GPR107, a G-protein-coupled Receptor Essential for Intoxication by Pseudomonas aeruginosa Exotoxin A, Localizes to the Golgi and Is Cleaved by Furin*♦

Background: Bacterial toxins, including P. aeruginosa exotoxin A (PE), are valuable tools to dissect biological processes. Results: A genome-wide genetic screen identifies several novel host factors used by PE, including GPR107. Conclusion: Bacterial toxins can help identify novel host components involved in key intracellular trafficking steps. Significance: GPR107 may be a receptor that associates with G-proteins at the Golgi to regulate membrane transport. A number of toxins, including exotoxin A (PE) of Pseudomonas aeruginosa, kill cells by inhibiting protein synthesis. PE kills by ADP-ribosylation of the translation elongation factor 2, but many of the host factors required for entry, membrane translocation, and intracellular transport remain to be elucidated. A genome-wide genetic screen in human KBM7 cells was performed to uncover host factors used by PE, several of which were confirmed by CRISPR/Cas9-gene editing in a different cell type. Several proteins not previously implicated in the PE intoxication pathway were identified, including GPR107, an orphan G-protein-coupled receptor. GPR107 localizes to the trans-Golgi network and is essential for retrograde transport. It is cleaved by the endoprotease furin, and a disulfide bond connects the two cleaved fragments. Compromising this association affects the function of GPR107. The N-terminal region of GPR107 is critical for its biological function. GPR107 might be one of the long-sought receptors that associates with G-proteins to regulate intracellular vesicular transport.

general screening approach were done as described previously (23). Briefly, about 100 million mutagenized KBM7 cells were exposed to 50 ng/ml PE for 10 days. The survivors were pooled and expanded for a few days. Genomic DNA was isolated, and inverse PCR was performed using primer sequences flanking the retroviral insertion sites followed by Illumina sequencing. The statistical significance of insertions at a given gene in the PE-treated population was calculated by comparing the number of inactivating insertions to those in the untreated control data set. To isolate the GPR107 GT clone, cells were FACSsorted in 96-well dishes and grown until confluent. Genomic DNA from the individual clones was extracted using a genomic DNA isolation kit (Qiagen). Genomic insertions were identified by inverse PCR using a forward primer located within the genetrap (5Ј-CTCGGTGGAACCTCCAAAT-3Ј) and a reverse primer designed to target the GPR107 gene. The gene-trap insertions were mapped by sequencing of the PCR product using the forward primer. RT-PCR analysis was performed to determine the absence of the GPR107 transcript in the isolated mutant cell line using SuperScript TM III first-strand synthesis kit (Invitrogen). The following primers were used: 5Ј-ATGGC-CGCTCTGGCGCCCGTCGGCT-3 and (5Ј-GGCCTTCTTG-GTCATCAGTGC-3Ј). As a positive control, RT-PCR analysis of the GAPDH gene was performed.
Cell Culture and Virus Transduction-KBM7 and HeLa cells were grown in Iscove's modified Dulbecco's medium or DMEM supplemented with 10% heat-inactivated fetal serum, respectively, at 37°C and 5% CO 2 . Cell lines stably overexpressing various versions of GPR107 constructs were generated by infecting with retroviruses expressing the corresponding cDNAs and were selected for G418 (0.8 mg/ml for HeLa and 1.2 mg/ml for GPR107 GT cells). Of the three reported splice variants of GPR107 (24), we detected only the expression of isoform 2 (UniProt accession number Q5VW38-2).
Designing CRISPR Target Sequence and Prediction of Off-target Effects-Target sequences for CRISPR interference were designed as detailed in Ref. 25. The target sequence preceding the PAM motif was obtained from the region of the exon of the indicated genes (Table 1). Potential off-target effects of the target sequence were confirmed using the NCBI Homo sapiens Nucleotide BLAST.
Genotyping of the CRISPR/Cas9-generated Mutant Cell Populations Using Surveyor Assay-Surveyor assay was performed as described previously (26,27). Genomic DNA from treated and control crude HeLa cells was extracted. PCR was performed using specific primers (Table 2) under the following conditions: 94°C for 2 min; 35ϫ (98°C for 10 s, 60°C for 30 s, and 68°C for 30 s); 68°C for 2 min; hold at 4°C. PCR product was loaded onto an ethidium bromide-stained agarose gel (3%) and purified. 500 ng of the purified PCR product were treated with Surveyor nuclease (Transgenomic) for 30 min and resolved using 3% agarose gel.
Pulse-Chase Experiments, Furin, and Glycosidase Digestion-Pulse-chase experiments were performed as described previously (28). Briefly, HeLa cells stably expressing C-terminally HA-tagged GPR107 WT , GPR107 R182A , GPR107 C109A/C228A , or GPR107 ⌬40 -182 were grown in 10-cm culture dishes. For a given time point, 1 ϫ 10-cm dish was used in all experiments. Cells were starved in methionine-and cysteine-free DMEM for 45 min, pulse-labeled with [ 35 S]methionine/cysteine at 0.77 mCi/ml for 30 min, and chased for different time points in complete media containing 1 mM cold methionine/cysteine. Where indicated, cells were treated with 100 g/ml brefeldin A (BFA) or 100 nM concanamycin A for 1 h prior to labeling and chased in the continuous presence of the drugs. At different time points during the chase, cells were harvested, lysed in Tris buffer (150 mM NaCl, 5 mM MgCl 2 , 25 mM Tris-HCl, pH 7.4) containing 0.5% Nonidet P-40 followed by immunoprecipitation with anti-HA-coupled beads. Typically, the immunoprecipitates were eluted with 100 l of PBS containing 1% SDS. Another 900 l of Tris buffer was added to get a 0.1% final concentration of SDS. Samples were then re-immunoprecipitated using anti-HA-coupled beads. Where indicated, immunoprecipitates were subjected to Endo H, PNGase F, or furin digestion according to the manufacturer's instructions (New England Biolabs). Immunoprecipitates were eluted with SDS sample buffer and resolved by SDS/-PAGE. Samples were visu-alized with autoradiography using DMSO/2,5-diphenyloxazole and exposed to Kodak XAR-5 film.
Microscopy-HeLa cells grown on coverslips were fixed with 4% paraformaldehyde in PBS for 30 min. Fixation was stopped by incubating the coverslips with 50 mM NH 4 Cl for 10 min. Samples were then blocked with binding buffer (0.1% saponin and 0.2% BSA in PBS) for 30 min and incubated with the first antibodies followed by secondary antibodies conjugated with Alexa Fluor. Images were captured using a confocal microscope with a 63ϫ 1.40 N.A. of the Carl Zeiss Plan Apo oil objective and processed using Velocity and Adobe Photoshop software.
Generation of GPR107-depleted HeLa Cells-Lentiviral plasmids containing the shRNA against human GPR107 and the control (shGFP) were purchased from Open Biosystems. To generate lentiviruses, low passage HEK293T cells were transfected with these plasmids using Lipofectamine 2000 following the manufacturer's protocol. GPR107-knockdown HeLa cell lines were generated by infecting the cells with the lentiviruses and then selecting them in the presence of puromycin (1 g/ml) for 1 week. In parallel, we transduced HeLa cells that overexpress HA-tagged GPR107 with the same viruses. Because there are no good antibodies, we determined the efficiency of the knockdown by Western blot analysis using anti-HA HRP.
Cell Viability Assay and Flow Cytometry-About 25 ϫ 10 3 cells per well were seeded in 96-well dishes with clear flat bottoms (Costar) and treated with different concentrations of PE for 18 h in the case of HeLa cells or 48 h for KBM7 cells. Cell viability assay was performed using the CellTiter-Glo luminescent cell viability assay kit (Promega) according to the manufacturer's protocol. This method determines cell viability based on quantitation of the ATP present, an indicator of metabolically active cells. ATP-based bioluminescence levels were measured using an EnVision plate reader (PerkinElmer Life Sciences).
For flow cytometry, cells were treated for 18 h with CDTs of different bacterial origin. The cells were harvested, washed with cold PBS, fixed with ethanol, stained with propidium iodide, and subjected to cytofluorometry (FACSCalibur, BD Biosciences). FlowJo was used to analyze the data. Intensity of fluorescence was measured, and the percent maximum was presented in the overlaid histograms.
Cholera Toxin Glycosylation Assay-Expression, purification, and generation of cholera toxin conjugated with the glycosylation motif (GGG-K(biotin)-NNSTG) was done as described previously (21). For the glycosylation assay, a similar number (about 20 ϫ 10 6 ) of KBM7 WT , GPR107 GT , and GPR107 GT ϩ GPR107 cells were washed once in PBS and resus- pended in 300 l of Opti-MEM containing a final concentration of 20 nM of CTx-GGGK(biotin)-NNSTG. Cells were incubated for 30 min at room temperature, and the medium containing the toxin was removed after a gentle centrifugation. The cells were resuspended in 3 ml of Iscove's modified Dulbecco's medium supplemented with 10% heat-inactivated fetal serum and chased for different time points at 37°C and 5% CO 2 . At the indicated time points during the chase, cells were collected and lysed in Tris buffer (150 mM NaCl, 5 mM MgCl 2 , 25 mM Tris-HCl, pH 7.4) containing 0.5% Nonidet P-40. Because the CTx also carries biotin probe, we recovered it from the lysate using NeutrAvidin beads. The toxin was eluted from the beads by boiling for 5 min in 1% SDS sample buffer. The glycosylation of CTx was detected by Western blotting using streptavidin-HRP and ECL substrate. The film was scanned, and the bands were quantified using ImageQuant software (GE Healthcare).

Haploid Genetic Screen Identifies Host Factors
Required for PE Intoxication-To identify host proteins important for PE intoxication, we performed a genetic screen that exploits KBM7 cells, haploid except for chromosome 8 (23,29). We mutagenized KBM7 cells using a gene-trap vector to obtain a collection of ϳ100 ϫ 10 6 mutants. We then intoxicated these mutant cells with 50 ng/ml PE, a concentration that kills the majority of wild type KBM7 cells. To identify genes that render cells resistant to PE, massively parallel sequence analysis was performed on genomic DNA isolated from the pooled clones. Gene-trap insertions in genes identified in the selected cell population were compared with gene-trap insertion events in a nonselected control population, and genes significantly enriched for mutations were thus identified. Among these, nine genes had already been implicated in PE intoxication, whereas the other eight genes were not previously known to be involved in the PE pathway ( Fig. 1, A and B).
Validation of the Haploid Genetic Screen Using CRISPR/ Cas9-gene Editing System-Because the screen was performed in the haploid KBM7 cell line, we extended these observations to another cell line to exclude cell type-specific effects. We performed PE intoxication assays on mutant HeLa cell populations, generated using CRISPR/Cas9 genome editing (25), for eight hits identified in the KBM7 screen. The introduction of mutations in the gene of interest was assessed by Surveyor assay that relies on digestion of amplified genomic DNA by an endonuclease that specifically cleaves at the site of the newly generated substitution. Accordingly, the region that flanks the sequence targeted by the sgRNA was amplified by PCR. The amplicons were then subjected to digestion and monitored by agarose electrophoresis. Unique cleavage products were observed in the mutant cell lines but not in the wild type control (Fig. 1D) showing the enrichment of gene knock-outs in these cell lines. As expected, these mutant cell populations showed increased resistance to PE as compared with the wild type control (Fig.  1E). Genes identified in the KBM7 screen as hits were thus confirmed by an independent genome editing approach in a different cell type.
In the following sections we describe the categories of mutations observed in our haploid genetic screen performed in KBM7 cells.
Mutations in the Diphthamide Biosynthetic Pathway Leads to PE Resistance-PE transfers an ADP-ribose group from NAD ϩ to diphthamide, a conserved, post-translationally modified histidine residue unique to eEF2 resulting in inhibition of protein synthesis (30). In eukaryotic cells, several genes are known to be involved in the biosynthesis of diphthamide (31). We observed a significant enrichment of mutations in DPH1, DPH2, DPH4, and DPH7 (WDR85) (Fig. 1A), indicating good coverage of the screen.
PE Trafficking in the Endocytic Pathway-PE enters cells via receptor-mediated endocytosis and reaches its final destination, the cytosol, via vesicular transport and translocation across an intracellular membrane (Fig. 8D). Prior to cytosolic delivery of its catalytic domain, PE undergoes proteolytic cleavage by furin, a protease localized in the TGN and in late endocytic compartments (32). The identification of furin in our screen is therefore consistent with the known requirement of PE for a proteolytic conversion. Moreover, LRP1, one of the known receptors for PE, undergoes a furin-mediated cleavage to generate the active form of the receptor (33). As we will show below, furin acts on yet other components essential for PE intoxication. The closely related proteins LRP1 and LRP1B are the known receptor(s) for PE (7,8). However, these proteins were not found in our screen. This might be because of the redundancy of their role as a receptor of PE or these proteins are essential for the survival of KBM7 cells.
A heterotetrameric tethering factor named the Golgi-associated retrograde protein (GARP) complex that comprises vacuolar protein sorting 51 (VPS51), VPS52, VPS53, and VPS54, promotes fusion of endosome-derived retrograde transport carriers with the TGN (34). In our screen, we identified VPS52, VPS53, and VPS54 of the GARP complex. HeLa cells lacking VPS53, resulting from CRISPR/Cas9-mediated inactivation of the corresponding gene, acquire resistance to PE (Fig. 1E), demonstrating that the GARP complex is essential for PE transport from endosomes to the TGN.
Trafficking of membrane proteins is also involves adaptor protein complexes (35). AP-1 mediates protein sorting between the TGN and early endosomes and is composed of four subunits, including AP1M1 (36). We identified AP1M1 as a novel host factor essential for PE intoxication (Fig. 1A); HeLa cells with mutations in AP1M1 become resistant to PE (Fig. 1E).
Retrograde Transport of PE from Golgi-to-ER-To modify eEF2, PE must cross a membrane barrier(s), but it is unclear whether PE travels through the Golgi to reach the ER as a point of escape. To address this, we performed microscopy studies in HeLa cells using PE conjugated with Atto647 (Fig. 2, A and B). We found that PE partially co-localizes with GPR107 (Fig. 2D), a protein that is localized at the TGN (see below). In our screen, we also implicated proteins involved in retrograde Golgi-to-ER transport, including the KDEL receptor 1 (KDELR1), which retrieves proteins from the Golgi for delivery to the ER through recognition of a C-terminal KDEL motif (Fig. 1A). To test whether PE reaches the ER, we installed a 3ϫGly-Lys-(biotin or Atto647) extended with the RDEL sequence or 3ϫGly-Lys-bi-otin without the RDEL sequence at the C terminus of PE via a sortase-mediated transacylation reaction (Fig. 2, A and B) (19,20). Both biotinylated and fluorescently labeled PE that carry the RDEL motif are toxic, whereas PE lacking the RDEL motif is not (Fig. 2C). The RDEL sequence immediately adjacent to the C terminus is thus required for PE intoxication. This is consis-   Novel genes identified in our screen are marked in red. C, schematic representation of the DNA amplicon used for genotyping of mutants generated using CRISPR/Cas9-gene editing strategy. D, surveyor assay for Cas9-mediated cleavage of the indicated genes in HeLa cells. Accordingly, genomic DNA was extracted, and the region that flanks the sequence targeted by the sgRNA was amplified by PCR.
The amplicons were then subjected to endonuclease digestion that specifically cleaves at the site of the newly generated substitution and monitored by agarose electrophoresis. Note that unique cleavage products were observed in the mutant cell lines but not in the control. E, PE intoxication assay of mutant HeLa cells generated using CRISPR/Cas9-gene editing system. Cell viability was determined and shown as percent relative to nonintoxicated cells. Error bars represent S.D. of three independent experiments performed in duplicate.
tent with a previous report suggesting that the KDEL receptor plays a role in the retrieval of PE to the ER (37). In addition to KDELR1, we also identified Sec1 family domain-containing protein 1 (SCFD1) and an oligosaccharyltransferase complex subunit (OSTC) as host factors used by PE, and neither of these proteins had previously been implicated in PE intoxication. Consistent with our KBM7 screen, pools of HeLa cells mutated for either SCFD1 or OSTC show increased resistance to intoxication (Fig. 1E). SCFD1 (also called rsly1) regulates Golgi-to-ER retrograde protein transport, possibly through its association with syntaxin 5 (38,39). OSTC is predicted to form stable complexes with the Sec61 complex, a protein-conducting channel that translocates nascent polypeptides across the ER membrane (40).
GPR107 Is Required for PE and Campylobacter jejuni CDT Intoxication-We identified GPR107 (synonyms are KIAA1624 and LUSTR1) as one of the hits with the highest enrichment of gene-trap insertions in the surviving cells (Fig. 3A). GPR107 was likewise identified in a screen performed with cytolethal distending toxin (cjCDT) secreted by C. jejuni (23). GPR107 is also required for intoxication of mouse cells by ricin, as identi-fied in a genetic screen performed in haploid mouse embryonic stem cells (41). Beyond its assignment to the family of GPCRs, nothing is known about GPR107 (42,43). CDTs produced by Escherichia coli, Aggregatibacter actinomycetemcomitans, and Haemophilus ducreyi retain the capacity to kill GPR107 null cells, whereas cjCDT no longer does so (23). CDTs are believed to exert their toxicity in the nucleus by triggering the cell cycle checkpoint, causing G 2 arrest, which eventually leads to cell death (44). Even though their mechanisms of intoxication are different, identification of GPR107 in the PE and cjCDT screens suggests that they share a common host factor at some stage en route to their final destination. Because GPR107-deficient and wild type KBM7 cells bind PE equally well (Fig. 2E), we can reasonably exclude a surface receptor function. The sensitivity of intra-Golgi trafficking to pertussis toxin (17) indicates involvement of a G-protein, and presumably this implies the existence of coupled GPCR(s) and effector(s). GPR107 may be a candidate for such a GPCR.
Ectopic Expression of GPR107 Restores Sensitivity to PE and cjCDT in GPR107-null Cells-To validate the conclusion that GPR107 is necessary for PE and cjCDT intoxication, a PE-re- HeLa cells expressing C-terminally GFP-tagged GPR107 that were intoxicated with PE-(Atto647)-RDEL. Note that PE partially colocalizes with GPR107. E, similar number (1 ϫ 10 6 ) of GPR107 null and wild type KBM7 cells were intoxicated with 50 ng/ml PE for 30 min, and the cell lysates (ϳ20 g) were analyzed by immunoblots using PE and GAPDH antibodies.
sistant mutant clone (GPR107 GT ) that carried a defined genetrap insertion was isolated (Fig. 3B). Treatment of these cells with cjCDT showed resistance of GPR107 GT cells to intoxication (Fig. 3D). Introduction of HA-tagged GPR107 into GPR107-deficient cells fully restored sensitivity to both PE and cjCDT, confirming that GPR107 is indeed required for intoxication (Fig. 3, C and D). As expected, GPR107 null cells are sensitive to CDTs from other species (Fig. 3D) (23). A pool of HeLa cells exposed to the appropriate CRISPR/Cas9 construct to target the GPR107 gene shows increased resistance to PE (Fig. 1E). GPR107 Is Localized in the Trans-Golgi Network-To determine the subcellular localization of GPR107, we established a HeLa cell line that stably expresses GPR107 with an HA tag at its C terminus. Confocal microscopy showed that GPR107 colocalizes with TGN46, a marker for the TGN, but not with the cis/medial Golgi marker, Giantin (Fig. 4, A and C). Upon addition of BFA, GPR107 redistributed over small vesicles as did TGN46, but GPR107 and TGN46 did not completely colocalize, placing GPR107 in a distinct Golgi sub-compartment (Fig. 4B). GPR107 also partially colocalizes with the endoprotease furin (Fig. 4D). We were unable to detect GPR107 at the cell surface (Fig. 4E), where we employed the palmitoylated/ myristoylated-tagged red fluorescent protein as a plasma membrane marker (45).
GPR107 Undergoes Proteolytic Cleavage-GPR107 contains seven transmembrane segments with a combined molecular mass of ϳ52 kDa for the protein backbone, which carries three predicted N-linked glycosylation sites (Fig. 5A). To assess maturation and turnover of GPR107, we performed pulse-chase analysis in HeLa cells that stably express C-terminally HA-tagged GPR107. We observed that GPR107 undergoes a proteolytic cleavage to yield two fragments of 17 and 35 kDa (molecular mass estimates made after complete deglycosyla-tion of immunoprecipitated material with PNGase F) (Fig. 5C). A single cleavage is therefore likely responsible for the generation of the two fragments. The larger fragment bears the HA . GPR107 is localized in the trans-Golgi network. A, confocal images of HeLa cells expressing C-terminally HA-tagged GPR107. Cells were costained for HA and for the trans-Golgi network marker TGN46. B, cells were treated with BFA for 3 h and co-stained as described in A. C-E, cells were costained with HA and the cis/medial-Golgi marker Giantin, furin, or cell surface marker palmitoylated/myristoylated-tagged red fluorescent protein (pm-RFP). In all images the nuclei were stained with DAPI (blue). Note that GPR107 colocalizes with TGN46 but not with Giantin, and the presence of both red and green vesicular structures in the BFA-treated cells is indicative of at least partial segregation of GPR107 and TGN46. AUGUST 29, 2014 • VOLUME 289 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 24011
epitope and therefore corresponds to the C terminus of GPR107 (Fig. 5C).

GPR107 Is Cleaved by Furin and the Cleaved Fragment Remains Associated with Disulfide Bond-Where in the cell does cleavage occur? What is the enzyme responsible for cleavage?
GPR107 cleavage is blocked by exposure of cells to BFA (Fig.  5D), whereas agents that compromise lysosomal function such as concanamycin A have little effect on cleavage (Fig. 5F, left panel). Because cleavage occurs relatively soon after synthesis and GPR107 is localized predominantly in the TGN, we explored the role of furin, the major processing protease of the secretory pathway, residing also in the TGN (46). GPR107 contains an extended furin recognition site that includes KSKR, a variant of the classical furin cleavage motif (Arg-Xaa-(Lys/Arg)-Arg) (Fig. 5, A and B).
Although not common among furin substrates, in GPR107 the Lys residue replaces the first conserved Arg, similar to the furin cleavage site found in the Ebola virus glycoprotein precursor (47). To assess its possible role as a cleavage site, we generated several mutants of GPR107 centered on the KSKR sequence and tested their sensitivity to cleavage in HeLa cell transfectants. A single mutation, R182A, abolished cleavage of GPR107 (Fig. 5C).
In an alternative approach, we pulse-labeled HeLa cells that express HA-tagged GPR107 with [ 35 S]methionine/cysteine for 30 min, immunoprecipitated GPR107, and then subjected the immunoprecipitate to digestion with furin. Under these condi- Proposed furin cleavage site Proposed furin cleavage site FIGURE 5. GPR107 is cleaved by furin and the disulfide bond that associates the cleaved fragments is essential for its activity. A, predicted topological structure of GPR107. N-Glycosylation sites, disulfide bond linkage, and the furin cleavage site are indicated. B, sequence alignment of the furin-like cleavage site present at the N-terminal region of GPR107 from different eukaryotes. C, HeLa cells expressing C-terminally HA-tagged GPR107 WT and GPR107 R182A were labeled with [ 35 S]methionine/cysteine for 30 min and chased for the indicated time points. Cells were lysed and immunoprecipitated with anti-HA coupled beads, subjected to PNGase F treatment, analyzed by SDS-PAGE, and autoradiography. D, BFA untreated and treated C-terminally HA-tagged GPR107 WTexpressing HeLa cells were labeled with [ 35 S]methionine/cysteine for 30 min and chased for different time points. Cells were lysed and immunoprecipitated with anti-HA coupled beads, subjected to Endo H treatment, and analyzed as described in C. E, HeLa cells expressing C-terminally HA-tagged GPR107 WT and GPR107 C109/228A were labeled with [ 35 S]methionine/cysteine for 30 min, chased, and immunoprecipitated as described in C. Unless otherwise indicated, immunoprecipitates were eluted and analyzed in reducing conditions. F, HeLa cells expressing C-terminally HA-tagged GPR107 WT were labeled with [ 35 S]methionine/cysteine for 30 min and chased in the presence or absence of the lysosomal inhibitor concanamycin A. Cells were lysed and immunoprecipitated with anti-HA coupled beads, and samples were processed as described in C (left panel). Alternatively, immunoprecipitate at the 0-h chase time point was subjected to furin digestion and analyzed by SDS-PAGE and autoradiography (right panel). G, KBM7 WT , GPR107 GT , and GPR107 GT cells reconstituted with GPR107 WT , GPR107 R182A , or GPR107 C109A/C228A were intoxicated with PE; cell viability was determined and presented as percent relative to nonintoxicated cells. Error bars represent S.D. of three experiments performed in duplicate.
tions, immunoprecipitated GPR107 was cleaved by furin and yielded a C-terminal fragment similar in mobility to that produced in living cells (Fig. 5F, right panel). GPR107 co-localized with furin at the TGN, consistent with furin's proposed role in the proteolytic conversion of GPR107 (Fig. 4D).
To determine whether the furin-processed fragments of GPR107 remain associated after cleavage, we examined the behavior of GPR107 by SDS-PAGE under nonreducing conditions. The predicted furin cleavage site is straddled by two cysteine residues, which could form the only possible disulfide bond in the predicted extracellular portion of GPR107. When analyzed under nonreducing conditions, the cleavage fragments indeed remained associated. The C109A/C228A mutation eliminates the possibility of forming this single disulfide bond and abolished the covalent association of the two cleavage fragments (Fig. 5E), further demonstrating that the two cleavage fragments are indeed disulfide-linked.
Is proteolytic cleavage of GPR107 important for PE and cjCDT intoxication? We introduced the cleavage-resistant form of GPR107 (GPR107 R182A -HA) into GPR107-deficient cells and examined its ability to restore sensitivity to PE and cjCDT. We found that GPR107 R182A -HA restored sensitivity to both PE and cjCDT to a level similar to that of wild type GPR107 (Fig. 5G). Furin cleavage of GPR107 is thus dispensable for PE intoxication. However, expression of the C109A/C228A mutant in GPR107-deficient cells only partially restored sensitivity to PE (Fig. 5G), indicating that the disulfide bond that preserves the interaction between the two cleavage products contributes to GPR107 function.
N-terminal Region of GPR107 Is Required for Its Function-We generated a truncation mutant (GPR107 ⌬40 -182 ) that lacks the N-terminal fragment of GPR107 to determine the role of this fragment in PE intoxication. This construct carries the N-terminal GPR107 signal peptide (amino acids 1-39) to allow proper ER insertion of the mutant GPR107 (Fig. 6A). GPR107 ⌬40 -182 localized to the TGN, similar to wild type GPR107 (Fig. 6B), establishing that the requisite TGN targeting signals are contained elsewhere in the GPR107 sequence. However, expression of GPR107 ⌬40 -182 in GPR107 GT cells failed to restore sensitivity to PE intoxication (Fig. 6C). The N-terminal extracellular region of GPR107 is therefore critical for its function in PE transport.
GPR107 Contributes to Retrograde Transport of Cholera Toxin-To gain more insight into the biological function of GPR107, we set out to test its involvement in both anterograde and retrograde trafficking. First, we examined by pulse-chase analysis whether or not the absence of GPR107 affects protein secretion. Media were collected at the different chase time points and analyzed by SDS-PAGE and autoradiography. No obvious differences were observed between the products released by GPR107-deficient and wild type KBM7 cells (Fig.  7A). The absence of GPR107 was also without effect on the maturation of class I MHC products, type I membrane glycoproteins that traffic via the constitutive secretory pathway to the cell surface. GPR107-deficient cells show maturation of class I MHC products at a rate similar to that seen in wild type cells (Fig. 7B), as assessed by rate and extent of acquisition of Endo H resistance. In our haploid genetic screen, we identified furin as one of the hits required for intoxication by PE. Furin is a protease that cleaves not only PE (5) but also its receptor (LRP1; Gu et al. (33)) and GPR107 (this study). Consequently, we examined the maturation/trafficking of furin in GPR107deficient cells. Furin not only recycles within the secretory pathway but it is also secretes (46). No difference was found between GPR107-deficient and wild type KBM7 cells in maturation or secretion of furin (Fig. 7C). Confocal microscopy on GPR107-depleted HeLa cells showed that the distribution of furin was similar to that of the control cells (Fig. 7D, left panel). GPR107 is thus dispensable for anterograde transport/secretion of furin.
Given its localization to the trans-Golgi network, does the lack of GPR107 affect the structure of the Golgi? Analysis by confocal microscopy on wild type and GPR107-depleted HeLa cells for the trans-Golgi marker TGN46 before and after BFA treatment (Fig. 7D) showed no difference in overall structure or in the recovery from Golgi fragmentation post-BFA treatment (Fig. 7D). GPR107 is therefore not obviously involved in maintaining the structure of this organelle.
To examine the role of GPR107 in retrograde transport, we used a modified version of cholera toxin (CTx), equipped with a glycosylation motif and biotin at the C terminus of the A1 subunit (Guimaraes et al. (21)). A similar number of KBM7 WT , GPR107 GT , and GPR107 GT cells reconstituted with GPR107 cDNA were intoxicated with CTx-GGGK(biotin)-NNSTG for  Supernatants were collected at the indicated time points and analyzed by SDS-PAGE and autoradiography. B, wild type and cells lacking GPR107 were pulse-labeled with [ 35 S]methionine/cysteine as in A and chased for the indicated time points. Cells were lysed, and class I MHC molecules were recovered using W6/32 antibody, treated with Endo H, and analyzed by SDS-PAGE and autoradiography. C, KBM7 WT and GPR107 GT expressing C-terminally HA-tagged furin were labeled with [ 35 S]methionine/cysteine for 20 min and chased for different time points. Both the cells and the supernatants were collected in parallel, lysed, and immunoprecipitated with anti-HA-coupled beads and where indicated subjected to PNGase F treatment and then analyzed by SDS-PAGE and autoradiography. furin i is intracellular furin; furin s is secreted furin. D, HeLa cells expressing HA-tagged GPR107 were transduced with the control shGFP or two sets of shRNA that targets GPR107. Cell lysates were prepared from these samples, and knockdown efficiency was examined by Western blotting using anti-HA HRP. E, confocal images of control and GPR107-depleted HeLa cells before or at different time points after BFA treatment. Cells were stained for the trans-Golgi network marker TGN46 or furin (left panel). In all images, the nuclei were stained with DAPI (blue). 30 min and then chased for 1 and 3 h. Glycosylation of CTx indicates arrival in the ER and is accompanied by an increase in apparent molecular weight, as detected by Western blotting using streptavidin-HRP. Cells that lack GPR107 showed a reduced level of CTx glycosylation compared with control cells (Fig. 8, A and B). GPR107 GT cells reconstituted with GPR107 cDNA (overexpressors) showed an increased level of CTx glycosylation (Fig. 8, A and B), suggesting that GPR107 contributes to retrograde transport of CTx and possibly of PE as well.
We generated a similarly engineered version of PE, equipped with a glycosylation motif extended with an ER retrieval signal sequence and modified with a biotin residue (PE-GGGK(biotin)-NNSTGKDEL), to probe the role of GPR107 in retrograde transport of PE. However, we were unable to detect any glycosylated PE even in the wild type or GPR107-null cells reconstituted with GPR107 cDNA overexpressors (data not shown). This is consistent with our microscopy results, where we were likewise unable to detect PE in compartments other than the endocytic pathway (data not shown). This is entirely consistent with the notion that only very few molecules of PE need to make it to the final destination, the cytosol, to achieve intoxication.

DISCUSSION
Bacterial toxins are valuable tools to dissect the physiology of mammalian cells. The coevolution of pathogens and their hosts results in strategies in which host factors are exploited to the advantage of the invader (48,49). We set out to expand our knowledge of how a bacterial toxin takes advantage of host cell machinery by performing a genome-wide genetic screen and molecular characterization of the hits.
Loss-of-function haploid genetic screens using human KBM7 cells have led to the identification of host factors essential for viral (29,50,51) and bacterial (52) pathogenesis as well as for other bacterial toxins (21,29,53). We used CRISPR/Cas9mediated gene editing in HeLa cells to support the results obtained in the KBM7 screen. GPR107 is one of the novel genes identified in our genetic screen in KBM7 cells as a host factor used by PE. GPR107 is ubiquitously expressed and localizes to the TGN. It is conserved in higher eukaryotes, including Caenorhabditis elegans, fruit fly, zebra fish, and Arabidopsis thaliana. GPR107 has the hallmarks of a GPCR, but if an endogenous ligand exists it remains to be identified, because GPR107 is unlikely to have evolved to enable intoxication by ricin or by bacterial toxin.
The GPR107 null cells obtained in our genetic screens are resistant to CDT from C. jejuni (cjCDT), but remain sensitive to other CDTs (E. coli and A. actinomycetemcomitans). How does GPR107 participate in intoxication by PE and cjCDT? Our inability to detect GPR107 at the cell surface makes it unlikely that GPR107 serves as the receptor, although the presence of a small number of surface-disposed GPR107 molecules would have escaped detection by the approaches used here. We consider a role for GPR107 as a shared receptor for PE and cjCDT less plausible also in view of the widely divergent molecular structures of PE and cjCDT. Finally, our finding that GPR107 null cells bind PE in quantities similar to the parental KBM7 cell line also argues against a surface receptor role for GPR107. There is a consensus that the closely related proteins LRP1 and LRP1B are receptors for PE (7,8). Although GPCR signaling has mostly been considered as confined to the cell surface, recent evidence points to the possibility of intracellular signaling as well (54).
Proteins involved in retrograde trafficking identified in the PE screen (GARP complex, KDELR1, AP1M1, SCFD1, and OSTC) did not appear as hits in the cjCDT screen (23), demonstrating that the pathways of intoxication by PE and cjCDT overlap only in part. Interestingly, such a high degree of specificity shown by individual toxins in co-opting trafficking pathways may represent an evolutionary adaptation of these bacterial products toward optimizing their ability to reach specific cellular targets in a fully active form.
Membrane vesiculation at the TGN requires the trimeric G-protein G␤␥ subunits. Addition of G␤␥ subunits stimulates Golgi vesiculation in vitro through activation of PKD (13,55,56). The heterotrimeric G-proteins G␣ i3 and G␣ s have been localized to the Golgi apparatus (57). Overexpression of G␣ i3 and the use of reagents that perturb the function of G-proteins demonstrate the involvement of heterotrimeric G-proteins in vesicular transport (17). PE and cjCDT might both require G-proteins for retrograde trafficking through the Golgi. To the best of our knowledge, the GPCR(s) associated with G-proteins at the TGN remain(s) unknown. GPR107 is therefore a candidate for a receptor that controls certain aspects of intra-Golgi trafficking.
GPR107 is cleaved by furin, and the cleaved fragments remain associated, suggesting a conceptual analogy with protease-activated GPCRs such as the thrombin receptor (58,59). In its role as a pro-protein convertase, cleavage by furin may activate the substrate, but proteolysis can also inactivate or modify the activity of the furin substrate (60). If GPR107 were a furin-activated GPCR that undergoes cleavage in the Golgi, then this cleavage might be required for GPR107 function. Our finding that expression of a cleavage-resistant form of GPR107 in GPR107-null cells restores sensitivity to both PE and cjCDT shows that cleavage is dispensable for intoxication. However, this does not rule out the possibility that cleavage is important for other cellular processes that might involve GPR107. Furthermore, the cleavage-resistant version of GPR107 may still allow a segment of its N-terminal domain to engage the ligandbinding pocket of GPR107 and activate it. Our attempts to reca-pitulate the GPR107 null phenotype through shRNA-mediated knockdown have failed, perhaps because only a few GPR107 molecules may suffice to exert its normal function. Consequently, if the cleavage-resistant version of GPR107 would show much reduced activity compared with its cleavable counterpart, it is uncertain whether we would have been able to accurately score such a difference.
Despite its presence in the TGN and seemingly normal localization pattern compared with intact GPR107, the mutant lacking the N-terminal region of GPR107 fails to restore sensitivity to PE when expressed in GPR107-deficient cells. Furthermore, compromising the disulfide bridge that links the two fragments together affects GPR107 function. This may be due to the fact that any modification at the N terminus affects its ability to bind its ligand and/or interactors that are necessary for its function. The disruption of the disulfide bond may lead to the assembly of an unstable or improperly folded form of GPR107 with a reduced ability to interact with its protein cargo.
Our biochemical data exclude an obvious role for GPR107 in anterograde trafficking. However, cells that lack GPR107 showed a reduced level of glycosylation of the engineered cholera toxin reporter. Reconstitution of GPR107-null cells with GPR107 cDNA, which presumably causes overexpression of GPR107 to levels that exceed wild type levels, resulted in an increased level of cholera toxin glycosylation. Combined, these data are consistent with the involvement of GPR107 in retrograde trafficking, supported also by the fact that GPR107 is localized at the Golgi and that PE uses the retrograde trafficking pathway to reach its final destination, the cytosol.
In conclusion, we identify several novel cellular components used by PE and thus provide a far more detailed map of the PE intoxication pathway than what was known until now (Fig. 8C). Bacterial toxins can thus help identify novel host components involved in key intracellular trafficking steps. The exact contribution of GPR107 to normal host physiology deserves exploration in depth. Given its ubiquitous patterns of expression and conservation among higher eukaryotes (61), GPR107 might belong to the GPCRs that likely orchestrate G-protein-dependent events at the Golgi apparatus (14,17).