Exosomes from hypoxic endothelial cells have increased collagen crosslinking activity through up‐regulation of lysyl oxidase‐like 2

Abstract Exosomes are important mediators of intercellular communication. Additionally, they contain a variety of components capable of interacting with the extracellular matrix (ECM), including integrins, matrix metalloproteinases and members of the immunoglobin superfamily. Despite these observations, research on exosome‐ECM interactions is limited. Here, we investigate whether the exosome‐associated lysyl oxidase family member lysyl oxidase‐like 2 (LOXL2) is involved in ECM remodelling. We found that LOXL2 is present on the exterior of endothelial cell (EC)‐derived exosomes, placing it in direct vicinity of the ECM. It is up‐regulated twofold in EC‐derived exosomes cultured under hypoxic conditions. Intact exosomes from hypoxic EC and LOXL2 overexpressing EC show increased activity in a fluorometric lysyl oxidase enzymatic activity assay as well as in a collagen gel contraction assay. Concordantly, knockdown of LOXL2 in exosome‐producing EC in both normal and hypoxic conditions reduces activity of exosomes in both assays. Our findings show for the first time that ECM crosslinking by EC‐derived exosomes is mediated by LOXL2 under the regulation of hypoxia, and implicate a role for exosomes in hypoxia‐regulated focal ECM remodelling, a key process in both fibrosis and wound healing.


Introduction
Lysyl oxidase-like 2 (LOXL2) is one of five members of the lysyl oxidase family, consisting of lysyl oxidase (LOX) and lysyl oxidase-like 1 -4. Like all LOX family members, LOXL2 facilitates crosslinking of collagens and elastin by catalysing oxidative deamination of lysine residues [1]. This process is crucial for giving the extracellular matrix (ECM) its tensile strength and load-bearing capabilities [2]. Although all LOX family members share a high degree of homology, only LOXL2, LOXL3 and LOXL4 possess four scavenger receptor cysteinerich domains at their N-termini [3]. The exact function of these domains is yet to be elucidated, but has been suggested to facilitate protein-protein interactions in the ECM [4]. Apart from its extracellular role, intracellular functions of LOXL2 have been reported in epithelial to mesenchymal transition (EMT) and breast cancer metastasis [1,5].
Lysyl oxidase-like 2 expression is regulated by hypoxia-inducible factor 1-alpha (HIF-1a), and is up-regulated in endothelial cells (EC) exposed to low oxygen concentrations [6]. Bignon et al. showed that LOXL2 produced by EC co-localizes with, and is required for proper assembly of, collagen IV in the basement membrane [7]. Overexpression of LOXL2 in EC resulted in increased migration and tubulogenesis, which was partially inhibited by the specific LOX enzymatic inhibitor b-aminopropionitrile (BAPN), indicating that the function of EC-derived LOXL2 at least partially relies on its enzymatic crosslinking activity [7].
In a previous study on EC-derived exosomes, we observed an approximate twofold increase of LOXL2 abundance when EC were cultured in 2% oxygen [8]. Exosomes are extracellular vesicles (EV) that are released by many cell types when the multivesicular body fuses with the plasma membrane and the intraluminal vesicles are released into the extracellular environment [9,10]. Exosomes play a role in intercellular communication through a variety of mechanisms [11,12]. Interestingly, exosomes have also been reported to interact directly with the ECM through integrins and matrix metalloproteinases [13][14][15]. Given the extracellular function of LOXL2 and the observations regarding interactions of exosomes with the ECM, we hypothesized that exosome-associated LOXL2 induces ECM crosslinking by deamination of lysine residues of ECM components, under the regulation of hypoxia.
Here, we demonstrate that exosome-associated LOXL2 is present on the exterior of the vesicles, allowing direct contact and interaction with the ECM. Furthermore, we demonstrate for the first time enzymatic LOX activity of intact exosomes, which is increased upon stimulation of EC with hypoxia. Using a fibroblast collagen gel contraction assay to study crosslinking of collagen I, a similar increase in the activity of EC-derived exosomes obtained under hypoxic conditions is observed. Lysyl oxidase activity of EC-derived exosomes was inhibited by lentiviral knockdown of LOXL2, as well as by the LOX enzymatic inhibitor BAPN, and increased by lentiviral overexpression of LOXL2. These data implicate a role for exosomes in hypoxia-regulated focal ECM remodelling, a key process in both fibrosis and wound healing.

Exosome isolation
Exosomes were isolated from conditioned medium by differential ultracentrifugation as described previously [8,18]. Exosome-free medium (prepared using FCS centrifuged for at least 1 hr at 200,000 9 g, followed by 0.2 lm filter-sterilization) was added to cells grown to~75% confluence for 24 hrs. During the 24-hr culturing period cells were exposed to hypoxia (2% O 2 ), or ambient oxygen level (20% O 2 ). The conditioned medium was centrifuged for 15 min. at 1500 9 g to remove cellular debris, after which larger microvesicles were removed by centrifugation for 30 min. at 10,000 9 g. Exosomes were isolated by centrifugation for 60 min. at 100,000 9 g, and subsequently washed twice by resuspending in PBS and centrifugation for 60 min. at 100,000 9 g. Centrifugation was performed using a Beckman LE-80K centrifuge with SW32-Ti and SW60-Ti rotors (Beckman Instruments, Indianapolis, IN, USA).

Immunoblotting
Exosomes were resuspended in RIPA buffer (Santa Cruz Biotechnology, Dallas, TX, USA), cell lysates were resuspended in 19 Laemmli buffer, and protein concentrations were determined by BCA protein assay (ThermoFisher). Equal protein amounts were subjected to SDS-PAGE using NuPAGE 4 -12% Bis-Tris gradient gels (ThermoFisher), and subsequently transferred to polyvinylidene difluoride (PVDF) membranes (ThermoFisher). Depending on the antibody, PVDF membranes were either blocked in 5% bovine serum albumin (BSA; Roche, Basel, Switzerland) or 5% low-fat dry milk powder (Campina, Woerden, The Netherlands) in TBS with 0.1% Tween-20 for 1 hr at RT. polyvinylidene difluoride (PVDF) membranes were incubated with primary antibodies o/ n at 4°C, washed in TBS with 0.1% Tween-20 (TBST), and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies in 5% low-fat dry milk powder in TBST for 1 hr at RT. Membranes were washed in TBS and proteins were visualized using Chemiluminescent Peroxidase Substrate (Sigma-Aldrich) and imaged on the Molecular Image ChemiDoc XRS system (Bio-Rad, Hercules, CA, USA). After imaging, PVDF membranes were stripped from their antibodies using ReBlot Plus Mild Antibody Stripping Solution (Millipore, Billerica, MA, USA) for 20 min. at room temperature, blocked and re-probed for bactin as a loading control. The primary antibodies used were b-actin

Proteinase K protection assay
Exosomes of approximately 1.5 9 10 7 cells were isolated by differential ultracentrifugation and resuspended in 20 ll PBS. Samples were incubated in either PBS or 10 lg/ml Proteinase K in PBS, with or without the presence of 1% Triton X-100, in a final volume of 20 ll per sample for 1 hr at 37°C. The assay was stopped by addition of 20 ll 95°C 29 Laemmli Buffer with 10 mM DTT. After 5 min. incubation at 95°C, samples were used for immunoblot analysis.

Electron microscopy
Transmission electron microscopy was performed as described [8]. After sucrose gradient centrifugation, fractions with a density of 1.10-1.12 g/ml were pooled and centrifuged at 100,000 9 g to collect exosomes. The pellet was resuspended in PBS and carbon-coated Formvar filmed grids were placed on top for 20 min. After three washes with 0.15% glycine in PBS and a 0.1% BSA in PBS wash grids were placed on 5 ll 1% BSA in PBS with 1:100 diluted anti-LOXL2 antibody for 20 min., washed four times with 0.1% BSA in PBS and incubated on a drop of 1% BSA in PBS containing rabbit anti-goat polyclonal antibody (Santa Cruz Biotechnology) for 20 min., and were labelled by incubation on PBS containing 1% BSA and 10 nm gold beads coupled to Protein A for 20 min. Vesicles were fixed by incubation in 1% glutaraldehyde in PBS for 5 min. washed twice with PBS and four times with distilled water. Subsequently, grids were placed on ice-cold 1.8% methylcellulose (25 Ctp)/0.4% Uranyl acetate for 5 min., and vesicles were visualized using a FEI Tecnai 12 (FEI, Hillsboro, OR, USA) transmission electron microscope.

Fluorescence microscopy
Human microvascular endothelial cells overexpressing eGFP were cultured for 48 hrs on Nunc Lab-Tek II 8-well chamber slides (Thermo-Fisher) at standard culturing conditions. Afterwards, cells were fixed for 30 min. at room temperature in 4% PFA-PBS. Cells were then washed twice with PBS and stained with 1 ug/ml DAPI (ThermoFisher) for 15 min. at room temperature, followed by two more washing steps with PBS. The Lab-Tek chamber slides were then removed and a cover-slip was applied using Fluoromount-G Slide Mounting Medium (Thermo-Fisher). Fluorescent images were taken using an Olympus BX53 microscope with a DP71 digital camera.

Fluorometric lysyl oxidase assay
Enzymatic activity of exosomal LOXL2 on 1,5-diaminopentane (MP Biomedicals, Santa Ana, CA, USA) was determined by measuring horseradish peroxidase type 2 (Sigma-Aldrich) mediated conversion of Amplex Red (ThermoFisher) to resorufin [19]. A substrate mixture [50 mM sodium borate buffer (pH = 8.2), 2.4 M Urea, 20 uM Amplex Red, 20 mM 1,5-diaminopentane] was mixed with an enzyme mixture [50 mM sodium borate buffer (pH = 8.2) containing 2 U/ml horseradish peroxidase type2] in a 1:1 ratio. Recombinant human LOXL2 (rhLOXL2) (Sino Biological Inc., Beijing, China) was resuspended in the same sodium borate buffer and used for a standard curve. Isolated exosomes were resuspended in PBS, protein concentrations were determined by BCA protein assay, and 5 lg exosomes was analysed per reaction. Assays were performed in a 100 ll volume on a Fluoroskan Ascent FL (ThermoFisher) in kinetics mode with excitation at 544 nm and emission at 570 nm. Measurements were made at a 1 min. interval for 2.5 hrs, at 37°C. The slope of the progress curve was determined over a 30 min. timeframe in the linear region of the curve, and corrected for background activity by deduction of the slope of the same samples analysed in parallel in the presence of 500 lM BAPN (Sigma-Aldrich).

Collagen gel contraction assay
Adult human dermal fibroblasts cells were cultured for 48 hrs in serumfree medium (DMEM containing 100 U/ml penicillin and 100 lg/ml streptomycin), and collected by incubation with TrypLE Express (Ther-moFisher) for 5 min. at 37°C. Cells were subsequently pooled using serum-free medium and pelleted by centrifugation for 8 min. at 400 9 g. After resuspension in serum-free medium, cells were added to a mixture of Rat Tail Collagen I (ThermoFisher), 29 modified eagle medium (MEM) (Life Technologies, Carlsbad, CA, USA), sterilized Milli-Q water and 1N NaOH, with a final concentration of 1 mg/ml Collagen I, 19 MEM, 0.2 million cells/ml, and physiological pH. After addition of 20 lg/ml exosomes in PBS, 2 lg/ml rhLOXL2, 500 lM BAPN, or a similar volume of PBS for control conditions, 0.5 ml gel matrices were made in 24-well plate wells. Gels were allowed to set for 30 min. at 37°C, and were then transferred to 12-well plate wells containing 1 ml 19 MEM. Collagen gels were incubated for 48 hrs at 37°C, 5% CO 2 , and pictures were taken every 12 hrs using the Molecular Image Chemi-Doc XRS system (Bio-Rad). Gel surfaces were determined with ImageJ software using the polygon selection tool, followed by an area measurement. Collagen contraction was calculated using the following equation: Collagen contraction ¼ 1 À Collagen gel size at t ¼ 36 hrs Collagen gel size at t ¼ 0 hrs

Statistical analyses
All data are expressed as mean AE SD. Student's t-test or one-way ANOVA with Sidak's multiple comparisons post-hoc analysis was used to compare groups, P < 0.05 was considered significant.

Hypoxia results in increased exosomal LOXL2 abundance
We have previously shown that cellular stress is represented in ECderived exosomes [8]. One of the exosomal proteins that was increased upon cellular hypoxic stress was the LOX family member LOXL2. To quantify the observed increase in exosomal LOXL2 abundance, HMEC-1 were cultured in control-and hypoxic conditions, as was confirmed by immunoblot analysis for HIF-1a and LOXL2 (Fig. 1A), after which exosomes from these cells were isolated and immunoblotted for LOXL2, using b-actin, which we previously showed to remain unchanged upon hypoxic stimulation in EC-derived exosomes [8], as a loading control (Fig. 1B). Densitometric analysis showed a twofold increase of LOXL2 (Fig. 1C), in line with the previously performed quantitative proteomics analysis [8]. To further illustrate that LOXL2 is indeed associated with exosomes, sucrose density gradient and subsequent immunoblot analysis were performed, showing that LOXL2 and the exosome marker Flotillin-1 are both detected at a density around 1.10 g/ml (Fig. 1D).

Exosomal LOXL2 is located on the exterior of exosomes
To determine the location of LOXL2, intact EC-derived exosomes were treated with proteinase K, with or without TX-100 to permeabilize exosome membranes ( Fig. 2A), using exosomal platelet endothelial cell adhesion molecule (PECAM-1) and GAPDH as controls for membrane and cytosolic proteins, respectively (Fig. 2B). If exosomes were left untreated, immunoblotting analysis showed presence of LOXL2, PECAM-1 (membrane protein control) and GAPDH (cytoplasmic protein control), all associated with exosomes as shown by sucrose density gradient analysis (Fig. 2C). Proteinase K treatment digested both LOXL2 and PECAM-1, whereas GAPDH, which is present in the interior of exosomes and thus protected by the lipid bi-layer, was not digested. Permeabilization of the exosome membrane by TX-100 makes GAPDH accessible and thus susceptible to proteinase K, resulting in its degradation. This demonstrates that exosomal LOXL2 is present on the exterior of the vesicle membrane, placing it in the same compartment as the ECM. These findings were confirmed by electron microscopy (Fig. 2D).

Hypoxia results in increased lysyl oxidase and collagen crosslinking activity of EC-derived exosomes
Using an in vitro fluorometric LOX activity assay with 1,5-pentanediamine as a substrate [19], the rate of LOXL2-catalysed oxidative deamination was measured. Exosomes from EC cultured in hypoxic conditions showed a 1.5-fold increase of LOX activity as compared to exosomes derived from EC cultured in control conditions (Fig. 3A).
To investigate exosomal LOXL2-mediated crosslinking of collagen, we used a collagen gel contraction assay in which increased ECM stiffness as a result of collagen crosslinking by LOXL2 results in fibroblast activation and subsequent gel contraction (Fig. 3B) [20]. In the presence of rhLOXL2, collagen gel contraction was increased, and this effect was fully inhibited by the LOX crosslinking inhibitor BAPN fumarate (Fig. 3C). We then analysed the effect of EC-derived exosomes in the collagen gel contraction assay (Fig. 3D). Exosomes derived from EC cultured in hypoxic conditions induced an increase in collagen contraction after 36 hrs (Fig. 3E), which was abolished by addition of BAPN (Fig. 3F).

LOXL2 overexpression results in increased exosomal lysyl oxidase activity
Human microvascular endothelial cells were transfected by with a pHAGE2 lentiviral vector containing a LOXL2 open reading frame (+LOXL2), or containing eGFP (+GFP) as a control. LOXL2 overexpression was confirmed by qPCR, showing an eightfold increase of LOXL2 mRNA (Fig. 4A), and immunoblotting confirmed 2.7-fold increase of LOXL2 protein abundance in exo-somes (Fig. 4B), which corresponded well with the approximately threefold higher in vitro LOX activity compared to control (+GFP)-EC derived exosomes (Fig. 4C). In control (+GFP) cells eGFP expression was confirmed by fluorescence microscopy (Fig. S1). LOXL2-high exosomes were also used for a collagen gel contraction assay (Fig. 4D). Similar to exosomes derived from EC cultured in hypoxic conditions, LOXL2-high exosomes increased collagen gel contraction, whereas GFP-high exosomes did not (Fig. 4E), and this effect was abolished by addition of BAPN (Fig. 4F).
Finally, to determine whether the increased activity observed in hypoxic exosomes was LOXL2-mediated, LOX activity of LOXL2-low exosomes isolated in normoxic (control) and hypoxic conditions were analysed (Fig. 5F). LOXL2 knockdown resulted in a 40% reduced LOX activity in EC-derived exosomes, which did not increase upon hypoxic stimulation of their producing cells. In a collagen gel contraction assay LOXL2 knockdown exosomes isolated in control and hypoxic conditions (Fig. 5G) did not significantly affect gel contraction (Fig. 5H). Altogether, these findings show that ECM crosslinking by EC-derived exosomes is mediated by LOXL2, under the regulation of hypoxia.
Effects of LOXL2 on processes like angiogenesis, fibroblast activation and pre-metastatic niche formation on the other hand mainly depend on its extracellular role in ECM crosslinking [7,20,25]. Regulatory roles for exosomes have been described in all these processes. Endothelial cell-derived exosomes increase angiogenesis by both regulating Notch-dependent tip cell formation and repression of senescence of ECs [18,27]. Upon hypoxic stimulation, epithelial cellderived exosomes are capable of activating fibroblasts by transformning growth factor-b signalling [28], and tumour exosomes affect the pre-metastatic niche by transfer of miRNAs [29]. Interestingly, LOXL2 has also been found in proteomic analysis of extracellular vesicles of squamous carcinoma cells [30] and human umbilical vein endothelial cell (HUVEC)-derived exosomes [31], and both LOXL2 and LOX were found in exosomes isolated from glioblastoma cells [32,33]. However, a potential role of exosomal LOXL2 in the aforementioned processes has thus far remained uninvestigated.
Here, we describe for the first time that exosomal LOXL2 is enzymatically active and can catalyse the crosslinking of collagen in the ECM. Although our assays measure total activity of all LOX family members the HMEC-1 derived exosomes make an excellent model system to investigate the role of exosomal LOXL2 as our previous proteomics analysis revealed that LOXL2 is the only member of the LOX family that can be detected in HMEC-1 derived exosomes [8]. Using this model system, we show a twofold increase of LOXL2 abundance in EC-derived exosomes. Since LOXL2 expression is at least in part regulated by HIF-1a [34], increased expression of LOXL2 as a response to hypoxia is not unexpected. The presence and upregulation of LOXL2 on the surface of EC-derived exosomes was particularly interesting, since LOXL2 is generally known as a soluble secreted enzyme [35]. A yeast-two hybrid screen performed by Hollosi et al. revealed that several ECM proteins may interact with LOXL2, including fibronectin [1]. Proteomic analysis of HMEC-1 derived exosomes revealed presence of a large number of ECM proteins, including fibronectin, which was also up-regulated twofold in exosomes in hypoxic conditions [8]. Additionally, it has been shown that proteolytic activation of LOX is regulated through interaction with fibronectin [36]. These findings suggest that exosomal LOXL2 is present on the exterior of exosomes as a result of interactions with exosome-associated ECM components.
Despite a twofold increase of LOXL2 abundance in hypoxic conditions, we observed an only approximately 1.5-fold increase of LOX activity. Given the complex nature of exosomes, it is possible that they contain several components that might interfere with the in vitro assays used to measure LOX activity. For example, ECderived exosomes also contain Semenogelin, a reactive oxygen species production inhibitor [37] that is also up-regulated in HMEC-1 derived exosomes in hypoxic conditions [8]. Since the Amplex Red based fluorometric assay used is based on reactive oxygen species production as a result of deamination of the substrate [19], it is possible that presence of such exosomal proteins results in an underrepresentation of enzymatic activity. However, in both assays, the observed increase in LOX activity under hypoxic conditions was reduced by LOXL2 knockdown in the exosome-producing cells, confirming that LOXL2 was indeed responsible for the observed LOX activity.
Altogether, these data indicate that crosslinking components of the ECM can be added to the functional repertoire of exosomes. Given the known role of LOXs in processes like angiogenesis, fibrosis and pre-metastatic niche formation these findings merit additional studies on exosomal LOX activity, as players in healing responses and in aberrant ECM remodelling in pathologies, as well as therapeutic applications and targets.