The C-terminal Helix 9 motif regulates cannabinoid receptor type 1 trafficking and surface expression

Cannabinoid type 1 receptor (CB1R) is only stably surface expressed in axons, where it downregulates neurotransmitter release. How this tightly regulated axonal surface polarity is established and maintained is unclear. To address this question, we used time-resolved imaging to determine the trafficking of CB1R from biosynthesis to mature polarised localisation. We show that the secretory pathway delivery of CB1R is axonally biased and that surface expressed CB1R is more stable in axons than in dendrites. This dual mechanism is mediated by the CB1R C-terminal and involves the Helix 9 (H9) domain. Removal of the H9 domain increases dendrite secretory pathway delivery and decreases in surface stability. Furthermore, CB1RΔH9 is more sensitive to agonist-induced internalisation and less efficient at downstream signalling than CB1RWT. Together, these results shed new light on how polarity of CB1R is mediated and indicate that the C-terminal H9 domain plays key roles in this process.

Introduction 18 CB1R is one of the most abundant G-protein-coupled receptors (GPCRs) in the CNS and 19 endocannabinoid signalling through CB1R is a neuromodulatory system that influences a wide range 20 of brain functions including pain, appetite, mood, and memory (Lu and Mackie, 2016;Soltesz et al., 21 2015). Furthermore, CB1R function and dysfunction are implicated in multiple neurodegenerative 22 disorders (Basavarajappa et al., 2017). Thus, modulation of endocannabinoid pathways is of intense 23 interest as a potential therapeutic target (Reddy, 2017). CB1R is present in both excitatory and inhibitory neurons, and also in astroglia, where it plays 25 important roles in synaptic plasticity and memory (Busquets-Garcia et al., 2018;Han et al., 2012;26 Robin et al., 2018). In hippocampal neurons, ~80% of CB1R is present in intracellular vesicular 27 clusters in the soma and dendrites (Leterrier et al., 2006). Strikingly, however, CB1R is not stably 28 surface expressed on somatodendritic plasma membrane. Rather, it has a highly polarised axonal 29 surface expression (Coutts et al., 2001;Irving et al., 2000) where it acts to attenuate neurotransmitter 30 release (Katona, 2009) and modulate synaptic plasticity (Lu and Mackie, 2016). 31 How this near exclusive axonal surface expression of CB1R is established remains the subject of 32 debate. One suggestion is that high rates of endocytosis due to constitutive activity selectively 33 remove CB1Rs from the somatodendritic compartment, resulting in an accumulation at the axonal 34 surface (Leterrier et al., 2006). These internalised somatodendritic CB1Rs may then be either sorted 35 for degradation or recycled to axons via a transcytotic sorting pathway (Simon et al., 2013). 36 Alternatively, newly synthesized CB1R may be constitutively targeted to lysosomes, but under 37 appropriate circumstances the CB1Rs destined for degradation are retrieved and rerouted to axons 38 (Rozenfeld, 2011;Rozenfeld and Devi, 2008). 39 Surprisingly, a direct role for the 73-residue intracellular C-terminal domain of CB1R (ctCB1R) in 40 axonal/somatodendritic trafficking or polarised surface expression has not been identified. It has, 41 however, been reported that motifs within ctCB1R are required for receptor desensitization and 42 internalization (Hsieh et al., 1999;Jin et al., 1999) (reviewed by (Mackie, 2008)). Interestingly, there 43 are two putative helical domains in ctCB1R (H8 and H9). H8 has been proposed to play a role in ER 44 assembly and/or exit during biosynthesis (Ahn et al., 2010;Stadel et al., 2011). The role of the 21-45 residue H9 motif is unknown, although analogous regions have been reported to act as a Gαq-binding 46 site in both squid rhodopsin (Murakami and Kouyama, 2008) and bradykinin receptors (Piserchio et 47 al., 2005). 48 Here we systematically investigated how axonal surface polarity of CB1R arises by tracking newly-49 synthesised CB1Rs through the secretory pathway to their surface destination. We demonstrate that 50 a population of CB1R is preferentially targeted to the axon through the biosynthetic pathway. CB1Rs 51 Moreover, although still significantly axonally polarised, the degree of polarisation was significantly 129 lower for CD4-ctCB1R ΔH9 , suggesting that H9 may also contribute to this process. 130 H9 restricts delivery of CB1R to the dendritic membrane. 131 To further explore the possibility that H9 is involved in the axonal surface polarity of CB1R, we used 132 RUSH to compare the forward trafficking of SBP-EGFP-CB1R WT and SBP-EGFP-CB1R ΔH9 . As in 133 Fig. 1, we labelled all the CB1R that had been surface expressed (surface+endocytosed) 0, 30, 60 134 and 90 min after biotin release from the ER (Fig. 3A-G). 135 Interestingly, significantly more SBP-EGFP-CB1R ΔH9 than SBP-EGFP-CB1R WT reached the surface 136 of dendrites during time course of our experiments (Fig. 3B), whereas trafficking to axons was similar 137 for both SBP-EGFP-CB1R WT and SBP-EGFP-CB1R ΔH9 (Fig. 3C). These altered properties resulted 138 in a significant difference in the surface+endocytosed polarity index after 90 min (Fig. 3D) and are 139 consistent with a role for H9 in restricting delivery of CB1R to the dendritic membrane. 140 H9 plays a role in the surface retention of CB1R. 141 Surprisingly, however, in contrast to the total amount of CB1R that had been surface expressed 142 during the time course (surface+endocytosed) (Fig. 3D) the polarity of the amount of CB1R on the 143 cell surface 90 min after biotin-mediated release was identical for SBP-EGFP-CB1R WT and SBP-144 EGFP-CB1R ΔH9 (surface; Fig. 3E). 145 Closer analysis revealed identical levels of axonal surface expression of both SBP-EGFP-CB1R WT 146 and SBP-EGFP-CB1R ΔH9 60 min after release from the ER. However, at 90 min there is significantly 147 less surface expression of ΔH9 mutant (Fig. 3F) suggesting that, although similar amounts of SBP-148 EGFP-CB1R WT and SBP-EGFP-CB1R ΔH9 reach the surface, surface expression of SBP-EGFP-149 CB1R ΔH9 is less stable than that of the wild-type. 150 Furthermore, in dendrites, the increased delivery and surface trafficking of the ΔH9 mutant is 151 counteracted by the fact that less is retained at the surface 60 min after ER release (Fig. 3G). 152 Taken together these results suggest that, separate from its role in restricting delivery to the dendritic 153 membrane, H9 also plays a role in membrane stability and retention at both axons and dendrites.

H9 stabilises CB1R at the surface. 155
To investigate the role of H9 in membrane stability, we next compared surface expression (Fig. 4A) 156 and endocytosis (Fig. 4B) of EGFP-CB1R WT and EGFP-CB1R ΔH9 in axons and dendrites at steady-157 state. EGFP-CB1R ΔH9 displayed lower levels of surface expression (Fig. 4C), as well as increased 158 endocytosis (Fig. 4D) in both axons and dendrites compared to EGFP-CB1R WT , suggesting H9 plays 159 a role in stabilising CB1R at the surface of both axons and dendrites. Moreover, similar to our findings 160 using RUSH, there was there was no difference in surface polarity between wild-type and EGFP-161 CB1R ΔH9 (Fig. 4E). These findings suggest that, while H9 plays a role in CB1R surface expression 162 and endocytosis, its potential to mediate surface polarity is masked in the context of the full-length 163 receptor. 164 CB1R ΔH9 is less efficient at activating downstream signalling pathways and more susceptible 165 to agonist-induced internalisation. 166 Because CB1R surface expression and polarisation has been linked to its activity (Ladarre et al., 167 2014;Leterrier et al., 2006), we next investigated if deleting H9 affects CB1R downstream signalling 168 pathways. We expressed EGFP-CB1R WT or EGFP-CB1R ΔH9 in HEK293T cells, which contain no 169 endogenous CB1R (Atwood et al., 2011), stimulated with the selective CB1R agonist ACEA 170 (arachidonyl-2'-chloroethylamide) (Hillard et al., 1999) and monitored ERK1/2 phosphorylation as a 171 measure of signalling downstream of CB1R (Daigle et al., 2008). There was no significant difference 172 in ERK1/2 phosphorylation in cells expressing EGFP-CB1R WT or EGFP-CB1R ΔH9 under basal 173 conditions in the absence of ACEA. However, upon ACEA stimulation, the level of ERK1/2 activation 174 was significantly reduced in EGFP-CB1R ΔH9 -transfected cells compared to EGFP-CB1R WT -175 transfected cells expressing equivalent amounts of receptor ( Fig. 5A-C), suggesting the ΔH9 mutant 176 is deficient in its ability to activate downstream signalling pathways. 177 We next monitored ACEA-induced internalisation of EGFP-CB1R WT and EGFP-CB1R ΔH9 in axons of 178 hippocampal neurons (Fig. 5D). ACEA-induced internalisation of EGFP-CB1R ΔH9 was significantly 179 greater than that observed for EGFP-CB1R WT (Fig. 5E). Taken together, these data indicate that 180 CB1R ΔH9 is less stable at the axonal surface under basal conditions and that it is more susceptible 181 to agonist-induced internalisation. 182 The role of H9 in polarity is revealed in the presence of inverse agonist. 183 Our data thus far have indicated that ctCB1R, and the H9 domain in particular, can mediate surface 184 polarity of a CD4 chimera (Fig. 2), and promote polarised surface delivery of CB1R (Fig. 3). In 185 contrast, deletion of H9 has no effect on CB1R surface polarity at steady-state (Fig. 4). However, 186 deletion of H9 does have a striking effect on the surface stability of CB1R -CB1R ΔH9 is less surface 187 expressed in both axons and dendrites and shows increased endocytosis (Figs. 3 and 4). 188 Furthermore, CB1R ΔH9 is more responsive to agonist-induced internalisation (Fig. 5). We therefore 189 wondered whether the difference between the CD4 chimeras and the full-length receptor, and the 190 difference between surface+endocytosed and surface polarity, may be due to the agonist binding 191 capability of the full-length receptor. Inverse agonist treatment, which prevents the receptor entering 192 an active conformation, has previously been shown to increase somatodendritic surface expression 193 similarly to treatment with an endocytosis inhibitor (Leterrier et al., 2006). We thus reasoned that in 194 this case, inverse agonist treatment may reveal a difference in surface polarity between EGFP-195 CB1R WT and EGFP-CB1R ΔH9 , like that observed with the CD4 chimeras and in surface+endocytosed 196 polarity. 197 We treated hippocampal neurons expressing either EGFP-CB1R WT or EGFP-CB1R ΔH9 with the 198 CB1R-specific inverse agonist AM281 (Leterrier et al., 2004) (Fig. 6A). In the DMSO control both 199 EGFP-CB1R WT and EGFP-CB1R ΔH9 displayed similar levels of surface polarity. In the presence of 200 AM281, however, EGFP-CB1R ΔH9 had significantly reduced surface polarity compared EGFP-201 CB1R WT (Fig. 6B) due to a significantly increased amount of dendritic surface expression (Fig. 6C). 202 These data suggest that in the absence of constitutive activity of the receptor, H9 plays a role in 203 mediating CB1R surface polarity. Furthermore, these data suggest that the increased internalisation 204 observed in dendrites with H9 deletion may be mediated by the presence of agonist. Finally, our 205 findings reaffirm the importance the state-dependent effect on CB1R trafficking. 206

Discussion 207
Our data indicate that axonal surface polarity of CB1R occurs as a result of two distinct, but 208 complementary, mechanisms. 1) Using time-resolved RUSH assays we demonstrate that more de 209 novo CB1R is delivered to the axon and that it is more rapidly surface expressed than in dendrites. 210 2) Once at the axonal membrane the newly delivered CB1R is more stably retained whereas in 211 dendritic membrane CB1R surface expression is transient and it is rapidly internalised. However, we 212 also note that our data do not specifically exclude the possibility that CB1R internalised into the 213 somatodendritic endocytosed compartment can be rerouted to the axon via the transcytosis 214 pathway, thus further facilitating axonal polarity (Simon et al., 2013). 215 Furthermore, since CD4-ctCB1R WT and CD4-ctCB1R ΔH9 chimeras cannot bind agonist, our results 216 are consistent with ctCB1R contributing to constitutive polarisation via a mechanism distinct from the 217 proposed continuous activation of CB1R by the presence of the endogenous agonist 2-218 Arachidonoylglycerol (2-AG) in the dendritic membrane (Ladarre et al., 2014). Our data suggest that 219 ctCB1R, especially H9, plays a role in constitutive preferential delivery of CB1R to the axonal 220 membrane. 221 Our results further demonstrate that ctCB1R is important for maintaining axonal surface polarity, in 222 part mediated by the H9 motif, which plays a role in both the preferential delivery and selective 223 retention of CB1R at in axons. We show that deleting H9 (CB1R ΔH9 ) has a range of effects on 224 trafficking, surface expression and signalling in hippocampal neurons. More specifically, these 225 include; i) CB1R ΔH9 lacks the preferential delivery to axons observed for CB1R WT , ii) CB1R ΔH9 is less 226 efficiently surface expressed, iii) CB1R ΔH9 that does reach the surface it is more rapidly endocytosed 227 in both axons and dendrites and iv) CB1R ΔH9 is more sensitive to agonist-induced internalisation and 228 less efficient at downstream signalling, monitored by activation of ERK1/2 phosphorylation. 229 Preferential axonal trafficking. 230 The mechanism behind polarised membrane trafficking in neurons is a fundamental question and 231 our data suggest a sorting mechanism at the level of the secretory pathway that preferentially targets CB1R to the axon. Since dendritic and axonal cargo are synthesized in the somatodendritic 233 compartment, selective sorting to the correct domain is crucial. While several sorting signals and 234 adaptors have been described for dendritic cargo, the mechanisms behind selective sorting to axons 235 are less well known (Lasiecka andWinckler, 2011, Bentley, 2016 #43663). For example, a recent 236 study in C. elegans has suggested that sorting of cargos to axons or dendrites depends on binding 237 to different types of clathrin-associated adaptor proteins (AP); axonal cargo bind to AP-3 whereas 238 dendritic cargo bind to AP-1 (Li et al., 2016). Interestingly, AP-3 binding has been associated with 239 CB1R trafficking to the lysosome in the dendritic compartment (Rozenfeld and Devi, 2008). One 240 possibility is that H9 may modulate CB1R binding to AP-3, reducing both preferential delivery to 241 axons and, perhaps, reducing sorting to lysosomes, causing an increase in dendritic membrane 242 CB1R. More studies are needed to examine the possibility of H9 influencing AP-3 and CB1R 243 interaction. 244 H9 and membrane retention. 245 Our data suggest that H9 stabilises CB1R at the membrane, regardless of compartment. While the 246 H8 domain is highly conserved in GPCRs, structural domains analogous to H9 have only been 247 reported in squid rhodopsin (Murakami and Kouyama, 2008)  Therefore, it is possible that H9 mediates the interactions between CB1R and SGIP1 and/or 260 selectively promotes β-Arrestin rather than CRIP1a binding. Further studies examining the 261 interaction between CB1R WT , CB1R ΔH9 , CRIP1a, β-Arrestin1/2, and SGIP1 are needed to examine 262 the mechanism by which H9 stabilises surface CB1R. 263 Given the increased interest in CB1R as a clinical target, understanding the fundamental cell biology 264 and trafficking behaviour of CB1R is an increasingly active and important area of research. Taken 265 together, our results reveal that the C-terminal domain, and H9 in particular, play important roles in 266 trafficking of CB1R. These findings provide important insight into the mechanisms of CB1R polarity 267 and highlight H9 as an important regulator of CB1R endocytosis and surface expression. 268

Cell culture and Transfection. 286
Dissociated hippocampal cultures were prepared from E17-E18 Wistar rats as previously described 287 (Martin and Henley, 2004). Glass coverslips were coated in poly-D-lysine or poly-L-lysine (1mg/mL, 288 Sigma) in borate buffer (10mM borax, 50mM boric acid) overnight and washed in water. Dissociated 289 hippocampal cells were plated at different densities in plating medium (Neurobasal,Gibco 290 supplemented with 10% horse serum, Sigma; 2 mM GlutaMAX, Gibco; and either GS21, 291 GlobalStem, or B27, Thermo Fisher) which was changed to feeding medium (Neurobasal 292 supplemented with 1.2 mM GlutaMAX and GS21 or B27) after 24 hours. For RUSH experiments, 293 cells were plated and fed in media containing GS21 instead of B27 because it does not contain 294 biotin. Cells were incubated at 37°C and 5% CO2 for up to 2 weeks. Animal care and procedures 295 were carried out in accordance with UK Home Office and University of Bristol guidelines. 296 Transfection of neuronal cultures was carried out at DIV 12 using Lipofectamine2000 (Invitrogen) 297 according to the manufacturer's instructions with minor modifications. Cells were left for 20-48 hours 298 before fixation. 299 Phospho-ERK assay. 300 HEK293T cells were transfected with EGFP-CB1R WT , EGFP-CB1R ΔH9 , or empty pcDNA3.1 and left 301 for 24 hours. The cells were serum-starved overnight and then treated with 1μM ACEA or 0.01% 302 EtOH for 5 min before being lysed in lysis buffer (50mM Tris-HCl; 150mM NaCl; 1% CHAPS, 303 ThermoFisher Scientific; protease inhibitors, Roche) with phosphatase inhibitors (Pierce, 304 ThermoFisher Scientific). SDS-PAGE and Western blotting procedures were carried out according 305 to standard protocols. 306 Live surface staining and antibody feeding. 307 To measure surface staining, cultured neurons were cooled at room temperature for 5-10 min, then 308 incubated with the appropriate antibody (chicken anti-GFP or mouse anti-CD4) in conditioned media 309 for 10-20 min at RT. The neurons were washed multiple times in PBS before fixation. 310 For agonist and inverse agonist experiments, the neurons were treated with 5μM ACEA (in EtOH) 311 or vehicle control (0.1% EtOH) for 3 hours or 10μM AM281 (in DMSO) or vehicle control (0.2% 312 DMSO) for 3 hours in conditioned media at 37°C and 5% CO2, and then subsequently surface 313 stained. 314 To measure endocytosed receptors, neurons were fed with chicken anti-GFP for 2h in conditioned 315 media at 37°C and 5% CO2. Neurons were washed several times in PBS and then surface antibody 316 was stripped by 2 quick washes with ice-cold pH 2.5 PBS followed by several washes in PBS before 317 fixation. 318

RUSH live labelling. 319
Neurons were transfected with RUSH constructs at DIV 12 for no longer than 24 hours to prevent 320 Axons were defined either as processes whose initial segment was positive for Ankyrin-G or as 351 processes negative for MAP2. The mean fluorescence was measured for each channel and the 352 dendritic values were averaged. "Surface" or "endocytosed" mean fluorescence values were 353 normalised to the "total" mean fluorescence value for each ROI to account for varying levels of 354 expression of transfected constructs. These values were then normalised to the axon value of the 355 control (WT, WT + vehicle, or CD4). 356 Because of the change in total mean fluorescence in axons throughout the different conditions, the 357 above image analysis was slightly modified for RUSH experiments. In these experiments, neurites 358 were traced using NeuronJ so that only the mean fluorescence of exactly the first 50μm of the axons 359 and 30-40μm of 2-4 primary dendrites for each channel was measured. All "surface" and 360 "surface+endocytosed" values (of both axons and dendrites) were normalised to the average total 361 dendritic value for each neuron. Axon total mean fluorescence was also normalised to the average total dendritic value within each cell. All values were then normalised to the WT 60 min axon value 363 within each experiment. 364 "Polarity indices (A/D ratio)" were calculated by dividing the axonal mean fluorescence value by the 365 average dendritic mean fluorescence value. 366 The scalebar for all images represents 20μm. 367

Statistics. 368
All statistics were performed using GraphPad Prism. The ROUT method was used to identify outliers 369 for all parameters measured before normalising to control. Neurons were removed from analysis if 370 any one parameter was found to be an outlier. To determine statistical significance between two 371 groups, a D'Agostino & Pearson normality test was performed. Unpaired t-tests were performed on 372 data that passed the normality test whereas the Mann-Whitney test was used if it did not. One-or 373 Two-way ANOVAs with Tukey's or Sidak's post hoc test were used to determine statistical 374 significance between more than two groups depending on the comparisons required. *p ≤ 0.05, **p 375 ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. All data are presented as mean ± SEM.