Team:Marburg/Results

Results

However beautiful the strategy, you should occasionally look at the results
-- Winston Churchill

After having conceived our projects, we went to the lab and tried to fulfill our expectations. Successes and failure were used to improve our projects strategy by iterating theoretical and practical working steps. On this page, we want to convince you that we managed to realize some of our visions and are able to show achievements in all of our subprojects.
Foundational experiments with V. natriegens
Growth and Selection of V. natriegens

Growth experiments

Two figures. The top figure is pH corellated to 3HPA concentration. The pH is lowest, 4.7 , for the concentration of 10mM, the highest concentration we tested. It continually rises to about 6.5 without 3HPA. 
		The second figure shows luminescence corelated to 3HPA concentration. Numinescence is very low for the highest 3HPA concentrations, climbing at medium concentrations and dropping again to about halve of the maximum at low to no 3HPA
Figure 1: The Graph shows four growth curves three times V. natriegens (blue, red and orange) and once E. coli (NEB Turbo)( green))

After reading papers about V. natriegens deciding to order the strain ATCC 14048, we were eager to make our first growth curves.

Because V. natriegens is a marine bacterium organism and from literature (et al), we knew that we had to supplement our standard LB media with an additional 1.5% NaCl.

Being inexperienced with the new organism, we only managed to get a doubling time slightly better than for E. coli. Not disheartened by this, we did some more research and found several points we wanted to try.

We tested the following media, LB 2.5 NaCl, LB v2 salts and Brainheart infusion v2 salts.

 The Graph shows four growth curves three times V. natriegens (blue, red and orange) and once E. coli (NEB Turbo)( green)
Figure 1: The Graph shows four growth curves three times V. natriegens (blue, red and orange) and once E. coli (NEB Turbo)( green)

v2 salts are: 204 mM NaCl, 4.2 mM KCl, 23.14mM MgCl2 Brainheart infusion can not be autoclaved together with the v2 salts. From this we learned that growth was most vigorous in BHIv2, closely followed by LBv2. So we decided to make LBv2 our standard media going forward. It is cheaper as well as just being standard LB media, supplemented after autoclaving with the three common salts NaCl, KCl and MgCl<sub>2</sub> .

Additionally, some aspects of handling could be improved. For instance reducing the time outside of the incubator for measuring, thereby avoiding cooling of the media, and using baffled flasks to improve aeration. Applying these, we managed to get much better results, even beating the reported doubling time of 9.8 min

(Eagon et al. 1962). The graph shows a growth curve of V. natriegens in BHIv2; V.natriegens in LBv2 and the fastest E.coli stain (NEB Turbo) in LB media.

We achieved an incredible growth rate of 7 min. Also, using baffled flasks, our overnight cultures routinely reached OD's from 14 to over 16. By using enpresso capsules, which slowly release glucose into the media, we could even best that and reach staggering OD's of around 26.

Antibiotic tolerance of Vibrio natriegens

This picture shows spotted colonies of V. natriegens on  gradients of the most common antibiotics. Except for cycloserine and penicillin there as alay a certain point when the wild type stopped growing. So selection is possible.
Figure 1: A) This picture shows spotted colonies of V. natriegens on gradients of the most common antibiotics. Except for cycloserine and penicillin there as alay a certain point when the wild type stopped growing. So selection is possible.

The required concentrations for selection in V. natriegens is still unknown for most standard antibiotics. Since these were a prerequisite for our project, we strove to find them early on.

An easy method to get a good idea of the effective concentrations is to make agar plates with a gradient , then set spots of your culture along that gradient. (Figure 1)

We found that the penicillins ampicillin and penicillin, the arguably most used antibiotics, seemed to be completely ineffective at inhibiting V. natriegens . Fortunately, V. natriegens susceptible for the penicillin derivative carbenicillin. Since the carbenicillin resistance is mediated through the same enzyme as ampicillin resistance, the ß-lactamase, all constructs carrying this resistance can still be selected for.

The concentrations in the following table are working fine in our lab for selecting transformants with the resistance cassettes from our Toolbox.

WT ingibition [µg/mL]

WT counter-selection

[µg/mL]

inhibition with resistance [µg/mL]

Chloramphenicol

2

2

<12

Carbenicillin

50

200

600

Kanamycin

65

200

400

Tetracycline*

7

11

<12

Streptomycin

80

/

/

Gentamycin

30

/

/

Glycloserine

<200

/

/

Penicillin

<100

/

/

(Figure 2 Scheme of the most common antibiotics in synthetic biology and their relevant concentration in correlation with selection V. natriegens with and without a resistance cassette )

The first four antibiotics, Chloramphenicol, Carbenicillin, Kanamycin and Tetracycline are the antibiotics we commonly used in our project. For Golden Gate cloning we routinely switch between Chloramphenicol and Kanamycin. Because, it is highly sensitive to light, and we are using inducible promoters that are activated by a derivative of tetracycline and also reacting to the antibiotic, tetracycline only is used as our emergency antibiotic.

Enabling transformation in Vibrio natriegens
The first transformations we tried were based on E. Coli standard protocols and did not work. In the paper of Weinstock et al. 2016 you can find a protocol for transformation with chemically competent cells and for electroporation. In the protocol for chemically competent cells it is stated that there must be a knockout of the DNAses for it to work. We tried it without the knockout with positive results, but for a re-transformation at least 550 ng of the desired DNA is needed.
Figure 1: On the left site the two plates top are positive controls, if the cells survived the procedure of making them competent and freezing at -80°C On the left side below are two negative controls without Plasmid for showing that the antibiotic plates have a selecting concentration. Then from left to right the concentration of DNA used of the transformations increases each time as triplicates in a row. The concentrations used are 140ng , 280ng and 560ng

We used the pYTK plasmid from the Dueber Toolbox pYTK which has a GFP dropout for visual differenciation, as well as a Chloramphenicol resistance.

We think the issue is that during the making and freezing of the competent cells, a lot of cells die, which releases the DNase from the wild type into the media. The DNases digest the DNA before it can get into the cells.

The reason why the transformation works without the knockout is that the DNases can't digest all 550 ng DNA and some is left over and can be taken up. You only get single colonies and this efficiency is too bad for cloning. The amount of DNA needed is too great for realistic applications and you would still need E. Coli for re-transformation. Since we think the DNase is the root of our problem, after we knock it out we want to give chemically competent cells and heat shock transformation another try. Next we tested the electroporation protocol, because it is reported to have a much higher efficiency. Akira Taket 1988 We made some minor adjustments to the protocol because we did not have the same equipment that was used in the paper and we wanted to use 3000xg instead of the rpm value for the centrifuge to make the protocol applicable to other labs.

Figure 2:

The picture shows four plates of Vibrio natriegens , transformed with different concentrations of pYTK. Top left: 50ng; top right: 5ng; down below left: 0.5ng; down below right: 0.1ng. For better visualising the GFP the plates are lightened by UV light.

We used an electroporator from Eppendorf whose only variable parameter was the voltage. This protocol worked fine with 100 ng of DNA. But the efficiency still left something to be desired. Especially when we think about cloning, so we optimized the protocol and our handling again. We made sure that all buffers and centrifugation tubes were cooled to 4 °C when making competent cells before they reached an OD600nm of around 0.5 and that the samples were kept on ice to inhibit further growth.

We increased the centrifugation strength because after diluting the cells to a fix OD of 16, we produced 30 aliquots which isn't enough when you do triplicates. We also froze the cells in liquid nitrogen instead of in dry ice. For the electroporation we tested with different gap sizes, ranging from 1 to 2 mm. 1mm proved to work better for Vibrio natriegens. We preheat the recovery media to 50 °C because mixing the 4 °C cells in the chilled cuvette lowers the temperature of the media. Preheating the media to 37 °C would result in a temperature approximating 30 °C after pipetting and mixing with the cold cells. Afterwards we put the electroporated aliquots in a 37 °C water bath to help them recover while we are treating the other samples. More cells survive, therefore the efficiency increases.

In the paper of Weinstock et al. 2016, a voltage between 700 and 900V is recommended. One of our advisers said that a higher voltage increases efficiency so we tested voltages of 700V up to 1600V in steps of 100V.

The difference between 1600V and 900V was not noteworthy. It might be that with an increasing voltage (efficiency)also more cells die during the procedure, resulting in a consistent final efficiency. Electroporating 24 samples takes time and during the first and the last electroporation the first cells have about 15 min more time to recover, directly i a water bath. After applying these new changes the efficiency increased we are now able to successfully transform with concentrations as low as DNA of 0.1ng. Which is nearly what is achieved when working with E. Coli with heat shock. E. Coli still performs better in this, but we are working with the Vibrio natriegens wild type and it is already good enough for cloning.

Flow cytometry

At many different occasions, at meetups or conferences, we showed the growth curve of V. natriegens compared to E. coli (Link). Other scientists were impressed by the extremely fast growth rate but even more by the high OD600 that we could show. We were asked many times if the high OD600 is really due to a high cell density or if it is rather caused by other components like cell debris or substances secreted to the medium which contribute to the measured absorbance.
To find out more about the growth dynamics, we decided to acquire a growth curve of V. natriegens in the most direct way, by counting cells in a flow cytometer. We inoculated three baffled flasks from stationary cultures and took samples in 15 minute intervals while the bacteria were incubated at 37 °C with shaking at 220 rpm. The OD600 of these samples was measured in a normal photometer and the cultures were then immediately analyzed by flow cytometry. The flow cytometer directs the samples through a thin capillary so that the cells pass a laser bean one by one and thus, can be counted and analyzed independently. A constant flow rate and time for data acquisition was set, which results in measuring a defined sample volume. Together with the counted events, the cells per volume can be calculated.

Figure 1: Comparison of OD600 and events/µL
OD600 is shown in red and events/µL measured with flow cytometry is shown in blue
A comparison of the OD600 to events/µL values are shown in figure xxxx.
When we planned this experiment, we were most curious about the composition of the culture in the stationary phase to answer the question if the high OD that V. natriegens can reach is the result of a high cell density or if it can be traced back to other substances. Interestingly, both values, OD600 and events/µL start to stagnate at a similar time point (165 min). We interpret this result as a confirmation that the high OD is indeed caused by bacterial cells.

By carefully comparing the shape of both growth curves, we realized that, in fact, the most striking data in this plot can be found at the beginning of the experiment. While exponentially increasing values can be seen right from the start for the curve created from the OD600 data, a short lag phase is apparent when events/µL are plotted (figure xxxx). We tried to find an explanation for this observation and realized that the absorbance of a culture does not necessarily correlate with the concentration of cells but rather with the biomass inside the flask.
Figure 1: Scatterpolt acquired by flow cytometry
A single sample after 45 minutes is shown as an example. Each dot represents one event
Fortunately, additional data can be obtained from the forward and side scatter of the laser beam in a flow cytometer which provide information about the size and inner complexity of the analyzed cells, respectively. Figure xxx exemplary shows one sample at t = 45 min. The side scatter (SSC-A) is plotted on the Y-Axis versus the side scatter (FSC-A) on the X-Axis. Each dot in this scatter plot represents one detected event and a heatmap can be used to visualize many events with the same properties (red = high event cont; blue = low event count). The population in the top right corner comprises roughly 98 % of all events and represents fully viable cells, since this is the population that increases in number throughout the growth experiment (peaking at >99% of all measured events at mid-log phase). The population in the bottom left corner most likely consists of sick or dormant cells and cell debris, since the number of events in this population remains low throughout the experiment.

Figure 1: Comparison of forward scatter versus events/µL
The measured events/µL are shown in blue and on the left Y-axis. The forward scatter is displayed in green and on the right Y-axis
We plotted the mean forward scatter values of all cells together with the events/µL. The mean side forward value dramatically increases during the first data points with a peak after 45 minutes. This is also the same time point which we identified as the beginning of the exponential growth phase. During the subsequent course of the experiment the forward scatter values decrease again, reaching a minimum when the culture goes into stationary phase.
Taking all three datasets into account, we suggest that the cells start to grow in size upon provision of fresh medium but initially without undergoing cell division. This results in an increase volume of individual cells and thus, an increase of the measured OD600 but without significant changes to the cell concentration in the culture. After 45 minutes, when the forward scatter peaks, we assume that a majority of cells reach maximum cell volume initiate rapid division, quickly entering the exponential phase. During the ensuing time points, exponential growth can be observed and the decrease of the forward scatter is a hint for a reduction in mean cell size.
Figure 1: Histograms of forward scatter during the course of the experiments
The histograms are plotted from top to bottom during the time course of the experiment
To additionally visualize the composition of the measured cells in regard to the forward scatter, we created figure xxx showing histograms of the forward scatter at different timepoints. It is apparent that the population is heterogeneous at the beginning of the experiment and at the end when the cultures again reaches stationary phase. During the period of exponential growth, the sample is more homogenous. The already discussed trend in the forward scatter curve can also be observed with these histograms which show a shift to the right when the forward scatter peaks and a shift to the left for the following time points.

We want to thank Dr. Max Mundt who carried out the experiments with us and who helped with analyzing the data.


Mutation rate in Vibrio natrigens

When we first presented our project, many people asked about the mutation rate of Vibrio natriegens, worrying its high growth speed would be accompanied by a higher mutation rate. Some were concerned that random mutations could undo their work in the lab, or lead to a pathogenic Vibrio natriegens mutant, others were hoping this could speed up their mutation experiments.

The mutation rate is the frequency with which new mutations appear in an organism. A mutation in a gene can be a silent mutation, meaning it has no effect, or lead to gene function loss, or alter the genes function. These mutations can be the result of DNA-damage, caused for example by radiation. They can also happen spontaneously through mistakes the DNA-polymerase makes during DNA-replication. The spontaneous mutation rate of most organisms is known, such as Escherichia coli’s and Vibrio cholerae’s. It was determined by selecting for mutants which had changed phenotype, (S. E. Luria and M. Delbrück 1943) or in recent years, as it became cheaper, by analyzing the genome trough genome sequencing. This method is more precise than estimating the mutation rate based on a phenotypical change (Patricia L. Foster 2015) (Marcus M Dillon 2017). The mutation rate plays a major role in multiple arears of biology, influencing for example the rate with which pathogenic microorganisms gain antibiotic resistances or adapt to a new environment (Christopher B Ford 2013).

We conducted mutation experiments to estimate the mutation rate (probability of mutation per cell per division or generation) of Vibrio natriegens by using the number of mutation events and the final number of cells in a culture (Patricia L. Foster 2006). To determine the number of mutation events, mutants have to first be identified. This can be done with rifampicin, an antibiotic which inhibits mRNA transcription by obstructing its elongation path through binding to the β-chain. Some microorganisms can gain a rifampicin resistance through a specific point mutation or deletion in the β-chain of the RNA-Polymerase (Wu and Hilliker, 2017). By plating Vibrio natriegens out on rifampicin plates, thes Vibrio natriegens mutants can be selected.

By testing sister colonies of Vibrio natriegens in this manner and counting the resistant colonies we estimated the mutation rate with the Lea-Coulson median estimator (Lea and Coulson, 1949). For this we inoculated 14 sister cultures of Vibrio natriegens with the same OD600 (About 0.0001) out of an exponential pre-culture. When the sister cultures almost reached the stationary phase, we platted them on rifampicin- and on plats without an antibiotic. After evaluating these plates, we could estimate the Median number of mutants in a culture, by counting all rifampicin resistant cultures and applying the Lea-Coulson median estimator. The final number of cells in the culture was estimated by counting the colony forming units on the plates without antibiotics.

The calculated number of mutations per culture with the Lea-Coulson median estimator was 5,7925 and the estimated final number of cells in the culture were 3165000000 cells. The number of mutations per culture is divided by the final number of cell (Patricia L. Foster 2006). The mutation rate we estimated was 1,83017E-09, and there for a bit lower than the mutation rate of Escherichia coli which was estimated by Luria and Delbrück in a comparable experiment for of about with was calculated to be 3,2E-09 (S. E. Luria and M. Delbrück 1943). Since our number of mutations per culture was estimated to be between 4 and 15 we determined the number of mutations per culture with the Lea-Coulson median estimator, unlike how the mutation rate of Escherichia coli was determined. For Escherichia coli the p0 method was used, because there were  between 0,3 and 2,3 mutations per culture (Patricia L. Foster 2006). Because Escherichia coli grows to a lesser OD600, cultures do not undergo as many cell divisions, thus fewer cells have the opportunity to mutate and less mutation events take place. However, our experiment suggests, that the number of mutants per cells are higher in Escherichia coli, meaning that the possibility to choose a mutated colony is higher when working with Escherichia coli.

We are now able to answer the question of our fellow researchers regarding Vibrio natriegens’ mutation rate. After these experiments we know that concerns about unwanted spontaneous mutations when switching from Escherichia coli to Vibrio natriegens are unwarranted. It can be assumed that Vibrio natriegens can speed up mutation experiments anyway, since there are more mutations occurring in the culture, simply because there is a higher density, increasing genetic variability within it. Our data is not conclusive but enables an estimation of the mutation rate of Vibrio natriegens. A more precisely calculated mutation rate can be obtained through whole genome sequencing of multiple Vibrio natriegens cultures as has been done for Escherichia coli (Heewook Lee et al. 2012).

When we first presented our project, many people asked about the mutation rate of Vibrio natriegens, worrying its high growth speed would be accompanied by a higher mutation rate. Some were concerned that random mutations could undo their work in the lab, or lead to a pathogenic Vibrio natriegens mutant, others were hoping this could speed up their mutation experiments.

The mutation rate is the frequency with which new mutations appear in an organism. A mutation in a gene can be a silent mutation, meaning it has no effect, or lead to gene function loss, or alter the genes function. These mutations can be the result of DNA-damage, caused for example by radiation. They can also happen spontaneously through mistakes the DNA-polymerase makes during DNA-replication. The spontaneous mutation rate of most organisms is known, such as Escherichia coli’s and Vibrio cholerae’s. It was determined by selecting for mutants which had changed phenotype, (S. E. Luria and M. Delbrück 1943) or in recent years, as it became cheaper, by analyzing the genome trough genome sequencing. This method is more precise than estimating the mutation rate based on a phenotypical change (Patricia L. Foster 2015) (Marcus M Dillon 2017). The mutation rate plays a major role in multiple arears of biology, influencing for example the rate with which pathogenic microorganisms gain antibiotic resistances or adapt to a new environment (Christopher B Ford 2013).

We conducted mutation experiments to estimate the mutation rate (probability of mutation per cell per division or generation) of Vibrio natriegens by using the number of mutation events and the final number of cells in a culture (Patricia L. Foster 2006). To determine the number of mutation events, mutants have to first be identified. This can be done with rifampicin, an antibiotic which inhibits mRNA transcription by obstructing its elongation path through binding to the β-chain. Some microorganisms can gain a rifampicin resistance through a specific point mutation or deletion in the β-chain of the RNA-Polymerase (Wu and Hilliker, 2017). By plating Vibrio natriegens out on rifampicin plates, thes Vibrio natriegens mutants can be selected.

By testing sister colonies of Vibrio natriegens in this manner and counting the resistant colonies we estimated the mutation rate with the Lea-Coulson median estimator (Lea and Coulson, 1949). For this we inoculated 14 sister cultures of Vibrio natriegens with the same OD600 (About 0.0001) out of an exponential pre-culture. When the sister cultures almost reached the stationary phase, we platted them on rifampicin- and on plats without an antibiotic. After evaluating these plates, we could estimate the Median number of mutants in a culture, by counting all rifampicin resistant cultures and applying the Lea-Coulson median estimator. The final number of cells in the culture was estimated by counting the colony forming units on the plates without antibiotics.

The calculated number of mutations per culture with the Lea-Coulson median estimator was 5,7925 and the estimated final number of cells in the culture were 3165000000 cells. The number of mutations per culture is divided by the final number of cell (Patricia L. Foster 2006). The mutation rate we estimated was 1,83017E-09, and there for a bit lower than the mutation rate of Escherichia coli which was estimated by Luria and Delbrück in a comparable experiment for of about with was calculated to be 3,2E-09 (S. E. Luria and M. Delbrück 1943). Since our number of mutations per culture was estimated to be between 4 and 15 we determined the number of mutations per culture with the Lea-Coulson median estimator, unlike how the mutation rate of Escherichia coli was determined. For Escherichia coli the p0 method was used, because there were  between 0,3 and 2,3 mutations per culture (Patricia L. Foster 2006). Because Escherichia coli grows to a lesser OD600, cultures do not undergo as many cell divisions, thus fewer cells have the opportunity to mutate and less mutation events take place. However, our experiment suggests, that the number of mutants per cells are higher in Escherichia coli, meaning that the possibility to choose a mutated colony is higher when working with Escherichia coli.

We are now able to answer the question of our fellow researchers regarding Vibrio natriegens’ mutation rate. After these experiments we know that concerns about unwanted spontaneous mutations when switching from Escherichia coli to Vibrio natriegens are unwarranted. It can be assumed that Vibrio natriegens can speed up mutation experiments anyway, since there are more mutations occurring in the culture, simply because there is a higher density, increasing genetic variability within it. Our data is not conclusive but enables an estimation of the mutation rate of Vibrio natriegens. A more precisely calculated mutation rate can be obtained through whole genome sequencing of multiple Vibrio natriegens cultures as has been done for Escherichia coli (Heewook Lee et al. 2012).

Results Part Collection
We created a simple and reliable workflow for the characterization of parts from the Marburg Collection in V. natriegens
(Link to Measurement). Experimental data for constitutive and inducible promoters, RBS strength, terminator readthrough, ori dependent plasmid copy number and the behavior of our newly designed connectors were obtained.

After creation of the Marburg Collection, we wanted to characterize the parts in V. natriegens. When we started with our project, we had no clue about the behavior of the genetic parts that were integrated into our toolbox. Previous research mainly focused on microbiological description rather than characterization of synthetic constructs as we already discussed in our V. natriegens review (Link to Description).
We decided to characterize the parts in our Marburg Collection and hence we did pioneering work to provide the scientific community the data that enable rational utilization of V. natriegens for various applications in synthetic biology.

Before acquiring the final data, we established a fast and convenient platereader workflow that is tailored to the fast growth rate of V. natriegens. We demonstrate its superior performance and discuss considerations in terms of plasmidal context and data analysis on our measurement page (Link to Measurement).

Promoter Characterization

After having established an experimental and data analysis workflow and after determining the optimal plasmidal context for reporter experiments, we started to apply our knowledge to characterize the parts in our Marburg Collection.

Figure 1: Relative promoter strength of Anderson promoters
Data were normalized over the strongest construct J23100. Error bars represent the standard deviation of the measurements of three independent experiments
We started by measuring the promoter strength of the Anderson Promoter library in V. natriegens. Firstly, we assembled 19 test plasmids with golden-gate-assembly and measured their expression strength, following our selfmade workflows. The results are shown in figure 1. We observed an even distribution of the tested promoters throughout the dynamic range. The strongest promoter (J23100) yielded 40 fold stronger signal than the promoter dummy and was used as a reference to calculate relative promoter strengths. The test constructs were built with dummy connectors which did not possess insulator elements. We assume that this resulted in additional expression caused by transcription throughout the rest of the plasmid, e.g. ori and antibiotic resistance. This is thought to add the same extent of signal to all measured promoters thus reducing the overall dynamic range. To further evaluate this assumption, we could repeat this experiment with one of our insulators instead of the dummy connector.

Characterization of pTet

In addition to constitutive promoters, the Marburg Collection contains two inducible promoters, pTet and pTrc. For all experiments with inducible promoters, we added the respective inducer concentration to the preculture as well as to the main culture to ensure constant expression.The first experiments were performed with the pTet promoter that can be induced by the tetracycline derivative anhydrotetracycline (ATc). ATc is much less cytotoxic but still capable of binding and altering the structure of the repressor TetR, leading to release of the promoter and enabling transcription.

To measure the dose response behavior of the pTet, we made a dilution series of ATc. Following the recommendation of our advisors (Stefano Vecchione), we started with the concentration commonly used in E. coli, started with the concentration (100 ng/mL). The starting concentration was diluted twofold in 20 subsequent steps. Our results are shown in figure 2. The absence of bars for the four highest concentrations is due to the fact that the cultures did not reach an OD600 of 0.2 in the six hours of the measurement.
Remarkably, we observed reasonable growth of those same cultures in the preculture already induced with the identical amount of ATc. Knowing that luminescence is produced at the end of an enzymatic cascade, starting with intermediates of the phospholipid metabolism ( Meighen 1991), we reckon that very strong induction could decrease the fitness of cells and that after dilution in room temperature medium, strained cells are not able to recover from the stationary phase.
However, we only observed this phenomenon in experiments with pTet, although we obtained higher signals for the strongest constitutive promoters as well as for the highly induced pTrc. We checked for toxicity of ATc but could not see a measurable effect. Another possibility is that TetR interacts with components inside the cell and that high ATc increases these interactions.
Blast searches of TetR against the genome of V. natriegens identified one protein that shares some homology with the N-terminal part of TetR which could result in cross talk between the host and the inducible promoter.
Figure 2: Dose response of pTet with ATc.
J23100 was used as positive control and for normalization. Error bars represent the standard deviation of the measurements of three independent experiments
All measured data were normalized to the strongest constitutive promoter J23100. Saturation occurred at a dilution of 2^6 (~ 1.6 ng/mL) and an exponential reduction of luminescence signal can be observed for higher dilutions. In the absence of ATc, the signal is twelve fold lower compared to saturation.
pTet allows relatively tight control of gene expression and is therefore well suited for driving the expression of potentially toxic proteins. On the other hand, we were not able to induce strong expression that can compete with strong constitutive promoters or the fully induced pTrc.

Characterization of pTrc

pTrc is the second tested inducible promoter. It contains lac operator sites and is therefore regulated by the repressor LacI which is constitutively expressed from a downstream gene.
pTrc can be induced Isoopropyl-β-D-thiogalactopyranosid (IPTG), a chemical derivative of lactose ( Camsund et al.2014). Similar to our experiments with pTet, we made a dilution series starting with the commonly used IPTG concentration for E. coli 0.5 mM. We observed a five fold induction and a saturation that occurred at a dilution of 2^5 (~15 µM). The strongest expression is similar to the expression gained from the strongest constitutive promoter J23100 while the expression in the absence of inducer equals medium strong promoters.
As a consequence, we do not recommend using pTrc in constructs where a tight control of gene expression is desired. Instead, pTrc is well suited when strong expression is required.
Figure 3: Dose response of pTrc with IPTG.
J23100 was used as positive control and for normalization. Error bars represent the standard deviation of the measurements of three independent experiments

Taking the results of both inducible promoters into account, we made two observation. In both cases, the dynamic range is smaller compared to E. coli and the inducer concentration that facilitates saturation is 32 and 64 fold lower for pTrc and pTet, respectively, than the concentration that is typically used for E. coli. A possible explanation could be found in the fast growth of V. natriegens which might result in a lower concentration of the repressor proteins in the cells, finally leading to a less restricted control of the negatively regulated promoters. However, we do not have experimental support for our idea.

Characterization of Connectors

One novel key feature of our toolbox are the connectors. They were designed in order to function as insulators to prevent crosstalk between neighboring transcription units (Link to Design). Therefore a perfectly insulating connector would prevent the readthrough from backbone sequences that most probably caused the notably high expression that was measured in the promoter experiment for the dummy promoter. In addition to blocking transcriptional readthrough, a good connector must not possess any cryptic promoter activity.

We focused on characterizing the 5’ Connector because we expect the stronger influence on signal strengths. For characterizing our connector parts, we created 20 test plasmids with the lux operon as the reporter.
In our toolbox we provide five short connectors, which solely possess the fusion sites for LVL2 cloning, and five long connectors which additionally harbor self-designed insulators (Link to Design).
Each of these ten connectors were cloned with the constitutive promoter J23100, to check for effects on an active promoter, and with the Promoter Dummy to quantify the extent of transcriptional activity that reaches the Promoter Dummy.
A
B
Figure 4:
Results of Connector measurments

A) Connector constructs built with J23100 as promoter part
B) Connector constructs built with the Dummy Promoter as promoter part
The acquired data are shown in figure 4. The data were normalized over the test construct J23100, that was used in the promoter experiment and constructed with the connector dummies. For the five constructs with the active promoter and the long connectors we observed extremely varying signals (figure 4, A). We measured a range from 0.2 to 2 fold change compared to the reference construct. It has been shown that the sequence directly upstream of small synthetic promoters can greatly impact the transcription efficiency ( Carr et al.2017). In case of the long connectors, the sequence upstream of the promoter forms the terminator and could affect the efficiency of RNA-polymerase binding to the -35 and -10 regions.
For the constructs built with small connectors, we also observed varying signals but to a lesser extent compared to the long connectors (figure 4, B).
For all ten connectors that are provided in our toolbox, we show a tenfold range in the measured luminescence/OD600 signal. As a conclusion, we recommend to carefully consider the combination of promoter and 5’ Connector for rationally designing constructs.

Taking a look at the constructs that were built with the Promoter Dummy, we also see a huge difference in the expression signals. For the long connectors we expected a negligibly low reporter expression which we observed for two out of five long 5’ Connectors resulting in a 14 fold signal reduction compared to the “Promoter Dummy” reference. We interpret this result as the confirmation that those two connectors function as insluators!
The remarkably strong signal observed for the remaining three connectors could be due to inefficient terminators or cryptic promoters in the pretended “neutral sequence”.

For the remaining five constructs possessing the five short 5’ connectors we observed a range from 0.3 to 5.5 fold compared to the “Promoter Dummy” reference. We are not able to give an experimental explanation for this observation but we could imagine that the LVL2 fusion sites, the only four bases that differ in these constructs, could constitute a weak promoter together with surrounding sequences.
Summarizing the connector characterization, we found that sequences upstream of short synthetic promoters greatly affect reporter expression, which is in accordance with literature ( Carr et al.2017). Moreover, we demonstrated that two of our five self-designed connectors efficiently reduce the signal resulting from other sources than the actual promoter. We additionally conclude that algorithms that predict the “neutrality” of sequences alone are not sufficient to create well functioning insulators.

Characterization of Terminators

Figure 5: Terminator test construct
LVL2 plasmids were created for these experiments consisting of a RFP transcription unit with the strong constitutive promoter J23100, followed by the lux operon with the promoter dummy. The terminator located at the 3' end of the RFP transcription unit is the part which is characterized in this experiment.
The Marburg Collection contains five terminators plus one terminator dummy that can be used as a placeholder. To obtain experimental data for this category of parts, we built a set of terminator test constructs to measure the extent of transcriptional readthrough and therefore the strength of a terminator. The terminator test constructs are built as LVL2 plasmid with our toolbox. The strongest constitutive promoter J23100 drives the expression of RFP which is the first transcription unit. The Lux operon is placed downstream with the promoter dummy instead of an active promoter. Both transcription units are separated by the terminator, which is the focus of this characterization, downstream of the RFP CDS

With this setup, transcriptional activity of the RFP reporter is blocked by the terminator. Therefore the measured luminescence signal can be seen as an indicator for the efficiency of the terminator.
As discussed previously, RFP is not suitable for precise quantitative characterizations (Link to Measurement). Therefore we did not calculate ratios of the reporter upstream and downstream of the tested reporter as was described in previous experiments ( Chen et al.2013). However, we used the existence of an RFP signal as control for the correctly assembled test constructs and for the activity of the promoter driving RFP.
Figure 6: Characterization of terminator read through
Displayed is the relative luminescence signal to the control construct J23100 obtained for constructs with the respective terminator. Error bars represent the standard deviation of the measurements of three independent experiments.

The data shown in figure 6 were acquired and analyzed following our novel workflow described on the measurement page (Link to Measurement). Like in all previous experiments, the obtained raw data for each sample were normalized over the construct J23100 from the promoter characterization.The strongest signal was observed for B1002 and B0010 with a relative signal 0.65 and 0.50 respectively, suggesting these two terminators as rather inefficient. In contrast, we could show a signal reduction four and eight fold for B0015 and B1006, respectively.

By comparing our data to the characterization provided in the iGEM registry for E. coli, we can show that the general trend is similar for both organisms. Exemplary, the terminator with the highest described efficiency in E. coli (B1006) also was found to reduce the luminescence signal most in our experiments.
In general, we found stronger signals for the reporter downstream of the terminator than what was described for E. coli. However, it has to be noted, that we used the highly sensitive reporter Lux instead of a fluorescent protein. Therefore we assume that we were able to detect a higher degree of transcriptional readthrough, which would not be distinguishable from the background when using a fluorescence reporter.
However, we cannot exclude species specific differences that cause a generally higher degree of transcriptional readthrough over terminators for V. natriegens compared to E. coli

In addition to the terminators shown in figure 6 we also measured a test construct possessing B1003. The, by far, lowest signal was found for B1003 with a 15000 fold reduction of the signal. To ensure that this extremely weak signal is authentic and not caused by cross talk from neighbouring wells, we tested this sample in an otherwise empty 96 well plate and confirmed the general existence of a signal. However, we can not exclude the presence of single mutations within the lux operon dramatically diminishing the overall generation of luminescence. This result was not displayed in figure 6 to omit extreme stretching of the Y-Axis thus loosing visual information for the other tested terminators.

Measuring RBS strength

For quantifying the RBS strengths, unfortunately, we could not use our favorite reporter the lux operon because in case of this operon, each CDS possesses its own RBS. Therefore, replacing the RBS upstream of the first CDS (LuxA) alone does not suffice to achieve a difference in reporter expression dependent on the RBS strength. Consequently we used sfGFP, the reporter that showed the second best performance in our initial reporter experiment (Link to Measurement).

Figure 7: RBS strength measured with sfGFP
Test constructs were built with the same parts except for the RBS part. The sample "Empty" represents V. natriegens with a plasmid without a sfGFP reporter. The error bars indicate the standard deviation of four technical replicates.
We built test constructs that are identical in all used parts except for the used RBS. The resulting data are shown in figure 7. The sample labeled with “empty” represents V. natriegens with a plasmid without a reporter. Apparently, the difference between the tested constructs expressing sfGFP in different amounts does not differ much from the non expressing control. However, it is possible to obtain some information about the order of the tested RBS. The strongest signal was observed for B0034 followed by B0030. These RBS are also described as strong in the iGEM registry. In case of B0031 and B0032, the measured signal is lower than for the non-sfGFP expressing strain. Therefore no conclusion can be provided for these two parts.

We expected that the order of the strength of the tested RBS should be similar to E. coli. Prior to the experiment, we created an alignment of the 16S rRNA of both organisms and found that the Anti-Shine-Dalgarno sequence, the bases that are responsible for binding the Shine-Dalgarno sequence on the mRNA, do not differ between both organisms (figure 8).
Figure 8: Alignment of 16S rRNA of E.coli MG1655 and V. natriegens ATCC14048
The Anti-Shine-Dalgarno sequences is indicated with an additional annotation. The 3' sequence of the 16S rRNA does not differ between both organism


On the measurement page (Link to Measurement). we suggested the lux operon as the reporter of choice for all experiments that focus on measuring transcriptional activity . We see the RBS characterization as the confirmation that using sfGFP as reporter does not yield reliable data for very weak expression.
In future experiments, enzymatic reporters such as LacZ or β-glucuronidase (GUS) could be tested for their suitability in experiments for the quantification of translational efficiency or post translational effects (e.g. degradation tags)

Analyzing degradation tags

Our toolbox contains the three degradation tags M0050, M0051 and I11012. Similar to our RBS experiments, we could not use the lux operon as a reporter because the degradation tag is only added to the C-terminal end of the last enzyme encoded in the operon. Therefore, we again used sfGFP as reporter and test constructs were designed with one of the tags fused to the CDS of sfGFP. A non-tagged sfGFP construct serves as the reference.

Figure 9: Fluorescence signal for degradation tag test constructs
The respective degradation tags were appended to the C-terminus of sfGFP. The error bars indicate the standard deviation of four technical replicates.


As expected, appending a degradation tag to a protein decreases its concentration. The strongest decrease and therefore the highest degradation was shown for M0050 with a signal that is not distinguishable from the non-sfGFP expressing control. The second strongest signal reduction was shown for I11012, followed by M0051, which showed the least efficient degradation (figure 9). Our results are in qualitative accordance with the description of these tags for E. coli, which are provided in the registry.
The tested degradation tags belong to the family of SsrA degradation tags. Naturally, they help to degrade incompletely translated proteins by labeling them for Clp proteases (Farrell et al. 2015). We checked for the presence of ClpP, one of the proteases involved in degrading SsrA tagged proteins in E. coli and found a highly homologous protein in V. natriegens. Therefore we assume that the mechanism of Clp mediated degradation is conserved between both organisms, which is in accordance with our results.

Influence of Codon optimization

Each organism has a preferred codon usage that affects the efficiency of translation through the abundance of tRNAs (Rocha, 2004). The tRNA composition differs between organism which can result in a loss of protein expression. The goal of our project is to replace E. coli in as many applications as possible. We know that scientists all over the world have been extensively working with E. coli for decades and have collected huge collections of plasmids for this organism. We already showed in many experiments that parts taken from E. coli are in general functional in V. natriegens.

Figure 10: Comparing codon optimized sfGFP with non-optimized sfGFP.
The error bars indicate the standard deviation of four technical replicates.
However, we wanted to test if optimizing the codons of a CDS can enhance the expression levels. Therefore, we ordered the DNA sequence of a sfGFP CDS with optimized codons for V. natriegens. In figure 10 we compare the signal of two constructs with and without an optimized sequence. Our experimental data suggest that codon optimization results in a considerable increase of the fluorescence signal. In conclusion, while an acceptable level of expression can be observed for parts without optimizing the codon usage, newly synthesized sequences optimized for V. natriegens can increase the expression levels. This information could prove important for industrial applications where even small changes in product yield can decide if an application is economically advantageous. We have to note that this conclusion is based on a single experiment with a single tested CDS. To further confirm the impact of codon optimization on expression levels, this experiment should be repeated with various sequences.

Characterization of origins of replication

Origins of replication (Oris) are genetic elements where DNA replication is initiated. In plasmids the Ori sequence is responsible for it’s maintenance and for the copy number inside the cell (Selzer et al., 1983; Brantl, 2014).

The origins of replication colE1, pMB1 and p15A belong to the same family. They do not code for any enzyme but are replicated by the hosts RNA polymerase (Cesareni et al., 1991; Brantl, 2014). The polymerase transcribes a region 508 bp upstream the Ori sequence (Tomizawa & Itoh, 1981; Selzer et al., 1983) synthesizing a pre-primer RNA called RNA II. During transcription the RNA II underlies conformation changes building secondary structures(Brantl, 2014). This structures contain typical loops (Cesareni et al., 1991) that binds to the plasmids’ Ori sequence building an RNA-DNA hybrid (Cesareni et al., 1991; Brantl, 2014). The RNA II is than cleaved by the hosts RNase H to become a mature primer (Cesareni et al., 1991; Brantl, 2014).

For our collection we characterized three Oris commonly used in molecular biology: colE1, pMB1 and p15A. We measured two different plasmids, one with and another without a LUX cassette. Both plasmids consist of a kanamycin resistance cassette and one of the three Oris described. The LUX expression plasmid contained a constitutively expressed LUX cassette of ~6kb. The other one contained a connector sequence to build an ‘empty’ plasmid. By comparing this constructs you may consider that the copy number is not only influenced by the LUX expression but also by the plasmids sizes. This Oris belong to the same family differing in mutations in the RNA I region (Tomizawa & Itoh, 1981; Selzer et al., 1983).

We measured the plasmids’ copy number by qPCR using the absolute quantification method.
A qPCR is set up the same way like a normal PCR but with addition of a DNA binding fluorophore in this case SYBR Green. SYBR Green binds double stranded DNA emitting a high signal while unbound SYBR Green shows only low fluorescence (Zipper et al., 2004). In every PCR cycle the number of double stranded DNA is duplicated emitting an increasing fluorescence signal. This signal is detected after every cycle by the qPCR machine and the value is saved. After the run finished, normally after ~40 cycles, a signal threshold is determined and the corresponding cycle when the threshold was reached is saved for further analysis.
For the qPCR run first total DNA from our host containing the plasmids of interest was isolated in the exponential phase (OD600 ~ 0.5), purified using the innuPREP Bacteria DNA Kit from Analytik Jena and all samples normalized to ~5ng/ul with the Qubit fluorometer from ThermoFisher scientific. Subsequently a dilution series was made in 1.5ml tubes diluting the DNA 7 times 1:2. This way the dilution series contained 8 steps reaching from 20 to 2-7. Two different primer pairs were used for the analysis: one matching the housekeeping gene dxs present once on the genome and the other matching the kanamycin resistance cassette on the plasmid. The DNA samples used for the amplification of the kanamycin cassette were the same used for the dilutions 2-4 and 2-5. The threshold cycles (Ct) acquired in triplicates from the dxs sequence were used for a standard curve. By comparing the Ct values from the resistance cassette with the corresponding standard curve the number of copies could be determined as multiples from the dxs sequence. It should be considered that the dxs sequence is coded on the first chromosome of V. natriegens at ~ one o’clock. Due to that probably the sequence is present more than once because of multifork replication of the genome.

To build the standard curve the Ct values were plotted on the y-axis and the corresponding dilution steps on the x-axis. The x-axis was set logarithmic and the standard curve was calculated with Excel. The curve’s formula was than used to calculate the corresponding x-value from the resistance cassette’s Ct values. Because the x-values describe a theoretical dilution the Ct values were multiplied with this value and with their corresponding dilution to obtain the final amount of multiplies compared to the genome. For every Ori an own standard curve was calculated.

In our experiments we showed that the plasmids’ copy number controlled by three different Oris differ a lot when comparing V. natriegens with E. coli.
One possible explanation might be different expression levels of RNA I and RNA II respecting the rate of RNA I – RNA II bounds (Cesareni et al., 1991) due to the divergent metabolism in V. natriegens and E. coli. Another plausible explanation might be the different methylation patterns in both organisms probable affecting the formation of the RNA II secondary structures and subsequently its binding affinity to the DNA (Russell & Zinder, 1987; Cesareni et al., 1991).

It was shown that mutations especially in the loop I structure might be responsible for Ori compatibility and copy number control (Selzer et al., 1983; Cesareni et al., 1991). The copy number is mainly determined by two factors: the binding efficiency of the RNA II to the DNA – specially controlled by the stabilization of stem-loop IV – (Cesareni et al., 1991) and the interference of the complementary RNA I to the RNA II pre-primer (Brantl, 2014). The RNA I is transcribed constitutively from the complementary strand from RNA II pre-primer (Brantl, 2014). Binding of RNA I to RNA II prevents the correct folding of the pre-primer (Brantl, 2014). This way the RNA-DNA hybrid can not be formed and subsequently the primer maturation can not take place (Brantl, 2014).
Figure 1: Quantification of plasmid copy number in dependency of different Oris.
The columns show the average of the calculated multiplies for the different plasmids. The blue columns show the numbers for the plasmids containing a ~6kb LUX cassette. The orange columns show the numbers for the ‘empty’ plasmids without reporter. For every column six measurements have been calculated. Looking at the ‘empty’ plasmids it is clearly shown that colE1 and p15A remain high copy plasmids like in E. coli with a copy number of ~200 copies per cell. For pMB1 the copy number is scaled down becoming a low copy number Ori in V. natriegens. Looking at the LUX plasmids it is clearly shown that the colE1 Ori remains at a high copy number while pMB1 and p15A drop down to a significantly lower level.

Metabolic engineering
Biosensor

In the first major experiment we tested induction of lux expression over a 3-HPA gradient. When we first saw the plate reader data, we were very optimistic that we had successfully cloned a functional 3-HPA sensor into V. natriegens. Just from the luminescence values, a clear trend to more emission at medium concentrations was apparent.

To explain this, we argued that a higher concentration of 3-HPA caused so much stress, the cells died, and that lower concentrations were insufficient for induction. The difference between the uninduced and the most highly expressing sample was about six-fold.

Two figures. The top figure is pH corellated to 3HPA concentration. The pH is lowest, 4.7 , for the concentration of 10mM, the highest concentration we tested. It continually rises to about 6.5 without 3HPA. 
		The second figure shows luminescence corelated to 3HPA concentration. Numinescence is very low for the highest 3HPA concentrations, climbing at medium concentrations and dropping again to about halve of the maximum at low to no 3HPA
Figure 1: A) The concentration of 3HPA plotted versus the pH
B) The concentration of 3HPA plotted versus the luminescence
This was lower than the values reported by Hanko et al. 2018.

What we also observed was a severe loss of vitality, materializing in form of significantly reduced growth which completely ceased at the two highest concentrations, 10 mM and 6.6 mM respectively. This would represent a significantly lower resilience to 3-HPA than previously reported for E. coli (Liu et al. 2016). There, the threshold at which first detrimental effects could be seen was 200 mM. If we want to make industrial production of 3-HPA in V. natriegens a reality, a way to increase its resistance is indispensable. Fortunately, strategies to implement such increased resilience have already been shown to work in E. coli (Liu et al. 2016). So, given the time, similar solutions for V. natriegens are not hard to imagine.

Several different factors could be responsible for the toxicity. At first, we thought that 3-HPA's structural similarity to cellular substrates, like for instance lactate, could cause inhibition of enzymes acting on these substrates. The mechanism of toxicity for 3-HPA is most likely its acidity. 3-HPA is a weak acid with a pKs of 4.50, and in our basic research (https://static.igem.org/mediawiki/2018/3/3a/T--Marburg--pH_Tolerance_Assay.pdf) we showed that a low pH does indeed slow the growth of V. natriegens, or can even be lethal. This could be compensated in part by choosing an alkaline starting pH of the culture media. A step that is favored, as V. natriegens is less susceptible to a high pH up to 9.5.

However, this does not explain the higher luminescence for medium concentrations, where growth does not appear to be impacted.

When we took a closer look at our control, the lux operon driven by J23107 ( http://parts.igem.org/Part:BBa_J23107 ), we found the same pattern of dead cells at high concentrations, high lux expression in the middle and background expression at lower concentrations. J23107 is a constitutive promoter of medium strength, chosen to be close to the expression levels we saw from our uninduced sensor. Regrettably, this shows that our biosensor is not functional, but presents us with an interesting enigma to solve.

It is plausible that 3-HPA , which is very similar to some substrates for enzymes of the lux operon, could in some way interact with said enzymes. Another explanation voiced by a team member is based on the fact that through division, cells dilute their cell content. This includes metabolites, DNA and proteins. If the cells divide more slowly, they have more time to accumulate these. To accumulate the proteins responsible for lux luminescence, which would result in a higher per OD luminescence. Our observations do not quite fit this, since we not only see a higher luminescence per OD, but a much higher luminescence overall. Also conceivable is, that resources that would normally go into cell growth are diverted into lux expression.

Luminescence is very low for the highest 3HPA concentrations, climbing at medium concentrations and dropping again to about halve of the maximum at low to no 3HPA
Figure 2: The concentration of 3HPA plotted versus the luminescence

There are no sources to substantiate that claim, but we found another promising lead in a source on the behavior of lux expression depending on the pH (Dorn et al. 2003). There, it was shown that pH is a major influence in lux expression. They observed the peak luminescence at a slightly acidic pH of around 6. This is remarkably close to the values we measured for the concentrations of 1.9 mM to 2.9 mM, where we also observed the strongest luminescence. Therefore, when luminescence is influenced that strongly by pH, that explains why our constitutively expressed lux reporter showed the same pattern. Amazingly, this effect was already apparent only five minutes after induction.

For the malonyl-CoA sensor, we had difficulties to clone the level 2 plasmids required for them to work. Once we solved this issue, testing may proceed.

Outlook:

There are several other sensors for 3-HPA as well as for malonyl-CoA as discussed in our design section.

One promising way to directly screen for productivity of our 3-HPA pathway would be to co-express the lux operon under a strong constitutive promoter. Since the substrates for the luminescence reaction are part of the fatty acid synthesis pathway, and malonyl-CoA is at the root of this pathway, a higher conversion rate of malonyl-CoA to 3-HPA should diminish luminescence output. This simple competitive assay could give a rudimentary indication of the Mcr activity.

Sources

Dorn, Jonathan G, Robert J Frye, and Raina M Maier. 2003. “Effect of Temperature, PH, and Initial Cell Number on LuxCDABE and Nah Gene Expression during Naphthalene and Salicylate Catabolism in the Bioreporter Organism Pseudomonas Putida RB1353.” Applied and environmental microbiology 69(4): 2209–16. http://www.ncbi.nlm.nih.gov/pubmed/12676702 (October 16, 2018).

Liu, Min et al. 2016. “Development of a 3-Hydroxypropionate Resistant Escherichia Coli Strain.” Bioengineered 7(1): 21–27. http://www.ncbi.nlm.nih.gov/pubmed/26709549 (October 15, 2018).

Hanko, Erik K. R., Nigel P. Minton, and Naglis Malys. 2017. “Characterisation of a 3-Hydroxypropionic Acid-Inducible System from Pseudomonas Putida for Orthogonal Gene Expression Control in Escherichia Coli and Cupriavidus Necator.” Scientific Reports 7(1): 1724. http://www.nature.com/articles/s41598-017-01850-w (October 14, 2018).

B. Marchal