Modulation of Lysyl Oxidase-like 2 Enzymatic Activity by an Allosteric Antibody Inhibitor

In this report, we assessed the steady-state enzymatic activity of lysyl oxidase-like 2 (LOXL2) against the substrates 1,5-diaminopentane (DAP), spermine, and fibrillar type I collagen. We find that both DAP and spermine are capable of activating LOXL2 to the same extent and have similar Michaelis constants (Km ∼ 1 mm) and catalytic rates (kcat ∼ 0.02 s−1). We also show that LOXL2 is capable of being inhibited by a known suicide inhibitor of lysyl oxidase (LOX), β-aminopropionitrile, which we find is a potent inhibitor of LOXL2 activity. The modality of inhibition of β-aminopropionitrile was also examined and found to be competitive with respect to the substrates DAP and spermine. In addition, we identified an antibody inhibitor (AB0023) of LOXL2 enzymatic function and have found that the inhibition occurs in a non-competitive manner with respect to both spermine and DAP. The binding epitope of AB0023 was mapped to the scavenger receptor cysteine-rich domain four of human LOXL2. AB0023 binds to a region remote from the catalytic domain making AB0023 an allosteric inhibitor of LOXL2. This affords AB0023 several advantages, because it is specific for LOXL2 and inhibits the enzymatic function of LOXL2 in a non-competitive manner thereby allowing inhibition of LOXL2 regardless of substrate concentration. These results suggest that antibody allosteric modulators of enzymatic function represent a novel drug development strategy and, in the context of LOXL2, suggest that inhibitors such as these might be useful therapeutics in oncology, fibrosis, and inflammation.

In this report, we assessed the steady-state enzymatic activity of lysyl oxidase-like 2 (LOXL2) against the substrates 1,5-diaminopentane (DAP), spermine, and fibrillar type I collagen. We find that both DAP and spermine are capable of activating LOXL2 to the same extent and have similar Michaelis constants (K m ϳ 1 mM) and catalytic rates (k cat ϳ 0.02 s ؊1 ). We also show that LOXL2 is capable of being inhibited by a known suicide inhibitor of lysyl oxidase (LOX), ␤-aminopropionitrile, which we find is a potent inhibitor of LOXL2 activity. The modality of inhibition of ␤-aminopropionitrile was also examined and found to be competitive with respect to the substrates DAP and spermine. In addition, we identified an antibody inhibitor (AB0023) of LOXL2 enzymatic function and have found that the inhibition occurs in a non-competitive manner with respect to both spermine and DAP. The binding epitope of AB0023 was mapped to the scavenger receptor cysteine-rich domain four of human LOXL2. AB0023 binds to a region remote from the catalytic domain making AB0023 an allosteric inhibitor of LOXL2. This affords AB0023 several advantages, because it is specific for LOXL2 and inhibits the enzymatic function of LOXL2 in a noncompetitive manner thereby allowing inhibition of LOXL2 regardless of substrate concentration. These results suggest that antibody allosteric modulators of enzymatic function represent a novel drug development strategy and, in the context of LOXL2, suggest that inhibitors such as these might be useful therapeutics in oncology, fibrosis, and inflammation.
Lysyl oxidase-like 2 (EC 1.4.3, LOXL2) 2 is one of five members of the lysyl oxidase family, which have a highly conserved catalytic region containing a copper binding domain as well as a quinone cofactor that is formed by highly conserved lysine and tyrosine residues (1,2). LOXL2 serves as an extracellular matrix metalloenzyme that can oxidatively deaminate the ⑀-amine group of specific lysine residues of collagen and elastin (3). This deaminated product then forms cross-links in fibrillar collagen and elastin (4). The amino-terminal region of the members of the lysyl oxidase family varies greatly and thus may impart distinct functions to these proteins in vivo (2). Four scavenger receptor cysteine-rich (SRCR) domains make up the aminoterminal catalytic domain of LOXL2. SRCR domains are found on secreted and cell surface-bound proteins and are known to be involved in cell adhesion and signaling (5,6). Previous studies on the expression of LOXL2 in CHO, MCF-7, and MCF-10A cell lines suggest that this protein is expressed as a polypeptide of ϳ100 kDa and a processed form of ϳ60 kDa (3,7).
Recent work on LOXL2 has shown that this extracellular enzyme is involved in several diseases, including Wilson disease and primary biliary cirrhosis (3). Vadasz et al. found elevated levels of LOXL2 associated with the fibrotic lesions from livers of patients suffering from these disorders (3) implicating LOXL2 in fibrotic diseases of the liver. Increased expression of LOXL2 has also been observed in various cancer types, including those of colon, esophageal, and breast tissue (8,9). LOXL2 has been implicated in epithelial-mesenchymal transitions associated with epithelial tumors via a Snail-dependent mechanism (10). In addition, it has been recently shown that LOXL2 is overexpressed in gastric cancer and that an antibody against LOXL2 significantly inhibited tumor growth and metastasis (11).
Lysyl oxidase (LOX) is the best characterized member of the family, with much known about its substrate specificity and inhibitors of enzymatic function (12)(13)(14)(15)(16)(17)(18)(19). In contrast, little is known about LOXL2. It has been shown that LOXL2 is capable of utilizing 1,5-diaminopentane and collagen I as substrates (3,20). However, the inhibitory effect of BAPN on LOXL2 is ambiguous as one group has shown that BAPN inhibits LOXL2 activity whereas another has shown that it has no effect on enzymatic activity (3,20,21). In this study, we characterize the steady-state kinetics of LOXL2. The inhibitory effect of ␤-aminopropionitrile was also investigated, and the mechanism of inhibition was determined. We also identified a novel antibody that specifically binds to LOXL2 and inhibits enzymatic function through a non-competitive inhibitory mechanism, which may serve as an important therapeutic in a variety of cancers and fibrosis-related diseases.
Source of LOXL2 Protein-Recombinant human LOXL2 was purchased from R & D Systems (Minneapolis, MN). LOXL2 was sent frozen at a concentration of 0.96 mg/ml in 25 mM MES, 0.5 M NaCl, pH 6.5. Purity was measured by SDS-PAGE 4 -12% BT with reduced samples and stained with Simple Blue Safe Stain. Identity was verified by Western blot analysis as well as by mass peptide fingerprinting. Western blot was performed by running 500 ng of LOXL2 on an SDS-PAGE 4 -12% BT under reducing conditions. The gel was then transferred to a nitrocellulose membrane using the iBlot apparatus. The membrane was blocked with 5% skim milk in PBST (10 mM sodium phosphate, 140 mM sodium chloride, 0.05% Tween 20, pH 7.4) at room temperature with rocking for 1 h. The membrane was washed three times with PBST. Washed membrane was probed with anti-LOXL2 antibody generated by Arresto at a concentration of 1 g/ml antibody in the 5% milk solution described above for 1 h at ambient temperature. Membrane was washed three times with PBST and then probed with anti-mouse secondary antibody at a 1:5000 dilution in PBST. Membrane was visualized using ChemiGlow reagent in a UVP (EC3) imaging system. Mass peptide fingerprinting was conducted by NextGen Sciences (Ann Arbor, MI). Briefly, 2 g of recombinant human LOXL2 was separated on an SDS-PAGE as described above and stained. The two bands corresponding to molecular masses of ϳ90 and ϳ60 kDa were excised and sent to Next-Gen for analysis. Recombinant protein was used without further purification.
Source of Active LOXL3 Protein-Recombinant human LOXL3 was purchased from R & D Systems. LOXL3 was sent frozen at a concentration of 0.204 mg/ml in 25 mM MES, 0.5 M NaCl, pH 6.0. Purity was assessed by SDS-PAGE 4 -12% BT with reduced samples and stained with Simple Blue Safe Stain.
Expression and Purification of Human LOXL1, LOXL3, and LOXL4 Proteins-Constructs for the recombinant human lysyl oxidases with a polyhistidine affinity tag were obtained from GeneCopoeia (Rockville, MD). All three constructs were transiently transfected into Hek293 cells using Lipofectamine 2000, transfection complexes formed in 8.8 ml Opti-Mem-I using 175 l of Lipofectamine and 70 g of DNA, and added to a T175 flask at 90% confluence containing 44 ml of complete Dulbecco's modified Eagle's medium. Transfection medium was changed after 4 h to Dulbecco's modified Eagle's medium plus 0.5% fetal bovine serum and harvested at 72 h. Cells were grown at 37°C, 5% CO 2 . Cell pellet was isolated and homogenized using a glass Dounce homogenizer in 0.1 M Tris, 2% SDS pH 8.0 at 4°C. After clarification the soluble material was discarded, and the resulting pellet was solubilized in 16 mM sodium phosphate, 8 M urea, 5 mM ␤-mercaptoethanol, pH 7.8. The mixture of cell pellet and solubilization buffer was rocked at ambient temperature for 45 min and then clarified. The soluble fraction was then loaded onto an Ni-Sepharose resin equilibrated with 16 mM sodium phosphate and 6 M urea, pH 7.8, and batch bound for 1 h at 4°C. The resin was washed with 16 mM sodium phosphate, 6 M urea, 20 mM imidazole, pH 7.8. The bound protein was eluted from resin with 16 mM sodium phosphate, 6 M urea, 0.4 M imidazole, pH 7.8. Eluted material was extensively dialyzed into 16 mM sodium phosphate, 6 M urea, 1 mM dithiothreitol, pH 7.8. Purity was assessed by SDS-PAGE. Identity of the purified proteins was assessed by mass peptide fingerprinting conducted by NextGen Sciences. Protein concentration was determined by absorbance at 280 nm using the calculated extinction coefficients based on the amino acid sequence of the constructs (22).
Expression and Purification of Human LOX-A recombinant human LOX bacterial construct was purchase from Open Biosystems (Huntsville, AL). DNA was ligated into the pET20b(ϩ) (Novagen) periplasmic expression vector. Transformation was carried out in BL21 (DE3). Positive colonies were selected and grown in the presence of 100 g/ml carbenicillin. Human LOX was expressed as an insoluble inclusion body after induction with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside at an optical density of 0.6 at 600 nm wavelength. The bacteria were harvested, and the insoluble pellet was resuspended with 8 M urea, 5 mM ␤-mercaptoethanol, and 16 mM NaPO 4 , pH 7.8. Resolubilization was centrifuged at 14,000 rpm for 30 min, and the soluble sample was then loaded onto Ni-Sepharose resin equilibrated with 16 mM sodium phosphate and 6 M urea, pH 7.8, and batch-bound for 1 h at 4°C. The resin was washed with 16 mM sodium phosphate, 6 M urea, 20 mM imidazole, pH 7.8. The bound protein was eluted from resin with 16 mM sodium phosphate, 6 M urea, 0.4 M imidazole, pH 7.8. Eluted material was extensively dialyzed into 16 mM sodium phosphate, 6 M urea, 1 mM dithiothreitol, pH 7.8. Purity was assessed by SDS-PAGE. Identity of the purified proteins was assessed by mass peptide fingerprinting conducted by NextGen Sciences. The concentration of protein was determined by absorbance at 280 nm using a calculated extinction coefficient (22).
LOXL2 Fragment His Tag Purification-Ni-Sepharose resin was equilibrated with 0.1 M Tris-HCl, pH 8.0. Condition medium was loaded onto equilibrated resin. After loading, the nickel affinity column was washed with 0.1 M Tris-HCl, pH 8.0, 0.25 M NaCl, and 0.02 M imidazole. Elution was carried out with 0.1 M Tris, pH 8.0, 0.150 M NaCl, 0.3 M imidazole. SDS-PAGE 4 -12% BisTris gel was run under reducing conditions on all samples of the purification to determine purity. Purified protein was then dialyzed overnight at 4°C in 0.05 M sodium borate, pH 8.0. Protein concentration was determined using the Bradford method using BSA as a standard.
Data Analysis-All curve fitting was conducted using GraFit 6.0. Data were processed or pre-processed using Microsoft Excel or SoftMax Pro. All experiments were performed at least in triplicate, and the means and standard deviations were obtained from a minimum of three independent experiments with each performed at least in duplicate.
Assay of LOXL2 Activity by Detection of Peroxide-The enzymatic activity of LOXL2 was measured by coupling horseradish peroxidase (HRP) activity to LOXL2 and using the conversion of Amplex Red to resorufin as our readout (23). All activity assays were conducted on a SpectraMax M5 from Molecular Devices (Sunnyvale, CA) in kinetics mode with the excitation wavelength at 544 nm and the emission wavelength set to 590 nm. Measurements were made at 30-s intervals for 1 h at 37°C. The slope of the progress curve, relative fluorescence units (RFUs) per second, was determined in the linear region (24). The RFU values were converted to a final concentration of peroxide produced using a mock run in which LOXL2 was inactivated by BAPN, in the presence of defined concentrations of hydrogen peroxide, and the resulting endpoint RFU measurements were used to construct a standard curve to back calculate the final concentration of peroxide based on the RFU reading. The enzymatic reaction was started by adding substrate mixture (50 mM sodium borate, pH 8.0, 100 M Amplex Red reagent, 30 mM DAP, or spermine and 1 ϫ 10 Ϫ4 % antifoam 204) to the enzyme mixture (50 mM sodium borate, pH 8.0, 2 units/ml HRP, 50 nM LOXL2, and 1 ϫ 10 Ϫ4 % antifoam 204). Assay mixtures using type I collagen as substrate were similarly composed except that the concentration of LOXL2 was increased to 100 nM. Collagen was polymerized according to the manufacturer's directions prior to use and kept on ice until added to the substrate mixture (see above).
Substrate Titrations-The Michaelis constant (K m ) and k cat were determined by measuring the activity of LOXL2 over a range of substrate concentrations. The data were plotted as rate (initial velocity) versus the concentration of substrate and fit to Equation 1, where K m is the Michaelis constant, k cat is the first order rate of product formation, E T is the total enzyme concentration, and S is the substrate concentration. The product of k cat and E T is the maximal velocity IC 50 Determinations-The inhibitory effects of the agents tested were performed in the following manner. In the case of the irreversible inhibitor BAPN a dilution series of the inhibitor was made in the substrate solution described above, and the reaction was initiated with the addition of enzyme solution. For the inhibitory antibody, a dilution series of the antibody was incubated with LOXL2 or LOLX3 at ambient temperature for 1 h to allow binding, and then the reaction was initiated with the addition of substrate. Data were collected as described above, and the observed rates were plotted as a function of inhibitor concentration. The IC 50 (the concentration of inhibitor that results in a 50% decrease in activity relative to no inhibitor) was determined by fitting this data to a four-parameter fit (see Equation 2). Where y is the observed rate, range is the fitted uninhibited value minus background, s is the slope factor, "background" is the background rate, and x is the concentration of inhibitor. Only partial inhibition was observed in the case of the inhibitory antibody so that the reported IC 50 is an apparent IC 50 for the magnitude of the effect observed.
Inactivation Kinetics-The inactivation of LOXL2 by BAPN was measure using the activity assay described above with the following modifications. A substrate solution containing BAPN to give the desired final concentration of BAPN and substrate concentration was mixed with enzyme solution (LOXL2 was at 75 nM final) to initiate the enzymatic reaction. The progress curve fluorescence versus time was monitored at 37°C for 3 h. The progress curves were then fit to Equation 3, which describes the formation of product during the inactivation process as a function of time to obtain the rate of inactivation (25), where P is product, v i is the initial velocity, k obs is the rate of inactivation, and t is time. The rates of inactivation were then plotted as a function of BAPN concentration and fit using linear regression in GraFit to obtain the slope, which is the second order rate constant k inact /K I . Mode of Inhibition by Steady-state Analysis-The mode of inhibition of the agents in this study was assessed using the kinetics scheme (Scheme 1), which does not assume a modality of inhibition. The experimental setup involves a substrate titration at differing concentrations of inhibitor. These data are then globally fit using Equation 4 to give the values relating to the kinetics scheme (Scheme 1). An ␣ value of 1 is a non-competitive inhibitor; a value of less than 1 is an uncompetitive inhibitor, and a value much greater than 1 is a competitive inhibitor. K m is the Michaelis constant, K i is the dissociation constant of the inhibitor, S is substrate concentration, I is the inhibitor concentration, V max is the maximal velocity, and ␤ is usually set to zero for linear inhibitors, but in the case of the inhibitory antibody ␤ was allowed to float as the inhibitory effect is only partial and nonlinear.
ELISA-based Binding Assay-Maxisorp plates were coated overnight with 1 g/ml of antigen in sodium borate buffer at 4°C (100 l per well). The following day the plates were washed three times with PBST (10 mM sodium phosphate, 140 mM sodium chloride, 0.05% Tween 20, pH 7.4) and blocked with BSA solution (5% BSA in 10 mM sodium phosphate, 140 mM sodium chloride, pH 7.4, 200 l per well) for 1 h at ambient temperature. Plates were then washed three times with 300 l of PBST, and varying dilutions of anti-LOXL2 antibody (100 l) were added to the blocked plates and incubated at ambient temperature for 1 h. Plates were washed again, and secondary antibody was added: 100 l of a 1:10,000 dilution of anti-mouse HRP secondary antibody diluted in 0.5% BSA solution (0.5% BSA in 10 mM sodium phosphate, 140 mM sodium chloride, pH 7.4) followed by incubation at ambient temperature for 1 h. Plates were washed again and developed using 100 l of 3,3Ј,5,5Ј-tetramethylbenzidine and allowing the resulting blue color to develop but not too intensely. The reaction was quenched with the addition of 100 l of 1 M hydrochloric acid. Care was taken to not exceed an optical density of one. Quantitation was carried out on a SpectraMax M5 in absorption mode at 450 nm. All readings were at ambient temperature. Dissociation constants were determined by plotting the absorbance values versus the concentration of antibody and fitting the data to Equation 5, where PL is the bound concentration, which is proportional to the absorbance reading, L is the antibody concentration, B max is the maximal binding, and K D is the dissociation constant.
Surface Plasmon Resonance Binding-The binding affinity of AB0023 to LOXL2 was determined using a ProteOn XPR36 instrument (Bio-Rad, Hercules, CA). LOXL2 was immobilized to a GLC sensor chip. The GLC sensor chip was activated with a 1:1 ratio mixture of 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride and N-hydroxysulfosuccinimide, prepared as per the manufacturer's directions, flowed over the chip at a flow rate of 30 l/min for 300 s. LOXL2, at 3 g/ml in acetate buffer, pH 4.5, was then flowed over the chip at a rate of 30 l/min for 300 s, and unreacted sites on the chip were blocked with 1 M ethanolamine flowed over the chip with a flow rate of 30 l/min for 300 s. A reference channel was created using the same procedure by flowing acetate buffer over the surface. This immobilization gave a coating density of ϳ2000 response units for LOXL2. Dilutions of AB0023 were flowed over the surface at a rate of 100 l/min for 150 s, and a buffer control was included in the series. Sensograms were analyzed using the ProteOn manager software, and data were fit to the Langmuir model within the software. The data represent the average and standard deviation of four separate experiments.
Generation of Inhibitory Antibodies-Mice were immunized with human recombinant LOXL2 to generate anti-LOXL2 antibodies (Antibody Solutions, Mountain View, CA). Hybridoma libraries were generated from mice testing positive for anti-LOXL2 antibodies. The libraries were screened for binding to LOXL2 using the ELISA assay described above. Antibodies found to bind LOXL2 were then further screened for their ability to inhibit the enzymatic activity of LOXL2. AB0023 is one such antibody that exerted an inhibitory effect on LOXL2. From a similar immunization we identified a murine monoclonal that is specific for the catalytic domain of human LOXL2 and found to be an excellent Western blot reagent and used in this study.
Antibody Production and Purification-Hybridoma cells were cultured in improved minimal essential media, 10% fetal bovine serum, low IgG containing 1:100 penicillin/streptomycin, 5% Hybridoma Cloning Factor, and hypoxanthine and thy-midine media supplement (1ϫ). Cells were transferred to Aragen Bioscience (Gilroy, CA) for the production of ascites fluid in BALB/c mice. Ascites fluid was then purified by batch mode on MabSelect resin. After batch binding, the flow-through was collected, and resin was washed with 10 column volumes of phosphate-buffered saline, pH 7.4. The antibody was eluted with 0.1 M citric acid, pH 3. The eluate was neutralized with a 1:10 volume of 1 M Tris, pH 8.0, and dialyzed overnight at 4°C in phosphate-buffered saline with 0.01% Tween 20. Purified AB0023 was then tested in multiple assays for release. Purity was measured by SDS-PAGE 4 -12% BT with reduced and nonreduced samples and stained with Simple Blue Safe Stain. A 10-g quantity of purified AB0023 was injected onto size exclusion chromatography high pressure liquid chromatography (Tosoh TSKgel G3000SWxl) to assess antibody stability and extent of aggregation. The identity of AB0023 between batches was determined by the banding pattern of isoelectric focusing gels (range of gels was pH 3-10) stained with Simple Blue Safe Stain. The concentration of purified antibody was determined by absorbance at 280 nm using the extinction coefficient of 1.4 ml mg Ϫ1 cm Ϫ1 . The molecular mass of the antibody was assumed to be 150 kDa.

RESULTS
Lysyl Oxidase Proteins-The purity of the recombinant human LOX protein family members used in these studies is shown in Fig. 1A. Human LOXL2 (hLOXL2) appears to be largely the unprocessed form, running at an apparent molecular mass of 87 kDa, and a faint band at 59 kDa, based on the relative mobility of the molecular mass markers relative to the dye front, is the processed human LOXL2. Also present are higher molecular mass proteins, which appear to be aggregates of LOXL2. Western blot analysis of the hLOXL2 protein (Fig. 1B) using an anti-LOXL2 antibody, yields a similar banding pattern to the Coomassie-stained SDS-PAGE in Fig. 1A with the larger (87 kDa) form being the more prominent species of hLOXL2 and larger aggregates appearing at higher apparent molecular masses. The Western blot does not show any bands below the 59 kDa form, indicating that there appears to be no degradation products. Fig. 1C shows an overdeveloped Western blot probed for the engineered polyhistidine tag on the proteins. The lower molecular weight bands suggest there are degradation products, likely due to the fact that the LOX, LOXL1, LOXL3, and LOXL4 were purified from inclusion bodies and stored under denaturing and reducing conditions. Also visible are the larger molecular weight bands seen for LOXL2 in the SDS-PAGE and the anti-LOXL2 Western blot. To confirm the identity of the LOX protein family members shown in Fig. 1A, mass peptide fingerprinting using liquid chromatography-tandem mass spectrometry following trypsin digestion was performed on the major bands identified by SDS-PAGE. Tryptic digestion of the 87-kDa hLOLX2 band yielded 43 peptides matched to 77 unique spectra, which gave sequence coverage of 64% for hLOXL2. Mass peptide fingerprinting performed on hLOXL1 (molecular mass ϭ 65 kDa by SDS-PAGE), gave 31 unique peptides with 56 unique spectra yielding 76% sequence coverage. hLOXL3 (molecular mass ϭ 86 kDa by SDS-PAGE) gave 46 unique peptides with 78 unique spectra yielding 59% sequence coverage. And hLOXL4 (molecular mass ϭ 84 kDa by SDS-PAGE) had a 75% sequence coverage with 55 unique peptides and 92 unique spectra. The identity of hLOX (molecular mass ϭ 33 kDa by SDS-PAGE) was not verified by mass peptide fingerprinting.
None of the enzymes purified from inclusion bodies had any measurable lysyl oxidase activity using DAP or spermine as sub- strates (data not shown); therefore, these proteins were only used for binding experiments. An active recombinant human LOXL3 protein was obtained commercially and used for enzymatic activity measurements. Substrate Specificity-Whereas LOX substrate specificity has been previously investigated, relatively little is known regarding the enzymatic activity of LOXL2. To that end, we examined the enzymatic activity of LOXL2 against three distinct substrates (1,5-diaminopentane, spermine, and fibrillar collagen I) using an HRP-coupled assay in borate buffer at 37°C. Fig. 2 shows the rate of peroxide production plotted as a function of substrate concentration and the data fit to the Michaelis-Menten equation. In the presence of DAP or spermine the maximal velocities (V max ) were nearly identical (Fig. 2) and the calculated Michaelis constant (K m ), defined as the concentration of substrate at which one-half maximal activity is achieved, was 1.01 Ϯ 0.18 mM for DAP and 1.05 Ϯ 0.32 mM for spermine. The viscosity of fibrillar collagen at increasing concentrations precluded us from completely defining the plot of rate versus collagen concentration and hence extracting the k cat and K m parameters for collagen I as a substrate. Nonetheless, activity against collagen I as a substrate was observed when collagen was used at a final concentration of 0.5 mg/ml. All substrates tested were also evaluated for their possible effects on HRP, carrying out similar experiments with HRP alone and in the presence of 1 mM hydrogen peroxide. Neither DAP, spermine, nor collagen exerted measurable effects on HRP enzyme activity when incubated separately with this peroxidase ensuring that the activity we measured is due to LOXL2 enzyme activity and not secondary effects of HRP.
Inhibition of LOXL2 by BAPN-BAPN has been shown to be an irreversible, covalent inhibitor of lysyl oxidase (15,16). However, the literature regarding the effect of BAPN on LOXL2 enzymatic activity is ambiguous. Some reports show BAPNmediated inhibition of LOXL2 (20), whereas others suggest that it has no effect on the activity of LOXL2 (3,7,21). We find that LOXL2 is inhibited by BAPN in a dose-dependent manner using either DAP or spermine as substrates (Fig. 3). These data were fit to a four-parameter equation to obtain the concentration of BAPN that leads to a 50% reduction in activity (IC 50 ).
BAPN inhibited LOXL2 regardless of whether DAP or spermine were used as substrates. An IC 50 of 5.0 Ϯ 1.4 M was obtained for DAP, and an IC 50 of 3.8 Ϯ 0.2 M was obtained for spermine. The data show complete inhibition at the highest concentrations of BAPN. BAPN also inhibited LOXL2 when fibrillar collagen I was used as a substrate (data not shown).
The kinetics of inactivation of LOXL2 by BAPN were also examined by analyzing the progress curves over a range of inhibitor concentration as described under "Experimental Procedures." The inactivation kinetics was examined at various concentrations of substrate to assess the modality of inhibition. Fig. 4A shows the rates of inactivation of LOXL2 by BAPN using DAP as substrate. There is a linear dependence of the observed rate of inactivation with increasing concentrations of BAPN, suggesting that inactivation results from a single step mechanism (25). As the concentration of substrate was increased we observed a decrease in the rates of inactivation relative to lower substrate concentrations. The slopes of the data in Fig. 4A represent the second order rate constants of inactivation (k inact /K I ). Fig. 4B shows the dependence of these rates on the concentration of DAP. The decrease in k inact /K I suggests that the modality of inhibition of BAPN is competitive with respect to DAP. BAPN has been previously shown to be a competitive inhibitor of LOX (16), and we observed a similar mechanism for LOXL2.
We also analyzed the inactivation kinetics using spermine as the substrate for LOXL2. Fig. 4C shows the dependence of the rate of inactivation on the concentration of BAPN. As in the case where DAP was used as substrate, we again observed a linear dependence of the rate of inactivation with BAPN suggesting that the inactivation occurs as a single step mechanism. The second order rate constant k inact /K I also decreased with increasing spermine concentrations (Fig. 4D) suggesting that the substrate protects LOXL2 from inactivation by the inhibitor.
Inhibition of LOXL2 by an Antibody Directed against LOXL2-A murine monoclonal antibody directed against LOXL2 was developed, which inhibited LOXL2 enzymatic activity. This anti-LOXL2 antibody, referred to as AB0023, exhibited high  specificity for LOXL2 and did not bind to the other related lysyl oxidases as demonstrated using an ELISA-based assay (Fig. 5). Using this assay, the dissociation constant of the AB0023 to LOXL2 was found to be 250 Ϯ 53 pM. The interaction of AB0023 with LOXL2 and with the other lysyl oxidases was also assessed using surface plasmon resonance on a ProteOn XPR36 instrument, which demonstrated that AB0023 binding was specific to LOXL2 (data not shown). In good agreement with the ELISA-based assay, AB0023 binding affinity was also determined by surface plasmon resonance and found to be 370 Ϯ 110 pM (Fig. 5B).
Peptide fragments of LOXL2 encompassing the individual SRCR domains and a minimal catalytic domain were generated. Binding of AB0023 to these regions of LOXL2 was assessed by ELISA. AB0023 only showed specific binding to the SRCR-4 domain of LOXL2 (Fig. 5C). In addition, AB0023 showed species cross-reactivity, because it also binds to mouse, rat, and cynomolgus monkey recombinant LOXL2 (data not shown).
The intact SRCR domains of human LOXL2, human LOXL3, and human LOXL4 as well as truncated forms of the domains were also evaluated against AB0023 for binding. These domains include SRCR1 through SRCR4 (SRCR1-4) and SRCR1 through SRCR2 (SRCR1-2) for each of the lysyl oxidases mentioned. Fig. 5 (D and E) shows that only the SRCR1-4 of human LOXL2 was bound by AB0023. No binding was detected against any of the other constructs. Fig. 6A shows the dependence of LOXL2 enzymatic activity on the concentration of antibody. AB0023 elicits a dose-dependent decrease in enzyme activity, though complete inhibition is not observed. Partial inhibitory effects are usually the result of the inhibitor's physical properties such as solubility. However, in the case of AB0023, we have concentrated the antibody in excess of 6 mg/ml with Ͻ5% aggregation, 3 suggesting that the observed partial inhibition is the result of the enzyme-substrate-inhibitor complex producing product at a greatly reduced rate compared with the uninhibited enzyme (25). The inhibitory effect is also seen in a dose-dependent manner when collagen I is the substrate (Fig. 6B). The apparent IC 50 values for AB0023 against DAP, spermine, and collagen I were 62 Ϯ 5.8 nM, 55 Ϯ 11 nM, and 61 Ϯ 3.8 nM, respectively. The enzymatic activity of recombinant LOXL2 from rat and cynomolgus monkey was also inhibited with AB0023. 3 The inhibition of human LOXL3 enzymatic activity was also evaluated. This analysis showed that, although BAPN was capable of inhibiting LOXL3 (Fig. 6D), AB0023 showed no inhibition over the dose range tested (Fig. 6C). This was not surprising, because AB0023 did not bind LOXL3.
This antibody was found to be a non-competitive mixedtype inhibitor of LOXL2. Fig. 7A shows the dependence of enzymatic function on DAP concentration at differing inhibitory antibody concentrations. Similar results were obtained when spermine was used as the substrate (Fig. 7B). This family of curves was analyzed by global fitting as described under "Experimental Procedures." The ␤ parameter was allowed to float to account for the observed partial inhibition of AB0023. Global fitting gave the following values: ␣ ϭ 1.02 Ϯ 0.05 and ␤ ϭ 0.53 Ϯ 0.04 for DAP and ␣ ϭ 1.13 Ϯ 0.26 and ␤ ϭ 0.51 Ϯ 0.06 for spermine. Lineweaver-Burke plot analysis of the AB0023 inhibition kinetics (data not shown) was FIGURE 5. Binding specificity of AB0023 against the five members of the lysyl oxidase family (A), LOXL2 using surface Plasmon resonance (B). The ordinate describes the signal, and the abscissa is time in seconds. The figure shows kinetic traces of AB0023 binding to LOXL2 with AB0023 at concentrations of 25 nM, 12.5 nM, 6.25 nM, and 3.125 nM at 25°C and the different SRCR domain fragments of LOXL2 (C). Binding against the lysyl oxidase family and SRCR fragments was detected using an ELISA-based assay and data fit as described under "Experimental Procedures." Surface plasmon resonance experiments were conducted on a ProteOn XPR system, and data were analyzed using the ProteOn Manager software. (D), binding assessment of AB0023 against the intact SRCR domains of hLOXL2, hLOXL3 and (E) hLOXL4 using an ELISA-based method. consistent with a mechanism by which the substrate-enzyme-inhibitor complex was capable of producing product (25,26). The ␤-value was ϳ0.5, which suggests that the substrate-enzyme-inhibitor complex forms product at half the catalytic rate. This mode of non-competitive-like inhibition was consistent with that expected of a partial inhibitor (26).

DISCUSSION
In this report, we identified an antibody inhibitor that binds specifically to human LOXL2 and modulates enzymatic function. Over 26,000 hybridoma clones, resulting from multiple immunizations with several protein and peptide constructs across the length of LOLX2, have been screened, and only seven inhibitory antibodies of LOXL2 have been identified. Five of these bind the enzymatic domain, but none of these inhibit LOXL2 as potently as AB0023. One very weak inhibitor bound to the linker region between SRCR-3 and SRCR-4. Various other antibodies that bind SRCR1-3, as well as other binders in the SRCR-4, did not inhibit LOXL2 enzymatic activity despite having binding affinities better than 10 nM.
AB0023 inhibits via a non-competitive like mechanism, which enables the inhibitor to bind to LOXL2 in either the substrate-bound or unbound state. This gives AB0023 the distinct advantage of exerting its inhibitory effect at high and low concentrations of substrate. Because LOXL2 has been impli-cated in fibrotic diseases and in cancer, this is an advantage because LOXL2 would be predicted to experience an environment with very high local concentrations of substrate. For instance, in liver fibrosis samples, Vadasz et al. found LOXL2 in fibrotic lesions that are rich in collagen. In situations such as this, orthosteric inhibitors would not effectively inhibit enzymatic activity because, at very high substrate concentrations, the competitive inhibitor loses its inhibitory effect. We observed this effect in vitro with BAPN against LOXL2. At high concentrations of DAP or spermine, the inhibitory effect of BAPN decreased. The same would be expected of an antibody competitively inhibiting an enzyme (27). AB0023, however, still exerted an inhibitory effect at very high substrate concentrations, making it a superior choice of inhibitor in an in vivo setting.
The binding epitope of AB0023 was mapped to the SRCR-4 domain of human LOXL2. This makes AB0023 an allosteric modulator of enzymatic function, because it binds to a region remote from the active site or the binding domain of the endogenous ligand, and suggests that AB0023 exerts its inhibitory effect, possibly, by propagating an inhibitory signal to the catalytic domain via some interaction through the SRCR-4 domain. The uncompetitive class of inhibitors is noted for inducing a conformational change that then inhibits the enzyme when substrate is bound. In the case of AB0023-mediated inhibition of LOXL2, the inhibitory effect can be exerted on the substrate-bound or free enzyme, effectively giving AB0023 two opportunities to inhibit LOXL2 enzymatic activity. LOXL2 might very well undergo a conformational change that influences the extent of activation by substrate in the substrate bound and unbound forms. SRCR domains are found on several proteins, including secreted enzymes, and while the exact role of these domains is not known, it is suggested that they may be involved in targeting or imposing substrate specificity (6). In this study, we have shown that AB0023 binds SRCR-4 of human LOXL2 and modulates the enzymatic activity. Because the inhibitory effect is present with two small molecule substrates as well as with collagen I, it seems unlikely that SRCR-4 imparts substrate specificity. Although the mechanism by which LOXL2 is regulated is not understood, it is possible that LOXL2 is regulated or sequestered via an interaction through the SRCR domains of this enzyme.
The use of antibodies as inhibitors of enzymatic function is not a new idea, and several such inhibitory antibodies have been created against enzymes. One approach has been to use dromedary antibodies (28), because the heavy chain homodimers isolated from them have the potential to recognize protein surfaces. Several reports have shown that these camel antibodies can also be inhibitors of enzymatic function and appear to be competitive inhibitors; these enzymes include carbonic anhydrase, ␣-amylase, ␤-lactamase, and HCV NS3 protease (27)(28)(29). A second approach has been to use full monoclonal antibodies as inhibitors. Several investigators have shown that it is possible to identify monoclonal antibodies that inhibit enzymatic function (30 -34). One of these investigators inserted an epitope from HIV-I gp120 protein V3 loop into different positions of bacterial alkaline phosphatase and showed that this insertion did not significantly affect the enzymatic activity of the phosphatase (30). When an anti-gp120 monoclonal antibody was introduced, inhibition was only seen in the designed enzyme that had the epitope introduced away from the active site. The enzymes with the epitope closer to the active site showed no inhibition, suggesting that inhibition is occurring via an allosteric mechanism. In this report we have similarly shown that antibody binding to a site remote from the active site can modulate enzymatic function.
In addition to identifying an allosteric inhibitor, we have also characterized the steady-state kinetics of human LOXL2. Both DAP and spermine behave similarly as substrates for LOXL2, achieving similar maximal velocities as well as nearly identical Michaelis constants. Collagen I, a putative substrate of LOXL2, is also capable of eliciting enzymatic activity from LOXL2. However, although it is a useable substrate, the increase in viscosity with increasing collagen concentration makes it impractical for use in the common Amplex Red detection system.
We demonstrated that BAPN is a potent inhibitor of LOXL2 with an apparent IC 50 in the micromolar range. We hope that the results of this study will help to clarify any confusion relating to the inhibitory effects of BAPN as well as presenting the in vitro characterization of purified LOXL2. We found that our results are consistent with those of Jung et al. (20).
Recombinant forms of the other lysyl oxidase family members were expressed in various systems, but we found them to be inclusion bodies. Although these proteins can be expressed and purified, we find that these proteins have no detectable enzymatic activity with the commonly used substrate DAP under our assay conditions. Attempted protein refolding and the introduction of copper did not produce active enzyme. To evaluate the binding specificity of AB0023, the SRCR domains of LOXL2, LOXL3, and LOXL4 were used. AB0023 was only found to bind the SRCR constructs that contained LOXL2 SRCR4, which is consistent with that seen for the individual LOXL2 SRCR domains. No binding was observed for the SRCR constructs of LOXL2 SRCR1-2, LOXL3, and LOXL4. LOXL3 is the most closely related lysyl oxidase to LOXL2 (1), and no inhibition was observed against the enzymatic activity of human LOXL3 demonstrating specificity of AB0023. LOX and LOXL1 do not contain SRCR domains (1), thus AB0023 would not be expected to bind to these two lysyl oxidase family members. The denatured human proteins LOX, LOXL1, LOXL3, and LOXL4 were also tested for binding activity, and none was observed. The lack of binding was not due to the denatured state of the proteins, because AB0023 was still capable of binding denatured human LOXL2 (supplemental Fig. 1).
As a cautionary note to other investigators, one must be rigorous in one's efforts to observe specific lysyl oxidase family member activity in the Amplex Red assay. For example, we have measured "monoamine oxidase" activity from various conditioned media and found that the measured activity is capable of being inhibited by BAPN. However, when we measured the activity of conditioned media from MCF-7 cells overproducing LOXL2, we found that the activity was incapable of being inhibited by our LOXL2-specific inhibitor AB0023. Importantly, when the overproduced LOXL2 was subsequently purified from the condition media, we did observe inhibition of the purified LOXL2 with AB0023. Thus, these data suggest that the measured activity is a sum of not only LOXL2 activity but also other monoamine oxidase activities or peroxide-producing activities. Additionally, we find that the introduction of serum to conditioned media results in a measurable activity that can be inhibited by BAPN. We suggest that investigators not rely on endpoint assays when attempting to measure activity, but rather use a continuous read assay to ensure that the measurements are within the linear range of the assay. Also, when characterizing the effects of various substrates or potential substrates it is crucial that the Michaelis constants be reported. Simply concluding that a substrate is more "specific" from a single endpoint read at a single substrate concentration is incorrect and should be avoided. One should also be mindful in the interpretation of activities measured from condition media

Antibody Inhibitor of LOXL2
JULY 2, 2010 • VOLUME 285 • NUMBER 27 as anything that modulates peroxide generation will lead to a signal.
Allosteric modulators of enzymatic function such as AB0023 represent a novel drug strategy. This is particularly true in the context of antibodies where specificity and selectivity can be incorporated within the antibody depending on the targeted allosteric site. These attributes of antibody inhibitors offer advantages over competitive inhibitors that most often mimic the endogenous substrate of the enzyme, leading to a lack of specificity, because other enzymes might also act on the same or similar substrates. Additionally, antibody allosteric modulators also benefit from having physical and pharmacokinetic properties superior to those of small molecules. Highly specific inhibitory antibodies can be identified by screening for high affinity binders and then subsequently characterized for inhibitory properties. AB0023 is one such antibody that provides a unique opportunity for assessing the role of LOXL2 in models of disease. Thus, antibodies such as AB0023 might represent new therapeutics for treating conditions such as liver fibrosis, cancer, and inflammation.