We performed reversion assays to investigate our mutagenic constructs. In the first step, we confirmed that we obtain functional revertants by sanger sequencing. Afterwards we quantified mutation rate of our Error Prone Polymerase I and our Genome Wide Mutator. At last we try to integrate the mutation feature into our complete system by assessing the mutation while continuous cultivation.

To quantify mutagenesis rate exactly is a very tedious task, because a multitude of factors affect the mutation rate. The most important are:

• the experimental setup
• the exact used method

Classical approaches to measure mutation rate are distinguished in reversion or forward assays, respectively. The mutation reverts a stop from a reporter (reversion assay) or mutates a protein in such a way that it gains a function e.g. the organism becomes resistant to antibiotics (mostly rifampicin forward assay). Since the rise of high-throughput sequencing, this technique starts to replace reversion assays as prime method for assessing the mutation rate due to higher sensitivity, resolution and additional information.
As the error prone polymerase I is mostly specific for ColE1 plasmids we used a reporter on a plasmid. Therefore, we decided to perform reversion assays with a stop beta-lactamase on a ColE1 plasmid.

Before going on and quantifying the mutagenesis rate we wanted to show that, our mutagenesis process yield functional revertants. Therefore, we applied a modified version of the mutagenesis protocol described in the literature (Alexander et al. 2014). In summary, we transformed the stop beta-lactamase plasmid pLA230 into E. coli JS200 cells carrying pHSG-EPPolI plasmids. After regeneration, the cells were used to inoculate a 37 °C prewarmed LB culture. After 24 hours, cells were plated in various dilution level at LB plates containing ampicillin. 10 colonies were picked, grown at 30 °C, the plasmids were isolated and sequenced.

The sequencing results show that the obtained revertants are really revertants, which mutated one base inside the stop codon to obtain fully functional beta lactamase. In some cases the reversion is not completely obvious and is only hinted as two different peaks at the same position in the interferogram.

The function of the error prone polymerase I in our experimental setup was validated by the identification of E. coli clones with different mutations. Further characteriation of the error-prone polymerase I involves the determination of the mutation rate and adapting for optimal use in the Evobody generation system.
A literature review revealed the technique of mutation accumulation, where the fraction of revertants f at two points with a large difference in cell count N is measured. Examples are measurement in the beginning and in the end of a cultivation.

However, critical points have to be considered, when calculating the mutation rate. The reporter is encoded on plasmid with copy numbers per cell ranging from 100 to 300. We postulate that one reversed beta-lactamase gene can confer ampicillin resistance to the cell. Therefore, the observed mutation rate is 100 to 300 fold higher than the true mutagenesis rate. Additionally, potential reversion sites on each plasmid contribute to the underestimation of the mutation rate. The stop codon TAA can mutate into nine different codons, which include two new stop codon (TGA, TAG). In total, there are seven mutation possibilities on every plasmid which could revert the stop beta-lactamase back to its functional state. These considerations lead to the modification of equation one to equation and.

To measure the mutation rate similar conditions for each replicate were choosen. The setup looked as following. JS200 cells carrying either pHSG-EP Pol I or pHSG-WT Pol I were transformed with 10 ng pLA230, regenerated for 60 min in SOC and transferred LB with chloramphenicol and kanamycin. After 24 h the cell count was determined by OD600, the cells were serial diluted and appropriate dilutions were plated on LB plates with ampicillin and grown at 30°C.
The reversion frequency f was determined by counting ampicillin resistant colonies and normalizing this value by the number of cells measured via OD600. Therefore, we assumed OD₆₀₀ as 5×10⁸ cells. This experiment was done in five biological replicates both with the error prone polymerase I as well as with the wild type polymerase I.

An increase in reversion frequency by a factor of 2.7×10³ when using the EP Pol I in contrast to WT Pol I was observed. In this evolution of the reversion experiment the controls pre mutagenesis were not plated, which prevents exact determination of the change in reversion frequency. This leads to us using plating to determine the viable cell count in addition to the revertant count.
Due to obvious difference between the error-prone and the wild type polymerase I, we reduced consumption of material by only characterizing the error prone polymerase in further detail. Major obstacles when trying to measure revertants in wild-type cultures were the trait offs in respect to the number of plated cells. Plating too few cell leads to no revertants, while plating too much cells leads to an unspecific growth of a strange bacterial lawn of not resistant cells on top of other dead cells. However, this experiment clearly showed that the number of mutations caused by the wild type polymerase is not measureable in this experiment.
Our next experiment was tailored to the Evobody generation process. In this proof of concept experiment Evobody-producing E. coli are selected based on their different growth rates. Simultaneously,the Evobody coding sequence was mutagenized. The recommended mutagenesis protocol for the EP Pol I includes periodic retransformation of the mutagenized plasmids. However, this is not suitable to our approach. We started with a library and therefore avoided transformations to keep the diversity as high as possible. Transformations are known bottle necks, while working with libraries. Therefore, we investigated the mutation accumulation rate by inoculating LB culture with colonies grown at 30°C and growing them to saturation at 37°C. To calculate the mutation accumulation according to equation 2 the reversion frequency were determined at two different points of the cultivation.

This cultivation experiment revealed two interesting results:

• The mutation frequency within JS200 colonies containing the EP Pol I grown at the permissive temperature of 30°C is still notable
• The reversion frequency increases after cultivation at 37°C by the factor of four.

Both results show striking similarity to a previously determined mutation rate at the permissive temperature (Camps, personal communication). Camps told us, that mutations still occur at 30°C, but the rate is reduced to only 25% of the mutation rate at 37°C.

Another observation is the huge variability in the reversion frequency at the starting point. This variability is much lower for the mutation rate after 24 h. This shows that mutations, which occur early into a mutagenic process, take an overproportional part of the whole culture. Therefore, exact determination of the mutation rate is only possible by measuring the mutation accumulation (f₂-f₁) and not the absolute mutation count (Pope et al. 2008). Using the results from Figure 3 for f₂- f₁ = (9.47±2.28)×10^-5 and Figure 2 for N_0h = 0.1 OD₆₀₀ and N_24h = 2.44 OD₆₀₀ and a plasmid copy number of 10 (Camps et al. 2003) in equation 2 this cumulates to a mutation rate of

This measures mutation rate is three orders of magnitude over the basal mutation rate of E. coli mutation rate (Lee et al. 2012). It strongly differs from the rate described by Camps of 8.1×10^-4 per bp which converts to 4.8×10^-5 per bp per generation under the assumption that Camps cultivation process involves ~16 generations.
Furthermore, Camps uses a completely different approach in quantifiying the mutagenesis rate by starting from one measurement of the reversion frequency f and using the plasmid copynumber (10) and reversion sites (7) and conversion between codon and base (3) (Camps et al. 2003). His equitation looks like

When using this formula we obtain a mutation rate of 1.11×10^-5 per bp, which is 72-fold lower than the one calculated by Camps. In our opinion, the method of mutant accumulation is a more reliable method to discern mutation rate. Especially our experiment where we used three independent cultures is a very descriptive example for this. Although the total reversion frequency at the start as well as at our end point differs strongly between the cultures, the change in reversion frequency between both measured time points is surprisingly low.

We performed reversion assays with our genome wide mutators dnaQ926(BBa_K2082116) and M6(BBa_K2082117) again using the stop lactamase plasmid pLA230.
Our first approach was to cotransform the reporter plasmid and pSB1C3:dnaQ into E coli Top10 and inoculating cultures containing arabinose or glucose directly with the transformed cells. This yielded a reversion frequency of 2.6×10^-5 induced and 1.3×10^-6 uninduced after 12 h. Because we did not plate the cells out for exact determination of the cell count this result is only a first hint at an increases mutation rate through activation of our mutator.
Our next step was to transform the mutator and the reporter plasmids in E coli Top10, plate this on LBCmKan, inoculate LBCmKan cultures and induced or repress this culture when reaching mid-log phase. The cells were plated after 24 h and counted. From this experiment, the plasmids for our high-throughput experiment were prepared.

For the observed smaller reversion frequency of the activated mutagenesis parts there can be two explanations.
The exact cell count was not determined by plating but only by OD₆₀₀. Because expression of the mutator proteins decreases cell viability (Badran and Liu 2015) the measured OD₆₀₀ value probably contains large amount of dead cells, therefore greatly inflating our supposed cell count. Badran found that induction of their M6 variant decreases cell viability to ~5 %. This would result that the accessed total cell count is 20x too high in the induced samples, resulting in a 20x to low reversion frequency.
Another possibility is that the arabinose promoter was active through very low levels of arabinose in LB. It is possible that even very low levels of leakiness can lead to some mutagenic proteins. For example in the case of dnaQ926 one protein is enough to create one proofreading deficient polymerase III complex. In normal E. coli strains this should be no problem but the Top10 strain is deficient for the arabinose metabolism and therefore low levels of arabinose are not metabolized but activate promoter activity.
We concluded that in order to obtain more reliable data we had to plate dilutions with low amounts of cells in order to determine the viable cell count in our samples. To combat the supposed problem of promoter leakiness in LB media we dicided to use LB supplemented with 20 mM glucose. This is analog to the protocol from Badran and coworkers who used a defined media with glucose.
Using all this experience, we designed another experiment to determine the mutation rate.
We transform E.coli Top10 with the reporter plasmid as well as the mutator plasmid (pSB1C3:BBa_K2082116 or BBa_K2082117), plate a portion and use another portion to inoculate a LBCmKan culture with 20 mM glucose, let this culture grow till mid-log phase and induce or repress the P_BAD. 24 h after induction the cells were plated on LBCmKan or LBAmp each with 20 mM glucose and counted revertants and the total viable cell number. This experiment was performed for each of our mutator in six replicates.

For the growth term in Equation 2 we measure the cell count right after transformation and while plating after 24 h. Therefore we could determine N₂/N₁. As plasmid copy number we used n=40 as determined in our NGS experiment.

		f2-f1	N2/N1	μ [bp-1×generation-1]
dnaQ926 (BBa_K2082116)	Induced	(3.92±1.91)×10-4	70.11	(2.3±1.12)×10-6
	Repressed	(2.30±1.41)×10-5	5530,43	(6.67±4.09)×10-8
M6 (BBa_K2082117)	Induced	(1.86±1.12)×10-3	660,85	(7.16±4.31)×10-6
	Repressed	(0.82±1.04)×10-5	3538,52	(2.51±3.18)×10-8

We determined the mutation rate of our genome wide mutator BBa_K2082117 to be (7.16±4.31)×10^-6 bp^-1×generation^-1 induced and (2.51±3.18)×10^-8 bp^-1×generation^-1 when repressed. This is almost exactly the same rate Badran and coworkers determined for their iteration of the M6 mutator (Badran and Liu, 2015), who determined a mutation rate of 6.2×10^-6 bp^-1×generation^-1 induced and 3×10_-9 bp^-1×generation^-1. Therfore our aim to reconstruct this mutator with definded iGEM parts was successful. Furthermore we showed the muation rate of BBa_K1333108 in our construct BBa_K2082116 to be (2.3±1.12)×10^-6 bp^-1×generation^-1 induced and (6.67±4.09)×10^-8 bp^-1×generation^-1. Thereby we show the functionality of this already existing part.

We showed mutation using the error prone polymerase I, which is 4-5 orders of magnitude higher than basal mutation rate.
We determined the mutation rate of the error prone polymerase I and our genome wide mutator systems. Both systems produced mutations with a frequency several orders of magnitude over the background. Therefore, our parts can be used be in vivo generation of protein libraries to optimze enzymes or proteins for various purposes.
After determining the mutation rate of our construct we also investigated them with high-throughput sequencing

• Alexander, David L.; Lilly, Joshua; Hernandez, Jaime; Romsdahl, Jillian; Troll, Christopher J.; Camps, Manel (2014): Random mutagenesis by error-prone pol plasmid replication in Escherichia coli. In: Methods in molecular biology (Clifton, N.J.) 1179, S. 31–44. DOI: 10.1007/978-1-4939-1053-3_3.
• Badran, Ahmed H.; Liu, David R. (2015): Development of potent in vivo mutagenesis plasmids with broad mutational spectra. In: Nature communications 6, S. 8425. DOI: 10.1038/ncomms9425.
• Camps, Manel; Naukkarinen, Jussi; Johnson, Ben P.; Loeb, Lawrence A. (2003): Targeted gene evolution in Escherichia coli using a highly error-prone DNA polymerase I. In: Proceedings of the National Academy of Sciences of the United States of America 100 (17), S. 9727–9732. DOI: 10.1073/pnas.1333928100.
• Lee, Heewook; Popodi, Ellen; Tang, Haixu; Foster, Patricia L. (2012): Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing. In: Proceedings of the National Academy of Sciences of the United States of America 109 (41), S. E2774-83. DOI: 10.1073/pnas.1210309109.