Reversible Non‐Stick Behaviour of a Bacterial Protein Polymer Provides a Tuneable Molecular Mimic for Cell and Tissue Engineering

Yersina pestis, the bubonic plague bacterium, is coated with a polymeric protein hydrogel for protection from host defences. The protein, which is robust and non-stick, resembles structures found in many eukaryotic extracellular-matrix proteins. Cells grown on the natural polymer cannot adhere and grow poorly; however, when cell-adhesion motifs are inserted into the protein, the cells proliferate.


DOI: 10.1002/adma.201304645
membrane "usher" protein [ 7 ] that provides a channel through to the extra-cellular surface. The nascent polymer grows out through this usher and across the outer membrane by the addition of monomeric subunits to its periplasmic end. Each monomer donates a single beta-strand to the preceding monomer thus joining the subunits by a strong but non-covalent link. Many CU proteins of pathogenic bacteria are so called adhesins, involved in binding to host cells, [ 8 ] but the bacterium Yersinia pestis , the etiologic agent of the bubonic plague, produces a unique form of CU polymer. After being injected into a warm blooded host via a fl ea bite, [ 9 ] a temperature sensitive promoter expresses the caf1 gene to produce a thick polymeric CU hydrogel coating on the surface of the cell. This effi ciently resists cell interactions and, by acting as a kind of anti-adhesin, inhibits macrophage attacks. [ 4 ] The formation of the Y. pestis Caf1 protein has been described by a series of elegant papers from Knight and coworkers who solved the high resolution structures of the chaperone (Caf1M-Caf1) and also revealed the energetics of the polymer formation. [ 10,11 ] We recently determined, by electron microscopy, the structure of Caf1 and revealed for the fi rst time the conformational fl exibility and large size of these poly mers. [ 12 ] They were up to 2 µm in length consisting of 800 monomers with molecular weights of up to 11 MDa. Usefully, these show no aggregation (apart from some inevitable knots) and remain soluble at >50 mg mL −1 . Furthermore the polymers are stable up to 90 °C, [ 13 ] protease resistant and easily purifi ed. [ 14 ] Models of Caf1 structure [ 12,15 ] and the structure of the similar Saf protein [ 16 ] predict it to be a polymer of immunoglobulin like domains ( Figure 1 A and B). As such it resembles the predominant family of extracellular proteins in humans (∼3% of human protein-coding genes) which includes fi bronectin type III repeats. [ 17,18 ] Since Caf1 also displays highly desirable properties (non-adhesion, stability and ease of production) that are diffi cult to design de novo into protein polymers, we investigated whether it could be a useful animal-free ECM substitute.
To imitate fi bronectin, we inserted into Caf1 the Arg-Gly-Asp-Ser peptide (RGDS) which has been shown to confer cell adhesion properties similar to the whole fi bronectin molecule. [ 19 ] The RGDS motif was incorporated into surface loops in order to expose the motif as in fi bronectin, [ 17 ] limit the changes to the structure of Caf1 and also to avoid important sites of Caf1: chaperone/ usher interactions. [ 10 ] Inserts were modeled using PyMOL [ 20 ] and the published coordinates for Caf1 (PDB fi le: 1Z9S) [ 10 ] and fi nally fi ve mutants were expressed and purifi ed (Figure 1 A).
Regenerative medicine demands the recreation of complex cellcell and cell-matrix interactions observed in vivo [ 1 ] and has led to the development of artifi cial biomaterials to mimic the protein network in the extracellular matrix (ECM). [ 2,3 ] Future developments would benefi t from an economic supply of protein polymers which closely match the molecular structure of the natural material. The polymeric Caf1 protein, from the plague bacterium Yersinia pestis , forms an enveloping hydrogel whose role is to inhibit interactions with host cells. [ 4 ] Furthermore, it shares a 3D structure with the largest class of human extracellular proteins. Here we show that recombinant Caf1 is a highly adaptable scaffold for a synthetic biology route to protein polymer engineering. It is robust, resisting both thermal and proteolytic degradation and when purifi ed it retains its biological ability to prevent mammalian cell attachment; three advantageous properties diffi cult to design into a protein de novo. We then reverse the "non-stick" phenotype by inserting a cell adhesion motif, express mixed polymers of different subunits and form hydrogels using a simple cross-linker. Such animal free proteins, economically produced in E. coli , offer a new family of tissue culture materials.
Gram-negative bacteria often form protein polymers on their surfaces via the chaperone-usher (CU) pathway, [ 5 ] so named because monomeric subunits, secreted into the bacterial periplasm, are initially stabilized by a specifi c chaperone protein. [ 6 ] Subsequently this binary complex interacts with an outer The expression of caf1 , from its own temperature dependent promoter, was revealed by the presence of a fl occulent layer (FL) above the cell pellet (CP) after centrifugation [ 14 ] (Figure 1 C). We analyzed this polymer rich layer by SDS-PAGE using three methods of preparation, no heat denaturation in which the polymer is intact (Figure 1 D lanes 3, 6, 9), limited heat denaturation (45 s at 95 °C) in which a ladder of oligomers is formed ( Figure 1 D lanes 2, 5, 8) and full heat denaturation (5 min at 95 °C) in which mostly monomers are observed [ 12 ] (Figure 1 D  lanes 1,4,7). In the heat denatured sample a signifi cant band ≈ 15 kDa was confi rmed as Caf1 monomer by western blot and peptide mass fi nger printing (Supporting Information, Figure  S1 and 2).
Of the fi ve mutants, L2RGDS (Loop 2 insertion) did not express Caf1, L5RGDS had the highest yield and thus we constructed a loop 5 Caf1 RGES mutant (L5RGES) as a noncell-adhesive control [ 21 ] and a high yield of L5RGES was also obtained (   [ 17 ] with sites of known cell adhesion motifs (RGDS and the accessory site PHSRN) highlighted in magenta. Lower molecule in green; Caf1 model based upon X-ray and EM structures [ 10,12 ] with RGDS insertion sites (loops) numbered. B) Linear, bead on a string, structure of Caf1 revealed by negative stain transmission electron microscopy. structure by near UVCD and electron microscopy. The spectra for near UV CD (Supporting Information, Figure S3A) for the Caf1 WT, Caf1 RGDS L5 and RGES L5 again show a similar structure. The main structural characteristics revealed on the near UV CD spectrum previously [12][13][14] such as the two minor (at 262 and 269 nm) and major peaks at 283 and 290 nm were observed in this study. Analysis of transmission electron microscopy images of negatively stained Caf1 fi bres was performed using Jmicrovision software. [ 22 ] We determined the length of Caf1 WT, Caf1 RGDS L5, Caf1 RGES L5 fi bres and the mean fi ber lengths were 401, 323, and 257 nm, respectively (Supporting Information, Figure  S3B). However it was evident, by inspecting the box charts, that there was a great heterogeneity of Caf1 fi ber length in all samples (Supporting Information, Figure S3B). Similar results for Caf1 WT fi ber length were observed by Soliakov and colleagues. [ 12 ] Next, to assess Caf1's suitability for cell culture, cell viability was assayed via calcein uptake. When cultured on surfaces coated with Caf1-RGDS, Caf1-RGES and Caf1-WT neither 3T3 fi broblasts nor PC12 cell lines revealed any toxicity within 48 h (Supporting Information, Figure S4). However, 3T3 cells proliferated more on fi bronectin. Fibronectin is a large protein which contains several other important peptides for cell proliferation such as PHSRN, LDVP and IDAP and so the challenge is to incorporate several different motifs in Caf1 fi bers as shown later.
We then studied cell morphology and adhesion in detail using scanning electron microscopy (SEM). Each cell line was grown on 12 mm glass cover slips pre-coated with either Caf1WT, L5RGDS, L5RGES, Fibronectin or Collagen IV, incubated for 24 h, fi xed and visualized by SEM ( Figures 2 , 3 ). Caf1WT inhibits adhesion of both PC12 and 3T3 cells whereas this effect is reversed on L5RGDS (Figure 2 A and B). To quantify cell adhesion on Caf1 protein-coated glass surfaces and on the control surfaces: collagen IV and fi bronectin, ten images, of each cell line were examined. The total number of cells counted varied between 558 for PC12 on collagen IV to 9 for 3T3 on WT Caf1 and the percentage of cell adhesion was determined (Supporting Information, Table S2). This revealed that approximately 80% of 3T3 cells bound to the L5RGDS and fi bronectin surfaces whilst only 3% bound to the WT Caf1 surface. However, the most striking differences were apparent when the shapes of the adherent cells were analyzed according to their interaction with the surface i.e., do they present one or more fi lopodia (cytoplasmic projections) (Figures 2 and 3). Thus Caf1WT provided a surface where the cells were invariably round and showed no projections. Critically, this behaviour of 3T3 and PC12 was largely reversed by the simple insertion of the RGDS motif, such that the results were identical to those on fi bronectin (Figures 2 and 3 and Supporting Information, Figure S5). Whilst the results for both 3T3 and PC12 cells on L5RGDS mimic those of fi bronectin, PC12 cells, as expected, grow rather better on collagen (Figure 3 ).
Previous work on other CU proteins has shown that when two subunit genes are expressed in the same cell they combine in the same polymer. [ 23 ] This offers a synthetic biology approach whereby a series of different monomers could be expressed in a single cell under the control of different promoters to create a range of mixed polymers on demand. To demonstrate this possibility for the Caf1 polymer system we expressed two different genes with two plasmids each carrying a different origin of replication, selection marker and promoter ( Figure 4 A); pAH34L containing the complete wild type Caf1 operon [ 14 ] and pBAD33 [ 24 ] which expressed only a mutant caf1 gene which included a FLAG epitope termed caf1-FLAG . The    (Figure 4 B lane  9). Finally, co-expression of the two plasmids produced FLAG labeled dimers and trimers when heated for 45s (Figure 4 C lane 2). Thus, caf1-FLAG must have assembled into polymers using the WT usher protein supplied by plasmid pAH34L. Since the corresponding lane 2 in Figure 4 C also reacts with anti-Caf1 antibodies it is thus reasonable to assume that we have produced mixed polymers composed of the products of both plasmids and not pure polymers of the different subunits. (Figure 4 B and C lane 2). Thus, Caf1-FLAG was detected outside the cell in the fl occulent layer (Figure 4 C) showing for the fi rst time the export of Caf1 hybrid polymers. In future the scaffold could more closely mimic the complexity of ECM by containing several functional monomers including different cell adhesion motifs e.g., PHSRN (Figure 1 A) or protease remodelling sites.
The Caf1 polymer can also be used to form a hydrogel with possible applications in 3D cell culture. WTCaf1 at a fi nal concentration of 30 mg mL −1 produces a viscous solution which we stabilized by the addition of amine reactive cross linkers of various lengths DTSSP (12.0 Å), NHS-PEG-NHS (197 Å), and 4-arm NHS-PEG (2 × 197 Å). These can react with any of the eight surface lysines on each Caf1 monomer. The Caf1 hydrogels were characterized using a simple tubeinversion assay in which non-cross linked solutions would fl ow down the sides of the plastic reaction vessel. The gelation time was visually estimated (Supporting Information, Table S3) to be within 24 to 27 min for NHS-PEG-NHS and 2 to 22 min for 4-arm PEG-NHS, depending on the concentration of the cross-linker. The higher the concentration of these two cross-linkers the quicker the gelation time. With DTSSP a solid gel was not observed. The increase in gelation rate of 4-arm PEG-NHS could be due to the structure of 4-arm PEG which infl uences its ability to react with the primary amine groups of Caf1. [ 25 ] The gelation time reported here is comparable with other studies using PEG hydrogels, for example Liu et al., (2012) [ 26 ] who examined the combination of PEG diacrylate (PEGDA) and acryloyl-PEG-RGD.
After cross-linking for 30 min, samples were analyzed on a 4-20% gradient gel. The band of approximately 15 kDa corresponding to the Caf1 monomer in the non-cross linked Caf1 sample (control) was used as a reference for the subsequent analysis. The relative density of the unreacted Caf1 monomer band, in Caf1 samples cross-linked with the different crosslinkers at various concentrations, was determined. Since the calculations of the relative density for the high molecular weight  bands were more complex, these were considered a single band and referred to as the "Caf1 cross-linked fraction". The 4-arm PEG-NHS showed the least amount of residual monomer and thus a higher degree of cross linking (Supporting Information, Table S4) Caf1 polymers cross-linked with a short arm length crosslinker such as DTSSP (12.0 Å) promote a closer contact between the Caf1 fi bres. Thus, the images obtained by TEM revealed a compact Caf1 hydrogel which refl ects the Caf1 fi bres proximity. A different result was obtained for Caf1 polymers cross-linked with a long spacer NHS-PEG-NHS (197 Å), which can better separate the Caf1 fi bres. Large Caf1 hydrogel meshes were observed by TEM. When we used the 4-arm PEG-NHS where each cross section is also 197 Å the interactions between Caf1 and the 4-arm PEG-NHS the structure was more condensed than that produced by NHS-PEG-NHS (Supporting Information, Figure S6).
When using a monomeric form of Caf1 made by circular permutation, cpCaf1, [ 13 ] these large networks were not seen (Supporting Information, Figure S7). A highly porous hydrogel (Supporting Information, Figure S7) could be advantageous for swelling and water uptake and also as a scaffold for cell culture that can allow the passage of nutrients, oxygen through the pores. The TEM images confi rmed that the formation of the Caf1 hydrogels depends on the cross-linker concentration and structure. The pore diameters of 4-arm PEG-NHS stabilized Caf1 hydrogels were assessed by SEM and ESEM (environmental SEM). [ 27 ] For SEM we used freeze-dried hydrogels which broke into fragments during analysis (Figure 4 D) and revealed a mean pore diameter of 8 ± 1.9 µm (Figure 4 E). ESEM analysis avoids the dehydration process and these images revealed a mesh-like network structure with a mean pore diameter of 300 ± 0.3 nm (Figure 4 F).
As shown by Lutolf and Hubbell (2005) [ 3 ] in tight gel networks, cells are required to use proteolytic strategies to degrade the surrounding matrix and be able to migrate through the gel. Cells can remain viable in stiffer gels, however they cannot spread and proliferate which is fundamental for obtaining cell-cell contacts and subsequent tissue formation. [ 28 ] The Caf1 polymer thus can be made to imitate different ECM by having varied sub-unit composition, different crosslinking ratios and possible protease cleavage sites. Its ease of production in an inexpensive and safe host bacterium means that it is a viable bulk product able to replace expensive cell culture reagents.

Experimental Section
For details of the experimental methods used please see the Supporting Information.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.