Introduction

Depression is a common and debilitating disorder. Patients with depression suffer from symptoms including amotivation, anhedonia, and social withdrawal, suggesting the involvement of the rewarding system in pathophysiology of depression. Ventral tegmental area (VTA) dopamine (DA) neurons are critical to processing salient signals [1] and modulating reward, aversion, and stress responses [2,3,4]. They respond differently to appetitive and aversive stimuli (acute or chronic), and mediate different responses [5,6,7,8]. For example, in response to chronic social defeat stress (CSDS), the activity of VTA DA neurons is reduced in resilient mice because of increased K+ currents, whereas K+ currents are hardly changed in susceptible mice [9]. However, underlying regulatory mechanisms of potassium (K+) channels are not well understood. Nevertheless, inability of VTA DA neurons to adapt to and cope with long-term stress has been implicated in depression pathology [9,10,11]. VTA DA neurons are tightly regulated by neurons of different brain regions. For example, in response to stress, VTA-projecting neurons of the ventral pallidum (VP) and norepinephrine (NE) neurons in the locus coeruleus (LC) increase in firing, which are required for susceptibility and resilience of depression-like behaviors, respectively [12,13,14]. On the other hand, chemogenetic inhibition of cholinergic neurons in LDTg or blockade of acetylcholine receptors (AChRs) in VTA inhibits the firing of VTA DA neurons and prevents the development of depression-like behaviors. However, how these circuits regulate the activity of VTA DA neurons remains unclear.

ErbB4 is a receptor tyrosine kinase of the EGF receptor family; it binds to and thus is activated by NRG1, a trophic factor [15,16,17,Behavioral tests

CSDS and subthreshold social defeat (subSD) were performed as previously reported [5, 7]. In CSDS, tested mice received 10 consecutive days of social defeat. Each day, the tested mouse was introduced into the home cage of an unfamiliar CD1 aggressor and received a 5-min physical attack. After physical attack, the tested mouse was maintained in the cage of aggressor, which was partitioned into two halves with a transparent and perforated plexiglass plate. The CD1 aggressors were prescreened and those showed an attack latency < 30 s in three tests were employed. Severe tissue damage was avoided. After social defeat procedure, tested mice were individually housed until all tests finished. The paradigm of subSD consisted of two rounds of 2-min physical attack and 10-min sensory contact within one day, intermitted by a 5-min rest period in the home cage of the tested mice.

Social avoidance and sucrose preference were tested 24 h post the final social defeat. Social avoidance was tested in a plastic box (50 cm × 50 cm × 20 cm), with an overhead camera monitoring the behaviors. The test included two 2.5-min sessions. In the first session, a mesh cup was located in the middle of one side of the box, tested mouse was allowed to freely explore the whole box. A 8-cm wide corridor surrounding the mesh cup was defined as interaction zone (IZ). Tested mice were put back to the home cage and an unfamiliar CD-1 male mouse was introduced into the mesh cup. After that, the tested mouse was put into the test box again and another 2.5-min activity was recorded. Tracking software (EthoVision; Noldus) was used to analyze the locomotion and the time spent in IZ. Interaction ratio (IR) was calculated by (time in IZ with presence of social target)/(time in IZ with absence of social target) × 100; after CSDS, mice with IR < 100 were grouped as susceptible, while other mice were grouped as resilient.

Sucrose preference test was performed at the home cage of the tested mice which were individually housed. Two bottles containing water and 2% sucrose, respectively were placed on the wire lid. Every 12 h, the two bottles were weighted and their positions were switched. Sucrose preference = (Consumed sucrose)/(Consumed sucrose + Consumed water) × 100%.

Statistical analysis

GraphPad Prism was used for data analysis. Data normality was determined by the D’Agostino-Pearson normality test. Difference between two groups was analyzed by unpaired Student’s t test; difference among three or more groups was analyzed by one-way ANOVA followed by Sidak’s or Tukey’s multiple comparisons test; difference in current injection-induced spikes or total outward K+ currents was analyzed by two-way ANOVA followed by Sidak’s multiple comparisons test (see Figure legends for details). Results for the spontaneous firing, current injection-induced spikes, and K+ currents were expressed as mean ± SEM; other data were expressed as box-and-whisker plots. P value < 0.05 for two-sided tests was considered significant.

Results

ErbB4 in VTA DA neurons is necessary for depression-like behaviors

To investigate the role of ErbB4 in depression-like behaviors, we determined whether its level or activity is changed in the VTA under stress. Mice were exposed to CSDS for 10 days [54] and subjected to social avoidance test (Fig. 1a, b) and sucrose preference test, two well established paradigms for CSDS-induced depressive-like symptoms [7], on day 11 and 12, respectively. Mice were scored for interaction ratio with a CD1 social target and sucrose preference, and were segregated into susceptible and resilient groups (Supplementary Fig. S1). On day 13, the VTA was dissected from freshly prepared brain slices (Fig. 1c) and probed for ErbB4 protein and its phosphorylation by western blotting. As shown in Fig. 1d, e, ErbB4 level was similar among susceptible, resilient and naive mice (without CSDS), suggesting CSDS had little effect on ErbB4 protein level. Interestingly, ErbB4 tyrosine phosphorylation (pErbB4) was increased in susceptible mice, compared with that of naive or resilient mice, indicative of increased ErbB4 activity in the VTA in CSDS-susceptible mice. Furthermore, pErbB4 level in VTA was negatively correlated with both social interaction and sucrose preference (Fig. 1f, g). These results suggest a potential involvement of ErbB4 signaling in depression-like symptoms. ErbB4 in the VTA is expressed in DA neurons [9]. 1NMPP1 increased the amplitudes of K+ currents at the peak and sustained phases. The effect of 1NMPP1 was reversible as it was diminished by medium change (Fig. 2o–q). These results indicated that K+ currents in VTA DA neurons was enhanced by 1NMPP1 and suggested that ErbB4 kinase activity maintains DA neuron activity by suppressing K+ channel activity. Further analysis of transient (IA) and sustained (IDR) components of K+ currents indicated that 1NMPP1 caused left-ward shift of IA activation curve, with little effect on IA inactivation curve or IDR activation curve (Supplementary Fig. S9). These results suggest that enhanced IA activation may be a mechanism of reduced DA neuron activity by 1NMPP1. Increased activity of VTA DA neurons could increase BDNF release from their terminals in the NAc, which is necessary for depression-like symptoms in response to social stress [54, 65, 66]. As previously reported, CSDS increased the BDNF in NAc (Supplementary Fig. S10c, d). Interestingly, BDNF levels in NAc were reduced by 1NMPP1 (100 nM) in the CSDS-susceptible mice (Supplementary Fig. S10c, d). However, 1NMPP1 had no effect on mRNA levels of BDNF in NAc (Supplementary Fig. S10e), in agreement with previous reports that BDNF is released from VTA-originated axon terminals [67,68,69,70]. Altogether, these results suggest that ErbB4 signaling in VTA is necessary for the firing of VTA DA neurons and the stress-induced BDNF release in NAc.

CSDS-induced expression of NRG1 in LDTg neurons for depression-like behaviors

Neuronal activity has been shown to increase NRG1 expression by increasing transcription and activity-dependent release [71,72,73]. CSDS induction of pErbB4 in VTA suggests an increase in NRG1. Indeed, western blot analysis showed an increase in NRG1 protein in CSDS-susceptible mice, compared with naive and resilient mice (Fig. 3a, b). However, this increase was not associated with an elevation in NRG1 mRNA in the VTA (Fig. 3b), suggesting that NRG1 may be delivered to VTA from upstream neurons. VTA DA neurons receive inputs from mPFC, VP, AMY, DRN, and LDTg (Fig. 3c), where the activity of projection neurons regulates social stress-induced depression-like behaviors [12, 74,75,76,77,78,79]. To identify neurons in which regions are critical, we analyzed NRG1 mRNA in these regions. As shown in Fig. 3d, NRG1 mRNA in mPFC, VP, or AMY was similar among naive, susceptible and resilient groups. In contrast, NRG1 mRNA was increased in DRN and LDTg of susceptible mice, compared with naive and resilient mice (Fig. 3d). To verify their contribution to increased NRG1 in VTA, we ablated NRG1 expression in LDTg and DRN by injecting vCre virus into LDTg (Fig. 3e, Supplementary Fig. S11a, b) and DRN (Fig. 3f, Supplementary Fig. S11c, d) of NRG1f/f mice. Three weeks after virus injection, NRG1 mRNA in LDTg and DRN was reduced (Fig. 3g). Remarkably, CSDS-induced increase in NRG1 protein in VTA was diminished by vCre injection in LDTg, compared with vGFP-injected mice (Fig. 3h, i). In accord, CSDS-induced pErbB4 in VTA was reduced by vCre injection into LDTg (Fig. 3k, l). vCre injection into to DRN seemed to have little effect on CSDS-induced increase in NRG1 protein (Fig. 3h, j) and pErbB4 (Fig. 3k, m) in VTA. Note that viral injection had little effect on total level of ErbB4 in VTA (Supplementary Fig. S12a–d). These results suggest that LDTg is required for CSDS-increased NRG1 protein and activation of ErbB4 in VTA.

Fig. 3: Increased expression of NRG1 in LDTg after CSDS and its requirement for the development of depression-like behaviors.
figure 3

a, b Increased NRG1 protein, but not mRNA, in VTA of CSDS-susceptible mice. a Representative western blots for dissected VTA; b Quantitative data of NRG1 protein and mRNA in VTA. Sus susceptible, Res resilient. Data were expressed as box-and-whisker plots. For western blot, one-way ANOVA, F(2, 12) = 6.785, **P = 0.0041; Sidak’s multiple comparisons test, **P(Naive vs. Sus) = 0.0065, *P(Res vs. Sus) = 0.0249; n = 10 (Naive), 9 (Res), or 11 (Sus) mice. c Diagram showing brain regions that send excitatory afferents to VTA DA neurons and are involved in social defeat-induced behavioral changes. mPFC medial prefrontal cortex, VP ventral pallidum, AMY amygdala, DRN dorsal raphe nucleus, VTA ventral tegmental area, LDTg laterodorsal tegmentum. d Increased NRG1 mRNA in LDTg and DRN, not mPFC, VP, or AMY in Sus mice. Data were expressed as box-and-whisker plots. One-way ANOVA. For LDTg, F(2, 27) = 8.185, **P = 0.0017; Tukey’s multiple comparisons test, **P(Naive vs. Sus) = 0.0028, **P(Res vs. Sus) = 0.0076. For DRN, F(2, 27) = 5.125, *P = 0.0130; Tukey’s multiple comparisons test, *P(Naive vs. Sus) = 0.0302, *P(Res vs. Sus) = 0.0227. n = 10 mice per group. e–j Deletion of NRG1 gene in LDTg reduced NRG1 protein in VTA in CSDS-exposed mice. e–g vCre injection into LDTg and DRN of NRG1f/f mice reduced local NRG1 mRNA. e, f vCre expression in LDTg and DRN; Left panels anatomic diagrams, right panels representative images of vCre expression, 2Cb 2nd cerebellar lobule, Aq aqueduct, Pa4 paratrochlear nucleus, mlf medial longitudinal fasciculus, xscp decussation of the superior cerebellar peduncle. Scale bars, 500 µm. g Reduced mRNA in LDTg and DRN detected by qRT-PCR. Data were expressed as box-and-whisker plots. Unpaired Student’s t test: t(6)(LDTg) = 4.112, **P = 0.0063; t(6)(DRN) = 5.056, **P = 0.0023; n = 4 mice per group. h–j Reduced NRG1 protein in VTA after injecting vCre into LDTg, but not DRN. Mice after indicated paradigms were analyzed. h Representative western blots for dissected VTA; i, j Quantitative data in h. Data were expressed as box-and-whisker plots. Unpaired Student’s t test; for LDTg, t(14)(Naive-vGFP vs. CSDS-vGFP) = 2.688, *P = 0.0177; t(14)(Naive-vCre vs. CSDS-vCre) = 1.352, P = 0.1979, n = 8 mice per group; for DRN, t(14)(Naive-vGFP vs. CSDS-vGFP) = 2.725, *P = 0.0164;, t(14)(Naive-vCre vs. CSDS-vCre) = 3.046, **P = 0.0087, n = 8 mice per group. k–m Reduced pErbB4 in VTA after vCre injection into LDTg, but not DRN. Mice after indicated paradigms were analyzed. k Representative western blots for dissected VTA; l, m Quantitative data in k. Data were expressed as box-and-whisker plots. Unpaired Student’s t test; for LDTg, t(14)(Naive-vGFP vs. CSDS-vGFP) = 2.554, *P = 0.0230; t(14)(Naive-vCre vs. CSDS-vCre) = 1.18, P = 0.2577, n = 8 mice per group; for DRN, t(14)(Naive-vGFP vs. CSDS-vGFP) = 2.629, *P = 0.0198; t(14)(Naive-vCre vs. CSDS-vCre) = 2.845, *P = 0.0130, n = 8 mice per group. n–y Requirement of NRG1 in LDTg for the development of depression-like behaviors. n–q Deletion of NRG1 in LDTg attenuated the development of social avoidance and reduction in sucrose preference. n Anatomical diagram showing injection of vCre or vGFP into LDTg of NRG1f/f mice. o Attenuation of CSDS-induced social avoidance in NRG1f/f;vCre mice. Data were expressed as box-and-whisker plots. Unpaired Student’s t test, t(34)(Naive-vGFP vs. CSDS-vGFP) = 5.371, ***P < 0.0001; t(34)(Naive-vCre vs. CSDS-vCre) = 3.409, # #P = 0.0017; t(34)(CSDS-vGFP vs. CSDS-vCre) = 2.718, *P = 0.0103; n = 18 mice per group. p Attenuated development of CSDS-induced reduction of sucrose preference in NRG1f/f;vCre mice. Data were expressed as box-and-whisker plots. Unpaired Student’s t test; t(34)(Naive-vGFP vs. CSDS-vGFP) = 3.08, **P = 0.0041; t(34)(CSDS-vGFP vs. CSDS-vCre) = 2.2, *P = 0.0347; n = 18 mice per group. q No change in locomotion in NRG1f/f;vCre mice. r–t Injection of AAV-NRG1-GFP (vNRG1) into LDTg increased local expression of NRG1. r Anatomic diagram showing injection of vNRG1 or vGFP into LDTg of WT mice; s Representative image showing expression of vNRG1 at LDTg. Aq aqueduct, 2Cb 2nd cerebellar lobule. Scale bar, 1 mm. t Increased NRG1 mRNA detected by qRT-PCR. Data were expressed as box-and-whisker plots. Unpaired Student’s t test, t6 = 4.865, **P = 0.0028. n = 4 mice per group. u–y Enhanced development of depression-like behaviors by overexpressing NRG1 in LDTg after exposure to subSD. u Diagram showing the procedure of subSD. v No alteration in time spent in social target-absent IZ. w Reduction in time spent in social target-present IZ after subSD in vNRG1-injected mice. Data were expressed as box-and-whisker plots. Unpaired Student’s t test; t(34)(Naive-vGFP vs. subSD-vGFP) = 1.686, P = 0.1009; t(34)(Naive-vNRG1 vs. subSD-vNRG1) = 4.088, # # #P = 0.0003; t(34)(subSD-vGFP vs. subSD-vNRG1) = 3.237, **P = 0.0027; n = 18 mice per group. x Reduction in sucrose preference after subSD in vNRG1-injected mice. Data were expressed as box-and-whisker plots. Unpaired Student’s t test, t(34)(Naive-vGFP vs. subSD-vGFP) = 1.565, P = 0.1269; t(34)(Naive-vNRG1 vs. subSD-vNRG1) = 3.553, # #P = 0.0011; t(34)(subSD-vGFP vs. subSD-vNRG1) = 2.591, *P = 0.0140; n = 18 mice per group. y No changes in locomotion in vNRG1-injected mice. (Also see Supplementary Fig. S11 for AAV expression at DRN and LDTg; Supplementary Fig. S12 for the unaltered expression of total ErbB4 in VTA; Supplementary Fig. S12 for increased NRG1 in VTA after subSD in vNRG1-injected mice.).

Full size image

Next, we determined whether LDTg-originated NRG1 is required for CSDS-induced depression-like behaviors by injecting NRG1f/f mice with vCre or vGFP into LDTg (referred to as NRG1f/f;vCre and NRG1f/f;vGFP mice, respectively) (Fig. 3n). Three weeks after viral injection, NRG1f/f;vCre and NRG1f/f;vGFP mice showed no difference in the time spent in social target-absent interaction zone before or after CSDS (Fig. 3o). CSDS;NGR1f/f;vGFP mice spent less time in the social target-present interaction zone and consumed less sucrose; however, this effect was diminished in CSDS;NRG1f/f;vCre mice (Fig. 3o, p and Supplementary Fig. S4b). As control, the viral injections had no effect on locomotion (Fig. 3q). These results indicate that LDTg-derived NRG1 is necessary for CSDS-induced depression-like behaviors. Next, we studied the impact of increasing NRG1 level in LDTg on CSDS-induced depression-like behaviors by injecting AAV-CMV-NRG1-GFP (vNRG1) into LDTg of wild type (WT) mice (Fig. 3r, s). Three weeks after viral injection, qPCR confirmed the increased expression of NRG1 (Fig. 3t). Mice were subjected to subSD paradigm as described previously (Fig. 3u) [5, 7]. Mice after subSD do not develop depression-like behaviors but are more susceptible to subsequent stress challenges [5, 7]. In agreement, mice injected with vGFP spent a similar amount of time in the social target-present interaction zone and consumed a similar amount of sucrose after subSD, compared to the naive mice (Fig. 3v–x). In contrast, vNRG1-injected mice spent less time in social interaction and consumed less sucrose after subSD, compared with the vNRG1-injected naive mice and vGFP-injected subSD mice (Fig. 3w, x, Supplementary Fig. S4c), suggesting that NRG1 overexpression in LDTg promotes the development of depression-like behaviors. After behavior tests, we analyzed NRG1 expression in VTA by western blot, which showed an increase in vNRG1-injected, subSD-treated mice (Supplementary Fig. S12e, f). Together, these results indicate that NRG1 in LDTg is necessary and sufficient to induce depression-like behaviors after social defeat stress and suggest that the LDTg-VTA circuit may regulate VTA DA neuron activity by releasing NRG1.

Our hypothesis predicts that NRG1, released from LDTg terminals, would regulate DA neuron activity. To test this, we recorded ErbB4+ neurons in VTA slices of ErbB4-CreER;Ai9 mice where ErbB4+ neurons were visualized by tdTomato [19]. Recorded neurons were back labeled by biocytin and stained with anti-TH antibody (Fig. 4a–c). We focused on neurons that were double positive for tdTomato and TH (Fig. 4a and Supplementary Fig. S13). As shown in Fig. 4b, c, NRG1 increased the number of APs in response to injected currents, suggesting increased activity of DA neurons. In addition, in WT mice, local injection of NRG1 increased both the firing rate and percentage of spikes in bursts, compared with Veh-treated group (Fig. 4d–g), indicative of increased VTA DA neuron activity by NRG1 in vivo. However, NRG1 had little effect on synaptic inputs onto VTA DA neurons (Supplementary Fig. S8) but inhibited the AHP and decreased the first spike latency (Supplementary Table S1). These results support the hypothesis that NRG1 regulates depression-like behaviors by increasing the firing of VTA DA neurons. In support of this notion, NRG1 was able to suppress K+ currents (such as IA activation) in VTA DA neurons (Fig. 4h–j; Supplementary Fig. S14). In addition, injection of NRG1 into VTA increased BDNF in NAc of subSD-exposed mice (Fig. S10c, d).

Fig. 4: VTA DA neuron firing increased by NRG1, via suppressing K+ currents.
figure 4

a–c In vitro recording performed in VTA slices from ErbB4-CreER;Ai9 mice showing increased VTA DA neuron firing after incubation with NRG1. a Representative images showing recorded neurons triple positive for ErbB4 (positive of tdTomato, referred to as tdT+), TH, and biocytin. Left, VTA slices with anatomic regions. SNc substantia nigra pars compacta, SNr substantia nigra pars reticulata, ml medial lemniscus, fr fasciculus retroflexus. Right, enlargement of the rectangle area in the left image. Arrowheads indicated the tdT, TH, and biocytin triple positive neurons. Scale bars, 200 μm in the left, 50 μm in the right. To avoid overexposure of fluorescent signal, low dose of tamoxifen was used to induced a small number of tdT+ cells. b Representative current injection-induced spikes. c Quantitative data in b. Data were expressed as mean ± SEM. Two-way ANOVA. Baseline vs. NRG1: F(interaction)(8, 288) = 2.081, *P = 0.0376; Sidak’s multiple comparisons test, *P(50 pA) = 0.0416, **P(100 pA) = 0.0085, **P(125 pA) = 0.0020, **P(150 pA) = 0.0012, ***P(175 pA) = 0.0008, ***P(200 pA) = 0.0003; NRG1 vs. Wash, F(interaction)(8, 288) = 1.0198, P = 0.3001; F(main effect of NRG1)(1,36) = 63.9, # # #P < 0.0001; Sidak’s multiple comparisons test, #P(125 pA) = 0.0206, #P(150 pA) = 0.0228, #P(175 pA) = 0.0279, #P(200 pA) = 0.0185; n = 19 cells from 8 mice. d–g In vivo recordings performed in WT mice showing increased firing rate (FR) and percentage of spikes in bursts (SB) after NRG1 injection into VTA. d Representative firing traces of VTA DA neurons shortly (20 min) after injection of Vehicle, or NRG1 into VTA. Arrowheads indicated some of the spikes that fired in bursts. e Representative APs of DA neurons. Arrowheads and Arabic numerals indicated the three phases. f Representative immunostaining image showing recorded neuron double positive of biocytin and TH. Scale bar, 20 µm. g Quantitative data in d, data were expressed as box-and-whisker plots. Unpaired Student’s t test: t(FR)(70) = 2.584, *P = 0.0119; t(SB)(70) = 2.861, **P = 0.0056; n = 36 cells from 12 (Vehicle) or 13 (NRG1) mice. h–j In vitro recordings performed in VTA slices from ErbB4-CreER;Ai9 mice showing inhibition of total K+ currents by NRG1. h Representative traces of total K+ currents in VTA DA neurons after NRG1 incubation. i, j Quantitative data showing inhibition of both the peak and sustained phases by NRG1. Data were expressed as mean ± SEM. Two-way ANOAVA. For the peak phase: Baseline vs. NRG1, F(interaction)(9, 414) = 3.403, ***P = 0.0005; Sidak’s multiple comparisons test, ***P(90 mV) < 0.0001, ***P(80 mV) = 0.0002, **P(70 mV) = 0.0021; NRG1 vs. Wash, F(interaction)(9, 414) = 1.851, P = 0.0577, F(main effect of NRG1)(1, 46) = 17.97, # # #P = 0.0001, Sidak’s multiple comparisons test, # #P(90 mV) = 0.0022, #P(80 mV) = 0.0260. For sustained phase: Baseline vs. NRG1, F(interaction)(9, 414) = 2.402, *P = 0.0116; Sidak’s multiple comparisons test, ***P(90 mV) = 0.0003, **P(80 mV) = 0.0012, **P(70 mV) = 0.0074; *P(60 mV) = 0.0431; NRG1 vs. Wash, F(interaction)(9, 414) = 1.449, P = 0.1649, F(main effect of NRG1)(1, 46) = 23.37, # # #P < 0.0001, Sidak’s multiple comparisons test, #P(70 mV) = 0.0417. n = 24 cells of 10 mice for each group. k–m pERK increased by NRG1, but decreased by NMP, injected into VTA. k Diagram showing acute injection of Veh, NRG1 or NMP into VTA. l Representative western blots for dissected VTA. m Quantification of l. Data were expressed as box-and-whisker plots. Unpaired Student’s t test, t(Veh vs. NRG1)(10) = 3.962, **P = 0.0027; t(Veh vs. 1NMPP1)(10) = 6.696, ***P < 0.001. n–q ERK inhibitor, PD98059 (referred to as PD), reduced the effects of NRG1 on DA neuron firing and K+ currents. n Representative traces of current injection-induced spikes in DA neurons after incubation with Veh+NRG1 or PD + NRG1. o Quantification of n showing reduced spikes in PD + NRG1-treated slices. Data were expressed as mean ± SEM. Two-way ANOVA, F(interaction)(8, 296) = 2.058, *P = 0.0398; Sidak’s multiple comparisons test, **P(100 pA) = 0.0087, **P(125 pA) = 0.0028, **P(150 pA) = 0.0054, **P(175 pA) = 0.0032, **P (200 pA) = 0.0063; n = 20 (Veh) or 19 (PD) cells from 8 mice. p Representative traces of total K+ currents after incubation with Veh+NRG1 or PD + NRG1. q Quantification of p showing increased K+ currents in PD + NRG1-treated slices. Data were expressed as mean ± SEM. Two-way ANOVA. For peak phase: F(interaction)(9, 396) = 2.292, *P = 0.0162; Sidak’s multiple comparisons test, ***P(90 mV) = 0.0002, **P(80 mV) = 0.0011, *P(70 mV) = 0.0154. For sustained phase: F(interaction)(9, 396) = 2.195, #P = 0.0216; Sidak’s multiple comparisons test, # # #P(90 mV) = 0.0003, # #P(80 mV) = 0.0022, #P(70 mV) = 0.0207; n = 22 cells of 8 mice for each group. (Also see Supplementary Fig. S13 for quantification of ErbB4-expressing DA neurons in VTA; Supplementary Fig. S10 for the increased BDNF release at NAc after injection of NRG1 into VTA in subSD-exposed mice.).

Full size image

NRG1 activation of ErbB4 activates the ERK pathway [15, 18]. ERK signaling in VTA has been shown to be required for DA neuron firing and behavioral sensitization to CSDS [80, 81]. Next, we investigated potential participation of the ERK pathway. As shown in Fig. 4k–m, NRG1 increased whereas 1NMPP1 reduced phosphorylation of ERK (pERK) in the VTA. Importantly, PD98059, an inhibitor of ERK activation, attenuated the effects of NRG1 on current injection-induced spikes (Fig. 4n, o) and K+ currents (Fig. 4p, q) of DA neurons. Together, these findings reveal that the NRG1-ErbB4 signaling of the LDTg-VTA circuit is critical to the activity of DA neurons via ERK-dependent inhibition of K+ currents.

Amelioration of depression-like behaviors by acute inhibition of VTA ErbB4

The above results demonstrate a necessary role of NRG1 and ErbB4 in CSDS-induced depression-like behaviors. To determine whether ErbB4 kinase activity is required for the expression of depression-like behaviors, we first injected afatinib and lapatinib, inhibitors of ErbB kinases, into bilateral VTAs (Supplementary Fig. S15a). Afatinib has IC50 values of 0.5 nM for ErbB1, 1 nM for ErbB4, and 14 nM for ErbB2 whereas lapatinib is more specific for ErbB1 and ErbB2 (IC50 values being 9.2 and 10.8 nM, respectively), but not ErbB4 (IC50 of 367 nM) [82,83,84]. In accord, pErbB4 was reduced in the VTA by afatinib (10 nM, 200 nl), but not by lapatinib (100 nM, 200 nl), and the inhibitory effect of afatinib occurred within 30 min, persisted 24 h, but recovered 48 h, after injection (Fig. S15b, c). To determine their effects on behavior expression, WT mice were subjected to CSDS, and susceptible mice were injected with the two inhibitors 30 min before social avoidance test, followed by sucrose preference test (Supplementary Fig. S15d). As shown in Supplementary Fig. S15e, without a social target, the time spent in the interaction zone by different groups of mice was similar. In contrast, compared with Veh- or lapatinib-treated mice, time in interaction zone with social target was increased in afatinib-injected mice, 30 min and 24 h, but not 48 h, after injection (Supplementary Fig. S15f). Note that afatinib or lapatinib had little effect on locomotor behavior (Supplementary Fig. S15g, h). These results suggest a necessary role of VTA ErbB4 kinase activity in expression of CSDS-induced social avoidance. Moreover, afatinib-injected mice increased sucrose preference 24 h after injection, compared with Veh- or lapatinib-injected mice; and again, this effect disappeared 48 h after injection (Supplementary Fig. S15i). We also compared the social avoidance and sucrose preference before and after afatinib injection, results of which indicate that both parameters are improved (Supplementary Fig. S15f, i). Considering the key role of increased DA neuron activity in the depression-like behaviors, we recorded the current injection-induced firing of DA neurons after incubation with afatinib or lapatinib (Supplementary Fig. S15j–l). The results showed attenuation of DA neuron hyperactivity after incubation with afatinib, but not lapatinib (Supplementary Fig. S15k, l), in agreement with the selective effect of afatinib in behaviors. Together, these results suggest a necessary role of VTA ErbB4 kinase activity in CSDS-induced depression-like behaviors.

Although afatinib is able to inhibit ErbB4, it has lower EC50 values for ErbB1 [82], and thus may have off-target effects. To convincingly demonstrate that ErbB4 kinase activity is critical to the expression of CSDS-induced depression-like behaviors, we studied T796G mice. T796G mice were subjected to CSDS, and susceptible mice were injected with 1NMPP1 into bilateral VTAs via cannulae (Fig. 5a). pErbB4 in VTA was reduced by 1NMPP1 (500 nM, 200 nl) within 30 min and recovered afterwards (Fig. 5b, c) [27]. Therefore, social avoidance in CSDS-susceptible mice was tested at 30 min after injection (Fig. 5d). As shown in Fig. 5e and Supplementary Fig. S4d, the time spent in the interaction zone with social target was increased in 1NMPP1-injected T796G mice, compared with Veh-treated mice. 1NMPP1 injection seemed to have little effect on locomotor activity (Fig. 5g). These results suggest a critical role of ErbB4 kinase activity in the expression of depression-like behavior. In agreement, the attenuating effect of 1NMPP1 was not observed 24 h after injection when ErbB4 kinase activity was already recovered (Fig. 5e). Note that sucrose preference was similar between 1NMPP1-injected and Veh-injected T796G mice (Fig. 5f), perhaps because the inhibition of ErbB4 by 1NMPP1 was transient and sucrose preference was a sum of 24 h. We have also investigated the effect of 1NMPP1 on VTA slices from CSDS-susceptible mice. As shown in Fig. 5h, i, the AP spikes in response to injected currents of VTA DA neurons were reduced by 1NMPP1. Together, these results demonstrate a critical role of VTA ErbB4 kinase activity in the expression of CSDS-induced social avoidance.

Fig. 5: Attenuated expression of depression-like behaviors after acute injection of 1NMPP1 into VTA of CSDS-susceptible mice.
figure 5

a Acute injection of NMP or Veh into VTA via implanted cannulae. b, c Acute inhibition of pErbB4 in VTA by single injection of NMP. b Representative western blots. c Quantitative data in b. Data were expressed as box-and-whisker plots. One-way ANOVA, F(6, 28) = 18.87, ***P < 0.0001, Sidak’s multiple comparisons test, ***P(Veh-30 min vs. NMP-30 min) < 0.0001; n = 5 mice for each time point. d Time scheme of behavioral studies. SI social interaction test, SP sucrose preference test, Sus susceptible. e Attenuation of CSDS-induced social avoidance by single administration of NMP into VTA. No difference was observed for time spent in social target-absent IZ. Data were expressed as box-and-whisker plots. At 30 min, unpaired Student’s t test, t38 = 3.488, **P = 0.0012. For comparison between different time points of NMP group, one-way ANOVA, F(2, 57) = 11.03, ***P < 0.0001, ***P(Pre vs. 30 min) = 0.0008. n = 20 mice per group. f No alteration by single injection of NMP in CSDS-induced reduction of sucrose preference (tested during two consecutive 24-h periods). g No alteration in locomotion. h, i Inhibition of DA neuron firing in VTA slices prepared from CSDS-susceptible mice by incubation with NMP. h Representative traces of current injection-induced spikes. i Quantification of h. Data were expressed as mean ± SEM. Two-way ANOVA, F(interaction)(8, 280) = 1.996, *P = 0.0470; Sidak’s multiple comparisons test, **P(100 pA) = 0.0068, **P(125 pA) = 0.0020, **P(150 pA) = 0.0040, *P(175 pA) = 0.0140, *P(200 pA) = 0.0191; n = 19 (Veh) or 18 (NMP) cells from 8 mice. (Also see Supplementary Fig. S15 for attenuation of depression-like behaviors by single administration of afatinib into VTA.).

Full size image

Discussion

Our study provides evidence that the NRG1-ErbB4 signaling in the LDTg-VTA circuit is necessary for social stress-induced depression-like behaviors. First, pErbB4 was increased in VTA in susceptible mice after CSDS. Accordingly, DA neuron- or VTA-specific deletion of ErbB4 caused a pro-resilient effect in the development of depression-like behaviors. Similar effects were observed by VTA-specific inhibition of ErbB4 kinase activity. These results suggest that the development of depression-like behaviors requires ErbB4 and its activity in VTA DA neurons. Second, CSDS increased the expression of NRG1 at LDTg; deleting NRG1 in LDTg reduced ErBB4 activation in VTA and CSDS-induced depression-like behaviors. Moreover, virus-mediated expression of NRG1 in LDTg enhanced subSD-induced depression behaviors, along with increased NRG1 protein in VTA. These results identify a novel function of the LDTg neurons that project to VTA DA neurons. Third, mechanistically, NRG1 suppressed K+ currents and thus increased the activity of VTA DA neurons, whereas inhibiting ErbB4 increased K+ currents and reduced DA neuron firing. Fourth, in accord, acute inhibition of ErbB4 in VTA attenuated the depression-like behaviors in CSDS-susceptible mice. These results demonstrate that the NRG1-ErbB4 signaling at the LDTg-VTA circuit promotes the response of VTA DA neurons to social stress and increases the susceptibility to depression-like phenotypes, revealing a novel mechanism that regulates the development of depression.

Abnormality of VTA DA neuron activity has been implicated in depression. The firing of DA neurons in the VTA is increased in rodents after CSDS and chronic restraint [7, 85, 86], and reduced in chronic mild stress model [8, 87]. Changes in DA neuron activity alter the development and expression of depression-like behaviors [5, 8]. The activity of VTA DA neurons is regulated by inputs from different brain regions. PV neurons of the VP region are increased in activity in susceptible mice after CSDS; optogenetic or chemogenetic inhibition of these PV neurons reduces CSDS-induced social avoidance [12]. On the other hand, the activity of NE neurons in the LC region is increased in resilient mice after CSDS; NE antagonism or depletion blocks the resilience [13, 14]. The burst firing of VTA DA neurons is implicated in encoding rewarding and aversive responses and regulated by neuronal activity in the LDTg [88]. Optogenetic activation of LDTg neurons projecting to VTA promotes reward-related behaviors [89,90,91]. Activity of LDTg neurons is increased by inescapable footshock, predator odorant, and CSDS [88, 92, 93]. In particular, chemogenetic inhibition or activation of LDTg cholinergic neurons suppresses or enhances, respectively, depression-like behaviors and associated DA neuron activity in the VTA [77]. We showed a novel mechanism for LDTg neurons to regulate the activity of VTA DA neurons. NRG1 protein, as well as pErbB4, was increased in the VTA in mice susceptible to CSDS (Figs. 3a, b and 1d, e). However, NRG1 mRNA was not increased in the VTA (Fig. 3b), but in LDTg and DRN regions (Fig. 3d). Specific ablation of the NRG1 gene in LDTg, but not DRN, reduced NRG1 protein and pErbB4 in VTA (Fig. 3e–m) and inhibited depression-like behaviors (Fig. 3n–q). On the other hand, overexpressing NRG1 in LDTg increased NRG1 protein in the VTA (Supplementary Fig. S12e, f) and enhanced depression-like behaviors after subSD induction (Fig. 3r–y). These observations suggest that in response to stress, NRG1 is utilized as a trophic factor for LDTg neurons to regulate the activity of VTA DA neurons. This notion is consistent with earlier findings that the transcription of the NRG1 gene and expression of its protein are dependent on neuronal activity [71,72,73].

K+ currents in VTA DA neurons are increased in resilient mice, which results in an attenuated activity of DA neurons and reduced depression-like behaviors in response to social stress [7, 9, 64]. In susceptible mice, DA neurons seemed to have a problem with increasing K+ currents [9]; increasing K+ currents and subsequent reduction in DA neuron activity in susceptible mice were shown to cause a pro-resilient effect [7, 64]. Nevertheless, regulatory mechanisms of K+ currents of VTA DA neurons in response to stress are not understood. Interestingly, NRG1 in LDTg and pErbB4 in VTA were increased in susceptible mice (Figs. 1 and 3). NRG1 and ErbB4 inhibition were able to suppress and increase, respectively, K+ currents in VTA DA neurons (Figs. 2o–q and 4h–j; Supplementary Figs. S9 and S14) and associated changes in neuronal excitability (increased by NRG1 and decreased by ErbB4 inhibition) (Figs. 2i–n and 4a–g). Notably, perturbing the NRG1-ErbB4 signaling attenuated the development of depression-like behaviors (Figs. 1–3, 5 and Supplementary Fig. S15). ERK activity in the VTA has been implicated in regulating responses to social stress [80, 81]; ERK blockade reduces the firing frequency of VTA DA neurons [80, 94] and increases IA K+ currents [95, 96]. We showed that the administration of NRG1 or 1NMPP1 into the VTA increased or decreased, respectively, ERK phosphorylation (Fig. 4k–m); pharmacological inhibition of ERK reduced the effect of NRG1 on DA neuron firing and K+ currents (Fig. 4n–q). A parsimonious interpretation of these results is that the NRG1-ErbB4 signaling promotes VTA DA neuron activity via an ERK-dependent inhibition of K+ channels.

In short, we demonstrate a role of the NRG1-ErbB4 signaling in the development and expression of depression-like behaviors in response to social stress, and reveal an unexpected mechanism by which the LDTg-VTA circuit regulates VTA DA neuron activity. NRG1 and ERBB4 were identified as risk genes for schizophrenia [15, 18, 97,98,99], although the association was not observed by a recent GWAS of large samples [100]. However, both NRG1 and ERBB4 were among the 269 putative genes identified by GWAS of 246,363 depression patients [36]. Responses to classical antidepressants were predicted by GWAS-identified genetic variants of NRG1 and ERBB4 [37,38,39]. Furthermore, ketamine, an antidepressant, was found to alter the levels of NRG1 or ErbB4 in the cortex and hippocampus [101,102,103]; exogenous NRG1 blocked the reactivation effect of ketamine on cortical plasticity [102], a potential mechanism for the antidepressant effect of ketamine [104]. Therefore, our studies provide insight into pathological mechanisms of relevant types of depression.