Bioengineered collagens

Abstract

Mammalian collagen has been widely used as a biomedical material. Nevertheless, there are still concerns about the variability between preparations, particularly with the possibility that the products may transmit animal-based diseases. Many groups have examined the possible application of bioengineered mammalian collagens. However, translating laboratory studies into large-scale manufacturing has often proved difficult, although certain yeast and plant systems seem effective. Production of full-length mammalian collagens, with the required secondary modification to give proline hydroxylation, has proved difficult in E. coli. However, recently, a new group of collagens, which have the characteristic triple helical structure of collagen, has been identified in bacteria. These proteins are stable without the need for hydroxyproline and are able to be produced and purified from E. coli in high yield. Initial studies indicate that they would be suitable for biomedical applications.

Recombinant Collagen: Establishing Commercial Availability

Collagen is the most abundant protein in mammals, where it plays critical roles in the structure and in molecular and cellular interactions in the extracellular matrix. For example, collagen is the major protein of many tissues, including skin, bone, ligament, tendon, and cartilage. Collagens are characterized by a unique molecular structure, the collagen triple-helix. This structure consists of a supercoiled triple-helix, which is made from three left-handed polyproline-like chains twisted together into a right-handed triple-helix.¹ A further, interesting feature of this structure is that the tight packing of the triple helix requires that every third residue in the primary sequence be Gly, because there is no space for any larger amino acid in the interior axis of the triple-helix. As a consequence, a second characteristic feature of collagens is the repetitive amino acid sequence pattern (Gly-Xaa-Yaa)_n.¹ A further feature, is that the non-Gly positions are often occupied by the imino acid proline. For animal collagens, Pro residues in the Yaa position, about 10% of all residues, are normally post-translationally modified by the enzyme prolyl-4-hydroxylase (P4H), to give 4-hydroxyproline (Hyp). This modification is very important for several reasons, particularly as an absolute requirement for good thermal stability of the triple-helix, but also to assist in self association, especially in the fibrillar collagens, and participation in certain receptor interactions.^1,2 Although collagen is used as a single all-embracing term, in humans and other mammals, there are at least 28 genetically distinct types,² all of which exemplify these various structural characteristics. The most abundant collagens are the interstitial, fibril forming collagens, particularly type I collagen, which are present in all the major connective tissues (Fig. 1). Type I collagen, which can be readily obtained and purified, has become the major collagen used in biomedical and tissue engineering applications.

Figure 1. A transmission electron micrograph of the collagen fibrils in a sheep anterior cruciate ligament, showing numerous collagen fibrils, all of similar diameter. Bacterial collagens do not form such fibrillar structures. Bar = 0.25 µm.

Although there are some early examples, it is only in the past 100 years that a range of collagen-based biomedical materials have emerged, starting from the use of human amnion in 1912, followed by collagen based dressings in the 1940s,³ with many subsequent products.⁴ Most recently, there has been a growing interest in using natural extracellular matrix-based, particularly collagen, supporting structures (scaffolds) in tissue engineering developments.⁵

In all these cases of medical applications, there are a number of concerns that have been expressed. These include, for example, the natural variability in preparations from animal tissues, where purity and predictability of performance are issues. Of more concern has been the possibility of transmission of animal based diseases, especially bovine spongiform encephalopathy (“mad-cow disease”). There have been two suggested approaches to this disease issue. One approach has been the examination of non-mammalian species, including jellyfish, fish, and chicken as material sources,⁶ while the other approach has been the production of collagen by recombinant technology.⁷ Recombinant approaches enable production of a disease-free product that is of uniform product quality. The approach also could allow for the production of less common collagen types, where extraction from tissue is not practical. There is also the opportunity to modify and improve on the natural protein structures by introduction of mutated residues, selection of specific domains, engineering products with multiple repetitive motifs, and development of chimeric structures.⁷

The potential disadvantage of recombinant technology, however, is that recombinant expression of type I collagen, or any mammalian collagen, will almost certainly also need the co- expression of functional P4H to enhance the triple-helical stability. The P4H enzyme consists of two different chain types, each with a molecular mass of around 65 kDa. If the stability of the collagen triple-helix is low, then the product is the denatured form of collagen, gelatin. Gelatin consists of only single chains that are not folded into in the native triple-helical structure. Also, while recombinant technology can be used easily for small scale production of laboratory samples, for commercial scale production there will be manufacturing issues that will need to be solved and process economics evaluated.⁸

Initial biotechnology developments focused on the production of full-length human collagens, including the heterotrimeric type I collagen and the homotrimeric type III collagen.⁷ Mammalian cell systems, such as HT1080 and HEK 293-EBNA cells, have been used, but yields are very low. The same is also true for production in an insect cell system, Baculovirus. Both systems, however, do have endogenous P4H activity (Table 1). Microorganisms provide alternative expression systems and are a natural choice because of their familiarity in the biotechnology industry.⁹ Unfortunately, the use of E. coli expression was initially not seen as practical, as there seemed to be difficulties in co-expressing all the necessary genes.⁷ On the other hand, yeast systems were considerably more promising, and several have been used successfully for the expression of full-length hydroxylated human collagens (Table 1).^10-12 The advantage of the yeasts is they can be optimized to integrate and co-express 3 or 4 genes concurrently. The most successful application, which has provided commercial opportunities, has emerged from expression in P. pastoris (Table 1).¹¹ Yeasts, as with other microorganisms typically require different codon preferences to those in mammalian systems for optimal expression. The development of rapid gene synthesis enables effective codon optimized sequences as well as specific variants of these sequences to be readily produced resulting in giving good yield. Other approaches to improve commercial yields are also possible and include modification of the propeptide domains, enhanced integration, increased copy number of introduced genes, and reduced production temperature and oxygen enrichment to increase the P4H activity.

Table 1. Examples of recombinant systems for human collagen production *.

Production systems	Yield	Comments
Mammalian cells:
HT1080	≤1 mg/L	Mainly type II collagen and variants
HEK293	≤80 mg/L	Used to produce collagen types V, VII, VIII, X, XVI
CHO	<0.5 mg/L	Type IV collagen
COS-1	<0.5 mg/L	Type III and XVII collagens
Insect cells
Baculovirus	≤40 mg/L	Best yield for homotrimer, type III collagen and less heterotrimer type I collagen
Bacterial cells
E. coli	≤14 g/L	No hydroxylation
Yeast cells:
S. cerevisiae	≤0.4% protein	Incomplete hydroxylation
P. pastoris	≤1.5 g/L	Fully hydroxylated. Commercially available
H. polymorpha	≤0.6 g/L	For a 14 kDa fragment. No hydroxylation
Transgenic animals
Mouse (milk)	≤8 g/L	For a 37kb fragment. Much lower yields for full-length constructs
Silkworm (Bombyx mori) (cocoon)	≤4.2 mg per cocoon	Probably not triple-helical, but rather as gelatin
Transgenic plants
Plant seeds (Barley, Rice, Maize)	≤140 mg/Kg of seed (barley)	For a 45 kDa fragment. Considerably less for full-length collagen. Low hydroxylation
Plant leaves (Tobacco)	≤20 g/L of soluble protein−	Yield can depend on hydroxylation status

*Compiled from data presented previously.⁷

Other systems have also emerged, based on transgenic organisms. For collagens, the mammary gland has been used for collagen production in mouse milk (Table 1).¹³ Another, more unusual system that is suited for fibrous proteins has been production of recombinant human type III procollagen in cocoons of transgenic Bombyx mori silkworms.¹⁴ Collagen production in transgenic plants has also been developed and seems to provide cost-effective, commercial outcomes. Initial reports produced unhydroxylated or poorly hydroxylated collagen from leaves¹⁵ and from seeds.¹⁶ Additional introduction of the mammalian P4H is necessary for a quality collagen product, even though plants contain a form of P4H. Most recently, heterotrimeric collagen has been produced with P4H also present, giving excellent protein yields, which can be commercialized.¹⁷

There may be further advantages of recombinant systems that possibly work only at a laboratory level. For example a range of constructs have been made for type II collagen, where specific “D-period” repeats, about 234 amino acids each, have been removed or substituted into the structure, allowing a better understanding of collagen structure and function.¹⁸ Another variation to this approach has been the design and production of small collagen fragments with embedded multiple repetitive domains (integrin) within the collagen chain.¹⁹

Recombinant Bacterial Collagen: An Emerging System

Collagens were seen historically as being associated with multicellular animal tissues. In more recent times the suggestion of collagen-like molecules in other species has emerged, with (Gly-Xaa-Yaa)_n repeating sequences present in a few fungi,²⁰ in viruses,²¹ and in phage.²² Most interesting, however, are the numerous collagen-like structures that are being identified in bacteria and in the emerging, large numbers of bacterial genomes. Thus, a gene with a collagen-like repeat, that is a virulence factor, was identified and characterized in the bacterium Streptococcus pyogenes.²³ Subsequently, a second, structurally distinct, collagen-like gene was identified from the same species.²⁴ Both were shown by biochemical and biophysical studies to have the characteristic collagen triple-helical structure.²⁵

The surprise that emerged was that these collagens were both stable around mammalian body temperature, with melting temperatures of 36.4 °C and 37.6 °C, without the presence of any hydroxyproline.²⁵

A decade ago, Rasmussen and colleagues reported an analysis of the available bacterial genome databases, searching for collagen-related structural motifs (CSM).²⁶ The study included 137 eubacterial genomes and 15 archaebacterial genomes and looked for sequences with homology to (Gly-Pro-Pro)_n where n was 7 or more. This identified 53 proteins in 25 bacterial genomes, with none of these in any of the 15 archaebacterial genomes. Several of the bacterial genomes contained several CSM, up to 9 in some cases in this analysis. These motifs showed a wide range in size, with a mean length of 76 Gly-Xaa-Yaa triplets, and the collagen-like sequences are always flanked by non-collagenous domains.²⁶ Although all show collagen like sequence motifs, it is very unlikely that they will all contain a stable collagen triple-helix, especially for those where as few as only 7 repeats are present. It has been suggested that a larger number of repeats, perhaps 35 or more, may be required for triple helix formation and stability.²⁷ Also, while the database search has identified many putative sequences in various genomes, the evidence for natural expression is limited to relatively few.²⁸ A more detailed analysis of amino acid compositions suggested different groups of collagen-like proteins. These were considered as a Thr-rich group, a Pro-rich group, and a group rich in charged residues. The collagen-like sequences from different bacteria all had a relatively high Pro content, generally in the Xaa position, rather than in the Yaa position which in animal collagens is then frequently found as Hyp. In contrast, in the bacterial sequences there is a higher proportion of threonine and glutamine in the Yaa position. The other interesting trend was the uneven distribution of charged residues; most frequently the X positions having a negative charge, the Y positions a positive charge.²⁶ Subsequently, the extent of the genomic information has increased many-fold and a large number of additional genomes are available for study. While the initial motifs that were identified were typically from pathogens, a wide range of collagen-like motifs are also emerging from non-pathogenic species.²⁷

It was suggested that bacterial collagen sequences arise from horizontal gene transfer from eukaryotes to bacteria.²⁶ However, whatever the genetic origins, it is clear that a wide range of structures, functions, and molecular interactions have evolved and become established.²⁸ The few collagen-like structures from pathogenic species that have been well studied suggest that the collagen motif is typically associated with the outer membrane of the bacteria, and may interact with the host to assist invasion or to help a pathogen evade the host immune system.²⁸ This binding can be, for example, to integrin receptors²⁹ or to other extracellular matrix molecules, mediating pathogen internalization by human cells.³⁰

The identification of this new group of collagens provides potential for development of new, recombinant biomedical materials.³¹ Several have now been studied after expression in E. coli.²⁸ They do not contain any hydroxyproline, and even though some native bacterial structures are glycosylated, this is not the case with the recombinant products. Also, although a large number of species contain a (Gly-Xaa-Yaa)_n sequence repeat, only eight have been examined in detail. In seven of these cases—two from S. pyogenes, and one from each of Bacillus anthracis, Clostridium perfringens, Solibacter usitatus, Rhodopseudomonas palustris and Methylobacterium sp 4-46 (Fig. 2)—the characteristic CD spectrum of the collagen triple-helix, with a maximum near 220 nm and a minimum near 198 nm, has been observed, and a melting temperature, T_m, has been established using CD with increasing temperature (Fig. 1), confirming the presence of the triple-helical structure.²⁸ Additionally, for several of these structures and one further example, Legionella pneumophila, the presence of a collagen triple-helix has been inferred from protease resistance to enzymes including trypsin, chymotrypsin and pepsin.²⁸

Figure 2. A schematic illustrating the structures and melting temperatures, T_m, of bacterial collagens, including the non-triple-helical terminal domains (blue) and the central triple helical domain (red) for which a triple-helical structure has been established.

Of great interest was the fact that although the bacterial collagens expressed in E. coli do not contain Hyp, they are all, nevertheless remarkably stable (Fig. 2) with T_m’s in the range of ~35–39 °C, similar to T_m ~37 °C for human collagens. The lack of Hyp residues may also account in part for bacterial collagens not forming fibrillar structures,³² unlike type I collagen (Fig. 1). For those studied, it seems that in most cases the presence of non-triple-helical N- or C-terminal domains had little effect on the stability of the triple-helix.²⁷ The relatively high content of Pro residues in all of these proteins is an important stabilizing factor for the triple-helix structure. However, additionally, other stabilization strategies are in place. Some bacterial collagens, such as those from S. pyogenes, are rich in charged residues and stabilized by electrostatic interactions, while polar residues may contribute to the stability of some of the other bacterial collagens.²⁷ It is interesting to speculate that the bacterial collagens from pathogenic species have high T_m values that are adapted for attachment to host mammalian tissues, but this reasoning may not translate to non-pathogenic species, where a different environmental temperature may be important.²⁸

The high stability of bacterial collagens is clearly important for both the production and application of a bacterial collagen for biomedical applications. Other key features have also been assessed, particularly in relation to safety. Thus S. pyogenes Scl2 derived collagen, lacking any terminal domains, has been shown to be non-cytotoxic in cell culture (Fig. 3) using mouse L929 cells and human WI-38 and HT1080 cells,³³ and to be non-immunogenic in both inbred and outbred mouse strains.³³ There was a marginal response when adjuvant was used, but the response was much less than previously observed with avian collagen and bovine medical collagen.⁶

Figure 3. Fibroblasts growing on and within pores of a bacterial collagen sponge, stained after 7 d with a Live/Dead - Viability/Cytotoxicity Kit assay (Molecular Probes). Live cells are shown as green; any dead cells would appear red. Bar = 250 µm.

A further positive for developing bacterial collagen as a new medical material is that it can be readily produced in good quantity.³⁴ No doubt, this is helped in part by the bacterial collagens being shorter than the human fibrillar collagen sequences. For example, the S. pyogenes Scl2 collagen domain comprises 78 Gly-Xaa-Yaa triplets compared with the 338 triplets found in human collagen type I. To date, production studies have used a pColdIII (Takara Bio Inc) vector for expression in E. coli, as this vector system had proved useful for small scale expression of bacterial collagens from various species.²⁷ However, it is possible that other vectors could give better commercial yields? Production in shake flask cultures gives low yields of recombinant product, <1 g/L, but significantly better yields are possible when production is transferred to fed batch stirred tank bioreactors, especially when a high cell density strategy is used.³⁴ The best yields, of up to 19 g/L, were reported for a S. pyogenes Scl2 construct using a high cell density strategy and an extended 24 h production time. The construct contained an N-terminal non-collagenous domain, but this still gives a calculated yield of the collagen only domain of around 14 g/L.³⁴ The observed protein yields compare well with the average yields reported for other bacterial expression studies of 14 g/L.⁹

A good fermentation yield still needs to be matched with a simple and efficient downstream purification protocol. For commercial production, laboratory purification protocols, such as immobilized metal affinity chromatography with a His₆-tag and other smaller-scale chromatography systems, are not suitable. This has been resolved, as described in a recent study.³⁵ This study showed that acid precipitation of host, E. coli, proteins under conditions where the bacterial collagen remains soluble, followed by a proteolysis step that exploits the stability of the collagen triple helix to proteolytic digestion, provides a simple cost effective methodology that should be suitable for large scale purification of recombinant collagens.³⁵

Together, the lack of cytotoxicity and of immunogenicity, the ease of production and of purification would be sufficient to indicate the promise of bacterial collagens for biomedical material applications. However, recombinant technology also gives the opportunity to readily adapt structures to give defined or new functions. Some bacterial collagens already have identified endogenous cell-binding domains. For example, the S. pyogenes Scl1 collagen structure can interact with integrins α2β1 and α11β1 through the sequence GLPGER,³⁶ whereas, in contrast, the S. pyogenes Scl2.28 structure seems to lack any known binding sites, and does not show any cell interactions and behaves as a “blank slate.” The “blank slate” means that molecules can be designed and made with a single or multiple defined function(s). This has been explored by various groups, either by alteration of the existing helical domain or by insertion of additional triple-helical motifs. For example, integrin and heparin binding domains have been replaced into the S. pyogenes CL domain sequence,^37,38 while an integrin site and a fibronectin site have each been inserted as extra residues into the CL domain.³⁹ In addition, the collagenase cleavage in type III collagen has been examined when inserted between two CL domains.⁴⁰ The concept of producing dimers, or small oligomers of the bacterial collagens, allows larger molecules to be produced—for example, similar in size to mammalian fibrillar collagens—but fermentation yields would be expected to be lower. The opportunities for addition of further function into the collagen is large, based on the large number of biological domains found just in human type I collagen,⁴¹ let alone those emerging from the numerous other collagen types and related triple-helical proteins. In addition, non-triple-helical binding motifs, such as RGD, could be added to the N- or C-terminal of the CL domain recombinantly (or throughout chemically), as has been done with other proteins.⁴² Similar non-triple helical sequence additions also allow the development of yet other new molecules with defined, specific biological functions. Recently, the addition of a silk-like motif, (GAGAGS)_n, to the C-terminal of the S. pyogenes CL domain has been shown to enable binding to a silk substrate, providing the opportunity to develop more complex composite structures.³⁹

A key biotechnology consideration is what products could be developed from the new bacterial collagen-based materials? There would certainly seem to be opportunities in biomedical products, as with mammalian collagens, either where the new collagens are used as coatings or fabricated into a specific device, to provide enhanced products for specific applications. For example, as a coating, bacterial collagen can be bioengineered to introduce discrete integrin-binding sequences, GLPGER, GFPGER, or GFPGEN, to target specific integrin domains (α1 I-domain or α2 I-domain).³⁷ This provides coating substrates that can discriminate between endothelial cell and smooth muscle cell adhesion while avoiding any platelet aggregation, which would have specific applications for vascular prostheses and other cardiovascular technologies.^37,43 The ability to produce bacterial collagens as designed bioactive hydrogels⁴⁴ and in other formats, such as sponges,³³ further extends the potential biomedical and tissue engineering applications. Recent studies have demonstrated that an adequate shelf life for this type of product can also be achieved when appropriate stabilization and dry storage strategies are used.⁴⁵

Conclusions and Future Trends

In addition to the present examples that show the promise of recombinant, bioengineered bacterial collagens in biomedical and tissue engineering applications, other exciting developments could emerge in future. These could include constructs where the intrinsic stability has been further enhanced. This could arise from the discovery of new, naturally more stable molecules, such as the high stability, T_m = 42 °C, collagen found in a bacteriophage collagen,²² or by modifications to the sequence that incorporate the existing understanding of how different sequence features, including electrostatic interactions, can enhance stability.⁴⁶ The other option to increase stability is to add Hyp, which is currently not included in the E. coli production system. Previously, it has been suggested that this could be achieved by adding the modified amino acid Hyp to the fermentation medium,⁴⁷ although the efficiency of incorporation may be limited and there would not be specificity to the Yaa position in the sequence. Recently, however, a system that allows Yaa-position specific post-translational hydroxylation of collagen domains during fermentation in E. coli has been described and shown to be effective for short, peptide length collagen substrates.⁴⁸ The use of this approach for bacterial collagens may be limited by their larger size and the high proportion of Pro in the Xaa position. A further potential area for development is that of chimeric structures. These could include composite structures of different bacterial collagen components, for example, using different species for the folding and collagen domains, or different collagen domains, or constructs that include bacterial collagens and different proteins, which, for example, could be silks, elastic proteins, or other collagens.⁴⁹ Opportunities also exist to further functionalize the bacterial collagen core protein to allow selective simple and complex tethering of one of more peptides or non-peptide molecules for off the shelf development of specific protein materials with a defined application.

Collagen has developed a well-established, and well-deserved, position as a useful material in biomedical devices, where a variety of different formats including manufactured tissue and solubilized collagen formats have been developed. These products have been used successfully in many different clinical applications. However, in certain cases, such as with a manufactured tissue-based dura mater replacement, unfortunate adverse consequences have been observed, in this example leading to death due to disease transmission.⁵⁰ This example, and other potential problems arising from disease transmission, mean that there is still a concern about using native collagen-based materials and devices. Notable progress has been made on commercial, recombinant production of human collagens.^7,10-17 However, the rapidly emerging technologies that have characterized and produced recombinant non-animal (bacterial) collagens in E. coli^23-25,27,28 have introduced an exciting new collagen option that can be readily customized for specific biomedical applications.^28,33 These customized collagens will pave the way for more diverse applications in medical devices, implantable materials and potentially as scaffolds for stem cells, to name but a few examples of uses for these recombinant bacterial collagens. To enable commercialization, future demonstration of the high fermentation yields and straightforward purification process at pilot-scale (e.g., 500 L batch size) is required to show scalability and evaluation of process economics. The pilot-scale batches will also facilitate production of large amounts of material that can then be used for more extensive biological evaluation and applications development in general animal models followed by defined functional animal models for specific biomedical applications.

10.4161/bioe.28791