Misfolding diverts CFTR from recycling to degradation : quality control at early endosomes

Multiple quality control mechanisms are required to prevent the cellular accumulation and cytotoxicity of misfolded, aggregation-prone polypeptides (Sherman and Goldberg, 2001; Arvan et al., 2002; Ellgaard and Helenius, 2003). Newly synthesized soluble and membrane proteins that are unable to fold or assemble are targeted for proteolysis by the ER-associated degradation mechanism via the cytosolic ubiquitin (Ub)–proteasome system (Brodsky and McCracken, 1999; Hampton, 2002; Ellgaard and Helenius, 2003). In addition, nonnative membrane proteins are also recognized and targeted for lysosomal proteolysis directly from Golgi compartments (Wolins et al., 1997; Reggiori and Pelham, 2002). Yet, misfolded membrane proteins can escape these quality control checkpoints and reach the plasma membrane. Conformationally unstable plasma membrane proteins, such as unliganded MHC I and mutant variants of cystic fibrosis transmembrane conductance regulator (CFTR), α₂-receptors, α-factor receptor, and transferrin receptor, as well as influenza HA are rapidly degraded (Ljunggren et al., 1990; Li et al., 1999; Zaliauskiene et al., 2000; Benharouga et al., 2001; Sharma et al., 2001; Wilson et al., 2001; Fayadat and Kopito, 2003). Although a variety of sorting signals have been described to account for the constitutive or ligand-induced down-regulation of fully folded cell surface receptors and transporters (Rotin et al., 2000; Sorkin and Von Zastrow, 2002; Bonifacino and Traub, 2003; Hicke and Dunn, 2003), the molecular machinery that prevents the accumulation of misfolded plasma membrane proteins is not known in mammalian cells (Arvan et al., 2002).

Impaired plasma membrane expression or function of the CFTR, a cAMP-activated chloride channel, accounts for the phenotypic manifestation of CF, the most common inherited disorder in the Caucasian population (Zielenski and Tsui, 1995). Previously, we have identified two naturally occurring pathogenic mutants that display thermosensitive conformational defects in post-Golgi compartments. The COOH-terminally truncated CFTR, lacking the last 70 amino acid residues (Δ70 CFTR), traverses the biosynthetic pathway and undergoes complex glycosylation with the efficiency of wild-type (wt) CFTR (Haardt et al., 1999; Benharouga et al., 2001). In contrast, deletion of Phe508 (ΔF508, found in 90% of CF patients) imposes an ER processing block on CFTR that could be partially overcome or “rescued” at permissive temperature (<30°C) or by chemical chaperones in cultured cells (Denning et al., 1992; Sato et al., 1996). The conformational defect of the complex-glycosylated rescued ΔF508 (rΔF508) and the Δ70 CFTR has been demonstrated by their increased protease susceptibility, accounting for the four- to sixfold faster metabolic turnover of the mutants in post-Golgi compartments (Zhang et al., 1998; Benharouga et al., 2001; Sharma et al., 2001).

The present work was undertaken to elucidate the retrieval mechanism of conformationally defective Δ70 and rΔ508 CFTR from the cell surface that may represent a paradigm for the peripheral quality control of membrane proteins. The results suggest that misfolding of CFTR dramatically augments the ubiquitination susceptibility of the channel in post-Golgi compartments. In turn, Ub modification serves as a recognition signal for the Ub-dependent endosomal sorting machinery that reroutes the channel from recycling toward the multivesicular body (MVB)/lysosomal degradation.

Three tandem hemagglutinin (3HA) tags were engineered into the 4th extracellular loop of CFTR variants and expressed stably in BHK to monitor the cell surface density, internalization, and recycling rates of CFTR variants. The processing defect of the ΔF508 CFTR and the marked decrease in the expression level of mature, complex-glycosylated rΔF508 and Δ70 CFTR-3HA were in line with previous data obtained on untagged channel or on CFTR variants with NH₂- or COOH-terminal HA tags (Fig. 1 a; Haardt et al., 1999; Heda et al., 2001; Sharma et al., 2001). The rΔF508 and Δ70 CFTR have nearly 10-fold decreased cell surface density compared with their wt counterpart, measured by radioactive anti-HA antibody binding (Fig. 1 b), which can be, at least in part, attributed to their rapid disposal rates (Fig. 1 c). The cell surface half-life (T_1/2) of rΔF508 and Δ70 CFTR was ≈90% shorter (T_1/2 ≈ 2 h and ≈ 1 h, respectively) than that of the wt CFTR (T_1/2 ≈ 16 h), determined by the disappearance of anti-HA antibody binding in the presence of cycloheximide (CHX; Fig. 1c and Fig. S1). Similar results were obtained by monitoring the decay kinetics of bound anti-HA antibody in the absence of CHX (unpublished data).

To assess whether the rapid turnover of the mutants is due to their heterologous overexpression in BHK cells by itself, the biochemical and functional stability of the channel was measured in polarized pancreatic duct (PANC-1) and primary respiratory epithelia, representing a heterologous and an endogenous expression system, respectively. PANC-1 cells were chosen because the tightest genotype–phenotype correlation was established for the pancreatic manifestation of CF (Zielenski and Tsui, 1995). PANC-1 cells were stably transfected with the wt and ΔF508 CFTR. The biochemical turnover of the complex-glycosylated CFTR was measured by CHX- and brefeldin A-chase and immunoblotting on polarized monolayers. The turnover of rΔF508 was four- to fivefold faster than the wt form by both methods (Fig. 2, a and b). Comparable data were obtained for the functional stability of rΔF508 CFTR in the apical membrane, monitored by the iodide efflux assay (Fig. 2 c). Importantly, an even faster functional removal of the rΔF508 CFTR occurred from the apical plasma membrane of differentiated respiratory epithelia, obtained from nasal polyps of ΔF508/ΔF508 patients (Fig. 2 d and Fig. S2). The cAMP-stimulated iodide release, mediated by rΔF508 CFTR, almost completely disappeared after 4 h of chase at 37°C, whereas it remained unaltered in case of the wt CFTR (Fig. 2 d). These observations indicate that the metabolic instability of the rΔF508 CFTR is an inherent characteristic of the mutant and is independent of the expression system used.

Accelerated internalization, attenuated recycling, and/or facilitated targeting toward lysosomal degradation may account for the 10-fold decrease in the cell surface T_1/2 of the mutant CFTR. To distinguish between these possibilities, first the rate of CFTR endocytosis was measured by the disappearance of cell surface–bound anti-HA antibody. No significant difference could be resolved in the endocytosis rates of Δ70 CFTR (5.3 ± 0.7%/min, n = 3), rΔF508 (4.0 ± 1.4%/min, n = 3), and wt CFTR (4.9 ± 1.1%/min, n = 3) (Fig. 3 a).

To assess whether impaired recycling contributes to the reduced cell surface T_1/2, the exocytosis of internalized CFTR and anti-HA antibody complex was monitored with biotinylated secondary antibody and ¹²⁵I-labeled streptavidin. Although 63.2 ± 5.8% (n = 4) of internalized wt returned to the cell surface, only ≈6% of rΔF508 and Δ70 CFTR recycled in 10 min (Fig. 3 b). The recycling defect may constitute the primary cause of the decreased cell surface density and stability of the Δ70 CFTR (Fig. 1), suggesting a novel cellular mechanism for the severe clinical phenotype of patients with large COOH-terminal truncations (Haardt et al., 1999). Impaired recycling may also diminish the cell surface expression of the ΔF508 CFTR in selected CF tissues, where ER retention of the ΔF508 CFTR appears to be incomplete (Kalin et al., 1999).

Although nonnative soluble and ER-associated polypeptides are known substrates of ubiquitination, the susceptibility of poorly folded plasma membrane proteins to Ub conjugation is poorly understood. To assess whether Ub modification is involved in the disposal of nonnative CFTR, the ubiquitination level of wt and mutant CFTR, confined to post-Golgi compartments, was determined. To this end, complete degradation of the core-glycosylated ΔF508, Δ70, and wt CFTR (and their ubiquitinated adducts) was ensured by treating the cells with CHX for 3 h (Fig. 4 a, lanes 1 and 2, bottom and top, respectively; Fig. S3). Then, CFTR was immunoisolated under denaturing conditions with anti-CFTR antibody, and the precipitates were probed with anti-Ub antibody. Detection of ubiquitinated rΔF508, Δ70, and wt CFTR in cells expressing exclusively the complex-glycosylated forms by three different anti-Ub antibodies demonstrated that CFTR is susceptible to ubiquitination in post-Golgi compartments (Fig. 4 a; unpublished data). Importantly, densitometry showed that ubiquitination of the complex-glycosylated rΔF508 and Δ70 CFTR was increased by ≈20-fold relative to the wt channel at 37°C (Fig. 4 b).

To explore a possible correlation between the unfolding and the ubiquitination propensity of the channel, the conformation of CFTR variants was modulated by temperature shifts. Favoring the native conformation at the permissive temperature (28°C) attenuated the ubiquitination of the mutants, whereas promoting unfolding at 40°C substantially enhanced the ubiquitination of rΔF508 CFTR and moderately affected the Δ70 CFTR (Fig. 4 c). In contrast, ubiquitination of wt CFTR was virtually independent of the temperature (Fig. 4 c), suggesting that Ub conjugation may be involved in the degradation of conformationally unstable CFTR from post-Golgi compartments.

If Ub conjugation plays a primary role in the recycling and cell surface stability defect of the mutants, fusing Ub to the wt channel (CFTR-Ub) may mimic the consequences of destabilizing mutations. In-frame fusion of Ub to the COOH terminus of CFTR indeed inhibited the channel recycling by ≈90% (Fig. 5 a), reduced its cell surface T_1/2 from ≈16 to ≈1.3 h (Fig. 5 b), and diminished the steady-state expression of the mature CFTR (Fig. 5 c). These observations indicate that the chimera reproduces the peripheral trafficking defects of the rΔF508 and the Δ70 CFTR. Fusing a Ub molecule harboring arginines in place of all its lysine residues resulted in a similar expression and recycling defect to that of the CFTR-Ub (unpublished data), suggesting that the lysine residues of Ub do not serve as acceptor sites for poly-Ub chain formation in the chimera. The unaltered internalization rates of CFTR-Ub relative to wt CFTR (4.7 ± 1.3%/min; Fig. 5 d) is consistent with the notion that the Ub-dependent recycling defect is the primary cause for the rapid down-regulation of the chimera from the cell surface, as is for Δ70 and rΔ508 CFTR.

To substantiate the role of Ub conjugation in the degradation of the rΔF508 and Δ70 CFTR, we measured the cell surface stability of mutants in ts20 CHO cells, harboring the thermosensitive E1 Ub-activating enzyme. Consistent with the results of Kulka et al. (1988), the E1 enzyme was inactivated at 40°C in ts20 cells, whereas the expression level of the wt E1 remained unaltered in E36 cells, as shown by immunoblot analysis (Fig. 6 a). The disappearance kinetics of the mutant CFTR were determined by the radioactive antibody-binding assay after the inactivation of the E1 enzyme. Although inactivation of the E1 enzyme completely prevented the elimination of the rΔF508 and Δ70 CFTR from the cell surface in ts20 cells, the degradation of the mutants resumed in E36 cells (Fig. 6 b). Importantly, inactivation of the E1 enzyme had no effect on the internalization rates of the mutants in ts20 cells (Fig. 6 c).

Because the internalization rates of mutants are independent of E1 activity as well as the ubiquitination level of the CFTR variants (Fig. 3 a and Fig. 5 d), it is conceivable that Ub modification serves as a post-endocytic sorting signal for rerouting the mutant from recycling toward the MVB/lysosomal degradation pathway. Therefore, we tested the association of mutants with constituents of the Ub-dependent endosomal sorting machinery.

It is well established that targeting of certain ubiquitinated cell surface receptors for lysosomal degradation requires the recruitment of Ub-binding adaptor proteins (e.g., Hrs, STAM, and eps15, or their yeast orthologues Vps27, HSE1, and Ede1) to the cytosolic surface of early endosomes (Katzmann et al., 2002; Bonifacino and Lippincott-Schwartz, 2003). Because the hepatocyte growth factor–regulated tyrosine kinase substrate (Hrs) and signal-transducing adaptor molecule (STAM) are the primary Ub-interacting motif-containing adaptors that form the sorting complex involving components of the endosomal sorting complex required for transport I (ESCRT I; Bilodeau et al., 2002, 2003; Bishop et al., 2002; Katzmann et al., 2002; Raiborg et al., 2002; Shih et al., 2002; Mizuno et al., 2003; Schnell and Hicke, 2003), the association of Hrs and STAM-2 with CFTR was assessed first. Immunoprecipitation of endogenous Hrs pulled down the complex-glycosylated rΔF508 and Δ70 CFTR as well as the CFTR-Ub, but neither the ER-associated ΔF508 CFTR nor the complex-glycosylated wt CFTR (Fig. 7 a). Selective association of STAM-2 (Fig. 7 b) and the TSG101 (Fig. 7 c), a component of ESCRT I, with the conformationally defective CFTR variants was also documented by the immunoprecipitation technique. Finally, similar results were obtained by probing for the association of CFTR variants with hVps25 and hVps32, components of the ESCRT II and III, respectively (Fig. 7 d). These observations are consistent with our hypothesis that the interaction of destabilized and preferentially ubiquitinated CFTR with the Ub-dependent sorting machinery is responsible for the recycling defect and accelerated degradation of the mutants.

To substantiate the inference that Ub-binding proteins have a role in rerouting misfolded CFTR from recycling to degradation, we took advantage of the observation that the hydrophobic surface of Ile₄₄ in Ub is necessary for the association of Ub-binding proteins (Shih et al., 2002). If Ub recognition is required for rerouting the chimera, disrupting this interaction should reverse the trafficking defect. First, GST-Ub pull-down assays verified that Ile₄₄Ala mutation (UbA) abolished the Hrs binding to Ub, as observed for vacuolar protein sorting 27 (Vps27; Shih et al., 2002), whereas combining Ile₄₄Ala and Phe₄Ala caused partial inhibition (Fig. 7 e). The Phe₄Ala mutation was introduced to eliminate the endocytic signal of Ub (Shih et al., 2000). The Ile₄₄Ala alone (CFTR-UbA) or in combination with Phe₄Ala (CFTR-Ub2A) rescued the recycling (Fig. 5 a) and cell surface stability of the chimeras (Fig. 5 b). Consistently, the steady-state expression of CFTR-UbA and CFTR-Ub2A was significantly augmented (Fig. 5 c) without altering their internalization rates (Fig. 5 d). These data not only ruled out an indirect effect of Ub fusion, but also demonstrated that the recycling efficiency of ubiquitinated CFTR was dramatically attenuated by association with Ub-binding protein(s) of the endosomal sorting machinery. This was underscored by the observation that single Ala replacement prevented the association of Hrs as well as TSG101, Vps25, and Vps32 with the CFTR-UbA (Fig. 7, a–c).

To demonstrate the conformation-dependent endosomal sorting of CFTR, the subcellular destination of the mutants was determined as a function of the ambient temperature. CFTR containing endocytic compartments were identified based on their characteristic pH and protein composition (Mukherjee et al., 1997). Anti–mouse F_ab fragments conjugated to the pH-sensitive fluorophore FITC were complexed to anti-HA antibody and internalized with CFTR to measure the lumenal pH of CFTR-containing vesicles by ratiometric fluorescence video image analysis (Gagescu et al., 2000). Although recycling endosomes have a pH_re of 6.5 ± 0.05 and accumulate transferrin in BHK cells, the lysosomes have a pH_ly <5.0 (measured by FITC-EGF) and harbor LAMP-1 (Fig. 8 a and Fig. S4).

At physiological temperature (37°C), internalized wt CFTR was confined to endosomes with lumenal pH_CFTR of 6.5 ± 0.1 (Fig. 8 a and Fig. S4), a characteristic of recycling endosomes (Mukherjee et al., 1997). In contrast, the rΔF508 as well as the Δ70 CFTR were delivered to vesicles with lumenal pH typical of late endosomes and lysosomes (pH_ΔF508 = 5.5 ± 0.05 and pH_Δ70 <5.0, respectively; Fig. 8 a), consistent with their rapid degradation (Fig. 1, b and c; Haardt et al., 1999; Benharouga et al., 2001). Attachment of Ub to CFTR efficiently diverted the channel from recycling endosomes into lysosomes with pH_CFTR-Ub <5.0 ± 0.05, whereas preventing the recognition of Ub by the endosomal sorting machinery rerouted the CFTR-UbA and CFTR-Ub2A chimera to recycling endosomes (pH_CFTR-UbA = 6.5 ± 0.02, pH_CFTR-Ub2A = 6.5 ± 0.1; Fig. 8 a). Promoting unfolding of the rΔF508 CFTR at 40°C accelerated its delivery to more acidic vesicles (pH_rΔF508 = 5.0 ± 0.1) already within 1 h of internalization (Fig. 8 a). Conversely, at the permissive temperature where the native conformer was the predominant form, both rΔF508 and Δ70 CFTR were targeted to recycling endosomes with lumenal pH of 6.2 ± 0.05 and 6.19 ± 0.06, respectively (Fig. 8 a). These observations demonstrate the conformation-dependent sorting capacity of early endosomes. The targeting of the conformationally stable CFTR-Ub chimera to lysosomes and the wt CFTR-Ub2A and transferrin to recycling endosomes was insensitive to temperature shifts (Fig. 8 a).

Immunostaining of internalized CFTR variants with markers of recycling endosomes (transferrin) and lysosomes (FITC-dextran and Lamp-1) confirmed that internalized rΔF508 and Δ70 CFTR were diverted to MVB/lysosomes, whereas the wt CFTR was associated with recycling endosomes at 37°C (Fig. 8 b; unpublished data). Lysosomal proteolysis of the wt and Δ70 CFTR was also supported by the detection of immunoreactive degradation intermediates in purified lysosomes (Benharouga et al., 2001).

Besides sequestration of the mutant CFTR at the early endosomes, accelerated targeting into MVB/lysosomes could enhance the efficiency of the channel degradation (Katzmann et al., 2002; Hicke and Dunn, 2003). We used multicompartment kinetic analysis to estimate the intercompartmental transfer rates of the wt and rΔF508 channels. Endocytic rate constants (k_en) were calculated from the internalization rates. Based on k_en, turnover of the cell surface and the complex-glycosylated pools, degradation (k_deg), and recycling rate (k_re) constants were computed by the SAAM II multicompartment analysis program. The k_deg of the rΔF508 CFTR was increased by twofold relative to wt CFTR (Fig. 8 c). The k_re of rΔF508 CFTR was attenuated by nearly fivefold as compared with the wt (Fig. 8 c). These results, together with the recycling measurements, suggest that misfolding has a major impact on the sequestration of CFTR at early endosomes. Endosomal retention in concert with modestly accelerated transfer rates into MVB/lysosomal compartments is sufficient to attenuate the cell surface stability of the rΔF508 CFTR by 10-fold (Fig. 1 c).

Major conformational defects account for the Ub-dependent degradation of newly synthesized wt and mutant CFTR at ER (Kopito, 1999). Comparison of the ER and peripheral quality control of CFTR suggests that both processes entail three consecutive steps: the recognition of misfolded CFTR, the delivery of the channel to the relevant proteolytic machinery, and the proteolysis itself. Although substantial progress has been made in our understanding of membrane protein quality control at the ER (Cyr et al., 2002; Hampton, 2002; Ellgaard and Helenius, 2003; McCracken and Brodsky, 2003), the present work represents the first attempt to characterize the peripheral counterpart of this process in mammalian cells.

Compelling evidence suggests that Ub conjugation is a prerequisite for the disposal of misfolded rΔF508 and Δ70 CFTR from the cell surface.

First, we demonstrated a correlation between the conformational destabilization and the ubiquitination propensity of rΔF508 and Δ70 CFTR in post-Golgi compartments. Rescuing the folding defect at the permissive temperature (28°C) diminished, whereas thermo-denaturation (40°C) further increased the ubiquitination of the mutants, leading to the preferential recycling or MVB/lysosomal targeting, respectively (Fig. 4 c). Neither accelerated biosynthesis nor the saturation of the degradation machinery can explain the nearly 20-fold increased level of ubiquitinated mutants as compared with that of the wt CFTR. The translational rates of the Δ70 and ΔF508 CFTR were comparable to that of the wt CFTR in BHK cells (Zhang et al., 1998; Haardt et al., 1999), and the copy number of wt CFTR at the cell surface is estimated to be only a few thousand molecules per cell.

Second, covalent attachment of Ub to the COOH-terminal tail of the wt CFTR reproduced the recycling as well as the cell surface stability defects of the mutants (Fig. 5). The genetically fused Ub mimics the cellular consequences of post-translational ubiquitination of CFTR rather than provoking misfolding. This conclusion is supported by the largely preserved processing efficiency of the CFTR-Ub (unpublished data), and the proportionally decreased PKA-stimulated whole-cell current and cell surface density of the chimera (Fig. S5). Furthermore, the Ile₄₄Ala point mutation in Ub was able to restore the recycling and stability of the chimera to that of the wt CFTR (Fig. 5).

Finally, the most direct evidence for the role of Ub-conjugation was provided by the fact that heat inactivation of the E1 Ub-activating enzyme prevented the disappearance of Δ70 and rΔF508 CFTR from the plasma membrane in ts20 cells without influencing their endocytosis rates (Fig. 6). Collectively, these observations support the pivotal role of Ub modification in the down-regulation of the misfolded rΔF508 and Δ70 CFTR from the cell surface via post-endocytic mechanism(s).

Ubiquitination is known to effectively down-regulate receptors and transporters from the plasma membrane of yeast and mammalian cells (Rotin et al., 2000; Sorkin and Von Zastrow, 2002; Bonifacino and Traub, 2003; Hicke and Dunn, 2003). Ligand-induced (e.g., EGF and β-adrenergic agonist) or constitutive (e.g., ENaC) ubiquitination of membrane proteins requires substrate recognition by Ub-conjugating enzymes and Ub protein ligases (E2/E3s; Pickart, 2001). This process is usually mediated by protein–protein interaction domains of the relevant E3s (e.g., SH2, WW, and PDZ domains in Cbl, Nedd4/Rsp5, and LNX, respectively) and their cognate binding sites (e.g., phospho-Tyr, PPXY, phospho-Ser/Thr, or PDZ-binding motif; Weissman, 2001; Hicke and Dunn, 2003). Although it cannot be precluded that exposure of similar signals is involved in the ubiquitination of the two CFTR mutations, we favor the scenario that Ub conjugation is promoted by the unfolding of the channel and involves solvent-exposed hydrophobic protein surfaces. A similar mechanism has been invoked in the degradation of misfolded polypeptides at the ER (Laney and Hochstrasser, 1999; Cyr et al., 2002; Ellgaard and Helenius, 2003). Structural destabilization of the rΔF508 and Δ70 CFTR nucleotide-binding domains was indeed demonstrated (Zhang et al., 1998; Benharouga et al., 2001). This mechanism would also be reminiscent of the degradation signal identified in random peptides and in the Matα2 transcription factor (Sadis et al., 1995; Gilon et al., 1998, 2000; Johnson et al., 1998).

Importantly, a small amount of ubiquitinated wt, complex-glycosylated CFTR was reproducibly detected by three different anti-Ub antibodies (unpublished data). This cannot be attributed to the presence of other ubiquitinated polypeptides because CFTR was isolated under denaturing conditions. It is tempting to speculate that the low level of ubiquitination of the wt CFTR is caused by its slow, physiological unfolding that eventually terminates its long residence time (T_1/2 ≈ 14 h) at the cellular periphery. The profound difference in the metabolic turnover of protease-resistant, structurally stable wt and the protease-susceptible, conformationally labile mutant CFTR is consistent with the previously proposed hypothesis that the structural stability of soluble polypeptides constitutes one of the determinants of their metabolic stability (McLendon and Radany, 1978; Parsell and Sauer, 1989; Kowalski et al., 1998; Klink and Raines, 2000).

Recycling of plasma membrane proteins protects polypeptides from degradation and allows them to undergo repeated cycles of endocytosis and exocytosis (Ghosh and Maxfield, 1995; Mellman, 1996). On the other hand, the sorting process at early endosomes offers an efficient mechanism to prevent the accumulation of misfolded membrane proteins at the cell surface. This is exemplified by the second step of the peripheral quality control of CFTR. Using Ub adducts of CFTR, obtained by post-translational ubiquitination or by genetic engineering, we presented three lines of evidence in support of the notion that ubiquitinated channels are selectively retrieved from recycling and are redirected for degradation into MVB/lysosomes. First, ubiquitination efficiently prevented the recycling and dramatically reduced the cell surface density, as well as the stability of the mutant and the chimera (Fig. 1 c, Fig. 3 b, and Fig. 5, a and b). Second, vesicular pH measurements verified that both the mutants and the CFTR-Ub are targeted into lysosomes after their internalization, in contrast to the wt CFTR, which traverses recycling endosomes (Fig. 8 a). The recycling defect provides a plausible explanation for the 4–22-fold faster metabolic turnover rates of the rescued ΔF508 CFTR at 37 and 40°C, respectively (Sharma et al., 2001). Rescuing the folding defect and thus reducing the ubiquitination of the rΔF508 and Δ70 CFTR restored their constitutive recycling (Fig. 8 a) in parallel to their metabolic stabilization (Benharouga et al., 2001; Sharma et al., 2001). Third, immunocolocalization of endocytosed CFTR with markers of the recycling compartment and lysosomes substantiated the notion that the native and ubiquitinated CFTR are segregated at early endosomes (Fig. 8 b).

Ub-binding adaptor proteins, including Hrs and STAM, have a critical role in the retrieval of ubiquitinated cargo for lysosomal degradation at sorting endosomes (Raiborg et al., 2002; Sachse et al., 2002). Selective binding of destabilized rΔF508, Δ70 CFTR, and CFTR-Ub variants, but not the wt CFTR to Hrs, STAM-2, TSG101, Vps25, and Vps32, is consistent with the hypothesis that transport of ubiquitinated CFTR from early endosome to MVB/lysosomes involves its successive association with Hrs/STAM-2 and the human homologues of the ESCRT I, II, and III (Katzmann et al., 2002). Importantly, the I₄₄A mutation in Ub not only prevented the recognition of CFTR-UbA and GST-UbA by Hrs and STAM-2, but also did so with downstream components of the ESCRT I, II, and III complexes (Fig. 7). As a result, CFTR-UbA resumed its constitutive recycling and escaped from MVB/lysosomal degradation (Fig. 5 b and Fig. 8 a), substantiating our working model that Ub recognition and the subsequent association of the misfolded CFTR with components of ESCRT complexes are required for MVB/lysosomal targeting.

Suppressor screens in yeast have identified several class E Vps mutants, including Vps23 and Vps27 (the yeast orthologues of TSG101 and Hrs, respectively), which are either directly or indirectly involved in the degradation of misfolded plasma membrane proteins (Li et al., 1999; Gong and Chang, 2001). Although the role of ubiquitination in the disposal of misfolded membrane proteins has not been established in yeast (Arvan et al., 2002), these observations suggest that some of the components of the peripheral quality control are evolutionarily conserved.

In summary, our analyses demonstrate that the folding state of CFTR is monitored not only during the early stage of its biogenesis in the ER (Brodsky and McCracken, 1999; Ellgaard et al., 1999; Arvan et al., 2002; Hampton, 2002), but is also surveyed in post-Golgi compartments. Our results provide direct evidence for the functional interplay between the ubiquitination machinery recognizing misfolded peripheral membrane proteins and the Ub-dependent endosomal sorting pathway in the elimination of misfolded CFTR, accounting for the cellular and clinical phenotype of the Δ70 CFTR mutation (Haardt et al., 1999). We propose that similar mechanisms may be involved in the recognition and degradation of other structurally destabilized membrane proteins that escape ER quality control or are generated by environmental stress. Thus, the peripheral quality control may have fundamental significance in the pathogenesis of conformational diseases and in the maintenance of cellular homeostasis.

BHK cells were stably transfected with wt, ΔF508, Δ70 CFTR, and the CFTR-Ub chimeras, harboring a single HA epitope at the NH₂ terminus (NHA; Haardt et al., 1999) or three tandem HA tags (3HA) in the 4th extracellular loop. After clonal selection in 500 μM methotrexate, 50–100 individual colonies were pooled and expanded for experiments. The HA tags preserved the biochemical and functional characteristics of CFTR (Haardt et al., 1999). Experiments have been performed on 3HA-tagged CFTR variants, if not indicated otherwise. Characterization of the 3HA-tagged CFTR will be described separately. PANC-1 cells (CRL1469; American Type Culture Collection) were stably transfected with the pMT/EP plasmid (provided by Dr. J. Ilan, Case Western Reserve University, Cleveland, OH) encoding the NHA-tagged wt or ΔF508 CFTR. Cells were seeded at confluence on plastic or on collagen-coated filters (Transwell-COL, 0.4-μm pore size; Costar) for iodide efflux and immunostaining. Differentiation was ensured by culturing the epithelia for more than three additional days. Expression of CFTR was induced by 50 μM ZnSO₄. CFTR-expressing ts20 and E36 cell lines were generated by retroviral infection (Benharouga et al., 2003). CFTR-Ub was constructed by fusing Ub in frame to the COOH terminus of CFTR using PCR (Shih et al., 2000). Mutant chimeras, incorporating UbF₄A,I₄₄A (CFTR-Ub2A), UbI₄₄A (CFTR-UbA), and the Lys-less Ub were generated by PCR mutagenesis. The cDNA of Lys-less Ub was provided by Dr. L. Hicke (Northwestern University, Evanston, IL). Similar expression levels of the CFTR chimeras were observed regardless whether CFTR harbored the NHA or the 3HA epitope.

Bacterial expression plasmids containing GST-Ub, GST-UbF₄A,I₄₄A (GST-Ub2A), and GST-UbF₄A (GST-UbA) were constructed by PCR mutagenesis in pGEX4T, and were expressed in HB101 cells. Recombinant proteins were purified, bound to glutathione-sepharose 4B (Amersham Biosciences), and incubated with HeLa cell lysate (3 mg, at 4°C for 2 h). Bound polypeptides were separated by SDS-PAGE and Hrs was visualized by immunoblotting with a polyclonal anti-Hrs antibody (Raiborg et al., 2002).

Nasal polyps were obtained from surgical materials of CF and non-CF individuals with informed consent of the family by a procedure approved by the Research Ethics Board of the Hospital for Sick Children. Two of the CF patients were homozygous ΔF508, and the third one had ΔF508/R1162X genotype. Explants were grown on collagen-coated dishes in DME/Ham's F12, 20% FBS, gentamycin, streptomycin, and amphotericin-B. Fibroblasts were removed by trypsinization, and epithelial cells were seeded at >80% confluence on collagen-coated filters and were cultured for >3 d. Polarization was demonstrated by the domain-specific CFTR-mediated anion conductance (Fig. S1 a).

Immunoblotting of CFTR and densitometric analysis was performed with NIH image 1.62 as described previously using ECL (Sharma et al., 2001). Antibodies were used as follows: FK1, FK2 (Affinity BioReagents, Inc.), and P4D1 (Santa Cruz Biotechnology, Inc.) anti-Ub antibodies; a6F anti-Na/K ATPase antibody (Developmental Studies Hybridoma Bank, University of Iowa, Iowa city, IA), M3A7 and L12B4 anti-CFTR antibody (Chemokine) and a rabbit polyclonal anti-CFTR antibody raised against the COOH-terminal tail of CFTR; 4A10 anti-TSG101 antibody (GeneTex); and anti-Hrs (Raiborg et al., 2002) and anti-E1 enzyme antibodies (Covance). Rabbit pAbs against human Vps25/EAP25 and human Vps32/CHMP4B were raised against maltose-binding protein fusion proteins and were affinity purified on Affi-Gel beads (Bio-Rad Laboratories) containing recombinant proteins. Whole-cell current measurements were performed as described previously (Haardt et al., 1999).

The cell surface density of 3HA-tagged CFTR was measured by the binding of the monoclonal anti-HA antibody (Covance; at 0°C for 1 h) and ¹²⁵I-labeled goat anti–mouse secondary antibody (3 μCi/ml, at 0°C for 1 h; Amersham Biosciences). Specific binding was calculated by correcting with the nonspecific antibody adsorption in the presence of 10 μg/ml nonimmune IgG (Santa Cruz Biotechnology, Inc.). Nonspecific antibody binding was usually 3–5% of the specific signal. Internalization of CFTR was calculated from the removal rate of anti-HA antibody from the cell surface during 2–5-min internalization at 37°C. Data are expressed as percentage of the specific radioactivity detected before internalization.

To measure CFTR recycling, first the endosomal CFTR was labeled with anti-HA antibody (at 37°C for 30 min). Then the remaining cell surface–associated antibody was blocked by biotinylated goat anti–mouse antibody (KPL) and 10 μg/ml streptavidin (Sigma-Aldrich) on ice. Recycling was initiated by shifting the temperature to 37°C for 5–10 min. The amount of exocytosed antibody–CFTR complex was measured by biotinylated secondary antibody and ¹²⁵I-streptavidin (Amersham Biosciences) at 0°C, and was expressed as the percentage of the anti-HA antibody associated with the endosomal compartment before exocytosis. The internalized anti-HA antibody in complex with CFTR was determined in parallel by monitoring the disappearance of cell surface anti-HA antibody with the biotin–streptavidin sandwich technique. The internalized anti-HA antibody was corrected for the degradation of mutant CFTR during the labeling. The radioactivity corresponding to internalized anti-HA antibody was ≈40,000 and ≈10,000–20,000 cpm for the wt and mutants, respectively. Each data point represents 3–4 independent experiments, consisting of triplicate measurements.

The pH of endocytic vesicles containing CFTR was measured by fluorescence ratio imaging of internalized anti-HA antibody (Covance) complexed with FITC-conjugated goat anti–mouse F_ab antibody (Jackson ImmunoResearch Laboratories). Cells were incubated with the primary and secondary antibodies in tissue culture medium at 37°C for 1–3 h, washed with NaKH medium (140 mM NaCl, 5 mM KCl, 20 mM Hepes, 10 mM glucose, 0.1 mM CaCl₂, and 1 mM MgCl₂, pH 7.3), and imaged on a microscope (Axiovert 100; Carl Zeiss MicroImaging, Inc.) at 35°C, equipped with a cooled CCD camera (Princeton Instruments) and a 63× NA 1.4 Planachromat objective. The metabolic stability of CFTR was not influenced by the continuous presence of anti-HA primary and secondary antibody (Fig. S2), indicating that antibody binding did not promote the degradation of CFTR.

Image acquisition and analysis were performed with the MetaFluor^® software (Universal Imaging Corp.). Images were acquired at 490 ± 5 nm, and 440 ± 10 nm excitation wavelength, using a 535 ± 25 nm emission filter. In situ calibration curves were obtained by clamping the vesicular pH between 4.5 and 7.0 in K⁺-rich medium (135 mM KCl, 10 mM NaCl, 20 mM Hepes or 20 mM MES, 1 mM MgCl₂, and 0.1 mM CaCl₂, containing 10 μM nigericin, 10 μM monensin, and 5 μM carbonyl cyanide p-chlorophenylhydrazon) and recording the fluorescence ratio of cells loaded with FITC-transferrin (Molecular Probes, Inc.), FITC-EGF (followed by a 2-h chase), or FITC-F_ab and anti-HA antibody in CFTR-Ub2A expressors. The fluorescence ratios as a function of extracellular pH provided the standard curve for the pH determination of CFTR-containing endosome. In addition, one point calibration was done on each coverslip by clamping the vesicular pH to 6.5. Mono- and multi-peak Gaussian fits for vesicular pH were performed with Origin 7.0 software (OriginLab^® Corporation). The average pH of each type of vesicle population was calculated as the arithmetic mean of the data and was identical to the Gaussian mean, based on single-peak distribution fitting, except for Δ70 CFTR, where two-peak Gaussian distribution analysis was done (Fig. S4).

The recycling (k_re) and degradation (k_deg) rate constants of CFTR were calculated by the SAAM II program (University of Washington, Seattle, WA), using a two-compartmental model (Fig. 8 c) and the Rosenbrock integration method. Data for fitting were obtained from the cell surface decay kinetics of wt and rΔF508 CFTR complexed to anti-HA antibody. The value of k_de was adjusted so that the model reproduced metabolic half-life of the complex-glycosylated rΔF508 and wt CFTR measured by the pulse-chase technique (Sharma et al., 2001). It was assumed that degradation of CFTR in the MVB/lysosomes is instantaneous. The rate constants are expressed as fractional transfer of molecules/min.

Experiments were repeated at least three times, and graphs include means ± SEM. Two-tailed P values were calculated at 95% confidence level with unpaired t test, using Prism software (GraphPad Software, Inc.).

Fig. S1 demonstrates that anti-HA antibody binding does not induce degradation of the wt CFTR. Fig. S2 illustrates the functional polarization of human respiratory epithelia. Elimination of the core-glycosylated CFTR during the CHX chase is documented in Fig. S3. Fig. S4 shows the pH distribution curves of vesicles containing CFTR variants. The electrophysiological characterization of the CFTR-Ub chimera is depicted in Fig. S5.

We are grateful to J. Gruenberg, L. Hicke, J. Ilan, and R. Haguenauer-Tsapis for providing anti-Lamp1 antibody and plasmid DNAs. We thank J. Rommens and R. Reithmeier for helpful comments, and L. Drzymala and M. Popov for the construction of CFTR-Ub chimeras.

M. Benharouga was supported by a Canadian Institutes of Health Research (CIHR) postdoctoral fellowship. J. So was awarded a summer studentship from the Canadian Cystic Fibrosis Foundation (CCFF). This work was supported by grants from CCFF and the CIHR to G.L. Lukacs. H. Stenmark was supported by the Research Council of Norway, the Novo Nordisk Foundation, and the Norwegian Cancer Society. M. Sharma is a CIHR doctoral student. F. Pampinella is a postdoctoral fellow of the CCFF.