Cooperativity and Flexibility of Cystic Fibrosis Transmembrane Conductance Regulator Transmembrane Segments Participate in Membrane Localization of a Charged Residue*

Polytopic protein topology is established in the endoplasmic reticulum (ER) by sequence determinants encoded throughout the nascent polypeptide. Here we characterize 12 topogenic determinants in the cystic fibrosis transmembrane conductance regulator, and identify a novel mechanism by which a charged residue is positioned within the plane of the lipid bilayer. During cystic fibrosis transmembrane conductance regulator biogenesis, topology of the C-terminal transmembrane domain (TMs 7–12) is directed by alternating signal (TMs 7, 9, and 11) and stop transfer (TMs 8, 10, and 12) sequences. Unlike conventional stop transfer sequences, however, TM8 is unable to independently terminate translocation due to the presence of a single charged residue, Asp924, within the TM segment. Instead, TM8 stop transfer activity is specifically dependent on TM7, which functions both to initiate translocation and to compensate for the charged residue within TM8. Moreover, even in the presence of TM7, the N terminus of TM8 extends significantly into the ER lumen, suggesting a high degree of flexibility in establishing TM8 transmembrane boundaries. These studies demonstrate that signal sequences can markedly influence stop transfer behavior and indicate that ER translocation machinery simultaneously integrates information from multiple topogenic determinants as they are presented in rapid succession during polytopic protein biogenesis.

Polytopic protein topology is established in the endoplasmic reticulum (ER) by sequence determinants encoded throughout the nascent polypeptide. Here we characterize 12 topogenic determinants in the cystic fibrosis transmembrane conductance regulator, and identify a novel mechanism by which a charged residue is positioned within the plane of the lipid bilayer. During cystic fibrosis transmembrane conductance regulator biogenesis, topology of the C-terminal transmembrane domain (TMs 7-12) is directed by alternating signal (TMs 7, 9, and 11) and stop transfer (TMs 8, 10, and 12) sequences. Unlike conventional stop transfer sequences, however, TM8 is unable to independently terminate translocation due to the presence of a single charged residue, Asp 924 , within the TM segment. Instead, TM8 stop transfer activity is specifically dependent on TM7, which functions both to initiate translocation and to compensate for the charged residue within TM8. Moreover, even in the presence of TM7, the N terminus of TM8 extends significantly into the ER lumen, suggesting a high degree of flexibility in establishing TM8 transmembrane boundaries. These studies demonstrate that signal sequences can markedly influence stop transfer behavior and indicate that ER translocation machinery simultaneously integrates information from multiple topogenic determinants as they are presented in rapid succession during polytopic protein biogenesis.
The topology of most eukaryotic polytopic proteins is generated in the endoplasmic reticulum (ER) 1 through the collective action of sequence determinants encoded within the nascent polypeptide. These determinants encompass hydrophobic transmembrane (TM) segments that, together with their flanking residues, interact with cytosolic and ER translocation machinery to initiate and terminate translocation and integrate the polypeptide into the lipid bilayer (reviewed in Refs. [1][2][3]. In the simplest model, topology can be established cotranslationally by alternating topogenic determinants that function as signal (anchor) and stop transfer sequences (4 -7). As the first signal sequence emerges from the ribosome, it targets the ribosome nascent-chain complex (RNC) to the ER and gates open an aqueous channel in the membrane (the Sec61 translocon) (8). Because the ribosome exit site is directly aligned with the axial pore of the translocon, newly synthesized polypeptide is cotranslationally directed into the aqueous environment of the translocon as it emerges from the ribosome (9 -11). Subsequent synthesis of a stop transfer sequence gates the translocon closed to the ER lumen, terminates translocation, and provides the growing nascent polypeptide access to the cytosol (12,13). Through sequential iterations of these events, signal and stop transfer sequences can alternately direct the polypeptide into the ER lumen or the cytosol and thus establish topology of transmembrane segments and lumenal and cytosolic peptide loops.
Not all native polytopic proteins utilize a simple cotranslational biogenesis pathway. For example, translocation of peptide loops may be initiated by either upstream (N-terminal) or downstream (C-terminal) signal sequences (14,15). In the latter case, topology of the preceding TM segment can be established in a retrograde manner (i.e. C terminus to N terminus) as the polypeptide moves from the cytosol into the translocon and then into the membrane. Moreover, synthesis of downstream topogenic determinants can also rearrange the topology of preexisting TM segments (16 -18). Thus initial, or cotranslational, topology is dynamic and can be subjected to constraints of folding imparted by distant peptide regions. In these latter cases, very little is known regarding how TM segments interact with the ER translocation machinery and, in particular, whether the presence of one segment may influence translocon interactions with preceding or subsequent segments. Studies of membrane integration, however, have demonstrated that TM segments may laterally exit the translocon and integrate into the lipid bilayer independently in pairs or even in groups (19 -23). Thus the translocon complex appears capable of associating simultaneously with multiple TM segments, raising the possibility that topogenic determinants may act synergistically in directing polytopic topology.
The cystic fibrosis transmembrane conductance regulator (CFTR) is an ideal substrate for examining the mechanism of polytopic protein biogenesis. CFTR exhibits a tandem repeat structure that contains two hydrophobic TM domains (TMDs) and two cytosolic nucleotide binding domains (24). A regulatory domain connects the two halves of the molecule. Both TMDs are predicted to exhibit a similar topology composed of six TM segments with the N and C termini facing the cytosol (25). CFTR is somewhat unusual, in that 7 of its 12 TM segments * This work was supported by National Institutes of Health Grants DK 51818 and GM 53457 and the Cystic Fibrosis Foundation. 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.
‡ contain a total of 10 charged residues (2 Lys, 4 Arg, 3 Glu, and 1 Asp), several of which have been implicated in contributing to a channel through which chloride ions are conducted (26 -29). Charged TM residues also influence early events of CFTR folding (14,30). For example, we previously showed that residues Glu 92 and Lys 95 markedly inhibit TM1 signal sequence activity (14). As a result, topology of TM1 is directed in part by signal sequence activity of TM2 via a ribosome-dependent posttranslational mechanism (31). Little is known about the detailed topogenic behavior of other CFTR TM segments or the effect of endogenous charged residues on the biogenesis process. However, numerous point mutations in TM segments have been implicated in causing cystic fibrosis (32), and defects in CFTR folding are the underlying cause of disease in the majority of CF patients (33). Understanding the mechanism by which CFTR assembles into the ER membrane is therefore of central importance.
To define the sequence of events that establish CFTR topology, we compared the properties of topogenic determinants encoded within TMD1 and TMD2. We now show that during biogenesis of TMD1, neither TM1 nor TM3 efficiently initiates translocation. This contrasts distinctly with corresponding segments in TMD2 (TM7 and TM9), both of which function as signal sequences. Detailed examination of TM7 and TM8 further revealed that a charged residue within TM8 (Asp 924 ) interferes with TM8 stop transfer activity and confers an unexpected flexibility on TM8 boundaries. This enables N-terminal flanking residues to extend a significant distance into the ER lumen. Moreover, TM7 was specifically required for TM8 stop transfer activity because TM8 failed to terminate translocation when engineered into an otherwise secretory protein containing a cleaved signal sequence. These studies suggest that in addition to functioning as a signal sequence, TM7 also enables ER translocation machinery to properly recognize TM8 as a transmembrane segment thus positioning D924V in the plane of the membrane. Moreover, they provide evidence that stop transfer activity can be regulated in a novel and complex manner as multiple transmembrane segments are presented to the translocon in rapid succession during polytopic protein biogenesis.
Oocyte Expression-In vitro transcription and Xenopus oocyte expression was carried out as described previously (38). Plasmids were linearized with PstI and transcribed with SP6 RNA polymerase (Epicentre, Madison, WI) (38). mRNA aliquots were used immediately or frozen and stored at Ϫ80°C. Tran 35 S-label (ICN Pharmaceuticals, Irvine, CA)) was concentrated 10-fold, added to 4 volumes of transcription mixture, and injected into stage VI Xenopus oocytes on ice (50 nl/oocyte, 50 Ci of Protease Protection, Immunoprecipitation-Oocyte homogenate was aliquoted on ice, and proteinase K (0.2 mg/ml final concentration) and Triton X-100 (1%) were added as indicated. Samples were incubated on ice for 1 h. Residual protease was inactivated by addition of phenylmethylsulfonyl fluoride (10 mM) and rapid mixing with 10 volumes of 1% SDS, 0.1 M Tris, pH 8.0, preheated to 100°C and incubating at 100°C for 5 min. Samples were then diluted in 10 volumes of Buffer A (0.1 M NaCl, 1% Triton X-100, 2 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, and 0.1 M Tris, pH 8.0). X. laevis oocytes samples were incubated at 4°C for 1 h and centrifuged at 16,000 ϫ g for 15 min to remove insoluble debris. Anti-prolactin antisera (ICN Biomedicals, Costa Mesa, CA) were added at 1:1000 -1:2000 dilution and preincubated for 10 -30 min, and 5.0 l of protein A Affi-Gel (Bio-Rad) was added. Samples were rotated at 4°C overnight and washed 3ϫ with Buffer A and twice with 0.1 M NaCl and 0.1 M Tris, pH 8.0, prior to addition of SDS sample buffer. Fidelity of protease protection was typically 85-95% as determined by a secretory control protein.
Endoglycosidase Digestion-Following immunoprecipitation, samples were eluted from protein A beads by incubation in 0.1% SDS, 0.1 M sodium citrate, pH 5.5, at 100°C for 2 min. Supernatant was split into 2 aliquots, and endoglycosidase H (100 units) was added to 1 aliquot prior to incubation at 37°C for 2 h. Samples were dried, dissolved in SDS loading buffer, and analyzed by SDS-PAGE.
Autoradiography and Quantitation-Samples were analyzed on 12-17% gradient gels by SDS-PAGE, EN 3 HANCE (PerkinElmer Life Sciences) fluorography, and autoradiography. Representative autoradiogram were scanned using a Umax PowerBook III scanner and Adobe Photoshop software. All quantitation was performed using a Bio-Rad Personal Molecular PhosphorImager Fx (Kodak screens, Quantity-1 software). Band intensities were calculated based on background-corrected volume-averaged pixel intensity. Translocation efficiencies were calculated from protease-protected, detergent-sensitive peptide relative to non-digested material. For protected fragments, signal intensity was corrected for the fractional methionine content based on fragment size. Calculations show average of at least three experiments Ϯ S.E.

RESULTS
To compare the behavior of topogenic determinants encoded within CFTR TMD1 and TMD2, coding sequences were truncated after each TM segment and tagged at the C terminus with a translocation reporter derived from bovine prolactin. This reporter has been extensively studied; it contains no intrinsic topogenic information, and it passively and faithfully follows the direction of upstream topogenic determinants (4,39,40). Predicted topology, location of methionine residues, and truncation sites of TMD1 and TMD2 are shown schematically in Fig. 1. Fusion proteins were expressed in Xenopus laevis oocytes as described under "Materials and Methods." The advantage of this system is that X. laevis oocytes efficiently process functional CFTR and thus recapitulate physiological translocation events. In addition, oocyte homogenates yield uniform right-side-out ER membranes amenable to standard protease protection assays (i.e. lumenal protein domains are protected from exogenously added proteinase K in the absence but not the presence of a non-denaturing detergent). A signal sequence N-terminal to the reporter would thus direct translocation into the ER lumen resulting in protease protection, whereas a preceding stop transfer sequence would terminate translocation and orient the reporter in the cytosol.
TMD1 and TMD2 Utilize Different Biogenesis Pathways-If CFTR utilized a simple cotranslational biogenesis mechanism, then TMs 1, 3, 5, 7, 9, and 11 should function as signal anchor sequences and initiate translocation of ECLs. TMs 2, 4, 6, 8, 10, and 12 would be expected to exhibit stop transfer activity. However, within TMD1, TM1 and TM3 were very inefficient at initiating translocation (21 and 10% of nascent chains, respectively, Fig. 2A, lanes 1-3 and 7-9). This is consistent with previous studies (14,30) and indicates that translocation of ECL1 and ECL2 must be facilitated by additional topogenic determinants. It has been shown previously that TM2 is both necessary and sufficient to direct topology of ECL1 (14), and it is likely that topology of ECL2 is similarly established by signal anchor activity(ies) of TM4 or TM3-4 acting in concert. As translation continues, translocation is cotranslationally reinitiated by TM5 and terminated by TM6 to properly orient ECL3 in the ER lumen.
In contrast to TMD1, translocation of TMD2 is reinitiated sequentially by signal anchor sequences within TMs 7, 9, and 11 and terminated by stop transfer sequences within TMs 8, 10, and 12 (Fig. 2B). Thus, each TM segment acquires its final topology in a simple cotranslational manner as it exits the ribosome. Several constructs, TM7-9 and TM7-11, give rise to multiple protease-protected fragments (Fig. 2B, lanes 7-9 and  13-15). This is likely due to the relative accessibility and cleavage of cytosolic loops by protease.
Effect of CF-related Mutations on TM8 Topology-One striking difference between TMD1 and TMD2 was the efficient signal and stop transfer activities of TM7 and TM8 as compared with their N-terminal counterparts TM1 and TM2. In addition, TM8 is particularly susceptible to disease-related CF mutations (Cystic Fibrosis Genetic Analysis Consortium, www. genet.sickkids.on.ca/cftr). Nine point mutations involving eight residues have been reported within TM8 in CF patients, and an additional seven mutations are located in TM8 flanking regions (Fig. 3A). We therefore focused attention on this region to better understand how TM7-8 contributes to CFTR biogenesis under normal and pathologic conditions. Initial studies examined the effect of five CF-related mis-sense mutations on the topology of TM8. The coding region of TM7-8 was isolated and engineered upstream of the translocation reporter, and topology was determined by protease protection. This analysis was simplified by the presence of two endogenous N-linked glycosylation consensus sites between TM7 and TM8 at residues Asn 894 and Asn 900 in ECL4 and by a third consensus site upstream of TM7. In the wild type construct, two sites are efficiently glycosylated consistent with translocation of the TM7-8 loop into the ER lumen (Fig. 3B). Glycosylation was confirmed by Endo H digestion (Fig. 3C). In addition, TM8 C-terminal residues were cytosolic and accessible to protease (Fig. 3D). Because TM7 translocates its C terminus and spans the membrane with N cyt /C lum topology (Fig.  2B), topogenic information encoded with TM7 and TM8 was thus entirely sufficient to establish proper topology of ECL4. Interestingly, none of the CF-related mutations significantly affected topology. In all cases, the nascent chain was doubly glycosylated, and the translocation reporter remained cytosolic (Fig. 3D). CFTR TM8 Lacks Independent Stop Transfer Activity-Although TM8 terminates translocation when it follows TM7, the P-reporter was protected from protease in ϳ14% of nascent chains. This is not excessive given the sensitivity of the assay. However, the D924V mutation consistently resulted in a 10fold decrease in the amount of reporter translocated (Fig. 3D). This was particularly interesting given that the hydropathy profile of TM8 reveals a weakly hydrophobic segment broadly extending from Ser 912 to Ser 945 (34 residues) (Fig. 4A). The D924V mutation replaces a charged residue near the middle of the predicted TM8 segment and thus significantly increases TM8 hydrophobicity.
We therefore tested the effect of D924V mutation on TM8 stop transfer activity by engineering TM8 together with its flanking residues into a chimeric cassette previously used to define stop transfer sequences derived from bitopic and polytopic proteins (Fig. 4B) (4,37,39,40,42). The cassette contains an N-terminal cleaved signal sequence followed by two passenger domains derived from ␣-globin (N-terminal to TM8) and prolactin (C-terminal to TM8). In this context, WT TM8 was surprisingly inefficient at terminating translocation; passenger domains on both sides of TM8 were protected from protease (70%) and hence translocated into the ER lumen ( Fig. 4B and Fig. 5). Note that this chimera contains an additional glycosylation site within the globin passenger and two glycosylation sites at CFTR codons Asn 894 and Asn 900 and therefore frequently migrates as a doublet consisting of triply glycosylated (top band) and doubly glycosylated bands (lower band, Fig. 4B). Glycosylation was confirmed by Endo H digestion (data not shown). Similar results were observed for TM8 containing mutations Y917C, R933S, H939D, and H939R (Fig. 5). The D924V mutation, however, markedly improved TM8 stop transfer activity and terminated translocation in ϳ90% of polypeptides, all of which acquired N lum /Ccyt topology (Figs. 4B and 5).

TM8 Exhibits Flexible Boundaries and Extends into the ER Lumen during Early Events of CFTR Biogenesis-Although
TM8 terminates translocation when it follows TM7, it does not terminate translocation when initiated by a cleaved signal sequence. Moreover, this unusual property is due in part to the presence of a charged residue, Asp 924 , within the hydrophobic core of the TM segment. This suggests that during biogenesis of TMD2, TM8 interacts with the Sec61 translocon in a manner different from usual stop transfer sequences. We therefore used the accessibility of N-linked glycosylation consensus sites as a means to measure these interactions.
Utilization of lumenal glycosylation consensus sites in membrane proteins is dependent on their accessibility to oligosac- charyltransferase (OST), which is related to distance of the site from the ER membrane. Lumenal loops of polytopic membrane proteins exhibit a relatively sharp decrease in glycosylation when consensus sites are closer than 12 residues from the C terminus and 14 residues from the N terminus of TM segments (43,44). CFTR residues Asn 894 and Asn 900 within ECL4 are thus close to the threshold needed for efficient glycosylation (see Fig. 6A). Conversion of Asn 900 to Gln and introduction of a new consensus site at codon 903 revealed that Asn 903 was glycosylated with the same efficiency as WT protein even though it was only nine residues from TM8 (Fig. 6B). Glycosylation was verified by Endo H digestion (Fig. 6C). In addition, a naturally occurring CF-related mutation T908N, which introduces a consensus site just four residues from TM8, was also efficiently glycosylated both in the WT context and in the N900Q mutant (Fig. 6B, lanes 7-12). This latter finding is consistent with previous studies (45) showing that T908N is also glycosylated in full-length CFTR expressed in mammalian cells. Alteration of glycosylation consensus sites did not affect the topological orientation of TM8 (Fig. 6B). Thus, even though TM8 terminates translocation, it appears to exhibit a marked flexibility by extending its N-terminal flanking residues significantly into the ER lumen, thus allowing glycosylation to take place at residues very close to the predicted TM segment boundary.
D924V Eliminates TM8 Flexibility-Finally, we tested whether strengthening TM8 stop transfer activity also influenced the extension of TM8 into the ER lumen. In the D924V mutant, both the native glycosylation sites (Asn 894 and Asn 900 ) as well as the engineered site, Asn 903 are efficiently utilized (ϳ90% glycosylation, Fig. 7A, lanes 1-3). In contrast to wild type TM8, however, the D924V mutation completely prevented glycosylation at residue Asn 908 (Fig. 7A, lanes 4 -6 and 10 -12). Thus, the improved ability of TM8 to terminate translocation in the D924V mutant limits the accessibility of TM8 N-terminal flanking residues to OST as would be expected for a bona fide stop transfer sequence. DISCUSSION The mechanisms by which ER translation and translocation machinery utilize topogenic information to direct folding and assembly of polytopic proteins is central to our understanding of membrane protein biology. Here we fused a C-terminal translocation reporter after each TM segment of CFTR to identify topogenic determinants responsible for generating the 12spanning topology characteristic of many mammalian ABC transporters. These results now provide a framework for describing the sequence of translocation initiation and termination events that give rise to CFTR transmembrane topology and identify a novel mechanism of cooperation between TM segments during early biogenesis events.
As CFTR first emerges from the ribosome, the RNC is targeted to the ER membrane by either of two alternate complementary signal sequences encoded within TM1 and TM2. Because TM1 is a weak signal, translocation of ECL1 is primarily mediated by TM2 which post-translationally positions TM1 (C terminus first) into its membrane spanning orientation (14). We now show that TM3 also exhibits poor signal sequence activity, and we propose that ECL2 topology is likely established in a similar manner via cooperative signal sequence activities of TM4 or TM3-4. The relatively weak signal sequence activity of TM3 is consistent with studies previously reported by Chen and Zhang (30), although TM3 function may also be context-dependent, as it efficiently initiated translocation (ϳ60% of chains) in a cell-free system when it replaced the TM segment of the transferrin receptor (46). As TM5 emerges from the ribosome, it initiates translocation of ECL3, and TM6 terminates translocation and allows NBD1 and the R domain to emerge into the cytosol. In the case of TM6, our results differ from those reported by Tector and Hartl (47), who found that

FIG. 2. Protease protection of TMD1 (A) and TMD2 (B) fusion proteins.
Constructs encoding increasing numbers of TM segments (ovals) were expressed in Xenopus oocytes as described under "Materials and Methods." Oocyte homogenates were digested with proteinase K (PK) in the presence and absence of nondenaturing detergent (Det) and immunoprecipitated with antisera against the C-terminal reporter domain (hatched rectangle). Translocation efficiency of the reporter is indicated below each autoradiogram. Band intensities of protected fragments (upward arrowheads) were corrected for relative methionine content and expressed as a percent of signal from each construct (as described under "Materials and Methods"). Note the intensity of the protected band for construct TM1-5 reflects methionine distribution as shown in Fig. 1. Translocation efficiencies were corrected for average protease protection as determined for a known secretory control protein (81%, n ϭ 9). Values represent average of at least three separate experiments Ϯ S.E. Experimentally derived topology for TMD2 constructs is shown at the bottom.
CFTR TM6 failed to terminate translocation when expressed as an invertase fusion protein in yeast. The reason for this discrepancy is unclear but may be related to the use of different expression systems and/or alternate truncation sites downstream of TM6 (14 versus 42 residues in this study). We favor the latter explanation because C-terminal flanking residues are known to influence stop transfer activity and because TM6 efficiently terminated translocation (Ͼ90% of chains) when examined independently in a secretory cassette (data not shown).
In contrast to TMD1, TMD2 topology is established in a sequential and cotranslational manner by alternating signal (TM7, -9, and -11) and stop transfer (TM8, -10, and -12) sequences. At present it is unclear whether the RNC remains bound to, or is released from, the Sec61 translocon during synthesis of the large cytosolic nucleotide binding domain and R domains. If released, TM7 would be required to re-target the RNC to a new translocon, open the translocon gate, reinitiate translocation of ECL4, and establish the ribosome-membrane junction (48). Topology of subsequent TM segments (i.e. TM8 -12) could then be established simply by alternate gating of the translocon and ribosome membrane junction into the ER lumen (for ECLs) and into the cytosol (for cytosolic loops). Thus, CFTR TMDs exhibit a common six-spanning topology but clearly encode different topogenic information and acquire their transmembrane topology through different translocation events.
A significant difference between CFTR TMDs is reflected by topogenic properties of TM1-2 versus TM7-8. Whereas TM1 signal sequence activity is decreased by charged residues (Glu 92 and Lys 95 ) (14), this is clearly not the case for TM7 which encodes a strong signal sequence that independently directs membrane targeting and translocation. Surprisingly, whereas TM8 is able to terminate translocation, span the membrane, and establish ECL2 topology during CFTR synthesis, topogenic behavior of TM8 is profoundly influenced by the presence of TM7. This observation raises questions regarding how multiple topogenic determinants might function in concert during protein biogenesis. In bitopic proteins for example, signal sequences facilitate tight ribosome binding to Sec61, gate open the translocation channel, and initiate translocation of the nascent chain into the ER lumen (8,49,50). Stop transfer sequences then terminate translocation, close the translocon, and allow the nascent chain access to the cytosol (12). Whereas signal and stop transfer sequences have traditionally been viewed to function independently, recent evidence indicates that this is not always the case. Indeed, signal sequences differ in how they regulate the ribosome-ER membrane junction (51,52) and thus can exert downstream effects on stop transfer activity and final protein topology (53).
Our observation that TM8 stop transfer function is dependent on the presence of TM7 significantly increases the complexity of stop transfer function during polytopic protein biogenesis. Given the relatively short length of ECL4 (31 residues), it is almost certain that TM7 remains within or at least adjacent to the translocon during the synthesis of TM8. Our results further indicate that TM7 interactions with translocon components influence the ability of TM8 to terminate translocation, gate the translocation, and thus properly direct CFTR residues into the cytosol. In a chimeric protein that contains a cleaved signal sequence, TM8 fails to gate the translocon closed and the C-terminal reporter continues translocating into the lumen. Whether this occurs because the signal sequence is cleaved and leaves the translocon or whether it remains associated with the translocon but through different interactions than those of TM7 remains unknown. In either event, the behavior of TM8, and hence its topologic outcome, is directly determined by the nature of its preceding sequences. Thus, the translocon appears capable of simultaneously integrating information from multi-ple topogenic determinants during polytopic protein biogenesis when these determinants are presented in rapid succession. An important next step will be to define how such determinants interact and regulate translocon behavior.
A second puzzling finding was the remarkable accessibility of N-linked glycosylation consensus sites within ECL4. Bitopic and polytopic proteins generally exhibit a narrow threshold distance from the ER membrane where lumenal consensus sites are inefficiently utilized (43,44). This threshold is thought to be due to steric constraints that limit the access of OST to Asn consensus sites (43) and possibly the length of time that the nascent chain resides within or adjacent to the translocon (54). The finding that OST has access to sites as close as four residues from the end of the TM segments strongly indicates that at least during the early stages of its biogenesis, TM8 exhibits an unusual flexibility in its TM boundaries and extends significantly farther into the ER lumen than would be expected. This flexibility is directly influenced by the presence of a charged residue (Asp 924 ) within the center of the predicted TM8 segment. Interestingly, the lumenal extension observed for TM8 is significantly greater than was observed by Monne et al. (55) when an Asp residue was engineered into an artificial TM segment (L23V).
Several possibilities could explain these results. First, the originally assigned TM8 boundaries extend from Ser 911 to Phe 932 . However, the region surrounding TM8 consists of a relatively long hydrophobic sequence (34 residues) and contains two charged residues Asp 924 and Arg 933 (see Figs. 3A and 4A). If instead of the initially predicted boundaries, TM8 actually spanned the membrane from Asp 924 to Ser 944 , then this alternate TM8 segment would exhibit a similar overall hydrophobicity, and Thr 908 would be positioned 16 residues from the membrane where it could be easily accessed by OST. The  Fig. 6 were engineered into a D924V background, and topology (A) and glycosylation (B) were determined by protease protection and endoglycosidase digestion. Consistent with its effect on TM8 ST activity, the D924V mutation prevented access of oligosaccharyltransferase to the T908N consensus site. Open circles indicate unused consensus sites. D924V mutant would then shift the TM boundaries by increasing hydrophobicity of the N-terminal region, thus sterically blocking glycosylation at residue 908. A second possibility is that TM8 might exhibit alternate or overlapping TM segment conformations that would position either a negative or positive charge within the plane of the membrane, perhaps during different states of channel gating. This is particularly intriguing because such a shift could have significant effects on the character of the pore, and Asp 924 is directly involved in pore architecture through formation of a salt bridge with residue Arg 347 in TM6 (28).
A third possibility is that as TM8 enters the translocon, its overall amino acid composition and relatively weak hydrophobicity enable it to transiently extend into the ER lumen prior to forming the TM segment. In this regard, nascent polypeptides are proposed to retain an extended conformation while exiting the large ribosome subunit (10,11), although early formation of helices may also be possible (56,57). TM helices might thus form within the aqueous environment of the translocon channel (58 -60) or, alternatively, as the polypeptide begins to interact with the translocon and/or lipids that compose the lateral gateway into the bilayer (23). The timing of membrane integration (i.e. primary interaction with lipids) appears to be dependent on the degree of TM segment hydrophobicity and, in particular, to the presence of charged residues. TM segments with charged residues partition more slowly into the bilayer and remain associated with the translocon complex (i.e. TRAM) for extended periods, even until the completion of protein synthesis (22,23). It is therefore plausible that during CFTR synthesis, the relatively polar nature of wild type TM8 favors an unstructured conformation as the nascent chain enters the translocon lumen. This would extend TM8 N-terminal residues ϳ30 Å further into the ER lumen than would a helical conformation and transiently increase accessibility to OST (55,60,61). In this model, residue Asp 924 would be expected to delay membrane integration (and possibly helix formation) by decreasing the tendency of the helix to move into the bilayer through the lateral translocon gate (23). In contrast, the D924V mutant would impose new folding constraints that favor membrane integration and helix formation within the lipid bilayer, thus decreasing the time for OST accessibility and increasing the apparent rigidity of TM8 membrane boundaries.
It is important to note that residue Asn 908 is also accessible to OST even in full-length CFTR expressed in mammalian cells (45,62). Thus the translocation events observed here likely reflect physiologic events of CFTR folding in its native environment and are not due to our use of CFTR fragments or heterologous expression in X. laevis oocytes. During synthesis of full-length CFTR, however, the presence of additional TM segments (i.e. TMs 1-6) likely exert additional cooperative folding interactions, e.g. helical packing and formation of a salt bridge (28). Until the relative location of TM segments within the translocon is defined, it is not possible to predict how and when such long range intramolecular CFTR interactions might influence the early folding events described here.
In summary, native polytopic proteins utilize a diverse repertoire of allowed folding pathways to acquire their final topology in the ER membrane. One feature that governs these pathways is the information content and organization of topogenic determinants encoded within TM segments and their flanking residues. Even subtle changes in sequence can markedly influence topogenic behavior. This study highlights the close relationship that exists between determinants of protein topogenesis and critical determinants of protein function. At present, relatively little is known about how specific residues influence and dictate polytopic protein folding pathways. Our study provides an example of this interplay by defining a novel mechanism by which CFTR achieves proper positioning of the charged residue, Asp 924 , within the plane of the lipid bilayer. This mechanism not only requires a precise arrangement of topogenic information but a cooperative interaction between TM7 and TM8 and the ER translocation machinery. Understanding the molecular nature of these interactions during polytopic protein biogenesis is the next challenge in our ability to predict folding pathways and folding outcomes induced by minor sequence perturbations in inherited disorders.