We aim to build a pathway for vanillin synthesis in E. coli starting from tyrosine. This involves the enzymes tyrosine ammonia lyase (TAL), coumarate-3-hydroxylase (C3H), caffeic acid O-methyl-transferase (COMT), feruloyl-CoA synthase (FCS), and enoyl-CoA hydratase (ECH). A summary of the different basic and composite Parts developed is given in the table below.

We will make four different constructs (Figure 1 and 2) to test the effectiveness of different microbial and plant genes at different points in the pathway. In all these constructs, the biosynthetic genes are divided into two transcriptional units, both under the control of cumate-inducible promoters (the cumate-responsive regulator CymR and the first promoter are provided by the cloning vector pUS250).

The starting sequences for most of our Parts were pre-existing Parts (see Table 1), with the exception of COMT-6714, for which the sequence was sourced from GenBank. We further improved these Parts as described in the steps below.

1. Removal of restriction sites.
To make the Parts as versatile as possible, we removed all the restriction sites from a set of 45 commonly-used Type II restriction enzymes (see end of file for list) - note that these include the iGEM BioBrick and GoldenGate enzyme sets. This was done by making 1-bp changes that do not lead to changes in the amino acid sequence, i.e. silent mutations. Changes were made to codons with frequencies as similar as possible to the original by consulting codon usage tables

2. Removal of cryptic promoters.
promoters were predicted using the Salis Lab website, and any predicted promoters with strength >2000 were removed (this was an arbitrary cutoff). This was done by making 1 bp changes at the most conserved positions in the promoter (-35 and -10 regions), while maintaining the protein sequence.

3. Addition of RBS.
The extended version of the BBa_B0034 strong RBS (26 bp) was added to all parts, then the strength of all the RBS's with their attached genes was checked with the Salis Lab tool, including 200 bp upstream and downstream mRNA. (Note that in the case of the first gene in the cluster, the upstream sequence was shorter (44 bp) since the first RBS is close to the transcription start point). The sequences were iteratively modified by 1 bp changes until the RBS strengths were standardised at 10000 ± 3000. Results are shown in Figure 2.

4. Golden Gate sites.
BsaI sites were added to each end of the Part, with six unique overhangs to enable assembly of the five-part gene cluster in the pUS250 cloning vector, which already provides two BsaI sites with ends GCCT, TGGC. The remaining overhangs were chosen by the NEBridge GetSet tool, and the complete set of overhangs, in order, was : GCCT, ATTA, CACC, AAGA, CGTC, TGGC

5. Final checks.
All the protein-coding sequences in the constructs were re-checked by BLAST-P analysis to ensure that no mutations had been introduced in the previous steps, and by in-silico restriction analysis to confirm all sites were as expected, before ordering the synthetic DNA Parts.

We performed four differnet Golden Gate reactions with the 5 gene parts and the pUS250 vector, then transformed the recombinant DNA into E. coli DH10B, and then plated on LB-Km.

The first round of constructs was tested solely by visual inspection of the transformants, and then sequencing of several representative clones. . Plasmid DNA was extracted from pmQ3A to D which clones was subjected to Nanopore sequencing to verify the presence and correct sequence of the inserted genes. This step was crucial to ensure that the genes have been accurately cloned and are positioned correctly within the plasmid constructs.

Phenotypic characterisation.
Some transformants, especially pMQ3C, yielded colonies of E. coli producing a brown diffusible pigment. We hypothesise that this results from the C3H enzyme acting on tyrosine instead of its intended substrate p-coumaric acid. C3H is known to have a very broad substrate range [1]. This reaction of C3H on tyrosine would yield L-DOPA, followed by spontaneous polymerization into a melanin-like compound (See Figure 4).

While this brown colour represents an unintended side reaction, it confirms successful expression of at least one of our foreign genes in E. coli, most likely C3H-HpaBC. This observation also indicates that our cumate regulation system is not performing optimally, as these biosynthesis genes should not be active on LB-Km agar without the presence of cumate.

Additionally, we observed that DH10B (pUS250) and RARE (pUS250) exhibited better growth at 30°C compared to 37°C, though this trend was not observed with the biosynthetic plasmids. Interestingly, the brown pigment seemed darker at 30 °C, but colony size remained small at both temperatures.

Genotypic characterisation
Sequencing showed that only one of the clones (pMQ3A_7) matched the expected design perfectly, while all others had either major deletions, or point mutations or both. To try to salvage some useful plasmids from some of the constructs that only had minor errors, we attempted to use PCR and re-ligation to repair these in the next DBTL cycle.

Primers were designed around the mutations in the biosynthetic genes (Figure 1) ; in this diagram, the red marked region is the mutation we are trying to correct.

PCRs to repair the biosynthetic plasmids were done, and where successful, these were purified, self-ligated and transformed E. coli DH10B and then plated on LB-Km.

The second round of constructs was again tested by visual inspection of the transformants and sequencing of several representative clones (as above).

The PCR approach for repairing the plasmids was not successful. In some cases, bands of the expected sizes were seen, but these were weak, and many other bands were present. In other cases, no products of the expected sized were obtained. The restrictions on the primer design in this experiment made it hard to make good primers, which aggravates the problem of having very long PCRs (>10 kb).

We decided to repeat the original GoldenGate reactions, and screen more clones to ensure we could find one despite the presence of deletions and mutations. It seems from our earlier DBTL cycles that using the ‘correct’ cycling conditions for GoldenGate helped to yield correct clones in earlier experiments compared to using the ‘cheat’ method (37C incubation only), so this approach (proper temperature cycling) was done for all further work.

We decided also to test the biochemical activity of some of the correct (or hopefully correct!) constructs via cumate induction in tyrosine-enriched broth, then LC-MS analysis. We also hope to be able to identify the brown diffusible pigment via LC-MS.

As described for cycle 1, we performed Golden Gate reactions with the 5 biosynthetic gene parts and the pUS250 vector, then transformed E. coli DH10B, plating on LB-Km. One difference to the earlier cycle is that some of the parts needed to be reamplified by PCR before cloning because stocks were running low.

The third round of constructs was again tested by visual inspection of the transformants and sequencing of several representative clones (as above).

The MS data from the 3C clone 11 indicates successful production of vanillin at a concentration of 0.2 ppm, as evidenced by the observed peak. However, the peak intensities do not correlate precisely with the quantities present, likely due to variations in detector response. In addition to vanillin, we also detected other intermediate substrates within our pathway, including p-coumaric acid, caffeic acid, ferulic acid, and vanillic acid.

These results demonstrate that the TAL and C3H enzymes are functioning effectively, as indicated by the efficient conversion of tyrosine to caffeic acid. However, the accumulation of caffeic acid suggests that the downstream enzymes COMT, FCS, and ECH may not be operating at optimal efficiency, given the presence of ferulic acid and only some vanillin in the MS data. Moreover, the construct included the HpaBC gene, implying that the C3H enzyme from HpaBC is outperforming the Sam5 C3H enzyme.