Recombinant protein scaffolds for tissue engineering Jerome A. Werkmeister and John A.M. Ramshaw CSIRO Materials Science and Engineering Bayview Avenue, Clayton 3169, Australia E-mail: [email protected] 1 Abstract: New biological materials for tissue engineering are now being developed using common genetic engineering capabilities to clone and express a variety of genetic elements that allow cost effective purification and scaffold fabrication from these recombinant proteins, peptides or from chimeric combinations of these. The field is limitless as long as the gene sequences are known. The utility is dependent on the ease, product yield and adaptability of these protein products to the biomedical field. The development of recombinant proteins as scaffolds, while still an emerging technology with respect to commercial products, is scientifically superior to current use of natural materials or synthetic polymer scaffolds, in terms of designing specific structures with desired degrees of biological complexities and motifs. In the field of tissue engineering, next generation scaffolds will be the key to directing appropriate tissue regeneration. The initial period of biodegradable synthetic scaffolds that provided shape and mechanical integrity, but no biological information, is phasing out. The era of protein scaffolds offers distinct advantages, particularly with the combination of powerful tools of molecular biology. These include, for example, the production of human proteins of uniform quality that are free of infectious agents and the ability to make suitable quantities of proteins that are found in low quantity or are hard to isolate from tissue. For the particular needs of tissue engineering scaffolds, fibrous proteins like collagens, elastin, silks and combinations of these, offers the further advantages of natural well-defined structural scaffolds as well as the endless possibilities of controlling functionality by genetic manipulation. Keywords: Tissue engineering, scaffolds, recombinant proteins, fibrous proteins, collagen, silk, elastin Running Title: Recombinant protein scaffolds 2 1. Introduction Tissue engineering and regenerative medicine continue to engage a plethora of multi- disciplinary scientists as an advanced strategy to restore or regenerate damaged or diseased tissue. In its most simplistic form it comprises culturing autologous or allogeneic cells, including stem cells, with or without appropriate cues and signals, in a range of scaffolds to form functional tissues for implantation. The scaffold plays a pivotal part providing a suitable support for cell adhesion, proliferation and in some cases differentiation. In addition scaffolds can be fabricated to provide the necessary biochemical and biomechanical cues that are required to mimic the natural extracellular matrix of the human body. The hope of regenerative medicine is to ultimately treat patients in the most natural way that does not simply add structural elements, but clearly is able to restore or induce tissue regeneration thus avoiding the need for organ or tissue replacement (Williams and Sebastine 2005). Initial approaches, which are still favoured by some, used synthetic polymers such as poly-lactic acid and poly-lactic/glycolic acid copolymers as scaffold materials. While such scaffolds can define the size and shape that is required from a scaffold construct, they present little information to the cells on the type and complexity of the extracellular matrix that is required. While adding growth factors and physical stimuli (Engler et al 2006) can assist in defining the tissue that is formed, they are probably still insufficient to fully define and produce a reasonable match to the native tissue that is being replaced. For example, to what extent do tissue engineered heart valves match the complexity of collagen distributions that are found in this tissue (White et al 2010) and which could be associated with the long-term durability and effective performance of an engineered device. Fibrous proteins are gaining increased attention in tissue engineering and, in particular, the versatility in structure and design is making recombinant forms of these materials the preferred option in advanced materials. Fibrous proteins are among the oldest medical materials. Keratin, probably as horse or other hair, may have been used as an early material and could possibly be the material mentioned for performing surgical suturing in the earliest known medical text, 5000 BP, the Edwin Smith Papyrus (Breasted 1930). Silk would undoubtedly have been used as an early medical material; for example, as sericulture, especially using Bombyx mori, dating back to about 2700 BC, with other silk cultivation possibly even earlier (Barber 1992). B. mori is the most well known silkworm, domesticated from B. mandarina and used for silk production for over 5000 years. Legend has it that the Chinese Princess Xi Ling Shi was the “true” inventor of the benefits of silk – the story is that a silkworm cocoon fell into her hot water being prepared for tea and the silk thread began to unravel itself (Wurm 2003). The oldest known suture is in a mummy from 1100 BC, while the first detailed description of a wound suture and the suture materials used in it is by the Indian physician Sushruta, written in 500 BCE. Development of collagen-based gut sutures has been attributed to Galen who was cited as using them for suturing wounded gladiators around 100 CE (MacKenzie 1973). The dissolving nature of catgut sutures was observed in the 10th-century by Andalusian surgeon al-Zahrawi, who reportedly discovered the dissolving nature of catgut when his lute's strings were eaten by a monkey. Within the last century, collagen in particular has been used successfully in a wide range of medical applications (Ramshaw et al 2009), perhaps more so than any other proteinaceous material. Proteins, and especially fibrous proteins, typically have „high‟ information content. The presence of different levels of information content in different cases of molecules has long been recognised, especially by those working in the field of molecular evolution (Zuckerkandl and Pauling 1965). The same concept is also true for molecules used as tissue engineering scaffolds. While synthetic, repetitive polymers have little information content, proteins with their structural complexity provide a wealth of potential information that cells 3 can recognise and respond to. For example, type I collagen alone has more than 50 documented molecular interactions (Sweeney et al 2008), and the other collagen types, now at least 28, contain even further content. Type IV collagen is a major component of basement membranes and so has a different set of interactions than the interstitial collagens, including interactions to different stem cell types. Although presently less numerous and less well defined, elastin also has specific interactions, including the role of the C-terminal in cell interactions (Kinikoglu et al 2011). The availability of recombinant fibrous proteins goes back a long way (Werkmeister et al 2003). For example, recombinant collagen production with hydroxylation of proline by introduced prolyl-4-hydroxylase was initially described in 1993 (Prockop et al 1993), although non-hydroxylated variants, including hybrid forms with other proteins were described in the late 1980‟s (Williams et al 1988). Even then, obtaining hydroxylated recombinant collagen in a commercial yield took around a further 10 years to achieve (Neubauer et al 2007). Similarly, recombinant elastin was available in very low yield in initial studies (Indik et al 1990), but it was only in 1995 that a system that could provide commercial yields was devised (Martin et al 1995). This involved constructing a fully synthetic gene of 2210 base pairs, in which the codons were optimised for expression in E. coli. Equally, recombinant silk-like proteins were available a decade ago, but only as partial constructs (Lazaris et al 2002), as full length constructs were too large and highly repetitive. The recent advances in all these proteins, therefore, have occurred in refinement of the various basic systems. Broadly, these include alternative production systems, designs based on key structural elements, development of new sources, production of hybrid structures with other proteins and composite materials, frequently with other fibrous proteins. It is these refinements that are discussed in the following review. 2. Collagens This section examines the production, efficiency and applications of recombinant collagens in the biomedical industry. The precedent for use of recombinant proteins already exists in the biopharmaceutical arena. The first recombinant protein biopharmaceutical, insulin, was approved nearly three decades ago, and this had led to an exponential surge in the so-called biotechnology-derived pharmaceutical products, largely of growth factors and humanised monoclonal antibodies. It is expected that the same tools can be used to engage in the biomedical and tissue engineering fields to open new pathways and systems to treat damaged or diseased tissues. 2.1 Background Collagens are the major protein of all connective tissues comprising a family of 28 genetically distinct types (Ramshaw et al 2009), each characterised by a common triple helical structure that forms the building block of tissue architecture. Collagen type I is the predominant interstitial fibril-forming collagen and for this reason has been the easiest and most abundant source for use as biomaterial and medical devices (Ramshaw et al 2009). All collagens are defined by 3 helical chains, each in a left handed polyproline II-like helix, that wind together to form a right handed super-coiled triple helix. At the molecular level it is defined by a unique three amino acid repeat, (GXY), where glycine is always found in every third position, and the Xaa and Yaa positions are often occupied by the imino acid proline. In the Yaa position the proline is modified post-translationally to 4-hydroxyproline (Hyp), which confers stability to the triple helix, and enables collagens, at least mammalian collagens, to be thermally stable at human body temperature. 4 The biosynthetic collagen pathway is quite complex and requires a number of secondary essential modification steps including some that are important for crosslinking and helical stability. For recombinant expression of collagen type I or any collagen, it is essential to mimic the part of the natural biosynthetic pathway that confers stability. In most expression systems this requires the co-expression of prolyl-4-hydroxylase (P4H; E.C. 1.14.11.2) which is translated from two genes that produce an active tetramer α β of P4H 2 2 that is essential for hydroxylation of proline in the Y position leading to helical stability. 2.2 Expression of mammalian collagen Almost every possible type of expression system has been used for recombinant expression of collagen, collagen-like proteins and selected domains. Largely, the wide array of expression systems has been used to further enhance the understanding of collagen structure- function in normal and pathological conditions. But the availability of gene sequences from all collagen types adds certain value to the future of using some of these as added value products for specific applications. For instance, collagen type XXVIII, a novel von Willebrand factor A domain-containing protein, is highly localised in the sciatic nerve at the basement membrane of certain Schwann cells surrounding nerve fibres (Veit et al 2006); collagen type XXII exhibits a strikingly restricted localization at tissue junctions, particularly the myotendinous junction in skeletal and heart muscle (Koch et al 2004); collagen type XVII is a transmembrane constituent of the epithelial anchoring complex involved in maintenance of cell adhesion at this interface (Van den Bergh et al 2006); collagen type XV, structurally homologous to collagen type XVII abundant in liver, is highly expressed in the skeletal muscle and regulates cell adhesion and migration (Hurskainen et al 2010). For commercial applications, as medical grade products for biomaterials or scaffolds in regenerative medicine, the recombinant approach offers distinct advantages including: availability and uniformity of product quality; ease of isolation and purification in certain systems; freedom from infectious agents associated with current animal-derived products; potential to produce any of the rare collagen types, not just the routine collagen type I; potential scope to modify and improve on natural protein structures, including selection of specific domains, engineering products with multiple advantageous repetitive motifs or chimeric molecules. In addition, cost-effectiveness is the key factor that will decide the future of these proteins. Recombinant collagens are now available commercially largely for in vitro applications using a very precise expression system aimed at low cost with reasonable yield (Baez et al 2005). 2.2.1 Prokaryotic expression E. coli is the most common bacteria used for expression of recombinant proteins including some extracellular matrix proteins. It is possible to use bacterial expression systems for small motifs, particularly those with inherent capacity to re-fold. There are multiple issues with this approach in general, at least for mammalian collagen proteins. Firstly it requires co- expression of the P4H enzyme which has been very problematic in bacteria; active P4H tetramer has been occasionally expressed in E. coli using a special strain of bacteria with a high oxidising cytosol (Neubauer et al 2005). In addition, the nature of these proteins means that most will end up as insoluble denatured proteins (inclusion bodies) within the bacteria. Production of full length mammalian collagen is highly unlikely in this system and most products reflect the denatured form of this protein, gelatine (Ferrari et al 1993, Cappello 1998). Smaller mammalian collagen fragments have been produced with limited success; for example, part of the collagen type V α1 chain comprising the heparin binding site (Ruggiero 5 and Koch 2008) has been expressed as a soluble functional small motif and production yields were increased using strain of bacteria that could be induced with salt rather than the standard lac promoter inducer IPTG (Ricard-Blum et al 2006). More recently, there have been a number of scattered reports on optimisation of E. coli for enhanced biomass production of recombinant human-like collagens. These include enhanced growth and expression by controlled CO pulsing (Xue et al 2009), increasing the 2 fermentor pressure and elevating the oxygen transfer rate (Ma et al 2010), medium optimisation of the carbon/nitrogen ratio (Guo et al 2010) and assimilation of acetic acid production by glucose starvation (Xue et al 2010). In these limited studies from these Chinese groups, the increase in biomass and product yield has been significant. With the control of oxygen transfer methods, the cell density and volumetric product yield increased by 2-fold to 77 gl-1 and 14 gl-1 respectively; with the medium optimisation methods, optimal carbon/nitrogen molar ratios for batch (2.36:1) and feeding media (5.12:1) yielded similar biomass of 67.2 gl-1 and collagen production of 10.8 gl-1. In all these studies it still remains questionable if the product is indeed recombinant collagen or merely gelatin as no evidence is given to the precise characterisation of the homotrimeric collagen type II that was produced. 2.2.2 Eukaryotic expression The lack of effective post-translational processing of collagens in bacteria has favoured a number of eukaryotic systems for collagen production. The choice here is large and there are certain advantages and disadvantages for each system. Certainly the simplest approach is to use a system that does not involve the co-expression of the essential P4H enzyme. Mammalian cell lines, including CHO, COS, HEK 293and HT1080, offer this distinct advantage – they all have the complete post translational machinery for correct folding and stability of collagen. Indeed given the complex biosynthetic pathway for collagen synthesis, it is not surprising that mammalian expression systems have always been the first choice for collagen production for purely scientific discovery and understanding collagen behaviour. The downside, however, is the very poor yield, generally less than 0.1-0.5 mgl-1 (Ruggiero and Koch 2008). The HEK 293-EBNA cell line was developed as an optimised mammalian expression system with very low background matrix production. This cell line has been used to express larger amounts of a number of matrix proteins including collagen types V, VII, VIII, X, XVI. Production yields have increased up to 80 mgl-1. Large scale production of another matrix protein, the globular LG4/5 domains of laminin 5, has been reported to be much higher, around 3 gl-1, using this system (Belin and Rousselle 2006). While offering a simple and easy approach to producing stable collagen, this methodology is not suitable for commercial production due to low yields. The most common expression system for large scale mammalian collagen production is yeast. While mammalian cells have P4H activity and insect cells have poor intrinsic P4H capacity, yeast cells do not possess this enzyme and so expression systems have been optimised to integrate and co-express at least 3 concurrent genes (at least 1 collagen gene for homotrimeric collagens and 2 genes for the P4H). Saccharomyces systems can be used and various designs have been adopted to integrate the 3 genes necessary for stable collagen production (Vaughan et al 1998, Toman et al 2000). Most of these studies, as well as those described below in Pichia systems, have been focused on large scale production of unmodified recombinant collagens. From a tissue engineering aspect, it is important to use these systems to optimise the genetic design to enhance performance. Fertala and colleagues have engineered a gene product comprising multiple repeats of a biological active complete D4 period (Majsterek et al 2003). We have recently shown it is possible to design and incorporate much smaller specific biological motifs to direct cellular function, and cell function can be enhanced by constructing genes with 2 or more repeating motifs encoding 6 these biological entities (Peng et al 2009). While product yields are still low in this system, these and other expressed proteins can be used to coat or add value to other types of scaffolds. P. pastoris is the most well developed yeast strain currently developed for effective commercial collagen production. It has been successfully engineered to co-express the P4H with the appropriate mammalian collagen gene. Fermentation systems have seen improvements of protein yield from 0.015 gl-1 to 1.5 gl-1 or greater for full length collagen, and up to 14 gl-1 for smaller defined collagen fragments (Baez et al 2005). There have been various approaches to improve biomass and product yields and it is the combination of all these, rather than one single improvement that has driven commercial viability. These include codon optimisation for P. pastoris; deletion of the N-propeptide which shortened the gene length and which was found to be unnecessary for correct folding; replacement of the C- propeptide with a shorter trimerizing foldon peptide from the phage T4 fibritin; optimisation of integration and copy number for P4H; fermentation at a reduced temperature of 32° C which increased likelihood of proline hydroxylation in the Y position; oxygen enrichment to increase the P4H activity that requires high concentrations of molecular oxygen; high biomass and subsequent product accumulation by designing an animal free fermentation broth with an initial glycerol feed containing hexametaphosphate and ammonium sulphate. Of all available yeast strains, P. pastoris has the potential for the highest expression levels with appropriate post-translational modifications. The system uses a strong promoter of alcohol oxidase (AOX1) which, after the initial glycerol feed and biomass optimisation, is induced by methanol which also acts as the sole carbon energy source. The combination of all these factors have led to high end point product yields (Baez et al 2005). This has involved patient and precise improvements along the way, but also initial upfront engineering costs to safely design large scale flame-proof fermentation facilities for production using high volumes of oxygen and methanol. Generally, optimisation of bioreactors uses simple fed- batch fermentation with controlled continuous addition of methanol. Improved production of human type II procollagen has recently been reported in P. pastoris in simple shake flasks using wireless-controlled fed batch (Ruottinen et al 2008), resulting in a 10-fold higher amount of mRNA and protein synthesis. Hansenula polymorpha, now re-classified to the Pichia genus, offers an alternate yeast strain that has been reported to have endogenous P4H activity and could possibly be used as a simpler expression system (de Bruin et al 2002). In this initial report only very small collagen fragments were expressed and some of the proteins may have been complicated by an endogenous collagen in the organism itself. A more recent study was unable to generate hydroxylated collagen and in addition failed to identify a gene encoding the P4H in the genome (Geerlings et al 2007). Downstream processing of fermented organisms need to be considered – generally these have to be simple, compliant with regulation and cost effective. For Pichia systems, the process is exactly that – the purification system does not involve laboratory chromatographic methods. Cross flow filtration, enzyme digestion and differential salt precipitation at defined pH is usually the method of choice. The issue for production of heterotrimeric collagen types is still challenging. Collagen type II and III are homotrimers and there have been no issues with production of thermostable products. But for collagen type I there is a preference to assemble the non- natural [α1(I)] homotrimer and not the natural heterotrimeric [α1(I)] α2(I) in the Pichia 3 2 system (Baez et al 2005). This, however, may not be unique to yeast systems as it appears to be a general phenomenon associated with heterologous expression systems. In a recent report, Roulet and colleagues used collagen type V as a model collagen product (Roulet et al 2010). Collagen type V, a fibrillar collagen like type I, II and III collagen, is naturally found as an abundant ubiquitous heterotrimeric collagen, [α1(V)] α2(V) and a less common 2 7 homotrimer form, [α1(V)] Surprisingly, using yeast, insect and even mammalian HEK-293 3. cells, the formation of the homotrimer [α1(V)] was considerably favoured over the 3 heterotrimer. There is only one company, Fibrogen (www.fibrogen.com/collagen) that produces highly purified fully characterised recombinant collagens for laboratory use and intended for medical applications. 2.2.3 Use of transgenic plants and animals A transgenic animal or plant is one that contains a deliberate modification to the genome such that the foreign introduced DNA is transmitted through the germ line to every cell. The simplest way of doing this is by using conventional recombinant DNA tools to microinject the foreign DNA directly into an ovum. Several transgenic animal species are capable of producing recombinant proteins from milk, blood, egg white, semen, urine and silk glands (Houdebine 2009). The mammary gland is the most explored system to express many recombinant proteins in transgenic animal bioreactors, followed by chicken egg white. Quite a number of mammalian species can be used including pigs, sheep, goats, cows, rabbits and mice. Each will have its own advantages and disadvantages. From a cost effective exercise, pigs or rabbits are seen to be the choice due to high levels of milk, high fertility and disease-resistance, although other options like the dwarf Nigerian goat have also been attempted for production of silks. There has been very little work on use of animal transgenics for mammalian collagen production. Expression of foreign genomic human α1(I) in mouse milk resulted only in severely under-hydroxylated procollagen that was thermally unstable suggesting inadequate endogenous P4H (John et al 1999). With the introduction of foreign α and β genes of P4H into the genome, reasonable levels of stable collagen, upto 0.2 gl-1 were produced (John et al 1999). A comparable report (Toman et al 1999, Berg 2000) produced much higher production yields of stable collagen around 8 gl-1 but these were truncated 37Kb molecular weight proteins and again all were homotrimeric in nature. While the use of animal transgenics has not been followed up from these early reports, the use of transgenic plants does offer some cost effective and commercial outcomes (Dyck et al 2003). Transgenic plants have a track record as bioreactors for pharmaceuticals (eg insulin, lactoferrin, and antibodies), vaccines, industrial processing enzymes (trypsin) largely due to the belief that these systems offer practical and cost effective solutions to recombinantly-expressed proteins. The first reports on transgenic expression of recombinant collagen-related proteins in plants used a model system of tobacco (Ruggiero et al 2000). These tobacco plants were able to produce fully processed triple helical human collagen type III (Olsen et al 2003) and collagen type I homotrimer (Ruggiero et al 2000). Of particular significance in these studies, was that while tobacco plants did provide a useful expression system for large scale production of recombinant collagen, hydroxylation of proline in the Y position did not occur effectively. So while there was some degree of triple helical folding occurring that allowed clues on the role of hydroxyproline in collagen folding and fibrillogenesis (Perret et al 2001), collagen expression in transgenic plants does require the additional genomic insertion of the genes for the P4H. Barley has also been used to produce full length collagen type I α chain by targeting to the endoplasmic reticulum using 3 different promoters (maize ubiquitin, endosperm-specific glutelin or a germination-specific α-amylase) (Eskelin et al 2009). In this study and others using Wave bioreactors to culture transgenic barley cells (Ritala et al 2008), the highest yields were obtained with the glutelin promoter (140 mg/kg seed) which is relatively still very low. More recently tobacco plants have again been tried as transgene for production of heterotrimer collagen type I (Stein et al 2009). In this study both collagen α chains were co-expressed with both genes encoding the P4H enzyme as well as the gene 8 encoding lysyl hydroxylase. Plants co-expressing all 5 vacuole-targeted proteins resulted in generation of full length collagen with extremely high yields around 2% (20 gl-1) of the extracted total soluble protein. This approach using tobacco for large scale collagen production has now been recently commercialised by CollPlant Ltd (www.meytavti.co.il). Transgenic corn has been extensively studied for the large scale production of recombinant industrial and medical pharmaceuticals. Corn, like other plant crops, offers advantages of large volumes (1,000‟s tons per year) and low cost around US $50/kg of grain (Zhang et al 2009a). Corn-grain production of a small human α1 collagen fragment (44 kDa) showed around 18% hydroxylation (2% hydroxyproline compared to maximum 11% expected in animal collagen type I) suggesting some endogenous P4H activity but the yields were only around 3 mg/kg (Zhang et al 2009a). Comparable low yields and low level hydroxylation also occurred with full length recombinant collagen type I using embryo specific promoters and a foldon to enhance trimerization (Zhang et al 2009b). The issue of improving product yields is currently under investigation and largely reflects methods of optimising protein extraction in a simple and cost effective approach without the cumbersome traditional analytical methods of purification (Zhang et al 2009b, Paraman et al 2010). Both dry and wet milling approaches are being optimised – both seem to be compatible with large scale up and purification. Dry milling is simple and can be performed on the farm minimising cost and potential inadvertent release; wet milling is largely more effective in yields and purity Transgenic silkworms have also been investigated as a convenient bioreactor system for production of recombinant collagen and chimeric products. Many recombinant proteins have been produced, particularly over the last decade, in silkworm larvae or pupae (Kato et al 2010). The benefits of transgenic silkworms is the convenience and cost effectiveness with increase product yields in most cases ranging from 20 to 10,000 fold compared with laboratory methods. Using a piggyBack transposon-derived vector system, recombinant collagen type III has been successfully produced in a stable form in the silk glands of B.mori larvae (Tomita et al 2003). The global annual production of silk is around 60,000 tons so the mathematics and logistics of this approach for foreign protein production is positive. As an additive to the natural silk product to improve cell function, the fibroin light chain gene from B. mori has been modified with a collagen insert - an 8 repeat of a 10 triplet GXY sequence with a C-C at the C-terminal end for stability (Yanagisawa et al 2007) which enhanced cell adhesion. More recently, full length collagen type I α chain has been produced in the middle silk glands and secreted into the cocoons of transgenic silkworms (Adachi et al 2010). The initial yields were poor (0.8%), but was improved significantly to around 8% (4.24 mg per cocoon) by cross breeding with moths carrying a gene of the transactivator IE1 from baculovirus. CD spectra and thermal transition curves show a lack of native triple helical structure, but the product was still capable of supporting cell growth; in particular the recombinant substrate was capable of supporting Cynomologous monkey embryonic stem cells in their undifferentiated state for a remarkable 30 passages. 2.3 Bacterial collagens Proteins comprising the repeating triplet of GXY with high proline content have the potential to form a left handed polyproline II and, with 3 chains, can then form a typical right-handed triple helical structure (Ramachandran 1988). Collagens are already known to be present in non-eukaryotes like sponges, mussels and worms. To some extent then, there was no great surprise that DNA encoding these collagen-like sequences was also found in the genome of bacteria (Ferretti et al 2001) and bacteriophages (Smith et al 1998). What was a surprise is that these genes can produce thermally stable collagen proteins even though these organisms 9 are unable to synthesize hydroxyproline (Figure 1). A more thorough bioinformatics BLAST search of available bacterial genome databases (137 eubacterial genomes; 15 archaebacterial genomes; 30 genomes of lower eukarya and against viral genomes) identified 103 collagen- related structural motifs (CSM) containing GXY repeats (Rasmussen et al 2003). The length of these CSMs varied from 7 to 745 continuous triplet repeats with a mean length of 76 GXY repeats. In particular, 53 CSMs were found from the 137 eubacterial genomes examined, and these contained a significantly higher proportion of threonine and glutamine in the Y position with a propensity for the X position to be proline, alanine or serine. The other important trend was the uneven distribution of charged residues; most frequently the X positions having a negative charge, the Y positions a positive charge. Molecular modelling suggested that the threonine in the Y position is able to form intermolecular hydrogen bonds to near-by carbonyl groups in the helix backbone, and the concept of charge, ionic pairing and sequence stability has been validated theoretically and experimentally (Persikov et al 2005). There is a large group of pathogenic bacteria that express these collagen-like sequences that are largely associated with host colonization and virulence in humans, including Group A Streptococcus (Lukomski et al 2000), Bacillus anthracis exosporium (Sylvestre et al 2002) and in insects, for example, the fungal pathogen Metarhizium anisopliae (Wang and St Leger 2006). Depending on the bacteria strain, the Group A Streptococcus presents multiple proteins including the streptococcal collagen-like proteins Scl 1(Lukomski et al 2000) and Scl 2 (Lukomski et al 2001), the former comprising the GLPGER sequence which can interact with the α β and α β integrins and induce 2 1 11 1 intracellular signalling (Humstoe et al 2005) or can interact with other extracellular matrix proteins including fibronectin and laminin (Caswell et al 2010). In a more in depth study, Caswell and colleagues used the “blank slate” from a non-cell binding integrin-lacking CSM sequence from Scl2.28 to introduce various G-Xaa-Yaa sequences to confirm the identity of the prokaryotic sequence for the human collagen receptors and the higher affinity for the α β integrin (Caswell et al 2008). There may be some immunological concerns with these 11 1 bacterial collagen-like proteins. By the nature of their natural function, these proteins could be thought to immunodominant proteins in their natural form as bacterial antigens. The collagen-like protein from B. anthracis exosporium can induce antibodies in mice with adjuvant and can further act as a booster to help protect mice against suboptimal amounts of infecting bacteria (Brahmbhatt et al 2007); these antibodies, however, did not cross react against human collagen types I, III or V. In a more recent study we have shown that the recombinant Scl2 protein is not immunogenic in mice either with or without the presence of adjuvant (Peng et al 2010). The presence of GXY repeats is also predicted in non-pathogenic bacteria (Rasmussen et al 2003) and recombinant proteins from a mixture of Gram-positive and Gram-negative bacteria (for example Methylobacterium sp 4-46, Rhodopseudomonas palustris and Solibacter usitatus) have been expressed in E .coli and characterized (Xu et al 2010). All of the expressed products were soluble and maintained similar triple helical thermal stability irrespective of the diverse variation in amino acid sequences. In the same study, the CSM from Clostridium perfringens was insoluble but could be coaxed into secreting soluble thermally stable products by flanking the CSM with heterologous folding trimerization domains at either the N or C-terminus. Assembly of stable human type I and III collagen molecules in P. pastoris has been shown to be dependent on the efficiency of trimerization; the substitution of the simple foldon sequence, derived from the native T4 phage fibritin, was able to generate 2.5 to 3-fold increase levels of collagen expression compared with the natural C-propeptides (Pakkanen et al 2003). The foldon sequence has also been used in E. coli expression systems to augment triple helical stability of collagens (Du et al 2008); in this instance the authors claim to express stable human collagen products 10