LSD1 modulates stress-evoked transcription of immediate early genes and emotional behavior

Abstract

Behavioral changes in response to stressful stimuli can be controlled via adaptive epigenetic changes in neuronal gene expression. Here we indicate a role for the transcriptional corepressor Lysine-Specific Demethylase 1 (LSD1) and its dominant-negative splicing isoform neuroLSD1, in the modulation of emotional behavior. In mouse hippocampus, we show that LSD1 and neuroLSD1 can interact with transcription factor serum response factor (SRF) and set the chromatin state of SRF-targeted genes early growth response 1 (egr1) and c-fos Deletion or reduction of neuro LSD1 in mutant mice translates into decreased levels of activating histone marks at egr1 and c-fos promoters, dampening their psychosocial stress-induced transcription and resulting in low anxiety-like behavior. Administration of suberoylanilide hydroxamine to neuroLSD1(KO)mice reactivates egr1 and c-fos transcription and restores the behavioral phenotype. These findings indicate that LSD1 is a molecular transducer of stressful stimuli as well as a stress-response modifier. Indeed, LSD1 expression itself is increased acutely at both the transcriptional and splicing levels by psychosocial stress, suggesting that LSD1 is involved in the adaptive response to stress.

Keywords: LSD1; SRF; epigenetics; immediate early genes; stress.

Fig. 1.

NeuroLSD1-mutant mice feature a low anxiety-like phenotype. Anxiety-like behavior is evaluated by comparing neuroLSD1^KO and neuroLSD1^HET mice with wild-type littermates. The anxiety trend is shown on the y axis. (A) Time spent in the open arm time (F_2,27 = 4.65, P = 0.018), entries into the open arm (F_2,27 = 6.16, P = 0.0063), and total entries (which did not show any difference among genotypes) evaluated in the EPM. (B) Number of marbles buried (F_2,27 = 10.06, P = 0.0002) and latency to the first burial (F_2,27 = 2.21, P = 0.12) evaluated in the MBT. (C) Latency to feeding (F_2,27 = 9.95, P = 0.0003) and total amount of food intake evaluated during the NSF test (n = 8–10 mice per genotype). Data are presented as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, one-way ANOVA, Tukey post hoc test.

Fig. S1.

(A) Schematic representation of the mouse LSD1 gene. (B) The strategy used to remove exon E8a, generating a mutated allele including a neomycin gene cassette. The mutated allele without neomycin is the result of a cre-driven recombination event. The two genotyping primers in the mutated allele without neomycin are indicated by arrows and referred to as genotyping primers in SI Methods. (Adapted from ref. .)

Fig. S2.

Motor activity does not change among the genotypes. Time course of locomotor activity evaluated for 120 min in terms of number of counts in wild-type, heterozygous, and LSD1-mutant mice (KO). (A, Left) Horizontal counts. (Right) Corresponding calculated area under the curve (AUC). (B, Left) Vertical counts. (Right) Corresponding calculated area under the curve. n = 10 for each group. Results are shown as mean ± SEM.

Fig. 2.

Hippocampal LSD1 and neuroLSD1 levels are modified in response to psychosocial stress and modulate stress-induced transcription of IEGs. Mice underwent a single session of the SDS test. A schematic representation of the experimental procedure is shown at the top of the figure. (A–C) Total LSD1 protein by Western blot (A), LSD1 total mRNA by qPCR (B), and LSD1/neuroLSD1 relative percentage by rqfRT-PCR (C) in the hippocampi of C57BL/6N wild-type mice. A schematic representation of the four LSD1 splicing isoforms is shown on the right (n = 5 mice per group). *P < 0.05, **P < 0.01, ***P < 0.001, Student t test. (D–G) SDS-induced transcription and protein expression of egr1 and c-fos in in the hippocampi of wild-type and neuroLSD1^HET mice challenged with SDS. (D and E) mRNA analysis of egr1 (treatment: F_1,44 = 3.561, P = 0.0657; genotype: F_1,44 = 6.190, P = 0.0167; treatment × genotype: F_1,44 = 1.339, P = 0.2535) (D) and c-fos (treatment: F_1,44 = 8.865, P = 0.0047; genotype: F_1,44 = 4.586, P = 0.0378; treatment × genotype: F_1,44 = 2.237, P = 0.1419) (E) (n = 10–14 mice per condition). (F and G) Western blot protein analysis of Egr1 (treatment: F_1,13 = 6.542, P = 0.0238; genotype: F_1,13 = 6.612, P = 0.0232; treatment × genotype: F_1,13 = 26,18, P = 0.0002) (F) and C-Fos (treatment: F_1,13 = 8.900, P = 0.0175; genotype: F_1,13 = 8.900, P = 0.0175; treatment × genotype: F_1,13 = 23.69, P = 0.0012) (G) (n = 3 or 4 mice per condition). Results are shown as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA, Bonferroni post hoc test.

Fig. 3.

LSD1, CoREST, and HDAC2 are previously unidentified SRF corepressors. (A and B) Hippocampus protein extracts immunoprecipitated with anti-SRF antibody (A) or with anti-LSD1 (B) were separated by SDS/PAGE and detected with the indicated antibodies. In addition to the 67-KDa band in A and B, the SRF antibody detected a 57-KDa band described as a neurospecific splicing isoform (28) (also see Fig. S3). (C) HeLa cells overexpressing HA-LSD1 or HA-neuroLSD1 immunoprecipitated with anti-SRF antibody were separated by SDS/PAGE and were immunodecorated with anti-HA. (D–F) qChIP experiments performed on hippocampal chromatin using anti-SRF, anti-panLSD1, and anti-HDAC2 antibodies together with IgG as mock treatment (dashed lines represent the highest mock treatment of the two genotypes) on egr1 (D) and c-fos (E) promoters and an unrelated control (the egr1 distal genomic region) (F). Data are shown as mean ± SEM; *P = 0.035, Student t test. (G and H) In neurons, LSD1/neuroLSD1 regulation of IEGs’ activity-dependent transcription requires SRF binding to DNA. (G) Reporter assay in primary wild-type and neuroLSD1^HET hippocampal neurons transfected with the pGL3-egr1(−370) construct (treatment: F_1,28 = 34.74, P < 0.0001; genotype: F_1,28 = 2.459, P = 0.1281; treatment × genotype: F_1,28 = 5.384, P = 0.0278) and the mutated version lacking the five SREs, pGL3-egr1(−370m) (treatment × genotype: F_1,20 = 0.2209, P = 0.6435). (H) Reporter assay in primary wild-type and neuroLSD1^HET hippocampal neurons transfected with the pGL3-egr1(−370) construct and the pCGN-HA vector (treatment: F_1,11 = 207.2, P < 0.0001; genotype: F_1,11 = 5.319, P = 0.0416; treatment × genotype: F_1,11 = 7.116, P = 0.0219) or pCGN-HA-neuroLSD1 (treatment × genotype: F_1,11 = 0.6388, P = 0.4411). Reporter activity was assayed in basal conditions and after treatment with BDNF. (n = 3–8 per condition). Results are shown as mean ± SEM; *P < 0.01, two-way ANOVA, Bonferroni post hoc test.

Fig. S3.

Western blot analysis showing that the additional 57-kDa SRF-related band is present only in hippocampal protein extracts and not in a nonneuronal cell line (HeLa). Total protein lysates from HeLa cells (lane1) and mouse hippocampus (lane2) were separated on SDS/PAGE and immunodecorated with anti-SRF antibody and anti-GAPDH antibody.

Fig. S4.

LSD1 does not interact with CREB. Mouse hippocampus protein extracts were immunoprecipitated with anti-CREB or preimmune IgG, separated on SDS/PAGE, and immunodecorated with anti-panLSD1 antibody and anti-CREB antibody.

Fig. S5.

SDS does not affect SRF association with target promoter regions. Histograms show SRF enrichment in wild-type mice and neuroLSD1^HET mutants in resting conditions (CTRL) and upon SDS at the level of the egr1 and c-fos proximal promoter regions including the SRE. As a negative control we also measured SRF enrichment at the level of an egr1 distal promoter region lacking SRE (unrelated region). (n = 4 mice per condition). Results are shown as mean ± SEM.

Fig. 4.

NeuroLSD1^KO and neuroLSD1^HET hippocampi display reduced levels of H3K4 dimethylation and H3K9/K14 acetylation at the egr1 and c-fos promoters. Hippocampal chromatin from neuroLSD1^KO or neuroLSD1^HET mice or wild-type littermates was immunoprecipitated with anti-H3K4me2 and anti-H3K9/K14ac antibodies. Enrichments are shown as percentage of input. Shown is qPCR analysis at the level of egr1 (anti-H3K4me2: F_2,23 = 14.07, P = 0.0001; anti-H3K9/14ac: F_2,15 = 1.667, P = 0.2220) (A) or c-fos (anti-H3K4me2: F_2,28 = 19.74, P < 0.0001; anti-H3K9/14ac: F_2,22 = 4.379, P = 0.0251) (B) proximal promoters and the unrelated control (egr1 distal genomic region) (C) (n = 5–13 per condition). Data are shown as mean ± SEM; *P < 0.05; **P < 0.01; ***P < 0.001, one-way ANOVA, Tukey post hoc test; ^§P = 0.0269, Bartlett’s test.

Fig. 5.

Treatment with SAHA normalizes the anxiety-like phenotype of neuroLSD1 mutants. (A–C) Anxiety-like behavior in neuroLSD1^KO mice compared with wild-type mice treated with SAHA or vehicle (VEH). The anxiety trend is shown. (A) Entries into the open arm (treatment: F_2,54 = 1.847, P = 0.1676; genotype: F_1,54 = 19.70, P = 0.001; treatment × genotype: F_2,54 = 3.643, P = 0.032) and time spent in the open arm (treatment: F_2,54 = 5.43, P = 0.007; genotype: F_1,54 = 6.979, P = 0.0108; treatment × genotype: F_2,54 = 6.969, P = 0.002) evaluated in the EPM. (B) The number of buried marbles (treatment: F_2,54 = 4.436, P = 0.0165; genotype: F_1,54 = 22.35, P < 0.0001; treatment × genotype: F_2,54 = 4.566, P = 0.0147), latency to first burial (treatment: F_2,54 = 5.45, P = 0.006; genotype: F_1,54 = 2.285, P = 0.1365; treatment × genotype: F_2,54 = 7.834, P = 0.001) evaluated in the MBT. (C) Latency to the first bite (treatment: F_2,54 = 0.08, P = 0.92; genotype: F_1,54 = 10.29, P = 0.0023; treatment × genotype: F_2,54 = 6.26, P = 0.004) evaluated in the NSF. (n = 8–10 mice per genotype for each condition). (D–F) Hippocampal chromatin from neuroLSD1^KO or wild-type mice treated with SAHA or vehicle and immunoprecipitated by anti-H3K4me2 and anti-H3K9/K14ac antibodies. Enrichment at specific loci is shown as percentage of input. (D) egr1 promoter (anti-H3K4me2, treatment: F_1,16 = 2.141, P = 0.1628; genotype: F_1,16 = 1.4405, P = 0.2468; treatment × genotype: F_1,16 = 9.080, P = 0.0082) (anti-H3K9/K14, treatment: F_1,14 = 0.0429, P = 0.8388; genotype: F_1,14 = 2.376, P = 0.1455; treatment × genotype: F_1,14 = 2.723, P = 0.1212; genotype). (E) c-fos promoter (anti-H3K4me2, treatment: F_1,16 = 11.57, P < 0.0036; genotype: F_1,16 = 3.018, P = 0.1015; treatment × genotype: F_1,16 = 2.938, P = 0.1058) (anti-H3K9/K14, treatment F_1,13 = 3.438, P < 0.0865; genotype: F_1,13 = 0,2726, P = 0.6104; treatment × genotype: F_1,13 = 0.3492, P = 0.5647). (F) Unrelated control (egr1 distal genomic region) (n = 3–6 per condition). Datasets referred to as “VEH” in D–F are different from those in Fig. 4 in which naive mice were used. Data are shown as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, two way ANOVA, Bonferroni post hoc test; ^§P < 0.05, Student t test.

Fig. S6.

Treatment with SAHA (25 mg/kg) or vehicle (VEH) for 10 d does not affect the body weight of wild-type or neuroLSD1^KO mice. Results are shown as mean ± SEM.

Fig. S7.

Treatment with SAHA restores egr1 and c-fos transcription in neuroLSD1 mutants. The experimental procedure is summarized graphically at the top of the figure. egr1 (A) and c-fos (B) mRNA levels in SAHA-treated wild-type and neuroLSD1^HET mice in resting conditions (CTRL) and exposed to SDS. For egr1, treatment: F_1,20 = 0.2393 P = 0.6300; genotype: F_1,20 = 0.4795, P = 0.4966; treatment × genotype interaction: F_1,20 = 0.7077, P = 0.4102. For c-fos, treatment: F_1,19 = 18.35, P = 0.0004; genotype: F_1,19 = 3.721, P = 0.0688; treatment × genotype interaction: F_1,19 = 3.923, P = 0.0623. As reference, egr1 and c-fos mRNA levels of CTRL and SDS mice that did not receive SAHA are reported also (data are from Fig. 2D in the main text).

Fig. S8.

SAHA treatment does not affect SRF association with target promoter regions. Histograms show SRF enrichment at the level of the egr1 and c-fos proximal promoter regions including the SRE in wild-type mice and neuroLSD1^HET mutants administered vehicle and upon chronic systemic SAHA treatment. Results are shown as mean ± SEM.

Fig. 6.

Graphical model of SRF-LSD1/neuroLSD1–mediated transcriptional modulation of the IEGs. Given the dimeric nature of the corepressor complex, it is conceivable that, depending on the relative amounts of LSD1 and neuroLSD1, both LSD1/LSD1 homodimers and LSD1/neuroLSD1 heterodimers could be present in vivo (11). (A) In wild-type mice, in resting conditions, IEG transcription is repressed but permits activity-induced transcription. (B) Upon stressful stimuli, IEG transcription is fully activated. (C and D) In neuroLSD1-mutant mice, in resting conditions (C), a more condensed chromatin structure at the IEG proximal promoters does not permit stress-induced transcription (D). (E and F) Chronic treatment with the HDAC inhibitor SAHA leads to normalization of anxiety in neuroLSD1 mice.

Fig. S9.

SDS-induced modification of H3K4me2 and H3K9/14ac. Hippocampal chromatin from C57BL/6 wild-type mice was immunoprecipitated by anti-H3K4me2 and anti-H3K9/14ac antibodies or IgG (not shown). Enrichments are shown as percentage of the input. qPCR analysis at the level of egr1 or c-fos or an unrelated regulatory region. (n = 8–10 mice per condition). Results are shown as mean ± SEM. *P < 0.01, Student t test. We did not detect a parallel H3K4 hypomethylation in the same time window in which we measured neuroLSD1 stress-induced down-regulation. This lack of immediate modification of the histone marks could be explained by the different kinetics of chromatin-based adaptations: A longer time interval after the cease of stressful conditions might be required to appreciate the effect of LSD1/neuroLSD1 stress on chromatin compaction in the IEG promoter.

Fig. S10.

A working model graphically showing the pathogenic perspective of chronic stress-induced neuroLSD1 down-regulation. (A) In normal conditions, the balance between the two LSD1 isoforms guarantees a chromatin environment that permits experience-induced transcription of plasticity-related genes. (B) Upon exposure to chronic stress, repeated neuroLSD1 down-regulation could translate into a chromatin compaction that hinders activity-evoked egr1 and c-fos transactivation. This molecular adaptation/maladaptation could contribute to the atrophy of apical hippocampal dendrites observed in stress-related psychiatric disorders.