Allosteric Modulation of the CB1 Cannabinoid Receptor by Cannabidiol—A Molecular Modeling Study of the N-Terminal Domain and the Allosteric-Orthosteric Coupling

molecules

Article Allosteric Modulation of the CB1 Cannabinoid Receptor by Cannabidiol—A Molecular Modeling Study of the N-Terminal Domain and the Allosteric-Orthosteric Coupling

Jakub Jakowiecki 1,* , Renata Abel 2,3 , Urszula Orzeł 1, Paweł Pasznik 1, Robert Preissner 2 and Sławomir Filipek 1,*

1 Biological and Chemical Research Centre, Faculty of Chemistry, University of Warsaw, 02093 Warsaw, Poland; [email protected] (U.O.); [email protected] (P.P.) 2 Structural Bioinformatics Group, Institute for Physiology, Charité–University Medicine Berlin, 10115 Berlin, Germany; [email protected] (R.A.); [email protected] (R.P.) 3 Department of Pharmaceutical Biology and Botany, Wroclaw Medical University, 50556 Wroclaw, Poland * Correspondence: [email protected] (J.J.); sﬁ[email protected] (S.F.)

Abstract: The CB1 cannabinoid receptor (CB1R) contains one of the longest N termini among class A G protein-coupled receptors. Mutagenesis studies suggest that the allosteric binding site of cannabidiol (CBD) involves residues from the N terminal domain. In order to study the allosteric binding of CBD

to CB1R we modeled the whole N-terminus of this receptor using the replica exchange molecular dynamics with solute tempering (REST2) approach. Then, the obtained structures of CB1R with the N terminus were used for ligand docking. A natural cannabinoid receptor agonist, ∆9-THC, was docked to the orthosteric site and a negative allosteric modulator, CBD, to the allosteric site Citation: Jakowiecki, J.; Abel, R.; positioned between extracellular ends of helices TM1 and TM2. The molecular dynamics simulations Orzeł, U.; Pasznik, P.; Preissner, R.; were then performed for CB1R with ligands: (i) CBD together with THC, and (ii) THC-only. Analyses Filipek, S. Allosteric Modulation of of the differences in the residue-residue interaction patterns between those two cases allowed us to the CB1 Cannabinoid Receptor by elucidate the allosteric network responsible for the modulation of the CB1R by CBD. In addition, Cannabidiol—A Molecular Modeling we identiﬁed the changes in the orthosteric binding mode of ∆9-THC, as well as the changes in its Study of the N-Terminal Domain and binding energy, caused by the CBD allosteric binding. We have also found that the presence of a the Allosteric-Orthosteric Coupling. complete N-terminal domain is essential for a stable binding of CBD in the allosteric site of CB R as Molecules 2021, 26, 2456. https:// 1 doi.org/10.3390/molecules26092456 well as for the allosteric-orthosteric coupling mechanism.

Academic Editor: Alessandra Bisi Keywords: CB1; cannabinoid receptor; GPCRs; allosteric modulation; CBD; THC; interaction ﬁnger- prints; molecular dynamics; REST2 Received: 7 March 2021 Accepted: 19 April 2021 Published: 23 April 2021 1. Introduction Publisher’s Note: MDPI stays neutral The cannabinoid CB1 receptor is the most abundant G protein-coupled receptor with regard to jurisdictional claims in (GPCR) in the human central nervous system [1] and, together with CB2 receptor, is a part published maps and institutional afﬁl- of the endocannabinoid system. Cannabinoid receptors participate in many biological iations. processes, including mood regulation, appetite and lipid metabolism, cognitive functions, nociception, as well as cell growth and proliferation [2]. In particular, CB1R is a very attractive pharmacological target for treatment of many disorders [3], such as: multiple sclerosis, Parkinson’s disease, Huntington’s disease, epilepsy, pain, appetite deregulation Copyright: © 2021 by the authors. and obesity [4,5]. Before the discovery of the allosteric binding site(s) in CB1R, design- Licensee MDPI, Basel, Switzerland. ing new ligands for this receptor was solely based on the orthosteric binding site [6]. A This article is an open access article long list of CB1R ligands including endocannabinoids, phytocannabinoids, and synthetic distributed under the terms and compounds have been discovered and developed over the last 20 years. In 2005, it was conditions of the Creative Commons discovered that the CB1 receptor contains allosteric binding site(s) that can be targeted by Attribution (CC BY) license (https:// small molecule ligands [7,8]. creativecommons.org/licenses/by/ 4.0/).

Molecules 2021, 26, 2456. https://doi.org/10.3390/molecules26092456 https://www.mdpi.com/journal/molecules Molecules 2021, 26, x FOR PEER REVIEW 2 of 20

Molecules 2021, 26, 2456 2 of 21

1.1. Advantages of Allosteric Over Orthosteric Ligands of CB1R

1.1. AdvantagesConsidering of Allosteric the psychomimetic over Orthosteric effects Ligands commonly of CB1 associatedR with CB1R agonists and

depressantConsidering effects the associated psychomimetic with CB effects1 antagonists/inverse commonly associated agonists, with allosteric CB1R agonists modulation and depressantmay offer effectsan alternative associated approach with CB to1 antagonists/inverseattain potential therapeutic agonists, benefits allosteric while modulation avoiding maythe unwanted offer an alternative side effects approach of orthosteric to attain ligands potential [7]. Allosteric therapeutic modulators benefits while do not avoiding possess theintrinsic unwanted efficacy, side effects but instead of orthosteric they enhance ligands or [ 7 diminish]. Allosteric the modulators receptor’s do response not possess to or- intrinsicthosteric efficacy, ligands. but In instead the absence they enhanceof any exogenous or diminish CB the1R receptor’sorthosteric response ligands tothe orthosteric maximum ligands.effect of In an the allosteric absence ofmodulator any exogenous is determined CB1R orthosteric by the level ligands of endocannabinoids, the maximum effect such of an as allostericanandamide modulator (AEA) isand determined 2-arachidonylglycerol by the level of (2 endocannabinoids,-AG) (Figure 1) [9,10] such. While as anandamide the exoge- (AEA)nous orthosteric and 2-arachidonylglycerol ligands may produce (2-AG) adverse (Figure1 effects)[ 9,10 ].through While thelong exogenous term over orthosteric-activation ligandsor down may-regulation produce adverseof a receptor effects [9,10] through, the long allosteric term over-activationmodulators of orCB down-regulation1R may not cause ofthose a receptor undesirable [9,10], effects the allosteric since their modulators efficacy ofdepends CB1R mayon the not presence cause those of orthosteric undesirable lig- effectsands. Endogenous since their efficacy orthosteric depends ligands on are the released presence on of demand orthosteric and ligands.removed Endogenous by enzymatic orthostericdegradation ligands (i.e. areprocessed released by on fatty demand acid and amide removed hydrolase by enzymatic (FAAH) degradation [11]) or cellular (i.e., processedreuptake bywhen fatty no acid longer amide needed hydrolase. Therefore, (FAAH) the [ 11allosteric]) or cellular modulators reuptake allow when for no a temper- longer needed.ing of cannabinoid Therefore, the receptor allosteric signalling modulators without allow desensitization, for a tempering tolerance of cannabinoid or dependence receptor signalling[12]. In addition, without desensitization, the GPCR allosteric tolerance binding or dependence sites are often [12]. less In addition, conserved the than GPCR or- allostericthosteric bindingsites, thus sites the are allosteric often less ligands conserved have potentially than orthosteric greater sites, specificity thus the for allostericparticular ligandsreceptor have types potentially and subtypes greater [7] specificity. for particular receptor types and subtypes [7].

FigureFigure 1. 1Chemical. Chemical structures structures of endocannabinoids: of endocannabinoids: 2-arachidonylglycerol 2-arachidonylglycerol (2-AG) and(2-AG)anandamide and anandamide (AEA). (AEA). The allosteric ligands can be classified based on their effects on receptor signalling. The allosteric ligands can be classified based on their effects on receptor signalling. There are positive allosteric modulators (PAMs), which enhance the orthosteric agonists signallingThere are efficacy positive (independently allosteric modulators of their effect(PAMs), on thewhich biding enhance affinity the of orthosteric the orthosteric agonists lig- and,signalling which canefficacy be either (independently positive, negative of their or effect null), on and the the biding negative affinity allosteric of the modulators orthosteric (NAMs),ligand, which which can attenuate be either the positive, orthosteric negative agonists or signalling.null), and the Apart negative from theallosteric modulators modu- therelators are (NAMs), also different which kinds attenuate of allosteric the orthosteric ligands, agonists such as allostericsignalling. agonists Apart from capable the of mod- en- hancingulators receptorthere are activity also different even in kinds the absence of allosteric of an orthosteric ligands, such agonist, as allosteric allosteric agonists antagonists capa- thatble canof enhancing supress receptor receptor activation activity even induced in the by absence orthosteric of an agonists orthosteric (in a non-competitiveagonist, allosteric manner)antagonists and, that finally, can neutralsupress allosteric receptor ligandsactivation (NALs), induced which by orthosteric bind to the allostericagonists (in binding a non- sitescompetitive without manner) affecting and, receptor finally, activity neutral and allosteric affinity ligand [13–15s]. (NALs), The activity which of bind an allosteric to the al- modulatorlosteric binding depends sites on without the orthosteric affecting probe receptor used. activity Both theand change affinity in [13 orthosteric–15]. The ligandactivity bindingof an allosteric affinity andmodulator the magnitude depends ofon the the allosteric orthosteric modulation probe used. effect Both will the vary change between in or- differentthosteric orthosteric ligand binding ligands affinity [10,16 and]. the magnitude of the allosteric modulation effect will vary between different orthosteric ligands [10,16]. 1.2. Cannabidiol—Its Properties and the Binding Site 1.2. Cannabidiol - Its Properties and The Binding Site Cannabidiol (CBD) is one of the major phytocannabinoids present in Cannabis plants. Cannabidiol (CBD) is one of the major phytocannabinoids present in Cannabis plants. Either alone or in a mixture with ∆9-tetrahydrocannabinol (∆9-THC) it is being intensively 9 9 studiedEither alone as a safe or in and a mixture efficacious with alternative Δ -tetrahydrocannabinol for the treatment (Δ of- variousTHC) it medicalis being conditionsintensively suchstudied as cancer, as a safe multiple and efficacious sclerosis, Parkinson’salternative for (PD) the and treatment Alzheimer’s of various disease medical (AD). Itcondi- has beentions recently such as cancer, approved multiple in some sclerosis, countries Parkinson’s for the treatment(PD) and Alzheimer’s of some specific disease types (AD). of It treatment-resistanthas been recently childhoodapproved epilepsyin some countries [17]. The propertiesfor the treatment of CBD of including some specific their possible types of biological targets and docking studies were recently reviewed in [18].

Molecules 2021, 26, x FOR PEER REVIEW 3 of 20

Molecules 2021, 26, 2456 treatment-resistant childhood epilepsy [17]. The properties of CBD including their possi-3 of 21 ble biological targets and docking studies were recently reviewed in [18]. Recently, it has been also proposed that CBD can have potential usage as a support drug for the treatment of COVID-19 disease [19–22]. There are several mechanisms (both CB1R/CBRecently,2R dependent it has beenand independent) also proposed that that substantiate CBD can havethe potential potential benefits usage asof aCBD sup- usageport drugin anti for-COVID the treatment-19 therapy. of COVID-19 Firstly, the disease well-known [19–22 anti]. There-inflammatory are several and mechanisms immuno- modulatory(both CB1R/CB activity2R of dependent CBD that andcan inhibit independent) an excessive that substantiate immunological the response potential leading benefits toof a CBD cytokine usage storm, in anti-COVID-19 which is a direct therapy. cause Firstly, of severe the well-known lung injury anti-inflammatory observed in many and COVIDimmunomodulatory-19 patients [23] activity. It has ofbeen CBD shown that canthatinhibit CBD can an ameliorate excessive immunological the symptoms of response acute respiratoryleading to dist a cytokineress syndrome storm, which in mice is athrough direct cause the upregulation of severe lung of injuryapelin observed [21] and studies in many onCOVID-19 humans have patients shown [23]. a Itreduction has been of shown some thatpro- CBDinflammatory can ameliorate monocytes the symptoms production of acuteand ofrespiratory CD4 and CD8 distress T-cells syndrome activation in mice frequencies through in the response upregulation to cannabinoids of apelin [21 [20,24]] and studies. Sec- ondly,on humans CBD can have possibly shown attenuate a reduction the ofcrucial some routes pro-inflammatory of SARS-CoV2 monocytes infection in production humans and of CD4 and CD8 T-cells activation frequencies in response to cannabinoids [20,24]. since cannabis extract with a high CBD content was able to down-regulate the expression Secondly, CBD can possibly attenuate the crucial routes of SARS-CoV2 infection in humans of two receptors needed for SARS-CoV2 entry into the cell: the angiotensin-converting since cannabis extract with a high CBD content was able to down-regulate the expression enzyme 2 (ACE2) and the transmembrane serine protease 2 (TMPRSS2) [19,25]. Thirdly, of two receptors needed for SARS-CoV2 entry into the cell: the angiotensin-converting CBD is a PPARγ (peroxisome proliferator-activated receptor-γ) agonist so it can display enzyme 2 (ACE2) and the transmembrane serine protease 2 (TMPRSS2) [19,25]. Thirdly, a direct antiviral activity. In addition, PPARγ agonists regulate fibroblast/myofibroblast CBD is a PPARγ (peroxisome proliferator-activated receptor-γ) agonist so it can display activation and thus they can inhibit the development of pulmonary fibrosis [19]. Despite a direct antiviral activity. In addition, PPARγ agonists regulate fibroblast/myofibroblast the fact that some of those reports sound very promising more data are needed to justify activation and thus they can inhibit the development of pulmonary fibrosis [19]. Despite the use of CBD in the treatment of COVID-19. the fact that some of those reports sound very promising more data are needed to justify Although CBD is an isomer of THC (Figure 1) it has no psychoactive properties. In the use of CBD in the treatment of COVID-19. fact, it even suppresses some of the undesired side effects induced by Δ9-THC such as Although CBD is an isomer of THC (Figure1) it has no psychoactive properties. In intoxication,fact, it even sedation suppresses and some tachycardia of the undesired while attenuating side effects its induced medical by effects∆9-THC such such as anal- as in- gesic,toxication, anti-emetic, sedation and and anti tachycardia-carcinogenic while effects attenuating [26]. Together its medical with effects terpenoids such as and analgesic, other phytocanti-emetic,annabinoids and anti-carcinogenic present in cannabis effects CBD [26 contributes]. Together withto the terpenoids synergistic and entourage other phyto- ef- fectcannabinoids [27]. CBD binds present to ina distinct cannabis binding CBD contributes site than the to orthosteric the synergistic ligands entourage since it effect antago- [27]. nises CB1R orthosteric agonists at concentrations far below its reported affinity [28]. The CBD binds to a distinct binding site than the orthosteric ligands since it antagonises CB1R NAorthostericM activity agonists of CBD at was concentrations also demonstrated far below by itsLapraire reported et al. affinity [10] who [28]. showed The NAM that activity CBD reducesof CBD the was potency also demonstrated and efficacy of by Δ Lapraire9-THC and et al.2-AG [10 ]in who a non showed-competitive that CBD manner. reduces Their the resultspotency also and indicate efficacy that of ∆the9-THC CBD and allosteric 2-AG inbinding a non-competitive site is located manner. within Theirthe CB results1R N-ter- also minal domain and is particularly close to residues C98 and C107, which form a disulfide indicate that the CBD allosteric binding site is located within the CB1R N-terminal domain bond.and isThey particularly performed close mutag to residuesenesis and C98 BRET and studies C107, which demonstrating form a disulfide that the bond.C98A mu- They tationperformed as well mutagenesis as the C107A and mutation BRET studies led to demonstrating a 50% loss of CBD that theNAM C98A activity, mutation versus as well both as agonists,the C107A Δ9- mutationTHC and led2-AG, to a when 50% loss compared of CBD to NAM WT activity,CB1R. Interestingly, versus both agonists,other mutations∆9-THC atand those 2-AG, residues, when C98S compared and C1 to07S, WT CBcaused1R. Interestingly, no change in other CBD mutationsNAM activity at those [10] residues,, which suggestsC98S and that C107S, other caused interactions no change can substitute in CBD NAM for the activity disul [10fide], whichbridge suggests and stabilize that other the localinteractions structure can of substituteCB1R. Presence for the of disulfide the Cys98 bridge-Cys107 and bond stabilize also the affects local the structure allosteric of CB ac-1R. tivityPresence of ORG27569 of the Cys98-Cys107 and PSNCBAM bond-1 also(Figure affects 2) [29] the and allosteric it is very activity likely of that ORG27569 CBD binds and toPSNCBAM-1 the same allosteric (Figure 2site)[29 as] andthose it istwo very compounds. likely that CBDHowever, binds in to contrast the same to allosteric ORG27569 site andas thosePSNCBAM two compounds.-1, for which However, the ligand in cooperativity contrast to ORG27569 is positive and for PSNCBAM-1, agonists and negative for which forthe antagonists ligand cooperativity and inverse is agonists positive [8,30] for agonists, CBD displayed and negative a negative for antagonists cooperativity and inversefor all testedagonists ligands [8,30 [10]], CBD. displayed a negative cooperativity for all tested ligands [10].

FigureFigure 2. 2. ChemicalChemical structure structuress of of CB CB1R1 Rallosteric allosteric ligands ligands ORG27569 ORG27569 and and PSNCBAM PSNCBAM-1.-1.

1.3. The Importance of N-Terminus for Allosteric and Orthosteric Binding

We have concluded that the role of the CB1R N-terminal domain in both allosteric and orthosteric ligand binding was signiﬁcantly underestimated during recent years, and for this reason we decided to determine its structure as a whole, dedicating a special effort to modeling of the membrane-proximal region (MPR) and the allosteric binding site of Molecules 2021, 26, x FOR PEER REVIEW 4 of 20

1.3. The Importance of N-terminus for Allosteric and Orthosteric Binding

We have concluded that the role of the CB1R N-terminal domain in both allosteric and orthosteric ligand binding was significantly underestimated during recent years, and for this reason we decided to determine its structure as a whole, dedicating a special effort Molecules 2021, 26, 2456 to modeling of the membrane-proximal region (MPR) and the 4allosteric of 21 binding site of CBD. The main inspiration for our project was the experimental work of Lapraire et al. on CBD binding [10]. In our research we have focused on exploring the CB1 receptor allosteric CBD. The main inspiration for our project was the experimental work of Lapraire et al. onbinding CBD binding site [(allosteric10]. In our researchsite 1) welocated have focused between on exploring the extracellular the CB1 receptor ends of helices I and II allosteric(colored binding in blue site (allostericand cyan site in 1) Figure located between 3, respectively). the extracellular The ends three of helices allosteric sites (Figure 1) Iwere and II proposed (colored in blueby Sabatucci and cyan in et Figure al. [31]3, respectively). who found The that three CBD allosteric very sites likely binds to the allo- (Figure1) were proposed by Sabatucci et al. [31] who found that CBD very likely binds to thesteric allosteric site site1. 1.

FigureFigure 3. 3CB1. CB1 cannabinoid cannabinoid receptor receptor with modelled with N-terminusmodelled and N-terminus the locations and of orthosteric the locations of orthosteric and andallosteric allosteric binding binding sites. sites. Cannabidiol Cannabidiol (CBD)—an (CBD) allosteric—an ligand, allosteric and ∆ 9-tetrahydrocannabinolligand, and Δ9-tetrahydrocannabinol (Δ9- (∆9-THC)—an orthosteric ligand of CB1R. According to Lapraire et al. and Sabatucci et al. CBD very THC)—an orthosteric ligand of CB1R. According to Lapraire et al. and Sabatucci et al. CBD very likely binds to the allosteric site 1 [10,31]. likely binds to the allosteric site 1 [10,31]. Sabatucci et al. performed docking studies in order to map the allosteric binding sites ofSabatucci CB1R[31]. However,et al. performed as they point docking out themselves, studies forin order docking to they map used the the allosteric binding sites crystal structures 5TGZ and 5U09, which lack most of the N-terminal residues, includ- of CB1R [31]. However, as they point out themselves, for docking they used the crystal ing C98, which has been shown to have an impact on allosteric modulation of the CB1 receptorstructures [10,29 ].5TGZBesides, and they 5U09, did not which perform lack any molecular most of dynamics the N- simulationsterminal residues, that including C98, wouldwhich confirm has been the stability shown of CBDto have binding an in impact the allosteric on siteallosteric 1. A study modulation on CBD al- of the CB1 receptor losteric binding that employed MD simulations has been done by Chung et al. [32]. They have[10,29] reported. Besides, an outward they rotation did not of TM1 perform and TM2 any upon molecular CBD binding dynamics to the allosteric simulations that would siteconfirm 1, resulting the instability the expansion of CBD of the binding orthosteric in binding the allosteric site. They also site revealed 1. A study that on CBD allosteric CBDbinding hydroxyl that group employed formed interactions MD simulations with D104 has (which been is in done agreement by Chung with our et al. [32]. They have results)reported while an its outward alkyl chain formedrotation hydrophobic of TM1 contactsand TM2 with upon F102, I105,CBD I267 binding and F268. to the allosteric site 1, However, it should be pointed out that they used very short MD simulations, 25–50 ns, whichresulting were very in likelythe expansion insufficient for of verification the orthosteric of CBD binding binding mode site. stability. They Besides, also revealed that CBD thosehydroxyl studies group relied on formed crystal/cryo-EM interactions structures with supplemented D104 (which with a small,is in proximalagreement with our results) partwhile of N-terminus its alkyl only,chain without formed considering hydrophobic the influence contacts of the completewith F10 N-terminal2, I105, I267 and F268. How- domain on the allosteric binding. ever,There it should is a large be difference pointed between out that lengths they of N-terminiused very of CBshort1 (111 MD residues) simulations, and 25–50 ns, which CBwere2 (28 very residues) likely receptors. insufficient We have for done verification a comparison of of N-terminiCBD binding of class mode A GPCRs stability. Besides, those excludingstudies odorantrelied receptorson crystal/cryo (Figure4). The-EM CB structures1R N-terminus su ispplemented one of 14 longest with N-termini a small, proximal part of of those receptors while for CB2R it has the most typical length. Because of this difference theN- allosteric-orthostericterminus only, couplingwithout in considering both cannabinoid the receptors influence is probably of the much complete different. N-terminal domain on the allosteric binding. There is a large difference between lengths of N-termini of CB1 (111 residues) and CB2 (28 residues) receptors. We have done a comparison of N-termini of class A GPCRs excluding odorant receptors (Figure 4). The CB1R N-terminus is one of 14 longest N-ter-

Molecules 2021, 26, x FOR PEER REVIEW 5 of 20

mini of those receptors while for CB2R it has the most typical length. Because of this difference the allosteric-orthosteric coupling in both cannabinoid receptors is probably much Moleculesdifferent.2021, 26, 2456 5 of 21

Figure 4. Distribution of N-terminus lengths within class A of GPCRs. On left the histogram with 10 amino acid interval; on Figureright the histogram4. Distribution for all class of AN GPCR,-terminus without lengths odorant within receptors, class with A 100 of amino GPCRs. acid interval. On left Number the histogram of receptors with 10in eachamino range acid is specified interval; above on each right bar. the The histogram ranges are given for all in parenthesisclass A GPCR, which are without open from odorant left side. receptors, Based on withGPCRdb 100 data amino [33]. acid interval. Number of receptors in each range is specified above each bar. The ranges are given in parenthesis which are open from left side. Based on GPCRdb data [33]. In our studies we have modeled the structure of the whole N-terminal domain of CB1R, docked CBD and THC to their binding sites, and run a set of long molecular dynamics In our studies (MD)we simulations,have modeled 10 × 1 µthes, to structure obtain statistically of the valid whole results. N All-terminal the above domain simulations of CB1R, docked CBDstarted and THC from the to sametheir structure binding of CB sites,1R with and a complete run a set N-terminus of long and molecular THC bound dy- to namics (MD) simulations,the orthosteric 10 × 1 site. µs, CBD to obtain was bound statistically to the allosteric valid site results. 1 in five All simulations, the above while sim- in the other five simulations the allosteric site was empty. Analysis of the residue contact ulations started fromnetwork the same and, especially, structure the differencesof CB1R inwith the interaction a complete patterns N-terminus between those and two THC sets bound to the orthostericof trajectories, site. CBD revealed was interactions bound thatto the are crucialallosteric for the site allosteric-orthosteric 1 in five simulations, coupling, while in the other fivecalled simulations here the ‘allosteric the network’. allosteric We alsosite investigated was empty. the relevanceAnalysis of CBof1 Rthe N-terminus residue for CBD allosteric binding by gradually truncating the N-terminus starting from its first contact network and,residue especially, and analyzing the differences the changes in in its the stability interaction and in the patterns ligand binding between energy. those All two sets of trajectories,performed revealed MD simulations interactions (including that REST2 are crucial simulations) for tookthe overallosteric 150 µs- (Tableorthosteric1). coupling, called here the ‘allosteric network’. We also investigated the relevance of CB1R N-terminus for CBD allostericTable 1. The binding MD simulations by gradually performed in thistruncating work. the N-terminus starting from MDits Simulationfirst residue Type and analyzing Average Length the changes [µs] Repeats in its stability REST2 Replicas and in the Cumulative ligand Length binding [µs] energy.CB All1R with performed MPR MD simulations 0.25 (including 8 REST2 simulations) 1 took over 2 150 μs REST2 of CB1R with complete N-terminus 1.2 6 16 115.2 CB(Table1R with CBD-THC1). (for contact map) 1 5 1 5 CB1R with THC-only (for contact map) 1 5 1 5 CB R N-terminus stability corroboration 1 2 × 3 3 1 18 Table(with 1. andThe without MD simulations CBD) performed in this work.

CB1R with truncated N-term (mutation studies) 3 × 1 2 1 6 Average Re- REST2 Cumulative TotalMD Simulation Type 151.2 Length [μs] peats Replicas Length [μs] CB1R with MPR 0.25 8 1 2 REST2 of CB1R with complete N-terminus 1.2 6 16 115.2 CB1R with CBD-THC (for contact map) 1 5 1 5 CB1R with THC-only (for contact map) 1 5 1 5 CB1R N-terminus stability corroboration (with 2 × 3 3 1 18 and without CBD) CB1R with truncated N-term (mutation studies) 3 × 1 2 1 6 Total 151.2

Molecules 2021, 26, x FOR PEER REVIEW 6 of 20

2. Results and Discussion

2.1. Construction of The N-Terminus of CB1R The so called membrane-proximal region (MPR) of cannabinoid receptor (∼ residues 90−110) has been proved to be crucial not only for the allosteric ligands binding, but also

forMolecules binding2021, 26, 2456of the orthosteric ligands [29]. According to Laprairie et al. [10] the 96–1126 of 21 fragment (stabilized by the C98-C107 disulfide bond) is essential for the NAM activity of CBD. Therefore, we started with modeling of this small fragment. The task was made easier by the fact that the 2.residues Results and 100 Discussion–112 were already present in the crystal structure of CB1R (PDB ID: 5U09 [34]2.1.), Constructionso only the of the 4 N-Terminusresidues of(96 CB–199)R were added. We chose this particular crystal structure becauseThe so calledit had membrane-proximal the longest N- regionterminal (MPR) part of cannabinoid from all receptoravailable (∼residues ex- 90−110) has been proved to be crucial not only for the allosteric ligands binding, but perimental structures ofalso that for receptor. binding of Also, the orthosteric it is very ligands likely [29 that]. According this structure to Laprairie contain et al.s [the10] the disulfide bridge C98–C10796–112 since fragment the(stabilized first residue by the occurring C98–C107 disulfide in the bond)crystallized is essential protein for the NAMis G89, although the visibleactivity part of of CBD. the Therefore,structure we starts started from with residue modeling E100. of this Therefore, small fragment. we The as- task was made easier by the fact that the residues 100–112 were already present in the crystal sume that the geometry of the 100–112 fragment is the most correct among all available structure of CB1R (PDB ID: 5U09 [34]), so only the 4 residues (96–99) were added. We chose CB1R experimental structuresthis particular and, crystal hence, structure both because the orthosteric it had the longest and N-terminal the allosteric part from binding all available sites are the most accurateexperimental and most structures resembling of that receptor. those Also, of itthe is very wild likely-type that thisreceptor structure ( containsFigure the disulfide bridge C98–C107 since the first residue occurring in the crystallized protein is G89, 5A,B). Choosing a correctalthough initial the visiblestructure part ofat the this structure stage starts is essential from residue since E100. the Therefore, completely we assume different residues constitutethat the the geometry particular of the 100–112binding fragment sites in is thedifferent most correct crystal among structures. all available CB1R experimental structures and, hence, both the orthosteric and the allosteric binding sites Different conformations of the binding site in the crystal structures of CB1R are not are the most accurate and most resembling those of the wild-type receptor (Figure5A,B). only a result of differentChoosing orthosteric a correct ligands initial structurebound but, at this most stage isimportantly, essential since thethey completely ensue from different the presence or absenceresidues of the constitute MPR and,the particular particularly, binding sites of inthe different C98– crystalC107 structures.disulfide bond (Figure 5B). Such claim is inDifferent agreement conformations with experimental of the binding siteresults in the showing crystal structures that mutations of CB1R are not only a result of different orthosteric ligands bound but, most importantly, they ensue within MPR weaken thefrom binding the presence of CP55 or absence,940 ofto the the MPR orthosteric and, particularly, binding of the site C98–C107 [35] while disulfide truncation of the wholebond N- (Figureterminus5B). Such at position claim is in 113 agreement abolishes with experimental the binding results completely. showing that However, if only MPRmutations is retained, within the MPR binding weaken the of binding the orthosteric of CP55,940 to ligands the orthosteric is maint bindingained site [ 35] while truncation of the whole N-terminus at position 113 abolishes the binding completely. [29]. However, if only MPR is retained, the binding of the orthosteric ligands is maintained [29].

Figure 5. The influence of C98–C107 disulfide bond on the orthosteric binding of CB R. (A) PDB id:5XR8 structure with Figure 5. The influence of C98-C107 disulfide bond on the orthosteric binding1 of CB1R. (A) PDB id:5XR8no disulfide structure bond; with (B) PDB no id:5U09 disulf structureide bond; with (B disulfide) PDB bondid:5U09 present. structure In 5XR8 structurewith disul the residuesfide bond F102 present. and M103 are outside of the orthosteric binding site; the ligand’s position is shallower and there is much more space available in the In 5XR8orthosteric structure binding the site residues so the ligand F102 has moreand freedomM103 are and isoutside in contact of with the many orthosteric water molecules. binding In 5U09site; structure the the ligand'sresidues position F102 and is M103 shallow are insideer and the orthostericthere is much binding more site; the space ligand isavailable positioned in deeper, the orthosteric surrounded by binding hydrophobic site soresidues the ligand and the bindinghas more is tighter. freedom and is in contact with many water molecules. In 5U09 structure the residues F102 and M103 are inside the orthosteric binding site; the ligand is positioned deeper, surrounded by hydrophobicWe first elucidatedresidues and the structure the binding of MPR is oftighter. CB1R (Figure 6)—see Section 3.1 of Meth- ods for details. The structure of CB1R with MPR was used in our previous work [36] for investigating the entry pathway into the orthosteric binding site of the CB1 cannabinoid We first elucidatedreceptor. the structure Next, the of remaining MPR of residues CB1R ( 1–95Figure of the 6)— N-terminalsee Section domain 3.1 were of Meth- added ac- ods for details. The structurecording toof theCB sequence1R with of MPR a wild-type was used human in CB our1R andprevious the conformational work [36] space for of the newly added part was extensively explored using the Replica Exchange Molecular investigating the entry pathway into the orthosteric binding site of the CB1 cannabinoid receptor. Next, the remaining residues 1–95 of the N-terminal domain were added according to the sequence of a wild-type human CB1R and the conformational space of the newly added part was extensively explored using the Replica Exchange Molecular Dynamics with Solute Tempering [37] simulations. Details of this procedure are specified in section 3.2 of Methods.

Molecules 2021, 26, x FOR PEER REVIEW 7 of 20 Molecules 2021, 26, 2456 7 of 21

Molecules 2021, 26, x FOR PEER REVIEW 7 of 20

Dynamics with Solute Tempering [37] simulations. Details of this procedure are speciﬁed in Section 3.2 of Methods.

Figure 6. The modelling process of MPR of the CB1R N-terminal domain. Residues 96–99 were added to the CB1R crystal structure (PDB ID: 5U09) and the C98-C107 disulf ide bond was created. Regions 96–100 and 107–108 were then sampled in Rosetta. Eight structures of CB1R with MPR FigureFigure 6. 6The. The modelling modelling process process of MPR ofof the MPR CB1R of N-terminal the CB1 domain.R N-terminal Residues domain. 96–99 were Residues added to the 96 CB–199R crystalwere structure (PDB ID: 5U09) andobtained the C98–C107 from disulﬁde Rosetta bond were was created. placed Regions in POPC 96–100 bilayer and 107–108 and were simulated then sampled for in 250 ns. added to the CB1R crystal structure (PDB ID: 5U09) and the C98-C107 disulfide bond was created. Rosetta. Eight structures of CB1R with MPR obtained from Rosetta were placed in POPC bilayer and simulated for 250 ns. Regions 96–100 and 107–108 were then sampled in Rosetta. Eight structures of CB1R with MPR 2.2. Structural Features of The CB1R N-terminal Domain obtained from Rosetta2.2. were Structural placed Features in POPC of the bilayer CB1R N-Terminal and simulated Domain for 250 ns. TheThe N-terminal N-terminal helix 3 (NTH3,helix 3 residues (NTH3, 80–94) residues is present 80 in-94) many is structurespresent obtainedin many structures obtained 2.2. Structural Featuresfromfrom of REST2T REST2he CB simulations1R simulations N-terminal and, once Dand, itomain forms, once it isit very forms, stable it (Figure is very7A). stable (Figure 7A). The N-terminal helix 3 (NTH3, residues 80-94) is present in many structures obtained from REST2 simulations and, once it forms, it is very stable (Figure 7A).

Figure 7. CB7. 1CBreceptor1 receptor N-terminal N-terminal domain conformation domain conformation depends on CBD presencedepends in theon allostericCBD presence in the allosteric site. (A) CB R with CBD and THC bound, (B) CB R with THC bound only; cyan—MPR, violet— site. (A) CB1 1R with CBD and THC bound,1 (B) CB 1R with THC bound only; cyan—MPR, violet— N-terminal helix 3 (NTH3, residues 80–94), red—a parallel β-sheet (residues 58–60 and 72–74), N-terminal helix 3 (NTH3, residues 80–94), red—a parallel β-sheet (residues 58–60 and 72–74), ma- Figure 7. CB1 receptor magenta—N-terminalN-terminal domain helix conformation 2 (NTH2, residues depends 61–69), salmon—N-terminal on CBD presence helix in 1 (NTH1, the allosteric residues genta—N-terminal helix 2 (NTH2, residues 61–69), salmon—N-terminal helix 1 (NTH1, residues 30– site. (A) CB1R with CBD30–35), and orange—N-terminal THC bound, ( capB) (residuesCB1R with 1–20). THC Proline bound residues only; are shown cyan as white—MPR, sticks. violet— N-terminal helix 3 (NTH3,35), orangeresidues— N80-–terminal94), red —capa parallel(residues β- 1sheet–20). (residuesProline residues 58–60 and are 72shown–74), asma- white sticks. Very likely this helix is a scaffold on which the rest of the N-terminus is arranged. The genta—N-terminal helixthree 2 (NTH2, glutamates, residues E91, E93, 61 and–69), E94, salmon located— inN this-terminal helix, form helix ionic 1 (NTH1, interactions residues with other 30– 35), orange—N-terminalparts cap of Very(residues N-terminus likely 1– (Figure20). this Proline 8helixA) constituting residuesis a scaffold aare hub shown essentialon which as forwhite thethe N-terminalsticks. rest of the domain N-terminus is arranged. stability.The three For glutamates, example, the 1–20 E91, residue E93, fragment,and E94, called located by us in the this N-terminal helix, form cap (or ionic interactions with Very likely thissimply otherhelix ‘N-cap’) partsis a scaffold of and N located-terminus on close which to (Figure TM1 the and rest 8 TM2A) of constituting extracellular the N-terminus ends, a hub is relatively is essential arranged. stable for the N-terminal do- due to a salt-bridge K2-E93 (and also K2-D104 as well as K2-E106) (see the interactions on The three glutamates,main E91, stability.E93, and ForE94, example, located in the this 1 –helix,20 residue form ionicfragment, interactions called bywith us the N-terminal cap other parts of N-terminus(or simply (Figure ‘N -8cap’)A) constituting and located a closehub essentialto TM1 and for TM2the N extracellular-terminal do- ends, is relatively sta- main stability. For example,ble due tothe a salt1–20-bridge residue K2 fragment,-E93 (and alsocalled K2 by-D104 us the as wellN-terminal as K2-E106) cap (see the interactions (or simply ‘N-cap’) andon Figurelocated 8 closeA) that to supportsTM1 and theTM2 conformation extracellular and ends, position is relatively of this sta- fragment. The residues ble due to a salt-bridgeM1 K2 and-E93 K2 (and are alsowithin K2 -theD104 allosteric as well siteas K2 1,- constitutingE106) (see the a interactionsbarrier separating that site from on Figure 8A) that supportsthe extracellular the conformation milieu. We and have position also ofobserved this fragment. a shor tThe parallel residues β-sheet forming repeat- M1 and K2 are withinedly the between allosteric the site 58 1,– 60constituting and 72–74 a strandsbarrier separatingwhile a very that short site helixfrom formed occasionally the extracellular milieu.within We the have 60 –also72 loop. observed The N a- terminusshort parallel contains β-sheet five formingprolines: repeat-P45, P49, P57, P68 and P72, edly between the 58–stacked60 and together72–74 strands within while a fragment a very shortof only helix 30 rformedesidues occasionally long. The stability of particular within the 60–72 loop.regions The N of-terminus N-terminus contains were five estimated prolines: by P45,RMSF P49, calculations P57, P68 and (Figure P72, 9A,B) based on ten 1 stacked together withinµs long a fragment MD simulations of only 30 for residues CB1R complexes long. The withstability and ofwithout particular CBD. The additional six regions of N-terminusMD were simulations estimated 3 by µs RMSF long performedcalculations for (Figure corro boration9A,B) based of N on-terminus ten 1 stability gave the µs long MD simulationsstructure for CB of1 Rincreased complexes stability with andfor the without CB1R complexCBD. The with additional THC and six without CBD (Figure MD simulations 3 µs long performed for corroboration of N-terminus stability gave the structure of increased stability for the CB1R complex with THC and without CBD (Figure

Molecules 2021, 26, 2456 8 of 21

Figure8A) that supports the conformation and position of this fragment. The residues M1 and K2 are within the allosteric site 1, constituting a barrier separating that site from the extracellular milieu. We have also observed a short parallel β-sheet forming repeatedly between the 58–60 and 72–74 strands while a very short helix formed occasionally within the 60–72 loop. The N-terminus contains ﬁve prolines: P45, P49, P57, P68 and P72, stacked together within a fragment of only 30 residues long. The stability of particular regions Molecules 2021, 26, x FOR PEER REVIEWof N-terminus were estimated by RMSF calculations (Figure9A,B) based on ten 1 µs 8 of 20

long MD simulations for CB1R complexes with and without CBD. The additional six MD simulations 3 µs long performed for corroboration of N-terminus stability gave the structure of increased stability for the CB1R complex with THC and without CBD (Figures7B and9C) where7B and two Figure additional 9C) where short helices two additional were formed short but helices helix NTH2 were was formed dissolved but helix in one NTH2 of was threedissolved simulations. in one of three simulations.

FigureFigure 8.8. (A––CC)) FlareFlare plots plots of of crucial crucial interactions interactions responsible responsible for for stabilizing stabilizing the N-terminalthe N-terminal do- domain mainin THC in- THC-onlyonly complex). complex). Line Line thickness thickness is proportional is proportional to tothe the frequency frequency of of the the interaction; interaction; (A) Ionic (interactions;A) Ionic interactions; highlighted highlighted are the are interactions the interactions of E91 of E91(pink), (pink), E93 E93 (dark (dark red) red) and and E94 E94 (orange), (orange), (B) Total (hydrogenB) Total hydrogen bonds bondsof N-terminus; of N-terminus; (C) The (C) The hydrogen hydrogen bonds bonds of of N N-terminus-terminus excluding excluding backbone- backbone-back- backbonebone H-bonds; H-bonds; (D, (ED), ELigand) Ligand-receptor-receptor dynamic dynamic interactioninteraction fingerprints fingerprints in THC-CBDin THC-CBD complex; complex; (D) (frequencyD) frequency fingerprints fingerprints o off interactions of of residues residues with with CBD CBD for each for simulationeach simulation separately; separately; (E) the (E) the average interactioninteraction frequencies frequencies of fingerprintsof fingerprints from from panel panel D; (F) D; the ( averageF) the average frequency frequency fingerprints fingerprints of interactionsinteractions of of residues residues with with THC THC for simulationsfor simulations with THC-CBDwith THC (green)-CBD (green) and with and THC-only with THC-only (blue) bound to the CB1R. (blue) bound to the CB1R.

Figure 9. RMSF plots of CB1R N-terminus (residues 1–113) measured during 1 µs long MD simulations starting from the same receptor structure (Figure 4A) bound with (A) CBD and THC; (B) THC-only; (C) for simulations from N- terminus stability corroboration (Table 1). The last 1 µs from 3 µs MD simulations were taken for RMSF calculations.

Molecules 2021, 26, x FOR PEER REVIEW 8 of 20

7B and Figure 9C) where two additional short helices were formed but helix NTH2 was dissolved in one of three simulations.

Figure 8. (A–C) Flare plots of crucial interactions responsible for stabilizing the N-terminal domain in THC-only complex). Line thickness is proportional to the frequency of the interaction; (A) Ionic interactions; highlighted are the interactions of E91 (pink), E93 (dark red) and E94 (orange), (B) Total hydrogen bonds of N-terminus; (C) The hydrogen bonds of N-terminus excluding backbone-backbone H-bonds; (D,E) Ligand-receptor dynamic interaction fingerprints in THC-CBD complex; (D) frequency fingerprints of interactions of residues with CBD for each simulation separately; (E) the Molecules 2021, 26, 2456 average interaction frequencies of fingerprints from panel D; (F) the average frequency fingerprints9 of 21 of interactions of residues with THC for simulations with THC-CBD (green) and with THC-only (blue) bound to the CB1R.

Molecules 2021, 26, x FOR PEER REVIEW 9 of 20

FigureFigure 9. 9.RMSF RMSF plots plots of of CB CB1R1 N-terminusR N-terminus (residues (residues 1–113) 1–113) measured measured during during 1 µs long1 µs MDlong simulations MD simulations starting starting from the from samethe same receptor receptor structure structure (Figure (FigureAs4A) we bound 4A) had with bound anticipated, ( A )with CBD ( andA) CBD the THC; fiveand (B) THC-only;THC;prolines (B) THC (-Ccontaining) for-only; simulations (C) for fragment simulations from N-terminus disrupts from N- a formation of stabilityterminus corroboration stability corroboration (Table1). The (Table last 1 1).µs fromThe last 3 µ s1MD µs from simulations 3 µs MD were simulations taken for RMSFwere taken calculations. for RMSF calculations. the larger secondary structure elements in that region and only transient structures can be formed.As weThe had whole anticipated, N-terminus the ﬁve prolines-containingis stabilized by fragmentinternal disrupts salt bridges a formation (Figure of 8A) and a largethe larger number secondary of internal structure hydrogen elements inbonds that region but also and forms only transient a few structuresinteractions can bewith the extra- cellularformed. ends The whole of TM1, N-terminus TM2, and is stabilized with th bye internalextracellular salt bridges loops (Figure (Figure8A) and 8B,C). a large number of internal hydrogen bonds but also forms a few interactions with the extracellular 2.3.ends The of O TM1,rthosteric TM2, and and with Allosteric the extracellular Ligand B loopsinding (Figure 8B,C).

2.3.The The Orthostericanalysis of and ligand Allosteric-receptor Ligand Binding interaction patterns in the complex of CB1R with CBD and Δ9-THC (Figure 10) during MD simulations demonstrates that the pose of CBD is very The analysis of ligand-receptor interaction patterns in the complex of CB1R with CBD stableand ∆ when9-THC it (Figure is bound 10) during to the MD allosteric simulations site demonstrates 1 of a wild- thattype the CB pose1 cannabinoid of CBD is very receptor with a wholestable when N-terminus it is bound which to the allosteric is indicated site 1 ofby a existence wild-type CB of1 severalcannabinoid ligand receptor-receptor with interactions witha whole a frequency N-terminus close which to is indicated1.0 for the by existenceresidues of D104, several I105, ligand-receptor E106 fro interactionsm MPR, Q1161.32 from with a frequency close to 1.0 for the residues D104, I105, E106 from MPR, Q1161.32 from 2.61 2.64 2.65 helixhelix I, I, and and residuesresidues F174 F1742.61, F177, F1772.64, H178, H1782.65 from helixfrom II helix (Figure II8D,E). (Figure The important8D,E). The important contributioncontribution toto CBDCBD binding binding comes comes also fromalso thefrom ﬁrst the residue first ofresidue N-terminus, of N M1,-terminus, which M1, which sideside chain chain partially protects protects CBD CBD from from the water the environment.water environment.

9 9 FigureFigure 10.10.CBD CBD and and THC THC binding binding to CB 1toR. CB CBD1R. is CBD bound is to bound the allosteric to the site allosteric 1 while ∆ site-THC 1 while is Δ -THC is boundbound toto the the orthosteric orthosteric site. Thesite. allosteric The allosteric site is located site veryis located close to very the orthosteric close to binding the orthosteric site and binding site andthey they partially partially overlap overlap one another. one The another. MPR (purple) The MP constitutesR (purple) a barrier constitutes between the a orthostericbarrier between the or- thostericand the allosteric and the binding allosteric sites. binding CBD bound sites. to the CBD allosteric bound binding to the site allosteric forms a hydrogen binding bond site with forms a hydrogen π π bondD104 andwith -D104interactions and π- withπ interactions F177. with F177.

The binding mode of THC in the orthosteric site is influenced by the presence of CBD in the allosteric site 1. The residues for which the frequencies of interactions with THC are not changed are V1963.32, T1973.33 of helix III, L2765.40, W2795.43 of helix V, L3596.51 of helix VI, and S3837.39, C3867.42 of helix VII (Figure 8F). CBD changes the average frequencies of interactions between THC and the receptor in such a way that some of them are increased and some decreased so the average binding pose of THC in the binding site is modified. The residues with increased frequencies are M103 of MPR, G1662.53, F1702.57 of helix II, and W3566.48 of helix VI. The residues with decreased frequencies are S1993.35, F2003.36 of helix III, and M3636.55 of helix VI (Figure 8F). The residue W3566.48 is a central residue of a transmission switch which is one of major molecular switches responsible for activation of GPCRs [38]. Molecular dynamics simulations for many GPCRs also suggest that the transmission switch is the first switch to be affected by the agonist binding [39,40]. Therefore, a modification of frequency of interactions between THC and the residues of the transmission switch (particularly F2003.36 and W3566.48 which are in tight proximity in CB1R) by CBD presence in the allosteric site most likely corresponds to the negative allosteric modulator properties of CBD. We have also compared the calculated ligand binding energies. The average THC- binding energy is lower (more negative, which means a more stable complex) by around 7.5 % in THC-only simulations (−53.9 kcal/mol) than in CBD-THC simulations (−50.1

Molecules 2021, 26, 2456 10 of 21

The binding mode of THC in the orthosteric site is influenced by the presence of CBD in the allosteric site 1. The residues for which the frequencies of interactions with THC are not changed are V1963.32, T1973.33 of helix III, L2765.40, W2795.43 of helix V, L3596.51 of helix VI, and S3837.39, C3867.42 of helix VII (Figure8F). CBD changes the average frequencies of interactions between THC and the receptor in such a way that some of them are increased and some decreased so the average binding pose of THC in the binding site is modified. The residues with increased frequencies are M103 of MPR, G1662.53, F1702.57 of helix II, and W3566.48 of helix VI. The residues with decreased frequencies are S1993.35, F2003.36 of helix III, and M3636.55 of helix VI (Figure8F). The residue W356 6.48 is a central residue of a transmission switch which is one of major molecular switches responsible for activation of GPCRs [38]. Molecular dynamics simulations for many GPCRs also suggest that the transmission switch is the first switch to be affected by the agonist binding [39,40]. Therefore, a modification of frequency of interactions between THC and the residues of the transmission switch (particularly F2003.36 and W3566.48 which are in tight proximity in CB1R) by CBD presence in the allosteric site most likely corresponds to the negative allosteric modulator properties of CBD. We have also compared the calculated ligand binding energies. The average THC-binding energy is lower (more negative, which means a more stable complex) by around 7.5% in THC-only simulations (−53.9 kcal/mol) than in CBD-THC simulations (−50.1 kcal/mol). Such observation is in agreement with experimental results indicating that allosteric modulation of CB1R by CBD decreases the binding affinity of THC and of the other orthosteric ligands [10]. In this paper we use for residues located in transmembrane helices the additional numbers in superscript which is the Ballesteros–Weinstein numbering [41] (extension of this scheme was proposed by Gloriam group [42]). This system was proposed for GPCRs to provide information about the relative positions of amino acids present in seven transmembrane helices. It contains two numbers separated by a dot: the first number indicates the number of the transmembrane helix (1–7) where the residue is located; the second number indicates a position of particular residue in relation to the most conserved residue in that helix, assigned number 50. Such numbers allow for direct comparison of residues among different GPCRs and especially in the most populated class A.

2.4. Stability of the N-Terminal Domain

The N-terminal domain of CB1 receptor was stable in all five CBD-THC simulations, however, it was significantly less stable in the THC-only simulations (Figure9A,B) and underwent a rearrangement (Figure9C). It should be noted that in the last stage of REST2 simulations we used a CBD bound CB1R and, therefore, we believe that the conformation of the N-terminal domain accommodated the allosteric ligand. Although initially the N-terminus was less stable in the THC-only simulations when we continued one of the simulations (sim7 in Figure9B) the N-terminus structure converged to more stable conformation also containing NTH3 (residues 78–92), β-sheet between 58–60 and 72–74, NTH1 (residues 29–35) and a very stable ‘N-cap’ (residues 1–20) (Figure7B). Therefore, we concluded that N-terminal domain is not only essential for stable binding of the allosteric ligands in ‘allosteric site 1’ but is also involved in allosteric modulation of the orthosteric binding. This hypothesis is supported by the mutagenesis studies of CB1R[29]. When this conformation has been achieved, we confirmed its stability in three independent 3 µs long MD simulations (Figure9C). In one of the three simulations (sim7C) the NTH2 helix together with the β-sheet (residues 58–74) were dissolved and therefore we observe high RMSF values for this region in simulation 7C. Molecules 2021, 26, 2456 11 of 21

2.5. The Effect of Truncation of the N-Terminus on the Energy of CBD Binding

The N-terminal domain of CB1R is essential for the stability of CBD in its allosteric binding site. In MD simulations of CBD-THC complex (5 × 1 µ and more) with a whole N-terminal domain none of the ligands exited its binding site (the average RMSD for each ligand was ~2.5 Å). However, in all simulations of ∆-88 (deletion of 1–88 residues) and ∆-98 (deletion of 1–98 residues) CB1R mutants (two simulations repeats for each mutant), the CBD binding mode was significantly distorted and the final pose was different for each of the four simulations (Figure 11). In two out of four cases (one simulation for ∆-88 and one simulation for ∆-98) CBD exited the allosteric site completely and within less than 500 ns it moved toward the orthosteric ligand binding channel between helices I and VII (this gate was described by us in our previous paper [36]) distorting the conformation of MPR significantly. In one of the two simulations for the ∆-8 (deletion of 1–8 residues) CB1R, CBD dissociated from the receptor completely after 350 ns and remained dissolved in the membrane for the rest of the simulation time. The average RMSD of CBD was 12.5 Å for ∆-8 receptor, 5.7 Å for ∆-88, and 7.7 Å for the ∆−98 CB1R mutant and the average binding energies of CBD were −63.29, −58.49 and −57.83 kcal/mol, respectively, while for the wild-type receptor the binding energy Molecules 2021, 26, x FOR PEER REVIEWwas −72.7 kcal/mol (Figure 12). The energies of CBD binding to mutants with a truncated 11 of 20 N-terminus (∆-8, ∆-88, ∆-98) are significantly higher than for the wild-type receptor, suggesting that a whole N-terminus is required for proper CBD binding.

FigureFigure 11.11.Comparison Comparison of CBD of bindingCBD binding modes (yellow modes surface (yellow for all poses)surface obtained for all after poses) 1 µs of obtained after 1µs of MD simulations for (A) WT CB R (with whole N-terminus), ﬁve CBD poses superimposed; and for MD simulations for (A) WT1 CB1R (with whole N-terminus), five CBD poses superimposed; and for (B) CB1R with truncated N-terminus—four CBD poses (two for ∆-88 and two for ∆-98 CB1R). All (B) CB1R with truncated N-terminus—four CBD poses (two for Δ-88 and two for Δ-98 CB1R). All poses were superimposed using receptor structure as a reference. Orthosteric agonist, THC, is shown posesin green. were superimposed using receptor structure as a reference. Orthosteric agonist, THC, is shown in green.

The average values of THC binding energy for Δ-8, Δ-88 and Δ-98 mutants (−51.1, −52.0 and −53.2 kcal/mol, respectively) are lower than those for THC bound to a wild-type receptor with CBD in the allosteric site (−50.1 kcal/mol) and higher than those recorded for THC bound to the wild-type receptor without CBD (−53.9 kcal/mol) (Figure 12). It suggests that even if CBD binds (at least to some extent) to CB1R with truncated N-terminus its impact on THC binding, and perhaps also its NAM activity, will be significantly reduced. Our results indicate that the larger fragment of the N-terminus is deleted the lower is the influence of CBD binding on the binding energy of THC (Figure 12).

Figure 12. Comparison of ligand binding energies for different mutants of CB1R with truncated N-terminus and a wild-type receptor for CBD-THC complex. The energies of CBD binding to the allosteric site (left) are shown in yellow while the energies of THC binding to the orthosteric site (right) are shown in green. All reported binding energies are the average values obtained from several (5 for WT, and 2 for mutants) 1 µs long MD simulations and they were performed for CBD- THC complex (except for the last rectangle). Δ-8 (deletion of 1–8 residues), Δ-88 (deletion of 1–88 residues) and Δ-98 (deletion of 1–98 residues).

Molecules 2021, 26, x FOR PEER REVIEW 11 of 20

Figure 11. Comparison of CBD binding modes (yellow surface for all poses) obtained after 1µs of MD simulations for (A) WT CB1R (with whole N-terminus), five CBD poses superimposed; and for (B) CB1R with truncated N-terminus—four CBD poses (two for Δ-88 and two for Δ-98 CB1R). All

Molecules 2021poses, 26, 2456 were superimposed using receptor structure as a reference. Orthosteric agonist, THC, is 12 of 21 shown in green.

The average values of THC binding energy for Δ-8, Δ-88 and Δ-98 mutants (−51.1, −52.0 and −53.2 kcal/mol,The respectively) average values are of lower THC binding than those energy for forTHC∆-8, bound∆-88 andto a∆ wild-98 mutants-type (−51.1, − − receptor with CBD52.0 in the and allosteric53.2 kcal/mol, site (− respectively)50.1 kcal/mol) are lowerand higher than those than for those THC boundrecorded to a wild-type receptor with CBD in the allosteric site (−50.1 kcal/mol) and higher than those recorded for THC bound to the wild-type receptor without CBD (−53.9 kcal/mol) (Figure 12). It sug- for THC bound to the wild-type receptor without CBD (−53.9 kcal/mol) (Figure 12). It gests that even if CBD binds (at least to some extent) to CB1R with truncated N-terminus suggests that even if CBD binds (at least to some extent) to CB1R with truncated N-terminus its impact on THCits binding, impact on and THC perhaps binding, andalso perhaps its NAM also activity, its NAM activity,will be willsignificantly be signiﬁcantly re- reduced. duced. Our resultsOur indicate results that indicate the larger that the fragment larger fragment of the N of-terminus the N-terminus is deleted is deleted the lower the lower is the is the influence ofinﬂuence CBD bin ofding CBD on binding the binding on the bindingenergy energyof THC of (Figure THC (Figure 12). 12).

Figure 12. Comparison of ligand binding energies for different mutants of CB R with truncated N-terminus and a wild-type Figure 12. Comparison of ligand binding energies for different1 mutants of CB1R with truncated receptorN- forterminus CBD-THC and complex. a wild-type The energiesreceptor of for CBD CBD binding-THC tocomplex. the allosteric The siteenergies (left) areof CBD shown binding in yellow to whilethe the energiesallosteric of THC bindingsite (left to) theare orthostericshown in siteyellow (right while) are shownthe energies in green. of All THC reported binding binding to the energies orthosteric are the site average values( obtainedright) are from shown several in(5 green. for WT, All and reported 2 for mutants) binding 1 µ senergies long MD are simulations the average and theyvalues were obtained performed from for CBD-THCsev- complexeral (except (5 for for WT, the lastand rectangle). 2 for mutants)∆-8 (deletion 1 µs long of 1–8 MD residues), simulations∆-88 (deletion and they of 1–88were residues) performed and ∆ for-98 CBD (deletion- of 1–98 residues).THC complex (except for the last rectangle). Δ-8 (deletion of 1–8 residues), Δ-88 (deletion of 1–88 residues) and Δ-98 (deletion of 1–98 residues). 2.6. The Dynamic Contact Networks The dynamic contact networks for CBD-THC and THC-only simulations are different. The most significant changes in frequencies of certain interactions in MD simulations are observed for hydrogen bonds, water mediated hydrogen bonds and van der Waals interactions networks (Figure 13). The interaction frequency and the interaction frequency difference (∆freq) is defined in Section 3.5 of Methods. Comparison of the contact networks of CBD-THC and THC-only MD simulations has revealed that several residues from the N-terminal domain (such as E91, E94, N95, I96, and C107) participate in the allosteric interaction network essential for the communication between the allosteric and orthosteric binding sites. Importance of disulfide bond C98–C107 for the allosteric-orthosteric coupling was strengthened since one of its residues, C107, formed a hydrogen bond with residue N95 in all MD simulations of CBD-THC complex while this bond was broken in all MD simulations of THC-only complex (Figure 13A,D). We have also found that the residues V110MPR, K3737.29 and T3777.33 constitute a switch that is involved in the allosteric modulation of CB1R by CBD. In the presence of CBD a hydrogen bond V110-T3777.33 is predominantly formed (Figure 14A) while in absence of CBD a hydrogen bond K3737.29-T3777.33 is formed much more often (Figure 14B). For those interactions we observed one of the highest average frequency differences between CBD-THC and THC-only systems, 0.78 and −0.74, for V110-T3777.33 and K3737.29-T3777.33, respectively (H-bonds in Figure 13A,D). Molecules 2021, 26, x FOR PEER REVIEW 12 of 20

2.6. The Dynamic Contact Networks The dynamic contact networks for CBD-THC and THC-only simulations are different. The most significant changes in frequencies of certain interactions in MD simulations are observed for hydrogen bonds, water mediated hydrogen bonds and van der Waals Moleculesinteractions2021, 26, 2456 networks (Figure 13). The interaction frequency and the interaction frequency13 of 21 difference (Δfreq) is defined in Section 3.5 of Methods.

Figure 13. 13Differences. Differences in contact in contact frequencies frequencies between CBD-THC between and CBD THC-only-THC and complexes; THC- (onlyA) Hydrogen complexes; bonds—both (A) Hy- drogenside views; bonds (B) Water—both mediated side hydrogenviews; (B bonds—the) Water mediated extracellular hydrogen and side view; bonds (C) van—the der extracellular Waals interactions—the and side view;extracellular (C) van and der side Waals view. Red interactions sticks represent—the contacts extracell occurringular moreand frequentlyside view. for Red CBD-THC sticks simulations, represent and contacts blue ones are the contacts occurring more frequently in THC-only simulations. The thickness of a stick as well as a color intensity occurring more frequently for CBD-THC simulations, and blue ones are the contacts occurring more correspond to the average frequency difference between those two populations of trajectories; (D) The interaction frequency frequentlyﬁngerprints forin CBD-THCTHC-only and simulations. THC-only complexes The thickness (5 simulations of a stick for each as complex).well as a Presented color intensity interactions correspond are very tolikely the participating average frequency in the allosteric-orthosteric difference between communication those network two populations since their frequencies of trajectories; differ signiﬁcantly (D) The between interac- tionsimulations frequency with and fingerprints without an allosteric for CBD ligand-THC (CBD). and Among THC all-only possible complexes interactions (5 only simulations the hydrogen for bonds, each water complex).mediated Presented hydrogen bonds,interactions and ionic are interactions very likely are shown. participating in the allosteric-orthosteric communication network since their frequencies differ significantly between simulations with and without an allosteric ligand (CBD). Among all possible interactions only the hydrogen bonds, water mediated hydrogen bonds, and ionic interactions are shown.

Comparison of the contact networks of CBD-THC and THC-only MD simulations has revealed that several residues from the N-terminal domain (such as E91, E94, N95, I96, and C107) participate in the allosteric interaction network essential for the communication between the allosteric and orthosteric binding sites. Importance of disulfide bond

Molecules 2021, 26, x FOR PEER REVIEW 13 of 20

C98–C107 for the allosteric-orthosteric coupling was strengthened since one of its residues, C107, formed a hydrogen bond with residue N95 in all MD simulations of CBD- THC complex while this bond was broken in all MD simulations of THC-only complex (Figure 13A,D). We have also found that the residues V110MPR, K3737.29 and T3777.33 constitute a switch that is involved in the allosteric modulation of CB1R by CBD. In the presence of CBD a hydrogen bond V110-T3777.33 is predominantly formed (Figure 14A) while in absence of CBD a hydrogen bond K3737.29-T3777.33 is formed much more often (Figure 14B). Molecules 2021, 26, 2456 For those interactions we observed one of the highest average frequency differences14 of be- 21 tween CBD-THC and THC-only systems, 0.78 and −0.74, for V110-T3777.33 and K3737.29- T3777.33, respectively (H-bonds in Figure 13A,D).

MPR 7.33 7.29 FigureFigure 14.14. TheThe molecular molecular switch switch formed formed by by residues residues V110 V110MPR-T377-T377-7.33K373-K373 as7.29 anas important an important element of the allosteric modulation. (A) In CBD-THC simulations the switch exists predominantly in element of the allosteric modulation. (A) In CBD-THC simulations the switch exists predominantly the configuration where a hydrogen bond V110MPR-T3777.33 (side chain-main chain) is formed; (B) in the configuration where a hydrogen bond V110MPR-T3777.33 (side chain-main chain) is formed; in THC-only simulations the hydrogen bond of side chain of T3777.33 is much more often formed B 7.33 (within) in THC-only the same simulationshelix VII to theK373 hydrogen7.29 (side chain bond- ofmain side chain). chain of T377 is much more often formed within the same helix VII to K3737.29 (side chain-main chain). The residue M103MPR, which is located between the allosteric and orthosteric sites MPR (FigureThe 10 residue), interacts M103 both with, which CBD is in located the allosteric between site the and allosteric with THC and in orthosteric the orthos sitesteric (Figuresite, and 10 the), interacts frequencies both withof these CBD interactions in the allosteric are high site and in CBD with- THC incomplex: the orthosteric 0.61 for site, in- andteraction the frequencies with CBD of (Figure these interactions 8E), and 0.80 are for high interaction in CBD-THC with complex: THC (Figure 0.61 for 8F). interaction We con- withclude CBD that (FigureM103MPR8E), could and be 0.80 relevant for interaction for the allosteric with THC-orthosteric (Figure8F). communica We concludetion thatsince MPR M103the frequencycould of be the relevant M103 for-THC the interaction allosteric-orthosteric is highly communicationdependent on the since presence the frequency or ab- ofsence the M103-THCof CBD in the interaction allosteric is site, highly 0.80 dependent in CBD-THC on thecomplex, presence and or 0.46 absence in THC of CBD-only in com- the allosteric site, 0.80 in CBD-THC complex, and 0.46 in THC-only complex (Figure8F). The plex (Figure 8F). The residue I105MPR is also located between the allosteric and orthosteric residue I105MPR is also located between the allosteric and orthosteric sites, however, the sites, however, the frequencies for I105-THC interaction are very low (Figure 8F), and frequencies for I105-THC interaction are very low (Figure8F), and therefore we cannot say therefore we cannot say if this residue participates in the allosteric-orthosteric coupling. if this residue participates in the allosteric-orthosteric coupling. The four ionic interactions (Figure 13D), R405H8-E1331.49, R14N-term-E100MPR, E2735.37- The four ionic interactions (Figure 13D), R405H8-E1331.49, R14N-term-E100MPR, E2735.37- K370ECL3, and E94MPR-K3737.29 were found significantly more often in the CBD-THC com- K370ECL3, and E94MPR-K3737.29 were found significantly more often in the CBD-THC plex than in THC-only, with the average frequency difference (Δfreq): 0.46, 0.53, 0.54 and complex than in THC-only, with the average frequency difference (∆ ): 0.46, 0.53, 0.54 0.75, respectively. Perhaps, the mutagenesis studies could elucidate thefreq influence of the and 0.75, respectively. Perhaps, the mutagenesis studies could elucidate the influence of above residues on the allosteric-orthosteric coupling. The extracellular loop 2 (ECL2), and the above residues on the allosteric-orthosteric coupling. The extracellular loop 2 (ECL2), particularly the residues N256, C257, Q261, S262, also participate in the allosteric-or- and particularly the residues N256, C257, Q261, S262, also participate in the allostericthosteric coupling mechanism especially that in some THC-only simulations we observed orthosteric coupling mechanism especially that in some THC-only simulations we observed thethe interactionsinteractions betweenbetween N-terminal N-terminal domain domain and and ECL2 ECL2 loop. loop. Interestingly, Interestingly, not not only only the the N- terminusN-terminus residues residues but but also also some some residues residues from from cytoplasmic cytoplasmic part part of the of receptorthe receptor(Figure (Figure7C) are7C) involvedare involved in the in allostericthe allosteric modulation modulation and canand influencecan influence the cellular the cellular response. response.

3.3. MethodsMethods 3.1.3.1. Modeling of theThe Membrane-Proximal Membrane-Proximal Region Region (MPR) (MPR) of of CB CB1R1R

TheThe crystalcrystal structurestructure ofof CBCB11R (PDB id: 5U09) was converted to thethe wild-typewild-type (T210A(T210A mutationmutation reversed),reversed), thethe residues 96–9996–99 were added using YASARA v.19.1.27 programprogram [[43]43] andand the regions regions 96 96–100–100 and and 107 107–108–108 were were then then sampled sampled using using the theRosetta Rosetta loopmodel loopmodel pro- protocoltocol [44] [44 with] with a condition a condition that that the the C98 C98–C107–C107 disul disulﬁdefide bond bond must must be maintained. FromFrom the 500 models obtained eight were selected based on Rosetta total score. Those eight selected structures were placed into a hydrated POPC bilayer and 250 ns of a full atom MD simulation was performed in AMBER v.18 program [45] for each of them. In ﬁve out of eight simulations the 96–112 fragment converged to a similar stable structure and the one with the lowest nonbonded energy of MPR was selected for further studies.

3.2. Building and Elucidating the Structure of Whole N-Terminus The remaining 95 residues of the N-terminal domain (residues 1–95) were added from the sequence of a wild-type human CB1R obtained from the Uniprot database [46]. The BuildLoop command from YASARA [47] was used to build the initial model of N- terminus and its conformation was sampled using the Rosetta loopmodel protocol [44]. The selected structures from the clustered Rosetta solutions were used for exploration of Molecules 2021, 26, x FOR PEER REVIEW 14 of 20

the 500 models obtained eight were selected based on Rosetta total score. Those eight selected structures were placed into a hydrated POPC bilayer and 250 ns of a full atom MD simulation was performed in AMBER v.18 program [45] for each of them. In five out of eight simulations the 96–112 fragment converged to a similar stable structure and the one with the lowest nonbonded energy of MPR was selected for further studies. 3.2. Building and Elucidating The Structure of Whole N-Terminus The remaining 95 residues of the N-terminal domain (residues 1–95) were added from the sequence of a wild-type human CB1R obtained from the Uniprot database [46]. The BuildLoop command from YASARA [47] was used to build the initial model of N- Molecules 2021 26 , , 2456 terminus and its conformation was sampled using the Rosetta loopmodel protocol15 of[44] 21 . The selected structures from the clustered Rosetta solutions were used for exploration of the conformational space of the CB1R N-terminal domain employing the Replica Exchange

theMolecular conformational Dynamics space with of the Solute CB1R Tempering N-terminal (REST2domain employing[37]) algorithm the Replica working Exchange within MolecularNAMD 2.13 Dynamics platform with [48] Solute using Tempering CHARMM36m (REST2 force [37]) field algorithm [49]. The working all-atom within REST2 NAMD sim- 2.13ulations platform were [ 48performed] using CHARMM36m for the whole forceCB1 cannabinoid ﬁeld [49]. The receptor all-atom (residues REST2 simulations 1–412) with werethe N performed-terminus (residues for the whole 1–98, unless CB1 cannabinoid specified otherwise) receptor selected (residues as 1–412)a hot region with the(Figure N- terminus15). (residues 1–98, unless speciﬁed otherwise) selected as a hot region (Figure 15).

FigureFigure 15. 15.Sample Sample structures structures of of CB CB1 receptor1 receptor with with complete complete N-terminal N-terminal domain domain at each at selectioneach selection step ofstep the of REST2 the REST2 MD simulations. MD simulation The spans. The of span hotregion of hot ofregion N-terminal of N-terminal domain anddomain number and ofnumber replicas of inreplicas each step in each are speciﬁed step are specified on the right. on the right.

TheThe receptor receptor was was immersed immersed in in a hydrated a hydrated 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho- 1-palmitoyl-2-oleoyl-sn-glycero-3-phos- cholinephocholine (POPC) (POPC) bilayer bilayer enriched enriched by 20% by ( n/n20%) cholesterol.(n/n) cholesterol. The whole The w systemhole system was generated was gen- inerated CHARMM-GUI in CHARMM webserver-GUI webserver [50] which [50] was which used was for used system for preparationsystem preparation and input and ﬁles in- generation. The standard system contained ~70,000 atoms and ~130 lipid molecules (cholesterol + POPC) including Na+ and Cl− counter ions in total ion concentration 0.15 M for system charge neutrality. The average dimensions of a periodic cell were 74 Å × 74 Å × 140 Å. The TIP3P water model, parametrized to use with CHARMM force ﬁelds, was employed. Each REST2 simulation was run in 16 replicas, all starting from the same initial structure. All simulations were performed using the 2 fs time step and all bond lengths to hydrogen atoms were constrained using SHAKE [51] algorithm allowing to use a longer time step than 1 fs. The actual temperature of the system was 300 K, while the interactions for the hot region were scaled for temperatures 300–600 K. Temperature exchange attempts were performed every 100 steps (200 fs). The electrostatic interactions cutoff was set to 12 Å with 10–12 Å switching function, while the cutoff for pairlist generation was set to 13.5 Å. Long-range electrostatic interactions were computed using the Particle Mesh Ewald [52] summation scheme with PMEGridSpacing parameter set to 1.0. In the above simulations the group pressure and the Langevin dynamics were used. The best structures of CB1R Molecules 2021, 26, 2456 16 of 21

were elucidated from REST2 simulations based on their lowest nonbonded energies and low local RMSF values of the N-terminus during the simulations. For the last step of REST2 simulation the residues 20–79 instead of 1–98 were selected as the hot region due to their high RMSF values and CBD was docked to the allosteric site (Figure 15).

3.3. Ligand Docking

The most stable structure of CB1R with the whole N-terminus obtained from REST2 simulations and selected based on RMSF and energy analysis, was used for ligand docking. CBD was redocked to the allosteric site 1 [31] (positioned between the extracellular ends of TM1 and TM2, defined by residues: I105, E106, C107) after the last step of REST2 set of simulations, while ∆9-THC was docked to the orthosteric binding site (defined by residues: I119, F174, L193, W356, F381, S383, M384) [34]. To perform docking of flexible ligands into protein binding sites we used Genetic Optimization for Ligand Docking (GOLD) v.5.7.3 [53]. Ligand poses were verified for stability by MD simulations. Only one stable pose was identified for CBD and it was stable for >1 µs of total MD simulation (in 5 simulation repeats). For each of the two ligand-receptor complexes (1st complex of CB1R with CBD and THC, and 2nd complex of CB1R with THC only) five independent all-atom and unbiased MD simulations were performed (2 × 5 × 1 µs).

3.4. All-Atom MD Simulations of CB1R Receptor with Ligands

Standard molecular dynamics simulations of the CB1 cannabinoid receptor with ap- propriate ligands immersed in a hydrated POPC bilayer (enriched by 20% (n/n) cholesterol), were performed in AMBER18 using CHARMM36m force field [49]. All systems were generated in CHARMM-GUI service [50] and the receptor orientation in the membrane was set according to the OPM database [54]. The standard system contained ~70,000 atoms, ~127 lipid molecules (cholesterol + POPC), 43 Na+ cations and 50 Cl− anions (total ion concentration was 0.15 M). The average dimensions of a periodic cell were 74 Å × 74 Å × 140 Å. TIP3P water model parametrized to use with CHARMM force fields, was employed. Before each MD simulation the system was submitted to a restrained energy minimization (5000 cycles). The first 2500 minimization cycles were performed with a steepest-descent method and after that conjugate gradient was switched on. Next, a six-step equilibration was performed (375 ps total) at a constant pressure and temperature (NPT ensemble; 310 K, 1 bar). During equilibration the restraints were released gradually, until the last step (last 100 ps of equilibration), in which no restraints were used. In the production simulations (as well as in the last three equilibration steps) all bond lengths to hydrogen atoms were constrained using SHAKE algorithm [51,55], which allowed us to use a longer time step (2 fs instead of 1 fs). Van der Waals and short-range electrostatic interactions cutoff =12 Å, with 10–12 Å switching function, were used. Long-range electrostatic interactions were computed using the Particle Mesh Ewald [52] summation scheme. The MD simulations of 1 µs lengths each were performed for the CB1 receptor with the complete N-terminus (model obtained from REST2 simulations) and the following ligands bound: (1) five simulations for ∆9-THC in the orthosteric binding site and CBD in the allosteric binding site, and (2) five simulations for ∆9-THC in the orthosteric binding site and without the allosteric ligand.

3.5. Contact Network Analysis The Pytraj tool (https://github.com/Amber-MD/pytraj (accessed on 6 March 2021)), a Python front-end of the popular cpptraj package [56], was used for trajectory processing. GetContacts software (https://getcontacts.github.io/ (accessed on 6 March 2021)) was used for analysis of the contact network within the receptor, and for calculation of the ligand- receptor and receptor-receptor interaction frequencies in each MD trajectory. The Flare plots (https://gpcrviz.github.io/ﬂareplot/ (accessed on 6 March 2021)) were used to visualize the residue-residue interactions. Based on the trajectory from the MD simulations using get_dynamic_contacts.py tool, we created a residue contact list for the MD simulation Molecules 2021, 26, 2456 17 of 21

range 100 ns–1 µs. Since all MD simulations started from the same structure of CB1R selected using the REST2 simulations, by skipping the first 100 ns we hoped to highlight the differences between the contact networks of receptors bound with a different set of ligands. Using get_contact_frequencies.py module we calculated contact frequencies for all interactions in each trajectory. The interaction frequency (Freq) ranges between 0 and 1: Freq = 1.0 means that a particular interaction is present in every single trajectory frame, Freq = 0.5 means the interaction is found only in 50% of the frames, while Freq = 0 means the interaction does not occur in the analyzed trajectory. By comparing contact frequencies between two sets of trajectories (1st for CB1R with CBD and THC, and 2nd for CB1R with THC-only; 5 trajectories for each set of ligands) we constructed the interaction fingerprints. The cutoff for residue-residue interaction fingerprint was set to 0.5 (which means that a particular interaction is considered only if its frequency is ≥0.5 in at least one of the trajectories) and for ligand-protein interaction fingerprints the cutoff was set to 0.3. Based on the analysis of numerical interaction fingerprints, we elucidated the interactions that are highly dependent on the presence or absence of CBD in the THC bound CB1 receptor. We performed a statistical analysis of the frequency distributions in two sets of trajectories and calculated the difference between average frequencies of a particular contact (∆freq, described by Equation (1)) in order to identify those contacts for which the frequency differed significantly between CBD-THC and THC-only CB1 receptor complexes. The Student’s t-test was performed to verify whether the distributions of frequencies are different between two sets of trajectories comparing the average frequencies of each contact between those sets. If the frequency distributions are classified as different (p value < 0.05) and the absolute value of the average frequency difference between the two populations of trajectories is higher than 0.4 (|∆freq| > 0.4) for a particular contact, then the contact is kept as a potentially meaningful for the allosteric-orthosteric coupling mechanism. We visualized the average frequency differences for these meaningful interactions as red and blue sticks linking the interacting residues (Figure 13A–C).

∆ f req = Freq(CBD+THC) − Freq(THC−only) (1)

3.6. Mutational Study of N-Terminus on CBD Binding to the CB1 Receptor In order to verify if the presence of the N-terminus is relevant for the allosteric binding of CBD to the CB1 receptor, we performed all-atom MD simulations of 1 µs lengths for CB1R with N-terminus truncated at various sites (∆-8, ∆-88, ∆-98) bound to CBD (in the allosteric site 1) and THC (in the orthosteric site). Then, we compared the binding energies and the RMSD values of ligands in all CB1R mutants with those values for ligand-bound wild-type receptor.

4. Conclusions

This is the first reported attempt of model the whole N-terminal domain of CB1R concluded by obtaining a relatively stable structure of this domain maintained throughout multiple MD simulations. Although for some parts of the N-terminus the RMSF values are in range 7–8 Å it also contains very stable regions with RMSF in range 1–3 Å. Besides, it cannot be ruled out that some regions of the N-terminus are flexible by nature and their conformation is highly sensitive to the allosteric ligand bound to allosteric site. This might explain why no experimental structure of the CB1R N-terminal domain has been determined so far. We observed that the particular conformation of N-terminus obtained for CBD-THC-CB1R complex was stable in all five simulations but it underwent a rearrangement in all five simulations of the THC-only-CB1R complex. Continuation of one of these MD simulations (4 µs in total) led to a different stable conformation of the N-terminus of the THC-only-CB1R complex. We found that CBD bound to the allosteric site is stable only when the whole N- terminus of CB1R is present. In the case of ∆-8, ∆-88 and ∆-98 mutants (deletion of 1–8 residues, 1–88 residues, and 1–98 residues, respectively), the binding energy of CBD is Molecules 2021, 26, 2456 18 of 21

significantly higher, indicating a less stable complex, and the ligand dissociates from the binding site in 50% of cases. By analyzing of MD simulations of CBD-THC and THC- only complexes, and comparing the residue contact networks within those systems, we elucidated the major interactions responsible for the allosteric modulation of CB1R by CBD: (i) the N-terminus residues with a possible molecular switch involving residues V110MPR, K3737.29 and T3777.33 with alternating hydrogen bonds between them, and (ii) the residues from the extracellular loop 2 (ECL2). The results of MD simulations indicate that the N-terminal domain of CB1R plays a critical role not only in the binding of CBD (and perhaps the other allosteric modulators) but also for the allosteric-orthosteric coupling mechanism. Better understanding of the role of the CB1R N-terminal domain, including elucidation of the allosteric-orthosteric interaction network, is extremely important for understanding the allosteric modulation mechanism in CB1R and can be potentially useful in a design of novel allosteric modulators. All of the above findings on the role of N-terminal domain need the experimental confirmation but they are also useful on their own for molecular modelers and also for drug discovery and design purposes.

Author Contributions: Conceptualization, J.J. and S.F.; methodology, J.J.; software, P.P.; validation, J.J., R.A. and R.P.; formal analysis, S.F.; investigation, J.J. and R.A.; resources, P.P.; data curation, J.J. and U.O.; writing—original draft preparation, J.J.; writing—review and editing, J.J. and S.F.; visualization, J.J. and U.O.; supervision, S.F.; funding acquisition, J.J. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by National Science Centre, Poland, grant PRELUDIUM 2016/23/N/NZ1/02960 for J.J. Part of computational work has been done at the Interdisciplinary Centre for Mathematical and Computational Modelling in Warsaw (grant no. G07-13 and GA71-27). Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data supporting reported results are deposited at Faculty of Chemistry, University of Warsaw, and are available upon request. Acknowledgments: We would like to thank Sunhwan Jo for consultations, priceless advice and a huge help with running REST2 simulations, Jana Selent and Tomasz St˛epniewskifor recommending GetContacts software used for contact network analysis, the whole CHARMM-GUI team for creating a great open access web service for MD input generation, Szymon Niewieczerzał for scientific consultations and Przemysław Miszta for help with software installation. The European COST Action CA18133 (ERNEST) is acknowledged for providing biannual forum for exchange ideas on GPCR research and for supporting mutual visits. Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Sample Availability: The structures and trajectories of CBD-THC-CB1R and THC-only-CB1R complexes are available from the authors.

Abbreviations

2-AG 2-arachidonylogycerol AEA N-arachidonoylethanolamine (anandamide) CB1R CB1 cannabinoid receptor CBD cannabidiol COVID-19 coronavirus disease 19 GPCR G protein coupled receptor MD molecular dynamics MPR membrane-proximal region NAM negative allosteric modulator Molecules 2021, 26, 2456 19 of 21

PAM positive allosteric modulator REST2 replica exchange. molecular dynamics with solute tempering RMSD root mean square deviation RMSF root mean square ﬂuctuation THC (−)-trans-∆9-tetrahydrocannabinol

References 1. Howlett, A.C.; Breivogel, C.S.; Childers, S.R.; Deadwyler, S.A.; Hampson, R.E.; Porrino, L.J. Cannabinoid physiology and pharmacology: 30 years of progress. Neuropharmacology 2004, 47 (Suppl. 1), 345–358. [CrossRef] 2. Stasiulewicz, A.; Znajdek, K.; Grudzien, M.; Pawinski, T.; Sulkowska, A.J.I. A Guide to Targeting the Endocannabinoid System in Drug Design. Int. J. Mol. Sci. 2020, 21, 2778. [CrossRef] 3. Alaverdashvili, M.; Laprairie, R.B. The future of type 1 cannabinoid receptor allosteric ligands. Drug Metab. Rev. 2018, 50, 14–25. [CrossRef] 4. Lazzari, P.; Pau, A.; Tambaro, S.; Asproni, B.; Ruiu, S.; Pinna, G.; Mastinu, A.; Curzu, M.M.; Reali, R.; Bottazzi, M.E.; et al. Synthesis and pharmacological evaluation of novel 4-alkyl-5-thien-2’-yl pyrazole carboxamides. Cent. Nerv. Syst. Agents Med. Chem. 2012, 12, 254–276. [CrossRef] 5. Mastinu, A.; Pira, M.; Pani, L.; Pinna, G.A.; Lazzari, P. NESS038C6, a novel selective CB1 antagonist agent with anti-obesity activity and improved molecular profile. Behav. Brain Res. 2012, 234, 192–204. [CrossRef] 6. Lu, D.; Immadi, S.S.; Wu, Z.; Kendall, D.A. Translational potential of allosteric modulators targeting the cannabinoid CB1 receptor. Acta Pharmacol. Sin. 2019, 40, 324–335. [CrossRef] 7. Nguyen, T.; Li, J.X.; Thomas, B.F.; Wiley, J.L.; Kenakin, T.P.; Zhang, Y. Allosteric Modulation: An Alternate Approach Targeting the Cannabinoid CB1 Receptor. Med. Res. Rev. 2017, 37, 441–474. [CrossRef] 8. Price, M.R.; Baillie, G.L.; Thomas, A.; Stevenson, L.A.; Easson, M.; Goodwin, R.; McLean, A.; McIntosh, L.; Goodwin, G.; Walker, G.; et al. Allosteric modulation of the cannabinoid CB1 receptor. Mol. Pharmacol. 2005, 68, 1484–1495. [CrossRef] 9. Wootten, D.; Christopoulos, A.; Sexton, P.M. Emerging paradigms in GPCR allostery: Implications for drug discovery. Nat. Rev. Drug Discov. 2013, 12, 630–644. [CrossRef] 10. Laprairie, R.B.; Bagher, A.M.; Kelly, M.E.; Denovan-Wright, E.M. Cannabidiol is a negative allosteric modulator of the cannabinoid CB1 receptor. Br. J. Pharmacol. 2015, 172, 4790–4805. [CrossRef] 11. Bracey, M.H.; Hanson, M.A.; Masuda, K.R.; Stevens, R.C.; Cravatt, B.F. Structural adaptations in a membrane enzyme that terminates endocannabinoid signaling. Science 2002, 298, 1793–1796. [CrossRef][PubMed] 12. Gado, F.; Meini, S.; Bertini, S.; Digiacomo, M.; Macchia, M.; Manera, C. Allosteric modulators targeting cannabinoid cb1 and cb2 receptors: Implications for drug discovery. Future Med. Chem. 2019, 11, 2019–2037. [CrossRef][PubMed] 13. Gentry, P.R.; Sexton, P.M.; Christopoulos, A. Novel Allosteric Modulators of G Protein-coupled Receptors. J. Biol. Chem. 2015, 290, 19478–19488. [CrossRef][PubMed] 14. Slivicki, R.A.; Xu, Z.; Kulkarni, P.M.; Pertwee, R.G.; Mackie, K.; Thakur, G.A.; Hohmann, A.G. Positive Allosteric Modulation of Cannabinoid Receptor Type 1 Suppresses Pathological Pain Without Producing Tolerance or Dependence. Biol. Psychiatry 2018, 84, 722–733. [CrossRef] 15. Hryhorowicz, S.; Kaczmarek-Rys, M.; Andrzejewska, A.; Staszak, K.; Hryhorowicz, M.; Korcz, A.; Slomski, R. Allosteric Modulation of Cannabinoid Receptor 1-Current Challenges and Future Opportunities. Int. J. Mol. Sci. 2019, 20, 5874. [CrossRef] 16. Christopoulos, A.; Kenakin, T. G protein-coupled receptor allosterism and complexing. Pharmacol. Rev. 2002, 54, 323–374. [CrossRef] 17. Morales, P.; Jagerovic, N. Novel approaches and current challenges with targeting the endocannabinoid system. Expert. Opin. Drug Discov. 2020, 15, 917–930. [CrossRef] 18. Mastinu, A.; Ribaudo, G.; Ongaro, A.; Bonini, S.A.; Memo, M.; Gianoncelli, A. Critical Review on the Chemical Aspects of Cannabidiol (CBD) and Harmonization of Computational Bioactivity Data. Curr. Med. Chem. 2021, 28, 213–237. [CrossRef] 19. Esposito, G.; Pesce, M.; Seguella, L.; Sanseverino, W.; Lu, J.; Corpetti, C.; Sarnelli, G. The potential of cannabidiol in the COVID-19 pandemic. Br. J. Pharmacol. 2020, 177, 4967–4970. [CrossRef] 20. Costiniuk, C.T.; Jenabian, M.A. Acute inflammation and pathogenesis of SARS-CoV-2 infection: Cannabidiol as a potential anti-inflammatory treatment? Cytokine Growth Factor Rev. 2020, 53, 63–65. [CrossRef] 21. Salles, E.L.; Khodadadi, H.; Jarrahi, A.; Ahluwalia, M.; Paffaro, V.A., Jr.; Costigliola, V.; Yu, J.C.; Hess, D.C.; Dhandapani, K.M.; Baban, B. Cannabidiol (CBD) modulation of apelin in acute respiratory distress syndrome. J. Cell. Mol. Med. 2020.[CrossRef] 22. Byrareddy, S.N.; Mohan, M. SARS-CoV2 induced respiratory distress: Can cannabinoids be added to anti-viral therapies to reduce lung inflammation? Brain. Behav. Immun. 2020, 87, 120–121. [CrossRef] 23. Huang, C.; Wang, Y.; Li, X.; Ren, L.; Zhao, J.; Hu, Y.; Zhang, L.; Fan, G.; Xu, J.; Gu, X.; et al. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet 2020, 395, 497–506. [CrossRef] 24. Manuzak, J.A.; Gott, T.M.; Kirkwood, J.S.; Coronado, E.; Hensley-McBain, T.; Miller, C.; Cheu, R.K.; Collier, A.C.; Funderburg, N.T.; Martin, J.N.; et al. Heavy Cannabis Use Associated with Reduction in Activated and Inflammatory Immune Cell Frequencies in Antiretroviral Therapy-Treated Human Immunodeficiency Virus-Infected Individuals. Clin. Infect. Dis. 2018, 66, 1872–1882. [CrossRef] Molecules 2021, 26, 2456 20 of 21

25. Wang, B.; Kovalchuk, A.; Li, D.; Ilnytskyy, Y.; Kovalchuk, I.; Kovalchuk, O. In Search of Preventative Strategies: Novel Anti- Inflammatory High-CBD Cannabis Sativa Extracts Modulate ACE2 Expression in COVID-19 Gateway Tissues. Preprints 2020. [CrossRef] 26. Russo, E.; Guy, G.W. A tale of two cannabinoids: The therapeutic rationale for combining tetrahydrocannabinol and cannabidiol. Med. Hypotheses 2006, 66, 234–246. [CrossRef][PubMed] 27. Russo, E.B. Taming THC: Potential cannabis synergy and phytocannabinoid-terpenoid entourage effects. Br. J. Pharmacol. 2011, 163, 1344–1364. [CrossRef][PubMed] 28. Pertwee, R.G.; Ross, R.A.; Craib, S.J.; Thomas, A. (-)-Cannabidiol antagonizes cannabinoid receptor agonists and noradrenaline in the mouse vas deferens. Eur. J. Pharmacol. 2002, 456, 99–106. [CrossRef] 29. Fay, J.F.; Farrens, D.L. The membrane proximal region of the cannabinoid receptor CB1 N-terminus can allosterically modulate ligand affinity. Biochemistry 2013, 52, 8286–8294. [CrossRef] 30. Baillie, G.L.; Horswill, J.G.; Anavi-Goffer, S.; Reggio, P.H.; Bolognini, D.; Abood, M.E.; McAllister, S.; Strange, P.G.; Stephens, G.J.; Pertwee, R.G.; et al. CB(1) receptor allosteric modulators display both agonist and signaling pathway specificity. Mol. Pharmacol. 2013, 83, 322–338. [CrossRef] 31. Sabatucci, A.; Tortolani, D.; Dainese, E.; Maccarrone, M. In silico mapping of allosteric ligand binding sites in type-1 cannabinoid receptor. Biotechnol. Appl. Biochem. 2018, 65, 21–28. [CrossRef][PubMed] 32. Chung, H.; Fierro, A.; Pessoa-Mahana, C.D. Cannabidiol binding and negative allosteric modulation at the cannabinoid type 1 receptor in the presence of delta-9-tetrahydrocannabinol: An In Silico study. PLoS ONE 2019, 14, e0220025. [CrossRef][PubMed] 33. Kooistra, A.J.; Mordalski, S.; Pandy-Szekeres, G.; Esguerra, M.; Mamyrbekov, A.; Munk, C.; Keseru, G.M.; Gloriam, D.E. GPCRdb in 2021: Integrating GPCR sequence, structure and function. Nucleic Acids Res. 2021, 49, D335–D343. [CrossRef][PubMed] 34. Shao, Z.; Yin, J.; Chapman, K.; Grzemska, M.; Clark, L.; Wang, J.; Rosenbaum, D.M. High-resolution crystal structure of the human CB1 cannabinoid receptor. Nature 2016, 540, 602–606. [CrossRef] 35. Murphy, J.W.; Kendall, D.A. Integrity of extracellular loop 1 of the human cannabinoid receptor 1 is critical for high-affinity binding of the ligand CP 55,940 but not SR 141716A. Biochem. Pharmacol. 2003, 65, 1623–1631. [CrossRef] 36. Jakowiecki, J.; Orzel, U.; Chawananon, S.; Miszta, P.; Filipek, S. The Hydrophobic Ligands Entry and Exit from the GPCR Binding Site-SMD and SuMD Simulations. Molecules 2020, 25, 1930. [CrossRef] 37. Jo, S.; Jiang, W. A generic implementation of replica exchange with solute tempering (REST2) algorithm in NAMD for complex biophysical simulations. Comput. Phys. Commun. 2015, 197, 304–311. [CrossRef] 38. Filipek, S. Molecular switches in GPCRs. Curr. Opin. Struct. Biol. 2019, 55, 114–120. [CrossRef] 39. Yuan, S.; Hu, Z.; Filipek, S.; Vogel, H. W246(6.48) opens a gate for a continuous intrinsic water pathway during activation of the adenosine A2A receptor. Angew. Chem. Int. Ed. 2015, 54, 556–559. [CrossRef] 40. Yuan, S.; Peng, Q.; Palczewski, K.; Vogel, H.; Filipek, S. Mechanistic Studies on the Stereoselectivity of the Serotonin 5-HT1A Receptor. Angew. Chem. Int. Ed. 2016, 55, 8661–8665. [CrossRef] 41. Ballesteros, J.A.; Weinstein, H. Integrated methods for the construction of three-dimensional models and computational probing of structure-function relations in G protein-coupled receptors. Methods Neurosci. 1995, 25, 366–428. 42. Isberg, V.; de Graaf, C.; Bortolato, A.; Cherezov, V.; Katritch, V.; Marshall, F.H.; Mordalski, S.; Pin, J.P.; Stevens, R.C.; Vriend, G.; et al. Generic GPCR residue numbers—Aligning topology maps while minding the gaps. Trends Pharmacol. Sci. 2015, 36, 22–31. [CrossRef] 43. Krieger, E.; Vriend, G. New ways to boost molecular dynamics simulations. J. Comput. Chem. 2015, 36, 996–1007. [CrossRef] 44. Mandell, D.J.; Coutsias, E.A.; Kortemme, T. Sub-angstrom accuracy in protein loop reconstruction by robotics-inspired conformational sampling. Nat. Methods 2009, 6, 551–552. [CrossRef][PubMed] 45. Salomon-Ferrer, R.; Case, D.A.; Walker, R.C. An overview of the Amber biomolecular simulation package. Wires Comput. Mol. Sci. 2013, 3, 198–210. [CrossRef] 46. UniProt, C. UniProt: A worldwide hub of protein knowledge. Nucleic Acids Res. 2019, 47, D506–D515. [CrossRef] 47. Canutescu, A.A.; Dunbrack, R.L., Jr. Cyclic coordinate descent: A robotics algorithm for protein loop closure. Protein Sci. 2003, 12, 963–972. [CrossRef] 48. Phillips, J.C.; Hardy, D.J.; Maia, J.D.C.; Stone, J.E.; Ribeiro, J.V.; Bernardi, R.C.; Buch, R.; Fiorin, G.; Henin, J.; Jiang, W.; et al. Scalable molecular dynamics on CPU and GPU architectures with NAMD. J. Chem. Phys. 2020, 153, 044130. [CrossRef][PubMed] 49. Huang, J.; Rauscher, S.; Nawrocki, G.; Ran, T.; Feig, M.; de Groot, B.L.; Grubmuller, H.; MacKerell, A.D., Jr. CHARMM36m: An improved force field for folded and intrinsically disordered proteins. Nat. Methods 2017, 14, 71–73. [CrossRef] 50. Jo, S.; Kim, T.; Iyer, V.G.; Im, W. CHARMM-GUI: A web-based graphical user interface for CHARMM. J. Comput. Chem. 2008, 29, 1859–1865. [CrossRef] 51. Ryckaert, J.-P.; Ciccotti, G.; Berendsen, H.J. Numerical integration of the cartesian equations of motion of a system with constraints: Molecular dynamics of n-alkanes. J. Comput. Phys. 1977, 23, 327–341. [CrossRef] 52. Essmann, U.; Perera, L.; Berkowitz, M.L.; Darden, T.; Lee, H.; Pedersen, L.G. A Smooth Particle Mesh Ewald Method. J. Chem. Phys. 1995, 103, 8577–8593. [CrossRef] 53. Jones, G.; Willett, P.; Glen, R.C.; Leach, A.R.; Taylor, R. Development and validation of a genetic algorithm for flexible docking. J. Mol. Biol. 1997, 267, 727–748. [CrossRef][PubMed] Molecules 2021, 26, 2456 21 of 21

54. Lomize, M.A.; Pogozheva, I.D.; Joo, H.; Mosberg, H.I.; Lomize, A.L. OPM database and PPM web server: Resources for positioning of proteins in membranes. Nucleic Acids Res. 2012, 40, D370–D376. [CrossRef][PubMed] 55. Miyamoto, S.; Kollman , P.A. SETTLE: An analytical version of the SHAKE and RATTLE algorithm for rigid water models. J. Comput. Chem. 1992, 13, 952–962. [CrossRef] 56. Roe, D.R.; Cheatham, T.E., 3rd. PTRAJ and CPPTRAJ: Software for Processing and Analysis of Molecular Dynamics Trajectory Data. J. Chem. Theory Comput. 2013, 9, 3084–3095. [CrossRef]