3D Light Sheet Microscopy Reveals Aging Osteoarthritis-associated Joint-innervated Nerve Remodeling in Mouse Knee Joints

Experimental animals

A total of n = 38 mice (both male and female) were used for the study. Wild-type (WT) C57BL/6 mice were grouped as young (3–5 months) and aged (20–22 months). Ethical approval was obtained by the University Committee on Animal Resources (UCAR Protocol #2019–008). For details, see Suppl. Methods.

Pressure application measurement (PAM) assay

Knee hyperalgesia was measured using a PAM device (Ugo Basile, Comerio, Italy) (19). Briefly, mice were scruffed and an increasing magnitude of force was applied to the hind limb knee joint at a rate of 30 g/s, up to a maximum of 450 gf. Knee joint withdrawal and/or vocalization were used as indicators of the pain bearing threshold. Results are expressed as knee withdrawal threshold (gf). The assessment was performed by an investigator blinded to the experimental groups.

Safranin O/Fast Green (Saf-O/FG) staining

Mouse knee joints were harvested, fixed in 10% neutral buffered formalin for 72 h, and decalcified in 14% ethylenediaminetetraacetic (pH 7.4) for 2 weeks at room temperature (20). Tissues were paraffin-embedded, sectioned sagittally at 7 μm, deparaffinized, rehydrated through graded ethanol to water, and stained with Saf-O/FG to visualize articular cartilage.

DRG isolation and immunolabeling

Lumbar dorsal root ganglia (L3–L5) were collected from young and aged mice following CO₂ euthanasia, fixed in 4% paraformaldehyde, cryoprotected, embedded, and sectioned at 10 μm. Sections were immunostained with primary and secondary antibodies (Supplementary Table1 and 2) and counterstained with DAPI. Images were acquired using a Zeiss Axio Imager 2 microscope and analyzed in ImageJ. PIEZO2 expression was quantified by threshold-based integrated density measurements of individual neuronal soma, normalized to the M_y group and reported fold changes represent back-transformed estimates from the log-scale analysis. The proportion of CGRP-immunoreactive neurons was calculated as a percentage of total PGP9.5⁺ neurons, with full methodological details provided in the Supplementary Methods.

iDISCO+ volume immunostaining of mouse knee joints

Whole-mount volume immunostaining and optical clearing of mouse knee joints were performed using an iDISCO+-based protocol with minor modifications (16). Following fixation and decalcification, samples were processed for methanol dehydration, autofluorescence bleaching, immunolabeling with primary and secondary antibodies, and organic solvent–based clearing. Cleared knee joints were rendered optically transparent in dibenzyl ether prior to imaging (Fig. 1B). For methodological details, see Supplementary Methods and Table 1 and 2.

Light sheet microscopy of iDISCO-cleared mouse knees

Mouse knees stained and cleared via the iDISCO method described above were imaged on a Luxendo Multi-View Selective-Plane Illumination Microscope (MuVi-SPIM, Bruker) in the single detection, 4X octagon. Samples were positioned in a 3D printed holder for imaging in the frontal plane. Dual-sided illumination was performed with a series of diode laser lines (488 nm, 561nm, 642 nm) and a pair of 4X illumination objectives (Nikon, 0.2 NA). The detection arm consisted of a 4X objective (Olympus, 0.28 NA), a 1X zoom lens, corresponding bandpass filters (BP500–530 nm, BP580–627 nm, BP 655–704 nm), and a CMOS camera (Hamamatsu Orca Flash 4.0 v3). Z-stack images (pixel size of 1.46 μm/pixel) were obtained at 3 μm increments (4.4X effective magnification), and tiled image acquisition was controlled by Lux Control software (Luxendo, v.4.4.4) set to ≥20% overlap (2400 μm) lateral step size.

3D image reconstruction and segmentation

Volumetric image reconstruction and 3D nerve segmentation were performed using Lux Processor and Imaris software. Tiled images were registered on the PGP9.5 (561 nm) channel, converted to .ims format, and rendered in three dimensions within a standardized region of interest encompassing the medial–lateral knee joint and proximal tibial epiphysis, including periarticular tissues. CGRP signal was quantified within reconstructed PGP9.5⁺ filaments and setting a fixed threshold to measure out the CGRP fibers from the overall nerve filaments. Filament metrics like length and branch points were exported for quantitative analysis. For details, see Suppl. Methods.

Computational geometries of reconstructed nerve networks and data normalization

Spatial coordinates from reconstructed PGP9.5⁺, CGRP⁺/PGP9.5⁺, and TH⁺ nerve networks were exported from Imaris and analyzed using custom MATLAB scripts. 3D α-shapes were generated to define the spatial boundaries of each nerve network, and network volumes were estimated after filtering spurious segmentation identified as spatial outliers using proximity-based metrics and multivariate analysis (21, 22). Network volume was then used to normalize total nerve length, while branch point counts were normalized to segment length. Medial–lateral nerve distribution was assessed by k-means clustering of segment coordinates, with full computational and statistical details provided in the Suppl Methods.

Statistical analysis

All statistical analyses were performed using GraphPad Prism v10 (GraphPad Software, Boston, Massachusetts, USA). Biological replicate numbers (n) are indicated in the figure legends. Data normality was assessed for each dataset using the Kolmogorov Smirnov normality test, and data were log10transformed when the assumptions of normality were not met prior to statistical analysis. Fold changes reported in the Results represent back-transformed estimates derived from the log-scale analysis. OARSI scores, PAM measurements, and imaging-derived x–y–z coordinate lengths were analyzed using Kruskal–Wallis tests followed by Dunn’s multiple-comparisons post hoc tests. All other outcome measures were analyzed using two-way ANOVA to assess the effects of age (young vs. aged), sex (male vs. female), and their interaction, with model assumptions evaluated by residual diagnostics. Post hoc comparisons were performed using Fisher’s LSD test. No statistical analyses were conducted for datasets with n ≤ 3. Medial–lateral nerve distribution was assessed using repeated-measures two-way ANOVA with compartment (medial vs. lateral) matched within animals, followed by Fisher’s LSD post hoc testing. Detailed ANOVA statistics are provided in Supplementary Table S3 and S4.

Aging enhances knee osteoarthritis severity and nociceptive signaling in male mice

The experimental design (Fig. 1A) included young male (M_y), aged male (M_a), young female (F_y), and aged female (F_a) mice. Structural joint changes were assessed and OARSI scoring Fig. 1B–C). The M_y group displayed intact cartilage, whereas the M_a group exhibited marked loss of SafO staining, indicating proteoglycan depletion and OA progression. OARSI scores differed significantly among groups (H = 16.56, p = 0.0009, ε² = 0.75), with the M_a group showing significantly higher scores than the M_y (adjusted p = 0.0048); no age-related differences were observed in females. Knee joint pain was evaluated using pressure application measurements (PAM) Fig. 1D. Withdrawal thresholds differed among groups (H = 11.59, p = 0.0089, ε² = 0.64), with M_a group exhibiting significantly reduced thresholds compared with M_y (adjusted p = 0.0048), consistent with increased mechanosensitivity. No significant effects of age or sex were observed in females.

To characterize DRG neurons associated with mechanical sensitivity and joint pain, we examined expression levels of CGRP and PIEZO2 on DRG sections isolated from lumbar segments L3-L5 of M_y, M_a, F_y, and F_a groups (Fig. 1E). The proportion of CGRP⁺ neurons was 1.31fold higher in aged males compared with young males (p = 0.0017). No age-related difference was observed in females (p = 0.8809). CGRP⁺ neuron proportions did not differ between males and females in young animals (p = 0.0741). However, aged males exhibited a 1.44-fold higher proportion of CGRP⁺ neurons compared with aged females (p = 0.0003) (Fig. 1F). Similarly, PIEZO2 expression in DRG neurons showed an age-dependent increase in males. Aged males exhibited approximately a 2.1-fold higher PIEZO2 expression compared with young males (p = 0.0057). In contrast, no significant age-related difference was observed in females (p = 0.5617). When comparing sexes, young females displayed ~1.9-fold higher PIEZO2 expression than young males (p = 0.0139), whereas no significant sex difference was detected in aged animals (p = 0.9304) (Fig. 1G).

Increased PGP9.5⁺ Nerve Innervation in Knee Joints of Aged Male Mice

Following iDISCO clearing and immunolabeling (Supplemental Fig. S1), knee joints were imaged by 3D light sheet microscopy. An anterior knee joint region of interest (ROI) spanning the medial–lateral width and extending from the inferior patellar pole to the proximal tibial epiphysis was selected (Fig. 2A). This region contained a dense PGP9.5⁺ nerve network distributed across the femur, tibia, infrapatellar fat pad, and adjacent soft tissues, primarily within the superficial joint capsule and closely associated with vasculature (Fig. 2; Supplementary Movies M1-4). ROI dimensions were consistent across all groups, with no age- or sex-related differences (Fig. 2B). Age-related changes in knee joint innervation were evaluated by labeling joint-innervating nerves with the pan-neuronal marker PGP9.5 and performing 3D filament reconstruction using Imaris (v10.1.1) (Fig. 2C). For the filament normalization, the 3D rendered PGP9.5+ nerve network defined in Imaris was imported to MATLAB, where computational geometry was used to calculate the volume of 3D boundary regions enveloping the PGP9.5+ nerve segments (Fig. 2D). Analysis of PGP9.5⁺ sensory nerve architecture revealed significant age and sex dependent differences in the knee joint. Aged males exhibited approximately a 2.0-fold higher total PGP9.5⁺ nerve fiber length compared with young males (p = 0.0003), whereas no age-related difference was observed in females (p = 0.7923). Similarly, PGP9.5⁺ nerve branching complexity was markedly increased in aged males, with approximately a 3.6-fold higher number of branch points compared with young males (p = 0.0001), while females showed no significant age-related change (p = 0.5369). No significant sex differences were observed in young animals for either total nerve fiber length (p = 0.9329) or branch points (p = 0.7452). However, aged males displayed significantly greater total nerve remodeling than aged females, including approximately a 2.1-fold higher total PGP9.5⁺ nerve fiber length (p = 0.0065) and approximately a 3.8-fold higher number of branch points (p = 0.0029).(Fig. 2 F-I). The PGP9.5⁺ nerve length and branch points with the boundary volume normalization exhibited similar trends (Supplementary Table 3).

Aged Male Mice Exhibit Increased CGRP⁺ Nociceptor Innervation in Knee Joints Compared to Young Male group

To assess age-associated changes in pain-sensory innervation in osteoarthritic knee joints, tissues were co-stained for the pan-neuronal marker PGP9.5 and the nociceptive neuropeptide CGRP (Fig. 3A). Given that CGRP is a secreted neuropeptide and is also present in vascular smooth muscle surrounding blood vessels, the complete nerve network was first reconstructed in 3D using PGP9.5⁺ filaments in Imaris (v10.1.1). A CGRP intensity threshold was then applied to selectively render colocalized CGRP⁺/PGP9.5⁺ sensory filaments. 3D imaging revealed dense peptidergic sensory innervation within and surrounding the knee joint, particularly in extra-synovial fat pads, including the infrapatellar fat pad, and the joint capsule (Fig. 5A, B). Quantitative analysis showed increased CGRP⁺/PGP9.5⁺ filament length and branching in M_a compared with M_y, whereas females showed minimal age-related differences (Fig. 3B–E). Normalization to the 3D PGP9.5⁺ network volume yielded the similar trend, confirming enhanced sensory innervation with aging in male mice.

The Density of TH⁺ Sympathetic Nerve Fibers in Knee Joints Remains Stable During Age-Related Osteoarthritis

We next examined sympathetic nerve innervation in the anterior knee region of young and aged mice using anti-tyrosine hydroxylase (TH) labeling (Suppl. Movies M5–8). Representative 3D micrographs of TH⁺ sympathetic networks are shown in Fig. 4A. Total TH⁺ filament length and branch points were quantified in Imaris and normalized in MATLAB (Fig. 4B). Notably, no age-related differences were observed in TH⁺ filament length or branching in either sex (Fig. 4C–G; Supplementary Table 3). These findings indicate that, unlike pain-sensory nerves, sympathetic nerve density and branching remain largely unchanged with aging in osteoarthritis. In both M_y, M_a, and F_a group, TH⁺ sympathetic fibers consistently innervated extra synovial fat pads and the joint capsule (Fig. 5C, D).

Spatial Distribution of Nerve Segments Between the Medial and Lateral Sides of the Knee Joint

To further characterize the spatial distribution of nerve network in knee joints, we segmented PGP9.5⁺, CGRP⁺/PGP9.5⁺, and TH⁺ filaments in medial and lateral sides using MATLAB (Fig. 6 and Fig. 7). In male mice, PGP9.5⁺ nerves and CGRP⁺/PGP9.5⁺ sensory nerves were distributed across the medial and lateral sides of the knee joint without trends (Fig. 6B, C). However, TH⁺ sympathetic nerves were more robustly innervated in the medial aspect of the knee joint in both M_y and M_a groups (Fig. 6D) (Suppl. Table 4). In contrast, female mice revealed that PGP9.5⁺ and TH⁺ nerves were distributed across the medial and lateral sides of the knee joint without trends (Fig. 7B, D); yet CGRP⁺/PGP9.5⁺ sensory nerves were more robustly innervated in the lateral sides than the medial sides of the F_y group (Fig. 7C). Together, these data suggest the region-dependent joint-innervating nociceptive and sympathetic nerve remodeling over OA progression.