Microarray Analysis of Gene Expression in the Supraoptic Nucleus of Normoosmotic and Hypoosmotic Rats

Abstract

1. Hypoosmolality produces a dramatic inhibition of vasopressin (VP) and oxytocin (OT) gene expression in the supraoptic nucleus (SON). This study examines the effect of sustained hypoosmolality on global gene expression in the OT and VP magnocellular neurons (MCNs) of the hypothalamo-neurohypophysial system (HNS), in order to detect novel genes in this system that might be involved in osmoregulation in the MCNs.

2. For this purpose, we used Affymetrix oligonucleotide arrays to analyze the expression of specific genes in laser microdissected rat SONs, and their changes in expression during chronic hypoosmolality. We identified over 40 genes that had three-fold or more greater expression in the SON versus total hypothalamus, and that also changed more than two fold in expression as a result of the chronic hypoosmolar treatment. These genes contained both novel as well as genes previously known to be present in the SON. All of the raw data for the genes that are expressed in the SON and altered by hypoosmolality can be found on the following NINDS website URL address: http://data.ninds.nih.gov/Gainer/Publications

INTRODUCTION

The hypothalamo-neurohypophysial system plays an essential role in the maintenance of body fluid and electrolyte homeostasis by secreting vasopressin and oxytocin (Antunes-Rodrigues et al., 2004; Robertson, 1995), peptide hormones that are synthesized in the supraoptic and paraventricular nuclei of the hypothalamus and transported to, stored in, and released from the posterior pituitary (Brownstein et al., 1980; Sladek, 1999). During chronic hyperosmolar conditions, VP and OT mRNA levels increase approximately two-fold (Van Tol et al., 1987; Sherman et al., 1988) in order to keep up with the increased secretion of these hormones into the blood for anti-diuretic and natriuretic functions. In contrast, during chronic hypoosmolar conditions, VP and OT mRNA levels in the HNS are decreased to 10–20% of those of normonatremic control animals (Robinson et al., 1990; Verbalis, 1993a), consistent with the inhibition of secretion of VP that results in maximum water diuresis.

While there have been many reports of changes in gene expression in the HNS during hyperosmolality (Sherman et al., 1986; Van Tol et al., 1987; Young et al., 1987; Sherman et al., 1988; Young et al., 1990; Young and Lightman, 1992; Hoffman et al., 1993; Kjaer et al., 1995; Burbach et al., 2001; Ghorbel et al., 2003; Somponpun and Sladek, 2003; Wong et al., 2003), and also during hypoosmolar conditions (Robinson et al., 1990; Verbalis, 1993a; Berghorn et al., 1995; Glasgow et al., 2000; Zhang et al., 2001; Somponpun and Sladek, 2003), most of these have focused on one or a few specific genes. The results of these studies indicate that the hyperosmolar and hypoosmolar states are accompanied by a global but selective change in expression of a wide variety of regulatory genes in the HNS, which could be involved in the magnocellular neurons’ adaptation to chronic, systemic osmotic changes. More recently, efforts have been made to use high-throughput methods, such as microarray analyses, to investigate the effects of these systemic osmotic perturbations on global gene expression in the HNS (Ghorbel et al., 2003; Mutsuga et al., 2004; Ghorbel et al., 2005; Mutsuga et al., 2005; Hindmarch, 2006).

These studies used different array platforms to analyze the gene profiles under these different chronic osmotic states, and these differences in platforms might make comparison of the gene expression profiles difficult (see discussion in Irizarry et al., 2005). The hyperoosmolar study used a high-density Affymetrix oligonucleotide array (Hindmarch, 2006), whereas the hypoosmolality study used a two color, spotted cDNA array (Mutsuga et al., 2005). In this paper, we reduce this complexity due to differences in the Array platforms used by using high-density Affymetrix oligonucleotide arrays to reanalyze the hypoosmotic condition. Here we compare gene expression profiles in the SON during hyperosmotic and hypoosmotic conditions, but using the same array platform. The resulting data are described in this paper, and confirm as well as extend conclusions drawn from the aforementioned previous studies (Mutsuga et al., 2005; Hindmarch, 2006).

METHODS

Animals

Adult male Sprague-Dawley rats weighing 260–320 g were housed individually in wire mesh cages in a temperature-controlled room (21–23°C) with lights on from 7:00AM to 7:00PM. All procedures were carried out in accordance within NIH guidelines on the care and use of animals and an animal study protocol approved by the Georgetown University Animal Use and Care Committee. To induce hyponatremia, male rats were given 1-desamino- [8-D-arginine]-vasopressin (dDAVP; Aventis Pharmaceuticals, Bridgewater, NJ) at a rate of 5 ng/h using osmotic minipumps (Alzet model 2002, Alza Palo Alto, CA) implanted subcutaneously, and by feeding the rats a dilute preparation (1.0 kcal/ml) of liquid formula (AIN76; Bioserv, Frenchtown, NJ) for 7 days (Verbalis and Drutarosky, 1988; Verbalis, 1993b). Rats fed with pelleted AIN-76 and allowed access to tap water ad libitum were used as normoosmolar controls. The body weights (g), plasma Na⁺ levels (mM), and plasma osmolalities (mOsm/kg H₂O) were measured in blood samples drawn via jugular puncture from each animal after 7 days of treatment.

Tissue Isolation and RNA Processing

In this study, we used laser microdissection techniques described previously (Mutsuga et al., 2004), with several modifications. These procedures were described earlier (Mutsuga et al., 2005) and are briefly recapitulated below.

Following decapitation, the rat's brains were quickly removed, immediately frozen on dry ice, and stored at −80°C until further processing occurred. The tissue was placed in a cryostat for 10 min at the cutting temperature (−18°C) for temperature equilibration, and seven μm thick coronal sections were cut at the SON level, placed on and thawed onto membrane-coated glass slides (Leica Microsystems Inc., glass foil PEN slides, Bannockburn, IL, USA), and immediately placed in a slide box embedded in dry ice. The sections were stored at −80°C until they were used. Before laser microdissection, the tissue was fixed and dehydrated with ethanol as described elsewhere (Luo et al., 1999). All of the solutions used for fixation were prepared with DEPC water. Laser microdissection was performed under dark-field illumination using a Leica Laser Microdissection Microscope (Leica Microsystems Inc., Bannockburn, IL) (Kolble, 2000) immediately after dehydration of the slides. Brains from three animals were microdissected. The dissected SONs were collected into 0.5 ml tubes with 70 μl of lysis buffer containing guanidine thiocyanate and 0.5 μl of β-mercaptoethanol.

Blocks of hypothalamic tissue were obtained from 10 control male rats as the reference tissue. After decapitation, the brains were removed from the skull and placed ventral side up on a rubber stopper. Coronal cuts were made rostral to the optic chiasm and caudal to the cerebral peduncle. The hypothalamus was removed from the resulting brain slices by making a cut at the top of the third ventricle and cuts at the lateral margins of the optic tracts. The tissue samples were frozen in liquid nitrogen and stored at −80°C until they were extracted.

RNA Extraction, Amplification and Biotin Labeling

RNA isolation from the laser microdissected SON tissue was performed according to the manufacture's protocol with the Absolutely RNA Microprep Kit (Stratagene, La Jolla, CA) (Dolter and Braman, 2001). After extraction, total RNA was measured using a Ribogreen RNA quantitation kit (Molecular Probes, Inc., Eugene, OR). The samples and the standard RNAs were excited at 485 nm and the fluorescence emission intensity was measured at 535 nm using a fluorescence microplate reader (VICTOR 2™ 1420 Multilabel counter, Wallac Oy, Turku, Finland). Fluorescence emission intensity of the samples was then plotted against an RNA standard curve. The average total RNA per animal (bilateral SON) was 193.2 ng and 90.6 ng for control and hypoosmolar animals, respectively, 60 ng of total RNA from each rat was used as the template for two rounds of linear cRNA and Biotin labeling (see below).

RNA extraction of the dissected hypothalamic blocks was done with Trizol (Life technologies, Gaithersburg, MD) (Chomczynski and Sacchi, 1987) followed by a DNaseI (GenHunter Nashville, TN) treatment. The product was cleaned up using an Absolutely RNA Microprep Kit. Approximately 200 μg of total RNA was obtained from each block, and 5 μg of RNA from each of 10 animals was pooled together. 60 ng and 5 μg of the pool were used for a two-round and a one-round cRNA synthesis and Biotin labeling, respectively.

Experiments were done to compare the effects of one-round versus two-round amplification on gene detection from the whole hypothalamus reference tissue. We found that the intra-group correlation coefficient for either the one-round or the two-round amplifications was 0.912, whereas the inter-group comparison between the one-round and two-round amplification groups was 0.89. The number of genes differentially expressed (at the p < 0.001 level of significance) between the one-round versus two-round amplification groups was 222 genes, or about 0.7% of the total 31,099 genes present on the chip. Thus, we conclude that the second round of amplification produced a very small impact on the mRNA representation of the sample on the chip. Similar observations have been reported by other laboratories (King et al., 2005; Schindler et al., 2005).

The quality and size distribution of purified total RNA was assessed on the Agilent Bioanalyzer (Agilent technologies, Palo Alto, CA) to ensure that the total RNA met the quality and purity criteria to be used as the templates for cRNA synthesis and Biotin labeling.

cRNA synthesis and Biotin labeling was performed according to the manufacture's protocol of the GeneChip^® Expression 3-Amplification One-cycle cDNA Synthesis kit (Affymetrix, Santa Clara, CA) by using 5 μg of total RNA as the templates, or the GeneChip^® Expression 3-Amplification Two-cycle cDNA Synthesis kit (Affymetrix, Santa Clara, CA) by using the 60 ng total RNA as the templates, then the total amount of cDNA from one-round or two-round amplification was used to synthesize the Biotin-labeled cRNA by using MEGAscript™ T7 kit (Ambion, Austin, TX). The purification of both cDNA and cRNA and the quantitation of cRNA were carried out at all stages by using GeneChip^® Sample Cleanup Module (Qiagen, Valencia, CA) and the NanoDrop-1000 (NanoDrop Technologies, Wilmington, DE), respectively. The quality and size distribution of purified labeled cRNA was assessed by using an E-Gel (Invitrogen, Carlsbad, CA), to ensure that the cRNA amplification was successful before proceeding to target fragmentation. This was achieved by incubation at 94°C for 35 min in fragmentation buffer (40 mM Tris-Acetate, pH 8.1, 100 mM KOAc, 30 mM MgOAc). The size distribution of the fragmented labeled transcripts was assessed by using an E-Gel.

High-density Oligonucleotide Array Properties

GeneChip^® Rat Expression Set 230 (Affymetrix, Santa Clara, CA) is comprised of more than 31,000 probe sets and 681,012 distinct oligonucleotide features, analyzing the expression level of over 30,200 transcripts and variants from over 28,000 well-substantiated rat genes. GeneChip^® Rat Expression Array 230A includes representation of the RefSeq data-base sequences and the GeneChip^® Rat Expression Array 230B contains probe sets from refined EST clusters.

GeneChip^® Rat Genome 230 2.0 Array (Affymetrix, Santa Clara, CA) is comprised of more than 31,000 probe sets, analyzing over 30,000 transcripts and variants from over 28,000 well-substantiated rat genes. The only difference between GeneChip^® Rat Expression Set 230 and GeneChip^® Rat Genome 230 2.0 Array is the former provides complete coverage of the rat genome on two arrays, and the latter is on single array.

Affymetrix GeneChip^® Analysis

The fragmented samples were subsequently prepared for hybridization using the Affymetrix hybridization control kit (Affymetrix, Santa Clara, CA). One sample from the control animals and two samples from the reference hypothalamic tissues were hybridized to GeneChip^® Rat Genome 230 2.0 arrays, the rest of the samples were hybridized to the GeneChip^® Rat Expression Set 230. The hybridization was performed in the Affymetrix GeneChip^® Hybridization Oven 640 (Affymetrix, Santa Clara, CA) for 16 hours. Following hybridization the GeneChip^® arrays were stained and washed on the GeneChip^® Fluidics Station 400 (Affymetrix, Santa Clara, CA). Following hybridization, washing and staining of fragmented cRNA to the GeneChip^® expression array, fluorescent signals were detected using the Affymetrix GeneChip^® Scanner 3000 (Affymetrix, Santa Clara, CA), which provides an image of the array and automatically stores high-resolution fluorescence intensity data. These data were initially documented using Affymetrix GeneChip^® Operating Software (GCOS, Affymetrix, Santa Clara, CA), which generates an expression report file that lists the quality control parameters. All of these parameters were scrutinized to ensure that array data had reached the necessary quality standards (Scaling Factor less than 3-fold, average background values 20–100, approximately 50–70% genes called Present, ratio of 3′: 5′ signal no more than 3 for housekeeping genes GAPDH & β-actin). All raw data have been put on the NINDS internet website URL address: http://data.ninds.nih.gov/Gainer/Publications.

Table I.

Continued

Probe set ID	Gene title	Gene symbol	GenBank ID	Entrez gene	Expression ratio	Fold change
1381012_at	Serine (or cysteine) proteinase inhibitor, clade F), member 1	Serpinf1	BI303349	287526	4.33	−2.91
1383032_at	Transcribed locus	—	BG666286		4.33	−2.05
1384921_at	EST	—	AW524713		4.13	−2.31
1371731_at	Phosphofructokinase, liver, B-type	Pfkl	AI408151	25741	3.80	−2.47
1384970_at	similar to RIKEN cDNA 5330437I02 gene	LOC361340	BM389861	361340	3.76	−3.75
1395563_at	EST	—	AA875586		3.72	−3.21
1369581_at	phosphatidylethanolamine N-methyltransferase	Pemt	NM_013003	25511	3.66	−4.39
1377165_at	Transcribed locus	—	BF420664		3.64	−2.74
1386681_at	similar to hypothetical protein FLJ20014	LOC497934	BF567176	497934	3.58	−2.11
1379025_at	retinoblastoma-associated factor 600	Rbaf600	BF392966	313658	3.55	−2.78
1394838_at	amine oxidase, copper containing 3	Aoc3	AI136390	29473	3.52	−2.99
1396621_at	EST	—	BF408910		3.48	−2.24
1371194_at	tumor necrosis factor alpha induced protein 6	Tnfaip6	AF159103	84397	3.39	−2.46
1393852_at	similar to Hypothetical protein MGC25614	RGD1304576	BE115334	288659	3.37	−3.24
1385813_at	similar to RIKEN cDNA 2010001M09	RGD1310251	AI069938	291675	3.35	−2.91
1369576_at	taste receptor, type 2, member 105	Tas2r105	NM_023999	78985	3.33	−2.34
1380703_at	Tyrosine kinase, non-receptor, 1	—	BM386391	303247	3.16	−2.99
1368923_at	endothelin converting enzyme-like 1	Ecel1	AB023896	60417	3.14	−2.47
1389248_at	galactokinase 1	Galk1_	AI548699	287835	3.14	−3.89
(B) Genes that increase in expression
1383794_at	EST	—	AI575277		8.23	2.36
1397583_at	Transcribed locus	—	BF394946		6.06	2.10
1368383_at	neuropeptide FF-amide peptide precursor	Npff	NM_022586	60337	5.33	3.97

^aThe genes shown in this table are identified as preferentially expressed since they are expressed at 3 fold greater levels in the SON versus total hypothalamus (expression ratio), and have about two fold changes in expression during chronic hypoosmolarity. A, Genes were decreased (as indicated by a negative sign) in the hypoosmolar compared to the control condition. B, Genes were increased 2 fold in hypoosmolar compared to the control condition. Differences in gene expression were measured by comparative analysis of laser captured control rat SONs (n=3/group) against a rat hypothalamus reference (n=10), with triplicate array hybridization for each. A significant change in differential expression was accepted only for genes that passed our quality criteria (see methods) including significant p values (p < 0.05) between control and hypoosmolar in the T-test analysis, and fold changes more than 2 fold. The gene names represent the annotations taken from the Affymetrix database.

Microarray Data Analysis

Separate microarrays (n=3) were hybridized using independently generated cRNA probes from each sample. In order to normalize between the Rat 230 A and B and Rat 230 2.0 chips, the 2.0 arrays were split into their A and B components. The A and B chips were normalized separately to the mean value of each gene using the Loess transformation (Workman et al., 2002) in statistical package R (www.cran.r-project.org). Then the A and B chips were then combined into a single file and overlapping genes were averaged. Pearson's correlation coefficient was used to check for data quality after normalization (all replicate array pairs had a correlation coefficient >0.91). T-tests and fold changes were calculated between groups and genes with a p-value <0.05 or with a fold change >2 were used for further analysis as indicated in the results section.

For correlations of data with gene function, gene annotation, etc, the following databases and URLs were used: 1) Rat Genome Database http://rgd.mcw.edu/, for searching for specific types of genes that we are interested in, such as transcription factors, signal transduction, nuclear factors and proteins, glia and astrocyte, secretion and cytoskeletal genes, and 2) DAVID http://david.niaid.nih.gov/david/version2, for gene annotation and function prediction.

RESULTS

Hypoosmolality and Gene Expression in the SON

The physiological effects of the hypoosmolar treatments on the rats that we studied were previously reported (see Table I in Mutsuga et al., 2005), and were, as expected, that there was a significant decrease in the plasma osmolality, sodium concentration and body weight in the hypoosmolar rats as compared to normoosmolar controls. We used the SON/hypothalamus expression ratio to identify genes that are preferentially expressed in the SON. The SON/hypothalamus expression ratio is defined as the normalized value of the expression level for a particular gene in the SON sample, divided by the normalized value of this gene in the hypothalamic reference sample (Mutsuga et al., 2004). This then compares the SON expression versus hypothalamus, a tissue that also is enriched in processes associated with peptidergic secretion, thus reducing the impact of this factor on the data. First, we evaluated all 31,099 genes to identify those with expression ratios greater than 3, then we determined which of these genes were down- regulated or up- regulated at least 2 fold. The genes that fit these criteria and also had statistically significant changes are shown in Table I. Table I(A) shows a list of 43 genes that were down regulated following hypoosmolality. Table I(B) shows 3 genes with >2 fold up-regulation in expression following hypoosmolar treatment. We also put 3 genes in Table I(A) that met only one criterion, they had very high expression ratios (>9), but were down regulated less than 2 fold. These are Galanin, similar to hypothetical protein FLJ20037 and an EST (BF 289229). Among the genes shown in Table I, we particularly note proprotein convertase subtilisin/kexin type 1, C1q domain containing 1, similar to encephalopsin, Rho, GDP dissociation inhibitor (GDI) beta, tyrosine hydroxylase and neuropeptide FF-amide peptide precursor, since they are consistently found changing in the SON in various array studies of osmotic perturbations (Mutsuga et al., 2005; Hindmarch, 2006).

Table I.

Changes in Gene Expression in Rat SONs during Chronic Hypoosmolarlity^a

Probe set ID	Gene title	Gene symbol	GenBank ID	Entrez gene	Expression ratio	Fold change
(A) Genes that Decrease in Expression
1368559_at	proprotein convertase subtilisin/kexin type 1	Pcsk1	NM_017091	25204	31.33	−4.23
1373260_at	C1q domain containing 1	—	AI412606	361303	20.50	−3.32
1391923_at	EST	—	BG376838		18.45	−5.22
1387088_at	Galanin	Gal	NM_033237	29141	12.97	−1.75
1387284_at	Dihydropyrimidinase	Dpys	NM_031705	65135	12.77	−10.29
1394940_at	similar to hypothetical protein FLJ20037	RGD1311381	BI294811	300870	12.72	−1.60
1376854_at	EST	—	BE110028		11.84	−2.81
1374375_at	CDNA clone IMAGE:7452995	—	AI406903		11.66	−4.64
1368548_at	Solute carrier family 12, member 1	Slc12a1	NM_019134	25065	10.62	−9.98
1376513_at	EST	—	BE098803		10.37	−2.88
1382659_at	Transcribed locus	—	BF289229		9.95	−1.91
1376836_at	EST	—	BF419655		9.86	−3.56
1393373_at	similar to encephalopsin	LOC498289	BI289640	498289	9.01	−5.74
1370055_at	RAB3D, member RAS oncogene family similar to methylenetetrahydrofolate dehydrogenase (NAD) (EC 1.5.1.15)/methenyltetrahydrofolate	Rab3d	M83681	140665	8.49	−2.07
1372808_at	cyclohydrolase (EC 3.5.4.9) precursor – mouse	LOC313410	AW251324	313410	7.48	−2.36
1387662_at	synaptotagmin 4	SYt4	L38247	64440	7.46	−4.94
1368416_at	integrin binding sialoprotein	Ibsp	NM_012587	24477	6.74	−3.23
1371212_at	neuregulin 1	Nrg1	U02315	112400	6.65	−2.14
1391095_at	matrix metalloproteinase 19	Mmp19	BI294977	304608	6.04	−2.02
1387075_at	tyrosine hydroxylase	Th	NM_012740	25085	5.80	−18.42
1373128_at	reticulocalbin 3, EF-hand calcium binding domain	Rcn3	BI278379	494125	5.36	−2.83
1369672_at	arachidonate 5-lipoxygenase activating protein	Alox5ap	NM_017260	29624	5.13	−4.44
1368162_at	cystatin E/M	Cst6	NM_133566	171096	5.12	−2.39
1396392_at	Dynactin 6 (predicted)	—	BF403099	290798	4.99	−2.07
1373881_at	Rho, GDP dissociation inhibitor (GDI) beta	Arhgdib	BF285771	362456	4.76	−4.73
1385691_at	EST	—	BI293900		4.69	−3.07
1369821_at	cyclic nucleotide gated channel beta 1	Cngb1	AF068572	83686	4.35	−2.33

Transcription Factors and other Genes Especially Expressed in SON Following Hypoosmolality

One of the major goals of this study was to identify genes that might regulate global transcription and translation in the MCNs of osmotically stressed animals, so we specifically searched for genes that were annotated as transcription/translation related genes in the Rat Genome Database (http://rgd.mcw.edu/). Table II lists 31 statistically significant transcription factors, nuclear factors and protein-related genes corresponding to the direction and extent of their fold changes. Among these 31 genes, we found 13 genes were up regulated, and 2 of them, Activating transcription factor 2 and cubilin (intrinsic factor-cobalamin receptor), had fold changes greater than 2. Among the18 genes that were down regulated, there are 7 genes had fold changes greater than 2. These 9 genes with fold change >2 might be involved in the regulation of the expression of neuropeptides in the MCNs.

Table II.

Changes in Gene Expression of Transcription Factors and Nuclear Factors in SON

Probe set ID	Gene title	Gene symbol	GenBank ID	Entrez gene	Fold change	Expression ratio
1395771_at	Activating transcription factor 2	Atf2	BF288181	81647	2.79	−2.99
1387037_at	cubilin (intrinsic factor-cobalamin receptor)	Cubn	AF022247	80848	2.19	−1.59
1377845_at	lymphoid nuclear protein related to AF4 platelet derived growth factor receptor, alpha polypeptide	Laf4	BM391291	363220	1.97	−1.51
1370941_at	Acidic (leucine-rich) nuclear phosphoprotein 32	Pdgfra	AI232379	25267	1.86	−2.69
1371986_at	family, member A	Anp32a	AI576652	25379	1.75	−2.11
1369007_at	nuclear receptor subfamily 4, group A, member 2	Nr4a2	L08595	54278	1.74	−3.66
1382558_at	transcription factor 3	Tcf3	AI045720	312451	1.58	1.34
1373032_at	musculoskeletal, embryonic nuclear protein 1	Mustn1	AW251450	290553	1.55	1.30
1371104_at	sterol regulatory element binding factor 1	Srebf1	AF286470	78968	1.55	−1.07
1370213_at	nuclease sensitive element binding protein 1	Nsep1	BI282111	29206	1.48	−1.59
1370224_at	signal transducer and activator of transcription 3	Stat3	BE113920	25125	1.45	3.04
1368835_at	signal transducer and activator of transcription 1	Stat1	AW434718	25124	1.36	−1.13
1370965_at	transcription factor 8 (represses interleukin 2 expression)	Tcf8	BG381660	25705	1.36	1.30
1384487_at	doublesex and mab-3 related transcription factor 2	Dmrt2	BF557192	309430	−1.35	1.37
1389733_at	methionine-tRNA synthetase	Mars	BM384125	299851	−1.40	1.30
1369289_at	hepatocyte nuclear factor 4, alpha	Hnf4a	NM_022180	25735	−1.46	−1.16
1381427_at	forkhead box C2	Foxc2	AI111819	171356	−1.53	1.07
1370983_at	POU domain, class 6, transcription factor 1	Pou6f1	AW523687	116545	−1.58	1.21
1383821_at	Interleukin enhancer binding factor 3	Ilf3	AA899489	84472	−1.59	1.35
1389506_at	Fas-associated factor 1	Faf1	BI298458	140657	−1.60	2.34
1374034_at	cysteinyl-tRNA synthetase	Cars	BG379410	293638	−1.70	2.25
1367624_at	activating transcription factor 4	Atf4	NM_024403	79255	−1.79	1.92
1391257_at	Zinc finger protein 148	Znf148	BF412175	58820	−1.82	1.55
1376418_at	isoleucine-tRNA synthetase	Iars	AA859497	306804	−1.99	1.74
1370777_at	eosinophil-associated, ribonuclease A family, member 11	Ear11	D88586	192264	−2.06	1.73
1368775_at	gonadotropin inducible ovarian transcription factor 1	Giot1	NM_133563	171090	−2.24	1.88
1369244_at	aryl hydrocarbon receptor nuclear translocator	Arnt	NM_012780	25242	−2.26	1.39
1368914_at	runt related transcription factor 1	Runx1	NM_017325	50662	−2.39	−1.02
1394068_at	Kruppel-like factor	Klf2	BF288243	306330	−2.95	1.46
1368416_at	integrin binding sialoprotein	Ibsp	NM_012587	24477	−3.23	6.74
1375043_at	FBJ murine osteosarcoma viral oncogene homolog	Fos	BF415939	314322	−3.55	1.40

Note. Genes shown in this table are transcription factors and nuclear factors that significantly changed in the hypoosmolar compared to the control condition, independent of considerations of expression ratios (which are also shown here). Differences in gene expression were measured by comparative analysis of laser captured control rat SONs (n=3/group) against a rat hypothalamus reference (n=10), with triplicate array hybridization for each. A significant change in differential expression was accepted only for genes that passed our quality criteria (see methods) and had significant p values (p < 0.05) between control and hypoosmolar in the T-test analysis. The fold changes that are negative, represent decreases in gene expression, and the expression ratios that are negative represent genes that are higher in expression in hypothalamus versus SON. The gene names represent the annotations taken from the Affymetrix database.

We also queried the signal transduction, secretion and glial/astrocyte-related genes in the Rat Genome Database and list these in Table III.

Table III.

Changes in Gene Expression of Selected Genes in SON

Probe set ID	Gene title	Gene symbol	GenBank ID	Entrez gene	Fold change	Expression ratio
(A) Signal transduction-related genes
1375043_at	FBJ murine osteosarcoma viral oncogene homolog	Fos	BF415939	314322	−3.55	1.40
1381051_at	Phospholipase D2	Pld2	AW524545	25097	−3.09	1.40
1371218_at	neuregulin 1	Nrg1	AF194443	112400	−3.05	2.60
1397014_at	Phosphodiesterase 9A	Pde9a	BF419776	191569	−1.56	1.15
1370035_at	Kirsten rat sarcoma viral oncogene homologue 2 (active)	Kras2	NM_031515	24525	−1.35	1.25
1398783_at	G protein pathway suppressor 1	Gps1	NM_053969	117039	−1.20	1.42
1391127_at	Cell division cycle 42 homolog (S. cerevisiae)	Cdc42	BF400818	64465	−1.19	−1.32
1370121_at	adducin 1 (alpha)	Add1	NM_016990	24170	1.47	−1.83
1374872_at	RAS guanyl releasing protein 2 (calcium and DAG-regulated)	Rasgrp2	AW532114	361714	1.69	−3.98
1392520_at	deleted in liver cancer 1	Dlc1	AI136944	58834	2.12	−2.27
1373081_at	Brain-specific angiogenesis inhibitor 1-associated protein 2	Baiap2	AI105000	117542	2.43	−8.74
1379396_at	engulfment and cell motility 1, ced-12 homolog (C. elegans)	Elmo1	AI101913	361251	2.47	−4.84
1379175_at	Protein phosphatase 3, catalytic subunit, alpha isoform	Ppp3ca	BF388224	24674	2.80	−5.98
1371237_at	Metallothionein	Mt1a	AF411318	24567	4.21	1.02
(B) Secretion-related genes
1387088_at	galanin	Gal	NM_033237	29141	−1.75	12.97
1387907_at	inositol 1,4,5-triphosphate receptor 1	Itpr1	J05510	25262	−1.68	−6.27
1370132_at	FK506 binding protein 1b	Fkbp1b	NM_022675	58950	−1.53	2.16
1374353_at	actin alpha cardiac 1	Actc1	AI711147	29275	−1.18	1.37
1367992_at	secretory granule neuroendocrine protein 1	Sgne1	NM_013175	25719	1.27	1.24
1369112_at	Cholinergic receptor, muscarinic 3	Chrm3	M18088	24260	2.20	−1.41
1387811_at	angiotensinogen	Agt	NM_134432	24179	2.25	1.52
1387595_at	gastric intrinsic factor	Gif	NM_017162	29319	2.47	−1.64
1368113_at	trefoil factor 2 (spasmolytic protein 1)	Tff2	NM_053844	116592	2.84	−2.11
1384240_at	angiotensin II receptor, type 1 (AT1A)	Agtr1a	BI290306	24180	4.82	−2.35
1368383_at	neuropeptide FF-amide peptide precursor	Npff	NM_022586	60337	5.33	3.97
(C) Glial/astrocyte-related genes
1368565_at	solute carrier family 1 (glial high affinity glutamate transporter), member 3	Slc1a3	NM_019225	29483	1.40	−1.06
1368104_at	tetraspan 2	Tspan2	NM_022589	64521	1.47	−4.43
1368353_at	glial fibrillary acidic protein	Gfap	NM_017009	24387	1.53	1.21
1381621_at	Hypothetical gene supported by NM_013150	Nrcam	BF386153	25691	2.19	−4.14

Note. Genes shown in this table represent signal transduction, secretion and glial or astrocyte related genes that changed significantly in the hypoosmolar as compared to the control conditions. Differences in gene expression were measured by comparative analysis of laser captured control rat SONs (n=3/group) against a rat hypothalamus reference (n=10), with triplicate array hybridization for each. A significant change in differential expression was accepted only for genes that had significant p values (p < 0.05) between control and hypoosmolar in the T-test analysis. The fold changes that are negative, represent decreases in gene expression, and the expression ratios that are negative represent genes that are higher in expression in hypothalamus versus SON. The gene names represent the annotations taken from the Affymetrix database.

Comparisons of Gene Expression Data Between the Affymetrix Array and the cDNA Array Analysis, and Between Chronic Hyperosmolar Versus Hypoosmolar Treatment

Table IV lists the genes that changed significantly following hypoosmolality both in this study that used an oligonucleotide-based Affymetrix array and from our previous use of a spotted cDNA array (Mutsuga et al., 2005). There are many similar findings, including C1q domain containing 1, tyrosine hydroxylase, similar to encephalopsin, cell growth regulator with EF hand domain 1, rho, GDP dissociation inhibitor (GDI) beta and moesin, which were found to be significantly down-regulated in response to chronic hypoosmolality.

Table IV.

Comparison of Expression Ratio and Gene Expression Changes of Selected Genes using Oligonucleotide Array versus Spotted Array Platforms

			Affymetrix Oligocucleotide Array-deriveda				Spotted cDNA array-deriveda
Probe Set ID	Gene Title	Gene Symbol	GenBank ID	Entrez Gene	Expression Ratio	Fold Change	Clone ID	Expression ratio	Fold Change
1373260_at	C1q domain containing 1	—	AI412606	361303	20.50	−3.32	BE946020	24.6	−4.76
1393373_at	similar to encephalopsin	LOC498289	BI289640	498289	9.01	−5.74	AW492784	6.2	−4.59
1387075_at	tyrosine hydroxylase	Th	NM_012740	25085	5.80	−18.42	C85951	3.5	−5.59
1370361_at	cell growth regulator with EF hand domain 1	Cgref1	U66470	245918	4.87	−1.84	AW045201	6.1	−3.31
1373881_at	Rho, GDP dissociation inhibitor (GDI) beta	Arhgdib	BF285771	362456	4.76	−4.73	AI842572	3.4	−4.63
1371575_at	Moesin	Msn	BF281185	81521	4.68	−1.44	C79581	9.8	−2.23
1367681_at	CD151 antigen	Cd151	NM_022523	64315	3.21	−1.30	AW554580	5.3	−2.64
1398871_at	ribosomal protein L17	RGD:1303019	BG671311	291434	1.81	1.43	AW537247	3.7	2.07
1387291_at	inter-alpha trypsin inhibitor, heavy chain 3	Itih3	NM_017351	50693	1.24	1.57	AW047821	4.5	2.06
1383181_at	DnaJ (Hsp40) homolog, subfamily C, member 9	Dnajc9	AI704947	364240	1.09	1.58	AU020082	3.5	2.29
1368418_at	ceruloplasmin	Cp	AF202115	24268	−1.52	2.59	BE984774	3.5	3.04
1370342_at	potassium channel, subfamily K, member 2	Kcnk2	AF385402	170899	−1.53	1.49	AI848993	4.4	2.90

^aGenes shown in the left part of this table are derived from this Affymetrix oligonucleotide array study, changed significantly in the hypoosmolar compared to the control condition. A significant change in differential expression was accepted only for genes that passed our quality criteria (see methods) including significant p values (p < 0.05) between control and hypoosmolar in the T-test analysis. The gene names represent the annotations taken from Affymetrix database

^bThe genes obtained from the spotted cDNA array study (Mutsuga et al., 2005) shown in the right part of this table are exactly the same genes as those that were identified in the Affymetrix analysis. A significant change in differential expression was accepted only for genes that passed our quality criteria (see methods) including significant p values (p < 0.05) between control and hypoosmolar in the ANOVA analysis with fold changes more than 2 fold. The gene names represent the annotation results from Blast searches.

Finally, we compared our data using the hypoosmolality paradigm with a recent report describing gene expression changes during hyperosmolality (Hindmarch, 2006). In this comparison, both of these studies used the Affymetrix oligonucleotide array platform. Seventeen genes changed significantly following both hyperosmolar and hypoosmolar chronic treatment and these are shown in Table V. Among these 17 genes, 15 were down regulated during hypoosmolar and up regulated during hyperosmolar conditions, and 2 were up regulated during chronic hypoosmolality and down regulated during chronic hyperosmolality. The fact that these genes responded in opposite directions under the two opposite experimental conditions indicated that these genes are specifically associated with osmoregulation in the MCNs, as opposed to a non-specific stress phenomenon.

Table V.

Comparison of Selected Gene Expression Changes in Hypo- and Hyperosmolar Rat SONs^a

					Hypoosmolar		Hyperosmolar
Probe set ID	Gene title	Gene symbol	GenBank ID	Entrez gene	Expression ratio	Fold change	Fold change
1387075_at	tyrosine hydroxylase	Th	NM_012740	25085	5.80	−18.42	3.09
1368359_at	VGF nerve growth factor inducible	Vgf	NM_030997	29461	1.38	−17.32	3.58
1387284_at	Dihydropyrimidinase	Dpys	NM_031705	65135	12.77	−10.29	14.68
1368548_at	solute carrier family 12, member 1	Slc12a1	NM_019134	25065	10.62	−9.98	4.16
1393373_at	similar to encephalopsin	LOC498289	BI289640	498289	9.01	−5.74	6.11
1373881_at	Rho, GDP dissociation inhibitor (GDI) beta	Arhgdib	BF285771	362456	4.76	−4.73	4.41
1369581_at	phosphatidylethanolamine N-methyltransferase	Pemt	NM_013003	25511	3.66	−4.39	3.96
1384970_at	similar to RIKEN cDNA 5330437I02 gene	LOC361340	BM389861	361340	3.76	−3.75	4.94
1375043_at	FBJ murine osteosarcoma viral oncogene homolog	Fos	BF415939	314322	1.40	−3.55	4.34
1373260_at	C1q domain containing 1	—	AI412606	361303	20.50	−3.32	3.52
1368491_at	DNaseII-like acid DNase	Dlad	NM_021664	59296	1.61	−2.79	3.38
1371194_at	tumor necrosis factor alpha induced protein 6	Tnfaip6	AF159103	84397	3.39	−2.46	12.74
1369164_at	transient receptor potential cation channel, subfamily C, member 4	Trpc4	AF288407	84494	1.82	−2.29	3.16
1368775_at	gonadotropin inducible ovarian transcription factor 1	Giot1	NM_133563	171090	1.88	−2.24	13.67
1384448_at	similar to RIKEN cDNA 1700045I19	LOC317486	AA963712	317486	6.31	−1.90	3.67
1369704_at	X transporter protein 3	RGD:621651	NM_133296	113918	−2.28	1.45	−2.73
1368641_at	wingless-related MMTV integration site 4	Wnt4	NM_053402	84426	−8.69	2.40	−2.69

^aAll data in this table were obtained using the Affymetrix oligonucleotide array platform. The hypoosmolar data are from this report, and the hyperosmolar data from Hindmarch et al. (2006).

DISCUSSION

In this study we use LMD and high-density oligonucleotide Affymetrix arrays to identify a set of SON-enriched, osmotically-regulated genes. Specifically our goal was to identify genes in the SON whose expression was altered by systemic hypoosmolar conditions. We selected the hypoosmolar rat model for this study because there is a dramatic reduction in OT and VP gene expression and secretion under this condition (Robinson et al., 1990; Verbalis, 1993a; Glasgow et al., 2000; Mutsuga et al., 2005), and hence, we anticipated that important regulatory genes in the SON would undergo similar dramatic changes. Several previously detected, as well as novel genes in the SON that were changed in expression during hypoosmolality were identified. Among these genes, VP, OT, dynorphin, galanin, and tyrosine hydroxylase were previously known to be present and osmotically regulated in the SON (Burbach et al., 2001), but most of the other genes found in this study that were preferentially expressed in the SON and greatly changed in expression during hyposmolality were not previously known to be expressed in the SON. Earlier reports of changed gene expression profiles in the SON during hypoosmolar (Mutsuga et al., 2005) and hyperosmolar (Ghorbel et al., 2003; Hindmarch, 2006) conditions have used different array platforms, and we chose to use a high density oligonucleotide rat array platform in this study, in part, to facilitate comparisons of data between these studies.

We identify preferentially expressed genes in the SON in this paper by using an approach that we previously described (Mutsuga et al., 2004; Mutsuga et al., 2005), in which the expression levels of genes in the SON are compared to their expression levels in total hypothalamus. The ratios of the SON to the hypothalamic signals are then taken as a measure of the preferential expression of specific genes in the MCNs in the SON relative to the hypothalamus, and we refer to this value as the “expression ratio”. We have previously validated selected genes identified on the array for their high expression ratios and also their changes in expression by quantitative in situ hybridization histochemistry (ISHH) (Mutsuga et al., 2004; Mutsuga et al., 2005). Many of these genes were also found in the present study (see Table IV), and this serves as a form of validation of the array data. Others were confirmed by their being previously known from the literature (Burbach et al., 2001) as being present in the SON and osmotically regulated. Overall, this strategy resulted in the identification of 46 preferentially expressed MCN genes in the SON, which also changed more than two-fold in expression during hypoosmolality (see Table I). These genes may be valuable candidates not only for the study of cell-specific gene expression in the magnocellular neurons, but also for future studies of their regulation by osmotic perturbations.

Using the Affymetrix array platform together with the hyposmolality experimental paradigm, we found many genes that were altered during this osmotic perturbation. Table I shows a list of those genes that were selectively expressed in the SON (i.e., had expression ratios >3) and that were robustly changed in expression (more than two fold) in hypoosmolar conditions. These included genes that had been described previously in studies using the hyposmolality paradigm and spotted cDNA arrays (Table IV, and Mutsuga et al., 2005), and also in studies employing a hyperosmolality paradigm and Affymetrix oligonucleotide array analysis (Table V, and Hindmarch, 2006). It is especially encouraging with respect to the latter genes, that the directions of the changes were in accordance with the directions of the osmotic perturbations, i.e., when the expression change was a decrease in expression in hyposmolality, the reverse was true for hyperosmolality, and vice versa (Table V). In this regard, it is interesting to note that the neuropeptide FF peptide precursor gene that was identified as having an expression ratio of 5.3, and an increase of about 4 fold in hypoosmolar conditions (Table I(B)I), had previously been reported by ISHH analysis to decrease during hyperosmolar conditions (Kalliomaki and Panula, 2004). We focused our attention on those genes which had high expression ratios (>3) since earlier studies using quantitative ISHH had validated their preferential expression in the SON (Mutsuga et al., 2004; Mutsuga et al., 2005) and those genes, which also significantly changed in gene expression during hypoosmolar conditions. Genes fulfilling these criteria are shown in Table I.

Of particular interest, are genes that were not previously known to be present in the SON prior to array studies. These include C1q domain containing 1, Rho GDP dissociation inhibitor (Rho GDI beta), encephalopsin, VGF nerve growth factor inducible, dihydropyrimidase, and solute carrier 12 (member 1). Each of these genes had either very high expression ratios and/or very large changes in gene expression in response to osmotic perturbation, and all of these genes decreased in expression in the SON in the hypoosmolar state. One of these genes, C1q domain containing 1 (also known as EEG-1), has been identified as a novel growth-inhibitor gene involved in terminal differentiation of human erythrocytes, where it is designated as EEG-1 (Aerbajinai et al., 2004). The fact that this gene's expression is suppressed in the hypoosmolar condition suggests that it may also be involved in regulation of the VP and/or OT MCNs. Bioinformatics analysis of its sequence has identified this gene as a new member of a family of genes containing a C1q globular domain, known as the C1q/TNF superfamily (Kishore and Reid, 1999). The family of genes containing a C1q globular domain includes the complement protein C1q itself, and several other proteins that are not related to complement function (Kishore and Reid, 1999, 2000). Other predicted motifs in the C1q domain containing gene 1 sequence include several putative posttranslational phosphorylation motifs, one tyrosine-kinase phosphorylation site, one cAMP- and cGMP-dependent protein kinase phosphorylation site, several casein kinase II and protein kinase C phosphorylation sites, and the N-terminal nuclear localization sequence (Aerbajinai et al., 2004).

Other genes of interest are the Rho GTP dissociation inhibitor beta (Rho GDI) regulates the GTP bound and GDP bound state cycle of Rho GTP binding protein by decreasing the rate of dissociation from Rho GTPases, which play important roles in neuronal morphogenesis (Ostrowski et al., 1992). Previous results from our laboratory showed that MCNs dramatically adjust their cell volumes during both hyper- and hypoosmolar conditions (Zhang et al., 2001). It is possible that the Rho GDI may be involved in these morphological transformations. Encephalopsin (also called Panopsin) is a member of the rhodopsin/ G-protein coupled receptor gene (GPCR) superfamily. It is specifically expressed in the mammalian brain (Bito, 2003; Aerbajinai et al., 2004). In a previous study, using ISHH we found a specific, strong hybridization in the SON and the subfornical region (SFO) in the rat brain, which confirmed the array's predictions of a prominent increase and decrease in expression in these nuclear during the hyperosmolar and hypoosmolar states, respectively (data not shown).

Many of the genes in Table I which have high expression ratios, appear to have relatively mundane functions. For example, dihydropyrimidase is known to be a pyrimidine degradation enzyme (van Kuilenburg et al., 2004), but may play other roles as well (van Gennip and van Kuilenburg, 2000). Similarly, solute carrier family 12, member 1 is known primarily as an electroneutral cation-choride co-transporter in the central neurons system (Mercado et al., 2004), and has also been identified as a susceptibility gene for diabetic neuropathy (Tanaka and Babazono, 2005). Another enzyme, phosphofructokinase-B was found expressed at an expression ratio of 3.8, and its expression decreased more than 2.47-fold during hypoosmolar conditions (Table I). This is notable in that another member of this gene family, phosphofructokinase-C, was found in our previous single cell-gene library differential hybridization screen study (Yamashita et al., 2002) to be selectively expressed in OT MCNs. PFK catalyzes the ATP-dependent conversion of fructose 6-phosphate to fructose 1, 6-bisphosphate, which is the rate-limiting reaction in glycolysis. Three PFK subunit isozymes have been identified (PFK-C, PFK-B (L), and PFK-M), and each isozyme subunit has distinct catalytic properties and specific sensitivity to hormonal regulators (Dunaway, 1983; Foe and Kemp, 1985; Gekakis et al., 1994; Gunasekera and Kemp, 2000). The specific catalytic properties of the holoenzyme depend on the specific isozyme subunit composition of the tetramer (Kemp and Foe, 1983; Dunaway and Kasten, 1985). PFK-C is predominantly located in the brain and in neuroendocrine tissues (Gekakis et al., 1994). PFK-M and PFK-B (L) are also located in the brain (Gunasekera and Kemp, 2000), resulting in a potentially complex mixture of homo- and heterotetramers. In view of the complexity of these enzymes’ regulation (Gekakis et al., 1994; Mhaskar and Dunaway, 1995; Gunasekera and Kemp, 2000), additional studies on the expression patterns of the other isozyme subunits present in the MCNs will be necessary to consider the possible tetrameric arrangements that might be present in these cells, and their potential metabolic consequences.

It should be recognized that the functional identification of the above genes, and most of the others on the array are based on PubMed, Gene Ontology, and other databases, which are general and may not reflect the actual functions of these genes in a specific neuron population, such as the MCNs in the SON. The above is true even when the gene, such as vgf (see Table V) is identified as a neuronal and endocrine peptide precursor and a neurotrophin (Sugaya et al., 1998; Levi et al., 2004). Other caveats to consider are that our criteria of >3 expression ratio and a two-fold change in expression to identify interesting genes may be too stringent, and that genes with smaller values for these parameters may also represent important genes. Note that the transcription factors we found in the SON (Table II), do not largely fall into this large expression ratio or fold-change category, but nevertheless might be of great importance in the regulation of OT and VP gene expression. Indeed, when we interrogated our array data for these values, for the known and believed to be important transcription factors in the SON, CREB and CREM, we found that their expression ratios were much less than 3 and their fold changes were barely above two (data not shown). With this in mind, one of the important values of this and other array studies is that they provide an archive of genes that are present in the tissues of interest. This is especially true for tissues that are very carefully isolated for RNA extraction, such as we have done here for the SON by using laser microdissection. For this reason, we have made the data from this study available to all investigators on the NINDS internet website URL address: http://data.ninds.nih.gov/Gainer/Publications

Even under these near optimum circumstances, where the dissected tissue contains primarily only two neuronal phenotypes, the OT and VP MCNs, it should be noted that there also are blood vessels, blood cells, and specialized astrocytes in the SON that whose genes will be represented in our data. Therefore, an important part of array analysis in the CNS is to do quantitative ISHH to identify the specific cell types that express the genes being identified.

In this regard, it should be noted that, unlike C1q, Rho, and encephalopsin, changes of expression of the other genes discussed above, and for most of the genes listed in Table I have not been confirmed by methods other than the arrays, and this validation remains to be done. Furthermore, while these genes represent potential candidates for being molecular participants in the regulation of OT and VP gene expression, there is no experimental evidence that these do in fact underlie osmoregulation mechanisms in the MCNs. To approach this issue it will first be necessary to obtain full-length clone information about these genes that appear to be so selectively expressed in the MCNs. For this purpose, we have generated a SON/PVN-enriched cDNA library that was derived from mRNAs obtained using the Palkovits punch technique on freshly dissected rat hypothalamus tissue, and we are currently isolating several of the SON-specific gene's cDNAs for sequence analysis and as a database for RNAi experiments.

We conclude that the using of laser microdissection together with microarray analysis provides a powerful and systematic approach for the identification of genes in the MCNs that are both preferentially expressed and substantially modified in their expression during osmotic perturbations. This information may provide important molecular candidates for gene regulation in the HNS. Future application of siRNA inhibitors of these candidate genes in the SON may begin to uncover the networks of intracellular molecular mechanisms that underlie the osmotic regulation of OT and VP gene expression in the HNS.

Abbreviations:

MCN

magnocellular neuron

SON

supraoptic nucleus

HNS

hypothalamoneurohypophysial

LMD

laser microdissection

OT

oxytocin

VP

vasopressin