an H3 binder retaining the interesting D2/D3 pattern. Notably, among the triple acting compounds, exemplified by the
7b-7c pair, adding a methylene to the side chain transferred the emphasis from 5-HT6 affinity to D3 while preserving
low D2 and moderate H3 affinities.
The collected dataset has been analysed from a different viewpoint to identify the structure-activity relationship
(SAR) patterns responsible for the multi-target nature of the investigated compound set. To do this, compound
promiscuity was plotted using three different threshold values in terms of the binding percentages: >40%, >65% and
>90% (Figure 1). This chart represents the SAR from an exceptional perspective, since it shows that this limited
chemical space is able to cover a wide range of promiscuity: from totally inactive, through single, dual, triple to fully
active ligands. The desirable impact on affinity - thus higher promiscuity - of the piperazine ring compared to
piperidine ring is also unequivocal from Figure 1. Promiscuity increases by growing the R substituents through Me-
Et-Pr at pos1, see 1b, 1c and 1d in Figure 1. A similar trend can be recognized in compound groups 6, 7, 8 and 10.
The highest multi-target effect was realized in compound group 6. In order to facilitate understanding of the cross
target effects, all pair-wise comparisons are available as Supporting Information Figure 2. As expected, the activities
on the different targets are independent, with the exception of D2-D3 pairing which shows binding site similarity.
In order to challenge the quality of the binding interactions, ligand efficiency11 (LE) values were calculated for all the
cases where Ki had been determined. In the case of 5-HT6 all the ligands could be characterized with high (LE>0.3)
LE (Supporting information Figure 3). Moreover, plotting LE versus pKi revealed that the investigated compound
collection shows a strong inverse correlation (R=0.92). Thus, the smaller the ligand the higher the LE, which might
suggest that the key pharmacophore features of the scaffold are exposed optimally and thus the quality of the binding
interactions should not be enhanced by further hydrophobic enlargement or solvation-mediated effects (the maximal
LE was 0.52, compound 1a). In the case of the LE values calculated from the D3 Ki a similar trend can be observed,
albeit with a weaker correlation and smaller slope (Supporting Information Figure 3). For the D2 and H3 targets the LE
values are also in the desirable range, although the small number of examples does not permit a deeper analysis. For
the 5-HT6-D3 pairings (the group with the largest number of measured Ki values) lipophilic ligand efficiency12 (LLE)
values were also calculated (Supporting Information Figure 4). This analysis showed that dual-acting ligands with
desirable lipophilic ligand efficiency (LLE>4) on both targets could be identified (mainly piperazines).
Having observed that the affinity change from H to the bulkier Pr group is different for 5-HT6 compared to D3 and D2,
we were curious to analyse this phenomenon in terms of the lipophilicity versus the affinity change (Supporting
Information Figure 5). Interestingly, in the case of 5-HT6 a weak inverse correlation was found between the two that
further underpins the previous observations as discussed above. This evaluation suggests that the systematic SAR
exploration of well-known pharmacophore elements is a fruitful approach to identify compounds with desirable
physicochemical and binding parameters on multiple targets.
Providing a different angle on the conceptual approach to selectivity design, the present study supplies evidence that
altering only the given spatial arrangement of relevant functionalities together with adjusting hydrophobic properties
around hot spot interacting key moieties, can have a substantial influence on a compounds’ binding profile even in the
intrinsically limited chemical space characterised by low molecular weight. Herein, we demonstrate that in addition to
high affinity selective ligands (e.g. 1a, 1b, 5a and 5b for 5-HT6; 4e for H3; 4d for D3), potent dual- (e.g. 8a and 8b for
5-HT6 and D3; 1e for 5-HT6 and D2; 4c for D3 and H3; 5e for D2 and H3) and triple- (e.g. 6d and 10d for 5-HT6, D3
and D2; 5c, 7b and 7c for 5-HT6, D3 and H3) acting agents could also be identified among this set of compounds,
implying the opportunity of MDL design. In conclusion, our detailed research offers a new insight into drug
promiscuity and has served as a starting point for various MDL programs that have strongly supported the validity of
the strategy presented. Further optimisation from selected multiple acting starting points revealed by this study and
the transfer of SAR to other core structures will be published under different cover.
Acknowledgement
We profoundly appreciate Derek Buckle for his contribution in the preparation of the manuscript. O. É. is deeply
indebted to N. Bús for preparative assistance. Contributions of Katalin Kovács, Éva Schmidt, Ferenc Horti, Károly
Fazekas and Sándor Kolok to the in vitro pharmacological measurements are gratefully acknowledged. Authors thank
Márta Sípos and Zoltán Szakács for their valuable help in the purity assessment and compound characterisation.
References and notes
1. (a) Cerejeira, J.; Lagarto, L.; Mukaetova-Ladinska, E. B. Front. Neurol. 2012, 3, 1; (b) Lyketsos, C. G.; Carrillo, M. C.;
Ryan, J. M.; Khachaturian, A. S.; Trzepacz, P.; Amatniek, J.; Cedarbaum, J.; Brashear, R.; Miller, D. S. Alzheimers
Dement. 2011, 7, 532; (c) Nowrangi, M. A.; Lyketsos, C. G.; Rosenberg, P. B. Alzheimers Res. Ther. 2015, 7:12.
2. (a) Roth, B. L.; Sheffler, D. J.; Kroeze, W. K. Nature Rev. Drug Discov. 2004, 3, 353; (b) Csermely, P.; Ágoston, V.;
Pongor, S. Trends Pharm. Sci. 2005, 26, 178; (c) Morphy, R.; Rankovic, Z. J. Med. Chem. 2005, 48, 6523; (d) Millan,