Careful analysis of the NMR structures of cyclo(4-10)[Ac-Delta(3)Pro(1),DFpa(2),DTrp(3),Asp(4),DNal (6), Dpr(10)]GnRH, dicyclo(4-10/5-8)[Ac-DNal(1),DCpa(2),DTrp(3), Asp(4), Glu(5),DArg(6),Lys(8),Dpr(10)]GnRH, and dicyclo(4-10/5, 5'-8)[Ac-DNal(1),DCpa(2),DPal(3),Asp(4), Glu(5)(Gly),DArg(6),Dbu(8), Dpr(10)]GnRH showed that, in the N-terminal tripeptide, a type II beta-turn around residues 1 and 2 was probable along with a gamma-turn around DTrp(3)/DPal(3). This suggested the possibility of constraining the N-terminus by the introduction of a cyclo(1-3) scaffold. Optimization of ring size and composition led to the discovery of cyclo(1-3)[Ac-DAsp(1),DCpa(2),DLys(3),DNal(6), DAla(10)]GnRH (5, K(i) = 0.82 nM), cyclo(1,1'-3)[Ac-DAsp(1)(Gly), DCpa(2),DOrn(3),DNal(6),DAla(10)]GnRH (13, K(i) = 0.34 nM), cyclo(1, 1'-3)[Ac-DAsp(1)(Gly),DCpa(2),DLys(3),DNal(6),DA la(10)]GnRH (20, K(i) = 0.14 nM), and cyclo(1,1'-3)[Ac-DAsp(1)(betaAla), DCpa(2), DOrn(3),DNal(6),DAla(10)]GnRH (21, K(i) = 0.17 nM), which inhibited ovulation significantly at doses equal to or lower than 25 microgram/rat. These results were particularly unexpected in view of the critical role(s) originally ascribed to the side chains of residues 1 and 3.(1) Other closely related analogues, such as those where the [DAsp(1)(betaAla), DOrn(3)] cycle of 21 was changed to [DOrn(1)(betaAla), DAsp(3)] of cyclo(1,1'-3)[Ac-DOrn(1)(betaAla), DCpa(2),DAsp(3),DNal(6),DAla(10)]GnRH (22, K(i) = 2.2 nM) or where the size of the cycle was conserved and [DAsp(1)(betaAla), DOrn(3)] was replaced by [DGlu(1)(Gly), DOrn(3)] as in cyclo(1, 1'-3)[Ac-DGlu(1)(Gly),DCpa(2),DOrn(3),DNal(6),DA la(10)]GnRH (23, K(i) = 4.2 nM), were approximately 100 and 25 times less potent in vivo, respectively. Analogues with ring sizes of 18 ¿cyclo(1, 1'-3)[Ac-DGlu(1)(Gly),DCpa(2),DLys(3),DNal(6),DA la(10)]GnRH (24)¿ and 19 ¿cyclo(1,1'-3)[Ac-DGlu(1)(betaAla),DCpa(2),DLys( 3),DNal(6), DAla(10)]GnRH (25)¿ atoms were also less potent than 21 with slightly higher K(i) values (1.5 and 2.2 nM, respectively). These results suggested that the N-terminal tripeptide was likely to assume a folded conformation favoring the close proximity of the side chains of residues 1 and 3. The dicyclic analogue dicyclo(1-3/4-10)[Ac-DAsp(1),DCpa(2),DLys(3),Asp (4),DNal(6), Dpr(10)]GnRH (26) was fully active at 500 microgram, with a K(i) value of 1 nM. The in vivo potency of 26 was at least 10-fold less than that of monocyclic cyclo(1-3)[Ac-DAsp(1),DCpa(2),DLys(3),DNal(6), DAla(10)]GnRH (5); this suggested the existence of unfavorable interactions between the now optimized and constrained (1-3) and (4-10) cyclic moieties that must interact as originally hypothesized. Tricyclo(1-3/4-10/5-8)[Ac-DGlu(1),DCpa(2), DLys(3),Asp(4),Glu(5), DNal(6),Lys(8),Dpr(10)] GnRH (27) was inactive at 500 microgram/rat with a corresponding low affinity (K(i) = 4.6 nM) when compared to those of the most potent analogues (K(i) < 0.5 nM).