Neuronal seipin knockout facilitates Aβ-induced neuroinflammation and neurotoxicity via reduction of PPARγ in hippocampus of mouse

Background A characteristic phenotype of congenital generalized lipodystrophy 2 (CGL2) that is caused by loss-of-function of seipin gene is mental retardation. Seipin is highly expressed in hippocampal pyramidal cells and astrocytes. Neuronal knockout of seipin in mice (seipin-KO mice) reduces the hippocampal peroxisome proliferator-activated receptor gamma (PPARγ) level without the loss of pyramidal cells. The down-regulation of PPARγ has gained increasing attention in neuroinflammation of Alzheimer’s disease (AD). Thus, the present study focused on exploring the influence of seipin depletion on β-amyloid (Aβ)-induced neuroinflammation and Aβ neurotoxicity. Methods Adult male seipin-KO mice were treated with a single intracerebroventricular (i.c.v.) injection of Aβ25–35 (1.2 nmol/mouse) or Aβ1–42 (0.1 nmol/mouse), generally a non-neurotoxic dose in wild-type (WT) mice. Spatial cognitive behaviors were assessed by Morris water maze and Y-maze tests, and hippocampal CA1 pyramidal cells and inflammatory responses were examined. Results The Aβ25–35/1–42 injection in the seipin-KO mice caused approximately 30–35 % death of pyramidal cells and production of Hoechst-positive cells with the impairment of spatial memory. In comparison with the WT mice, the number of astrocytes and microglia in the seipin-KO mice had no significant difference, whereas the levels of IL-6 and TNF-α were slightly increased. Similarly, the Aβ25–35/1–42 injection in the seipin-KO mice rather than the WT mice could stimulate the activation of astrocytes or microglia and further elevated the levels of IL-6 and TNF-α. Treatment of the seipin-KO mice with the PPARγ agonist rosiglitazone (rosi) could prevent Aβ25–35/1–42-induced neuroinflammation and neurotoxicity, which was blocked by the PPARγ antagonist GW9962. In the seipin-KO mice, the level of glycogen synthase kinase-3β (GSK3β) phosphorylation at Tyr216 was elevated, while at Ser9, it was reduced compared to the WT mice, which were corrected by the rosi treatment but were unaffected by the Aβ25–35 injection. Conclusions Seipin deficiency in astrocytes increases GSK3β activity and levels of IL-6 and TNF-α through reducing PPARγ, which can facilitate Aβ25–35/1–42-induced neuroinflammation to cause the death of neuronal cells and cognitive deficits. Electronic supplementary material The online version of this article (doi:10.1186/s12974-016-0598-3) contains supplementary material, which is available to authorized users.


Background
Congenital generalized lipodystrophy (CGL) is an autosomal recessive disorder that is characterized by a neartotal loss of adipose tissue [1]. Genome-wide linkage analysis has identified two loci related to CGL, i.e., CGL1 mutations in the 1-acylglycerol-3-phosphate O-acyl transferase 2 (AGPAT2) gene and CGL2 mutations in the Berardinelli-Seip congenital lipodystrophy 2 (BSCL2) gene that encodes seipin [2]. CGL2 patients with loss-offunction mutations in the seipin gene exhibit much higher rates of mental retardation than CGL1 patients [3,4]. Seipin is highly expressed in the neuronal cells of the cortex, cerebellum, hippocampus, and hypothalamus [5,6]. The seipin knockout in rats or mice causes spatial cognitive deficits through the synaptic dysfunction in hippocampal CA1 regions without loss of pyramidal cells [5,7].
Seipin, an exclusive endoplasmic reticulum-residing N-glycosylated protein, can affect the generation of peroxisome proliferator-activated receptor gamma (PPARγ) [8]. The level of PPARγ is reduced in the embryonic fibroblasts of seipin-deficient mice [9]. We have recently reported that the neuronal knockout of seipin in mice (seipin-KO mice) reduces the expression of the hippocampal PPARγ [7,10,11]. The PPARγ is expressed in astrocytes and microglia and exerts an anti-inflammatory effect [12]. PPARγ agonists can inhibit the activations of microglia and astrocytes [13] and reduce the production of pro-inflammatory cytokines [14]. PPARγ deficiency has been reported to increase the neuroinflammation in allergic encephalomyelitis [15] and multiple sclerosis [16]. The levels of pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), are elevated in Alzheimer's disease (AD) brains [17]. The activation of PPARγ can enhance the cognitive reserve in humans with AD and in mouse model for AD amyloidosis [18]. The activation of PPARγ can improve hippocampus-dependent cognitive deficits in AD mouse models [19]. The PPARγ activation has been recently shown to mitigate the neuronal inflammation in chronic and acute neurological insults [20]. The PPARγ has received increasing attention in AD due to its anti-inflammatory function [21]. The hippocampal astrocytes express the seipin protein [10]. Thus, it should be interesting to examine whether seipin deficiency in astrocytes through reducing PPARγ affects Aβ-induced neuroinflammation and neurotoxicity.
The beneficial effects of PPARγ ligands potentially involve the suppression of signal transducer and activator of transcription (STAT) [22]. Li et al. have reported the elevation of STAT3 activity in the hippocampus of seipin-KO mice [10]. The STAT family of transcription factors has a central role in inflammatory reactions through regulating the expressions of multiple cytokines. In addition, the activation of STAT3 is known to be crucial for the differentiation of astrocytes [23]. Adult seipin-KO mice showed an increase in the astrocytic differentiation of progenitor cells in the hippocampal dentate gyrus, which can be corrected by the activation of PPARγ [10]. The STAT3 activation is highly dependent on the phosphorylation of glycogen synthase kinase-3 (GSK3) [24]. GSK3 is a constitutively active Ser/Thr kinase consisting of GSK3α and GSK3β isoforms. GSK3β has been identified as a strong promoter of pro-inflammatory cytokines, including TNF-α and IL-6 [25,26].

Behavioral examination
The behavioral performances on days 5-14 after Aβ [25][26][27][28][29][30][31][32][33][34][35] injection were captured via video recording (Winfast PVR; Leadtek Research, Fremont, CA) and analyzed using TopScan Lite 2.0 (Clever Sys., Reston, VA). Morris water maze task: The water maze task was consecutively performed to examine the spatial memory [29]. A pool (diameter = 120 cm) made of black-colored plastic was prepared. The water temperature was maintained at 20 ± 1°C with a bath heater that was used before the sessions. On days 1-2 of training, a cylindrical blackcolored platform (diameter = 7 cm) was placed 0.5 cm above the surface of the water. A mouse was randomly released into one of the four different quadrants and allowed to swim for 90 s. The latency to reach the visible platform was measured. On days 3-7 of the training, the platform was moved to the opposite quadrant of the previously visible platform and submerged 1 cm below the surface of the water. Four trials were conducted each day with an intertrial interval of 30 min. The average swimming speeds (m/s) and latencies (s) to reach the platform were scored for all trials. If the mouse was unable to reach the platform within 90 s, the experimenters gently assisted it onto the platform and allowed it to remain there for 15 s. Each mouse began in one of the four quadrants, which was selected in a random manner. On day 8 of the training, a probe trial was performed by removing the platform. The mouse was released from the opposite quadrant relative to the previous location of the platform and allowed to swim for 90 s. The percentages of swimming time spent in the target quadrant, opposite quadrant, and right and left adjacent quadrants were determined. Y-maze task: Spatial working memory performance was assessed by recording spontaneous alternation behavior in a Y-maze [30]. A Y-maze task was performed 48 h after the probe trial task. The Y-maze was constructed of black-painted wood. Each arm was 40-cm long, 13-cm high, 3-cm wide at the bottom, and 10-cm wide at the top, and the arms converged at equal angles. Each mouse was placed at the end of one arm and allowed to move freely through the maze during an 8-min session. The series of arm entries were recorded visually, and arm entries were considered to be completed when the hind paws of the mouse were completely placed in the arm. Alternations were defined as successive entries into the three arms on overlapping triplet sets. The percentage of alternations was calculated as the ratio of actual to possible alternations (defined as the total number of arm entries minus two). The scorers were blinded to the treatment groups.

Histological examination
The mice were anesthetized with chloral hydrate (400 mg/kg, i.p.) and perfused with 4 % paraformaldehyde. The brains were removed and immersed in 4 % paraformaldehyde at 4°C overnight.
Toluidine blue staining: The brains were processed for paraffin embedding, and coronal sections (5 μm) were cut. The pyramidal cells in the hippocampal CA1 region were identified using a conventional light microscope (Olympus DP70, Tokyo, Japan) with a 60× objective. The healthy pyramidal neurons exhibited round cell bodies with plainly stained nuclei. Stereological counting CA1 pyramidal cells: The brains were transferred into 30 % sucrose. After gradient dehydration, the coronal sections of the hippocampus (40 μm) were cut with a freezing microtome (Leica, Nussloch, Germany) and stained with toluidine blue. Every fourth section was obtained for consecutive cell quantification analyses. Stereological cell counting was performed by a stereological system, consisting of a light microscope with a CCD camera (Olympus DP70), a motorized specimen stage for automatic sampling (MicroBrightField, Williston, USA), and a computer running Microbrightfield Stereo Investigator software (Microbrightfield, Williston, VT, USA) [36]. All counts were performed by the same investigator, who was blind to the treatment. The total numbers of healthy pyramidal cells throughout the hippocampal CA1 region were counted using the optical fractionators' method. The section thickness was measured using a dissector height of 10 μm. The layer of pyramidal cells was defined according to the terminology of Blackstad [37]. The boundary between the CA3 and CA1 subregions was identified by the differential thickness of the cell layers and by the smaller cell somas of the CA1 subregion. The boundary between CA1 and the subiculum was defined as the line separating the contiguous cells of CA1 from the more widely spaced cells of the subiculum.
Hoechst staining: Coronal paraffin sections (5 μm) were incubated in Hoechst 33342 (1 μg/ml, Cell Signaling Technology, Inc., Boston, MA, USA) for 2 min. Quantitative analyses of Hoechst-positive cells: Hoechst-positive (Hoechst + ) cells in hippocampal CA1 region were observed by a conventional light microscope (Olympus DP70, Japan) with 60× objective and counted. The density of Hoechst + cells was expressed as the number of cells per millimeter length along the pyramidal cells layer.
Counting GFAP-and Iba1-positive cells in the hippocampal CA1: The GFAP-positive (GFAP + ) cells and Iba1positive (Iba1 + ) cells in the hippocampal CA1 radiatum layer were counted using a conventional light microscope (Olympus DP70). The radiatum layer was outlined at ×5 magnification. Systematic random sampling was achieved with a uniform sampling grid superimposed randomly over the region. A counting frame was placed in each square of the grid, and the cells were counted within that frame at ×40 magnification. A cell was counted if the nucleus was stained and was (a) in focus, (b) within the counting frame or crossed the green counting frame inclusion line, and (c) within the optical disector. Total estimated cell number per section = number of counted cells × (1/ssf) × (1/asf) × (1/hsf), where ssf is the section sampling fraction, asf is the area sampling fraction, and hsf is the height sampling fraction [38]. The densities of GFAP + cells and Iba1 + cells are expressed as the mean numbers per mm 3 , which were normalized to the control value obtained from WT mice [39].

Data analysis/statistics
The data were retrieved and processed with the Microcal Origin 8.0 software. The group data are expressed as the means ± standard errors, and significance was tested with Student's t tests or ANOVAs with or without repeated measures, followed by Bonferroni post hoc analysis for multiple comparisons. The results of the Morris water maze test were analyzed using repeated-measures ANOVA. The statistical analyses were performed using Stata 7 software (STATA Corporation, College Station, TX, USA). Differences were considered statistically significant at P < 0.05.

Seipin deficiency enhances Aβ 25-35 -induced cognitive deficits
Spatial memory was examined using place learning in a Morris water maze task from day 5 after the injection (i.c.v.) of Aβ [25][26][27][28][29][30][31][32][33][34][35] . The mean latency required to find the visible platform on days 1-2 of training and the subsequent escape latencies to reach the hidden platform on days 3-7 of training are illustrated in Fig. 1a. First, the latency to reach the visible platform was not affected by genotype (F (1, 60)  Spatial working memory performance was assessed with a Y-maze task. There was a main effect of genotype on the alternation ratio (F (1, 60) = 18.472, P < 0.001; Fig. 1c), but no effect on the number of arm entries was observed (F (1, 60) = 0.165, P = 0.686; Fig. 1d). The alternation ratio of the seipin-KO mice was reduced compared to that of the WT mice (P < 0.05). Additionally, there was a significant interaction of genotype × Aβ 25-35 for the alternation ratio (F (1, 60) = 7.214, P = 0.009). Notably, Aβ 25-35 -KO mice exhibited a significant reduction in the alternation ratio compared with the seipin-KO mice (P < 0.05).
Consistent with the report by Zhou et al. [7], the treatment of the seipin-KO mice with the PPARγ agonist rosiglitazone (rosi, 5 mg/kg, p.o.) for consecutive 17 days corrected the prolongation of escape latencies to reach the hidden platform (P < 0.05, n = 16; Fig. 2c-i) and the decrease in the swimming time of target quadrant (P < 0.05, n = 16; Fig. 2d) and the alternation ratio of Y-maze (KO: P < 0.05, n = 8; Aβ 25-35 -KO: P < 0.05, n = 8; Fig. 2e). To test the involvement of the reduced PPARγ level in the Aβ 25-35induced cognitive deficits, the rosi treatment was given for consecutive 17 days starting from 2 days before the Aβ 25-35 of the probe test. The left panels illustrate typical tracks. **P < 0.01 vs. WT mice; ## P < 0.01 vs. seipin-KO mice (two-way ANOVA with Bonferroni's test). c, d The bar graphs display group means in the alternation rate (%) and the numbers of arm entries (8 min) in the Y-maze (Y-M) task. *P < 0.05 and **P < 0.01 vs. WT mice; # P < 0.05 vs. seipin-KO mice (two-way ANOVA with Bonferroni's test) injection (Fig. 2b). In comparison with the vehicle-treated Aβ 25-35 -KO mice, the rosi treatment in the Aβ 25-35 -KO mice could perfectly reduce their escape latencies reaching the hidden platform (P < 0.01, n = 16; Fig. 2c-ii) and increase the swimming time in the target quadrant (P < 0.01, n = 16) and alternation ratio in the Y-maze (P < 0.05, n = 8).
The activation of GSK3β is required to activate tyrosine phosphorylation of STAT3 in astrocytes and microglia [24]. Similarly, the level of hippocampal STAT phosphorylation (phospho-STAT) in the seipin-KO mice was elevated compared with that in the WT mice (P < 0.01, n = 8; Fig. 5c). The level of phospho-STAT3 in the seipin-KO mice was further increased by the rosi treatment (P < 0.01, n = 8) but was not affected by the Aβ 25-35 injection (P > 0.05, n = 8). Additionally, the rosi treatment increased the level of phospho-STAT3 in the WT mice (P < 0.05, n = 8).
Seipin deficiency enhances GSK3β activity to increase pro-inflammatory cytokines GSK3β has been identified as a strong promoter of proinflammatory cytokines, including IL-6 and TNF-α [25,26]. The inhibition of GSK3β increases inflammatory tolerance and reduces inflammatory sensitization in the brain [45]. GSK3β inhibition in glial cells reduces pro-inflammatory responses by blocking STAT3 signaling [46]. The inhibition of GSK3 reduces STAT3 activation, IL-6 production, and GFAP up-regulation by LPS-stimulated primary glia [47]. The activation of PPARγ potentially suppresses the activator of STAT [48]. The level of STAT3 phosphorylation was significantly increased in the seipin-KO mice [10]. STAT3 activation is highly dependent on GSK3β in mouse primary astrocytes [24]. Notably, the catalytic activity of GSK3β in seipin-KO mice was enhanced as indicated by the elevation of Tyr216 phospho-GSK3β and the reduction of Ser9 phospho-GSK3β. The activation of PPARγ regulates negatively the expressions of GSK3β [49], but the level of GSK3β protein failed to be increased in the seipin-KO mice. Strangely enough, the administration of rosi to the seipin-KO mice could suppress the increase of GSK3β activity but further enhanced the phosphorylation of STAT3. Therefore, it is indicated that seipin deficiency in astrocytes through reducing PPARγ enhances the GSK3β activity rather than the STAT3 signaling to increase the production of TNF-α and IL-6. On the other hand, GSK3β regulates interferon-γ (IFN-γ) signaling and is involved in IFN-γ-induced inflammation [50]. Thus, further studies are required to examine whether the GSK3β activity in seipin-KO mice through synergistically facilitating IFN-γ-induced STAT1 activation increases the production of TNF-α [41].

Seipin deficiency enhances Aβ-neurotoxicity via the reduction of PPARγ
A principal finding in this study is that seipin deficiency in hippocampal neuronal cells enhances the Aβ 25-35/1-42 neurotoxicity as indicated by a massive death of pyramidal cells in the seipin-KO mice treated with the Aβ [25][26][27][28][29][30][31][32][33][34][35] (1.2 nmol/mouse) or Aβ 1-42 (0.1 nmol/mouse), generally non-neurotoxic doses in the WT mice or control mice [29]. The level of the hippocampal PPARγ was significantly reduced in the seipin-KO mice. The decline of PPARγ in the seipin-KO mice [11] or the neuronal specific PPARγ knockout in mice [51] did not cause the loss of neuronal cells in the hippocampus. The rosi treatment could significantly reduce the Aβ 25-35/1-42 -induced death of pyramidal cells in the seipin-KO mice. Activated microglia and reactive astrocytes can produce cytokines, reactive oxygen species, and other neurotoxic substances to cause neuronal apoptosis and other series of pathologic events. The progressive neuroinflammation and neuronal apoptosis in AD is considered to be a consequence of the Aβneurotoxic properties [52,53]. Our data support the notion that Aβ 25-35/1-42 -induced neuroinflammation in the seipin-KO mice causes the death of neuronal cells. Zhao et al. [51] reported that neuronal PPARγ deficiency increased susceptibility to brain damage after cerebral ischemia through suppressing the expressions of Cu-Zn superoxide dismutase (SOD1), catalase (CAT), and glutathione S-transferase (GST). However, the levels of hippocampal SOD1, CAT, and GST expression in seipin-KO mice did not differ from WT mice (data not shown). Additionally, Ito et al. reported that mutations (N88S/S90L) of the seipin gene can cause the formation of cytoplasmic inclusions and enhance ubiquitination, which leads to endoplasmic reticulum (ER) stress [54]. However, this idea is not supported because the levels of the ER stress makers BiP and CHOP in seipin-KO mice are not increased [11].

Conclusions
Using the seipin-KO mice with reduction of PPARγ, we observed that seipin deficiency elevated the activity of GSK3β to enhance the production of TNF-α and IL-6, which in turn triggered and strengthened the Aβ 25-35 -induced inflammatory responses leading to the death of neuronal cells and the spatial cognitive deficits. This is particularly true of neuroinflammation, which contributes to a broad range of neurodegenerative diseases [55]. Although much more work needs to be performed in the future, this is the first report to demonstrate that the expression of seipin in hippocampal astrocytes is required for regulating the inflammatory responses and preventing the neurodegeneration.