Receptor and subtype selectivity is an important criterion for minimizing side effects, therefore tremendous efforts go into the development of compounds with designed binding profile.

Keserű et al. have developed a fragment based docking protocol to design specific receptor ligands. Based on the docking results, they have synthesized several compounds and demonstrated the usefulness of the method for the designing D₂/D₃, 5-HT_1B/5-HT_2B and H₁/M₁ receptor ligands with improved selectivity (Figure 5). In the first two cases, the selectivity of the PP was reversed using the SP moiety, while in the third case, a selective compound was designed and synthesized for a receptor pair with very similar PP (Egyed et al., 2021).

The importance of bitopic compounds in the inhibition of dopamine receptors is demonstrated by second and third generation antipsychotics, including aripiprazole (Burris et al., 2002) and cariprazine (Ágai-Csongor et al., 2012). 2,3-dichlorophenyl-piperazine, that serves as PP in these compounds was changed to 2-methoxyphenylpiperazine (1) PP. Although this PP exhibits weak D₂R selectivity, combined with a suitable SP group (2,3) its profile has been changed to mild D₃R selectivity. The efficacy of this methodology was further tested on serotonin receptors. The LSD-like PP of ergotamine (Figure 3B) did not show subtype selectivity at the two selected serotonin receptors, but the designed compounds (4,5) with the modified SP already had significantly higher affinity at 5-HT_2BR over 5-HT_1BR. Although ergotamine was more potent, compounds 4 and 5 had much greater selectivity over it. Among the first-generation antihistamines, muscarinic acetylcholine M₁ activity was a major problem due to side effects. Therefore, huge efforts were dedicated to the development of compounds with significant H₁R receptor selectivity. Starting from amitriptyline having only 7-fold selectivity, bitopic compounds (6,7) were designed that demonstrated 50–80 fold selectivity over M₁R (Egyed et al., 2021). The proposed protocol detailed in the design section of this review may be applied to other targets to achieve designed selectivity with bitopic compounds.

Tan et al. have exploited the basic 2-phenylcyclopropylmethylamine (PCPMA) scaffold (8, 9), whose analogues are known 5-HT_2CR agonists (Cheng et al., 2015; Cheng et al., 2016a; Cheng et al., 2016b; Zhang et al., 2017), to design new bitopic compounds (Tan et al., 2020) (Supplementary Table S3). Here we discuss only a subset of these compounds. As secondary pharmacophore, 1,2,4-triazolylthiol ethers were used and a propyl chain was employed as a linker. The introduction of SP alone improved D₃R activity 3-fold. A major leap forward was the realization that the alkyl side chain introduced on the amino group of PCPMA significantly improves subtype selectivity and D₃R affinity. Next, they investigated the substituents of the aromatic ring of PCPMA. First, the ortho positioned 2-fluoroethoxy group was changed, whereby methoxy was found to be the optimal one, thus significantly improving the D₃R affinity. The replacement of the fluorine atom by chlorine resulted in a moderate selectivity towards D₂R, D₄R, 5-HT_2CR and a strong selectivity towards D₁R and D₅R (10–12). As these results could only approximate the values of the reference compound 13 (BP-897) (Supplementary Table S3) the strategy was changed and a buthylene linker was used instead of the propylene group, the SP was replaced by other aromatic rings (naphthyl, indolyl, and 4-pyridylphenyl) and an amide bond between the linker and the SP was introduced instead of thioether (14–20). For these compounds, only N-alkyl substituted variants have been prepared and the effects of several PPs have been investigated. When examining the racemic compounds, the compound containing 4-pyridylphenyl SP and dichlorophenyl PP (20) has more than 1000-fold selectivity towards the other DRs, with milder but still significant selectivity in the range of 17, 18, and 19 (Tan et al., 2020) (Table 1).

Battiti and co-workers performed a SAR analysis combining two PPs for the synthesis of bitopic compounds; one is a selective dopamine agonist PF-592379 (Allerton et al., 2005; Ackley, 2008) and the other is PD-128907, which is a D₂R/D₃R agonist. ( Supplementary Table S4) They concluded that the structural features of PD-128907 avoided the construction of bitopic compound. Therefore, they focused to PF-592379 to synthesize D₂R/D₃R active bitopic compounds. Here we discuss a representative example for different SPs (Supplementary Table S4). D₂R and D₃R binding data clearly show that the (S,S) enantiomer of the PP is more favourable for receptor binding. The (S,S) enantiomer already plays a prominent role in PP (22, 23), with a 3-fold activity difference between the enantiomers. The same effect can be observed when using tetrahydroisoquinoline (24,25) or indole (26,27) SP, although here the difference in activity at the D₃ receptor is about 100-fold (Table 1, Supplementary Table S4). Compound 27 show a 22.5-fold subtype selectivity towards D₃R that is due to the SP moiety. Next the authors investigated the effect of the linker. Changing the original cyclopropylethyl linker (Supplementary Table S4) for the racemic derivative (rac-trans-28) resulted in 37.3-fold selectivity towards D₃R. Separating the enantiomers, (1S, 2R)-trans-cylopropyl stereochemistry (28a) showed D₃R K_i of 1.85 nM and an unprecedented 47.5-fold selectivity for D₃R over D₂R (D₂R Ki = 87.8 nM), while the other enantiomer (28b) has much weaker activity coupled with poor selectivity (Battiti et al., 2019). Finally, two additional linkers were used (31, 32) that are widely used among D₃R bitopic compounds including several high selectivity partial agonists or antagonists (Kumar et al., 2016; Michino et al., 2017; Verma et al., 2018). Compound 31 showed reduced affinity compared to 28a, inferring that the hydroxyl group on the linker is optimal for antagonism but cannot be directly transferred to the agonist binding mode due to different receptor conformations in the active and inactive states. Compound 32 shows good affinity but neither affinity nor selectivity reaches that of 28a. Compounds 31, 32, 28a were tested at MOR. For 32 there is a decrease in affinity at the dopamine receptor but the weak subtype selectivity is retained, however there is a 22.9-fold increase in activity at the MOR receptor (Supplementary Table S4) (Battiti et al., 2020). The same group synthesized a number of eticlopride analogues using different SPs in the 2-N or 4-C position of pyrrolidine via lycerol (Battiti et al., 2020). They found that O-alkylated analogues had better affinity for D₂ and D₃ receptors than the N-substituted derivatives. In BRET assays, these compounds exhibited antagonist or very weak partial agonist behaviour. Docking studies revealed that the SPs of the O-alkylated analogues form aromatic stacking interactions with conserved residues His6.55 and Tyr7.35 both in the D₂ and D₃ receptors, while the SPs of the N-alkylated derivatives extend towards the extracellular site that is less conserved (Shaik et al., 2021).

N-phenylpiperazine analogues were used extensively for constructing bitopic ligands against dopamine receptors. Lee et al. synthesized and evaluated a series of N-phenylpiperazine analogues substituted with 3-thiophen and 4-thiazolylphenylfluoride (Supplementary Table S5). They identified several ligands that bind with high affinity to D₃R and exhibit considerable selectivity towards D₂R. Comparison of the binding results of compounds 33–38 and 39–44 suggests that 39–44 binds to D₃R but not to D₂R. The replacement of the thiophene ring by a thiazole ring (45–50) led to a decrease in receptor binding selectivity. Compound 39 (Table 1) possessed the highest D₃R affinity (K_i = 1.4 nM) and 450-fold selectivity that nominated this compound for in vivo testing. Intraperitoneal administration of 39 led to a significant reduction in DOI-dependent head twitch response in mice and a reduction in AIM scores in dyskinetic hemiparkinsonian rats. These data suggest that compound 39 is able to cross the blood-brain barrier and achieves therapeutic concentrations (Lee et al., 2021).

Starting from the 5-HT_2A receptor-bound structure of aripiprazole and cariprazine Chen et al. designed D₂/D₃ receptor ligands with no significant 5-HT_2A affinity (Chen et al., 2021). The authors suggested that the unusal “upside-down” binding mode (Figure 3A) might affect the observed selectivity. According to the structural rearrangements, the location of the SP of aripiprazole in the exosite is important for its signal transduction efficiency. In the interest of identifying residues critical for efficacy, the exosite sequence of the 5-HT_2A and D₂ receptors was aligned, with an important difference between the two found at position 5.51, which is Leu in 5-HT_2AR and Phe in D₂R. Mutations in D₂R demonstrated that substitution of F202^5.51 with Leu or Ala reduces the G-protein activity and β-arrestin2 recruitment of aripiprazole. In addition, a derivative of aripiprazole substituted with benzothiazole for the dihydroquinoline ring of D₂R had reduced efficiencies of both G protein activity and β-arrestin2 recruitment. Substitution of L247^5.51F in 5-HT_2AR did not increase the efficacy of aripiprazole. The results suggest that aripiprazole may stabilize different conformations of TM5 and TM6 between the two receptors. Alignment of 5-HT_2AR and D₂R structures (active and inactive) shows that activation of 5-HT_2AR requires a larger downstream swing of W6.48 from the CWxP motif than that observed for D₂R activation. In the 5-HT_2AR, dihydroquinoline is located deeper in the binding pocket interacting with W336^6.48, restricting its movement, whereas it can move gently upon D₂R receptor activation. Similar observations were made for cariprazine. Here, the dynamic coupling between F/L5.51, W6.48 and the PIF motif by the exosite may partly explain why the compounds tested have different efficacies at 5-HT_2AR and D₂R receptors. Compared with inactive and active D₂R constructs, the 5-HT_2AR-aripiprazole complex in the extracellular compartment shows inward movement of TM6, TM7, and ECL2 toward the seven transmembrane cores. These rearrangements suggest that the 5HT_2AR affinity of the bitopic compounds can be reduced by increasing the size of the PP. Changing the arylpiperazine PP to aza-ergoline the authors identified IHCH7009 (D₂R K_i = 33.65 nM, 5-HT_2AR K_i = 3639.15 nM), IHCH7010 (D₂R K_i = 9.03 nM, 5-HT_2AR K_i = 906.78 nM) and IHCH7041 (D₂R K_i = 50.64 nM, 5-HT_2AR K_i = 2371.37 nM) all with very weak 5-HT_2AR affinity. IHCH7041 retains partial agonism of D₂R while IHCH7009 and IHCH7010 are full D₂R agonists (Chen et al., 2021).

Kling et al. investigated the neurotensin receptor type (NTS) 1 receptor crystal structures (White et al., 2012; Egloff et al., 2014) and found that an allosteric binding site was saturated at the C-terminus of NT (8–13). Following sequence analysis, they confirmed that there is a difference between NTS₁R (Arg149^3.32) and NTS₂R (His115^3.32) that may allow for subtype selectivity. Several bitopic ligands of type NT (8–13) were synthesized (Table 2) and compounds (51–56) showed a promising trend in the NTS₁R selectivity. The best compound (54) has K_i value of 1.3 nM associated with 26-fold selectivity towards NTS₂R. Homology modelling and MD simulations confirmed that the compounds bind in a bitopic mode, with NT (8–13) occupying the orthosteric binding site and the amino acid extension occupying the secondary binding site. These results provide a promising starting point for the design of NTS₁R selective agonists (Kling et al., 2019).

Advances in GPCR structural biology and pharmacology have opened up new opportunities for functional drug design. Modulation of GPCRs through allosteric binding sites can alter receptor structure, dynamics and function, resulting in increased spatial and temporal variation. One important aspect of these changes is functional selectivity or otherwise termed biased signalling. Biased signalling can contribute to the enhancement of the intended effect, but can also cause side effects, so one of the most intriguing areas of current research is investigating the functional character of the ligands in different signalling pathways (Hauser et al., 2017).

Egyed at al. reported a systematic study exploring the extracellular SBP to fine-tune the functional profile of D₂R and D₃R ligands. Introduction of the SP increased affinity at both D₂ and D₃ receptors for each ligand. The study demonstrated that the G_i/o and β-arrestin pathways can be specifically modulated from the extracellular vestibule incorporating different SPs to the ligands. Molecular dynamics simulations revealed that G-protein signalling could be linked to the orientation of the PP that is influenced by the SBP binding part of the bitopic compounds (Figure 6). Three PPs and two SPs (Figure 6) were tested using an ethylcyclohexyl linker in analogy to cariprazine. In the G_i/o-mediated signalling pathway, dichlorophenylpiperazine (57) (PP 1) was a partial agonist on both D₂R and D₃R (Table 3). Application of N,N-dimethylurea (SP 1) (cariprazine) also resulted in a partial agonist with significantly increased potency (D₂R pEC₅₀ = 8.85 nM, E_max = 77.4%, D₃R pEC₅₀ = 8.58 nM E_max = 27%). The use of the OtBu motif (SP 2) (61) led to a full agonist, the potency on D₂R was superior to that on D₃R. For 2-methoxyphenylpiperazine (2, 58, 62) (PP 2), no prominent change was observed, all were partial agonists. The 3-(piperazin-1-yl)-5-(trifluoromethyl)benzonitrile (59) (PP 3) with the N,N-dimethylurea SP (60), showed antagonist effects on the G protein coupled signalling pathway of D₂R and D₃R, with an increase in potency. Interestingly, incorporating SP 2 (63) turned the function of PP to a weak partial agonist at both receptors. These results suggest that PP and SP affect functionality together. In the β-arrestin signalling pathway, compounds with SP 2 achieve the largest increase in E_max values, while this was lower for cariprazine. (Table 3). This suggests that cariprazine shows a significant bias towards the G-protein controlled pathway on D₂R. In all cases, the bitopic compounds with 2-methoxyphenylpiperazine PP (2, 58, 62) exhibited antagonist behaviour in contrast to the partial agonism observed in the G-protein coupled signalling pathway. The antagonistic behaviour of 59 was also preserved in the β-arrestin signalling pathway; following the previous trends introduction of any SP led to an increase in pIC₅₀ values here as well. In general, the efficacy data measured at both receptors followed similar trends in both modalities as the receptor affinities (Egyed et al., 2020).

High affinity binders, such as 39, 40, 41, 42, 49 (Supplementary Table S5) were also tested for their efficacy on D₃R, both by examining forskolin-dependent inhibition of adenylyl cyclase and by measuring β-arrestin binding. Compounds 42 and 49 were found antagonists in both assays. Compound 41 display functional selectivity, being a weak partial agonist in the adenylyl cyclase assay and a very weak partial agonist/antagonist in the β-arrestin binding assay. Compounds 39 and 40 exhibit weak partial agonism in both the adenylyl cyclase inhibition and β-arrestin binding assays (Lee et al., 2021).

Investigating pure enantiomeric forms of compounds 17–20 (Supplementary Table S3) Tan et al. showed that the (R,R) enantiomers (17b-20b) have a better affinity for D₃R than (S,S) (17a-20a), with the exception of compound 17, which had an identical affinity for both of the enantiomers (17a, 17b) (Tan et al., 2020). The (R,R) isomers (17b-20b) showed weaker affinity (3–20-fold) towards 5-HT_2CR than their (S,S) counterparts (17a-20a). The data suggest that D3R is less sensitive to conformational changes than the 5-HT_2C receptor. Functional studies were also performed with the 17a,b-20a,b (Table 4). Compounds 18–20 were all full or partial agonist on D₃ receptors, whereas for 5-HT_2CR the (S,S) enantiomers (18a-20a) are weak partial agonists, whereas the (R,R) enantiomers (18b-20b) are weak antagonists. Compared to the binding assay, functional results indicate greater selectivity towards D₃R. Furthermore, these compounds showed only very weak partial agonism at 5-HT_2AR and no affinity at 5-HT_2BR. The two enantiomers of compound 17 exhibit opposite behaviour, while (1R,2R)-17b was a potent agonist (EC₅₀ = 3.6 nM, E_max = 77.9%), (1S,2S)-17a was an antagonist on D₃R with a K_i of 16.7 nM, and both derivatives were weak antagonists with micromolar activity on 5-HT_2C receptor. Docking studies suggested a difference between the two compounds (17a,17b) in the orientation of PP. In the case of the agonist (1R,2R)-17b, the 2-methoxy group is deep in the OBP and forms hydrophobic interactions with residues C114^3.36, S196^5.46, and F346^6.52. In the case of the antagonist (1S,2S)-17a, the 2-methoxy group flips out to the extracellular side and the cyclopropane linker between the benzene ring and the protonated N overlays perfectly with the amide linker of eticlopride, which is not present in the agonist. Compounds (1S,2S)-17a, (1R,2R)-18b, (1R,2R)- 19b, and (1R,2R)-20b were inactive in the Tango assay on D₃R, indicating their preference for the G-protein signalling pathway. For further profiling (1R,2R)-17b and (1R,2R)-19b were tested on 29 other aminergic GPCRs that confirmed their good selectivity for D₃R (Tan et al., 2020).

Yan et al. also used PCPMA analogues as PP, with propyl, butyl, pentyl, or cyclohexylethyl linkers, and SP groups taken from aripiprazole, brexipirazole, and cariprazine, respectively. The synthesized library was measured in D₂R binding, D₂R G_i and D₂R β-arrestin BRET assays (Table 5). The starting compound (64) exhibits good affinity (K_i = 61.9 nM) and partial agonist activity in both G_i (EC₅₀ = 49.0 nM, E_max = 25%) and β-arrestin (EC₅₀ = 67.6 nM, E_max = 30%) BRET assays. In comparison, replacement of SP with quinolone (65) increased the potency two-fold with unchanged binding. Changing the linker to propyl (66,67) led to a small decrease in binding affinity but an increase in efficacy (∼10 nM EC₅₀ values and E_max values higher than 50%). Lengthening the linker to 5C units (68,69) led to a decrease in binding affinity and functional activity. The cariprazine-like SP (dimethylamine) and linker (cyclohexyl) with this PP did not show significant activity. The best compound from this series (70) has very potent partial agonist character in both G_i BRET (EC₅₀ = 8.45 nM, E_max = 68%) and β-arrestin2 recruitment assays (EC₅₀ = 9.49 nM, E_max = 16%), with a much lower E_max in the latter. The significant difference between binding affinity and potency for many of these compounds likely reflects the use of an antagonist radioligand [(³H)-N-methylspiperone] in the competitive binding assay, from which an agonist ligand tends to show much lower apparent binding affinity. Attempts have been made to use several PPs but these have been shown to give significantly worse results than the methoxy derivative. In the case of isoquinoline and tetrahydroisoquinoline SP, it was not practical to use the dichlorophenyl motif in the PP (71,72). The best results were obtained with derivatives containing halogen in the meta position on the phenyl group of PP and methoxy in the ortho position (73a,b-76a,b). Pure forms of the enantiomers were also investigated. The majority of the fluorinated derivatives ((1S,2S)-42a, (1R,2R)-73b, (1S,2S)-74a) showed K_i values below 50 nM on binding assay and EC₅₀ values below 20 nM in both G_i and β-arrestin2 BRET assays. The same trend was observed for the chlorinated derivatives [(1S,2S)-75a,(1S,2S)-76a]. Higher E_max was observed for the halogenated derivatives in the G_i signal transduction than in the β-arrestin. After separation of the enantiomers, it was confirmed that the (S,S)-isomers were more efficient in D₂R binding and functional assay. The (R,R) compounds exhibit partial agonist behaviour and the E_max values are higher for G_i signaling. The selectivity of the compounds [(1S,2S)-73a, (1S,2S)-74a, (1S,2S)-75a, (1S,2S)-76a] was investigated on D₁R, D₂R, D₄R, D₅R, 5-HT_1AR, 5-HT_2AR, and 5-HT_2CR, with low selectivity observed towards the D₃ receptor and potent activity on the 5-HT_1A receptor, and good or acceptable selectivity on the other receptors (Table 6). In the case of D₃R, these compounds showed weak partial agonist activity in both G_o and β-arrestin2 BRET assays, albeit with different efficacies. For the 5-HT_1A receptor, all four compounds ((1S,2S)-73a, (1S,2S)-74a, (1S,2S)-75a, (1S,2S)-76a) were similar partial agonists in G_i BRET assays. The lack of selectivity over D₃R and 5-HT_1AR should not be a concern for these compounds, as both D₃R and 5-HT_1AR have been shown to be involved in the therapeutic effects of some antipsychotics. Overall, these four compounds have shown an interesting pharmacological profile (Yan et al., 2021).

Schramm et al. investigated the effect of bitopic compounds on muscarinic acetylcholine receptors. Carbachol (CCh) PP was cross-linked to allosteric ligands by linkers of different lengths (1C, 3C, 5C, 8C). The benzoimidazole-piperidine moiety of TBPB [1-(1′-(2-tolyl)-1,4′-bipiperidin-4-yl)-1H-benzo(d)imidazol-2(3H)-one], a known selective bitopic M₁R agonist, and BQCA (benzyl quinolone carboxylic acid) derivatives, that are PAMs, were used as allosteric modulators (Table 7). It was found that BQCA-CCh bitopic compounds act as agonists. The highest potency and efficacy was observed for the compound containing BQCA moiety 81. Comparing with reference compound 86, which does not contain a CCh moiety but only the linker, revealed that the CCh moiety provides some of the agonist activity. In contrast, the TBPB-CCh bitopic ligand (78) showed partial agonism, while the reference 84 was a full agonist. The binding mode of 81 was investigated by docking to an active receptor model. The ammonium group of the CCh moiety forms a charge-assisted hydrogen bond with D105^3.32, while the carbamate carbonyl group serves as a hydrogen bond acceptor for the hydroxyl group of Y408^7.43. This is different from the carbachol binding mode, in which the carbamate structure has a different orientation. The BQCA moiety, located in the region of the extracellular loop, is stabilized by hydrophobic contacts with L174^ECL2 and Y179^ECL2 and a charge-assisted H-bond with K392^ECL3. They concluded that partial agonism through bitopic compounds can be achieved not only by quenching orthosteric receptor activation by an allosteric moiety as in 81 but also by quenching bitopic activation of the receptor by an orthosteric moiety such as CCh in 78 (Schramm et al., 2019).

Holze et al. have shown that allosteric coupling of the M₁R can induce conformational changes that affect intracellular signalling. They investigated two groups of M₁R bitopic agonists and varied the length of the linker. Iperoxo, a known agonist, was selected as the PP motif, while two negative allosteric modulators, phtp (89–91) and naph (92–94), were incorporated as SP. (Figure 7) The latter differs from the phth derivative in two main respects: naph contains a larger and branched aliphatic linker. The two pharmacophores were linked by alkyl chains of different length (6–8C) (89–94). While the ligand affinities for the allosteric binding site were very similar within a ligand set, the ligand affinities for the orthosteric binding site depended on the length of the linker, where increasing linker length was correlated with increasing ligand affinity. From this information, it was concluded that the same binding mode was adopted by iperoxo in a series of bitopic compounds driven by its high affinity, and this was confirmed by MD simulations. Therefore, a series of bitopic ligands differing only in the length of the linker may be suitable to investigate the effect of allosteric coupling on signal transduction with subnanometer accuracy. Whereas the longest bitopic agonist, 91, was able to stimulate all three G-protein families, 90 activated G_q/11 and G_s proteins, 89 promoted signal transduction only via G_q/11. 93 and 94 only activated G_q/11 protein signalling, while 92 did not activate any signalling pathway, unlike 89. None of the naph-based ligands were able to activate G_s and G_i/o signalling. These data suggest that different G-proteins show different sensitivities to M₁R activation by these bitopic compounds. While G_q/11 coupling is conserved in almost all bitopic ligands, G_s signalling is promoted only by two members of the phth series. G_i/o activation is particularly sensitive to the bitopic ligand structure with only 91 showing weak M₁R/G_i/o coupling among the compounds tested. MD simulations show that binding of iperoxo results in a complete contraction of the extracellular parts of the ligand binding pocket. In contrast, the bitopic ligands of the phth series bind in such a way that they sterically inhibit the closure of the binding pocket. The extent of the conformational interference depends on the length of the linker and hence the position of the allosteric building block. Since the phth part of 89 is located close to the orthosteric binding site, it inhibits closure, resulting in a more open extracellular conformation. Elongation of the linker with additional methylene groups allowed for subnanometer regulation of the position of the allosteric building block, thereby progressively reducing the closure of the binding pocket, ultimately resulting in greater G-protein binding capacity. FRET measurements have demonstrated that the more closed ligand-binding pocket is associated with greater receptor conformational changes at the G-protein binding surface via an allosteric coupling mechanism. Consistent with this idea, 92, a bitopic ligand with a branched and larger allosteric motif, did not induce conformational changes in M₁R (Holze et al., 2020).

Wang et al. investigated two naltrexone derivatives substituted with isoquinoline at MOR. The isoquinoline moiety of these bitopic compounds is the SP that interacts with the allosteric site of MOR, and the epoxymorphinan moiety is the PP (Table 8). NAQ has a high affinity for MOR (K_i = 0.55 nM) and high selectivity for κ-opioid receptor (KOR) (48-fold) and δ-opioid receptor (DOR) (241-fold). Compared to DAMGO, it acts as a MOR antagonist in the ³⁵S-GTP [γS]-binding assay with CHO cell lines expressing MOR. It showed less significant withdrawal effects compared to the well-known opioid antagonists naloxone and naltrexone. Similar properties were observed for the compound NCQ (K_i = 0.55, 40-fold KOR, 62-fold DOR selectivity), which shares the same PP part as NAQ and differs only in the SP. NCQ contains a methoxy at position 1 and a chloro functional group at position 4 of isoquinoline. However, in ³⁵S-GTP (γS)-binding assay, NCQ behaved as a partial agonist. MD simulations and free energy calculations proposed that the allosteric part of NAQ and NCQ bind differently in the inactive structure and in the active structure, respectively. Docking studies have shown that the SP parts of NAQ and NCQ may occupy two different subdomains of the allosteric site of MOR, named ABD1 and ABD2. MD simulations were performed with three poses (NAQ inactive, NCQ active and inactive) obtained from the docking calculations and showed that the SP part of NAQ was bound to ABD1 in the inactive MOR. Although the SP motif occupied an allosteric site, no significant modulatory effect was observed on the binding of the PP, similar to the function of a silent allosteric modulator. In the inactive and active MOR the SP of NCQ showed positive allosteric modulation through binding to ABD2. Molecular modelling combined with interaction energy and distance analyses unravelled the molecular mechanisms of allosteric modulation of NAQ and NCQ and emphasized the importance of the chlorine and methoxy substituents of the isoquinoline ring for the allosteric modulatory function of NCQ (Wang et al., 2020).

Although ligand-receptor binding kinetics might have a fundamental role in the development of drug candidates, it is still often overlooked in the early phase of drug discovery. In line with the increased interest in the field, more and more kinetics data (among others association and dissociation rate, residence time, etc.) have been published in the literature, however the magnitude still lags behind the amount of affinity and selectivity data available especially regarding only the allosteric and bitopic ligands. Furthermore, the interpretation of the kinetic data might be hindered by the probe dependence as observed in a prototypical competitive radioligand binding assay for H₁ receptor antagonists, although that aspect is often not considered (Bosma et al., 2019). In line with the relatively limited amount of recent papers, first we refer the readers to recent general review articles on binding kinetics (Sykes et al., 2019; Hoare et al., 2020; Rafael et al., 2020; van der Velden et al., 2020). Very recently a book chapter collecting available kinetic data of GPCR ligands together with experimental evidence for properties that influence the residence time were published (Potterton et al., 2022). The repository enables researchers to analyse the relationship between the structure and the kinetic parameters as well as provides data for the development of predictive algorithms. The authors also outline machine learning workflows to predict residence time. Sykes et al. reviewed recently the literature related to the binding kinetics of GPCR ligands (Sykes et al., 2019). They discussed the theoretical aspects, the experimental methods and their limitations, detailed several factors influencing binding kinetics among others they explored the role of allosteric modulators, that by definition act through the modulation of the binding kinetics of the endogenous or orthosteric ligands. The authors also discuss some molecular level features including shielding the hydrogen bonds from water that affects the binding kinetics.

Although shielding the hydrogen bonds was thought to decrease residence time, in a recent case study on CCR2 receptor, MD simulations of Magarkar et al. suggested that even shielding an intra protein hydrogen bond can enhance the residence time of ligands through the preservation of the binding site rigidity (Magarkar et al., 2019). The ECL2 loop, that is regularly engaged with bitopic compounds, was also proposed to modulate the binding kinetics (Sykes et al., 2019; van der Velden et al., 2020). Already one of the seminal works in the field of modelling the binding pathway to GPCRs, which investigated the binding of three antagonists and an agonists to the β2-adrenoreceptor and one agonist to the β1-adrenoreceptor with MD simulations, highlighted the role of the ECL2 loop and the extracellular vestibule. Interestingly, even the highest barrier of binding often corresponds to the association with the extracellular vestibule even though the binding requires conformational change of the receptor and the ligand has to enter through a narrow passage (Dror et al., 2011). In several receptors, ECL2 were proposed to function as a lid facilitating the entrance and exit of the ligands (Thomas et al., 2016; Wacker et al., 2017; Frank et al., 2018). One of these studies investigated the binding kinetics of cariprazine and aripiprazole. As a prototypical bitopic compounds we exemplify here the effect of the SBP on the binding kinetics through them (Frank et al., 2018). At the D₃ receptor, aripiprazole exhibits a slow monophasic dissociation, while cariprazine displays a rapid biphasic behaviour. Interestingly, in the D₂ receptor both compounds display a slow dissociation. These differences may influence the in vivo action of the drugs. Interactions with ECL2 residues influence the residence time in other receptors like in the β₂ and A_2A receptors, as well (Guo et al., 2016; Masureel et al., 2018). Gaussian accelerated molecular dynamics revealed the role of the ECL2 loop in the formation of allosteric sites for PAMs in the adenosine A₁ receptor (Miao et al., 2018) and unveiled an intermediate binding site between ECL2 and TM1 for caffein in the adenosine A_2A receptor. The authors analysed the effect of more general features like physicochemical properties of the ligand (e.g., lipophilicity) and close contact residue numbers on the drug-receptor dissociation.

Van der Velden et al. summarized structural considerations in relation to binding kinetics presenting the results through four case studies (van der Velden et al., 2020). They showcased the role of the ECL2 loop in the regulation of the ligand kinetics through tiotropium binding to the M₃R and M₂R receptors (Kruse et al., 2012; Tautermann et al., 2013). The more open, flexible ECL2 loop conformation was linked to the shorter residence time observed in the M₂R receptor. Through the example of ZM241385, an A_2A receptor antagonists they highlighted the role of molecular dynamics and mutation experiments in providing structural background for observed kinetics behaviour (Guo et al., 2016). Another example was focused on the β₂ adrenoreceptor. Salmeterol, a bitopic compound displays a 5–7 fold higher residence time compared to salbutamol and epinephrin, both binding only to the orthosteric site (Figure 3C). As salmoterol and salbutamol share the orthosteric binding motif, the interactions in the extracellular site are linked to the increased residence time (Masureel et al., 2018). They also discussed other aspects, like the effect of natural receptor variants, ligand variants and probe dependency.

Riddy et al. investigated the binding kinetics of H₃ receptor antagonists/inverse agonists (Riddy et al., 2019). Although the binding mode of the compounds were not investigated experimentally, they likely form interactions outside the orthosteric pocket, too therefore can be considered bitopic. The different pharmacological profile and the residence time of the compounds might be linked to their preclinical and clinical efficacy. Furthermore, H₃ and off-target sigma-1 receptor occupancy may contribute to paradoxical efficacy of some compounds. In the study of Pedersen et al. (2020) the differential binding kinetics profile of the agonists were not linked to the functional bias, as the bias profile of the selected agonists were not time-dependent and despite the difference in their binding kinetic properties they can display the same degree of bias.

Bitopic compounds and allosteric modulators may directly bind to the secondary binding pocket, however, during the association and dissociation process the secondary site plays a crucial role for the appropriate positioning of all compounds. While experiments rarely shed light on the structural details of binding, molecular dynamics simulations can explore the atomistic process and are useful to predict residence time (Potterton et al., 2019; Decherchi and Cavalli, 2020; Lamim Ribeiro et al., 2020; Salmaso and Jacobson, 2020; Bekker et al., 2021; Kokh and Wade, 2021). Ribeiro and co-workers recently used machine learning and infrequent metadynamics to efficiently predict kinetic rates, transient conformational states, and molecular determinants of drug dissociation on the MOR (Lamim Ribeiro et al., 2020). While both investigated compounds bind to the orthosteric pocket, the transient conformational state for the dissociation was identified around the secondary binding pocket suggesting a key role of the secondary site in the association/dissociation process. In dynamic docking simulations Bekker et al. investigated β₂-adrenoreceptor antagonists identifying several stable and metastable conformational states for the compounds along their association/dissociation path (Bekker et al., 2021). Based on these simulations they propose a way to develop allosteric modulators to inhibit the receptor by blocking the path of the endogenous ligand to the orthosteric site. Metastable binding sites play a crucial role in the study of Gaiser et al. as well (Gaiser et al., 2019). They developed homobivalent bitopic ligands for β₂AR to target the OBP and a previously identified metastable binding site as an allosteric site. Among others the residence time of orthosteric and bitopic A_2A receptor binders was predicted with ensemble based steered molecular dynamics (Potterton et al., 2019). Analysis of the pathways revealed dominant interactions, residues influencing the dissociation time and the calculations proposed that changes in water-ligand energy from the ligand in the binding pocket to the extracellular vestibule was the main factor in the determination of residence time. While hydrophilic ligands are expected to access the orthosteric binding site, that is deeply embedded in the center of the receptor, from the aqueous phase, hydrophobic compounds were proposed to entry through lipid pathways. The examples detailed in this part explore the traditional pathway, however cholesterol and other ligands might enter the receptor from the membrane. As an exciting study we refer to the work of Guixá-González et al. who investigated the cholesterol access to the A_2AR with combined computational and experimental methods. They showed that cholesterol’s impact on A_2AR-binding affinity goes beyond pure allosteric modulation and unveils a new interaction mode between cholesterol and the A2AR (Guixà-González et al., 2017). Similar findings were collected and analysed in a recent review dedicated to the role of the lipid bilayer in the binding of the ligands to the orthosteric and allosteric sites (Szlenk et al., 2019). Even though in this review we focused mainly on the secondary binding pocket in the extracellular vestibule that is accessible through the aqueous phase, some allosteric sites on the receptor surface can only be targeted through the membrane fortifying that investigation of the binding pathways through the membrane is also crucial.