Lipid Composition and Thermotropic Properties of Meibum of Animal Models and Humans with Meibomian Gland Dysfunction

Abstract

Meibum—a lipid-rich secretion produced by holocrine Meibomian glands (MG)—plays a central role in maintaining ocular surface homeostasis in humans. Previously, changes in MG lipidomes induced by inactivation of critical genes of meibogenesis, such as Elovl3, Soat1, Awat2, Sdr16c5/Sdr16c6, and others were shown to cause MG dysfunction (MGD) and dry eye in experimental animals. Here, we describe the impact of the changes in the lipid composition of meibum on its protective properties, specifically physiologically relevant thermotropic characteristics, using various mutant and wild-type animal models, and comparing them with healthy human subjects and patients with MGD. Meibum samples were analyzed using liquid chromatography/mass spectrometry (LC/MS) and differential scanning microcalorimetry (DSC). We found that any change in the balance between major lipid classes in meibum—wax esters, cholesteryl esters, triacylglycerols, and free cholesterol—cause detrimental changes in its thermotropic properties, loss of cohesiveness, and abnormal expressibility from MG, resulting in MGD-like phenotypes of the eyes and adnexa. We conclude that tested knockout mice can be valuable models for modeling and studying MGD. A combination of LC/MS and DSC can be a powerful diagnostic tool and may help to diagnose MGD and other pathologies, as well as determine their molecular mechanisms.

1. Introduction

Sebum and meibum are two types of lipid-rich secretions that are produced by related, but not identical types of exocrine/holocrine sebaceous and Meibomian glands (SG and MG, correspondingly). Sebum creates a protective hydrophobic layer that covers the skin, human hair, and animal fur [1], while meibum [2] is one of the main components of the preocular tear film [3]. The main functions of sebum and meibum are similar: sebum is a lubricant that protects the skin from desiccation and exogenous pathogens, while meibum performs similar functions in the eye. To work, sebum and meibum need to be expressed from their corresponding glands as a first step. This mechanism involves the (mostly) passive transport of secretions through a system of ducts before being released onto the skin or the ocular surface. The ability of the secretions to passively travel through the ducts heavily depends on their composition: secretions that are too thick may severely reduce the flow and cause stagnation of sebum and meibum in the ducts and glands, while secretions that are abnormally fluid will cause lipid overflow—conditions that are detrimental to both the ocular surface and the skin homeostasis.

Indeed, deviations in the chemical composition and/or quantity of biosynthesized meibum from the norm due to pathologies known as hypo- and hypersecretory MG dysfunctions (MGD) have been reported to trigger the onset, or exacerbate the consequences, of evaporative dry eye (EDE) [4,5,6]. Similar effects are expected to play a role in the release of abnormal sebum from free and hair-associated SG.

Previously, targeted changes in MG lipidomes induced by inactivating mutations in critical genes of meibogenesis [7], such as Elovl3, Soat1, Awat2, Cyp4f39, Sdr16c5/Sdr16c6, Scd1, Far1/Far2, and others, were shown to almost universally cause MGD- and EDE-like signs in mice [8,9,10,11,12,13,14,15,16]. The aim of this paper is to describe the impact of the differences in the lipid composition of mouse, canine, and rabbit MG secretions on their biophysical/physiological properties, specifically thermotropic/melting characteristics, and compare them with meibum of healthy human volunteers and patients with abnormal meibum.

2. Results

2.1. Untargeted, Unsupervised LC/MS and Multivariate Statistical Analyses of Human and Animal Meibum

The chemical composition of meibum varied significantly between the species and genotypes. Results of the initial evaluation of the human and animal Meibomian lipidomes were reported earlier [17]. However, for the purpose of this study, we re-analyzed human and animal meibum using liquid chromatography/high resolution time-of-flight mass spectrometry (LC/MS) and compared the samples using unbiased, untargeted, unsupervised Principal Component Analysis (PCA) and supervised Partial Least Squares Discriminant Analysis (PLS-DA) approaches included in Progenesis QI version 2.3 and EZinfo version 3.0.3 software packages (all from Waters Corp., Milford, MA, USA). The data are shown in Figure 1A as a Loadings Bi-Plot. Note a rather tight intra-group clustering of the samples of the same origin (shown as circles of different colors), and clear inter-group separation of the samples. The black dots represent the lipid analytes, whose proximity to specific groups of samples depends on their associations with specific species or genotypes. The chemical composition of meibum varied significantly between the species and genotypes, but the close proximity of human and canine groups on the plot is indicative of their close biochemical similarities. Wild-type (WT) mice also demonstrated rather tight grouping of the samples of the same type and were the second closest group to human samples. Meibum samples of Sdr16c5^−/−/Sdr16c6^−/− double knockout (DKO) mice were the most distant from both the WT mouse and human meibum. The rabbits are not included in the PCA/PLS-DA analyses because of their fundamental differences from other species: rabbit meibum is based primarily on dihydrolanosteryl esters (DiHLE) as major nonpolar lipid components, and much smaller amounts of cholesteryl esters (CE) and wax esters (WE), which are major Meibomian lipids in other studied animals [17].

Importantly, the Loadings Bi-Plot also revealed the type of lipids that were associated with specific groups of samples (discussed below), while the Hotelling’s T² Range Plot demonstrated that there were no significant outliers in the datasets: all specimens were well below the T² Crit(99%) threshold, with only two samples barely exceeding the T² Crit(95%) threshold (Figure 1B).

2.2. Mouse Meibum

2.2.1. Characterization of the Chemical Composition of Wild-Type and Mutant Mouse Meibum

Then, more targeted comparative liquid chromatography/mass spectrometry (LC/MS) analyses of WT mouse and normal human meibum samples were performed to ensure consistency of the results with previously published data [7,8,10,12,17,18,19].

First, the effects of the mutations on mouse Meibomian lipidomes were studied (Figure 2). The WT meibum of C57BL/6J mice was used as a baseline for comparing the effects of mutations on the composition and properties of meibum of genetically modified mice.

Inactivating mutations in Elovl3, Soat1, Awat2, and Sdr16c5/Sdr16c6 genes in the mice on the same background led to fundamental shifts in their Meibomian lipidomes, affecting the ratios between individual lipids within their respective groups, and the balance between different lipid groups and classes. As meibum is predominantly based on nonpolar lipids, such as WE, CE, free cholesterol (Chl), and triacylglycerols (TAG), these lipids were profiled in the study samples as shown in Figure 2A–E. The levels of Chl were severely upregulated by inactivation of Soat1, while inactivation of other genes had only a minimal impact on the compound. On the other hand, CE were almost nonexistent in Soat1^−/− mice, considerably upregulated in Awat2^−/− mice, but very close in WT and DKO mice. The levels of TAG, which are the main sources of carbon and energy for lipogenesis in cells, were almost unaffected in Awat2^−/− and Soat1^−/− mice but somewhat suppressed in other mutants.

Inactivation of Awat2 in mice has been previously reported to lead to a decline [16] or an almost complete suppression [10,20] of the biosynthesis of Meibomian WE in mouse tarsal plates. This observation was reconfirmed for expressed Awat2^−/− meibum in our current experiments (Figure 2B)—a multifold decrease in the levels of mono- and di-unsaturated Meibomian WE (MUWE and DUWE, correspondingly) was observed, which, together with a concomitant increase in Chl and CE, caused a fundamental change in the MG lipidome. Therefore, the remaining major components of the Awat2^−/− meibum were Chl, CE, and related compounds.

Aside from an almost complete suppression of the formation of CE and upregulation of Chl, inactivation of Soat1 lowered the expression levels of the tested WE, making Chl the major lipid in meibum (Figure 2A,C).

While the overall ratios of Chl, CE, and TAG in meibum of DKO mice underwent only modest alterations, the pool of their nonpolar lipids became highly enriched with sebaceous-type lipids, such as diunsaturated and shorter-chain WE (Figure 2D). Note that the total amount of biosynthesized meibum in the DKO tarsal plates was ~3× that of WT mice. Importantly, the total fraction of Meibomian-type WE did not change much, while the increase in WE was caused by accumulation of sebaceous-type WE [19,21].

Inactivation of Elovl3 changed the balance between MUWE and DUWE toward the latter and led to a ~7× increase in the abundance of shorter-chain CE with fatty acid (FA) chains ≤ C₂₀, a ~30% increase in their longer chain counterparts (FA > C₂₀), and a ~40% decline in TAG (Figure 2E).

2.2.2. Effects of the Mutations on Thermotropic Properties of Wild-Type and Mutant Mouse Meibum

For the reader’s convenience, the results of differential scanning microcalorimetry (DSC) experiments with mouse meibum are summarized in

Table 1. Thermotropic properties of human and animal meibum (see list of abbreviations at the end of the paper).

Animals and Human Subjects	Ocular and Skin Features and Abnormalities	Major Features of MG Lipidomes	Tm, °C (Mean ± SD) *	T1/2, °C (Mean ± SD) *
C57BL6/J mice	None	Similar to human	28 ± 1	19 ± 3
Awat2-knockout mice	Severe ocular and Meibomian gland phenotype; solid meibum	WE ↓, free Chl ↑, and CE ↑	Multiple melting peaks between 20 °C and 70 °C	Extremely wide, >40 °C
Soat1-knockout mice	Severe ocular and Meibomian gland phenotype; solid meibum	Lacks CE, free Chl ↑↑↑	Biphasic melting curve; 30 ± 2 (1st) 50 ± 1 (2nd)	Each peak wider than >20
Elovl3-knockout mice	Severe ocular and Meibomian gland phenotype; liquefied meibum	Abnormal WE and CE elongation and unsaturation patterns; ELC WE ↓	Very low, <0	n/d
Sdr16c5/Sdr16c6 double knockout mice	Severe ocular and Meibomian gland phenotype; semi-liquefied meibum	Transformed into a sebum-like lipidome; TAG ↑ and short WE ↑	19 ± 1	30 ± 5
Healthy canines	None	Similar to human	29 ± 1	13 ± 2
White New Zealand rabbits	Extremely stable tear film	Based primarily on DiHL, DiHLE, DiAD, steryl esters (not CE), and OAHFA	33 ± 1	8 ± 1
Normal human subjects	None	Based primarily on ELC saturated and mono-unsaturated WE, CE, Chl, TAG, OAHFA, Chl-OAHFA, and DiAD	29 ± 1	10 ± 1
Abnormal human Subject #1	Lard-like meibum; abnormal tear film with lipid rafts	TAG ↑, short WE ↑, and ELC WE ↓	Biphasic melting curve; 21 ± 2 (1st) 57 ± 2 (2nd)	~17 (1st) ~15 (2nd)
Abnormal human Subject #2	Dry eye, skin, and solid meibum	Saturated WE ↑↑↑	39 ± 1	12 ± 1

* The values were rounded up or down to the nearest integer; averaged numbers for multiple samples are shown. The following signs were used: ↑—upregulated metabolites; ↑↑↑—highly upregulated metabolites; ↓—downregulated metabolites.

and Figure 2F. Close similarities between human and WT mouse meibum were expected to result in close thermotropic properties of both secretions, with, possibly, slightly lower transition temperatures (T_m) of the mouse meibum. Indeed, mouse meibum produced a well-defined, rather symmetrical thermogram, and had a major apparent T_m between 26 °C and 28 °C, which was ~2 °C lower than the human one, and a somewhat broader major melting range with a T_1/2 (the width of the transition at half-height of the melting peak) of 19 ± 3 °C. The broadening of the T_1/2 was caused mainly by the ascending part of the melting curve which showed a first sign of a phase transition between −5 °C and 0 °C. Interestingly, its upward section had a few faint shoulders which were either much less pronounced or non-existent in human samples, and might indicate either heterogeneity of the lipid mixture, or minute phase transitions in its structure. Another distinctive feature of mouse meibum was a shoulder, or in some cases, a minor transition peak, with a T_m of around 38 °C to 40 °C. However, the downward section of the curve had almost the same shape and slope as the human one. The most distinctive feature of the WT mouse meibum was a secondary, though minor, transition peak, or a shoulder in some samples, with a T_m of around 40 °C. WT mouse meibum became completely liquefied at temperatures above 45 °C. The significance of these features for the ocular surface physiology will be discussed later in the paper.

Changes in the chemical composition of meibum caused by the mutations radically affected its melting properties (Figure 2F) and, hence, expressibility from MG orifices.

The Awat2 inactivating mutation had an unexpectedly complex effect on the thermotropic properties of the mutant meibum, with multiple transitions/calorimetric domains being observed. Aside from a low-temperature transition peak with T_m of ~3 °C, multiple transitions between T_m of ~35 °C and ~58 °C were observed with an overall T_1/2 of more than 35 °C—the widest among all animal models tested in this study.

The thermotropic properties of Soat1^−/− meibum were also fundamentally different from those of WT mice. The melting curves were of a bimodal nature, though three apparent transition temperatures were observed: 14 ± 1 °C (minor), 32 ± 1 °C (major), and 51 ± 1 °C (major). The estimated value of T_1/2 was also wide at ~30 °C.

Commensurate with the shorter-chain and less saturated nature of DKO Meibomian lipids, their melting started at below −5 °C, with a major apparent T_m of ~18.2 °C being about 10 °C lower than the T_m of WT meibum (Figure 2F). Interestingly, a shoulder with a T_m of around 3 °C was visible and close to the similar low-temperature transition peaks in WT and Soat1^−/− mice. The T_1/2 of DKO meibum (28.5 ± 1.5 °C) was considerably wider than the T_1/2 of WT mice (19 ± 3 °C).

2.3. Canine Meibum

Canine MG secretions are studied to a much lesser extent than the human and mouse ones, though, being house pets, their ocular conditions are closely monitored and treated by veterinarians [22,23,24,25,26]. Results of initial characterization of the chemical composition of canine meibum by means of ion trap mass spectrometry (MSⁿ) had been described earlier [17,27]. The secretions were re-evaluated in the current study using a more accurate high-resolution LC/MS (Figure 3). Total ion chromatograms of both secretions are shown in Figure 3A, while their corresponding mass spectra are shown in Figure 3B. The major canine WE were present with a very distinctive pattern—m/z 605 ↑, m/z 619 ↓, m/z 633 ↑ (major peak), m/z 647 ↓, m/z 661 ↑—while human and mouse WE showed an opposite trend—m/z 605 ↓, m/z 619 ↑, m/z 633 ↓, m/z 647 ↑ (major peak), m/z 661 ↓. The individual analytes were compared using extracted ion chromatograms (EIC) as shown in Figure 3C, and their identities were established on the basis of their m/z values, chromatographic retention times, and MS/MS fragmentation patterns. Canine and human WE have similar saturation/desaturation ratios, being based primarily on C_16:0, C_18:0, C_16:1, and C_18:1 fatty acids and saturated fatty alcohols. Also, canine and human CE and free Chl were present in comparable (but not identical) amounts and ratios [17,27].

Surprisingly, despite the reversed WE expression patterns and other differences, the major apparent transition temperature of canine meibum—about 29 ± 1.5 °C—was extremely close to that of a representative human sample, while the T_1/2 value was ~12 ± 1 °C—again, almost the same as T_1/2 of normal human meibum (Figure 3D). A minor shoulder at around 22–23 °C, detected in some canine samples, was a feature that differentiated them from human meibum, though its low intensity and rather low T_m (which is well below physiological levels) seems to be less relevant to the ocular surface physiology.

2.4. Rabbit Meibum

Rabbits are laboratory animals that are frequently used in biomedical experiments and pharmaceutical trials due to the anatomical similarities of their ocular structures with the human eye [26,28,29,30]. However, the lipidomic profile of rabbit meibum is radically different from that of humans, canines, and mice [17]. Rabbit meibum is composed primarily of a rather complex mixture of dihydrolanosteryl (DiHL) esters (DiHLE), steryl esters (SE) (Figure 4A–C), and some other compounds of polar and nonpolar nature. Because of its complexity, a comprehensive characterization of rabbit meibum has not been performed yet and is left for future studies. Interestingly, most of the major lipids of rabbit meibum are not found in humans; however, despite being fundamentally different from human meibum, the T_m of the rabbit secretion is only 3 to 4 °C higher than the human one (Figure 4D). The T_1/2 value of the rabbit meibum was between 7 and 8 °C, a few degrees less than the T_1/2 values of human and canine secretions, implying higher cooperativity of melting and sharper responses to the changes in temperature.

2.5. Normal and Abnormal Human Meibum

Then, meibum from human subjects was studied using untargeted and targeted LC/MS experiments. A truly quantitative analysis of all, or even major only, Meibomian lipids is yet to be completed due to the lack of necessary lipid standards for a wide variety of Meibomian lipids. Therefore, for the purpose of this study, the apparent LC/MS abundances of major detected analytes were used, compared, and discussed instead.

2.5.1. Normal Human Meibum

Normal human meibum has been qualitatively and semi-quantitatively characterized on the levels of lipid classes and individual lipid species by LC/MS and gas chromatography/MS (GC/MS) in previous studies performed by our laboratory and by independent teams. It has been firmly established that the major components of meibum are numerous WE and CE, free Chl, free fatty acids (free FA), TAG, (O)-acylated ω-hydroxy fatty acids (OAHFA), cholesteryl esters of OAHFA (Chl-OAHFA), diacylated α,ω-diols (DiAD), small amounts of phospholipids and sphingomyelins, and other minor compounds of lipid nature [31,32,33,34,35,36]. Yet, its melting properties remain poorly understood.

For our current experiments, human meibum samples were collected from normal, non-MGD, non-EDE volunteers and individually analyzed using LC/MS. Untargeted analyses of normal human meibum samples revealed their rather tight grouping with no clear differentiating factors—note large SD values even for the most influential components of the secretions, and lack of outliers, which correlated well with their highly similar DSC thermograms (Figure 5A–D).

Then, targeted analyses of specific lipids and lipid classes were conducted using EIC of specific lipids, as shown in Figure 6 for Chl, CE, some of the major Meibomian WE, and a major TAG triolein. Two different LC methods were used—a gradient LC on a C18 BEH reverse phase column and isocratic LC on a reverse phase C8 BEH column. The first approach was used for detailed analyses of Meibomian lipidomes, while the second—for lipid quantitation (Figure 6A,B) [19]. All LC/MS experiments were conducted in the atmospheric pressure chemical ionization positive ion mode (APCI PIM). A representative observation mass spectrum of a normal human meibum sample is shown in Figure 6C, which provided the data for analyzing the major components of meibum—WE, CE, Chl, DiAD, and other lipids as depicted in Figure 6D.

In total, eleven normal meibum samples were analyzed by DSC (

Table 2. Interdonor variability of the thermotropic properties of normal human meibum.

Normal Human Subject	Gender (Age, in Years)	Meibum Tm, °C, *	Meibum T1/2, °C *
N-1	M (41)	29.2	10.9
N-2	M (56)	28.6	9.9
N-3	F (48)	30.9	10.2
N-4	M (25)	29.1	9.3
N-5	F (41)	27.8	11.2
N-6	not recorded	29.6	10.7
N-7	F (42)	28.5	9.3
N-8	F (27)	28.3	9.9
N-9	F (28)	27.7	11.0
N-10	F (40)	28.6	9.62
N-11	F (60)	28.1	10.6
Average temperature (Mean ± SD)		28.8 ± 0.9	10.2 ± 0.7

* The values were rounded up or down to the nearest decimal point. M—males; F—females.

). Due to the sensitivity limits of the instrumentation, only samples that exceeded 0.2 mg were selected for the experiments. Normal meibum produced thermograms that had a main average T_m at around 28.8 ± 0.9 °C and a T_1/2 of 10.2 ± 0.7 °C (Figure 5D). Their melting started at around 12 ± 2 °C. Compared with pure lipids [37], meibum thermograms showed rather broad melting peaks with a slightly shallower ascending and steeper descending portions. As evidenced by the thermograms, meibum became solidified at <10 °C, and completely liquefied at a physiological body temperature of 37 ± 2 °C, being in a partially melted state at a physiological cornea temperature of 32 ± 2 °C [38,39,40,41,42]. We consider this melting behavior in the ~30 °C to ~37 °C temperature range to be absolutely critical for proper expression of meibum from MG orifices and its protective properties (see the Discussion section for details).

2.5.2. Abnormal Human Meibum

Abnormal meibum samples collected from human Subjects #1 and #2 (see Table 1) were studied using the same LC/MS and DSC approaches as described above. Initially, both types of abnormal samples were compared with normal human meibum using unbiased, untargeted PLS-DA approaches. The three types of samples formed tight, but clearly separated groups, of which none of the samples was an outlier—all samples fell below the T² Crit(95%) threshold (Figure 7A,B).

Abnormal Subject #1. The Meibomian lipid profiles and clinically relevant ocular surface features of Subject #1 have already been partially characterized in our earlier publication [43]. Subject #1’s meibum samples were collected on three different occasions, a year apart from each other, and analyzed in this study alongside other human and animal specimens. The experiments produced the same result as before (Figure 8A,B): a severe imbalance between WE, CE, Chl, and TAG. Specifically, the lipid profile of Subject #1’s meibum samples resembled that of human sebum, being highly enriched with shorter-chain and less saturated WE compared to normal meibum (Figure 8C and ref. [5,44]), among other changes.

DSC experiments with Subject 1’s meibum samples produced a clearly abnormal biphasic/bimodal pattern with two major apparent melting peaks with T_m of 21 ± 1 °C and 57 ± 2 °C, and several shoulders (Figure 8D). Notably, melting of the samples started at below −5 °C, which was at least 15 °C lower than the starting temperature of melting of normal human meibum. Also, the first major peak of the thermogram had two shoulders at T = 8 °C and 12 °C.

Abnormal Subject #2. Ocular features of the abnormal Subject #2 with a c.1579C>T (p.Arg527Cys) heterozygous mutation in the SREBF1 gene and the preliminary findings on the subject’s abnormal meibum were described in our recent paper [44]. Here, we present the results of a more detailed analysis of the biochemical and biophysical properties of the subject’s MG secretion. Superficially, the meibum of Subject #2 looked similar to normal meibum shown in Figure 9A,B. However, the sample differed from the normal human meibum in the relative balance of saturated and unsaturated WE, which was shifted toward the former, and in the WE/(Chl + CE) ratio (Figure 9B), which was lower in the abnormal meibum (Figure 9C). The mass spectra of the three major groups of saturated WE (SWE, marked with red asterisk), MUWE, and DUWE are shown in Figure 9D,E. Once the LC/MS data for individual WE were integrated, a repeating pattern emerged—the abnormal meibum had a considerably higher proportion of SWE, and a lower proportion of MUWE and DUWE regardless of the WE carbon chain length (Figure 9F).

Cumulatively, these changes resulted in a considerable shift in the melting curve of abnormal meibum to higher temperatures with the main T_m of 40 ± 1 °C, but virtually the same T_1/2 of 11.5 ± 1 °C (Figure 9G). Unlike abnormal meibum from Subject #1, no visible shoulders or peak separation was observed in Subject #2’s thermograms, except for a minor feature at T_m of ~3.5 °C—the secretion remained cohesive regardless of the temperature.

2.6. Deconvolution of Thermograms

Experimentally obtained thermograms were numerically deconvoluted as described earlier for human meibum samples and WE model mixtures [37]. Deconvolution of the thermotropic peaks can be a useful tool for detecting and visualizing distinctive lipid domains within study samples [45,46]. Depending on the complexity of the thermograms, the routine allowed for a reasonably accurate approximation of the meibum melting curves using multi-component models with 3 to 8 calorimetric/lipid domains (Figure 10), though calculation of true ΔH_v and ΔH values was impossible because of the current lack of data on average molecular weights of animal meibum, and precise physical weights of the samples. Rather complex melting behavior of meibum produced by Awat2^−/−, Soat1^−/−, and DKO mice and abnormal human Subject #1 with several clearly visible transition peaks, and very broad (in comparison with pure lipids) melting peaks of other types of samples, can indicate either multistep transitions in the samples upon heating:

Crystal (solid) → Liquid crystal (several forms; partially melted) → Liquid (fully disorganized)

or the existence of separate poorly mixing lipid domains (rafts, pools, microcrystals, etc.) enriched with different types of lipids that melt at distinctively different temperatures and independently from each other. Establishing the nature of these transitions and/or lipid aggregates goes beyond the scope of this paper and should be addressed in future studies. However, the changes in the apparent melting characteristics of meibum induced by mutations, differences between studied species, and normal and abnormal human meibum are discussed below.

2.6.1. WT Mouse Meibum

A rather wide, but cohesive, melting peak of WT mouse meibum with an average T_m of 28.2 °C and several shoulders (Figure 10A) implied either a fairly good miscibility of mouse Meibomian lipids that melted uniformly as a whole, or the existence of several calorimetric domains that had close T_m’s. To approximate the shape of WT thermograms, up to six T_m were needed with three major transitions, rounded to the first decimals, at 19.6 ± 0.5, 26.8 ± 0.2, and 32.3 ± 0.2 °C.

2.6.2. Awat2^−/− Mouse Meibum

Meibum of Awat2^−/− mice produced the most complex thermograms of all tested animals being highly incoherent (Figure 10B), which required at least seven different T_m’s 3.5 ± 0.1, 21.9 ± 0.1, 30.9 ± 0.1 (major), 35.9 ± 0.1, 44.0 ± 0.1 (major), 53.1 ± 0.1 (major), and 59.8 ± 0.1 °C. The temperature at which Awat2^−/− meibum completely melted exceeded 65 °C.

2.6.3. Soat1^−/− Meibum

Deconvolution of thermograms of Soat1^−/− meibum produced three transition peaks with T_m1, T_m2, and T_m3 of 13.8 ± 0.2 (minor), 32.74 ± 0.1 °C (major), and 50.2 ± 0.1 °C (major), respectively (Figure 10C). Interestingly, the first T_m1 was close to the shoulder visible in the thermograms of WT mice, while T_m2 was close to the main T_m of the latter and that of normal human meibum. The most obvious differentiating feature was the third main transition peak with an extremely high T_m3 of ~50–51 °C, which is likely to be caused by large amounts of free Chl in the abnormal meibum. The Soat1^−/− meibum became completely liquefied only at temperatures above 60 °C.

2.6.4. DKO Meibum

The experimental thermotropic peaks of DKO meibum were adequately deconvoluted to produce four main superimposed transitions with T_m1 to T_m4 of ~3.0, 18.6, 32.6, and 37.0 °C (Figure 10D). The samples were completely melted by ~41 ± 2 °C.

2.6.5. Canine Meibum

Deconvolution of the canine thermograms resulted in three major transition peaks: T_m1 = 3.4 ± 0.4 °C, T_m2 = 23.6 ± 0.1 °C, and T_m3 = 29.1 ± 0.1 °C. Thus, close lipidomic profiles of canine and human meibum correlated well with their close melting characteristics (this paper and ref. [37]).

2.6.6. Human Meibum

Deconvolution of the thermograms of normal human meibum was achieved with three different T_m [37], while abnormal meibum #1 required seven (Figure 10E). Numeric deconvolution of the thermogram produced five major components with T_m of 5.6 ± 0.2, 13.9 ± 0.2, 22.2 ± 0.1, 47.7 ± 0.5, and 57.1 ± 0.2 °C (all major peaks). The first experimental transition peak with an apparent T_m of 21.4 °C could be described well as a superimposition of two peaks with T_m values of 13.9 and 22.2 °C. The second largest peak with T_m of ~56.8 °C was also a combination of at least two peaks with T_m1 of 47.7 °C and T_m2 of 57.1 °C. The nature of these thermograms will be discussed later in the manuscript.

Numerical deconvolution of the thermograms of abnormal sample #2 resulted in three T_m of 3.2 ± 0.2 (minor), and two major peaks at 35.4 ± 0.1 and 40.3 ± 0.1 °C (Figure 10F). Generally, the only difference between this sample and normal human meibum was a considerable increase in the major T_m of the abnormal sample, though the overall shape of the peak and its T_1/2 remained quite similar. Establishing the mechanisms of these reversible rearrangements of lipid aggregates upon heating of normal and abnormal meibum goes beyond the scope of this manuscript and should be attempted in future experiments.

2.7. Transcriptomic Analysis

Comparative transcriptomic analysis of mouse MG and corneas revealed that most of the genes of interest were barely expressed in the latter. Specifically, the gene expression levels (all shown as the MG/cornea ratios on a log 2 scale) were as follows: Awat2 (16.8/4.6; p < 10⁻¹⁶; FDR < 10⁻¹²), Elovl3 (17.0/4.5; p < 10⁻¹⁵, FDR < 10⁻¹¹), Soat1 (17.3/7.5; p < 10⁻⁸; FDR <10⁻⁶), Sdr16c5 (13.1/4.6; p < 10⁻¹⁴; FDR < 10⁻¹⁰), and Sdr16c6 (16.8/4.4; p < 10⁻¹⁴; FDR < 10⁻¹⁰), while a panel of house-keeping and reference genes [47]—Chmp2A (11.6/12.0), Emc7 (12.2/10.9), Gpi1 (13.6/12.4), Psmb4 (8.6/8.1), Rab7 (13.8/14.0), Reep5 (13.0/13.5), Vcp (14.7/13.6), and Vps29 (8.2/8.1)—showed much smaller differences.

3. Discussion

Thermotropic properties of meibum determine whether it can fulfill its physiological roles, such as being a protector of the eye from the environment, among other functions. To protect the eye, meibum must first be expressed from the orifices of the exocrine MG (which is mostly a passive process, possibly facilitated by blinking), and, second, spread across the ocular surface to form the protective tear film lipid layer (TFLL; Figure 11). This two-step process critically depends on several factors, among which the eyelid temperature, the ocular surface temperature (OST), and the melting characteristics of meibum, which determine its expressibility from MG’s ducts and ductules, will be discussed below. Under normal conditions, the human eyelid temperature is between 33 °C and 37 °C [48,49,50], and can be somewhat lower or higher, depending on the subject’s health status and environmental conditions. The OST in various corneal regions was reported to be between 30 and 31 °C [51], 34.3 ± 0.7 °C [38], 35–35.5 °C [52], and 37.1 ± 0.1 °C [53], with a tendency to decline due to tear film evaporation [54,55,56]. Thus, prolonged periods of non-blinking [38,52,54] and high and low air temperatures [57], among other factors that cause cooling of the ocular surface and adnexa below physiological limits, may facilitate meibum solidification on the ocular surface and/or in the MG ducts, and hinder its expressibility from MG. Higher body and OS temperatures, in contrast, would make meibum more liquid than normal. Similarly, a drop in OST would hamper its spreading across the ocular surface and inhibit the formation of the normal TFLL. The in vivo effects of the eyelid temperature on the human meibum expressibility were tested by Nagymihalyi et al. [48], who reported a considerable decline in the delivery of meibum due to low eyelid temperatures, and an increase in meibum delivery/expressibility caused by elevated temperatures. Later, we examined the direct impact of temperature on the solidification and melting of normal human meibum using a Langmuir trough and found that Meibomian lipid films become highly elastic upon cooling below the physiological ranges [57]. Our current data on thermotropic properties of meibum fully supports those observations.

However, changes in the eyelid and OST are just two factors that regulate meibum delivery onto the ocular surface; quality of meibum (or, in the context of this paper, its melting characteristics) is another factor that is critically important for ocular surface health. In case of abnormal mouse meibum, the ocular surface phenotypes of Awat2, Soat1, Elovl3, and DKO mice are highly abnormal (Figure 11 and ref. [8,9,10,11,12,13,58]). The visible changes in eye ellipticity, signs of inflammation of the eye and adnexa, neovascularization of the cornea, accumulation of lipid deposits on and around the eye, abnormal tearing, and other pathologies were observed in the knockout mice, replicating those seen in human patients with MGD and/or EDE. Importantly, these genes, which are major genes of mouse and human MG, are not expressed in the cornea (Section 2.7 and Figure 11), making a direct impact of abnormal meibum on ocular surface homeostasis—due to its altered physical properties and biochemical composition—a highly plausible mechanism. As DSC experiments demonstrated (Figure 5D and Figure 11E, Table 2, and ref. [37]), normal human meibum is partially melted at the normal corneal temperature of around 30 °C to 34 °C, and almost completely— at the normal body temperature of 36 °C to 37 °C. In these conditions, meibum is easily expressible from the MG orifices and freely flows across the ocular surface. On the other hand, meibum with abnormally high T_m (such as that from human Subject #2) will not be easily expressible from the MG orifices and will stagnate in the MG acini and ducts, lessening the protection of the eye. Moreover, even partially melted meibum with a large pool of unmelted components (as shown in Figure 11E) will not spread evenly across the ocular surface and will form irregular patches/droplets of lipids that exist, e.g., in the tear film of abnormal patient #1 [43]. Thus, the bimodal nature of thermograms of abnormal Subject #1 may explain the subject’s normal tear film breakup times, due to the components that melt at ~21 °C, but unusual morphology of the TFLL with large lipid rafts that are clearly visible upon slit lamp examination and do not melt at normal physiological corneal temperatures.

Close similarities between human, canine, and WT mouse meibum were expected to result in close thermotropic properties of the three secretions, with, possibly, slightly lower T_m of the mouse meibum due to slightly shorter carbon chain lengths of the major mouse wax esters (such as C₄₄H₈₆O₂ in humans and C₄₂H₈₂O₂ in mice [7]). However, the apparent T_m of mouse meibum, 28 ± 1 °C, is virtually the same as the human one, but its T_1/2 is a somewhat broader 19 ± 2 °C (Figure 2 and Table 1). The broadening of the T_1/2 was caused mainly by the ascending part of the melting curve, which showed a first sign of a phase transition at around 0 °C. Interestingly, the upward section of the mouse thermograms showed a few shoulders that were either much less pronounced or non-existent in human samples, whereas the downward section of the curve had almost the same shape and slope as the human one. The most distinctive feature of mouse meibum was a secondary, though minor, transition peak, or a shoulder in some samples, with a T_m of around 40 °C. The body temperature of WT mice ranges from 36 to 38 °C [59,60,61], with the OST being between 36.5 °C to 36.9 °C [53]. This temperature is in the same range as the human corneal temperature [53], which may imply that mouse meibum, once deposited onto the ocular surface, may behave the same way as the human one. However, if the actual human corneal temperature is closer to 30 °C to 34 °C [38,51], then human meibum deposited onto the ocular surface can become less fluid and more liquid crystal-like.

The canine meibum was found to be close to the human meibum in terms of lipid composition, T_m, and T_1/2 (Figure 3). Furthermore, the OST in dogs, depending on the experiment and environment, ranged from ~33 °C to ~38 °C [62,63,64], being the same as human and WT mouse OST. Thus, canines could have been considered a suitable model for human meibum studies, if not for the obvious ethical considerations. However, close similarities between human and animal ocular surface physiology and biochemistry can be used for developing proper treatments for veterinarian use, as ophthalmic diseases similar to human ones are a major concern in veterinarian practice [65].

Inactivating mutations in Awat2, Soat1, and Sdr16c5/Sdr16c6 genes caused massive changes in the thermotropic properties of mouse meibum (Figure 11D). Aside from shifting major T_m to lower or higher temperatures than the norm, the mutations caused considerable broadening of the meibum melting peaks and/or their disintegration into several clearly visible calorimetric domains. Consistent with visual characteristics of Soat1^−/− and Awat2^−/− meibum as being solid or wax-like at physiological temperatures, DSC demonstrated that both secretions undergo complete melting only at temperatures above 60 °C, which means that at physiological temperatures of 30 °C to 37 °C a large portion of meibum exists in a solidified state and cannot effectively interact with the tear film. Furthermore, the bi- and multimodal melting of mutant meibum is indicative of abnormal molecular interactions between Meibomian lipids. The complex modality of melting can be caused either by phase separation within the mixture (i.e., due to the formation of separate domains of different types enriched with specific lipids that melt at different temperatures):

Crystal 1 → Liquid crystal 1 → Liquid 1                                            

Crystal 2 → Liquid crystal 2 → Liquid 2…     (Mechanism I)

Crystal n → Liquid crystal n → Liquid n                                            

or complex, multistep transitions with formation of several liquid-crystal mesophases:

Crystal → Liquid crystal 1 → … → Liquid crystal n → Liquid   (Mechanism II)

or a combination of both. In normal meibum, these transitions are seamless and produce smooth, cohesive melting curves. Earlier, we demonstrated that the existence of a proper number of lipids of proper types in a mixture is important for generating smooth and cohesive lipid melting curves [37]. It is plausible that the mutation-induced elimination of certain types of lipids from meibum (such as obliteration of all wax esters from Awat2^−/− mice, or cholesteryl esters from Soat1^−/− animals), or abnormal accumulation of certain lipids (such as Chl in the latter, and Chl and short chain and unsaturated wax esters in abnormal human meibum #1) disrupts the normal lipid–lipid interactions in the secretions, causing visible segregation of their calorimetric domains. Though obviously important, establishing the actual mechanisms of these changes necessitates experiments that go beyond the scope of this manuscript and is left for future studies.

Paradoxically, an unexpected result of rather close, but still different, thermotropic properties of human and rabbit Meibomian gland secretions, which are fundamentally dissimilar in terms of their lipid compositions, illustrated that common rules may be applicable to different species: despite the fact that rabbit meibum has a T_m that is ~4 degrees higher than the human one (Figure 4D), their OST is also several degrees higher—about 39.1 ± 0.2 °C [53]. Thus, the melting state and behavior of rabbit meibum on their ocular surface could be similar to those of human meibum, despite human and rabbit secretions being biochemically different.

Inactivation of Soat1, Awat2, Sdr16c5/Sdr16c6, and Elovl3 genes in mice led to fundamental and complex shifts in their thermograms, which implied selective effects of specific lipids on meibum melting. Soat1^−/− and Awat2^−/− meibum samples had more complex thermograms and higher T_m than WT meibum, while Elovl3^−/− and Sdr16c5/Sdr16c6 DKO meibum displayed an opposite trend. Due to a very low melting point of Elovl3^−/− meibum (T_m < 20 °C) and a broad melting curve [9], obtaining their DSC was unsuccessful because of the VP-DSC‘s lowest practical temperature of −5 °C, which was not low enough for recording full melting curves using this instrument.

Finally, MGD gradation scales from 1 to 4, which are widely used in the clinical ophthalmic practice, represent a rather crude approach suitable only for superficial characterization of meibum expressibility and quality. Therefore, meibum of abnormal human subjects studied in this project was not graded using this approach. DSC, on the other hand, provide quantitative melting characteristics, such as T_m and T_1/2, with high accuracy and precision. Another important experimental parameter is the shape of the thermogram, which is directly related to the integrity and quality of meibum. The latter can be further explored by LC/MS to provide detailed information on the chemical composition of meibum. Thus, our data demonstrated that the combination of LC/MS and DSC is a powerful diagnostic tool that may help to diagnose MGD and other pathologies, and determine their biochemical origins and molecular mechanisms.

4. Materials and Methods

4.1. Chemical Reagents

The lipid standards (such as saturated and unsaturated wax esters, cholesteryl esters, cholesterol, triacylglycerols, and others) were purchased from Nu-Chek Prep, Inc. (Elysian, MN, USA), Avanti Polar Lipids (Alabaster, AL, USA), and MilliporeSigma (St. Louis, MO, USA). Chromatography/mass spectrometry grade solvents were from Sigma-Aldrich, ThermoFisher (Waltham, MA, USA), or Honeywell-Burdick & Jackson (Muskegon, MI, USA).

4.2. Animal and Human Meibum Samples

The founder Sdr16c5^+/−/Sdr16c6^+/− knockout mice on a C57BL/6J background were provided by Drs. Natalia Kedishvili and Olga Belyaeva (University of Alabama at Birmingham, Birmingham, AL, USA) [13] and a colony was established at the UT Southwestern Medical Center (Dallas, TX, USA) [19]. The founder Soat1^+/− mice on a C57BL/6J background were purchased from the Jackson Laboratory (B6.129S4-Soat1^tm1Far/Pgn; stock #007147; Bar Harbor, ME, USA) [12]. The heterozygous Elovl3^burf/GrsrJ mice were purchased from the same manufacturer (stock no. 024182), and cross-bred with C57BL/6J mice [8]. The Awat2^+/− mice on a C57BL/6N background (stock number 043484-UCD) were purchased from MMRRC (https://mmrrc.ucdavis.edu) and transferred onto a C57BL/6J background through multiple generations [10]. All animals used in this study were maintained on a 12 h light/dark cycle with ad libitum access to a standard commercial laboratory diet (Teklad 2016; Envigo, Somerset, NJ, USA) and free access to water.

Human meibum was collected by expressing the secretion from eyelids of male and female non-EDE/non-MGD human volunteers (Table 2) as described before [43]. However, we reported earlier that there were no noticeable sex-related differences between the lipid composition of meibum of human and mouse males and females [66,67].

Meibum of euthanized mice, canines, and rabbits was expressed from the tarsal plates (surgically excised or otherwise) by applying light pressure and collecting the expressed meibum with a metal spatula or a syringe needle with a flat tip, for waxy and solid secretions, or pre-cleaned cellulose tissues for more liquid specimens. If the cellulose wipes were used, the lipids were extracted from the wipes with n-hexane/isopropanol (1:1, vol/vol) solvent mixture, the extracts were brought to dryness with a stream of nitrogen. Then, the material was dissolved in iso-propanol (IPA), sealed in glass vials and stored at −80 °C until the analytical experiments. Initially, meibum samples were analyzed using LC/MS, and then compared using the PCA approach. Next, meibum samples were analyzed using DSC, and their T_m, melting ranges (ΔT), and other parameters were determined as described earlier [37]. In case of complex DSC thermograms, they were numerically deconvoluted to reveal individual phase transitions/calorimetric domains in the samples. The details of LC/MS and DSC procedures were provided in previous publications, e.g., ref. [19,37], and will be briefly explained in the following sections.

4.3. Liquid Chromatography–Mass Spectrometry

Meibum samples were analyzed using reverse phase LC/MS using an M-Class UPLC/Synapt G2-Si system equipped with C8 or C18 BEH Acquity chromatographic columns (all from Waters Corp., Milford, MA, USA) as described earlier [19] and shown in Figure 6. The elution of the analytes was conducted using a ternary gradient of IPA, acetonitrile, and 10 mM aqueous ammonium formate with a C18 BEH column, or an isocratic elution with a C8 BEH column. The lipids were identified on the basis of their m/z values, fragmentation patterns in MS/MS or MS^E mode, LC retention times, and comparison with authentic chemical standards, such as WE, Chl, CE, TAG, and other lipids, if available. The analytes were detected in either APCI, or electrospray ionization modes [7,19]. The raw data were acquired in the m/z 100 to m/z 2000 range with 5 to 10 mDa accuracy, and individual lipids were analyzed as EIC. After their integration, relative abundances of the individual analytes were compared as fractions of the total lipid peaks, or as fold changes in meibum samples of different types.

Then, the data were exported to the Progenesis QI version 2.3 and EZinfo version 3.0.3 software packages (also from Waters) for final analyses using an unbiased, untargeted multivariate PCA and/or PLS-DA approaches.

4.4. Differential Scanning Microcalorimetry

The meibum samples dissolved in IPA were loaded into the sample cell of the differential scanning microcalorimeter (VP-DSC from MicroCal/GE, Piscataway, NJ, USA) using a microsyringe, and the solvent was evaporated under a stream of dry ultrapure nitrogen, while the reference cell was left empty. Typically, between 200 μg and 300 μg of dry lipid material were loaded for each DSC experiment. Knowing the exact amount of loaded material was not essential for this study as the results were compared after normalization of DSC thermograms of all samples (the DSC response from 0% to a maximum of 100%). However, larger amounts of the samples radically improved the signal-to-noise ratio and provided more reproducible results. Then, the cells were sealed, and the heating/cooling cycles from −9 °C to +80 °C were started at a typical scan rate of 2 °C/min. Fifteen consecutive thermograms obtained for a representative canine sample are shown in Figure 12A as overlaid curves, 14 of which (from #2 to #15) demonstrated high reproducibility of the experiments. The first heating/cooling cycle is considered to be a preparatory one [37] and has been excluded from further analyses. For actual meibum melting experiments, heating/cooling cycles were repeated 5 times or more. Then, the scans from #2 and up were averaged using MicroCal/Origin 7 software version 7.05 supplied with the VP-DSC instrument, the baseline correction was performed using the baseline correction algorithm of the software, and the parameters of the melting curves, such as T_m, peak areas A, and T_1/2, were determined (Figure 12B). These parameters were calculated directly from the thermograms using the “Pick Peaks” and “Integrate from Zero” commands of the Origin/MicroCal software. Other approaches and/or programs can be used for the same purposes [68,69]. Where applicable, the thermograms were deconvoluted using the “MN2State: Cursor Initiated” routine of the Origin/MicroCal software, and the T_m values of the deconvoluted melting peaks were determined. Also calculated were apparent van’t Hoff enthalpy changes (ΔH_V) and calorimetric transition enthalpies (ΔH). The use of apparent ΔH_V and ΔH was dictated by uncertainty in the estimation of the average molecular weights of all but human study samples and are provided here only to illustrate the ability of the tested approach to simulate the actual thermograms of meibum.

4.5. Transcriptomic Analysis of Mouse Tarsal Plates and Corneas

The expression levels of Awat2, Soat1, Elovl3, and Sdr16c5/Sdr16c6 genes in mouse corneas and MG were determined at the UTSW Microarray and Immune Phenotyping Core using Clariom D microarrays (from Affymetrix, Santa Clara, CA, USA).

4.6. Statistical Analyses

LC/MS data were initially prepared using the Progenesis QI software package and exported to EZinfo for final analyses using an unbiased, untargeted multivariate PCA and PLS-DA approaches. All programs were obtained from Waters Corp. To compensate for the uneven physical sizes of the study samples, each dataset was normalized against the total ion count for that sample. The following PLS-DA model was used: Pareto scaling with no transformation. Two-group targeted comparisons of individual Meibomian lipids in wild-type vs. mutant mice were conducted using Student’s t-test (Prism 8, version 8.0.2, from GraphPad Software, Boston, MA, USA). The statistical significance of multiple t-tests for multiple types of samples was determined using the Holm–Sidak method (α = 0.05) and the two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli (Q = 5%). Each row was analyzed individually, without assuming a consistent SD. DSC data are shown as Means ± Standard Deviation (n ≥ 3 in Table 1, or n = 11 in Table 2).

Transcriptomic data were exported to the Expression Console and Transcriptome Analysis Console v. 4.0.3 (from ThermoFisher) for quantitation as described earlier [58]. An empirical Bayes ANOVA (eBayes) algorithm was used for ranking differentially expressed genes. Based on industry-standard settings, the linear fold changes in gene expression patterns that satisfied the following criteria—(−2) ≥ LFC ≥ (+2) and p ≤ 0.05—were considered to be statistically significant.

5. Conclusions

Our data demonstrated that the high complexity of the (bio)chemical composition of normal meibum serves the same goal in all tested species—maintaining its proper physico-chemical properties, including, but not limited to, its melting characteristics. Major alterations in the balance between main classes of Meibomian lipids, or even changes within individual classes of lipids, may lead to abnormal expressibility, the loss of meibum cohesiveness and, generally, its ability to protect the eye. However, even completely different lipid secretions—such as human and rabbit meibum—may demonstrate comparable thermotropic behavior that is adjusted to provide proper eye protection. DSC proved to be a valuable tool that, together with other approaches, such as LC/MS, can be used in research and clinical practice to detect, characterize, and classify different forms of Meibomian gland dysfunction and other related pathologies, uncover their molecular mechanisms, and help to devise strategies for their treatment.