N-glycosylation of α1D-adrenergic receptor N-terminal domain is required for correct trafficking, function, and biogenesis

G protein-coupled receptor (GPCR) biogenesis, trafficking, and function are regulated by post-translational modifications, including N-glycosylation of asparagine residues. α1D-adrenergic receptors (α1D-ARs) – key regulators of central and autonomic nervous system function – contain two putative N-glycosylation sites within the large N-terminal domain at N65 and N82. However, determining the glycosylation state of this receptor has proven challenging. Towards understanding the role of these putative glycosylation sites, site-directed mutagenesis and lectin affinity purification identified N65 and N82 as bona fide acceptors for N-glycans. Surprisingly, we also report that simultaneously mutating N65 and N82 causes early termination of α1D-AR between transmembrane domain 2 and 3. Label-free dynamic mass redistribution and cell surface trafficking assays revealed that single and double glycosylation deficient mutants display limited function with impaired plasma membrane expression. Confocal microscopy imaging analysis and SNAP-tag sucrose density fractionation assays revealed the dual glycosylation mutant α1D-AR is widely distributed throughout the cytosol and nucleus. Based on these novel findings, we propose α1D-AR transmembrane domain 2 acts as an ER localization signal during active protein biogenesis, and that α1D-AR N-terminal glycosylation is required for complete translation of nascent, functional receptor.


Results and Discussion
n-terminal glycosylation is required for complete α 1D -AR biogenesis. The α 1D -AR N-terminal contains two putative N-glycosylation sites (N65, N82) with both serving as theoretical acceptors for N-glycans within the ER lumen 47 . N-glycosylation is the covalent attachment of an N-glycan sugar moiety to an asparagine residue within the consensus sequence N-X-S/T, where X is any amino acid except proline 48,49 . Thus, we sought to examine the glycosylation state of full length α 1D -AR using PNGase F deglycosylation assays. To test this possibility, HEK293 cells were transiently transfected with N-terminal SNAP-epitope tagged α 1D -AR cDNA constructs (SNAP-α 1D ). We have previously demonstrated the SNAP epitope-tag facilitates visual analysis of GPCR protein bands directly within polyacrylamide gels, and do not require nitrocellulose paper transfer or antibody staining, thus removing all potential false positive bands 22,28,50 . Incorporating this powerful technology, HEK293 cell lysates expressing SNAP-α 1D were lysed, denatured, and incubated with PNGase F, then subjected to polyacrylamide gel electrophoresis and near-infrared imaging (PAGE NIR). Similar to previous reports 27, 46 , our results were inconclusive, likely due to instability of α 1D -AR in the required buffer conditions ( Supplementary  Fig. S1A). To overcome this technical issue, we utilized lentil lectin affinity purification. Lentil lectin recognizes complex glycans containing α-(1 → 6)-linked fucose on the core GalNAc as well as glucose and/or α-mannose residues, and is active in a variety of buffer conditions 51,52 . HEK293 cells were transiently transfected with SNAPα 1D and lysates were incubated with lentil lectin sepharose beads. Samples were eluted and subjected to PAGE NIR analysis. Shown in Fig. 1A are the results. In agreement with our previous studies 22,28 , the input lane demonstrates full length SNAP-α 1D is robustly expressed as a monomeric band at ~80 kDa, a larger, more intense band at ~90 kDa (Fig. 1A, arrow, Supplementary Fig. S1B), as higher order oligomers (MW > 180 kDa), and as multiple NTD cleavage products (MW = ~30-35 kDa). Remarkably, both ~90 kDa monomeric and multimeric SNAP-α 1D species were detected in the lectin bound lane (Fig. 1A, bound). Although faint, the largest α 1D NTD cleavage product 28 was also observed in the lectin-bound sample. Thus, this experiment clearly demonstrates, for the first time, that the α 1D NTD is N-glycosylated.
Towards our goal of addressing the importance of each NTD glycosylation site for α 1D -AR function, we created single (N65Q or N82Q) and double (NQQ) glycosylation deficient SNAP-α 1D mutants using PCR site-directed mutagenesis (see Fig. 1B for schematic). To ensure each α 1D -AR NTD mutant was expressed as protein, cDNA constructs were transfected into HEK293 cells and subjected to PAGE NIR analysis. Both the N65Q and N82Q SNAP-α 1D NTD mutants display equivalent protein band patterns as SNAP-α 1D (Fig. 1C). Unexpectedly, the NQQ SNAP-α 1D double mutant did not produce monomeric or higher order oligomeric bands. Instead, NQQ SNAP-α 1D was primarily expressed as a single, robust band of ~43 kDa in size. Subtracting the size of the SNAP-epitope tag plus linker (25 kDa) yields a polypeptide of 18 kDa,roughly equivalent in size to the α 1D NTD, transmembrane domain (TM) 1, intracellular loop 1, and TM2. Subsequent lectin-purification assays reveal full length, N65Q and N82Q, but not NQQ, SNAP-α 1D species are glycosylated ( Supplementary Fig. S1C). To ensure this unexpected NQQ product was due to inhibition of glycosylation, and not a by-product of mutation, cells expressing WT SNAP-α 1D were treated with tunicamycin -an inhibitor of N-glycosylation 53 . 24 hours after transfection with WT SNAP-α 1D , HEK293 cells were treated with fresh media supplemented with 5 μg/mL tunicamycin or EtOH vehicle followed by PAGE NIR (Fig. 1D). Interestingly, though faint, the same ~43 kDa species observed in the NQQ SNAP-α 1D is also present in the tunicamycin treated samples ( Fig. 1D; circle). Thus, this initial round of experiments demonstrates that (A) α 1D -AR is glycosylated at N65 and N82; (B) only a single glycosylation site needs to be available for the NTD to become glycosylated and full length α 1D protein processing to occur; and (C) removal of both α 1D NTD glycosylation sites not only prevents glycosylation, but produces an abnormally short, previously unreported α 1D -AR peptide species.
We next tested two potential explanations for this serendipitous, intriguing result: (A) the NQQ double mutation introduces a destabilizing effect, causing the α 1D -AR to be targeted for degradation, with the observed 43 kDa band representing the major degradation product; or (B) NQQ is inhibiting proper translation of α 1D -AR, causing an early termination after TM2. These hypotheses were tested using a dual epitope-tagging approach. InFusion PCR was used to add C-terminal CLIP-epitope tags to WT SNAP-α 1D (S-WT-C) and NQQ SNAP-α 1D (S-NQQ-C). CLIP is a homolog of SNAP that covalently interacts with benzylcytosine conjugates, displaying no cross-reactivity for the SNAP substrate, benzylguanine 54 . We reasoned that if A is true, CLIP substrate fluorescence in the 700 channel (red) would be observed in both the S-WT-C and S-NQQ-C PAGE NIR lanes. Conversely, we would expect to detect no 700 signal in the S-NQQ-C lane if B were true, as the CLIP tag would not be transcribed if α 1D -AR translation was halted at TM2. Thus, S-WT-C and S-NQQ-C α 1D -AR cDNA www.nature.com/scientificreports www.nature.com/scientificreports/ constructs were expressed in HEK293 cells and subjected to PAGE NIR analysis. Fig. 1E shows that overlapping CLIP (red) and SNAP (green) substrate signals are detectable in the S-WT-C lane (left). Contrarily, no CLIP signal is observed in the S-NQQ-C lane, and only the previously observed 43 kDa SNAP-α 1D species ( Supplementary  Fig. S1D).
As an orthogonal approach, HEK293 cells expressing either S-WT-C or S-NQQ-C were incubated with bortezomib (BTZ)a proteasomal inhibitor (Fig. 1F,G)or protease inhibitor (PI) cocktail (Fig. 1I,J) for 24 hours followed by PAGE NIR analysis. As expected, significant increases of S-WT-C and S-NQQ-C protein bands were observed with BTZ treatment ( Fig. 1H; S-WT-C = 166.3 ± 4.3%, mean ± SEM of vehicle; S-NQQ-C = 186.3 ± 6.0% mean ± SEM of vehicle; Unpaired t test; p < 0.001), but not with PI cocktail treatment ( Fig. 1K; S-WT-C = 96.7 ± 0.3%, mean ± SEM of vehicle; S-NQQ-C = 97.0 ± 2.5%, mean ± SEM of vehicle; Unpaired t test, p > 0.05). However, neither BTZ nor PI cocktail had any discernable effect on the molecular weight of the NQQ band; nor were CLIP signals observed in either condition. Taken together, these findings indicate that the NQQ α 1D -AR species is not created by proteolytic cleavage and/or degradation of full-length α 1D -AR.
To further confirm the identity of this unexpected NQQ species, HEK293 cells were transiently transfected with either WT SNAP-α 1D or NQQ SNAP-α 1D , lysed, and SNAP-fusion proteins were isolated with SNAP-Capture beads. Due to the covalent nature of the SNAP-Capture:SNAP-tag, an on-bead digest was performed using Trypsin and Glu-C proteases. Samples were subjected to MS/MS analysis (SNAP MS/MS). As shown in Fig. 2A, identified peptides spanned the entirety of the WT SNAP-α 1D ( Fig. 2A, underlined). Contrarily, only peptides in the N-terminal domain were identified in NQQ SNAP-α 1D samples (Fig. 2B, underlined). Furthermore, previously reported α 1D -AR interactors syntrophin 21,24,25 , members of the dystrophin-associated protein complex 23 , and scribble 22,25 were identified in the WT, but not NQQ samples (Supplementary Datas S1, S2). Together, these data provide compelling evidence that glycosylation of both N65 and N82 are necessary for proper biogenesis of full-length α 1D -AR, and disruption of these essential glycosylation sites results in early termination of α 1D -AR processing after TM2.
Glycosylation imparts α 1D -AR function and plasma membrane insertion. The effects of NTD glycosylation on GPCR function and trafficking are highly divergent. Mutating N-terminal glycosylation sites decreases functional responses of the FSH 55 , dopamine D2 56 , and neurokinin 1 receptor subtypes 57 , while loss of glycosylation has no effect on the function of the histamine H2 receptor 58 . Conversely, blocking N-terminal glycosylation increases binding site density of the human oxytocin receptor 59 , and signaling efficacy of the vasopressin 1A receptor 60 . To understand how N-glycosylation impacts α 1D -AR function, label-free dynamic mass redistribution (DMR) assays were used to quantify the efficacy of the α 1 -AR agonist phenylephrine for stimulating α 1D NTD glycosylation site mutants. HEK293 cells expressing WT, N65Q, N82Q, or NQQ SNAP-α 1D were seeded in 384-well DMR plates and incubated with increasing concentrations of phenylephrine to facilitate concentration-response curve analysis (Fig. 3A). Surprisingly, phenylephrine maximal responses for N65Q (24.99 ± 11.35 pm, mean ± SEM), N82Q (46.64 ± 9.96 pm, mean ± SEM), and NQQ (45.20 ± 8.35 pm, mean ± SEM) were significantly lower than WT SNAP-α 1D (112.5 ± 9.27, mean ± SEM; p < 0.01, One-way ANOVA with Tukey's multiple comparisons post-hoc test). www.nature.com/scientificreports www.nature.com/scientificreports/ Glycosylation has been shown to facilitate plasma membrane trafficking of the angiotensin II type 1 61 , GPR30 62 , rhodopsin 1 63 , δ-opioid receptor 40,64,65 , and P2Y 2 receptor subtypes 66 . Therefore, one possible explanation for the reduced function of α 1D NTD glycosylation mutants may be aberrant cellular trafficking, leading to a decrease in cell surface expression. This was examined by quantifying WT, N65Q, N82Q and NQQ SNAP-α 1D plasma membrane expression levels in fixed HEK293 cells treated with the cell-impermeable SNAP substrate, BG-782 (Fig. 3B,C) 22,28,50 . Cells were also treated with nuclear stain TO-PRO-3 to normalize for cell number. We observed significant reductions in N65Q (13.40 ± 4.65%, mean ± SEM change from SNAP), N82Q (9.49 ± 5.95%, mean ± SEM change from SNAP), and NQQ (13.64 ± 5.76%, mean ± SEM change from SNAP) cell surface expression in comparison to WT SNAP-α 1D (46.13 ± 5.61%, mean ± SEM change from SNAP; p < 0.01, One-way ANOVA with Tukey's multiple comparisons post-hoc test). Combined, these data strongly indicate both N65 and N82 must be glycosylated to facilitate α 1D -AR plasma membrane insertion and agonist-stimulated functional responses in cultured human cells.
TM2 of α 1D -AR triggers ER translocation during protein synthesis. TM1 domain is thought to provide the ER localization signal for myriad polytropic integral membrane proteinsincluding some GPCRsduring protein synthesis 38 . Though, synthesis of TM2 has also been shown to trigger ribosomal translocation to the ER for some multi-pass transmembrane proteins, such as Cig30 67,68 and ProW 69 . Because the α 1D -AR NQQ mutant appears to cause early termination after TM2 (Figs. 1C,E,G,J and 2), we hypothesized that TM2 acts as the ER localization signal for α 1D -ARs. To test this, we utilized two orthogonal, but complementary, approaches: sucrose density gradient and confocal imaging.
Previous studies examining α 1 -AR subcellular localization used cell fractionation/sucrose density gradient to sequester distinct cellular compartments, and then radioligand binding to quantify the number of receptors present in each compartment sample 18 . Although useful, this method is only able to detect properly folded, functional receptors that are able to bind ligand; and has non-optimal signal-to-noise ratios 18 . Thus, sucrose density centrifugation protocols were modified to incorporate the sensitivity of SNAP-epitope tag PAGE NIR imaging analysis. This novel experimental approach allows accurate detection of poorly expressing α 1D -AR peptide species, regardless of their structural or functional state. Furthermore, the use of the SNAP epitope tag displays increased sensitivity compared to traditional immunoblotting techniques, which can be limited by the inability of antibodies to detect low expression levels of endogenous protein markers 70 . Thus, HEK293 cells were transfected with SNAP-α 1A -AR, which we have previously shown expresses readily at the plasma membrane 18 , or SNAP-Sec61β, an ER integral membrane protein 71 . Cells were lysed in detergent free buffer then conjugated to SNAP substrate BG-782. Labelled lysates were then fractionated in a discontinuous gradient (see methods for details), collected, and subjected to PAGE NIR analysis. In each case, the detectable SNAP signal from each isolated fraction was normalized to input. Data were analyzed by area under curve (AUC) to quantify the distribution of each SNAP protein in specific fractions. Figure 4 displays the PAGE NIR band pattern for SNAP-α 1A (Fig. 4A, Supplementary  Fig. S2A) and SNAP-Sec61β (Fig. 4B, Supplementary Fig. S2B). Subsequent AUC analysis revealed SNAP-Sec61β to be primarily distributed in fractions 1 through 4 with a peak in fraction 2 (91.50% total AUC; Fig. 4C), which is considered to be the ER fraction 18 . Conversely, SNAP-α 1A is significantly concentrated in fractions 6 through 9 with the maximum signal in fraction 7 (100% total AUC). Next, WT, N65Q, N82Q and NQQ SNAP-α 1D cDNA constructs were examined (Fig. 5A-D). As expected based on the findings of previous studies performed by us and others [16][17][18] , WT SNAP-α 1D (Fig. 5A, Supplementary  Fig. S3A) displayed a similar distribution pattern as SNAP-Sec61β, with a major peak spanning from fractions 1 to 4 (92.47% total AUC; Fig. 5E) and a minor peak in fractions 6 to 8 (7.53% total AUC). Similarly, N65Q SNAP-α 1D (Fig. 5B, Supplementary Fig. S3B) was bi-modally distributed, with peaks in fractions 1 to 3 (46.33% total AUC) and fractions 5 to 7 (53.67% total AUC; Fig. 5E). N82Q SNAP-α 1D (Fig. 5C, Supplementary Fig. S3C) was largely concentrated in fractions 1 through 4 with the maximum signal in fraction 2 (89.13% total AUC). A minor peak was also observed in fractions 6 to 7 (10.87% AUC; Fig. 5E). Remarkably, NQQ SNAP-α 1D (Fig. 5D,  Supplementary Fig. S3D) formed a single, strong peak spanning fractions 1 to 3, with the majority of the protein concentrated to the first fraction (100% total AUC; Fig. 5E), which corresponds with a primarily cytosolic localization.

conclusion
Our findings support a model in which TM2, not TM1, triggers ribosomal translocation to the ER during α 1D -AR synthesis [67][68][69]75 . Upon docking with the ER, the N-terminus is translocated into the ER lumen -possibly via the ER protein complex 38,76-78 -where glycosylation occurs. This event prevents the N-terminus from retrotranslocating back to the cytosol, which anchors the nascent peptide in the ER membrane in the proper membrane topology, such that the N-terminal will be within the extracellular matrix upon plasma membrane insertion. This event is required before complete translation of the nascent polypeptide continues. However, if glycosylation is prevented, the immature receptor does not anchor in the ER membrane, thus terminating receptor translation after TM2 (see Supplementary Fig. S4 for schematic); and presumably this degenerate polypeptide is degraded via ERAD 73 or other cytosolic degradation mechanisms 40,79 . Furthermore, we show that glycosylation of both N65 and N82 is required for proper function and plasma membrane expression of α 1D -ARs.

Materials and Methods
Plasmids and chemicals. Molecular cloning was performed using inFusion HD cloning technology www.nature.com/scientificreports www.nature.com/scientificreports/ rocking. Beads were washed 3X with lysis buffer, transferred to new 1.5 mL tubes and washed 3X with 20 mM Tris-HCl pH 8.0 and 2 mM CaCl 2 . After the final wash the saturated beads were incubated with 20 mM Tris-HCl pH 8.0 supplemented with 5 mM DTT for 30 min at 60 °C with agitation, followed by incubation with 15 mM iodoacetamide for 10 min at RT. Denatured protein was then incubated with 1.5 μg of Trypsin (Sigma, St. Louis, MO) and 1.5 μg of Glu-C endoprotease (Thermo Fisher Scientific, Waltham, MA) for 16 hr at 37 °C with vigorous agitation. Peptides were collected and acidified using formic acid (FA) to a final concentration of 1% FA and desalted using StageTips 80 . Peptides were eluted from StageTips using elution buffer (40% acetonitrile, 1% FA), dried down and re-suspended in 8% acetonitrile, 1% FA. Samples were then loaded on a self-pulled 360 µm OD x 100 µm ID 15 cm column with a 7 µm tip packed with 3 µm Reprosil C18 resin (Dr. Maisch, Germany). Peptides were analyzed by nanoLC-MS in a 90 minutes linear gradient from 6% to 38% buffer B (buffer A: 0.1% acetic acid; buffer B: 0.1% acetic acid, 80% acetonitrile) on an EASY nLC 1200 (Thermo Scientific, Rockford, IL) and Orbitrap Fusion Lumos Tribrid Mass Spectrometer (FTMS; Thermo Scientific, Rockford, IL). Orbitrap FTMS spectra (R = 60 000 at 200 m/z; m/z 350-1600; 7e5 target; max 20 ms ion injection time) and Top Speed data-dependent acquisition with 3 second cycle time; HCD MS/MS spectra (R = 30 000 at 200 m/z; 31% CE; 5e4 target; max 100 ms injection time) were collected with an intensity filter set at 2.5e4 and dynamic exclusion for 45 second. Mass spectra were searched against the UniProt human reference proteome downloaded on February 20th, 2020 with the addition of SNAP-tag-ADRA1D sequence using MaxQuant v1.6.10.43. Detailed MaxQuant settings: samples were set to fraction 1 and 5 for NQQ mutant and WT, respectively, to allow within-group "match between run"; Trypsin/P and Glu-C were selected in digestion setting. Other settings were kept as default.
Label free dynamic mass redistribution (DMR) assay. DMR assays were performed in 384-well Corning Epic microsensor plates (Corning, Corning, NY) using previously described protocols 22,28,50,81 . Data were analyzed using GraphPad Prism (La Jolla, CA). cell surface assay. Cell surface assay was performed as described previously 22,28,50 . Sucrose density centrifugation. Cells (~6.7 M cells/mL) were suspended in detergent-free lysis buffer (1 mM Tris-HCl pH 7.4, 140 mm NaCl, 10% sucrose) on ice for 20 min with vortexing every 5 min. 19 μL of lysate (~125,000 cells) was labeled with BG-782 at 37 °C. Reacted lysate was gently layered on top of a discontinuous sucrose gradient. Gradient consisted of equal volumes of 65%, 62.5%, 60%, 57.5%, 55%, 52.5%, 50%, and 15% sucrose dissolved in 1 mM Tris-HCl pH 7.4 and 140 mM NaCl. Samples were centrifuged at 134,633 × g at 4 °C for 65 min using a TH-660 rotor (Thermo Fisher Scientific, Waltham, MA). 400 μL fractions were collected and subjected to PAGE NIR analysis. Fluorescence of each fraction was quantified (NIR: λ = 800 nm) using Image Studio software (LI-COR, Lincoln, NE) and analyzed using area under the curve (AUC) analysis with a cut-off of 10% and minimum change in height of 5% from minimum to maximum in GraphPad Prism (La Jolla, CA). confocal microscopy. 48 hours after transfection, cells were fixed with 4% paraformaldehyde/PBS solution for 10 min. at room temperature, washed with PBS, and permeabilized in 0.1% TritonX-100/PBS for 1 min. Cells were incubated with 1 μM of SNAP Surface Alexa Fluor 488 (New England BioLabs #S9129S, Ipswich, MA) and 1:1000 ER Staining Kit-Red Fluorescence-Cytopainter (Abcam #139482, Cambridge, MA) at 37 °C for 30 min. protected from light. Hoechst 33342 was used for nuclear staining. Cover slips were mounted using ProLong Glass antifade reagent (Thermo Fisher #P36982). Confocal fluorescence microscopy was performed using Leica SP8X laser scanning confocal microscope equipped with a 40x oil immersion objective (Leica Camera, Wetzlar, Germany). The detection pinhole was set to 1 Airy unit, light collection configuration was optimized according to the combination of chosen fluorochromes (Alexa Fluor 488, Texas Red, and Hoechst), and sequential channel acquisition was performed to minimize the risk of bleed-through. The intensity gain was adjusted for each channel before capture in order to avoid saturated pixels. 8 bit, 1024 × 1024 pixel images were collected as Z-stack acquisition. All microscopy was performed in collaboration with the W.M. Keck Microscopy center on the University of Washington School of Medicine campus. colocalization analysis. The Alexa Fluor 488 (SNAP) and Texas Red (ER) channels were analyzed for colocalization using Coloc2 plugin for Fiji 82 . Pearson's coefficients for a cell were averaged over each slice in a Z-stack. Data were analyzed using GraphPad Prism (La Jolla, CA).