Recombinant Small Leucine-Rich Proteoglycans Modulate Fiber Structure and Mechanical Properties of Collagen Gels

Abstract

Collagen is a key extracellular matrix protein found in connective tissues. The structure and organization of collagen fibers play a crucial role in determining tissue function and how tissues respond to mechanical loads. Small leucine-rich proteoglycans (SLRPs) are well-known facilitators of collagen fibrillogenesis in connective tissues. While the role of SLRPs has been extensively documented in tissues such as tendon and skin, their functions are primarily inferred from changes observed in knockout models. Additionally, their specific roles and influences of their addition to a system, particularly in collagen gel-based materials, remain underexplored. Previous in vitro studies of SLRPs have been partly limited by the challenges associated with obtaining pure SLRPs in sufficient quantities and with appropriate glycosylation. Therefore, novel methods to reliably produce SLRPs at the required quality and scale are needed. In this study, we first evaluated the feasibility of producing recombinant decorin, biglycan, and fibromodulin using HEK293-F cells. Subsequently, we investigated the effect of SLRP supplementation on high-density collagen gels using scanning electron microscopy and assessed the impact on tensile properties. Our findings demonstrated that each SLRP uniquely influenced collagen structure at both the fibril and fiber levels, consequently modifying the tissues’ mechanical response to load. Decorin, in particular, exhibited significant differences in tensile properties compared to biglycan and fibromodulin, underscoring its distinct role in promoting a structurally and mechanically robust response under tensile load.

Graphical Abstract

1. INTRODUCTION

Collagen is a key extracellular matrix protein found in mammalian connective tissues, where its collagen fiber structure and organization critically influence function and mechanical properties.^1,2 Given its essential role in these tissues, collagen-based systems have been widely explored as models for tissue engineering.^3–7 However, the intricate organization of collagen fibers in native tissues presents a significant challenge in replicating both their structural integrity and functional performance in engineered constructs. In native tissues, the organization of collagen is governed by a self-assembly process known as fibrillogenesis. Fibrillar collagens have nearly 50 binding partners in vivo, including proteoglycans, minor collagens, and fibronectin, which contribute to a large diversity of architectures found throughout these tissues.⁸ The wide variety of binding partners enables tissues to range from containing nanoscale fibers to those with large fiber bundles. In particular, small leucine rich proteoglycans (SLRPs) are well-known facilitators of the collagen fibrillogenesis process.^9–11 These proteoglycans bind to ECM molecules like collagen, aiding in the organization of these networks. While the role of SLRPs has been well documented in a variety of tissues such as tendon⁹ and skin,^12,13 their influence and specific roles in fibrocartilage tissues, and especially in collagen-gel based systems, have not been well studied.

The proteoglycans decorin, biglycan, and fibromodulin are all part of the SLRP family and are believed to play an important role in collagen fiber assembly in connective tissues.^10,14–16 Structurally, these proteoglycans consist of core proteins featuring leucine-rich repeats flanked by cysteinerich regions.¹⁷ The core proteins vary by SLRP and are glycosylated with one or more glycosaminoglycan (GAG) side chains.¹¹ These glycosylated core proteins interact with collagen fibrils, finely tuning and regulating fibrillogenesis.¹⁸ Decorin is a key player in this regulation with a structure that allows it to align either parallel to or at a right angle to collagen, facilitating both fusion and axial alignment of fibrils.¹⁹ Decorin has been shown to manage the lateral growth of fibrils, controlling spacing between fiber bundles.^12,19,20 Decorin is found in almost all tissues and has been well studied in skin and tendon, where it has been associated with the lateral and linear growth of fibrils, which regulates the tensile properties of these tissues.^13,21,22 In knockout models, decorin-free mice exhibited fibrils in their tendons and skin that were irregular in size and shape, accompanied by a decrease in tensile strength.¹² While the role of biglycan is less well-known than decorin, biglycan null mice show irregular collagen fiber formation with broader fibril sizes.²³ Fibromodulin delays fiber formation and modulates the assembly of collagen monomers into fibrils.^24–26 In early development of many tissues, decorin, biglycan, and fibromodulin are all present and contribute to both the establishment of the collagen fiber network as well as their homeostasis.

Knockout models studying skin and tendon have been used to try and elucidate the specific roles of these SLRPs in collagen fibrillogenesis during development.^12,23,26 While this is a useful tool, often it is hard to specifically probe the individual effects of these SLRPs due to compensatory mechanisms or developmental abnormalities that occur in these knockouts.^26,27 For example, in fibromodulin null mice, the expression of the SLRP lumican is highly upregulated, making it difficult to determine whether the effects of the knockout are due to the loss of fibromodulin, the increase of lumican, or some mixture of both.^26,27 Collagen fibrillogenesis is an entropy driven process that can be mediated by these proteoglycans. Studying this process in vitro would help minimize the complications of studying the effects of these SLRPs on this self-assembly process in vivo.²⁸ Although in vitro fibrillogenesis of collagen is more simplified than what occurs in vivo, a collagen-based system would provide a model to more specifically probe the roles individual SLRPs and how they affect this fibrillogenesis process. Additionally, most studies examining SLRPs have previously focused on what occurs in their absence, as in knockout models. Few have looked at what happens due to their specific addition. Several studies have used in vitro collagen based systems to try to elucidate the effects of these SLRPs on this process, however, these studies use very low concentrations of collagen (1–3 mg/mL) and have not explored the effects of all three SLRPs.^19,29–31 Tissue engineering applications using collagen-gel based systems often require much higher concentrations of collagen (~20 mg/mL) to enhance mechanical properties, decrease contraction compared to low density gels, and promote the formation of larger fiber bundles.^4,5,32–35

In addition to the role that these SLRPs play on the collagen fiber formation, the resultant networks that form are largely responsible for their mechanical properties. Understanding how each of these SLRPs individually affects the fiber network and that network’s contribution to the mechanical properties of the tissue as a whole is essential for fine-tuning the complex architecture of collagen dense tissues with complex and regionally varied structures. This knowledge is key to developing tissue engineered replacements and collagen-based biomaterials that accurately replicate the composition, structure, and zonally dependent mechanical properties of native tissues. The role of decorin, biglycan, and fibromodulin on fiber formation, fiber properties, and resultant effects on mechanical properties have not been evaluated in a high density collagen-based tissue engineered model. Previous works have investigated various biomechanical and biochemical tools to manipulate the fiber structure of collagen gels to achieve a desired fiber structure.^33,36,37 Additionally, we have demonstrated that fibril formation in these collagen-gel based systems is enhanced either by specifically decreasing the production of aggrecan or degrading the GAG side chains.^38,39 However, it is unknown how these SLRPs can facilitate the development of fiber networks in vitro and how their individual contributions can affect the collagen architecture and subsequent mechanical properties of collagen gel-based tissues. We hypothesize that decorin, biglycan, and fibromodulin will all have distinct effects on the collagen fiber formation and architecture of high density collagen gels and that these fiber structures will affect the mechanical response of the tissue.

Controlling fibrillogenesis and collagen fiber assembly is crucial for tailoring final fiber properties. Taking inspiration from in vivo observations, SLRPs offer a potential avenue for manipulating and controlling the production of collagen-based biomaterials. Although progress has been limited, several promising examples demonstrate the use of SLRPs to regulate fibrillogenesis in vitro.^19,29–31 However, a significant challenge in scaling up their application lies in the limited availability of high-quality, high-volume SLRPs, which hinders both further research and large-scale manufacturing.⁴⁰ Previous in vitro studies of SLRPs have been partly limited by the challenges of obtaining purified SLRPs. Commercially available sources can be contaminated with other SLRPs, complicating efforts to isolate their individual effects.³⁰ While tendon- or skin-derived SLRPs can be collected and purified, this approach is labor-intensive and yields limited amounts of protein.⁴⁰ An alternative approach based on recombinant production in suspension-adapted HEK293-F was demonstrated at smaller scales for several SLRPs that were expressed using episomal vectors.⁴¹ HEK293-F has been shown to preserve human glycosylation patterns for some glycoproteins, such as lubricin, which is densely grafted with mucin-type O-glycans.^42,43 However, the capacity of HEK293-F or similar cell platforms to construct appropriate GAG chains on SLRPs during stable production has not been evaluated. The incorporation of appropriate GAG chains on SLRPs is crucial to achieve the full biological activity of SLRPs in collagen fibrillogenesis.^44,45

Therefore, in this work, we examined how recombinant decorin, biglycan, and fibromodulin influence fibril and fiber diameter, porosity, and mechanical properties in a collagen-gel based tissue engineered construct. In this study we first demonstrated the feasibility of production of recombinant decorin, biglycan, and fibromodulin with appropriate GAG chains via HEK293-F cells. We then supplemented collagen gels with these recombinantly produced SLRPs and investigated collagen fiber formation and mechanical performance. Examining the role of SLRPs in collagen fibrillogenesis within an in vitro collagen-gel system provides valuable insights into their specific functions in regulating fiber size, porosity, and the mechanical properties of these networks, while also offering a deeper understanding of their roles in vivo.

2. EXPERIMENTAL SECTION (MATERIALS AND METHODS)

2.1. Cell Lines and Culture.

Freestyle 293-F cells (Thermo Fisher), a suspension-adapted variant of HEK 293 cells, were maintained according to the manufacturer’s guidelines in glass Erlenmeyer flasks (VWR, Kimax, Pyrex) with silicone sponge closures (Chemglass). Cells were maintained between 0.3 × 10⁶ and 3 × 10⁶ cells/mL at 130 rpm, 37 °C, and 8% CO₂ in 293 Freestyle Expression Medium (Thermo Fisher). Viability and cell counts were obtained utilizing a hemocytometer with trypan blue exclusion (Life Technologies).

2.2. cDNAs and Expression Vectors.

The cDNAs encoding decorin (NCBI Reference Sequence 1634), biglycan (NCBI Reference Sequence 633), and fibromodulin (NCBI Reference Sequence 2331) with C-terminal polyhistidine tags (GGHHHHHH) for purification were generated by custom gene synthesis (Twist Bioscience). The proteoglycan cDNAs were each inserted into a previously reported Piggybac expression vector that included a cassette for inducible expression of the desired protein and a bicistronic fluorescent reporter of expression under the control of a minimal CMV promoter with seven tetO sequences.⁴⁶ In these studies, mNeonGreen was used as the fluorescent protein reporter. The “all-in-one” vector also included a second cassette for constitutive expression of the reverse tetracycline transactivator, rtTA-M2, and a bicistronic neomycin resistance marker under the control of an EF1α promoter.

2.3. Generation of Stable Cell Lines.

Stable 293-F cell lines were generated by cotransfection of Piggybac plasmids for decorin, biglycan, or fibromodulin with a hyperactive transposase plasmid.⁴⁷ Transfection was performed with the use of polyethylenimine (PEI) as previously described (Figure 1A – 1).⁴⁸ Transfected cells were selected with G418 (500 μg/mL) for 2 weeks (Figure 1A – 2). Cells expressing high levels of the mNeonGreen reporter were enriched by one round of fluorescence-activated cell sorting (FACS). Before FACS, cells were expanded to 0.5–1 × 10⁶ cells/mL and induced with doxycycline (1 μg/mL) for 24 h. FACS was performed on a FACSMelody (BD Biosciences) by gating the top 25% mNeonGreen expressing cells (Figure 1A – 3). Following expansion, the sorted cells were frozen at a density of 1 × 10⁷ viable cells/mL in a controlled rate cooling device (Corning CoolCell) in cryopreservation media composed of 45% fresh 293 Freestyle Expression Medium, 45% conditioned 293 Freestyle Expression Medium, and 10% DMSO.

Figure 1.

Methods for short leucine rich protein (SLRP) production and assessment of SLRP production. (A) Flowchart of stable cell line generation and SLRP production via (1) transfection of plasmids for the SLRP with bicistronic mNeonGreen reporter and hyperactive transposase, (2) cell selection, (3) fluorescence-activated cell sorting (FACS), (4) cell-line expansion and induction, (5) media clarification, and (6) purification leading to purified SLRPs. Partially generated with a premium Biorender account. (B) Western blots of the supernatant of cultures of decorin, biglycan, and fibromodulin for assessment of protein production.

2.4. Production of Recombinant SLRPs.

Cryopreserved cells were thawed in 125 mL flasks containing 30 mL fresh media. Cultures were further expanded in 100 mL of media to 1 × 10⁶ cells/mL in 500 mL flasks. Cultures were induced with 1 μg/mL doxycycline. Cell viability was monitored daily and additional doxycycline (1 μg/mL) was added to the culture every 48 h. Production was terminated when cell viability dropped below 80%, typically around 4–5 days (Figure 1A – 4). Cells were centrifuged at 200g for 5 min and the clarified supernatant was collected for further purification (Figure 1A – 5).

2.5. Purification.

Recombinant SLRPs were purified by immobilized metal affinity chromatography (IMAC) taking advantage of the polyhistidine tags on the recombinant proteins. Hispur Ni-NTA resin (Thermo Fisher) was used per the manufacturer’s guidelines. Briefly, 0.1 mL of resin equilibrated in equilibration buffer (300 mM NaCl, 50 mM K₂HPO₄, 10 mM imidazole, pH 8) per 10 mL sample was added to the clarified culture supernatant with 10X Ni-NTA binding buffer (1.5 M NaCl, 0.5 M K₂HPO₄, 0.1 M Tris, 100 mM imidazole, pH 8.5) and incubated with end-overend mixing at 4 °C for 20 min. This supernatant-resin mixture was loaded onto spin columns, washed three times with wash buffer (300 mM NaCl, 50 mM K₂HPO₄, 20 mM imidazole, pH 8), and eluted in elution buffer (300 mM NaCl, 50 mM K₂HPO₄, 250 mM imidazole, pH 8) (Figure 1A – 6). The elution buffer was exchanged for phosphate-buffered saline (PBS) using Zeba spin desalting columns, 7K MWCO, according to manufacturer recommendations (Thermo Fisher).

2.6. SDS-PAGE Analysis and Immunoblot.

Unpurified and purified samples were denatured by heating at 95 °C for 5 min in NuPAGE LDS sample buffer with NuPAGE sample reducing agent (Invitrogen). The denatured samples were separated on NuPAGE 4–12% Bis-tris gels (Invitrogen) with MES-SDS running buffer according to the manufacturer’s guidelines. Gels were transferred onto PVDF (Millipore) membranes to assess protein production by Western blot. The PVDF membranes were blocked with 5% (w/v) bovine serum albumin (BSA) in tris-buffered saline +0.1% Tween 20 (TBST) for 30 min at room temperature. Rabbit antibiglycan (Monoclonal ab231297, Abcam), antidecorin (Monoclonal ab277636, Abcam) and antifibromodulin (Monoclonal ab267465, Abcam) IgG primary antibodies were each diluted 1:1000 in TBST, incubated on membranes overnight at 4 °C, and washed three times with TBST for 5 min. The goat antirabbit IgG (H+L) DyLight 800 4X PEG (Invitrogen) secondary antibodies were diluted 1:2000 in TBST, incubated on membranes for 1–2 h at room temperature protected from light, and washed three times with TBST for 5 min. Blots were imaged on a Chemidoc HP Imaging System (Bio-Rad). Image processing was performed using Fiji ImageJ software (ImageJ).

2.7. RNA Extraction and Library Preparation.

HEK293-F cells were lysed using TRIzol (Thermo Fisher Scientific) and stored at −80 °C for >24 h to improve RNA quality and yield. Thawed Trizollysates (on ice) were treated with chloroform (Sigma-Aldrich) to partition cellular RNAs into the aqueous layer and RNA extraction was performed using the RNeasy Mini Kit (QIAGEN) following the manufacturer’s instructions. Quantity and purity of the isolated RNAs were estimated through absorbance at 260 nm, and 260/280 and 260/230 ratios, respectively, on a NanoDrop Spectrophotometer (Thermo Fisher Scientific). RNA quality was assessed through ribosomal 28S/18S peak ratios on electropherograms obtained using Tapestation QC assay (Agilent). Only samples that scored 10.0 for RNA Integrity Number equivalent (RINe), indicating little or no degradation, were used to construct libraries by the Genomics Facility at Cornell Institute of Biotechnology using the QuantSeq 3′ mRNA-Seq Library Prep Kit (Lexogen) following the manufacturer’s protocol. 3′ mRNA-seq library preparation was selected for its ability to generate one fragment per transcript to minimize length-biased expression reads for downstream differential gene expression (DGE) analysis.

2.8. RNA Sequencing and Glycosylation Pathway Analysis.

Sequencing of the QuantSeq 3′ mRNA libraries was performed by the Genomics Facility at Cornell Institute of Biotechnology as high-throughput 1 × 86 bp single-end reads on an Illumina NextSeq500 sequencer to yield at least 333–400 M reads. The quality of the generated FastQ files was checked with the fastqc tool (v.0.12.1).⁴⁹ FastQ files were right-end-trimmed by 10 bases and filtered for a minimum Phred quality score of 20 using the BBDuk tool from BBtools (v.39.08).⁵⁰ Processed reads were aligned to the human reference genome GRCh38.p13 using STAR aligner (v.2.7.11b) with default alignment parameters.^50,51 Aligned read counts were generated using the featureCounts tool from Subread (v.2.0.3).⁵² TPM-normalized gene counts were submitted for glycosylation pathway analysis with GlycoMaple (GlyCosmos Portal v.4.1.1).⁵³

2.9. GAG Characterization.

Purified recombinant SLRP samples were incubated with enzymes known to degrade certain GAG groups. To 20 μL of each purified sample, Chondroitinase ABC (Sigma, C3667), Heparinase III (New England Biolabs, P0737S), or both, were added to the samples. Chondroitinase ABC was reconstituted in a Tris buffer (50 mM Tris, 60 mM NaOAc, 0.02% BSA, pH 8.0^54,55 to a stock concentration of 10 U/mL. In Chondroitinase ABC treated samples, 1 μL of the enzyme was added to a final concentration of 0.5 U/mL. The manufacturer’s guidelines for Heparinase III treatment were followed: 2 μL of 10X enzyme reaction buffer (provided with enzyme) and 0.2 μL of enzyme were added to each sample, leading to a final concentration of 7 U/mL. All samples were incubated at 37 °C with constant orbital mixing at 120 rpm for 2 h, alongside untreated samples as controls. Samples were prepared and loaded into SDS-page immunoblot gels exactly as described in 2.6 for all three SLRPs.

2.10. Collagen Gel Prep.

Type I collagen was extracted from Sprague−Dawley rat tail tendons (Pel-Freez Biologicals, Rogers, AZ). Tendon bundles were removed from skinned tails and collagen was solubilized in 0.1% (v/v) acetic acid. Once solubilized, the suspension was centrifuged and the supernatant was removed. Finally, the supernatant was lyophilized and reconstituted in 0.1% (v/v) acetic acid at a concentration of 30 mg/mL, as previously described.^32,34,56

To assess the effects of the addition of SLRPs on collagen fiber formation and mechanical properties in a collagen-based tissue engineered construct, two model systems were used: a disc and a linear (dog-bone) geometry. Using a 3-way stopcock, a working solution of 1x PBS, 10x PBS, and 1 M NaOH was mixed with the reconstituted collagen solution to bring the pH to 7 to initiate gelation. The collagen gel was then mixed with the SLRPs at a final concentration of 20 mg/mL collagen and 5 nM decorin, biglycan, or fibromodulin.^33,34 Disc gels were formed by immediately injecting collagen/SLRP mixture between two glass plates set apart at 2 mm and allowed to gel before 8 mm punches were taken. For linear samples, collagen/SLRP mixtures were injected into 25 × 5 × 3 mm PDMS molds. Gels were polymerized for 30 min at 37 °C and either fixed for SEM or immediately mechanically tested.

2.11. Fiber Analysis via Scanning Electron Microscopy.

To assess the effects of the addition of SLRPs on the collagen structure at the micro and nanoscale, scanning electron microscopy (SEM) was performed.⁵⁷ After polymerization, samples from each group were fixed in 2% glutaraldehyde in HEPES buffer for 24 h. Samples were then washed with HEPES [x2] for 10 min each and subsequently dried with a series of dilution of water into ethanol (25, 50, 70, 90, and then 100% [x2]) for 10 min each. After soaking in 100% ethanol for 24 h, samples were dried using critical point CO₂ until complete ethanol removal. Samples were fixed to 18 mm aluminum specimen mounts using double-sided adhesive and silver conductive paint.⁵⁷ Prior to imaging, samples were sputter-coated with gold/palladium alloy for 15 s at a target current of 20 mA. Prepared construct samples were imaged on a Tescan Mira3 field emission SEM (Cranberry, Township, PA).⁵⁷ Collagen fibril diameters, porosity, and contour-to-end ratio were measured manually using ImageJ. Images were divided into a 4 × 4 grid and the diameters of fibrils were measured until the average remained constant (after measurement of ~ 100 fibrils per micrograph). Porosity was measured by thresholding the SEM images using ImageJ and using the Analyze Particles function. Contour-to-end ratio was calculated by tracing the length of the fiber, tracing the area between the two “ends”, and finding the ratio between them.^58,59

2.12. Mechanical Analysis via Tensile Testing.

Tensile tests were performed on the linear (dog-bone) tissue engineered constructs on an Enduratec Electroforce 5500 System (Bose, Eden Prairie, MN), as previously described.⁶⁰ Constructs were clamped at the dog-bone edges, and loaded into the tensile grips with sandpaper at the ends to prevent slippage. Samples were pulled to the full gauge length of the testing frame (12 mm) or until failure at a rate of 0.15 mm/s for quasi-static loading.⁶¹ Once tensile testing was complete, samples were released from the clamps and then fixed in 2% glutaraldehyde in HEPES buffer for 24 h. Samples were then prepared for SEM as described above. Length dimensions for mechanical property calculations were taken as the distance between the clamps before testing. Ultimate load, Young’s modulus, toughness, and ultimate tensile strength are calculated from measured stress and strain curves.⁶¹

2.13. Statistics.

Data was assessed for normality and if normal, values were reported as mean ± standard deviation. If normal, data were analyzed using one-way analysis of variance with Tukey’s honestly significant difference for post hoc analysis. If data was not normal, values were reported as mean ± 95% CI. If not normal, data were analyzed using a Kruskal-Wallis test using Wilcoxon for pairwise comparisons. p < 0.05 was considered significant. Statistical analysis was performed using R Statistical Software (R Foundation for statistical computing, Vienna, Austria) and GraphPad Prism (GraphPad Prism Software INC., San Diego, CA).

3. RESULTS

3.1. Recombinant Production of SLRPs.

HEK293-F cells enable high titer, scalable production of proteins in mammalian cells and recapitulate many aspects of native human glycosylation.^42,43 We therefore selected HEK293-F cells as the cell production platform for small leucine rich proteoglycans (SLRPs). Stable expression of polyhistidine-tagged decorin, biglycan, and fibromodulin was achieved through transposon-mediated integration of the SLRP gene with a bicistronic mNeonGreen reporter under the control of a tetracycline-inducible promoter. The bicistronic mNeonGreen reporter enabled isolation of the stable expressing HEK293-F populations through fluorescence activated cell sorting (FACs). Production of the SLRPs was confirmed through Western blot analysis (Figure 1B). We observed a smeared pattern of high-molecular weight products indicative of extensive glycosylation. Total yields following SLRP purification via immobilized metal affinity chromatography exceeded 10 mg per liter of 293-F culture media. Although we did not attempt to optimize productivity, in principle, the bicistronic mNeonGreen reporter could be used for enrichment of high expressing subpopulations of the stable HEK293-F cells.⁴² Together, our results demonstrated the suitability of HEK293-F for stable production of glycosylated SLRPs.

3.2. GAG Characterization of Recombinant SLRPs.

Given the importance of GAGs to proteoglycan function, we considered the capacity of HEK293-F to decorate recombinantly produced SLRPs with appropriate GAG chains. We noted that decorin and biglycan are exclusively chondroitin/dermatan sulfate-substituted proteoglycans, whereas fibromodulin is substituted with keratan sulfate.⁶² We first assessed the chondroitin/dermatan sulfate content of the purified SLRPs through enzymatic digestion with chondroitinase ABC, which digests chondroitin sulfate and dermatan sulfate but does not act on keratan sulfate, heparin or heparan sulfate. The addition of chondroitinase ABC led to a clear decrease in the molecular weight of recombinant decorin and biglycan and eliminated most of the GAG-dependent smearing of the two proteoglycans on an SDS-PAGE gel (Figure 2A). Treatment of fibromodulin with chondroitinase ABC did not noticeably alter its molecular weight when analyzed by SDS-PAGE, consistent with expectations based on native fibromodulin being exclusively keratan sulfate substituted. Treatment with Heparinase III did not reduce the molecular weight of any of the recombinant SLRPs when analyzed by SDS-PAGE, indicating that heparan sulfate chain substitution was minimal, mirroring the glycosylation of native SLRPs (Figure 2A).

Figure 2.

Reconstruction of the biosynthetic pathway for glycosaminoglycans (GAGs) using GlycoMaple. (A) Analysis of GAG content in recombinant decorin, biglycan, and fibromodulin. Purified proteins were untreated or treated with 0.5 U/mL of chondroitinase ABC or 7 U/mL of heparanase III for 2 h at 37 °C. (B) GAG biosynthetic pathway in suspension-adapted HEK293-F. Reaction arrows are scaled and color-coded based on the normalized gene expression (transcripts per million; TPM) of the governing enzymes analyzed by RNA-Seq If several enzymes have overlapping function in a reaction, the maximum TPM value among the values of overlapped genes was used. When several gene products make a complex for a reaction, the minimum TPM value of the subunit genes was used. GAG structures are drawn using the standard symbol nomenclature for glycans (SNFG). (C) Analysis of pathways for additional modification of chondroitin sulfate and dermatin sulfate in 293-F.

Precise structural determination of extended GAG chains is an unresolved analytical challenge. As an alternative, GAG biosynthetic pathways in 293-F were computationally reconstructed based on gene expression analysis using the glycosylation mapping tool called GlycoMaple.⁵³ The 293-F transcriptome was profiled with RNA-Seq and analyzed to yield gene expression data normalized as transcripts per million (TPM). Computational mapping of the GAG biosynthetic pathways predicted that 293-F have the necessary capacity to form the common tetrasaccharide linker of chondroitin sulfates, dermatan sulfates, and heparan sulfates. Interestingly, we noted relatively low expression of CSGALNACT1 and CSGALNACT2, which governs the funneling of the common tetrasaccharide linker down the biosynthetic pathway for chondroitin/dermatan sulfate (Figure 2B). However, our experimental analysis of the recombinant proteins with chondroitinase ABC indicated sufficient capacity in HEK293-F to generate chondroitin/dermatan sulfate chains (Figure 2A).

Further assessment of the GAG pathways predicted that 293-F have full capacity for modification of nascent chondroitin, dermatan, and heparan sulfate chains into more complex and extended structures (Figure 2C& Supplemental Figure 1). The reconstructions for keratan sulfate biosynthesis predicted a capacity to initiate chain construction in 293-F. However, we noted only modest expression of B3GNT7, which encodes a β1,3 N-acetylglucosaminyltransferase that is critical for elongation of the keratan sulfate backbone (Supplemental Figure 2). Consistent with this prediction, recombinant purified fibromodulin produced in 293-F exhibited a molecular weight of approximately 50 kDa, which is larger than the expected molecular weight of the unsubstituted protein core (43 kDa), but less than the ~ 59 kDa molecular weight that has been reported for native fibromodulin.⁶³ Taken together, our experimental and computational analyses demonstrate the potential of the 293-F system as a platform for production of GAG-substituted SLRPs. In particular, the molecular weights of GAG-substituted decorin and biglycan were comparable to those previously reported for the native proteoglycans (Figure 1B&2A).^64,65

3.3. Assessing Fiber and Fibril Organization in SLRP-Supplemented Constructs via Scanning Electron Microscopy.

Collagen gel discs were imaged using SEM to evaluate the effects of SLRPs on fiber formation at both the fibril and fiber levels. Constructs included controls as well as those supplemented with decorin, biglycan, or fibromodulin. At low magnification, distinct fiber architectures were observed (Figure 3). Control constructs showed minimal organization at the fiber level, with a mean fiber bundle size of 566 ± 67 nm (95% CI) (Figure 3A,E&F). Similarly, decorin constructs exhibited little organization, with a mean size of 546 ± 36 nm (95% CI) (Figure 3B,E&F). In contrast, biglycan-supplemented constructs displayed significantly larger fiber bundles, with a mean size of 1034 ± 78 nm (95% CI)—1.83, 1.89, and 2.95 times larger than those of control, decorin, and fibromodulin constructs, respectively (Figure 3C,E&F).

Figure 3.

Assessment of fiber and fibril organization in SLRP-supplemented constructs. (A–D) Representative scanning electron microscopy (SEM) micrographs of collagen, decorin, biglycan, and fibromodulin-supplemented gels at low magnification (view field = 100 μm). Scale bars = 20 μm. (E) Normalized overlaid histograms of the fiber diameters measured from low-magnification (view field = 100 μm) SEM micrographs with means and 95% CIs reported. (F) Mean fiber diameters of SLRP-supplemented and control gels measured from low magnification (view field = 100 μm) SEM micrographs. Error bars are ± 95% CI, **** = p < 0.0001, * = p < 0.05.

To assess the process of fibril assembly, probability density functions (PDFs) of fiber size were generated for all groups (Figure 3E). All PDFs displayed a primary peak centered around ~ 400 nm and tails at higher diameters with varying number and magnitude of secondary modes. PDFs were analyzed by fitting a Gaussian curve centered around the primary peak of the data (Supplemental Figure 3). To quantify how much data deviated from this Gaussian peak, the areas under the Gaussian fit and the original PDF were calculated, and their ratio was determined. To quantify the deviation of the data from the Gaussian peak, the ratio was subtracted from one to calculate the Peak Deviation Index (PDI). The PDI represents the proportion of data that is not captured by the primary peak, providing a measure of how much the distribution deviates from a “normal” Gaussian distribution and hence how much of the collagen has been assembled into larger structures. The biglycan group exhibited a much wider spread in fiber diameter compared to other groups. In addition to the increased spread and skewed data distribution the PDFs revealed a prominent peak, along with additional modes indicating larger fiber diameters. The biglycan data exhibited the highest PDI at 0.56 (95% CI: 0.50–0.62), indicating that 56% of the data fell outside the Gaussian peak, making it the most deviated from a “normal” distribution. Furthermore, biglycan data displayed multiple modal peaks, with diameters close to multiples of the primary peak and reaching up to 5 μm. The presence of these higher modal peaks suggests that some initial base structure, represented by the primary peak, undergoes bundling and lateral aggregation. In contrast, fibromodulin had the lowest PDI at 0.01 (95% CI: −0.21–0.24), indicating minimal bundling of fibers. The control and decorin data sets showed similar PDIs of 0.25 (95% CI: 0.09–0.42) and 0.23 (95% CI: 0.15–0.31), respectively, suggesting comparable levels of bundling. Although these groups also showed additional modes, they were fewer in number and smaller in magnitude compared to biglycan. Where these additional modes did appear, they seemed to occur at multiples of the primary peak, potentially reflecting a base structural unit with subsequent modes corresponding to the bundling of these collagen structures.

Fibromodulin constructs exhibited the smallest fiber bundles, with a mean size of 350 ± 21 nm (95% CI) (Figure 3D,E&F). To assess fibril level structure, higher magnification SEM images were acquired (Figure 4). Interestingly, trends in fibril diameter differed slightly from those observed at the fiber level. Control constructs had the largest mean fibril diameter at 160 ± 3.8 nm (95% CI), followed by fibromodulin constructs at 134 ± 3.2 nm (95% CI). Biglycan and decorin constructs had mean fibril diameters of 126 ± 4.3 nm (95% CI) and 99 ± 3.2 nm (95% CI), respectively (Figure 4E&F). The fibril structures were more distinct at this scale, with thicker but less organized, rope-like fibrils in the control group, a disordered arrangement in the decorin constructs, a more webbed architecture in biglycan constructs, and a more open architecture in the fibromodulin group.

Figure 4.

Higher magnification images of SLRP-supplemented constructs. (A–D) Representative SEM micrographs of collagen, decorin, biglycan, and fibromodulin-supplemented gels at high magnification (view field = 25 μm). Scale bars = 5 μm. (E) Normalized overlaid histograms of the fiber diameters measured from high magnification (view field = 25 μm) SEM micrographs with means and 95% CIs reported. (F) Mean fiber diameters of SLRP-supplemented and control gels measured from high magnification (view field = 25 μm) SEM micrographs. Error bars are ± 95% CI, **** = p < 0.0001.

To quantify the complex architectures facilitated by these SLRPs, the contour-to-end ratio of the fibers was calculated (Figure 5A,B). This ratio, analogous to persistence length, is determined by dividing the trace distance of a fiber by its end-to-end distance. Fibromodulin constructs exhibited the highest mean ratio at 1.23 ± 0.05 (95% CI), indicating the greatest fiber undulation. Decorin and biglycan constructs showed similar ratios of 1.07 ± 0.02 (95% CI) and 1.11 ± 0.03 (95% CI), respectively, while the control group had the lowest ratio at 1.03 ± 0.01 (95% CI) (Figure 5C). These measurements were performed at the fibril level to capture the local fibril structure within the constructs. Looking at the PDI, the control group had a very low deviation of data from the primary peak, with a PDI of 0.03 (95% CI: 0.027–0.031) (Figure S4). Decorin and biglycan had slightly higher deviations with 0.23 (95% CI: 0.219–0.246) and 0.35 (95% CI: 0.346–0.358) PDIs, respectively. Fibromodulin had a large deviation, with a PDI of 0.56 (95% CI: 0.552–0.567). This high deviation in the fibromodulin group corresponded to a broader data spread and a greater number of modes compared to the other groups, with contour to end ratios reaching close to 2. The porosity of the fiber networks was also analyzed (Figure 6), with fibromodulin constructs showing the highest porosity, followed by biglycan, decorin, and finally the control group. Collectively, these findings suggest that all SLRPs contribute to more intricate fiber networks compared to controls, characterized by increased undulation and a more open, porous architecture.

Figure 5.

(A) Example of SEM image with overlaid contour-to-end ratios with trace length and end to end lengths. Scale bar = 5 μm. (B) Probability density functions of the contour-to-end ratios measured from high magnification (view field = 25 μm) SEM micrographs with means and 95% CIs reported (* = p < 0.05, *** = p < 0.001). (C) Mean contour-to-end ratios. Error bars are ± 95% CI.

Figure 6.

(A–D) Representative SEM micrographs at high magnification (view field = 25 μm). (E) Weighted pore size from SLRP-supplemented and control gels measured from high magnification SEM micrographs. *** = p < 0.001.

3.4. Tensile Properties.

To explore how differences in fiber diameter, contour-to-end ratio, and porosity influence the mechanics of tissue engineered menisci, bulk tensile testing was performed (Figure 7). The mechanical assessment revealed a 55% increase in ultimate load (Figure 7A) and 51% increase in ultimate tensile strength (UTS) (Figure 7B) in decorin-supplemented constructs compared to controls, with a 2.6-fold increase in toughness (p = 0.001, Figure 7C). Decorin-supplemented constructs also outperformed biglycan and fibromodulin groups in ultimate load and toughness, though UTS values were comparable, however the differences between SLRP groups in ultimate load were not statistically significant. Young’s modulus was also comparable between groups with decorin and biglycan being slightly higher than control and fibromodulin, although these differences were not statistically significant (Figure 7D). Strain at failure was also elevated by 2-fold in decorin treated constructs compared to controls (Figure 7E). Qualitative evaluation of SEM images near failure or defect sites (Figure 8 A&B) in the tensile-tested constructs revealed intriguing fiber network responses, providing additional insights into the structural behavior under mechanical stress. While all constructs had areas where there were tears post-tensile testing, control constructs showed areas between larger pieces of samples that were torn, with very thin and less interconnected networks of collagen fibers (Figure 8C&D). Mechanical testing seemed to have pulled the fibers in the decorin constructs into alignment. Frayed edges appear but with more structural integrity compared to controls (Figure 8E&F). Fibers in the biglycan constructs were frayed, but the structure was fairly compact compared to the others (Figure 8G&H). Fibromodulin’s structure was similar to biglycan at higher magnifications but lower magnification images showed longer frayed fibers with a slightly more open structure (Figure 8I&J).

Figure 7.

Tensile properties of SLRP-supplemented and control linear constructs (n = 5–6). (A) Ultimate load. (B) Ultimate tensile strength. (C) Toughness. (D) Young’s modulus. (E) Strain at failure. Statistical error bars are ± 95% confidence interval (* = p < 0.05). Analyzed using 1-way ANOVA with Tukey’s multiple comparisons test. (F–I) All stress strain curves of control, decorin, biglycan, and fibromodulin samples. The darkest curve represents the median example.

Figure 8.

(A) Image of linear (dog-bone) constructs in tensile test grips before testing. (B) Image of linear construct in tensile test grips after test. Dashed red box indicates where SEM images were taken after testing. Scale bars = 10 mm. (C–J) Representative SLRP-supplemented and control gels imaged using SEM after tensile testing. All images were taken at the failure site or at a site of fracture if no full failure site was present in the sample. Left column micrograph scale bars = 5 μm. Right column micrograph scale bars = 20 μm.

4. DISCUSSION

The objective of this study was to investigate how decorin, biglycan, and fibromodulin influence the fiber structure in a collagen gel-based system, and how these structural changes induced by SLRP supplementation affect bulk tensile properties. Previous studies have explored various biomechanical and biochemical tools to manipulate fiber architecture development in vitro. Here, we hypothesized that decorin, biglycan, and fibromodulin would each have distinct effects on the collagen fibril assembly and subsequent mechanical performance of these collagen gel-based constructs. To test this, we used our previously established high-density collagen-based tissue engineered model using both disc and linear (dog-bone) geometries for imaging and mechanical testing. Our findings demonstrated that each SLRP uniquely influenced collagen structure at both the fibril and fiber levels, thereby altering the samples’ mechanical response to load. Decorin, in particular, exhibited significant differences in tensile properties compared to other groups, highlighting its distinct role in promoting a structurally and mechanically robust response under tensile load.

Small leucine rich proteoglycans are known to play an important role in the collagen fibrillogenesis process. While their functions have been well documented in tissues like tendon and skin, their roles and effects in high density collagen based models remain unclear.^14–16 Additionally, the function of SLRPs has been mostly inferred by changes that happen due to their removal, as in knockout models. Constitutive knockouts have been used to study SLRPs in mice, but it is challenging to determine whether the observed differences are due to abnormal tissue development caused by the absence of SLRPs during development, or due to altered tissue responses resulting from their absence during normal tissue function.⁶⁶ Inducible knockouts provide more precise insights into the effects of SLRP absence during tissue function without disrupting development.⁶⁶ However, the knockout of SLRPs across all tissues complicates interpretation due to the structural complexity of joints, which comprise multiple tissues containing SLRPs. Tissue-specific knockouts could help disentangle the effects of SLRPs on individual tissues. While studying these knockout models could elucidate the physiological importance of each SLRP within specific tissues, validating the molecular activities of each SLRP requires well-defined in vitro systems under the modulation of each SLRP and its interacting counterparts, such as collagens or other signaling molecules. In summary, studying the roles of individual SLRPs remains challenging due to the complexities introduced by in vivo factors.

Suspension-adapted HEK293-F cells were demonstrated as a platform for recombinant generation of SLRPs substituted with GAG chains. Chondroitin/dermatan sulfate substituted decorin and biglycan were recombinantly produced with molecular weights comparable to their native counterparts, suggesting that the 293-F cells had sufficient biosynthetic capacity for GAGs, at least at the modest scales of production reported here. However, analysis of the GAG biosynthetic pathways in 293-F with RNA-Seq revealed relatively low expression of CSGALNACT1 and CSGALNACT2, which encode enzymes that are crucial for chondroitin sulfate chain initiation and elongation. These results suggest a potential opportunity to further enhance the 293-F biosynthetic capacity through glycoengineering efforts that enhance the expression of these or other critical enzymes. For instance, 293-F cells also expressed relatively low levels of B3GNT7, which encodes a β1,3 N-acetylglucosaminyltransferase that is critical for elongation of the keratan sulfate backbone.

Although specific attempts to optimize recombinant protein productivity were not made, glycoprotein titers of >10 mg/L were achieved in the 293-F cultures. In this work, a bicistronic reporter of expression was included in the SLRP expression cassette to sort for cell populations with stable SLRP expression. Subsequent sorts could be attempted to isolate higher-expressing subpopulations.⁴² Alternatively, given that the SLRP expression cassette is stably integrated into the cell genome through transposition, high-expressing clones could, in principle, be isolated through clonal expansion.⁶⁷ Recombinant production of SLRPs in human cell lines is a potential alternative to natural sourcing from animal tissues. The purification of animal-sourced proteins can be challenging. Moreover, natural tissue contains a minimal amount of SLRPs, and animal-sourced proteins require multistep processes that generate many impurities and can present biocompatibility issues for in vivo applications. The application of recombinant SLRPs presents a promising and versatile strategy for modulating biological processes in a temporally, spatially, and dose controlled manner. Compared to other strategies to increase SLRPs in vivo or in vitro, such as adeno-associated viruses (AAVs) or lentiviruses, delivery of recombinant proteins offers a more controlled, tunable strategy that avoids immunogenic responses, allows for more precise dose control, and the ability to fine-tune temporal supplementation. While gene therapy offers the potential to provide sustained endogenous expression, its application in tissue engineered systems is limited by challenges in reduced transduction efficiency in 3D matrices and difficulty in supplementing these proteins temporally. As such, recombinant SLRP supplementation allows for a more controllable means to study SLRP functions without the complexities associated with genetic modification.

Supplementation of SLRPs into collagen gels revealed distinct structural changes at both the fiber (250 nm-20 μm)⁶⁸ and fibril (5–250 nm)⁶⁹ levels, as observed in SEM images at high and low magnifications (Figure 3&4). While each SLRP uniquely influences collagen fibrillogenesis, isolating their individual effects is often challenging due to the complexity of the in vivo environment.^{12,19,20,23–26} Interestingly, these SLRPs had different effects at the fiber and fibril level, a trend also reported in a previous study using low-density collagen gels supplemented with decorin.³⁰ At low magnifications, decorin-supplemented gels displayed fiber diameters similar to the control (Figure 3). The overall fiber structure resembled that of the control group, lacking a pronounced hierarchical organization and showing minimal aggregation of fibrils into larger bundles. Decorin is thought to regulate the lateral aggregation of fibers, and studies of decorin-null mice have reported more irregular collagen fiber size and shape.^12,19,20 In decorin-supplemented gels in this work, the bounds of the 95% confidence intervals for fiber diameters were 46% and 54% narrower than those in control and biglycan-supplemented gels, respectively (Figure 3). This reduction suggests that decorin promotes more uniform fiber formation, contributing to a homogeneous tissue structure at the macro level. At the fibril scale, decorin exhibited the smallest fibril diameters among the groups and maintained consistency with the fiber level with a narrower distribution (Figure 4E). Additionally, the contour-to-end ratio for decorin was comparable to the control, indicating straighter fibers with fewer bends. The porosity of decorin-supplemented gels was higher than control but lower than the other two SLRP-supplemented groups. These data suggest that decorin supports the formation of more regular and homogeneous fibril structures, which may enhance overall tissue organization.

Decorin is highly expressed in load bearing tissues, where it plays many roles in maintaining tissue homeostasis, including in fibrillogenesis, fiber maintenance, and retention of aggrecan.^70,71 This is particularly relevant in the context of osteoarthritis (OA) where the loss of matrix proteins such as aggrecan is a hallmark of disease progression. Notably, loss of decorin, but not loss of biglycan, has been shown to accelerate surface fibril remodeling and exacerbate OA progression following destabilization of the medial meniscus in mice.^72,73 These findings highlight decorin’s unique function in maintaining the superficial zone and protecting against matrix degradation. Our findings, which demonstrate that supplementing recombinant decorin in collagen gel based constructs promotes more uniform collagen fiber morphology, suggest potential utility for recombinant decorin as a therapeutic agent. In early OA, where abnormal remodeling and loss of matrix organization are common, recombinant decorin may help stabilize the collagen network and retain proteoglycans, slowing disease progression. Further studies are needed to examine the efficacy of decorin supplementation in in vivo models of OA.

In contrast to what decorin-supplemented gels showed, biglycan and fibromodulin were comparable at the fibril level, with fibril diameters falling between those of control at the high end, and decorin at the low end. Biglycan-supplemented gels in particular at low magnification displayed a denser fiber network that organized more groupings of fibrils into small fiber bundles. While these structures were not particularly organized in terms of period or alignment, there were distinct areas where fibrils aggregated together to form a more hierarchical structure. Biglycan-supplemented gels showed an averaged 2.1-fold increase in mean fiber/fibril bundle diameter compared to the other SLRP-supplemented and control gels. This 2x increase in the presence of large fibrils in biglycan-supplemented gels is notable. In addition, the shape of the PDF with a larger spread of data with multiple modes suggests that some initial base structure, represented by the primary peak, undergoes bundling and lateral aggregation, especially in this biglycan group (Supplemental Figure 3). Fibromodulin, on the other hand, displayed very little hierarchical organization and very homogeneous fiber diameters with the smallest spread of all four groups (Figure 3). This more pronounced effect at the larger scale was not as prominent in more magnified images. While the roles of biglycan and fibromodulin are less clear than those of decorin, biglycan deficient mice also show irregular formation of collagen fibers and fibromodulin is thought to promote the formation of thinner collagen fibers.^23,24 Biglycan was originally isolated from bovine bone and in fibrocartilage, biglycan and fibromodulin are found most prevalently in the parts of the tissue that experiences the most compressive load.^13,17,74 The more webbed like structure of biglycan and the thinner fibrils found in both biglycan and fibromodulin, lend themselves to the theory that these SLRPs are helping to create meshed like structures that can disperse compressive loads. Interestingly, both the contour-to-end ratio and the pore size were largest in fibromodulin-supplemented gels. Fibromodulin is thought to act on collagen by inhibiting the axial addition of triple-helical collagen monomers, leading to delayed fiber formation and the formation of thinner fibrils.²¹ In immature tendons, intermediate fibril structures are semiflexible and lack significant orientation, allowing them to reorganize and adapt as the tissue matures.²¹ This flexibility coincides with the expression of fibromodulin during tendon maturation,²¹ suggesting that fibromodulin’s inhibition of axial addition of collagen monomers may facilitate the formation of flexible precursor structures. These intermediates may be crucial for adapting to the mechanical demands of the maturing tissue as it begins to bear load.

Connective tissues can possess zonal architectures designed to accommodate complex loading environments. When designing a tissue engineered material, evaluating the tensile properties is critical. Decorin-supplemented gels demonstrated a 2.61-fold increase in toughness compared to the control gels (p = 0.001). Decorin exhibited a 1.7-fold increase in toughness compared to biglycan (p = 0.04) and although the difference between decorin and fibromodulin was not statistically significant (p = 0.17), there was a 1.47-fold increase in the toughness of decorin-supplemented gels compared to fibromodulin gels. Among the SLRP-supplemented gels, biglycan led to a statistically significant increase in UTS compared to controls (1.59-fold, p = 0.03), while decorin and fibromodulin showed nonsignificant trends toward higher UTS (1.51- and 1.42-fold, p = 0.09 for both). High tensile strength and toughness largely depend on the ability of the collagen fiber network to align with the direction of load.⁷⁵ Although decorin-supplemented gels did not show larger fiber or fibril diameters compared to the other SLRP groups or the control, the overall organization of the decorin gels revealed regular and homogeneous fibril structures. This improved organization likely enhanced tissue alignment and response to load. In addition, decorin samples displayed the lowest porosity of the SLRP-supplemented gels. While porosity and fiber/fibril diameter size are structural features that may influence mechanical performance, these results suggest they may not be the primary driver of the enhanced mechanical properties of these decorin supplemented gels. Additional mechanisms, such as enhanced interfibrillar interactions, SLRP-mediated cross-linking, or ability of the fibers to align under load, may contribute to improved mechanical outcomes. SEM images after tensile testing revealed fibers that were highly aligned with the direction of load, with some fibrillation and tearing, but larger groups of fibrils oriented in the same direction. This reorganization of decorin supplemented fibers under load likely contributed to their increased ability to withstand tensile forces. Decorin displayed a 1.9 (p = 0.01), 1.2-, and 1.4-fold increase in strain at failure compared to control, biglycan, and fibromodulin-supplemented gels, respectively. Other work in gels with decorin supplementation displayed higher strain at failure compared to controls and were stiffer and more tough.⁷⁶ This data shows that decorin-supplemented gels were able to be strained more than the other groups, which allowed further reorganization of collagen fibrils during tensile testing. Notably, decorin is the predominant SLRP found in skin, a tissue which has high toughness and tensile strength.⁷⁷ In addition, decorin null mice exhibited reduced tensile strength, further underscoring decorin’s critical role in enhancing this property.¹²

Fibromodulin and biglycan exhibited similar properties in ultimate load, UTS, and toughness, with biglycan showing a slightly higher but nonsignificant Young’s modulus. Fibromodulin exhibited a 1.13-fold increase in strain at failure compared to biglycan-supplemented gels, but again this was nonsignificant. At a larger scale, the fiber structures of the biglycan- and fibromodulin-supplemented gels appeared quite different; however, their fibril properties and weighed pore sizes were more comparable. The primary distinction at higher magnifications was the increased undulation of fibromodulin fibers, as shown by the contour-to-end ratios, compared to biglycan. Higher magnification SEM micrographs post-tensile testing revealed similar features for both groups, including frayed edges and minimal realignment, consistent with their reduced toughness and strain at failure relative to decorin. At the fiber level, neither group demonstrated significant global alignment, and both remained qualitatively unorganized with frayed collagen fibers. This reduced toughness, coupled with fiber structures resembling the compressive zones of the meniscus, suggest that biglycan and fibromodulin-supplemented gels may perform better under compressive loads.^13,74 This work primarily focused on the tensile properties of these tissues, leaving compressive properties, such as aggregate modulus, unexplored.

In this study, we investigated the supplementation of SLRPs at only a single concentration. The selected dose of 5 nM is on the lower end of concentrations reported in the literature.^30,76 As a result, it is unclear whether we observed the maximum potential effect of these SLRPs on fibrillogenesis in the gels. Notably, the data in this study suggests that SLRPs have potent modulatory effects on collagen self-assembly and subsequent mechanics, even at subphysiologic levels. Since the binding constants for each individual SLRP may vary, a dose−response study for all three SLRPs is warranted. The contour-to-end ratio data indicate a rightward shift in the fibromodulin-supplemented gels, while decorin and biglycan show results comparable to the control, albeit with some fluctuations at higher ratios. Exploring higher doses of these SLRPs, particularly decorin and biglycan, may reveal additional effects that are not apparent at the lower doses used in this study. In addition, with the ability to produce recombinant SLRPs in larger quantities more easily than tissue extraction methods, future studies will be able to explore native tissue SLRP supplementation levels.

The differences in structural organization and mechanical performance observed with the supplementation of decorin, biglycan, and fibromodulin, may have important implications for tissue-specific scaffold design. The enhanced toughness and more uniform collagen distribution found in decorin-supplemented gels suggests potential suitability for load-bearing tissues such as tendon and ligaments, which require the ability to withstand higher tensile loads. In contrast, the presence of thicker, more bundled structures in the biglycan group may better mimic the collagen architecture found in cartilage and fibrocartilage, where such features contribute to resistance under compressive loading. These findings suggest that incorporation of specific SLRPs and possibly a combination of them could be used to tune scaffold properties to better match the functional demands of different target tissues.

CONCLUSIONS

In conclusion, decorin, biglycan, and fibromodulin were successfully produced via recombinant expression, demonstrating the feasibility of using HEK293-F cells to generate small leucine-rich proteoglycans in sufficient quantities for in vitro studies. SEM images showed that all SLRPs facilitated the formation of unique and distinct collagen structures at both the fiber and fibril levels. Decorin-supplemented gels showed homogeneous fibril structures which lent themselves to being able to rearrange parallel to the direction of load, leading to increased toughness in these gels. The more webbed fibrils found in biglycan and thinner diameters found in fibromodulin and biglycan gels are consistent with the fact that these SLRPs are often found in zones of fibrocartilage tissues, that undergo compression, for example, where this distinct architecture can facilitate the distribution of compressive loads. Distinct collagen architectural changes occur in constructs supplemented with SLRPs, resulting in differences in their response to tensile loads. These findings suggest that supplementing high-density collagen gels with small leucine rich proteoglycans during polymerization and early collagen fibrillogenesis promotes the formation of distinct, tunable fiber networks, enhancing their ability to respond to mechanical loads. Further investigation into the role of SLRPs in in vitro systems is essential to facilitate the development of tissue engineered constructs with zonally dependent architectures. Optimizing collagen fiber formation and mechanical properties at the zonal level will be critical to replicating the varying loads experienced throughout the variety of connective tissues.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsbiomaterials.5c00732.

Supporting Information Figure 1: pathways for heparan sulfate modification; Supporting Information Figure 2: pathways for biosynthesis of keratan sulfate; Supporting Information Figure 3: probability density functions of fiber diameters overlaid with Gaussian fits centered on the primary peak; Supporting Information Figure 4: probability density functions of contour-to-end ratios overlaid with Gaussian fits centered on the primary peak (PDF)