For each of our three subgroups, Bioprinting, Mussel Foot Protein and Spinal Cord Injury, we have developed extensive design, build and test cycles. We designed our engineering cycles to demonstrate how we built on our Phase I project design, developed our protocols and tested our parts in the lab.
In Phase I we identified five possible macro-architecture designs that could be suitable for our project: channelled, cylindrical, tube, open-path with a core, and open-path without a core (Wong et al., 2008). Unfortunately, due to COVID-19 restrictions, we were unable to build any of these designs in the lab for testing, however, we did perform Finite Element Analysis of all five designs, investigating variables relating to solid von Mises stress and strain, displacement and applied force. We subsequently discovered that although all five designs did not exceed the yield strength, the two open-path designs were mechanically superior. Following literature reviews, we learnt that the open-path with core structure was more beneficial for regeneration, and hence this design was selected. In regards to the micro-architecture, we designed a scaffold consisting of an iterated gyroid unit cell fitted within the macro-architecture, thus generating a homogeneous porous medium for axonal regeneration. We designed an algorithm to generate porosity and found that the chosen unit cell design was suitable for application within our project due to its properties—such as high interconnectivity. For Phase II of our project, we planned to test and validate these design choices and evaluate whether the micro- and macro-architectures selected were optimal for their purpose. In the case that the design was flawed, we planned to rectify these issues through a number of engineering cycles.
In Phase II when we, alongside Professor Trevor Coward, attempted to print the scaffold using this design, it became apparent that the gyroid unit cell was too complex to print due to two factors. Firstly, our chosen software was unable to slice the mesh design sufficiently since it attempted to create too many faces that required unavailable computational power. Secondly, there were no available printers of a high enough resolution to print this design accurately. This meant that we were not able to print our scaffold using a gyroid unit cell design, which led us to explore alternatives.
We contacted Dr Jia An from Singapore Centre for 3D Printing to discuss the issues we were facing with slicing and printing our current gyroid unit cell design. He highlighted specific aspects of our proposed fabrication process that could be altered slightly to result in a successful print.
We recognised that we would have to either alter the method of introducing pores in our scaffold or amend the pore structure to be simpler in order to allow for a successful print. We researched the different methods of fabricating porous scaffolds — salt leaching and gas-forming for example — and evaluated the pros and cons of each method to identify if there was a better way to achieve the result we wanted. A key drawback of these methods, as opposed to 3D printing, is that there is a lack of control over the pore distribution and size (Loh and Choong, 2013), hence affecting reproducibility. As the location, size, and homogeneity of our pores are important to our scaffold design, we were able to conclude that 3D printing our scaffold would be the most suitable option for our proposed treatment.
Regarding 3D printing methods, Dr An proposed that selective laser sintering (SLS), may have a higher chance of resulting in a successful print for more complex microscale geometries than fused deposition modelling (FDM). Further, SLS has the added benefit of not requiring support material.
Taking Dr An’s advice, we researched the SLS printing method, starting with the literature he recommended to us. We learned that the SLS method had many benefits; for example, the printing method would naturally create micropores throughout the scaffold (Sudarmadji, 2010), which would increase the interconnectivity to further promote axonal regrowth. Despite these benefits, the post-processing of the smaller pores would be difficult due to the likeliness of excess powder remaining in the printed scaffold (May, 2006). We also were conscious that selective laser sintering printers would be more expensive (Choudhari and Patil, 2016) and therefore not as accessible as fused deposition modelling printers.
To further consolidate our knowledge of scaffold micro-architecture, we contacted more researchers. The author of a paper regarding mathematical modelling suggested the complexity and linearity of our pore structure were possible issues. Specifically, he postulated that the linearity of our pore structure may cause axonal bundling at the entry point of the scaffold pores, which would prevent a significant percentage of axons from regrowing beyond the scaffold and bridging the gap in the spinal cord. To explore this notion, we conducted a literature review to learn about the effect of having a linear pore structure in our scaffold and its effect on axonal regrowth and regeneration. From this, we learnt that Ehsanipour et al. (2019) discovered that a linear-pore scaffold had a packing density that was 10% lower than its non-linear counterpart. This meant the scaffold with a linear pore structure had more space available for the respective cells to enter and regrow. Literature also suggested if pore structures were nonuniform, it could result in regenerating tissues receiving an uneven distribution of nutrients (Choi et al., 2010), therefore, resulting in a nonhomogeneous regenerated tissue. This would be detrimental to the functionality of the spinal cord post-implantation and bridging. We were able to conclude that our implementation of a linear and homogeneous pore structure into our scaffold design would be vital for our treatment to have its desired effect.
To decide on our pore structure redesign, we considered what we learnt from the experts that we contacted. Specifically, Dr An mentioned that convex pore structures are easier to print than non-convex ones. Professor Trevor also presented us with a few pore structures that may be more successful during the scaffold fabrication process. As such, we set about devising a new design for the scaffold micro-architecture. After performing extensive literature reviews and consulting Dr Lorenzo Veschini, we determined that a simple log-based cross-hatch lattice could provide an appropriate solution due to its printability and easy control over dimensions, mechanical properties, and porosity (Kelly et al., 2018).
In Phase II, we decided to further investigate the macro-architecture of the scaffold to ensure that our therapy is optimal. Therefore, following research into printing methods, we discovered that printability is an important factor to consider. Hence we decided to, again, compare the two open-path designs. We concluded that an open-path design without a core would be a much simpler design to print due to greater leniency regarding printer resolution. Further, Wong et al. (2008), discovered that an open-path with core design slightly outperformed that without, thus we decided to prioritise the printability of the scaffold. A simpler design such as this improves accessibility to our therapy in regions where high-resolution printers may not be easily available. In order to test this macro-architecture design, we 3D printed the scaffold using Fused Deposition Modelling (FDM).
First, we printed a test scaffold in PLA to determine whether the architecture would be able to slice in pre-processing and printed with the available FDM printer resolution limitations. We used PLA because it is a reliable plastic that is more widely used for 3D printing, since we wanted to understand whether the architecture is feasible in the first place we wanted to work only with 1 variable. After a successful print, we confirmed that our new geometry file is working and compatible with printer resolution. Then we moved on to producing a PCL scaffold
PCL ‘logs’ are printed vertically with a 500 μm gap between each log, the printer is then rotated 90° and prints the same logs again, thus creating an intersecting lattice of parallel lines. Given the complexity of 3D printing, there are various adjustable printer parameters that alter the quality of the scaffold. Initially, the PCL material was melting at a temperature much lower than expected, resulting in cooling issues. As each layer was being printed, it would not cool sufficiently, causing adjacent layers fusing together. The printer’s fans were switched to 100% and the printing speed was decreased to 6 mm/s. Thus the decreased printing speed allowed the filament to cool down with the trade-off of a longer printing time. Through a trial and error approach, we identified that a print temperature of 85°C and a bed temperature of 45°C ensured consistent and replicable results.
With the aim of gaining further insight into and testing our scaffold macro- and micro-architecture computationally, we began research into the different types of modelling we could implement into our project. Upon carrying out literature reviews into the different models available to quantify axonal regrowth or the effect our treatment would have on the environment of the spinal cord post-implantation of the scaffold. The axonal regrowth model is a mathematical model that would be implemented on MATLAB to show the growth path of regenerating axons along our scaffold in a post-SCI environment. It requires the use of three different numerical methods (Lattice Boltzmann method, central difference and interpolation scheme and Euler’s method) on a number of different equations to result in the growth path of each regenerating axon (Zhu et al, 2018). However, when discussing this model with Dr Jack Lee, it became apparent that the model was not feasible as outlined in the paper; a surface area equation was required for the scaffold, which was not possible considering our scaffold’s complex geometry. Therefore, we decided to explore other avenues of modelling, namely Computational Fluid Dynamics.
Subsequent to our discussion with Dr Jack Lee, we decided to implement a Computational Fluid Dynamics (CFD) simulation to assess the permeability of the scaffold and the wall shear stress that the scaffold endures from the cerebrospinal fluid. Here, Dr Lee suggested we use ANSYS to perform our simulations, as it is a widely available commercial software and has less of a learning curve than other CFD software.
We then discussed our model and its aims with Dr Bryn Martin, who suggested that we make a number of assumptions and simplify our model as much as possible. Additionally, Professor Sundarajan Madihally gave us technical advice regarding running CFD simulations; helping us identify how to choose the appropriate values and settings to generate a mesh suitable to both our scaffold and modelling aims. Finally, we researched methods to reduce the simulation’s complexity (such as symmetry and reducing the length of the scaffold) and hence, we were able to create an efficient design, at a lower computational expense.
Following expert guidance, we decided to use ANSYS Fluent Student to model our system. Upon the import of our scaffold into SpaceClaim via the Geometry module of the Fluid Flow (Fluent) system as an STL file, we were met with many errors in the geometry. These errors were unresolvable utilising the repair mechanism on SpaceClaim which meant the scaffold had to be refined pre-import into the ANSYS software. To do this, we had to fix ‘dirty geometry’, a term referring to geometries that are flawed. We decided to repair these errors on Meshmixer by altering the mesh density and offset distances between faces. We also convert our scaffold file format from STL to STEP on Fusion 360 as we found that this file format saves without diminishing the accuracy of the geometry (O’Connell, 2020). Following these refinements, we reimported the scaffold file into SpaceClaim and repaired any final geometry errors.
When attempting to mesh the whole length of our scaffold, we found that the meshing process would never complete running. We hypothesised that this may be due to the process being too computationally expensive for the computing power available to us. To resolve this we decided to only model an approximately quarter length of our scaffold which went along with advice from Dr Bryn Martin and Professor Sundarajan Madihally, to simplify simulations as much as possible. We also confirmed this decision would not affect our results as we were working out the permeability coefficient using Darcy’s Law, which requires the input of the length of what we are modelling.
We also found that the Mesh module had difficulty processing our cylindrical log structure as the cross-hatched nature of the geometry meant that there was only one small point of contact between each log and another. Making our scaffold logs cuboidal instead resolved this as it created larger points of contact between the logs making it easier to mesh the geometry as a whole.
When meshing was successful, depending on how fine the mesh was the Setup module would reject our generated mesh due to licensing issues. As we were using ANSYS’ student package we had a limit of 512,000 cells or nodes in any mesh we generated. This meant we had to find the optimal mesh size through trial and error which would be able to be accepted by the Setup module while not compromising accuracy. We found that this was a mesh element size of 0.0015 m.
When running our initial simulation, it was apparent that many of the pathlines were reversing back into the scaffold, or not flowing through the scaffold at all. Furthermore, the permeability of the scaffold was unexpectedly high. Therefore, we decided to reassess our boundary conditions and geometry.
Following a review of the literature we had previously looked at, we obtained additional values which we could use as boundary conditions. We found a value for inlet velocity which prompted us to change our inlet from being pressure-reliant to velocity-reliant. We also added the effect of gravitational acceleration on the fluid flow to our simulations.
Concerning the geometry of the model, we also decided to reduce the size of the enclosure surrounding the scaffold (i.e. the area representing the spinal cord). This is because the flow appeared to avoid the scaffold and traverse through the empty surrounding space, as can be expected (being the path of least resistance), which positively skewed the permeability value significantly. Conversely, there was a trade-off between the tightness of the enclosure to the scaffold, and its ability to produce a viable mesh. Through trial-and-error, we manually selected the smallest distance value between the scaffold and enclosure.
Once the aforementioned boundary conditions were incorporated into the simulation we were able to successfully run a simulation with fluid pathlines flowing through the scaffold without reversing back. Our wall shear stress values that were taken from this simulation were within the expected range. To confirm that our results were mesh-independent, we generated meshes with varying element sizes and reran the simulations with identical boundary conditions. Using the pressure drop values from these simulations, we were able to calculate the permeabilities resultant from each one. These values all fell within a 10-8 range which validated our results as per Dr Bryn Martin's suggestion.
Building off of our research from Phase I, we conducted another literature review to learn more about the use of Mussel Foot Proteins as bioadhesives in a medical setting. We continued researching and learning about our chosen Mussel Foot Protein, PVFP-5, from the species Perna viridis. PVFP-5 is one of the very first proteins to be secreted by the mussel foot during formation of the mussel byssus, and contains a disproportionate number of tyrosine residues (24) (Guerette et al., 2013) which are the primary contributors in the adhesion mechanism, through their post-translational modification (hydroxylation) into L-3,4-dihydroxyphenylalanine (L-DOPA) (Ahn, 2017).
We looked into the various advantageous biomedical properties of PVFP-5 which include non-cytotoxicity, promotion of cell proliferation and elevated adhesive strength (Santonocito et al., 2019) We also continued learning about the main source of adhesion in our class of proteins, L-DOPA.
After a brief collaboration with previous team members of the Shenzhen_SFLS 2019 iGEM team, we designed a method in which we would express PVFP-5 as a chimeric protein, after fusion with MaSp1 (from Nephila clavipes) to increase its solubility (Aich et al., 2018). Building upon these findings, we contacted Dr Byeongseon Yang, a postdoctoral researcher at the University of Basel. Dr Yang has significant experience in the use of mussel-based adhesives and mussel protein-fusions, and therefore advised us that while a PVFP-5-MaSp1 fusion will increase the solubility of our protein, it will dramatically increase its degradation rate.
We learnt from this feedback and made the decision to design and build a PVFP-5 construct without a fusion partner.
We updated our PVFP-5 DNA sequence from Phase I by designing a new one based on UniProtKB: U5Y3S6 (Guerette et al., 2013). This sequence contains 7 additional tyrosine residues, which are post-translationally modified into L-DOPA, and therefore increase the adhesive potential of our protein. A Lac regulatory system, an N-terminal 6xHis tag and TEV Cleavage site were added to this new sequence to aid in protein expression and purification. It was also codon optimised for expression in E. coli K12. Subsequently, we built and synthesised this sequence as a new composite part (BBa_K3794001) through IDT.
Following the update in our sequence for PVFP-5, we built a new structural model to characterise our protein and also provide a deeper understanding of how best to navigate our wet-lab work. This structural model was built using our structural modelling method from Phase I, which involved homology modelling through the Phyre2 web application (Kelley et al., 2015), and a molecular dynamics simulation using GROMACS. We tested our model to confirm it contained the correct number of disulphide bonds we had predicted.
After learning not all our predicted disulphide bonds had formed, we contacted Dr Andrew Beavil regarding improvements to our modelling workflow. Dr Beavil suggested that we individually model the three EGF-like domains our protein structure is composed of through homology modelling to generate all disulphide bonds correctly, and then to subsequently run a molecular dynamics simulation on the overall structure after bringing the domains together.
After building a new modelling pipeline and testing our method, we were able to build a new structural model that contained all predicted disulphide bonds.
We built our PVFP-5 expression system after conducting PCR on our DNA inserts of BBa_K3794001, and ligating our composite part (BBa_K3794001) into plasmid pSB1A3. Our expression system consisted of expressing our recombinant protein using both BL21 (DE3) and Rosetta-gami B (DE3) competent cells at 37°C and induction with IPTG.
We tested the success of our protein expression from BBa_K3794001 by conducting an SDS-PAGE gel analysis using 1mL aliquots from our expression medium as well as an SDS-PAGE gel analysis following purification of our protein using a Ni-NTA Spin Column (Qiagen). We learnt that a large volume of our protein derived from BBa_K3794001 from both cell lines was aggregating within the cell-lysate pellet and adhering to other proteins in the culture due to its intrinsic adhesive ability, and therefore our total yield of purified protein was very low. We also observed that there were no clear and distinct bands of PVFP-5 in our SDS-PAGE analysis, therefore we learnt we need to re-express our protein under different conditions.
We conducted further literature searches to review the best conditions for expression and purification in order to maximise the yield of our purified protein from BBa_K3794001. We spoke to Dr Paul Brown regarding our protein expression and purification system, who advised us to re-express our construct at a lower temperature with auto-induction media. We learnt this advice was given to encourage sufficient time for our protein to fold in culture, as well as increase its concentration in the soluble fraction of the cell-lysate.
Primary literature regarding the expression and purification of PVFP-5 and its analogues, have expressed the proteins in inclusion bodies, then conducted denaturing/unfolding steps during purification to extract it from the insoluble cell fraction, before re-folding it into its native conformation. Throughout our engineering cycles and project, we have attempted to express PVFP-5 derived from BBa_K3794001 directly into its soluble form, to prevent the need for unfolding, and refolding. To achieve this throughout our engineering cycles, we have experimented with various E. coli cell lines which are optimised for disulphide bond formation, to test if this would increase the yield of our soluble protein.
We designed and built a new expression system using competent Rosetta-gami B (DE3) and SHuffle (DE3) to express PVFP-5 from our composite part (BBa_K3794001). We induced our protein expression at 21°C for Rosetta-gami B (DE3) cells, however continued to use a temperature of 37°C for SHuffle (DE3) cells. Despite Dr Brown’s advice on protein expression at a lower temperature, issues with our incubator’s cooling system meant this was not entirely feasible. We also continued with the use of IPTG-induction, rather than auto-induction as we had seen no significant issue within our previous method of induction.
We tested the results of our PVFP-5 expression from SHuffle (DE3) cells by conducting an SDS-PAGE analysis with samples taken during various stages of the protein induction. We learnt that expression of PVFP-5 in SHuffle (DE3) resulted in a slightly improved yield of soluble protein (produced from BBa_K3794001).
In the future, after synthesising sufficient PVFP-5, we aim to build a visual binding model of PVFP-5 to a single unit of polycaprolactone (PCL) - the biomaterial that our scaffold is made out of. We recognised the importance of proving adhesion between our protein and the scaffold we designed for the success of our project. After building our model, we predicted that the catechol moiety of the L-DOPA residue in the protein sequence would hydrogen bond with the ester group of the PCL. To test this prediction we reached out to Professor Herbert Waite, who validated this. We designed a visualisation of this binding using PyMol and documented our findings on a video as there is no existing literature regarding PVFP-5 / PCL binding currently.
After learning our prediction model was incomplete through Dr Sarah Barry, we built and designed a new binding model between PVFP-5 and PCL, which involves hydrophobic interactions between the non-polar regions of our polypeptide, and the straight aliphatic chain of PCL - as well as the hydrogen bonding that had been previously predicted.
In Phase I, we characterised the inhibitory properties of the SCI microenvironment (specifically the syringomyelia) to better understand the differences between a healthy and injured spinal cord. From this research, we began looking at the targets for various biological agents across the glial scar and local environment. After identifying multiple biological agents, namely growth factors, enzymes, and IL-6, we looked into prior literature evidence of their implementation within a plethora of studies and trials. Eventually, ChABC was chosen because of the wealth of information available online - relative to the other groups - regarding its benefits for functional recovery in spinal cord injury (Lee, McKeon and Bellamkonda, 2009).
The first component we had to decide was the way in which it would be most optimal to administer ChABC. We looked into 5 main methods: hydrogel, PCL scaffold as a drug delivery system, nanoparticles of various materials, microinjections with hydrogel, and microinjections alone. After much deliberation and conversing with Dr Jerry Silver, we decided to stick with the very last option where we would insert the MFP-bioadhesive-coated PCL scaffold via surgery and then inject ChABC at the rostral and caudal ends. Microinjections are minimally invasive, would not induce a secondary inflammatory response, and are compatible with the rest of our proposed therapy. This was corroborated by supplementary sifting through literature (Tom and Houlé, 2008). Moreover, there is limited literature evidence on how ChABC-loaded hydrogels would interact with MFP and/or PCL making number 5 the most feasible route. Dr Jerry Silver also recommended that we make the scaffold slightly larger than the lesion to ensure that axonal regeneration occurs across it.
The next hurdle came with the fact that ChABC is not stable at physiological temperatures. Again, we turned to literature and found multifarious methods which could help increase its stability: trehalose, gene therapy, self-assembling peptides, fibrin gel, aromatic ring fusion, and computational mutagenesis. In terms of feasibility within the context of the iGEM competition, we chose computational mutagenesis - specifically mutations from a paper by Dr. Marian Hettiaratchi (Hettiaratchi et al., 2020). After choosing the mutations, there were 250 wild-type ChABC sequences found within the protein database and selecting the most ideal sequence involved finding the one closest to conserved residues of all 250.
To do this, we utilised EMBL Clustal Omega software which performed a multiple sequence alignment of all 250 sequences. Visualisation of the results and identification of the conserved sequence was achieved through JalView. We then conducted a BLAST of the resultant conserved sequence against all sequences to find the closest match. Towards ensuring that there was functional significance when choosing one sequence over another, we compiled the functional domains of all 250 sequences into a Google Sheets and compared the reviewed protein. The ideal sequence that we ended up using has the PDB code 1HN0.
After identifying the correct mutations, we then built and thermostabilised - through computational mutagenesis - our chosen ChABC sequence (RCSB PDB: 1HN0). Consequently, we pinpointed an avenue for collaboration with another iGEM team as Phystech Moscow kindly agreed to test our mutated ChABC in silico using the GROMACS software which produced promising results.
Towards validating the in silico model of thermostabilised ChABC, we expressed the thermostabilised ChABC (BBa_K3794005) in BL21(DE3) cells using IPTG-induction at 37°C. The cell-lysate was then purified using an Ni-NTA spin column under native conditions. Protein elutions resulting from the purification were tested using an SDS-PAGE analysis, which indicated low purified protein yield. Following steps to improve our purification steps, we loaded non-heated samples onto our SDS-PAGE gels, which improved the yield of our eluted ChABC. We confirmed the presence of thermostabilised ChABC by the use of Western Blotting. Towards improving our overall soluble protein yield, we re-expressed thermostabilised ChABC (BBa_K3794005) using BL21 E. coli. We conducted SDS-PAGE analysis following re-expression, and our subsequent re-purification. Both analyses showed an increase in yield of ChABC, however it largely remained insoluble.
Having slightly improved our yield of soluble ChABC (from BBa_K3794005). We concentrated our protein elutions, and used a buffer exchange reaction to prepare them for an enzymatic assay. To demonstrate functionality of our recombinantly expressed ChABC (BBa_K3794005), we conducted an enzymatic assay at 37°C after 10 minutes of incubation of thermostabilised ChABC (0.014mg/mL) with a Chondroitin Sulfate sodium salt substrate (3mg/mL). Results from this enzymatic assay demonstrated BBa_K3794005 activity by showing an increase in absorbance at 232nm.
Onto surgical considerations, we looked at different parts of the spinal cord and re-evaluated the initial target of solely cervical spinal cord lesions. After conversing with Dr. Jerry Silver, he told us that this location would not be ideal due to its correspondence with hand function and vital respiratory function. Accidentally injuring any of the remaining axonal tracts during scaffold implantation could be eradicating any residual function for patients that are already distressed with their present physical state. Consequently, he directed our attention to the thoracic vertebral region where we could more safely insert the scaffold without any undue consequences. To corroborate this information, we also got into contact with Mr. Gordon Grahovac who informed us that C6 - C7 would be avoiding this ever-important motor phrenic nerve pool. Furthermore, both individuals encouraged us to centre around patients with complete SCIs (AIS A) to prevent any detrimental effects during our preliminary stages of clinical trials.
In terms of the timeline, we have identified that chronic phase SCI is when the lesion has stabilised and is no longer expanding or growing. This makes it the most conducive stage for diagnostic imaging and segmentation towards enabling individualisation of the SCI scaffold for each patient. Hence, we have decided to treat patients with chronic and complete SCIs anywhere from C6 - T12.
For the procedure itself, it is best to resect the glial scar to allow for the scaffold to be inserted and we have confirmed with both Dr. Jerry Silver and Dr. Elizabeth Bradbury that this would be an ideal approach. The scar constitutes a large majority of the inhibitory environment as chondroitin sulphate proteoglycans (CSPGs) are highly upregulated within this region so effectiveness of our therapy would be greatly impacted should we decide not to remove the scar.
The final obstacle that we came across during literature review, when tying up any and all loose ends, was the fact that ChABC degrades the glycosaminoglycan side chains of CSPGs. These GAG chains have to be removed so that the environment can become more growth-permissive however they are also a component of the extracellular matrix that MFP binds to. After doing further research, we discovered that there are ‘stubs’ of the GAG chains left behind after ChABC digestion (Crespo et al., 2007) and theorised that this would be sufficient for the MFP to adhere to. Upon contacting Professor James Fawcett, he verified that these ‘stubs’ have a similar chemical make-up and structure as the fully intact GAG chains and would consequently be enough for MFP adhesion.
Overall, we have decided to implement a combinatorial therapy with ChABC for patients with chronic spinal cord injuries (minimum 4 weeks after injury) from C6 - T12. Each scaffold will be personalised to the respective patient and will be covered in the MFP bioadhesive prior to implantation. Each surgery will commence with a glial resection once the site of the lesion has been exposed, the scaffold set in place and ChABC microinjections consequently administered in the rostral and caudal ends of the lesion. To potentiate recovery of the patient, we will follow up with rehabilitative protocols and this collectively synergistic approach will hopefully lead to functional recovery of all patients under Renervate Therapeutics.