RAMP2 Influences Glucagon Receptor Pharmacology via Trafficking and Signaling

Endogenous satiety hormones provide an attractive target for obesity drugs. Glucagon causes weight loss by reducing food intake and increasing energy expenditure. To further understand the cellular mechanisms by which glucagon and related ligands activate the glucagon receptor (GCGR), we investigated the interaction of the GCGR with receptor activity modifying protein (RAMP)2, a member of the family of receptor activity modifying proteins. We used a combination of competition binding experiments, cell surface enzyme-linked immunosorbent assay, functional assays assessing the Gαs and Gαq pathways and β-arrestin recruitment, and small interfering RNA knockdown to examine the effect of RAMP2 on the GCGR. Ligands tested were glucagon; glucagonlike peptide-1 (GLP-1); oxyntomodulin; and analog G(X), a GLP-1/glucagon coagonist developed in-house. Confocal microscopy was used to assess whether RAMP2 affects the subcellular distribution of GCGR. Here we demonstrate that coexpression of RAMP2 and the GCGR results in reduced cell surface expression of the GCGR. This was confirmed by confocal microscopy, which demonstrated that RAMP2 colocalizes with the GCGR and causes significant GCGR cellular redistribution. Furthermore, the presence of RAMP2 influences signaling through the Gαs and Gαq pathways, as well as recruitment of β-arrestin. This work suggests that RAMP2 may modify the agonist activity and trafficking of the GCGR, with potential relevance to production of new peptide analogs with selective agonist activities.

1. The authors need to alter the wording relating to "internalisation" throughout the manuscript. This has not actually been measured in this paper. They should use cellular distribution or similar instead. The confocal data for example show the presence of receptor/RAMP inside cells but this was measured at a single point in time and not in response to ligand. Therefore the authors cannot conclude that there is internalisation, rather than the receptor simply not reaching the cell surface. i.e. the authors could equally conclude that the receptor is trapped inside the cell by the RAMP.
As suggested, internalisation has been rephrased throughout the manuscript. 2. The binding data are now much clearer. However, the authors should not refer to binding data as "total binding", if in fact it is specific binding. The relevant text and figures need to be changed to deal with this.
As suggested, total binding has been corrected to specific binding. The references have been corrected and rereviewed. 4. The wording for G(X) on page 5 is unclear. The motivation for using this ligand is less important than being clear on its actual sequence. Is this 1-15 glucagon together with 16-34 of exendin-4? Or is there more of a mix. Either reference the original sequence, explain more clearly or show the sequence in supplementary information. In order for any reader to reproduce the work, this is essential.
The sequence of G(X) has been further clarified, however, the actual sequence cannot be disclosed for intellectual property reasons.

Amendments: Methods page 5 line 112-119
5. There still seems to be some confusion around curve fitting parameters and inconsistency in the paper. The data from signalling assays appear to have been fit with a variable slope (4 parameter) but the equation implies that this was 3 parameter with a fixed slope of 1. This needs to be clarified.
The curve fitting for the biphasic curves was done using a variable slope (four parameters) model. This has been clarified in the methods section.
Amendments: Methods page 7 lines 201-203 6. All of the cell experiments use different conventions for reporting the cell number used, which is very confusing and does not allow The reporting of cell numbers has now been harmonised and numbers for cells/ml have been comparisons between assays. The actual cell number should be specified. Cells per mL is not acceptable if the volume pipetted into the well is not specified. recalculated to give cells/well. Amendments: Methods page 6 line 176, page 7 line 187 7. Do not use "u" for micro. Use the correct symbol. This occurs in numerous places throughout the document. This has been corrected throughout. 8. On page 14, the speculation about the work showing bias should be removed and left until the description on page 15, which is adequate. As it stands, the work does not show bias.
This has now been removed as suggested. 9. Table 1 should include statistical analysis.
Statistical analysis was performed however none of the comparisons were statistically significant. This is described further in the results section.

Amendments: Results Page 10 Lines 297-302
10. Table 3. If some of these experiments are less than n=3 as is suggested by the legend, then there should not be statistical analysis on these data.
Statistical analysis was only carried out for experiments where n≥3.

Reviewer 2
1. The additional data that has been added improves the work and helps somewhat to support the interpretations. It will still be very important to quantify the GCGR on the cell surface by a method other than "total binding relative to control". Since you have performed competition binding of GCG radioligand and included IC50 data in Table 1, the homologous competition curves should be analyzed for Ki and Bmax values. This will be much more important and informative than what is shown in Figure 1.
As explained above, Bmax has now been calculated for a second independent cell line. Ki is proportional to IC50 and therefore will be comparable with and without RAMP2. 2. Essentially all of the functional data presented can be explained by lower surface GCGR expression in the RAMP2 positive cell line than in the control cell line. It will be important to prepare and characterize several clonal lines with RAMP2 coexpression to be certain that the very low levels of GCGR in this line is reflective of the coexpression itself and not an artifact of the single clonal line studied.
To ensure that these findings were attributable to co-expression of RAMP2 with the GCGR, rather than artefactual, a second cell line with RAMP2 stably upregulated was investigated (CHO-K1-GCGR-CFP-RAMP2) and compared to a cell line transfected in parallel with a control (pcDNA3.1) plasmid. Importantly, these findings corroborate those described in the first cell line. 3. It would be quite interesting to start with a stable RAMP2-expressing CHO cell line and a non-RAMPexpressing CHO cell line and make stable GCGR-The reviewer makes a good point, although we feel that this extensive work-not requested during the expressing lines from both via independent transfection to try to achieve clonal lines with similar surface GCGR expression. That would be quite interesting to study functionally. I understand that this would represent a substantial new effort. first round of review-is better served in another manuscript. Ideally, we would use CRISPR-Cas9 to delete/replace the endogenous loci in a beta or hepatocyte cell line, thus leading to stable and physiological GCGR expression levels in the presence or absence of RAMP. We have discussed this in the revised manuscript.

Amendments: Discussion page 15 line 471-474
4. The data set that best seems to support a mechanism for the reduction in cell surface receptor when RAMP gets coexpressed is figure 6 in which morphology is done on transfected HEK cells, and the RAMP-transected cells clearly had fluorescent receptor internalized. I fear that there could be glucagon in the serum used to culture these cells and the agonist-stimulated internalization response could be amplified in the presence of the RAMP. This could be quite interesting, but would require additional experiments. It would be helpful to perform this and other experiments in the absence of serum that could contain glucagon or other agonist. Another way to achieve this would be to perform the work in the presence of glucagon antagonist.
The GCGR experiments performed in HEK cells were designed to corroborate the radioactive binding assays, rather than a complete work-up of receptor internalization mechanisms. In any case, we would not expect meaningful glucagon levels in FBS, since the fetus secretes high levels of insulin to counteract elevated maternal glucose concentration, and this would strongly inhibit alpha cell function. Additionally, glucagonstimulated GCGR internalization is known to be a high dose phenomenon, occurring at ≥1 µM (Roed et al JBC 2015), far above the known glucagon concentration in calves of 30-40 pM (Bloom et al, J Physiol, 1974). Thus the results are likely due to the interaction between non-bound/non-activated receptor and RAMP. The reviewer makes a reasonable point, however, and we discuss the requirement for future experiments with a GCGR antagonist or serum-free medium, or alternatively excess glucagon, in the revised manuscript.

Amendments: Discussion page 15 line 451-454
5. I am concerned that the differences being reported here from the literature may be a function of low level receptor expression that reflects clonal choice and/or hormone-stimulated internalization. I would like to see three additional sets of data to be convinced this is real. 1) independent clonal lines for the CHO-GCGR-RAMP2 expression to be sure the low level of surface receptor is real; 2) these lines need not only IC50 data for receptor binding, but also Bmax data to quantitatively determine receptor density; and 3) use of non-serum-containing medium or use of GCG antagonist to block agonist effect to downregulate the receptor.
As described above, we have completed additional experiments to address points 1) and 2). 1) Independent clonal lines for CHO-GCGR-RAMP2 expression confirm that the low level of surface receptor is real; 2) these lines provide IC50 data for receptor binding, but also Bmax data to quantitatively determine receptor density. In terms of 3), we have inserted a caveat into the discussion re. GCGR antagonist, although as explained in response to the point above, the glucagon levels in FBS are expected to be far below the levels which have been demonstrated to be necessary to observe agonist-downregulation of the glucagon receptor.  We have used a combination of competition binding experiments, cell surface ELISA, functional 46 assays assessing the Gαs and Gq pathways and β-arrestin recruitment, and siRNA knockdown to 47 examine the effect of RAMP2 on the GCGR. Ligands tested were glucagon, glucagon-like peptide-1 48 (GLP-1), oxyntomodulin and analogue G(X), a GLP-1/glucagon co-agonist developed in-house. 49 Confocal microscopy was employed to assess whether RAMP2 affects the subcellular distribution of 50

GCGR. 51 52
Here we demonstrate that co-expression of RAMP2 and the GCGR results in reduced cell surface 53 expression of the GCGR. This was confirmed by confocal microscopy, which demonstrated that 54 RAMP2 co-localises with the GCGR and causes significant GCGR cellular redistribution. 55 Furthermore, the presence of RAMP2 influences signalling through the Gαs and Gαq pathways, as 56 well as recruitment of β-arrestin. This work suggests that RAMP2 may modify the agonist activity 57 and trafficking of the GCGR, with potential relevance to production of new peptide analogues with 58 selective agonist activities. Gut and pancreatic hormones involved in appetite regulation are an attractive target for the 64 development of drugs that aim to cause effective weight loss with minimal side effects. Glucagon has 65 been shown to potently increase satiety and acutely reduce food intake in humans (1). Additionally, 66 glucagon significantly increases energy expenditure in man (2-4). This, in association with the 67 anorectic effects of glucagon (1), enhances its usefulness as an anti-obesity therapy. 68

69
The glucagon receptor (GCGR) is a 7 transmembrane class B G-protein coupled receptor (GPCR). It 70 classically activates adenylyl cyclase through Gαs with subsequent activation of protein kinase A 71 (PKA) signalling (5,6). In hepatocytes, elevated PKA activity suppresses glycolysis and glycogen 72 synthesis, and enhances gluconeogenesis and glycogenolysis (7,8). However activation of GCGR also 73 stimulates the phospholipase C-inositol phosphate pathway in hepatocytes via Gq, inducing 74 intracellular calcium (Ca 2+ ) signalling and stimulating glycogenolysis and gluconeogenesis (6,9). 75 Although work to unpick glucagon signalling pathways has been underway since the 1970s, it has 76 focussed primarily on understanding the interactions involved in the downstream effects in the liver 77 and the pancreas. Less attention has been paid to the role of specific pathways in the extrahepatic 78 roles of glucagon, namely in appetite regulation and control of energy expenditure. As a prototypical 79 class B GPCR, the GCGR is desensitised and sequestered in the cytosol following activation (10-12). 80 The internalised receptor is then either recycled to the cell surface or targeted for degradation. Krilov 81 et al recently demonstrated that the GCGR recycles to the plasma membrane in a β-arrestin-dependent 82 manner, and that downregulation of β-arrestins significantly reduces recycling (13,14). 83

84
Understanding the interaction of these pathways may allow 'biasing' of signalling to exploit desirable 85 downstream effects (15,16). A particularly well characterised example of an accessory protein that 86 clearly alters the pharmacology of GPCRs is a family of single transmembrane proteins known as 87 Receptor Activity Modifying Proteins (RAMPs). RAMPs were discovered as proteins that interact 88 with the calcitonin receptor-like receptor (CRLR) and calcitonin receptor (CTR) to give rise to 89 RAMPs to influence downstream signalling pathways is an exciting concept, as it may enable the 96 creation of biased agonists that fully exploit the therapeutic potential of clinically important receptors. 97

98
The functional impact of RAMPs on GCGR pharmacology is not clearly understood. Over ten years 99 ago, the Christopoulos group showed that the GCGR may interact with RAMP2 (27). Recently, one 100 study has found that RAMP2 may alter GCGR ligand selectivity and G protein preference using yeast 101 reporter systems (29). The work presented here is concerned with further understanding the effect of 102 RAMP2 on the pharmacology of the GCGR in mammalian cells. Human GCG, GLP-1 and OXM were purchased from Bachem, Ltd. (UK). GLP-1(7-36)NH2 was the 109 form used in all experiments, and will now be referred to simply as GLP-1. A dual glucagon/GLP-1 110 analogue, G(X), was designed in the Department of Investigative Medicine, Imperial College London 111 and custom synthesised using solid-phase peptide synthesis (Bachem Ltd). G(X) contains identical 112 amino acid sequences to glucagon from positions 1 to 15 as the N-terminal of glucagon has been 113 shown to be critical for glucagon receptor binding and activation (30). To create a dual agonist that is 114 also effective at the GLP-1 receptor, G(X) has been modified to resemble exendin-4. This peptide, 115 first isolated from the venom of the lizard Heloderma species, has been found to be a potent agonist at 116 the human GLP-1 receptor (31,32). Also favorable is its prolonged pharmacokinetic profile compared 117 to native GLP-1. Therefore, from positions 16-34, amino acid substitutions have been made to 118 resemble exendin-4. 119 120

Establishing a cellular co-expression system for RAMP2 and GCGR 121
Chinese hamster ovarian (CHO-K1 cells; GeneBLAzer® GCGR-CRE-bla CHO-K1 cells; K1855A)) 122 (Invitrogen) cells expressing the GCGR were cultured in DMEM supplemented with 10% FBS, 0.1 123 mM non-essential amino acids, 25 mM HEPES (pH7.3), 100 IU/ml penicillin, 100 μg/ml 124 streptomycin and 5 μg/ml blastocidin. This cell line expressed no background RAMP2, as confirmed 125 using QPCR (CT values >32). The human RAMP2 DNA construct (pCMV6-AC-RAMP2) (Origene, 126 USA) was transfected into CHO-K1 cells expressing the human GCGR using polyethylenimine (PEI,127 Sigma) (33). The cells were transfected with pCMV6-AC-RAMP2 (containing a neomycin resistance 128 gene) and 9 nitrogen equivalents of PEI. Forty-eight hours later, media was supplemented with 800 129 μg/ml Geneticin to select cells containing the construct. 130 To establish a second independent cell line stably expressing RAMP2, CHO-K1 cells expressing the 131 human GCGR were co-transfected with C-terminally CFP-tagged RAMP2 (Tebu-bio Ltd, UK) and a 132 plasmid conferring puromycin resistance using lipofectamine 2000 (Thermo Fisher). Forty-eight 133 hours later, media was supplemented with puromycin 10 μg/ml to select cells containing the construct. Cells were grown up to 70% confluence and resuspended in 1.5 ml assay buffer (25 mM HEPES (pH 143 7.4), 2 mM MgCl2, 1% BSA, 0.05% (w/v) Tween 20, 0.1 mM diprotin A and 0.2 mM PMSF). 50 µl 144 of I 125 -glucagon dissolved in assay buffer at 1000 counts per second (final concentration 5.6 nM), 145 unlabelled peptide made up in 400 µl of assay buffer and 50 µl of the cell suspension was added to 146 each microtube, vortexed and incubated at room temperature for 90 minutes. Microtubes were then 147 centrifuged (15781 x g, 4°C, 3 minutes), supernatant removed, 500µl of assay buffer added, and then 148 re-centrifuged. The supernatant was again discarded and the pellets measured for γ radiation for 240 149 seconds (Gamma counter NE1600, NE Technology Ltd, UK). The specific binding (maximal specific 150 binding minus the non-specific binding) was calculated for each cell line. The binding data was 151 normalised so that the maximal specific binding (i.e. when no unlabelled peptide was present) was 152 100%. The percentage specific binding was calculated for each peptide concentration as a percentage 153 of the specific binding. The half-maximal inhibition concentrations (IC50), a measure of binding 154 affinity, were then calculated and compared for CHO-K1-GCGR and CHO-K1-GCGR-RAMP2 cells. 155 IC50 values were calculated using the Graphpad Prism 5.01 (GraphPad Software Inc., USA) using the 156 following regression fit line: 157 Where Y=% specific binding and X=concentration of the agonist. 159 To calculate receptor density (Bmax), binding data was normalised to protein content of the cell 160 samples, as determined by a bicinchoninic acid assay (Sigma). Bmax was then calculated for using 161 GraphPad Prism 7.0b (GraphPad Software Inc., USA) using the following regression fit line: were incubated for 30 minutes with the test peptide after which media was replaced with 110 µl lysis 179 buffer (0.1M HCl with 0.5% Triton-X). The lysate was assayed using a direct cyclic AMP ELISA kit 180 (Enzo Life Sciences, UK), as described in the assay manual. The cAMP response was corrected for 181 well protein levels (Bradford reagent, Sigma) and expressed as a percentage of response to 10 µM 182 forskolin. 183 184 Human hepatoma 7 cells overexpressing the human GCGR (Huh7-GCGR) were cultured in DMEM 185 supplemented with 10% FBS, 100 IU/ml penicillin, 100 μg/ml streptomycin and 10 μg/ml geneticin 186 (standard maintenance media). They were plated onto 96 well plates at 20,000 cells/well in standard 187 maintenance media with transfection reagents for gene silencing (see details below). After 72 hours, to determine the effect of RAMP2 on the potency of GCGR ligands for recruitment of β-Arrestin-1 to 218 the GCGR. The CHO-K1-βArr-GCGR cells are engineered to detect the interaction of β-arrestin with 219 the activated GCGR using β-galactosidase fragment complementation. CHO-K1-βArr-GCGR cells 220 were stably transfected ± RAMP2, as described above. Cells, plated at 100 µl/well into a 96-well plate 221 were incubated with glucagon, GLP-1, oxyntomodulin, or G(X) (10 µl) for 90 minutes at 37°C and 222 5% CO2. 55 µl of the PathHunter TM detection reagents was added to each well and the microplate was 223 incubated at room temperature for 60 minutes. Ltd, UK). A first set of experiments was carried out using a Crest X-Light spinning disk system 255 coupled to a Nikon Eclipse Ti microscope and a 63x 1.4 NA oil immersion objective. GFP was 256 excited using a solid-state at λ = 491 nm laser (Cobalt) and emitted signals collected at λ = 525/25 nm 257 using a highly-sensitive Orca-Flash4.0 Digital CMOS camera. Due to bleedthrough of the intense 258 GFP signal into the CFP channel at λ = 440 nm, the latter fluorophore was instead excited slightly off-259 peak using a solid-state 405 nm laser and emitted signals collected at λ = 525/25 nm. A second set of 260 experiments was performed using a Zeiss LSM780 confocal microscope and a 63x 1.2 NA water 261 immersion objective. GFP and CFP were excited using a λ = 488 nm argon laser and emitted signals 262 collected at λ = 510 -550 nm using a GaAsP spectral detector. CFP was excited using a λ = 405 nm 263 diode laser and emitted signals collected at λ = 455-490. Images were post-processed using Zen 264 software (Zeiss, UK) and subjected to Gaussian smoothing (1.3) to remove noise. Uniform linear 265 adjustments were applied to contrast and brightness to improve image quality for analysis and 266 presentation purposes, while preserving the pixel dynamic range and the intersample intensity 267 differences. Cell surface expression of GCGR-GFP was calculated using the threshold plugin for 268 ImageJ (NIH).

RAMP2 reduces specific glucagon binding at the GCGR 292
When specific glucagon binding to the GCGR was compared in RAMP2 positive and negative CHO-293 K1 cells, it was found to be 10-fold lower in the presence of RAMP2 (see Figure 1A). This was 294 despite the protein content being similar in both groups (see Figure 1B). 295 296 Glucagon bound to the GCGR with an IC50 of 1.403 nM. This was not significantly altered when the 297 GCGR was co-expressed with RAMP2 ( Figure 1C, table 1). As expected, GLP-1 had poor affinity for 298 the GCGR with an IC50 of >10000 nM ( Figure 1D). Oxyntomodulin and analogue G(X) showed a 7-299 fold and 2.5 fold lower affinity for the GCGR than the native peptide, respectively ( Figures 1E and F). 300 Similar to glucagon, the presence of RAMP2 had no effect on the binding affinity at the GCGR for 301 GLP-1, oxyntomodulin or analogue G(X). 302

303
To ensure that these findings were attributable to co-expression of RAMP2 with the GCGR, a second 304 independent cell line with RAMP2 stably upregulated was investigated (CHO-K1-GCGR-CFP-305 RAMP2) and compared to a cell line transfected in parallel with a control (pcDNA3.1) plasmid. As 306 with the first cell line (CHO-K1-GCGR-RAMP2), the binding affinity of glucagon for its receptor 307 was not altered with the upregulation of RAMP2 (IC50 4.377nM with CFP-RAMP2 vs 5.123nM 308 without (p=0.16)) (Supplemental Figure 2A), however the density of GCGR binding sites (Bmax) was 309 significantly lower in the cell line with upregulated RAMP2 (p=0.0069) (Supplemental Figure 2B). 310

RAMP2 reduces cell surface expression of the GCGR 312
Using an in-cell ELISA, surface GCGR expression was detected in non-permeabilised CHO-K1-313 Figure 3). GCGR cell surface expression was significantly 314 reduced in cells expressing RAMP2. 315

RAMP2 reduces potency and increases efficacy of the Gαs pathway at the GCGR 317
To assess whether RAMP2 affected the Gαs pathway, cAMP accumulation was measured in its 318 presence/absence in CHO-K1 cells (Figure 2, Table 2). In control cells, the highest concentrations of 319 peptide resulted in cAMP accumulation lower than the Emax, which is a well described desensitisation 320 effect (14). In the presence of RAMP2, glucagon, oxyntomodulin and analogue G(X) increased the 321 EC50 i.e. RAMP2 reduced the potency of these ligands for GCGR Figure 2A, C and D). When the 322 GCGR was stimulated by oxyntomodulin or analogue G(X), the Emax (efficacy) was increased in the 323 presence of RAMP2. The EC50 and Emax were not calculable for GLP-1 response at the concentrations 324 used ( Figure 2B). There was no significant difference in cAMP responses to forskolin between control 325 and RAMP2 expressing cells

RAMP2 reduces efficacy of the Gq pathway at the GCGR 337
To assess the effect of RAMP2 on the Gαq pathway, intracellular Ca 2+ flux was measured in real time 338 in CHO-K1 cells. For glucagon and oxyntomodulin, the Ca 2+ response was attenuated when cells 339 expressing the glucagon receptor were co-expressed with RAMP2, as demonstrated by a significantly 340 lower Emax ( Figure 3A, B, C and E). RAMP2 also appeared to lower the response to G(X), however, 341 as the maximal Ca 2+ response was not achieved with cells expressing GCGR alone, and Emax could not 342 be determined ( Figure 3F). Similarly, the EC50 and Emax were not calculable for GLP-1 response at the 343 concentrations used ( Figure 3D). EC50 was unchanged in the presence of RAMP2 for all ligands 344 (Table 2). There was no significant difference in Ca 2+ responses to ATP between control and RAMP2 345 expressing cells (RFU fold increase from baseline 1.81 (± 0.08) vs. 1.84 (± 0.10) respectively; 346 p=0.77) (Figures 3A and B). 347

RAMP2 knockdown partially restores GCGR functioning for the Gαs and Gq pathways 353
Efficient siRNA knockdown of RAMP2 was achieved with both 10 nM and 50 nM siRNA pools 354 ( Figure 5A). siRNA knockdown of RAMP2 in CHO-K1-GCGR-RAMP2 cells resulted in a trend 355 toward restoration of cAMP EC50 and Emax to levels seen with control cells (CHO-K1-GCGR cells), 356 however, they were not significantly different to control or RAMP2 (without siRNA) cells ( Figure  357 5B). A similar finding was demonstrated for Ca 2+ fluxes ( Figure 5C). The EC50 and Emax data is 358 summarised in Table 3. 359

360
The GCGR and RAMP2 colocalise and the GCGR is internalised in the presence of RAMP2 361 High resolution confocal microscopy showed that GCGR-GFP and RAMP2-CFP co-localised as 362 puncta within the cytosol of HEK293 ( Figure 6A). In cells where RAMP2 was not overexpressed, 363 GCGR-GFP remained predominantly at the cell surface/membrane ( Figure 6B). This was not due to 364 bleedthrough of GCGR-GFP fluorescence into the RAMP2-CFP channel, since signal could not be 365 detected in RAMP2 negative/GCGR positive cells ( Figure 6C). Overexpression of non-native protein 366 (pcDNA3.1) did not interfere with the distribution of the GCGR-GFP, which remained almost 367 exclusively at the membrane ( Figure 6D), whereas non-tagged RAMP2 led to a significant decrease in 368 receptor at the cell membrane (Fig. 6E). This demonstrates that protein expression per se is unlikely 369 to interfere with GCGR localisation. Thus, overexpression of RAMP2-CFP or RAMP2 consistently 370 leads to a decrease in cell surface GCGR-GFP (Fig. 6F). It has previously been demonstrated by immunofluorescence confocal microscopy that RAMP2 may 375 interact with the glucagon receptor. We have investigated the functional effect of this possible 376 interaction by looking specifically at the effect of RAMP2 on: 1) ligand binding at the GCGR; 2) 377 GCGR cell signalling; and 3) GCGR subcellular distribution. Co-expression of RAMP2 with GCGR 378 did not alter the binding affinity of glucagon or its related peptides. However, the presence of RAMP2 379 had a marked effect on signalling via the Gαs and Gq pathways, as well as β-arrestin recruitment. 380 Furthermore, RAMP2 appears to co-localise with the GCGR and influence its subcellular distribution. 381

382
Interaction between calcitonin family receptors and the individual RAMP proteins alters both ligand 383 binding affinity and the intracellular signalling pathways engaged (17,35,36). By contrast, we found 384 that expression of RAMP2 with the GCGR did not cause a significant alteration in the binding affinity 385 of glucagon and its related peptides in whole cells. However, competition binding experiments using 386 125 I-glucagon as the radioligand revealed that co-expression of RAMP2 resulted in a ten-fold 387 reduction in GCGR binding sites when compared with those determined in the absence of RAMP. 388 This reduction in specific binding of glucagon may be due to reduced receptor expression at the cell 389 surface. This could have been a direct effect of the interaction of RAMP2 and the GCGR resulting in 390 internalisation. Alternatively, it might be an indirect effect if, for example, RAMP2 influences GCGR 391 cell surface expression via its effect on β-arrestin recruitment. 392

393
The presence of RAMP2 completely abolished β-arrestin recruitment. This finding was consistent for 394 glucagon as well as GLP-1, oxyntomodulin and G(X). One possible explanation is that RAMP2 395 interacts with the GCGR at the same site as β-arrestin binds or causes steric hindrance, thereby 396 disrupting β-arrestin recruitment. Krilov et al have shown that β-arrestins are crucial for the recycling 397 of the GCGR (13) and, therefore, loss of β-arrestin recruitment may result in reduced cell surface 398 expression of the GCGR when RAMP2 is present. Alternatively, reduced cell surface expression of 399 GCGR may be the primary effect of RAMP2 and this may in turn prevent β-arrestin recruitment. 400 401 Co-expression of RAMP2 with the GCGR also altered the intracellular signalling properties of the 402 receptor in CHO-K1-GCGR cells, with the same effects seen for all agonists tested. With regards to 403 the Gαs pathway, the presence of RAMP2 caused a reduction in potency and increase in efficacy. In 404 Huh7-GCGR cells, the knockdown of RAMP2 resulted in no change in potency and a trend towards 405 decreased efficacy. Whether this is a result of a change in availability of binding sites is yet to be 406 determined. In contrast to our findings, Weston et al found that RAMP2 increases potency of the 407 cAMP response at the GCGR (29). One possible explanation for these different findings could be the 408 different cell lines used. Weston et al overexpressed RAMP2 in HEK cells that already express 409 endogenous RAMP2, whereas we overexpressed RAMP2 in CHO-K1 cells that do not express 410 RAMP2. It has previously been shown that interaction of the CTR with RAMPs, especially RAMP2, 411 is sensitive to the cellular background in which it is expressed, suggesting that other cellular 412 components, such as G proteins, are likely to contribute to RAMP-receptor interactions (36). 413

414
The increase in efficacy of cAMP production observed with RAMP2 is intriguing. This enhancement 415 in cAMP response is all the more striking as it is in the face of an apparent reduction of cell surface 416 expression of GCGR. The simplest interpretation is that by some mechanism, RAMP2 increases the 417 accessibility of the receptor to the G-protein (37). Alternatively, RAMP2 may inhibit the 418 desensitisation response that is classically seen with the GCGR, involving phosphorylation of 419 receptors by GPCR kinases (GRKs) and binding of β-arrestins, which uncouple receptors from G-420 proteins (38). We speculate that the GCGR-RAMP2 interaction causes loss of desensitisation, which On examination of the Gq pathway, intracellular Ca 2+ fluxes were found to be attenuated in 426 the presence of RAMP2. Interestingly, preferential coupling to Gαs versus Gq has been reported for 427 AMY1 and AMY3 receptors, but not AMY2 (39). The finding that cAMP signalling is specifically 428 augmented and Ca 2+ signalling attenuated by RAMP2 at the GCGR is important because the classic 429 coupling pathway associated with GCGR activation has always been thought to be the stimulation of 430 cAMP accumulation. Moreover, the presence or absence of endogenous RAMP2 may account for 431 discrepancies in previous studies examining the signalling mechanisms engaged by the GCGR. 432 Whether this is tissue-specific and dependent on the prevailing physiological conditions is yet to be 433 seen. 434

435
Visualisation of RAMP2 and the non-ligand bound GCGR using confocal microscopy revealed two 436 key findings. Firstly, it is demonstrated that RAMP2 and the GCGR show some co-localisation, 437 although super-resolution approaches will be needed to confirm this, as well as delineate the 438 compartment(s) involved. Secondly, in the presence of RAMP2, there was reduced GCGR cell 439 surface expression. This is consistent with the competition binding and ELISA experiments, which 440 found reduced binding of 125 I-GCG in the presence of RAMP2. These findings appear to be at odds 441 with the work done by Christopoulos et al which reported that, when co-expressed with GCGR, 442 RAMP2 translocates to the cell surface. A number of differences exist in the experimental approach 443 between this current study and that of Christopoulos. Firstly, in their study only the RAMPs, and not 444 the GCGR, were tagged so it was not possible to comment on where the receptor was trafficked to. 445 Secondly, in the Christopoulos study, RAMP2 was N-terminally tagged with haemagglutinin whereas 446 here both C-terminally CFP-tagged and native RAMP2 was utilised. It is the N-terminal that contains 447 the natural, predicted signal peptide sequence of RAMP2 and therefore this may have had a bearing 448 on expression of RAMP2. In line with our findings, using C-terminal receptor-fluorescent protein 449 fusion constructs and cell surface ELISAs of myc-tagged receptors, Weston et al found that 450 expression of RAMP2 caused a non-significant decrease in cell surface expression of GCGR (29). To 451 ensure that the agonist-stimulated internalization response is not due to glucagon in the serum used to 452 culture these cells, with amplification in the presence of RAMP2, further experiments could be 453 performed with a GCGR antagonist or serum-free medium, or alternatively excess glucagon. access to all the data, and takes full responsibility for the integrity of data and the accuracy of data 506 analysis. 507 508