Synthetic or Engineering Biology is a field of science that involves redesigning organisms for useful purposes by engineering them to have new abilities. By breaking the problem apart into smaller ones and following iterations of the DBTL cycle (Design, Build, Test, Learn) to focus individually on each one, we can simplify the development of a functional complex biological system. So with the goal of eutrophicated water bioremediation, we engineered a device that could help absorb more nutrients in the presence of eutrophication. Moving down the abstraction hierarchy, we came up with a 2-module system; a detection module and a nutrient uptake module. The first consists of the TetR-KRAB silencing complex, which is under the regulation of a synthetic riboswitch, able to detect the presence of microcystin-LR (biomarker). The second module is a TetR-regulated PHT1 transporter overexpression cassette.

Figure 1. Representation of our synthetic device in the absence (left) and presence (right) of microcystin-LR.

Here we illustrate the characterization of both of these constructs by performing 3 different DBTL cycles, with several iterations.

Figure 2. Iterative Design-Build-Test-Learn (DBTL) cycle of our Engineering Cycles

Our goal was to design a microcystin-LR (MC-LR) responsive, synthetic riboswitch. This is a process that usually involves in silico engineering, in vitro and/or in vivo testing of final molecules. Given the narrow time frame and limited resources we had to conduct experiments, we nonetheless managed to establish the computational design. Despite the scarcity of tools available for RNA riboswitch engineering (compared to protein engineering), we have successfully demonstrated that this pipeline can produce theoretically verified RNA receptor – ligand couples for any small molecule, if an aptamer is at hand.
This work required several iterations of the Design/Build – Test – Learn cycle. We started with the sequences of the plant thiamine pyrophosphate (TPP) riboswitch and the microcystin-LR aptamer, combined them to yield hundreds of random sequences to choose from, screened the candidates according to secondary and tertiary structure formation, 3’ splicing site functionality and successful ligand docking. During this process, at multiple occasions, we had to take a step back to correct and optimize the pipeline, thus ending up with six DBTL cycle iterations.

First Iteration: Studying the TPP riboswitch

Design & Build

To achieve sensing of exogenous signals by plants, one can think of many ways to engineer them. We went with the design of a new riboswitch because a MC-LR targeting aptamer was described in a recent study1. We chose the plant TPP riboswitch as a backbone, as it is the only riboswitch present in plants2. Another advantage was the size of the ligands. TPP (MW of 424, 31) 3 and MC-LR (MW of 995,2) 4 are far more comparable than MC-LR and other small molecules targeted by bacterial riboswitches, like theophylline 5 or adenine 6. To ensure that the secondary and tertiary structure prediction methods we would use could provide as accurate results as possible, we tested them with the sequence of the TPP riboswitch 7 and compared the result with the experimentally determined 3D structure.

Test

The .fasta and .pdb files from RCSB PDB (id: 3D2G) were obtained and the .fasta file was fed into the RNAfold 8 algorithm to produce a secondary structure prediction. The result was given to the RNAcomposer program 9 to calculate the tertiary structure. The final .pdb file was compared with that of 3D2G.

Learn

About 35% (including the part where TPP binds to) of the prediction was identical to 3D2G, with an alignment from RNA-align 10. We now know that a combination of RNAfold and RNAcomposer can produce accurate enough predictions for our purpose.

Figure 3. Comparison of the TPP riboswitch RNAcomposer - determined 3D structure and the experimentally calculated one.

Second Iteration: Designing the new riboswitch – Secondary Structure (First try)

Design & Build

Not only is the TPP riboswitch special in that it is present in eukaryotes, but contrary to other switches, it functions by affecting intron splicing in pre-mRNA 11. When TPP binds to its aptamer, it promotes the formation of a long and unstable mRNA, by causing a poly-A signal to be spliced out of the 3’ UTR. TPP aptamer folding, prior to ligand binding, includes both 5’ and 3’ splicing-site (SS) sequestering elements, by the secondary and tertiary structure respectively, with further complex structural rearrangements occurring upon binding 12. With this level of complexity, splicing regulation through this particular mechanism appears to be far from amenable to rational replication with simple bioinformatic tools. Our approach, then, was confined to the replacement of the TPP aptamer with the MC-LR one, to achieve 3’ (SS) sequestration by its tertiary structure. Similar attempts have been reported in the past with the Tetracycline Repressor (TetR) protein and theophylline aptamers 13, 14 in mammalian cells, but never in plants.

Test

The randomseq.py program 15 was used to calculate multiple random sequences of overhangs for the MC-LR aptamer. These are small sequences that will be flanking the aptamer region inside the riboswitch, to “protect” it from unpredictable structural arrangements. Hence, we paid close attention to this RNA element’s modularity prospects, one of the foundational ideas of Synthetic Biology. The outcome of this algorithm was well over a thousand sequences. These were printed in a .txt file which was used as an input for RNAfold.

randomspythoneq.py MS ‒n=1000 ‒cap_start=False ‒cap_stop=False 
‒fasta=True ‒prefix=’Test’ ‒statement=’v(15,15,False,False,False,False); 
c (GGCGCCAAACAGGACCACCATGACAATTACCCATACCACCTCATTATGCCCCATCTCCGC); 
V (20,20,False,False,False,False)’ 
> results.txt

Learn

The result was a file with pictures of these sequences’ secondary structure. A quick look over them was enough to see that most were obviously problematic and could not be used, because they had no structural elements (like loops and junctions) for the tertiary structure to form.

Figure 4. Examples of problematic secondary structures.

Third Iteration: Designing the new riboswitch – Secondary Structure (Second try)

Design & Build

According to the manual pages of RNAfold, to force constraints on the algorithm, one must manually create a file with the nucleotides that should be paired or unpaired, or create loops etc. To do that, a clear picture of the aptamer’s 3D structure is required. This was calculated with the iFoldRNA web server 16, and the .pdb files were visualized with UCSF Chimera 17. Nucleotides positioned both in the middle of the sequence and inside “pockets” of the structure, were used to apply constraints 18.

Figure 5. The aptamer domains that were used to apply folding constraints. Created with Molstar Viewer.

Test

The algorithm ran again with the constraint file.

RNAfold –constraint[=constraint.txt] > sec_str_corrected.fasta

Learn

Indeed, most secondary structures now looked better and could be used for the next step.

Fourth Iteration: 3’ Splicing Site Functionality test

Design & Build

The .fasta file of the A.thaliana thiC gene 3’ UTR (where the TPP riboswitch naturally occurs) was obtained (NCBI Reference Sequence: NM_001336219.1). It was time to replace the TPP aptamer with the MC-LR aptamer candidates. The whole scheme now should work as a normal intron, with a 5’ SS and a branch point site from the natural sequence, and a 3’ SS from the new aptamer.

Test

To test that, some sequences were randomly selected to look for splicing sites by the NetGene2.0 web server 19. Only one of them yielded a satisfying result. This meant that at least a few of the sequences were appropriate and now they had to be found among the hundreds. A simple bash script utilizing the grep command and regular expressions targeting the 3’ SS consensus sequence for A.thaliana 20, enabled precise selection of the best candidates. NetGene2.0 ran again with these sequences, and all worked as expected.

grep -E ‘Τ(A|G|Τ)CAG(G|Α)(A|Τ)’ results.txt > new_results.txt

Learn

Starting with a thousand sequences, we were now left with 15 that could be used for 3D structure prediction.

Fifth Iteration: Designing the new riboswitch – Tertiary Structure

Design & Build

While secondary structure calculation is a crucial step for the functional characterization of an RNA molecule, it is the three-dimensional structure that determines interactions with ions and ligands inside the cell. For this purpose, the 15 2D structures from the previous step were used as input for RNAcomposer. Our goal was to produce 3D structures with a good RMSD score for the docking step. The structures were also compared with the one predicted with iFoldRNA (see 3rd iteration).

Test

RNAcomposer ran, and the output files were tested with RNAlyser 21, to check for RMSD and similarity with the iFoldRNA predicted .pdb file.

Learn

As a result of this screening, 6 of the structures made it to the final step; docking.

Sixth Iteration: Docking

Design & Build

Our goal in this final step was to obtain a secondary structure, to which MC-LR could bind in a reasonable position. There are very few options for tools to perform in silico RNA-ligand docking experiments. Whereas we could use common docking applications like AutoDock Vina, ZDOCK or DOCK 6.5, these are better suited for protein design. With further research, we found RLDOCK 22 and AnnapuRNA 23. These are RNA specific tools, utilizing scoring functions to predict RNA-small molecule interactions. Because RLDOCK needed too much computational power, something we did not possess (at least 64 GB RAM), we went with AnnapuRNA.

Test

The receptor (RNA) and ligand (MC-LR) PDB files were prepared for the program to run.

./annapurna.py -r cluster1.pdb -l MCLR_ligand.titles.sdf -m kNN_basic -m kNN_modern -o testresults/output  -s --overwrite --groupby --merge --cluster_fraction 1.0 --cluster_cutoff 2.0 --clustering_method AD

Learn

At least half of our sequences bind the ligand in a protected “socket”, created by the aptamer and appear to be suitable for use as a new, MC-LR responsive riboswitch. This would be a 3’ UTR Phytobrick with B6-C1 overhangs 24. However, the functionality of such a system remains to be validated with in vitro and, most importantly, in vivo experiments in plants. We hope that we have inspired future iGEM teams to follow a similar process for synthetic splicing riboswitch design or even test our own.

Figure 6. Cartoon of the aptamer - microcystin docking that was calculated with AnnapuRNA. Created with Molstar Viewer.

Repressors have been widely used in synthetic biology as they allow for a precise control of gene expression, producing a desired phenotype. Amid the field, are repressors that can function properly in plants and amongst the most prevalent ones are the Tet Repressors 25, as was suggested to us by Prof. Kalliope Papadopoulou. The registry of biological parts already provided a standardized part (BBa_C0040) with the coding region for the TetR protein without a Ribosome Binding Site (RBS) so we based our design on this sequence. Before we continued with the TetR design, we needed first to establish the basic parts of our transcriptional unit and test their functionality.

Figure 7. Design of a Transcriptional Unit; Cloning Overhangs, Promoter, 5'UTR, Coding Sequence, 3' UTR, terminator respectively.

First Iteration: Deciding on Regulatory elements

Design

The Reporter System

Starting with the promoter of choice, we opted for the Nopaline Synthase promoter (NOS promoter) as it is a promoter proven to work in plants 26 and a standardized Phytobrick 27 that is weaker than the obvious counterpart (CaMV 35s promoter) 28. Lower expression meant less possibility for toxicity symptoms caused by high expression of TetR 29 and for better responsiveness of the system to tetracycline. Continuing with the reporter system, mVenus 30 was the protein of choice as this is an autofluorescent protein (AFP) (Excitation λ: 515 nm, Emission λ: 527 nm, 31) that is well characterized in Nicotiana benthamiana. We paid careful attention to the Kozak frame of the protein and given that, in green plants, a characteristic motif has been observed $$R_{(-3)} \;M_{(-2)}\; –_{(+1)}\; A_{(+1)} \;U_{(+2)} G_{(+3)} G_{(+4)} C_{(+5)} $$32, we made two substitutions (4:T>C, 5:G>T) to the ORF leading to a mutation (V2A) in the amino acid (aa) sequence, which should not cause folding and activity mishaps due to the position and size of the two aa. Lastly, what remains is the terminator. Settling, again, on Nopaline Synthase (NOS terminator), this is a strong terminator that is a standardized Golden Braid part 33. Putting it all together, we created a promising Reporter System, consisting of pNOS:mVenus:tNOS (BBa_K4213029).

Experimental Design-Plant Selection

While our end-product was already decided to be the plant P. australis, we needed a stepping stone, a model plant to perform the first characterization and optimization of our constructs. For this, we selected Ν. benthamiana since it allows for high and rapid expression of transgenes, by agroinfiltration, as the species is quite susceptible to plant viral vectors 34. N. benthamiana leaf samples are also able to be observed under UV lumination for expression of fluorescent proteins 34.

Experimental Design-Cloning method

Among the prevalent cloning methods used today, one that stands behind its simplicity, effectiveness and rapidness is the Golden Braid method 33. Based on the Golden Gate system, this technology relies on the Type IIS restriction enzymes to alternate between level Alpha and Omega states, resulting in the assembly of one or more Transcriptional Units (TUs) One of the main benefits of Golden Braid is the four nucleotide cutting sequence of the restriction enzymes, which can be completely random, equipping it with the ability to completely fix the order of insertion between different parts by designing different complementary four-nucleotide overhangs at each end. For our assembly, we used the proposed sequences of the technique.

Figure 8. Proposed overhang sequences of the Golden Braid method for the different parts of a TU (A1-C1).

Build

Next came the cloning experiments that lead to the assembly of our Reporter System. The procedure was conducted on a special strain of E. coli, K-12 DH5a, engineered to maximize transformation efficiency. With the help of diagnostic digestion and analysis by gel electrophoresis we were able to confirm the assembly of our construct.

Figure 9. Diagnostic Digestion of pDGB3α2_pNOS-Venus-tNOS with HindIII and expected bands (bp): 6345, 1033 and 509.

Test

As mentioned above, we selected N. benthamiana for our experiments, so what followed was the transient expression of our constructs in planta through agroinfiltration. This method allows for noticeable expression of transgenes within 4 days of infiltration39 and, with the fluorescent mVenus, we planned to observe transgene expression under UV luminescence using fluorescent microscopy.

Figure 10. Final Steps to insert our built constructs and observe their Fluorescence.

Learn

After examining our samples under the confocal microscope and taking our pictures (Figure 11) we concluded that our transcriptional unit did manage to express mVenus in great numbers and, therefore, we moved on to the second iteration of this DBTL cycle, to test our design with the repressor.

Figure 11. Agroinfiltration experiment: positive (pDGB3α2_pNOS-Venus-tNOS) control. The red arrows show intercellular bridges that confirm that Venus protein is in the cytoplasm.

Second Iteration: Establishing Repressor Expression

Design

Having learnt that the above-mentioned syntax works as intended, it was time to substitute the fluorescent protein with our repressor of choice. As already mentioned, the registry gave as the base sequence for TetR (BBa_C0040), but this part was intended for expression in prokaryotic organisms and not eukaryotic ones. For that reason, we needed to anticipate the presence of a nucleus, possible splicing sites and unoptimized expression on account of less frequently used synonymous codons present in the coding sequence (cds).

Tet Repressor for plants

Starting with the first, it seemed that due to the size of TetR (48 kDa) diffusion into the cell nucleus was more than likely 40. Nevertheless, we decided to add a Nuclear Localization Signal (NLS), specifically Simian Virus 40 NLS 41, to the sequence since this would ensure its passage into the nucleus. Moving on to the possible splicing sites, the now NLS-containing sequence was run through a neural network based program called NetGene2. This method is proven to predict possible splicing sites within given sequences, based on analysis made on the plant Arabidopsis thaliana 42. Analysis showed no potential splicing sites within the cds so we moved on with no modifications. Concluding with codon optimization, this is a tool revolving around the frequency that each synonymous codon occurs within a certain organism’s genome. We used IDT’s codon optimization tool for expression in Nicotiana benthamiana.
It is worth pointing out that, for the visualization of expression, we, again, decided to incorporate the modified mVenus into the sequence, leading to a construct consisting of pNOS:Venus:TetR:NLS:tNOS (BBa_K4213030).

Build

Following the same steps as before and making the necessary optimizations to our protocols, needed to properly and efficiently get through the cloning procedures, we managed to confirm the assembly of the level alpha construct using diagnostic digestion (Figure 12).

Figure 12. Diagnostic Digestion of pDGB3α2_pNOS-Venus-TetR-tNOS with HindIII and expected bands (bp): 6345, 1678 and 513.

Test

The testing of the produced construct was, again, done with model plant Nicotiana benthamiana by agroinfiltration. The sample was prepared with a special cell-permeant nuclear counterstain that emits blue fluorescence when bound to dsDNA (Hoechst 33342, Excitation λ: 350 nm, Emission λ: 460 nm) and can be distinguished from mVenus, when observed under both fluorescent microscopy and confocal microscopy, in order to confirm localization of our construct in the nucleus.

Learn

After examining our samples under the confocal microscope and taking our pictures we concluded that our transcriptional unit did manage to express mVenus:TetR:NLS in great numbers and the fused protein was mostly present within the nucleus, confirming the functionality of the NLS (Figure 13).

Figure 13. Agroinfiltration experiment: pNOS-TetR-tNOS. The red arrows show the nuclei, where the Venus-TetR protein is located.

Third Iteration: Upgrading repressing abilities

Design

It was pointed out to us by Prof. Aikaterini Kalliampakou that tetracycline-controlled systems can be quite leaky. After careful literature review, we learnt that ‘leaky’ gene expression is a common and rather persistent problem in synthetic biology and often contributes to the poor performance of a genetic circuit 43. This led us to the investigation and utilization of the Tet toolbox 44.

Improving our repressor with a transregulator

In more detail, we decided to use a part fitter for eukaryotic systems, the TetR-KRAB transregulator. As the name suggests, TetR has been fused with the Krüppel Associated Box protein (KRAB) 45, enhancing the inhibition abilities of the complex by adding another mechanism of silencing 46. Since our design already incorporated TetR, we just needed to add KRAB next to the repressor as this is the supposed working syntax 47. For this we had to remove the already present stop codon by inducing a mutation. Furthermore, after the conducted optimization of the part, we had to domesticate it by inducing a silent mutation, in order to remove an unwanted recognition site for BsaI.
The final syntax contained mVenus as, again, we wanted to spot protein expression with the help of fluorescent microscopy. This led to a final construct containing pNOS:Venus:TetR:NLS:KRAB:tNOS(BBa_K4213033).

Build

After almost three weeks of cloning experiments, we managed to confirm the assembly of the abovementioned construct by diagnostic digestion and visualization through gel electrophoresis.

Figure 14. Diagnostic Digestion of:
1. and 3 “pNOS-Venus-TetR-KRAB-tNOS“ with HindIII (expected bands in bp: 6345, 2056 and 513)
2. and 5 “pNOS-TetR-KRAB-tNOS“ with HindIII (expected bands in bp: 6345, 1342 and 513)
Positive results: 2, 4 and 5 1 MW.

Test

We repeated agroinfiltration in Nicotiana benthamiana leaves and our samples were prepared with the previously mentioned counterstain for better visualization of the nucleus for observation under the confocal microscope.

Learn

With the images presented below, we confirm the expression of our transregulator as well as its localization in the nucleus.

Figure 15. Agroinfiltration experiment: pNOS-Venus-TetR-KRAB-tNOS.

As mentioned above, the purpose of our entire synthetic system is the increased absorption of phosphorus, which will only be induced by the presence of microcystin-LR in the roots of the plant. Now, we come to the second part of our system. This module consists of a transcriptional unit that contains the pht1 gene with a constitutive promoter, whose structure is engineered to be inhibited in the presence of the TetR protein. So, again, we had to start with the basic parts of the system.

First Iteration: Deciding on Regulatory Elements

Design

The Second Reporter System

Beginning with the promoter, we opted for the Cauliflower Mosaic Virus 35S promoter (p35s) as this is arguably the most well-studied and experimentally used regulatory component with activity in plant cells 48. With abundant information available on its individual functional domains, we could incorporate Tet Operator (TetO) sequences within the promoter, placing the second construct under the influence of TetR. Therefore, we introduced three repeats of TetO to p35s and set them in close proximity to the TATA box, based on the design of pTriple Op promoter 49. Given that we planned to test both TetR and TetR-KRAB for their repressive abilities, we also introduced seven direct repeats of the Tet Response Element (TRE) upstream of the distal part of p35s, based on the design of pTight (also known as pTetO7)50. With the distance between the TIS and the modified TRE being under 3.000 bases, this was the best fit for TetR-KRAB’s long range repressive power 50. Continuing on, both the protein and the terminator of choice would be the same as the first reporter system so we ended up with a final construct containing mod-p35s:Venus:tNOS (BBa_K4213041).

Build

After two weeks of cloning experiments, we managed to confirm the assembly of the abovementioned construct by diagnostic digestion and visualization through gel electrophoresis.

Figure 16. Diagnostic Digestion of “TetO7-pTriple-Venus-tNOS(s)” with EcoRI and HindIII and expected bands in bp: 6345, 1829, 674 and 387.

Test

We repeated agroinfiltration in Nicotiana benthamiana leaves but this time we were not able to visualize our constructs in our University’s confocal microscope, due to malfunctions. Fortunately, we were able to measure the fluorescence of our prepared samples with the help of the Plate Reader.

Learn

With the results presented below we confirm the expression mVenus and, therefore, the functionality of our modified p35s. Next step was the expression of the pht1 gene.

Figure 17. Fluorescence of different leaf samples measured under the Plate Reader.

Second Iteration: Establishing Transporter Expression

Design

After the confirmation of expression, we could move on to our phosphorus transporter. We decided to test the genes of PHT1;5 from Arabidopsis thaliana 51 52 and PHT1;6 from Oryza sativa 53 54 Pi transporters due to their evolutionary relatedness of each source organism to our end-product plant (P. australis) and their potential to be overexpressed without showing significant levels of toxicity, leading to even higher Pi uptake regardless of its concentration 55. So after finding the cds of each gene, we did two rounds of modifications. Firstly, we searched for the same Kozak Frame motif around the TIS, leading to two substitutions in the OsPHT1;6 sequence (+5: G>C, +6: A>T) and a mutation in the aa sequence (G2A) which should not cause a problem to the functionality of the transporter due to the size and the place of the two aa. Secondly, we decided to test the effects of codon optimization on these transporters by ordering both codon optimized and not. Sadly, the high complexity of the OsPHT1;6 sequence did not allow us to order it from neither IDT nor TWIST, therefore we partially optimized the spots characterized as “highly complex” by the tool. Finally, as again we planned to visualize expression through fluorescent microscopy, we included a linker sequence and our reference gene (mVenus) downstream of the part. To summarize, our final versions of construct 2 were:

1. mod-p35s:(optimized)AtPHT1;5:Venus:tNOS (BBa_K4213035)
2. mod-p35s:(non-optimized)AtPHT1;5:Venus:tNOS (BBa_K4213036)
3. mod-p35s:(optimized)OsPHT1;6:Venus:tNOS (BBa_K4213040)
4. mod-p35s:(partially-optimized)OsPHT1;6:Venus:tNOS (BBa_K4213043)

Build

Repeating the same steps and after several weeks of cloning, together with sequencing, we managed to confirm the assembly of the abovementioned constructs.

Figure 18. Diagnostic Digestion of:
1. 2 and 3 “TetO7-pTriple-AtPHT1;5 opt-Venus-tNOS(s)” with EcoRI and HindIII (desired bands in bp: 6345, 1978, 1492, 674 and 387)
2. 4 and 5 “TetO7-pTriple-AtPHT1;5 non-Venus-tNOS(s)” with EcoRI and HindIII (desired bands in bp: 6345, 1978, 908, 674, 584 and 387)
3. 6 and 7 “TetO7-pTriple-OsPHT1;6 opt-Venus-tNOS(s)” with EcoRI and HindIII (desired bands in bp: 6345, 2041, 1405, 674 and 387)
4. 8 and 9 “TetO7-pTriple-OsPHT1;6 part-Venus-tNOS(s)” with EcoRI and HindIII (desired bands in bp: 6345, 3449, 674 and 387)
Positive results: 2, 4 and 5 1 MW.

Test

Again, after agroinfiltration and preparation of samples like before, we tried to observe our constructs under the confocal microscope.

Learn

Unfortunately, we encountered a problem with the setup of our confocal microscope and due to the time constraints we were not able to rerun the experiment in order to confirm the expression and localization of the transporters.

After 3 different DBTL cycles, with several iterations, we can present to you our two new and optimized expression platforms for N.benthamiana (BBa_K4213029, BBa_K4213041) and 2 optimized versions of TetR for expression in N.benthamiana (BBa_K4213044, BBa_K4213045).