Regulation of actions and habits by ventral hippocampal trkB and adolescent corticosteroid exposure

Abstract

In humans and rodents, stress promotes habit-based behaviors that can interfere with action-outcome decision-making. Further, developmental stressor exposure confers long-term habit biases across rodent-primate species. Despite these homologies, mechanisms remain unclear. We first report that exposure to the primary glucocorticoid corticosterone (CORT) in adolescent mice recapitulates multiple neurobehavioral consequences of stressor exposure, including long-lasting biases towards habit-based responding in a food-reinforced operant conditioning task. In both adolescents and adults, CORT also caused a shift in the balance between full-length tyrosine kinase receptor B (trkB) and a truncated form of this neurotrophin receptor, favoring the inactive form throughout multiple corticolimbic brain regions. In adolescents, phosphorylation of the trkB substrate extracellular signal-regulated kinase 42/44 (ERK42/44) in the ventral hippocampus was also diminished, a long-term effect that persisted for at least 12 wk. Administration of the trkB agonist 7,8-dihydroxyflavone (7,8-DHF) during adolescence at doses that stimulated ERK42/44 corrected long-lasting corticosterone-induced behavioral abnormalities. Meanwhile, viral-mediated overexpression of truncated trkB in the ventral hippocampus reduced local ERK42/44 phosphorylation and was sufficient to induce habit-based and depression-like behaviors. Together, our findings indicate that ventral hippocampal trkB is essential to goal-directed action selection, countering habit-based behavior otherwise facilitated by developmental stress hormone exposure. They also reveal an early-life sensitive period during which trkB-ERK42/44 tone determines long-term behavioral outcomes.

Fig 1. Effects of CORT exposure during adolescence: Validation of the procedure.

Top: timeline of experimental events. See also Table 1. (a) Blood serum CORT levels at the end of an 11-d CORT exposure period (from P31–42) did not differ between groups at the beginning of the active cycle (following sleep) but were elevated in CORT-drinking mice (relative to control mice) at the end of the awake, active cycle. n = 5–10/group. (b) Adrenal and thymus gland weights also atrophied following exogenous CORT exposure (left), but glands recovered with a washout period (right). n = 5–6/group. (c) In a progressive ratio test, a history of CORT exposure reduced break point ratios. n = 5–6/group. (d) Forced swim stress in adolescence also increased blood serum CORT (though this effect appeared to habituate with repeated exposure). n = 5–11/group. (e) Further, break point ratios were also reduced, as with CORT exposure. n = 5–10/group. (f) CORT exposure from P31–42 also eliminated dendritic spines on excitatory neurons in the anterior mPFC (prelimbic subregion) of thy1-YFP expressing transgenic mice. n = 7 mice/group. (g) CORT increased overall spine volume. (h) Dendritic spine head diameter did not, however, change. Representative dendrites are adjacent (unprocessed image at top, followed by reconstruction). (i) Finally, in the forced swim test, a history of subchronic CORT exposure did not impact mobility but modified reactivity to an acute stressor: Acute stress induced mobility in control mice, but mice with a history of CORT exposure remained immobile. n = 5–7/group. * p ≤ 0.05, ** p < 0.001 compared to control (red or white bars). Scale bar = 2 μm. Raw data for this figure can be found in S1 Data. CORT, corticosterone; FST, forced swim test; mPFC, medial prefrontal cortex; P, postnatal day; YFP, yellow fluorescent protein.

Fig 2. Greater vulnerability to CORT-induced habits in adolescents than adults.

Experimental timelines are positioned above the response acquisition curves associated with each experiment. Response acquisition curves represent both responses/min, and breaks in the curves represent tests for sensitivity to instrumental contingency degradation. (a) Instrumental response acquisition was intact 2 wk following adolescent CORT exposure. Text below the x axis notes the schedules of reinforcement used throughout (FR1 before test 1, followed by an RI schedule). (b) Sensitivity to instrumental contingency degradation was also initially intact (test 1) in that mice inhibited a response that was unlikely to be reinforced (“degraded” condition), but CORT-exposed mice then developed habit-based response strategies, failing to differentiate between responses (test 2). n = 8–9/group. (c) In a separate group, test 2 was conducted in a distinct environment (“context shift”). In this case, all mice preferentially engaged the response most likely to be reinforced in a goal-directed manner, indicating that CORT-induced habits (in b) are context dependent. n = 7/group. (d) A history of subchronic CORT exposure in adulthood also did not impact instrumental response acquisition. (e) Unlike with adolescent CORT exposure, however, both groups inhibited a response that was unlikely to be reinforced in a goal-directed fashion. n = 7/group. (f) Additionally, all mice inhibited responding following prefeeding with the reinforcer pellets (“devalued”), relative to prefeeding with chow (“non-devalued”), regardless of age of CORT exposure. (g) Another group of adolescent CORT-exposed mice was allowed a longer (4-wk) washout period in order to match the timing of testing in adult CORT-exposed mice. Mice acquired the responses without group differences. (h) Response preferences were intact in test 1, as above. During test 2, mice were initially able to differentiate between the responses that were likely versus unlikely to be reinforced, but response preference decayed in the CORT-exposed mice. n = 11–12/group. Bars/symbols = means + SEMs, * p < 0.05, ** p < 0.001 versus nondegraded or main effects, as indicated. Raw data for this figure can be found in S1 Data. CORT, corticosterone; FR1, fixed ratio 1; RI, random interval.

Fig 3. Adolescent CORT exposure regulates cortico-limbic trkB and ERK42/44 phosphorylation.

(a) Based on our findings, we profiled the neurobiological effects of CORT in a mPFC–vHC–amygdala circuit. (b) Brains were first collected at the end of the CORT exposure period. Top: Adolescent CORT exposure decreased the ratio of full-length to truncated trkB in the mPFC, which would decrease the ability of full-length trkB to initiate intracellular signaling events, illustrated at right. Bottom: trkB.t1 levels alone did not significantly differ. (c) Top: Adolescent CORT exposure also decreased the ratio of full-length to truncated trkB in the vHC and amygdala. Bottom: In the vHC, this was accompanied by an increase in overall trkB.t1 levels. (d) Top: p-ERK42 levels were higher overall in the vHC than in the amygdala, and CORT reduced vHC p-ERK42 (in planned comparisons, see also Table 3). Bottom: The same pattern was detected for p-ERK44. (e) To determine whether this effect was long lasting, we assessed ERK42/44 phosphorylation 12 wk following adolescent CORT exposure. Top: p-ERK42 was reduced in the vHC. Bottom: p-ERK44 was also blunted. (f) Next, we tested the effects of subchronic CORT in adult mice. Top: CORT decreased trkB:trkB.t1 in the amygdala and vHC, as in adolescent mice. Bottom: trkB.t1 was also elevated. (g) Despite these modifications, ERK42/44 phosphorylation was not impacted. Representative blots are adjacent throughout. These and additional analyses are summarized in Table 3. n = 4–10/group throughout. Bars/symbols = means+SEMs, * p < 0.05 versus control within the same brain region. When in the legend, asterisks indicate main effects of CORT. Raw data for this figure can be found in S1 Data. CORT, corticosterone; ERK42/44, extracellular signal-regulated kinase 42/44; mPFC, medial prefrontal cortex; p-ERK, phosphorylated ERK; trkB, tyrosine kinase receptor B; trkB.t1, truncated trkB; vHC, ventral hippocampus.

Fig 4. Durable blockade of CORT-induced habits and depression-like behavior.

(a) Repeated 7,8-DHF treatment increased p-ERK42 in the vHC (b) but had no effects on p-ERK44. (c) At the highest dose tested, levels of the synaptic marker PSD95 were also increased relative to the control and 3 mg/kg groups. Representative blots are below. n = 9–14/group. (d) Experimental timeline. (e) Mice developed food-reinforced instrumental responses without differences between groups. Response acquisition curves represent both responses/min, and breaks in the acquisition curves represent tests for sensitivity to instrumental contingency degradation. (f) As shown in the prior figures, CORT-exposed mice developed habit-based response strategies. 7,8-DHF at 3 mg/kg blocked these habits, as indicated by preferential engagement of the response most likely to be reinforced in the CORT+7,8-DHF group. n = 5–6/group. (g) 7,8-DHF–treated mice were also less immobile in the forced swim test, a durable antidepressant-like effect. n = 7/group. (h) Furthermore, 7,8-DHF dose-dependently blocked CORT-induced deficiencies in break point ratios (while having no effects on responding when a fixed ratio 1 schedule was applied; symbols at right). n = 6–13/group. Bars/symbols = means + SEMs, * p < 0.05, ** p <0.001 following CORTx7,8-DHF interaction, ^ p < 0.001 following response choice interactions, & p = 0.002 versus CORT alone, # p < 0.05 main effect of 7,8-DHF (no interaction). Raw data for this figure can be found in S1 Data. 7,8-DHF, 7,8-dihydroxyflavone; CORT, corticosterone; ERK42/44, Extracellular signal-Regulated Kinase 42/44; p-ERK, phosphorylated ERK; PSD95, postsynaptic density 95; vHC, ventral hippocampus.

Fig 5. Overexpression of truncated trkB induces habits, makes them harder to “break,” and causes depression-like behavior.

(a) We infused a lentivirus expressing truncated trkB (Trkb.t1) or GFP into the vHC and CeA. Large (red) and small (white) infusion sites are represented on images from the Mouse Brain Library [27]. All infusion sites are described in S3 Fig. At bottom right, representative lentiviral GFP expression in the CeA is shown (at gray arrows). The external capsule is highlighted in red, and the image is intentionally overexposed. (b) In Trkb.t1-expressing mice, p-ERK42/44 immunoreactivity was diminished at the infusion site. n = 7–8/group. (c) All mice acquired the instrumental responses. Response acquisition curves represent both responses/min. (d) Trkb.t1 in the vHC and CeA blocked sensitivity to instrumental contingency degradation. Thus, selective Trkb.t1 overexpression recapitulated the long-term effects of adolescent CORT exposure. n = 6–8/Trkb.t1 group; total control = 23. (e) Another group of mice was first trained using a fixed ratio schedule of reinforcement. Then, to build on our findings reported in (d), an RI schedule of reinforcement was applied, with no interruption in training and no differences in responding between groups. (f) In reaction to repeated instrumental contingency degradation training, control mice inhibited a response that was unlikely to be reinforced, their habits “breaking” (“degraded contingency,” right). By contrast, mice with vHC Trkb.t1 failed to inhibit responding. Response rates associated with an intact contingency were unaffected (left). Response rates are represented on 2 plots in the interest of clarity but were compared together by ANOVA. n = 9–10/group. (g) Finally, vHC Trkb.t1 overexpression also decreased responding on a progressive ratio schedule of reinforcement in adulthood, again recapitulating the long-term effects of adolescent CORT exposure. n = 9/group. Bars/symbols = means + SEMs, * p < 0.05, ** p ≤ 0.004. Raw data for this figure can be found in S1 Data. CeA, central nucleus of the amygdala; ERK42/44, Extracellular signal-Regulated Kinase 42/44; p-ERK, phosphorylated ERK; trkB, tyrosine kinase receptor B; trkB.t1, truncated trkB; vHC, ventral hippocampus.