of the 5-HT 3B R, we have examined the assembly and surface expression of 5-HT 3A ⅐5-HT 3B subunit chimeras and truncation mutants of 5-HT 3B . We demonstrate that 5-HT 3B possesses an ER retention signal capable of preventing homomeric cell surface expression. This "CRAR" signal requires masking by subunit interactions with the 5-HT 3A subunit, regardless of whether it is present natively in the 5-HT 3B subunit or recombinantly within the homologous position in 5-HT 3A . In addition, the export of 5-HT 3B from the ER appears to require the masking or exposure of other signals downstream from this ER retention signal.
Antibodies-Anti-Myc and anti-HA monoclonal antibodies (mAbs) were harvested from 9E10 or 12CA5 hybridoma cell lines, respectively, and used direct as supernatant (20 g/ml) or purified on immobilized protein A. For enzyme-linked immunosorbent assays with tsA201 cells, mAbHA-7 was obtained from Sigma. Rabbit anti-Myc antibody was purchased from Santa Cruz Biotechnology and used at 2 g/ml in co-immunofluorescence experiments. The secondary antibodies, goat anti-mouse Alexa Fluor 568, goat anti-mouse Alexa Fluor 488, and goat anti-rabbit Alexa Fluor 488 were purchased from Molecular Probes and used at 1 g/ml. The secondary antibody, sheep anti-mouse horseradish peroxidase, was purchased from Amersham Biosciences and used at 1/1000.
Immunofluorescence-Immunofluorescence was performed as described previously (21). Briefly, COS-7 cells cultured on coverslips were fixed in 3% paraformaldehyde (15 min in phosphate-buffered saline) and quenched twice in 50 mM NH 4 Cl (in phosphate-buffered saline). Subsequent washes and antibody dilutions were performed in phosphate-buffered saline containing 10% fetal bovine serum and 0.5% bovine serum albumin (block). When appropriate, cells were permeabilized with 0.5% Nonidet P-40 for 15 min, quenched, blocked, and processed as above. Coverslips were examined using a confocal microscope (Zeiss LSM510).
Enzyme-linked Assay of Cell Surface and Intracellular Expression Levels-Cell surface antibody binding was assayed on tsA201 cells grown on poly-L-lysine and collagen-coated glass coverslips. Coverslips were blocked (2% bovine serum albumin), incubated in primary antibody, and fixed as described previously (32,33). To measure total (cell surface and internal) antibody binding, fixed cells were exposed for 15 min to 0.1% Triton X-100. Coverslips were processed as above, but 0.1% Triton X-100 was included in all incubation buffers and in buffer used for the first two washing steps. Antibody solutions contained, additionally, 5% fetal calf serum. When surface and total antibody binding levels were compared, coverslips (with permeabilized or intact cells) were fixed before addition of primary antibody and assayed in parallel. Additionally, lysine (25 mM) was added to buffers to reduce nonspecific binding. In all cases, coverslips were incubated with horseradish peroxidase-conjugated goat anti-mouse IgG (Pierce) and washed five times before incubation with 750 l of 3,3Ј,5,5Ј-tetramethylbenzidine (Sigma) for 30 min. The supernatant was transferred to a cuvette, and absorbance was determined at 655 nm.
Immunoprecipitation-Cells were L-methionine-starved for 30 min before being labeled with [ 35 S]methionine (0.5 mCi per 10-cm dish, Translabel ICN/Flow) for 4 h. Cells were lysed in 10 mM sodium phosphate buffer containing 5 mM EDTA, 5 mM EGTA, 50 nM sodium fluoride, 50 mM sodium chloride, 1 mM sodium orthovanadate, 5 mM sodium pyrophosphate, 2% Triton X-100, 0.5% deoxycholate, 0.1 mM phenylmethylsulfonyl fluoride, 10 mg/ml leupeptin, 10 mg/ml antipain, 10 mg/ml pepstatin, and 0.1 mg/ml aprotinin (lysis buffer). Immunoprecipitations were performed as described previously (21) and analyzed by SDS-polyacrylamide gel electrophoresis, followed by autoradiography. Cross-linking experiments were performed by the treatment of cells with dimethyl pimelimidate, 7.6 mg/ml) followed by quenching in 100 mM Tris (pH 8) prior to cell lysis. 3 Rs-The subcellular distribution and assembly of 5-HT 3 Rs was examined by heterologous expression of 5-HT 3A and 5-HT 3B subunit cDNAs in COS-7 cells. To facilitate biochemical and morphological analyses, 5-HT 3 R subunits were tagged using the Myc or hemagglutinin (HA) epitopes. These epitope tags were added at the amino terminus of 5-HT 3 subunits between amino acids 5 and 6 of the mature polypeptides (downstream from the predicted signal sequence cleavage site) and appear to be silent (9).

5-HT 3 Receptor ER Retention
The subcellular distribution of the tagged 5-HT 3 Rs expressed in COS-7 cells was determined by immunofluorescence and confocal microscopy. As shown previously (9), 5-HT 3A -Myc when expressed alone and assessed for Myc immunofluorescence revealed surface labeling, as determined by secondary antibody Alexa Fluor 488 staining in the absence of detergent (Fig. 1A, left panel). The same cells were then permeabilized and processed for immunofluorescence a second time, using an Alexa Fluor 568-conjugated secondary antibody to detect intracellular subunits. It should be noted that surface receptors would also be labeled after permeabilization because of the presence of free antibody-binding sites present on the secondary antibody (capable of binding the second round of primary antibody). However, significant additional staining is observed after permeabilization, indicating the presence of intracellular 5-HT 3A subunits. The intracellular staining pattern is consistent with an ER distribution, suggesting the presence of unassembled receptor subunits.
In contrast, upon the expression of 5-HT 3B -Myc, no surface expression is observed (Fig. 1A, middle panel). However, substantial staining was evident after permeabilization, suggesting that the 5-HT 3B -Myc subunit cannot reach the cell surface when expressed alone but is retained intracellularly within the ER as reported previously (9). Upon co-expression of 5-HT 3A -HA, the 5-HT 3B -Myc subunit was rescued and expressed on the cell surface as a heteromeric complex (Fig. 1A, right panel).
To identify the region responsible for the ER retention of 5-HT 3B , two Myc-tagged chimeric receptor subunits, 5-HT 3A ⅐5-HT 3B and 5-HT 3B ⅐5-HT 3A , were analyzed (Fig. 1B). These constructs were expressed in COS-7 cells and analyzed by immunofluorescence. Despite being expressed at high levels within the ER, no surface staining was observed for the 5-HT 3A ⅐5-HT 3B chimera (Fig. 1B, left panels). In contrast, the 5-HT 3B ⅐5-HT 3A chimera exhibits cell surface staining (Fig. 1B, right  panels). Furthermore, upon [ 35 S]methionine labeling and cross-linking, high molecular weight oligomers were evident when separated by SDS-PAGE (Fig. 1C), suggesting the ability of the 5-HT 3B ⅐5-HT 3A construct to assemble into homomeric receptors. Thus, elements downstream from the first transmembrane domain are either inhibitory (in 5-HT 3B ) or promote (in 5-HT 3A ) cell surface delivery.
To probe for the existence of regions influencing cell surface expression within the 5-HT 3B -Myc subunit, a series of truncated 5-HT 3B subunits was constructed. Subunits were truncated after the transmembrane domain I, between transmembrane domain 1 and II (following residue 267), after transmembrane domain II, and between transmembrane domains III and IV (following residues 349, 369, 380, 393, 411, and 418) ( Fig. 2A). The rationale for this approach is that if inhibitory signals exist within 5-HT 3B and are responsible for ER retention, then once all such signals are removed the resulting protein will be released from the ER and expressed on the cell surface. To confirm the fidelity of the truncated constructs, COS-7-transfected cells were [ 35  In addition, 5-HT 3B -Myc has five potential N-linked glycosylation sites, each expected to contribute ϳ2-3 kDa to the molecular mass. The apparent molecular mass of the 5-HT 3B -Myctruncated subunits (Fig. 2B) is consistent with the presence of multiple sites for N-linked glycosylation and the presence of degradation products, as observed previously (9).
When expressed in COS-7 cells, truncations of 5-HT 3B at positions 418, 411, 393, 380, 369, 349, and TMII all had no effect on ER retention and did not result in surface expression (Fig. 2C, lower panels). In contrast, when 5-HT 3B was truncated to residue 267, found between TMI and TMII, surface expression was observed. Similarly, a further truncation of 5-HT 3B , such that the construct terminated at TMI, also resulted in cell surface expression. Quantification by a cell-based enzyme-linked immunosorbent assay confirmed the results observed by immunofluorescence. No significant cell surface detection of 5-HT 3B or when the construct was truncated to TMII was observed. In contrast, a significant enhancement of cell surface receptors (Ͼ14-fold, compared with TMII truncation) could be detected for the TMI truncation. These findings suggest the existence of an ER retention signal within 5-HT 3B within the first cytoplasmic domain, between TMI and TMII. Interestingly, this cytoplasmic domain possesses a potential ER retention motif "RAR" analogous to those found in other receptors (23,25,26,34).
Comparison of the homologous regions between 5-HT 3A and 5-HT 3B (Fig. 3A) revealed that these subunits differ in only four residues incorporating the RAR motif. To assess the capacity of this putative motif to function in ER retention we generated a mutant 5-HT 3A R containing CRAR in place of SGER, termed 5-HT 3A (CRAR). In keeping with a role in ER retention, this signal is capable of preventing the cell surface expression of 5-HT 3A (CRAR) (Fig. 3B). To eliminate the possibility that the

5-HT 3 Receptor ER Retention
mutant receptor subunit was unable to correctly fold and assemble, the 5-HT 3A (CRAR) was co-expressed with wild-type 5-HT 3A . Under these circumstances the 5-HT 3A (CRAR) was rescued from the ER in a similar manner to the wild-type 5-HT 3B subunit. Furthermore, the 5-HT 3A (CRAR) could not be rescued by wild-type 5-HT 3B (not shown). To assess the role of this ER retention signal within 5-HT 3B , a reciprocal mutation was constructed, yielding 5-HT 3B (SGER). The subcellular distribution of this construct was indistinguishable from wild-type 5-HT 3B , in that it was efficiently retained within the ER when expressed alone yet could be rescued upon the co-expression of 5-HT 3A , but not 5-HT 3B (Fig. 3B).
The failure of the 5-HT 3B (SGER) to access the cell surface might be explained by the existence of other ER retention signals. To address this issue, the 5-HT 3B (SGER) construct was used to generate truncations at residue 270 and TMII. The truncation of 5-HT 3B (SGER) at residue 270 (immediately following SGER) resulted in cell surface expression (Fig. 3C). Furthermore, and in direct contrast to the findings of the 5-HT 3B (Fig. 2C), when the 5-HT 3B (SGER) was truncated at TMII significant cell surface expression was observed (Fig. 3C). These findings confirm the ability of the RAR signal to function as an ER retention signal in both 5-HT 3A and 5-HT 3B .
In an attempt to understand how the identified ER retention signal might operate, we sought to determine whether a direct masking of the ER retention signal might occur between the two homologous domains in 5-HT 3A (SGER) and 5-HT 3B (CRAR) subunits. To address this possibility, 5-HT 3A (CRAR) and 5-HT 3B (SGER) were co-expressed and assessed for cell surface expression. Our findings indicate that these two subunits are unable to complement each other, with both being retained within the ER upon co-expression (Fig. 4A). Therefore, other regions are likely required for masking to occur. Furthermore, such regions cannot be supplied by the same subunit but must be provided by another neighboring subunit in trans.
To determine whether direct subunit interactions are required for the masking of the ER retention signal, a construct was generated expressing only the TMI-II region (inclusive of transmembrane domains) of 5-HT 3A and 5-HT 3B (termed 5-HT 3A TMI-II and 5-HT 3B TMI-II, respectively). To appropriately orient these subunit fragments within the ER membrane and provide an immunological handle on their expression, a signal sequence (from GABA A ␤3 subunit) followed by a Myc tag was positioned upstream (and in-frame). When co-expressed with the 5-HT 3B , neither construct was capable of masking the ER retention signal(s) and promoting cell surface delivery. To examine whether the 5-HT 3B TMI-II construct is capable of interacting with full-length 5-HT 3A -HA and/or 5-HT 3B -HA, these constructs were co-expressed in COS-7 cells, labeled with [ 35 S]methionine, cross-linked with dimethyl pimelimidate (to preserve molecular interactions), and immunoprecipitated via the 5-HT 3A TMI-II fragment (9E10) or the fulllength subunit (12CA5). Significant interactions (Fig. 4B) are observed when the immunoprecipitation is performed using the 12CA5 antibody directed against the full-length subunits. Much less dramatic, yet still evident, is an interaction detected via the 9E10 antibody against the subunit fragment, 5-HT 3B TMI-II. Although these results do not prove direct masking of the TMI-II region, it does show that subunit interactions occur within this region. Identical results were observed for 5-HT 3A TMI-II with respect to interactions with both 5-HT 3A and 5-HT 3B (Fig. 4B). In addition to identifying subunit-subunit interactions, this approach may identify other molecular interactions. In this respect, potentially novel interacting proteins with molecular weights of ϳ69 kDa, 78 kDa, 97 kDa, and 102 kDa are observed to co-purify when immunopurification was performed with either 9E10 or 12CA5 (Fig. 4B, asterisks) but not in mock-transfected cells. To emphasize nonspecific bands, the control lanes (mock-transfected) are enhanced digitally (Fig. 4B, right gel). The identity of the 78 kDa band is most likely BiP, because it has been identified previously to interact with 5-HT 3 and GABA A receptors (9, 21). The identity (and significance) of the other co-purifying proteins is unknown.

DISCUSSION
To date, two subunits of the human 5-HT 3 Rs have been cloned, 5-HT 3A (4, 5,) and 5-HT 3B (5,8). 5-HT 3A is capable of forming functional homomeric receptors exhibiting a very low single channel conductance (sub-picosiemens) (5,35,36). In contrast, recombinant 5-HT 3B does not form functional ion channels (5) because of a failure to exit the ER and reach the cell surface (9). Upon co-expression of the human 5-HT 3A and 5-HT 3B subunits, receptors with a large (16 pS) single channel conductance, low relative permeability to calcium ions, and a current-voltage relationship similar to that observed for many neuronal 5-HT 3 Rs are produced (5). From these observations, it is evident that 5-HT 3R expression is regulated by processes that control receptor assembly. Such regulation could occur at several stages, including subunit oligomerization and export from the ER. Consistent with the failure of the 5-HT 3B subunit to produce functional receptors (5,8), we have shown previously (9) that the 5-HT 3B subunit does not reach the cell surface but is efficiently retained within the ER as identified by the characteristic "lace-work" reticular staining pattern.
There are two possible explanations for the localization of 5-HT 3B in the ER (unless rescued by the 5-HT 3A subunit). First, the 5-HT 3B may not be capable of attaining the appropriate tertiary or quaternary structure unless co-assembled with 5-HT 3A . In this scenario, the 5-HT 3B would remain bound to ER chaperone proteins, BiP and calnexin (9,21), and be retained within this compartment by virtue of the ER retention signals residing on these chaperone molecules. By analogy with the GABA A receptors, the most likely failure to achieve a functional maturity is the lack of appropriate assembly signals within the large extracellular amino terminus (13)(14)(15)(16)(17)(18)(19)(20). A second possibility is the exposure of cytoplasmic ER retention signals that actively retain the subunit within the ER until masked by the assembly with another subunit, interactions with other proteins, or phosphorylation (23,24,26,28,29,34,37,38).
To discriminate between these possibilities, the subcellular distribution of 5-HT 3 R chimeras was investigated. Interestingly, when the extracellular region of 5-HT 3A was fused (beyond TMI) to the downstream (including both cytoplasmic) regions of 5-HT 3B (5-HT 3A ⅐5-HT 3B ), no cell surface expression was observed. This is despite the presence of all the predicted extracellular assembly signals likely to be required for the formation of 5-HT 3A homomeric receptors. Thus, other regions (provided by 5-HT 3B ) may be lacking that are also required for surface expression. Alternatively, inhibitory regions within the 5-HT 3B sequence may be present. Analysis of the 5-HT⅐5-HT 3A chimera that revealed significant cell surface expression supported the latter possibility. Thus the apparent lack of ligandbinding sites evident on 5-HT 3B (9) does not limit subunit folding and release from the ER. Indeed, this chimera is capable of assembling into high molecular weight oligomers, indicative of assembly. The results with these chimeras suggest that the failure of 5-HT 3B to exit the ER and reach the cell surface is not because of a lack of appropriate assembly signals or protein misfolding. We determined, therefore, to assess the potential for the existence of inhibitory sequences, such as ER retention signals, within 5-HT 3B .
The rationale for this approach was that if 5-HT 3B were progressively truncated, all ER retention signals would be removed. If successful, this approach would identify the most amino-terminal ER retention signals if multiple signals exist. Consistent with this hypothesis, progressively truncated 5-HT 3B constructs remained ER-localized until the truncation reached the region between TMI-II. This truncation mutant (5-HT 3B stop267) and a further truncation (at TMI) were efficiently expressed on the cell surface. Interestingly, a potential arginine-based (RXR) ER retention signal homologous to that identified in other proteins (23,25,26,34) is present within this region of 5-HT 3B , but not of 5-HT 3A .
In support of a role of this signal as an ER retention signal within the environment of a 5-HT 3 R, when CRAR is transferred to 5-HT 3A homomeric cell surface receptors are no longer observed. Like 5-HT 3B , the 5-HT 3A (CRAR) mutant can be rescued by wild-type 5-HT 3A , but not 5-HT 3B , and expressed on the cell surface, suggesting that protein misfolding is not responsible but that this signal may function to retain the 5-HT 3 R subunit within the ER until masked. To identify whether the CRAR signal functions similarly within its native 5-HT 3B subunit, we generated the reciprocal mutation, 5-HT 3B (SGER). This mutant was retained in the ER and could be rescued by the co-expression of 5-HT 3A , but not 5-HT 3B , as observed for the wild-type 5-HT 3B subunit, implicating the presence of other ER retention signals. To provide direct evidence that the CRAR signal functions within 5-HT 3B to retain the subunit within the ER, new truncation mutants using 5-HT 3B (SGER) as the template were generated. Termination of 5-HT 3B (SGER) following the SGER signal, or after TMII, resulted in cell surface expression of 5-HT 3B (SGER) (Fig. 3). This is in contrast to the identical TMII truncation generated on the wild-type 5-HT 3B (Fig. 2). Thus, the CRAR does function as an ER retention signal within 5-HT 3B , at least in the TMII-truncated version of 5-HT 3B . Furthermore, it demonstrates that the SGER homologous sequence in 5-HT 3A does not function as an ER retention signal that may have been masked by intramolecular or inter-molecular homomeric subunit interactions.
To determine whether the SGER (5-HT 3A ) and the CRAR (5-HT 3B ) might cooperate by direct binding to mask the presence of the ER retention signal in heteromeric receptors, the distribution of both mutant (5-HT 3A (SGER) and 5-HT 3B (CRAR)) subunits when co-expressed was examined. As observed when expressed alone (Fig. 3), no cell surface expression is evident, implying that a different mode of masking is operating. This is supported by the use of 5-HT 3 R chimeras (Fig. 1). Paradoxically, the 5-HT 3B ⅐5-HT 3A chimera can access the cell surface despite the presence of the CRAR signal. The most likely explanation for this discrepancy is that the major intracellular loop of 5-HT 3A is responsible for the masking of the CRAR signal. No evidence of a significant interaction with endogenous proteins was evident upon immunoprecipitation with the TMI-II constructs or wild-type 5-HT 3 Rs. However, this is not surprising because such interactions may be of low affinity or transient. Furthermore, although dilysine ER retention signals appear to interact with components of COPI, RXR motifs do not appear to do so (24), suggesting a different mechanism may be operating. Indeed, a COPI-independent route between the Golgi and the ER also exists (40). In support of a masking interaction involving the TMI-II region of 5-HT 3 Rs, constructs expressing this domain are capable of interacting with wild-type 5-HT 3A and 5-HT 3B subunits when co-expressed within COS-7 cells. That the mutation of this ER retention signal in 5-HT 3B to the homologous region of 5-HT 3A does not prevent ER retention implies the existence of multiple ER retention signals. Alternatively, 5-HT 3B may lack essential forward signals necessary for export from the ER (29,30,38,39,(41)(42)(43). We are currently investigating the presence of other ER transport signals within the 5-HT 3 Rs.