Rimonabant:CannabinoidPharmacology and the Rise and Fall of the Obesity Wonder-Drug AbstractRimonabant(SR141716) was a novelantagonist/inverse agonist at the CB1 receptor developed to aide weight loss inobese patients. Despite showing efficacy for its indication in phase III and IVstudies, rimonabant caused significant increases in adverse psychiatric effectsin patients, including an increase in suicide ideation, and was subsequentlywithdrawn from the market. Rimonabantwas born out of exploratory research into the cannabinoid signalling system,and was the first pharmacological modulator of this system to be approved inEurope for human use. The FDA never approved rimonabant for the US market.
Besides the intrinsic safety issues of the drug itself, the story of rimonabantis also associated with faults in trial design and an inferred lack ofappreciation for arguably predictable adverse central effects. Whilst thefailure of rimonabant was a significant blow to the development of CB1antagonists, recent research has offered the potential of a second life forrimonabant and its analogues. IntroductionRimonabant was a first-in-class anorectic drug developedby Sanofi-Aventis (now Sanofi) and indicated to treat obesity.
Rimonabant was the first specific cannabinoidreceptor 1 (CB1) inverse agonist to be licenced for human use, and, under thetrade name Acomplia, received approval for sale in Europe by the EuropeanMedicines Agency (EMA) in June 2006 to be used as an adjunct to diet andexercise for the treatment of obese patients (BMI ? 30kg/m2), or overweightpatients (BMI > 27 kg/m2) with other associated risk factors suchas type 2 diabetes or dyslipidaemia (EuropeanMedicines Agency, 2009) (Fong andHeymsfield, 2009). The application for rimonabant to be sold in the UnitedStates under the trade name Zimulti was not approved by the US Food and DrugAdministration (FDA) (FDA, 2007).Obesity is a significant worldwide health issue. A reportpublished in 2016 by researchers from the UK Parliament House of CommonsLibrary found that in England 27% of people are obese (BMI ? 30kg/m2)and a further 36% are overweight (BMI = 25.0 – 29.9kg/m2) (figure 1).
(House of Commons Library, 2017). A writer from the World Health Organisation (WHO) coined the term”Globesity” in 2002 to reinforce the scale of the problem, and the WHOestimates that in 2016 1.9 billion adults were overweight, with more than 650 millionof these qualifying as obese. The global prevalence of obesity has almosttripled since 1975. Obesity is a major cause of preventable death, and a raisedBMI is major risk factor for other non-communicable major causes of prematuredeaths worldwide, including stroke, heart disease, diabetes mellitus and sometypes of cancer (World Health Organisation, 2017). In the US alone theestimated annual cost of obesity-related illness has been estimated to be over$200 billion. (Hammond and Levine, 2010)(Cawley and Meyerhoefer, 2012).
Therefore the need and market for effectiveanti-obesity drugs is huge.Despite displaying efficacy for thereduction of weight in pre-clinical and clinical studies (Muccioli and Lambert, 2005), rimonabant (Acomplia) was suspendedfrom the European market in 2008 before being withdrawn in 2009 after post-marketingsurveillance by the European Union’s Committee for Medicinal Products for HumanUse (CHMP) discovered that the use of rimonabant doubled the risk ofpsychiatric disorders in overweight and obese patients who were taking the drug(European Medicines Agency, 2009)(Moreira and Crippa, 2009). TheEndocannabinoid SystemThe elucidation of the endocannabinoid system(ECS) is a result of research stemming from the initial study of therecreational drug now known as cannabis. The recreational use of preparationsfrom the plant Cannabis sativa fortheir psychotropic effects has been documented for thousands of years, but theprimary psychoactive chemical of cannabis was not isolated and synthesised until1964, when a research group led by the Israeli organic chemist RaphaelMechoulam identified it as ??-tetrahydrocannabinol (??-THC) (Mechoulam and Gaoni, 1967) (Mechoulam,Braun and Gaoni, 1967) (Howlett et al.,2004).The two identified cannabinoid receptor subtypes, CB1 andCB2, are components of the ECS, along with their endogenous ligands and thesynthesis and degradation pathways of these ligands.
The CB1 receptor is widelyexpressed in the brain and peripheral tissues, whilst the CB2 is mostlyexpressed in the peripheral tissues of the immune system (Simon and Cota, 2017). There has been debate as to whether or notCB2 receptors were also present in the brain, and for over a decade the generalconsensus was that they were not, and were exclusively expressed peripherally.More recent research has determined that this is not the case, and CB2receptors are indeed expressed in the brain (Onaivi,2006.) It is, however, the CB1 receptor that is involved in the centralregulation of feeding behaviour through which rimonabant mediates its effect. CB1 Receptor Extracellular The CB1 receptor (figure2) is a Class A (Rhodopsin-like) G-protein coupled receptor (GPCR), and consistsof 7 transmembrane ?-helices.
These transmembrane domains are linked by 3extracellular and 3 intracellular loops. The N terminal is located outside ofthe cell and the C terminal is inside. The CB1 receptor is among the mostabundantly expressed GPCRs in the brain (Huaet al., 2016) (Shao et al., 2016). The receptor is coupled to the G?i/oguanine nucleotide-binding protein (G protein) (Howlett et al., 2004).
GProtein Signalling Intracellular The coupling of CB1 to G?i/owas proven when a team led by Allyn Howlett reported that the binding of ??-THCcauses a specific inhibition of cyclic adenosine monophosphate (cAMP) inneuronally-derived cells by inhibiting of basal and hormone-stimulatedadenylate cyclase activity. This effect was found to be inhibited by pertussistoxin, which disables the G?i subunit. (Howlett, 1984). CB1 activation inhibits the adenylate cyclase isoforms 5and 6 to reduce the activity of the cAMP-stimulated protein kinase A (PKA). PKAnormally acts to phosphorylate and therefore inactivate A-type potassiumchannels, so the deactivation of this enzyme enhances the activity of thepotassium current to decrease the duration of presynaptic action potentials.The ultimate effect of this is to decrease the amount of neurotransmitterexocytosed into the synaptic cleft by speeding up the repolarisation of thecell after an action potential has fired (Deadwyleret al., 1995) (Howlett, Blume and Dalton, 2010).CB1 receptor activation also leads to the phosphorylationand subsequent activation of the mitogen activated protein kinases 1 and 3(MAPK1 and MAPK3), also called ERK2 and ERK1.
Derkinderen et al (2003) demonstrated that this happensin vivo using an immunocytochemicalassay involving phosphorylation-state specific antibodies in rat hippocampalslices to assess the degree of ERK phosphorylation. A strong phosphorylatedERK-like immunoreactivity was seen in pyramidal cell layers of the hippocampus10 minutes after the live rats were injected with ??-THC. ERK responses areinvolved in complex signalling pathways which ultimately function to regulatetranscription factors in cell nuclei to alter gene transcription (Derkinderen et al.,2003) (Howlett, 2005).
EndogenousLigands for the CB1 ReceptorThe discovery of the CB1 receptor and some of itssignalling pathways prompted the search for its naturally occurring ligands.The first 2 endogenous CB1 ligands to be identified were the eicosanoidsanandamide (N-arachidonoylethanolamide, AEA) and 2-arachidonyl glycerol (2-AG).These molecules are synthesised on demand rather than stored for release, andare degraded primarily by the enzymes fatty acid amide hydrolase (FAAH) andmonoacylglycerol lipase (MGL) respectively (figure3) (Pertwee, 2006) Petrocellis,Cascio and Marzo, 2004). Figure 3: the synthesis and breakdown pathways of AEA and 2-AG, the first identified and most-studied endocannabinoid molecules Roleof the ECS in Food Intake TheECS has been well characterised as a modulator of feeding behaviour in humansand other animals.
Early controlled human studies found that smoking cannabissignificantly increased calorific intake and weight gain (Greenberg et al.,1976) (Foltin, Fischman and Byrne, 1988) due to an increase in appetite knowncolloquially as the ”munchies” (Devlin, 2015). The??-THC-mediated increase in appetite and feeding behaviour is also seen in rats(Koch, 2001).TheCB1 receptor is widely expressed in the hypothalamus of the brain, an areaassociated with the control of homeostasis within the body.
Animal studies havedemonstrated that the injection of ??-THC increases the feeding behaviourinduced by electrical stimulation of the hypothalamus (Trojniar and Wise, 1991), suggesting that thesereceptors in this specific brain area are involved in the control of foodintake. Injection of the endocannabinoid AEA into the ventromedial hypothalamusalso stimulates the intake of food (Jamshidiand Taylor, 2001), providing more evidence of the involvement of thehypothalamus in the ECS-mediated feeding behaviour. In genetic mouse models ofobesity (ob/ob and db/db mutants) there are elevated levels of endocannabinoidsin the hypothalamus compared to normal-weight controls whilst there are were nodifferences between endocannabinoid levels in the cerebellum (Di Marzo et al., 2001). In rats, levels of 2-AG in thehypothalamus and limbic forebrain decrease when upon feeding, and AEA and 2-AGlevels in the limbic forebrain decrease when fasting (Kirkham et al.
, 2002). Intheir review, Gamage and Lichtman interpret these animal data to suggest thatthe general mechanism through which cannabinoids induce feeding behaviour is anegative feedback loop whereby CB1 stimulation mediates a disinhibition oforexigenic (appetite-stimulating) neurones in the lateral hypothalamus whilstinhibiting anorectic (satiety-inducing) neurones in the paraventricular nucleus(Gamage and Lichtman, 2011). Rimonabant,initially codenamed SR141716, was first used in animal experiments to inhibit CB1signalling during investigations of the actions of cannabinoids on CB1receptors. At the same time, the ECS became associated with leptin, anappetite-suppressing hormone whose gene is knocked out in mice (ob/ob mutants, figure 4) to model excessive foodintake and resultant obesity. Research groups began to shift their focus onto rimonabantas a potential therapeutic agent in obesity after it was found to reduceobesity in this model (Ravinet Trillou etal., 2002) as well as in the alternative diet-induced obesity (DIO) model (Liu et al.
, 2004).The underlying pharmacology therefore suggested that SR141716could be a useful therapeutic for the reduction of weight in obesity, andSR141716 entered the development pipeline as a novel drug for this indication. Rimonabant: Clinical Trials and Market ApprovalRimonabantreceived market authorisation from the EMA in Europe in 2006 on the back of itspivotal RIO (Rimonabant in Obesity) phase III trials; these were RIO NorthAmerica, RIO Europe, RIO Diabetes, and RIO Lipid (Sam, Sale and Ghatei, 2011).TheRIO North-America study found that rimonabant, in addition to lifestylemodification, produced sustained reductions in weight and other favourableimprovements in cardiovascular and metabolic endpoints over a 2 year period.However there was an unusually high drop-out rate, with 47% of patientswithdrawing from treatment (Pi-Sunyer etal, 2006). RIO Europe reported similar successes in weight loss whilst alsorecording that rimonabant was ”generally well tolerated with mild andtransient side-effects” (Van Gall etal., 2005).
TheRIO Lipids trial found that rimonabant at 20mg was associated with asignificant decrease in weight, waist circumference and triglyceride levels,and an increase in the proportion of HDL cholesterol, in overweight or obesepatients with dyslipidaemia. This study reported that depression, anxiety andnausea were the primary reasons for patients discontinuing the trial (Després et al., 2005). The RIO Diabetestrial tested rimonabant in overweight or obese patients with type 2 diabetes,and found a similar result to the lipid trial; a clinically ‘successful’efficacy result where the drug improved weight loss and cardiovascular andmetabolic risk factors, but there was a presence of central nervous systemadverse events including depressive mood disorders (Scheen et al., 2006).Thusit seems that there was already evidence of the ability of rimonabant to inducea potentially-dangerous increase in mood disorders. Two meta-analyses put forwardworries about the safety of rimonabant soon after it was approved in Europe; aCochrane review by Curioni and André (2006) found that 20mg per day rimonabantcaused significant nervous-system and psychiatric adverse effects, and calledinto question the unusually high drop-out rate of approximately 40% after 1year of dosing (Curioni and André, 2006)- the average drop-out rate in phase III studies is normally closer to 30% (Alexander, 2013). Christensen et al.
(2007) also found that 20mg per day rimonabant increased the risk ofpsychiatric adverse events including depression and anxiety disorders, and evenrecommended physicians have ”increased alertness” to the potentialpsychological effects of Rimonabant (Christensenet al., 2007).Theconclusions of these meta-analyses alongside the rejection of rimonabant by theFDA on the grounds of worrying and poorly-characterised safety data wouldforeshadow the demise of rimonabant in the years following its approval inEurope. RimonabannedSoon after rimonabant was made available on the Europeanmarket, the manufacturer and regulator were inundated with complaints of theincreased incidence of depressive and anxiety symptoms in patients, as well asreports of an increase in suicide ideation. Rimonabant was eventually suspendedby the EMA in November 2008 once it had become clear that the psychiatric risksof the drug outweighed the therapeutic benefit, and in January 2009 it was formallywithdrawn from the European market (Simon and Cota, 2017).
The ban on Rimonabant also ended the ongoing CRESCENDO(Comprehensive Rimonabant Evaluation Study of Cardiovascular Endpoints andOutcomes) trial. Despite early termination of the studies, four participantstaking rimonabant had committed suicide compared to only one in the placebocontrol group, and an analysis of the data revealed that rimonabant producedsignificantly more neuropsychiatric and serious psychiatric adverse effectsthan placebo (Topol et al., 2010). Thewithdrawal of Rimonabant by the EMA was therefore proven to be appropriate andjustified, and the caution of the FDA vindicated. Could the problems of Rimonabant havebeen predicted? Depression andanxiety-related adverse effects can be predicted of CB1 antagonism/inverseagonism simply by considering the common effects of recreational cannabis use.
Therationale of rimonabant was to reduce appetite in an opposite manner to theincreased appetite observed with recreational cannabis use (CB1 agonism). Giventhat cannabis (e.g. when smoked) can also induce states of relaxation,well-being, reduced anxiety, happiness and even increased enjoyment of visualand auditory stimuli (Moreira and Crippa,2009) (nhs.uk, 2017), it is logical to presume that rimonabant may causethe converse psychiatric effects of stress, depression and anxiety.
Indeed, this wasconfirmed in preclinical animal studies. Rimonabant was found to cause”anxiety-like responses” in rats (Navarroet al., 1997) as well exacerbating stress responses (Moreira, Grieb and Lutz, 2009). Drawing on such animal data,Moreira and Lutz (2008) even suggested that a disruption in endocannabinoidfunctioning may result in depression and anxiety (Moreira and Lutz, 2008). The potential for rimonabant to causesignificant psychiatric adverse effects should have therefore been anticipated,and this risk should have been integrated into all stages of rimonabant’sdevelopment. RIO:mistakes with patient selection and data stratificationThe RIO trials excluded individuals with depression orother neuropsychiatric conditions (Sam,Sale and Ghatei, 2011). The reason for this is likely to bethat Sanofi-Aventis were aware of the propensity of rimonabant to inducepsychiatric adverse effects, and therefore did not want to worsen the conditionof those with pre-existing depression.
This would also allow them to see if,and to what extent, rimonabant would induce its psychiatric adverse effects inotherwise mentally healthy participants. However,this meant that the safety data from the RIO studies was not generalisable tothe true patient population. Depressivedisorders are common and are co-morbid with obesity (Carey et al.
, 2014), so for phase III safety data to beappropriate they must come from a participant population reflective of thepatient population. The participant design of the RIO studies also meant thatthere were no data examining the effect of rimonabant on existing depressivedisorders; it would have been invaluable and potentially lifesaving if thedrug-induced increase in suicidal ideation had been appropriately characterisedbefore the market authorisation and CRESCENDO trial.Thereforeif the trials had included participants with depressive and anxiety disorders,the unacceptable frequency of psychiatric adverse effects, including theincreased risk of suicide ideation and completion, would have been demonstratedbefore the drug was approved and distributed to a much wider population ofat-risk patients.Anotherissue with the RIO trial data highlighted by the FDA is that the commonlyreported adverse effect of ‘irritability’ was not classed as psychiatric. Thismay have led to an underestimation in the ability of rimonabant to causepsychiatric symptoms because irritability, a symptom indicative of depressiveand anxiety disorders, was excluded from the pool of psychiatric complaints (FDA, 2007). Aftermath and the Future of CB1 AntagonistsThe failureof rimonabant severely hindered the study of CB1 antagonists and many researchprojects and in-progress trials were terminated. However, revisitation ofpreclinical data showed that the therapeutic benefits of CB1 antagonistsinvolve a peripherally-mediated component, whilst the psychiatric side-effectsthat brought about the downfall of rimonabant are centrally-mediated.
Thisprompted novel research into peripherally selective CB1 inhibitors to treatobesity (Klumpers et al., 2013). Tothis end, TM38837 is a newer CB1 antagonist which, unlike rimonabant, actsselectively in the periphery. In a phase I trial in healthy volunteers, TM38837at a dose of 100mg had no effect on the central nervous system and did notcause any psychiatric side effects, suggesting that this dose does not crossthe blood-brain-barrier to antagonise central CB1 receptors. The authors ofthis trial report that animal models predict this molecule to have an equalpotency to rimonabant in terms of treating metabolic disorders, suggesting thatthis molecule has future potential for treating obesity without the psychiatricadverse effects associated with rimonabant (Klumpers et al., 2013). However, there has been little news and nopublished data from the parent company of TM38838, (‘7TM Pharma’) since 2013,which suggests that the molecule has not performed as well as expected infurther trials.
In apaper published only months ago, rimonabant was found to produce positivechanges in the gut microbiota of obese mice modelled with the Diet-InducedObesity (DIO) model, as well as inducing an anti-inflammatory state in theadipose tissue by attenuating the trafficking of M1 macrophages into it. Thereforethis study revealed that the therapeutic mechanism of rimonabant in treatingobesity may involve anti-inflammatory and microbiological components (Mehrpouya-Bahrami et al., 2017). Withthis information, researchers may be able to identify novel targets or pathwaysupon which to design new drugs to treat obesity.Finally, rimonabant may yetexperience success for a different indication entirely. Low (µM ) dose rimonabant has been found todirectly affect the function of the kappa opioid receptor (KOR) in-vitro, andto produce anxiolytic effects in mice performing the elevated plus maze.
(Zádor et al., 2015). Therefore, counterintuitively,low-dose rimonabant may have potential as a novel anxiolytic by acting througha completely different mechanism to that behind its therapeutic efficacy inreducing weight in obesity. Conclusions and Lessons Learned Rimonabant revealed the dangersof the use drugs which act on metabolic and neurological systems which are notfully understood.
Whilst the clinical failure of rimonabant may have beeninevitable from the start, the suffering of patients due to its intrinsicpsychiatric adverse effects was not, and could have been avoided had themanufacturer treated the CNS liability of the drug with greater caution andrespect, and designed their trials accordingly.The story of rimonabant alsohighlights the importance of regulatory vigilance, and serves as a lesson forEuropean regulators in that they should be more stringent in their reviewprocesses, especially in the case of novel drugs such as rimonabant, wherethere are not comparable phase IV safety data from other approved drugs thatwork by similar mechanisms. Perhaps an increase in harmonization between theEMA and the FDA could have prevented the entry of rimonabant into the European marketand the harm to patients that followed.