Human adipose tissue-derived stromal cells act as functional pericytes in mice and suppress high-glucose-induced proinflammatory activation of bovine retinal endothelial cells

Aims/hypothesis The immunomodulatory capacity of adipose tissue-derived stromal cells (ASCs) is relevant for next-generation cell therapies that aim to reverse tissue dysfunction such as that caused by diabetes. Pericyte dropout from retinal capillaries underlies diabetic retinopathy and the subsequent aberrant angiogenesis. Methods We investigated the pericytic function of ASCs after intravitreal injection of ASCs in mice with retinopathy of prematurity as a model for clinical diabetic retinopathy. In addition, ASCs influence their environment by paracrine signalling. For this, we assessed the immunomodulatory capacity of conditioned medium from cultured ASCs (ASC-Cme) on high glucose (HG)-stimulated bovine retinal endothelial cells (BRECs). Results ASCs augmented and stabilised retinal angiogenesis and co-localised with capillaries at a pericyte-specific position. This indicates that cultured ASCs exert juxtacrine signalling in retinal microvessels. ASC-Cme alleviated HG-induced oxidative stress and its subsequent upregulation of downstream targets in an NF-κB dependent fashion in cultured BRECs. Functionally, monocyte adhesion to the monolayers of activated BRECs was also decreased by treatment with ASC-Cme and correlated with a decline in expression of adhesion-related genes such as SELE, ICAM1 and VCAM1. Conclusions/interpretation The ability of ASC-Cme to immunomodulate HG-challenged BRECs is related to the length of time for which ASCs were preconditioned in HG medium. Conditioned medium from ASCs that had been chronically exposed to HG medium was able to normalise the HG-challenged BRECs to normal glucose levels. In contrast, conditioned medium from ASCs that had been exposed to HG medium for a shorter time did not have this effect. Our results show that the manner of HG preconditioning of ASCs dictates their immunoregulatory properties and thus the potential outcome of treatment of diabetic retinopathy.


Introduction
Diabetic retinopathy is the most common microvascular complication of diabetes and remains a leading cause of blindness worldwide [1]. The prevalence of diabetic retinopathy increases with diabetes duration and develops from non-proliferative diabetic retinopathy to proliferative diabetic retinopathy (PDR) and macular oedema [2]. Hyperglycaemia in the retina activates four main biochemical pathway-related changes: (1) polyol pathway flux; (2) accumulation of advanced glycation end-products (AGEs); (3) activation of protein kinase C (PKC); (4) activation of hexosamine pathway flux [3]. Together, these molecular changes induce formation of reactive oxygen species (ROS), which induces impaired retinal blood flow and increased vascular permeability [4]. Retinas or vitreous from diabetic individuals with PDR contain increased levels of proinflammatory mediators including TNF-α, IL-1β [5], IL-6 [6], IL-8 [7] and chemokine (C-C motif) ligand 2 (CCL2) [8,9]. NF-κB is the main transcription factor that regulates expression of proinflammatory genes. Activation and nuclear translocation of NF-κB promotes forward feedback that augments the proinflammatory state of activated cells [10]. Diabetes is related to the upregulation of cyclooxygenase-2 (COX2) encoded by the prostaglandin-endoperoxide synthase 2 gene (PTGS2), both in macro-and microvessels [11]. In retinas of diabetic animals, increased COX2 is followed by increased production of prostaglandin E 2 (PGE2) [12]. At the onset of diabetic retinopathy, pericyte dropout and the subsequent loss of retinal endothelial cells through apoptosis causes vasoregression. This is the driving force for the pathological angiogenesis of PDR [13]. In the normal retina, pericytes protect and regulate endothelial cell survival and proliferation, vessel integrity and the susceptibility of vascular cells to environmental stimuli [14]. Cell therapy with adipose tissue-derived stromal cells (ASCs) were promising in diabetic animal models. Adipose tissue is easy to acquire and is rich in ASCs. These ASCs are the endogenous mesenchymal stem cells in fat and they harbour a multipotent capacity to differentiate, for example, into adipocytes, chondrocytes and osteoblasts [15].
Recent studies suggested a direct role for ASCs in retinal microvascular support [16]. In rodent models of diabetic retinopathy, ASCs acquired pericytic features, which dampened proliferative angiogenesis [17,18]. Recently, we showed that the pericytic nature of ASCs depends on neurogenic locus notch homolog protein 2 (NOTCH2)-based juxtacrine interaction between ASCs and endothelial cells [18]. Moreover, ASCs may home in on sites of inflammation [19]. The prime immunosuppressive factors produced by mesenchymal stem cells and ASCs include indoleamine 2,3-dioxygenase [20], PGE2 [21], nitric oxide (NO) [22], IL-10 [23] and antioxidant enzymes such as haem oxygenase-1 [24]. Individually or in combination, these factors also reduce oxidative stress in target cells [25,26]. We hypothesised that ASCs preconditioned in high glucose (HG) can rescue a dysfunctional retinal endothelium through suppression of the inflammation that was caused by glucose-induced oxidative stress. The aim of this study was to investigate the immunomodulatory paracrine function of ASCs in decreasing the production of endothelial ROS and in the normalisation of dysfunctional endothelial cells in diabetic retinopathy.

Methods
Cell isolation and culture Human subcutaneous adipose tissue samples from healthy individuals (white females aged 25 to <75 years and BMI <30 kg/m 2 ) were obtained after liposuction surgery (Bergman Clinics, Zwolle, the Netherlands). Anonymously donated samples were obtained with informed consent as approved by the ethical board of the University Medical Center Groningen, following the guidelines for 'waste materials'. Lipoaspirates were enzymatically digested to obtain the ASCs, as described in our previous study [27]. The pooled ASCs from passages 3 to 6 were used for experiments. The ASCs were cultured in DMEM (Lonza, Basel, Switzerland) either with normal glucose (NG; 5 mmol/l D-glucose) or HG (30 mmol/l D-glucose) at normoxia (21% O 2 ). Continuous HG maintenance of ASCs (more than three passages; ≥21 days) was considered to be chronic HG, while short-term exposure to HG (7 days) was considered to be acute HG. Conditioned medium from ASCs (ASC-Cme) was collected from confluent monolayers after culturing for more than three passages in HG-DMEM. The percentage of FBS was reduced from 10% to 2% (vol./vol.) for 24 h prior to collection [28].
For the co-culture of ASCs and HUVECs, single-cell suspensions of HUVECs were seeded on top of confluent ASCs monolayers or, as a control, on gelatin-coated wells at 10,000 cells/cm 2 and vascular networks were allowed to form for at least 7 days [29]. Bovine retinal endothelial cells (BRECs) were isolated from freshly-enucleated cow eyes obtained from the slaughterhouse as described previously [30]. First-passage BRECs were used in all experiments. A purity of >99% was routinely achieved in BREC culture, which was checked microscopically, and by immunofluorescence staining of von Willebrand factor (not shown). Confluent monolayers of BRECs were incubated in three different groups (NG-DMEM, HG-DMEM and ASC-Cme) for 7 days to study the effects of each set of conditions.
Animals and the retinopathy of prematurity model All animal experiments in this study adhered to the association for research in vision and ophthalmology (ARVO) Statement for the use of animals in ophthalmic and vision research. Male C57BL/6J mice (Charles River, Frankfurt, Germany) were housed with free access to standard chow and water under a 12 h light-dark rhythm. To study the effect of ASCs on hypoxia-driven angiogenesis, newborn mice were subjected to the model of retinopathy of prematurity (ROP) [31]. Mice at postnatal (P) day 7 (P7) were exposed to an atmosphere of 75% oxygen with their nursing mother for 5 days and then returned to room air at P12. Directly after their return to room air, randomly selected mice were intravitreally injected under anaesthesia with either 1 μl of PBS containing approximately 10,000 ASCs (passage 1) or 1 μl of PBS as a control. Eyes were enucleated under deep anaesthesia at P13 for immunofluorescence analysis; at P13 and P19 for quantitative real-time PCR (qPCR) analysis; and at P17 for quantification of neovascularisation. After collection, eyes were immediately fixed in buffered formalin or stored at −80°C for the following analysis.
Quantification of hypoxia-driven neovascularisation Neovascularisation in retinas was assessed in paraffin sections of P17 animals injected at P12 with ASCs or a control. To this end, sections (6 μm) were stained with periodic acid-Schiff's reagent (Sigma-Aldrich). Nuclei of neovessels at the vitreous side of the inner limiting membrane of the retinas were counted as described previously [32].

Assessment of ASCs in vivo
Whole-mount retinas from P13 animals were permeabilised with 0.5% (wt/vol.) Triton-X100 at room temperature for 1 h.
Gene expression analysis Total RNA was extracted from ASCs and BRECs in TRIzol reagent (Life Technologies, Carlsbad, CA, USA) following the manufacturer's protocol. Retinas were isolated from frozen eyes of ROP and control mice at P13 (n = 5) and ROP mice at P19 in the presence of ASC injection or PBS (n = 5). Afterwards, 1 μg of total RNA from each sample was reverse transcribed using the First Strand cDNA Synthesis Kit (Fermentas, Vilnius, Lithuania) according to the manufacturer's instructions. The cDNA equivalent of 10 ng RNA was used for amplification in 384-well plates in a TaqMan ABI 7900HT thermal cycler (Applied Biosystems, Foster City, CA, USA) in a final reaction volume of 10 μl containing 5 μl SYBR Green Universal PCR Master Mix (BioRad, Hercules, CA, USA) and 6 μmol/l primer mix (forward and reverse). The cycle threshold (C t ) values were normalised to GAPDH/ACTB as a reference gene using the ΔΔC t method [33].
Assessment of cell viability Viability was assessed using the Apoptosis & Necrosis Kit (Promokine, Heidelberg, Germany) as recommended in the manufacturer's instructions. In short, BRECs were incubated with 5 μl fluorescein-conjugated annexin V (a marker of apoptosis) and 5 μl ethidium homodimer III (at a concentration of 2 × 10 6 cells/ml) at room temperature for 15 min. Fluorescence was recorded on a BD FACSCalibur (BD Biosciences, Franklin Lakes, NJ, USA) within 1 h of staining.
Quantification of ROS Cellular ROS production was determined using the dye 2′,7′-dichlorofluorescein diacetate (DCFH-DA, Sigma-Aldrich). Two-electron oxidation of DCFH-DA results in the formation of a fluorescent product, dichlorofluorescein (DCF) [34]. Experimental cells were suspended in 20 μmol/l DCFH-DA in the dark at 37°C for 15 min. The general ROS scavenger N-acetyl-L-cysteine (NAC, 10 mmol/l, Sigma-Aldrich) and H 2 O 2 (3 μmol/l, Merck Millipore, Darmstadt, Germany) were used as negative and positive controls, respectively. Samples were analysed directly using a FACSCalibur within 15 min of staining.
Screening for immune stimulation The supernatant of HGstimulated BRECs cultured with or without ASC-Cme and the control group (BRECs in NG) were collected after 7 days. To assess the immunogenicity of these samples, THP1-XBlue-MD2-CD14 cells (InvivoGen, Toulouse, France) were used as described previously [35]. THP1-XBlue-MD2-CD14 cells were plated, then each well was stimulated with samples of supernatants and cultured overnight at 37°C in 5% CO 2 . Lipopolysaccharide (LPS, 10 μg/ml, Sigma-Aldrich) was used as a positive control. Production of SEAP in the supernatant was quantified using QUANTI-Blue (InvivoGen). An aliquot of QUANTI-Blue (200 μl) was dispensed into a new flatbottomed 96-well plate with 20 μl of supernatant from the stimulated cell-lines for 45 min at 37°C. Secreted embryonic alkaline phosphatase (SEAP) activity, representing activation of NF-κB, was then measured at a wavelength of 650 nm on a VersaMax microplate reader (Molecular Devices, Biberach an der Riss, Germany) [35].
ELISA Culture medium was collected from different experimental conditions. The concentrations of PGE2 and CCL2 in the medium were quantified, respectively, with the Prostaglandin E2 Human ELISA Kit (ThermoFisher) and the Human CCL2/ MCP-1 DuoSet ELISA (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's protocol. LPS (200 ng/ ml) and TNF-α (50 ng/ml, Sigma-Aldrich) were used as positive controls. The COX2 inhibitor celecoxib (10 mmol/l, Sigma-Aldrich) was used to inhibit PGE2 production. Results were normalised to the number of cells in each experimental condition and presented in pg/ml as fold change relative to their respective experimental controls.
Monocyte adhesion assay One of the cardinal steps of inflammation is the infiltration of immune cells such as monocytes across the endothelial cell layer [36]. THP-1 monocytes were stained using the Vybrant CFDA SE Cell Tracer Kit (ThermoFisher) according to the manufacturer's instructions. Treated groups of BRECs (under HG with or without ASC-Cme), a positive-control group (pre-incubated with medium containing TNF-α [10 ng/ml] for 8 h) and a control group (NG) were plated at a density of 5 × 10 3 cells/cm 2 in standard 96-well culture plates and allowed to adhere at 37°C for 4 h. Labelled THP-1 cells (2 × 10 5 cells/ml) were added to each well. After incubation for 1 h, wells were washed to remove non-adhered cells. Fluorescence of adherent cells was recorded on a Varioskan spectrofluorometer (Thermo Scientific, Waltham, MA, USA) at excitation/emission = 492/520 nm [37].
Scratch wound healing assay BRECs in three different groups (NG, HG and ASC-Cme) were cultured until they reached confluency. The straight, width-limited scratch was made in all the wells, simulating a wound. The recovery of both wound edges was recorded simultaneously using the Solamere Nipkow confocal live cell imaging microscope (Solamere Technology Group, Salt Lake City, UT, USA) for 30 h. The percentage of covered area between the edges was analysed by ImageJ 1.8.0-172 software (imagej.nih.gov/ij/download/).
Statistical analysis Data are expressed as mean ± SEM and relative to vehicle controls of at least three independent experiments in triplicate. Statistical evaluation was performed using unpaired t tests and ANOVA followed by Bonferroni post hoc analysis. p values <0.05 were considered statistically significant.

Results
Ultrastructure of ASC-induced vascular networks The 0.5 μm cross-sections showed that interconnected laminar structures had been formed by co-cultures of endothelial cells on ASC monolayers at day 5 and subsequent days (Fig. 1a). Transmission electron micrographs of longitudinal sections showed the build-up of 3D structures that comprised endothelial-cell-derived (Fig. 1c) vessel-like structures with a lumen (Fig. 1a-c), surrounded and tightly aligned by ASCs (Fig. 1c). At high magnification, these vessel-like structures appeared as intermittent structures between the ASCs (Fig. 1d). The vascular networks formed were reminiscent of genuine vessels with respect to the slanted intercellular junctions between endothelial cells (Fig. 1e, h, i). Peg-and-socket processes [38,39] that extended from ASC-derived pericytes to endothelial cells had also formed (Fig. 1f, g), as well as an extracellular matrix-based membrane between pericytes and (e) Specific cell-cell connections with tight junctions between endothelial cells (white arrowheads), with lumen formation on top of the ASCs. (f) ASCs deposit extracellular matrix (black asterisks), which forms a basement membrane-like structure between the endothelial cells and the ASCs. (f, g) Peg-and-socket connections are shown by the lightning symbols, and the inset in (g) shows intracellular filaments (white asterisks), indicative of contractility, similar to smooth muscle cells, i.e. hinting at the maturation of ASCs to pericytes. (h, i) Detailed views of the endothelial cell-cell connections and basal membrane formation around the endothelial cells, i.e. the vascular structure with connected ASCs. Scale bar, 5 μm. Lumen (L); endothelial cells (arrows); ASCs (black arrowheads); endothelial cell-cell connections (white arrowheads); extracellular matrix formation (black asterisks); intracellular filaments representative for smooth muscle cell phenotype (white asterisks); peg-and-socket connections of ASCs with endothelial cells (lightning symbols) endothelial cells (Fig. 1f, i). ASC-derived pericytes, i.e. those cells in close contact with endothelial cells, had lost their typically abundant vesicular contents compared with more distal ASCs (Fig. 1d). These ultrastructural results indicate that ASCs promoted formation of vascular networks by endothelial cells and that ASCs had acquired a functional pericytic phenotype in vitro (Fig. 1g).
ASCs stabilise hypoxia-driven angiogenesis but also engraft at pericytic sites in the animal model of ROP ASCs were injected intravitreally at P12 in the eyes of ROP-model mice (Fig. 2a). At P13, injected ASCs were present in angiogenic sprouts and had attached to endothelial cells of maturating capillaries at a pericyte-like position (arrows, Fig. 2b, c). Further analyses showed that dye-labelled-ASCs (green, Fig. 2d-f) had aligned and attached to maturing intraretinal capillaries (red, Fig. 2d-f) at pericyte-specific perivascular positions and in contact with the endothelium.
Quantification of hypoxia-induced retinal neovascularisation at P17 showed that injection of ASCs in the animal models of ROP increased neovascularisation by 54% (Fig. 2g- ASCs modulate the inflamed ROP micro-environment ROP and control retinas (P13) were assessed by qPCR. At P13, i.e. 24 h after returning the animals from hyperoxia to ambient oxygen, Angpt1 (encoding angiopoietin [ANGPT] 1) expression had slightly decreased in ROP retinas, while Angpt2 had slightly increased (although neither significantly) compared with the non-ROP controls (Fig. 3a). This was accompanied by increased expression of angiogenesis-related genes i.e. Vegfa (encoding vascular endothelial growth factor [VEGF] A; 2.0-fold change), Fgf2 (3.2-fold change), and Col4a1 (1.4-fold change) (Fig. 3a). The analysis showed increased expression of proinflammatory genes i.e. Il1b (2.0-fold change) and Ccl2 (2.1-fold change) compared with P13 controls. Cxcl15 (mouse orthologue of human CXCL8) and Tnf changed non-significantly (Fig. 3a). Comparison of ROP retinas with and without ASC intervention showed that injection of ASCs at P12 caused a 1.5-fold increase of Angpt1 and a 0.6-fold decrease of Angpt2, suggesting induction of vascular quiescence. Injection of ASCs into the ROP eyes caused increased expression of inflammatory genes such as Tnf, Cxcl15 and Ccl2 by 2.2-, 1.8-and 2.5-fold, respectively. A slight decrease was observed in the expression of proangiogenic Vegfa and a 1.3-fold increase of Fgf2 was observed in ROP at P19 with ASCs, compared with the control ROP at P19 without ASCs (Fig. 3b). No differences in the expression of Pdgfb, Col4a1 and Il1b were detected.
Chronic or acute exposure to HG influences ASC gene expression The expression of 36 relevant genes was measured in ASCs after acute exposure to HG (i.e. maintained in NG medium, followed by 7 days in HG medium) or chronic exposure to HG (i.e. always maintained in HG medium and for more than 21 days). Data are presented as the fold change compared with (chronic) maintenance in NG medium (Fig. 4a, b). The acute exposure of ASCs to HG upregulated the proinflammatory genes TNF (1.8-fold change), IL1A (6.0-fold change), IL6 (4.5-fold change), CCL2 (4.3-fold change) and CXCL8 (6.0-fold change); as well as the proangiogenic genes VEGFA (4.8-fold change), ANGPT2 (1.8-fold change) and MMP1 (4.2-fold change). Expression of the angiopoietin receptors TIE1 and TEK (also known as TIE2) was downregulated (0.52-fold and 0.47-fold change, respectively). The glucose transporter SLC2A1 (encoding GLUT1) was upregulated 2.5-fold. Interestingly, acute HG exposure induced a 4.9-fold change in the mesenchymal pericyte marker RGS5, while expression of other mesenchymal markers such as ACTA1, TAGLN and PDGFRB changed only marginally. Endothelial marker PECAM1 was also upregulated (1.3-fold change). The expression of two immunoregulatory genes, PTGS2 (encoding COX2) and IDO1, were upregulated (3.4-fold and 3.6-fold change, respectively). The expression of all genes in ASCs was similar upon culture under chronic HG and chronic NG conditions, which indicated that ASCs adapt to different glucose concentrations in the culture medium during long-term culture. The inflammatory response after acute exposure to HG resulted in an increased release of PGE2 (11.0-fold change) and CCL2 (1.6-fold change) by ASCs, compared with chronic exposure to HG (Fig. 4c, d). Interestingly, chronic exposure to HG medium slightly decreased PGE2 secretion (0.3-fold change) and secretion of CCL2 (0.8-fold change) compared with controls maintained in NG. LPS and TNF-α both induced the secretion of PGE2 and CCL2, while celecoxib inhibited the production of PGE2 completely and had less effect on secretion of CCL2 in acute-HG-stimulated ASCs.
The monocytic NF-κB reporter (SEAP) cell line THP1-XBlue-MD2-CD14 was used to assess the proinflammatory capacity of conditioned medium obtained from BRECs that were co-treated with HG and ASC-Cme. The strongest activation of NF-κB occurred after stimulation with LPS (29.0-fold change, Fig. 5f), while neither NG nor HG medium (DMEM) had any influence. Compared with NG controls, the activation of monocyte-expressed NF-κB increased (1.4-fold change) upon exposure to conditioned medium from HGstimulated BRECs. In contrast, NF-κB activation was normalised to NG control levels (0.6-fold change) by conditioned medium from BRECs that were HG induced and simultaneously treated with ASC-Cme.
ASC-Cme downregulates principal inflammatory factors in HG-challenged BRECs HG stimulation of BRECs for 7 days upregulated the gene expression of the relevant proinflammatory genes TNF, IL1B, IL1A, IL6, CXCL8 and CCL2, as well as leucocyte adhesion-related genes SELE, ICAM1 and VCAM1. In addition, the proangiogenic genes ANGPT1, ANGPT2, VEGFA, VEGFB and PDGFB, and the endothelial NO synthases, NOS2 and NOS3, were upregulated, as well as PTGS2. The upregulation of these genes was normalised to NG control level in ASC-Cme-treated HG-challenged BRECs. Only KLF4 did not change after HG challenge of BRECs. Interestingly, expression of both KLF4 and NOS2 were upregulated in ASC-Cme-treated HG-challenged BRECs. The expression of ANGPT1, a vessel quiescenceassociated factor, remained increased upon ASC-Cme treatment of HG-challenged BRECs (Fig. 6a-r).

Discussion
The main results of our studies are that human ASCs stimulate retinal angiogenesis in ROP mice via acquisition of a pericytic position and function. This was confirmed by ultrastructural analyses showing that in vitro ASCs promote vascular network formation of endothelial cells and stabilise these networks via a pericytic function. Extracellular matrix was deposited between ASC-derived pericytes and endothelial cells in vitro, while typical peg-and-socket connections were also formed [38,39]. Acute exposure of ASCs to HG caused a proinflammatory and proangiogenic activation compared with ASCs that were propagated chronically in HG. In fact, conditioned medium from ASCs that had been chronically exposed to HG reduced ROS and proinflammatory activation of BRECs challenged with HG. Activated Values are mean ± SEM (n = 6). *p < 0.05, **p < 0.01, ***p < 0.001 vs NG control; †p < 0.05, † †p < 0.01 vs HG genes were reduced to normal expression levels, while endothelial protective genes such as KLF4 were upregulated. The immunomodulatory properties of ASCs [40][41][42] may vary upon passaging in culture [41]. Early passages of ASCs express markers, e.g. ligands that stimulate cells of the adaptive immune system such as MHCII, CD80 and CD86, which are lost upon passaging [41,43]. Our results show that ASC injection into ROP mice, a model of pathological angiogenesis in PDR [44], led to an increased expression of inflammatory genes such as Tnf, Cxcl15 and Ccl2, which might relate to the early passaging of ASCs. The balance of ANGPT1 and ANGPT2 and the expression of VEGFA are among the most important regulators of vascular development, maturation and maintenance [45]. Reduced expression of pericyte-derived ANGPT1, and concomitant upregulation of its antagonist ANGPT2 and increased VEGFA levels induce pericyte dropout, vessel destabilisation and sprouting angiogenesis [46]. In ROP retinas of ASC-treated mice, expression of Angpt1 and Fgf2 were increased, while levels of Angpt2 and Vegfa were reduced. Injected ASCs adhered to neovessels at a pericytic position. This corroborates the role of NOTCH2 in the juxtracrine function of ASCs [18]. Earlier work from our group showed that under acute HG culturing, ASCs largely maintain their pericytic role as an endothelial supporting cell, despite an upregulated ROS production accompanied by increased apoptosis in ASCs [29].
In general, most papers use DMEM as constitutive culture medium to propagate ASCs. This medium contains 25 mmol/l D-glucose, which is, in fact, similar to hyperglycaemic conditions in vivo. A main finding of our study was that chronic exposure of ASCs to HG medium normalised gene expression to levels seen under NG conditions. Interestingly, in comparison to chronic propagation in HG, ASCs that were shifted from NG-DMEM to HG-DMEM for 7 days had upregulated gene expression of proinflammatory cytokines and chemokines, but also of proangiogenic factors. Moreover, secretion of PGE2 and CCL2 were also strongly increased. COX2 is responsible for the diabetes-induced retinal PGE2 production, and inhibition of COX2 inhibited the diabetes-induced upregulation of retinal VEGF, which links COX2 expression to angiogenesis in diabetic retinopathy [12,47,48]. PGE2 activates the extracellular signal-regulated kinases 1 and 2 (ERK1/2) and thus promotes angiogenesis through endothelial upregulation of the secretion of e.g. VEGF, chemokines and activation of cell cycle genes [49,50]. Thus, the acute HG exposure of ASCs likely renders these cells in a proangiogenic state. Their upregulated expression of proinflammatory genes likely also activates endothelial cells. Yet the exposure of retinal endothelial cells to HG alone activated NF-κB and increased expression of downstream IL1B, VEGFA, TNF and ICAM1 to name a few of the upregulated main proinflammatory genes. ASC-Cme suppressed these proinflammatory genes such as PTGS2 in BRECs plus VEGFA as a proangiogenic gene. Acute HG exposure of ASCs would induce endothelial dysfunction rather than counteract it. However, the continuous propagation of ASCs in HG medium not only suppresses their proinflammatory status, but also augments their capacity to normalise endothelial cell function. This was corroborated in HGactivated BRECs by the suppression, by ASC-Cme, of ROS production, the reduction of monocyte adhesion and improved cell survival (Fig. 8). Part of the immunosuppression resides in the suppression of endothelial ROS production by ASC-secreted factors, because ROS functions, via activation of TAK1, as an activator of NF-κB. Taken together, our results show that HG preconditioning of ASCs potentially improves their therapeutic efficacy as compared with short-term exposure to environmental conditions such as HG. Fig. 7 (a) THP-1 cell adhesion to BRECs exposed to HG, with and without ASC-Cme. The intensity of fluorescence of adherent THP-1 cells (mean ± SEM, n = 3, fold change of NG-treated control) in resting or activated BRECs was measured. A significant decrease in adhered THP-1 cells occurred after treatment of HG-challenged BRECs with ASC-Cme, compared with HG treatment alone. (b) Scratch wound healing assay to study interaction in the BREC monolayer under NG, HG and ASC-Cme conditions. The percentage of covered area between the wound edges was analysed. The percentage signifies the remaining gap size 30 h after making the scratches, compared with the initial gap size. The gap width decreased in a similar pattern in all three groups (NG, closed circles; HG, open circles; ASC-Cme, open diamond) and was not significantly delayed in HG conditions. Values are mean ± SEM. **p < 0.01, ***p < 0.001 vs NG control; †p < 0.05 vs HG Duality of interest The authors declare that there is no duality of interest associated with this manuscript.
Contribution statement GH designed the study, acquired data, analysed/ interpreted data, wrote and revised the manuscript. EP, FP, SB and JK designed the study, acquired data, analysed/interpreted data and reviewed/revised the in vivo part of the manuscript. GAPJ and KM designed the study, acquired data, analysed/interpreted data and reviewed/ revised the in vitro part of the manuscript. IK, MJAvL, GK, HPH and MCH designed the study and reviewed/revised the manuscript. All authors are fully responsible for all content and editorial decisions and were involved at all stages of manuscript development, and have approved the final version of the manuscript to be published. MCH is responsible for the integrity of the work as a whole.