Physiological Effects of Soat1 Inactivation on Homeostasis of the Mouse Ocular Surface

Abstract

Purpose: Soat1/SOAT1 have been previously reported to be critical for the biosynthesis of cholesteryl esters (CEs) in the mouse Meibomian glands (MGs) as the loss of function led to an arrest of CE production and a substantial accumulation of nonesterified cholesterol in the meibum, causing an increase in its melting temperature. The purpose of this study was to further investigate the role of Soat1 in meibogenesis and ocular surface physiology.

Methods: The mouse ocular features of knockout Soat1^−/− and wild type (WT) mice were studied using various ophthalmic and histological techniques, mouse lipidomes were monitored using liquid chromatography/mass spectrometry, whereas their transcriptomes were compared to characterize the effects of the mutation on the gene expression profiles in the MG and cornea.

Results: Soat1^−/− mice displayed increased tear production and severe corneal abnormalities, such as corneal thinning, (neo)vascularization, ulceration, and opacification that progressed with aging. Transcriptomic analyses led to identification of a range of significantly disrupted pathways, which included general and specific lipid metabolism-related pathways, keratinization, angiogenesis/(neo)vascularization, muscle contraction, and several other pathways. In addition, histological and histochemical experiments revealed morphological changes in the MG, cornea, and conjunctiva in Soat1^−/− mice. Notably, the mRNA microarray expression level of Soat1 in WT MGs (log2 17.5) was 1000 × of that in the mouse cornea (log2 7.5).

Conclusions: These findings suggest a direct involvement of Soat1/SOAT1 in MGs in maintaining ocular surface homeostasis, in general, and corneal health, specifically.

Meibomian gland dysfunction (MGD) and dry eye disease (DED) are widespread pathological conditions that affect the world population regardless of ethnicity, sex, and habitat. In the United States, the prevalence of MGD was estimated to be 21% of the entire population,¹ while, in Africa, it is close to 46%.² In the elderly (85+ years) population in Russia, the MGD prevalence was reported to be 69%,³ while DED was diagnosed in 27% of those surveyed with a mean age of 58.5 ± 10.5 years.⁴ In New Zealand, MGD was diagnosed in 61% of recruited subjects,⁵ while, in Japan, symptomatic and asymptomatic MGD were found in 18% and 47% of the study subjects.⁶ In India, between 15% and 32% of the population have been reported to have MGD and/or DED.^7–9 

MGD and DED are multifactorial diseases in which ocular homeostasis is disrupted leading to a variety of signs and symptoms, such as ocular desiccation, inflammation, swollen eyelids, corneal abrasions and ulceration, Meibomian gland (MG) dropout, accumulation of lipid deposits on the eyelid margins, and a range of other abnormalities. Routinely, DED is associated with either insufficient or excessive production of aqueous tears (hypo- or hypersecretory DED, respectively), obstructed MG orifices (also known as obstructive MGD), excessive production of meibum (i.e. hypersecretory, or seborrheic, MGD), and other ocular pathologies associated with MG abnormalities. 

The exocrine MGs reside within the tarsal plates (TPs) of the eyelid synthesizing and excreting a unique mixture of lipids called meibum. Meibum is important for protecting the ocular surface by forming the tear film lipid layer (TFLL) which prevents tear evaporation, aids visual acuity, and lubricates the ocular surface, among other functions. Changes in the quantity and/or quality of meibum result in MGD, leading to alteration of the tear film (TF), eye irritation, and various ocular pathologies. Multiple studies, for example,^10–18 demonstrated that alterations in the lipid profiles of meibum caused by inactivating mutations in specific genes of meibogenesis^19,20 led to MGD and DED-like signs in laboratory animals. 

We previously reported that the gene Soat1 that encodes enzyme sterol O-acyltransferase-1 (SOAT1) is highly expressed in the mouse MGs and is an essential enzyme of meibogenesis, converting free (i.e. nonesterified) cholesterol (Chl) and fatty acids (FAs) into very and extremely long chain cholesterol esters (VLC/ELC CEs). The latter are vastly different from regular, much shorter, CEs found in plasma and other tissues.¹¹ Notably, the VLC/ELC CEs dominate the CE pool of normal meibum in humans and mice.¹⁹ Inactivation of Soat1 in the gene knockout (Soat^−/−) mice resulted in a massive shift in the balance of Meibomian lipids (MLs) with a complete loss of VLC/ELC CEs and a massive increase in Chl. Although CEs are the second most abundant lipids in meibum in humans and mice, trailing only wax esters (WEs), their specific roles in maintaining MGs and ocular surface homeostasis are yet to be evaluated. The purpose of this study was to correlate changes in the ML profiles and MG and cornea transcriptomes with associated pathological changes in the MGs and the ocular surface morphology of homozygous Soat1^−/− mice. 

Methods

Experimental Animals

All animal procedures used in this study were approved by the Institutional Animal Care and Use Committee of UT Southwestern Medical Center (protocol # 2016-101549, approved July 14, 2023) and were conducted in accordance with the Association for Research in Vision and Ophthalmology Guidelines. Heterozygous Soat1^+/− mice were purchased from the Jackson Laboratory (B6.129S4-Soat1^tm1Far/Pgn; stock #007147; Bar Harbor, ME, USA) and bred to yield Soat1^+/+ (WT) and homozygous Soat1^−/− knockout mice. Mice were genotyped using 3 primers described in the Jackson Laboratory protocol #24675. Common forward primer 18958 (TGC TGA CGT CTT CCT GTG TC), WT reverse primer 18959 (GAG CTG TTG GG AGT AGG TG), and mutant reverse primer oIMR6218 (CCT TCT ATC GCC TTC TTG ACG) provided the expected bands 400 bp (mutant) and 240 bp (WT). All mice were maintained on a 12-h light/dark cycle with ad libitum access to food and water. Age matched WT and mutant mice were used for all experiments. Two to 3-month-old mice were used for transcriptomic analysis of the TPs and corneas, as described before for mouse MG.²¹ 

Mouse Ocular Evaluation

Mice were anesthetized using ketamine/xylazine cocktail (120 mg/kg and 16 mg/kg, respectively) and mouse ocular features were assessed using a slit lamp model BQ 900 (Haag-Streit USA, Mason, OH, USA). The ocular surfaces of mice were photographed using a digital camera. Sodium fluorescein staining was used to visualize the anterior surface of the eye. Approximately 0.25 µL of 2% sodium fluorescein solution (Fluorescein-PF 2%, from Greenpark Compounding Pharmacy, Houston, TX, USA) was administered at the medial canthus, the eye was gently closed to distribute the dye, and subsequently washed with Hanks balanced salt solution with calcium chloride and magnesium chloride (HBSS; Gibco, Grand Island, NY, USA) to remove residual dye. The degree of cornea staining was measured using the cobalt blue filter and graded following the NEI/Industry workshop guidelines,²² and the total fluorescein score for each eye was recorded. One week later, tear production was measured using Zone-Quick phenol red thread (PRT; from Ocusoft, Rosenberg, TX, USA). Mice were lightly anesthetized using ketamine/xylazine cocktail and the thread was placed in the inferior conjunctival fornix near the lateral canthus for 15 seconds using jeweler's forceps. Tear production was measured using the scale provided on the packaging of the thread. 

Mouse Tissue Collection

Mice were euthanized by isoflurane followed by cervical dislocation. The TPs were excised from the eyelids using a Zeiss Stemi 508 Stereo dissecting microscope (Carl Zeiss, Oberkochen, Germany) and extraneous tissue, such as muscle and conjunctiva, were carefully removed, as described earlier.¹⁹ The TP specimens were placed in 200 µL of RNAlater (Qiagen, Germantown, MD, USA) and stored at −20°C until the samples were taken to the Genomics and Microarray Core facility of the University of Texas Southwestern Medical Center (UTSW) for mRNA extraction and analysis. Only samples with RNA Integrity Numbers of ≥8 were analyzed. For cornea mRNA collection, the anterior chamber was punctured above the limbus using a 25g needle. To isolate the cornea, we used Vannas scissors to cut around the circumference of the globe above the limbus. The corneal button was briefly rinsed in HBSS to remove any residual iris material and/or fur, placed in 200 µL of RNAlater, and stored at −20°C. The eyes were enucleated and the extra tissue was removed and weighed to obtain eyeball weights. 

Histological Evaluation of Mouse Tarsal Plates and Meibomian Glands

TPs and eyeballs were excised using the Zeiss dissecting microscope, fixed in Carson's formalin for at least 24 hours, and dehydrated in successive passages through 50%, 70%, 95%, 100% ethanol, and xylene prior to paraffin embedding. Tissue sections were cut at 4 µm using a microtome (model Shandon Finesse 325; from Thermo Shandon, Cheshire, UK) and stained with hematoxylin and eosin (H&E; StatLab, McKinney, TX, USA) or periodic acid Schiff's (PAS; StatLab, McKinney, TX, USA). 

For Oil Red O (ORO; Thermo Scientific, Waltham, MA, USA) staining, fresh TPs were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight at 4°C. Samples were passed through 10%, 20%, and 30% sucrose in PBS before embedding in the Tissue-Tek optimal cutting temperature (OCT) compound (Sakura Finetek USA, Torrance, CA, USA). Then, 10 µm sections were cut using a cryostat (model Leica CM3050S; from Leica Biosystems, Deer Park, IL, USA) and stored at −20°C until staining. The 0.5% ORO stock solution was prepared by dissolving 0.5g ORO in 100 mL iso-propanol (IPA) and heated to 56°C for 1 hour. The working ORO solution was made prior to staining by diluting stock solution with de-ionized water (6:4, vol/vol), which sat for 10 minutes before filtering. Tissue slides were removed from the freezer, air-dried for 15 minutes, washed with PBS thrice for 5 minutes, and then rinsed with de-ionized water. The slides were stained with the freshly made ORO working solution for 6 minutes, washed with water twice and mounted with VectaShield with DAPI (4′,6-diamidino-2-phenylindole; Vector Laboratories, Newark, CA, USA). ORO staining was examined by fluorescent microscopy using the Texas Red stain with a 595 nm excitation filter, and DAPI with a 360 nm ultraviolet excitation filter. 

Conjunctival flat mounts were prepared from the upper and lower eyelid conjunctival sheets. The lateral canthus was cut to access the conjunctiva at the upper and lower eyelids joint. The conjunctiva was held on to as scissors were used to make a cut right below the limbus and proceeding along the palpebral conjunctiva to the medial canthus as one upper lid sheet that comprised of the bulbar conjunctiva, fornix, and palpebral conjunctiva. Then, the tissue specimen was placed flat on a slide with the conjunctiva side down and left to dry at room temperature. The lower conjunctival sheet was collected in the same manner. Each mouse produced four conjunctival flat mounts. Once the conjunctival tissue dried, the tissue was peeled off leaving behind a thin layer of cells that adhered to the slide. The samples were fixed with Carson's formalin for 15 minutes, washed with water, and then stained with PAS using the manufacturer's protocol. Imaging of tissue specimens (WT, n = 8 and Soat1^−/−, n = 10) was conducted using a Zeiss AXIO Observer D1 microscope (Carl Zeiss, Oberkochen, Germany). To eliminate any bias, blind tests were conducted for image acquisition and goblet cell counting. Goblet cells were counted using ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.net/ij/). Only goblet cells with well-defined cell bodies were counted. 

Lipidomic Analyses

Lipidomic analyses of the study ML samples were conducted using previously described ultra-high-pressure isocratic liquid chromatography—high resolution mass spectrometry (LC-MS) protocols^11,23 with no modifications, unless indicated otherwise. Briefly, 2 to 4 excised mouse TPs from individual mice were extracted with 3 × 1 mL of methanol : chloroform = 1 : 2 (vol/vol) solvent mixture at 36°C. The three lipid extracts were pooled and the solvent was evaporated under a gentle stream of ultra-high pure nitrogen at 36°C. The specimens were redissolved in 1 mL of IPA and stored in sealed 2-mL autoinjector vials at −80°C until the analyses. 

All LC-MS equipment – a µSample Manager-FL, a µBinary Solvent Pump, an Auxiliary Solvent Pump, a Trap Valve Manager, a Synapt G2-Si high resolution quadrupole Time-of-Flight mass spectrometer, and all LC-MS analytical software – were manufactured by Waters Corp. (Milford, MA, USA). The LC-MS experiments were carried out using a MassLynx version 4.1 software package (from Waters). The lipids were separated on a C8 BEH Acquity column (1.7 µm, 2.1 mm × 100 mm) in isocratic mode. The eluent was composed of a mixture of 95% IPA and 5% of 10 mM ammonium formate in water. To reduce the backpressure of the LC system and improve the shapes of the peaks, the LC column was thermostated at 40°C. The flow rate was maintained at 16 µL/min by the µBinary Solvent Pump. Between 0.4 µL and 1 µL of samples were injected. The analytes were detected in the atmospheric pressure chemical ionization (APCI) sensitivity mode at R ≥ 10,000 full-width-at-half-maximum (FWHM) resolution using an IonSabre II APCI ion source and a ZSpray/LockSpray housing unit (from Waters). The 0.5 µg/mL leucine-enkephalin LockSpray solution in acetonitrile : water = 1:1 (vol/vol) with 0.1% formic acid as additive was pumped by the Auxiliary Solvent Pump at a flow rate of 1.5 µL/min. The analytes were detectable in positive ion mode (PIM) almost exclusively as major (M + H)⁺, (M + H – H₂O)⁺, or (M + H – FA)⁺ adducts, unless stated otherwise. Formation of sodium, potassium, and ammonium adducts of the analytes was minimal, or nonexistent. The chemical nature of the analytes was derived from their mass spectra using the EleComp and MS^E DataViewer software packages and confirmed in tandem mass spectrometry (MS/MS) fragmentation experiments. 

The corneal lipids were analyzed as follows. The cornea buttons (2 from each mouse; 2 mm in diameter) were excised and extracted the same way as TPs. The dry lipid material was dissolved in 0.5 mL of IPA and stored in sealed 2-mL glass HPLC vials at −20°C until the analyses. 

The cornea lipidomes were evaluated using APCI and electrospray ionization (ESI) approaches. The LC-MS-APCI experiments were conducted as described above for TP lipids. The LC-MS-ESI experiments were performed using an ESI Low-Flow ion source (from Waters). A C18 BEH Acquity column (1.7 µm, 1 mm × 100 mm) was used instead of the C8 column. The analytes were eluted with a ternary gradient of IPA, acetonitrile, and a 5% aqueous solution of ammonium formate as described before for TP lipids.^11,23 

The data were analyzed in the MassLynx (version 4.1) and Progenesis QI (version 2.3) software packages. Statistical analyses were performed in Progenesis QI, SigmaStat (version 3.5) and SigmaPlot (version 11.0) statistical analysis suites from Systat Software Inc. (San Jose, CA, USA). 

Transcriptomic Analyses

Transcriptomic analyses of TPs and cornea samples were performed at the UTSW Genomics and Microarray Core facility using MTA-1 mRNA microarray chips (from Affymetrix/Applied Biosystems, purchased from Thermo Fisher Scientific, Santa Clara, CA, USA). The RNA Integrity Numbers (RINs) were between 8 and 10 for all study samples. The raw data was initially processed in the Expression Console software version 1.4 (from Affymetrix) and then exported to the Transcriptome Analysis Console version 4.02 (from the same manufacturer). The following parameters were used: Array Type: MTA-1_0; Analysis Type: Expression (Gene); Analysis Version: version 2; Positive versus Negative area under the curve (AUC) threshold: 0.7; Genome Version: mm10 (Mus musculus); Annotation: MTA-1_0.r3.na36.mm10.a1.transcript.csv; Map File: MTA-1_0_MappingFile.r1.map; Gene-Level Fold Change <−2 or >2 (unless stated otherwise); Gene-Level P value <0.05; and ANOVA Method: Ebayes. 

Results

Soat1^−/− Mouse Ocular Phenotype

Under slit lamp examination, WT mice demonstrated no changes in their ocular health and cornea clarity with aging (Figs. 1A–1C). Conversely, even young Soat1^−/− mice displayed corneal pathologies, such as moderate corneal opacification, irregular corneal surface, excessive pooling of tears, and crusty eyelid margins, all of which progressed with aging (Figs. 1D–1F). Using the slit beam to evaluate the location of the corneal haze was difficult due to the hyper-reflective cornea and the slit eye phenotype of the Soat1^−/− mice, but the cornea appeared thinner with hazing near the anterior stroma and epithelium. These abnormalities progressed to severe corneal opacification, neovascularization, and ulcerations that began to appear in 5-month-old mice with half of the mice developing severe cornea damage by 12 months of age (Fig. 2). In addition, the eyelids showed progressive changes with age with features such as irregular lid margin, lashes that arise from the lid margin, upturning of the lashes toward the cornea, loss of lashes, and ulcerated eyelids (see Figs. 2C, 2D). The eye lashes and lid margins had crusty lipid accumulations and thick meibum similar to those observed in human MGD associated with chronic blepharitis,²⁴ Meibomian seborrhea,²⁵ and/or seborrheic dermatitis.²⁶ 

Figure 1.

 

Ocular phenotypic abnormalities of Soat1^−/− mice at 3 months, 7 months, and 12 months of age. WT slit lamp images at ages 3 months (A), 7 months (B), and 12 months (C) with no abnormalities to the ocular surface. (D, E) Soat1^−/− slit lamp images of corneal abnormalities. (D) Three-month-old Soat1^−/− mouse with an uneven corneal surface (white arrow), hyper-reflective cornea, and lipid accumulation around the eyelid margin (red triangle). (E) Seven-month-old Soat1^−/− with corneal haze (white arrow). (F) Twelve-month-old Soat1^−/− cornea with loss of clarity (white arrow), rough corneal surface, lipid particles (red triangle), and an irregular lid margin.

Figure 2.

 

Progressive destruction of the cornea and eyelid of Soat1^−/− mice with age. (A) Four-month-old Soat1^−/− mouse with advanced corneal haze (yellow star). (B) Six-month-old mutant mouse with a corneal ulcer, neovascularization (red arrow), loss of clarity, and lipid debris (red chevron). (C) Eleven-month-old Soat1^−/− with deterioration of the lower eyelid with lash loss and large crusty accumulation (black arrow). (D) Soat1^−/− 12-month-old ulcerated cornea (yellow arrow) with changes to the lid margin and lash growth direction (white chevron).

Fluorescein staining of the cornea was used to identify epithelial damage, as epithelial defects are associated with increased risk of corneal ulcers, haze, and decreased vision. WT mice had minimal to no punctate staining of the corneal epithelial cells (Figs. 3A, 3B). Soat1^−/− corneas revealed substantial coalescent patchy staining of the epithelium, most significant in the central cornea (Fig. 3C). In addition, there was a significant uptake of fluorescein in the conjunctiva and MG orifices (see Fig. 3C). In some instances, the Soat1^−/− mice fluorescein staining diffused into the stroma and anterior chamber, indicating severe damage to the corneal epithelium or tight junctions because the stain was able to penetrate the epithelial barrier (Figs. 3D). Soat1^−/− mice had a significant increase in their fluorescein scores compared to WT mice (Fig. 3E). 

Figure 3.

 

Fluorescein staining and ocular characteristic. (A) Three-month-old WT mouse with sparse epithelial staining (white arrowhead). (B) Sparse punctate staining (white arrowhead) of the corneal epithelium in a 12-month-old WT mouse. (C) Three-month-old Soat1^−/− mouse with numerous coalescent staining (white arrowhead) of the epithelium. (D) Twelve-month-old Soat1^−/− cornea with significant staining of the epithelium and diffusion of the dye into the stroma (white arrowhead). (E) Mutant mice (n = 20) had significant fluorescein staining of the cornea compared to WT controls (n = 18, P = <0.001). (F) Tear production of Soat1^−/− mice (n = 24) was increased compared to WT mice (n = 20, P = <0.001). (G) Eyeball weights were not significantly different between Soat1^−/− (n = 10) and WT (n = 10) mice (P = 0.588).

During the slit lamp examination, we observed excessive tear pooling around the eyelid margins of Soat1^−/− mice. The tear production of Soat1^−/− mice, measured using the PRT test, was increased compared to WT mice (Fig. 3F). Eyeballs of WT and Soat1^−/− mice were weighed to determine if the slit eye phenotype was a result of a reduced eyeball size, but no significant differences between the two genotypes were observed (Fig. 3G). 

Altered Meibomian Gland Morphology of Soat1^−/− Mice

To examine whether structural or functional changes underlie the observed phenotype, TPs of WT and Soat1^−/− mice were processed for histological evaluation. WT mice had the typical MG characteristics with large abundant acini that secrete lipid into the central duct through narrow orifices. A magnified view of the WT acini shows large foamy meibocytes with cytoplasm rich in lipid droplets. The latter meibocytes undergo maturation and disintegration within the connecting duct leading to the central duct (Figs. 4A–4C). However, close examination of the TPs of Soat1^−/− mice revealed significant changes in gland morphology caused by the mutation. Soat1^−/− meibocytes were abnormally small and irregular with the loss of the uniform foamy lipid-laden cytoplasm and had variously sized lipid droplets with cytoplasmic clefts (Figs. 4D–4F). While the acini of mutant mice were smaller than those in WT mice, the Soat1^−/− MGs were large and the number of acini was still high. The central duct and MG orifices of Soat1^−/− mice were dilated and had inflammatory cells within the duct and ductal epithelium. In addition, the mutant ductal epithelium was thickened and hyperplastic. There was also accumulation of eosinophilic debris within the central duct. Although Soat1^−/− MG acini appeared atrophic, the inspissated meibum did not result in MG dropout or shortening of the glands. 

Figure 4.

 

Morphologic alterations to the meibomian glands due to Soat1 inactivation. (A) WT tarsal plate at 3 months with the typical appearance of the MGs in which the plump acini (a) are connected to the central duct by connecting ducts and small orifices at the lid margin (red circle). (B) The WT MGs had a narrow central duct (red line) and thin ductal epithelium (pink arrow). (C) A magnified view of the WT acini shows large meibocytes with cytoplasm that are rich with lipid droplets (yellow arrow). (D, E, F) Soat1^−/− 3-month-old tarsal plate with an altered morphology of the MGs. (D) The MG orifices were enlarged (red circle) with infiltration of inflammatory cells of the central duct (yellow arrowheads) and prominent branching of the connecting ducts (black chevron). (E) The ductal epithelium was thickened (pink arrow) with inflammation (yellow arrowhead) and enlarged central ducts (red line). The MGs had numerous acini (a) that were smaller and lacked the normal foamy cytoplasm of the acini. (F) Mutant meibocytes were small and irregular with varying sizes of lipid droplets (yellow arrow) throughout the acini and some meibocytes had fine needle-like spaces of the cytoplasm (green arrow). Hematoxylin and eosin (H&E) staining. Scale bars in panels A, B, D, and E = 100 µm; in panels C and F = 20 µm.

To correlate the altered meibum production with the morphological changes of the MGs, TPs of WT and Soat1^−/− mice were stained with ORO. TPs of controls displayed lipid-rich MGs with a uniform distribution of lipid staining from the acini to the central duct (Fig. 5A). The lipid droplets were visible in the meibocytes of the acini until the connecting duct in which they released their contents as meibum. The lipid staining of Soat1^−/− MG acini was significantly reduced with a majority of the staining occurring in the connecting ductules and the central duct. The size of Soat1^−/− meibocytes varied widely, and so did lipid accumulations between neighboring cells and non-staining areas of the cytoplasm (see Fig. 5). The maturation of the Soat1^−/− meibocytes also appeared to proceed differently from WT meibocytes as the former retained their cellular shape and nuclei further into the connecting ductules, as shown in Figure 5. The lipid staining of Soat1^−/− meibocytes in the ductules was significantly more pronounced than the staining of WT meibocytes that underwent cellular degradation earlier in the duct. 

Figure 5.

 

Oil Red O lipid staining of the Meibomian glands. (A) WT meibomian glands had ample ORO staining throughout the acini and ductal system (green arrow). (B) Soat1^−/− meibomian glands with sparse ORO lipid staining of the acini with most lipid staining within the central duct and connecting ducts (green arrow). (C) WT meibocytes underwent disintegration within the duct (white chevron). (D) Soat1^−/− meibocytes appear diverse with variations of lipid droplet accumulations (yellow arrow) and non-staining clefts in the cytoplasm (white arrow). Individual meibocytes could be seen in the connecting duct as the lipid highlighted cell shape and retained nuclei (white chevron). Scale bars = 50 µm.

Histopathological Changes of the Ocular Surface Due to the Inactivation of Soat1

Corneas of WT and Soat1^−/− mice were processed for histology to investigate if morphological changes underlie the phenotypic changes seen in vivo with the slit lamp examination. WT mice maintained their ocular health and displayed the normal morphology of the epithelium and stroma (Figs. 6A, 6B). Low power images of corneal sections revealed the reduced corneal thickness of Soat1^−/− mice compared to the WT corneas (Fig. 6C). The thickness of the stroma and epithelium varied as the epithelial/stroma interface was irregular with a jagged appearance compared to the uniform thickness of the WT corneas. The average WT central total corneal thickness was 128.15 µm and epithelial thickness was 32.82 µm, while Soat1^−/− cornea was significantly thinner with the total corneal thickness of 90 µm and epithelial thickness of 19.30 µm (P <0.001). The epithelium of the Soat1^−/− corneas was abnormal with flattened basal epithelial cells with necrotic cells and epithelial vacuoles. The stroma appeared condensed and hyper eosinophilic and was prone to large stromal clefts even though they were processed and stained alongside WT corneas (Fig. 6D). These corneal abnormalities progressed to severe deterioration of the cornea with epithelial vacuoles, stromal vacuoles, infiltration of inflammatory cells, hydropic change of basal cells, and focal edema in the area of epithelial loss (Figs. 6E, 6F). Evaluation of corneal ulcers showed complete loss of the central epithelium, neovascularization, fibrosis, and significant inflammation that lacked visible keratinization (Fig. 6G). 

Figure 6.

 

Histopathologic changes to the cornea. (A) Three-month-old WT cornea with the typical morphology of the stroma and epithelium. (B) WT cornea with a smooth junction between the stroma and epithelium (red dashed line) and healthy basal epithelial cells (black arrow). (C) Soat1^−/− 3-month-old cornea is noticeably thinner than the WT cornea (black box). (D) The basal epithelial cells are flat (red arrow) with areas of apoptotic cells (black arrows) and a jagged interface between the stroma and epithelium (red dashed line). (E) Eight-month-old Soat1^−/− cornea with epithelial vacuoles (red arrows), focal edema (star), and epithelial deterioration (black arrowhead). (F) Epithelial degeneration (red chevron), hydropic change of the basal cell (yellow arrowhead), and sub epithelial vacuoles (green arrowhead). (G) Twelve-month-old ulcerated (yellow circle) cornea with neovascularization (red arrowhead), fibrosis, and inflammatory cells (black star). Hematoxylin and eosin (H&E) staining. Scale bars in panels A, C, and G = 100 µm. Scale bars in panels B, D, E, and F = 50 µm.

Based on the phenotypic changes seen in the cornea, we stained the tissue samples with PAS to visualize the basement membrane (BM). The BM in WT corneas was a continuous band across the peripheral and central cornea at the epithelium and stroma junction. However, Soat1^−/− corneas did not have uniform staining of the BM, with frequent breaks and/or accumulations of PAS positive material in the central cornea, but maintained BM staining of the peripheral cornea (Figs. 7A, 7B). Conjunctival goblet cells were also evaluated with PAS staining. Due to the variations in goblet cell density and distribution in the conjunctiva, the middle fornix area of whole flat mounts were analyzed for a more accurate quantitation of the conjunctival goblet cells. As shown in Figures 7C and 7D, the goblet cell density increased in Soat1^−/− mice compared to 3-month-old WT mice. 

Figure 7.

 

Periodic acid—Schiff staining of the corneal basement membrane and conjunctival goblet cells. (A) WT basement membrane was uniform across the entirety of the cornea (white arrow) compared to ( B) the Soat1^−/− cornea with frequent breaks (red arrow) in the BM and areas with accumulations of PAS positive material (white arrow). (C) WT conjunctival goblet cells stained with PAS. (D) Soat1^−/− conjunctival goblet cells that had increased goblet cells in the conjunctival fornix. (E) WT conjunctiva had less goblet cells compared to the goblet cells in the Soat1^−/− conjunctiva. PAS staining. Scale bars = 50 µm.

Comparative Analyses of Free Cholesterol and Cholesteryl Esters in Mouse Corneas

A detailed analysis of Chl and CEs of WT and Soat1^−/− TP lipidomes was performed earlier.¹¹ For this follow-up paper, Chl and CEs of mouse corneas have been evaluated. The LC-MS experiments demonstrated that WT corneas were highly enriched in free, nonesterified Chl, whereas CEs were represented only by small amounts of C₁₈₋, C₂₀₋, and C₂₂–CEs, and even smaller amounts of VLC/ELC CEs. Four major cornea CEs were identified as C_18:1−, C_18:2−, C_20:4−, and C_22:6–CEs in both genotypes (Figs. 8A, 8B). The apparent Chl-to-total CE ratio in WT mice, calculated from the LC-MS abundances of their common (Chl – H₂O + H)⁺ and (CE – FA + H)⁺ fragment C₂₇H₄₅⁺ with m/z 369.351, was 30 : 1, while the apparent molar ratio of plasma-type CEs-to-meibomian VLC/ELC CEs in the same samples varied from 6.5 : 1 to 100 : 1, or higher. For comparison, the latter molar ratio in mouse TPs was 1: 6 or lower. 

Figure 8.

 

LC-MS analysis of cornea cholesteryl esters (CEs) of wild type (WT) and Soat^−/− mice. Extracted ion chromatograms of the Chl and CE common analytical ion with m/z value of 369.35 are shown. (A) Overall CE profile of WT corneas. Non-esterified cholesterol (Chl) is detected as the major species in APCI PIM. (B) Normal length plasma-type CEs detected in all WT and Soat^−/− corneas in ESI PIM: (1) C_18:1−, (2) C_18:2−, (3) C_20:4−, and (4) C_22:6–CEs. Note that ESI PIM is more sensitive for CEs than free Chl, producing a much smaller peak of Chl. (C) The fraction of very and extremely long chain CEs (VLC/ELC CEs) in WT mice is present, but widely varies from sample to sample. (D) The fraction VLC/ELC CEs in Soat^−/− mice is non-existent in all tested Soat1^−/− corneas.

Importantly, the VLC/ELC CEs (i.e. C₂₂₋ to C₃₄–CEs) were virtually undetectable in all Soat1^−/− cornea specimens (Figs. 8C, 8D), which came as no surprise due to the complete suppression of the VLC/ELC CE biosynthesis in the MGs of Soat1^−/− mice.¹¹ Remarkably, no differences among the profiles of 4 major corneal CEs of WT and Soat1^−/− mice were observed – the same C_18:1−, C_18:2−, C_20:4−, and C_22:6–CEs were dominant species in the corneas of both genotypes, and their apparent ratios were almost identical (Fig. 9). The only CE that differed between WT and Soat1^−/− mouse corneas was C_18:2–CE, but the difference, although statistically significant with P = 0.007, was very small and is doubtful to have any significant physiological impact. 

Figure 9.

 

Relative abundance of normal length plasma-type C_18:1−, C_18:2−, C_20:4−, and C_22:6–CEs in corneas of wild type (wt; n = 10) and Soat^−/− (ko; n = 8) mice.

The molecular profiles of VLC/ELC CEs in WT corneas closely replicated the profiles of those in meibum (Figs. 10A, 10B). Notably, the overall presence of meibomian-type VLC/ELC CEs in WT mouse cornea samples varied from sample to sample – in some cases, they were undetectable, whereas, in the others, meibomian-type CEs comprised up to 4% of the total CE pool. For these reasons, corneal ELC CEs were tentatively attributed to the remnants of MLs that originated from the TFLL, but not the corneas themselves. This hypothesis was further tested in transcriptomic experiments (see below). 

Figure 10.

 

Comparative analysis of very and extremely long chain cholesteryl esters (VLC/ELC CEs) observed in meibomian glands and corneas of wild type mice. Extracted ion chromatograms of an analytical ion m/z 369.35 are shown. (A) An LC/MS ESI PIM profile of VLC/ELC CEs obtained directly from meibomian glands. (B) VLC/ELC CEs detected in the corneas of wild type mice. Representative chromatograms are shown. Note a very close similarities of the CE profiles in both types of samples.

Transcriptomic Analysis of Tarsal Plates and Corneas of Wild Type and Soat1^−/− Mice

Comparison of Tarsal Plate and Cornea Transcriptomes of Wild-Type Mice

Next, the entire transcriptomes of WT TPs and corneas were compared using scatter and Volcano plots (Figs. 11A, 11B). There were about 66,000 unique transcripts identified, of which approximately 39,000 passed the selection criteria of a P <0.05 and a linear fold change (LFC) of −2 ≥ LFC ≥ 2. Of these TP transcripts, 4625 were found to be upregulated, while 34,327 were downregulated. Clearly, the transcriptomes of TPs and corneas bared no resemblance to each other and were further evaluated using more selective criteria. To reduce the massive number of transcripts that were to be analyzed, the non-coding, micro-RNA, tRNA, small RNA, ribosomal, and unassigned transcripts were filtered out, resulting in approximately 16,000 known protein-coding, multiple-complex, and pseudogenes, which were included in the subsequent, more focused, analyses. 

Figure 11.

 

Comparative analysis of mouse tarsal plates (MTP; shown in red) and cornea (MC; shown in blue) transcriptomes of wild type (WT) mice. (A) A scatter plot reveals high dissimilarity between MTP and MC tissues. (B) A Volcano plot was used to estimate the statistical significance of fold changes in specific genes between MTP and MC tissues. (C) The expression levels of Soat1 in MC and M are significantly different. (D) A heat map of the expression levels of major genes of meibogenesis in MTP and mouse corneas MC.

As a first step, the Soat1 and Soat2 expression levels in TPs and cornea were determined (Fig. 11C). It was found that Soat2 was present in both tissues at identically low levels (around 4.35 ± 0.05 on a log2 scale). In contrast, Soat1 was highly expressed in TPs (17.6 ± 0.3), but much less in the cornea (7.4 ± 0.2), with a TP-to-cornea LFC of over 1000. Importantly, a related gene – Lcat/Slc12a4 – that encodes enzyme lecithin-cholesterol acyltransferase (LCAT), was found in mouse TPs and corneas only at a very low log2 level of 4.7 ± 0.1 for both tissues. 

In an attempt to connect lipid profiles of TPs and corneas with their transcriptomes, we conducted targeted evaluation of TPs and cornea mRNA profiles focusing on the rest of the main genes of meibogenesis.^19–21 A heatmap of approximately 50 main genes that are involved in meibogenesis and/or differ significantly between TPs and corneas are shown in Figure 11D (data from Supplementary Table S1). The vast differences in the expression levels of these genes between TPs and corneas corroborated the results of our lipidomic experiments as no genes related to the generation of typical VLC/ELC ML (such as WE, CEs, [O]-acylated omega-hydroxy FAs [OAHFAs], and CEs of OAHFAs [Chl-OAHFAs] that require Elovl3, Elovl4, Far1/Far2, Awat1/Awat2, and others to be synthesized) were found to be expressed at substantial levels, except for a few genes related to cholesterol biosynthesis, such as Dhcr24, Hmgcs1, Idi2, Fdps, and Fdft1. 

Tarsal Plate Transcriptomes of Wild Type and Soat1-Knockout Mice

Previously, Soat1 inactivation was shown to result in an almost complete arrest of VLC/ELC CE production in Soat1^−/− MGs, but the glands retained their ability to supply shorter chain serum-like CEs to meibum with only minor changes in their WE or triacylglycerol profiles.¹¹ These specific effects were accompanied by dramatic changes in the mouse ocular phenotype, which affected not only TPs, but also the mouse cornea and other ocular structures (see Figs. 2–7). Importantly, the Soat1-inactivation mutation did not have any effect on the expression levels of a related gene Soat2 in either tissue, which retained its log2 value (4.4 in WT and 4.3 in Soat1^−/− mice; P = 0.9). 

When the mouse TP transcriptomic data was analyzed using the Integrated Molecular Pathway Level Analysis (IMPaLA) version 13.0 web tool (found at http://impala.molgen.mpg.de), specifically its pathway over-representation and Wilcoxon pathway enrichment analyses tools, and the PANTHER Knowledge Database (http://pantherdb.org), a number of up- and downregulated pathways with P-genes and Q-genes values below 0.05 were identified (Table and Supplementary Table S2). 

Table.

 

Most Up-and Down-Regulated Pathways in Tarsal Plates of Soat1-Null Mice.

 

Among pathways downregulated with the highest statistical significance, there were pathways responsible for general metabolism, keratinization, FA metabolism, lipid metabolism, cholesterol metabolism, muscle contraction, striated muscle contraction, glucose metabolism, O-linked glycosylation of mucins, among other functions, whereas highly upregulated pathways included RNA metabolism, mitosis, and cell division in general. In addition, notably upregulated genes were responsible for formation, modification, and trimerization of collagen, SUMOylation, extracellular matrix reorganization, and, controversially, de novo FA and squalene/cholesterol biosynthesis, among other pathways (see Table and Supplementary Table S2). Note that the number of genes for the last two processes were not as large as those that were downregulated, and a heavy reliance on the genes of the NUP family (nucleoporins) in making those conclusions leaves room for alternative explanations. 

To determine whether inactivation of Soat1 had any effect on the lipid homeostasis in MGs, expression levels of major genes of meibogenesis in TPs of WT and Soat1^−/− mice were compared (Fig. 12). It appeared that almost all genes of interest underwent either up- or downregulation, but the magnitude of the effects varied widely. Of 360 genes tested, 86 genes were downregulated in the Soat1^−/− TPs with a WT-to-Soat1^−/− LFC of > (+2), whereas 21 genes were upregulated with LFC < (−2) (Supplementary Table S3). The rest of the changes in these genes were outside of the default LFC criterion of ≤ (−2) or ≥ (+2). Moreover, less than 50% of the genes underwent statistically significant changes with P ≤ 0.05, with most of them being expressed at relatively low levels. Interestingly, two key genes of WE biosynthesis – Awat1 and Awat2 – were significantly downregulated in the MGs of Soat1^−/− mice (see Supplementary Table S3). 

Figure 12.

 

Expression levels of 360 major genes of meibogenesis in tarsal plates of wild type (WT) and Soat1^−/− (KO) mice.

Cornea Transcriptomes of Wild-Type and Soat1-Knockout Mice

Human and mouse cornea transcriptome have been described and discussed earlier.^27–30 Inactivation of Soat1 in mice caused a noticeable perturbation in the cornea gene expression profiles (GEP; Supplementary Table S4). When the data for Soat1-null and WT corneas were compared using the IMPaLA Pathway Analysis Web tool, among the most upregulated pathways with P-gene and Q-gene values of <0.05 in Soat1-null mice were those related to: (1) activation of angiogenesis via VEGFA-VEGFR2 signaling pathway^31–33; (2) glycolysis, gluconeogenesis, Cori cycle, and carbohydrate metabolism in general (https://pubchem.ncbi.nlm.nih.gov/pathway/WikiPathways:WP1946); (3) keratinization, ossification, collagen formation, and extracellular matrix organization; and (4) various cell signaling pathways that are involved in hypoxia, stress, apoptosis, and immune responses, among others (Supplementary Table S5). The number of downregulated pathways with P-gene values of <0.05 was considerably smaller, with none of them producing Q-gene values of <0.05 (Supplementary Table S6). The only downregulated pathway that approached the level of statistical significance of a Q-gene value of 0.07 was the SLC-mediated transmembrane transport pathway, whereas, for the rest of them, the Q-gene values were >0.3. To better understand some of these changes, a more targeted analysis of the GEP in WT and Soat1-null corneas was conducted as follows. 

Neovascularization/Angiogenesis

An overview of corneal neovascularization in experimental transgenic mouse models can be found in a review by Kather and Kroll.³⁴ Consistent with neovascularization of the corneas of Soat1-null mice observed in our histological experiments (see Fig. 6), VEGFA-VEGFR2 signaling pathway was seemingly activated. Specifically upregulated from log2 (7.61) to log2 (9.36) was Vegfa (LFC = 3.36; P = 2.8 × 10⁻⁶), with no changes in the expression levels of Vegfb and Vegfc genes. However, no correlation of the changes in the GEP with an earlier publication of Wang et al.³⁵ was found, except for downregulation of Tgm2, Sorbs2 (Sorbs1 in the paper³⁵), and Gsta4 [ibid], and upregulation of S100a4/S100a7a (S100a9/S100a8 [ibid]). Yet another gene to consider is Il36a/IL36a, which is a promoter of angiogenesis in humans³⁶ and an ortholog of the mouse gene El1f8.³⁷ Importantly, the latter was shown to participate in the activation of the Nf-κb pathway (discussed below). Also noticeable was an increase in Mmp2 (log2 values of 11.39 in WT and 12.38 in mutants; LFC = 2; P = 0.0002), which plays multiple roles in various processes, including angiogenesis and inflammation (discussed below).^38,39 

Opacification and Keratinization of Cornea

Although our histological observations did not provide a direct evidence of cornea (hyper)keratinization in young mice, it became visible with aging mice. Transcriptomic data demonstrated clear and significant upregulation of the relevant pathway(s) (see Supplementary Table S5). Among genes that are involved in (hyper)keratinization, there were Sprr1a and Sprr1b genes which encode small proline-rich proteins SPRR1A and SPRR1B.⁴⁰ These genes were highly upregulated in the corneas of Soat1-null mice: the WT-to-mutant LFC of Sprr1a and Sprr1b genes were 198 (P = 4.5 × 10⁻⁶) and 170 (P = 1.9 × 10⁻¹⁰), correspondingly (see Supplementary Table S4). Other genes of the keratinization pathway, such as Krt17, Ivl, Krt19, Tgm1, Cds, Krt10, Krt14, and Spink5, were also notably upregulated (ibid). 

Hypoxia, Apoptosis, and Inflammation

The Plk3/c-Jun signaling pathway is known to be involved in hypoxic stress and apoptosis in the cornea. In particular, activation of polo-like kinase 3 Plk3/PLK3 induces c-JUN phosphorylation in the central part of the cornea but not in the limbal region.⁴¹ We observed upregulation of at least 3 genes of this pathway – Plk3, Jun, and Hif1a – in Soat1-null corneas. The log2 expression level of Plk3 was increased from 5.69 in WT corneas to 8.54 in the mutant ones (LFC = 7.2; P = 2.5 × 10⁻⁸). Other members of the Plk family (Plk1, Plk2, Plk4, and Plk5) were also increased, but to a much lower extent with log2 values of 1.07 to 1.64. The Jun levels were also highly upregulated from log2 8.52 in WT mice to 12.46 in the mutants (LFC = 15.3; P = 6.2 × 10⁻⁸). Finally, Hif1a was also increased from 13.3 to 13.9 (LFC = 1.5; P = 2 × 10⁻⁴). 

Another cluster of relevant genes are MMPs that encode matrix metallopeptidases, which are implicated in various processes, including apoptosis and inflammation.^42,43 In our experiments, only 4 MMP genes out of 19 detected were found to be noticeably increased in Soat1^−/− corneas with P values approaching, or exceeding, the statistically significant levels of P = 0.05: Mmp2 (log2 values of 11.39 in WT and 12.38 in mutants; LFC = 2; P = 0.0002), Mmp3 (5.26 in WT and 7.36 in mutants; LFC = 4.3; P = 0.06), Mmp13 (4.8 in WT and 5.76 in mutants; LFC = approximately 2; P = 0.05), and Mmp19 (5.76 in WT and 6.45 in mutants; LFC = 1.6; P = 8 × 10⁻⁵). All four genes have been clearly linked to inflammatory processes in various tissues. Interestingly, Mmp9 – a reportedly potent player in cornea neovascularization, inflammation and ulceration^44,45 – was not changed in the Soat1^−/− corneas (4.4 in WT mice and 4.5 in the mutant mice, respectively; LFC = −1.03). 

An important marker of inflammation to consider was Nlrp3,⁴⁴ a member of a large Nlr family of genes. Contrary to Shimizu et al.,⁴⁴ neither its levels, nor levels of other Nlrp genes, were changed in the Soat1 mutant mice remaining at very low log2 values of 4 to 5 regardless of the genotype. Similarly unaffected by the mutation were Il18 and functionally related genes Il18bp, Il18r, and Il18rap.⁴⁶ Interleukin 18 (IL-18) is a potent cytokine that, among other functions, regulates innate and acquired immune response,^47,48 and is considered a suitable diagnostic marker for various disorders. However, its expression levels remained steady at log2 values of 8.2 with no fold changes (P <0.3). Conversely, Il1f8⁴⁹ – an ortholog of a human IL36b gene (see Reference 37 and https://www.ncbi.nlm.nih.gov/gene/69677) – demonstrated a 4.5-fold increase (P = 0.1) in response to the Soat1 inactivating mutation, and so did the Nfkbia gene that encodes a transcription factor NF-κB⁵⁰ (also a part of inflammasome regulation), which was increased from 9.62 to 11.26 with an LFC of 3.1 (P = 4 × 10⁻⁷). 

Discussion

Effects of Soat1 Inactivation on Meibomian Glands and Meibum

Inactivation of Soat1 – one of the major genes of meibogenesis that is highly expressed in TPs of mice and humans (for both species, 17.5 ± 0.5 on a log2 scale) – was previously shown to dramatically reduce biosynthesis of VLC/ELC CEs in MGs of mice, while preserving some amounts of shorter chain CEs and leading to the accumulation of nonesterified Chl in their TPs.²¹ At the same time, the expression levels of Soat1 and Soat2 in the WT corneas, reported in this study, were rather low with log2 values of approximately 7.4 and approximately 4.4, correspondingly (see Fig. 11), and so were the expression levels of Lcat (log2 = approximately 4.6). The levels of the final lipid products of meibogenesis in mouse MGs positively correlated with the levels of expression of corresponding metabolically related genes. For example, the levels of Soat1 and CEs correlated in a quasi-linear fashion, and so did the levels of WE and Awat2 and Elovl3 genes. Importantly, it was not until the Soat1 expression level exceeded 10 on a log2 scale that VLC/ELC CEs started to accumulate in MGs in large quantities.²¹ Also noteworthy are the extremely low corneal expression levels of Elovl3 (log2 of 4.5) and Elovl4 (log2 of 5.9; see Fig. 11, Supplementary Table S1), which are essential for the biosynthesis of VLC/ELC FA – essential precursors for meibomian-type lipids, including VLC/ELC CEs,^10,16 and which are absent from the cornea lipidome. Therefore, it seems possible to predict an impact of a gene that encodes a major metabolically important enzyme on the basis of its expression level in a given tissue and conclude that VLC/ELC CEs that were randomly observed in WT mice were not the products of in situ biosynthesis in the mouse cornea. 

Characteristically, the VLC/ELC CEs profiles in WT and Soat1^−/− corneas were almost identical (see Fig. 10). This fact, in combination with lack of expression of essential genes of meibogenesis in corneas (see Fig. 11), signifies that VLC/ELC CEs in the WT cornea specimens originated, in fact, from the MGs, and were uncontrollably deposited on the corneas as remnants of the TF lipids. Furthermore, LC-MS analysis of shorter chain CEs in WT and Soat1^−/− corneas revealed only negligible differences in their profiles (see Fig. 9). Some other organs and tissues, for example, the liver, colon, cerebral cortex, and cerebrospinal fluid, plasma, and serum, and others, express and/or contain considerable amounts of LCAT (GeneCards, https://www.genecards.org/cgi-bin/carddisp.pl?gene=LCAT), which can then circulate with blood across the body.^51,52 Together with LCAT, SOAT2 (https://www.genecards.org/cgi-bin/carddisp.pl?gene=SOAT2) produces shorter-chain, plasma-like CEs,⁵³ and, therefore, both enzymes can be considered to be legitimate sources of regular CEs for the cornea. Thus, the shorter chain CEs, such as C_18:1−, C_18:2−, C_20:4−, and C_22:6–CEs, could have been biosynthesized either in situ (i.e. within MGs) by SOAT2 and/or LCAT, or produced in distant loci and transported to the corneas with the blood stream. 

There could be numerous pathophysiological consequences of the Soat1 inactivation. Two of the most obvious outcomes is solidification of abnormal meibum at physiological temperatures¹¹ and dramatic changes in the rheological properties of ML once they are augmented with nonesterified Chl.⁵⁴ A sharp increase in the melting temperature of Soat1^−/− meibum is mostly due to the accumulation of nonesterified Chl (which melts at approximately 145–150°C), at the expense of CEs with much lower melting temperatures.⁵⁵ Moreover, in our previous study,⁵⁴ we tested the effects of free Chl on rheological properties of human meibum using the Langmuir trough and determined that incremental increases in the Chl-to-total ML ratio led to progressive stiffening of the lipid films and increased their collapsibility once the Chl-to-meibum ratio exceeded 5% (w/w). In homozygous Soat1-null mice, nonesterified Chl becomes the most prominent lipid dominating the entire lipid pool. In addition, solidification of meibum has a detrimental impact on meibum expressibility from the orifices and cause dilation of central ducts and MG orifices. Taken together, these factors are likely to be responsible for lessening the protective properties of meibum on the ocular surface and may cause stagnation of abnormal lipid in the ductules of MGs of Soat1^−/− mice. 

Thus, it was reasonable to conclude that the major, if not all, effects of Soat1 inactivation on the cornea and MGs homeostasis could have arisen from the effects of meibomian-type VLC/ELC CEs ablation and accumulation of free, nonesterified Chl in meibum. 

The vast changes in the ocular phenotype of Soat1^−/− mice homozygous for the Soat1 mutation included severe anatomic abnormalities in their eyelids and corneas. One of the most visible features of the eyes of knockout mice was their abnormal shape – a slit-eye phenotype with the ellipticity of 0.92 ± 0.01, which clearly differentiated Soat1^−/− mice from their WT counterparts with virtually round eyes. This abnormal mouse phenotype has been found to be a characteristic outcome of inactivating mutations in other genes involved in meibogenesis, for example, Elovl3,^10,56 Elovl4,¹⁶ Awat2,^13,14 Sdr16c5/Sdr16c6,^12,57 Far1/Far2,^58,59 and Cyp4f39.¹⁸ Considering that these mutations result in abnormal meibum with deficiencies in clearly different groups of lipids, it was surprising to come to a conclusion that any tested major alteration in meibomian lipidome causes similar reshaping and minimization of the palpebral fissure of mutant mice regardless of their genotypes. It is not clear whether these changes are voluntary (or spontaneous) happening in an attempt to minimize the ocular surface irritation, or if they were caused by a slew of detrimental structural changes in the eyelids and adnexa. Our analysis of the TP transcriptome of Soat1-knockout mice revealed strong evidence of downregulation of the Muscle Contraction and Striated Muscle Contraction Pathways, compared to their WT littermates. Specifically, genes Tnni2, Myom1, Tnnt3, Ttn, Des, Tcap, Actn2, Tnnc2, Myl1, Actn3, Acta1, Mybpc1, Mybpc2, and Neb, among others, were affected. As striated Riolan's muscles that are found in MGs are believed to help and/or regulate meibum secretion,⁶⁰ downregulation of these pathways will likely impede the muscle contraction and relaxation cycle, and, hence, dysregulate meibum delivery onto the ocular surface. Simultaneously, the striated orbicularis oculi and levator palpebrae superioris muscles that control eye blinking and squeezing may also be affected by the mutation, which could play a significant role in the congenital and/or acquired dysmorphism of the palpebral fissures of the Soat1-knockout mice. 

Effects of Soat1 Inactivation on Mouse Cornea

For the reader's convenience, a diagram that summarizes the effects of Soat1 inactivation on ocular surface homeostasis in mice is shown in Figure 13. The 10 most highly upregulated genes in Soat1^−/− corneas were Sprr1a and Sprr1b followed by Ptgs2, Jun, Dusp1, Cdsn, Kprp, Spink5, Gm5941, and S100a7a (see Supplementary Table S4). Until now, there was only limited information on their functions in the cornea, which prompted us to briefly summarize some of their known activities in the cornea and other tissues. However, due to the high magnitude of their dysregulation in mutant corneas, we expect these genes and proteins to be evaluated in future studies in more detail. 

Figure 13.

 

The effects of Soat1 inactivation on the mouse ocular surface homeostasis. SC – short chain, MC – medium length chain, LC – long chain, ELC – extremely long chain, CEs – cholesteryl esters.

Sprr1a and Sprr1b. Small proline-rich proteins SPRR1A and SPRR1B, called cornifins, are encoded by genes Sprr1a and Sprr1b. These keratinocyte proteins participate in the formation of cell envelopes in cornifying stratified epithelia,⁶¹ one of whose functions is to increase mechanical resistance of tissues to deformation stress. In addition, an increase in SPRR1A was reported in mouse corneas that were subjected to desiccating stress.⁶² Indicatively, the Jun gene was also upregulated in Soat1-null corneas, similarly to its increase in desiccating stress conditions studied by de Paiva et al. [ibid]. Coincidentally, Sprr1b was found to be a valid biomarker for squamous metaplasia in human subjects with Sjogren's syndrome and mouse models of dry eye,⁶³ whereas Sprr1a is considered to be a regeneration-associated protein. 

Interestingly, Newsome et al.⁶⁴ described an open eyelids with cleft palate (oel) mouse with hyperkeratinized ocular surface that exhibited poor wetting characteristics, thickened and vacuolated BM, and progressive opacification of the cornea. To the best of our knowledge, this mutation is currently not linked to any specific chromosome or gene, but the phenotype of oel mice partially resembles the ocular features of Soat1^−/− mice. 

Cdsn. Corneodesmosin (CDSN) – a secreted glycoprotein that plays an important role in cohesion of cells, such as corneocytes,⁶⁵ – is encoded by the gene Cdsn. Its known functions are related almost exclusively to human epidermis and other cornified squamous epithelia (https://ncbi.nlm.nih.gov/gene/1041). In our experiments, its expression in mouse corneas went up considerably in response to Soat1 inactivation from relatively low log2 values of 6.9 to 10.1 (P <0.08; see Supplementary Table S4). 

Ptgs2 and Vegfa. Another frequent abnormality in Soat1^−/− mice was neovascularization of their corneas (see Figs. 2B, 6G). Our transcriptomic data implies that PTGS2 and VEGFA-VEGFR2 signaling pathways are significantly upregulated in Soat1^−/− corneas (see Supplementary Tables S4, S5). The Ptgs2 gene encodes inducible cyclooxygenase-2 (COX2)⁶⁶ which is a key enzyme in the biosynthesis of prostaglandins (PGs). One of its products, specifically PGE2, is an important signaling molecule for activation of the VEGF signaling pathway.^67,68 At the same time, Ptgs2/COX2, unlike its constitutive counterpart COX1, participates in wound healing of the cornea and is upregulated in stromal keratocytes.⁶⁶ 

S100a4. The S100 calcium-binding protein A4 (S100A4) is involved in inflammatory processes and thought to have pro-angiogenic activity in tumor development and inflamed cornea.⁶⁹ These functions may be related to the development of the Soat1^−/− phenotype in mice and correlate well with our observation of cornea neovascularization in the mutants. 

Jun. Our results demonstrated that the expression levels of Jun in mouse cornea was highly increased in Soat1-null mice with LFC of >15 (see Supplementary Table S4). The c-JUN N-Terminal kinases (JNKs) are the most important mitogen-activated protein kinases (MAPKs) which have crucial roles in several cellular processes, such as apoptosis, autophagy, and inflammation.^70,71 The c-JUN, a member of activator protein-1 (AP-1), is the main substrate of JNKs. The c-JUN has been thought to participate in two fundamental mechanisms: apoptosis and autophagy acting via the JNK signaling pathway. Activated JNKs in mitochondrial matrix migrate to nucleus and phosphorylate c-JUN and other transcription factors. The c-JUN/AP-1 dependent activation results in and increase in the transcription of pro-apoptotic genes which trigger apoptosis. Phosphorylation of c-Jun also mediates the transcription of autophagic genes.⁷¹ The other important signaling pathway, Polo-like kinase 3 (Plk3), causes the corneal epithelial cell death in response to hypoxic stress and hyperosmotic stress conditions (see Reference 72 and https://doi.org/10.1074/jbc.M116.725747). It has been shown that the increased Plk3 kinase activity has a direct effect on the phosphorylation of c-JUN corneal epithelial cells [ibid]. 

Dusp1. Dual specificity phosphatase 1 (DUSP1) is a mediator in the resolution of inflammation⁷³ and has a protective anti-inflammatory role in osteoarthritis. It also participates in cellular responses to environmental stress (https://ncbi.nlm.nih.gov/1843). In cultured keratoconic corneal stromal cells, its expression level is five times of that in the norm.⁷⁴ 

S100a7a. Another S100 calcium binding protein A7A (S100A7A) was demonstrated to have antimicrobial activity against Staphylococcus aureus and Haemophilus influenze, but not E. coli, and suggested to be a component of the ocular surface innate immune system.⁷⁵ 

Kprp. Keratinocyte proline rich protein KPRP participates in keratinocyte differentiation and was reported to be expressed in human skin (https://ncbi.nlm.nih.gov/448834). In mice, KPRP plays a role in the epidermal barrier function⁷⁶ as Kprp-knockout mice demonstrated an increase in transepidermal loss of water. 

Spink5. The functions of serine peptidase inhibitor Kazal type 5 (SPINK5) in the cornea are also largely unknown. In a recent publication,⁷⁷ Kim et al. suggested a role for SPINK5, in conjunction with its effector, STAT3, in maintaining skin barrier homeostasis. Loss of SPINK5 function due to a mutation results in a genetic disease called Netherton syndrome,^78,79 whose characteristic signs are an impaired skin barrier (which was linked to lipid abnormalities), allergic reactions, inflammation, and stratum corneum detachment.⁸⁰ 

Il1f8. The of IL1F8/IL36B is a key part of a IL-36 signaling pathway (see References 37, 49, and 81 and https://www.ncbi.nlm.nih.gov/gene/27177) which activates NF-κB³⁷ and is related to neutrophilic inflammation and angiogenesis.³⁶ 

Finally, the predicted gene Gm5941 has no known functions at this time. 

Conclusions

Summarizing our observations and data published by other laboratories, one can suggest a direct involvement of Soat1/SOAT1 of meibomian glands, and extremely long chain cholesteryl esters they produce, in maintaining ocular surface homeostasis, in general, and the health of the cornea, specifically. It is also reasonable to assume that a sharp accumulation of free, nonesterified cholesterol in Soat1^−/− meibum is responsible not only for its abnormally poor expressibility and spreadability due to the high melting temperature of the secretion, but can also trigger, promote and/or mediate inflammation in the eye and adnexa similarly to the mechanisms related to the onset of atherosclerosis and obesity.⁸² It is also likely that upregulation of marker genes of apoptosis, inflammation, neovascularization, keratinization, and muscle contraction in the meibomian glands and/or cornea are downstream effects of Soat1 inactivation, similar to those in Elovl3- null mice,^10,56 despite being caused by depletion of entirely different types of meibomian lipids. 

Acknowledgments

The authors would like to express their gratitude to Matthew Petroll and the Core grant P30 EY030413 for providing access to the Zeiss fluorescent microscope that was used for taking images of stained tissue sections. Figure 13 was created using BioRender. 

Funded by the U.S. National Institutes of Health R01 grant EY027349 (to I.A.B.). 

Disclosure: A. Wilkerson, None; S. Yuksel, None; R. Acharya, None; I.A. Butovich, None