Are BALB/c mice relevant models for understanding sex-related differences in gene expression in the human meibomian gland?

Abstract

Background:

A compelling feature of dry eye disease is that it occurs predominantly in women. We hypothesize that this female prevalence is linked to sex-related differences in the meibomian gland. This gland plays a critical role in maintaining the tear film, and its dysfunction is a major cause of dry eye disease. To understand the factors that underlie meibomian gland sexual dimorphism and promote dry eye in women, we seek to identify an optimal model for the human meibomian gland. Our goal was to determine whether a murine meibomian gland is such a model. Towards that end, we examined whether sex differences in meibomian gland gene expression are the same in BALB/c mice and humans.

Methods:

Eyelid tissues were collected from humans (n = 5–7/sex) and BALB/c mice (n = 9/sex). Meibomian glands were isolated and processed for the evaluation of gene expression by using microarrays and bioinformatics software.

Results:

Our analysis of the 500 most highly expressed genes from human and mouse meibomian glands showed that only 24.4% were the same. Our comparison of 100 genes with the greatest sex-associated differences in human and mouse meibomian glands demonstrated that none were the same. Sex also exerted a significant impact on numerous ontologies, KEGG pathways and chromosomes, but these effects were primarily species-specific.

Conclusions:

Our results indicate that BALB/c mice are not optimal models for understanding sex-related differences in gene expression of the human meibomian gland.

Introduction

Significant sex-related differences exist in the physiology and pathophysiology of the eye. Indeed, sex influences multiple aspects of ocular tissues, including their anatomy, gene expression, protein synthesis, secretion, sensitivity, cellular density, permeability, immune activity and visual acuity.^{1, 2} Moreover, sex differences have been linked to the development and/or progression of many ocular pathologies, including dry eye disease (DED), vernal keratoconjunctivitis, refractive errors, myopia, cataracts, glaucoma, age-related macular degeneration, diabetic retinopathy, low vision and blindness.^1–5 We believe that recognition of these sex-related differences, as well as the determination of their basis, are extremely important. We also believe that such understanding may be translated into new insights into the physiological regulation of ocular tissues, as well as the development of novel and unique therapeutic strategies to treat diverse disorders of the eye.

However, despite this influence, scant information exists concerning the precise mechanisms mediating the effects of sex on the eye. As an example, we and others have discovered that sex-related differences exist in the morphology, gene expression, function and secretion (i.e. meibum) of the meibomian gland (MG).² The MG, through its release of meibum, promotes the stability and prevents the evaporation of the tear film, and plays a critical role in the health and well-being of the ocular surface.⁶ Conversely, MG dysfunction (MGD), and the associated meibum deficiency, destabilize the tear film, increase its evaporation, and serves as a major cause of DED.^6–8 DED affects hundreds of millions of people throughout the world and occurs predominantly in women.⁹ We hypothesize that the sex-related differences in the MG contribute to this female DED prevalence. But, if our hypothesis is correct, the underlying mechanisms are unknown.

To understand the factors that underlie human MG sexual dimorphism and possibly promote DED in women, we seek to identify a relevant model for the human MG. We hypothesized that the BALB/c mouse MG would be such a model, and especially its expression of sex differences in gene expression. Our rationale was three-fold. First, murine models have been invaluable for providing insight into human physiological and pathophysiological processes.^{10, 11} Second, as in humans, sex-related differences are present in the MGs of BALB/c and other strains of mice in both health and disease.^12–16 Third, approximately 99% of mouse genes have a homologue in the human genome.¹⁷ As one additional consideration, the meibomian gland is a large sebaceous gland, and sex-related variations in non-ocular sebaceous glands have been linked to changes in gene expression.¹⁸ Accordingly, to begin to address our hypothesis, we examined whether sex differences in MG gene expression are the same, or almost the same, in BALB/c mice and humans.

Materials & Methods

Human subjects and tissue collections

Human eyelid tissues were obtained from seven females (age = 68.9 ± 7.3 years; age range = 32 to 88) and five males (age = 69.6 ± 6.6 years; age range = 44 to 79) after lid resection surgery at the Massachusetts Eye and Ear Infirmary, as previously described.¹⁹ The surgeries were performed to correct ectropion (n = 6 subjects), entropion (n = 3 subjects), lid laxity (n = 1 subject), lid retraction (n = 1 subject) or floppy eyelid syndrome (n = 1 subject).¹⁹ Subjects were excluded if they wore contact lenses, had active infection, or used topical anti-inflammatory or anti-glaucoma medications. Eight subjects were taking medicines for high blood pressure (n = 3), high cholesterol (n = 2), hyperthyroidism (n = 1), osteoporosis (n = 1) or anxiety (n = 1).¹⁹ Subjects with MGD were not excluded, and four females and two males had varying degrees of this condition, with secretion qualities equal to 2 (n = 3 subjects) or 3 (n = 3 subjects) according to a published classification system.²⁰ Tissue segments were placed in RNA later (Ambion, Austin, TX) and frozen at −80ºC. MGs (2 – 5 glands/tarsal plate) were later isolated under a dissecting microscope (Bausch & Lomb, Rochester, NY) by removing adherent skin, subcutaneous tissue, muscle and palpebral conjunctiva, and processed for molecular biological procedures. The use of human eyelid segments, that would otherwise have been discarded, was approved by the Institutional Review Boards of the Schepens Eye Research Institute and the Massachusetts Eye and Ear Infirmary, and followed the tenets of the Declaration of Helsinki.

Mice and tissue collections

Ten-week old female and male BALB/c mice (n = 9/sex) were obtained from Jackson Laboratories (Bar Harbor, ME) and maintained in constant temperature rooms with 12-hour light/dark intervals. At 12 weeks of age animals were sacrificed by CO₂ inhalation and MGs were removed from the upper and lower lids under direct visualization with a biomicroscope. This surgical technique involved making a small incision near the inner corner of the eyelid, separating skin and subcutaneous tissue from the inner to outer aspect of the lid, and excising skin from the MGs by cutting at the mucocutaneous junction. The palpebral conjunctiva was then removed from the MGs, and the glands were dissected from the residual tissue by starting at the outer lid corner and carefully avoiding an adjacent vein. MG samples were prepared by combining tissues from 3 mice/sex/group. Three such pooled samples were made for each tissue/sex/group and then processed for RNA analysis. All studies were approved by the Institutional Animal Care and Use Committee of the Schepens Eye Research Institute and adhered to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

Molecular biological procedures

MGs were processed for gene expression analyses, as previously described.^15,19 In brief, total RNA was extracted from MGs using RNeasy Mini Kits (Qiagen, Inc., Valencia, CA), according to the manufacturer’s instructions. The RNA concentrations and 260/280 nm ratios were determined by using a NanoDrop 1000 spectrophotometer (Thermo Scientific, Waltham, MA), and RNA integrity was examined by using a RNA Nano 6000 Series II Chip with a Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA). The RNA samples were further processed by Asuragen (Austin, TX) for quantitation of mRNA levels by utilizing microarray analyses (HumanHT-12 v.3 and MouseWG-6 v.2 Expression BeadChips; Illumina, San Diego, CA).^21,22

The non-log-transformed, background subtracted and cubic spline normalized data were analyzed with GeneSifter.net software (PerkinElmer, Boston, MA). This comprehensive program generated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, ontology and z-score reports. Standardized hybridization intensity data were adjusted by adding a constant, such that the lowest intensity value for a sample equaled 16. ^23,24 BeadChip data were evaluated with Student’s t-test (two-tailed, unpaired). All human data are available for download through the National Center for Biotechnology Information’s Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) via series accession number GSE17822. All BALB/c data are accessible in Supplemental files 1 through 6.

Results

Gene expression in human and mouse MGs

For our comparative evaluation of gene expression in human and BALB/c mouse MGs, we used species-specific Illumina Expression BeadChips. These arrays contained 48,803 and 45,281 probes that encoded 35,325 and 33,622 annotated transcripts for human and mouse genes, respectively. Of these transcripts, 21,101 coded the same genes on both the human and mouse Illumina BeadChips. After sample processing, a total of 11,099 and 18,671 gene transcripts from the experimental human and mouse arrays met or exceeded Illumina’s quality value of 0.95. These ‘quality’ genes were used for our analyses.

The eleven most highly expressed non-ribosomal genes in human and mouse MGs are shown in Table 1. The most highly expressed ribosomal genes are listed in Supplemental Table 1. Our data show that 90.9% of the non-ribosomal, and all of the ribosomal, most highly expressed genes are the same in human female and male MGs (Table 1 and Supplemental Table 1). Similarly, 90.9% of the non-ribosomal, and all of the ribosomal, most highly expressed genes are the same in the mouse female and male MGs (Table 1 and Supplemental Table 1). The high intensity expression of these BALB/c mouse genes were not unique to this murine strain, as the same genes were also very highly expressed in female and male MGs of 5-month old C57BL/6J mice (data are available from the National Center for Biotechnology Information’s Gene Expression Omnibus via series accession number GSE5878). In addition, age and MG health differences did not appear to be major factors in determining the human genes listed in Table 1 and Supplemental Table 1. These genes were all highly expressed in the youngest and oldest women and men, as well as in individuals without MGD (data not shown).

Table 1.

Highest intensity genes in meibomian glands of female and male humans and BALB/c mice

Accession Number	Gene	Female Intensity	Male Intensity
Human
NM_021706	Leukocyte-associated immunoglobulin-like receptor 1	29709 ± 1309 (1)	25989 ± 5052 (1)
NM_001402	Eukaryotic translation elongation factor 1 α1	26849 ± 871 (2)	24851 ± 1209 (2)
NM_024793	Clusterin associated protein 1	23311 ± 996 (3)	19962 ± 3857 (3)
NM_001992	Coagulation factor II receptor	22403 ± 966 (4)	17893 ± 3583 (4)
NM_058173	Mucin-like 1	20620 ± 4483 (5)	14122 ± 4496 (7)
NM_005063	Stearoyl-coenzyme A desaturase	17565 ± 1600 (6)	15673 ± 4420 (5)
NM_001001851	Inter-α inhibitor H5, transcript variant 3	14526 ± 1248 (7)	11385 ± 2152 (8)
NM_005687	Phenylalanyl-tRNA synthetase-like, β subunit	13616 ± 1251 (8)	11322 ± 2000 (10)
NM_021109	Thymosin β4	13502 ± 1491 (9)	11326 ± 2619 (9)
NM_000146	Ferritin	5193 ± 1608 (10)	4426 ± 887
NM_002652	Prolactin-induced protein	2390 ± 870 (11)	14473 ± 6139 (6)
NM_000239	Lysozyme	192 ± 42	9219 ± 6088 (11)
Mouse
NM_007793	Cystatin B	60693 ± 433 (1)	59223 ± 2290 (2)
NM_007830	Diazepam binding inhibitor	59960 ± 1339 (2)	59490 ± 2500 (1)
NM_010362	Glutathione S-transferase ω1	58397 ± 892 (3)	55548 ± 2010 (4)
AK018753	NADH dehydrogenase subunit 1	57059 ± 1373 (4)	58921 ± 2495 (3)
NM_016958	Keratin 14	53407 ± 3280 (5)	52790 ± 3298 (7)
NM_145942	3-hydroxy-3-methylglutaryl-coenzyme A synthase 1	51180 ± 2264 (6)	54655 ± 2028 (6)
NM_019422	Elongation of very long chain fatty acids-like 1	49793 ± 3071 (7)	51248 ± 735 (9)
NM_019738	Nuclear protein 1	47460 ± 1631 (8)	55537 ± 458 (5)
NM_016740	S100 calcium binding protein A11	46363 ± 992 (9)	44614 ± 1616
NM_011664	Ubiquitin B	46316 ± 1477 (10)	52102 ± 1230 (8)
NM_008496	β-Galactoside-binding protein galectin-7	45377 ± 2300 (11)	47826 ± 1428 (11)
NM_009127	Stearoyl-coenzyme A desaturase 1	45122 ± 1158	48852 ±1802 (10)

The accession number is the sequence identity of the gene fragment expressed on the microarray and is listed in the NCBI nucleotide database. The intensities of the top eleven genes/sex/species are presented in this Table, and were obtained after normalizing all meibomian gland data and calculating the mean (± SE) of all samples. Values were generated from non-transformed data. The rank intensity (i.e. one through eleven) of each gene is listed in the parenthesis. Although not in the top eleven, the comparative intensities of human female lysozyme, human male ferritin, mouse female stearoyl-coenzyme A desaturase, and mouse male S100 calcium binding protein A11 are also shown.

Cross species comparisons demonstrated that almost all of the most highly expressed ribosomal and non-ribosomal genes in human and BALB/c mouse MGs are not the same. Only one ribosomal gene, ribosomal protein L41, and one non-ribosomal gene, stearoyl-coenzyme A desaturase, were among the most highly expressed transcripts in both human and mouse MGs. These expression differences could not be attributed solely to intrinsic variations between human and mouse microarray platform designs. All of the most highly expressed mouse MG genes were present in human MGs, but with the exception of stearoyl-coenzyme A desaturase, at much lower levels. In the same way, a majority of the most highly expressed non-ribosomal and ribosomal human MG genes were present in female and male mouse MGs, but typically in lower quantities.

To extend these findings, we compared the 500 most highly expressed genes in human and BALB/c mouse MGs. Our analysis showed that 24.4% of the human and mouse genes are the same (Supplemental Table 2). If ribosomal genes were excluded, only 15.2% of the genes are identical in human and mouse MGs.

Sex-related differences in gene expression in human and mouse MGs

Sex had a significant influence on gene expression in human and BALB/c mouse MGs. We found that 112 genes are more highly expressed in human females, and 130 genes more highly expressed in human males. We also discovered that 1,608 genes are more highly expressed in female, and 2,067 genes more highly expressed in male, BALB/c mice.

Genes with the highest sex-related differences in expression in human and BALB/c mouse MGs, and not located on sex chromosomes, are shown in Table 2. None of these differences are the same between humans and mice. Genes located on human sex chromosomes with the highest sex-related differences were family with sequence similarity 122B, lysine (K)-specific demethylase 6A, plastin 3 and shroom family member 4. All are located on the X chromosome. A number of mouse genes originating from X and Y chromosomes also expressed high sex-linked differences, including inactive X specific transcripts (female > male); and on the Y chromosome, male > female ratios were found for eukaryotic translation initiation factor 2, subunit 3, DEAD (Asp-Glu-Ala-Asp) box polypeptide 3, lysine demethylase 5D, and ubiquitously transcribed tetratricopeptide repeat gene. None of these differences are the same between humans and mice. Indeed, comparison of the 100 genes with the greatest sex-associated differences in human and mouse MGs also demonstrated that none are the same. Only 27 (11.2%) of all the genes exhibiting sex-related variations in the human MG were also altered in the BALB/c mouse MG. None of these genes with similar sex differences were located on sex chromosomes.

Table 2.

Highest sex-related differences in gene expression in human and mouse meibomian glands

Accession Number	Gene	Ratio	P value	Ontology
Human F > M
NM_006121	Keratin 1	6.98	0.0150	complement activation
NM_002275	Keratin 15	2.42	0.0051	epidermis development
NM_006390	Importin 8	2.15	0.0067	intracellular protein transport
NM_001017920	Death associated protein-like 1	2.07	0.0065	apoptosis
NM_001069	Tubulin, β2A	1.99	0.0263	protein folding
Human M > F
NM_000239	Lysozyme	18.18	0.0422	inflammatory response
NM_002652	Prolactin-induced protein	6.02	0.0422	protein binding
NM_199462	Dual serine/threonine and tyrosine protein kinase	2.05	0.0011	protein phosphorylation
NM_001042472	Abhydrolase domain containing 12	1.81	0.0018	response to stimulus
NM_017864	Integrator complex subunit 8	1.71	0.0124	snRNA processing
Mouse F > M
NM_177446	Androgen binding protein ζ	109.24	0.0416	binding
NM_207262	Androgen binding protein ε	89.85	0.0419	binding
NM_194338	Androgen binding protein γ	20.11	0.0136	binding
NM_020563	Androgen-binding protein η	18.48	0.0068	binding
NM_013809	Olfactory-specific steroid hydroxylase	9.04	0.0020	oxidation-reduction process
Mouse M > F
NM_148942	Serine (or cysteine) peptidase inhibitor, clade B, member 6c	3.13	0.0015	negative regulation of endopeptidase activity
NM_026682	Cleavage and polyadenylation specific factor 4-like transcript variant 1	2.51	0.0036	metal ion bimdog
NM_020277	Transient receptor potential cation channel, subfamily M, member 5	2.49	0.0071	Transport
NM_009022	Aldehyde dehydrogenase family 1, subfamily A2	2.19	0.0040	metabolic process
NM_145419	Hexokinase domain containing 1	2.18	0.0002	carbohydrate metabolic process

Genes with the highest sex-related differences in expression in human and mouse MGs, and not located on sex chromosomes, are shown in this Table. Relative ratios were calculated from Illumina non-transformed data by comparing the degree of gene expression in meibomian glands from females (F) and males (M). Genes listed had a comparative P value between groups of < 0.05 and a known identity. The data for several listed genes originate from a previous study.¹⁹

Of particular interest, over 96.5% of the significant sex-associated differences in gene expression in human and BALB/c mouse MGs were modest, and were equal to, or less than, a 2.0-fold amount.

Sex-associated differences in the gene expression of ontologies, KEGG pathways and chromosomes in human and mouse MGs

Sex exerted a significant impact on numerous biological process, cellular component, and molecular function gene ontologies in human and BALB/c mouse MGs (Table 3). However, most of these effects were highly species-specific (Tables 4 and 5). When sex did influence analogous gene ontologies in both human and mouse MGs, the direction was not necessarily the same in both species. For example, while the translational elongation, RNA binding and acid-amino acid ligase activity ontologies were significantly increased in the same manner in both humans and mice, the translation and ribosome ontologies were not. These latter ontologies were upregulated in human female and mouse male MGs.

Table 3.

Sex-related differences in the expression of gene ontologies (≥ 3 genes/ontology) in human and mouse MGs

Ontology	Human		Mouse
	Female ↑	Male ↑	Female ↑	Male ↑
Biological processes	40	19	866	506
Molecular functions	6	9	150	164
Cellular contents	26	13	144	141

The number of biological process, molecular function and cellular component ontologies with a z-score ≥ 2.0 are listed. A z-score is a statistical rating of the relative expression of genes, and shows how much they are over- or under-represented in a specific gene list.²³ Positive z scores reflect a higher number of genes meeting the criterion than is expected by chance, whereas negative z scores represent fewer genes meeting the criterion than expected by chance.23 Z-scores with values ≥ 2.0 or less than ≤ −2.0 are significant. Criteria for inclusion in this Table were an ontology containing ≥ 3 genes and a z-score ≥ 2.0. Most of the ontologies were different between males and females, but some (e.g. lipid metabolic ontologies) were upregulated in both sexes.

Table 4.

Influence of sex on the expression of gene ontologies in human MGs

Ontology	M Genes ↑	F Genes ↑	M z-score	F z-score
Biological Processes
Translational termination	1	4	0.7	4.81
Cellular protein complex disassembly	1	5	0.23	4.63
Translational elongation	1	4	0.58	4.46
Cellular macromolecular complex subunit organization	3	11	−0.06	4.44
Translation	2	9	−0.17	4.42
Golgi vesicle transport	5	1	4.9	0.24
Nucleoside metabolic process	3	1	3.78	0.76
Establishment of organelle localization	3	0	3.75	−0.7
Post-translational protein modification	4	0	3.25	−0.98
Nucleobase metabolic process	2	1	3.06	1.2
Molecular Functions
Nucleoside-triphosphate diphosphatase activity	1	2	5.43	10.96
Integrin binding	3	2	4.49	2.77
RNA binding	7	9	1.46	2.43
Structural constituent of ribosome	1	5	0.2	4.61
Structural molecule activity	1	9	−1.24	3.32
Ribonuclease activity	2	1	2.79	1.08
Ubiquitin-protein ligase activity	4	1	2.68	−0.16
Binding	70	66	2.62	1.28
Small conjugating protein ligase activity	4	1	2.53	−0.22
Acid-amino acid ligase activity	4	2	2.23	0.5
Cellular Contents
Catalytic step 2 spliceosome	0	4	−0.68	5.5
Ribonucleoprotein complex	3	11	0.11	5.25
Ribosome	1	6	−0.03	5.19
Macromolecular complex	23	33	1.11	4.19
Nuclear lumen	17	22	1.51	3.42
Ubiquitin ligase complex	4	0	3.66	−0.86
Endomembrane system	16	5	2.74	−1.08
DNA-directed RNA polymerase II, holoenzyme	2	1	2.39	0.94
Nuclear membrane-endoplasmic reticulum network	8	3	2.33	−0.23
Condensed chromosome, centromeric region	2	1	2.32	0.9

Biological process, molecular function and cellular component ontologies with some of the highest z-scores were selected following the analysis of non-transformed data. Criteria for inclusion in this Table were an ontology containing ≥ 3 genes and a z-score > 2.0 in a male or female sample. Those ontologies with a z-score > 2.0 are highlighted in bold. Terms: M Genes ↑ - number of genes up-regulated in male tissues; F Genes ↑ - number of genes up-regulated in female tissues; z-score - specific score for the up-regulated genes in the male and female tissues. The data for two listed ontologies originate from a previous study.¹⁹

Table 5.

Effect of sex on the expression of gene ontologies in mouse MGs

Ontology	M Genes ↑	F Genes ↑	M z-score	F z-score
Biological Processes
Regulation of catalytic activity	130	147	1.66	6.87
Regulation of metabolic process	347	340	1.13	6.37
Immune response	34	72	−2.03	6.25
Cell death	125	133	1.92	6.05
Cytoskeleton organization	34	66	−1.49	5.96
Oxidation-reduction process	123	47	8.09	−0.18
Macromolecule catabolic process	81	41	5.74	1.19
Intracellular transport	93	52	5.36	1.53
Cellular response to stress	91	47	4.51	0.33
Heterocycle metabolic process	122	79	3.83	1.44
Molecular Functions
Protein binding	457	491	1.35	9.84
Enzyme binding	78	88	1.64	5.35
Enzyme regulator activity	67	86	0.36	5.26
Gtpase regulator activity	30	49	−0.22	5.16
Actin binding	12	40	−2.57	5.07
Oxidoreductase activity	102	41	6.77	−0.34
Hydrolase activity	238	150	5.8	1.55
Cofactor binding	42	14	4.94	−0.58
Ligase activity	56	28	4.36	0.51
Adenyl nucleotide binding	155	104	4.11	1.57
Cellular Components
Cell periphery	193	336	−6.31	8.76
Plasma membrane	191	325	−6.1	8.33
Plasma membrane part	81	172	−5.41	6.62
Cytoskeleton	106	143	−0.51	6.57
Intracellular non-membrane-bounded organelle	204	205	1.72	5.8
Mitochondrion	257	76	14.11	−1.63
Mitochondrial part	117	30	11.75	−0.7
Mitochondrial inner membrane	72	15	9.67	−1.13
Organelle inner membrane	73	17	9.37	−0.88
Organelle envelope	111	50	9.32	1.99

Biological process, molecular function and cellular component ontologies with some of the highest z-scores were chosen after analyses of non-transformed data. Criteria for inclusion in this Table were a biological process ontology containing ≥ 100 genes, or molecular function and cellular component ontologies containing ≥ 50 genes, and a z-score > in a male or female sample. Those ontologies with a z-score > 2.0 are highlighted in bold. Terms are as described in the legend to Table 4.

Sex also significantly influenced the expression of KEGG pathways in human and BALB/c mouse MGs, but again the effect was primarily species-specific (Table 6). Of the 43 KEGG pathways upregulated in male mice, only 1 was the same as in humans. Of the 43 KEGG pathways upregulated in female mice, only 2 were identical in humans.

Table 6.

Impact of sex on the expression of KEGG pathways in human and mouse MGs

Ontology	M Genes ↑	F Genes ↑	M z-score	F z-score
Human
Pyrimidine metabolism	3	1	4.09	0.57
N-Glycan biosynthesis	2	0	3.91	−0.55
Base excision repair	1	1	2.35	1.86
Ribosome	1	4	1.01	4.78
Spliceosome	0	4	−0.75	3.73
Purine metabolism	2	3	1.64	2.15
Alzheimer’s disease	0	3	−0.85	2.1
Mouse
Metabolic pathways	161	70	6.8	−0.69
Peroxisome	23	3	6.17	−1.07
Oxidative phosphorylation	27	0	5.35	−2.92
Proteasome	13	3	4.69	0
Lysosome	26	7	4.68	−0.47
PPAR signaling pathway	16	5	3.54	−0.11
Leukocyte transendothelial migration	6	21	−1.49	4.93
NOD-like receptor signaling pathway	3	13	−1.13	4.59
Toll-like receptor signaling pathway	5	18	−1.49	4.43
Focal adhesion	13	26	−1.24	3.68
RIG-I-like receptor signaling pathway	5	12	−0.56	3.51
Adherens junction	5	12	−0.63	3.4

KEGG pathways were selected after analyses of non-transformed data. Human pathways contained ≥ 2 genes, and those of mice ≥ 16 genes. Pathways with z-scores > 2 are highlighted in bold.

In addition, sex had a significant impact on chromosomal gene expression in the MG (Supplemental Table 3). The sex influence was species-specific.

Discussion

In the present study we sought to determine whether sex-related differences in MG gene expression are the same, or almost the same, in BALB/c mice and humans. We found that none of the 100 genes with the greatest sex-associated differences in human and mouse MGs were the same. Indeed, less than 11.5% of the genes exhibiting sex-related variations in the human MG were altered similarly in the BALB/c mouse MG. In addition, most of the sex-linked differences in the expression of ontologies, KEGG pathways and chromosomal genes were species-specific. This lack of correlation also extended to the expression of the 500 most highly expressed, non-ribosomal genes, where fewer than 16 % were the same in human and mouse MGs. Overall, our results indicate that BALB/c mice are not optimal models for understanding sex-related differences in gene expression of the human MG.

Our observation that less than 11.5% the genes exhibiting sex-associated differences in the human MG were similarly altered in the BALB/c mouse MG was unanticipated. We also did not expect that the sex-linked expression of most gene ontologies and KEGG pathways would be so species-specific. However, our findings are not unique. Researchers have reported that only 13% of the sexually dimorphic genes in the human kidney (n = 67 genes) could be identified in the mouse kidney, and that only 0.8% of the sex-associated mouse renal genes (n = 1,162 genes) were the same as in humans.²⁵ In the same way, significant differences between humans and mice in their sex-related gene expression have been found during development, ^26,27 as well as in the liver and heart.^28,29 These findings have led some investigators to speculate that distinct mechanisms for the regulation of sexually dimorphic gene expression exist in humans as compared to mice.²⁸

Of particular interest, researchers have just reported that the gene expression profile in the tarsal plate of male and female C57BL/6J mice shows no statistically significant differences in the expression levels of the main protein-coding genes related to lipid metabolism, storage and meibogenesis.³⁰ In addition, they report no discernible sex-specific variations in the expression levels of the vast majority of the genes in these tarsal samples.³⁰ One interpretation of their results is that they represent the gene expression pattern in mouse MGs. In contrast, we have identified significant sex-related differences in the expression of over 2,400 MG genes in adult C57BL/6J mice,¹⁵ as well as over 3,600 genes in adult BALB/c mice (this study). These sex-related differences include many genes related to lipid dynamics, such as acetyl-coenzyme A acetyltransferases 1, 2 and 3, acyl coenzyme A reductase 1, acyl-coenzyme A wax alcohol acyltransferase 2, diacylglycerol o-acyltransferases 1 and 2, elongation of very long chain fatty acids like 4, fatty acyl-coenzyme A desaturase, lanosterol synthase, lipoamide acyltransferase component of branched-chain alpha-keto acid dehydrogenase complex, and stearoyl-coenzyme A desaturases 1, 2 and 4. Furthermore, we have identified significant (i.e. z-scores between 2.09 to 8.38) sex-associated differences in 36 lipid-related biological process ontologies (5 to 135 genes/ontology) in MGs of C57BL/6J¹⁵ and/or BALB/c mice (this study). These gene ontologies include such categories as lipid biosynthetic and metabolic processes, lipid homeostasis, lipid localization, lipid oxidation, lipid storage, and neutral and phospholipid biosynthetic and metabolic processes.

In effect, there are a multitude of significant sex-related differences in the expression of meibogenesis-associated genes in mouse MGs. A question, then, is why are our genetic data so different than those of Butovich et al?³⁰ There are a number of reasons, some of which include: [a] we isolated MGs from eyelids in our experiments, and did not use whole tarsal plates as in the Butovich et al study;³⁰ [b] we collected and then processed our BALB/c MG samples, as well as our C57BL/6J MG samples,¹⁵ at the same time, so as to control for day of collection and for timing of microarray procedures; the Butovich et al study³⁰ appeared to involve 3 separate tarsal collections and processings over a 6-month period [c] we used Illumina MouseWG-6 v.2 Expression BeadChips to quantitate gene expression in this study, and Amersham Biosciences/GE Healthcare CodeLink UniSet Mouse 20K I Bioarrays and Affymetrix GeneChip Mouse Genome 430A 2.0 Arrays for the C57BL/6J mouse studies,¹⁵ where as Butovich et al³⁰ used a Transcriptome Analysis Console. There are significant differences in the capabilities of microarray platforms to identify differences in gene expression;^15,24,31,32 and, very importantly, [d] we identified differences in gene expression by using a significance level of p< 0.05 and did not use a fold-change cutoff; Butovich et al³⁰ used a default setting of a fold-change ≥ 2. Our rationale for no fold-change cutoff was that the vast majority of sex-associated differences in gene expression in somatic tissues are modest (i.e. ≤ 1.2-fold).³³ This is true for 71.4% of sexually dimorphic genes in adipose tissue, 81.3% in liver, 82.5% in muscle and 94.4% in brain.³³ Similarly, over 96.5% of the sexually dimorphic genes in the BALB/c mouse MGs display modest differences in gene expression (i.e. ≤ 2.0-fold). Consequently, we, as others,³⁴ utilize unbiased gene expression signatures, because the use of a fold-change cutoff eliminates the possibility of finding small genetic differences that are biologically valid.

There are limitations in our study, given that age and ocular health differences may have contributed to the sex-related differences in MG gene expression between humans and BALB/c mice. However, the highest intensity genes were similar in the youngest and oldest women and men, as well as in 3- and 5-month-old BALB/c and C57BL/6J mice. Further, none of the highest sex-related differences in human gene expression in this study are the same as those reported between MGD patients and healthy controls.¹⁹ Similarly, almost none of the highest sex-associated differences in the expression of biological process, molecular function and cellular component ontologies are the same as those found between MGD patients and healthy controls.¹⁹

Many of the sex-related variations in gene expression in the MG may be due to the action of sex steroids.^2,6 Estrogens, progestogens and/or androgens are known to modulate the levels of numerous genes in human MG epithelial cells³⁵ and mouse MGs.^36–40 In addition, other factors may also contribute to the sex-associated variations in MG gene expression, including hypothalamic-pituitary hormones, glucocorticoids, insulin, insulin-like growth factor 1 and thyroid hormones, X chromosome gene dosage (e.g. X-inactivation), epigenetics and gender.² However, the impact of these factors in humans and mice may not be the same. For example, major differences are known to exist between the sex steroid systems of mice and humans.⁴¹ As another example, the control of X-inactivation is substantially different between species.²⁷ And gender, in turn, is distinctly human, and reflects socially constructed characteristics such as behaviors related to being masculine or feminine.² Our MG findings, then, are consistent with the recent ENCODE project report, which concluded that many gene expression patterns are not shared between mice and humans.^42,43

The lack of similarity in sex-related gene profiles also extended to the identities of the 500 most highly expressed, non-ribosomal genes in human and BALB/c mouse MGs. Less than 16% were the same. Such differences in gene transcription levels between humans and mice have also been discovered in other tissues, and may possibly reflect fundamental physiological differences between the two species.⁴³ Such genetic differences may also become amplified in disease states, where mice may be less reliable as models of human disease.⁴⁴ For instance, murine genomic responses to inflammatory challenges (e.g. trauma, burns, and endotoxemia) show very poor correlation to those of humans.⁴⁵ These findings underscore the need to demonstrate whether a mouse model mimics, or fails to mimic, humans in health or disease. ^45–47

It is possible that intrinsic differences in Illumina platform designs for the HumanHT-12 v.3 and MouseWG-6 v.2 Expression BeadChips contributed to the low concordance of gene expression between human and mouse MGs. Such microarray differences could be in probe length and content, deposition technology, labeling procedures, hybridizing protocols, image segmentation and signal detection, as well as in background correction, and data standardization and mining.^24,31,32,48. However, these possible differences may be limited, given that all array platforms were made by Illumina and over 21,000 transcripts coded for the same genes on both the human and mouse BeadChips. In addition, all data were generated by Asuragen, and all data were analyzed with the same GeneSifter software programs. Overall, as has been found in other studies, it is likely that most gene expression changes revealed by the human and mouse platforms are biologically correct, and that the differences are not due to technological variations.^15,24,32,49.

In summary, our data demonstrate that many significant sex-related differences exist in the gene expression of human and mouse MGs, but that most of these differences are not the same. Consequently, our results indicate that BALB/c mice are not optimal models for understanding sex-related differences in gene expression of the human MG. The question, then, is whether there is another species or murine strain that might serve as a relevant model to help clarify the factors underlying human MG sexual dimorphism? Such a model might help to explain not only sex-linked MG gene expression, but also why sex-associated differences exist in the morphology of the human MG,^50–54 and in the quality, expressibility, lipid profiles and apparent secretion of human meibum.^55–58 Such a model might also shed light on the question as to whether sex-related differences do ^52,59–62, or do not^55,63–66, exist in the prevalence and/or severity of MGD or MG loss. This understanding is very important, given that MGD is a major cause of DED,^6–8 and that DED occurs predominantly in women.⁹.

Supplementary Material

Supplemental Data File 1

Supplemental file 1, Female mouse sample 1

Supplemental Data File 2

Supplemental file 2, Female mouse sample 2

Supplemental Data File 3

Supplemental file 3, Female mouse sample 3

Supplemental Data File 4

Supplemental file 4, Male mouse sample 1

Supplemental Data File 5

Supplemental file 5, Male mouse sample 2

Supplemental Data File 6

Supplemental file 6, Male mouse sample 3

Supplemental Table 1

Supplemental Table 1. Highest intensity ribosomal genes in meibomian glands of female and male humans and BALB/c mice

Supplemental Table 2

Supplemental Table 2. Highest intensity genes that are the same in meibomian glands of female and male humans and BALB/c mice

Supplemental Table 3

Supplemental Table 3. Effect of sex on chromosomal gene expression of in human and mouse MGs