Overview

    We have accomplished Design-Build-Test-Learn (DBTL) cycles in Production Module, Oscillator Module, Directed Evolution Module, Safety Switch Module and Hardware.
    In Production Module, we used DBTL cycle to improve the enzymatic activity of SAMe synthetase opSam2 (BBa_K4144011) and to improve membrane localization of SAMe transporter opPet8p (BBa_K4144004) by designing a new part (BBa_4144010) considering the structural disruption caused by sfGFP.
    In the Oscillator Module, we used the parts of Repressilator (BBa_K3482025) to test their function. And we failed many times but finally made it to observe green fluorescence and explored the suitable way of testing the oscillation.
    In the Directed Evolution Module, to develop a lactose-tolerant LacI protein is our goal. After several rounds of selection, we have gained a colony with expected LacI in a certain way.
    In the Safety Switch Module, we struggled to construct double switches using Holin and EGFP. We met great difficulties during the process, but we successfully got positive results in a part of test experiments in the end.
    In hardware, we improved efficiency of Portable Dual Port Filter based on results of previous experiments. We have characterized several new basic parts in engineering this year: BBa_4144004, BBa_K4144011, and BBa_K4144012.

DBTL cycle of SAMe Production Module

Design of the First Cycle

    Sam2 is a S-adenosyl methionine (SAMe) synthetase from Saccharomyces cerevisiae which won’t be inhibited by its enzymatic reaction product SAMe [1]. We chose it as the enzyme to our produce SAMe in our engineered bacteria. In past iGEM competitions, there are few teams constructing such protein to generate SAMe. They all worked on gram-negative bacteria Escherichia coli (E. coli), while we tend to express such protein in gram-positive bacteria model bacteria Bacillus subtilis (B. subtilis) firstly and then in Bifidobacterium longum (B. longum). Thus, we used a codon optimization tool ExpOptimizer to optimize the coding sequencing, and the optimized sequence is called opSam2 (BBa_K4144011). Then to activate its expression, we utilized a lactose-inducible promoter Pgrac which is endogenous for B. subtilis. We expect that with induction of IPTG, opSam2 can be expressed successfully and catalyze SAMe synthesis in B. subtilis.

Build of the First Cycle

    We used a shuttle vector pHT01K to insert our fragment opSam2 to construct recombined plasmid pHT-opSam2. Through molecular clone procedure, like PCR, enzyme cut, ligation, and transformation, we have successfully transferred our pHT-opSam2 into E. coli and B. subtilis. We have demonstrated the results in Result Page.

Test of the First Cycle

    In order to verify opSam2 expression in B. subtilis, we utilized the 6xHis tag fused to N-terminal of opSam2 to perform Western Blot. Since the opSam2 is downstream of Pgrac, a lactose induced promoter, we used IPTG to induce its expression in E. coli first. To enlarge the expression of opSam2, we choose the induction condition as 1mM IPTG and 37˚c according to the literature. Then, we prepared loading sample at different time after induction. We firstly ran SDS-PAGE and used Coomassie staining to verify the lysis of bacteria and protein input (Fig. 1.1A). Then to compare the induction efficiency and detect opSam2 expression, we used BCA quantification method to measure the concentration of total protein (Fig. 1.1B). Then after the concentration measurement, we diluted samples into almost the same concentration, and used such diluted solution to perform western blot using antibody against His tag (Fig. 1.1C and D). In this case, we can correspond the band brightness of protein of interest to the concentration directly.

Figure. 1.1 Expression verification of opSam2 in E. coli and B. subtilis through WB.
     A. Coomassie staining to verify the lysis efficiency and protein input.
     B. BCA quantification standard curve before WB.
    C. Western Blot in E. coli DH5a. Lane 1-2: Protein pre-stained marker, Lane 3-8: lysate pellet induced for 0h (negative control), 1h, 2h, 4,h, 8h, 24h, respectively.
     D. Western Blot in B. subtilis. Lane 1: Protein marker, lane 2, 3, 4: lysate pellet induced for 0h (negative control), 1h, and 4h, respectively.

    As the result shows, we can figure out that the band between 40 and 55 displays the existence of 6xHis::opSam2 whose molecular weight can reach 45.80 kDa. As the induction during increases, the band brightness becomes darker, which means that the expression of 6xHis::opSam2 is increased. Meanwhile, we may find some shifting phenomenon in rear lanes like lane 6, 7 and 8. This could be caused by the degradation of opSam2 due to the much longer induction. The similar result can be found in B. subtilis (Fig. 1.1D). The target band obviously becomes darker when induction time increases. So we can conclude that the expression of opSam2 in B. subtilis succeeded.

Figure 1.2 Enzymatic activity verification of opSam2 in B. subtilis
     A. ELISA principle depiction
     B. Standard cruve of SAMe quantification ELISA kit.
     C. Absolute concentration change of SAMe versus induction time.
     D. Relative concentration change of SAMe versus induction time.

    Since we have verified the expression of opSam2, we next measured its catalytic capacity of synthesizing SAMe from ATP and methionine. We induced opSam2 expression in B. subtilis, then aspirated the supernatant, and took the pellet to measure the SAMe concentration through SAMe quantification ELISA kit (Fig. 1.2A). The standard curve demonstrated our result was credible. In this experiment, we set a control group with empty vector (EV), which did not contain the recombined protein gene but contain the primary plasmid pHT01K, so we can use this group to standardize the concentration change of experiment group (Fig. 1.2C and D).
    We can figure out that the after induction, the concentration of SAMe definitely increased compared to EV group, so we can conclude that opSam2 can function normally in gram-positive bacteria B. subtilis, which means that it would be greatly possible to work in our final target probiotic B. longum.

Learn of the First Cycle

     In the first cycle, we can conclude that we have achieved the engineering success since we have verified the expression and function of opSam2.
    As we can see, IPTG induction may cause the toxicity to bacteria and high strength of induction expression also brings metabolic stress. So we can observe the shark declines of SAMe concentration in both EV group and pHT-opSam2 group. So in the final version of design, we tend to put opSam2 downstream of pTet promoter of Oscillator Device, we will not need toxic IPTG to induce, so the decline existing in above experiments might not disrupt our primary design.
    However, we still have several to improve. To have a better understanding of opSam2 and to improve its activity, we used Alphafold2 and SWISS-Model to analyze this homodimer’s structure (Fig 1.3). We can easily find that the C-terminus remains unordered and free, while the N-terminus is hid inside. So in the first version, in the case where we added the 6xHis tag, the high order structure may be disrupted or the enzymatic activity may be reduced.
    According to the prediction result, we can figure out the essential domain adjacent to SAMe. In the next cycle, we can use directed evolution to improve its enzymatic ability.

Figure 1.3 Structural analysis of opSam2 through AlphaFold2 and SWISS-Model.
     A. Sequence coverage of opSam2 when searching in Alphafold2.
     B. Structure prediction of opSam2 by AlphaFold2, colored by N -> C.
     C. Structure prediction of opSam2 by SWISS-Model, colored by confidence.
    D. Structure prediction of opSam2 by SWISS-Model, using Trace model to display the substrate SAMe binding site.

Design of the Second Cycle

    According to what we learned in last DBTL cycle, we decided to improve enzymatic ability of opSam2 and to optimize the catalytic condition.

1. Directed Evolution of opSam2.

    We can figure out that opSam2 has several ligands, such as magnesium ion, potassium ion, PPK, adenosine and SAM. The binding of such ion’s residues can be found according to the structural prediction model. For example, the SAMe is adjacent to 20 residues within 4å. We decided to collect the relative residue sites and perform point mutation to construct a mutation library (Fig 1.4A). Through SAMe production ability selection, we can gain a m-opSam2 with a better enzymatic ability.

2. Tag Addition Optimization

     We found that the C-terminus remains unordered and free, while the N-terminus is hid inside. So in the first version, in the case where we added the 6xHis tag, the high order structure may be disrupted or the enzymatic activity may be reduced. In this cycle, we will add the tag into the free C-terminus which might affect the folding and structure of opSam2 (Fig 1.4B).

3. Condition optimization of SAMe production

     Since the reagents of SAMe synthesis reaction are Met and ATP, and the binding to magnesium and potassium ions are demonstrated, we will performed orthogonal experiment to explore the best SAMe production condition for such molecules when induction (Fig 1.4C).

Fig 1.4 Improved design of the Second DBTL cycle of SAMe production.
    A. All residues adjacent to ligands (Down) and several sites marked red to be mutated (Up).
    B. Fused terminus change of His tag
    C. Orthogonal experiment table for exploration of production condition of opSam2

Build of the Second Cycle

    We have sent required fragments to synthesize in Tsingke Biology. However, due to the lack of time, we don’t have enough time to do more experiments to build in second DBTL cycle.

DBTL cycle of SAMe Secretion

Design of the First Cycle

Purpose

    To secrete the natural depressant SAMe, we need a SAMe transporter to localize in the membrane of our engineered bacteria. However, we didn’t find such transporter in prokaryotic cells since they almost don’t have any organelles. In eukaryotic cells, we search SAMe transporter in yeast, and found Pet8p.
     Pet8p is one transporter of mitochondrial carrier protein family of Saccharomyces cerevisiae. It is also called Sam5p, since it functions to transport S-adenosyl methionine (SAMe). We planed to express it in gram-positive model bacteria Bacillus subtilis firstly and then in Bifidobacterium longum, and then to verify its transportation function.However, we need to consider lots of problems since we plan to insert such eukaryotic multi-span transmembrane protein into bacterial plasma membrane.
     Firstly, some researches have reported that they had successfully expressed Pet8p in E. coli [2].

Design

    In our first-version design, to ensure the expression efficiency in Bacillus subtilis (B. subtilis), we use a codon optimization tool to optimize its coding sequence, and the optimized coding sequence is named as 6xHis::opMistic::opPet8p::sfGFP (Fig 1.5A). Then to verify its expression, we added a 6xHis tag to its N-terminus. Since it is a transmembrane protein, we have to observe its membrane localization, so we fused a sfGFP into its C-terminus. Meanwhile, to facilitate the membrane localization, we insert a short peptide Mistic derived from B. subtilis to the N-terminus, which is reported to help transmembrane protein target at bacterial plasma membrane. With a high-magnification fluorescent microscope, we can figure out whether the recombined proteins are enriched at the “peripheral” compared to cytosol.

Build of the First Cycle

    Then we constructed such part in a shuttle vector pBE2 for cloning in Escherichia coli DH5α, and for expression in B. subtilis. We firstly inserted a P43-promoted XylR fragment into the plasmid with PstI/KpnI double cut, and then we inserted another fragment Pxyl-promoted opPet8p into the first recombined plasmid using KpnI/EcoRI double cut. XylR expressed onstitutively and promoter Pxyl constructed a xylose-inducible device to activate the expression of opPet8p. The XylR fragment was PCR from genome of B. subtilis, and the other fragments are synthesized. pBE-XylR-opPet8p (10046bp) has been constructed and transferred into E. coli and B. subtilis (Fig 1.5 B, C, D, E).

Figure 1.5 Successful cloning result of pBE-XylR-opPet8p in E. coli and B. subtilis.
     A. Graph description of plasmid construction from pBE2 to pBE2-XylR, and then to pBE2-XylR-opPetp.
     B. Enzyme cut of XylR fragment and pBE2. Lane1: GL DNA 5000 marker, Lane2, 4: pBE2, Lane3, 5: pBE cut by PstI/KpnI, Lane6: XylR PCR fragment, Lane7: XylR cut by PstI/KpnI, Lane 8: GL DNA 2000 marker.
     C. Enzyme cut of pBE-XylR and PCR Pxyl-opPet8p. Lane1: pBE-XylR, Lane2, 3: pBE-XylR cut by KpnI/EcoRII, Lane4, 5, 6: Pxyl-opPet8p PCR fragment, Lane7: GL DNA 2000 marker.
     D. Colony PCR of pBE-XylR-opPet8p in B. subtilis.
     E. Partial opPet8p sequencing result in E. coli.
     F. Partial opPet8p sequencing result in B. subtilis.

Test of the First Cycle

     And the expression of E. coli and B. subtilis are verified by green fluorescence (Fig 1.6A, B), fluorescent image (Fig 1.6C), Western Blot (Fig 1.6D, E, F). We can concluded that we have successfully expressed recombined protein 6xHis::Mistic::opPet8p::sfGFP, since the N-terminus tag and C-terminus tag are both verified (Fig 1.6G).

Figure. 1.6 Expression verification of opPet8p in E. coli and B. subtilis.
     A. Plate spread with E. coli containing plasmid pBE-Pxyl::6xHis::opMistic::opPet8p::sfGFP can emit green fluorescence when excited by ultraviolet light (UV).
     B. Visual green shows the expression of opPet8p in E. coli after 1%xylose induction for different periods of time.
     C. Fluorescent image of E. coli and B. subtilis with pBE-XylR-opPet8p::sfGFP after induction.
     D. BCA quantification standard curve before WB.
     E. Western blot of opPet8p using anti-His-HRP. The front 5 lanes show the induction expression of assumed fusion protein in E. coli, and the rear 5 lanes show those in B .subtilis, induced for 0h, 1h, 2h, 4h and 8h, respectively.
     F. Western blot of opPet8p using anti-GFP. The front 5 lanes show the induction expression of assumed fusion protein in E. coli, and the rear 5 lanes show those in B .subtilis, induced for 0h, 1h, 2h, 4h and 8h, respectively.
     G. Graph demonstrates that we used anti-His antibody to target the N-terminus of opPet8p and uesd anti-GFP antibody to target the C-terminus of opPet8p as well in WB, whose results are showed in E and F above.

    Meanwhile, we also verified the membrane localization of opPet8p (Fig 1.7). According to the distribution of fluorescent signal, we can conclude that the recombined proteins can target to the bacterial membrane. However, if the expression rate is too high to conduct the proper folding so that they might form the particle-like inclusion body, they might gather together to precipitate. To better understand the membrane integration of our recombined protein opMistic::opPet8p::sfGFP, we used a prediction tool Phobius to predict. The results are showed below (Fig. 1.7A).

Learn of the First Cycle

     As we can see in Fig 1.7A, the addition of Mistic can result in the addition of transmembrane segment which can facilitate the membrane insertion of opPet8p, while the sfGFP does the opposite thing. So after using sfGFP to verify its expression and membrane localization, if we want to improve its membrane localization efficiency, we need to replace the sfGFP with other tag, like some small particular fluorescent group. According to the integration pattern prediction, we draw the model graph to display the difference (Fig 1.8A).

Figure 1.7 Membrane localization verification of opPet8p in E. coli and B. subtilis.
     A. Membrane integration prediction of primary opPet8p (Left) and recombined fusion protein opMistic::opPet8p::sfGFP (Right)
    B. Fluorescent image of Bacillus subtilis induced by different concentration of xylose and a constitutive expression strain.

Design of the Second Cycle

     In above results, we found that there are some particles enriched with fluorescent signals in bacteria, except for the average distribution. We supposed that the formation of the particle could be due to wrong protein folding. So we performed membrane insertion prediction of coding sequence of opPet8p and the recombined protein, and we found that the addition of sfGFP induce a large hydrophilic segment which is unfriendly to the membrane insertion (Fig 1.7A). It could be possible that sfGFP disrupts the folding of opPet8p, so to eliminate such particles, we designed a new part in which a linker (Gly–Gly–Gly–Gly–Ser) was inserted between the opPet8p and sfGFP (Fig. 1.8A, B). With such a flexible linker, sfGFP is supposed to keep a distance from the opPet8p so that the folding of opPet8p won’t be less disrupted and the membrane insertion could be better.

Figure 1.8 Graph description of different recombined proteins

Build of the Second Cycle

    We have sent required fragments to synthesize in Tsingke Biology.

Test of the Second Cycle

    Then we performed the same operation as the last cycle to observe the membrane localization of 6xHis::opMistic::opPet8p::linker::sfGFP, and set the old strain who expressed opPet8p without fusing with linker. We can obviously found that strains expressing opPet8p fused with linker have absolutely fewer particle-like inclusion bodies in cytosol compared with strains expressing opPet8p without linker (Fig 1.9). So we can conclude that the linker can work to facilitate the folding of proteins and reduce the formation of inclusion body. However, as the large copy number of vector and decrease of protein precipitation, large amount of proteins are distributed in cells whose fluorescent signal is too large to observe the membrane localization. Due to the lack of time, we haven’t explored more proper condition to observe, but we are optimistic to observe its membrane localization after decreasing the induction time.

Figure 1.9 Comparison of fluorescent signals in strains expressing opPet8p fused with linker and without linker.

DBTL cycle of Oscillator model

Design of the First Cycle

    During the time when we assembled our project, we decided that it would be a great idea to simulate the normal sicario of drug uptake, to be more precise, we want our engineered bacteria to give off drug at due time intervals rather than a constant secretion that would last the whole time, in hope that this can achieve fluctuation of the serum drug concentration to aid in the avoidance of adverse effects led to by elevated serum drug concentration while still maintaining a minimum concentration that can have a curing effect.
    To simulate this sicario of drug uptake, we decided to introduce an oscillator that can be exploited to control the production, and subsequent secretion of the drug SAMe, by controlling the expression of the enzyme sam5p that is responsible for the conversion of Met into SAMe and which is the only step in the pathway leading to the synthesis of this compound.
     There have already been many mature solutions in E. coli for constructing an oscillator, but when it comes to Gram positive bacteria, we found none that is sure to work. So, after looking up previous publication and works done by previous iGEM teams, we decided on changing an existing design that has already shown to be working robustly in E. coli, namely, the Repressilator. Published in 2001 on Nature, the Repressilator consists of three transcription factors and their corresponding promoters.

Figure 2.1 A graphic representation of the Repressilator

Build of the First Cycle

    And in the experimental design of our group, we focused on determining whether these three components can function as expected in B. subtilis as has been characterized in E. coli, respectively.
     In order to determine the function of these transcription factors, we cloned these transcription factors and their promoters onto shuttle vectors that can self-replicate and express in B. subtilis as well as in E. coli.

Test of the First Cycle

    And we intended on using microplate reader to obtain data on the fluorescence intensity of GFP, whether is driven by the promoter, or is combined with the expression of the corresponding transcription factor.

Figure 2.2 The plasmid we are planning on constructing

Learn of the First Cycle

    However, due to failure in the process of molecular cloning, we didn’t obtain these plasmids, and as a result, no further experiments were carried out.
    But nevertheless, we do believe results strongly in favor of our hypothesis will result due to the fact that the three transcription factors and there corresponding promoters are thoroughly studied, and the fact that there has been publication confirming the success of protein expression in the chassis.
    And taking the possibility of failure into account, we designed another set of experiments from the beginning, and we think that this can be taken as the second cycle of the DBTL cycle

Design of the Second Cycle

     Another thing we would like to try is to reconstruct the intact Repressilator into our engineered bacteria, and see if it works as expected. In this way, we can verify the three component that constitute the Repressilator in one experiment.

Build of the Second Cycle

     As mentioned, we are planning to reconstruct the intact Repressilator in B. subtilis, and we first tried to reconstruct it in E. coli.
    So, we transformed the Repressilator plasmid into E. coli, and found a reporter plasmid from iGEM repository that constitute an GFP under the expression of tetR promoter.

Test of the Second Cycle

     After co-transformation, we verified the result via colony PCR, and after synchronizing the bacteria by adding and removing IPTG, we measured the fluorescence intensity using microplate reader. But the result showed no significant oscillation.

Figure 2.3 Results of colony PCR

Learn of the Second Cycle

     We think that this is due to some bacteria not fully synchronized, and due to the fact that the bacteria will gradually become out of pace in respect of each other. So, we planned on measuring the fluorescence intensity of single bacteria.

Design of the Third Cycle

     As mentioned, we planned on measuring the fluorescence of the single bacteria. So, we first simply planned on videoing the same bacteria without removing it from the microscope.
    Before we took the video, we took a photo of the bacteria, to see if we can observe fluorescence, and the result showed that there are indeed fluorescence in the bacteria co-transformed the reporter plasmid and the Repressilator plasmid.

Figure 2.4 Photo of bacteria that contains both the reporter plasmid and the Repressilator plasmid taken under a fluorescence micrpscope.

Build of the Third Cycle

     The plasmid and the bacteria strain we used in this cycle were the same as the last, but the modifications we made is in the protocol that we used to photograph the bacteria.
     We learnt that we can cover the slide we used to image with a thin slice of solid LB medium, in this way, the bacteria can survive for a relative period, while for the whole duration we were able to image it.

Test of the Third Cycle

     We set the system up as described, consisting of the fluorescence microscope that we used for imaging, the special glass slide we used for supporting the bacteria, and the bacteria itself.

Learn of the Third Cycle

     But when doing imaging, we found some big issues. The biggest one being photobleaching, and there are also minor ones such as difficulty in image processing. So, we started searching for other ways around this problem. And fortunately, we found one that worked, and the result we obtained made its way into the result section of our project.

Design of the Fourth Cycle

     The way around the problem is by photographing the same region of the slide at given time intervals. So that in this way, we do not have to place the bacteria under the excitation light for the whole duration, and instead of processing a video that can last for hours, we only need to process a few photos.

Build of the Fourth Cycle

     We set the imaging system up the same way as in the last cycle, but this time we didn’t take video, but instead, took photo once every 10 minutes.

Test of the Fourth Cycle

     Using the procedure described above, we took various photos, as shown down below.

Figure 2.5 Photo of the same region of the slide taken at an time interval of 10 minutes.

     Then, we measured the light intensity of the whole photo, to serve as the background of the fluorescence intensity (making a hypothesis that the light intensity of the bacteria with fluorescence is so much less than the background that the light intensity of the bacteria can be neglected, and as can be seen below, this hypothesis is correct)

Figure 2.6 The measurement of the light intensity of the whole photo, to see the background.

    The first box indicates the average light intensity per pixel of the whole photo, and the second box indicates the number of pixels of the image. The whole light intensity can be obtained by multiplying the two numbers together.
    Then, we cycled out the same bacteria of each photo, making sure that they are all at the same region in each photograph, numbered them 1-10, as shown in Fig 2.7a, and measured the light intensity in each of the small regions, the result is represented in Fig 2.7b, and the data obtained is shown below in the table.

Figure 2.7 The numbering of each bacterium, and the measurement of the light intensity in regions around each bacterium.

Total number of pixels around each bacterium and the whole picture.

Average light intensity around each bacterium and the whole picture.

Total light intensity around each bacterium and the whole picture without background.
Figure 2.8 Analysis workflow of light intensity around each bacterium and the whole picture.

    To visualize these data, we imported them into Prism, and performed non-linear regression using sine waves as the model, to find out the expected oscillation cycle of the Repressilator plasmid. The result is shown below.

Figure 2.9 The analyzed result of the oscillation cycle of the co-transformed bacteria.

    But as can be felt from the regression curve, the data point at 30 minutes violated the whole trend, and the goodness of fit is quite low. So, we decided to delete this data point, and do the regression once more.

Figure 2.10 The analyzed result of the oscillation cycle of the co-transformed bacteria after deleting the data obtained from the photo taken at 30 minutes.

    And as can be seen, the result is much better, and the oscillation cycle settled around 60 minutes, which agrees with the result shown in the original article.

Learn of the Fourth Cycle

     Through these experiments, we found the mostly likely oscillation cycle of the Repressilator, and we think that this can wrap up the experiments we need to do in E. coli. Then, we think that its time to move on to the verification of the function of the Repressilator plasmid in B. subtilis.

Design of the Fifth Cycle

     In this cycle, we planned to reconstruct the Repressilator plasmid in B. subtilis.
     The protocol we decide to follow and the plasmid we decided to transform into the bacteria strain is the same as the last cycle.

Build of the Fifth Cycle

     Due to the lack of time, we didn’t manage to get into this part of the experiments, but we do think that positive result will come into being if we get our hands on this.

DBTL cycle of selection plasmid construction

    We set up the Directed Evolution module to develop our LacI protein, making it only responsive to high lactose in a way gut environment would hardly contain. By lifting the threshold value of lactose toleration, our engineered bacteria can stably produce periodical SAMe, achieving better therapeutic effect. According to the previous research, we chose three residues as our mutagenesis positions - I79, F161 and L296 - since they have been proved to function in substance binding.

Design of the First Cycle

    At first, we planned to utilize fluorescence intensity as result instruction, using EGFP gene controlled by Lac promoter as reporter (Fig 3.1). However, as we are selecting LacI protein based on one single quantitative, screening fluorescence could be quite complicated and inaccurate. So, we developed our design and added two more genes – sacB and KanR – so that we can conduct the evolution procedure on rounds of selective culture of engineered bacteria. The positive selector sacB eliminate variants that couldn’t tolerate present level of lactose, while negative selector making sure the LacI protein still obtain its function to release from the promoter. More details are illustrated in Design of the Directed Evolution Module.

Figure 3.1 Diagram of plasmid construction. A: Initial design of selection element; B: Present design of selection element;

Build of the First Cycle

     During the process, we first used overlap extension PCR to construct the selection element including mRFP1, sacB and KanR into one giant fragment. By using flanking primers, we successfully observed the ideal bands at first. However, the products turned out to be quite unstable and would easily break into two after a second amplification (Fig 3.2A). Also, the whole PCR process is quite complicated and time-consuming.

Test of the First Cycle

    We improved our protocol by dividing the construction into two steps, connecting 2-3 fragments at a time, using only Gibson Assembly to construct the entire plasmid. This alteration significantly increased the successful rate. The final plasmid then underwent MEGAWHOP to build mutagenesis library and was transformed into BL21 and carried on for further selection. DpnI digestion contrast experiments proves the mutagenesis being effective (Fig 3.2B).

Figure 3.2 Plasmid and mutagenesis library construction. A: Overlap extension PCR product break in the second amplification; B-C, plasmid transformation results after DpnI digestion. B: MEGAWHOP without megaprimer; C: Normal MEGAWHOP;

Learn of the First Cycle

    In the first cycle we successfully achieved overlap extension PCR of adjacent fragment. However, perhaps it’s because the fragments are rather too long (sacB-1437bp) or too short (terminator+Pc-164bp), the product isn’t quite stable.
    By concluding the fragments into occlusive plasmid and deriving target ones through antibiotic selection, the percentage of desired plasmid has risen, even though the fragment loss still occasionally occurred. Using transformation to derive the desired product is relatively time-consuming, while the results are also more promising.

DBTL cycle of variant selection

Design of the First Cycle

    If the LacI responds to certain level of lactose and release from the promoter, sacB will be expressed and cause bacterial death. By applying gradually lifted lactose, we were able to establish rounds of selection based on sucrose or kanamycin plates until the ideal LacI proteins were derived.

Build of the First Cycle

    We first built up solid medium with IPTG gradient addition. Containing chloramphenicol, sucrose, and IPTG, theoretically only desired mutants can grow on such medium. The reason why we chose IPTG is because natural metabolism of lactose would result in the inconsistency of concentration during the experiment, and its product is immeasurable. Since IPTG is more stable and has been widely used in lactose-related research, we used IPTG as an alternative of lactose to maintain a reliable result.

Test of the First Cycle

    The first round of selection was carried out on 2% sucrose plate with IPTG at 0.1mM and 0.2mM respectively, by spreading bacteria culture in different collections of plates. However, it turned out to have bare difference between different concentrations of IPTG and fail to separate the improved variants (Fig 3.3A). We wondered whether the sucrose content wasn’t high enough, so we lifted the IPTG and sucrose concentration at the same time in the next round, setting IPTG gradient of 0.2, 0.4, 0.6, 0.8, 1mM and sucrose content of 8%. Unfortunately, we still failed to observe the expected difference (Fig 3.3B).
    We instead stepped forward to explore the role of IPTG induction duration played in results. Adding gradient of IPTG when OD600 equals to 0.6, we sampled each group every hour, with the duration reaching to around 6 hours. However, the results are still unappreciated, with bare difference observed (Fig 3.3C), either among IPTG levels or among induction duration. Our selection based on sacB ended in failure.

Figure 3.3 Series of exploration toward sacB.
   A: Comparison between different IPTG concentration on 2% sucrose plate.
   B: Comparison between different IPTG concentration on 8% sucrose plate.
   C: Comparison between different IPTG concentration and inducing time on 8% sucrose plate.

Learn of the First Cycle

    After investigation, we found that the inefficiency of sacB is quite common, influenced by many factors such as sucrose content, media component and even bacteria species. Our followed-up experiment also revealed the timing to apply IPTG induction or solid media spreading being vitally important as well. When spreading the culture at OD600=0.2 approximately, that is when turbidity can just be observed, the killing effect from sacB expression is remarkably significant (Fig 3.4). On regular antibiotic plate with 15ug/ml chloramphenicol transformed bacteria grew into lawn, while on Cml plate with 8% sucrose and 0.8mM IPTG no colony is witnessed. In previous selection, we rather added the inducer IPTG at OD600=0.8 as in regular protein experiments or spread the culture when it has grown overnight. This may be the reason why the first round of our selection failed.
    However, this discovery appeared in late stage of our experiments when a new round of selection is not available. Before it we had to change our mind and establish the selection in a new path. This is how the next round of selection based on KanR was put forward.

Figure 3.4 Contrast between 8% sucrose and non-sucrose plate.

Design of the Second Cycle

    Even though KanR gene was first added to exclude the possibility that mutant LacI cannot release from the promoter, however, since sacB selection didn’t work out, we decided to utilize KanR to carry out the next round of selection. Due to its property, the expression of KanR survives the bacteria. Thus, we need to compare one mutant’s behavior in multiple inducer concentration to make the conclusion. If the mutant can only growth on high-IPTG plates rather than lower ones, then it’s still promising to say that the LacI can only respond to high-level lactose.

Build of the Second Cycle

    We set up IPTG gradient in kanamycin plates of 0, 0.1, 0.5 and1 mM, and each mutant culture needs to spread plate in such combination group. In total 30 colonies were cultured and undergo selection.

Test of the Second Cycle

    Even though the procedure is complicated, the results turn out to be quite satisfying. Different mutants showed different tolerance to IPTG, which once again illustrates the successful construction of our lacI mutagenesis library (Fig 3.5). Among them, we derived the desired variant that only responds to high-level of IPTG (0.5mM) with little growth at low concentration (Fig 3.5). More surprising, a few growths also appeared on 0.8mM IPTG plate, suggesting that the mutant still contains its ability to release from the promoter and thus can be utilized in oscillator (Fig 3.5).

Figure 3.5 Selection results.
     A: Control group 1. Original strain and random mutant are applied on empty antibiotic plate with IPTG contrast.
    B: Control group 2. Original strain is applied on Kan+ plate with IPTG gradient.
    C: Desired mutant, Strain #29. No colony growth is observed at 0 and 0.1mM IPTG, suggesting that the LacI can tolerate high-level IPTG.
    D: Strain #27. The mutagenesis causes its binding to lactose (IPTG) to be easier, and lawn are witnessed at all concentration, even without IPTG existence. Its behavior is quite familiar to performance on non-antibiotic medium.
    E: Strain #9. The mutagenesis slightly reduces the leakage of the promoter, however, barely difference occurred between different inductor (IPTG) concentration.
    F: Strain #24. Little up-regulation of the promoter takes place as IPTG concentration lifting. Its response to low-level inductor (IPTG) is still relatively high.
    G: Strain #2. Few colonies occurs at 0mM IPTG and no growth at higher level.
    H: Strain #6. No growth at all. The sensation to IPTG is totally lost.
    I: Strain #13. Expression is activated only at 0.1mM IPTG.
    J: Strain #14. Few colony are witness, expect one at 0.5mM IPTG, suggesting its affinity to IPTG has greatly dropped.

    In order to evaluate the behaviors of different mutants for better judgement, we designed an algorithm to access the IPTG responsive ability of the strains. We use 1-7 to measure the intensity of the colonies on each plate, and integrated them into a heatmap for visualization (Fig 3.6). Our goal is to select a LacI variant that cannot sense the low concentration IPTG and at the same time that normally functions at high concentration of IPTG. Thus, reflecting on different colors, the ideal strain is supposed to have lighter color between darker ones at high range of IPTG. Comparing the data, Strain #29 is exactly the one that we are looking for (Fig 3.5 C). Meanwhile, a few growths under 0.8mM IPTG guarantees that the mutant LacI still keeps the ability to release from promoter and it could function normally in our oscillatior - “Repressilator”. Selected mutant also appeared to solve the leakage problem that original LacI contains, showing a more outstanding performance.

Figure 3.6 Visualization of arithmetic LacI evaluation result. The triangle points out the desired variant strain #29.

Learn of the Second Cycle

    In the second cycle we managed to achieve our goal of deriving improved LacI mutant using KanR. The drawback and advantage of this plan are all distinct. The biggest drawback is the great amount of works because the experiment requires cultures of single colony and each one requires groups of spreading. However, this also keeps its consistency in specificity, with no need to re-culture the post-selective colony.
     Although we successfully separated the ideal variant, the experiment still has several faultiness and could be improved in the future. For example, the concentration gradient can be divided into more interval, which can identify a more detailed section that LacI functions in. Besides, such actions with high labor and time cost can be solved depending on computer using high throughput equipment, such as micro-fluidic chip or microplate reader. We are looking forward to such attempt afterwards.
     During the experiment, the leakage of the lac promoter is still an unignorable problem (Fig 3.5B). Though the selected variant seems to have improved this situation, some improvements to solve this problem in the first place are still highly appreciated and will undoubtedly benefit in selection efficiency. We believe that methods such as adding a sponge could better improve the performance of the plasmid and thus of the selection process.

DBTL cycle of Safety Switch

Design of the First Cycle

    In our design, we use chemical switch for active termination. CinR is an autoinducer binding protein found in Rhizobium leguminosarumis. In complex with O3-C14 HSL, which is synthesized by the inducer CinI, CinR protein activates transcription of genes downstream the pCin promoter. Thus, when the bacteria accept this chemical signal and activate the pCin promoter, the downstream gene of holin will express. Holin is a protein that will make holes on cell membrane and cause death of bacteria. Next, we use temperature switch for leakage protection. TlpA36 is a temperature sensitive protein which is capable of repressing the pTlpA promoter under temperature of below 36 degrees Celsius, thus repress the expression of PenI repressor. As a result, holin will express below 36 degrees Celsius and cause the death of bacteria (Fig 4.1).

Figure 4.1: design of safety switch.

Build of the First Cycle

    We used restriction enzyme and DNA ligase to construct our recombined plasmids (Fig 4.2). And we transformed the plasmids into E. coli and B. subtilis for examination.

Figure 4.2: The result of plasmid construction and sequencing.
    A, B: Construction and sequencing of chemical switch.
     C, D: Construction and sequencing of temperature switch. )

Test of the First Cycle

    We chose EGFP as its downstream reporter. Firstly, we tested the function of the plasmids in E. coli DH5α. For chemical switch, we used recombined E. coli BL21 to produce autoinducer and extract the supernate. We tested the function of chemical switch by inducing the expression of EGFP with the supernate. For temperature switch, we induced it under 25℃. As a result, in E. coli the fluorescence of temperature switch is observed under 25℃, but unfortunately, no fluorescence was observed among different groups of chemical switch. Also, no significant changes in fluorescence were observed in both groups in B. subtilis (Fig 4.3).
    Lastly, we tested the function of PBSX holin by measure the cleavage of the E.coli. In this construction, the expression of holin is induced by IPTG. We found that after inducing, the bacteria were killed or inhibited to some extent though not fully (Fig 4.4).

Figure 4.3: The fluorescence in E.coli with chemical or temperature switch and the fluorescence in B. subtilis with chemical or temperature switch.

Figure 4.4: Inducing the expression of PBSX holin in E. coli.

Learn of the First Cycle

    Thus, we still need to improve our design in the future to make it a complete safety switch. Though by many research and proof, we have confirmed that most basic parts in our design should be able to work in B. subtilis, further examination of ever single part is still needed. For we met difficulties when tested the function of the whole composite parts. For example, TlpA36 promoter should be tested in B. subtilis whether it have its normal function. We simplify the construct and can use these plasmids to test the function of TlpA36 and PenI with their promoter respectively. Due to the lack of time, we did not carry out the next round of construction. But we planned to make them more available based on the lessons of previous experiments.

Design of the Second Cycle

    What is more, as soon as we find the problems of that single part and improve the switch part, we should bind the composite part and holin together to achieve the function of the safety switch. By now, they are designed as follows: (Fig 4.5).

Figure 4.5: Construction of composite part with holin.

    Although to fully achieve these composite parts is challenging, we still believe that some methods can be used to direct our design and increase the success rate. We still hope to try to construct these safety switch.

DBTL cycle of Hardware

Design of the First Cycle

     For the hardware, we have accomplished the systematical work of turning it from design into reality. We wanted to explore the specific parameters of the capsule so that the drug could be degraded in the intestine. Now we have completed part of the gut simulation.

Build of the First Cycle

     At the beginning of the experiment, we tried to observe the dissolution of various substances in a stationary solution of different pH values, but this was quite different from the real human environment, so we modified and designed a simple device to achieve intestinal peristalsis and liquid flow by manual pushing.

Figure 5.1

Test of the First Cycle

     However, we found that various chemicals took a relatively long time to dissolve, and anthropogenic vibration was not a good idea. To make matters worse, human operation could easily lead to various problems such as water leakage or instrument damage.

Learn of the First Cycle

     Therefore, we needed to design an automatic instrument that could operate stably for a long time.

Design of the Second Cycle

     Considering that the dissolution time took about 2 hours, we modified the device to include a motor and a microfluid pump to make the operation of the equipment more convenient.

Build of the Second Cycle

     We repurchased the microfluid pump and motor and assembled it together.

Figure 5.2

     The coarse catheter was so short that it caused the liquid to flow differently from the real situation, so we lengthened the coarse catheter. In addition, in order to enable the coarse catheter to be disassembled repeatedly, we used thermoplastic hoses, which would gradually shrink after heating, to prevent the deformation of the coarse catheter caused by repeated disassembly, which would lead to water leakage.

Figure 5.3

Test of the Second Cycle

     After that, we conducted experiments to explore the solubility of carboxymethyl cellulose particles of different radii in different solutions, and the experimental results are as follows.

Learn of the Second Cycle

     In the end, we found that the granular material is made of carboxymethyl cellulose and the radius should be less than 0.3cm to ensure that it completely dissolves in the intestine and can stay in the stomach for at least 2 hours.

     In this DBTL cycle, we used the last device to conduct the experiment and obtained the dissolution rate of different substances. According to the experimental results and the residence time of the capsule in various parts of the human body, we designed the parameters of the capsule. Although such a set of parameters may not be completely in line with the human environment, it basically provides a reasonable search range, and further confirmation of parameters requires human clinical trials.

Figure 5.4

Reference

     [1] Joshua S. Klein, Siduo Jiang, Rachel P. Galimidi, Jennifer R. Keeffe, Pamela J. Bjorkman, Design and characterization of structured protein linkers with differing flexibilities, Protein Engineering, Design and Selection, Volume 27, Issue 10, October 2014, Pages 325–330, https://doi.org/10.1093/protein/gzu043
     [2]Elowitz, M., Leibler, S. A synthetic oscillatory network of transcriptional regulators. Nature 403, 335–338 (2000). https://doi.org/10.1038/35002125
     [3] Wu J, Jiang P, Chen W, Xiong D, Huang L, Jia J, Chen Y, Jin JM, Tang SY. Design and application of a lactulose biosensor. Sci Rep. 2017 Apr 7;7:45994. doi: 10.1038/srep45994. PMID: 28387245; PMCID: PMC5384092.
     [4] Piraner, D., Abedi, M., Moser, B., Lee-Gosselin, A., & Shapiro, M. Tunable thermal bioswitches for in vivo control of microbial therapeutics. Nature Chemical Biology, 2016, 13(1), 75-80. http://dx.doi.org/10.1038/nchembio.2233
     [5] Calatrava-Morales, N.; McIntosh, M.; Soto, M.J. Regulation Mediated by N-Acyl Homoserine Lactone Quorum Sensing Signals in the Rhizobium-Legume Symbiosis. Genes 2018, 9, 263. https://doi.org/10.3390/genes9050263
     [6] Susanne Krogh, Steen T. Jørgensen, Kevin M. Devine. Lysis Genes of the Bacillus subtilisDefective Prophage PBSX. Journal of Bacteriology. 1998, Vol. 180, No. 8. https://doi.org/10.1128/JB.180.8.2110-2117.1998
     [7] Bowtle W J. Materials, process, and manufacturing considerations for lipid-based hard-capsule formats[M]//Oral Lipid-Based Formulations. CRC Press, 2007: 101-128.
     [8] Duke G E. Alimentary canal: secretion and digestion, special digestive functions, and absorption[M]//Avian physiology. Springer, New Y