The lysyl oxidase like 2/3 enzymatic inhibitor, PXS‐5153A, reduces crosslinks and ameliorates fibrosis

Abstract Fibrosis is characterized by the excessive deposition of extracellular matrix and crosslinked proteins, in particular collagen and elastin, leading to tissue stiffening and disrupted organ function. Lysyl oxidases are key players during this process, as they initiate collagen crosslinking through the oxidation of the ε‐amino group of lysine or hydroxylysine on collagen side‐chains, which subsequently dimerize to form immature, or trimerize to form mature, collagen crosslinks. The role of LOXL2 in fibrosis and cancer is well documented, however the specific enzymatic function of LOXL2 and LOXL3 during disease is less clear. Herein, we describe the development of PXS‐5153A, a novel mechanism based, fast‐acting, dual LOXL2/LOXL3 inhibitor, which was used to interrogate the role of these enzymes in models of collagen crosslinking and fibrosis. PXS‐5153A dose‐dependently reduced LOXL2‐mediated collagen oxidation and collagen crosslinking in vitro. In two liver fibrosis models, carbon tetrachloride or streptozotocin/high fat diet‐induced, PXS‐5153A reduced disease severity and improved liver function by diminishing collagen content and collagen crosslinks. In myocardial infarction, PXS‐5153A improved cardiac output. Taken together these results demonstrate that, due to their crucial role in collagen crosslinking, inhibition of the enzymatic activities of LOXL2/LOXL3 represents an innovative therapeutic approach for the treatment of fibrosis.

development of effective therapeutic avenues that halt or reverse disease progression is urgently required.
In healthy tissues, collagens play a key role by providing strength, stability and integrity. 6 These properties are based on a highly organized molecular structure in which three collagen chains are interwoven into a triple helix. 6 In addition, collagen fibrils are further strengthened by covalent crosslinks formed enzymatically between the collagen molecules. 6,7 This process has been confirmed to be driven by lysyl oxidases, as when enzymatically active LOX was incubated (in vitro) with purified collagen it resulted in dihydroxylysinonorleucine (DHLNL) as well as pyridinoline (PYD) formation. 8 Specifically, lysyl oxidases oxidatively deaminate lysine and hydroxylysine residues in the telopeptide domains of the collagen molecule to form the corresponding aldehydes (allysines or hydroxyallysines) ( Figure S1). 7,[9][10][11] These aldehydes react spontaneously, either with other aldehydes or with unmodified lysine or hydroxylysine residues, thereby resulting in the formation of crosslinks. This process results in dimeric immature crosslinks, such as DHLNL (hydroxyallysine-derived crosslink) as well as hydroxylysinonorleucine (HLNL) (allysine-derived crosslink). The reaction of immature crosslinks with an additional aldehyde results in the formation of very stable trimeric, mature crosslinks PYD and deoxypyridinoline (DPD) (hydroxyallysine and allysine-derived crosslink, respectively). During fibrosis, the disproportionate accumulation of collagen leads to excessive maturation of crosslinks, resulting in the key characteristics of the disease: tissue stiffening, scarring and irreversibility.
Lysyl oxidases constitute a family of copper-dependent amine oxidases comprised of five members, lysyl oxidase (LOX) and lysyl oxidase-like 1-4 (LOXL 1-4). 2 They show a high degree of homology in the catalytic carboxy terminal end and more divergence in the rest of the sequence. 12 Lysyl oxidases regulate many biological processes including extracellular matrix (ECM) stabilization, cellular growth and homeostasis. Their protein expression has been positively correlated with fibrotic diseases in many different tissues including liver, lung and kidney. [13][14][15][16] In addition, certain members of the family -in particular LOX and LOXL2-have been widely associated with cancer progression and metastasis. 13,17,18 Early research with a functional antibody for LOXL2 (AB0023), demonstrated efficacy in various pre-clinical models of fibrosis and cancer. 13 Recently, PAT-1251 a selective small molecule inhibitor of LOXL2 also demonstrated potential as an anti-fibrotic agent in pre-clinical models. 19 Although the primary role of this family of enzymes is ECM remodelling, a number of extra-and intracellular functions have also been reported that are independent of the enzymatic activity. Extracellularly, LOXL2 has been shown to signal through β-integrin in cancer-associated fibroblasts 20 and intracellularly, LOXL2 and LOXL3 have been associated with epithelial to mesenchymal transition (EMT). [21][22][23] Additionally, LOXL2 is a key player for heterochromatin formation via Snail-dependent mechanisms. 21,22,24 Moreover, one study showed that LOXL2 is a negative regulator of Notch1 transcription, thereby attenuating epidermal differentiation. 25 Despite the extensive body of work on lysyl oxidases and the considerable therapeutic potential of LOXL2 and LOXL3 inhibition, no studies have provided unequivocal evidence for the relationship between the enzymatic activities of LOXL2/LOXL3, and the reduction of crosslinks resulting in therapeutic benefits in models of fibrosis. Herein, we set out to develop PXS-5153A, an innovative small molecule inhibitor with complete LOXL2/LOXL3 enzymatic inhibition -unlike the currently available antibody (simtuzimab)-with faster onset and higher potency for LOXL2/ LOXL3 than the small molecule inhibitor racemate of PAT-1251.
Using PXS-5153A, we were able to dissect the enzymatic function of LOXL2 and LOXL3 on the formation of different crosslink subtypes during fibrosis and examine the relevance of LOXL2/LOXL3 inhibition on disease severity as well as organ function recovery.

| Fluorometric enzymatic activity assays
The measurement of the enzymatic activity of all lysyl oxidase family members was based on the detection of hydrogen peroxide with an Amplex-Red oxidation assay, as described in Zhou et al 26 Recombinant human semicarbazide sensitive amine oxidase (SSAO/VAP1), diamine oxidase (DAO) and monoamine oxidases A and B (MAO-A and MAO-B) assays were performed as previously described. 27 For detailed enzymatic procedures, see supporting information.

| Off-target activity
PXS-5153A was tested at Eurofins Cerep Panlabs Taiwan, Ltd in the "Hit Profiling Screen", which tested 30 different targets.

| Collagen oxidation assay
Collagen oxidation assay was based on the release of hydrogen peroxide as previously described 27 and detailed in the supporting information. Briefly, collagen was combined with rhLOXL2 with or without the pan-lysyl oxidase inhibitor BAPN (β-aminoproprionitrile, 100 μmol/L, Sigma-Aldrich) or PXS-5153A. An Amplex reaction mixture was added into each well. The slope per minute of the kinetic curves for each sample was calculated using MARS data analysis software (BMG labtech) in the linear phase (between the 20 and 40 minute time points).

| In vitro crosslinking assay
About 200 μL of 3 mg/mL collagen (rat tail, type I, Thermo Fisher) was combined with 800 μL of 50 mmol/L sodium borate buffer (pH 8.2) and 20 nmol/L of rhLOXL2 (with or without inhibitor, PXS-5153A 200 nmol/ L). Enzyme/inhibitor were replenished daily for 5 days. Samples were incubated at 37°C and crosslinks were extracted on day 7.

| Protein, hydroxyproline and collagen crosslinking analysis
About 10mg of freeze-dried samples were reduced with NaBH 4 . Pellet then underwent acid hydrolysis in 6 mol/L HCl at 100°C for 24 hours. Hydroxyproline and crosslinks were extracted from the hydrolysate using an automated solid phase extraction system (Gilson GX-271 ASPECA system). After extraction and drying, hydroxyproline and crosslinks were analysed by UHPLC-ESI-MS/MS on a Thermo Dionex UHPLC and TSQ Endura triple quad mass spectrometer. Total protein was quantified in the samples using a commercially available kit (QuickZyme Biosciences, Leiden, The Netherlands).
For detailed procedures, see supporting information.

| CCl 4 -induced liver fibrosis
The study was performed by Pharmalegacy with approval from local ethics committee. Sprague Dawley rats were orally administered with 0.25 μL/g Carbon tetrachloride (CCl 4 ) in olive oil solution, starting from day 0, 3 times per week for 6 weeks. Animals were killed 48 hours after the last CCl 4 administration. PXS-5153A was given by oral gavage after 3 weeks of CCl 4 administration and continued throughout the remainder of the study at 3 mg/kg (low dose) or 10 mg/kg (high dose) once a day or 10 mg/kg (high dose) three times a week. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were assessed in the plasma. One lobe of the liver tissue was fixed in 10% formalin, stained for Sirius red and the percentage coverage area was measured. The remainder of the liver was snap frozen and used for protein, hydroxyproline and crosslink analysis.

| NASH-induced liver disease
The study was performed by Stelic MC, Inc. (Tokyo, Japan) with approval from local ethics committee. NASH was established in male C57/BL6 mice by a single subcutaneous injection of 200 μg streptozotocin (Sigma-Aldrich) after birth and with a high fat diet (CLEA Japan) ad libitum after 4 weeks of age (day 28 ± 2) until 14 weeks of age. Mice were orally administered with 10 mg/kg PXS-5153A once daily from 8 to 14 weeks of age. ALT levels were assessed in the plasma. One lobe of the liver tissue was fixed in 10% formalin, stained for Sirius red and the percentage coverage area was measured; HE staining was performed to estimate non-alcoholic fatty liver disease (NAFLD) activity score according to the criteria of Kleiner et al 28 The remainder of the liver was snap frozen and used for protein, hydroxyproline and crosslink analysis.

| Myocardial infarction
The study was performed by CL Laboratory (Baltimore, USA) with approval from the Institutional Animal Care and Use Committee.
Myocardial infarction (MI) was induced in C57/BL6 mice by occluding the left coronary artery. The same surgery but without occluding the left coronary artery was used as a sham control. At 24 hours post-surgery, animals received echocardiography.
Infarcted mice with high cardiac function (FS > 40%) or low cardiac function (FS < 10%) were excluded from the study. The remaining mice were treated q.d., p.o., with 25 mg/kg of PXS-5153A for 4 weeks. At the end of the experiment, echocardiography was performed on mice to assess left ventricular function and remodelling, followed by heart collection. The heart was fixed with 10% formalin. Fibrosis was assessed in the non-infarct area. Fibrotic blue area and whole non-infarct area were measured using computerized planimetry (Image J). The fibrotic area was presented as a percentage of the whole non-infarct area. Three random fields per heart were counted, averaged and a total of 30-45 fields per group were measured.

| RNA isolation and real-time PCR analysis
Total RNA was extracted with the PureLink RNA Mini Kit according to the manufacturer's instructions (Ambion), followed by the cDNA synthesis using SuperScript VILO cDNA Synthesis Kit (Life technologies). Gene expression was measured by the 2 −ΔΔCT method using ABI7500 (Applied Biosystems) with the ABI TaqMan primer sets as specified in the supporting information.  (Table 1). PXS-5153A also inhibited human LOXL3 with an IC 50 value of 63 nmol/L. The compound is >40-fold selective for LOXL2 over both LOX and LOXL1 and >700fold selective over other related amine oxidases. PXS-5153A was found to have little to no activity against a number of additional targets (Table S1), with the exception of Adrenergic α2A (97%, when tested at 10 μmol/L) and calcium channel L-type, dihydropyridine receptor, rat (80%, when tested at 10 μmol/L).
PXS-5153A was designed to interact with the LTQ (lysine tyrosylquinone) cofactor in the enzymatic pocket of LOXL2 and LOXL3, which, upon elimination of the fluoride-leaving group, leads to a covalently bound enzyme-inhibitor complex. The importance of the leaving group was highlighted by the 20-fold reduced potency displayed by the corresponding des-fluoro analogue (Table S2) PXS-5153A is a fast acting inhibitor, with enzymatic activity almost entirely blocked within 15 minutes, as judged by the shift in the concentration response curves ( Figure 1A). In contrast, the racemate of the selective LOXL2 inhibitor of PAT-1251-requires approximately 4 hours to achieve complete inhibition ( Figure S3B).
The pharmacokinetic (PK) properties of PXS-5153A were investigated in rats and mice and are reported in    Total immature collagen crosslink as well as total mature crosslink quantities were then analysed. There was a significant increase of immature as well as mature crosslinks upon CCl 4 -stimulus compared to healthy animals (4.6 and 2.7-fold increase, respectively; Figure 3C).

| Inhibition of NASH by PXS-5153A
Although the CCl 4 Figure 5H). PXS-5153A treatment significantly reduced hepatocyte ballooning demonstrating the hepatoprotective effects of the compound and was paralleled by a reduction in NASH disease score ( Figure 5I).

| Correlation between collagen crosslinks and liver function
To investigate whether there was a correlation between collagen crosslink amounts and liver function, a simple linear regression model was applied to directly compare the individual crosslinks with various liver functional readouts.
In the CCl 4 model, there was a significant positive correlation between mature crosslinks and all liver functional readouts, while for the immature crosslinks there was only significant correlation with In the NASH model, both PYD and DHLNL showed a significant positive correlation compared to the percentage fibrotic, and a modest correlation with the NASH score ( Figure S7A-D). Sham and n = 12-16 NASH. Data are compared using Student Two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001 relative to the NASH group; # P < 0.05, ## P < 0.01, #### P < 0.01 compared with sham control. Effect of PXS-5153A on NASH: (H) representative photomicrographs of the HE stained liver sections (50 and 100X); (I) NAFLD activity score (NAS), calculated according to the criteria of Kleiner (30). Data are presented as mean values ± SEM for n = 6 Sham and n = 12-16 NASH. Histological scoring analysed using nonparametric Mann-Whitney U test. Data are compared using Student Two-tailed t test. ***P < 0.001, ****P < 0.001 relative to NASH. #### P < 0.0001 compared with sham control Overall, these results indicate that PYD and DHLNL can be used as predictors of disease severity during liver fibrosis, with levels of PYD showing more significance than those of DHLNL.

| Inhibition of heart fibrosis by PXS-5153A
A number of studies have indicated that inhibition of LOX family members, in particular LOXL2, can positively influence myocardial remodelling, 33,34 with LOXL2 mRNA expression being highly upregulated upon cardiac disease. 34   On a more fundamental level, the findings of this study confirm that hydroxyallysine-derived crosslinks are a direct result of LOXL2/ LOXL3 enzymatic function. The mechanism by which collagen crosslinks are formed is based on the reactions of allysine or hydroxyallysines present on collagen side-chains with other aldehydes or with unmodified lysine or hydroxylysine residues, resulting in the formation of crosslinks. 7,[9][10][11] The availability of hydroxyallysine is the direct result of lysine hydroxylation, through a process driven by lysyl hydroxylases. 35 In bone, tendon, ligaments and cartilage, 36 Even though crosslink maturity as well as allysine/hydroxyallysine pathway ratios have been previously correlated with tissue mechanical properties, 49,50 this is the first study to correlate hydroxyallysine pathway with liver organ function during fibrosis. ALT and AST are commonly used as predictors of liver function, as the release of these enzymes from liver cells to the bloodstream parallels hepatocellular damage or death and provide an indirect measurement of liver and hepatocyte impairment. 31 In the CCl 4 model, there was a positive correlation between mature crosslinks and all functional readouts (ALT, AST and percentage fibrotic coverage area), clearly
A number of studies have been useful in confirming the role of lysyl oxidases in controlling tissue stiffness and collagen extractability, 14,51,52 however, they did not provide definite evidence for the correlation between lysyl oxidase enzymatic action and collagen crosslinking. In our current study, analysis of the liver revealed that the content of immature (DHLNL and HLNL) and mature (PYD and DPD) crosslinked collagen was increased upon the induction of liver fibrosis and that dual inhibition of the enzymatic functions of both LOXL2/ LOXL3 reduced formation of these crosslinks. Although a direct comparison of individual crosslinks and liver stiffness was not performed in the current study, the results generated support the concept that difficulty in extractability and increases in tissue stiffness 14,52 are due to collagen crosslink maturation. Altogether, these results suggest that LOXL2/LOXL3 enzyme-dependent crosslink formation leads to remodelling during fibrosis and, overall, tougher tissue.
Taken together, these results show that LOXL2 and LOXL3 enzymatic functions are key players in the formation of hydroxyallysine derived collagen crosslinks during fibrosis. This study highlights the potential of inhibiting LOXL2/LOXL3 enzymatic activities, for example by use of PXS-5153A, as a novel therapeutic tool for the treatment of diseases that are characterized by abnormal increases in collagen crosslinking.

ACKNOWLEDG EMENTS
We thank Dr. Lucy Cao for her technical expertise regarding crosslinks.