Androgen Control of Gene Expression in the Mouse Meibomian Gland

Abstract

purpose. In prior work, it has been found that the meibomian gland is an androgen target organ, that androgens modulate lipid production within this tissue, and that androgen deficiency is associated with glandular dysfunction and evaporative dry eye. This study’s purpose was to test the hypothesis that the androgen control of the meibomian gland involves the regulation of gene expression.

methods. Meibomian glands were obtained from orchiectomized mice that were treated with placebo or testosterone for 14 days. Tissues were processed for the analysis of differentially expressed mRNAs by using gene bioarrays, gene chips, and real-time PCR procedures. Bioarray data were analyzed with GeneSifter software (VizX Labs LLC, Seattle, WA).

results. The results show that testosterone influenced the expression of more than 1590 genes in the mouse meibomian gland. This hormone action involved a significant upregulation of 1080 genes (e.g., neuromedin B), and a significant downregulation of 518 genes (e.g., small proline-rich protein 2A). Some of the most significant androgen effects were directed toward stimulation of genes associated with lipid metabolism, sterol biosynthesis, fatty acid metabolism, protein transport, oxidoreductase activity, and peroxisomes.

conclusions. These findings demonstrate that testosterone regulates the expression of numerous genes in the mouse meibomian gland and that many of these genes are involved in lipid metabolic pathways.

Recently, researchers have demonstrated that androgens regulate the meibomian gland, ^{1 2 3} which is the primary tissue involved in maintaining tear film stability and preventing tear film evaporation. ^{4 5 6} Androgens modulate meibomian gland function, improve the quality and/or quantity of lipids produced by this tissue, and promote the formation of the tear film’s lipid layer. ^{1 7 8} Moreover, androgen deficiency, such as occurs during menopause, aging, Sjögren’s syndrome, complete androgen insensitivity syndrome, and the use of antiandrogen medications, ^{9 10 11} is associated with meibomian gland dysfunction, altered lipid profiles in meibomian gland secretions, tear film instability, and evaporative dry eye. ^{2 3 12 13 14} These findings are very significant, given that scant information exists concerning the physiological control of this tissue and that meibomian gland dysfunction is the major cause of evaporative dry eye syndromes throughout the world. ¹⁵  

However, the mechanism(s) underlying this androgen influence on meibomian gland lipogenesis and function is unknown. It is possible that the hormonal regulation of this tissue is analogous to that of sebaceous glands, given that the meibomian gland is a large sebaceous gland and that androgens control the development, differentiation, and lipid production of these glands throughout the body. ^{16 17} Androgen effects on sebaceous glands are mediated primarily through hormone binding to androgen receptors within acinar cell nuclei. ^{17 18 19} This receptor interaction leads to increased gene transcription and the elaboration of proteins that stimulate the synthesis and secretion of lipids. ^{17 18 19 20} In many sebaceous glands, androgen activity is also enhanced by, or dependent on, the presence of 5α-reductase, an enzyme that converts testosterone into the potent androgen, 5α-dihydrotestosterone. ¹⁷  

Consistent with this possibility are the findings that meibomian glands of males and females contain androgen receptor mRNA, androgen receptor protein within acinar epithelial cell nuclei, and the mRNAs for both types 1 and 2 5α-reductase. ^{21 22} Given these observations, we hypothesized that the androgen control of the meibomian gland, as with other sebaceous glands, involves the regulation of gene expression. The purpose of the present study was to test this hypothesis. 

Materials and Methods

Animals and Hormone Treatment

Young adult BALB/c mice (n = 5–22/group), which had been orchiectomized at 8 to 9 weeks of age by veterinary surgeons, were purchased from Taconic Laboratories (Germantown, NY). Animals were housed in constant-temperature rooms with fixed light–dark intervals of 12 hours. Mice were allowed to recover from surgery for at least 9 days, were anesthetized intraperitoneally with ketamine and xylazine, and received subcutaneous implants of placebo (cholesterol, methyl cellulose, lactose)- or testosterone (10 mg)-containing pellets in the subscapular region. These pellets were obtained from Innovative Research of America (Sarasota, FL) and were designed for the slow, but continual, release of vehicle or physiological amounts of hormone over a 21-day period. After 2 weeks of treatment, mice were killed by CO₂ inhalation, and the upper- and lower-lid meibomian glands were removed under direct visualization with a biomicroscope and immediately frozen in liquid nitrogen. All studies with experimental animals were approved by the Institutional Animal Care and Use Committee of The Schepens Eye Research Institute and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Molecular Biological Procedures

To determine the effect of testosterone on meibomian gland gene expression, total RNA was isolated from tissues by using TRIzol reagent (Invitrogen Corp., Carlsbad, CA). When indicated, samples were also exposed to RNase-free DNase (Invitrogen), analyzed spectrophotometrically at 260 nm to determine concentration and examined on 6.7% formaldehyde/1.3% agarose (Gibco/BRL, Grand Island, NY) gels to verify RNA integrity. The RNA samples were then processed by utilizing several different technical approaches. 

The principle method to evaluate differential gene expression involved the use of CodeLink Uniset Mouse I Bioarrays (∼10,000 genes; Amersham, Piscataway, NJ). Before array studies, the integrity of glandular RNA preparations was further assessed with a RNA 6000 Nano LabChip with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). The RNA samples were then processed for CodeLink Bioarray hybridization, as previously described. ²³ Briefly, cDNA was synthesized from RNA (2 μg) with a CodeLink Expression Assay Reagent Kit (Amersham) and isolated with a QIAquick purification kit (Qiagen, Valencia, CA). After sample drying, cRNA was generated with a CodeLink Expression Assay Reagent Kit (Amersham), recovered with an RNeasy kit (Qiagen) and quantified with an UV spectrophotometer. Fragmented, biotin-labeled cRNA was incubated and agitated (300 rpm shaker) on a CodeLink Bioarray at 37°C for 18 hours. The Bioarray was then washed and exposed to streptavidin-Alexa 647. Bioarrays were scanned by using ScanArray Express software and a ScanArray Express HT scanner (Packard BioScience, Meriden, CT) with the laser set at 635 nm, laser power at 100%, and photomultiplier tube voltage at 60%. Scanned image files were examined by using CodeLink image and data analysis software (Amersham), which generated both raw and normalized hybridization signal intensities for each array spot. The ∼10,000 spot intensities on the microarray image were standardized to a median of 1. Normalized data, with signal intensities exceeding 0.50, were analyzed with GeneSifter software (VizX Labs LLC, Seattle, WA; vizxlabs.com). This program also produced gene ontology and z-score reports. These ontologies, which were organized according to the guidelines of the Gene Ontology Consortium (http://www.geneontology.org/GO.doc.html), ²⁴ included biological processes, molecular functions and cellular components. Statistical analysis of individual gene expression data was conducted with Student’s t-test (two-tailed, unpaired). Data were evaluated with and without log transformation. 

The data from the individual Bioarrays (n = 6) are accessible for downloading through the National Center for Biotechnology Information’s Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) via series accession number GSE1582. The data will also be available for evaluation through GeneSifter (http://genesifter.net/datacenter/). 

Differentially expressed mRNAs were also analyzed by using GEM 1 (>8,000 genes) and GEM 2 (>9,500 genes) gene chips (Incyte Genomics, Inc., St. Louis, MO). Poly(A) mRNA was isolated from meibomian gland RNA samples by using the MicroPoly(A) Pure mRNA Isolation Kit (Ambion, Inc., Austin, TX). The mRNA concentration was determined with a RiboGreen RNA Quantitation Kit (Molecular Probes, Eugene, OR), according to Incyte’s protocol. After designating mRNA samples (800 ng) for use with either cy3 or cy5 probes, preparations were suspended in TE buffer, placed in siliconized RNase-Free Microfuge Tubes (Ambion) and shipped on dry ice to Incyte for hybridization. Microarray data were sent electronically to the Harvard Center for Genomic Research (Cambridge, MA) and results were downloaded into the Resolver Gene Expression Data Analysis System, version 3.1 (Rosetta Inpharmatics, Kirkland, WA). This system displayed the sequence identification and description of all chip nucleotides, the signal strength of the treatment (i.e., testosterone) and control (i.e., placebo) channels, the relationship between the two channels in terms of ratio and fold change, the comparative P-value and information concerning various quality control fields. In addition, this system determined the error-weighted average ratios for each chip, and normalized data across chips, thereby permitting the combination of GEM 1 and 2 microarray results to achieve a stronger analysis of gene expression. The error model applied by Rosetta Resolver on Incyte’s microarrays has been described in the addenda of recent literature reports. ^{25 26}  

To verify the differential expression of selected mRNAs, quantitative real-time PCR (qPCR) was utilized. cDNAs were transcribed from mRNA samples by employing SuperScript II Reverse Transcriptase (Invitrogen) and oligo dT priming (Promega, Madison, WI). Primers were designed by using Primer Express Software, version 1.5 (Applied Biosystems, Inc., Foster City, CA). Specificity of the primers was verified by performing BLASTn searches on all relevant NCBI nucleotide databases. Particular focus was placed on identifying primers with a 16- to 40-bp length, 20% to 80% GC content, and a melting temperature between 58°C and 60°C, that would generate amplicons between 140 bp and 160 bp. The qPCR was performed by utilizing the specific primers at optimal concentrations (Table 1)and Applied Biosystems’ SYBR Green PCR Master Mix, MicroAmp Optical 96-Well Reaction Plate, ABI PRISM Optical Adhesive Covers and GeneAmp 5700 Sequence Detection System, according to the manufacturer’s protocol. The instrument’s dissociation protocol did not show any secondary PCR products in any of the amplifications. Gene expression was determined by using either the Relative Standard Curve Method or the Comparative C_T Method ²⁷ and standardizing levels to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or tubulin, δ1 mRNA. 

Results

To determine the effect of testosterone on meibomian gland gene expression, tissues were obtained from placebo- and androgen-treated, orchiectomized mice (n = 7/group/experiment) and processed for analysis (CodeLink Uniset Mouse I Bioarrays, GE Healthcare; GeneSifter software, VizX Labs). 

Evaluation of non- and log-transformed data from three separate experiments demonstrated that testosterone influenced the expression of more than 1590 genes in the mouse meibomian gland. This hormonal action involved a significant upregulation of 1080 genes (e.g., neuromedin B) and a significant downregulation of 518 genes (e.g., small proline-rich protein 2A; Table 2 ). Of particular interest was the finding that androgen treatment increased the activity of genes encoding various steroid receptors (e.g., types of estrogen, progesterone, and retinoic acid-binding sites), steroidogenic enzymes (e.g., 17β-hydroxysteroid dehydrogenase 7), and endocrine factors (e.g., insulin-like growth factor 1; Table 3 ). Moreover, testosterone altered the expression of several immune-associated genes (e.g., caspase 7; Table 3 ). 

Overall, testosterone treatment had a considerable impact on a diverse array of biological processes, molecular functions, and cellular components in meibomian tissue. Androgen influence extended to such processes as cell growth, metabolism, communication and transport, binding, catalytic activity, signal transduction, and receptor activity (data not shown). Most notable were the effects of testosterone on genes related to lipid dynamics (e.g., monoglyceride lipase, stearoyl-coenzyme A desaturases), protein transport (e.g., adaptor protein complexes), and intracellular vesicles (e.g., peroxisomal biogenesis factors; Table 4 4 ). Indeed, as shown by z-score analyses, androgen exposure led to the up (↑)- and down (↓) regulation of numerous genes associated with lipid metabolism (46↑, 14↓), lipid transport (8↑, 1↓), sterol biosynthesis (5↑, 1↓), fatty acid metabolism (16↑, 4↓), intracellular protein transport (31↑, 10↓), oxidoreductase activity (69↑, 16↓), peroxisomes (15↑, 0↓), mitochondria (76↑, 13↓), and early endosomes (5↑, 0↓; Tables 4 4 5 ). For comparison, testosterone’s effects were least directed toward processes such as cell adhesion, actin binding, and cytoskeleton (Table 5) . 

In contrast, the meibomian gland gene ontologies with the highest z-scores in the placebo-treated group were those related to mRNA metabolism, cell growth, endonuclease activity, and the cytoskeleton. The lowest scores were associated with processes such as proteolysis, G protein signaling, transport, and mitochondrial activity (Table 5) . 

To verify in part the bioarray (CodeLink; GE Healthcare) results, additional meibomian gland mRNA samples (n = 22 mice/group/experiment) were processed for gene chip (GEM 1 and 2; Incyte Genomics, Inc.) analyses. The gene chips and bioarrays have 4717 sequences in common. This approach showed that 474 genes were up (n = 319)- or downregulated (n = 155) by testosterone on both the bioarrays (P < 0.05) and gene chips (androgen/placebo ratio = > or <0; data not shown). If comparisons were restricted to those gene-chip genes that had expression ratios >1.5 (↑ or ↓), then 83 genes were identified as being similarly influenced by androgen on both platforms (e.g., Table 6 ). 

For further partial verification of bioarray and gene chip results, selected genes were analyzed by qPCR. As shown in Table 7 , this method confirmed the androgen-induced differential expression of several meibomian gland genes, including those identified by both bioarrays and gene chips (i.e., 17β-hydroxysteroid dehydrogenase 7, Abcd3, elongation of very long chain fatty acids-like 3, fatty acid transport protein 4, insulin-like growth factor 1, and monoglyceride lipase) or by GEM chips alone (i.e., Abca1, fatty acid synthase, cyclin D1, and odorant-binding protein Ia). 

Discussion

Our results demonstrate that androgens regulate the expression of numerous genes in the mouse meibomian gland. Testosterone administration to orchiectomized mice led to a significant increase in the transcriptional products of 1080 genes, including those related to lipid metabolism, sterol biosynthesis, fatty acid metabolism, protein transport, oxidoreductase activity, and peroxisomes. Androgen exposure also suppressed the levels of 518 mRNAs in this glandular tissue. These findings support our hypothesis that the androgen control of the meibomian gland involves the regulation of gene expression. 

The mechanism by which testosterone influences meibomian gland gene expression undoubtedly involves an association with saturable, high-affinity, and steroid-specific receptors in acinar epithelial cell nuclei. Androgen receptors are members of the steroid/thyroid hormone/retinoic acid family of ligand-activated transcription factors and appear to mediate the classic actions of androgens throughout the body. ^{28 29} After androgen binds to the receptor, the monomeric, activated hormone–receptor complex invariably associates with an androgen response element in the regulatory region of specific target genes; typically dimerizes with another androgen-bound complex; and, in combination with appropriate coactivators and promoter elements, controls gene transcription. ^{28 29} In support of this hypothesis, it has been shown that androgen receptors exist in sebaceous gland epithelial cells ^{17 18 19} and androgen activity in these cells may be compromised by androgen receptor defects or antagonists. ^{30 31 32} Similarly, androgen receptors exist in meibomian gland epithelial cells, ^{21 22} and androgen receptor disruption or the use of antiandrogen medications is associated with significant meibomian gland dysfunction and striking alterations in the neutral and polar lipid profiles of meibomian gland secretions. ^{2 3 12 13 14} In addition, we have recently found that many androgen-regulated genes in the meibomian gland appear to depend on the presence of functional androgen receptors. ³³  

It is important to note, though, that other processes may also be involved in, or mediate, androgen influence on meibomian gland gene expression. For example, the apparent androgen control of transcriptional activity may actually reflect hormone-induced alterations in mRNA stability, ³⁴ a possibility that remains to be explored. Another possibility is that testosterone’s impact on the meibomian gland is not direct, but rather is mediated through estrogen activity. The meibomian gland contains the mRNA for aromatase cytochrome P-450 (Schirra F, Suzuki T, Dickinson DP, Townsend DJ, Gipson IK, Sullivan DA, manuscript submitted) an enzyme that transforms testosterone into 17β-estradiol. ³⁵ Moreover, the meibomian gland harbors estrogen receptor mRNA and protein. ^{21 36} However, an estrogen mediation of androgen effects in the meibomian gland is highly unlikely. Recent research has shown that 17β-estradiol treatment of ovariectomized mice elicits a pattern of gene expression in meibomian tissue that is dissimilar from that induced by testosterone (Suzuki T, Schirra F, Jensen RV, Richards SM, Sullivan DA, manuscript submitted). ³⁷ Furthermore, unlike androgens, estrogens appear to decrease sebaceous gland function, ^{32 38} and this effect has been proposed to be due to an antagonism of androgen action. ^{39 40}  

Our finding that androgens modulate gene expression in the mouse meibomian gland is consistent with our earlier preliminary observations in rabbits. ⁴¹ Thus, by using RNA arbitrarily primed polymerase chain reactions, sequencing gels, and autoradiography, we were able to identify 58 differentially expressed mRNAs in the meibomian glands of orchiectomized rabbits treated topically with testosterone- or vehicle. However, analysis of 22 of the corresponding cDNA bands demonstrated that the majority had no significant homology to sequences in the GenBank database (http://www.ncbi.nlm.nih.gov/Genbank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD), presumably due to limited data for rabbit sequences. Consequently, to identify genes regulated by androgens in the meibomian gland, we selected mice as an experimental model because of the extensive genetic information available for this species. 

Considering that androgens modulate meibomian gland lipogenesis, the genes upregulated in this tissue by testosterone and the proteins they encode are particularly intriguing. Fatty acid synthase is a critical lipogenic enzyme that is known to be regulated by androgens in other tissues ^{42 43} and is expressed in meibomian gland epithelial cells (Richards SM, et al. IOVS 2002;43:ARVO E-Abstract 3150). Fatty acid transport protein 4 facilitates the cellular uptake and metabolism of long- and very-long-chain fatty acids, ⁴⁴ whereas elongation of very-long-chain fatty acids-like 1 and 3 promote the tissue-specific synthesis of very-long-chain fatty acids and sphingolipids. ^{45 46} These proteins could be involved in the androgen-induced increase of long-chain fatty acids in the total lipid fraction of rabbit meibomian glands. ¹ Monoglyceride lipase hydrolyzes tri- and monoglycerides to fatty acids and glycerol. ⁴⁷ Abca1 and Abcd3, which are members of the adenosine triphosphate (ATP)-binding cassette family, transport various molecules across extra- and intracellular membranes. Abca1 functions as a cholesterol efflux pump in the lipid-removal pathway ⁴⁸ and thereby serves as a key regulator of cholesterol distribution. ^{49 50} Abcd3 modulates the importation of fatty acids and/or fatty acyl-CoAs into peroxisomes. ⁵¹ 3-Hydroxy-3-methylglutaryl-coenzyme A reductase is the rate-limiting enzyme of sterol biosynthesis. ⁴⁶ Oxysterol binding protein-like 1A, sterol carrier protein 2, liver, lipocalin 3, and phosphatidylcholine transfer protein are involved in the binding and/or transfer of phospholipids. ⁴⁶ However, whether these proteins play a definitive role in androgen-meibomian gland interactions has yet to be determined. 

Androgens also were shown to control a series of genes that may be very important in the endocrine regulation of the meibomian gland. Thus, testosterone increased the mRNA levels of 17β-hydroxysteroid dehydrogenase 7, a member of the enzyme family that regulates the interconversion of 17-ketosteroids with their corresponding 17β-hydroxysteroids. ⁵² This enzymatic activity is essential for the metabolism of all active androgens and estrogens in peripheral sites ⁵² and may mediate the local, intracrine synthesis of androgens from adrenal precursors in the meibomian gland. Testosterone also enhanced the mRNA content of insulin-like growth factor 1, a pleiotropic protein that stimulates DNA synthesis and differentiation in sebaceous cells. ⁵³ Insulin-like growth factor 1 may also promote steroidogenesis ⁵⁴ and has been shown to be regulated by androgens in other tissues. ⁵⁵ Moreover, androgen treatment increased the expression of the gene encoding estrogen receptor 2 (β). This receptor, which is upregulated by androgen in the prostate, ⁵⁶ may inhibit the activity of estrogen receptor 1 (α). ⁵⁷ Testosterone also elevated the mRNA levels of 11β-hydroxysteroid dehydrogenase 1, an enzyme that catalyzes the conversion of cortisol to the inactive metabolite cortisone. ⁴⁶ Of interest, testosterone downregulated the gene expression of aldehyde dehydrogenase family 1, subfamily A3 (also called retinaldehyde dehydrogenase 3), an enzyme that stimulates retinoic acid biosynthesis. ⁴⁶  

In addition to these actions, androgens promoted the expression of genes involved in the sorting (e.g., adaptor protein complexes, RAB9), trafficking (e.g., ADP-ribosylation factor 5, sorting nexin 2), and hydrolysis (e.g., cathepsin B) of proteins in various cellular locations, including the endosome, Golgi apparatus, endoplasmic reticulum, lysosome, proteasome, nucleus, and mitochondrion. ⁴⁶ Androgens also increased the mRNA levels of epimorphin (an extracellular protein that directs epithelial cell morphogenesis), neuromedin B (a bombesin-like peptide that stimulates epithelial cell proliferation), ⁵⁸ and phospholipases C-β3 and -β4 (mediators of the production of the second-messenger molecules diacylglycerol and inositol 1,4,5-trisphosphate). ⁴⁶ Testosterone decreased the mRNA amounts of Ia-associated invariant chain, which plays an essential role in major histocompatibility (MHC) class II antigen processing. ⁴⁶  

In summary, our results show that testosterone regulates the expression of a number of genes in the mouse meibomian gland and that many of these genes are involved in the production, metabolism, transport, and release of lipids, as well as in steroidogenic pathways. We are currently attempting to determine the meibomian gland distribution of these genes (e.g., by in situ hybridization), in order to identify the cellular targets for androgen action. This procedure will also verify mRNA location within the gland, compared with the conjunctiva, given that very small parts of this latter tissue adhered to the meibomian gland during isolation. In concert with these studies, we are endeavoring to determine whether a variety of these hormone-regulated genes are translated (e.g., by immunohistochemistry and Western blot analysis). This combined information may help to explain, at least in part, the mechanism by which topical androgens reportedly stimulate the synthesis and secretion of meibomian gland lipids, prolong the tear film breakup time, and alleviate dry eye. ^{7 8}  

 Supported by Grants EY05612, K16, and EY12523; Allergan, Inc., (Irvine, CA), and German Research Society (Deutsche Forschungsgemeinschaft) Grants SCHI 562/1-1 and 1-2.

 Submitted for publication April 4, 2005; revised May 25, 2005; accepted August 15, 2005.

 Disclosure: F. Schirra, None; T. Suzuki, Allergan, Inc. (F); S.M. Richards, None; R.V. Jensen, None; M. Liu, Allergan, Inc. (F); M.J. Lombardi, None; P. Rowley, None; N.S. Treister, None; D.A. Sullivan, Allergan, Inc. (C, F)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Oligonucleotide Primers for Real-Time PCR Confirmation of Selected Genes

Accession No.	mRNA	Orientation	Nucleotide Sequence (5′→3′)
Y15733*	17β-hydroxysteroid dehydrogenase 7	Sense	TTTGTAAATGCGCTCACTGTGA
		Antisense	TTTTGGCCCGTGACGTAATT
NM_013454	Abca1	Sense	CCCTGCTTCCGTTATCCAACT
		Antisense	GGACCTTGTGCATGTCCTTAATG
NM_008991	Abcd3	Sense	CAGTCGCCCCTTCCTAGATCT
		Antisense	TCACGCCCAGCCAAAACTATAC
S78355	Cyclin D1	Sense	CTGACACCAATCTCCTCAACGA
		Antisense	CTCACAGACCTCCAGCATCCA
NM_007703	Elongation of very long chain fatty acids-like 3	Sense	CTATGAAGGCTGCCAAACTGAAG
		Antisense	TTGTTGTGTGGCATCCTTTCTC
AF127033	Fatty acid synthase	Sense	TCCTGGAACGAGAACACGATCT
		Antisense	GAGACGTGTCACTCCTGGACTTG
AF072759	Fatty acid transport protein 4	Sense	GGTGTTGAGGTGCCAGGAACT
		Antisense	GCAAGAAGCGCAGGAAGATG
BC049235	Glutathione peroxidase 3, †	Sense	GTTCCAAATGAGCCCAAAGG
		Antisense	TAGTGTGGGCATGTGGGAGAT
M32599	Glyceraldehyde-3-phosphate dehydrogenase	Sense	CATGGCCTTCCGTGTTCCTA
		Antisense	CTGGTCCTCAGTGTAGCCCAA
NM_010512	Insulin-like growth factor 1	Sense	ACCTCAGACAGGCATTGTGGAT
		Antisense	TGAGTCTTGGGCATGTCAGTGT
NM_011844	Monoglyceride lipase	Sense	GCTCCCCTGAAGCAGTGAAAC
		Antisense	GGCCCTCCGTAAAGATAGAAGTG
BC028490	Neuromedin B	Sense	GGGACAGCACCCCCTAACA
		Antisense	TTTCTTTCGCAGGAGGATCCT
Y10971	Odorant-binding protein Ia	Sense	ACAATCACTGGGTATTTGCAAGAAG
		Antisense	TCAGCTCTGTCCACGTTCTCA
AF323080	Resistin, †	Sense	GCCACCCGTTCCTGTAGCT
		Antisense	GGATTCGCGCACGTGAGT
AK007928	Small proline-rich protein 2A	Sense	CATGCCCTCCTGTGCAATTT
		Antisense	CACTTCCCCTGTTCCTGATGA
NM_019756	Tubulin, δ1	Sense	GGAACGGGTGAGGTCATTGT
		Antisense	TGAATGAGAAGGGCATCTGAAG

 The accession number refers to the sequence identity of the gene fragment used on the gene chip or bioarray. This sequence, which is listed in the nucleotide database of NCBI, served as the basis for primer design. The primers were used to determine mRNA levels with real-time PCR by the relative standard curve method or the comparative C_T Method.

*  Also identified with Unigene cluster code of Mm.12882. Additional names for these mRNAs include: elongation of very long chain fatty acids-like 3 is (FEN1/Elo2, SUR4/Elo3, yeast)-like 3 (Elovl3), cold inducible glycoprotein 30 is CIN-2, Cig30; Abca1 is ATP-binding cassette, sub-family A member 1; Abcd3 is ATP-binding cassette, sub-family D member 3.

†  Comparative C_T method.

Table 2.

 

Androgen Influence on Gene Expression Ratios in the Mouse Meibomian Gland

Accession Number	Gene	Ratio	P	Ontology
Testosterone > placebo
NM_026523	Neuromedin B	3.2	0.0105	Signal transduction
AF218416	Tocopherol (α) transfer protein	3.0	0.0024	Fat-soluble vitamin metabolism
NM_010898	Neurofibromatosis 2	2.6	0.0013	Phosphate metabolism
NM_008161	Glutathione peroxidase 3	2.5	0.0036	Response to biotic stimulus
NM_007588	Calcitonin receptor	2.3	0.0063	Signal transduction
NM_009998	Cytochrome P450, family 2, subfamily b, polypeptide 20	2.3	0.0001	Electron transport
NM_010220	FK506-binding protein 5	2.3	0.0032	Protein metabolism
NM_019501	Trans-prenyltransferase	2.2	0.0003	Isoprenoid biosynthesis
NM_017370	Haptoglobin	2.2	0.0274	Response to biotic stimulus
NM_013874	Neuronal d4 domain family member	2.1	0.0003	Transcription
NM_009848	Ectonucleoside triphosphate diphosphohydrolase 1	2.1	0.0018	Ribonucleotide catabolism
NM_009947	Copine VI	2.1	0.0106	Lipid metabolism
NM_007884	Dermatan sulphate proteoglycan 3	2.1	0.0322	Extracellular space
NM_008288	Hydroxysteroid 11-β dehydrogenase 1	2.1	0.0123	Steroid metabolism
AK006085	Dipeptidase 3	2.1	0.0089	Protein catabolism
NM_016870	SA rat hypertension-associated homolog	2.0	0.0030	Carboxylic acid metabolism
Placebo > testosterone
NM_011468	Small proline-rich protein 2A	4.0	0.0418	Morphogenesis
NM_011339	Chemokine (C-X-C motif) ligand 15	3.3	0.0094	Response to biotic stimulus
NM_007894	Eosinophil-associated, ribonuclease A family, member 1	2.7	0.0148	Ribonuclease activity
AK013634	Cyclin T2	2.4	0.0027	Transcription
NM_022984	Resistin	2.4	0.0136	Receptor binding
NM_011585	Cytotoxic granule–associated RNA binding protein 1	2.3	0.0468	Programmed cell death
AF250135	Splicing factor, arginine/serine-rich 2	2.2	0.0066	RNA splicing, via transesterification reactions
NM_011478	Small proline-rich protein 3	2.2	0.0263	Structural molecule activity
NM_008476	Keratin complex 2, basic, gene 6a	2.1	0.0251	Cytoplasm organization and biogenesis

 Relative ratios were determined by comparing the degree of gene expression in meibomian glands from placebo and testosterone-treated orchiectomized mice. Genes listed had a comparative P < 0.05 (between glands) and a known identity. These ratios, which were generated from nontransformed data, were analogous to those found after log transformation.

Table 3.

 

Testosterone Regulation of Various Endocrine- and Immune-Related Genes in the Meibomian Gland

Upregulation	Downregulation
Endocrine
Estrogen receptor 2 (β)
Membrane progestin receptor α
Progesterone receptor membrane component 1
Progesterone receptor membrane component 2
Retinoic acid receptor, α
Retinoid X receptor γ
Insulin-like 6
Insulin-like growth factor 1
Insulin receptor
Insulin-degrading enzyme
Thyroid hormone receptor–associated protein 2
Arginine vasopressin receptor 1B
Arginine vasopressin receptor 2
Cholecystokinin
Cholecystokinin A receptor
Hydroxysteroid (17-β) dehydrogenase 7
Retinol dehydrogenase 11
Retinol dehydrogenase 6
Immune
IL-5	IL-1 receptor, type I
IL-12b	IL-4 receptor, α
IL-21	IL-6 signal transducer
IL-1 receptor-like 1 ligand	IL-10 receptor, β
Chemokine (C-C motif) receptor 4	IFN regulatory factor 1
Chemokine-binding protein 2	IFN-γ receptor
	Caspase 7
	Chemokine (C-C motif) ligand 5
	Chemokine (C-C motif) ligand 19
	Chemokine (C-X3-C motif) ligand 1
	Chemokine (C-X-C motif) ligand 16

 The selected genes were significantly (P < 0.05) up- or downregulated by testosterone.

Table 4.

 

Androgen Control of Meibomian Gland Genes Associated with Lipid Metabolism and Transport, Sterol Biosynthesis, Protein Activity, and Cellular Components

Upregulation	Downregulation
Lipid metabolism
1-Acylglycerol-3-phosphate O-acyltransferase 3	3-hydroxy-3-methylglutaryl-coenzyme A synthase 2
3-Hydroxy-3-methylglutaryl-coenzyme A reductase	Acetyl-coenzyme A dehydrogenase, long chain
Acetyl-coenzyme A dehydrogenase, medium chain	Aldehyde dehydrogenase family 1, subfamily A3
Acyl-CoA synthetase long-chain family member 4	Arachidonate 12-lipoxygenase, 12R type
Acyl-CoA synthetase long-chain family member 5	Arachidonate 5-lipoxygenase-activating protein
Acyl-coenzyme A oxidase 3, pristanoyl	Copine III
Adipose differentiation-related protein	Fatty acid desaturase 2
Alcohol dehydrogenase 5 (class III), χ polypeptide	Glucosamine
Cellular nucleic acid–binding protein 1	Glycerol phosphate dehydrogenase 2, mitochondrial
Colipase, pancreatic	GM2 ganglioside activator protein
Cytochrome P450, 51	Peroxiredoxin 6
Cytosolic acetyl-CoA hydrolase	Protein kinase, cAMP dependent regulatory, type II beta
Cytosolic acyl-CoA thioesterase 1	RIKEN cDNA 5133401H06 gene
Elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 1		Sortilin-related receptor, LDLR class A repeats-containing
	Sphingosine phosphate lyase 1
Elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 3
Emopamil binding protein-like
Hydroxysteroid 17-β dehydrogenase 7
Hydroxysteroid 11-β dehydrogenase 1
Gastric lipase
Longevity assurance homolog 4 (S. cerevisiae)
Membrane-interacting protein of RGS16
Monoglyceride lipase
NAD(P) dependent steroid dehydrogenase-like
Oxysterol binding protein-like 1A
Peroxisome proliferator activator receptor δ
Phosphatidylserine synthase 2
Phospholipase A2, group IIC
Phospholipase A2, group IIE
Phospholipase A2, group IIF
Phospholipase A2, group XIIA
Phospholipase C, β3
Phospholipase C, β4
Phytanoyl-CoA hydroxylase
SA rat hypertension-associated homolog
Sodium channel, voltage-gated, type XI, alpha polypeptide
Sphingomyelin phosphodiesterase 1, acid lysosomal
StAR-related lipid transfer (START) domain containing 5
Stearoyl-coenzyme A desaturase 1
Stearoyl-coenzyme A desaturase 2
Stearoyl-coenzyme A desaturase 3
Sterol carrier protein 2, liver
Sterol-C4-methyl oxidase-like
Sulfotransferase family 1A, phenol-preferring, member 1
Triosephosphate isomerase 1
Lipid transport
Adipose differentiation-related protein	Sortilin-related receptor, LDLR class A repeats-containing
Apolipoprotein A-V
Apolipoprotein E
Niemann pick type C1
Oxysterol binding protein-like 1A
Phosphatidylcholine transfer protein
StAR-related lipid transfer (START) domain containing 5
Sterol carrier protein 2, liver
Intracellular protein transport
Adaptor protein complex AP-1, μ2 subunit	A kinase (PRKA) anchor protein (gravin) 12
Adaptor protein complex AP-1, σ1	CDNA sequence BC003281
Adaptor-related protein complex 3, β1 subunit	Ia-associated invariant chain
Adaptor-related protein complex 3, μ1 subunit	Nuclear factor of κ light chain gene enhancer in B-cells inhibitor, α
Adaptor-related protein complex 3, σ2 subunit	Nuclear RNA export factor 1 homolog (S. cerevisiae)
Adaptor-related protein complex AP-4, σ1	Ras-GTPase-activating protein SH3-domain binding protein
ADP-ribosylation factor 5	Secretory carrier membrane protein 3
B-cell receptor-associated protein 29	Splicing factor, arginine/serine-rich 1 (ASF/SF2)
Cathepsin B	Transforming growth factor, β1
Choroidermia	Transforming, acidic coiled-coil containing protein 3
Coatomer protein complex, subunit ε
Epimorphin
Histocompatibility 47
(continues)

Table 4A.

 

(Continued). Androgen Control of Meibomian Gland Genes Associated with Lipid Metabolism and Transport, Sterol Biosynthesis, Protein Activity, and Cellular Components

Upregulation	Downregulation
Intracellular protein transport (continued)
Karyopherin (importin) β1
PDZ domain containing 11
Peroxisomal biogenesis factor 13
RAB1B, member RAS oncogene family
RAB2, member RAS oncogene family
Rab38, member of RAS oncogene family
RAB4B, member RAS oncogene family
RAB9, member RAS oncogene family
RIKEN cDNA 1110034E15 gene
RIKEN cDNA 5830417C01 gene
RIKEN cDNA A430019L02 gene
Sec61 β subunit
Signal recognition particle 54
Sortilin-related VPS10 domain containing receptor 2
Sorting nexin 2
Translocase of inner mitochondrial membrane 8 homolog b (yeast)
Translocase of outer mitochondrial membrane 20 homolog (yeast)
Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, β pol
Peroxisome (also microbody)
2-4-Dienoyl-coenzyme A reductase 2, peroxisomal
Acyl-coenzyme A oxidase 3, pristanoyl
ATP-binding cassette, sub-family D (ALD), member 3
Catalase
Mpv17 transgene, kidney disease mutant
Peroxisomal biogenesis factor 13
Peroxisomal membrane protein 3
Peroxisomal membrane protein 4
Peroxisomal membrane protein 4
Peroxisome biogenesis factor 19
Phytanoyl-CoA hydroxylase
Polyamine oxidase
Solute carrier family 25
Sterol carrier protein 2, liver
Early endosome
CD1d1 antigen
RAB34, member of RAS oncogene family
RAB5A, member RAS oncogene family
Tumor protein D52-like 1
Vesicle-associated membrane protein 8

 The selected genes, which are examples of each category, were significantly (P < 0.05) up- or downregulated by testosterone treatment.

Table 5.

 

High and Low Expression of Gene Ontologies in the Meibomian Glands of Placebo- or Testosterone-Treated Mice

Ontology	Gene List	T Genes ↑	P Genes ↑	Array Genes	T z-Score	P z-Score
Biological process
Lipid metabolism	60	46	14	265	3.56	0.47
Fatty acid metabolism	20	16	4	77	2.87	0.21
Lipid transport	9	8	1	33	2.51	−0.45
Morphogenesis	90	52	38	652	−2.4	1.45
Cell adhesion	34	19	15	293	−2.4	0.36
Protein modification	94	56	38	728	−2.81	0.72
mRNA metabolism	24	9	15	95	−0.4	5.16
Regulation of cell growth	11	3	8	38	−0.57	4.79
Cell growth	12	3	9	46	−0.93	4.79
Proteolysis and peptidolysis	49	43	6	344	1.09	−2.65
G-protein-coupled receptor signaling pathway	52	47	5	337	1.96	−2.86
Catabolism	75	66	9	495	1.95	−3.14
Molecular function
Oxidoreductase activity	85	69	16	396	4.39	−0.82
Lyase activity	21	16	5	79	2.73	0.59
Carboxylic ester hydrolase activity	12	12	0	57	2.51	−1.72
Binding	668	435	233	4340	−2.57	2.3
Actin binding	11	3	8	118	−2.91	0.95
Protein binding	214	125	89	1540	−3.8	1.81
S-adenosylmethionine-dependent methyltransferase activity	11	3	8	38	−0.57	4.62
RNA binding	49	24	25	246	−0.52	3.89
Endonuclease activity	12	5	7	41	0.29	3.62
Rhodopsin-like receptor activity	30	29	1	178	2.4	−2.72
Transporter activity	114	91	23	788	0.74	−2.73
G-protein coupled receptor activity	36	34	2	228	2.05	−2.86
Cellular component
Microbody	15	15	0	45	4.82	−1.5
Peroxisome	15	15	0	45	4.82	−1.5
Mitochondrion	89	76	13	485	3.44	−2.22
Extracellular matrix	10	7	3	130	−2.06	−1.32
Nucleus	273	166	107	1751	−2.31	3.12
Cytoskeleton	56	24	32	340	−2.37	4.15
Cytoskeleton	56	24	32	340	−2.37	4.15
Intracellular	618	409	209	3744	−0.12	3.6
Nucleus	273	166	107	1751	−2.31	3.12
Extracellular space	250	185	65	1705	−0.17	−2.1
Mitochondrion	89	76	13	485	3.44	−2.22
Integral to membrane	322	240	82	2108	0.75	−2.23

 Specific ontologies, with some of the highest and lowest z-scores, were selected after the analysis of log-transformed data. Criteria for inclusion in the table were an ontology containing ≥9 genes and having a z-score >2.0 or < −2.0. High and low values for the placebo (P) and testosterone (T) groups in designated ontologies are highlighted in bold print.

Table 6.

 

GEM Chip Confirmation of Selected Bioarray Results

Accession Number	Gene	Ontology
Testosterone > placebo
NM_010476	Hydroxysteroid (17β) dehydrogenase 7	Steroid metabolism
NM_008288	Hydroxysteroid 11β dehydrogenase 1	Steroid metabolism
NM_013754	Insulin-like 6	Receptor binding
NM_026058	Longevity assurance homolog 4 (S. cerevisiae)	Sphingoid biosynthesis
AK009450	Membrane progestin receptor α	Signal transducer activity
NM_010784	Midkine	Cell cycle
NM_010898	Neurofibromatosis 2	Phosphate metabolism
NM_020573	Oxysterol binding protein-like 1A	Lipid transport
NM_026438	Pyrophosphatase	Hydrolysis of diphosphate bonds
NM_021557	Retinol dehydrogenase 11	Oxidoreductase activity
M62361	Sterol carrier protein 2, liver	Carboxylic acid metabolism
Placebo > testosterone
NM_007606	Carbonic anhydrase 3	One-carbon compound metabolism
K02782	Complement component 3	Vesicle-mediated transport
NM_011469	Small proline-rich protein 2B	Structural molecule activity

 Separate meibomian gland mRNA samples from placebo- or testosterone-treated mice were processed for GEM 1 and 2 gene chip evaluation. Differentially expressed genes were then compared to those identified by CodeLink bioarrays. This table lists known genes with known functions that showed greater than a 1.5-fold up- or downregulation in both CodeLink and GEM arrays.

Table 7.

 

Verification of Selected Bioarray and Gene Chip Results

Gene	CodeLink Ratio	GEM Ratio	qPCR Ratio
Testosterone > placebo
17β-Hydroxysteroid dehydrogenase 7	1.77	2.80	2.21
ATP-binding cassette, sub-family A, member 1	/	1.55	2.17
ATP-binding cassette, sub-family D, member 3	1.41	1.82	2.19
Elongation of very long chain fatty acids-like 3	1.40	1.91	1.65
Fatty acid synthase	/	2.10	2.08
Fatty acid transport protein 4	1.25	2.20	1.57
Glutathione peroxidase 3	2.52	−	2.09*
Insulin-like growth factor 1	1.36	1.90	1.51†
Monoglyceride lipase	1.39	2.09	2.07
Neuromedin B	3.21	−	2.10
Placebo > testosterone
Cyclin D1	−	2.49	1.57†
Odorant-binding protein Ia	−	29.41	55.0
Resistin	2.42	−	2.97*
Small proline-rich protein 2A‡	3.96	3.41	4.30

 The expression of selected genes, that were shown to be differentially upregulated in meibomian glands of placebo- or testosterone-treated mice by using CodeLink Bioarrays or GEM chips, were re-evaluated with qPCR procedures. Unless otherwise noted (i.e., n = 2†), all qPCR ratios equal the average of 3 measurements. The mRNA levels were standardized to those of GAPDH or, where designated,* tubulin, δ1. The “–” symbol indicates that the gene sequence was not present on either the CodeLink Bioarray or GEM chip, or that the normalized data were below the CodeLink sensitivity limit of 0.5 signal intensity units (i.e., cyclin D1). The “/” symbol indicates that the hormone effect did not reach significance on the CodeLink Bioarrays. All other CodeLink data were significantly (P < 0.05, two-tailed) increased or decreased relative to control, with the exception of fatty acid transport protein 4 (P < 0.05, one-tailed). Differences between CodeLink and GEM results may be due, at least in part, to variations (e.g., up to 30%) in signal intensities that may occur among corresponding elements on different microarrays. ⁵⁹

‡  The UniGene Cluster identification code for small proline-rich protein 2A is now apparently the same as for 2B.

The authors thank Christian B. Wade (Seattle, WA) for assistance with the GeneSifter software, and the Harvard Center for Genomic Research for help with processing the GEM 1 and 2 gene chip data.