Anterior cruciate ligament overuse increases collagen fibril compliance: collagen unraveling and elevated glycosaminoglycan activity in a murine model

Abstract

Data increasingly indicates that many anterior cruciate ligament (ACL) injuries are likely due to tissue damage accumulation and the gradual decrease in ACL strength and stiffness over time. Previous ex vivo experiments found molecular-level ACL damage (collagen unraveling) to be associated with increased tissue compliance at the collagen fibril-level. However, it remains unclear if this biomechanical effect is present in vivo, or what effect other modulators of tissue mechanics (e.g., glycosaminoglycans (GAGs)) have in fatigued-damaged ACL matrices. Herein, we investigated the in vivo relationship between collagen unraveling and fibril-level mechanics in response to ACL fatigue loading in a mouse model and determined whether GAGs are also involved in altering tissue mechanics in this injury scenario. Mouse ACLs were submaximally fatigue loaded in vivo and harvested after one hour of tissue recovery. Compared to their contralateral controls, fatigued ACLs showed increased collagen unraveling and elevated GAG signatures within damaged extracellular matrix (ECM) regions. These regions also exhibited higher fibril-level compliance. Spectral data indicated colocalization of unraveled collagen and GAG-rich domains. Versican, a large aggregating PG, emerged as a candidate contributor to increased GAG presence in these regions. Together, these findings suggest that fatigue-induced collagen unraveling promotes local GAG accumulation – mediated by PGs such as versican – which likely contributes to reinforcement of the ECM.

1. Introduction

There are 4 major ligaments in the knee that serve to stabilize the joint while maintaining a degree of flexibility to guide normal joint movements [1]. The mechanical requirements include the ability to withstand both creep (static) and fatigue (dynamic, cyclic) loading throughout daily activities. Ex vivo [2] and in silico [3] data estimate that the anterior cruciate ligament (ACL), which is frequently injured, spends much of its time operating within the toe and early linear regions of the stress-strain curve throughout most routine tasks. However, athletic movements (e.g., sudden deceleration, cutting, landing) increase the strain transferred to the knee ligaments as they stiffen to resist stretch and maintain joint stability [4] This scenario can result in extracellular matrix (ECM) damage and functional decline as the movement is repetitively performed, resulting in an overuse injury [5,6]. Knee overuse injuries are prevalent among young active individuals; however, they are disproportionally skewed towards females [7]. Structurally, submaximal cyclic (i.e., fatigue) loading of ligaments and tendons induces ECM damage across collagen length scales, including collagen unraveling (molecular-scale), fibril kinking (nano-scale), and fiber kinking and disorganization (micro-scale) [8–14]. As damage accrues over time, it is accompanied by a progressive decline in gross tissue function, including reductions in tissue creep, stiffness, elastic modulus, and strength [15]. Despite this established structure-function relationship and knee injury risk, little attention has been given to the gradual generation and accumulation of ECM damage that likely precedes many noncontact knee ligament injuries, particularly those to the anterior cruciate ligament [16,17].

Previously, we reported that ex vivo submaximal fatigue loading of the human ACL results in increased collagen unraveling as compared to controls [12,13,18]. Additionally, tissue domains with larger concentrations of unraveling were more mechanically compliant (i.e., less stiff) than regions containing less damage. This association between collagen unraveling and increased tissue compliance was greatest within the more fibrocartilaginous matrices comprising the ACL enthesis [18]. These molecular and mechanical findings support earlier reports of others concluding that mechanically-induced unraveling of collagen alpha-chains allows for further strain energy absorption [19,20], providing a temporary fail-safe under continued loading. Another factor that may also be associated with an increase in tissue compliance during fatigue loading are proteoglycans (PGs). Of these PGs, decorin, a small leucine-rich proteoglycan (SLRP) with an affinity for type I collagen binding, is the most prevalent in ligament midsubstance [21,22].

Mechanically, decorin provides lubrication to collagen fibrils to facilitate fibril sliding under tensile forces and thus increase viscoelasticity [23]. There are also large-aggregating proteoglycans, such as aggrecan and versican, which are most abundant in ligament. Here, they provide biomechanical resistance to multi-axial forces while maintaining tissue elasticity. The mechanical contributions of PGs to the ECM arise from their negatively charged glycosaminoglycan (GAGs) side chains, that allow for the sequestration of water and divalent cations [24]. Most GAGs present in ligament are sulfated and include keratan, chondroitin, and dermatan sulfates. Hyaluronic acid is the lone non-sulfated GAG. Importantly, an increase in chondroitin sulfated GAGs has been found to be associated with submaximal tendon injuries, implicating them in mechanical decreases in elastic modulus and ultimate tensile strength relative to healthy connective tissue [25,26]. However, outside of full thickness ligament injury or ex vivo GAG depletion experiments [27], there has been little if any investigation into the relationship between submaximal ligament overuse injuries, the resulting alteration of mechanical properties, and GAG activity.

To investigate the relationship between fatigue-induced ECM damage (i.e., unraveled collagen), GAG activity, and increased tissue compliance, we employed a mouse model to submaximally fatigue the ACL in vivo [14]. We tested the hypothesis that PGs and associated GAG side chains contribute to the increased tissue compliance in domains of unraveled collagen following loading, and that this is most prevalent in the more fibrocartilaginous matrix of the femoral ACL enthesis, wherein we previously reported increased tissue compliance following in situ submaximal fatigue loading in cadaveric human ACLs [18]. Following submaximal loading of the mouse ACLs, immunofluorescence was used to identify unraveled collagen at the microscopic level, and atomic force microscopy infrared spectroscopy (AFM-IR) was used to map the unraveled to intact collagen ratio, the GAG to intact collagen ratio, tissue topography, and tissue compliance at the ACL femoral enthesis and midsubstance in loaded and contralateral control knees. Chromogenic immunohistochemistry was performed to identify post-fatigue changes in PG density, and thus that of their associated GAGs.

2. Methods

2.1. In vivo induction of fatigue in mice model

In compliance with our Institutional Animal Care and Use Committee approved protocols, five female C57BL/6 J (B6) inbred mice were acquired from The Jackson Laboratory at 9 weeks of age and allowed one week to acclimate with food and water ad libitum. At 10 weeks of age, mice were anesthetized with isoflurane, provided heat support, and had ocular lube applied. Mice were then positioned into a custom loading jig that incorporates knee valgus and internal tibial rotation to apply peak strain to the ACL in vivo by leveraging its primary role in restricting anterior tibial translation (Fig. 1) [14,28]. Using an electromechanical testing system (5944; Instron), the right knee of each mouse was then fatigue loaded in vivo for 5000 cycles at a rate of 0.75 mm/s. The applied load to repeatedly strain the ACL was 60% of the maximum load required to fail the ACL in vivo, or 8.5 N out of a median failure load of 14.2 N, which was previously determined from our previous studies [14, 29]. The contralateral knee served as an internal control. After one hour of recovery following fatigue loading, all mice were euthanized, and the left and right knees were immediately harvested. All harvested knees were trimmed of excess muscle and fixed in 4% paraformaldehyde for 72 h, rinsed, and decalcified in 10% ethylenediamine tetraacetic acid at 4 °C for 14 days. Demineralized tissues were then dehydrated in graded alcohols and embedded in paraffin. Paraffin blocks were then sagittally sectioned using a microtome (Fisher Scientific) at a thickness of 5 μm. Each slide contained an average of four tissue sections. A total of three slides per knee were analyzed (~20 μm of sequential tissue per slide. The first slide was used for quantifying denatured collagen using immunofluorescent and enhanced darkfield hyperspectral microscopic techniques. The second slide was used for AFM-IR analyses. All remaining slides that contained at least the proximal third of the ACL, wherein mouse and human data consistently demonstrate the most fatigue-induced damage, were used for proteoglycan antibody staining. Serial sections were taken across the entire ACL spanning the femoral origin and tibial insertion.

Fig. 1.

In vivo testing setup. Right knee held in valgus (θ_M), with the tibia internally rotated (θ_IR). Cyclic (fatigue) or maximal (failure) compressive force (F) applied to femur to tension ACL.

2.2. Microscopic identification of unraveled collagen

Consistent with our previous work [12–14], a subset of tissue sections (n = 3–5) per each mouse that included >75% of the ACL were then used to microscopically identify and spatially quantify unraveled collagen using immunofluorescence. Tissue sections were first deparaffinized and rehydrated, followed by incubation of a 1.5-μM dilution of a single sulpho-cyanine3 fluorophore conjugated collagen hybridizing peptide (R-CHP, 3Helix Inc.), specific to unraveled collagen of all types, at 4 °C overnight. Tissue sections were then rinsed, cover slipped and imaged at 10x magnification using a fluorescence microscope with a 250 ms exposure and 1x gain (Eclipse Ni-U, Nikon Instruments Inc.). Images collected were normalized to the fluorescence intensity of a thermally denatured tissue section. The percent area of fluorescence signal was quantified using ImageJ (NIH) and a custom thresholding macro provided by the peptide manufacturer [30], providing a measure of total collagen unraveling across the ACL. The resulting images were then used to identify 3 tissue regions within the femoral ACL enthesis and mid-substance on adjacent tissue sections for AFM-IR analysis.

Using the same tissue sections, higher resolution spectral data of bound Cy3 fluorophore was captured using enhanced darkfield hyperspectral microscope (CytoViva Hyperspectral Microscope, CytoViva Inc.). This system utilizes a tungsten halogen light (Dolan Jenner) set to the full 150 W output. Light was directed to the microscope using a liquid light guide (Newport Inc.), which transmits the illumination wavelengths from 420 nm – 2200 nm. The light guide connects directly to the enhanced darkfield illuminator system mounted onto the microscope condenser, and consists of an annular cardioid oil condenser to produce highly collimated light at oblique angles with a numerical aperture of 1.2 – 1.4 NA. The light scattered from the sample was collected using a 60x oil iris objective (Olympus Inc.) and projected onto a visible and near-infrared (VNIR) diffraction grating spectrograph (Specim). Diffraction grating separates the light into distinct wavelengths from 400 nm – 1000 nm at high spectral resolution (~ 2 nm). The spectrally resolved light was then projected through a 30 μm slit onto the pixels of a charge-coupled device (CCD) video camera (PCO, Kelheim, Germany). Hyperspectral images were captured using a pushbroom method to capture pixels across rows, with each pixel containing the optical spectral response from 400 nm – 1000 nm of that pixel’s spatial area at 2 nm spectral resolution.

After image capture, image analysis was conducted using a customized version of ENVI 4.8 hyperspectral image analysis software (Harris Geospatial Solutions, Inc), using a custom Peak Location Classifier algorithm. All pixels with a peak wavelength equal to 570 nm (± 10 nm) were identified and mapped as green in the hyperspectral image file. This allows spectral mapping of the Cy3 emission spectrum full width half maximum across the hyperspectral image.

2.3. AFM-IR mapping of unraveled collagen, GAG sugar chains, and fibril-level mechanics

For AFM-IR imaging, adjacent tissue sections (~20–40 μm nearer the femoral notch midline) to the ones used for the above microscopy techniques, and that also included >75% of the ACL, were deparaffinized in 20 mL of xylene for 5 min (x2). Immediately following and prior to the tissue drying, sections were rehydrated through graded alcohols and submerged in deionized water. Individual rehydrated tissue sections were then transferred to and mounted on AFM pucks using double-sided adhesive tape.

AFM-IR data was collected using a NanoIR3 system (Bruker). The system was equipped with a contact-mode gold-coated silicon nitride probe with a spring constant of 0.07 – 0.4 N/m and resonant frequency of 13 ± 4 kHz. The affixed Quantum Cascade Laser (QCL) and a photo-detector were capable of spectroscopic detection ranges of 800 – 1880 cm⁻¹. Using this system, we were able to simultaneously collect topographic data, single wavenumber IR intensity maps, and PLL frequency maps (relative stiffness). Imaging on each hydrated tissue section was in three regions of the ACL, one near the femoral enthesis and two within the mid-substance (Fig. S1). For each location, topography, PLL frequency maps and wavenumber IR intensity maps were simultaneously collected at three different 5 × 25 μm regions of interest (ROI) ~30 μm apart at 100 nm pixel density. The ROIs were oriented so that the long axis runs parallel to the ACL. Three single wavenumber IR intensity maps associated with Amide I intact collagen triple helix (1680 cm⁻¹), the unraveled collagen triple helix (1740 cm⁻¹), and glycosaminoglycan sugar chain (1070 cm⁻¹) were collected using 1.93% laser power and 200 ns pulse width (Fig. S2). Simultaneous mechanical data collection was performed by implementing a Phase-Locked Loop (PLL) frequency feedback loop on the probe-sample contact resonance frequency with an I gain of 6, a P gain of 10, and a range between 60 – 80 kHz. Hyperspectral data was collected at a 1 μm pixel density over the same ROI in which each pixel contains a full IR spectrum from 800 – 1880 cm⁻¹.

Hyperspectral data were analyzed using MATLAB to generate ratio maps of 1740/1680 and 1070/1680 from the individual IR spectrum in each pixel. Plane fitting of the topography maps and IR ratio maps calculated from the single wavenumber maps were done using Analysis Studio v3.15 (Bruker). All data was then flattened into table format in which a single row of observation contains information such as PLL frequency, and IR ratio values. We established 0.3 as a cut-off ratio based on our library of collected IR spectra for human and mouse ACL samples, which show that an unraveled/intact collagen ratio (1740/1680) <0.3 in both control and tested groups is consistent with a typical shoulder seen in IR spectra collected from healthy collagenous tissue. Only when the ratio is greater than 0.3 is a distinct peak observed at 1740 cm− 1, thereby deviating from what is considered normal, healthy collagenous tissue spectra [18]. Thus, ratios above this threshold were considered representative of fatigue-induced collagen unraveling. Additional details on the combined application of topographic, infrared imaging, and PLL analysis of stiffness domains have been previously reported [31]

2.4. Microscopic identification of primary ACL PGs

To investigate proteoglycans (PGs) that may correspond with AFM-IR detected GAG activity, all remaining sections that included at least the proximal third of ACL tissue were deparaffinized and rehydrated with graded alcohols (~3–5 slides). Epitope retrieval was performed using hyaluronidase from bovine testes (H3506; Sigma-Aldrich) and endogenous peroxidases were blocked (BLOXALL; Vector Laboratories). Primary rabbit antibodies included decorin (PA-13538; Invitrogen), aggrecan (NB100–74350; Novus Biologicals), and versican (MA-42721; Invitrogen). For each slide, tissue sections were incubated with one of each antibody (1 antibody/section) for 1 hour and detected using an anti-rabbit peroxidase kit (Vectastain Elite ABC Kit; Vector Laboratories). Negative controls were also included for each antibody to ensure antibody specificity without non-specific binding Following this, tissue sections were incubated 3,3′-diaminobenzidine (DAB) for chromogenic detection, and counterstained using 0.01% Fast Green. Stained tissue sections were then dehydrated, cleared, and cover slipped. All tissue sections were imaged using an Eclipse Ni-U microscope (Nikon Instruments Inc.) at 10x magnification. Following, the entirety of the ACL present was cropped from the original image and saved as a different file. For each cropped ACL the percent area of tissue with DAB present was quantified using ImageJ (NIH) and the Immunohistochemistry Image Analysis Toolbox plugin.

2.5. Statistical analysis

All statistical and data analysis were performed using Prism (v10; GraphPad Software Inc.) and RStudio v3.6 (Posit). Normality was assessed with a Shapiro-Wilk test. If normality was not met, the non-parametric Kolmogorov-Smirnov test was used. When normality was confirmed, the homogeneity in variances was assessed using an F-test. A paired two-sided t-test was performed on all data assuming equal variances. An alpha of 0.05 was considered significant.

3. Results

3.1. Fatigue loading induces highly localized unraveled collagen

Fatigued ACLs had ~78% more fluorophore-detected unraveled collagen by area compared to controls (Fig. 2). Most fatigue-induced collagen unraveling occurred with the ACL midsubstance (Fig. S3). Spectral imaging of adjacent unstained tissue sections with AFM-IR also indicates a higher ratio of unraveled to intact collagen. Fig. 3A-B illustrates the localization of the unraveled collagen domain in the femoral enthesis of fatigued ACLs as compared to controls using the unraveled/intact (1740/1680) collagen ratio maps generated from both hyperspectral and single wavenumber imaging. The change in unraveled/intact collagen ratio observed in the hyperspectral and single wavenumber generated ratio maps was quantified using a kernel density distribution and demonstrates a greater degree of unraveled collagen post-fatigue in these ROIs, at the hyperspectral (K-S test, p < 4.6 × 10⁻⁰⁶) (Fig. 3C) and single wavenumber (K-S test, p < 2.2 × 10⁻¹⁶) (Fig. 3D). However, even though localized domains of unraveled collagen were found in all fatigued ACLs, there was not a significant difference in the overall amount of collagen unraveling upon summing all ROI data across all mice and plotting the unraveled/intact collagen as a set of kernel density distributions (Fig. 3E). We presumed the discrepancy between immunofluorescent and spectral quantification of unraveled collagen was attributable to large differences in spatial resolution (~100 nm vs ~3–5 μm) between techniques.

Fig. 2.

Percent of ACL Unraveled Collagen Post-Fatigue. Means and standard deviations of the amount of unraveled collagen detected from ACL tissue sections incubated with a Cy3 fluorophore-conjugated collagen hybridizing peptide (CHP) for contralateral control (C) and fatigued (F) ACLs (n = 5 paired knees).

Fig. 3.

AFM-IR Spectral Detection of Unraveled Collagen. Representative topography and unraveled/intact collagen ratio maps from the control (A) and fatigue tested (B) ACLs of mouse sample I acquired at the femoral enthesis. 1740/1680 ratio maps are shown for both hyperspectral and single wavenumber scans. Kernel density distributions of derived from the unraveled/intact collagen ratio map overlayed with cumulative density plots are provided for both control (A) and fatigued (B) hyperspectral (C) and single wavenumber (D). (E) Overall kernel density distribution of unraveled/intact collagen ratio for all ROI measured maps across all five mice samples in the femoral enthesis (top) and midsubstance (bottom) tissue regions for both control and fatigued ACLs.

To further investigate this discrepancy, enhanced darkfield hyperspectral microscopy was used to verify the amount of fluorophore present in the tissue sections. At the fluorophore excitation peak of 570 nm wavelength (Fig. 4A), the generated electromagnetic (Em) spectral maps indicate that ~2.5% of pixels within the fatigued ACL midsubstance contained unraveled collagen (Fig. 4B). Taking into consideration both the AFM-IR and enhanced darkfield hyperspectral microscopy data, collagen unraveling appears highly localized and only present in ~2.5% of the tissue. Therefore, when employing a method such as AFM-IR with small ROIs, one often encounters ROIs in which no unraveling has occurred leading to kernel densities that include all AFM-IR data across all mice not yielding a significant difference (Fig. 3E). Our methods that employed much larger imaging regions and therefore are more likely to include unraveling in each ROI, fluorescence microscopy (Fig. 2) and darkfield hyperspectral microscopy (Fig. 4), did indicate a significant increase in the amount of unraveled collagen.

Fig. 4.

Darkfield Hyperspectral Microscopy Detected Unraveled Collagen. (A) Spectral profile of CHP-Cy3 shows 570 nm Em wavelength unique to the dye, which is not present in ACL tissue. (B) Darkfield hyperspectral mapping at 570 nm Em wavelength collected from the mid-substance of tested ACL tissue section stained with CHP-Cy3. Total fluorescent pixels indicative of unraveled collagen account for 2.5% of the ECM.

3.2. Increase in overall GAG content due to repetitive in vivo loading

Spectral analysis of the GAG/intact collagen ratio (1070/1680 cm⁻¹) indicated a greater presence of GAG sugar chains in fatigued ACLs with both hyperspectral and single wavenumber imaging (K-S test, p < 2.2 × 10⁻¹⁶) (Fig. 5A and B). The change in GAG/intact collagen ratio observed in the hyperspectral and single wavenumber generated ratio maps is quantified using a kernel density distribution and demonstrates an increase in GAGs in fatigued ACLs in these ROIs (Fig. 5C and D). When looking across all ROIs obtained for all mice, there was also an overall greater presence of GAGs (K-S test, p < 2.2 × 10⁻¹⁶) (Fig. 5E). The ROIs in the ACL midsubstance (Fig. 5E, bottom) had a 14% greater abundance of GAGs compared to the femoral enthesis (Fig. 5E, top). This is consistent with the distribution of unraveled collagen of fatigued ACLs in Fig. S3. Fig. 6 indicates that unraveled collagen domains with ratios ≥ 0.3 also include significantly more GAG compared to those domains with ratios < 0.3. Specifically, across all the hyperspectral maps containing an unraveled/intact collagen ratio ≥ 0.3, we observed damaged domains (red) colocalizing with high ratio (>0.5) GAG domains (green). However, the higher GAG domains are not necessarily always colocalized with domains of unraveled collagen. Regardless of this, colocalization of unraveled collagen and GAGs was apparent across both fatigued and control ACLs.

Fig. 5.

AFM-IR Detected Glycosaminoglycan. Representative topography and GAG/intact collagen ratio map collected from the control (A) and fatigue (B) knees of mouse sample I acquired at the femoral enthesis. Ratio maps are shown for both hyperspectral and single wavenumber scans. Kernel density distributions of the GAG/intact collagen ratio map overlayed with cumulative density plot are provided for both control (A) and tested (B) hyperspectral (C) and single wavenumber (D) data. (E) Overall kernel density distribution of GAG/intact collagen ratio maps across all five mice in the femoral enthesis (top) and midsubstance (bottom) tissue regions for both control and fatigued ACLs.

Fig. 6.

Relationship Between Unraveled Collagen and GAG. (A) Hyperspectral unraveled/intact collagen ratio maps and GAG/intact collagen ratio maps simultaneously collected at the same location from all the mouse samples for both control and tested at the femoral enthesis and midsubstance ligament region, consisting of damaged domains with unraveled/intact collagen ratio greater than 0.3. (B) Overall boxplot of GAG/intact collagen ratio map across all mice samples for both control and tested at the femoral enthesis and midsubstance ligament, where each point represents a pixel from the map with unraveled/intact collagen ratio <0.3. (C) Overall boxplot of the GAG/intact collagen ratio map across all five mice for both control and tested ACLs at the femoral enthesis and midsubstance, where each point represents a pixel from the map with unraveled/intact collagen ratio greater than 0.3.

To identify proteoglycan candidate(s) from which the greater GAG abundance arises post-fatigue, immunohistochemistry experiments on adjacent tissue sections of each mouse ACL were performed (Fig. 7). There was no distinguishable or significant difference in decorin or aggrecan abundance between control and fatigued ACLs. However, versican was consistently more abundant across all fatigued ACLs compared to their contralateral controls, with an overall 47% increase in this large-aggregating proteoglycan (p = 0.0002) (Fig. 8). Representative images of decorin and aggrecan staining are provided as supplemental material (Fig. S4).

Fig. 7.

Percent Proteoglycan Staining Across ACLs Post-Fatigue. Mean, standard deviation, and p-value for percent-stained tissue area of decorin, aggrecan, and versican for contralateral control (C) and fatigued (F) ACLs (n = 5 paired knees).

Fig. 8.

Representative Images of Versican IHC. Paired ACLs incubated in Vcan antibody detected with 3,3′-diaminobenzidine (DAB), counterstained with Fast Green. Percent Vcan is obtained using ImageJ to isolate the DAB signal.

3.3. Highly unraveled domains are compliant and colocalized with GAGs

Analysis of PLL frequency maps acquired simultaneously from the same ROI from which spectral data was captured indicate that in domains wherein unraveled collagen and GAGs colocalize, the tissue compliance is higher at the collagen fibril-level, reducing matrix stiffness. Specifically, domains of unraveled/intact collagen ratios greater than 0.3 had greater mechanical compliance (Fig. 9). This is consistent across all the control and fatigued ACLs with unraveled/intact ratios above 0.3. Surprisingly, there is also a shoulder at the higher frequency range for the domains with high collagen unraveling, which is not present in the domains with ratios below 0.3 (Fig. 9G). This provides evidence of GAG colocalization with unraveled collagen-rich and compliant matrix domains following fatigue loading.

Fig. 9.

Mechanical Effects of Colocalized Unraveled Collagen and GAG. Representative single wavenumber data collected from sample IV right fatigued ACL at the femoral enthesis showing the (A) topography, (B) unraveled/intact collagen ratio map, (C) GAG/intact collagen ratio map, (D) PLL frequency or relative stiffness map. (E) Kernel density distribution of GAG ratio map (C) showing distribution of co-localized pixel with unraveled/intact collagen ratio map <0.3 (green) and colocalized pixel with unraveled/intact collagen ratio map greater than 0.3 (lavender). (F) Kernel density distribution of PLL frequency map (C) showing distribution of colocalized pixel with unraveled/intact collagen ratio map <0.3 (green) and colocalized pixel with unraveled/intact collagen ratio map greater than 0.3 (lavender). Domains with unraveled/intact collagen greater than 0.3 have greater density of high GAG/intact collagen ratio and more compliant. (G) Overall kernel density distribution of PLL frequency maps at the control and tested knee in each of the five mouse samples, showing distribution of colocalized pixel with unraveled/intact collagen ratio map <0.3 (green) and colocalized pixel with unraveled/intact collagen ratio map greater than 0.3 (lavender).

4. Discussion

We investigated the in vivo relationship between collagen unraveling and fibril-level mechanics in response to ACL fatigue loading in a mouse model and determined whether GAGs are also involved in altering tissue mechanics in this injury scenario. With our model, we confirmed previous ex vivo findings from fresh human ACL cryosections that suggested an increase in tissue compliance in ECM regions containing fatigue-induced collagen unraveling [18]. However, there were discrepancies in findings between our ex vivo and in vivo studies. First, the overall mean unraveled/intact collagen ratio for both the control and fatigued ACLs (0.14 average ratio, 0.6% of measured pixels ratio > 1) is well within the 0.30 ratio threshold that, based on earlier observations in healthy collagenous tissue [18], we consider to be representative of normal ratios of unraveled collagen. Whereas our ex vivo spectroscopic findings from fatigued cadaveric ACL tissue found that 29% of pixels exhibited a ratio >1 consistent with unraveled collagen [18]. By way of contrast, the ex vivo cadaveric loading is dynamic, allows for free knee movement, and includes a ground reaction force [28], whereas the mouse loading is quasi-static and the knee has controlled movement [14]. Moreover, the cadaveric knee is laxer than the live mouse knee. Combined, these factors likely translate to greater ACL forces and tissue damage ex vivo compared to what is experienced in vivo under controlled loading conditions. Second, the degree of spectroscopically detected unraveled collagen within the physiological range in the fatigued ACLs does not correspond with the fluorescence microscopy of R-CHP-incubated tissue sections that indicate the fatigued ACLs have ~80% more unraveled collagen by area than controls. Darkfield hyperspectral microscopy indicates that only ~2.5% of the ACL contains fluorophore bound to unraveled collagen triple helices. This degree of collagen unraveling corresponds with ex vivo findings reported by Zitnay et al. [11] in cyclically loaded tendon. However, given that only 2.5% of the midsubstance collagen matrix is positively bound with the R-CHP and sparsely distributed, the AFM-IR data is consistent in showing variations in the unraveled/intact collagen ratio maps where only certain locations show unraveled collagen. This difference is likely reflective of the low spatial resolution of epifluorescence, wherein an AFM-IR detection of the 1740 cm⁻¹ wavenumber assigned to unraveled collagen at ~100 nm scale fluoresces to a ~3–5 μm scale under optical microscopy. This outcome also helps explain why a bulk IR spectroscopy technique such as Fourier Transform Infrared (FTIR) spectroscopy, with a resolution of roughly 5 – 10 μm, has difficulty identifying and discerning subtle increases in unraveled collagen. Hence, the repetitive in vivo fatigue loading of the mouse ACL does induce collagen unraveling but it is difficult detect from the overall unraveled/intact collagen ratio. Moreover, that ~2.5% of collagen unraveled in mouse ACLs is consistent with findings by Lin et al. [9] in tendon fascicles uniaxially loaded in tension. Although Lin et al. observed that this occurred upon tissue yielding, whereas in our model the load experienced was in the elastic region of the load-displacement curve. A similar findings between studies is interesting considering the loading models employed completely differ from one another since in our study the loading is cyclic and the ACL is in situ, wherein the ligament is anchored to bone, obliquely oriented, and its anteromedial and posterolateral bundles crossed. Thus, the in our model the ACL not only experiences tension, but also compressive, shear, and frictional forces not accounted for in ex vivo uniaxial tendon fascicle loading. Multiaxial strains generated via repetitive loading are distributed heterogeneously across the ACL [32–34], thus certain tissue regions may experience greater strains and likely localized plastic deformation.

We also confirmed our hypothesis that there is an increase in GAG content in unraveled collagen domains, which collectively contribute to an increase in fibril-level compliance. However, we did not find the relationship between tissue compliance and unraveled collagen to be greatest in the more mineralized fibrocartilaginous regions of the femoral enthesis as previously reported in human ACLs [18]. Instead, the majority of colocalized unraveled collagen and GAGs, and their combined mechanical effect on tissue compliance, was predominately found within the ACL midsubstance. This could reflect a difference in loading, as the mouse knee is held in a fixed position and quasi-statically fatigued, while the human knee is free to rotate and displace while dynamically fatigued using a drop-weight system [12,14]. Moreover, the cadaveric loading encompasses a ground reaction force, while the in vivo mouse model does not. Nevertheless, the relationship between collagen unraveling and increased tissue compliance is present within the femoral enthesis of the fatigued mouse ACLs, confirming our ex vivo findings, which may be a potential mechanism for ACL overuse damage to accrue over time. Our finding of significant colocalization between fatigue-induced unraveling, increased GAG activity, and increased tissue compliance differs from a rat tail tendon fascicle study reported by Abreu et al. [26]. They found that GAG content was associated with bulk tissue viscoelasticity changes but not bulk tissue collagen unraveling, as they did not detect a significant increase in unraveling relative to GAGs. The differences between study approach could explain this discrepancy, since Abreu et al. reported on the GAG-collagen unraveling relationship generated via in vitro low-force static loading experiments devoid of a physiological response, compared to our study that investigated this relationship following in vivo high intensity, high force cyclic loading. Moreover, our focus was localized to small tissue regions rather than the entire tissue, allowing us to detect more nuanced changes at higher resolution within specific ROIs. It is worth noting that the changes in unraveled/intact collagen ratio and the increase in GAG/intact collagen ratio are consistent for both the hyperspectral data and the single wavenumber data. This consistency is important as the imaging process is substantially different for the two imaging methods. Specifically, the hyperspectral ratio map is generated from pixel-by-pixel IR spectra collection (1800 – 800 cm⁻¹) while the single wavenumber ratio map is generated by detecting the response across the surface from fixed IR wavenumber scans (unraveled/intact collagen 1740/1680; GAG/intact collagen 1070/1680). Here we show that the two different AFM-IR collection techniques closely match one another in the structural domains discerned, regardless of pixel density (1 μm vs 0.1 μm), as exhibited in the representative data (Fig. 6A and B). Both imaging methods indicate GAGs are abundant in the ACL shortly following fatigue loading.

Collagen unraveling occurs when fibrils are stretched beyond the yield point [9]. Ex vivo experiments have shown that overloaded tendon fibrils (i.e., yield point exceeded) kink in discrete regions of unraveled collagen [8]. Importantly, these damage regions remain stable, but become more mechanically compliant, suggesting unraveling allows for the dissipation of further post-yield strain through energy absorption [18,19]. However, there is a limit to this safety factor as we previously showed in human and mouse ACL, wherein the repeated accumulation of unraveled collagen contributes to a progressive decline in structural integrity and thus function [12–14,18,35]. Finding an increase in sulfated GAG content in mechanically-unraveled collagen domains within an hour of loading suggests a complimentary mechanism, one that can further dissipate strain while assisting in the stabilization and restoration of the ACL’s structural and functional integrity. Since local synthesis of GAGs is unlikely within an hour of loading, it is more likely that GAGs are being recruited locally by resident PGs. Versican, a large hyaluronan-binding chondroitin-sulfate proteoglycan (~500 kDa), is the dominant large chondroitin sulfate PG in ligament and may be a primary vehicle for these GAGs [36]. Distinctive from other PGs, versican also directly binds to collagen through its G3 domain irrespective of GAG modifications and retains a functional G1 domain longer to maintain robust hyaluronan-binding [36,37]. Versican mainly resides in the fascicular and interfascicular matrices of the ACL exposed to dynamic shear forces [38]. The ACL midsubstance is exposed to high multi-axial forces where the anteromedial and posterolateral bundles cross or where the ACL and posterior cruciate ligament (PCL) make contact. Mechanically, versican assists in collagen bundle lubrication, and like aggrecan, acts as molecular shock absorber through its hyaluronan backbone and abundant negatively-charged GAG side chains that attract positively-charged hydrogen atoms in unbound water. Interactions between versican, hyaluronan, and collagenous matrices provide have been shown to reduce matrix stiffness both in vitro and in vivo [37,39]. This suggests that its presence in fatigued ACLs would also contribute to an increase in tissue compliance, particularly as versican is most abundant in the ACL midsubstance, wherein collagen damage was shown to be greatest post-fatigue. Based on our current and previous findings, along with mechanistic data reported by others, we hypothesize that tropocollagen unraveling within fibrils during fatigue loading drives the release of positively charged ions [40,41], attracting resident GAGs and their versican host to physically penetrate the increasingly porous ECM [12,42] and directly bind to collagen to stabilize the disrupted extracellular matrix [37].

Initially, the recruitment of nearby GAGs to disrupted tissue regions would facilitate strain dissipation and matrix reinforcement. However, versican, with its hyaluronan interactions, is also integral to the inflammatory response post-injury and subsequent ECM remodeling [43–46]. Early in the wound response to soft tissue injury versican and hyaluronan, in addition to interactions with fibronectin, rapidly form a provisional matrix in the wound bed to reinforce the structure [47]. Within this provisional matrix, versican interacts with cell surface receptors and additional ECM molecules to in part facilitate leukocyte migration into the impacted tissue, modulate chemokine availability, and contribute to cell proliferation and migration [44,46,48,49]. Following the proliferative phase, versican and hyaluronan have a major role remodeling the provisional ECM, predominately by maintaining the myofibroblast phenotype integral to wound closure. Due to the importance of this interaction to matrix stability and repair, further work is needed to elucidate versican and hyaluronan structural and functional roles in response to submaximal fatigue loading, including which of the 4 versican isoforms that bind chondroitin sulfated GAGs are active in this process.

This study has significant limitations. First, fluorescence microscopy, dark field hyperspectral microscopy, and AFM-IR were not performed on the same tissue sections but on adjacent sections, which limits the ability to ascertain the colocalization of these measurements. However, the AFM-IR analysis did simultaneously confirm the colocalization of unraveled collagen and GAGs, and an increase in tissue compliance within the exact same tissue domains evaluated. Second, this study has only tested decorin, aggrecan, and the larger V0/V1 versican isoforms, as they are the most abundant PGs in ligament proper and the fibrocartilaginous zones of the enthesis [36]. However, other small leucine-rich PGs present in the ACL, like biglycan and fibromodulin, require further investigation as they may also be part of the biological response and influence shifts in tissue mechanics. Third, this investigation was limited to 5 adolescent females. Adolescent females were chosen to compliment previously published data from our group [14], however to evaluate generalizability of our outcomes, males and additional ages need to be included in the future.

5. Conclusions

Based on these in vivo mouse outcomes, and those resulting from our previous ex vivo human studies, our working hypothesis is that submaximal fatigue loading drives the unraveling of collagen triple helices, resulting in an increased GAG distribution in these damaged domains. Collectively, the loss of collagen structure and the reinforcement of these matrix defects by GAGs results in increased collagen fibril-level compliance in these domains. Versican is a likely candidate for attracting GAGs in these unraveled regions of the sub-fibril network, which can both reinforce and mount a reparative response to restore fibril structure and function.

Supplementary Material

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.actbio.2026.08.035.

Statement of significance:

Outcomes from this anterior cruciate ligament (ACL) fatigue loading study provide high-resolution in vivo evidence for a structural and biochemical mechanism for the development of chronic overuse injuries. Our findings suggest that fatigue-induced tropocollagen unraveling increases collagen fibril-level tissue compliance, and that glycosaminoglycans rapidly colocalize with these damaged regions, further influencing ACL mechanics. The large aggregating proteoglycan, versican, was immunohistochemically identified as a candidate for glycosaminoglycan recruitment to the damaged extracellular matrix. Versican may be preferentially retained in domains of tropocollagen damage to facilitate the tissue remodeling response, though further research is needed to confirm this potential link.