Proteoglycans and proteoglycan mimetics for tissue engineering

Abstract

Proteoglycans play a crucial role in proper tissue morphology and function throughout the body that is defined by a combination of their core protein and the attached glycosaminoglycan chains. Although they serve a myriad of roles, the functions of extracellular proteoglycans can be generally sorted into four categories: modulation of tissue mechanical properties, regulation and protection of the extracellular matrix, sequestering of proteins, and regulation of cell signaling. The loss of proteoglycans can result in significant tissue dysfunction, ranging from poor mechanical properties to uncontrolled inflammation. Because of the key roles they play in proper tissue function and due to their complex synthesis, the past two decades have seen significant research into the development of proteoglycan mimetic molecules to recapitulate the function of proteoglycans for therapeutic and tissue engineering applications. These strategies have ranged from semisynthetic graft copolymers to recombinant proteoglycan domains synthesized by genetically engineered cells. In this review, we highlight some of the important functions of extracellular proteoglycans, as well as the strategies developed to recapitulate these functions.

INTRODUCTION

Proteoglycans (PGs) are a ubiquitous class of biomolecules characterized by a core protein decorated with one or more glycosaminoglycan (GAG) chains. GAGs are separated into two categories, sulfated and nonsulfated GAGs. Sulfated GAGs (sGAGs) include chondroitin sulfate (CS), dermatan sulfate (DS), heparan sulfate (HS), keratin sulfate (KS), and heparin, with these sGAGs primarily found in vivo as part of PGs. In contrast, the nonsulfated GAG hyaluronic acid (HA) is not attached to a core protein; however, HA is a major component of the extracellular matrix (ECM) (1).

PGs exist as both intracellular/cell-membrane-bound and extracellular molecules, although the majority are extracellular. Extracellular PGs can be further divided into several categories: modular HA-binding PGs, modular PGs that do not bind to HA, and small leucine-rich PGs (SLRPs) (1, 2). Modular HA-binding PGs, known as hyalectans, contain core proteins composed of three domains: an HA-binding domain, a central domain to which the GAG chains are primarily attached, and a lectin interacting domain. This central domain can have between three and one hundred GAG chains attached, depending on the PG (1). In contrast, modular non-HA-binding PGs are more varied in their form and reside primarily in the basement membrane of tissues (1). The smaller SLRPs all contain leucine-rich repeat units and are divided into five classes, three canonical and two noncanonical, based on their genetic lineage and protein homology. Eight of the noncanonical SLRPs do not contain GAGs, but are included in the classification due to their structural and functional similarity to GAG-containing SLRPs (2).

Proteoglycan research is still a relatively nascent area of study within the field of tissue engineering and most of the focus has been on extracellular PGs to mimic the environment cells would experience in native tissue. As such, this review will also primarily focus on extracellular PGs and efforts to recapitulate their function. PG form, both intracellular and extracellular, has been extensively reviewed by other authors. For further reading, we direct the reader to reviews by Iozzo et al. (2) and Schaefer et al. (1).

FUNCTIONS OF PROTEOGLYCANS

Although the structure of PGs varies greatly depending on the size of the core protein and the type and number of conjugated glycans, PG functions can be generalized into four categories: modulation of tissue mechanical properties, structural regulation and protection of the ECM, immobilization of growth factors, and mediation of cellular signaling and interactions (Fig. 1). Although PGs have a myriad of functions throughout the body, this section will focus on a select number of examples relevant to the field of tissue engineering.

Figure 1.

Categories of proteoglycan functions in the ECM (created with Biorender.com with permission). ECM, extracellular matrix.

Modulation of Tissue Mechanical Properties

Because of the negative charge conferred by sulfate and carboxylate groups, GAGs are more hydrophilic than most other ECM components and promote water intake into the tissue. Water uptake is prevalent in cartilage, where the proteoglycan aggrecan is found in large amounts (3, 4). Aggrecan comprises a core protein grafted with ∼60 KS chains and ∼100 CS chains. Aggrecan binds to HA through a link protein, immobilizing it in the cartilage ECM and forming large negatively charged aggregates, which create an osmotic gradient that supports water uptake and substantial tissue compressive stiffness (3).

One aspect of degenerative cartilage diseases, such as osteoarthritis (OA), is the loss of the articular surface due to uncontrolled inflammation and overexpression of catabolic enzymes including matrix metalloproteases (MMPs), hyaluronidases, and aggrecanases (5). Aggrecanases cleave the aggrecan core protein near the linkage between it and HA, severing it from the network and allowing it to diffuse out into the synovial fluid (6). This loss of aggrecan results in a reduction in the cartilage’s ability to retain water and is characteristic of early-stage osteoarthritis. Aggrecan loss further exposes other ECM components to hyaluronidases and MMPs facilitating lass of cartilage (5, 6).

The contributions of aggrecan to the biology and mechanical properties of cartilage has been extensively reviewed by others. For more information, we direct the reader to these reviews (3, 7).

Regulation and Protection of the ECM

The major component of the ECM is fibrillar collagen, which governs the tissue structure and provides support for cells. Proper regulation of collagen fibril formation and organization is necessary for the correct assembly of the ECM and for the function of the tissue. For example, SLRPs, including decorin, biglycan, fibromodulin, and lumican play a large role in ECM assembly (8). During collagen fibril formation, the core protein of the SLRPs bind to the collagen to modulate fibril growth (9). Although the core protein is primarily responsible for the PG-collagen interaction, the attached GAGs interact with adjacent collagen fibrils via charge interactions (10). This network of carefully organized collagen fibrils and proteoglycans contribute to ECM structure and properties such as hydration and mechanical stability (11, 12).

In addition to regulation of collagen formation and organization, SLRPs also protect collagen from proteolytic degradation. By decorating the exterior of collagen fibrils, SLRPs block proteases from interacting with collagen and prevent degradation (13). Aside from SLRPs, other proteoglycans have been shown to protect against matrix proteolytic degradation. Aggrecan, with its large bottlebrush structure and excluded volume, can limit MMP access to collagen II and protect against its degradation (14). In cases of degenerative diseases where proteases degrade PGs, the newly unprotected collagen will be subject to degradation, resulting in loss of the network structure and a reduction in the mechanical strength of the matrix (15).

For further reading regarding PG’s effects on matrix structure, the reader is directed toward reviews by Chen et al. (8) and Pang et al. (16).

Sequestering of Proteins

Aside from interacting with matrix components, another role of the sGAGs is binding to and modifying the biological activity of a wide range of proteins, including proteases, chemokines, and growth factors. This is often mediated through a cationic domain on the protein, which binds to the anionic GAGs. Though the specific amino acids of the cationic domain will vary between proteins, they often take the form of clusters of basic residues (e.g., arginine and lysine) flanked by one or two nonbasic residues (17–19). Though this interaction commonly occurs between proteins and heparin or HS PGs, the cationic domain also allows for binding to other sGAGs (20). The sulfation pattern of the GAG can also have an effect on protein binding to GAGs. For example, antithrombin III is a heparin-binding protein that binds primarily with a specific pentasaccharide sulfation pattern found on a subset of heparin molecules (21). Similarly, it has been found that 6-O-sulfation promotes the binding of fibroblast growth factor (FGF) to HS, which regulates the FGF signaling in the body (22).

Binding to GAGs immobilizes the protein, thus constraining and regulating its biological activity. For example, perlecan, a HS PG found in basement membranes throughout the body, binds to many growth factors, including the those from the FGF (23), the platelet-derived growth factor (PDGF), and vascular endothelial growth factor (VEGF) families (24). Through growth factor binding, perlecan establishes growth factor gradients necessary for the proper formation of new tissue. This was demonstrated in knockout models for perlecan, where the absence of perlecan resulted in tissue abnormalities including defects in endochondral ossification and cardiovascular development (25–27). The removal of perlecan resulted in improperly formed tissues and the premature death of the embryo, highlighting the importance of this PG for proper organ function.

Further reading regarding the binding of proteins to PGs can be found in the reviews by Whitelock et al. (24) and Muñoz et al. (17).

Regulation of Cell Signaling

PG-rich layers, most notably the endothelial glycocalyx found in the vasculature, can prevent cell-cell interactions from occurring through steric hindrance and the anionic charge of the PG GAG chains (28, 29). This prevents direct cell interactions between the endothelial cells lining the vessels and the circulating cells in the blood, including erythrocytes, platelets, and leukocytes. In cases where the glycocalyx is disrupted, either through injury or enzymatic degradation of the GAGs, endothelial surface proteins such as selectins, ICAM, and VCAM become exposed (30). This allows for leukocyte capture, activation, arrest, and migration into the surrounding tissue. Under normal conditions, this is necessary for the proper vessel healing, but poor reconstruction of the glycocalyx can result in uninhibited leukocyte activation, leading to uncontrolled inflammation at the wound site. Further reading about the glycocalyx can be found in the review by Reitsma et al. (28).

BIOMIMETIC STRATEGIES TO REPLACE PROTEOGLYCANS

Given then ubiquity and importance of the contributions of PGs to proper tissue function, researchers have sought to develop methods to recapitulate these functions. For mimicking PG function, many groups have developed strategies ranging from the synthesis of semi-synthetic PG mimetics to the use of recombinant PG domains (Fig. 2). In this section, we will highlight some of the strategies used to recapitulate the four general functions of PGs.

Figure 2.

Biomimetic strategies to recapitulate proteoglycan function (created with Biorender.com with permission).

Glycosaminoglycan Bottlebrush Graft Copolymers

One approach to recapitulate function involves the use of graft copolymers to mimic the bottlebrush-like structure of hyalectans including aggrecan. An aggrecan mimetic designed by grafting amine-terminal CS to a synthetic poly(acryloyl chloride) backbone (31, 32) exemplifies this approach (Fig. 3A). Using this approach, both large and small proteoglycan mimetics, ranging from a 10 kDa polyacrylate core with ∼7–8 CS chains attached (32) to a 250 kDa core polymer with ∼60 CS chains attached were synthesized (31). The molecule swelling compared favorably to that of aggrecan and unconjugated CS (32). In addition to swelling, these bottlebrush polymers regulated collagen fibril formation when mixed into a collagen gel, as do SLRPs (33). In contrast to SLRPs, the regulation of fibril formation was attributed to CS structure, since there was no change in collagen fibril morphology in the presence of negatively charged poly(acrylic acid). The researchers hypothesized that the interactions between CS, either in its free-floating or PG form, and collagen modulated the kinetics of fibril formation, leading to the observed changes in diameter and band spacing. Because this strategy involves the use of biomimetic molecules, rather than a bulk material, the bottlebrush copolymers can be directly injected to the site of injury (31).

Figure 3.

Graft-copolymer proteoglycan biomimetic strategies. A: CS grafted to poly(acryloyl chloride) backbone. Adapted from Prudnikova et al. with permission (32). B: CS or heparin grafted to HA backbone. Adapted from Place et al. with permission. (34). C: HA-binding peptides (GAH) grafted to oxidized CS backbone. Reprinted from Bernhard and Panitch with permission (36). CS, chondroitin sulfate; HA, hyaluronic acid.

Another approach to mimicking aggrecan function involves a graft copolymer composed of CS grafted to an HA backbone (Fig. 3B) (34, 35). Molecule synthesis was achieved by first converting the HA carboxylic acid groups to hydrazides, then conjugating the graft polymer following reductive amination. The addition of the CS graft copolymer increased the modulus of an agarose hydrogel, though the addition of free CS had a similar effect (35). It was hypothesized that the two additives affected the mechanical properties by different mechanisms, with free CS increasing the modulus through chain entanglement and the CS graft copolymer increasing the modulus through osmotic pressure. However, despite being entirely composed of biopolymers, inclusion of the CS graft copolymer in agarose gels resulted in decreased cell viability, demonstrating the need for further optimization of the mimetic (35).

Although these mimetic strategies recapitulate the brush structure of hyalectan PGs through the grafting of GAGs to a core polymer, mimicking the targeted bioactive properties of the PGs, the functions of the core protein are not included. Most notably, the brush PGs are not designed to bind to HA to prolong localization in a tissue, resulting in the eventual diffusion of these molecules from their target area. From a therapeutic perspective this may increase the dosing frequency required to maintain function over time. However, the substitution of the protein core with another polymer results in resistance to proteolytic degradation, which may outweigh the downsides of the loss of protein core functionality. Future work is needed to better tease out this balance between functional mimicry and therapeutic efficacy.

Peptide-Based Mimetic Strategies

An alternative strategy to synthesize functional mimetics involves grafting substrate-binding peptides to polymers. These peptides often either bind to HA or collagen and aim to mimic the binding domains of hyalectans and SLRPs, respectively. By grafting these peptides to polymers, the ECM organization or protection provided by PGs can be recapitulated.

One example of this approach includes the peptide-glycan molecules consisting of ECM-binding peptides grafted to a GAG backbone (Fig. 3C). Although PGs often have a single ECM-binding domain, mimetics were generated by grafting 5–15 peptides onto the GAG backbone, taking advantage of avidity to enhance target binding. This approach supported the protection of the ECM from proteolytic degradation as demonstrated using an aggrecan mimetic, made with HA-binding peptides grafted to a CS backbone (36–39). By adding both HA-binding and collagen II-binding peptides to a CS backbone, a lubricin mimetic was created which lowered the coefficient of friction when applied to the surface of articular cartilage (40, 41). Similarly, a decorin mimetic (collagen-binding peptides grafted to a DS backbone) modulated collagen fibril formation and protected against proteolytic degradation (42, 43). In addition to ECM regulation, this approach was used to restore the glycocalyx following vascular endothelial denudation by using the collagen-binding decorin mimetic (44) and to protect inflamed endothelium by using DS grafted with selectin-binding peptides (45), both of which localized mimetics to the injured vessel and recapitulated the GAG barrier by blocking vessel interactions with circulating platelets and neutrophils.

Building on earlier work demonstrating lubricin mimetics composed of hyaluronic acid- and collagen-binding peptides grafted to CS could reduce friction at the articular cartilage surface (41, 46), a lubricin mimetic consisting of HA-binding peptides and collagen-binding peptides grafted to poly(ethylene glycol) (PEG) was generated (46, 47). Following intra-articular injection of the PEG-based lubricin mimetic in a model of OA, this molecule bound to both exposed collagen II and HA, tethering HA to the surface of the cartilage and decreasing the joint friction (46). Treatment with this molecule led to reduced progression of OA in a mouse model, perhaps due to the increased localization of HA to the cartilage surface (47).

Although the structure of these molecules differs from the PGs they aim to mimic, the use of substrate-binding peptides allows for these molecules to mimic the function of the binding domains present on the cores of PGs. This promotes local retention, extending the time of their therapeutic effect. However, their smaller size compared with the PGs they aim to mimic can limit their ability to fully recapitulate PG function, and the use of peptides makes them to susceptible to degradation by peptidases.

Recombinant Protein Domains

Another approach to recapitulate the form of PGs while maintaining the ability to tune their properties is using recombinant PG domains. This is commonly done with the terminal recombinant perlecan domains (rPlnD), where the HS chains are natively attached (24). These rPlnDs can be obtained through transfecting mammalian cells, resulting in the synthesis of the core protein and the attachment of the HS chains (48, 49). Using this approach, several groups have immobilized the rPlnDs to various substrates to control the binding, presentation, and release of heparin binding GFs.

Using recombinant perlecan domain I (rPlnDI), the Farach-Carson group demonstrated the sustained release of bone morphogenic protein 2 (BMP-2) from a poly(ϵ-caprolactone) (PCL) electrospun scaffold (48). This was achieved by first covalently immobilizing the rPlnDI to the scaffold fibers, then loading the scaffold with BMP-2. The immobilization of rPnlDI not only increased the amount of BMP-2 loaded into the scaffold but released the BMP-2 over a longer period of time compared with a control PCL scaffold. The group also used a custom 3D-printed microfluidic system to generate gradients of rPlnDI, which in turn were used to create gradients of FGF-2 within the hydrogel (49). The gradient of FGF-2 led to an increase in cell migration compared with hydrogels where the FGF-2 was uniformly distributed throughout the gel. In addition to cell scaffolds, rPnlDI has also been immobilized to microparticles for controlled delivery of growth factors such as bone morphogenetic protein 2 (BMP-2), providing better release kinetics over free GF delivery (50).

Recombinant domain V on the C-terminus of perlecan has also been studied. One reported form encompassing the domain from Glu3687 to Ser4391, referred to as endorepellin, has been reported to be a strong anti-angiogenic molecule. However, a large form encompassing the amino acids Leu3626 to Ser4391 (rPlnDV) has been reported to promote angiogenesis (51). Although both possess the α2β1 integrin binding site, one significant difference between these two reported recombinant domains is the latter contains a HS or CS chain attached to it, whereas endorepellin does not, demonstrating the importance of the GAG chain to the potentiation of angiogenic growth factors (51). Lin et al. tested the larger, GAG bound molecule both in its soluble form and immobilized to silk fibroin scaffolds, where the presence of the mimetic promoted angiogenesis through the potentiation of available growth factors (52). Evidence that the GAG chains of rPlnDV were necessary for GF potentiation came from the removal of the GAG chains, which resulted in no proangiogenic signals to the cells.

The use of recombinant PG domains results in molecules closest in form and function to their respective PGs, as they contain both the appropriate GAG chains and portions of the core protein. However, this comes at the cost of being the most complex mimetic method, short of the use of full recombinant PGs, due to the need for genetically engineered cells to synthesize the molecules. Advances in genetic engineering and mammalian cell culture coupled with streamlined isolation methods will make these forms more competitive in the future.

GAG Functionalized Materials

To recapitulate the GF-binding ability of PGs, many groups have opted to incorporate sGAGs into their biomaterials, rather than synthesizing specialized molecules. This is commonly done with heparin (53–57) and, to a lesser extent, CS (58–60), given their commercial availability. In addition, sulfated HA has also been used to sequester growth factors, allowing for sulfation patterns similar to heparin while also allowing for control over degree of sulfation, thus decreasing the risk of antithrombotic activity inherent in heparin (61, 62). By incorporating these polymers into materials, GFs including FGF-2 (59), VEGF (54, 57), transforming growth factor β (TGF-β) (56, 60), BMP-2 (53, 54), and nerve growth factor (NGF) (63) have been sequestered and released over time or presented to nearby cells. For example, Levinson et al. developed a HA/heparin hydrogel which was loaded with TGF-β and chondrocytes. In this experiment, one hydrogel was made with HA crosslinked to heparin and the other having uncrosslinked heparin. Both gels were loaded with TGF-β, with the crosslinked heparin gel demonstrating sustained release and the uncrosslinked heparin gel having a larger burst release of the GF. This resulted in improved collagen II and GAG deposition by the encapsulated cells, comparable to culture in TGF-β supplemented media (56). Similarly, Kim et al. functionalized a gelatin cryogel with heparin, which was used to sequester VEGF for the treatment of ischemic wounds (57). Increasing the concentration of heparin incorporated increased the amount of VEGF retained overtime. NIH-3T3 fibroblasts and VEGF-loaded cryogels were then implanted in a rat ischemic hindlimb model, where the construct yielded improved angiogenesis over cell only and VEGF only loaded gels, demonstrating the effect of sequestered GFs on tissue engineering outcomes. Similar trends were seen by other groups, where the sequestration of GFs through binding to sGAGs resulted in the engineering of more robust tissue.

Finally, incorporating GAGs can allow for the spatial patterning of GFs through methods including controlled deposition (63) and 3D printing (55). By focusing on only the GAG component of PGs, this method is less complex compared with other PG mimetic strategies while still recapitulating certain functions of PGs. However, this simplicity comes at the cost of not replicating the form of PGs or the functionality of the core protein. As such, this method only recapitulates the functions PGs resulting from the biochemical activity of their GAG chains.

CONCLUSIONS AND PERSPECTIVES

Current PG mimetic strategies have largely focused on ECM PGs, and to a lesser extent membrane anchored PGs, to recapitulate contributions of PGs to the cellular environment that have not traditionally been captured in artificial scaffolds. Because PGs exhibit structural variety and serve a myriad of functions, PG mimetic strategies have also come in various forms, with almost all approaches incorporating sGAGs in some manner. However, given the complex structure of proteoglycans, no mimic replaces all the functions of the PG it mimics. Although bottlebrush copolymers can modulate swelling through their charged side chains, these approaches lack the HA binding sites of hyalectans. Conversely, approaches involving substrate-binding peptides can interact with the ECM in a manner similar to PGs, but these approaches do not fully mimic the structure of PGs. Functionalizing materials with GAGs supports GAG presentation to cells, but this approach is inherently different to how GAGs would be presented to cells natively. The closest approach to mimicking native PGs is exhibited by the use of recombinant PG domains, but the use of single domains omits functions present in other domains. In addition, recombinant technology is currently the most difficult to scale up due to the complexity of PG biosynthesis, which requires genetically engineered cells.

To be effective, the PG mimetic does not need to fully recapitulate all the functions of a given PG. Given the complexity of PGs, to perfectly mimic their form and function would be prohibitively difficult, short of using fully recombinant PGs. Instead, each mimetic approach presented here identified and tailored molecules with a specific function. As such, a one-size-fits-all solution to mimicking PGs may not be practical, with future approaches possibly utilizing a combination of specialized molecules. The development of PG mimetics continues to be a critical area of research, with current and new technologies allowing for better recapitulation of PG function and the engineering of more robust tissue.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grant T32 HL086350.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

This article is part of the special collection “Deciphering the Role of Proteoglycans and Glycosaminoglycans in Health and Disease.” Dr. Liliana Schaefer served as the guest editor of this collection.

AUTHOR CONTRIBUTIONS

M.N. prepared figures; M.N. and A.P. drafted manuscript; M.N. and A.P. edited and revised manuscript; M.N. and A.P. approved final version of manuscript.