Research Article

Restoring auditory cortex plasticity in adult mice by restricting thalamic adenosine signaling

See allHide authors and affiliations

Science  30 Jun 2017:
Vol. 356, Issue 6345, pp. 1352-1356
DOI: 10.1126/science.aaf4612

Reopening a critical period

Young brains, compared with adult brains, are plastic. This phenomenon has given rise to the concept of critical periods, during which acquisition of certain skills is optimal. In mice, an auditory critical period is only open in early postnatal days. The youthful brain tunes circuits to sounds in its environment in a way that the adult brain does not. This facility may form the basis for childhood language acquisition in humans. Blundon et al. show that by manipulating adenosine signaling in mice, some plasticity of the adult auditory cortex can be regained (see the Perspective by Kehayas and Holmaat). Disruption of adenosine production or adenosine receptor signaling in adult mice leads to improved tone discrimination abilities.

Science, this issue p. 1352; see also p. 1335

Abstract

Circuits in the auditory cortex are highly susceptible to acoustic influences during an early postnatal critical period. The auditory cortex selectively expands neural representations of enriched acoustic stimuli, a process important for human language acquisition. Adults lack this plasticity. Here we show in the murine auditory cortex that juvenile plasticity can be reestablished in adulthood if acoustic stimuli are paired with disruption of ecto-5′-nucleotidase–dependent adenosine production or A1–adenosine receptor signaling in the auditory thalamus. This plasticity occurs at the level of cortical maps and individual neurons in the auditory cortex of awake adult mice and is associated with long-term improvement of tone-discrimination abilities. We conclude that, in adult mice, disrupting adenosine signaling in the thalamus rejuvenates plasticity in the auditory cortex and improves auditory perception.

Acoustic experiences change cortical representations in the auditory cortex. These changes are a substrate for auditory cognition (1, 2). During a postnatal critical period in rodents [postnatal day 11 (P11) to P15], passive exposure to an acoustic tone expands the cortical representation of that frequency in the primary auditory cortex (37). In adulthood, passive exposure is insufficient to induce plasticity. Cortical map plasticity during the critical period helps construct a stable, adapted representation of the auditory world. Similar phenomena in humans are likely necessary for language acquisition (8, 9). The mechanism by which cortical plasticity becomes restricted in adults has been extensively debated (2, 3, 1017). Long-term synaptic plasticity (e.g., long-term potentiation and long-term depression) at excitatory thalamocortical projections is restricted to the same critical period as cortical map plasticity in auditory cortex (1820). In adults, both auditory cortex map plasticity (11, 12, 21) and thalamocortical long-term potentiation and long-term depression are restored when thalamocortical activation is paired with activation of cholinergic input from the nucleus basalis (18, 19). This cholinergic modulation of thalamocortical synaptic plasticity is mediated by adenosine, which inhibits neurotransmitter release via presynaptic A1 adenosine receptors (A1Rs) (22, 23). Blocking or deleting A1Rs removes this inhibition of neurotransmitter release and thus enables the expression of thalamocortical long-term potentiation and long-term depression in the auditory cortex of mice matured beyond the early critical period in vitro (18, 19).

Thalamic adenosine signaling restricts auditory cortex map plasticity

We tested whether auditory cortex map plasticity can be reestablished after the early postnatal critical period by restricting A1R-adenosine signaling in vivo. We compared auditory cortex map plasticity in mature (P45 to P56) wild-type (WT) mice and mice lacking A1Rs (A1R–/–) exposed to pure tones for 5 to 14 days (fig. S1). We mapped auditory cortex by recording sound-evoked responses of neurons situated 300 to 400 μm from the pial surface [corresponding to layer (L) 3/4 thalamorecipient neurons] in anesthetized mice. The characteristic frequency (the frequency that evokes the maximal response at the lowest intensity) of each recording site was determined from responses to a range of tone frequencies and intensities. Auditory cortex maps in mature WT mice were unaffected by tone exposure; however, the percentage of sites with a characteristic frequency equal to the exposure frequency increased in sound-exposed A1R–/– mice (Fig. 1, A and B). This increase was specific to the exposure tone frequency: A1R–/– mice exposed to 7.9 or 16.4 kHz showed a 2.6- or 2.9-fold increase in the percentage of recording sites with characteristic frequencies of 7.9 or 16.4 kHz, respectively, compared to naïve (not exposed to tone) A1R–/– mice (Fig. 1B). We found no effect of tone exposure on tonotopic maps in the auditory thalamus [i.e., the ventral division of the medial geniculate body of the thalamus (MGv)] of mature A1R–/– mice (fig. S2).

Fig. 1 Deletion of A1Rs in auditory thalamus is sufficient for passive tone exposure to induce auditory cortex map plasticity.

(A) Representative auditory cortex maps from WT or A1R–/– mice exposed or unexposed (naïve) to a 16.4-kHz tone. Black and gray depict blood vessels. D, dorsal; V, ventral; R, rostral; and C, caudal. 0, unresponsive sites; X, sites that do not match the tonotopic criteria of primary auditory cortex. (B) Percentage of recording sites versus characteristic frequencies for naïve (black circles, five mice per genotype) or tone-exposed (7.9 kHz: red open circles, five mice per genotype; 16.4 kHz: red filled circles, five mice per genotype) WT (top) and A1R–/– (bottom) mice [two-way repeated measures analysis of variance (ANOVA): genotype WT, F1,5 = 0.644, P = 0.446; two-tailed Student’s t test: genotype A1R–/– (7.9 kHz), t8 =10.516, *P < 0.001; A1R–/– (16.4 kHz), t8 = 5.452, *P < 0.001]. (C to E) Auditory thalamus (MGv) injections of siRNAs and recordings in auditory cortex: percentage of recording sites versus characteristic frequencies in auditory cortex of WT mice exposed to 9.8 kHz (red circles) or naïve mice (black circles) after injections of Adora1 siRNA1 (123) (naïve, four mice; exposed, six mice; t8 = 4.839, *P = 0.001) (C); Adora1 siRNA2 (789) (naïve, four mice; exposed, four mice; t6 = 3.498, *P = 0.01) (D); or control siRNA (naïve, six mice; exposed, four mice; t8 = 0.802, P = 0.4) (E). (F) Percentage of recording sites versus characteristic frequencies in Adora1fl/fl mice exposed to 9.8 kHz and injected with AAV-CamKIIα-GFP (black squares, eight mice) or AAV-CamKIIα-Cre-GFP (red squares, nine mice) into MGv (t15 = 4.612, *P = 0.0003). (G) (Top) Schematic of the prepulse-inhibition acoustic-startle experiment. Prepulse frequencies are 9.800, 9.604, 9.408, 9.016, 8.232, or 6.664 kHz, which are, respectively, 0, 2, 4, 8, 16, or 32% different from the frequency of the background tone. (Bottom) Mean inhibition of the acoustic startle response in Adora1fl/fl mice exposed to 9.8 kHz after injection of AAV-CamKIIα-GFP (black squares, 10 mice) or AAV-CamKIIα-Cre-GFP (red squares, nine mice) into MGv (F1,5 = 42.382, *P < 0.001). CF, characteristic frequency; WN, white noise; PPI, prepulse inhibition.

We tested multiple methods of reducing A1R signaling and found that each enabled cortical map plasticity in adult mice when paired with tone exposure. First, we injected lentiviruses containing small interfering RNAs (siRNAs) against the Adora1 gene, which encodes A1Rs, into MGv or auditory cortex of P45 to P56 mice. Knocking down A1Rs in MGv only (figs. S3 and S4) was sufficient to induce auditory cortex map plasticity in mice exposed to a 9.8-kHz tone (Fig. 1, C to E). Neither naïve mice injected with these siRNAs in MGv nor tone-exposed mice injected with control (scrambled) siRNA showed cortical map plasticity (Fig. 1, C to E). By contrast, tone exposure produced no cortical map plasticity in mice injected with the same Adora1 siRNAs into auditory cortex (fig. S5), suggesting that plasticity at thalamocortical synapses, but not corticocortical synapses, is essential for the auditory cortex map plasticity that we observed. Auditory cortex map plasticity could also be induced in aged (7- to 8-month-old) mice after knocking down A1Rs in MGv (fig. S6).

Second, we conditionally deleted Adora1 in thalamic excitatory neurons of mature (10- to 13-week-old) mice by injecting an adeno-associated virus (AAV) encoding Cre-GFP (AAV-CamKIIα-Cre-GFP) into MGv of newly generated Adora1fl/fl mice (GFP, green fluorescent protein) (fig. S7). After exposure to 9.8 kHz, the number of sites with a 9.8-kHz characteristic frequency increased in auditory cortex of Adora1fl/fl mice injected with Cre-GFP (Adora1fl/fl;Cre mice), compared to Adora1fl/fl mice injected with GFP alone (Adora1fl/fl;GFP mice) (Fig. 1F). Adora1fl/fl;Cre mice exposed to the 9.8-kHz tone were better at distinguishing changes in sound frequencies around 9.8 kHz than were Adora1fl/fl;GFP mice (Fig. 1G). This improvement in auditory perception was tone specific, as Adora1fl/fl;Cre mice exposed to another tone did not show improved performance around 9.8 kHz (fig. S8). Exposure to 9.8 kHz did not affect frequency-discrimination performance in WT mice (fig. S9). Neither exposure to 9.8 kHz nor in vivo injection of Cre-GFP into MGv affected hearing, as measured by auditory brain stem responses or startle response (fig. S10).

Third, we crossed Adora1fl/fl and Gbx2creER mice to produce a line of mice with an inducible Adora1 deletion in thalamocortical neurons. Tamoxifen injection reduced Adora1 mRNA in Gbx2creER;Adora1fl/fl mice (fig. S11). In situ hybridization in WT mice showed that Adora1 is enriched in MGv (fig. S11), suggesting that thalamocortical neurons are especially sensitive to adenosine. Tone exposure produced cortical map plasticity in Gbx2creER;Adora1fl/fl mice. The number of sites with a 9.8-kHz characteristic frequency was increased in these mice, but not in control mice (fig. S12).

Fourth, pharmacologic inhibition of A1R signaling enabled long-lasting auditory cortex map plasticity. We paired tone exposure with a selective, blood-brain barrier–permeable A1R antagonist FR194921 {2-(1-methyl-4-piperidinyl)-6-[2-phenylpyrazolo(1,5-a)pyridin-3-yl]-3(2H)-pyridazinone}, which, after intraperitoneal injection, persists in the brain for several hours (fig. S13). Pairing tone exposure with FR194921 for 5 to 9 days induced auditory cortex map plasticity when measured the following day (fig. S14A). Replacing FR194921 with vehicle or injecting naïve mice with the drug did not induce auditory cortex map plasticity (fig. S14A). Auditory cortex map plasticity induced by pairing tone exposure with FR194921 persisted for weeks after pairing (fig. S14, B and C), as did improved auditory perception (fig. S14, D to F). Cortical map plasticity was malleable. Mice exposed to a second week of FR194921 treatment, paired with a different frequency, exhibited an increased number of auditory cortex sites responding to the second frequency and showed no residual increase in cortical sites responding to the first frequency (fig. S14G).

NT5E and adenosine are increased in adult mice

We then tested auditory cortex map plasticity after reducing thalamic adenosine production. Of the ectonucleotidases that produce adenosine in the brain (24), only ecto-5′-nucleotidase (Nt5e) mRNA was increased in mature versus neonatal thalamus (Fig. 2A). NT5E protein expression was higher in the thalamus and, to a lesser degree, in auditory cortex of adults versus neonates (Fig. 2B). Consistent with this developmental NT5E increase, total (intracellular and extracellular) adenosine was higher in the thalamus and auditory cortex of adults than of neonates (Fig. 2C). In Nt5e–/– mice, this age-dependent increase of adenosine was absent in the thalamus but evident in auditory cortex (Fig. 2C). Auditory cortex map plasticity was induced in mature Nt5e–/– mice, but not in WT littermates exposed to a pure tone (Fig. 2, D and E). Mature tone-exposed Nt5e–/– mice distinguished frequencies better than naïve Nt5e–/– mice (Fig. 2F), without any effect on the auditory brain stem response or startle response (fig. S15). We also knocked down Nt5e in the MGv by injecting two different Nt5e siRNAs that reduced thalamic Nt5e (fig. S16) and produced auditory cortex map plasticity after tone exposure (Fig. 2, G and H). Metabotropic glutamate receptor (mGluR) antagonists inhibit synaptic plasticity at mature thalamocortical synapses (18, 19). CTEP, an mGluR5 inhibitor that crosses the blood-brain barrier (25), blocked auditory cortex map plasticity in tone-exposed Nt5e–/– mice (fig. S17).

Fig. 2 Age-dependent increase of adenosine production underlies cortical map plasticity restrictions in adult mice.

(A) Mean Nt5e, Acpp (prostatic acid phosphatase), and Tnap (liver-bone-kidney alkaline phosphatase) mRNA levels [normalized to Gapdh (glyceraldehyde-3-phosphate dehydrogenase)] in the auditory thalamus of neonatal (P5 to P12, gray bars, seven mice) or mature (P45 to P56, black bars, six mice) WT mice (Nt5e: t11 = 2.325, *P = 0.038; Acpp: t11 = 1.145, P = 0.276; Tnap: t11 = 0.101, P = 0.921). (B) Representative Western blots of MGv extracts (top) and mean NT5E protein levels in the auditory thalamus (t15 = 7.857, *P < 0.001) and auditory cortex (t15 = 3.342, *P = 0.005) (bottom) of neonatal (gray bars, eight mice) or mature (black bars, eight mice) WT mice. (C) Total adenosine concentrations in the auditory thalamus and auditory cortex of neonatal (gray bars, 10 mice) or mature (black bars, 10 mice) WT mice (thalamus: t18 = 11.257, *P < 0.001; auditory cortex: t18 = 10.036, *P < 0.001) and neonatal (pink bars, 9 mice) or mature (red bars, 10 mice) Nt5e–/– mice (thalamus: t17 = 0.961, P = 0.35; auditory cortex: t17 = 3.218, *P = 0.005). (D and E) Percentage of recording sites versus characteristic frequencies in auditory cortex of naïve (black circles) or 11.4-kHz tone–exposed (red circles) WT mice (naïve, four mice; exposed, five mice; t7= 0.283, P = 0.785) (D) and Nt5e–/– mice (naïve, five mice; exposed, six mice; t9 = 2.148, *P = 0.03) (E). (F) Mean prepulse inhibition as a function of prepulse frequency in Nt5e–/– mice exposed to 9.8 kHz (red circles, 17 mice) or naïve (black circles, 20 mice) (F1,6 = 13.643, *P < 0.001). (G and H) Percentage of recording sites versus characteristic frequencies in auditory cortex of naïve (black circles) or 9.8-kHz tone–exposed (red circles) WT mice after injection of Nt5e siRNA1 (1264) (naïve, three mice; exposed, four mice; t5 = 3.33, *P = 0.02) (G) or Nt5e siRNA2 (1366) (naïve, five mice; exposed, seven mice; t10 = 4.79, *P = 0.0007) (H) into MGv. ACx, auditory cortex; CF, characteristic frequency; PPI, prepulse inhibition.

A1R agonist inhibits cortical map plasticity in juvenile mice

In a reverse experiment, we exposed pups to 9.8 kHz during the critical period (P11 to P15), while treating them with the blood-brain barrier–permeable A1R agonist 2-chloro-N6-cyclopentyladenosine (CCPA) or vehicle. We mapped auditory cortex 6 to 10 weeks later (fig. S18A). Tone exposure in these pups produced auditory cortex map plasticity when paired with vehicle, but not when paired with CCPA (fig. S18B).

Auditory plasticity can be reestablished in individual neurons

To test for auditory cortex map plasticity in individual neurons, we used two-photon imaging of sound-evoked calcium transients in thalamorecipient neurons of auditory cortex in mature awake mice before and after tone exposure. We infected L3/4 excitatory neurons in auditory cortex (fig. S19) with a recombinant AAV expressing genetically encoded calcium indicator GCaMP6f (26) under control of the excitatory neuron–specific promoter CamKIIα (AAV-CamKIIα-GCaMP6f) (Fig. 3A). We determined each neuron’s best frequency (i.e., the frequency that evokes the greatest response across all intensities) from calcium transients (fig. S20) evoked by a range of tone frequencies and intensities (Fig. 3, B and C). We measured tuning changes by repeatedly imaging the same auditory cortex neurons before and after as many as 5 days of exposure to 9.8 kHz (Fig. 3D and figs. S21 to S23).

Fig. 3 Restricting thalamic adenosine signaling or production is sufficient for passive tone exposure to cause a tuning shift in individual neurons in auditory cortex of awake mice.

(A) Representative image of GCaMP6f-expressing L3/4 excitatory neurons in auditory cortex. (B and C) Sound-evoked GCaMP6f fluorescence transients (B) and respective receptive field map (C) of an L3/4 neuron. Gray vertical bars in (B) represent sound stimuli. (D) Examples of receptive field heat maps (normalized to maximum intensity) in the same neurons before (day 0) and after (day 5) three experimental conditions: Adora1 siRNA1 (123) in MGv with no tone exposure (two left panels), Adora1 siRNA1 (123) in MGv with 9.8-kHz exposure (four middle panels), and control siRNA in MGv with 9.8-kHz exposure (two right panels). Vertical dotted lines indicate 9.8 kHz. (E to G) Cumulative histograms of best frequencies of recorded neurons in auditory cortex before (black solid lines) and after 9.8-kHz exposure (green lines) under the same three experimental conditions as in (D) (three mice per condition) [(E): 19 neurons, P > 0.05; (F): 34 neurons, *P < 0.05; (G): 25 neurons, P > 0.05]. (H to J) Heat maps (ΔBF maps) depicting shifts in neuron numbers (normalized to percent sampled) as a function of best frequencies between days 0 and 5, under three experimental conditions as in (D), from data in (E) to (G) [(H): P > 0.05; (I): *P < 0.001; (J): P > 0.05). Diagonal lines represent no change in best frequencies. Vertical lines represent 9.8 kHz. (K to N) As described above for (E) to (J), under two experimental conditions (three mice per condition): Nt5e siRNA1 (1264) in MGv with no tone exposure [(K) and (M): 24 neurons, P > 0.05] or Nt5e siRNA1 (1264) in MGv with 9.8-kHz exposure [(L) and (N): 23 neurons; (L): *P < 0.05; (N): *P < 0.001]. BF, best frequency.

All tested approaches that reduced adenosine signaling or production in MGv enabled plasticity in individual neurons when paired with tone exposure. A1R knockdown in MGv with Adora1 siRNA paired with 9.8-kHz exposure shifted best frequencies of individual L3/4 neurons toward 9.8 kHz but not in naïve mice with A1R knockdown or when control siRNA was paired with 9.8-kHz exposure (Fig. 3, D to J). Nt5e siRNA injected into MGv also shifted best frequencies of individual neurons toward 9.8 kHz in mice exposed to this frequency but not in naïve mice (Fig. 3, K to N). Pairing tone exposure with pharmacologic blockade of A1Rs also caused L3/4 neurons to shift best frequencies toward the exposure frequency. We measured receptive fields in the same population of neurons at days 0, 1, 3, and 5 during the 9.8-kHz tone exposure combined with FR194921 administration. L3/4 neurons in auditory cortex shifted their best frequencies toward 9.8 kHz starting at day 1 and maintained this shift during all the days of tone exposure (fig. S24, A to F). By contrast, pairing 9.8-kHz exposure with vehicle did not shift best frequencies of L3/4 neurons toward 9.8 kHz on any tested day (fig. S24, G to L).

Similar experiments in Adora1fl/fl;Cre mice (lacking A1Rs in MGv neurons) showed that L3/4 neurons in auditory cortex also shifted best frequencies toward the exposure frequency. Cortical neurons of Adora1fl/fl mice, which received MGv injections of Cre-tdTomato (AAV-CamKIIα-Cre-tdTomato) (fig. S25) and were exposed to 9.8 kHz, shifted best frequencies toward 9.8 kHz at day 1 and maintained this shift at days 3 and 5 of tone exposure (fig. S24, M to R). By contrast, this best frequency shift was not seen in L3/4 neurons of WT littermates injected with Cre-tdTomato into MGv and exposed to 9.8 kHz for the same periods of time (fig. S24, S to X). A switch in best frequency after tone exposure did not depend on a neuron’s receptive field properties or GCaMP6f-response amplitudes (fig. S26).

These findings suggest that an NT5E-mediated, age-dependent increase in adenosine production in the auditory thalamus terminates the juvenile critical period of auditory cortex map plasticity via A1R activation. A1R activation inhibits virtually every neurotransmitter in the brain, and its strongest inhibitory action is on excitatory glutamatergic synapses (22). A1R activation reduces presynaptic Ca2+ influx and hyperpolarizes presynaptic membrane potential by activating inwardly rectifying K+ channels (22). During the critical period, when adenosine levels are low, sustained activity in auditory thalamocortical synapses triggers enough glutamate release to activate mGluRs and produce long-term potentiation or long-term depression. In adults, higher NT5E expression leads to increased production of adenosine, which limits sustained glutamate release via A1R activation, blocking thalamocortical synaptic plasticity and thereby preventing cortical map plasticity. Restricting thalamic adenosine signaling is sufficient to restore higher levels of glutamate release and synaptic plasticity at thalamocortical synapses in brain slices from adult mice (18, 19, 27) and reestablish cortical map plasticity in adult mice in vivo.

Experimental manipulations in adult mice that were expected to enable thalamocortical synaptic plasticity (i.e., knockdown or deletion of A1Rs or knockdown of NT5E in thalamic neurons) also facilitated auditory cortex map plasticity. Manipulations expected to inhibit thalamocortical synaptic plasticity (i.e., mGluR5 antagonist treatment in Nt5e–/– adult mice and A1R agonist treatment during the postnatal critical period) blocked cortical map plasticity. Knockdown of A1Rs in cortical neurons failed to potentiate cortical map plasticity in adult mice. These results point to thalamocortical long-term plasticity as the necessary step that leads to cortical map plasticity and enhanced auditory perception. However, our results do not exclude the possibility that the improvement in auditory perception may result from additional changes at other synapses in the auditory cortex as a downstream consequence triggered by thalamocortical synaptic plasticity.

Implications for the human context

Pairing tone exposure with administration of the specific A1R antagonist FR194921 resulted in cortical map plasticity and improved auditory perception in adult mice. These results imply that other A1R inhibitors, such as caffeine, could be used to improve auditory perception in adult humans. However, because caffeine, especially at high concentrations, inhibits other adenosine receptors and affects intracellular cyclic adenosine 3′,5′-monophosphate (cAMP) and calcium levels (28), its effectiveness at improving auditory perception could be limited, compared to that of FR194921 or other specific A1R inhibitors.

Cortical plasticity in adults holds potential as a strategy to restore normal neural activity in patients with neurological diseases, such as tinnitus and stroke (2932). One current method for inducing therapeutic targeted cortical map plasticity in patients is the pairing of sensory stimulation (i.e., a sound) or motor training with repeated electrical stimulation of the vagus nerve by using surgically implanted electrodes. Our work offers an alternative, noninvasive method for inducing lasting cortical map plasticity, namely, by pairing a sound with administration of a specific A1R inhibitor that has a short half-life.

Adult cortical plasticity is associated with improved tone-frequency discrimination. In humans, tone discrimination is required for acquiring language and musical skills. Tonal representation in auditory cortex is larger in musicians than in nonmusicians (33). This difference is observed primarily in musicians who began practicing their instruments before 9 years of age (34); it is also associated with improved linguistic skills (35, 36). Here we show that disruption of thalamic adenosine signaling in the adult mouse that restores auditory cortex map plasticity also improves auditory perception. A similar mechanism may improve the ability of adult humans to learn linguistic, musical, and other auditory skills.

Supplementary Materials

www.sciencemag.org/content/356/6345/1352/suppl/DC1

Materials and Methods

Figs. S1 to S26

References (3761)

References and Notes

Acknowledgments: This work was supported by NIH grants DC012833, MH097742 (S.S.Z.), and DC015388 (T.A.H.), and by American Lebanese Syrian Associated Charities (ALSAC) (S.S.Z.). The funding sources had no role in the study design, data collection, data analysis, decision to publish, or preparation of the manuscript. We thank the Vector Core Laboratories (St. Jude Children’s Research Hospital and University of Tennessee Health Science Center) and J. Westmoreland for assistance in producing AAVs and lentiviruses; F. Du for assistance in generating Adora1fl/fl mice and pharmacokinetic studies; J. Min and A. Mayasundari for help with FR194921; A. Onar-Thomas for statistical assistance; A. Uptain and K. Anderson for assistance with imaging experiments and viral injections in vivo; and V. Shanker and A. McArthur for editing the manuscript. The supplementary materials contain additional data. S.S.Z. and J.A.B. are inventors on patent application PCT/US2016/018377 submitted by St. Jude Children’s Research Hospital, which covers the subject of improved learning through inhibition of adenosine signaling in the thalamus.
View Abstract

Navigate This Article