Relationship between altered knee kinematics and subchondral bone remodeling in a clinically translational model of ACL injury

1 INTRODUCTION

Anterior cruciate ligament (ACL) injury is a debilitating condition that accelerates the onset of posttraumatic osteoarthritis (PTOA), a disease plaguing more than 50% of ACL injured patients.1 While most patients opt to receive surgical reconstruction, restabilizing the joint has been largely unsuccessful at reducing the risk of developing PTOA.2 Abnormal joint kinetic and kinematics are commonly reported in the acute and chronic stages of recovery3, 4 and have long been mechanistically implicated as a primary driver of PTOA onset.5 These biomechanical adaptations are thought to cause a nonuniform redistribution of mechanical load on articular cartilage,5 which can lead to degenerative joint changes (identified by superficial fraying or splitting of cartilage and maladaptive subchondral bone architecture) associated with PTOA onset. Though strongly theorized,5 it is unclear to what extent biomechanical adaptations after ACL injury culminate in the development of PTOA, as data that directly connects these factors does not exist. Work that can clearly establish this relationship represents a significant contribution to the literature, as it can be used to substantiate and develop empirically driven research programs.

The majority of work has used ACL transection models to study measures of PTOA.6-13 Though practical, transection models lack in clinical translation as they violate the joint capsule creating a cascade of negative neuromuscular events that confound observations. For this reason, noninvasive models have been developed that more closely replicate the human mechanism of ACL injury and are particularly advantageous to fully characterize disease progression without the confounding effects of surgical transection.14-16 Despite this evolution, the concurrent investigation of joint kinematics with PTOA progression remains virtually unexplored,17 even with it being a fundamental area of focus for those with a history of joint injury.

To overcome many of these challenges that limit insight into disease progression, the goal of this study was to use our noninvasive rodent model of ACL injury to explore the direct effect of an isolated ACL injury on longitudinal joint kinematics and the pathogenetic mechanisms involved in the development of PTOA. To do this, we isolated observations directly to the injury by withholding all analgesics that are known to interfere with gait analysis18 and bone remodeling.19, 20 Then, we used marker-less deep learning software, DeepLabCut,21 to track animal movement and performed comprehensive kinematic analyses to better understand the longitudinal changes in knee kinematic profiles. We then investigated subchondral bone architecture via micro computed tomography (μCT) to determine the association between hallmark measures of joint kinematics and PTOA after ACL injury. As a supplementary analysis, we also investigated sex and regional differences in bone health across the distal femur and proximal tibia.

2 METHODS 2.1 Experimental animal models

40 Long-Evans, 16-week-old male and female rats were obtained from Envigo Laboratories and were randomly assigned to 5 groups: 1 control and 4 ACL-injury (n = 4 male/4 female per group). All animals were housed in individual cages within the vivarium, had a 1-week acclimatization period, and were allowed food and water ad libitum for the duration of the study. Before injury and during the 1-week laboratory acclimatization period, all rats were trained to walk on the motor-driven treadmill with a stable walking pattern. Animals were maintained in accordance with the National Institutes of Health guidelines on the care and use of laboratory animals. This study was approved by our Institutional Animal Care and Use Committee (IACUC approval #A17-042).

2.2 Noninvasive ACL injury

A single load of tibial compression was used to induce knee injury to the right limb of all ACL injury rats. Briefly, rats were anesthetized (5% induction, 2% maintenance isoflurane/500 ml via a nose cone) and the ACL injury was induced in a custom built device (Figure 1) that was instrumented with a linear accelerator (8 mm/s loading rate, model: DC linear actuator L16-63-12-P; Phidgets), load cell (PW6D; HDM Inc), and custom-written software program (LabVIEW; National Instruments) that monitored the release of tibial compression during the load cycle, signifying an ACL tear. After the ACL rupture, a Lachman's test was performed to clinically confirm an ACL injury had occurred by detecting excessive anterior tibial translation while under the plane of anesthesia. The hindlimb was also palpated to detect any gross bone damage. If no contraindications (fractures, damage to other ligaments) were identified, the animal was transferred back to its cage and allowed to recover.

image Mechanism of Injury. Custom built device used to tear a rat anterior cruciate ligament through a single load of tibial compression. The bottom plate holds the flexed knee, while the ankle is fixated into approximately 30° flexion. The knee is then loaded at a rate of 8 mm/s, increasing the compressive forces on the tibia and causing anterior subluxation of the tibia relative to the femur. The compressive force imposed on the tibia releases when the anterior cruciate ligament is torn, signaling the device to retract [Color figure can be viewed at wileyonlinelibrary.com]

All 32 injured rats had a positive Lachman's test at the time of injury. ACL injuries were then confirmed in all 32 injured rats upon dissection following μCT scanning (with no fractures, ruptures or avulsions of other tissues). Mean compressive force at the knee injury for the females and males was 84.81 ± 11.9 N and 96.11 ± 28.2 N, respectively. Mean tearing rate of the knee injury for the females and males was 193.49 ± 32.3 N/s and 235.81 ± 51.3 N/s. Notably, in accordance with IACUC approval, analgesics were withheld throughout the duration of the study to ensure all findings were isolated to the injury, as the use of nonsteroidal anti-inflammatory drugs and opioids are known to directly interfere with the natural native biological response.22, 23

2.3 Time points, gait collection, and analyses

To detail the time course of knee biomechanical adaptations after ACL injury associated with PTOA a longitudinal study design was utilized, where gait analyses were conducted at one of 4 time points (7, 14, 28, 56 days after injury). Before each session, the right hindlimb was shaved to better define the ankle, knee, and hip in each recording. Walking gait was then collected on a level-ground motor-driven treadmill (EXER 3/6 treadmill) set to 16 m/min24 and recorded at 250 frames per second using OptiTrack Prime Color cameras and Motive software (version 2.1.1).

DeepLabCut, markerless pose-estimation software previously shown to achieve accuracy levels comparable to human labeling,21 was used to track the hip, knee, and ankle in the sagittal plane. Videos were clipped to 5–10 s durations. Frames were then extracted and labeled by one single investigator (MSW) using DeepLabCut's graphical user interface. We used a ResNet-50 based neural network and trained it on 95% of the labeled images. A reiterative process was used to refine the model where frames with poor tracking results, through visual inspection or labels that jumped a pixel distance between frames, were extracted, relabeled, and added back to the model for retraining. The network was considered satisfactory when the training and test errors reached 2.7 and 5.39 pixels, respectively (660,000 iterations; image size: 1920 × 1080 pixels). We used a p-cutoff of 0.9 to condition the (X,Y) coordinates for future video analyses.

The trained network was applied to our unseen videos for analysis and used to predict kinematic landmarks of the hip, knee and ankle joints (X,Y coordinates). The DeepLabCut-tracked trajectories were then imported into Python (3.7), where custom-written software was used to process all kinematic data. Knee sagittal plane angular displacement was extrapolated from the (X,Y) coordinates of each labeled landmark. Three consecutive step cycles for each rat were then extracted and each stride was time normalized to 100% of the step cycle that included both swing and stance phases. Average peak knee flexion angle was extracted from the stance phase of each rat from the time normalized strides. Knee flexion angle throughout the entire step cycle and the average peak knee flexion angle were both utilized for statistical analysis.

2.4 μCT collection and analyses

To quantitatively assess bone health associated with PTOA, the ACL injured rats were euthanized (via carbon dioxide overdose) following gait analysis at one of 4 time points (7, 14, 28, 56 days after injury). Uninjured control rats were euthanized on day 56. ACL-injured knee joints were scanned using μCT (Zeiss XRM Xradia 520 Versa; Zeiss Microscopy) with respective rodent bone settings (beam setting = 70 kV/6 W, pixel size = 12.25 μm) and subsequently analyzed using Dragonfly image analysis software (version 4.1; Object Research Systems).

μCT data were imported in DICOM format and a normalization filter was applied to uniformly distribute subject specific grayscale histograms. Image stacks were reoriented such that two-dimensional view scenes were aligned with their respective anatomical views (axial, sagittal, and coronal). Images were then manually segmented by one investigator (RJB), blinded to time point and injury, and then further subdivided into specific regions of interest (ROI's; Figure 2).14, 15, 25 Subchondral trabecular bone and the subchondral bone plate of the distal femur and proximal tibia were analyzed in the medial and lateral load bearing regions. Femoral load bearing regions were separated at the intercondylar notch and excluded the trochlear region, and tibial load bearing regions were separated at the intercondylar eminence.14

image Micro computed tomography segmentations and regions of interest used in analyses. First, trabecular (white) and subchondral bone plate (blue) were segmented in the femur (A) and tibia (D). The segmentations were then subdivided into regions of interest for the medial (blue) and lateral (yellow) compartments for trabecular bone (B and E) and subchondral bone plate (C and F). The enthesis region of interest for trabecular bone (yellow) and subchondral bone plate (blue) are shown in (G) with a 3D representation in (H) [Color figure can be viewed at wileyonlinelibrary.com]

To quantify microarchitectural changes, multiple bone related measures of the subchondral trabecular and subchondral bone plate were calculated. Trabecular bone measures were sampled from the entire load bearing regions (Figures 2B and 2E) and included bone volume fraction (BV/TV, %), trabecular number (Tb.N, 1/mm)25 and tissue mineral density (TMD; mg HA/cm3). Subchondral bone plate measures were sampled from a 1.04 × 1.04 mm2 region of the subchondral bone plate centered in each of the medial and lateral regions of the femur and tibia (Figures 2C and 2F)26 and included subchondral bone plate thickness (SB.Pl.Th, µm) and porosity (Sb.Pl.Po, %). To determine TMD, grayscale units were transformed to hydroxyapatite concentration (mg HA/cm3) using a bone phantom (model: Micro-CT HA D10, Quality Assurance in Radiology and Medicine) which was scanned with the same settings as the samples.

Clinically meaningful measures of PTOA were also assessed from μCT including the medial and lateral joint space width and anterior tibial translation (i.e., position of tibia relative to femur). Joint space width was measured as the minimum distance between the distal edge of the femoral condyle and the proximal edge of the tibial plateau. Anterior tibial translation was extracted from sagittal plane µCT images and measured as the perpendicular distance between a line tangent to the most posterior point of the femoral condyle to a line tangent to the most posterior point of the respective tibial plateau27 on both the lateral and medial sides of the joint. To further extend insight, the femoral ACL enthesis was also explored using the aforementioned microarchitecture measures. This ROI was included, as emerging data indicate that this region of bone is particularly susceptible to PTOA, and thus is an important clinical consideration to more fully illustrate disease progression.28, 29 To capture this ROI, a 1.2 mm diameter cylinder was placed at a 55° angle30 from the tibial articular surface in the frontal plane and at a 30° angle from the femoral articular surface in the axial plane such that the center of the cylinder intercepted the center of the enthesis (Figure 2G). Then, both trabecular and cortical bone were analyzed at the bottom third, or aperture of the cylindrical volume.31 Supportive, qualitative images that characterize the extent of bone remodeling are depicted in Figure 3.

image

Representative bone structures by micro computed tomography at each time point following injury

2.5 Statistical analyses

To detail the time-course of knee biomechanical adaptions after ACL injury, functional analyses of variance (ANOVA) were performed using the functional data analysis package in R (version 1.1.5). This approach treats knee joint angles as a polynomial function thereby allowing for the detection of between-group (ACL-injured and control) differences at any percent of the step cycle. This allows for a more comprehensive analysis by comparing the timing and magnitude of knee flexion throughout the entire step cycle rather than at a single percent (i.e. peak knee flexion angle).32 Significant differences were considered to exist when 95% confidence intervals did not overlap zero and were plotted on top of ensemble averages (Figure 4). To detail the time-course of PTOA after ACL injury, two-way ANOVA were performed on all μCT measures with factors of time since injury and sex (male/female) entered into the model. If there was no interaction effect, main effects of time since injury and/or sex were evaluated. In the case of a significant sex by time interaction post hoc one-way ANOVAs and independent t-tests were performed. Fisher's least significant difference post hoc analyses were used to assess pair-wise effects for significant main effects relative to the control group and interactions. To determine regional differences in the femur and the extent of anterior tibial translation, relative difference (relative difference (x,xreference) = Δ/|xreference|) scores were calculated at each time point after ACL injury and two-way ANOVA were performed. For regional differences in the femur, factors of region (enthesis, lateral, medial) and sex were entered into the model. For anterior tibial translation, factors of time and sex were entered into the model with the post-hoc analysis as described above. Finally, to explore biomechanical factors that were associated with bony markers of PTOA after ACL injury, Pearson correlations were conducted to investigate the relationship between peak knee flexion angles during the stance phase of gait and porosity, an early-hallmark measure of PTOA33 that is closely related to articular cartilage degeneration.34, 35 Unlike the functional data analysis described above, peak knee flexion angle is a discrete measure that allows for statistical comparisons. All analyses were performed using RStudio (version 1.1.5), with α levels of ≤.05 required for statistical significance.

image

Longitudinal female (Panels A, B, C, D) and male (Panels E, F, G, H) kinematics and functional analyses of variances between anterior-cruciate-ligament-injured and control rats. The dashed vertical line indicates the percent of the step cycle where average peak knee flexion occurred. Average knee flexion angles across the step cycle are shown for each group with the shaded regions indicating significant differences between limbs. p < .05

3 RESULTS

The primary goal of our work was to directly explore the effect of an isolated ACL injury on longitudinal joint kinematics and the pathogenetic mechanisms of PTOA via detailed analyses of the subchondral bone. Hence, outcomes that directly illuminate this sequala and test this relationship are described in full detail below. Supportive measures that fully characterize noninvasive ACL injury model are described in Tables 2 and 3, and Figures 6-9.

Table 1. Subchondral bone plate remodeling was observed as a function of time and sex, with evidence in the medial and lateral femur and tibia Subchondral bone plate Control Day 7 Day 14 Day 28 Day 56 Femur Sb.Pl.Th (μm) Laterala Females 202.86 ± 23.99 170.02 ± 6.67 164.46 ± 6.49 199.41 ± 13.01 184.55 ± 11.96 Males 218.69 ± 8.51 206.22 ± 3.08 211.97 ± 21.02 198.70 ± 21.83 225.24 ± 8.67 Mediala Females 222.48 ± 17.67 170.60 ± 4.81 198.37 ± 5.82 200.36 ± 5.94 186.45 ± 4.43 Males 233.12 ± 22.18 215.90 ± 12.64 231.65 ± 9.95 248.05 ± 10.29 241.83 ± 18.89 Sb.Pl.Po (%) Lateral Females 4.86 ± 1.06 12.34 ± 0.51b 20.48 ± 2.30b 18.34 ± 2.13b 13.54 ± 1.70b Males 4.14 ± 0.50 15.03 ± 0.87b 19.55 ± 2.20b 22.20 ± 1.96b 13.34 ± 1.40b Medial Females 5.05 ± 0.36 9.59 ± 0.40b 10.87 ± 0.60b 14.53 ± 0.57b 10.33 ± 0.95b Males 4.03 ± 0.24 10.74 ± 0.41b 12.06 ± 1.41b 16.06 ± 0.59b 11.84 ± 0.65b Tibia Sb.Pl.Th (μm) Laterala Females 217.98 ± 13.43 176.49 ± 12.08b 178.51 ± 5.45b 168.66 ± 12.51b 179.49 ± 11.47b Males 257.59 ± 11.41 205.11 ± 11.02b 220.84 ± 14.57b 233.77 ± 22.69b 231.47 ± 8.48b Mediala Females 218.52 ± 15.42 165.01 ± 8.77 155.67 ± 9.89 154.12 ± 10.16 170.21 ± 5.33 Males 243.97 ± 12.68 236.16 ± 15.81 244.66 ± 21.17 289.60 ± 42.72 245.78 ± 17.15 Sb.Pl.Po (%) Lateral Females 7.93 ± 0.94 13.06 ± 0.99b 17.72 ± 1.64b 19.09 ± 1.92b 13.90 ± 1.03b Males 4.73 ± 0.93 11.73 ± 0.41b 17.73 ± 2.60b 19.21 ± 0.36b 12.50 ± 0.43b Medial Females 6.20 ± 0.71 9.57 ± 0.42b 11.71 ± 0.94b 15.99 ± 1.59b 11.62 ± 0.96b Males 4.43 ± 0.63 10.30 ± 0.54b 11.32 ± 0.55b 18.03 ± 0.84b 11.38 ± 1.06b Enthesis Sb.Pl.Th (μm) Females 200.91 ± 8.45 175.89 ± 5.74b 166.18 ± 6.80b 205.45 ± 9.43 221.29 ± 11.75 Males 232.41 ± 8.74 187.38 ± 14.92b 188.36 ± 8.29b 215.26 ± 14.85 204.00 ± 24.44 Sb.Pl.Po (%) Females 4.26 ± 0.29 11.02 ± 1.07b 15.26 ± 0.89b 17.92 ± 0.70b 16.27 ± 1.67b Males 5.74 ± 0.22 11.63 ± 0.72b 14.24 ± 0.83b 19.55 ± 0.83b 16.30 ± 0.63b Note: Most notably, Sb.Pl.Po was found to be significantly elevated at every time point following injury in all regions of interest (p < .05). Data are represented as mean ± SE. Alpha level, p < .05. Abbreviations: Sb.Pl.Po, subchondral bone plate porosity; Sb.Pl.Th, subchondral bone plate thickness; Sb.Pl.Po, subchondral bone plate porosity. Table 2. Femoral and tibial trabecular bone remodeling was observed as a function of time and sex, and significant changes primarily resided in the medial compartments starting 14 days after injury Epiphyseal trabecular bone Control Day 7 Day 14 Day 28 Day 56 Femur

留言 (0)

沒有登入
gif