Differentiation Patterns of Immortalized Human Meibomian Gland Epithelial Cells in Three-Dimensional Culture

Abstract

Purpose: To establish a simplified three-dimensional (3D) meibomian gland culture model using a meibomian gland epithelial cell (HMGEC) line that might be a useful tool to gain deeper insights into meibomian gland dysfunction. For this purpose, 3D differentiation patterns and growth characteristics of HMGECs were studied on various membranes/scaffolds as well as in hanging drops.

Methods: Several types of inserts consisting of different materials (Millicell-HA, Millicell-PCF, ThinCert, and Alvetex) as well as hanging drop culture were analyzed. Culture conditions were optimized employing exposure to air (air-lift) and different cell culture media for a maximum of 28 days. To characterize cell differentiation in the developed 3D model, the expression pattern of cytokeratins was investigated by immunohistochemistry. Sudan III staining was performed for detection of lipid formation and transmission electron microscopy (TEM) was used for ultrastructural analysis.

Results: Only Alvetex scaffolds and the hanging drop method revealed satisfactory results with regard to 3D culture. Continuous use of proliferation medium (serum-free keratinocyte medium containing epidermal growth factor and bovine pituitary extract) and air-lift were important steps for HMGEC differentiation in 3D culture. However, HMGECs only reached a differentiating state and never became mature or hypermature. When cultured in hanging drops, HMGECs showed serum-induced keratinization processes.

Conclusions: HMGECs have the capability to differentiate in a long-term 3D culture, especially when adapted to an air-rich environment. However, even in the 3D format, HMGECs only reach a state of differentiating meibocytes.

Tissue engineering approaches, including in vitro disease models, benefit from continuously available, functional, and pure cellular phenotypes. In contrast to the standard method of two-dimensional (2D) culture on a plastic substrate, 3D cell cultures establish cell–cell contacts and cell–matrix interactions in a three-dimensional (3D) scaffold system that more closely mimics the biochemistry and mechanics of the microenvironment in vivo. Hence, they constitute a practical alternative in their capacity as more physiological tissue models for the investigation of biological processes in a controlled laboratory environment.^1,2 Thus, 3D cell culture provides an important stepping stone between 2D cell culture in vitro and tissues in vivo. It is a challenging task to constantly simulate the physiological 3D environment throughout the entire course of cell-based studies. This, however, is one of the most critical prerequisites for an accurate translation to animal models and clinical investigations.^2,3 

Meibomian glands are holocrine, modified sebaceous glands that secrete a special lipid (meibum) onto the ocular surface. There, it becomes part of the tear film, increases tear-film stability, decreases aqueous tear evaporation, and provides a smooth optical surface.^4–6 Meibomian gland dysfunction (MGD; the term refers to a diffuse abnormality of the meibomian glands) is considered to be the most common cause of evaporative dry eye,⁷ but can also be responsible for eyelid inflammation that may not manifest as classic dry eye disease (DED).^8–10 Ophthalmologists today recognize the meibomian gland as a key component in the etiology of DED, contributing to the evaporative status of the tear film.^11,12 However, research related to the pathogenesis of MGD is difficult due to various factors. First, meibomian glands are highly specialized glands with a relatively long excretory duct (in relation to sebaceous glands) and a characteristic 3D structure that is embedded in a hard connective tissue plate called tarsus. Secondly, the secretory product of meibomian glands (meibum) develops from meibocytes that regularly undergo mitosis and subsequent apoptosis in the context of holocrine secretion. Therefore, primary culture of meibocytes is challenging due to the short cellular life span and because of its technical complexity to handle. Thirdly, no satisfactory animal model currently exists for MGD. This is of particular significance for MGD-related pharmacologic research, which consequently has yet to produce promising pharmaceuticals. These aspects suggest that effective in vitro models are required to better understand the complex biology of MGD. Thus far, the biological behavior of meibocytes in MGD is still unknown. 

In 2010 Liu et al.¹³ established an immortalized human meibomian gland epithelial cell line (HMGEC), which is suitable for studying the (patho)physiology of human meibocytes in vitro.¹⁴ Previous studies, including some from our group, showed the functional influence of cultivation conditions,^15,16 neurotransmitters,^17,18 growth factors,^19,20 ω-3 fatty acids,²¹ sex hormones,^22,23 and pharmacologic substances²⁴ on HMGEC. However, all of these studies rely on 2D cell culture conditions or limited animal experiments. Traditional 2D monolayer cultures are well established for analyzing cellular responses involving specific signal transduction pathways. However, in vivo meibocytes move toward the center of acini in shifting their microenvironment and progressively accumulate lipid, suggesting that meibocytes dynamically differentiate in response to a heterogeneous environment. Such spatial conditions can be essential to exploit their ability in dynamic meibocyte differentiation, and therefore 3D culture is suitable for providing a more physiologically relevant environment compared to homogeneous 2D monolayer. 

Interestingly, 3D culture was successfully studied using sebocytes that are associated with hairs in the skin.^25,26 These sebocytes share many similarities with meibocytes. In the report of Barrault et al.,²⁷ immortalized sebocytes spontaneously differentiated and produced lipids when 3D cultured at the air–liquid interface. Moreover, Yu et al.²⁸ recently passaged primary rat meibocytes and cultured them successfully as planar sheets under air–liquid interface conditions and in Matrigel matrices where the cells formed spheroids. 

In our quest for a model that could identify the molecular mechanisms underlying MGD, we considered that the development of a 3D culture model using HMGEC might be a rewarding enterprise. The primary aim of this study was therefore to establish a simplified 3D meibomian gland culture model. For this purpose, we tested different types of membrane matrices as scaffolds and optimized the growth conditions for our 3D culture system. We also investigated cytokeratin (CK) expression patterns and lipid accumulation and analyzed cell morphology. The present study is supposed to serve as a first step in building a 3D model using HMGEC, thus representing an important bridge for the combination of our current knowledge of the cell structure and holocrine lipid production in meibomian glands. 

Materials and Methods

Cell Culture

Immortalized human meibomian gland epithelial cells (HMGEC, kindly provided by David Sullivan, Schepens Eye Research Institute, Boston, MA, USA) were cultivated in proliferation medium (serum-free keratinocyte medium supplemented with epidermal growth factor and bovine pituitary extract). To induce differentiation of HMGECs, the cells were incubated in serum-containing differentiation medium as previously described.¹⁴ Medium was refreshed every second day throughout. 

Membrane-Based 3D Cell Culture

HMGECs were seeded onto Millicell-HA, Millicell-PCF (Millipore, Billerica, MA, USA), or ThinCert (Greiner Bio-one, Frickenhausen, Germany) 24-well culture inserts at a density of 2 × 10⁵ cells/cm², followed by 7-day culture until confluent in the proliferation medium. Subsequently, proliferation medium in the apical chamber was removed in order to expose the cells to the air, and the cells were then cultured for further 10 days. Thereafter, cells were processed for conventional histology. Additionally, to investigate long-term growth characteristics, HMGECs were cultured on ThinCert with or without air–liquid interface for up to 28 days in the predetermined medium combinations and monitored under a phase-contrast microscope throughout the experiment. 

Highly Porous Scaffold-Based 3D Cell Culture

According to the manufacturer's protocol, Alvetex 3D scaffold membranes (Alvetex 12-well inserts, AMS.AVP005-34; Amsbio, Abingdon, UK) were activated with 70% ethanol and subsequent medium wash. The scaffolds were then placed into six-well culture plates and HMGECs were seeded at a density of 5 × 10⁵ cells in 50 μL culture medium per insert. The cells were incubated for 1 hour to facilitate cell–scaffold attachment before proliferation media were added. Afterward, the HMGECs were incubated for 7 days under submerged condition to infiltrate the 3D structure of Alvetex and disperse. Inserts were then moved to deep petri dishes (well insert holder in a deep petri dish, AMS.AVP015-2; Amsbio) at the medium position and maintained at the air–liquid interface with approximately 34 mL proliferation medium for up to 5 weeks. In another approach, HMGECs were first cultivated with proliferation medium at the air–liquid interface for 3 weeks and then medium was switched to differentiation medium. Some cells were fixed in 4% paraformaldehyde (PFA) once weekly for 4 weeks after the air–liquid interface culture for immunohistochemistry and lipid staining. The remaining cells were processed at time points of 1, 3, and 5 weeks for ultrastructural analysis. 

Histology

HMGECs cultivated on Millicell-HA, Millicell-PCF, or ThinCert membranes were rinsed with PBS twice and fixed with 4% PFA. Cell conglomerates were embedded in paraffin together with the membranes. Cross sections (7 μm) were created, stained with hematoxylin and eosin, and examined under a Keyence Biorevo BZ 9000 microscope (Keyence, Osaka, Japan). 

Immunohistochemistry

Immunostaining for CK was performed based on the previously described methods with minor modifications.²⁹ In brief, after fixation, 3D-cultured HMGECs in Alvetex were washed with PBS three times. The membrane scaffolds were removed using a scalpel and subsequently embedded in optimal cutting temperature (OCT) compound (Tissue-Tek OCT Compound, Sakura Finetek USA, Inc., Torrance, CA, USA) and frozen at −20°C. Afterward, OCT-embedded membranes were cross-sectioned (thickness: 10 μm) with a cryostat. All sections were mounted on poly-L-lysine–coated slides and cryopreserved. After thawing, the sections were washed with PBS, then blocked with 20% normal serum (corresponding to the same species as the secondary antibody) in PBS for 2 hours at room temperature. The sections were then incubated with primary antibodies against CK1, CK6, CK8, CK10, and CK14 diluted in PBS containing 2% bovine serum albumin and 0.2% Triton-X 100 overnight in a humidified chamber at 4°C. The following day, the sections were rinsed with Tris buffered saline with Tween 20 (TBST) and incubated with the secondary antibody diluted in blocking solution for 30 minutes at room temperature. Visualization was achieved by incubation with aminoethylcarbazole substrate for at least 5 minutes, resulting in reddish-brown reaction product. After counterstaining with hemalum, sections were coverslipped with Aquatex (Merck Millipore, Darmstadt, Germany). Sections of human eyelids taken from donor cadavers served as positive control²⁹ and were used in accordance with the Declaration of Helsinki. Two negative control sections were used: One was incubated with the secondary antibody only, and the other with the primary antibody only. All slides were examined using the BZ 9000 microscope (Keyence). Antibody information can be found in Supplementary Table S1. 

Sudan III Lipid Staining

To visualize lipid droplets in 3D-cultured HMGECs, OCT-embedded cryosections were counterstained with hemalum for 7 minutes and washed for 10 minutes in tap water. After incubation in 50% ethanol for 30 seconds, samples were stained with filtered Sudan III working solution (Merck Millipore; 0.3 mg Sudan III in 70% ethanol) for 15 minutes. To remove excessive staining, samples were washed with 50% ethanol, then coverslipped. Lipid droplets were visible as red color and observed under the BZ 9000 microscope (Keyence). 

Transmission Electron Microscopy (TEM)

Three-dimensional–cultured HMGECs were rinsed with PBS twice and fixed with freshly prepared Ito's fixative (2.5% glutaraldehyde, 2.5% PFA, and 0.3% picric acid dissolved in PBS adjusted to pH 7.3) for at least 24 hours at room temperature. The Alvetex membranes were then cut out as described above. Samples were postfixed in 1% osmium tetroxide for 1 hour. Afterward, samples were dehydrated in an ascending series of ethanols and embedded in Epon (Carl Roth, Karlsruhe, Germany). Semithin sagittal sections (thickness: 1 μm) were created with a microtome (Ultracut E; Reichert Jung, Vienna, Austria) and subsequently stained with toluidine blue. Semithin sections were viewed under the BZ 9000 microscope and photographed. Ultrathin sections were stained with uranyl acetate and lead citrate and analyzed using TEM (JEM 1400 Plus; JEOL Germany GmbH, Freising, Germany). 

Hanging Drop Culture

Hanging drop culture was performed according to the previously established protocol.³⁰ Briefly, HMGECs were maintained under 2D standard conditions and subsequently seeded in a 20-μL drop of proliferation medium containing approximately 1000 cells. Approximately 50 drops were placed on the inner side of petri dish lids. The lids were re-placed on their respective petri dishes containing PBS to generate a humidified atmosphere. Hanging drops were cultured for 48 hours and rinsed with proliferation medium in 0.1% gelatin-coated petri dishes.³¹ Afterward, hanging drops were fixed with Ito's fixative for TEM as previously described. 

Results

Membrane-Based 3D Cell Culture

Using Millicell-HA or Millicell-PCF membranes, HMGECs formed only a few cell layers (data not shown). Moreover, the cell layers were detached from the membrane in almost all cases due to low cell adhesion. In contrast, if cultivated on ThinCert, the HMGECs differentiated into a multilayered epithelium of varying height during air-lift for 7 days (Fig. 1). Basal HMGECs that were in direct contact with the ThinCert membrane were connected to each other via cell protrusions. They were surrounded by circular spaces forming a symmetrical 3D network filled with culture medium. Overlying cells contacted each other and the basal cells directly and along the entire cell surface. Notably, the cells cultivated on ThinCert tended to detach from the membranes. It was therefore vital to handle the cells/membranes carefully. 

Figure 1

 

Vertical histologic section through multilayered HMGECs on ThinCert after air-lift. HMGECs were exposed to air (day 0) and cultured for a further 10 days at the air–liquid interface. The ThinCert membrane occurs as a symmetrical structure at the bottom consisting of a thin dark gray baseline, a thicker pale middle part, and an overlying dark violet thin line to which basal HMGECs are attached. Basal HMGECs stay in contact with each other via protrusions of the cytoplasm and are surrounded by circular spaces forming a symmetrical three-dimensional network. These basal cells are covered by HMGECs that form a multilayered epithelium of varying height. Staining: hematoxylin/eosin. Scale bar: 100 μm.

Having identified the ThinCert membrane as the most suitable matrix for epithelial stratification, we optimized the 3D culture conditions by testing possible combinations of proliferation medium, differentiation medium, and air-lift. Of the four different combinations, air-interface culture with proliferation medium showed the best performance (Figs. 2A, 2E). The cell layers did not develop uniformly on the ThinCert membrane. Instead, they formed mountain-like structures varying in height. The other combination of air-lift with differentiation medium did not illustrate culture success compared with proliferation medium (Figs. 2C, 2G). Submerged culture with proliferation medium alone or combined with differentiation medium led to impaired growth (Figs. 2B, 2D) or even cell death most notably after 4 weeks (Figs. 2F, 2H). An issue highlighted across all conditions was the detachment of cells from the membrane during or after fixation with 4% PFA (data not shown). Furthermore, we sought more suitable conditions with regard to membrane size and medium volume, as 3D culture generally requires more nutrition than conventional 2D culture. The results revealed that the combination of bigger well size with smaller insert size led to better stratification with more “mountain-like” structures (see Supplementary Methods, Supplementary Fig. S1). 

Figure 2

 

Microscopic top view on HMGECs cultured for different times and under different culture conditions on ThinCert. HMGECs were cultured under different conditions on ThinCert until confluent. Growth characteristics and differentiation process were observed using a phase-contrast microscope. Of the four different combinations tested, air-lift combined with proliferation medium (A, E) provided the best performance after 24 hours and after 4 weeks. Also, the combination of air-lift with differentiation medium yielded some success (C, G). However, performance with proliferation medium was clearly better. All other combinations often showed detachment from the membrane during or after fixation. Cultivation without an air–liquid interface either using proliferation medium alone or combined with differentiation medium led to impaired growth (B) or cell death, especially in the course of the 4-week culture period (D, F, H). Where present, stratification was not evenly distributed but was pronounced in some areas and less pronounced in others. This resulted in mountain-like structures with “hills and dales” (A, C, E, G).

Highly Porous Scaffold-Based 3D Cell Culture

On our quest for the ideal 3D culture system for the experimental differentiation of meibocytes, we also tested Alvetex 3D scaffolds. Their highly porous characteristics enable cells to proliferate within the scaffold and allow for sufficient exchange of gases, nutrients, and waste products during cell culture. Usage of noncoated Alvetex scaffolds enabled HMGECs to successfully proliferate, although this approach required 10 times as many cells to infiltrate the entire scaffold as the ThinCert approach. Cytokeratin expression was determined every week over 4 weeks by immunohistochemistry to investigate a possible time-dependent change (Fig. 3). No CK1 reactivity was observed throughout the experiments (data not shown). CK6 reactivity was generally detected prior to air-lift (at week 0), but gradually increased especially on the apical surface of the membrane where the cells had been constantly exposed to air. Intensive CK6 expression was observed within the cytoplasm of HMGECs after 4 weeks. CK8 reactivity clearly increased dependent on the culture period. Interestingly, CK10 and CK14 expression was undetectable or barely present at week 0, yet slightly intensified from week 1 onward especially on the apical surface. To investigate whether differentiation stimuli have an impact on the CK expression and/or morphology in 3D-cultured HMGECs, proliferation medium was replaced by differentiation medium in the basal chamber at week 3, and the air–liquid interface culture was continued for 1 further day (Fig. 4). Three weeks was chosen as HMGECs had formed a continuous epithelium by then. One day after medium exchange was selected based on previous investigations with 2D-cultured HMGECs.¹⁵ In that study, lipid production in HMGECs reached a maximum 1 day after switching to differentiation medium and gradually declined thereafter. Histologic examination revealed that 1-day culture with differentiation medium promoted the apical surface cells to become flattened and transformed, resulting in a thin overlying cell layer. HMGECs within the scaffold exhibited a globular shape during cultivation with proliferation medium, while they displayed a more polygonal morphology after medium switch. No CK1 or CK10 reactivity was observed. There was no clear difference in cytoplasmic CK6 and CK8 expression between before and after the switch to differentiation medium. CK14-positive cells inside the scaffold decreased following the switch to differentiation medium, whereas CK14 staining in the overlying cell layer was comparable regardless of the medium exchange. 

Figure 3

 

Expression patterns of cytokeratins 6, 8, 10, and 14 during long-term 3D culture in Alvetex. HMGECs were cultured at the air–liquid interface and supplied with proliferation medium for 4 weeks. The time course of cytokeratin expression was examined by immunohistochemistry. CK6 and CK8 showed immunoreactivity at week 0 that increased over time up until week 4. No CK10 staining was detectable at week 0. From week 1 onward, CK10 reactivity was visible around the apical surface and intensified up to week 3 and diminished at week 4. Interestingly, CK14 expression showed a time-dependent and continuous increase, especially on the apical surface. All immunohistochemical procedures were performed on 4% PFA-fixed cryosections. Scale bar: 100 μm.

Figure 4

 

Cytokeratin expression and morphologic change in 3D culture following differentiation stimuli. After 3 weeks of 3D culture at the air–liquid interface, the culture medium was replaced by differentiation medium. The cells were then incubated for 1 further day, maintaining the air–liquid interface. Thereafter, cytokeratin expression was visualized by immunohistochemistry. Histology revealed that the differentiation medium led to morphologic changes from a more spheroid to a more polygonal shape. No reactivity was obtained for CK1 and CK10, regardless of the medium used. CK6 and CK8 were positive in the cytoplasm in proliferation medium. No change in the expression pattern was visible after change to differentiation medium. CK14 expressions were clearly reduced inside the scaffold, whereas cells in the overlying cell layer revealed comparable staining intensity after change to differentiation medium. Scale bar: 100 μm.

Lipid Accumulation During Scaffold-Based 3D Cell Culture

Intracellular lipid accumulation was analyzed by Sudan III staining to further characterize HMGECs cultured in Alvetex (Fig. 5). No intracellular lipid accumulation was detected at week 0. However, the scaffold itself bound Sudan III probably due to the dye reagent trapped by cross-linking sites. To test this, we stained the scaffold alone (without HMGECs) with medium pretreatment. This confirmed the dye–scaffold binding (data not shown), and unfortunately it was impossible to wash the stain out completely without breaking down the scaffold structure. Inevitable artifacts were observed in other staining techniques such as Sudan black and Oil red (see Supplementary Methods, Supplementary Fig. S2). Despite this artificial binding to the Alvetex scaffold the number and size of intracellular lipid droplets increased gradually over time with a notable peak at week 3. 

Figure 5

 

Lipid droplet accumulation during long-term 3D culture in Alvetex. HMGECs were cultured at the air–liquid interface, supplied with proliferation medium only for 4 weeks. Intracellular lipid droplets are stained by Sudan III (red). Staining within the scaffold represents an inevitable artifact due to reagent binding. 3D culture promoted intracellular lipid accumulation with a peak at 3 weeks. Scale bar: 100 μm.

Ultrastructure of HMGECs During Scaffold-Based 3D Cell Culture in Alvetex

Ultrastructural analysis revealed that HMGECs in Alvetex did not build a dense cell conglomerate over the whole culture period. Instead, they were only attached to the Alvetex scaffold and connected to each other by small cell protrusions. Around the cells, large extracellular spaces were present, filled with culture medium (Fig. 6A). Over time, HMGECs in the Alvetex scaffold underwent morphologic changes. Twelve hours after dissemination and culture in Alvetex, HMGECs displayed a globular morphology with many large vacuoles. It remained unclear whether these vacuoles actually contained something. In some of them cell detritus was observed. Most of them, however, had an empty appearance (Fig. 6B). After 1 week, most of the cells had become more spindle-shaped and were more closely connected to the Alvetex membrane via cell protrusions or along their entire basal surface. Some of these cells still contained empty vacuoles. However, the vacuoles had shrunk considerably and some even appeared to contain lipids (Fig. 6C). After 3 weeks, most cells contained several lipid droplets, typically located around the cell nucleus (Fig. 6D). After 5 weeks, the number of lipid vacuoles had further increased but never filled the whole cytoplasm (Fig. 6E). Hence HMGECs closely resembled differentiating meibocytes. 

Figure 6

 

Electron microscopic images of the HMGECs during long-term 3D culture in Alvetex. HMGECs were cultivated until confluent in Alvetex. Then, air–liquid interface culture was performed for up to 5 weeks with proliferation medium. 3D-cultured cells were analyzed using transmission electron microscopy. (A) Overview, (B) 12 hours after cultivation onset in Alvetex, (C) 1 week, (D) 3 weeks, (E) 5 weeks. AS, Alvetex matrix; c, cytoplasm; s, space; ld, lipid droplets; stars, “empty” intracellular vacuoles.

Ultrastructure of HMGECs During Culture in Hanging Drop

Hanging drop cultures displayed small lipid droplets localized in close proximity to the cell nucleus (Figs. 7A, 7B). Cells were connected to each other by desmosomes in some places (Fig. 7B). Except for these connection sites the cells were surrounded by small intercellular spaces (Figs. 7B, 7C). After 48 hours of hanging drop culture, cells accumulated more lipid droplets in the cytoplasm (Figs. 7D, 7E). Besides desmosomes, also the number of CK filaments increased in the cells (Figs. 7D–F). Overall, the cells exhibited the morphology of differentiating meibocytes after hanging drop culture. 

Figure 7

 

HMGECs during culture in hanging drop. HMGECs were cultivated in hanging drops for up to 48 hours in proliferation medium. Hanging drop cultures were analyzed using transmission electron microscopy. (A) Section through a meibocyte that had been cultured for 16 hours. Close to the cell nucleus some small lipid droplets (stars) have formed. (B) Section through meibocytes that had been cultured for 16 hours. Small lipid droplets (stars) are visible in close proximity to the cell nucleus. (C) Area between two meibocytes that had been cultured for 16 hours showing two desmosomes (arrows) bridging the intercellular space. (D) Section through a meibocyte that had been cultured for 48 hours. Several lipid droplets are visible within the cytoplasm. Next to the droplets, cytokeratin filaments are visible in the cytoplasm. (E) Meibocyte after 48 hours cultivation time. The cytoplasm contains a lot of small lipid vesicles (stars) beside other vesicles and cytokeratin filaments in the cytoplasm. (F) Contact area of two meibocytes that had been cultured for 48 hours. The cells are connected via desmosomes. Cytokeratin filaments (arrowheads) are increased in this area. c, cytoplasm; ics, intercellular space; ld, lipid droplets.

Discussion

Since 3D culture is expected to bridge the gap between monolayer cell culture and living tissues, a wide variety of 3D culture methods have been developed to mimic the physiological microenvironment.³² In this study, we firstly tried to establish a membrane-based 3D culture system with reference to previous reports on immortalized sebocytes (SB662) spontaneously differentiating into a sebaceous-like phenotype²⁷ and primary rat meibocytes being successfully cultured three-dimensionally when cultivated as planar sheets.²⁸ Similar to the findings derived from sebocytes and primary meibocytes, our results demonstrated that HMGECs are capable of yielding multiple cell layers, if subjected to air-interface culture with proliferation medium. The so cultured HMGECs did not grow evenly on the membrane, but formed mountain-like structures, meaning that cell layer formation was pronounced at distinct locations and reduced in the surrounding areas. Moreover, the basal cells did not seem to develop a strong contact with any type of membrane, probably due to lack of cell adhesion. This finding intrigued us as recently it was shown that maturing meibocytes are connected by desmosomes and HMGECs increase cell contacts when treated with serum.^15,33 Apparently, the basally located HMGECs in direct contact with the membranes were unable to develop strong contact points comparable to desmosomes or hemidesmosomes. Interestingly, in sebocytes, extracellular matrix precoating was not necessary to form a stable multilayer.²⁷ Moreover, no detachment has been described in the course of the further procedure for histologic examination (Barroult C, oral communication, ARVO 2014). As previously stated, the most probable explanation is that HMGECs are unable to form cell contacts with the membranes tested but form cell contacts between each other. 

As HMGEC cultured on ThinCert did not yield convincing results, we further tested Alvetex scaffolds allowing cells to proliferate three-dimensionally, thereby communicating more dynamically,³⁴ to differentiate in response to distinct stimuli^35,36 and to function in an appropriate microenvironment.^37,38 Histologic analysis demonstrated that HMGECs exhibited the same appearance regardless of their location within Alvetex. We expected that air–liquid interface culture encouraged the epithelial stratification upon Alvetex; however, HMGECs did not form multiple cell layers, but rather differentiated into sphere-shaped colonies on the apical surface as already observed in membrane-based 3D air-lift culture. It is tempting to speculate that spheroid 3D culture is better suited for the investigation of the entire differentiation process as already indicated using rat primary meibocytes.²⁸ Moreover, a high-throughput 3D human sebocyte cell model was recently developed using adult primary sebocytes and an immortalized sebocyte cell line.²⁵ The authors demonstrated that 3D spheroid culture promotes a higher expression of terminal differentiation markers and genes involved in lipid synthesis in sebocytes. It is likely that this also holds true for primary meibocytes and HMGECs. However, this requires further research in the future. 

In one of our previous studies, CK1 was highly expressed in skin epidermis of human eyelid as a positive control.²⁹ However, it remained undetectable in Alvetex-cultured HMGECs, even after long-term air-lift culture. Since CK1 is one of the best-known markers of fully keratinized epithelium and detected in the orifices of meibomian duct, this suggests that air–liquid interface culture alone does not lead to keratinization in HMGECs. CK10 is generally associated with CK1 as the coexpressed partner of CK1 and has been described in suprabasal epidermal epithelial cells.³⁹ Recent studies have shown that CK10 is upregulated by air-lift culture in conjunctiva⁴⁰ and corneal limbus⁴¹ but remains absent from peripheral cornea,⁴¹ sebaceous gland,⁴² or acini of meibomian glands.⁴³ In 2016, Call et al.⁴⁴ demonstrated that CK10 and CK4 are colocalized within the central duct near the ductile openings in fully developed meibomian glands of mice. Accordingly, CK10 was detected only in the ductal epithelium of human meibomian gland, but not in the meibocytes.²⁹ Hence, several bodies of evidence indicate that CK10 has the potential to serve as a biomarker to identify human ductal epithelial cells and to determine their differentiation status. In this study CK10 expression was mildly inducible on the apical surface in response to air exposure. This suggests that air exposure to HMGECs induces a state similar to that of epithelial cells of the meibomian excretory duct. This is also in line with previous findings demonstrating that HMGECs can differentiate to meibomian gland excretory duct lining cells.¹⁵ CK14 is generally found in the basal cell layer of stratified epithelium³⁹ as well as in acini and ductal epithelium of human meibomian glands.²⁹ Aside from some cytoprotective functions of CK14, very little is known about CK14-related functional mechanisms, since the homozygous CK14 knockout mice show postnatal lethality of 100%.⁴⁵ Nevertheless, it is presumed that CK14 expression is tightly regulated by the differentiation program in stratified epithelium. When epithelial cells move upward and differentiate on their way to the apical surface of a stratified epithelium, CK14 levels diminish gradually and a new pair of CKs is induced.⁴⁶ Based on this, we expected that CK14 would decrease during cell differentiation or at least be confined to the basal layer after air–liquid interface culture. However, CK14 immunoreactivity strongly increased over time in Alvetex. Maybe this is due to the lack of a proper stratification in 3D-cultured HMGECs. Another possible reason could be the lack of mechanical stress and the necessity to actively maintain the cell shape in the scaffold. Nevertheless, additional work is required to explore the precise mechanisms and to confirm this finding. 

CK6, a marker for noncornified epithelium, revealed a clear positive reactivity in our 3D-cultured HMGECs. This finding is consistent with previous reports describing CK6 in epithelial duct cells of meibomian glands.^13,47 In this study, intensive CK6 expression increased over time. It is therefore possible to speculate that culture conditions mimicked the microenvironment around meibomian gland ducts and hence elicited differentiation toward keratinization at least in meibocytes at the air–liquid interface as already discussed for CK10. 

Hampel et al.¹⁵ demonstrated that serum-induced differentiation leads to keratinization. In long-term serum treatment, CK1 and CK6 expression showed a time-dependent increase on the standard 2D format. In our 3D culture system, 1-day serum treatment did not alter the overall expression levels of CKs. Instead, treatment with differentiation medium significantly changed cell morphology on the apical side of the Alvetex scaffold leading to the formation of a sheet-like, stretched cell layer, resembling ductal epithelial cells. In addition, intensive CK6 expression was detected inside the layer, but no CK1 expression was induced. The notable differences between 2D format and 3D culture in CK1 expression and cell appearances could raise the important hypothesis that HMGECs may function as ductal epithelial cells rather than acinar cells as long as they are cultured in a 2D format regardless of serum treatment. Liu et al.²⁰ reported that HMGECs exhibited lower cell proliferation and keratinization when cultured with serum-containing medium.²⁰ In particular, CK6 gene expression was significantly downregulated when cultured with proliferation medium. Taken together, these unique morphologic cell changes may be triggered by specific CK filament reorganization. Interestingly, this was found only on the apical surface of the Alvetex scaffold, suggesting that air exposure boosts morphologic differentiation. Reduced spatial restriction on the apical surface of the scaffold may represent an alternative inductor for HMGEC transformation. In contrast, HMGECs inside the scaffold may be granted or forced to maintain a more natural round shape even after differentiation stimuli. Intriguingly, the cells underwent keratinization processes during hanging drop culture as when cultured two-dimensionally with serum. 

Lipid secretion from the meibomian glands is vitally important for tear film stability.⁴⁸ To exert this function, meibocytes undergo several differentiation steps until they become hypermature and ultimately die by apoptosis to become meibomian oil. During the differentiation process meibocytes accumulate more and more secretory vesicles and subsequently fill up with lipid droplets. In our 3D-cultured HMGECs, lipid accumulation was increased over time, which could be advantageous for retaining steady state for a longer period of time, compared with the 2D culture setting, where lipid production reached a peak at 1 day when treated with serum.¹⁵ However, cellular shape and size did not change and HMGECs only reached a state that is best described as “differentiating.” Further differentiation states such as “mature” or “hypermature” were not reached. This became especially obvious by TEM investigation of 3D-cultured HMGECs in Alvetex scaffolds or hanging drops. Here, the ultrastructural analysis also revealed lipid droplets that are characteristic of “differentiating” meibocytes but not for mature or hypermature cells. Our results support previous reports of HMGECs being unable to differentiate to an advanced state.^15,23 This is most probably due to the fact that the cells are derived from an immortalized cell line that is subjected to a cell cycle program very different from the physiological situation. Since meibocyte apoptosis is considered a key factor involved in the mechanisms of hypermaturation and becoming meibum, the immortalization itself may represent an inherent obstacle for those cells on their way toward an advanced differentiation status. Another possibility is that HMGECs used in this study might have originated from wrong or mixed progenitors for acinar differentiation. It was recently proposed that renewal of meibocyte could rely on a limited number of progenitor meibocytes, whereas meibomian gland duct would be derived from multiple origins.^49,50 This can explain why HMGECs induced CK expression in a pattern closely resembling that of ductal cells in response to an air-rich environment. Even if appropriate progenitor populations are found in HMGECs, they may have been exhausted because HMGECs were established from donors ranging from 35 to 85 years. 

Recently, Mauris et al.⁵¹ demonstrated that CD147 regulates meibocyte proliferation and lipid biosynthesis along with matrix metalloproteinase-9 (MMP-9) upregulation. Gidfar et al.⁵² showed that Notch signaling plays a critical role for meibocyte homeostasis. The holocrine turnover was investigated in mouse meibomian glands by tracking several different CK progenitors and reconstructing them three-dimensionally using a novel technique called immunofluorescent computed tomography.⁵³ To our knowledge, there have been no previous reports of establishment of 3D cell culture using human immortalized meibocytes; however, molecular biomarkers as described above should be incorporated into future 3D studies to ensure utility. 

In conclusion, our findings indicate that HMGECs have the capability to differentiate in a long-term 3D culture even without serum treatment, especially when exposed to an air–liquid interface, but only reach a state of “differentiating meibocytes.” The present study is designed to contribute to the optimization of an in vitro 3D culture model suitable for basic research on meibomian glands. 

Acknowledgments

The authors thank Gerti Link, Elke Kretschmar, and Hong Nguyen for their expert technical assistance in the fields of electron microscopy and immunohistochemistry, and Marco Gößwein for the photographic work. 

Supported by Deutsche Forschungsgemeinschaft (DFG) Grant PA 738/9-2. 

Disclosure: N. Asano, None; U. Hampel, None; F. Garreis, None; A. Schröder, None; M. Schicht, None; C.M. Hammer, None; F. Paulsen, None