Aging impairs the osteocytic regulation of collagen integrity and bone quality

Poor bone quality is a major factor in skeletal fragility in elderly individuals. The molecular mechanisms that establish and maintain bone quality, independent of bone mass, are unknown but are thought to be primarily determined by osteocytes. We hypothesize that the age-related decline in bone quality results from the suppression of osteocyte perilacunar/canalicular remodeling (PLR), which maintains bone material properties. We examined bones from young and aged mice with osteocyte-intrinsic repression of TGFβ signaling (TβRIIocy−/−) that suppresses PLR. The control aged bone displayed decreased TGFβ signaling and PLR, but aging did not worsen the existing PLR suppression in male TβRIIocy−/− bone. This relationship impacted the behavior of collagen material at the nanoscale and tissue scale in macromechanical tests. The effects of age on bone mass, density, and mineral material behavior were independent of osteocytic TGFβ. We determined that the decline in bone quality with age arises from the loss of osteocyte function and the loss of TGFβ-dependent maintenance of collagen integrity.


RNA isolation and analysis of gene expression
Humeri from mice were cleaned of soft tissue with metaphysis and periosteum dissected away, and the marrow removed via centrifugation.The remaining cortical bone of two humeri per mouse were snap-frozen in liquid nitrogen and combined in a single tube with 1.5 mL of QIazol (Invitrogen).While frozen, bone samples in QIazol solution were mechanically homogenized with a rotor-stator homogenizer (GLH, Omni).Lysate was then stored at -80 °C until RNA extraction [3][4][5].
For RNA extraction, QIAzol lysate was thawed, and 220 μL chloroform per mL of lysate was added and mixed well via pipette.After solution began to separate (2-3 mins on benchtop), samples were then centrifuged for 15 minutes at 12,000 x g at 4°C, and the aqueous supernatant carefully transferred to a clean tube.mRNA was purified using the miRNeasy Mini Kit (Qiagen), following manufacturer's instructions.On-column DNase (Qiagen) was applied for 15 minutes, and the RNA was eluted in 60 μL RNase-free water.RNA concentration was quantified with a NanoDrop spectrophotometer and the product stored at -80 °C.
RNA Sequencing for the aged WT mice (C57BL/6) was completed as described on the Illumina HiSeq 4000 at the UCSF Functional Genomics Core [6].Single-end 50 bp RNA-seq reads were aligned to the Ensembl mouse GRCm38.87 reference genome using STAR 2.5.2baligner [7].Analysis of the aged mice bone resulted in 745 million total reads with an average of 78.7% of these reads aligning uniquely to the mouse genome.The DESeq2 package in R Statistical Computing Environment [8] was used to find differentially expressed genes with a false discovery rate of 0.1.Read count fold change for individual gene comparisons were normalized with reference to young (2 month) WT mice average read count.Statistical comparisons utilized a Bonferroni corrected student's T-test for 6 multiple comparisons among the 4 groups.
For RT-qPCR, RNA was reverse-transcribed using the iScript cDNA synthesis kit (BioRad) for a final concentration of 10 ng cDNA / µl.RT-qPCR of cDNA utilized iQ SYBR Green Supermix (BioRad) with 18s as the housekeeping gene for comparison of target genes Loxl2, loxl3, and Postn (Supp.Table S1) on a BioRad Thermocycler (Hercules, CA).In total, 60 ng of cDNA was used per reaction for 18s and ~60ng of cDNA per gene was used for amplification between 20 and 30 cycles.Expression was then quantified by the ΔΔCt method.

TGFβ ELISA
Protein from bone lysates was harvested as previously described [9].Briefly, for collection of bone specific protein, the distal half of tibia were stripped of periosteum and connective tissue then centrifuged to remove marrow before being snap frozen in liquid nitrogen.Frozen bones were transferred to chilled pre-filled Bead Mill Tubes (ThermoFisher Scientific, Waltham, MA) containing radioimmunoprecipitation buffer [RIPA -10 mMTris, pH 8, 1 mM EDTA, 1 mM EGTA, 140 mM sodium chloride, 1% sodium pyrophosphate, 100 mM sodium fluoride, 500 μM PMSF, 5 mg/mL eComplete Mini protease inhibitor tablet (Roche, Basil Switzerland)].Samples were disrupted using a benchtop bead mill homogenizer (Omni International, Kennesaw, GA) at 4°C.Following centrifugation supernatants were sonicated for 8 times (15s/45s rest) in an ice-cold water bath.Serum was collected from blood as described [10].Protein concentrations were quantified using the Pierce Coomassie (Bradford) Assay Kit (ThermoFisher Scientific, Waltham, MA) before hydrolysis in 6N HCL.Total levels of TGFβ ligand were quantified from hydrolyzed lysates using the Bio-Plex Pro TGF-β Assay Kit (Bio-Rad, Hercules, CA) on a Bio-Plex 200 (Bio-Rad, Hercules, CA).TGFβ ligand levels were normalized to each samples' original total protein concentration.

Histological Analysis
Proximal femurs were stripped of periosteum and connective tissue and fixed for 48 hours in 10% Neutral Buffered Formalin (NBF) and then decalcified in 0.5M disodium and tetrasodium EDTA (pH 7.3-4) for 2 weeks at 4°C and prepared for cryosectioning as described [11][12][13].For imaging of large sections of the osteocyte network, femurs stained using the lipophilic dye 1,1'-Dioctadecyl-3,3,3',3'-Tetramethylindocarbocyanine Perchlorate (DiI, ThermoFisher) dissolved in 50% PBS/DMSO and Phalloidin-conjugated AlexaFlor 488 (Invitrogen) for labeling of osteocyte membranes and cytoskeleton, as described [5,11].Sections for 3D imaging were taken from the midpoint of the diaphyseal femur halfway between the growth plates.Given the ages and sexes analyzed, the absolute distance changed per animal and so this method was used to obtain consistent regions of interest.Each sample was sectioned at a size of 50 mm, moving from the distal end of the bisected proximal femur midshaft toward the femoral head.Each digitally analyzed region was 35 mm in thickness and excluded ~7.5 mm on either edge of the image stack to avoid imaging artifacts.Histologic TUNEL assays were completed on 8-10 μm sections using the Fluorescein In Situ Cell Death Detection Kit from Roche (Basel, Switzerland).Cell death was measured as the percent of identified cell nuclei co-stained with DAPI and Fluorescein against total number of DAPI-positive nuclei.Fluorescent confocal imaging was completed using a Lecia DMi8 (Leica Microsystems, Wetzlar, Germany) inverted confocal microscope running LAS X software, while brightfield imaging was completed with a Nikon Eclipse E800 microscope (Tokyo, Japan).3D structural analysis of the osteocyte LCN was completed as described [11].Briefly, canalicular tortuosity was measured utilizing the Simple Neurite Tracer plug-in for FIJI [14] while Imaris (Oxford Instruments) was used to count individual canaliculi per osteocytes in 3D [13].μCT Unfixed femora were scanned using a Scanco μCT50 specimen scanner (Scanco Medical, Wangen-Brüttisellen, Switzerland) with an X-ray potential of 55 kVp, current of 109 μA, and voxel size of 10 μm at room temperature while hydrated in Hanks Balanced Sodium Salts (HBSS, ThermoFisher Scientific, Waltham, MA).Following 3point bend mechanical testing, distal femur halves containing femoral condyles were scanned again for trabecular bone.Samples displaying damage near the growth plates from the mechanical tests were excluded from trabecular analysis.3D measurements of cortical and trabecular parameters were obtained using Scanco analysis methods as described [15], while measurements of cortical porosity were collected through manual segmentation and quantification of Scanco generated DICOM files utilizing Dragonfly (Object Research Systems, Montreal, Canada.)

Mechanical Testing
Whole hydrated femurs were loaded to failure in three-point bending in the direction of primary physiological bending (posterior compression) using a Bose Electroforce 3200 test frame with an 8-mm span at a fixed displacement rate of 10 μm/s.Structural parameters were extracted from load-displacement curves using a custom MATLAB script and material properties were acquired by transforming specific structural properties by their individual parameters from μCT as described [16].Following testing, fracture surface cross-sections were imaged by scanning electron microscopy on a Sigma 500 VP FE-SEM (Zeiss) at an excitation voltage of 15 kV and a partial pressure of 35 Pa.Endosteal and periosteal cross-sectional diameter and thickness were measured with ImageJ and used to calculate moment of inertia assuming an elliptical cross-section for calculation of material properties from structural parameters.

Synchrotron in situ Tensile testing
Collagen fibril and hydroxy appetite mineral crystal strains were recorded during uniaxial tension testing of isolated, hydrated ulnae using synchrotron small-angle x-ray scattering (SAXS) and wide-angle x-ray diffraction (WAXD) at beamline 7.3.3 at the Advanced Light Source (LBNL, Berkeley, CA) [17].This strategy permits simultaneous, real-time measurement of specific strains in the composite materials as compared to bulk tissues as captured by CCD camera and measured with a Digital Image Correlation (DIC) MATLAB package.A two-part dental bonding adhesive (OptiBond, Kerr Corporation, Orange, CA) and resin (Aeliteflo, Bisco Dental, Schaumburg, IL) was used to create a "dumbbell" shape around the epiphyses for clamping without crushing the bone or harming the diaphysis exposed to the X-ray beam.First, the bonding adhesive was applied to the ends of the bone and cured with blue light then the ends were submersed in uncured resin in a 3D printed dumbbell mold and then cured again to solidify the resin around the bone ends.In situ tensile tests (N = 7-12 per group) were performed with a TST350 Tensile Testing Stage (Linkam Scientific Inc.) on hydrated samples at RT at a displacement rate of 2.5 μm/s.The mid-diaphysis of hydrated ulnae were exposed to a perpendicular X-ray beam with spot size 750 μm x 250 μm and an energy of 10 keV for 0.5 s every 5 s until failure or until total radiation dose reached 30 kGy to mitigate effects of radiation exposure on the material performance of the bone [18].
Samples that slipped from clamps, removed from resin prior to fracture, or showed a non-diaphyseal fracture were excluded from final analysis.The changes in collagen d-spacing and mineral crystal lattice spacing from the unloaded condition, as seen as a shift in the location of the Bragg scattering peak from SAXS pattern and the Bragg diffraction peak in the WAXD pattern, were used to calculate collagen and mineral associated strains, respectively.Conversion from 2D scattering/diffraction patterns to 1D curves for fitting and extraction of parameters was completed via a custom LabVIEW program [19].Material specific strains were determined from changes in the d-or lattice spacings over time and were time-matched to their tissue level strain at each point as determined from CCD images of the bone during deformation.

Non-enzymatic Collagen Crosslinking Quantification
Decalcified, mid-diaphyseal femurs were hydrolyzed in HCl (24 hours, 110°C) and fluorescence of neutralized lysates (excitation 370 nm, emission 440 nm) were referenced to a quinine sulfate standard and then normalized to the collagen content calculated from the amount of hydroxyproline, as measured by a colorimetric assay with absorbance at 560nm, for quantification of Advanced Glycation End-products (AGEs) within bone [20].

Fourier-transform Infrared Spectroscopy
Cortical diaphysis from femurs of male and female TβRII ocy−/− and littermate controls at 4 months of age ( 16weeks) (N=5-7 per group) were dissected and marrow removed via centrifugation.Samples were fixed overnight in 70% ethanol at 4°C prior to an ethanol series for dehydration in 80%/90%/100% ethanol, for 1 hour each at room temperature.Remaining sample moisture was removed via desiccation in a vacuum chamber containing Dri-Rite (Chicago, IL) for 1 hour.Dried cortical bone samples were homogenized by grinding with a polished mortar and pestle to a fine powder.Fourier-transform Infrared (FTIR) spectrums were collected in Attenuated Total Reflection (ATR) mode on a PerkinElmer Spotlight 200i FT-IR Microscopy System (Waltham, MA) made available by the UCSF CCMBM.To account for the possibility of genotype-dependent changes in bone mineral densities, uniform amounts (10 mg) of powdered bone material, enough to cover the ATR crystal, were analyzed per sample.Spectra were background corrected, normalized to mineral content at the peak of phosphate signal (~1010-1030 cm -1 ) [21,22] and ATR corrected via Spectrum software (PerkinElmer).Prior to normalization, raw intensities of the phosphate peak were compared to ensure that a comparable amount of mineralized bone was analyzed per sample and normalization to this feature would not skew results elsewhere in the collected spectra.Powdered samples were ensured to be in "perfect" contact with the ATR anvil/crystal for collection in ATR mode [23,24].Additional grinding and drying was applied if needed.Enzymatic and non-enzymatic collagen quality and crosslinking were quantified by measuring signal intensity at defined locations within the Amide I band [25][26][27].

Protein Extraction and Mass Spectrometric Quantitative Analysis
Protein Extraction: Femurs had soft tissue and periosteum stripped off, ends dissected at the metaphysis, and the marrow removed via centrifugation.Bones were demineralized by rotating overnight in 1 mL of 1.2 M HCl at 4°C [28].Bones were subsequently transferred to a new Eppendorf tube and kept on dry ice.All components of the homogenizer including sample tubes and homogenizing plates were cooled using liquid nitrogen.The frozen bones were subsequently pulverized with the SPEX SamplePrep 1600 MiniG tissue homogenizer in polycarbonate tubes with a 9.5 mm steel grinding ball for 2 minutes at 1500 strokes/minute.After the first 2 minutes, the samples were removed from the homogenizer and allowed to cool in liquid nitrogen for 3 minutes and were then homogenized for another 2 minutes at 1500 strokes/minute.Pulverized samples were transferred into a fresh Eppendorf tube using 800 µL of extraction buffer (6 M guanidine hydrochloride GuHCl; 10 mM Tris-HCl; 50 mM EDTA) and incubated by rotating at 4 °C for 72 hrs [29].Subsequently, the samples were spun for 3 minutes at 15,000 x g to pellet bone matrix and the supernatants, containing the proteins, were collected.The supernatant was buffer exchanged to remove GuHCl using Amicon 3 kDa Centrifugal Filters.Samples were washed via spinning through filters at 12,000 x g for 20 minutes and resuspended in 500 µL of 10 mM Tris-HCl (pH 7) three times.Final protein aliquots were resuspended in 20 µL 10 mM Tris-HCl.Extracted protein was quantified using a bicinchoninic acid assay (BCA).
Proteolytic Digestion: 50 µg of protein lysate per sample was brought up to 4% SDS using a 10% SDS solution and then reduced using 20 mM dithiothreitol in 50 mM triethylammonium bicarbonate buffer (TEAB, pH 7) at 50 °C for 10 min.Samples were cooled to and held at RT for 10 min, and alkylated using 40 mM iodoacetamide in 50 mM TEAB (pH 7) at RT in the dark for 30 min.Samples were acidified with 12% phosphoric acid to obtain a final concentration of 1.2% phosphoric acid.S-Trap buffer consisting of 90% methanol in 100 mM TEAB at pH ~7.1, was added and samples were loaded onto the S-Trap mini spin columns (Protifi, Farmingdale, NY).The entire sample volume was spun through the S-Trap mini spin columns at 4,000 x g at RT, binding the proteins to the mini spin columns.Subsequently, S-Trap mini spin columns were washed twice with S-Trap buffer at 4,000 x g at RT and placed into clean elution tubes.Samples were incubated for one hour at 47 o C with sequencing grade trypsin (Promega, San Luis Obispo, CA) dissolved in 50 mM TEAB (pH 7) at a 1:25 (w/w) enzyme:protein ratio.Finally, additional trypsin solution was added at the same w/w ratio, and proteins were digested overnight at 37 o C. Peptides were sequentially eluted from the mini S-Trap spin columns with 50 mM TEAB, 0.5% formic acid (FA) in water, and 50% acetonitrile (ACN) in 0.5% FA.After centrifugal evaporation, samples were resuspended in 0.2% FA in water and desalted with Oasis 10 mg Sorbent Cartridges (Waters, Milford, MA).The desalted elutions were subjected to centrifugal evaporation were re-suspended in 0.2% FA in water at a final concentration of 1 µg/µL.Finally, indexed Retention Time Standards (iRT, Biognosys, Schlieren, Switzerland) were added to each sample according to manufacturer's instructions [30].
Mass Spectrometric DIA Data Processing: The DIA data files were processed in Spectronaut (version 15.6.211220.50606,Biognosys) using directDIA for both the protein and peptide level.Data was searched against the Mus musculus reference proteome with 58,430 entries (UniProtKB-TrEMBL), accessed on 01/31/2018.Data extraction parameters were set as dynamic and non-linear iRT calibration with precision iRT was selected.Trypsin/P was set as the digestion enzyme and two missed cleavages were allowed.Cysteine carbamidomethylation was set as a fixed modification while methionine oxidation and protein N-terminus acetylation were set as dynamic modifications.For the protein level, identification was performed requiring a 1% q-value cutoff on the precursor ion and protein level.Unique protein groups were reported with at least 2 unique peptide identifications.The protein level quantification was based on the peak areas of extracted ion chromatograms (XICs) of 3 -6 MS2 fragment ions, specifically b-and y-ions, with local normalization and qvalue sparse data filtering applied (Supp.Table S3).In addition, iRT profiling was selected.For the posttranslational (PTM) hydroxyproline-containing peptide level, proline oxidation was additionally set as variable modification, and identification was performed requiring a 1% q-value cutoff on the precursor ion and protein level.The PTM site localization score was selected with a probability cutoff of 0.75.Quantification was based on XICs of 3 -6 MS2 fragment ions, specifically b-and y-ions, without normalization as well as data filtering using q-value sparse.Grouping and quantitation of PTM peptides were accomplished using the following criteria: minor grouping by modified sequence and minor group quantity by mean precursor quantity.(Supp.Table S4).

Statistical Analysis and Comparison
Prism v.9 (Graph-pad, San Diego, CA) was used for all statistical comparisons except for ELISA and Proteomic results.Analysis of WT aging male cohorts utilized a one-way ANOVA with an initial alpha of 0.05 with a single factor of age at four different levels (2 mo, 1 yr, 2 yrs, 2 ½ yrs), and post-hoc comparisons used a Bonferroni corrected alpha of 0.008 (0.05 / 6) for differences in transcriptional activity across the 4 age groups compared.Two-Way ANOVA comparisons for TβRII ocy-/-and Cre-negative controls used factors of age and genotype.
Significance (p) values for all interaction terms and main effects of age and genotype can be found in Supplemental Table S5.For Two-Way ANOVA analyses, factors of age and genotype were considered at an alpha level of 0.05.While sex differences are noted throughout, they were not statistically compared to each other due to their difference in biologic mechanism and each sex was run in independent Two-Way ANOVA comparisons.Within each Two-Way ANOVA, age had up to three independent levels (4, 12 and 15 months depending on ages included) and genotype had 2 (Cre-negative control, TβRII ocy-/-).A priori planned pairwise post hoc comparisons were completed with the Fisher's Least Significance Difference Test that does not correct p-value for multiple comparisons, and instead the significant level (alpha = 0.05) was Bonferroni corrected for the number of comparisons being made.In a planned manner, not all possible pairwise comparisons are scientifically meaningful and power was not sacrificed for excessive comparisons.In post hoc analysis for factors that violated the Fisher's LSD assumption that all groups are sampled from populations with the same standard deviation, standard Student's T-tests were completed for post hoc assessment of significance.Because ELISA analysis had additional group complexity (age, sex, genotype, tissue source, and ligand sub-type), full ensemble results were first analyzed with univariable and multilevel Linear Mixed Models of the log of concentration to determine what factors most strongly predicted ligand concentration (Supplemental Table S6).Afterward, pairwise comparisons were completed with the non-parametric Wilcoxon signed-rank test with significant alpha values Bonferroni corrected for the specific number of comparisons being completed in each pairwise comparison.Synchrotron tissue vs material strain regressions were completed with a linear least squares method to generate linear slopes for statistical comparisons via an extra sum-of-squares F test that compares independent fits of a pair of regressions with a global fit.Regressions were considered statistically different if the p-value for the comparison of slopes of independent regressions to the global fit was below alpha = 0.05/4 = 0.0125 for a 4-way Bonferroni correction for multi-comparisons.P-values considered 'trending' between p<0.1 and Bonferroni corrected alphas are noted on figures.Differential expression analyses of mouse femur proteomes were performed using a paired t-test, and pvalues were corrected for multiple testing using the Storey method [35].Specifically, group wise testing corrections were applied to obtain q-values.For the protein level, protein groups with at least two unique peptides, q-value < 0.05, and absolute Log2(fold-change) > 0.58 were considered significantly-altered (Supp.Table S3).For the quantification of the hydroxyproline-containing peptides, post translational modification (PTM) fold-change values were normalized by the corresponding protein fold-change values, and hydroxyprolinecontaining peptides with p-value < 0.01 and absolute Log2(normalized fold-change) > 0.58 were considered significantly-altered (Supp.Table S4).Volcano plots were generated using the ggplot2 package [36] in R (version 4.0.5;RStudio, version 1.4.1106).