NO may mediate synaptic plasticity changes in brain regions such as the hippocampus in order to influence both long-term potentiation (LTP) and memory formation3,4,5,6,7, but this idea is controversial9,10. In rodents, the inhibitory GABAergic granule and periglomerular cells of the primary sensory processing region for airborne odours, the olfactory bulb, contain large amounts of nNOS11, which converts arginine to citrulline and NO3,4. We investigated whether this could also be the case in sheep, and whether the increased sensitivity of mitral-to-granule cell synapses that results from learning the odours of their lambs might involve potentiation of glutamate release by NO through modulation of cGMP.

We cloned and sequenced the ovine nNOS gene and found that it was 88% homologous with human12, rat13 and mouse14 sequences (EMBL accession number, X99042). In situ hybridization and immunocytochemistry indicated that mitral as well as periglomerular and granule cells contained nNOS messenger RNA and protein (Fig. 1a). The dendrodendritic connections between the granule and mitral cells in particular contained large amounts of nNOS immunoreactivity. Only mitral cells expressed the mRNAs for the αa and β1 subunits of soluble guanylyl cyclase (Fig. 1a), which are both required for NO to cause cGMP formation8. The granule and periglomerular cells expressed low levels of only the β1 subunit. Thus NO probably only stimulates cGMP formation in mitral cells.

Figure 1: a, Photomicrographs of nNOS protein expression in olfactory bulb (OB) mitral (top arrow in top panel) and granule (bottom arrow in top panel) cells, and emulsion autoradiography (bottom two panels) showing the mitral cell layer (arrow) expressing mRNA for the α1 and β1 subunits of guanylyl cyclase.
figure 1

Scale bar, 200 μm. b, Effects of SNAP, NMDA and AMPA on OB levels of glutamate, GABA, nitrite and cGMP alone, and after 6 h of treatment with ODQ or L-NARG or 1 h of treatment with selective NMDA, D,L-2-amino-5-phosphopentanoic acid (AP5) or AMPA/kainate, 6,7-dinitroquinoxaline2,3-dione (DNQX) antagonists. Values are given as mean ± s.e.m. percentage change, and drug concentrations are in μM. * P < 0.05 versus baseline (100%) and # P < 0.05 versus the same dose within antagonist/inhibitor treatment (two-tailed Wilcoxon test). L-NARG treatment reduced basal nitrite levels by 40% and ODQ basal cGMP levels by 50%. c, Left, OB organization. Arrows show direction of inputs; and the plus and minus signs show excitatory and inhibitory influences. The mitral-to granule cell synapses circles are those where NO is proposed to act to potentiate glutamate release. Glu, glutamate; NA, noradrenaline; PG, periglomerular cell; VCS, vaginocervical stimulation. Right, proposed interactions between glutamate receptors and the NO/cyclic GMP pathway at the mitral-to-granule cell synapses during memory formation. AMPA-R, AMPA receptor; G, G protein; mGluRs, metabotropic receptors; NMDA-R, NMDA receptor.

In brain, NO is released by glutamate acting on both NMDA (N-methyl-D-aspartate) and AMPA ((2-aminomethyl)phenylacetic acid) receptors3,4,15. In the mouse accessory olfactory bulb, activation of both receptor types is required for the formation of pheromonal olfactory memory16 and both are present on the main olfactory bulb mitral and granule cells17. We used in vivo microdialysis to determine whether glutamate acts on both types of receptors to evoke NO and cGMP release in the olfactory bulb. Local retrodialysis infusions of NMDA or AMPA dose-dependently increased glutamate, GABA, NO and cGMP, although NMDA was more potent. These actions were blocked by specific receptor antagonists (Fig. 1b) and the NOS inhibitor L-nitroarginine (L-NARG). A selective inhibitor of NO-induced guanylyl cyclase, 1H-[1,2,4]oxadiazolo[4,3,-a]quinoxalin-1-one (ODQ)18 also prevented stimulation of cGMP, but not NO release (Fig. 1b). Both L-NARG and ODQ significantly reduced agonist-induced increases in glutamate and GABA, showing that agonist effects are partly mediated by NO and cGMP. Infusions of an NO donor, S-nitrosoacetylpenicillamine (SNAP), dose-dependently increased cGMP, glutamate and GABA; these effects were blocked by ODQ (Fig. 1b). Thus NO can act as a retrograde or intracellular messenger in the olfactory bulb by stimulating glutamate release through modulation of cGMP.

The importance of NO release in the olfactory bulb for plasticity changes underlying olfactory memory was investigated in post partum animals given bilateral local infusions by microdialysis probes of drugs targeting the NO signalling pathway. In control animals receiving either no treatment or an inactive enantiomer of L-NARG (D-NARG), which did not inhibit NOS, it was found that glutamate, GABA, noradrenaline, NO and cGMP all increased during the first 30 min after birth (Figs 2 and 3). Behavioural tests showed that all these animals formed an olfactory memory that allowed selective recognition of lambs (Fig. 4a). However, all animals treated with the ionotropic glutamate receptor antagonist γ-D-glutamylglycine (DGG), and NOS inhibitor L-NARG, or the guanyl cyclase inhibitor ODQ, accepted their own and strange lambs equally, indicating that these agents completely prevented olfactory memory formation (Fig. 4a). The quality of maternal care shown by ewes towards their lambs during the two hours immediately post partum was unaffected by the drugs, confirming that normal olfactory bulb function is not important for maternal behaviour per se in multiparous animals19. Both DGG and L-NARG prevented birth-induced increases in NO, cGMP, glutamate and GABA (Figs 2 and 3). With ODQ treatment, there was the normal increase in NO at birth, but there was no increase in cyclic GMP, glutamate or GABA (Fig. 2). Drugs targeting the NO signalling pathway therefore prevent both memory formation and the normal birth-induced potentiation of glutamate release. The birth-induced increase in noradrenaline release was unaffected by DGG, L-NARG and ODQ, even though memory formation was prevented. Thus noradrenergic action on olfactory memory formation20 may be effected by the NO signalling pathway.

Figure 2: Glutamate, GABA, noradrenaline, citrulline, nitrite and cGMP levels before, during, and after birth in control animals (dotted line), and in DGG (500 μM; dashed line)- and ODQ (200 μM solid line)-treated animals.
figure 2

*P < 0.05, compared with mean levels in the four samples before birth (two-tailed Wilcoxon test). Values are mean ± s.e.m. percentage change relative to the time point 75 min before birth (set at 100%).

Figure 3: Glutamate, GABA, noradrenaline, citrulline, nitrite and cGMP levels before, during, and after birth in control D-NARG (500 μM solid line)- and L-NARG (500μM dashed line)-treated animals.
figure 3

*P < 0.05 compared with mean levels in the four samples before birth (two-tailed Wilcoxon test). Values are mean ± s.e.m. percentage change relative to the time point 75 min before birth (set at 100%).

Figure 4: a, The effects of drug infusions in the olfactory bulb on acceptance (low-pitch bleating and suckling) and rejection behaviour (butting), as shown by maternal ewes towards their own (white bars) and strange (black bars) lambs.
figure 4

For memory formation, values are mean ± s.e.m. of 4 tests (2, 4, 6 and 8 h post partum), and for recall, from 2 tests (44 and 46 h post partum). * P < 0.05 versus own-lamb (two-tailed Wilcoxon test). DGG (50 μM) and L-NARG (100 μM) had no effects. b, Mean ± s.e.m. percentage change in levels of glutamate (white bars) and GABA (hatched bars) in 15-min microdialysis samples taken from the olfactory bulb during 10-min exposures to own-lamb odours in the tests. * P< 0.05 (two-tailed Wilcoxon).

If NO does mediate plasticity changes in the olfactory bulb at birth, blockage of its release and olfactory memory formation by a NOS inhibitor might be reversed by local NO infusions. When a nitric oxide donor, SNAP, was co-infused with L-NARG for 1.5 h post partum, then memory formation was unaffected (Fig. 4a); however, the NO donor did not reverse ODQ-induced blockade of memory formation, indicating that NO released artificially must also act by stimulating cGMP accumulation in mitral cells. As infusing NO into the olfactory bulb for 1.5 h allowed a specific olfactory memory to form, then the NO must promote plasticity changes only at those synapses between mitral and granule cells that are activated by both birth and specific lamb odours. If NO caused all synapses to be potentiated non-selectively, this would simply increase the overall level of noise in olfactory bulb circuits. Activity-dependent plasticity changes are also a characteristic of hippocampal synapses showing LTP21 and the latter can be induced by NO infusion5.

None of the drugs that blocked olfactory memory formation did so irreversibly. After treatments ended at 8 h post partum, all animals learned to recognize their lambs selectively within 16 h. To test whether the NO signalling pathway is important for memory recall, we re-infused the drugs for a further 20 h. The ewes continued to recognize their own lambs' odours selectively (Fig. 4a). Thus this NO signalling pathway is not critical for memory recall and cannot be important for odour perception per se. Our finding that even DGG failed to block memory recall confirms that in the olfactory bulb, as in the hippocampus22, NMDA-receptor activation is required for memory formation but not for recall. It also indicates that the DGG infusions must have been restricted to the mitral and granule cell layers, because if they had spread into the glomerular layer and blocked ionotropic receptors on the mitral cell apical dendrites, then the cells could not have responded to inputs from the olfactory nerve and odour perception would have been impaired.

After memory formation, own-lamb odours selectively increase glutamate and GABA1 release, demonstrating that there is enhanced excitation of both the glutamatergic mitral cells and their associated inhibitory GABAergic granule-cell interneurons. If GABA is prevented from acting on GABAA receptors on mitral cells, sheep are unable to recognize their lambs selectively after memory formation2. If NO acts by promoting synaptic changes during memory formation that allow own-lamb odours to stimulate glutamate and GABA release more effectively, then drugs targeting this pathway should prevent potentiation of transmitter release, which is borne out by our results. Thus, whereas a 10-min exposure to own-lamb odours significantly increased glutamate and GABA in the olfactory bulb after memory formation in control animals, it did not do so in DGG, L-NARG or ODQ-treated animals in which memory formation was blocked (Fig. 4b). But when the NO donor SNAP prevented blockade of olfactory memory by L-NARG, then own-lamb odours did increase glutamate and GABA release (Fig. 4b). When memory formation occurred at the end of the drug treatment, own-lamb odours were able to evoke increased glutamate and GABA release, which was not prevented by their re-administration (Fig. 4b). Own-lamb odours never significantly increased NO or cGMP in these experiments, confirming their lack of direct involvement in memory recall.

The fact that lamb-odour recognition after memory formation is not associated with NO release or affected by the ionotropic receptor antagonist DGG suggests that the NMDA and AMPA receptors are not stimulated much. Odour-evoked increases in glutamate and GABA release may involve glutamate acting on metabotropic receptors (mGluRs) whose sensitivity has been increased by NO-induced plasticity changes during memory formation. In particular, mGluR1 is localized presynaptically on mitral cells as well as postsynaptically on granule cells23.

Our findings show that NO can mediate those plasticity changes in the brain that underlie memory formation through cGMP-dependent potentiation of glutamate release. We propose (Fig. 1c) that olfactory memory formation induced at birth involves several steps. First, vaginocervical stimulation during birth evokes noradrenaline release at the centrifugal synapses with olfactory bulb granule cells, which acts at β-noradrenergic receptors20 to reduce GABA release at the granule-to-mitral cell synapses. This relieves the inhibition of the mitral cells, and glutamate released at their reciprocal synapses with granule cells acts on NMDA and/or AMPA receptors to stimulate NO release from both pre- and postsynaptic sites. Nitric oxide further potentiates glutamate release through modulation of cGMP, resulting in a long-term increase in the sensitivity of the mitral-to-granule cell synapses to glutamate. This positive-feedback cycle has parallels with hippocampal LTP5,21,22,24,25. When mitral cells responding to the lamb's odours are activated after memory formation, glutamate becomes more potent, acting by means of autoreceptors both to increase the excitation of these cells and subsequently to inhibit them further through GABA release potentiated from reciprocal granule-cell synapses. This results in an enhanced pattern of phasic firing discharges from the cells in response to own-lamb odours. In this way, their output is more easily decoded by subsequent secondary and tertiary olfactory processing regions in the brain so that only own-lamb odours evoke maternal acceptance.


Gene sequencing and histochemistry. To identify the sheep nNOS gene, we generated a polymerase chain reaction (PCR) using specific primers to a conserved region of an exon-2 alignment between mouse, rat and human sequences. We sequenced 218 bases of the gene and used a digoxigenin-labelled ribroprobe to show which olfactory bulb cells expressed it, by using in situ hybridization. We used 45-nucleotide antisense probes from α1 (bases 1,197–1,240 and 1,377–1,423) and β1 (bases 1,168–1,213) bovine soluble guanylyl cyclase gene26 subunits to localize their mRNA by autoradiography. The olfactory bulb cells and processes containing nNOS protein were revealed immunocytochemically using a polyclonal antibody (from Affiniti, Exeter; NC3000).

In vivo microdialysis. In the first experiment, 9 non-pregnant multiparous sheep were treated for 6 weeks with vaginal sponges containing oestradiol and progesterone to induce lactation and to allow maternal behaviour to be stimulated by vaginocervical stimulation. Microdialysis probes (4-mm membrane CMA-10 from CMA Microdialysis, Sweden) were inserted in the olfactory bulb by using surgically implanted guide tubes with Kreb's ringer solution (pH 7.4) pumping through them at 2 μl min−1. Samples were collected at 15-min intervals. The effects of 15 min retrodialysis of NMDA, AMPA and SNAP from the microdialysis probes on NO (citrate, nitrite and nitrate15), cGMP and transmitter concentrations were measured alone or after 6 h retrodialysis infusion of L-NARG or ODQ, or 1-h infusions of AP5 or DNQX. Citrate, nitrite and nitrate and transmitters were measured by HPLC15, and cGMP was measured by ELISA (Cayman). Animals were tested at weekly intervals for six weeks, with drug treatments given in a random order. For all experiments, concentrations were expressed as mean ± s.e.m. percentage change from control (100%). Overall means ± s.e.m. baseline levels measured were: cGMP, 122 ± 26 pM; glutamate, 974 ± 107 nM; GABA, 66 ± 7 nM; nitrite, 669 ± 33 nM; nitrate, 1,534 ± 104 nM; noradrenaline, 421 ± 70 pM. Bilateral microdialysis sampling was also carried out in the olfactory bulb of multiparous pregnant ewes as described1, starting 12–36 h pre partum. Neurochemical and behavioural assessments of olfactory memory formation and/or recall in control (n = 7) or D-NARG (n = 6)-treated animals were compared with those receiving retrodialysis infusions of DGG (500 μM, n = 6; or 50 μM, n = 4), L-NARG (500 μM, n = 6; or 100 μM, n = 4) or ODQ (200 μM, n = 5). Drugs were given bilaterally for 6–24 h before birth and up to 8 h after it. Additional animals were given L-NARG (n = 4) or ODQ (n = 3) but SNAP was added for 1.5 h after birth. Drug treatments were repeated for 20 h starting 24 h post partum. Maternal responses to lambs were recorded for 2 h post partum (licking duration, low-pitched bleats, udder acceptance, suckling latency and duration). Behavioural assessments of olfactory memory formation were made at 2-h intervals starting 2 h after birth. First the ewe's own lamb was removed for 15 min, then a strange one was placed in her pen for 10 min, and maternal acceptance (low-pitch bleats, licking, acceptance of suckling, and suckling duration) and rejection (butting and refusing suckling attempts) behaviours scored. Her own lamb was returned 5 min later and behaviour scored for 10 min. Memory formation was considered to have failed if ewes showed full maternal acceptance (low-pitched bleating, licking and allowing suckling) of strange lambs and did not reject them by butting or refusing their suckling attempts on more than a single occasion; and if the frequency and duration of these behaviours towards strange lambs did not differ significantly from those shown towards their own lambs. Observers were >95% in agreement on scores and treatments given blind.

Tissue NOS levels. After the experiments, animals were killed and their brains removed to confirm that the probes were correctly placed in the olfactory bulb and to quantify NOS activity by measuring the conversion of radiolabelled arginine to citrulline27. After L-NARG treatment (500 μM), NOS activity was reduced by 62% (P < 0.01, t-test versus D-NARG controls). Levels in the hippocampus, amygdala, frontal cortex and cerebellum were unaffected. L-NARG (100 μM) only reduced NOS activity in the olfactory bulb by 23% (P < 0.05); DGG and ODQ had no effect in any region. A 10-min assay time is used in this standard method of assessing NOS activity27 and may underestimate control NOS activity levels28.