Real-Time Fluorescence Detection of ERAD Substrate Retrotranslocation in a Mammalian In Vitro System

Summary Secretory proteins unable to assemble into their native states in the endoplasmic reticulum (ER) are transported back or “retrotranslocated” into the cytosol for ER-associated degradation (ERAD). To examine the roles of different components in ERAD, one fluorescence-labeled ERAD substrate was encapsulated with selected lumenal factors inside mammalian microsomes. After mixing microsomes with fluorescence-quenching agents and selected cytosolic proteins, the rate of substrate efflux was monitored continuously in real time by the decrease in fluorescence intensity as cytosolic quenchers contacted dye-labeled substrates. The retrotranslocation kinetics of nonglycosylated pro-α factor were not significantly altered by replacing all lumenal proteins with only protein disulfide isomerase or all cytosolic proteins with only PA700, the 19S regulatory particle of the 26S proteasome. Retrotranslocation was blocked by antibodies against a putative retrotranslocation channel protein, derlin-1, but not Sec61α. In addition, pro-α factor photocrosslinked derlin-1, but not Sec61α. Thus, derlin-1 appears to be involved in pro-α factor retrotranslocation.

the ∆gpαf-BOF molecules that were degraded in the cytosol; and the few ∆gpαf-BOF molecules that were adsorbed to the outer surface of the microsomes after gel filtration (<10%). Clearly, it is essential to identify, quantify, and thoroughly characterize the biochemical origin of any spectral changes before interpreting them (Johnson, 2005).
The fraction of ∆gpαf-BOF that is retrotranslocated from RRMs in our samples can be calculated using the above intensities for different species. Assuming that 10% of the ∆gpαf-BOF is adsorbed to the outer surface of the microsomes and that 10% of the remaining ∆gpαf-BOF (= 9%) is exposed to the cytosol by the background release of contents from microsomes ( Figure S2A), a total intensity decrease of 46% (Figure 2C) in cptRRM + cptcyto samples under our conditions would result from 36% of the originally encapsulated ∆gpαf-BOF being retrotranslocated from the lumen to the cytosol and then digested within 2000 sec. As noted in the text, this magnitude of retrotranslocation is similar to the fraction of ∆gpαf-BOF found in the cytosol after the microsomes have been sedimented ( Figure 2D).

Origin of Background Fluorescence Quenching
Is the slow, steady increase in αBOF-dependent quenching in the absence of cytosolic proteins or ATP ( Figure 2B) caused by encapsulated material leaking from the microsomes? RRMs were reconstituted with ATP, total lumenal proteins, and BOF-labeled glutathione to determine if the release of a small molecule occurs spontaneously from the RRMs over time. Other RRMs were prepared with ATP and either BOF-labeled BiP or PDI because these proteins are located in the ER lumen and should not be substrates for retrotranslocation. When incubated in cptcyto, the rates and extents of quenching, and hence BOF exposure to αBOF in the cytosol, were the same for both large (BiP-BOF, PDI-BOF) and small (glutathione-BOF) encapsulated molecules ( Figure S2A). Thus, any openings in the RRMs had to be large enough to release PDI and BiP at the same rate as glutathione. Yet we showed previously that microsomal membranes are impermeable to iodide ions for more than 4 hours at 4°C (Crowley et al., 1994). Thus, holes large enough to release PDI or BiP from microsomes are unlikely, and the slow increase in αBOF-dependent quenching in the absence of retrotranslocation apparently does not occur due to either glutathione-BOF or ∆gpαf-BOF leakage from RRMs.
Instead, this signal loss most likely results from a low constant rate of RRM breakage at 30°C that simultaneously exposes glutathione-BPF and the larger ∆gpαf-BOF, PDI-BOF, and BiP-BOF to αBOF. Whatever its origin, this αBOF-dependent emission intensity decrease appears to constitute a "background" signal change because it is observed under conditions in which retrotranslocation does not occur (compare Figure 2B "-cyto" and "-ATP" traces with Figure S2A). Moreover, after very long time periods (~50 min), the rate of αBOF-dependent ∆gpαf-BOF quenching was the same in the presence and absence of cytosolic proteins. Thus, only background quenching was observed at long times ( Figure S2B). The quenching due to retrotranslocation was complete within ~50 min under our conditions in the sample containing cytosolic proteins, and the net intensity reach a plateau by 50 min ( Figure 2C). We have therefore routinely subtracted the αBOF-dependent quenching observed with samples lacking cytosol to accurately portray the retrotranslocation-dependent fluorescence change.

Temperature Effects
CptRRMs were incubated in parallel with or without cytosolic proteins at different temperatures. Since the -cyto background quenching was temperature-dependent, the -cyto signal was subtracted from the +cyto signal to yield the net αBOF-dependent quenching at each temperature ( Figure S3). These data revealed that both the rate and extent of mammalian retrotranslocation are temperature dependent under our conditions. No retrotranslocation was detected at 4°C after 1 hr. Hence, no retrotranslocation occurs while our samples are on ice prior to raising their temperature at t 0 to initiate retrotranslocation. The net αBOF-dependent quenching (retrotranslocation) was essentially the same at 37°C and 30°C ( Figure S3). Thus, our experiments were done at 30°C because the extent of non-retrotranslocation release of ∆gpαf-BOF, apparently by microsomal rupture, was lower.

Photocrosslinking of Lumenal, Cytosolic, and Membrane Components
Microsomes containing photoreactive [ 35 S]∆gpαf were incubated in cptcyto at 30°C for 0, 15, or 30 min (Experimental Procedures; Figure 6). At each time point, the radioactive protein species in the total sample were detected using SDS-PAGE (Fig, 4A). Following sedimentation to separate the microsomal pellet from the soluble protein supernatant, the radioactive species in the cytosol (Figure 4B) and in the microsomes ( Figure 4C) were visualized by SDS-PAGE. Even prior to the 30°C initiation of retrotranslocation, many membrane-bound photoadducts with apparent molecular masses that exceed that of ∆gpαf are visible ( Figure 4C, lanes 1,4). The intensities of these bands are not greatly altered at later times ( Figure 4C, lanes 2, 3, 5, 6). In contrast, the extent of ∆gpαf photocrosslinking to cytosolic proteins increases significantly as retrotranslocation proceeds and the number of photoreactive ∆gpαf molecules transported into the cytosol increases ( Figure 4B, lanes 4-6). While likely photocrosslinking targets can be identified based on the apparent masses of the expected photoadducts, it is clear from the multiplicity of bands that the extent of ∆gpαf photocrosslinking to Derlin-1, Sec61α, TRAM, and other membrane, lumenal, and cytosolic proteins is best determined by immunoprecipitation.

Supplemental Experimental Procedures Proteins
Hemin-free rabbit reticulocyte lysate was prepared as described (Carlson et al., 2005), and hemoglobin was removed by passing the lysate through HisTrap HP resin (General Electric) in Buffer A [50 mM HEPES (pH 7.5), 40 mM KOAc, 5 mM MgCl 2 ] at 4°C, adding EDTA (pH 7.5) to 1 mM, and dialyzing overnight at 4°C against Buffer A + 4 mM reduced glutathione. Bovine 26S proteasomes and PA700 were purified as detailed elsewhere (DeMartino et al., 1994;Liu et al., 2006). p97, Npl4, and Ufd1 were purified as a complex (GND, in preparation). To inhibit proteolysis, 60 µM epoxomicin (Calbiochem) was incubated (0°C, 30 min) with 26S proteasomes or cyto. ATPase activity of PA700 was inactivated by incubation with Nethylmaleimide (NEM) as described previously (DeMartino et al., 1994). Human intein-tagged PDI was purified using a chitin column according to Novagen specifications, while hamster BiP with C-terminal His tags was purified as before (Alder et al., 2005). PDI and BiP were each then additionally purified on a Q-Sepharose (GE Healthcare) column using a linear salt gradient. BiP, PDI, and ∆gpαf are each stored in 50 mM HEPES (pH 8.0), 250 mM sucrose; the BiP solution also contained 1 mM ATP. Before use, PDI was usually incubated at 30°C for 30 min in Buffer A containing either 5 mM DTT or 5 mM G-S-S-G. αBOF was purchased from Invitrogen, affinitypurified αTRAM and αSec61α from Research Genetics (Huntsville, AL), and affinity-purified αDer1 from Novus Biologicals (Littleton, CO). Canine SRP and salt-washed ER microsomes were prepared as before (Flanagan et al., 2003;Walter and Blobel, 1983a;Walter and Blobel, 1983b).

∆gpαf-BOF
The following alterations were made in the wild-type S. cerevisiae pαf sequence: (i) N23Q, N57Q, and N67Q mutations prevented glycosylation; (ii) Y165C permitted the site-specific attachment of a single fluorescent dye; and (iii) a hexameric histidine tag at the C-terminus facilitated purification. No signal sequence was present because the protein was encapsulated into RRMs biochemically, not via SRP. The primary sequence of ∆gpαf was confirmed by DNA sequencing of the plasmid. The DNA encoding ∆gpαf was cloned into a heat-inducible expression vector (Bush et al., 1991) and ∆gpαf was overexpressed in E. coli BL21 (DE3) cells. After cell lysis, ∆gpαf was bound to an XK 16/10 Chelating Sepharose Fast Flow column (GE Healthcare) loaded with Co 2+ , equilibrated in Buffer C [8 M urea, 50 mM HEPES (pH 8.0), 100 mM NaCl], and eluted at 3 ml/min with 300 mM imidazole, 50 mM HEPES (pH 8.0).
The ∆gpαf solvent was changed to Buffer D [50 mM Hepes (pH 8.0), 2 mM EDTA, 1 mM dithiothreitol (DTT), 2 M urea] by gel filtration through Sephadex G-25 (30 cm x 2.5 cm i.d.). ∆gpαf was further purified by ion exchange chromatography on FPLC Q-Sepharose using a linear salt gradient (100-1000 mM NaCl in Buffer D without urea); ∆gpαf eluted in a single peak near 300 mM NaCl. After ∆gpαf was transferred into 20 mM Hepes (pH 8.0), 50 mM NaCl, and 2 mM EDTA by gel filtration to remove DTT prior to labeling, a 4-fold molar excess of 4,4difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-S-indacene-3-propionyl)-N'-iodoacetylethylenediamine (Invitrogen) dissolved in DMSO was added dropwise to ∆gpαf (10-20 µM; ~50 ml) at 4°C with stirring (final DMSO concentration ≤ 3% v/v). After at least 12 hr in the dark at 4°C,the reaction was quenched (30 min, 4°C) by the addition of 10 mM DTT. Unreacted dyes were removed by first loading a volume of Buffer C equal to the protein reaction volume on a Sephadex G-25 gel filtration column (25 cm x 2.5 cm i.d.) that had been equilibrated in 50 mM HEPES (8.0), 250 mM sucrose. After the Buffer C had fully entered the resin, the reaction mix was loaded on the column so that the protein was eluted through the urea-containing buffer C, thereby releasing any non-covalently bound dye before the protein was finally eluted in 50 mM HEPES (8.0), 250 mM sucrose. The resulting ∆gpαf-BOF (5-15 µM) was stored at -80°C in aliquots appropriate for reconstitution. No non-covalently bound BODIPY dyes were detected in ∆gpαf-BOF preparations after SDS-PAGE electrophoresis and fluorescence detection using the Bio-Rad FX imager.
Gel filtration was used throughout to exchange the solvent because: (i) diluting the protein reduced aggregation; (ii) the BOF labeling efficiency was higher with gel-filtered ∆gpαf than with dialyzed ∆gpαf; and (iii) the labeled protein aggregated during a long dialysis at high concentration. Protein aggregation was also a problem when the pH was lowered below 7.5.
When examined by analytical HPLC using a C1 column to separate unlabeled and BOFlabeled ∆gpαf proteins, 48-50% of the ∆gpαf was found to be labeled with BOF using the above conditions. Since purifying ∆gpαf-BOF from ∆gpαf proved difficult, the ∆gpαf-BOF used here was only about half-labeled.

BOF Labeling of BiP, PDI, and Glutathione
BiP and PDI were reacted with BOF using the same procedures used for labeling ∆gpαf and were purified by gel filtration from the unreacted dye. Reduced glutathione (1 mM) was reacted with 10 µM BOF reagent; unreacted glutathione was not separated from glutathione-BOF.