Lysyl oxidase-mediated collagen crosslinks may be assessed as markers of functional properties of tendon tissue formation.

2.1. In ovo culture and tendon harvest All animal procedures received prior approval from the university institutional animal care and use committee board. All reagents were from Sigma-Aldrich Co. (St. Louis, MO) unless otherwise specified. White leghorn chick embryos (University of Connecticut Poultry Farm, Storrs, CT) were cultured in a humidified rocking incubator at 37.5° C. Embryos were sacrificed and staged according to Hamburger and Hamilton (HH) [ 17 ] at HH 28, 35, 40 and 43, equivalent to approximately days 5.5, 9, 14, and 18 out of a 20-day gestation period, respectively. At 24 h before each timepoint, embryos we re injected with 200 μL of (β-aminopropionitrile (BAPN; inhibitor of LOX activity) in saline equivalent to either 0, 5 or 15 mg/g of dry embryo mass [ 18 ] into the chorioallantoic membrane [ 19 ]. The shell hole was sealed with liquid paraffin and embryos were cultured in ovo for an additional 24 h. For the stages and BAPN doses tested, the viability rate 24 h after injection was 97.9%. After sacrifice, the calcaneus tendon was dissected from skin and muscle tissues. At HH 28, since tendon is not visible in the gross, lower limbs were removed at the hip, the feet were removed, and the lower two-thirds of the remaining limb were used for analysis based on tenascin-positive immunohistochemical staining in this region [ 20 ] and our previous histological analysis [ 4 ]. Tissues were then prepared for either mass spectrometry or imaging analysis. 2.2. Sample preparation for tandem mass spectrometry (LC-MS/MS) Samples were prepared following the method of Gineyts et al. [ 21 ]. Briefly, freshly excised chick tendon tissues, including an adult porcine Achilles tendon control group, were minced, washed in dH 2 O, and lyophilized for 1 week to obtain dry mass. Lyophilized tendons were then suspended in DPBS and reduced by addition of 10 mg/mL NaBH 4 in 1 mM NaOH to yield a 1:30 reagent-to-tendon dry mass ratio. The reaction was allowed to proceed for 2 h and was terminated by addition of acetic acid to pH 3. Samples were then washed in dH 2 O and lyophilized for 1 week. Weighed, dried samples were added to 6 M HCl at 10 mg/mL and hydrolyzed for 20 h at 110° C. Acid was evaporated and hydrolysate were resuspended at 10 mg/mL in LC-MS grade H 2 O and subsequently passed through 0.2 μm filters. Samples were stored at -20° C until use. Degradation of HP and LP during storage or freeze thaw was not expected due to an estimated stability of 99.9% after 25 years at -20° C and an absence of degradation after 10 freeze-thaw cycles [ 22 ]. Bovine cortical diaphyseal femur bone samples from 24 month-old calves were processed for LC-MS/MS as a control tissue. Femur samples were obtained from a local abattoir (Research 87, Boylston, MA) the same day as slaughter and frozen at -80° C. Diaphyseal femur sections were excised and split longitudinally with a diamond rotary saw. Marrow and soft tissue were removed, and samples were immersed in liquid nitrogen for 10 min and pulverized. Bone fragments were then demineralized in 0.5 M EDTA and 50 mM Tris (pH 7.4) for 21 days at 4° C with daily solution changes [ 21 ]. Demineralized bone samples were subsequently processed in an identical manner as tendon samples for LC-MS/MS analysis. 2.3.

Sample preparation for multiphoton microscopy

Freshly harvested tendon and demineralized bone samples were washed and immersed in OCT cryosectioning media (Sakura Finetek, Torrance, CA) for multiphoton microscopy analysis. Tissues were oriented for longitudinal sectioning and frozen at -80° C until use. On the same day as imaging, tissues were placed in a -19° C cryostat, sectioned longitudinally at 50 μm, and immersed in saline to remove OCT. Sectioned samples were washed with saline and stored at 4° C until imaged. This sectioning protocol was designed to minimize differences in optical properties resulting from tissue thickness. 2.4. LC-MS/MS analysis and analyte quantification A triple quadrupole linear ion trap mass spectrometer (3200 QTRAP, AB SCIEX, Foster City, CA) coupled to a binary pump HPLC (1200 Series, Agilent, Santa Clara, CA) was used for the identification and quantification of HP, LP, pyridoxine and hydroxyproline following the method of Gineyts et al. [ 21 ] with minor modifications. Prior to sample analysis, MS-specific parameters were optimized for each target analyte to identify the multiple reaction monitoring transition (precursor/product fragment ion pair) with the highest intensity under positive mode. For each compound, the following precursor-product fragment transitions were monitored: 429.2 to 82.0 m/z for HP, 413.2 to 84.0 m/z for LP, 170.1 to 152.1 m/z for pyridoxine, and 132.05 to 68.0 m/z for hydroxyproline [ 23 ]. Chromatographic separation was achieved using a Waters Corp. (Milford, MA) Atlantis T3 4.6 × 100 mm reversed phase column with 3 μm particle size, in series with a Waters 4.6 × 20 mm guard cartridge at 22° C. The LC solvents used were Solvent A: 0.12% heptafluorobutryic acid (HFBA) in LC-MS grade H 2 O; Solvent B: 50% acetonitrile. The LC gradient used was as follows: t=0, 10% B; t=60 min, 70% B, t = 61 min, 10% B, t = 70 min, 10% B at a flow rate of 400 uL/min. Tendon hydrolysates were mixed with 10 μM pyridoxine as an internal standard and 1% HFBA to yield, in a 50 μL injection, 100 pmol of pyridoxine and an appropriate dry tissue mass (50-500 μg) to maintain signals within the linear range of the standards. The elution times of HP and LP were confirmed by spiking hydrolysates with an HP and LP standard (Quidel Corp., San Diego, CA). Peak areas from chromatograms were quantified using the Applied Biosystems Analyst software v1.5.1 employing automatic peak integration. Standard curves were developed for HP, LP and hydroxyproline analytes by injecting varying molar amounts of these compounds in triplicate with 100 pmol of pyridoxine and calculating the analyte-to-pyridoxine peak area ratio. The number of collagen moles in the sample was calculated assuming 300 moles of hydroxyproline per mole collagen [ 24 , 25 ]. HP- and LP-to-collagen ratios were calculated as the molar ratio of HP- or LP-to-collagen. HP- and LP-to-dry mass ratios were calculated as the ratio of HP and LP moles measured normalized to the amount of hydrolyzed dry tissue mass injected onto the LC-MS/MS. Collagen content was calculated as the mass ratio of measured hydroxyproline mass to hydrolyzed mass injected on the LC-MS/MS [ 23 , 24 ]. 2.5.

Multiphoton microscopy and data analysis

The imaging system was composed of a Leica (Wetzlar, Germany) TCS SP2 laser scanning confocal microscope equipped with a 100 femtosecond pulsed Ti:sapphire laser (Mai Tai, Spectra Physics, Irvine C A) delivering either 720 or 800 nm light onto the specimen. A two-photon excited fluorescence (TPEF) image was acquired at 720 nm excitation in epi-detection mode with a 63×, 1.2 NA water immersion objective using a descanned photomultiplier tube (PMT), collecting light between 380 and 425 nm. SHG images were acquired in the forward direction (F-SHG) with a 400 ± 10 nm bandpass filter using linearly polarized illumination at 800 nm. Incident laser power and photomultiplier tube (PMT) gain were adjusted to optimize fluorescence signal collection (PMT offset was kept constant at zero). To compare images acquired with different laser power and PMT gain settings, image pixel intensities were calibrated using a fluorescein solution, as described previously [ 26 ]. The incident power varied between 20 mW to 23.5 mW for 720 nm excitation, and from 7.6 to 31.8 mW at 800 nm. Z-stacks of images were acquired through the full depth (50 μm) of the tissue at 7 μm increments from 4-5 locations (stacks) over each cryosectioned tissue specimen. Background noise was assessed at each tissue imaging location by taking images 20 μm below the tissue section, focused in the saline medium. The mean intensities of these images were subtracted from the mean tissue image intensities at each respective location to remove background contributions. Relative crosslink density was estimated from multiphoton images by taking the ratio of power-normalized background-subtracted mean TPEF intensity over the corresponding power-normalized mean F-SHG intensity at each imaging location. The latter was used as an optical measure of collagen density since we found the intensity of these images to be well correlated to tendon collagen content of corresponding tissues measured by LC-MS/MS (r 2 =0.89; Fig. S1 ). Therefore, while the F-SHG intensity is not expected to depend on the collagen content via a simple linear relationship in general [ 27 , 28 ], this was a reasonable assumption for this set of samples. To analyze the potential fluorescence efficiency of HP and LP when assessed with imaging, a high performance liquid chromatography (HPLC) approach was employed with fluorescence detection at 295 nm ex./400 ± 10 nm em. to examine the fluorescent peak area of HP and LP as a function of moles injected on the column [ 25 ]. Here an isocratic methanol-based elution buffer was employed as described previously with fluorescence detection of HP and LP [ 25 ]. 2.6.

Statistical analysis

The number of biological replicates measured with each method is summarized ( Table 1 ). For tandem mass spectrometry, each biological replicate consisted of pools of chick tendons from multiple embryos to acquire sufficient material for measurement. For multiphoton microscopy, each biological replicate was from a single chick tendon. Statistical significance testing for LC-MS/MS data was performed using a two-way ANOVA between drug treatment and developmental stage with Bonferroni's post-hoc test. For multiphoton microscopy data a one-way ANOVA with Tukey's post-hoc test was performed between drug treatments for each stage. All statistical calculations including linear regression analyses were performed with Graphpad (La Jolla, CA) Prism v.5.03.

3.2. Validation of LC-MS/MS method using adult tissue controls To validate the accuracy of the mass spectrometry method, we used adult porcine Achilles tendon and bovine diaphyseal femur as control tissues. The chromatograms of these two tissues showed clear differences in their HP-to-LP peak area ratios, where this ratio was on average larger in adult tendon (19.1; Fig. 2 ) than adult bone (4.1; Fig. 2 ), which was consistent with prior reports [ 24 , 25 ]. Specifically, the magnitudes of HP and LP density of adult tendon were 0.47 ± 0.05 mol/mol for the HP-to-collagen ratio and 0.02 ± 0.009 mol/mol for the LP-to-collagen ratio ( Fig. 2 ), which were similar to previously reported values ranging from 0.16-0.95 mol/mol for HP-to-collagen and 0.01-0.1 mol/mol for LP-to-collagen depending on age, species and tendon location [ 2 , 3 , 13 , 14 ]. The measured bovine femur values were 0.37 ± 0.04 mol/mol for the HP-to-collagen ratio and 0.09 ± 0.02 mol/mol for the LP-to-collagen ratio ( Fig. 2 ). These values were similar to reported values in human and bovine femurs of various ages, which ranged from 0.1-0.4 mol/mol for HP-to-collagen and 0.03-0.08 mol/mol for LP-to-collagen [ 29 - 31 ], and were not significantly different from previously reported values for the same bovine femur tissue of the same age used in this study ( p > 0.05; [ 21 ]). Taken together, these results demonstrate that the LC-MS/MS method in our study was sensitive to HP and LP crosslinks.

3.5.

Analysis of collagen crosslink density via multiphoton imaging

To assess the potential of multiphoton imaging as a non-invasive means to assess collagen crosslink density, we acquired F-SHG and TPEF images of 50 μm thick tendon tissue sections ( Fig. 7 ). In HH 35, 40 and 43 tendons, the spatial distribution of the TPEF signal showed in most cases similar features as the corresponding F-SHG images, suggesting that the fluorescence was originating from collagen fibers. To obtain a multiphoton-based crosslink density measure (crosslinks/collagen), the mean fluorescence of a tissue area was normalized to the corresponding mean F-SHG signal, since for our studies the F-SHG signal intensity correlated strongly with collagen content ( Fig. S1 ). Comparison of the multiphoton-based crosslink density estimates of normal vs. BAPN-treated tissue showed significant changes due to LOX activity inhibition in a manner similar to that found with LC-MS/MS. Specifically, optical crosslink density was significantly reduced with BAPN treatment at HH 35 and 43 ( p < 0.01; Fig. 7A,C ), but not at HH 40 ( p > 0.05; Fig. 7B ). This was the same pattern found with LC-MS/MS values of HP-, LP-, and total HP+LP-to-collagen ratios, where the largest reductions from BAPN treatment were at HH 35, and the smallest reductions were at HH 40 ( Fig. 4A-C ). To confirm the two-photon microscopy data was representative of total HP+LP crosslinks, we evaluated the relative fluorescence intensities of HP vs. LP crosslinks using HPLC. Fluorescence HPLC analysis found that the fluorescent peak areas of HP were 27% larger than LP on a molar basis (data not shown), indicating that the TPEF imaging method is sensitive to both HP and LP crosslinks, rather than being specific to either HP or LP.

Supplementary Material 01 Fig. S1 . Relationship between mean power-normalized SHG image intensity and hydroxyproline-to-dry mass (mass-to-mass) ratios of embryonic tendon, from HH 35, 40 and 43. Each data point represents the mean of image stacks from N = 6 chick tendons from chicks treated with 0, 5 or 15 mg/g BAPN (9 points total). An approximately linear relationship was found (r 2 = 0.89).