Subendothelial Matrix Stiffening by Lysyl Oxidase Enhances RAGE-Mediated Retinal Endothelial Activation in Diabetes

Diabetic retinopathy (DR) is a common microvascular complication of diabetes that remains the leading cause of preventable vison loss in the working-age population (1,2). In the early stages of DR, hyperglycemia leads to retinal inflammation that activates retinal vascular endothelial cells (ECs). Activated retinal ECs, in turn, upregulate intercellular adhesion molecule-1 (ICAM-1), a crucial proinflammatory EC adhesion molecule that binds circulating leukocytes and, thereby, contributes to the development of early vascular lesions of DR (namely, vascular atrophy and hyperpermeability) (3,4). However, precisely how retinal ECs become activated and express ICAM-1 in diabetes remains insufficiently understood.

Recent findings have revealed that, under high-glucose (HG) conditions, retinal ECs increase the expression of lysyl oxidase (LOX) (5,6), a copper-dependent amine oxidase that cross-links and stiffens extracellular matrix (5,7). These findings correlate with LOX upregulation observed in the retina and vitreous of diabetic animals and humans, respectively (5,6,8). Crucially, our studies showed that LOX-mediated subendothelial matrix stiffening alone is sufficient to promote retinal EC activation (ICAM-1 expression) in vitro (5), a finding that aligns with the proinflammatory effects of matrix stiffening on cardiovascular and lung ECs in atherosclerosis and sepsis, respectively (7,9). This ability of subendothelial matrix stiffening to control EC behavior is attributed to endothelial mechanotransduction, a process by which ECs transduce matrix-based mechanical, structural, and adhesive cues into intracellular biochemical signals that regulate cell behavior at both translational and transcriptional levels (5,10–12). Given this newly identified role of LOX in the mechanical control of retinal EC activation, it becomes important to elucidate both the regulation and proinflammatory role of retinal LOX in diabetes.

LOX is known to be upregulated in inflammatory conditions such as sepsis and atherosclerosis (7,13). Because HG also promotes proinflammatory signaling by upregulating receptor for advanced glycation end products (RAGE), we asked whether HG-induced LOX upregulation in retinal ECs is mediated by RAGE. RAGE is the chief receptor for advanced glycation end products (AGEs), which are formed by nonenzymatic glycation of proteins and lipids in diabetes and aging. Notably, the proinflammatory AGE and RAGE (hereafter, AGE/RAGE) signaling is significantly enhanced in the retinas of patients with diabetes, where it contributes to retinal vascular inflammation and degeneration by inducing NF-κB–dependent ICAM-1 upregulation, leukostasis, breakdown of the blood-retinal barrier, and EC and pericyte apoptosis (14). Despite their important implications in DR pathogenesis, direct inhibition of AGE or RAGE has proven challenging in clinical trials, due to drug safety issues and low efficacy of the drug (15,16). Thus, determining the role of AGE/RAGE signaling in retinal endothelial LOX expression may not only provide mechanistic insights into LOX regulation in diabetes but, crucially, also implicate LOX as an alternative (downstream) target to block AGE/RAGE signaling.

Here, we report that HG and methylglyoxal (MGO), precursors of AGE formation and subsequent RAGE activation, increase LOX expression and activity in retinal EC cultures in a RAGE-dependent manner. Importantly, inhibiting LOX alone blocked the ability of MGO and RAGE (hereafter, AGE/RAGE) to cause EC activation (ICAM-1 expression and monocyte–EC adhesion) and subendothelial matrix remodeling (namely, increased matrix stiffness and organization). Finally, using our mouse model of diabetes and retinal EC cultures, we demonstrate that LOX feeds back to increase RAGE expression, thus providing a potential mechanism for both LOX autoregulation and its observed proinflammatory effects (5).

All animal procedures were performed in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and approved by the University of California Riverside Institutional Animal Care and Use Committee. Induction of diabetes in adult mice and treatment of diabetic mice with LOX inhibitor β-aminopropionitrile (BAPN) are described in the Supplementary Research Design and Methods.

Human retinal endothelial cells (HRECs), mouse retinal ECs, and human THP-1 monocytes were purchased from commercial vendors and cultured per vendor recommendation, which is described in the Supplementary Research Design and Methods, as are all in vitro treatments.

mRNA expression was assessed using gene- or species-specific TaqMan primers per our standard reverse transcription-quantitative PCR (RT-qPCR) protocol (17), which is described in the Supplementary Research Design and Methods.

Expression levels of target proteins were assessed using target-specific primary antibodies per our standard Western blotting protocol (17), which is described in the Supplementary Research Design and Methods.

Levels of MGO adduct methylglyoxal-hydroimidazolone (MG-H1) were measured in retinal and cell-culture lysates using a competitive ELISA (catalog ab238543; Abcam, Cambridge, MA) per the manufacturer protocol.

Monocyte–EC adhesion and ICAM-1 clustering index were assessed according to our previously reported protocols (5,12), which are described in the Supplementary Research Design and Methods.

LOX activity was measured from HREC culture supernatant using our previously reported protocol (5), which is described in the Supplementary Research Design and Methods.

HRECs (passage 5–6) were transfected with LOX siRNA (catalog hs.Ri.LOX.13, Integrated DNA Technologies, Inc.) using Lipofectamine RNAiMax transfection reagent (catalog 13778-030, Invitrogen, Carlsbad, CA), per the manufacturer’s protocol. Nontargeting control siRNA was used as a control, and transfection efficiency was evaluated by quantifying LOX mRNA.

For mRNA measurement, retinal vessels were isolated from fresh (unfixed) eyes using an established protocol (18), which is described in the Supplementary Research Design and Methods.

An HREC permeability assay was performed per a published protocol (19), which is described in the Supplementary Research Design and Methods.

Decellularized subendothelial matrices were obtained from HREC cultures grown in normal glucose or MGO ± BAPN medium, using our previously reported protocols (5,20) that are described in the Supplementary Research Design and Methods.

A NanoWizard 4 XP BioScience atomic force microscope (Bruker Nanotechnologies, Santa Barbara, CA), fitted with a precalibrated PFQNM-LC-A-CAL probe (spring constant 0.075 N/m) containing a 70 nm–radius hemispherical tip (Bruker AFM Probes) and coupled with a Zeiss Axiovert phase-contrast microscope, was used for stiffness measurement and topographical scanning of the subendothelial matrix, as described in the Supplementary Research Design and Methods.

Fibronectin-coated, polyacrylamide-based synthetic matrices of tunable stiffness were fabricated according to published protocols (5,11,12) that are described in the Supplementary Research Design and Methods.

All data were analyzed using GraphPad Prism 6.01 (GraphPad Software, San Diego, CA), as described in the Supplementary Research Design and Methods.

The data sets generated during this study are available from the corresponding author on reasonable request. No applicable resources were generated or analyzed during this study.

We have previously shown that diabetes and HG increase LOX expression in the retina and HRECs, respectively, which, in turn, promotes EC activation (5). To test the hypothesis that this LOX upregulation in retinal ECs is mediated by AGE/RAGE, we first assessed the degree to which HG increases AGE/RAGE levels in these cells. As shown in Fig. 1A, HG (but not mannitol, which is used as an osmotic control) treatment of HREC cultures caused a 1.5-fold (P < 0.001) increase in RAGE expression, which was consistent with an ∼1.5-fold increase (P < 0.05) in retinal RAGE levels in diabetic mice (Fig. 1B). Notably, this HG-induced RAGE upregulation was associated with a 1.3-fold increase in the formation of MG-H1, an AGE that is implicated in DR pathogenesis and that activates RAGE (21), in both HREC cultures (P < 0.01; Fig. 1C) and retinas of diabetic mice (P < 0.05; Fig. 1D). Furthermore, consistent with the proinflammatory effects of AGE/RAGE interactions, we observed a concomitant increase in ICAM-1 expression in both HG-treated HRECs (approximately twofold; P < 0.001; Fig. 1E) and retinas of diabetic mice (∼1.7-fold; P < 0.01; Fig. 1F).

To directly implicate RAGE in HG-induced LOX upregulation and associated retinal EC activation, we treated HRECs with varying doses of MGO, a highly reactive AGE precursor that forms MG-H1 (AGE) adducts and thereby selectively binds to and upregulates RAGE (22). Our findings revealed that 10 μmol/L MGO maximally increases the levels of MG-H1 by 1.45-fold (P < 0.0001; Supplementary Fig. 1A), RAGE mRNA by approximately twofold (P < 0.05; Supplementary Fig. 1B), protein by 1.75-fold (P < 0.01; Fig. 1G), and ICAM-1 protein by 2.5-fold (P < 0.01; Fig. 1H) in HRECs. Consistent with an increase in ICAM-1 expression, 10 μmol/L MGO-treated HRECs exhibited an approximately fourfold greater (P < 0.001) monocyte adhesion than did untreated ECs (Fig. 1I).

Importantly, this trend in MGO-induced RAGE and ICAM-1 upregulation was mirrored by LOX expression. Specifically, treatment of HRECs with varying doses of MGO produced maximal increase in LOX mRNA (P < 0.0001; Fig. 1J), protein (P < 0.001; Fig. 1K), and activity (P < 0.01; Fig. 1L) at the same 10 μmol/L dose that maximally increased RAGE expression (Fig. 1G and Supplementary Fig. 1B). This MGO dose also increased, by twofold (P < 0.01), the levels of LOX propeptide, which is released during LOX processing and contributes to retinal EC death associated with DR (23) (Supplementary Fig. 2).

LOX expression increases under proinflammatory conditions (7,13). Because HG and MGO simultaneously upregulate LOX and proinflammatory RAGE (Fig. 1) (5), we asked whether HG and MGO increase retinal endothelial LOX via RAGE. Indeed, addition of a RAGE-specific inhibitor, FPS-ZM1 (24), to HG- and MGO-treated HRECs prevented the significant LOX upregulation seen in these conditions (Fig. 2A and B). Because the AGE/RAGE signaling activates the proinflammatory transcription factor NF-κB, we further asked whether RAGE increases LOX via NF-κB. As Fig. 2C shows, pharmacological inhibition of NF-κB in MGO-treated HRECs prevented the expected increase in LOX mRNA expression.

AGE/RAGE (Fig. 1) and LOX (5) activate retinal ECs, leading to ICAM-1 upregulation. Because AGE/RAGE enhances LOX (Figs. 1 and 2), we asked whether they promote retinal endothelial activation via LOX. Our findings revealed that simultaneously treating HRECs with MGO and BAPN, a pharmacological and irreversible inhibitor of LOX activity (25), prevents the MGO-induced increase in ICAM-1 mRNA (Supplementary Fig. 3) and protein expression (Fig. 3A). The same trend was seen with mouse retinal EC cultures (Supplementary Fig. 4A). To rule out any potential nonspecific effects of BAPN, we alternatively reduced LOX in MGO-treated HRECs using siRNA (Supplementary Fig. 5). Our RT-qPCR measurements revealed that LOX-specific siRNA (but not control siRNA) inhibits the MGO-induced increase in ICAM-1 mRNA by 75% (P < 0.01; Fig. 3B).

Importantly, and consistent with these in vitro findings, LOX inhibition using BAPN almost completely blocked the 1.6-fold increase (P < 0.05) in ICAM-1 mRNA in the retinal vessels of diabetic mice (Fig. 3C). Given this requirement of LOX in both MGO- and diabetes-induced increase in ICAM-1 expression in HRECs and mouse retinal vessels, respectively, BAPN predictably reduced, by 75% (P < 0.0001), the ICAM-1–dependent adhesion of THP-1 monocytes to MGO-treated HRECs (Fig. 3D). Immunolabeling of the monocyte–EC adhesion sites with anti-phospho–ICAM-1, which labels activated ICAM-1, revealed that LOX inhibition also decreases, by 68% (P < 0.0001), MGO-induced “clustering” of activated ICAM-1 around adherent monocytes (Fig. 3E). Finally, consistent with the ability of LOX to induce retinal EC activation, we show that LOX also mediates MGO-induced increase in HREC permeability (Supplementary Fig. 6).

MGO-associated AGEs (26) and LOX (5) are implicated in the crosslinking and stiffening of subendothelial matrix. Because MGO enhances LOX (Fig. 2), we asked whether MGO-induced subendothelial matrix crosslinking and stiffening are mediated by LOX. Decellularization of the 15-day HREC cultures (Supplementary Fig. 7), followed by immunofluorescent anti-LOX labeling and fluorescence intensity analysis of the residual subendothelial matrix revealed that MGO markedly increases, by threefold (P < 0.001), matrix-localized LOX levels (Fig. 4A), which was reduced by 70% (P < 0.01) with BAPN treatment. Notably, this MGO-induced increase in matrix LOX caused a predictable 2.7-fold increase (P < 0.0001) in subendothelial matrix stiffness that was nearly completely inhibited by BAPN (Fig. 4B). Atomic force microscopy (AFM) imaging also revealed that MGO increases matrix fiber density by 2.2-fold (P < 0.0001) and aligns fibers in a preferential orientation (40–90°), which are prevented with BAPN treatment (Fig. 4C).

We next asked whether matrix-localized LOX and associated matrix stiffening can mediate AGE/RAGE-induced EC activation independent of any potential effects of soluble LOX. To this end, we plated untreated HRECs on decellularized matrices obtained from preceding untreated or MGO ± BAPN–treated EC cultures and assessed ICAM-1 expression (Supplementary Fig. 7). Under these conditions in which elevated LOX is present only in the predeposited matrix, untreated HRECs plated on decellularized matrix from preceding MGO-treated cultures exhibited a substantial (P < 0.01) increase in ICAM-1 mRNA expression, which was significantly inhibited (P < 0.05) in cells plated on LOX-inhibited matrices (Fig. 5A).

Because MGO can simultaneously alter matrix stiffness (Fig. 4B) and density or composition (27), the aforementioned effect of MGO-treated matrix on fresh HRECs may result from changes in both mechanical (stiffness) and biochemical (ligand density or composition) cues from the matrix. To evaluate the independent effect of matrix stiffness on HREC activation, HRECs were plated on synthetic matrices of tunable stiffness that were fabricated to mimic the subendothelial matrix stiffness of untreated (1 kPa) or MGO-treated (2.5 kPa) conditions (Fig. 4B) and assessed for ICAM-1 expression and monocyte–EC adhesion in the presence or absence of NF-κB inhibitor. When compared with cells plated on the normal 1 kPa matrix, those plated on the stiff (2.5 kPa) matrix exhibited a 1.7-fold increase (P < 0.001) in ICAM-1 mRNA levels (Fig. 5B) and a 1.2-fold increase in monocyte adhesion (Fig. 5C). Importantly, NF-κB inhibition in HRECs prevented these stiffness-induced activation responses (Fig. 5B and C).

Past studies have shown that inhibition of LOX activity reduces LOX expression, thus indicating that LOX can autoregulate its expression (28,29). Our findings agree with these observations: inhibiting LOX activity in HRECs with BAPN (Supplementary Fig. 8A) completely blocked the MGO-induced increase in LOX mRNA (Supplementary Fig. 8B) and protein (Supplementary Fig. 8C) expression. The same trend was seen with mouse retinal EC cultures (Supplementary Fig. 4B). Consistent with these in vitro data, BAPN also blocked the diabetes-induced increase in LOX mRNA in mouse retinal vessels (Fig. 6A). Because RAGE promotes LOX expression (Fig. 2), we wondered whether LOX autoregulates its expression by feeding back to increase RAGE. Indeed, inhibiting LOX activity in HRECs with BAPN completely blocked the MGO-induced increase in RAGE mRNA (Fig. 6B) and protein (Fig. 6C) expression, a trend that was mirrored in mouse retinal ECs (Supplementary Fig. 4C). Notably, siRNA-based silencing of LOX had the same inhibitory effect on RAGE expression, which was not seen with control siRNA (Fig. 6D).

The ability of LOX to feedback and promote RAGE expression was validated in vivo where LOX inhibitor BAPN completed inhibited the diabetes-induced increase in mouse retinal vascular RAGE mRNA expression (Fig. 6E).

To determine whether LOX-dependent matrix stiffening is sufficient alone to promote autoregulation (via enhancement of RAGE expression) independent of any potential effects of soluble LOX, we plated fresh HRECs on decellularized matrices obtained from preceding untreated or MGO ± BAPN–treated EC cultures and assessed RAGE and LOX expression (Supplementary Fig. 7). Our findings revealed that untreated HRECs plated on decellularized matrix from preceding MGO-treated cultures exhibit a significant increase in mRNA levels of both RAGE, by 1.7-fold, (P < 0.001; Fig. 6F) and LOX, by 1.6-fold (P < 0.001; Supplementary Fig. 9A), an effect that is completely blocked on LOX-inhibited matrices of normal stiffness.

To unequivocally confirm the role of subendothelial matrix stiffening in RAGE expression and, consequently, LOX autoregulation, HRECs were plated on the aforementioned normal (1 kPa) or stiff (2.5 kPa) synthetic matrices and assessed for RAGE and LOX mRNA. Similar to ICAM-1 expression (Fig. 5B), mRNA levels of both RAGE (Fig. 6G) and LOX (Supplementary Fig. 9B) increased significantly (by ∼1.6-fold and ∼1.5-fold, respectively; P < 0.001) on the stiff matrix, which was prevented by NF-κB inhibition.

Taken together, these data from decellularized and synthetic matrices indicate that LOX-dependent subendothelial matrix stiffening alone can feed back to enhance RAGE expression via NF-κB and, thereby, achieve LOX autoregulation.

Despite the crucial role of retinal EC activation in inflammation-mediated DR pathogenesis (3,4,30), the underlying mechanisms remain poorly understood. We previously discovered that LOX-mediated subendothelial matrix stiffening contributes significantly to retinal EC activation under HG conditions (5). The present study further extends our understanding of this mechanical control of retinal EC activation, by shedding light on both the regulation and proinflammatory role of LOX. Specifically, we show that retinal endothelial LOX expression under hyperglycemic conditions is regulated by RAGE, whose own expression, in turn, is enhanced by LOX-mediated subendothelial matrix stiffening. This sets in motion a mechanically regulated feed-forward loop that sustains retinal EC activation in diabetes. This is an important finding because it provides new insights into retinal endothelial LOX regulation in diabetes, further highlights the role of mechanical cues in retinal EC activation, and identifies LOX as a potential alternative (downstream) target to block AGE/RAGE signaling that is otherwise difficult to achieve (15,16).

LOX-mediated subendothelial matrix stiffening is strongly implicated in vascular inflammation associated with debilitating conditions such as atherosclerosis and sepsis (7,13,31). In line with these findings, we recently reported that LOX-mediated subendothelial matrix stiffening contributes significantly to HG-induced retinal EC activation (ICAM-1 expression) (5), a rate-limiting step in the development of retinal vascular lesions in DR (3,4,30). Our finding has been supported by an independent study in which LOX knockout mice did not develop the inflammation-dependent vascular lesions of DR (32). Despite these strong implications of retinal endothelial LOX in DR, a mechanistic understanding of its regulation and proinflammatory role in diabetes remains absent.

Past and current findings reveal that, under HG and diabetic conditions, LOX-mediated retinal EC activation (5) is associated with upregulation of proinflammatory RAGE, which promotes the development of inflammation-mediated retinal vascular lesions (33). LOX-mediated vascular inflammation in other inflammatory conditions, such as sepsis and atherosclerosis (7,13), is also associated with RAGE overexpression (34,35). Thus, we asked whether RAGE could upregulate retinal endothelial LOX. Although RAGE is a multiligand receptor, its primary ligand are the AGEs that result from nonenzymatic glycation of proteins and lipids caused by HG and the highly reactive glycolysis metabolites such as MGO. Indeed, MG-H1, an MGO-derived AGE, is abundant in the retinas of diabetic rodents (33), which is consistent with our in vitro and in vivo findings. These findings, coupled with past reports of elevated MGO levels in patients with diabetes (36,37), led us to choose the AGE precursor MGO as the RAGE ligand for the present study.

We found that MGO treatment maximally increases HREC LOX expression at the same 10 μmol/L dose that causes peak RAGE expression and ICAM-1 expression. That MGO failed to increase RAGE-dependent LOX expression at higher doses may be attributed to an increase in the levels of soluble RAGE, which is produced from alternative splicing of the RAGE gene or proteolytic cleavage of membrane-bound RAGE (38,39) and competitively binding to AGEs, thus inhibiting membrane RAGE signaling. Subsequent studies confirmed that the MGO-dependent increase in LOX levels is mediated by RAGE, because treating HRECs with a RAGE inhibitor blocked the MGO-induced LOX upregulation. This is, to our knowledge, the first direct evidence of LOX upregulation by MGO and RAGE and aligns with previous findings that NF-κB, a downstream effector of RAGE, increases LOX transcription in ECs and cancer cells (40,41). Interestingly, contrary to these findings, AGE-modified albumin suppressed LOX expression in mouse preosteoblasts (42), indicating that the LOX-enhancing effect of AGEs may depend on cell type and/or their degree of differentiation.

We have previously shown that LOX promotes retinal EC activation under HG conditions (5). Our present findings indicate that LOX achieves this effect, in large part, by mediating the proinflammatory effects of RAGE that is upregulated by HG treatment. That LOX mediates the proinflammatory effects of RAGE was further confirmed when LOX inhibition blocked the MGO-induced increase in HREC ICAM-1 expression, monocyte–EC adhesion, and permeability in vitro, as well as diabetes-induced retinal vascular ICAM-1 expression in vivo. Independent studies have shown that LOX inhibition blocks the proinflammatory effects of lipopolysaccharide on pulmonary ECs (7,31) and ApoE knockout on aortic ECs (13). LOX activity is also significantly increased in aortas of atheroprone animals (43). Thus, LOX appears to play a central role in vascular inflammation caused by multiple risk factors. Interestingly, we show that LOX inhibition also blocked the MGO-induced ICAM-1 “clustering” at the site of monocyte–HREC adhesion. We have previously shown that endothelial ICAM-1 clustering around adherent monocytes can be induced by subendothelial matrix stiffening (12). Consistent with this finding, our AFM measurements show that LOX inhibition also significantly reduced MGO-induced subendothelial matrix stiffening. That LOX inhibition did not return subendothelial matrix stiffness to the basal (untreated) level indicates that MGO also, albeit to a lesser extent, directly crosslinks and stiffens matrix by forming AGE adducts. Overall, these findings demonstrate that LOX is a crucial mediator of both the biochemical (retinal EC activation) and mechanical (subendothelial stiffening) effects of AGE/RAGE.

LOX exists in both matrix-localized and soluble forms. Although soluble LOX has primarily been implicated in cell migration (44), the aforementioned proinflammatory effects of LOX have been attributed to its matrix-localized form, where it promotes subendothelial matrix stiffening and, thereby, alters EC mechanotransduction (5,7,13,31). Consistent with the latter reports, here we found that MGO-induced LOX upregulation is also significantly localized to the subendothelial matrix, where it causes matrix stiffening. To unequivocally confirm the role of LOX-mediated subendothelial matrix stiffening in MGO-induced retinal EC activation, we cultured HRECs on decellularized matrices obtained from long-term HREC cultures or on polyacrylamide-based synthetic matrices of tunable stiffness that confer independent control of matrix stiffness versus composition (5,12). These studies revealed that subendothelial matrix stiffening actively promotes MGO-induced HREC activation. We had previously shown that HG-induced HREC activation is also similarly dependent on LOX-mediated subendothelial matrix stiffening (5). Thus, these findings underscore the pivotal role that LOX plays in the matrix-mediated mechanical control of retinal EC activation associated with key DR risk factors. Notably, in vitro studies indicate that LOX may also promote retinal vascular basement membrane (BM) thickening (5), which contributes to the development of retinal vascular lesions of DR (45). The excess fibronectin in thicker retinal vascular BM can potentially enhance LOX-dependent collagen IV crosslinking (46,47), thereby further increasing BM thickness. Thus, we speculate that LOX contributes to DR pathogenesis via both stiffening and thickening of the retinal vascular BM.

Although our findings reveal the RAGE-mediated regulation and proinflammatory roles of retinal endothelial LOX in diabetes, they do not explain precisely how LOX promotes retinal EC activation. In this regard, we interestingly found that inhibiting LOX activity inhibits its own mRNA expression in MGO-treated retinal EC cultures and retinal vessels of diabetic mice, thereby indicating that LOX can autoregulate its expression. This finding is consistent with previous reports of LOX autoregulation in animal models of tissue fibrosis (28,29). Notably, in contrast to transcription factors, proteins and enzymes such as LOX require the involvement of additional factor(s) for their autoregulation (48). We found that additional factor for LOX autoregulation is RAGE. Specifically, we found that LOX autoregulates its expression by feeding back to increase RAGE expression, because LOX inhibition completely blocked RAGE upregulation in both MGO-treated retinal EC cultures and retinal vessels of diabetic mice.

We further show that this LOX–RAGE feedback signaling is dependent on NF-κB activation by LOX-dependent subendothelial matrix stiffening. Collectively, these findings indicate that RAGE-regulated retinal endothelial LOX exerts proinflammatory effects in DR by increasing subendothelial matrix stiffness that, in turn, feeds back to enhance RAGE expression via activation of mechanosensitive NF-κB (Fig. 7).

Our finding that LOX-dependent subendothelial matrix stiffening both regulates and is regulated by RAGE is significant because it offers fresh mechanistic insights into the regulation and proinflammatory role of LOX in diabetes. Perhaps more crucially, the ability of LOX to regulate RAGE has important translational implications because LOX might serve as an alternative (downstream) target to block AGE/RAGE signaling in retinal ECs, thus overcoming the limitations of direct AGE/RAGE targeting in DR, such as drug toxicity and low efficacy (15,16). Furthermore, because subendothelial matrix stiffening can increase endothelial ρ/ROCK activity (9,12), which mediates the angiogenic and vascular hyperpermeability effects of VEGF (49,50), LOX inhibition may also suppress VEGF-mediated DR pathogenesis, thereby offering an alternative therapeutic approach for patients who do not respond well to anti-VEGF therapies. These therapeutic implications make LOX a promising drug target for the future clinical management of DR.

Funding. This work was supported by the National Eye Institute, NIH grants R01EY028242 (to K.G.), and R01EY033002 and R01EY022938 (to T.S.K.); Start-up Funds provided by the Doheny Eye Institute (to K.G.); The Stephen Ryan Initiative for Macular Research Special Grant from W.M. Keck Foundation (to Doheny Eye Institute); and the Ursula Mandel Fellowship and UCLA Graduate Council Diversity Fellowship (to I.S.T.). This work was also supported by an unrestricted grant from Research to Prevent Blindness, Inc., to the Department of Ophthalmology at UCLA.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.C. designed and performed experiments, analyzed data, and wrote the manuscript. I.S.T., M.A., and N.M. performed experiments and analyzed data. T.S.K. designed experiments, and K.G. conceived the idea, designed experiments, analyzed data, and wrote the manuscript. All authors reviewed, edited, and approved the manuscript. K.G. is the guarantor of this work and, as such, takes responsibility for the integrity and accuracy of the reported data.

Prior Presentation. A non–peer-reviewed version of this article was submitted to the bioRxiv preprint server (https://doi.org/10.1101/2022.08.31.505952) on 3 September 2022, and some of the original findings reported in this article were published in meeting abstracts (Invest Ophthalmol Vis Sci 2020;61:5409 and Invest Ophthalmol Vis Sci 2021;62:3028).