We present a novel method to experimentally visualize in vivo the topology of transmembrane proteins residing in the endoplasmic reticulum (ER) membrane or passing through the secretory pathway on their way to their final destination. This approach, so-called redox-based topology analysis (ReTA), is based on fusion of transmembrane proteins with redox-sensitive GFP (roGFP) and ratiometric imaging. The ratio images provide direct information on the orientation of roGFP relative to the membrane as the roGFP fluorescence alters with changes in the glutathione redox potential across the ER membrane. As proof of concept, we produced binary read-outs using oxidized roGFP inside the ER lumen and reduced roGFP on the cytosolic side of the membrane for both N- and C-terminal fusions of single and multi-spanning membrane proteins. Further, successive deletion of hydrophobic domains from the C-terminus of the K/HDEL receptor ERD2 resulted in alternating localization of roGFP and a topology model for AtERD2 with six transmembrane domains.
The secretory pathway is an essential system of functionally inter-connected organelles. Approximately one-third of all newly translated proteins in a typical eukaryotic cell are translocated into the endoplasmic reticulum (ER) to be delivered to their final destination along secretory pathway routes (Boyce and Yuan, 2006). A daunting task for molecular analyses of membrane proteins in the secretory pathway is to determine their topology with respect to the membranes in which they are embedded. Despite the general biological importance of transmembrane (TM) proteins, tools for the analysis of protein topology in vivo are limited. MPtopo, a curated database of TM proteins with experimentally validated transmembrane domains (TMDs) (http://blanco.biomol.uci.edu/mptopo/; Jayasinghe et al., 2001), currently contains only a total of 185 proteins with 1089 TMDs from all species. For other proteins, the most frequently used approach is the use of bioinformatics tools to predict the protein topology based on the known hydrophobicity of amino acids and experimental analysis of the primary protein sequence (Elofsson and von Heijne, 2007). However, the number, location and orientation of TMDs as predicted by bioinformatics algorithms are often highly contradictory (Elofsson and von Heijne, 2007). Many predicted topology models also differ with respect to the orientation of N- and C-terminal ends.
In yeast, the topology of TM proteins targeted to the secretory pathway can be investigated in vivo using fusions with essential metabolic enzymes as topology reporters and expression of these fusions in a mutant background (Kim et al., 2006). This approach, however, is not feasible for most other eukaryotic cell types because suitable mutants are often lacking. An advanced method to analyse the topology of TM proteins in vivo is the fluorescence protease protection assay (Lorenz et al., 2006). This assay relies on fusion of TM proteins with GFP and the fluorescence read-out in response to trypsin-induced destruction of the fused GFP before and after plasma membrane permeabilization (Lorenz et al., 2006). Drawbacks of the fluorescence protease protection assay are the requirement for selective plasma membrane permeabilization, and subsequent in-cell digestion of the GFP residue, which may result in progressive alteration of the cellular environment. The use of bi-molecular fluorescence complementation as a topology reporter avoids the need for external interference (Zamyatnin et al., 2006), but requires duplicate co-transformation of a membrane protein tagged with a non-fluorescent YFP fragment together with a second non-fluorescent YFP fragment targeted to either the cytosol or the ER lumen. The pH sensitivity of YFP has also been exploited as a topology reporter for plasma membrane proteins (Swarup et al., 2004). In this case, however, the reporter is fluorescent only on the cytosolic side of the membrane, and the assay cannot be used in the absence of a steep pH gradient across the membrane. To overcome the limitations of the various topology assays, we have devised a more straightforward redox-based topology assay (ReTA) that allows analysis of the topology of TM proteins in the secretory pathway in living, unperturbed cells. This method relies on the difference in the redox potential of glutathione (EGSH) across the ER membrane, which is specifically sensed by ratiometric redox-sensitive GFP (roGFP) (Gutscher et al., 2008; Hanson et al., 2004; Meyer et al., 2007). After expression of fusion proteins constructed using TM proteins and roGFP, the ratiometric fluorescence read-out provides direct information as to which side of the membrane the probe is localized.
The ER lumen is a more oxidizing environment than the cytosol
In the early secretory pathway, EGSH in the ER lumen is far more oxidizing than in the cytosol due to ratios of reduced glutathione (GSH) to glutathione disulfide (GSSG) of between 1:1 and 1:3 in the ER lumen compared with a ratio of approximately 100:1 or more in the cytosol (Cuozzo and Kaiser, 1999; Hwang et al., 1992). We thus reasoned that the difference in the degree of oxidation of the glutathione pool across the ER membrane is more than sufficient to produce significantly different fluorescence read-outs when roGFP is targeted to either the cytosol or the ER lumen (Figure 1a). To demonstrate this, roGFP was targeted to the ER using a chitinase signal peptide and the HDEL retention signal. Labelling of the ER resulted in a typical reticulate network, and no labelling of the cytosol or nucleoplasm was observed (Figure 1b) (Boevink et al., 1998). When targeted to the cytosol, the fluorescence ratio of roGFP was 0.26 ± 0.07, which is equivalent to an EGSH of approximately −320 mV (Figure 1b). The ratio for ER-targeted roGFP was 1.68 ± 0.39, which indicates an EGSH that is less negative than −240 mV (Figure 1b). The large dynamic range of roGFP thus allows a binary read-out for localization of the probe from the 405/488 nm fluorescence ratio values.
Method for rapid identification of transmembrane protein orientation
To test whether the physical properties of roGFP between the two compartments can be exploited to provide a quick and reliable read-out for the orientation of TM proteins in the ER membrane, roGFP was first fused to either the cytosolic N- or luminal C-terminus of AtSEC22, a vesicle-SNARE protein involved in anterograde protein trafficking at the ER–Golgi interface (Chatre et al., 2005). Expression of both fusion proteins in tobacco leaves resulted in strong labelling of the ER network (Figures 2b,c and 3). Ratiometric analysis of the observed fluorescence, however, revealed significant differences. Consistent with our hypothesis, a low 405/488 nm ratio for roGFP fused to the N-terminus of SEC22 indicated a reduced probe, and thus localization to the cytosolic side of the ER membrane (Figure 2b). In contrast, fusion of roGFP to the C-terminus of SEC22 resulted in a high 405/488 nm fluorescence ratio, indicative of an oxidized roGFP tag localized to the luminal side of the membrane (Figure 2c). In addition to the reticulate ER structure, SEC22:roGFP was also distributed on Golgi bodies (Figure 3), consistent with previous observations of punctate structures using SEC22:YFP fusions that were identified as Golgi stacks (Chatre et al., 2005). Incubation with brefeldin A (BFA) as an inhibitor of secretion (Boevink et al., 1998; Nebenführ et al., 2002) caused redistribution of fusion proteins to the ER membrane (Figure 3). The fluorescence ratio of the roGFP moiety was not affected by BFA incubation.
The results were independently corroborated using a biochemical protease protection assay on microsomes isolated from tobacco leaves that had been transiently transfected with plasmids encoding N- or C-terminally tagged SEC22. Digestion of protein residues exposed to the medium confirmed our microscopy results indicating that the N-terminus was localized on the cytosolic side and the C-terminus was localized in the lumen (Figure 4).
roGFP was further validated as a suitable reporter for determination of the orientation of membrane proteins by fusion to TMcCCASP, a synthetic type I protein (Hanton et al., 2005). Consistent with previous observations based on protease protection assays (Hanton et al., 2005), use of roGFP directly reported the N-terminus of this protein being localized in the ER lumen (Figure S1). These data support the feasibility of use of ratiometric analysis for determination of the orientation of TM proteins in vivo. To test the applicability of ReTA for experimental analysis of multi-spanning TM proteins, we tagged AtRER1B, a highly conserved and well characterized protein that is required for retrieval of various ER membrane proteins (Sato et al., 2003). Consistent with previous observations (Sato et al., 2003), both termini are reported by ReTA to reside on the cytosolic side of the ER and Golgi membrane (Figure 2d–g).
Topology analysis of AtERD2, a multi-spanning transmembrane protein
After proof of concept, we next tested the applicability of ReTA for detailed analysis of the more complex K/HDEL receptor homologue ERD2, which is suspected to be responsible for retrograde Golgi–ER transport of soluble ER resident proteins (Lee et al., 1993). Despite knowledge of the protein function and identification of seven hydrophobic domains in the primary sequence (Figure 6a), the topology of the protein is still unknown. Various algorithms used by the plant membrane protein database ARAMEMNON (http://aramemnon.botanik.uni-koeln.de/; Schwacke et al., 2003) predict between none and seven TMDs for AtERD2 (Figure 5a). A six-TMD topology was suggested for the highly homologous mammalian KDEL receptor based on fusion of N-terminal fragments with an N-glycosylation reporter (Singh et al., 1993). On the other hand, a seven-TMD topology was proposed for human ERD2.1 and the K/HDEL receptor of Saccharomyces cerevisiae based on site-directed mutagenesis studies (Buck and Skach, 2005; Townsley et al., 1994). In light of this apparent contradiction, we decided to use the six-TMD topology model predicted by the TmPred algorithm (Figure 5b) as a template, and tagged AtERD2 at both termini using roGFP. Expression of these roGFP fusion proteins in tobacco resulted in low fluorescence ratios, indicating that both the N- and C-termini of AtERD2 are located on the cytosolic side of the ER membrane (Figure 6b). We also constructed truncated versions of AtERD2 with successive deletion of all TMDs predicted by TmPred from the C-terminus, and fused roGFP to the C-termini of these truncated sequences. Ratiometric imaging clearly showed alternating ratio values, indicating localization of roGFP within the ER lumen when roGFP was fused behind TMD1, TMD3 and TMD5. Conversely, fusions behind TMD2 and TMD4 resulted in a cytosolic localization of roGFP (Figure 6b). The observation that fusions of roGFP behind TMD1 and TMD2 were oriented towards opposite sites of the ER membrane strongly suggests that no extra TMD exists between amino acids 25 and 60 as proposed previously for the human homologue (Scheel and Pelham, 1998). Instead, ReTA supports a topology model with six TMDs and both termini on the cytosolic side of the membrane.
The experimental evidence presented in this work shows that we have developed a novel method for reliable and rapid analysis of the topology of TM proteins based on ratiometric imaging of roGFP tags fused to these proteins. This method appears to be particularly reliable for membrane proteins in the secretory pathway. ReTA presents several key advantages over previously established techniques. First, ReTA is simple in that it only requires the construction of fusion proteins with roGFP of either full-length proteins or N-terminal fragments thereof and transient expression of the fusion protein. Second, ratiometric analysis is easy to perform, and, due to the large dynamic range of roGFP, does not require sophisticated digital image processing for quantitative analysis of fluorescence images. Direct ratiometric analysis instead provides a binary read-out of the protein orientation. Third, the in vivo approach overcomes possible limitations of other imaging-based assays such as the fluorescence protease protection assay (Lorenz et al., 2006), which depends on selective permeabilization of membranes, or assays that produce fluorescence on one side of the membrane only (Swarup et al., 2004; Zamyatnin et al., 2006). The roGFP probe used in the ReTA assay is directly responsive to EGSH in the respective subcellular compartments (Meyer et al., 2007). Thus, the prerequisites for using ReTA are (i) sufficient expression levels for detection at both wavelengths, (ii) that the fluorescent tag does not obscure localization and stability of the membrane protein, and (iii) a sufficiently steep gradient in EGSH across a membrane. In some cases, it might thus be necessary to generate a number of different fusion proteins in which the point of fusion of roGFP is changed by only a few amino acids. If the stability and topological maturation of TM proteins is affected by truncation (Buck and Skach, 2005), insertion of roGFP into specific extra-membrane loops is straightforward. In principle, dual-excitation ratiometric analysis of roGFP can also be performed using epifluorescence microscopes equipped with a fast filter changer for the excitation light (Dooley et al., 2004). However, particularly in plant cells, fluorescence from membrane-localized roGFP fusions in this case may be severely obscured by background autofluorescence. The demand for an EGSH gradient is clearly fulfilled for the ER membrane, and it may be assumed that the same type of gradient exists across membranes of the secretory pathway downstream of the ER. In addition, the EGSH gradient of the ER membrane may also be exploited for proteins destined for distal compartments. This may be achieved by treating cells with BFA, which, by inhibiting ER–Golgi transport, induces ER retention and does not affect roGFP ratiometric measurements. In systems that are insensitive to BFA (Richter et al., 2007; Teh and Moore, 2007), it may be possible to co-express the SAR1 guanine exchange factor SEC12. Increased levels of SEC12 can block the secretory pathway through inhibition of COPII-mediated ER export (da Silva et al., 2004). Because of the difference in EGSH across the ER membrane, ReTA is immediately applicable to several thousand proteins that pass through the secretory pathway. The EGSH gradient across the ER membrane is a general feature of all eukaryotic cells, and the opposite response of roGFP between cytosol and ER lumen has also been shown for human nasal epithelial cells (Schwarzer et al., 2007). We therefore consider ReTA to be broadly applicable to any eukaryotic species of interest and the most simple-to-use method currently available for the topological analysis of TM proteins in the secretory pathway.
Plant growth and transient transfection
Tobacco plants (Nicotiana tabacum L. cv Samsun or Nicotiana benthamiana Domin.) were grown in the greenhouse under controlled conditions. Transformation of tobacco leaf epidermal cells was performed as described previously (Sparkes et al., 2006) using Agrobacterium tumefaciens strain AGL-1 (Lazo et al., 1991) containing binary vectors with roGFP or roGFP fusion proteins. Transfected cells were imaged by confocal laser scanning microscopy 2–4 days after inoculation.
Targeting of roGFP to the ER lumen
For ER targeting, roGFP was PCR-amplified using the primers 5′-ACCATGGTGAGCAAGGGCGAGGAG-3′ (forward; NcoI site underlined) and 5′-TCTAGACTTGTACAGCTCGTCCAT-3′ (reverse; XbaI site underlined) to introduce restriction sites and remove start and stop codons. The amplified product was blunt end-ligated into pCAPS (Roche, http://www.roche.de), then further cloned as an NcoI/XbaI fragment into pWEN81 between a chitinase signal peptide and the HDEL retrieval signal. Subsequently, the entire cassette consisting of roGFP with the N-terminal chitinase targeting signal and the C-terminal HDEL motif was sub-cloned into the binary vector pWEN22 under the control of a strong constitutive CaMV 35S promoter using XhoI and SacI.
Construction of roGFP fusions with transmembrane proteins
To obtain N- and C-terminal fusions of full-length SEC22 with roGFP2, fusions were generated by splicing the roGFP2 cDNA sequence in-frame upstream or downstream of SEC22 using suitable restriction sites of the multiple cloning site of the binary vector pVKH18En6 (Batoko et al., 2000) as described previously for SEC22:YFP (Chatre et al., 2005).
The roGFP:TMcCCASP fusion protein was generated by replacing the GFP at the N-terminal end of TMcCCASP (Hanton et al., 2005) using SalI and BamHI.
Fusion of roGFP to the N-terminus of AtRER1B was achieved through fusion PCR using primers P1, P2 and P3 (see Table S1 for a list of all primers P1–P21). In the first round, roGFP was amplified using primers P1 and P2, the latter of which also partially overlaps with RER1B. The resulting PCR product was mixed with RER1B template and amplified using primers P1 and P3. The resulting PCR product was cloned into pBinCM using BamHI and SalI. For C-terminal fusions, RER1B was PCR-amplified using primers P4 and P5 and cloned into pVKH18En6-SEC22:roGFP using BamHI and SalI, thereby replacing SEC22.
A C-terminal fusion of ERD2 with roGFP was constructed by assembly PCR. In a first PCR reaction, full-length ERD2 was amplified using primers P6 and P7 (Table S1) to introduce a KpnI site at the 5′ end and a 3′ overhang complementary to the 5′ end of roGFP. Simultaneously, roGFP was amplified using primers P8 and P9 to generate a fragment lacking the start codon and including a BamHI restriction site at the 3′ end. In a second PCR, both products were combined in a 1:1 ratio and amplified using primers P6 and P9. The resulting fusion construct was subcloned into pCAPS, digested using KpnI and BamHI, and finally ligated into the binary vector pBinAR (Höfgen and Willmitzer, 1990). For N-terminal fusion of ERD2 full-length protein, roGFP was amplified using primers P10 and P11 to generate a KpnI site at the 5′ end and a 3′ overhang complementary to the N-terminus of the ERD2 sequence. Concurrently, ERD2 was amplified using primers P12 and P13 to produce a fragment with a 3′XbaI site. In a second PCR, the products were combined and amplified using primers P5 and P8. The resulting PCR product was subcloned into pCAPS, cut out using KpnI and XbaI, and ligated into pBinAR.
For analysis of N-terminal fragments of ERD2 containing only two, three or four predicted N-terminal TMDs, the fragments were cloned as XbaI/SalI fragments in front of roGFP by replacing SEC22 in the plasmid pVKH18En6-SEC22:roGFP used previously. The fragment TMD2 was amplified using primers P14 and P15, and the fragment TMD3 was amplified using primers P14 and P16. The fragment TMD4 was cut out using XbaI and SalI from vector pVKH18En6-ERD2:GFP, exploiting an endogenous SalI restriction site, and then inserted in front of roGFP. roGFP fusions of truncated ERD2 sequences with only the first TMD (TMD1) or the first five predicted TMDs from the N-terminus were generated in a Gateway-compatible system (Invitrogen, http://www.invitrogen.com). To construct a suitable destination vector, roGFP was amplified by PCR using primers P17 and P18, and, after restriction with AvrII and PacI, was ligated into the vector pEarleyGate100 (Earley et al., 2006). The new destination vector pSS01 contained roGFP in-frame behind the Gateway cassette. The remaining N-terminal ERD2 fragments were amplified using Gateway-compatible primers P19 and P20 for TMD1 and P19 and P21 for TMD5. The resulting fragments were purified and mixed with pDONR201 for the BP reaction. Positive clones were linearized using PacII, and recombined in the LR reaction with the destination vector pSS01.
Ratiometric fluorescence imaging
All experiments were performed using roGFP2 (Hanson et al., 2004), because this offers the largest dynamic range at the fixed excitation wavelengths used on a confocal microscope. Ratiometric confocal imaging was performed on a Zeiss LSM510 META laser scanning microscope (Zeiss, http://www.zeiss.com) with either a Zeiss 63 × C-Apochromat 1.2 NA water immersion lens or a Zeiss 40 × C-Apochromat 1.2 NA water immersion lens. roGFP fluorescence was excited at 405 and 488 nm in multi-tracking mode with line switching. For improvement of the signal-to-noise ratio, all images were collected using a mean of 4. roGFP fluorescence was collected using a band-pass filter at 505–530 nm. All samples were additionally labelled using 50 μm propidium iodide for cell-wall staining and to prove the viability of the cells (Meyer et al., 2007). The propidium iodide signal was collected after excitation at 543 nm and emission filter set to 560–615 nm. Images were collected either as single plane images or as z-stacks with optimal resolution along the z axis. Correct ratiometric analysis of roGFP2 is mainly limited by the low excitation efficiency of the reduced form at 405 nm. In order to minimize the contribution of noise to the final ratio image, only cells in which the fluorescence excited at 405 nm was at least five times higher than the background were used for further analysis. Comparisons with recombinant protein in buffer imaged under identical instrument settings indicate that a local concentration of membrane-bound roGFP of approximately 1 μm was necessary to obtain a binary read-out that was not obscured by background noise. Ratiometric analysis of non-saturated images was performed using a custom-written MATLAB script (MathWorks Inc., http://www.mathworks.com/). The MATLAB script is available on request from A.J. Meyer. Briefly, after background subtraction, the 405 nm image was divided by the 488 nm image. A Gaussian fit was applied to the ratio value distribution of the calculated ratio image in order to determine the mean ratio value. Fluorescence values varied slightly with the use of different lenses for collecting the raw images, but the general change in redox-dependent fluorescence was not affected. The ratio images were illustrated using a self-created colour map. To achieve better contrast between background and blue pixels, the background was converted from black to white. Fluorescence ratio values were converted to redox potentials as described previously (Meyer et al., 2007). Due to its midpoint potential of −280 mV (Dooley et al., 2004), roGFP2 is almost completely oxidized in the ER lumen and almost completely reduced in the cytosol (Meyer et al., 2007; Schwarzländer et al., 2008). Ratio values measured using free roGFP2 in these two compartments thus set the upper and lower limit of the dynamic range.
Biochemical protease protection assay
For biochemical protease protection assays, tobacco leaves were used 3 days after transfection. Transfected leaf areas excluding the major veins were cut out, homogenized in extraction buffer (10 mm KCl, 1 mm MgCl2, 0.4 mm sucrose, 0.4% polyvinylpyrrolidone (average mol wt 40 000; PVP40), 40 mm HEPES-KOH, pH 7.5), filtered, and centrifuged at 2000 g for 5 min. The resulting supernatant was centrifuged at 20 000 g for 20 min, and the resulting supernatant was centrifuged at 100 000 g for 60 min. The resulting total membrane pellet was re-suspended in homogenizing buffer. For the protease protection assay, total membranes were incubated at a final concentration of 1 mg ml−1 in the presence of 1 mm CaCl2, and in the presence or absence of 300 mU proteinase K (Sigma-Aldrich, http://www.sigmaaldrich.com) per mg of total membrane protein, and in the absence or presence of 1% Triton X-100 for 30 min at 32°C. The reaction was terminated by addition of 5 mm EGTA, 5 mm PMSF, and subsequent CHCl3/methanol precipitation. The samples were analysed using protein-gel blots. Precipitated proteins were dissolved in SDS loading buffer, subjected to 11% SDS–PAGE, and transferred onto nitrocellulose membrane by electroblotting. After blocking in TBS, 0.05% Tween, 3% milk powder and 1% BSA, the membranes were then incubated overnight in TBS, 3% BSA with primary anti-GFP antibody (ab290, dilution 1:3000; Abcam, http://www.abcam.com/). Secondary antibodies were coupled to horseradish peroxidase (Sigma-Aldrich). Visualization of the bound antibodies was achieved using an enhanced chemiluminescence kit (Pierce, http://www.piercenet.com/).
We thank J. Remington (University of Oregon, OR, USA) for providing the roGFP sequence, B. Kost (University of Heidelberg, Germany) for providing plasmids pWEN22 and pWEN81, C. Hawes (Oxford Brookes University, Oxford, UK) for plasmid pVKH18En6:ERD2, and A. Nakano (Molecular Membrane Biology Laboratory, RIKEN Discovery Research Institute, Wako, Japan) for the RER1B cDNA. We also thank W. Khan for constructing the N-terminal fusion of roGFP:SEC22 and A. Speiser for help with generating roGFP:TMcCCASP. Critical reading of a manuscript draft by M. Fricker (Department of Plant Sciences, University of Oxford) and D. Robinson is gratefully acknowledged. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG) to A.J.M. (ME1367/3-2).