Global transcriptome analysis of rat hypothalamic arcuate nucleus demonstrates reversal of hypothalamic gliosis following surgically and diet induced weight loss

The central mechanisms underlying the marked beneficial metabolic effects of bariatric surgery are unclear. Here, we characterized global gene expression in the hypothalamic arcuate nucleus (Arc) in diet-induced obese (DIO) rats following Roux-en-Y gastric bypass (RYGB). 60 days post-RYGB, the Arc was isolated by laser-capture microdissection and global gene expression was assessed by RNA sequencing. RYGB lowered body weight and adiposity as compared to sham-operated DIO rats. Discrete transcriptome changes were observed in the Arc following RYGB, including differential expression of genes associated with inflammation and neuropeptide signaling. RYGB reduced gene expression of glial cell markers, including Gfap, Aif1 and Timp1, confirmed by a lower number of GFAP immunopositive astrocyte profiles in the Arc. Sham-operated weight-matched rats demonstrated a similar glial gene expression signature, suggesting that RYGB and dietary restriction have common effects on hypothalamic gliosis. Considering that RYGB surgery also led to increased orexigenic and decreased anorexigenic gene expression, this may signify increased hunger-associated signaling at the level of the Arc. Hence, induction of counterregulatory molecular mechanisms downstream from the Arc may play an important role in RYGB-induced weight loss.


Materials and Methods
Animals. The Danish Animal Experiments Inspectorate approved all experiments which were conducted using internationally accepted principles for the use of laboratory animals under the personal license #2013-15-2934-00784. Male Sprague-Dawley rats (8 weeks old) were obtained from Taconic (Lille Skensved, Denmark). Upon arrival, rats were single-housed and maintained in controlled environmental conditions (12 h light/12 h dark cycle; 22 ± 1 °C; 50 ± 10% relative humidity). Rats were fed a two-choice diet consisting of chow (Altromin 1324, Brogaarden Denmark) and a high-palatable high-fat diet (HPHF diet; Nutella, peanut butter and powdered chow), as described previously 17 . Rats had ad libitum access to water and the two-choice diet regimen for 4 (4w DIO, n = 16) or 12 (12w DIO, n = 30) weeks to induce mild or severe obesity, respectively, prior to surgical intervention. For 12w DIO rats, body weight and food intake was monitored daily throughout the study. Body weight and food intake in the 4w DIO rats has been reported previously 16 . Surgical procedures. Rats were assigned to RYGB (4w DIO, n = 8; 12w DIO, n = 10) or sham surgery (4w DIO, n = 8; 12w DIO, n = 10). Sham-operated and weight-matched (WM) rats (n = 10) served as an additional control group to RYGB in the 12-week pre-feeding study. Three days prior to surgery, animals were placed on a liquid diet (Osmolite 1 Cal, Abbott Nutrition, Chicago, IL; Fresubin Original, Mediq Danmark, Broendby, Denmark). On the day of surgery, animals underwent whole-body composition analysis by non-invasive EchoMRI scanning (EchoMRI-900 Analyzer, EchoMRI, USA). The RYGB surgical procedure were carried out as previously described in detail 16 . In brief, the abdomen was exposed using a midline laparotomy after induction of surgical anesthesia with an isoflurane/O 2 mixture. The jejunum was transected 30 cm distal to the ligament of Treitz, and a longitudinal anti-mesenteric incision was made 10 cm distal to the transected bowel and connected to the afferent limb of the jejunum with a running absorbable suture (Ethicon, Somerville, NJ, USA). The stomach was exposed, the fundus was excised and a staple line was placed across the waist of the stomach, creating a gastric pouch approximately 10% the size of the normal stomach. The distal remnant was returned to the peritoneal cavity and an incision was made on one side of the gastric pouch considering the vascular architecture. The efferent limb of the transected jejunum was connected to the gastric pouch with a running suture. After repositioning the gastric pouch into the peritoneal cavity, the abdominal wall was closed. The sham procedure followed the steps of the RYGB but gastrointestinal surgery only included a transection of the jejunum 30 cm distal to the ligament of Treitz, which was immediately re-sutured. Animals were subcutaneously administered with warm saline (37 °C, 20 ml/kg), enrofloxacin (0.5 mg/kg) and carprofen (0.5 mg/kg) to prevent post-operative infection, lethargy and pain relief. The wire grate was kept in place post-operatively for up to 10 days where only liquid diet was offered in this period whereupon the two-choice diet was reintroduced. The food ration for 12w DIO-WM rats was adjusted on a daily basis to promote a body weight change equal to that of 12w DIO-RYGB rats. A daily check was made to ensure that the previous food ration had been consumed.
Termination and tissue sampling. 60 days post-surgery, animals were euthanized by decapitation under CO 2 /O 2 anesthesia. All animals were euthanized in the morning 2-3 hours after lights on. RYGB and sham-operated animals were euthanized in a randomized order. Prior to euthanasia, animals underwent an additional whole-body echo-MRI scan. Following decapitation, brains were quickly removed and snap-frozen on crushed dry ice and stored at −80 °C until further processing.
Laser-capture microdissection and RNA sequencing. The brains were sectioned on a cryostat (CM 3050, Leica, Germany) into 12 μm thick coronal sections using systematic uniform random sampling (Fig. 1A). Serial sections covering the entire rostro-caudal axis of the Arc were collected at every 180 μm (i.e. every 15 th section) on Pen membrane microscope slides (Life Technologies, Carlsbad, CA) for LCM and on super-frost plus slides for immunostaining, respectively. For LCM, sections were allowed to thaw for 5 min at 4 °C and then stained for 5 min in a 0.1% cresyl violet acetate (Sigma-Aldrich, St. Louis, MI) solution dissolved in 70% ethanol. Sections were then dehydrated in 96% and 100% ethanol at 4 °C and finally allowed to dry in fume hood for 2 min at room temperature. Using the Arcturus XT LCM Instrument (AXT2148, Life Technologies, Carlsbad, CA) the www.nature.com/scientificreports www.nature.com/scientificreports/ Arc was identified and captured by a combination of infrared capturing and UV laser cutting. The Arc tissue was captured on CapSure ® Macro LCM Caps (Applied Biosystems, Foster City, CA). The Arc was collected on 8-10 sections representing the entire rostro-caudal axis of the nucleus (Fig. 1B). Lysis of cells and RNA extraction was performed using the PicoPure RNA isolation kit (Life Technologies, Carlsbad, CA) as recommended. RNA integrity was assessed using the Bioanalyzer 2100 system (Agilent Technologies, Waldbronn, Germany). RIN (RNA integrity number) values ranged between 7.3-8.1 (data not shown). RNA amplification and sequencing was performed on all Arc samples by Cofactor Genomics (Saint Louis, MO). In brief, RNA sequencing (RNAseq) libraries were prepared with the Ovation ® RNAseq system (NuGen, Redwood City, CA) and sequenced on the HiSeq platform (Illumina, San Diego, CA) (single-end, approximately 20 M 75 bp reads). Reads were aligned to the Rnor_5.0 Ensembl Rattus norvegicus genome using STAR v.2.3.0 and feature counts were obtained with HTseq v. 0.6.1, both with default parameters 18,19 . Known orexigenic/anorexigenic genes (AgRP, NPY, CART and POMC) were selected a priori for analysis of the orexigenic state in the Arc following RYGB. For network analysis a raw p-value < 0.01 was used as threshold and protein-protein interaction annotations were obtained from the String database 20 and imported to Cytoscape 21 for visualization. To identify putative associations of gene expression patterns, the Cytoscape plug-in BINGO 22 was used to extract enriched GO terms for each network.
Immunohistochemical analysis of astrocyte profiles. In all individual rats, an additional series of cryo-sections was used for glial fibrillary acidic protein (GFAP) immunohistochemistry. Sections were fixated in 10% neutral buffered formalin for 10 min followed by rinsing in water. Endogenous peroxidase activity was blocked for 10 min in 1% H 2 O 2 + tris buffered saline (TBS) + 0.3% Triton-X (TX). Blocking of unspecific binding was obtained with 5% swine serum in TBS + 0.3% TX + 1% BSA for 20 min followed by addition of rabbit anti-GFAP antibody (1:300; Z0334; Dako, Denmark) diluted in TBS-T + 1% BSA for 60 min. The reaction was amplified for 30 min using MACH 2 rabbit HRP-polymer (RHRP520, BioCare Medical, USA). Sections were rinsed in TBS before subjected to 3,3′-diaminobenzidine as chromagen (K3467; Dako, Denmark). The evaluation of astrocyte profiles was performed on digitalized slides using the newCAST software (Visiopharm, Hørsholm, Denmark). A region of interest was delineated and a counting frame was placed using systematic uniform random sampling principles. Astrocyte profiles were counted when located within the counting frame. A total number of approximately 50 counting frames was evaluated.
Statistics. Graphical presentations, calculations and statistical analyses were carried out using GraphPad software (GraphPad Prism version 5, San Diego, California, USA). Statistical analysis of in-vivo data, plasma analyses and mRNA levels of a priori selected genes were performed using either one-way analysis of variance www.nature.com/scientificreports www.nature.com/scientificreports/ (ANOVA) followed by Tukey's post-hoc test or student's unpaired t-test where appropriate. A p value less than 0.05 was considered statistically significant. Differential gene expression analysis was performed using edgeR with raw p-values 23 . Combined p-values were calculated using Fisher's combined probability test applied separately to genes up-regulated or down-regulated in both cohorts. Table 1. We have previously reported significantly reduced terminal body weight and whole-body fat mass in 4w DIO-RYGB rats compared to 4w DIO-sham controls 16 . Correspondingly, 12w DIO-RYGB rats showed significant reductions in terminal body weight relative to baseline (F(2,27) = 4.52, p < 0.001). Also, 12w DIO-RYGB rats exhibited lowered terminal whole-body fat mass as compared to 12w DIO-sham controls (F(2,27) = 5.41, p < 0.05). The magnitude of absolute and relative weight loss as well as whole-body fat mass reductions in 12w DIO-RYGB rats were comparable to that of Sham-WM rats. Terminal lean mass was unaffected by RYGB in 12w DIO rats and slightly reduced in Sham-WM rats (F(2,27) = 3.67, p < 0.05). Cumulated energy intake was similar between RYGB and Sham rats in 12w DIO rats whereas Sham-WM rats were significantly food restricted (both chow and HPHF diet) in order to match the weight loss of RYGB rats (F(2,27) = 24.81, p < 0.001). Furthermore, 12w DIO-RYGB rats had an increased cumulated chow intake as compared to 12w DIO-sham controls (F(2-27) = 18.54, p < 0.001). Terminal plasma hormone levels in 12w DIO rats were significantly changed by RYGB (Table 1). Leptin levels were significantly reduced in 12w DIO-RYGB, but not Sham-WM rats, compared to 12w DIO-sham controls (F(2,27) = 6.08, p < 0.01). C-peptide levels were significantly reduced in both 12w DIO-RYGB and 12w DIO-WM rats (F(2,27) = 9.72, p < 0.001). Plasma total GLP-1 were significantly increased in 12w DIO-RYGB rats, as compared to both 12w DIO-sham and Sham-WM rats (F(2.27) = 16.18, p < 0.001). Active GLP-1 levels were significantly increased in 12w DIO-RYGB rats compared to both 12w DIO-sham controls (t = 5.37, df = 17, p < 0.01). While total GLP-1 levels were unchanged in 12w DIO-WM rats, active GLP-1 concentrations were below assay detection level (0.24 pg/ml) in 12w DIO-WM. Total PYY levels were significantly increased in 12w DIO-RYGB rats compared to both 12w DIO-sham and 12w DIO-WM rats (F(2.27) = 57.46, p < 0.001). Total PYY levels were unchanged in 12w DIO-WM rats. Plasma ghrelin levels were only increased in 12w DIO-WM rats (F(2.27) = 4.23, p < 0.05). These metabolic effects are in agreement with the consistent findings in clinical studies on RYGB [24][25][26] , supporting the translational value of the present DIO rat model of RYGB.

RYGB induces significant weight loss and increases plasma GLP-1 and PYY levels. Metabolic parameters are indicated in
Global gene expression analysis in the arcuate nucleus. LCM on the Arc was used to determine a region-specific transcriptional profile of gene regulatory events following RYGB vs. sham-operated and sham-WM controls, respectively (Fig. 1). To increase confidence in the observed differential expression, arcuate RNAseq data sets from DIO-RYGB rats (4w DIO-RYGB, 12w DIO-RYGB) and DIO-sham rats (4w DIO-sham, 12w DIO-sham) were pooled and analyzed specifically for differentially expressed genes (DEGs) regulated in the same direction in both 4w and 12w DIO-RYGB rats. Using this procedure, a total of 21 DEGs were identified (combined FDR <0.05), including 13 downregulated and 8 upregulated genes, respectively ( Table 2). Most significantly down-regulated genes were associated with astroglial (Gfap, Aqp4, Fabp7) and immune cell (Serping1, Fabp4) function. In addition, RYGB promoted significant downregulation of genes associated with extracellular matrix (ECM) remodeling and cell migration (Timp1, Tmbim1, Capn3, Ucma). Significantly upregulated genes represented transcription factors (Egr1, Egr2), nuclear receptors (Nr4a1, NR4a3) as well as the orexigenic peptide, agouti-related peptide (Agrp). The set of 21 genes were more robustly regulated in 4w DIO-RYGB rats compared to 12w DIO-RYGB rats ( Table 2).

Network analysis and histological validation. Subsequent network analysis of arcuate DEGs in 4w
DIO-RYGB rats confirmed several significantly downregulated genes (raw p-value below 0.01) being associated with glial function, see Fig. 2A. Most marked reductions in expression levels (RPKM values) were observed for glia cell-specific genes, including Gfap and Aqp4 (Fig. 2B). Also, the network analysis indicated lowered Gfap expression was linked to reduced expression of the microglial marker Aif1 (also known as Iba1), see Fig. 2A,B. In addition, the network analysis confirmed reduced expression of genes functionally linking glial cell and complement function, as represented by significant downregulation of Timp1 ( Fig. 2A,B). 12w DIO-RYGB and 12w DIO-WM rats showed similar degree of downregulation of Gfap, Aqp4, Aif1 and Timp1 expression relative to 12w DIO-sham controls (Fig. 2B). Compared to 12w DIO-sham rats, Egr1 expression was significantly down-regulated in 12w DIO-RYGB rats, but not 12w DIO-WM rats (Fig. 2C). To further confirm reduced expression of astroglial markers following RYGB, we performed quantitative immunohistochemical analysis of GFAP-positive astrocyte profiles in the Arc of 4w DIO rats. Compared to 4w DIO-sham rats, 4w DIO-RYGB rats showed significantly lowered numbers of GFAP-positive cell profiles (199 ± 11 vs. 169 ± 8, p < 0.05), see Fig. 2D.

Discussion
Here, we employed LCM in combination with RNAseq for a highly sensitive analysis of gene expression changes in the hypothalamic Arc following RYGB surgery in the DIO rat. Global gene sequencing indicated that the most marked transcriptional changes in the hypothalamic arcuate nucleus following RYGB were associated to markers of immune function. Notably, Gfap expression and number of GFAP-positive cell profiles were significantly downregulated compared to sham-surgery, suggesting reversal of reactive gliosis after RYGB-induced weight loss. Moreover, we extend previous findings based on quantitative ISHH technologies and confirm upregulation of arcuate orexigenic gene expression markers after RYGB surgery.
The initial identification of downregulated Gfap and Timp1 expression suggested altered glia cell function upon RYGB surgery. Whereas Gfap is a well-known marker of astrocytes 27 , Timp1 is known to be expressed by both neurons and astrocytes in response to insults such as ischemia, demyelination and kainate-induced seizures [28][29][30] . It has previously been demonstrated that high fat diet and obesity induce a state of inflammation in the hypothalamus, as indicated by an increased expression of proinflammatory marker genes (such as IL-1β, IL-6, TNF-α, and Gfap) 31,32 , supported by reports of increased number and ramifications of microglia 33 and increased numbers of GFAP positive cells 34 . To further address potential inflammatory changes after RYGB, we performed a series of targeted bioinformatics analyses. A network analysis supported reduced expression of several glia cell marker genes, as well as genes related to the complement innate immune system, following RYGB. Astrocytes and microglia are dynamic cells that are important in maintaining the microenvironment surrounding neurons  www.nature.com/scientificreports www.nature.com/scientificreports/ and in co-modulating neuronal functions, such as synaptic plasticity and metabolism 35,36 . In response to CNS insults, both cell types respond rapidly and develop a reactive phenotype termed reactive gliosis characterized by upregulation of specific structural proteins like GFAP in astrocytes 37 and allograft inflammatory factor 1 (AIF-1) in microglia 38 . The reduction in both Gfap, Aqp4 and Aif1 expression after RYGB therefore suggests ongoing processes towards reversal of gliosis. Since data from WM controls demonstrated similar alterations, the reversing of gliosis appears to be an effect of weight loss per se, rather than specific to RYGB surgery. In line with this, Grayson and coworkers 39 have reported reduced microglia infiltration in the hippocampus following RYGB, and pair-feeding leads to a similar improvement. RYGB also resulted in increased expression of the transcription factors Egr1 and Egr2. Interestingly, this effect was not observed in WM animals. Although the implications of regulated Egr1/Egr2 expression in the Arc is not fully understood, Egr1 expression has been associated with appetite signaling. For example, central infusion of NPY increases Egr1 protein levels in the Arc, and downregulation of Egr1 results in halted post-translational processing of POMC into α-MSH 40    The number of GFAP-positive astroglia profiles was significantly downregulated in 4w DIO-RYGB relative to 4w DIO-sham controls. Representative micrographs of GFAP-immunostained sections from sham-operated and RYGB-operated animals with arrowheads identifying astrocytes; scalebar equal to 20 µm in inserts. GFAPpositive cell counts are presented as mean ± SEM; *p < 0.05 (Student's unpaired t-test). Gene names: Aif1, Allograft inflammatory factor 1; Anxa2, Annexin A2; Anxa5, Annexin A5; Arpc1b, Actin-related protein 2/3 complex subunit 1B; Aqp4, Aquaporin 4; C1s, Complement component 1s; C1qb, Complement component 1q B chain; Egr1, Early growth response protein 1 (also known as Zif268); Gfap, Glial fibrillary acidic protein; S100B, S100 calcium-binding protein B; Lgals1, Galectin-1 (lectin, galactose binding, soluble 1); Serping1, Serpin peptidase inhibitor, clade G (C1 inhibitor) member 1; Timp1, Tissue inhibitor of metalloprotease 1.
ghrelin also upregulates arcuate Egr1 protein expression in both fasted and fed rats 41 , this suggests a complex role of Egr1 in both orexigenic and anorexigenic signaling.
Using quantitative ISHH, we have previously reported that RYGB increases NPY and AgRP mRNA expression in the Arc without affecting POMC and CART mRNA levels in 4w DIO-RYGB rats, suggesting increased orexigenic tone at the level of the mediobasal hypothalamus 16 . Using combined LCM-RNAseq, we confirmed upregulation of arcuate orexigenic signaling in both 4w and 12w DIO-RYGB rats. In addition, LCM-RNAseq analysis indicated reduced Pomc and Cart expression in 4w DIO-RYGB rats, but not 12w DIO-RYGB rats. The discrepancies may be related to differences in sensitivity using RNAseq compared to ISHH as well as differences in pre-surgical body weight. Nonetheless, taken together our data suggest enhanced arcuate orexigenic signaling in a DIO rat model of RYGB, and extend findings reported in other rat models of bariatric surgery, including biliopancreatic diversion 14 , vertical sleeve gastrectomy 42 and gastric banding 43 . NPY/AgRP and POMC/CART expressing neurons are well-known metabolic sensors highly responsive to changes in feeding status and circulating levels of nutrients and hormones, including glucose, leptin and insulin (for review see 44 ). As RYGB reduced plasma leptin and insulin levels, it may be speculated that improved metabolic status played a role in the adaptive changes in the Arc. RYGB also markedly increased plasma levels of GLP-1 and PYY levels, which are gut-derived Orexigenic AgRP (A) and NPY (B) mRNAs were elevated after RYGB in both studies relative to sham controls. Anorexigenic CART (C) and POMC (D) mRNAs were reduced in RYGB following 4 weeks, but not 12 weeks of pre-feeding. Statistical analysis was performed using EdgeR with raw p-values; *p < 0.05, **p < 0.01, ***p < 0.001 compared to corresponding sham control group. (E) The increased ratio between the expression level of orexigenic and anorexigenic genes suggests an orexigenic signaling state following RYGB compared to sham surgery; ***p < 0.01 compared with sham (student's unpaired t-test or one-way ANOVA with Tukey posthoc test). Data is presented as mean ± SEM. anorectic hormones 45,46 . We have previously reported RYGB-induced hypertrophy of the rat gut mucosa in the food exposed regions of the small intestine, coupled with a doubling in the total number of L-cells and marked increases in intestinal preproglucagon and PYY mRNA expression 47 . Hence, the intestinotrophic effect of RYGB likely explains the markedly elevated plasma levels of GLP-1 and PYY in the DIO-RYGB rat. Because WM did not increase circulating levels of GLP-1 and PYY, this supports the notion that these hormones are not responsible for RYGB-induced weight loss, at least at the level of the Arc 48 .
conclusion For the first time, we here present global gene expression analysis of the Arc in a rat model of RYGB surgery. By use of highly sensitive cutting-edge technologies (the combination of LCM and RNAseq), we demonstrate a marked modulation of immune signaling pathways following RYGB. In particular, mRNAs related to astrocyte and microglia function were reduced, suggesting lowered glia cell activity. These findings support that weight loss, whether induced by bariatric surgery or caloric restriction, may lead to improved regulation of immune function within the Arc.