Increased regional activity of a pro-autophagy pathway in schizophrenia as a contributor to sex differences in the disease pathology

Abstract

Based on recent genome-wide association studies, it is theorized that altered regulation of autophagy contributes to the pathophysiology of schizophrenia and bipolar disorder. As activity of autophagy-regulatory pathways is controlled by discrete phosphorylation sites on the relevant proteins, phospho-protein profiling is one of the few approaches available for enabling a quantitative assessment of autophagic activity in the brain. Despite this, a comprehensive phospho-protein assessment in the brains of schizophrenia and bipolar disorder subjects is currently lacking. Using this direction, our broad screening identifies an increase in AMP-activated protein kinase (AMPK)-mediated phospho-activation of the pro-autophagy protein beclin-1 solely in the prefrontal cortex of female, but not male, schizophrenia subjects. Using a reverse translational approach, we surprisingly find that this increase in beclin-1 activity facilitates synapse formation and enhances cognition. These findings are interpreted in the context of human studies demonstrating that female schizophrenia subjects have a lower susceptibility to cognitive dysfunction than males.

Graphical abstract

Figure 1

Activity of AMPK and its regulatory targets in the DLPFC of female schizophrenia and female bipolar disorder subjects (A) Blots show phospho and total levels of AMPK, beclin-1, and ULK1 in female DLPFC homogenates. Gender (M = male, F = female), diagnosis (C = control, S = schizophrenia, B = bipolar disorder), and absence (−) or presence (+) of psychosis indicated. Molecular weight makers are indicated (kDa). (B) 65 kDa (upper band) P-AMPK T172 in DLPFC homogenates. ∗p = 0.043 compared to control with Bonferroni correction. (C) 60 kDa (lower band) P-AMPK T172 in DLPFC homogenates. (D) 65 kDa (upper band) total AMPK in DLPFC homogenates. (E) 60 kDa (lower band) total AMPK in DLPFC homogenates. (F) P-beclin-1 S93 in DLPFC homogenates. ∗p = 0.031 compared to control with Bonferroni correction. t = p value of 0.01 compared to control with direct comparison, p value of 0.06 with Bonferroni correction. (G) P-beclin-1 S15 in DLPFC homogenates. (H) Total beclin-1 in DLPFC homogenates. (I) 140 kDa (lower band) P-ULK1 S317 in DLPFC homogenates. (J) 140 kDa (lower band) P-ULK1 S467 in DLPFC homogenates. (K) 140 kDa (lower band) P-ULK1 S555 in DLPFC homogenates. (L) 140 kDa (lower band) P-ULK1 S638 in DLPFC homogenates. (M) 140 kDa (lower band) total ULK1 in DLPFC homogenates. Summary data are the mean ± SEM. n = 9 control, 9 schizophrenia, 18 bipolar disorder (5 no psychosis, 12 with psychosis, 1 no psychosis determination). See Table S2 for statistical details for each protein. See also Figures S1–S8.

Figure 2

Activity of AMPK and its regulatory targets in the DLPFC of male schizophrenia and male bipolar disorder subjects (A) Blots show phospho and total levels of AMPK, beclin-1, and ULK1 in male DLPFC homogenates. Gender (M = male, F = female), diagnosis (C = control, S = schizophrenia, B = bipolar disorder), and absence (−) or presence (+) of psychosis indicated. Molecular weight makers are indicated (kDa). (B–E) No differences in P-AMPK T172 or total AMPK were detected between any groups and controls. (F–H) No differences in P-beclin-1 S93, P-beclin-1 S15, or total beclin-1 were detected between any groups and controls. (I–M) No differences in P-ULK1 S317, P-ULK1 S467, P-ULK1 S555, P-ULK1 S638, or total ULK1 were detected between any groups and controls. Summary data are the mean ± SEM. n = 24 control, 25 schizophrenia, 16 bipolar disorder (7 no psychosis, 8 with psychosis, 1 no psychosis determination) See Table S2 for statistical details for each protein. See also Figures S1, S7, and S9.

Figure 3

S91/94D in the mPFC increases the biochemical signatures of autophagy in females (A) Schematic of experimental design in female mice. (B) LC3B-I and LC3B-II in the mPFC of three representative female mice (molecular weight in kDa). Lower blot is a longer exposure to reveal the weaker LC3B-II bands. (C) No change in LC3B-I levels in the S91/94D-GFP hemisphere vs. the GFP hemisphere. t(14) = 0.8169, p = 0.4277 via paired t test. n = 15 mice. (D) Per mouse, LCB-I levels in the HSV-S91/94D-GFP mPFC hemisphere were divided by that of the HSV-GFP hemisphere. S91/94D-GFP does not alter LCB-I levels above a chance level ratio of 1.0. t(14) = 1.263, p = 0.2273 with one-sample t test. n = 15 mice. (E) A significant increase in LC3B-II levels in the S91/94D-GFP hemisphere vs. the GFP hemisphere. t(14) = 2.263, ∗p = 0.0401 via paired t test. n = 15 mice. (F) Per mouse, LCB-II levels in the HSV-S91/94D-GFP hemisphere were divided by that of the HSV-GFP hemisphere. S91/94D-GFP increases LCB-II levels above a chance level ratio of 1.0. ∗∗p = 0.01 with one-sample t test. n = 15 mice. (G) Schematic of experimental design in male mice. (H) LC3B-I and LC3B-II in the mPFC of three representative male mice (molecular weight in kDa). Lower blot is a longer exposure to reveal the weaker LC3B-II bands. (I) No change in LC3B-I levels in the S91/94D-GFP hemisphere vs. the GFP hemisphere. t(11) = 0.9527, p = 0.3612 via paired t test. n = 11 mice. (J) Per mouse, LCB-I levels in the HSV-S91/94D-GFP hemisphere were divided by that of the HSV-GFP hemisphere. S91/94D-GFP does not alter LCB-I levels above a chance level ratio of 1.0. t(11) = 0.2992, p = 0.7704 with one-sample t test. n = 11 mice. (K) No change in LC3B-II levels in the S91/94D-GFP hemisphere vs. the GFP hemisphere. t(11) = 0.3093, p = 0.7629 via paired t test. n = 11 mice. (L) Per mouse, LCB-II levels in the HSV-S91/94D-GFP hemisphere were divided by that of the HSV-GFP hemisphere. S91/94D-GFP does not alter LCB-II levels above a chance level ratio of 1.0. p = 0.2661 with one-sample t test. n = 11 mice. Summary data are the mean ± SEM. See also Figure S10.

Figure 4

Enrichment of beclin-1 in postsynaptic regions (A) Blots show indicated proteins in mPFC tissue that underwent subcellular fractionation. (B) Beclin-1 immunoprecipitated (IP) from young adult mouse mPFC tissue homogenates and the pull-down immunoblotted (IB) for PSD-95. PSD-95 was recovered from the beclin-1 immunoprecipitation. Input shows total amounts of beclin and PSD-95. (C) Beclin-1 phosphorylated at S93 immunoprecipitated (IP) from young adult mouse mPFC tissue homogenates and the pull-down immunoblotted (IB) for PSD-95. PSD-95 was recovered from the P-beclin-1 S93 immunoprecipitation. (D–G) SIM images of endogenous total beclin-1 and phospho-beclin-1 S93 in dendritic spines. GFP signal is shown for some images. Orange endogenous nanodomains of total and phospho-beclin-1 are present in many spines. White spine traces are the outline of the spine perimeter as determined from the GFP signal. (H) Quantification shows the percentage of dendritic spines that contain at least one beclin-1 and phospho-beclin-1 S93 nanodomain. See also Figure S10.

Figure 5

S91/94D in the mPFC increases dendritic spine density in female, but not male, mice (A) Experimental design in female mice. (B) Dendrite segments from female layer 2/3 mPFC pyramidal neurons. Scale bar, 10 μm. (C) S91/94D-GFP increases total spine density in females, ∗∗∗ (t(41) = 4.390, p < 0.0001; Mann-Whitney U = 82, p = 0.0003). n = 18 neurons from 4 GFP mice (4–5 neurons/mouse) and 25 neurons from 5 S91/94D-GFP mice (4–6 neurons/mouse). (D) S91/94D-GFP increases thin spine density in females, ∗∗(t(41) = 3.501, p = 0.0011; Mann-Whitney U = 102, p = 0.0020). n is same as in (C). (E and F) S91/94D-GFP in females does not affect the density of stubby spines (t(41) = 0.5966, p = 0.5541; Mann-Whitney U = 168, p = 0.1659) or mushroom spines in females (t(41) = 1.734, p = 0.0904; Mann-Whitney U = 154, p = 0.0826). n is same as in (C). (G–I) No differences in the cumulative head diameter curve for thin, stubby, or mushroom spines in females. Thin spines: χ2 = 0.2576, p = 0.6118, and n = 1,200 GFP and 2,012 S91/94D. Stubby spines: χ2 = 0.02214, p = 0.8817, and n = 210 GFP and 271 S91/94D. Mushroom spines: χ2 = 0.03674, p = 0.8480, and n = 364 GFP and 455 S91/94D. (J) Experimental design in male mice. (K) Dendrite segments from male layer 2/3 mPFC pyramidal neurons. Scale bar, 10 μm. (L) S91/94D-GFP does not affect total spine density in males, (t(30) = 0.1584, p = 0.8752; Mann-Whitney U = 122, p = 0.9698). n = 13 neurons from 3 GFP mice (4–5 neurons/mouse) and 19 neurons from 4 S91/94D-GFP mice (4–5 neurons/mouse). (M–O) S91/94D-GFP in males does not affect the density of thin spines (t(30) = 0.08819, p = 0.9303; Mann-Whitney U = 119, p = 0.8798), stubby spines (t(30) = 0.3472, p = 0.7309; Mann-Whitney U = 116, p = 0.7913), or mushroom spines (t(30) = 0.2098, p = 0.8353; Mann-Whitney U = 117, p = 0.8206). n is same as in (L). (P–R) No differences in the cumulative head diameter curve for thin, stubby, or mushroom spines in males. Thin spines: χ2 = 3.712, p = 0.0540, and n = 1,726 GFP and 2,125 S91/94D. Stubby spines: χ2 = 0.2117, p = 0.6454, and n = 210 GFP and 259 S91/94D. Mushroom spines: χ2 = 0.01055, p = 0.9182, and n = 164 GFP and 217 S91/94D. Unless otherwise indicated, summary data are the mean ± SEM.

Figure 6

S91/94D in the mPFC enhances recognition memory in female mice (A) Schematic of object-in-place experimental design in females. During trial 2, mice should prefer to investigate the objects that swapped location between trials (purple star and green triangle). (B) Image shows targeting on HSV-S91/94D-GFP to the mPFC. Scale bar, 500 μm. (C) Graph depicts time exploring objects throughout trial 1. No differences between groups (t(24) = 0.3449, p = 0.7331). n = 13 GFP and 13 S91/94D. (D) Graph depicts the trial 2 object-in-place recognition ratio. Differences between groups were identified at the 0–3 min time bin. 0–3 min bin, Bonferroni ∗p = 0.0433; 3–4 min bin, Bonferroni p > 0.999; 4–5 min bin, Bonferroni p = 0.7894; 5–6 min time bin, Bonferroni p > 0.999; 6–7 min time bin, Bonferroni p > 0.999. n is the same as in (C). (E) Graph depicts an increased recognition ratio across all time bins of trial 2 in the S91/94D-GFP group vs. the GFP group (t(24) = 2.151, ∗p = 0.0418). n is same as in (C). (F) Schematic of Y-maze experimental design in females. (G) Graph depicts the cumulative spontaneous alternation percentage per time bin. Direct comparisons revealed a significant increase in alternation percentage in the S91/94D group at the indicated bins (3 min, p = 0.0370; 5 min, p = 0.0652; 7 min, p = 0.0348; 10 min, p = 0.0126). n = 20 GFP and 22 S91/94D-GFP mice. Summary data are the mean ± SEM.

Figure 7

S91/94D in the mPFC does not affect recognition memory in male mice (A) Schematic of object-in-place experimental design in males. (B) No differences between groups were observed for time spent exploring objects during trial 1 (t(21) = 0.1973, p = 0.8455). n = 13 GFP and 10 S91/94D. (C) No differences in the trial 2 recognition ratio were identified between groups at any time bins; Bonferroni p > 0.999 for 0–3, 3–4, and 4–5 min time bins; p = 0.4042 for 5–6 min time bin; and p = 0.5743 for 6–7 min time bin. n is the same as in (B). (D) No differences in the recognition ratio for trial 2 across all time bins between groups were identified (t(21) = 0.5552, p = 0.5847). n is same as in (B). Summary data are the mean ± SEM. See also Figures S11 and S12.