In order to gain trust from the population, the technology should belong to everyone. In a way similar to the open-source software industry, people should be able to improve our project or create their own versions of it. This idea of openness is very common among the community of synthetic biologists, but a lot of pitfalls have to be overcome to make it a sustainable reality.

A parallel could be drawn with electronics in the 1960's, when computer programming was extremely low-level and belonged to the realm of academia. Since then, it has reached a way wider population, thanks to the creation of frameworks allowing for abstraction of the most technical parts. How could the same principles be applied to synthetic biology, in the context of metabolic engineering and vitamins production?

Even though a lot of lab strains designed for easier modification have been designed in the past, they usually have a very general purpose and biotechnology remains a matter of specialists where every modification has to be made from scratch. We imagined a repurposed organism made especially for the quick construction of these self-replicative tiny factories, that could be easily used by startups, community labs or just by enthusiasts. In the following section we discuss the constraints associated with it, and what such an organism could look like.

For a biological product to leave the benches and actually reach the population, it's essential to foresee its life in the hands of the people who will cultivate it and make sure it stays alive all along. Our design must therefore provide strategies to create an durable, usable product. On paper, the plan is simple: the manufacturers grow the micro-organism, distribute it and save a little fraction to start a new culture with. This could in principle last forever, but in reality the universal rules of biology soon kick back in.

Let's consider the following scenario: a wild type organism sneaks into the incubator and starts to replicate along with the engineered organism. Our microbe cannot compete: this contaminant has been selected precisely for its ability to sneak into environments and replicate, during hundreds of years, while our microbe has the burden of producing tons of enzymes to make the precious vitamins. Additionally, unnatural proteins and metabolites can have toxic effects when their production rate is high. After a couple of growth cycle, the worst seems unavoidable: the micro-organism that will be distributed will not be the right one. Not only this one doesn't produce nutrients, but it might not ferment the rice well or even be pathogenic.

These contamination events bring a lot of hassle for the manufacturer, so our design must provide solutions for making them as rare as possible.

Our approaches is based on two strategies:

• Reducing the fitness burden: To make our micro-organism more resistant to contamination, we need to design it so our modifications come with a minimal fitness cost.
• Identifying the contamination: If a contamination occurs, it is essential that it does not go unnoticed. Our design must allow the manufacturer to detect contamination, and check that what he is growing is exactly what he wants to grow.

The quality control of our system is possible thanks to an ID gene, which consists in the fluorescent protein mCherry expressed to low levels. We chose this protein because its fluorescence colour is easily distinguishable from the media that will be used for growth, hence a better signal.

The aim of this gene is to provide a quick and reliable way to determine whether the strain that will be distributed is the one intended. After growing the micro-organism in a bioreactor, a sample is taken in order to start a new culture from it. We suggest that, while doing so, the sample has to pass a quality check where its fluorescence is measured.

If the sample displays fluorescence, the culture is sent for packaging and a new culture can be launched. If the fluorescence in not sufficient, the sample is discarded and a new blister stock of the original strain is used to start the culture.

This fluorescence measurement is a good example of real-life use for the DIλ spectrophotometer, a low-budget device that is developed by our neighbours at the Openlab. Heavy development is currently ongoing to make it capable of fluorescence measurements.

Our differentiation system is inspired by the Brainbow system, initially developed for tracking the axons of neurons in mammalian's brain. We modified it so it becomes extendable (A).

This system is randomized on a single-cell level, so each cell produce one —and only one—, vitamin pathway. In most research work, metabolic engineering has been done only one target compound at a time, and little is known about what happens when production pathways are used simultaneously in the same cell (B).
Having one cell expressing only one pathway should theoretically preclude unexpected interactions between different pathways, thus making an extendable framework where every synthesis function is decoupled (C).
The different vitamin-producing pathways can be prototyped separately on a classical lab strain, and it is then easy to put them all together in the same chassis for a multi-functional organism.

Starting from this chassis, up to four metabolic pathways can be added by using the attB sequence as a landing pad. Like the Lox sites, this sequence comes from a bacteriophage: the PhiC31 phage uses it to integrate itself in the genome of the host. To insert a new sequence in this landing pad, all you need to do is build a plasmid with the matching "attP" site and express the PhiC31 integrase.

• attB: GTGCGGGTGCCAGGGCGTGCCCTTGGGCTCCCCGGGCGCGTACTCCA
• attP: AGTGCCCCAACTGGGGTAACCTTTGAGTTCTCTCAGTTGGGGGCGT

In summary, a new gene to be added in the system should have the following standard structure (A):

• An attP sequence different from the one that was used just before,
• A Lox sequence (Lox sequences should be added in the same order they come in the Lox Array),
• The operon to be expressed,
• An attB sequence, orthogonal to the attP used for integration,
• A selection system (not depicted here for clarity).

• GCGGGTGCCAGGGCGTGCCCTTGGGCTCCC for killing cells who have not integrated anything in the first attB version,
• GCGGGTGCCAGGGCGTGCCCCCGGGCTCCC, for the second attB version.

Now that the different genes have been added to the chassis, it is time to see it in action.

The CRE recombinase will cut the LoxP sites in the middle, remove the region in-between, and join the two remaining halves of LoxP sites together (Nagy 2000). This only occurs if the overlap sequence are exactly identical (Missirlis 2006). This means that, in the picture of the right, only LoxP sites of same colour would recombine. Given the configuration of this system, any LoxP recombination event would result in the loss of several other LoxP site, in a such way that further recombinations are not possible. The pair of LoxP sites that undergo recombination is therefore chosen randomly by each cell.

Depending on which region is excised, one random coding region settles next to the promoter and starts to be expressed. For a chassis containing four different operons, the mother cells differentiates in four different daughter cells, each of them expressing one operon.

Even if the chassis is not completely filled, it still works: the number of different daughter cells is always equal to the number of inserted cassettes, and the probability of each is adjusted accordingly.

Chromosomal integration

This cassette was constructed by gene synthesis and Gibson assembly and assembled in a self-integrating plasmid vector (Saint-Pierre, 2013). This vector uses the integrase of the phage HK022 to integrate itself in E. coli's chromosome. This plasmid was electroporated in the bacteria and the HK022 integrase was induced.

To check that the cassette has correctly been integrated in the right locus, we performed an analytical PCR on the whole genome of the transformants, with a set of four primers which allows for amplification of the junction between the vector and the chromosome.

Here is presented the result of this PCR using the genome of six clones of transformants as a template.

It tells us that the vector was succesfully integrated at the right locus. It also shows that there have been only one integration and no tandem integrations, which would have resulted in an additional band.

The Lox array was the most difficult region to construct. As it is a very repetitive sequence with numerous dyad repeats, it is tedious to synthesize, amplify and assemble. That's why we created biobrick BBa_K1678005. It contains the promoter followed by the four orthogonal Lox sites. We sequenced this biobrick to confirm that it contains no mutation.

We characterized this new biobrick's function by assembling it in pSB1C3 with the part BBa_K516030 which contains a RBS, the mRFP coding sequence and a double terminator. For comparison, the biobrick BBa_J23119 was assembled with the same mRFP cassette on the same vector.

As in prokaryotes the 30S subunit of the ribosome binds directly to the RBS, the LoxP array does not theoretically interfere with translation. It can however interfere with the transcription.
During the transcription, the RNA polymerase has to go through the LoxP array, which is made of repetitive sequences that are likely to form a hairpin. We show that this has an impact on the transcription efficiency (Mann-Whitney-Wilcoxon test, p-value < 10^-6). However, it still allows for strong protein expression as the average expression level was equal to 91% of the expression level of the BBa_J23119 promoter. The fraction of RNA polymerases that go through the Lox array should be more than enough for our design.

We then aimed to trigger the differentiation of our newly constructed strain. A strain carrying the plasmid pFHC2938, that allows for expression of the CRE recombinase upon arabinose induction (Nielsen 2006), was aquired.
Unfortunately, all our attempts at transforming our strain with it have been unsuccesful, even when resorting to very efficient techniques such as electroporation. It always lead to either nothing, or a lawn of bacteria that did not seem to carry the antibiotic resistance using for selecting the plasmid. To troubleshoot this transformation, we made three hypothesis:

• Is there something in the cell that interferes transformation?
• Is the strain not suitable for CRE expression, e.g. there are LoxP sites somewhere that result in deletions in the genome?
• Is the plasmid the wrong one?

To figure this out, we transformed our strain with the CRE recombinase plasmid along with a standard pSB1C3-mRFP plasmid. The transformation of the CRE-recombinase gave a lawn, while the control plasmid gave clear colonies. We then performed the same 4-primer PCR that was used to check for integration on the bacteria present on the plates. The bacteria transformed with pSB1C3-mRFP still contained the integrated cassette, while the bacteria on the pFHC2938 plate did not display any band, meaning that they were contaminants. This means that the transformation process is not the problem.

We then transformed a non-modified strain without our integrated cassete with the pFHC2938 plasmid. At the same time, the integrative plasmid was transformed into another E. coli strain (STBL). None of the transformations yielded any colonies, meaning that the problem came from the plasmid and not from the strain.