WET LAB

We spread our engineering strategy such that it addressed the most mechanistically plausible causes for EPS secretion, which has not yet been determined in literature.

• Our first theory was that the non-specific nature of the truncated Hoc1 enzyme results in less branching of glycans and more free mannans in the extracellular space.
• Secondly, a putative reparative cell wall stress response as a result of the thinner cell wall phenotype may drive excess mannans into the Golgi apparatus.
• Finally, we accounted for the scenario where neither of our assumptions were true, in which case we planned to reduce the total mannose available in the cell by inhibiting or downregulating the production of mannose within the cell.

Why we target exopolysaccharides

• Their relevance as an impurity for downstream processing was a primary goal. Through both our HP Magnify Bio and recent literature, we could establish the likely relevance and presence of these impurities in industrial K. phaffii strains. This impurity matters as it complicates downstream processing greatly, exacerbating processing unit steps and thus increasing costs to reach the same purity ^{[10],[11],[12]}.
• This happens because mannans reside in similar mass bands as a lot of proteins of interest like growth factors or dairy proteins such as casein, complicating size-based separation techniques like micro- and ultrafiltration, which are most affordable and scalable ^[12].
• Further, they may agglomerate with some proteins and form protective “cake” layers on filtration membranes, contributing to fouling and reducing yield by trapping proteins ^[12].
• Alternative, higher-effort recovery techniques like anion exchange may not always work due to their reliance on charge-distribution differences of the protein and mannans, which are not reliably present ^[12].
• These complications for downstream processing are only starting to gain attention in literature as there is a recent shift towards using K. phaffii for more affordable, non-therapeutic proteins in the broader bioeconomy ^[10].
• Though there are no full-scale analyses yet, our literature searches and outreach with leading scale-up experts like Magnify Bio and Liberation Bioindustries determined that the cost range for downstream processing, which is determined by the reliability of each unit step (e.g., filtration), can make or break production economics of most proteins.
• This is aside from energetic losses that come along with impurities composing up to 4 % of dry cell weight ^[10],[12].
• To our knowledge, which may be limited due to academia–industry disconnects, there have been no appreciable efforts to resolve this issue on the strain level.
• This means that our goal, while likely aiding production of our target growth factors, could also impact the economics of other proteins.

We decided on one protein target for each plausible theory, which resulted in the three targets we chose to engineer for our experiment.

• Hoc1 — a 1,6-mannosyltransferase enzyme responsible for the branching of long poly-mannose arms on already glycosylated proteins.
• PMI — phosphomannose isomerase, the enzyme responsible for converting fructose-6-phosphate into mannose-6-phosphate, the bridge between glycolysis and GDP-mannose production.
• Rho1 — a key signalling protein in the cell wall integrity pathway whose overexpression can positively or negatively affect cell wall integrity based on basal cell wall integrity.

Since the Hoc1 truncation arose during an industrial directed-evolution experiment and has maintained broad application due to enhanced protein secretion and transformability properties associated with the frameshift, we do not aim to restore full cell wall integrity, but instead combat EPS impurity formation while maintaining high secretion and enzyme specificity by downregulating the enzyme.

However, restoring cell wall integrity and altering glycosylation pathways may carry trade-offs: tighter cell walls could reduce secretion efficiency, and modifying central metabolism could affect growth rate or introduce unforeseen downstream effects. This is why we spread our engineering more broadly.

To modify Hoc1 and PMI, we chose to use the CRISPR/Cas9 system as it represented a relatively straightforward and well-characterised method of gene editing that would allow us to perform all of our desired modifications without major issues. However, we note that going forward, CRISPR may not be the optimal choice for genetic engineering, as the commercial red tape and patents surrounding the technology make it undesirable for strain engineering in scale-up.

We established in silico that split-marker integration would result in a similar effect to that of CRISPR/Cas9 with substantially fewer restrictions on commercial usage. We therefore determined that it would be mechanistically possible to reproduce our intended modifications using this method. Our contribution therefore maintains commercial viability as it does not inherently rely strictly on CRISPR, thus avoids its licensing troubles.

We designed 10 sgRNA devices following a protocol from Gassler et al. that would allow us to efficiently edit the K. phaffii genome and insert our own sequences using homologous recombination ^[1]. These devices have 5′ hammerhead (HH) and 3′ hepatitis delta virus (HDV) ribozymes flanking the sgRNA construct to allow for efficient sgRNA self-excision once the plasmid is inserted. We targeted our sgRNAs around the start and stop codons of the Hoc1 and PMI genes where possible in order to minimise the chance of any unwanted engineering.

Regarding the engineering of Rho1, we chose to overexpress this protein using a weak constitutive promoter, pPGK1, in the hopes that it would upregulate the expression levels to partially repair cell wall integrity, but not so much that it would negatively affect cell growth and proliferation ^[2].

Our growth factor experiment planned to demonstrate the ability of K. phaffii to produce recombinant mammalian growth factors at high titres and to characterise the production of HGF and IGF-1 in K. phaffii for the first time in iGEM.

We chose to target four growth factors based on their roles in muscle growth and tissue regeneration:

• Epidermal Growth Factor (EGF) — stimulates cell proliferation and differentiation^[3].
• Basic Fibroblast Growth Factor (bFGF) — promotes strong proliferation and migration of myocytes^[4].
• Hepatocyte Growth Factor (HGF) — activates muscle satellite cells and supports regeneration^[5].
• Insulin-like Growth Factor 1 (IGF-1) — promotes cell proliferation in muscle and bone^[6].

We engineered these growth factors to be secreted from the cells via the pre-Ost-pro-alpha mating factor fusion secretion tag.

Typically, the alpha-mating factor of S. cerevisiae is commonly used to target recombinant proteins for secretion in K. phaffii. However, when secreting recombinant proteins, the alpha-mating factor runs into two primary issues. Firstly, the signal promotes post-translational translocation into the ER, which can negatively affect the translocation rate of larger proteins like FGF2 and HGF. Furthermore, the signal also can cause aggregation in self-associating proteins, which negatively affects ER export if the production volume is high, like in our case with growth factors.

Barrero et al., 2018, found that by fusing the pre-factor of the Ost1 signal sequence from S. cerevisiae, these issues could be overcome and secretion was several-fold more efficient than the alpha-mating factor in full, up to 20-fold for some proteins [7].

Additionally, a C-terminal poly-histidine (6xHis) tag was fused to the protein using a G4Sx3 flexible linker in order to facilitate Nickel spin column-mediated affinity chromatography for purification of recombinant growth factors. We chose to use His-tag affinity chromatography knowing the potential immunogenic effects for two key reasons:

• We do not plan on transitioning this into a therapeutic, which means there is very likely no severe effect of the His-tag on humans.
• His-tag is the combination of a reliable, affordable, and low-manpower requirement protein purification method, making it quite ideal to isolate and purify recombinant proteins at the lab scale.

Strain Map

The proposed path for our experiments leaves us with several distinct strains of K. phaffii to characterise.

Firstly, we have our CRISPR transformants, which include the downregulated and/or repaired Hoc1 gene and the downregulated PMI gene.

Secondly, we have our Rho1 engineering, which is dependent on the growth factor being produced, where one version of the plasmid for each growth factor has a copy of Rho1 to further increase cell wall integrity, and the other has a copy of mCherry to act as a control, as the proteins are of similar enough size to exhibit a similar burden.

Finally, we have our growth factor engineering, for which we planned to produce all four of the growth factors previously listed. We chose to focus on FGF-2 for our initial engineering as it is a previously characterised growth factor that has a large enough size and enough glycosylation for our modifications to show some sort of effect. We planned to also use HGF for these experiments if the production of HGF without the upregulation of chaperone proteins proved viable in K. phaffii.

We planned to validate the functionality of our growth factors by comparing their ability to proliferate bovine muscle cells against the current commercially used alternatives, foetal bovine serum (FBS) and recombinant growth factors of bacterial or mammalian cell origin (commonly E. coli or CHO). However, we were unable to do this due to time constraints.

Following the production of recombinant growth factors in K. phaffii, we planned an experimental workflow that would allow us to quantify and compare the production of both recombinant proteins, our desired end product, and exopolysaccharides, our side product, from our engineered K. phaffii strains.

We derived a method to approximate the proportion of polysaccharides, most of which we established to be mannans ^[11], by quantifying the total concentration of all sugars and then quantifying the concentration of reducing sugars. The difference between the two concentrations would represent the concentration of polysaccharides, namely, polymannose EPS.

We planned to do this through two colorimetric assays that could be parallelised for high-throughput quantification of sugars:

• The phenol-sulfuric acid (PSA) assay ^[13], which quantifies the concentration of all sugars by first hydrolysing carbohydrates.
• The dinitrosalicylic acid (DNS) assay ^[14], which quantifies the concentration of reducing sugars through a redox reaction with 3,5-dinitrosalicylic acid.

We designed an experiment that would be conducted both in silico and in vitro to determine the usefulness of BioKernel batch Bayesian optimisation (BO) workflow for optimising inducer concentrations to fine‑tune enzyme expression in a metabolic pathway.

To validate this experiment, we designed a framework that would utilise the Marionette‑Wild MG1655 strain of E. coli[8] and the EcoFlex MoClo kit[9] to assemble a plasmid containing ten distinct promoters, each regulating the expression of one enzyme in the astaxanthin bioproduction pathway.

EcoFlex Kit

The EcoFlex MoClo kit is a well‑characterised plasmid kit for the assembly of expression vectors for E. coli, developed at Imperial College London by the Paul Freemont lab. We chose this kit because it was immediately available on campus and uses negative selection markers (lacZ and RFP) to provide quick, colourimetric confirmation of correct assembly.

The kit uses two distinct negative selection strategies: blue/white selection (lacZ) and RFP loss. We used RFP negative-selection vectors at Stage 1, 2, and 3 for this experiment. Cloning was planned so that we would make 10 transcriptional units (one per Stage 1 plasmid), assemble five into each of two Stage 2 plasmids, and then combine those into a single Stage 3 plasmid.

Astaxanthin

We chose astaxanthin as the target because its synthesis is well characterised downstream of the β‑carotene pathway, and its concentration can be colourimetrically determined without cell lysis, offering a practical readout for high‑throughput screening.

The planned protocol used 96‑well plates: transformed Marionette‑Wild E. coli would be incubated overnight and astaxanthin quantified directly by absorbance at 490 nm. We were unfortunately unable to complete this experiment in time due to DNA order delays, but the workflow remains suitable for future validation of BioKernel.