Functional Divergence in the Role of N-Linked Glycosylation in Smoothened Signaling

Abstract

The G protein-coupled receptor (GPCR) Smoothened (Smo) is the requisite signal transducer of the evolutionarily conserved Hedgehog (Hh) pathway. Although aspects of Smo signaling are conserved from Drosophila to vertebrates, significant differences have evolved. These include changes in its active sub-cellular localization, and the ability of vertebrate Smo to induce distinct G protein-dependent and independent signals in response to ligand. Whereas the canonical Smo signal to Gli transcriptional effectors occurs in a G protein-independent manner, its non-canonical signal employs Gαi. Whether vertebrate Smo can selectively bias its signal between these routes is not yet known. N-linked glycosylation is a post-translational modification that can influence GPCR trafficking, ligand responsiveness and signal output. Smo proteins in Drosophila and vertebrate systems harbor N-linked glycans, but their role in Smo signaling has not been established. Herein, we present a comprehensive analysis of Drosophila and murine Smo glycosylation that supports a functional divergence in the contribution of N-linked glycans to signaling. Of the seven predicted glycan acceptor sites in Drosophila Smo, one is essential. Loss of N-glycosylation at this site disrupted Smo trafficking and attenuated its signaling capability. In stark contrast, we found that all four predicted N-glycosylation sites on murine Smo were dispensable for proper trafficking, agonist binding and canonical signal induction. However, the under-glycosylated protein was compromised in its ability to induce a non-canonical signal through Gαi, providing for the first time evidence that Smo can bias its signal and that a post-translational modification can impact this process. As such, we postulate a profound shift in N-glycan function from affecting Smo ER exit in flies to influencing its signal output in mice.

Fig 1. Identification of Smo N-linked glycosylation sites.

A. A multiple sequence alignment of Smo proteins from different phyla are shown. Consensus sequences for N-linked glycosylation are highlighted in gray and the Asn acceptor residues are bold. Sites conserved across vertebrate proteins are indicated as N1-N4. The predicted D. melanogaster sites are not tightly conserved with vertebrates. B. Drosophila Smo is N-glycosylated. Cell lysates prepared from Cl8 cells expressing Hh with wild type or NQ5 dSmo proteins were treated with the indicated deglycosylating enzymes. Wild type Smo demonstrated ER (arrow) and post-ER (arrowhead) glycosylation species. NQ5 migrated similarly to the fully deglycosylated species under all conditions (arrow). C. Mouse Smo is N-glycosylated. Cellular lysates from Smo-/- cells stably expressing mSmoWT or mSmoNQ4 were treated with deglycosylating enzymes and subjected to SDS-PAGE and western blot. mSmoWT resolves as two distinct forms (lane 2). The arrow marks the ER form and the arrowhead indicates the post-ER form. mSmoNQ4 migrates in SDS-PAGE similarly to the PNGase-treated wild type protein (lanes 4–5). C’. mSmoNQ4 is O-glycosylated. Lysates were prepared from NIH3T3 cells expressing mSmoNQ4 and subjected to lambda phosphatase, PNGase and O-glycosidase/neuraminidase treatments. The upper band collapsed upon O-glycosidase/neuraminidase treatment. D. Expression of individual N to Q dSmo mutants. The indicated N to Q dSmo mutants were expressed in Cl8 cells and cell lysates were analyzed by SDS-PAGE and western blot against the Myc tag. Kin serves as loading control. E. Extracellular mSmo consensus sites are N-glycosylated. Mutation of individual extracellular mSmo glycosylation sites induced faster mobility on SDS-PAGE. mSmoN450Q migrated similarly to SmoWT. For western blots, mSmo was detected using anti-Smo and dSmo with anti-Myc. Kinesin (Kin) and Tubulin (Tub) were blotted for loading controls.

Fig 2. N-linked glycans are required for dSmo trafficking and activity.

A. dSmoNQ5 does not signal in vitro. Cl8 cells were transfected with control or smo 5’UTR dsRNA, the ptcΔ136-luciferase reporter, pAc-renilla control, pAc-myc-smoWT or NQ5, and pAc-hh or empty vector control. Hh-induced reporter activity (gray bars) was ablated by knockdown of endogenous smo and rescued by dSmo cDNA lacking UTR sequence for wild type, but not for dSmoNQ5. B-B’. dSmoNQ5 demonstrates altered sub-cellular localization. Cl8 cells expressing Calreticulin-EGFP-KDEL ER marker (GFP-ER, green) and Myc-SmoWT or NQ5 in the presence or absence of Hh were imaged by immunofluorescence microscopy. Wild type dSmo (anti-Myc, magenta) localized to puncta that did not overlap with the ER marker in the absence of Hh, and translocated to the plasma membrane in response to Hh. The NQ5 mutant overlapped with the ER marker under both conditions. DAPI (blue) marks the nucleus. Scale bar is 5 μm (upper right box). B’. GFP-ER colocalizes with V5 tagged BiP, Calnexin (Cnx) and Calreticulin (Crc). DAPI marks the nucleus. C-H. dSmoNQ5 does not signal in vivo. Transgenes encoding wild type (G) or NQ5 (H) dSmo proteins were expressed in the nubbin>dicer;smo ^3’UTR background (E). Whereas wild type Smo could rescue the loss of function phenotype induced by smo ^3’UTR, dSmoNQ5 could not (G-H compared to F and C-D, control). UAS-EGFP was expressed in the nubbin>dicer;smo ^3’UTR background and serves as a control for normalized transgene dosage (F). I. Wild type and NQ5 dSmo proteins are present at similar protein levels in wing imaginal disc tissue lysate. The dSmoN213Q,N336Q protein level is higher.

Fig 3. N336Q impacts dSmo trafficking and activity.

A. N336 possesses an essential glycan modification. The rescue reporter assay was performed in smo knockdown Cl8 cells as in Fig 2. Each of the single N to Q mutants was able to rescue reporter gene activity in the smo knockdown background with the exception of N336Q. This mutant rescued to ~50% of the wild type activity. B. dSmoN336Q is mislocalized. Cl8 cells expressing Hh and the indicated dSmo proteins were imaged by immunofluorescence microscopy. Whereas each of the active single glycosylation site N to Q mutants (anti-Myc, magenta) reached the plasma membrane (indicated by F-actin stain, green) in Hh-expressing cells, dSmoN336Q failed to do so. DAPI (blue) marks the nucleus. Scale bar is 5 μm (upper right). C. SmoN336Q is retained in the ER. Myc-SmoN336Q expressed in Cl8 cells overlaps with the ER marker protein Calreticulin-KDEL-GFP. Scale bar is 5 μm (upper right). D. Treatment with deglycosylating enzymes demonstrates that the majority of dSmoN336Q is present in the EndoH sensitive, ER resident fraction (black arrowhead). WT and N213Q proteins are equally distributed between ER and post-ER fractions (white arrowhead).

Fig 4. N213 glycosylation partially compensates for N336 N-glycan loss.

A. The indicated mutants were expressed in Cl8 cells. Each of the mutant proteins migrated more quickly in SDS-PAGE than wild type Smo, but not as quickly as SmoNQ4 or NQ5. B. Glycosylation at N213 partially compensates for N336 glycan loss. The rescue reporter assay was performed as described in Fig 2A. Each of the indicated double mutants, with the exception of N213Q,N336Q, was able to rescue reporter gene expression in the smo knockdown background to a level similar to that of N336Q. dSmoN213Q,N336Q demonstrated a level of activity similar to dSmoNQ5. Significance was determined using Student’s t-test. C. N336Q-containing double mutants are retained in the ER. Cl8 cells expressing the indicated Smo proteins (anti-Myc, magenta), the Cal-EGFP-KDEL marker (green) and Hh were examined by immunofluorescence microscopy. Whereas wild type Smo reached the plasma membrane, the double mutants overlapped with the ER marker. ActinRed (red) marks F-actin. DAPI (blue) marks the nucleus. Scale bar is 5 μm (upper right). D. Treatment of lysates from WT or N213Q,N336Q expressing cells with deglycosylating enzymes reveals that the N213,336Q mutant is present in the EndoH sensitive ER fraction, arrowhead. E-E’. N336Q and N213Q,N336Q mutants have disulfide bond defects. Biotin-maleimide was used to tag free thiol groups in cellular lysates prepared from Cl8 cells expressing WT, N336Q, N213Q,N336Q and C320A dSmo proteins. WT dSmo is not well captured on NeutrAvadin beads (lane 2, bound). N-glycan mutants are captured similarly to the disulfide bond mutant C320A (lanes 3–5, bound), indicating that at least one disulfide bridge is disrupted by N-glycan loss. E’ shows the ratio of bound to unbound dSmo proteins normalized to kinesin. Relative binding was determined by densitometry analysis of two independent binding assays. C320A, which has an established disulfide bond defect served as positive control. It’s binding ratio was arbitrarily set to 1.0 and other values are shown relative to it. Error bars are provided to show the standard deviation between the two experiments. F. dSmoN213Q,N336Q fails to rescue smo knockdown in vivo. UAS-dsmoN213Q, N336Q was co-expressed with UAS-dicer and UAS-smo ^3’UTR using the nubbin-Gal4 driver. Its expression did not modify the smo knockdown phenotype (compare to Fig 2F). Multiple progeny were analyzed over two crosses and a representative wing is shown.

Fig 5. mSmo trafficking and dimerization are unaffected by N-glycan loss.

A. mSmoNQ4 reaches the cell surface. Cell surface biotinylation analysis was performed on Smo-/- cells stably expressing WT or NQ4 mSmo proteins. Biotinylated proteins were collected on streptavidin beads, and analyzed by western blot and densitometry. Arrowheads indicate post-ER species that were used for quantification (black, WT and white, mutant). The ratio of extracellular to intracellular post-ER mSmo is similar for WT and NQ4 proteins (quantified in A’). The experiment was performed 5 times and all data pooled. Error bars indicate s.e.m. B. mSmoNQ4 localizes to the primary cilium. Smo-/- cells stably expressing WT or NQ4 Smo proteins (green, anti-Smo) were treated with SAG. Both WT and mutant mSmo proteins enrich in the primary cilium in response to SAG. γ-tubulin marks the basal body (magenta, anti-γ-tubulin) and DAPI marks the nucleus (blue). B’ shows quantification of ciliary localization (~100 cells over 3 independent experiments). C. NQ4 forms homo- and hetero-dimers. Differentially tagged WT and NQ4 mSmo proteins were expressed in NIH3T3 cells as indicated (left panel). mSmo complexes were immunoprecipitated (IP) from cellular lysates using anti-GFP antibody and immunocomplexes were analyzed by SDS-PAGE and western blot against the V5 epitope tag (right panel). mSmoNQ4-V5 co-purified with both WT and NQ4 YFP-tagged proteins.

Fig 6. mSmoNQ4 binds ligands and activates canonical signaling.

A-C. NQ4 binds bodipy cyclopamine. NIH3T3 cells expressing WT, NQ4 or M2 Smo proteins were treated with bodipy cyclopamine (5nM) for 4 hours, then fixed and immuno-stained for Smo (anti-Smo, magenta). Bodipy cyclopamine (green) accumulated in primary cilia in cells expressing WT or NQ4 Smo proteins. Bodipy cyclopamine failed to enrich in primary cilia in cells expressing oncogenic SmoM2. Higher resolution images of primary cilia with shifted channel overlays are shown (right). DAPI (blue) marks the nucleus. Scale bar is 5 μm (upper right). D. SmoNQ4 binds 20(S)-OHC. Lysates from cells expressing WT, NQ4, M2 or M2-NQ4 Smo proteins were incubated with 20(S)-OHC sepharose beads, and proteins purifying on the beads were analyzed by western blot using anti-Smo. Both WT (lane 2) and NQ4 (lane 5) Smo proteins were captured on 20(S)-OHC beads. Binding was specific, as it could be disrupted by 50 μM cold competitor (+). Neither SmoM2 (lane 8) nor SmoM2-NQ4 (lane 11) were captured on 20(S)-OHC beads. I indicates input. E. mSmoNQ4 signals to Gli. Smo-/- cells stably transfected with vector control (pMSCV-puro), pMSCV-mSmoWT or pMSCV-mSmoNQ4 were transfected with the 8xglibs-luciferase reporter and tk-renilla control. Both wild type and NQ4 Smo proteins induced reporter gene activity in response to Shh conditioned media (pink), SAG (green, 100 nM) and 20(S)-OHC (purple, 10 μM). Reporter assays were performed 2 times in triplicate and all data pooled. Error bars indicate s.e.m.

Fig 7. Glycosylation is required for non-canonical mSmo signaling.

A-B. Glycosylation alters mSmo DMR. HEK293T cells expressing wild type (B-B’), NQ4 (B”) or NQ3 (B*) mSmo proteins were subjected to label-free DMR analysis. SAG- and cyclopamine-induced changes in cellular responses are displayed as refractive index alterations (Δ picometer). DMR responses were collected for cells pretreated with vehicle (red line, all panels), cyclopamine (B’, green line, 2μM), cholera toxin (CTX, blue line, B, B”, B*, 10 ng/mL) or pertussis toxin (PTX, green line, B, B”, B*, 10 ng/mL) prior to addition of 200 nM SAG. DMR experiments were performed 6–12 times. Representative graphs are shown. C-D. cAMP modulation is attenuated by NQ4 mutation. HEK293T cells expressing WT (black bars) or NQ4 (gray bars) mSmo proteins were pretreated with 0.5 μM forskolin prior to addition of SAG or 20(S)-OHC. cAMP levels were measured using HTRF with the LANCE Ultra cAMP kit. The forskolin induced cAMP level was set to 100% (dashed lines). 1x = 100 nM of the indicated ligand. Experiments were performed 6–8 times and all data pooled. Error bars represent s.e.m. E-F. Shh-regulated cell migration is attenuated by NQ4 mutation. Scratch assays were performed using Smo-/- cells stably transfected with pMSCV-puro, pMSCV-mSmoWT or pMSCV-mSmoNQ4. Migration into the scratch zone is shown at T = 0 and T = 8 hours. Shh conditioned media was added at T = 0. The non-canonical signal-regulated migration response was attenuated by N-glycan loss (quantified in F). SmoWT-induced migration was attenuated by cyclopamine treatment (F).

Fig 8. N-glycosylation status correlates with signal bias.

A model for modulation of mSmo signal bias. SAG (yellow) binding to N-glycosylated mSmo stabilizes a conformation that effectively signals through both canonical and non-canonical routes. SAG binding to the mSmo mutant stripped of N-glycans (red) silences non-canonical signaling, but is highly permissive for canonical signaling.