Enhancement of lacrimal gland cell function by decellularized lacrimal gland derived hydrogel

Sustainable treatment of aqueous deficient dry eye (ADDE) represents an unmet medical need and therefore requires new curative and regenerative approaches based on appropriate in vitro models. Tissue specific hydrogels retain the individual biochemical composition of the extracellular matrix and thus promote the inherent cell´s physiological function. Hence, we created a decellularized lacrimal gland (LG) hydrogel (dLG-HG) meeting the requirements for a bioink as the basis of a LG model with potential for in vitro ADDE studies. Varying hydrolysis durations were compared to obtain dLG-HG with best possible physical and ultrastructural properties while preserving the original biochemical composition. A particular focus was placed on dLG-HG´s impact on viability and functionality of LG associated cell types with relevance for a future in vitro model in comparison to the unspecific single component hydrogel collagen type-I (Col) and the common cell culture substrate Matrigel. Proliferation of LG epithelial cells (EpC), LG mesenchymal stem cells, and endothelial cells cultured on dLG-HG was enhanced compared to culture on Matrigel. Most importantly with respect to a functional in vitro model, the secretion capacity of EpC cultured on dLG-HG was higher than that of EpC cultured on Col or Matrigel. In addition to these promising cell related properties, a rapid matrix metalloproteinase-dependent biodegradation was observed, which on the one hand suggests a lively cell–matrix interaction, but on the other hand limits the cultivation period. Concluding, dLG-HG possesses decisive properties for the tissue engineering of a LG in vitro model such as cytocompatibility and promotion of secretion, making it superior to unspecific cell culture substrates. However, deceleration of biodegradation should be addressed in future experiments.


Introduction
Dry eye disease affects 11%-22% of the world population, with 17% of patients developing aqueous deficient dry eye (ADDE) [1,2].This is usually the result of loss of lacrimal gland (LG) function due to trauma, radiation therapy, autoimmune diseases like Sjögren's Syndrome, or simply but usually less severely by aging [3][4][5].Resulting LG insufficiency can lead to severe ADDE accompanied by irritation, foreign body sensation, corneal defects, and even blindness, therefore massively impairing patients' quality of life [6].
To date, treatment options have mainly focused on palliative methods comprising substitute lubricants in the form of eye drops or ointments.So far, curative or regenerative strategies have only been evaluated sporadically.For instance, regenerative cell therapy using mesenchymal stem cells (MSCs) has been successfully applied in a mouse model and a clinical trial [7,8].LG substitution by autologous salivary gland transplantation was also reported but is limited by several aspects as saliva differs from tears in composition and osmolarity and, in case of autoimmune involvement, salivary gland function is often also impaired [9,10].Thus, there is a need for substantial treatment approaches aiming for regeneration of damaged LG tissue and sustainable relief for patients.
The availability of reproducible in vivo 3D model platforms made from tissue-specific cells could contribute to understanding the diverse pathomechanisms of ADDE and enable an ethically uncritical, animal-free, broad-based testing of cell therapeutics and pharmaceuticals.3D model platforms have already been created for many tissues such as lung, kidney, heart, and liver (reviewed in [11]).In most cases, tissue-specific cell types are combined with a three-dimensional synthetic or organic scaffold and cultured under flow perfusion conditions to imitate the in vivo situation and to promote native cell function and morphology.Hydrogels (HGs), composed of the decellularized extracellular matrix (ECM) of the original organ, can serve as a scaffold material.Thanks to the native ECM composition and structure, these hydrogels ideally reproduce the physiological niche of the tissue cells and can promote their differentiation, morphology, and function more effectively than synthetic substitutes [reviewed in [12]].Tissue-derived ECM hydrogels have already been established for many organs such as the cornea [13], pancreas [14], and salivary gland [15] and their cytocompatibility has been established.
Spheroids consisting of several tissue-specific cells representing functional units already provide an organ-like 3D architecture that is superior to the application of a single cell type [16,17] and have been implemented for a broad range of organs [18][19][20], including the LG [21][22][23].LG-spheroids consisting of LG epithelial cells (EpC), LG-MSC, and dermal microvascular endothelial cells were able to recapitulate an acinar structure and were secretioncompetent upon parasympathetic stimulation [23].However, in these experiments spheroid assembly was achieved by combining cells with Matrigel (MG), a commonly used cell culture substrate of cancerous origin, which is, therefore, inappropriate for clinical use [24].Moreover, these spheroids are too large for bioprinting applications and the precise deposition within a complex structure.
Another approach considered for patients with severely damaged LG tissue to which regenerative strategies are not applicable, is the tissue engineering of a transplantable LG analogue.However, reconstruction of a natural 3D tissue, just like the generation of an exact in vitro model, remains a challenge to date, as the macro-and microscopical architecture, alignment of cell-cell and cell-ECM polarity, immunological adaptation, innervation, vascularization, and many other aspects of the natural LG must be implemented.In vitro reconstructed tissue should be able to adapt to the new in vivo environment and continue to support differentiation, proliferation, and functional organization of organ-specific cells.The 3D-bioprinting allows spatially controlled deposition of cells and scaffold materials, called bioinks, thereby enabling the manufacturing of tissue-engineered whole organs [25].Decellularized ECM derived hydrogels of different tissues such as heart [26,27], lung [28], pancreas [29], and cornea [13,30] have been used as bioinks and successfully characterized in the context of 3D bioprinting and cell cultivation.
Here we established a decellularized LG-based hydrogel (dLG-HG) and characterized its composition, gelling behavior, and impact on cell viability and function.Considering that ECM digestion duration immensely influences hydrogel characteristics [28], we compared varying hydrolysis times to achieve the optimal properties regarding a bioink for the generation of a LG in vitro model.

LG extraction
Fresh LG from 8 month-old domestic pigs were obtained from a local abattoir.All experiments were conducted in accordance with the Association for Research and Vision in Ophthalmology statement for the use of animals in ophthalmic research.

LG decellularization
LG were decellularized as described previously [31].In brief, LG were cut into pieces of 3 mm diameter and washed in cold phosphate-buffered saline solution (PBS; Sigma-Aldrich, St. Louis, MO, USA) containing 5% penicillin/streptomycin (P/S; Sigma-Aldrich) overnight (o/n).Cellular components were removed by incubation in 1% (w/v) sodium deoxycholate monohydrate solution (Sigma-Aldrich) for 36 h with three changes, followed by DNase solution (200 U ml −1 in PBS; Roche, Penzberg, Germany) for 24 h, and washed in PBS + 5% P/S.The dap-dry dLG was stored at −80 • C until further use.All incubation steps were performed at 4 • C under continuous agitation with interposed washing in PBS + 5% P/S.

Soluble collagen assay, sulfated glycosaminoglycan assay, hyaluronan assay, and dry weight
For comparison, lyophilized native LG tissue was completely dissolved in 0.01 M HCl containing 4000 U ml −1 pepsin at 10 mg ml −1 by blending (TissueRuptor II; Qiagen, Hilden, Germany) and digested for 4 h at room temperature.Collagen type I-V content was determined by a colorimetric assay (sircol soluble collagen assay; biocolor, Carrickfergus, UK).dLG-HG pregels and native tissue digest were diluted 1:10 in 0.01 mM HCl and 10 µl per sample used to conduct the assay according to manufacturer´s instructions.Sulfated glycosaminoglycan content was determined by Blyscan sGAG assay (biocolor) applying 50 µl samples of dLG-HG pregels and native tissue digest according to man-ufacturer´s instructions.For hyaluronic acid quantification dLG-HG pregels and native tissue digest were lyophilized, and 10 mg samples were employed according to manufacturer´s instructions (Purple-Jelley Hyaluronan Assay, biocolor).For dry weight estimation, 1 ml dLG-HG pregels were dispensed, lyophilized, and pellets weighed.

Mass spectrometry
For mass spectrometric analysis, triplicate samples of dLG-HG pregels after hydrolysis times from 4 to 96 h and native LG lysate (10 mg ml −1 ) were separated by a short 4%-12% polyacrylamide gel.After Coomassie brilliant blue staining, the protein containing bands were excised and processed as described elsewhere [32].Briefly, bands were destained, washed, reduced with dithiothreitol, alkylated with iodoacetamide and digested with trypsin (Serva, Heidelberg, Germany) in 50 mM NH 4 HCO 3 overnight at 37 • C. Tryptic peptides were extracted with 0.1% trifluoroacetic acid and subjected to MS-coupled liquid chromatography.For peptide separation over a 120 min LCgradient with 300 nl min −1 an Ultimate 3000 Rapid Separation liquid chromatography system (Thermo Scientific, Bremen, Germany) equipped with an Acclaim PepMap 100 C18 column (75 µm inner diameter, 25 cm length, 2 µm particle size, Thermo Scientific) was used.MS analysis was carried out on an Orbitrap Lumos mass spectrometer (Thermo Scientific) with FAIMS operating in positive mode and equipped with a nano electrospray ionization source.FAIMS was operated with a carrier gas flow of 4.5 l min −1 and the compensation voltages were set to −40, −55 and −70 V. Capillary temperature was set to 250 • C and source voltage to 1.5 kV.Survey scans were carried out over a mass range from 200-2000 m/z at a resolution of 120 000.The target value for the automatic gain control was 250 000 and the maximum fill time 60 ms.Within a cycle time of 2 s the most intense peptide ions (excluding singly charged ions) were selected for fragmentation.Peptide fragments were analyzed in the ion trap using a maximal fill time of 50 ms and automatic gain control target value of 10 000 operating in rapid mode.Already fragmented ions were excluded for fragmentation for 60 s.
Acquired spectra were searched using Sequest HT within Proteome Discoverer version 2.4.1.15against the UniProt database (sus scrofa, 291 231 sequences) applying a precursor mass tolerance of 10 ppm and a mass tolerance of 0.6 Da for fragment spectra applying a Protein FDR ⩽ 0.01.Methionine oxidation was considered as variable modification, carbamidomethylation as static modification as well as tryptic cleavage specificity with a maximum of two missed cleavage sites.
Proteins with less than 4 hits across all analyzed samples were excluded.Remaining proteins were assigned to cellular compartment clusters by UniProt ID mapping within the UniProKB database.Extracellular and basement membrane clusters were visualized as heat maps with GraphPad Prism 9 (GraphPad Software Inc., San Diego, CA, USA).

Rheology
Rheological characterization was performed with an oscillatory rheometer (Kinexus PRO; Malvern Panalytical, Malvern, UK) using a plate-cone geometry (diameter: 20 mm, angle: 1 • ).dLG-HG pregels and Col were neutralized immediately prior to measurement.Ice-cold 40 µl samples were placed and a temperature ramp was applied from 10 • C to 42 • C (heating rate: 5 K min −1 ) with constant oscillation frequency and shear strain (1 Hz, 1%).Storage and loss modulus were recorded every two seconds and plotted against temperature.The maximum storage modulus value (G ′ max ) was extracted for every sample as a measure for the maximally achieved gel stiffness.The gelation point of each sample was defined as the temperature at which the storage modulus value exceeded the loss modulus value for the fifth time in successive measurement points.

SEM imaging and quantification of porosity, fiber interconnectivity and diameter
For scanning electron microscopy of dLG-HG and Col, 100 µl gelled samples were fixed for 2 h in 2.5% glutaraldehyde (Serva)/4% paraformaldehyde (Merck, Darmstadt, Germany) in 0.1 M cacodylat buffer pH 7.4 (Serva) and dehydrated in an ascending acetone series from 50% to 100%.Samples were then critical point dried (Leica EM CPD030; Leica Biosystems, Wetzlar, Germany) and chrome sputtered (Q150T ES; Quorum Technologies, Laughton, East Sussex, UK).Six representative images of each sample were taken at 5800x magnification with constant contrast setting and exposure time (Zeiss Crossbeam 550; Carl Zeiss Microscopy, Munich, Germany) and analyzed according to [28].
For porosity estimation, images were converted into binary using ImageJ [33] and the proportion of black pixels was calculated.Fiber diameter and number of interconnections were determined in the upper fiber layer.Underlying layers were excluded by threshold.

Isolation and cultivation of LG associated cells
LG-EpC were isolated using the explant culture technique as previously described [31].Briefly, porcine LG were chopped into small pieces and placed onto a 3T3 fibroblast feeder layer (CCL-92; ATCC, Manassas, VA, USA) which has been growth inhibited with 0.01 mM mitomycin C (Sigma-Aldrich).
LG-EpC were used in passage 2 for all experiments.

Viability assay
Cells were seeded at a density of 5 × 10 4 (LG-EpC and HUVEC) or 2.5 × 10 4 (LG-MSC) cells per well in 96-well plates coated either with 100 µl dLG-HG, Col, or MG (Corning, Corning, NY, USA; diluted to 5 mg ml −1 with serum-free DMEM) and cultured for 72 h in respective culture medium with a medium change after 48 h.Medium was replaced with 100 µl 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma-Aldrich; 0.5 mg ml −1 in respective culture medium) followed by incubation for 4 h.The MTT solution was removed and 200 µl isopropanol were added.After o/n incubation under panning at 150 rpm (Mini Shaker, VWR, Radnor, PA, USA) to fully resolve the formed formazan, 150 µl were transferred to a fresh well plate and the absorption at 570 nm was determined (Victor X4 Multilabel Reader, PerkinElmer, Solingen, Germany).Viability was measured in duplicates of six biological replicates.In case of HUVEC, independently cultured cells were used as biological replicates.

β-hexosaminidase activity assay
Quantification of β-hexosaminidase activity was performed as described previously [31].LG-EpC were seeded at a density of 3 × 10 5 cells per well in 48-well plates, coated either with 200 µl dLG-HG, Col, or MG and overlaid with epithelial cell culture medium.After 72 h cells were washed with serum-free DMEM and incubated with 300 µl serum-free DMEM for 2 h (baseline value).Carbachol (Sigma-Aldrich) was added for 30 min at a final concentration of 100 mM, and a stimulated sample was removed.For measurement of β-hexosaminidase activity, 4-methylumbelliferyl N-acetyl-b-D-glucosaminide (Sigma-Aldrich) was used as a substrate.The fluorescence intensity was determined at 360 nm excitation and 450 nm emission (Victor X4 Multilabel Reader).EpC including substrates were fixed in 4% formaldehyde (Roti Histofix, Carl Roth, Kalrsruhe, Germany) for 30 min at RT and paraffin embedded.β-hexosaminidase activity was accessed in duplicates of six biological replicates.

Histology and immunohistochemistry
For documentation of the macroscopic morphology of LG-EpC, LG-MSC, and HUVEC growing on dLG-HG, Col, or MG, photographs were taken of the formazan-dyed cells after MTT incubation prior to dissolution with isopropanol.
For haematoxylin and eosin (H/E) staining, LG-EpC paraffin sections were deparaffinized, rehydrated in an ascending EtOH series, and stained in H/E solutions (Carl Roth) for 5 min each.After dehydration, sections were embedded in mounting medium (Roti-Histokit 2; Carl Roth) and imaged (200x magnification; Leica DM4000 B and camera system Leica DFC450 C).Epithelial thickness was quantified in 5 images per sample (n = 6) at 3 sites per image defined by a template using ImageJ.
LG-MSC growing on top of dLG-HG, Col, MG, or cell culture plastic (2.5 × 10 4 cells well −1 , 96-well plate) were fixed after 3 d in culture and used for immunofluorescence wholemount staining.Samples were permeabilized and blocked in 1% v/v triton x-110/2.5% v/v donkey serum/PBS for 30 min at RT. Anti-nestin (rabbit polyclonal, 1:100, ABIN388764; antibodies-online) was applied in 2% v/v donkey serum/PBS o/N at 4 • C. Alexa Fluor 488 AffiniPure donkey anti-rabbit IgG was diluted 1:500 in PBS including 1 µg ml −1 Hoechst 33 342 (Thermo Scientific) as nuclear staining and applied o/N at 4 • C. Image acquisition and processing was performed with a confocal laser scanning microscope (100x, SP8) and LAS X software.

Tube formation assay
GFP expressing HUVEC were seeded in 24-well plates (6 × 10 4 cells well −1 ) coated with dLG-HG, Col, or MG (300 µl well −1 ) in six replicates of independently expanded cell populations and cultured for 24 h.Tube formation was documented with a fluorescent microscope (DM 4000B) and the number of formed lumina per area was determined in 6 images each.

Hydrogel degradation upon matrix metalloproteinase (MMP)-inhibitor treatment
LG-EpC were seeded at 5 × 10 4 in 96-well plates coated with 100 µl dLG-HG, Col, or MG (5 mg ml −1 ) and overlaid with epithelial cell culture medium supplemented with either 20 µM of the MMP-inhibitor GM6001 (dissolved in dimethyl sulfoxide (DMSO); both Sigma-Aldrich) or an equivalent of DMSO.Photographs of the plate bottom side were taken on days 3 and 14 in culture.Gel-covered well area was quantified in duplicates of 6 biological replicates using the ImageJ free hand tool.To do so, the remaining hydrogel area and the respective well area were encircled and measured in square pixels and the percentage of hydrogel area per well area was calculated.For statistical analysis, two-way analysis of variance (ANOVA) with Tukey post hoc test was carried out for statistical analysis using GraphPad Prism 9.

Statistical analysis
The data are presented as mean values ± standard deviation.If not stated otherwise, one-way ANOVA with Tukey post hoc test was carried out for statistical analysis using GraphPad Prism 9. Statistical significance was declared when the p value was less than 0.05.

dLG hydrolysis for 48 h was most efficient
dLG-HG was generated by pepsin mediated hydrolysis of LG-ECM for 4, 12, 24, 48, and 96 h to define the optimal time for LG-ECM hydrolysis (figure 1(A)).The time dependent pepsin hydrolysis efficiency of LG-ECM was evaluated by measuring the dry weight of solubilized ECM, which increased from 7.95 ± 0.19 mg ml −1 after 4 h to 9.55 ± 0.10 mg ml −1 after 48 h (p < 0.0001) with no further benefit after 96 h of hydrolysis (figure 1(B)).
Size distribution of hydrolyzed proteins was evaluated by SDS-PAGE and subsequent Coomassie staining and compared to pure Col (figure 1(C)).Col demonstrated bands corresponding to its monomeric α1 and α2 chains, β12 or β11 dimers and γ112 trimers [35].Equal bands were also present in dLG-HG samples, corresponding to monomeric, dimeric, or trimeric molecules of putatively several collagen types.Besides, pepsin appears as a double band with approximately 41 kD, prominent in dLG HG-4 h and −12 h, hardly visible in dLG-HG-24 h, and no longer detectable in dLG-HG ⩾ 48 h.Furthermore, bands corresponding to overdigested collagen molecules and/or additional components contained in LG-ECM or LG basement membrane are present in dLG-HG samples and absent in Col.

(F)).
Mass spectrometry was performed to clarify the detailed protein composition of dLG-HG in comparison to native LG tissue.Among the extracellular components, the alpha 1 and alpha 2 subunits of collagen type I were by far the most abundant proteins in all samples (figure 2(A)).Besides, collagen types III, V, VI, VII, and XXI were found in native LG as well as in all dLG-HG variants (figure 2(B)), all in accordance with the results of SDS-PAGE.Additionally, the glycoproteins proteoglycan 3, c-type lectin, thrombospondin 1, tenascin, and heparan sulfate proteoglycan 2, as well as the secreted components lipocalin, lactotransferrin, lysozyme, BPI fold-containing family A member 1, and mucin 5B were identified within the extracellular cluster.Among basement membrane proteins, laminin gamma 1 and laminin beta 2 were the most abundant proteins in all samples (figure 2(C)).In addition, several further laminin subunits as well as fibronectin and collagen type IV alpha 2 were identified besides others (figure 2(D)).
In some cases (lactotransferrin, complement C3, collagen type VI alpha-1,inter-alpha-trypsin inhibitor heavy chain, cathepsin D, galectin, annexin, and laminin gamma 2) protein abundance was below detection level in dLG-HG-4 h/12 h suggesting yet incomplete ECM hydrolysis.Collectively, most efficient LG-ECM hydrolysis was obtained after 48 h of pepsin digestion.

Hydrolysis for 48 h resulted in the densest fiber network
After all applied hydrolysis times between 4 and 96 h, gelation competent hydrogels were achieved as estimated by visual evaluation.The gelation kinetics of dLG-HG was compared to Col by oscillatory rheometry whereby a temperature ramp was applied, and storage and loss modulus were recorded (figure 3(A)).The maximal storage modulus (G ′ max ) as a measure for maximal stiffness did not differ between dLG-HG samples (195 ± 46.79 Pa to 227.5 ± 54.24 Pa; p ⩾ 0.80) but was lower for Col (86.83 ± 39.01 Pa; p ⩽ 0.0035) (figure 3(B)).The temperature when storage modulus initially exceeded loss modulus, defined as the gelation point, similarly was equal for all dLG-HG samples (32.06 ± 0.92 • C to 33.19 ± 0.71 • C; p ⩾ 0.59), but was lower for Col (28.31 ± 2.39; p < 0.0001) (figure 3(C)).In the context of the applied measurement setting, a lower temperature corresponded to an earlier time point, meaning that the gelation of dLG-HG samples occurred slower than that of Col.
The ultrastructure of dLG-HG and Col was visualized by electron microscopy and images were used to quantify porosity, fiber diameter and fiber interconnection density (figures 3(D)-(G)).Besides fibrillary structures, dLG-HG samples additionally comprised spherical structures that were absent in Col probes reflecting further components of the LG-ECM (figure 3(D)).Col showed highest porosity (69.11 ± 7.15%), whereas dLG-HG-48 h demonstrated the lowest porosity (61.91 ± 2.33%; p = 0.046 vs. Col), what correlated to the highest fiber density (figure 3(E)).Accordingly, the number of interconnections per square micrometer was highest for dLG-HG-48 h (2.45 ± 0.845 µm-2; p = 0.008 vs. dLG-HG-4 h) (figure 3(F)).Fiber diameter also increased with longer hydrolysis from 66.68 ± 2.37 nm after 4 h to 74.43 ± 1.98 nm after 48 h (p = 0.0006) and decreased again after further hydrolysis to 64.24 ± 3.15 nm after 96 h (p < 0.0001).Col had the smallest fiber diameter (61.96 ± 1.73; p < 0.0001 vs. dLG-HG ⩽ 48 h).Col fibers were rather equally thick, whereas dLG-HG samples demonstrated more heterogeneous diameter distribution (figures 3(D) and (G)).In summary, rheological properties of dLG-HG samples did not differ with hydrolysis time, but with regard to hydrogel (ultra)structure, 48 h hydrolysis led to the densest fiber network with thickest fibers and most interconnections.

LG-EpC viability was enhanced by dLG-HG ⩾ 24 h
Regarding a future LG in vitro model comprising different LG-associated cell types, we analyzed cell behavior of LG-EpC, LG-MSC, and HUVEC, which are crucial for tear secretion, regeneration and immunomodulation, and nutrient and oxygen supply, respectively.Cell behavior on dLG-HG was compared to Col as a tissue-unspecific single component hydrogel and MG.
Macroscopically, cell growth was confluent on Col, whereas LG-EpC contracted to spherical aggregates on MG.On dLG-HG, the whole cell lawn including the gel substrate aggregated to the center of the  well with reticular structures on dLG-HG ⩾ 24 h (figure S1).
LG-MSC viability was highest on Col (0.15 ± 0.014 A 570nm ) and lowest on MG (0.11 ± 0.0081 A 570nm ; p < 0.0001 vs. Col) (figure 5(A)).LG-MSC viability on dLG-HG increased with longer hydrolysis times and was higher   on dLG-HG ⩾ 48 h than on MG (48 h: 0.13 ± 0.031 A 570nm ; p = 0.02; 96 h: 0.14 ± 0.017 A 570nm ; p = 0.0053).Comparable to LG-EpC, LG-MSC grew macroscopically confluent on Col and formed small aggregates on MG.The dLG-HG substrate contracted while overlying LG-MCS grew rather aggregated than as a lawn (figure S1).Immunocytochemical staining of Nestin, a marker distinguishing MSC from fibroblasts, revealed that LG-MSC formed cell clusters on all hydrogel substrates including Col and MG with most dense aggregates on MG, in contrast to confluent growth on cell culture plastic (figure 5(B)).
To evaluate angiogenic stimulation potential of dLG-HG, a tube formation assay was performed and the number of formed lumina was quantified.HUVEC formed typical filigree reticulated structures on MG, so that hereon most lumina were counted (23.94 ± 3.01 mm −2 ) (figures 6(B), (C) and S1).On dLG-HG, tube formation was less pronounced than on MG, but more distinct than on Col (7.41 ± 2.46 mm −2 ).Besides on MG, second most lumina were formed on dLG HG 96 h (14.96 ± 2.21 mm −2 ; p < 0.0001 vs. Col).

LG-EpC grown on dLG-HG expressed basal membrane proteins
Laminin and fibronectin, which are localized at the basal membrane of the LG acini, were expressed by LG-EpC cultured on dLG-HG and Col and less on MG (figure 7(B)).However, an oriented disposition directed to the cell-gel interface was not evident after 3 d in culture.The EpC markers pan-Cytokeratin and Aquaporin-5 are expressed by LG-EpC in the acini and ducts of the LG and were also present in cultured LG-EpC on all substrates proving their identity (figure 7(C)).However, a directed Aquaporin-5 expression to the apical cell part was not seen at this time in culture.
Col area also reduced to 51.78 ± 11.20% after day 14.On MG, no area reduction was measurable within the observed period.
LG-MSC and HUVEC also degraded dLG-HG and CoI, however, the degradation by HUVEC was less pronounced (data not shown).His degradation was not due to dissolution, but cell dependent, as no area reduction of neither hydrogel occurred within two months under the same conditions without seeded cells (data not shown).
We further excluded cell dependent pH-shifts which could have led to acidic collagen dissociation (data not shown).Next, we added the broadband MMP inhibitor GM6001 to the culture medium which led to significantly decelerated area reduction of dLG-HG with 4-48 h hydrolysis after 3 d (p ⩽ 0.038) and of all dLG-HG samples after 14 d (p ⩽ 0.0004), so that terminally 29.92 ± 4.65% (4 h) to 44.38 ± 3.59% (24 h) of the dLG-HG area were still present.Col area reduction, in contrast, was not prevented by MMP inhibition.

Discussion
To date, ADDE is primarily treated symptomatically using lubricants or punctal occluders, thus remaining without lasting relief for the patient.The development of curative or regenerative therapies can be driven forward by appropriate platforms to explore disease mechanisms and drug impact, which approximate physiological conditions and thereby minimize animal experiments (reviewed in [11]).Hydrogels are the ideal scaffold for multicellular model platforms as they feature tunable morphology, stiffness, and composition to match the requirements for each individual organ or tissue (reviewed in [36]).A variety of synthetic and organic hydrogels is available, each of which having individual advantages and drawbacks.While synthetic polymers like polyethylene glycol, polyhydroxy acid, and pluronic acid feature easily controllable rheology and stiffness, they lack cellular interaction and bioactivity (reviewed in [37]).Hydrogels consisting of single ECM components including collagen [38], gelatin [39], and hyaluronic acid [40] are well recognized by cells and can additionally be chemically modified to overcome rheological limitations.However, they still lack the unique mixture of structural and functional proteins provided by the native tissue ECM (reviewed in [41]).
Aiming to construct a physiological LG analogue in the long-term, our goal was to establish a hydrogel derived of LG-specific ECM reflecting the native niche of LG cells, thereby supporting their physiological function better than alternative unspecific materials.Our decellularization and hydrolysis method successfully preserved the complex native LG-ECM composition in dLG-HG.Mass spectrometry analysis revealed that laminin is the predominant basement membrane protein in the porcine LG, supported by [31], with laminin gamma 1, laminin beta 2, laminin alpha 5, and laminin alpha 1 being the most expressed subunits.Therefore, it can be inferred that laminin-11 or laminin-3, which are trimers of these subunits, represent the most abundant basement membrane proteins in the LG.Many studies on decellularized tissuederived hydrogels often overlook the identification of optimal manufacturing methods and may fail to consider the ideal parameters.In our study, we found that hydrolysis for a minimum of 24 h is necessary to achieve the best stimulation of LG-EpC tear secretion, a crucial parameter for producing a functional LG construct.However, Pouliot and colleagues demonstrated that pepsin hydrolysis of decellularized lung ECM for less than 24 h offers advantages in terms of mechanical properties and cytocompatibility [28], underscoring the distinct characteristics of different tissue sources.In our experiments, longer hydrolysis times not only improved cytological features but also significantly increased dry weight and hyaluronan content.
We were able to produce a gelled dLG-HG with structural integrity from dLG-ECM hydrolyzed for 4-96 h.Despite observing distinct ultrastructural variations based on the duration of digestion, wherein 48 h of hydrolysis resulted in noticeably reduced porosity, enhanced interconnectivity, and thicker fibers in comparison to Col in the resulting dLG-HG, no changes were observed in terms of rheological properties.The gelation occurs by self-assembly when pH and temperature are adjusted to physiological values without the necessity of chemical crosslinkers and UV-light or the occurrence of toxic byproducts.That offers ideal conditions to incorporate cells prior to gelation in the future with regard to a 3D construct.However, gelation kinetics and storage modulus are not yet ideal for its application as a bioink, although all dLG-HG variants were stiffer than Col in rheological measurements.A bioink should solidify rapidly and provide enough rigidity to stay in shape while still recapitulating the mechanical characteristics of the original organ [42].In order to address this issue, several options are available and to be considered in future experiments.Gelation kinetics and storage modulus can be tuned by addition of either natural or synthetic components like alginate [43], nanofibrillated cellulose [44] or polyethylene glycol [45] retaining the advantages of the tissue specific dLG-HG.
Our findings revealed the lowest viability of LG-EpC, LG-MSC, and HUVEC on MG, which is in accordance with others who showed decreased proliferation of human LG-EpC on MG as compared to Col [46].The authors explained this with MG promoting differentiation as opposed to proliferation.Similar results were obtained regarding bone marrow derived MSC cultured on MG, which enhanced their osteogenic differentiation while reducing proliferation compared to polystyrene [47].LG-EpC proliferation was significantly enhanced by dLG-HG ⩾ 12 h compared to Col and MG, as indicated by Ki67 + nuclei, supporting the concept of dLG-HG as the ideal cell niche.Substrate stiffness is known to influence cell proliferation and function [48,49], so that reduced Col stiffness may partially account for detrimental effects on cell behavior compared to dLG-HG besides the compositional differences.Sun et al showed that proliferation of umbilical cord derived MSC was higher on softer substrates whereas growth of bone marrow derived MSC in contrast increased with greater substrate stiffness [50,51].That demonstrates differential substrate stiffness preferences of MSC depending on their tissue origin.Consequently, we can assume that LG-MSC prefer a rather soft substrate based on in vivo measurements of LG stiffness [52].This can in part explain enhanced LG-MSC proliferation on Col compared to all dLG-HG variants.
Noteworthy, dLG-HG ⩾ 24 h did promote tear secretion by LG-EpC more than dLG-HG ⩽ 12 h, Col and even MG.The simultaneously reduced dLG-HG area may even have led to an underestimated impact of dLG-HG on LG-EpC tear secretion, highlighting its promoting effect.This is in accordance with results of others who showed significantly improved functionality of human primary kidney cells on methacrylated kidney ECM hydrogel compared to methacrylated gelatin [53].Furthermore, esophageal ECM retained tissue specific characteristics of esophageal stem cells better than heterologous ECM [54], and heart ECM supplemented collagen hydrogel was able to induce cardiac differentiation of embryonic stem cells more than growth factor supplemented collagen [26].MG is known to improve functional and morphological properties of a variety of cell types [55], including salivary gland EpC [56] and sweat gland EpC [57], and hepatocyte function was equally supported by liver derived ECM hydrogel and MG [58].These effects are reported to be due to the complex MG ingredients comprising multiple growth factors besides the basement membrane proteins collagen-IV and laminin, and proteoglycans among others [55,59] originating from the murine Engelbreth-Holm-Swarm sarcoma, thus excluding its clinical application.Consequently, it is remarkable that dLG-HG ⩾ 24 h significantly exceeded MG's secretion promoting effect on LG-EpC emphasizing the impact of the tissue specific ECM composition on inherent cell's function.
The morphological alterations of LG-EpC and HUVEC grown on dLG-HG in contrast to Col, leading to multilayered epithelia and rudimental tube formation, respectively, point to differentiation promoting effects of dLG-HG, foremost after hydrolysis times ⩾24 h, while still enabling proliferation more than MG.Rudimental angiogenic potential of HUVEC on dLG-HG can be linked to laminin content, as laminin is thought to be a crucial component for the tube formation promoting effect of MG.Thus, laminin enriched Col was able to induce HUVEC differentiation to a limited extend in contrast to pure Col in a study by Kubota and colleagues [60].However, the ratio of the dLG-HG components, as well as the overall protein concentration and the substrate stiffness altogether should be improved to reflect ideal angiogenic conditions.
Morphological changes of LG-MSC on all substrates, besides cell culture plastic, are rather due to the structural features of hydrogels per se in contrast to the less intricate and much stiffer surface of plastic on which LG-MSC maintained their original morphology and distribution.Accordingly, it was shown that substrate stiffness immensely influences MSC morphology [61].
Not only proliferation, but also differentiation potential of MSC depends on substrate stiffness.In this regard, adipogenic MSC differentiation is more effective on softer substrate whereas osteogenic differentiation is enhanced by a more rigid substrate [62].Furthermore, substrate stiffness is known to regulate the MSC secretome.Specifically, a softer surface has been shown to enhance the secretion of immunomodulatory and regenerative factors [63].Of note, tissue specific effects of hydrogels derived from decellularized ECM on human bone marrow derived MSC were already proven, as hydrogels derived from the nucleus pulposus and the annulus fibrosus led to respective tissue specific differentiation of the MSC [64].The release of secreted factors by LG-MSC grown on dLG-HG should be addressed in the future.
As dLG-HG-24 h, −48 h, and −96 h altogether had a positive impact on LG-EpC tear secretion and viability as well as HUVEC viability, the ultrastructural differences between these samples most likely had no cytological implications.Improved hydrolysis efficiency and higher concentration of tissue specific compounds after at least 24 h rather account for these beneficial cytological effects.
The observed cell-dependent dLG-HG degradation, as displayed by area reduction, indicates vital cell-matrix interaction and is desirable regarding future in vivo applications and the process of constructive remodeling [65].While degrading the provided dLG-HG, the cells are enabled to shape and restructure their environment by simultaneously secreting new ECM proteins (reviewed in [66]).Furthermore, the degradation products of ECM derived scaffolds were shown to have antibacterial activity in vitro [67] as well as chemotactic and mitogenic activity towards multipotent progenitor cells [68].dLG-HG degradation was mitigated, but not fully prevented, by a broadband MMP-inhibitor, that inhibits MMP-1 (fibroblast collagenase), MMP-2 (72 kDa gelatinase), MMP-3 (stromelysin), MMP-8 (neutrophil collagenase), and MMP-9 (92 kDa gelatinase).This could either be explained by incomplete MMP inhibition, as for example the collagenase MMP-13 is not inhibited, or a synergistic shrinkage mechanism, combining MMP mediated degradation and collagen lattice contraction [69].MMPs are crucial contributors to constructive ECM remodeling and thus essential for tissue homeostasis and maintenance (reviewed in [70]).MMP mediated ECM degradation can expose bioactive recognition sites, termed matricryptic sites, modulating favorable cell responses like proliferation or migration [70,71].
In contrast, Col area decreased equally despite MMP-inhibition pointing to a MMP independent process caused by mechanical contraction and water loss [69].Thus, Col is either less prone to MMPmediated digestion than dLG-HG, or LG-EpC growing on Col are less stimulated to express MMP, implying less interaction with this substrate.Degradation of MG was not detectable within the monitored period, what can be explained equally.In contrast to Col and dLG-HG, the main component of MG are not fibrillar collagens like collagen type I, but basement membrane components like laminins [55] so that MG is less vulnerable to MMP digestion.
Although desired, biodegradation of dLG-HG should be alleviated, enabling cells to secrete a corresponding amount of new ECM to allow for an appropriate cultivation period of an in vitro model.Degradation rate can be reduced, for example, by MMP inhibition [72].In that study, in vivo MMP inhibition by doxycycline slowed down biodegradation without reducing biocompatibility.Others observed similar area reduction in case of urinary bladder and dermal ECM derived hydrogels by incorporated fibroblasts, which was dependent on both ECM concentration and tissue source [73].Accordingly, increasing the ECM content in dLG-HG is a further option to decrease its biodegradation accompanied by potentially higher bioactivity.Furthermore, there are several crosslinking options available that increase hydrogel rigidity and reduce susceptibility to enzymatic degradation [74,75].The direct methacrylation of hydrolyzed ECM enables photo-crosslinking and facilitates printability as was shown for ECM-derived hydrogels from kidney [53] and liver [76].Alternatively, ECM-derived hydrogels can be supplemented with methacrylated collagen [77] and gelatine [78,79], or other photocrosslinkable additives like poly(ethylene)glycol derivates [80,81].The addition of chemical crosslinkers like proanthocyanidin [82], genipin [83], or the ethyl(dimethylaminopropyl)carbodi-imide/Nhydroxysuccinimide system are further options [84].Besides, alginate [43] or nanofibrillated cellulose [85] can be combined with ECM hydrogels to form hybrid hydrogels with tunable stiffness.The applicability of these potential modifications on dLG-HG should be considered and evaluated in terms of cytocompatibility, biodegradation and rheological properties to obtain an optimal bioink for the creation of an LG in vitro model and further tissue engineering applications.

Conclusions
In conclusion, we were able to develop a hydrogel based on decellularized porcine LG tissue approaching the complex composition of native LG-ECM.dLG-HG formed from ECM hydrolyzed for at least 24 h did not compromise viability of LG-MSC and promoted viability of HUVEC and LG-EpC more than Col and MG.Most outstandingly, dLG-HG from ECM hydrolyzed for at least 24 h amplified tear secretion by LG-EpC compared to Col and MG, indicating suitability as a ground substance of an LG in vitro model reflecting in vivo conditions, or a transplantable LG analogue in the long term.Despite the necessity of rheological adaptations and regulation of biodegradation, dLG-HG could contribute to exploring the pathomechanisms underlying ADDE as well as appropriate curative and regenerative treatments.

Figure 2 .
Figure 2. Mass spectrometry analysis of native LG (nat) and dLG-HG after 4, 12, 24, 48, and 96 h hydrolysis.(A) Mean abundance of extracellular proteins over all groups.(B) Extracellular protein abundance normalized to row means.(C) Mean of basement membrane proteins abundance over all groups.(D) Basement membrane protein content normalized to row means.White boxes with x denote measurements below the detection threshold.

Figure 3 .
Figure 3. Rheological and ultrastructural characterization of dLG-HG.(A) dLG-HG and collagen type I (Col) demonstrated typical gelation kinetics as shown by temperature sweep rheology with increasing loss modulus (G ′′ ) and storage modulus (G ′ ) while G ′ exceeded G ′′ when the gelation point was reached.(B) Maximal G ′ indicating hydrogel stiffness of dLG-HG was significantly higher than of Col. (C) dLG-HG gelation point temperature was significantly higher than that of Col. (D) Scanning electron microscopy of critical point dried Col and dLG-HG (scale upper row: 2 µm; scale lower row: 500 nm).Porosity as the percentage of black pixels after conversion to binary images (E), the number of interconnections (F) and fibre diameter (G) were determined in 6 representative images per group.* p < 0.05, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001.

Figure 4 .
Figure 4. Cell response of LG-EpC on dLG-HG.(A) LG-EpC viability after 3 d in culture on collagen (Col), Matrigel (MG), or dLG-HG with 4-96 h hydrolysis was determined via MTT assay and showed highest viability on dLG-HG with ⩾24 h hydrolysis.(B) Proliferating Ki67 + cells were quantified by means of immunohitochemically stained paraffin sections (D) showing significantly more proliferation on dLG compared to Col and MG.(C) Epithelial thickness of LG-EpC after 3 d in culture was quantified by means of H/E stained paraffin sections (E).LG-EpC formed a monolayer on Col, multilayered aggregates on MG, and multilayered epithelia on dLG-HG with increasing thickness corresponding to longer hydrolysis.(A)-(C) n = 6; * p < 0.05, * * p < 0.01, * * * p < 0.001.(E) and (F) scale 50 µm.

Figure 5 .
Figure 5. Cell response of LG-MSC on dLG-HG.(A) LG-MSC viability after 3 d in culture on collagen type I (Col), Matrigel (MG), or dLG-HG with 4-96 h hydrolysis was determined via MTT assay and was least on MG ( * p < 0.05).(B) Immunocytochemical staining of the MSC marker nestin shows cell clusters on all hydrogel substrates with most dense aggregates on MG in contrast to evenly distributed cells on cell culture plastic (scale 100 µm).

Figure 6 .
Figure 6.Cell response of GFP-HUVEC on dLG-HG.(A) GFP-HUVEC viability after 3 d in culture on collagen type I (Col), Matrigel (MG), or dLG-HG with 4-96 h hydrolysis was determined via MTT assay showing least viability on MG and highest viability on dLG-HG increasing with hydrolysis time.(B) In vitro lumen formation by GFP-HUVEC was quantified after 24 h in culture as lumina per image area.(C) Most efficient lumen formation was observed on MG, on Col only sporadically small lumina were formed while on dLG-HG more distinct lumina were visible especially on dLG-HG-96 h (scale 100 µm).

Figure 8 .
Figure 8. MMP-mediated dLG-HG degradation.LG-EpC were seeded on collagen type I (Col), Matrigel (MG), and dLG-HG coated 96-well plates and cultured with DMSO (ctrl.) or MMP-inhibitor supplemented culture medium.Hydrogel area reduction was photographically documented for 14 d (A) and quantified (B).MG area did not change during the observed period, but cells contracted on top.Col area was unchanged after 3 d, but halved within 14 d.dLG-HG area was already reduced after 3 d and decreased further until 14 d.MMP-inhibitor treatment had no impact on Col area reduction, but significantly decelerated dLG-HG degradation.* p < 0.05, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001.