Discussion
Classic genomic modification to knockout out genes is usually conducted by inserting antibiotic resistance markers. This is simple and easy to implement. However, the existence of unwanted antibiotic resistance marker genes may leave some unexpected effects, which is not expected by researchers in metabolic engineering and functional genome analysis. Through two-step selection/counter-selection method, target genes can be edited without alerting other parts of the chromosome. This unmarked mutation, known as “seamless modification”, overcomes the above-mentioned shortcomings.
Although many counter-selection markers have been developed, such assacB , ccdb , I-SceI , galk, mazF (Z. Chen, Ling, & Shang, 2016; DeVito, 2008; Muyrers et al., 2000; Van Zyl, Dicks, & Deane, 2019; Wang et al., 2014; Warming, Costantino, Court, Jenkins, & Copeland, 2005) etc., excellent markers are still limited. Counter-selection genes suffer from some problems which limit their application in genomic modification, among which high background is a serious obstacle. For example, although sacB is most widely used, researchers have to face low positive rates due to the spontaneous mutations that inactivate sacB (Khetrapal et al., 2015; Y. Zhang, F. Buchholz, J. P. Muyrers, & A. F. J. N. g. Stewart, 1998). Therefore, low background is an important consideration when counter-selection genes are to be choose.
Selection stringency frequency is an indicator that shows the level of background during counter-selection. It has a crucial role of determining the ratio of correct recombinants in counter-selection (Wei Chen et al., 2019). Selection stringency is reflecting by calculating the fraction of viable colonies that escape the selection stress. The smaller the number is, the higher the stringency is (Khetrapal et al., 2015). So far, inducible toxins system performs best among all the developed counter-selection system, followed by tetA-sacBcassette. As for toxins system, the best selection stringency frequency achieved in E. coli is 1.11×10−8, which is exceeds the next best reported tetA–sacB system up to 60- fold. The selection stringency frequency even approaches that of kanamycin resistance gene (2.4×10−9) (Khetrapal et al., 2015). Excellent counter-selection markers like inducible toxins system will greatly improve the positive rates and avoid wasting time on redundant work during genomic seamless modification.
Inspired by the inducible toxins system, in this paper we developed a powerful generic counter-selection system for genomic modification, especially for some species which lack efficient genetic tools. Lysis gene E is of general killing effect on Gram-negative bacteria. Given the speculation that counter-selection based on a universal lethal gene would be applicable generally. We focused our attention on the developing and optimizing of lysis E as a counter-selection marker. Expression of lysis gene E under the control of pLpromoter efficiently killed its host. dsDNA mediated seamless modification were also successfully performed to substitute target region, and the ratio of correct recombinants is higher than kilcounter-selection marker constructed previously in our lab (Wei Chen et al., 2019). In S. marcescens , lysis gene E also can be used for counter-selection marker both through ssDNA and dsDNA mediated recombination. However, the ratio of correct recombinants through dsDNA mediated recombination is lower than that in E. coli . Maybe the difference in ratio of correct recombinants can be attributed to different gene expression regulation system used for controlling lysis gene E . After all, more lysis protein E could be produced to kill bacteria under the control of strong promoter pL compared with rigorous PBAD promoter.
Lysis gene E could be used as a counter-selection marker both inE. coli and in S. marcescens. Since it has general lethal efficiency in Gram-negative strains, it is also applicable in other Gram-negative strains as long as promoter pL andPBAD could work. Although a universal counter-selection gene based on E was developed by us, there still existed escaping colonies during seamless modification, especially in S. marcescens. This leads to relative low ratio of correct recombinants during PCR identification when dsDNA mediated recombination was conducted in S. marcescens . Of course escaping colonies also existed in ssDNA mediated recombination, high ratio of correct recombinants is attributed to the high recombination efficiency of ssDNA itself (Ellis, Yu, & DiTizio, 2001). It should be pointed out that in order to make the colonies exhibit clearly in the LB plate, only 5μl of recovered cells after transportation was plated in ssDNA recombination, while 100 μl was plated in dsDNA recombination.
To provide a good alternative tool for seamless modification, counter-selection system based on lysis gene E need to be optimized to improve its selection stringency. In our previous work, we developed a counter-selection cassette based on kil gene of lambda phage. It performs well in E. coli and its selection stringency is comparable to the next best counter-selection systemtetA -sacB . Both kil gene and E gene are short in length: the CDS of kil and E is 144 bp and 273 bp separately. We believe that combining kil and E gene in a certain way would be a good idea. An obvious benefit is that the selection stringency would be greatly improved. What is more, it will not bring any additional difficulties to the experimental process. After all, CDS length of the counter-selection cassette after combining the two genes is less than 500 bp, which is shorter than most of the counter-selection markers. We have tried to combine the two gene through fusion expression, however no selection stringency improvement was observed (data now shown). Maybe the function of proteins was affected by the changes in spatial conformation after fusion expression. We think co-express the two genes in the form of bi-cistron should be a better alternative solution from the perspective of reducing the background. Because nonsenses mutation in the first gene would not inactivate the second gene since independent products are translated, which is not applicable for fusion expression. Co-express kil and Egene through a classic RBS between their CDS indeed greatly improved the selection stringency to the level of the best reported toxins inducible system in E. coli (4.9×10−8 at ack locus and 3.2×10−8 at araB locus). Surely, it is possible to obtain better selection stringency through combiningkil and E in other ways, such as improving translation initiation of the second gene using “bicistronic design” (BCD) reported by Mutalik (Mutalik et al., 2013). Many other strategies could also be used, such as combining more counter-selection markers with short size in length. Our work, in this paper, mainly aims to provide an example for improving selection stringency based on existing markers.
Other than developing the high selection-stringent kil-sd-Ecounter-selection cassette under the control of pL promoter in this paper, we also constructed AraC/PBAD controlling kil-sd-Emarker which could be used in bacteria where high temperature is not suitable for growth. In S. marcescens , it does not perform as excellent as pL promoted kil-sd-E . By introducingaraC gene harboring plasmid, the selection stringency was improved 4- to 170- fold. In particular, the selection stringency at these loci tested in this paper is all comparable to the best counter-selection system, inducible toxins system. At araB locus in E. coli and marR-1 locus in S. marcescens , it reaches the level of 10−9, the lowest selection stringency frequency for counter-selection so far.
This gives us an enlightenment: we could integrate the regulating elements of counter-selection markers, such as araC gene, into Red recombinases genes harboring plasmid. Therefore, counter-selection cassette will be reduced greatly in size. Take AraC/PBAD-kil-sd-E-GmR double selection cassette as an example, if araC gene was integrated into the Red recombination system providing plasmid, only 1,327 bp PBAD-kil-sd-E-GmR cassette was needed for seamless modification. This will not only facilitate PCR amplification, but also improve recombination efficiency because longer substrate dsDNA decreases the chances of a λ Red recombination event (Wei Chen et al., 2019).