Triacylglycerol Lipidome from Human Meibomian Gland Epithelial Cells: Description, Response to Culture Conditions, and Perspective on Function

Abstract

Preliminary work has shown that select triacylglycerols (TAGs) are upregulated in a preclinical model of MGD, suggesting that TAGs may be an important outcome variable in research involving human meibomian gland epithelial cells (HMGECs). The purpose of this study was to explore the HMGEC TAG lipidome in culture conditions known to influence differentiation. HMGECs were differentiated in DMEM/F12 with 10 ng/ml EGF, FBS (2% or 10%), and rosiglitazone (0, 20, or 50 μM) for two or five days. Following culture, lipids were extracted, processed, and directly infused into a Triple TOF 5600 mass spectrometer (SCIEX, Framingham, MA) with electrospray ionization. MS and MS/MS^ALL spectra were acquired in the positive ion mode and performed with the SWATH technology. Only the TAGs that were present in all 48 samples were included in the analysis. Multiple regression techniques were utilized to assess the effects of each factor (FBS, rosiglitazone, and culture duration) on each expressed TAG. The HMGEC TAG lipidome consisted of 115 TAGs with 42 to 62 carbons and zero to 10 double bonds. Fatty acyl chains had 14 to 26 carbons and zero to five double bonds. C18:1 (oleic acid, 25/115, 21.7%) and C16:0 (palmitic acid, 16/115, 13.9%) were the most common fatty acids. FBS, rosiglitazone, and culture duration were significant predictors for 93 TAGs (80.9%) with R² values ranging from 0.20 to 0.77 (p < 0.05). FBS and rosiglitazone achieved significance (p < 0.05) for 80 (69.6%) and 67 TAGs (58.3%), respectively. Rosiglitazone demonstrated a selective upregulation of TAGs containing 16 or 18 carbons. Culture duration reached significance (p < 0.05) for only 36 TAGs (31.3%). When comparing the 10 most abundant C18:1-containing TAGs in meibum, FBS was a negative predictor for five TAGs (mean standardized coefficient [SC] = −0.58, p < 0.001), rosiglitazone was a positive predictor for six TAGs (mean SC = 0.41, p ≤ 0.03), and culture duration weakly influenced one TAG (SC = 0.27, p = 0.008). FBS and rosiglitazone, unlike culture duration, are powerful modulators of the TAG profile. Rosiglitazone induces changes that could be consistent with fatty acid synthesis, suggesting that quantifying the TAG lipidome could be an indirect measure of lipogenesis. Though both have been described as differentiating agents, FBS and rosiglitazone induce opposing effects on meibum-relevant TAGs. Culturing with rosiglitazone is associated with a TAG profile that is more consistent with the expected outcome of lipogenesis and with the profile observed in normal human meibum.

1. Introduction

Immortalized human meibomian gland epithelial cells (HMGECs) have immense potential in meibomian gland dysfunction (MGD) research. The use of preclinical models permits better assessment of cause-and-effect relationships but only to the extent that they replicate normal physiology. Early reports of HMGECs revealed increased gene expression for lipogenic enzymes in response to androgen exposure^1,2—an expected behavior of lipid-producing meibocytes. Manifesting the expected phenotype, HMGECs have also been shown to increase nonpolar lipid production following exposure to a variety of differentiating agents, a finding that is often assessed by vital dyes and fluorescent microscopy.^1,3-10 This method provides strong visual evidence of nonpolar lipid upregulation, but it lacks differentiation among the various lipid classes (nonetheless, species), complicating our understanding of whether the produced lipidome truly mirrors that from the meibomian gland in vivo.

Meibomian gland secretion, termed meibum, is a complex mixture of primarily nonpolar lipids. The composition of this mixture has been extensively analyzed by mass spectrometry, a tool that permits the identification and quantification of specific lipid species.¹¹ Previous reports have shown that the nonpolar wax and cholesteryl esters dominate the lipid pool, accounting for nearly 90% of all lipids.¹² Triacylglycerols (TAGs), another nonpolar lipid family, have been cited to represent between 0.05% to 6%.^13-17 Although TAGs do not represent a major lipid class in meibum, our preliminary studies have suggested that select TAGs are upregulated in a preclinical disease model of MGD,¹⁸ suggesting that they may be an important indicator of pathology. To date, however, a description of the TAG lipidome produced by HMGECs has not been published, nor is it known how it responds to culture conditions known to influence HMGEC differentiation.

Although several differentiating agents for HMGECs have been described, the original mechanism for inducing differentiation utilized 10% serum in culture media.^1,3 This method has since been challenged, however, in more recent reports that failed to detect a significant increase in lipid production.^9,19 More recently, the use of rosiglitazone, a peroxisome proliferator activator receptor-γ (PPARγ) agonist, was introduced in a compelling series of experiments that has recently been expanded to include whole transcriptome analysis.^9,20 PPARγ, a key player in both sebocyte and adipocyte differentiation, is a nuclear receptor that upon binding to one of its many ligands, such as rosiglitazone, associates with other nuclear receptors and cofactors to ultimately regulate transcription of a variety of genes involved in cellular differentiation and lipid metabolism.^21,22 We previously evaluated the effects of rosiglitazone on the expression of cholesteryl esters by HMGECs and found that the lower serum concentration (2%) and the highest rosiglitazone concentration (50 μM) after two days of culture yielded a more meibum-like profile.²³

In this study, we sought to define the TAG lipidome, specifically related to its fatty acid (FA) composition and assess its viability as an outcome measure in HMGEC lipidomic research. To this end, we evaluated the effects of serum, rosiglitazone, and culture duration in a 2 x 3 x 2 experimental design using multiple regression techniques to explore the collective and individual effects of each factor on each expressed TAG.

2. Material and methods

2.1. Cell culture

Immortalized HMGECs were maintained in proliferating conditions in keratinocyte serum-free media (KSFM), 5 ng/ml epidermal growth factor (EGF), and 50 μg/ml bovine pituitary extract (BPE) until 80% confluence.¹ HMGECs were split (passages 25-26) into 6-cm glass petri dishes at a density of one million cells per dish and exposed to various differentiating conditions, all of which included Dulbecco’s modified Eagle’s medium (DMEM) and Ham’s F12 (1:1) with 10 ng/ml EGF. All differentiation media also consisted of fetal bovine serum (FBS, 2% or 10%) and rosiglitazone (0 μM, 20 μM, or 50 μM, Cayman Chemical, Ann Arbor, MI). Stock solutions (4 mM or 10 mM) were made by dissolving rosiglitazone into sterile-filtered dimethyl sulfoxide (DMSO, Hybri-Max™, Sigma-Aldrich, St. Louis, MO) and stored under nitrogen at −20°C. Rosiglitazone was added to each media immediately prior to introducing it to the cells. The concentration of DMSO was held constant at 0.5% across all samples. Differentiating conditions were maintained for either two days or five days; media changes were performed every other day. All experiments consisted of two technical replicates and two experimental replicates for a total of 48 samples across the 12 conditions

2.2. Lipid extraction

After incubation for two days or five days, HMGECs were washed twice and then simultaneously harvested and extracted by the direct application of 3 ml chloroform-methanol (2:1, v/v), an adaptation of the Folch technique,²⁴ to increase extraction efficiency from cultured cells.²⁵ The surface of the glass dish was scraped with a sterile stainless-steel scraper, and the sample was transferred to a glass vial. Ammonium acetate (0.75 ml, 50 mM) in molecular biology-grade water was added to each sample, and the resulting emulsion was agitated on ice at 350 rpm for 20 minutes. To facilitate phase separation, the samples were centrifuged at 1600 x g for five minutes. The lower, nonpolar phase was withdrawn by glass pipet and stored at −80°C until analysis. All materials that contacted organic solvents were made of glass, polytetrafluoroethylene (PTFE), or stainless steel.

2.3. Analysis by mass spectrometry (MS)

Samples were analyzed by mass spectrometry using methods previously described.²³ Briefly, dried lipids were reconstituted in 2:1 methanol/chloroform (v/v) with 5 mM ammonium acetate, and the solution was directly infused into a Triple TOF 5600 mass spectrometer (SCIEX, Framingham, MA) with electrospray ionization at a flow rate of 7 μl/min. The direct infusing syringe was cleaned before and after each sample run with two flushes of 100% methanol, 100% acetonitrile, 100% isopropyl alcohol, and 2:1 methanol-chloroform. MS and MS/MS^ALL spectra were acquired in the positive ion mode and performed with the SWATH technology (Sequential Window Acquisition of all THeoretical mass spectra, SCIEX, Framingham, MA). Product ion MS/MS spectra for all parent ions were acquired between 200 to 1200 m/z at every one Dalton step. Collision energy was fixed at 35 eV. The acquisition time was six minutes per sample.

2.4. Data analysis

All acquired MS data were processed using LipidView 1.2 (SCIEX, Framingham, MA), which assigns lipid identities based upon known ion fragmentations. The mass tolerance window was set to 5 mDa. Only the TAG peaks with a signal-to-noise ratio greater than three that were present in all samples were included in the analysis. Each analyzed TAG was normalized to the sum intensity and is therefore reported as percent of the overall TAG pool. The labeling convention for TAGs is TAG n_c:db (FA or TAG n_c:db), where n_c is the total carbon count of the fatty acyl chain(s) and db is the number of double bonds. The parent ion is provided first, and the product ion (either a FA or a TAG) is provided second. For example, TAG 54:3 (FA 18:1) denotes that a fatty acid with 18 carbons and one double bond was detected in the product ion scan from the parent TAG of 54 carbons and three double bonds.

To determine the response of the TAG lipidome to each of the three variables—FBS, rosiglitazone, and duration of culture—all data were analyzed by multiple regression techniques using SPSS v26 (Armonk, NY). Rosiglitazone was analyzed in two concentrations (20 μM and 50 μM), rather than as a continuous variable. This technique is valuable in multiple regression when data are not assumed to be linear, an important consideration since Kim et al. observed a non-linear relationship between lipid production and rosiglitazone between 0 and 50 μM.⁹ When the assumptions of normality (Kolmogorov-Smirnov) and equal variance (Levene’s test) were violated, the data were transformed by ranking prior to performing the multiple regression analysis. A p-value of 0.05 was considered statistically significant.

3. Results

3.1. Description of the TAG lipidome

Across all samples, 2,131 unique TAGs were identified; however, only 115 met the inclusion criterion of being present in all 48 samples. The total carbon count from the three acyl chains, excluding the glycerol backbone, ranged from 42 to 62 with the majority (83/115, 72.2%) falling within the range of 48 to 54 (Figure 1A). An even carbon count was significantly more common (90/115, 78.3%) than an odd carbon count (25/115, 21.7%).

Figure 1.

HMGECs were treated with varying concentrations of FBS and rosiglitazone for two days or five days. Lipid extracts (n = 48) were analyzed by tandem mass spectrometry in the positive ion mode to identify triacylglycerols (TAGs). (A-C) There were 115 identified TAGs present in all 48 samples with carbon counts varying from 42 to 62 and double bonds varying from zero to 10. LipidView 1.2 detected the neutral loss of 17 unique fatty acyl chains. FA 18:1 and FA 16:0 were the most frequently observed. (D-L) Multiple regression analyses were performed to assess the individual contribution of each factor (FBS in Panels D-F, rosiglitazone in Panels G-I, and culture duration in Panels J-L) to the variance observed in each TAG. Each condition consisted of two experimental replicates and two technical replicates. Notable observations include a narrow range of carbon counts (48 to 56, Panel G) and FA-containing TAGs (16 or 18 carbons, Panel I) that increase with rosiglitazone. Further descriptions of the distributions are detailed in the text.

The carbon count includes the total number of carbons present in the three acyl chains. Fatty acids are labeled as two numbers separated by a colon, where the first number is the number of carbons and the second number is the number of double bonds.

HMGEC = human meibomian gland epithelial cell

TAG = triacylglycerol

FBS = fetal bovine serum

Rosi = rosiglitazone

FA = fatty acid

The number of double bonds in the acyl chains of the TAGs varied from zero to 10. Very few TAGs were fully saturated (4/115, 3.5%). The degree of unsaturation followed a bimodal distribution (Figure 1B). TAGs were primarily of low unsaturation (49/115, 42.6%, one to three double bonds) or of high unsaturation (61/115, 53.0%, six to 10 double bonds). Only five of 115 (4.4%) had a moderate degree of unsaturation (four to five double bonds).

The LipidView 1.2 software identified the neutral loss of 17 unique fatty acyl chains from the 115 TAGs (Figure 1C). Their individual carbon counts varied from 14 to 26 with double bonds ranging from zero to five. Similar to the parent TAG molecules, an even carbon count (84/115, 80.0%) was much more common than an odd carbon count (15/115, 15.2%). The most frequently observed fatty acyl group was oleic acid (FA 18:1), which was present in 25 of 115 TAGs (21.7%). The second most common was palmitic acid (FA 16:0), which was present in 16 (12.8%) TAGs. Very few TAGs, only seven of 115 (6.1%), consisted of very long-chain fatty acids (at least 20 carbons).

3.2. FBS, rosiglitazone, and culture duration—collectively—on the TAG profile

Multiple regression was performed for each of the 115 TAGs to determine the predictive capacity—or contribution of effect—of each of the factors (FBS, rosiglitazone, and culture duration). Of the 115 TAGs analyzed, the factors significantly predicted 93 (80.9%) TAGs. Of these 93, zero were saturated, eight (8.6%) were monounsaturated, and the remaining 85 (91.4%) were polyunsaturated. The eight that were monounsaturated represented 50% of the overall monounsaturated TAGs produced by HMGECs. Statistically significant R² values ranged from relatively weak (0.20) to strong (0.77) (Figure 2).

Figure 2.

Each dot represents one of 115 identified TAGs that were present in all 48 samples of HMGECs treated with varying concentrations of FBS and rosiglitazone for two days or five days. Multiple regression was performed to determine the predictive potential of all three factors—FBS, rosiglitazone, and culture duration—on the variance observed for each TAG. The model was able to significantly predict 93 of 115 TAGs, as denoted by the black dots that fall beneath the p = 0.05 reference line. Significant R² values ranged from 0.20 to 0.77, and p-values ranged from < 0.001 to 0.978.

TAG = triacylglycerol

HMGEC = human meibomian gland epithelial cell

FBS = fetal bovine serum

Only 22 (19.1%) of 115 TAGs showed no association with the statistical model. Of these 22, their profile was largely consistent with the overall TAG profile, possessing similarities in both carbon counts and double-bond counts. Among these TAGs, the range of their carbon numbers varied from 44 to 62, and the double-bond count varied from zero to eight. Of note, all four (100%) of the saturated TAGs detected from the samples were not predicted by the factors in the model: TAG 46:0 (FA 16:0), TAG 48:0 (FA 16:0), TAG 48:0 (FA 14:0), and TAG 52:0 (FA 16:0).

3.3. Influence of FBS on TAG profile

To determine the individual contribution of FBS on each of the 115 TAGs analyzed, the standardized coefficients (SC) from the multiple regression analysis were evaluated. Standardized coefficients are a ratio of the amount of change in standard-deviation units for the outcome variable (here, each TAG) per corresponding change in the predictor variable (here, FBS). Controlling for both rosiglitazone and culture duration, FBS was a significant predictor for 80 of 115 TAGs (69.6%), where 49 TAGs (61.3%) were positively associated and 31 TAGs (38.7%) were negatively associated. In general, TAGs with nearly all carbon counts and double-bond counts were affected, ranging from 42 to 62 and one to 10, respectively (Figure 1D-E). Of the 25 TAGs containing FA 18:1, 10 (40.0%) were not associated with FBS, five (20.0%) were positively associated, and 10 (40.0%) were negatively associated (Figure 1F). Of the 10 that were inversely associated, six of them (60%) were among the most responsive to changes in FBS (Table 1).

Table 1:

Most Responsive TAGs (90^th Percentile) Per Factor

	Most Positively Associated			Most Negatively Associated
	TAG	Std Coeff	p-value	TAG	Std Coeff	p-value
FBS	TAG 48:7 (TAG 36:1)	0.75	<0.001	TAG 56:3 (FA 20:1)	−0.78	<0.001
	TAG 54:8 (FA 16:4)	0.73	<0.001	TAG 48:2 (FA 18:1)	−0.74	<0.001
	TAG 54:9 (FA 17:0)	0.72	<0.001	TAG 56:3 (FA 18:1)	−0.71	<0.001
	TAG 54:1 (FA 18:0)	0.72	<0.001	TAG 50:3 (FA 18:1)	−0.65	<0.001
	TAG 44:6 (TAG 32:0)	0.72	<0.001	TAG 50:2 (FA 18:1)	−0.64	<0.001
	TAG 52:6 (FA 14:2)	0.70	<0.001	TAG 51:10 (FA 16:1)	−0.64	<0.001
	TAG 52:8 (FA 17:0)	0.68	<0.001	TAG 48:2 (FA 16:1)	−0.64	<0.001
	TAG 54:9 (FA 18:1)	0.67	<0.001	TAG 54:3 (FA 18:1)	−0.62	<0.001
	TAG 58:1 (FA 16:0)	0.66	<0.001	TAG 53:10 (FA 18:1)	−0.61	<0.001
	TAG 53:8 (FA 18:0)	0.65	<0.001	TAG 52:3 (FA 16:1)	−0.60	<0.001
	TAG 48:9 (TAG 36:3)	0.65	<0.001	TAG 46:1 (FA 16:0)	−0.55	<0.001
Rosiglitazone (50 μM)	TAG 56:5 (FA 18:1)	0.81	<0.001	TAG 42:6 (TAG 30:0)	−0.71	<0.001
	TAG 53:10 (FA 16:0)	0.71	<0.001	TAG 48:6 (FA 14:2)	−0.64	<0.001
	TAG 52:3 (FA 18:2)	0.63	<0.001	TAG 50:6 (FA 14:2)	−0.63	<0.001
	TAG 54:4 (FA 18:1)	0.62	<0.001	TAG 42:7 (TAG 30:1)	−0.62	<0.001
	TAG 54:3 (FA 18:2)	0.61	<0.001	TAG 50:6 (FA 16:5)	−0.61	<0.001
	TAG 53:9 (FA 16:1)	0.60	<0.001	TAG 50:7 (FA 14:2)	−0.58	<0.001
	TAG 54:2 (FA 18:0)	0.59	<0.001	TAG 44:7 (TAG 32:1)	−0.58	<0.001
	TAG 53:9 (FA 18:0)	0.59	<0.001	TAG 52:9 (FA 16:4)	−0.56	<0.001
	TAG 54:4 (FA 18:2)	0.59	<0.001	TAG 50:8 (FA 17:0)	−0.56	<0.001
	TAG 56:4 (FA 18:1)	0.56	<0.001	TAG 53:6 (FA 17:5)	−0.55	<0.001
	TAG 51:9 (FA 16:0)	0.55	<0.001	TAG 52:8 (FA 16:4)	−0.55	<0.001
Culture Duration	TAG 52:9 (FA 16:0)	0.49	<0.001	TAG 53:8 (FA 17:5)	−0.70	<0.001
	TAG 60:3 (FA 24:1)	0.48	<0.001	TAG 50:1 (FA 14:0)	−0.61	<0.001
	TAG 54:10 (FA 18:1)	0.45	<0.001	TAG 49:6 (FA 17:5)	−0.53	<0.001
	TAG 52:9 (FA 18:1)	0.44	0.001	TAG 45:8 (TAG 33:2)	−0.52	<0.001
	TAG 60:3 (FA 18:1)	0.39	0.004	TAG 53:9 (FA 17:5)	−0.50	<0.001
	TAG 51:9 (FA 16:1)	0.38	<0.001	TAG 51:6 (FA 17:5)	−0.49	<0.001
	TAG 51:10 (FA 16:1)	0.35	<0.001	TAG 51:7 (FA 17:5)	−0.49	<0.001
	TAG 52:3 (FA 16:1)	0.29	0.001	TAG 53:7 (FA 17:5)	−0.48	<0.001
	TAG 50:2 (FA 18:1)	0.27	0.008	TAG 45:7 (TAG 33:1)	−0.46	<0.001
	TAG 51:9 (FA 16:0)	0.27	0.014	TAG 47:7 (FA 14:0)	−0.40	0.006
	TAG 54:2 (FA 20:1)	0.27	0.04	TAG 48:6 (TAG 36:1)	−0.36	0.004

The TAG labeling convention is described under Data Analysis in the Material and Methods section.

3.4. Influence of rosiglitazone on TAG profile

Although the statistical model does not assume linearity, tests of linearity were performed to determine the relationship between rosiglitazone and each of the TAGs. Only nine of 115 (7.8%) deviated from linearity: TAG 46:6 (TAG 34:0), TAG 48:8 (TAG 36:2), TAG 53:10 (FA 16:0), TAG 54:4 (FA 18:1), TAG 54:4 (FA 18:2), TAG 54:3 (FA 18:2), TAG 54:2 (FA 18:0), TAG 56:4 (FA 18:1), and TAG 58:2 (FA 24:1). There was no individual effect due to the intermediate concentration of rosiglitazone on any of the expressed TAGs, as all p-values were greater than 0.05. When controlling for both FBS and culture duration, however, the highest concentration of rosiglitazone (50 μM) was a significant predictor for 67 of 115 TAGs (58.3%). Of these, 25 TAGs (37.3%) were positively associated and 42 (62.7%) were negatively associated. Of the 25 TAGs that were positively associated, only a narrow range of carbon counts (48 to 56) varied with rosiglitazone; the large majority (20/25, 80.0%) of these were between 50 and 54 (Figure 1G). All 25 were polyunsaturated, primarily bearing two to four double bonds or nine to 10 double bonds (Figure 1H). None of the saturated or monounsaturated TAGs were positively associated with rosiglitazone. Of particular interest, TAGs with only a small subset of FAs (carbon number 16 or 18) were positively associated with rosiglitazone (Figure 1I): FA 18:1 (11/25, 44.0%), FA 16:1 (5/25, 20.0%), FA 16:0 (4/25, 16.0%), FA 18:2 (3/25, 12.0%), and FA 18:0 (2/25, 8.0%). The most responsive TAGs to rosiglitazone supplementation are provided in Table 1.

Of the 42 TAGs that were negatively associated with rosiglitazone, the profile showed less selectivity and was more similar to the overall TAG profile for HMGECs. The carbon count ranged from 42 to 60, and the double-bond distribution was bimodal with peaks at one to two double bonds and six to nine double bonds (Figure 1G-H). TAGs with 12 different fatty acids consisting of individual carbon counts varying from 14 to 26 inversely varied with rosiglitazone (Figure 1I). Of note, only one FA 18:1-containing TAG decreased (TAG 58:2).

3.5. Influence of culture duration on TAG profile

Controlling for both FBS and rosiglitazone, culture duration was a significant predictor for 36 (31.3%) of 115 TAGs, where 14 (38.9%) were positively associated and 22 (61.1%) were negatively associated. Of the 14 TAGs that were positively associated, their carbon counts ranged from 48 to 60, and their double-bond counts were bimodally distributed with peaks at two to three and nine to 10 (Figure 1J-K). Only select FA-containing TAGs were positively associated with culture duration (Figure 1L): FA 18:1 (6/14, 42.9%), FA 16:1 (4/14, 28.6%), FA 16:0 (2/14, 14.3%), FA 20:1 (1/14, 7.1%), and FA 24:1 (1/14, 7.1%).

Of the 22 TAGs that were negatively associated with culture duration, their carbon counts ranged from 42 to 56, and their double-bond counts ranged from one to two and five to nine (Figure 1J-K). A similar pattern of selectivity for certain fatty acids was seen among the negatively associated TAGs, though the selectivity was different. TAGs containing FA 17:5 (7/17, 41.2%), FA 18:0 (4/17, 23.5%), FA 14:0 (3/17, 17.7%), FA 18:1 (2/14, 11.8%), and FA 14:2 (1/17, 5.9%) were negatively associated with culture duration. All seven of the FA 17:5-containing TAGs that were present in the overall TAG lipidome were inversely associated with culture duration (Figure 1L). The most responsive TAGs to culture duration are provided in Table 1.

3.6. Opposing effects of FBS and rosiglitazone on meibum-relevant TAGs

The collective and individual contributions of FBS, rosiglitazone, and culture duration were evaluated on the 10 most abundant TAGs containing FA 18:1 in meibum (Table 2).²⁶ The model statistically significantly predicted six of the 10 TAGs with R² values ranging from 0.32 to 0.58 (p < 0.01 for all). All three variables added significantly to the prediction for TAG 50:2 (p < 0.01). FBS and rosiglitazone—but not culture duration—added significantly to the prediction for four TAGs: TAG 54:3, TAG 52:2, TAG 54:4, and TAG 56:3 (p < 0.05 for all). Rosiglitazone was the only significant predictor in the model for TAG 54:2 (p = 0.001). Of note, FBS and rosiglitazone had opposing effects on the meibum-relevant TAGs that were statistically significant (Figure 3). FBS was always negatively associated, while rosiglitazone was always positively associated.

Table 2:

Predictive Capacity of Each Factor for Common Meibum-Relevant TAGs

Meibum-Relevant TAG	Overall Model		FBS		Rosiglitazone (50 μM)		Culture Duration
	R2	p-value	Std. Coeff.	p-value	Std. Coeff.	p-value	Std. Coeff.	p-value
TAG 54:3 (FA 18:1)	0.58	<0.001	−0.62	<0.001	0.43	0.001	0.14	0.18
TAG 52:2 (FA 18:1)	0.39	<0.001	−0.53	<0.001	0.31	0.03	0.15	0.21
TAG 54:2 (FA 18:1)	0.32	0.002	−0.20	0.13	0.53	0.001	0.02	0.90
TAG 52:3 (FA 18:1)	Not detected
TAG 53:2 (FA 18:1)	Not detected
TAG 52:1 (FA 18:1)	0.11	0.26	0.26	0.08	−0.07	0.67	−0.21	0.16
TAG 50:2 (FA 18:1)	0.58	<0.001	−0.64	<0.001	0.34	0.004	0.27	0.008
TAG 54:4 (FA 18:1)	0.56	<0.001	−0.41	<0.001	0.62	<0.001	0.006	0.95
TAG 56:3 (FA 18:1)	0.58	<0.001	−0.71	<0.001	0.25	0.03	0.16	0.115
TAG 50:1 (FA 18:1)	0.01	0.98	0.08	0.61	−0.01	0.96	−0.06	0.68

The standardized coefficients for FBS, rosiglitazone, and culture duration for each of the 10 most abundant FA 18:1-containing TAGs from Chen et al²⁶ were evaluated to determine the predictive capacity of each factor on meibum-relevant TAGs. Standardized coefficients are a ratio of the amount of change in standard-deviation units for the outcome variable (here, each TAG) per corresponding change in the predictor variable (here, either FBS, rosiglitazone, or culture duration). Graphical data are provided in Figure 3. Bolded values denote statistical significance. The TAG labeling convention is described under Data Analysis in the Material and Methods section.

Figure 3.

The standardized coefficients for FBS, rosiglitazone, and culture duration for each of the 10 most abundant FA 18:1-containing TAGs from Chen et al²⁶ were evaluated to determine the predictive capacity of each factor on meibum-relevant TAGs. Standardized coefficients are a ratio of the amount of change in standard-deviation units for the outcome variable (here, each TAG) per corresponding change in the predictor variable (here, either FBS, rosiglitazone, or culture duration). Numeric data are provided in Table 2. Two of the 10 TAGs were not detected in all of our samples and therefore not included in the analysis. Rosiglitazone, FBS, and culture duration were significant predictors for six, five, and one TAG(s), respectively. Rosiglitazone was always positively associated with meibum-relevant TAG expression, while FBS was always negatively associated. TAGs are labeled by two numbers separated by a colon, where the first number is the total number of carbons in the three acyl chains and the second number is the number of double bonds.

FBS = fetal bovine serum

TAG = triacylglycerol

FA 18:1 = oleic acid

*p ≤ 0.001

#p ≤ 0.01

^{^} p ≤ 0.05

4. Discussion

This report describes the TAG lipidome produced by HMGECs and explores its response to common differentiating conditions. The TAG lipidome from HMGECs is robust, diverse, dynamic, and responsive to culture conditions. Multiple regression techniques were used to assess both the combined and individual effects of serum, rosiglitazone, and culture duration on TAG expression. Serum and rosiglitazone, both purported as HMGEC differentiating agents, were shown to strongly modulate the TAG profile, yet they elicited opposing effects on many meibum-relevant TAGs. The highest concentration of rosiglitazone (50 μM) and the lower concentration of serum (2%) promoted a more meibum-like TAG profile. Culture duration, however, had minimal effect. Together, these findings suggest that culture conditions affect lipid metabolism and that supplementation with the ideal concentrations of serum and rosiglitazone should be carefully considered in preclinical models involving HMGECs. Our results further support previous assertions that rosiglitazone could have a therapeutic role in meibomian gland dysfunction and should be further investigated.^9,27

4.1. Effects of serum on the HMGEC TAG lipidome

The effects of serum on HMGEC differentiation have been a source of debate in recent literature. It has been stated that serum is a strong inducer of HMGEC differentiation;³ however, serum’s effects have been poorly replicated among different research groups.^3,9,19 Some reports describe a significant upregulation of lipid production;^1,3 other reports describe minimal change in lipid production yet an increase in keratinization.^9,19 This keratin upregulation implies that serum-treated HMGECs may be a better disease model of the hyperkeratinization that occurs in MGD pathophysiology.²⁸ Using mass spectrometry, we—and others—have reported production of meibum-relevant cholesteryl esters following 10% serum supplementation, albeit in low relative abundance.^3,19,25 We later found that reducing the serum content to 2% resulted in a more optimized and more similar CE profile to that present in normal human meibum.²³ The exact mechanism for how HMGECs modulate TAG expression in response to serum supplementation, however, remains largely unknown. In the present study, FBS induced significant changes to the TAG profile produced by HMGECs where nearly 70% of all detected TAGs varied in some way. The characteristics of the TAGs that increased and those that decreased were similar to the TAG profile as a whole, complicating the ability to draw any conclusions regarding how serum affects TAG production. Because of this overlap, it is likely that there are several pathways involved in lipid metabolism, both lipogenesis and lipolysis, occurring simultaneously that are influenced by serum.

One intriguing observation is the negative association between serum and many of the FA 18:1-containing TAGs. Specifically, 2% serum, when compared to 10% serum, yielded higher production of 10 TAGs bearing oleic acid (FA 18:1), the most predominant fatty acid in meibum-relevant TAGs.²⁶ Further, six of these 10 TAGs were among those that showed the highest response to serum (Table 1). HMGECs exposed to low-serum environments have less access to serum-borne lipids, proteins, and growth factors that are abundant in high-serum environments. Serum-free or low-serum environments may, therefore, be analogous to a relative state of exogenous lipid starvation. It has been shown that culture media supplemented with lipid-deficient FBS halts cell growth and promotes lipogenesis,²⁹ both of which are believed to be hallmarks of HMGEC differentiation.³⁰ In culture conditions completely devoid of serum altogether, a similar increase in lipid production has also been observed.^31,32 Lipogenesis, or fatty acid synthesis, yields the production of palmitic acid (FA 16:0), which can be further elongated and desaturated to produce oleic acid.³³ Indeed, both palmitic acid and palmitoleic acid (FA 16:1), additional signs of fatty acid synthesis, were also strongly upregulated with the lower serum concentration. Lastly, the observation that cellular oleic acid is higher in a low-serum environment argues against oleic-acid uptake or contamination from the serum-containing media, further supporting the conjecture of increased fatty acid synthesis in low-serum (and likely serum-free) environments. Considering that oleic acid is present in approximately 60% to 75% of all wax esters,^15,34,35 a class that in itself represents 40% to 50% of all meibum lipids,^12,17,34 the production and storage of oleic acid in TAG depots could be of physiologic relevance. Additional work is needed to evaluate the mechanism of serum deficiency on fatty acid synthesis.

The possibility that residual lipids from the serum lipidome could have contributed to the extracted TAG profile from HMGECs, despite careful washing steps, was considered. Had this occurred, HMGECs cultured in 10% serum would have had higher levels of serum-relevant TAGs relative to those cultured in 2% serum. Our results reveal the opposite: the higher serum concentration was typically associated with a decrease in serum-relevant TAGs. We detected 30 variations (based on the fatty acid composition) of the 16 TAGs reported to be abundant in FBS (Table 3).²⁹ Of these, 11 had no significant associations with FBS, 14 were negatively associated, and only five were positively associated. This pattern—or lack thereof—rules out the possibility of trace amounts of the serum lipidome contaminating our HMGEC lipid extracts and reinforces the complex influence that serum has on lipid metabolism.

Table 3:

Standardized Coefficients for FBS and TAGs Reported in Serum

TAG	Standardized Coefficients
	FA 14:0	FA 16:1	FA 16:0	FA 18:2	FA 18:1	FA 18:0	FA 20:2	FA 20:1
TAG 48:3	nd
TAG 48:2	nd	−0.64 p < 0.001	−0.48 p < 0.001	nd	−0.74 p < 0.001	nd	nd	nd
TAG 48:1	−0.18 p = 0.21	−0.15 p = 0.30	−0.28 p = 0.06	nd	−0.21 p = 0.15	nd	nd	nd
TAG 50:3	nd	nd	nd	nd	−0.65 p < 0.001	nd	nd	nd
TAG 50:2	nd	nd	nd	nd	−0.64 p < 0.001	nd	nd	nd
TAG 50:1	0.00 p = 1.00	0.02 p = 0.92	0.34 p = 0.02	nd	0.08 p = 0.61	0.49 p < 0.001	nd	nd
TAG 52:4	nd
TAG 52:3	nd	−0.60 p < 0.001	nd	−0.23 p = 0.05	nd	nd	nd	nd
TAG 52:2	nd	nd	nd	nd	−0.53 p < 0.001	nd	nd	nd
TAG 52:1	nd	nd	0.60 p < 0.001	nd	0.26 p =0.08	nd	nd	nd
TAG 54:4	nd	nd	nd	−0.44 p < 0.001	−0.41 p < 0.001	nd	nd	nd
TAG 54:2	nd	nd	−0.40 p < 0.01	nd	−0.20 p = 0.13	0.24 p = 0.05	nd	−0.31 p = 0.02
TAG 56:4	nd	nd	nd	nd	−0.40 p < 0.01	nd	nd	nd
TAG 56:3	nd	nd	nd	nd	−0.71 p < 0.001	nd	nd	−0.78 p < 0.001
TAG 56:2	nd	nd	nd	nd	−0.05 p = 0.76	0.41 p < 0.01	nd	−0.07 p = 0.63
TAG 56:1	nd

The standardized coefficients for FBS with respect to each serum-relevant TAG²⁹ were evaluated to determine their relationship. Standardized coefficients are a ratio of the amount of change in standard-deviation units for the outcome variable (here, each TAG) per corresponding change in the predictor variable (here, FBS). Positive values describe a positive association, while negative values describe a negative association. Multiple isomers were detected for each TAG provided in Brovkovych et al.²⁹ Fourteen of 19 isomers that reached significance were negatively associated with FBS, ruling out the possibility for FBS contamination or spillover into our extracted HMGEC lipidome. Bolded values denote significance.

FBS = fetal bovine serum

TAG = triacylglycerol

HMGEC = human meibomian gland epithelial cell

4.2. Effects of rosiglitazone on the HMGEC TAG lipidome

The effects of rosiglitazone and the downstream events of PPARγ activation have been well-studied in many different cell types, including HMGECs.^9,20,23 Upon PPARγ activation, transcriptional activity is upregulated for many genes involved in lipid metabolism.²¹ Specifically, lipoprotein lipase, an enzyme that catalyzes the release and facilitates the internalization of fatty acids from extracellular TAGs, is increased.³⁶ Fatty acids are then shuttled through the hydrophilic cytosol via fatty acid transport proteins to areas where they can undergo oxidation to harness their energy or be converted into other long-chain fatty acids through enzyme systems.^36,37 Additionally, new fatty acids are synthesized de novo from acetyl coA in the cytosol.³⁸ Ultimately, the end product of this process is lipid assembly, often taking the form of bundling fatty acids into TAGs for storage in lipid droplets.³⁸ Importantly, gene targets of PPARγ are integrally involved in every step of this process.³⁹

Although our study was not designed to interrogate the mechanisms of PPARγ activation and sequelae, our findings provide a remarkably consistent snapshot of how these concerted actions manifest in the TAG lipidome. We found that 25 TAGs, all possessing 48 to 56 carbons, were increased with 50 μM rosiglitazone. All 25 of these TAGs consisted of a 16- or 18-carbon fatty acid, which likely represents the product of increased fatty acid synthesis, as discussed in Section 4.1. While we cannot rule out the possibility of selective internalization of 16- and 18-carbon fatty acids from the media, past experiments assessing the substrate specificity of lipoprotein lipase have concluded that there is no fatty acid specificity.⁴⁰ All fatty acids are capable of being cleaved from TAGs by lipoprotein lipase and internalized, albeit at different rates.⁴⁰ We propose that after the fatty acids were non-discriminately absorbed and shuttled through the cell, they likely proceeded toward one of three fates, depending on which would have been the most energetically favorable. They would either be unprocessed and rapidly sequestered into TAGs (in the case of 16- and 18-carbon fatty acids), modified by simple enzymatic steps into 16- or 18-carbon FAs, or—especially if several steps would have been necessary—degraded into their components via beta oxidation. The fragments could subsequently be assembled into palmitic acid (FA 16:0) via fatty acid synthase, which is highly expressed in HMGECs and the human meibomian gland,^1,30,41 followed by elongation and/or desaturation to convert into a combination of 16- to 18-carbon fatty acids that are either fully saturated or bearing just one or two double bonds.³³ Additional pathways are likely, if not probable, as comprehensively described in a recent review article on TAG metabolism.⁴² These three pathways, however, may represent the most common pathways and would explain the accumulation of the 16- and 18-carbon fatty acids in response to rosiglitazone. Of note, these fatty acids (FA 18:1, FA 16:0, FA 16:1, and FA 18:2) are among the primary fatty acids found in TAGs in human meibum.¹³ The preference for differentiated HMGECs to accumulate 16- or 18-carbon fatty acids could be a sign of increased demand, particularly by wax esters in need of FA 18:1,^15,34,35 which we found to be present in nearly half of the elevated TAGs associated with rosiglitazone in our study. Taken together, as in other cell types, TAGs likely serve as an interstage in HMGECs to support production of other lipid classes, and evaluating the TAG lipidome may serve as an indirect assessment of lipid metabolism.⁴²

The observation that long-chain fatty acyl groups are not abundantly expressed in HMGEC TAGs may seem counterintuitive, considering that the major cholesteryl esters (CEs) and wax esters found in normal human meibum highly express long-chain fatty acyl groups. Further, we recently reported that CE 24:1, 28:1, and 26:1 are the three most responsive CEs produced by HMGECs in response to rosiglitazone supplementation.²⁵ This seemingly discordant observation actually aligns well with the TAG lipidome of normal human meibum. Chen et al previously reported that FA 18:1, 16:0, 16:1, 17:0, and 18:2 are the most abundant fatty acyl groups found in meibum TAGs.¹³ Four of these five fatty acyl groups were also among the major fatty acyl groups found in HMGEC TAGs. Additionally, they were further upregulated with rosiglitazone supplementation. In 2016, Butovich and colleagues proposed a series of metabolic steps that comprise “meibogenesis,” a term used to describe the major final products of meibum (such as cholesteryl esters and wax esters).⁴¹ The most basic building blocks of meibogenesis are FA 16:0 to FA 18:0—the common end products of fatty acid synthesis. We propose, based upon the patterns observed herein and the work of Kim et al,^9,20 that rosiglitazone stimulates fatty acid synthesis, leads to the selective upregulation of these fatty acid building blocks, and stores them in TAG molecules until they are subsequently needed for ongoing meibogenesis.

The highest concentration of rosiglitazone had a negative association with 42 TAGs. Less selectivity was observed among this cohort compared to those that were positively associated, suggesting that, generally, rosiglitazone stimulates the cell to use any fatty acid sources available without much discrimination. There did appear to be a minor preference to decrease the TAGs that included 14- and 15-carbon fatty acids, as well as highly polyunsaturated 16-carbon fatty acids with four or five double bonds. This same pattern was reflected among the parent TAGs also. Those with a high number of double bonds (six to nine) were much more likely to decrease in response to rosiglitazone treatment. The significance of this observation is unknown, but it may represent an attempt by the differentiated cell to tighten its spectrum of fatty acid-containing TAGs (stated differently, to decrease fatty-acid diversity) toward a profile that contains the building blocks that are in greatest demand for ongoing lipogenesis in the meibomian gland.

4.3. Effects of culture duration on the HMGEC TAG lipidome

The last variable that we assessed was culture duration. Compared to serum and rosiglitazone, culture duration demonstrated an overall weaker effect. Fewer TAG molecules varied with respect to culture duration, and the magnitude of change, when change was observed, was also comparatively smaller (Table 1). Approximately 70% of TAGs were unchanged between two days and five days of culture. Of the 30% that did change, we observed that slightly shorter-chain fatty acids (14 to 18 carbons) decreased over time, while slightly longer-chain fatty acids (16 to 24) increased over time. The peaks in the bimodal distribution of unsaturation level also shifted slightly. Those TAGs bearing just one to two or six to nine double bonds decreased, while those with two to three or nine to 10 increased. These observations may provide hints to the kinetics of the enzymes involved in elongation and desaturation within the differentiated HMGEC. Perhaps these enzyme systems experience a greater degree of upregulation between two and five days of culture. Ultimately, however, the overall weaker effect of culture duration is worthy of re-emphasis. Despite the observation of some mild trends in the TAG lipidome, the vast majority of TAGs did not vary with time.

In the literature, across a wide variety of differentiating approaches, HMGECs have been exposed to differentiating conditions for as little as one day and for up to 28 days.^3,10,19 Lipid production by HMGECs is a frequent outcome variable; however, its response to serum differentiation is poorly replicated among different laboratories. An early report stated that serum-differentiated HMGECs dramatically increase lipid production over time up to 13 days.³ Another report stated that lipid production was maximized after one day and waned thereafter.¹⁹ A third report failed to find a significant increase in lipid production following serum differentiation through a period of two to six days.⁹ Using mass spectrometry, our previous work has shown that meibum-relevant cholesteryl esters are detected after just two days of culture in serum-containing media ²⁵ and that extending culture duration to five days had a negligible effect on these CEs²³. Our present work focuses on TAGs and is consistent with our previous results: duration of culture has less of an impact on the TAG lipidome compared to other differentiating factors.

Rosiglitazone-induced differentiation, however, has indeed shown a time-dependent increase in lipid production over two to six days of culture using vital dye and fluorescent microscopy.⁹ The authors of this same paper also performed transcriptome analysis on rosiglitazone-differentiated HMGECs and found that genes involved in cellular differentiation and lipid metabolism were already upregulated after just 24 hours and likely sustained for up to six days.²⁰ Our decision, therefore, to assess the lipidome at two days and five days is supported by the kinetics of PPARγ activation by rosiglitazone: the necessary genes are upregulated during this interval. One important distinction worth mentioning between our methodology and that of these other reports is that our methods allow us to evaluate changes in the relative abundance of specific lipid species but do not allow us to detect absolute changes in overall lipid production as a whole. A direct comparison of lipid production, therefore, would not be appropriate. A strength of our approach, however, is the ability to identify and quantify the exact lipids that are being produced to ensure consistency with the human meibum profile obtained in vivo. Based on our work, we conclude that the duration of culture has a relatively weak effect on the TAG lipidome (and the CE lipidome²³) and that the decision to use a certain culture duration should be guided by research objectives, methodology, and outcome variables.

4.4. Comparison of the TAG lipidomes between HMGECs and human meibum

A common criticism of the HMGEC line is the discrepancy between its lipidome and that of meibum’s, a topic we have previously discussed.^3,25,43 It is therefore of utmost importance to evaluate and establish culture conditions that yield a lipidomic profile that is as close to human meibum as possible. We previously reported that 2% FBS and 50 μM rosiglitazone in DMEM/F12 supplemented with EGF optimized the cholesteryl ester profile of HMGECs.²³ In our present paper, we have focused not only on the TAG lipidome, but also on defining what contribution each factor (FBS, rosiglitazone, and culture duration) has. To do this, we specifically evaluated each factor’s contribution to the 10 most abundant 18:1-containing TAGs reported in human meibum by Chen et al.²⁶ Two of the 10 did not meet our inclusion criteria and were therefore not included in our analysis. Six of the remaining eight varied significantly with serum, rosiglitazone, and/or culture duration. Consistent with our discussion in Section 4.3, culture duration provided little contribution to the observed changes and only reached significance with a low magnitude of effect for one meibum-relevant TAG. FBS and rosiglitazone, however, were significant predictors for five and six meibum-relevant TAGs, respectively. Interestingly, though, their effects were in opposite directions. Across all significant meibum-relevant TAGs, the highest concentration of rosiglitazone and the lower concentration of serum yielded a more meibum-like profile. Although both substances have been purported to be differentiating agents, their divergent effects suggest otherwise. Our findings support the growing body of literature that serum alone is a poor differentiating agent for HMGECs, at least when evaluating variables related to lipid production and/or expression.^19,27,43 Conversely, rosiglitazone appears to induce a genotype and phenotype that are more consistent with what would be expected from differentiated HMGECs.^9,20,23

Recently, Butovich and Suzuki described a male patient with abnormal meibum, tear, and sebum lipid profiles compared to normal controls.⁴⁴ The normal TAG profile consisted of TAGs with fatty acyl chain lengths varying from 12 to 20 carbons and zero to five double bonds per full TAG molecule. The abnormal TAG profile had a six-fold decrease in triolein (TAG 54:3, FA 18:1/18:1/18:1, 32.9% of all TAGs in normal human meibum) and a ten-fold increase in shorter-chain TAGs. (The reader should note that the labeling convention used by Butovich and Suzuki includes the three-carbon glycerol moiety, resulting in a three-carbon difference in labeling between TAGs in this manuscript compared to their manuscript.) Interestingly, the TAG profile from HMGECs supplemented with rosiglitazone was associated with an increase in TAG 54:3 (FA 18:1), whereas the profile from HMGECs supplemented with higher FBS was associated with a decrease in TAG 54:3 (FA 18:1). Further, the TAG profile from HMGECs supplemented with a higher level of FBS revealed a higher level of unsaturation (double bond counts ≥6), representing a significant shift away from normal. These comparisons further illustrate the positive role of rosiglitazone and the negative role of FBS in the induction of expected phenotype of HMGEC differentiation. Although TAGs were once considered a minor component of the human meibum profile, our past¹⁸ and present work, in combination with the work of Butovich and Suzuki,⁴⁴ help to illustrate the growing awareness of TAGs as a potential biomarker of meibomian gland disease.

4.5. Limitations of the present study

To maximize output, we utilized an efficient, one-step harvesting and extraction method involving the direct application of chloroform-methanol to cells in culture.²⁵ This technique afforded us the opportunity to evaluate more parameters, more conditions, and more replicates, yielding a more robust measurement of the TAG lipidome. Our previous work has shown that the lipid profiles from extracts acquired using this methodology are equivalent to those from trypsin- or EDTA-harvested extracts.²⁵ A shortcoming of this method, however, is the inability to normalize lipid extracts to cell count,²⁵ limiting the ability to detect absolute changes in lipid production. Future work specific to rosiglitazone and serum-free environments with regards to their ability to stimulate absolute changes in lipid production is needed.

Owing to our robust dataset and our comprehensive analytical approach, several patterns emerged related to rosiglitazone and serum, specifically that many FA 18:1-containing TAGs are positively associated with rosiglitazone but negatively associated with FBS, as described in the preceding sections. These observations are hypothesis-generating and are begging for further experimentation. Future work will address other outcome parameters with respect to variations in rosiglitazone, serum content, azithromycin, or other proposed differentiating agents.

4.6. Perspective on the role of TAGs in meibum and the human tear film

Likely owing to their structural complexity and lower abundance relative to some of the major lipid classes (e.g., wax esters and cholesteryl esters), TAGs, as a meibum lipid class, are largely understudied. Systemically, TAGs are known to be dense energy stores, vehicles for fatty acid mobilization through the vasculature, depots of fatty acids for intracellular lipid metabolism, and scavengers of free fatty acids to reduce oxidative stress.⁴² In the human tear film, however, little is known about their function. TAGs have been stated to comprise about 0.05% to 6% of the entire lipid pool in human meibum,^13-17 which begs the question of whether a minor lipid class with such low abundance could confer a structural role to the tear film lipid layer (TFLL). Granted, (O-acyl)-ω-hydroxy fatty acids (OAHFAs), accounting for just 0.69% to 3.1% of lipids in human meibum, are thought to be responsible for TFLL stability;^45-47 however, their amphipathic properties allow them the unique capacity to help form a monolayer at the aqueous-lipid interface to facilitate nonpolar lipid spreading.^48-50 Despite having some degree of polarity due to its glycerol backbone,⁵¹ TAGs, overall, are regarded as nonpolar lipids and lack this amphipathic property. Indeed, they are capable of interacting with this polar lipid monolayer by intercalating two of their fatty acid chains with those of the polar layer, leaving its third chain embedded within the thicker, nonpolar lipid layer. Whether this preferential organization confers any structural effect, as previously proposed by McCulley and Shine, remains to be seen.⁵¹ It could just be the natural positioning of a nonpolar lipid bearing very weak polar activity seeking to minimize a variety of intermolecular forces.

Another hypothesis could link the unique structure of triacylglycerols (a glycerol backbone esterified to three fatty acids) with TFLL fluidity. Indeed, the structure of TAGs could interfere with ordered packing of other major lipid species, possibly decreasing TFLL viscosity. A greater proportion of TAGs relative to wax esters and cholesteryl esters would likely be needed to elicit such an important physiological role. Notably, cholesteryl esters have a five-fold increase in saturation compared to wax esters, conferring the ability to influence phase transition temperatures and meibum viscosity.^12,52 This degree of saturation likely positions cholesteryl esters as a primary modulator of lipid layer viscosity in the tear film.⁵²

It is possible that the presence of TAGs in the TFLL is a result of passive release rather than intentional secretion. Perhaps TAGs in the tear film represent spilled contents from the intracellular compartment of meibocytes during holocrine secretion. Our findings support that under circumstances of induced differentiation due to PPARγ activation by rosiglitazone, the fatty-acid profile of TAGs in HMGECs shifts toward the common end products of fatty acid synthesis (palmitic acid, FA 16:0) elongated and/or desaturated into oleic acid (FA 18:1), stearic acid (FA 18:0), linoleic acid (FA 18:2), and palmitoleic acid (FA 16:1). In vivo, we suggest that these fatty acids, which are in high demand in the lipid-producing meibocyte,^12,15 are bundled into TAGs and stored in lipid droplets until ultimately needed for processing (e.g., further elongation and/or desaturation) and assembly into other lipid molecules. While waiting in queue, the stimulus for holocrine secretion may be received, and the cell may dump these fatty acid depots into meibum, effectively serving as a timestamp of lipid metabolism at the point of holocrine secretion. Once secreted in the tear film, the TAGs likely have biological activity, but the activity conferred by each individual TAG species may be secondary to its primary role of serving as a fatty acid reservoir intracellularly. As such, the parent TAG molecule may have less relevance than its fatty-acid constituents. As the field continues to pivot toward a better understanding of the role of individual lipid classes in the tear film, attention should be given to not just the TAG parent identities (i.e., the sums of carbons and double bonds in the three fatty acyl chains) but also to their specific fatty acids.

5. Conclusions

In conclusion, this study reveals that the triacylglycerol lipidome produced by HMGECs is affected by culture conditions. We found that both serum and rosiglitazone are strong, yet opposing, modulators of TAG expression. Serum independently altered nearly 70% of all TAGs, while rosiglitazone independently altered nearly 60% of all TAGs. Rosiglitazone positively influences many meibum-relevant, oleic acid-bearing TAGs; serum negatively influences many of them. Culture duration, however, elicits a minor effect on a minority of TAGs and is deemed a low contributor, at least when evaluating relative TAG expression. To optimize HMGEC’s ability to serve as a preclinical model, culturing with 50 μM rosiglitazone and 2% serum appears to promote a more meibum-like TAG (and cholesteryl ester²³) profile. Additional work is needed to evaluate serum-free conditions on HMGEC differentiation. Lastly, we proposed the hypothesis that any function associated with TAGs in the TFLL may be incidental or perhaps secondary to its primary role, which may be to serve as a fatty acid depot in meibocytes. TAGs present in the TFLL may represent a snapshot of intracellular lipid metabolism and lipid storage at the time of holocrine secretion. This conjecture was reported as a perspective, rather than a research conclusion, and is an area ripe for further scientific inquiry.

Highlights.

• FBS and rosiglitazone strongly influence the triacylglycerol lipidome from HMGECs.

• Rosiglitazone promotes a TAG lipidome consistent with fatty acid synthesis.

• FBS negatively influences the meibum-relevant TAG lipidome.

• Rosiglitazone appears to be a better differentiating agent than FBS.