Concentration-Dependent Effects of Fibroblast-Like Synoviocytes on Collagen Gel Multiscale Biomechanics and Neuronal Signaling: Implications for Modeling Human Ligamentous Tissues

Abstract

Abnormal loading of a joint's ligamentous capsule causes pain by activating the capsule's nociceptive afferent fibers, which reside in the capsule's collagenous matrix alongside fibroblast-like synoviocytes (FLS) and transmit pain to the dorsal root ganglia (DRG). This study integrated FLS into a DRG-collagen gel model to better mimic the anatomy and physiology of human joint capsules; using this new model, the effect of FLS on multiscale biomechanics and cell physiology under load was investigated. Primary FLS cells were co-cultured with DRGs at low or high concentrations, to simulate variable anatomical FLS densities, and failed in tension. Given their roles in collagen degradation and nociception, matrix-metalloproteinase (MMP-1) and neuronal expression of the neurotransmitter substance P were probed after gel failure. The amount of FLS did not alter (p > 0.3) the gel failure force, displacement, or stiffness. FLS doubled regional strains at both low (p < 0.01) and high (p = 0.01) concentrations. For high FLS, the collagen network showed more reorganization at failure (p < 0.01). Although total MMP-1 and neuronal substance P were the same regardless of FLS concentration before loading, protein expression of both increased after failure, but only in low FLS gels (p ≤ 0.02). The concentration-dependent effect of FLS on microstructure and cellular responses implies that capsule regions with different FLS densities experience variable microenvironments. This study presents a novel DRG-FLS co-culture collagen gel system that provides a platform for investigating the complex biomechanics and physiology of human joint capsules, and is the first relating DRG and FLS interactions between each other and their surrounding collagen network.

Introduction

Joint pain is a leading cause of chronic pain [1], affecting 27 million adults in the U.S. with an annual cost of $100 billion expected by 2020 [2]. Synovial joints like the motion segments in the spine, jaw, knee, and hip can become painful with aging or from trauma due to repeated and/or supra-physiologic loading, all of which can initiate tissue damage and degeneration [3]. For example, neck and low back pain are among the most prevalent chronic syndromes [4] and can be due to pathology of the spinal facet joints which are susceptible to trauma [5,6] and degeneration [7,8]. Abnormal loading of a joint's ligamentous capsule can initiate pathophysiological pain cascades by activating the nociceptive fibers that innervate the capsule [9–11], which have cell bodies in the dorsal root ganglia (DRG) and synapse with spinal dorsal horn neurons to transmit pain [12,13]. The extracellular matrix (ECM) of the joint capsule is composed primarily of collagen, with type I collagen making up 80–99% of the ECM network [14]. The capsular network has subregions with parallel and irregular collagen fiber orientations [12,15,16]. Along with afferent fibers, fibroblast-like synoviocytes (FLS), also known as synovial fibroblasts or type B synoviocytes, reside in the capsule's ECM and in the lining of the synovium of synovial joints [12,15–18]. For the spinal facet capsules, FLS density is greater in the inner capsule than the outer capsule [15,16], and collagen network organization also varies, with regions that are randomly aligned and regions that are differentially aligned depending on the anatomical location [12,15,16].

Our group has developed an in vitro model using either dissociated neurons or DRGs seeded in a three-dimensional (3D) collagen gel to replicate both the sensory innervation and the network microstructural organization of the ligamentous capsule of the synovial spinal facet joint [19–21]. Using that model, the strain threshold for collagen fiber realignment was defined and found to be concurrent with the threshold for elevated expression of the neuronal injury marker, phosphorylated extracellular signaling kinase (pERK) [20,22]. In addition, regional strain directly relates to the increased expression of pERK and the neurotransmitter substance P (SP) in DRG axons [19,21]. Further, failure properties and network reorganization under tension are altered by a degradation-induced loss of collagen fibers [23]. Fibroblasts cultured in 3D collagen gels have also been extensively studied to define fibroblast–matrix interactions, with tension and network parameters found to regulate fibroblast mechanobiology [24–26]. For example, in free-floating matrices, fibroblasts exist in a “low-tension” environment and reorganize collagen fibers circumferentially around the edge of the gel, if at all [24,26]; in anchored matrices, fibroblasts exist in a “high-tension” environment and reorganize collagen fibers according to gel geometry and the direction of applied tension [25–27]. Yet, no culture system has integrated neurons or DRGs together with FLS in a co-culture system to capture the anatomy and physiology of human joint capsules. Accordingly, despite their co-existence in capsular ligaments, very little is known about the interactions of afferent fibers and FLS with each other and their surrounding collagen network.

Fibroblasts are ubiquitous stromal cells that play crucial roles in both normal and pathological physiologic functions, including development, repair, wound healing, and ECM remodeling [28]. In healthy and disease states, fibroblasts model and remodel their ECM via mechanotransduction mechanisms that convert mechanical cues into biological events [24–26,29]. Although fibroblasts are defined broadly by their morphology, adherence characteristics, and lack of lineage-specific markers [28,30], they are functionally and phenotypically diverse, with distinct gene expression profiles depending on their anatomical origin in the body [28], even across synovial joints [30]. Although the role of FLS in inflammation and degradation, particularly in rheumatoid arthritis, has been described [17], little is known about the effect of FLS on regulating matrix mechanics, either in an unloaded state or during loading, despite reports that fibroblast-collagen mechanobiological relationships exist for fibroblasts not derived from capsular ligaments [24–26]. Defining if, and how, FLS alter matrix mechanics and/or microstructure is critical to understand load-induced cell signaling and afferent–FLS interactions in the synovial lining and the capsular ligaments in which FLS reside.

Matrix-metalloproteinase-1 (MMP-1), a proteolytic enzyme in the interstitial collagenase family, is a likely mediator of joint pain given its role in nociception and mediating ECM degradation. For example, since collagen degradation alters the biomechanics and microstructure of joint tissues [31,32], and MMP-1 can degrade type I collagen [33], it is possible that MMP-1 may alter the local microenvironment and initiate mechano-regulated responses in afferent fibers [19,20,34,35]. MMP-1 also binds to several receptors implicated in nociception and pain [36–39], in both catabolically active and inactive states, and can bind to several non-ECM substrates that are also involved in pain signaling, including SP, which regulates nociception in culture [13,21,40] and in joint pain [41]. Although the elevated levels of MMP-1 are reported in the joint capsule and synovial fluid of painful joints [42–45], its role in joint pain is poorly understood. Since mechanical loading increases the secretion of MMP-1 by dermal fibroblasts and upregulates the MMP-1 gene expression in patellar tendon fibroblasts [46,47], it is possible that abnormal loading of a joint's ligamentous capsule modulates FLS regulation of MMP-1.

Regional strains of cell-embedded culture constructs directly relate to neuronal nociceptive signaling [19–21] and ECM remodeling by fibroblasts, including collagen synthesis, collagen deposition, and protease secretion [27,47,48]. Collagen fiber organization and reorganization also regulate these same load-induced pathological responses in cell-embedded networks [19,20,27,35]. Under tension, fibers reorganize in the direction of loading, and the extent of fiber reorganization increases with strain [20,23,49–53]. On the cellular level, a “switch-like” response has been observed for fiber alignment whereby fibers reorganize along with neuronal dysfunction occurring for loading that generates strains above 11.3% [20]. However, that work does not have any consideration of the physiologically relevant effects of fibroblasts on either the ECM and/or the neuronal function.

This study integrated primary FLS cells into our existing DRG-collagen model [19,21] to more closely mimic the multicellular environment of the capsular ligament of synovial joints [12,15–18]. In doing so, studies also used that integrated co-culture system to understand the effects of FLS on multiscale collagen gel mechanics and neuron physiology in response to distraction to failure. The overall objectives of this study were to (1) determine if FLS alter macroscopic failure properties, regional strains, and/or collagen microstructural kinematics and (2) measure if the presence and extent of FLS in the culture system alters MMP-1 expression and/or neuronal nociceptive signaling. Since capsular ligaments exhibit variable densities of FLS in different regions of the capsule [15] and the mechanical properties of fibroblast-embedded collagen gels depend on the initial fibroblast concentration [54], FLS were seeded in collagen gels at two concentrations to simulate the extremes of the physiologic range. Further, since active MMP-1 stabilizes by 6–8 days in ligament fibroblasts embedded in gels [55], the effects after 7 and 9 days of culture were tested. Day-in-vitro (DIV) 7 was chosen because DRGs exhibit neurite outgrowth nearly twice the diameter of their soma by that time [19,21]; DIV9 was selected because total active and inactive MMP-1 reach a steady-state before DIV10 [55]. To determine if the longer time in culture affects potential FLS-induced gel organization and/or baseline cell responses, expression of MMP-1 or substance P before any gel loading was compared between those 2 days (DIV7 and DIV9). Macroscopic gel mechanics were quantified by failure properties and gel stiffness, and regional strains and collagen alignment and organization were analyzed to assess the effects of FLS on regional kinematics. Fiber alignment data were measured during loading by quantitative polarized light imaging, which quantifies dynamic fiber reorganization [56–59]. Using fiber angle distributions, collagen organization was quantified by circular variance (CV) of the spread of fiber angles, describing the clustering of angles, with a lower CV indicating a tighter clustering and a higher degree of fiber alignment [19,20]. After failure, gels were assayed for total MMP-1 and neuronal substance P protein in the context of multiscale mechanical outcomes.

Materials and Methods

Cell Isolation and Co-Culture System.

All cells were harvested from Sprague-Dawley male rats under approved conditions and using sterile procedures. DRGs were harvested from all spinal levels of embryonic day 18 rats (from the CNS Cell Culture Service Center of the Mahoney Institute of Neuroscience) and stored in Hibernate-E medium supplemented with 1% GlutaMAX and 2% B-27 at 4 °C until plating [19,60]. FLS were harvested from both hind knees of a sexually mature adult rat (384 g) by removing the capsular tissue surrounding the knee joints, dicing the isolated capsular tissue as finely as possible, and incubating the diced tissue from both knee capsules together in Dulbecco's Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS), 1% Penicillin–Streptomycin (P–S), and 2 mg/mL bacterial collagenase (C0130, Sigma-Aldrich, St. Louis, MO) for 6 h at 37 °C under gentle agitation [61]. Digested tissue was filtered with a 70 μm cell strainer, spun down at 300 g for 5 min, and resuspended in feeding medium made up of DMEM with 10% FBS and 1% P–S. All cells were pooled together in culture, medium was changed every other day, and cells were passaged at 90% confluence. By passages 3 and 4, when cultures reach ≥95% purity for the FLS cell type [17,61], cultures exhibited an elongated and dendritic morphology indicative of FLS morphology [62]. As such, on passage 3 or 4, FLS cultures were trypsinized, rinsed, and resuspended in rat tail Type I collagen solution (2 mg/mL, Corning, Inc., Corning, NY) cast in 12-well plates (1 mL/well), allowed to gel at 37 °C [19], and cultured in DMEM with 10% FBS and 1% P–S. FLS were resuspended in two separate groups with different concentrations, based on densities found in the capsule [15]: 5 × 10⁴ cells/mL (low; n = 10) to simulate regions with low FLS cell density and 1 × 10⁵ cells/mL (high; n = 9) to simulate denser regions. Collagen gels without any FLS were included as controls (none; n = 8) for their effects on mechanics, MMP-1 and/or neuronal SP.

On DIV1, DMEM medium was removed from the top of the gels, and DRGs were seeded on the gel surface for all samples (6–10/gel) in 100 μL of Neurobasal feeding medium supplemented with 1% GlutaMAX, 2% B-27, 5% FBS, 10 ng/ml 2.5S nerve growth factor, 2 mg/ml glucose, 10 mM FdU, and 10 mM uridine [19,60] (Fig. 1). After 12–24 h, fresh Neurobasal feeding medium was added to all gels and changed every other day; all gels were cultured in the supplemented Neurobasal feeding medium for the remainder of the study. On DIV6, an additional layer of collagen (150 μL) was added to encapsulate the DRGs in half of the samples (none n = 4, low n = 4, and high n = 3). For the remaining samples (n = 4/group), DRGs were encapsulated with collagen using the same protocol but on DIV8 (Fig. 1).

Fig. 1.

Experimental timeline for co-culture conditions, setup for mechanical testing, and analyses of elemental strain and collagen organization. FLS were seeded into a collagen gel solution on day-in-vitro (DIV) 0 at either a low or high concentration, or were omitted (none), followed by seeding DRGs onto all gels at DIV1. On either DIV7 or DIV9, gels were loaded in a planar test device with an integrated polarized light imaging system. During loading, each gel was affixed in grips for uniaxial tension to failure and marked with a grid of dots for strain tracking. The corresponding strain map of MPS is shown extracted from three elements on the representative low FLS gel. The corresponding vector map is also shown displaying the raw fiber alignment data for an element, which is used to calculate CV.

Mechanical Testing and Data Acquisition.

Gels underwent distraction to failure on DIV7 (none n = 4, low n = 4, and high n = 3) or DIV9 (n = 4/group) (Fig. 1). On the day of testing, gels were stamped into a strip (21 mm × 8 mm), and a 4 × 4 grid of markers was drawn on the surface in the center of each gel to establish regional elements and enable strain tracking [19]. Using the grid of markers, each group of four nodes was designated as an element (Fig. 1) and used for data analyses. Using a planar testing machine (574 LE2, TestResources, Shakopee, MN), gels were loaded into grips attached to actuators equipped with 500 g load cells attached to each grip and immersed in a 37 °C phosphate buffered saline (PBS) bio-bath (Fig. 1). The mechanical system was integrated with a polarized light imaging system [20,56,63,64] and high-speed cameras (Phantom-v9.1, Vision Research, Inc., Wayne, NJ) that acquired collagen alignment maps and tracked marker locations and displacements during loading. Force and displacement data (200 Hz) were synchronized with high-speed imaging (500 Hz). Gels were preloaded until taut (< 2 mN) in either arm and then distracted at 0.5 mm/s to failure. Immediately following the distraction, gels were removed from the grips and fixed for 2 h in 4% paraformaldehyde, washed, and stored in 30% sucrose at 4 °C. In order to assess the effect of days in culture and FLS concentration on protein expression, additional gels were constructed for all three FLS concentrations (none, low, and high). These control gels were not loaded (n = 2/group/DIV) and were removed from culture on DIV7 or DIV9 for fixation and storage as described.

Immunolabeling of Matrix-Metalloproteinase-1 and Neuronal Substance P.

Gels were immunolabeled after loading for MMP-1, substance P, and βIII tubulin to evaluate the effects on MMP-1 and neuronal SP. Gels were blocked in PBS with 10% normal goat serum (Vector Laboratories, Burlingame, CA) and 0.3% Triton-X100 (Bio-Rad Laboratories, Hercules, CA) for 2 h at room temperature and incubated overnight at 4 °C with primary antibodies to MMP-1 (anti-rabbit, 1:200, Proteintech, Rosemont, IL), substance P (anti-guinea pig, 1:200, Neuromics, Inc., Minneapolis, MN), and βIII tubulin (anti-mouse, 1:300, Biolegend, San Diego, CA). Gels were then washed in PBS and incubated with the secondary antibodies goat anti-guinea pig Alexa Fluor 633, goat anti-rabbit Alexa Fluor 555, and goat anti-mouse Alexa Fluor 488 for 2 h at room temperature (all 1:1000, Life Technologies, Carlsbad, CA). Finally, gels were incubated in DAPI solution (1:200, ThermoFisher, Waltham, MA) at room temperature for 15 min to stain cell nuclei, washed in PBS, washed in distilled water, and then cover-slipped. Labeled gels were imaged using the 40× objective of a Leica TCS SP8 confocal microscope (1024 × 1024 pixels, Leica Microsystems, Wetzlar, Germany). Stacks of 6 confocal images were acquired for each gel (5 stacks/gel) at 1 μm increments up to 5 μm depth. Since MMP-1 and substance P can localize to DRG axons and cell bodies (somas) [21,65,66], images were acquired from both regions, with at least n = 2/region of the 5 stacks acquired for each gel and each stack acquired from a distinct region, with no DRG soma or its axons imaged twice. The location of each image was registered with the elements from each gel in order to relate cellular outcomes with the strain and collagen organization data. Image stacks were also acquired for control gels that did not undergo any loading but were fixed at either DIV7 or DIV9 for all groups (n = 2/group/day) in order to evaluate whether FLS concentration alters baseline expression of immunolabeled proteins.

Data and Statistical Analyses.

Force data were filtered using a ten-point moving average filter [19] and the maximum force detected was extracted and taken as failure (Fig. 2(a)). Stiffness was calculated using force–displacement curves and defined as the slope of the curve at between 20% and 80% of the maximum force [67] (Fig. 2(a)). The locations of the fiducial markers captured by high-speed imaging were digitized with fiji software (NIH) [68] for the unloaded image before any distraction (reference) and in the image immediately prior to failure. Grid position data were processed in ls-dyna (Livermore Software Technology Corp., Livermore, CA) to calculate the maximum principal strain (MPS) for each element for each loaded gel (Fig. 1(b)). Pixel-wise fiber alignment maps were created using 20 consecutive high-speed images acquired both before distraction (reference) and immediately prior to failure using a custom script based on a harmonic equation in matlab (R2018, MathWorks, Inc., Natick, MA) [56,63,64]. The CV was quantified from the spread of fiber angles detected for each element separately (Fig. 1(b)) [20]; CV at failure was normalized to the reference CV for all elements analyzed.

Fig. 2.

Analysis of macromechanics after tensile distraction to failure shows no differences between FLS concentration groups. (a) Force–displacement data show representative gels with high (gel 14), low (gel 18), and no (none; gel 21) FLS included. For each gel, peak force defines the failure and stiffness (k) that is calculated as the linear slope of the force–displacement curve between 20% and 80% of the peak force. (b) There are no differences between the FLS concentration groups for any of force at failure (p = 0.63), displacement at failure (p = 0.42), or stiffness (p = 0.30). DIV7 and DIV9 data are pooled in bar plots and show mean and standard deviation (SD).

To quantify the amount of positive protein labeling in the immunolabeled gels, the average intensity projection of each stack was generated using Fiji and a custom matlab script quantified the number of positive pixels above a threshold for positive MMP-1, substance P, and βIII tubulin labeling, separately [20,69]. Thresholds were determined from pilot studies with naïve FLS and DRG cultures, and samples with no primary antibodies added to control for labeling procedures were included as controls to verify the specificity of each antibody. The overall percentage of positive MMP-1 labeling was quantified without discriminating between cell types in order to account for total MMP-1 in the overall culture system, from both DRG and FLS cell sources. In order to compare neuronal SP, the co-localization of βIII tubulin and SP was computed; the co-localization of pixels positive for each was normalized to the total βIII tubulin for each image, separately, to account for differences in neuronal labeling.

All statistical analyses were performed in jmp (Pro 14; SAS Institute; Cary, NC) with α = 0.05. In order to test the effect of day in vitro on collagen fiber organization, separate t-tests compared CV at reference between DIV7 and DIV9 for each concentration. Further, a one-way analysis of variance (ANOVA) was used to assess the effect of FLS concentration on reference CV separately for each day in vitro. Similarly, the total MMP-1 and neuronal SP quantification in nondistracted control gels were compared across 2 days (DIV7 and DIV9) for each concentration using separate t-tests, to determine if the difference in length of time in culture altered baseline protein expression. An ANOVA assessed the effect of FLS concentration on baseline expression of total MMP-1 and neuronal SP quantification. Statistical comparisons between day in vitro and across groups for baseline protein expression were calculated with each quantified image as a statistical unit; this justification was rationalized since control gels that were used for baseline protein expression were nondistracted, and as such, there was no variation in the local biomechanical environment as is the case with stretched gels. The effect of FLS concentration on force at failure, displacement at failure, and stiffness was tested with a one-way ANOVA comparing different FLS concentration groups (none, low, and high). The MPS values and normalized CV quantification for those elements in which confocal images were acquired (n = 5/gel) were extracted and compared at failure between gels with different FLS concentrations (none, low, and high); the effect of concentration on elemental MPS, elemental CV, total MMP-1, and neuronal SP was tested with separate one-way ANOVAs and post hoc Tukey tests for each outcome.

Results

Overall, neither FLS concentration nor time in culture affects collagen microstructure before undergoing distraction. Although gels with FLS at either concentration spontaneously released from the culture plate wall by DIV2, the presence of FLS at either concentration does not change the collagen organization compared to gels without any FLS (Table 1). This is true regardless of the length in culture, with the CV at reference not different between no (none), low, and high at either DIV7 (p = 0.56) or DIV9 (p = 0.45) (Table 1). Further, the reference CV is also not different between DIV7 and DIV9, regardless of group (Table 1); there is no difference between the two DIV culture times detected in the microstructure of collagen for gels without any FLS (p = 0.41), with a low FLS concentration (p = 0.35), or with the high FLS concentration (p = 0.28) (Table 1).

Table 1.

Summary of macroscale biomechanics, microstructure before distraction and after failure, and protein expression in unloaded controls

Macro- and microscale mechanics
Group	Gel ID	Unloaded CV (×103)a	Force (mN)	Displacement (mm)	Stiffness (mN/mm)	Normalized CVa,b
DIV7
High FLS	1	0.13 ± 0.07	13.60	4.47	4.14	17.48 ± 15.13
	2	N/A	20.43	6.34	3.57	N/A
	3	N/A	16.82	6.06	5.02	N/A
	4	0.39 ± 0.15	11.68	5.97	1.56	2.37 ± 1.35
	Mean	0.28	15.63	5.71	3.57	8.85
SD	0.18	3.83	0.84	1.47	11.93
Low FLS	5	0.34 ± 0.08	20.42	2.00	9.82	1.07 ± 1.18
	6	0.31 ± 0.17	37.13	4.63	8.83	2.25 ± 1.08
	7	1.86 ± 2.20	65.61	4.26	18.53	2.83 ± 1.39
	8	0.27 ± 0.22	16.02	5.47	3.78	5.66 ± 2.17
	Mean	0.62	34.79	4.09	10.24	3.15
SD	0.99	22.47	1.47	6.13	2.24
None	9	0.62 ± 0.51	4.74	2.53	2.00	1.37 ± 0.48
	10	0.12	11.70	6.06	2.42	3.59
	11	0.33 ± 0.12	22.40	4.40	4.41	4.97 ± 2.06
	12	0.29 ± 0.28	11.48	5.71	3.64	7.04 ± 7.28
	Mean	0.37	12.58	4.68	3.12	4.14
SD	0.29	7.30	1.60	1.11	3.69
DIV9
High FLS	13	1.43 ± 0.27	15.96	6.00	2.92	16.30 ± 17.29
	14	0.20 ± 0.12	15.91	6.08	3.32	43.28 ± 34.12
	15	0.28 ± 0.10	15.20	4.95	3.89	24.10 ± 12.01
	Mean	0.50	15.69	5.68	3.38	30.89
SD	0.54	0.42	0.63	0.48	25.64
Low FLS	16	0.13 ± 0.10	6.49	5.36	1.10	13.74 ± 9.60
	17	1.97	11.91	4.26	1.89	1.91
	18	0.13 ± 0.11	20.71	5.92	6.47	7.42 ± 3.34
	19	0.19 ± 0.16	15.84	7.94	4.34	36.33 ± 21.42
	Mean	0.28	13.74	8.33	2.44	25.57
SD	0.52	6.03	3.41	1.22	5.96
None	20	1.04 ± 0.58	72.46	8.09	13.63	9.93 ± 7.54
	21	N/A	12.32	4.43	3.92	N/A
	22	0.41 ± 0.20	21.65	3.42	7.70	2.76 ± 0.86
	23	0.25 ± 0.14	5.12	2.14	2.77	2.45 ± 1.36
	Mean	0.49	27.89	4.52	7.01	4.23
SD	0.42	30.48	2.56	4.89	4.56

Protein expression
Group	Total MMP-1 (% positive pixels)a	Neuronal substance P (% positive pixels)a
DIV7
High FLS	8 ± 15	0.9 ± 0.7
Low FLS	4 ± 3	0.6 ± 0.6
None	4 ± 2	0.1 ± 0.1
DIV9
High FLS	1 ± 1	0.3 ± 0.3
Low FLS	8 ± 1.3	0.2 ± 0.2
None	5 ± 6	0.2 ± 3.6

N/A data were not collected due to technical problems with data capture.

^aElemental data are presented as mean  ±  standard deviation (SD) for all elements for that sample.

^bNormalized CV is calculated by dividing raw CV at reference by raw CV at failure.

As with the collagen microstructure, neither the concentration of FLS nor the time in culture alters baseline expression levels of MMP-1 or neuronal substance P in gels. The time in culture from DIV7 to DIV9 does not influence the baseline protein expression of either MMP-1 or neuronal substance P in unloaded gels. Specifically, the amount of MMP-1 on DIV7 is not different from the quantification of MMP-1 on DIV9 for none (p = 0.62), low (p = 0.39) or high (p = 0.15) FLS gels (Table 1). The same relationship exists for neuronal substance P between DIV7 and DIV9 for the none (p = 0.14), low (p = 0.06), and high (p = 0.07) FLS concentrations (Table 1). Furthermore, MMP-1 expression in gels is not different across groups (none versus low versus high) (p ≥ 0.31), regardless of DIV (Table 1); the same is evident for neuronal SP expression (p ≥ 0.05) with no difference based on concentration of FLS (Table 1). Since overall time in culture does not influence any outcome of the gel properties between DIV7 and DIV9, those groups were merged separately for each concentration to investigate the effects of three different concentrations on the biomechanical and physiological responses during and after failure.

Neither the presence nor the amount of FLS in the culture system alters the macromechanical responses of gels at failure (Table 1 and Fig. 2). None of the force or displacement at failure, or gel stiffness (p ≥ 0.30) is different across three concentration groups (none, low, and high) (Fig. 2(b)). In contrast, there are differences in the elemental responses; there are FLS concentration-dependent differences in both elemental MPS and collagen organization (Fig. 3). Despite undergoing similar deformations and forces at failure (Fig. 2), groups with FLS experience higher strains at failure than gels without any FLS, with both the low (p < 0.01) and the high (p = 0.01) FLS concentration gels undergoing strains that are nearly twice the strains sustained in gels with no FLS (Fig. 3). Further, for the greater FLS concentration (high FLS group), the collagen network shows a higher degree of reorganization at failure, with a graded normalized CV increasing with increasing FLS concentration (Fig. 3). However, the difference in normalized CV at failure is only significant (p < 0.01) between the high FLS concentration (21.2±23.1) and the group with no FLS (none; 4.3±4.2) (Fig. 3).

Fig. 3.

Regional strain and collagen microstructure at failure are FLS concentration-dependent. Stretch-induced MPS in gels with FLS at both concentrations (low and high) are significantly higher than the MPS at failure in gels without FLS (*p  ≤  0.01). CV normalized to the corresponding reference CV is also significantly greater (*p < 0.01) in gels with high FLS concentration than in those without FLS (none). Bar graphs show summary data with DIV7 and DIV9 pooled and overlaid with individual data points, each representing an element (open circles). Histograms show the probability that collagen will orient at a given angle for reference and at failure for representative elements from each of the high (gel 14), low (gel 18), and no FLS (gel 22). A larger spread of angles is evident with the high FLS gel element at failure and corresponds to a larger CV value, indicating a large degree of fiber reorganization relative to the network microstructure prior to loading.

As with the regional micromechanics outcomes (Fig. 3), the physiological effects of the presence, and amount, of FLS cells after tensile gel failure are differentially altered (Fig. 4). In fact, the stretch-induced expression of both total MMP-1 in the culture system and neuronal substance P generally follow each other with the low concentration of FLS having the greatest expression (Fig. 4). MMP-1 expression is observed in both the DRG somas and axons and appears to label in concentrated or compacted masses in contrast to a more diffuse label (Fig. 4(a)); there is little MMP-1 in gels with only DRGs (Fig. 4(a)). Despite positive labeling in gels with either concentration of FLS, those with a low FLS concentration (35.4±27.4%) express significantly greater levels (p < 0.01) of MMP-1 after stretch to failure than gels without FLS (none) (13.1±16.3%) (Fig. 4(a)). In addition, neuronal substance P in the low FLS group is higher than expression in gels with no FLS and only DRGs (none) (p = 0.02) and those with the high FLS concentration (p < 0.01) (Fig. 4(b)). Neuronal substance P appears to be more evident in the soma and surrounding many cell bodies (Fig. 4(b)).

Fig. 4.

Stretched gels with a low FLS concentration exhibit the greatest expression of total MMP-1 and neuronal substance P, with representative images from gels in all three groups including regions with soma and/or axons. Images are shown for gels 14, 19, and 11 for high FLS, low FLS, and none, respectively, for MMP-1 images. Neuronal substance P images are of gels 3, 18, and 11 for high FLS, low FLS, and none, respectively. (a) Total MMP-1 labeling is greatest in gels with the low FLS concentration, but is only significantly higher than the expression in gels with no FLS (*p < 0.01). (b) Neuronal substance P labeling is also observed in both DRG soma and axons, but is more diffused than MMP-1 labeling. The co-localization of positive substance P labeling with the marker βIII tubulin for neurons is significantly greater in the low FLS group than in the other two groups (*p = 0.02 versus none; #p < 0.01 versus high FLS). The scale bar is 100 μm in both panels and applies to all images. Bar graphs show summary data for each group with DIV7 and DIV9 pooled, overlaid with individual data points, each representing confocal images from a single element (open circle).

Discussion

Given that human ligamentous joint capsules include fibroblast-like synoviocytes and afferent nerve fibers in a collagenous matrix [12,14–16], the novel DRG-FLS co-culture collagen gel system here provides a platform to better mimic the human anatomy, physiology, and biomechanics. The FLS in this culture system do exhibit a dendritic morphology like those in native ligamentous capsules and distinct from morphology when on two-dimensional (2D) culture plates (Fig. 5(a)). Not surprisingly, because fibroblasts exert mechanical forces on their surrounding microenvironment [24,25,27], including FLS in the DRG-collagen gel model increases regional strains and the extent of microstructural reorganization at failure (Fig. 3), despite not altering fiber organization before loading nor changing macroscale failure properties (Table 1 and Fig. 2). Although total MMP-1 and neuronal substance P are the same regardless of FLS concentration before any load (Table 1), both of those proteins are mediated after gel failure by FLS concentration (Fig. 4). In fact, the increased protein expression exhibits differential patterns from both strain responses and fiber reorganization (Figs. 3 and 4), which suggests that either or both cell types (FLS and DRG) are sensitive to its local surroundings in this system during failure loading.

Fig. 5.

Cellular physiologic responses relevant to pain and physiology depend on the biomechanics and phenotypes of cells and also can be mediated by chemical factors. (a) Confocal images of FLS in 2D and 3D cultures show labeling for the intermediate filament vimentin and nuclear marker DAPI. FLS exhibit a flattened phenotype in 2D culture compared to an elongated and dendritic phenotype in 3D culture (in a 2 mg/mL collagen gel at DIV9). The inset image in the 2D culture shows co-labeling of vimentin with the cell surface protein expressed specifically on FLS-type fibroblasts, CD90. CD90 labeling confirms the FLS phenotype (arrowheads). (b) The co-localization of vimentin or βIII tubulin and MMP-1 shows MMP-1 co-localized with FLS and DRG axons in a co-culture gel from DIV9. The insets show a cell labeled positively (arrows) for MMP-1 that is clearly co-localized with vimentin, not βIII tubulin, suggesting that MMP-1 is from the FLS cell type. (c) In pilot experiments, increased calcium signaling is induced in DRG neurons exposed to active exogenous MMP-1. In those studies, neurons were transduced with a GCaMP6f adeno-associated virus that fluoresces with increases in intracellular calcium after MMP-1 exposure on DIV7. Raw fluorescence traces (shown for two example neurons; arrows) are acquired over time, normalized to baseline, and analyzed. The scale bar in images in (a) and (b) is 100 μm; the scale bar in the insets is 50 μm.

The behavior of fibroblasts in fibrous networks and their subsequent effect on gel mechanics varies with the culture conditions (i.e., free-floating or constrained boundary conditions) [26,70,71]. Fibroblasts in anchored collagen gels weaken the gel under tension after 6 days [72], whereas fibroblasts in free-floating collagen gels strengthen it under tension after 1 day [54]. In the current study, gels were cultured in 12-well plastic nontreated tissue-culture plates and could freely contract on their own. Indeed, gels with the low and high FLS concentrations contracted over the culture period, spontaneously detaching from the well wall and demonstrating contractile function throughout their time in culture. In fact, in pilot studies using low and high FLS concentrations and the same culture conditions as presented in the current studies, gels decreased their diameter by approximately 46–50% after 7 days in culture. In addition, although FLS were not visualized explicitly in each individual gel in this study, separate pilot studies confirmed the presence, phenotypic morphology, and viability of low and high FLS embedded in 2 mg/mL collagen gels with immunolabeling and live-dead cell assays. Dermal fibroblasts cultured with the same collagen concentration (2 mg/mL) and seeded at the same concentrations as used here increase the gel modulus by between 1.12- and 3-times depending on the concentration [54]. Although that finding does not corroborate the lack of effect on stiffness observed for the same FLS concentrations (Fig. 2), the difference may be due to functional differences between FLS and dermal or 3T3 fibroblasts, as well as different culture times since the short culture time of 1 day used in the dermal fibroblast study has been hypothesized to not allow for collagen synthesis or degradation [54]. In contrast, gels seeded separately with either human dermal or mouse 3T3 fibroblasts that decrease the failure force under uniaxial tension [72], for a similar loading rate (∼0.16 mm/s) and time in culture (6 days) but a lower cellular concentration (2.5 × 10⁴ cells/mL). Since in that study the dermal and 3T3 fibroblast-seeded gels were anchored in culture [72] and the cell-mediated gel contraction would be minimal, any decrease in failure properties would likely be due to fibroblast-mediated remodeling via enzyme degradation [72]. Although gels with the higher FLS concentration do exhibit a lower failure force and stiffness than gels without FLS, these differences are not significant (Fig. 2). Taken together, these studies suggest that early in culture strengthening of free-floating gels may be driven by cell contraction [54], and the weakening of stretched gels at later times in culture may be due to degradation [72]. It is possible that by the time of mechanical testing (DIV7, DIV9), FLS-mediated collagen degradation may have occurred, which would explain why the FLS do not change macroscale biomechanics (Fig. 2). Further, despite similar morphology and ability to synthesize and degrade ECM components across FLS, dermal, and 3T3 fibroblasts [28], their differences in contractile behavior may differentially regulate their surrounding local microstructure.

Both experimental [20,50] and computational [49,53] investigations of collagen networks under tension demonstrate fiber realignment in the direction of loading, with load redistributed as fibers parallel to the applied tension align and those perpendicular to it buckle, facilitating a transition from a bending-dominated to a stretching-dominated response. The distribution of fibers in all groups regardless of the amount of FLS exhibit a distribution with more variance at failure than at reference (Fig. 3), which likely captures the redistribution as fibers reorient toward the direction of loading (Fig. 3). Reorganization of collagen fibers along the loading direction likely also reorients the DRGs embedded in that network. Since neurons are more compliant than collagen fibers, any change in fiber kinematics can compress and/or stretch the neurons [20,73,74]. In fact, for this same neuron-collagen gel system under tension, the greatest extent of fiber alignment and elongation corresponds to the largest reorientation of neurons toward the loading direction and increased neuronal expression of pERK, an indicator of neuronal activation [20]. That finding also suggests that DRGs in all of the gel constructs used here are likely being deformed since all gels exhibit some extent of fiber reorganization, with the greatest in the high FLS gels (Fig. 3). However, the protein expression is not increased in the gels with high FLS in which fibers reorganize the most (Figs. 3 and 4); this is contrary to the finding that pERK increases when fibers reorganize the most in the collagen gels with neurons only [20]. Although fiber reorganization may be responsible for some degree of the increases in protein that are evident in all gels (Fig. 4), the disconnect between fiber reorganization and protein increases brings to the forefront the notion that different concentrations of FLS must be interacting with their collagen microstructure differently.

FLS do not alter the organization of the collagen network during culture at the concentrations and days-in vitro tested (Table 1), but under tension the collagen matrix reorganizes to different extents in a concentration-dependent manner (Fig. 3). This may be due to FLS contracting the network by different mechanisms depending on their concentration, leading to differential load distribution and fiber reorganization. For free-floating circular gels, fibroblasts contract their matrix either by tractional-force locomotion during cell migration, in which the fibers become aligned parallel to the fibroblasts around the circumference of the gel, and not in its center, or by elongation/spreading, which has no effect on the fiber organization [24–26]. During tractional-force locomotion, cytoplasmic microfilaments in fibroblasts contract and pull on cell surface integrin-collagen fibril complexes through a myosin ATPase-dependent mechanism [24]. In the cell elongation/spreading mechanism, fibroblasts reach out in all directions as they are introduced into a collagen matrix and pull collagen fibrils toward them [24]. Since at a low, but not high, fibroblast concentration, cells transition from the elongation/spreading to the locomotion mechanism [24,26], it is likely that below a critical cell concentration threshold, fibroblasts start to reorganize fibers circumferentially. Although that concentration threshold is not defined for FLS, it is possible that the same mechanisms are at play and that the high FLS concentration is orienting collagen via the elongation and spreading mechanism. If this is the case, it may explain the graded reorganization response at failure that is concentration-dependent (Fig. 3) and would imply that the distribution of forces across fibers and the embedded cells also differs between groups. Differential force distribution could affect the soft embedded DRGs differently, and it is even possible that the greater fiber recruitment in the high FLS gels might lower the loads experienced by the DRGs. Although this could explain the disconnect between fiber reorganization and MMP-1 and neuronal substance P increases (Figs. 3 and 4), quantification of the gel contracture and the microstructure across the entire gel is needed to more fully elucidate the FLS–collagen fiber interactions that may also directly and/or indirectly affect neurons.

Matrix-metalloproteinase-1 is expressed by both FLS and neurons in response to mechanical stimuli [46,47,75] and by FLS in inflammatory states [17,61]. Yet, since total MMP-1 was measured here (Fig. 4), it is not known which proportion is attributable solely to FLS or to DRGs. Since MMP-1 was quantified only in regions containing DRG somas and/or axons (Fig. 4), it is probable that positive MMP-1 immunolabeling represents extracellular MMP-1 that is co-localized to DRG cells, or MMP-1 that is sequestered in the cytosol of DRG cells [76,77]. Given the quick fixation of cells following stretch, it is likely that differences in MMP-1 protein between groups indicate differential MMP-1 regulation via cellular localization, cell sequestration, endocytic/exocytotic processes, or even stretch-induced cell rupture [33,76,77], not regulation on the transcriptional or post-translational levels, which take hours to days [46,47]. Although MMP-1 is quantified in DRG-biased images (Fig. 4), it may be FLS-secreted and could depend on FLS proliferation throughout the gel [47,54]. In fact, pilot studies co-labeling MMP-1 either with the intermediate filament vimentin, which structurally identifies fibroblasts, or with the neuronal microtubule βIII tubulin, show that MMP-1 co-localizes with both FLS and neurons, although MMP-1 co-localization appears more abundant to FLS than neurons (Fig. 5(b)), supporting the notion that stretch increases FLS-produced MMP-1 in addition to the MMP-1 that co-localizes to DRGs (Fig. 4). Although neither cell-specific MMP-1 nor FLS proliferation analyses were included in the present study since the primary goal was to investigate neuronal outcomes in the context of nociception, that work highlights the need to evaluate MMP-1 expression by cell type in order to understand if it is being secreted by FLS [47] and/or neurons [75]. Furthermore, along with neurons, DRG somas may contain Schwann cells, microglia, and resident macrophages [78], which may contribute to stretch-induced changes in cell physiology, although these cell populations were not investigated in this study.

Although MMP-1 is elevated after even a single load to failure in the presence of FLS, it is only different from gels with no FLS at the lower (5 × 10⁴ cells/mL) FLS concentration (Fig. 4), despite similar force and strains at failure in both groups with FLS (Figs. 2 and 3). This disconnect between the biomechanical and MMP-1 outcomes in the low and high FLS gels supports the assertion that the force distribution and deformations on the embedded cells are concentration-dependent and contribute to differential MMP-1 expression (Fig. 4). An alternative hypothesis is that FLS-mediated degradation occurs by the time of the mechanical testing [72], resulting in a less dense collagen network at the time of gel stretch; such an altered gel composition could lead to cell-secreted MMP-1 that is trapped in the collagen matrix to be released to the gel surface where the DRGs are, making it more available for antibody detection via immunolabeling; conversely, it could be released out of the gel into the testing bath, making it nondetectable via immunolabeling [55]. If either or both mechanisms occur, then the levels of secreted and/or trapped MMP-1 would depend on the degree of matrix degradation, an effect which might be FLS concentration-dependent since fibroblasts model and remodel their surrounding ECM [24–26,29]. Assaying secreted MMP-1 in the culture medium, together with quantifying MMP-1 proximal to cells (Fig. 4), would provide insight(s) into the effects of such factors on the local biochemical environment of joint tissues. Finally, although neither overall time in culture nor FLS concentration influences baseline protein levels (Table 1), assessing cell localizing, gel entrapped, and secreted proteins, and in larger sample sizes, would provide a more complete understanding of the cellular environment prior to stretch.

Elevated total MMP-1, regardless of the cellular source, can increase neuronal excitability [39,79,80] and/or contribute to receptor- or cytokine-mediated nociceptive signaling [33,36–38,81]. Exposure to exogenous MMP-1 alone causes an influx of intracellular calcium in DRG and cortical neurons (Fig. 5(c)) [39,79,80,82,83], which indicates more action potentials and increased neuronal excitability, and suggests MMP-1 that is increased by stretch (Fig. 4) may also excite DRG neurons independent of the stretch itself. Since increased afferent activity and neuronal hyperexcitability are characteristics of pain from joint trauma [84], it is possible that MMP-1 may have a role in vivo by potentiating afferent hyperexcitability. Further, both pro-MMP-1 and MMP-1 bind to the α₂β₁-integrin complex [37] and the PAR-1 receptor [36,39] on neurons, which are both involved in nociception and painful facet joint injury [21,85,86]. This suggests that the elevated MMP-1 (Fig. 4) in low FLS gels may increase neuronal excitability or propagate nociceptive signaling (via increased substance P) through an α₂β₁-integrin and/or PAR-1 receptor mediated mechanism. In fact, inhibiting the α₂β₁-integrin complex prevents the strain-induced increases in axonal substance P in this same DRG-collagen model under equibiaxial stretch injury (∼20% strain) [21]. Since total MMP-1 and neuronal substance P together increase in low FLS gels (Fig. 4), it is possible that the increase in neuronal substance P may be due to the concomitant increase in MMP-1 via integrin interactions. Probing MMP-1 co-expression with other pain-related neuronal receptors to which MMP-1 can bind [33,36–38] would provide important physiological insight.

The higher neuronal substance P expression evident in the low FLS gels (Fig. 4) provides a proxy for indicators of a heightened state of nociceptive signaling [19,21,66] and may also contribute to a further upstream neuronal signaling role of MMP-1 [33,36]. In addition to its ability to bind to the α₂β₁-integrin complex and affect downstream signaling cascades [37], MMP-1 stimulation of release and/or activation of other MMPs is also directly involved in substance P signaling [33,36]. Most notably, MMP-1 stimulates neuronal MMP-9 secretion [36] and activates pro-MMP-9 [33]; since MMP-9 directly cleaves substance P [87], it may also contribute to neural regulation in vivo [88]. MMP-9, like MMP-1, can sensitize peripheral neurons [89] and is necessary for early phase acute pain [89]. Whether the increased neuronal substance P (Fig. 4) is MMP-1-independent or MMP-1-dependent, that finding implies that the incorporation of FLS not only better mimics the anatomy and physiology of ligaments in a culture system but also alters cell–cell interactions to dysregulate neuronal signaling. Although MMP-1 may act independent of regulating the ECM, it is also possible that MMP-1 may alter the microstructural environment via collagen degradation [33], and in doing so, trigger afferent signaling, including upregulation of substance P [19,20,35,90]. However, MMP-1-mediated degradation depends on MMP-1 enzyme kinetics and diffusion [91], which likely occurs over longer time scales than MMP-1 signaling. Since the timing of mechanical tests and subsequent cellular assays occurred within minutes here, it is more likely that if MMP-1 mediates neuronal substance P, it is through an ECM-independent mechanism.

The finding that the mechanical effect of FLS on the collagen network is not resolved at the macroscale (Fig. 2) and varies with FLS concentration on smaller scales (Fig. 3) has implications for injury thresholds. This is further complicated for DRGs in a stretched heterogeneous network, and given that the microenvironment in capsular ligaments can vary with anatomical location [12,15,16]. Even in simple cases, uniaxial tension produces spatial variability in strain fields and fiber organization [20], such that the local microenvironment of embedded cells, and subsequently their mechano-regulated responses, also varies. This notion is further supported by computational models of the ligamentous capsule reporting altered strains and fiber kinematics experienced by neurons in a stretched fibrous network not being predicted by macroscale mechanics [35,90]. Furthermore, those models find that embedded cells can sustain strains much greater than the bulk gel stretch [90], and that “strain amplification” depends on the direction and angle of loading, cell geometry and orientation, and fiber volume and organization [35,90]. Of note, the effects of FLS have not been considered in those models. Our finding showing the (differential) effects of FLS on regional strains and microstructure (Fig. 3) supports the need to include FLS or other contractile cells into computational models to fully capture their physiological effects and improve model fidelity and perhaps help elucidate FLS–fiber interactions that are not yet able to be probed experimentally. The concentration-dependent effect of FLS on collagen gel microstructure is particularly relevant based on the regional variance and that the ligamentous capsule regions have different FLS densities [15]. Together, these findings imply there is a high likelihood that variable mechanosensitive properties may be conferred based on the regional anatomical FLS cell density. Since microstructure reorganization at failure and loading-induced protein expression are concentration-dependent (Figs. 3 and 4), it is likely that nerve fibers in FLS-dense regions reside in a microstructural and cellular environment that is unique from those in FLS-sparse regions. Although failure loading was used here, imposing subfailure stretch to this co-culture model will provide further quantitative measures of how FLS may modulate biomechanical thresholds for nociceptive signaling previously defined in the DRG-collagen gel model in the absence of FLS [19,21].

In summary, this study presents considerations for in vitro modeling of the complex biomechanical and physiological anatomic structures to better understand their multiscale behavior. Addition of FLS into an existing DRG-collagen gel model [19,21] not only alters the regional microenvironment in a concentration-dependent manner but also modulates physiologic cellular responses (Figs. 3 and 4). Certainly, there are also cell-specific variations that must be considered when incorporating fibroblasts with neurons and an ECM. The biomechanical and physiological effects of cell–cell and cell–fiber interactions are themselves mediated by the presence and extent of FLS present. Nevertheless, this study demonstrates that FLS mechanically and chemically alter the local DRG microenvironment and are critical in understanding nociceptive mechanisms in capsular ligaments. Certainly, expanding experimental techniques and applications to better visualize both the collagen microstructure and the cells within the matrix, along with quantifying bulk gel contracture, is needed to fully describe the FLS-collagen fiber interactions and how they relate to cellular and microstructural function and dysfunction, especially under load. Nevertheless, this study provides the first information about mechanical and physiological interactions between afferent fibers and FLS, with each other and their surrounding collagenous network, emphasizing the importance of considering the FLS cell type in modeling ligamentous capsules.

Funding Data

• NCCIH, National Institutes of Health (AT010326-07).

• Penn Center for Musculoskeletal Disorders: NIAMS, National Institutes of Health (T32AR007132).

• National Science Foundation (Center for Engineering MechanoBiology (CEMB): CMMI15-48571, Funder ID: 10.13039/501100008982).