Team:Paris Bettencourt/SebaTemplate

   As antibiotic-resistant bacterial strains are rising and pharmaceutical pipelines are drying up, we decided to invest in a new strategy based on specific silencing of the genes responsible for resistance through bio-engineered stealth bacteriophages. The silencing of the genes is obtained with a simple and modular system of tailor-made small RNAs. The spreading of this construct in a bacterial population is based on an autonomous phagemid/helper phage system. We demonstrated the validity of this trojan horse strategy by converting back to a sensitive state populations of bacteria initially resistant to antibiotics as chloramphenicol or kanamycin.

   Designing a silencing device

Most antibiotic-resistance result from specific bacterial proteins such as efflux pump that extract the antibiotics from the cells or enzymes that quickly metabolize the antibiotic molecules. We choose a straightforward strategy to force bacteria to stop producing those proteins by targeting their cognate mRNA with tailor-made small RNAs. In order to design our silencing device, we used the protocol described in Na et al (2013) and inserted a 24bp sequence complementary to the RBS area of the target genes inside the published scaffold sequence that allow the stabilization of the hybridization through recruitment of the Hfq protein. We designed three sRNA respectively targeting Kanamycin resistance gene located on a pCOLA Duet Vector, Chloramphenicol resistance on pACYC Duet Vector and lacZ on E. coli's chromosome.

   Designing a genetic element that spread in a population
We adopted non-lytic filamentous bacteriophages as our vectors of choice as they are good at spreading genetic elements in bacterial populations. However, being infected by a phage represents a huge burden for an individual bacteria which we thought would be detrimental for our construct to be able to maintain itself long enough in a population for this system to be clinically relevant. Therefore we chose a phagemid/helper system which is composed of two mobile interacting elements. The “helper” elements of this system is an M13 phage mutant that provides packaging proteins but largely lacks the capacity to be packaged; the “light” element called phagemid is a normal plasmid harbouring the packaging signal and can thus propagate from hosts carrying the “helper” phage. As a result a cell containing both elements will produce and secrete many encapsulated phagemids together with a small minority of helper phages.

  



Design of a sequential killing strategy
To perform our proof of concept experiments, our targeted genes were different antibiotic resistance genes on commercial plasmids that were transformed in E. coli MG1655 cells. Just as in nature where resistance genes are often found on mobile elements, our model respects the constraint of having multiple copies of the target gene to silence which would not have been the case with a chromosome located target. The experimental set up consisted of a sequential protocol: (I) Phage infection leads to the silencing of the antibiotic-resistance genes and (II) cells are plated on antibiotics to kill the sensitive phenotype and quantify the efficiency of the silencing.

   The phagemid Helper system (kindly provided by Monica Ortiz of D. Endy lab) consists of a helper plasmid (M13K07) bearing Kanamycin resistance and a phagemid (Litmus-28) bearing GFP and an Ampicilin resistance cassette. Using those different markers, we characterized the system by infecting the cells with a ratio of phage of 1 to 100 and plating appropriate dilutions on different medias (LB,Kan, Amp, Kan/Amp). We show that the phagemid is spread a thousand fold more than the helper plasmid. There are as many cells infected with the helper as cells infected by both helper and phagemid, that is to say, phages producing cells. Those results show the validity of using this system to spread a genetic element in a bacterial population. Most of the cells bear the desired genetic element and the spread is carried on by a minor quantity of cells producing the vector.

   In principle if burden of the phage system is too high the cells infected would get out-competed by the natural population. We have constructed a simple model of phage infection for our system, where we have assumed conditions similar to a chemostat (constant dilution). The evolution of the system is described the following ODE:

where R are the recipient cells, P infected with phagemid, P infected with helper, HP infected with both, p is the phagemid, h the helper, a_R, a_P, a_H, a_HP are the corresponding growth rates for each cell.

We assumed that a 0 fitness cost a_R = a_P > a_H > a_HP. Then we simulated an decrease in fitness by multiplying all infected cells growth rate by a constant factor (one minus fitness cost). On figure 4 we can see that for the phage system to be stable (at equilibrium) it requires a critical value of cell growth.

   We used our designed phagemid, bearing the sRNA anti-Cm, to infect a population of chloramphenicol resistant cells. By applying the sequential killing strategy described in Figure 3, we efficiently killed 99,1% of the bacterial population (chloramphenicol concentration 1000ug/ml). We tested this strategy with different concentrations of chloramphenicol. We show that the rise of chloramphenicol concentration directly lead to drastic reduction of the survival of the infected cells.

   In order to show the general application of our system, we then applied the same strategy on MG1665 transformed with pCOLA Duet. Those kanamycin resistant cells were infected with phages bearing our anti-Kan sRNA. 99,87% of the population was killed (standard deviation 0,23 %) at 1000ug/mL of kanamycine

   One of the main limits of our system could be the development of resistances. Indeed, there are at least two ways to become resistant to our system: by avoiding infection or by developing a resistance to the silencing. In order to get a better idea of what is the main cause of resistance we screened the survivor colonies for GFP expression. GFP positive colonies correspond to infected cells with malfunctioning silencing. We show that 70% of resistance is caused by a resistance towards the sRNA. This resistance could be due to a mismatch between the sRNA sequence and the target sequence resulting from a single point mutation.

   Several sRNA on the same phagemid : Although our silencing small RNAs were assembled in phagemids with Gibson assembly, the Biobricks cutting sites included in our BBG adapter method provide full modularity. In our case a direct application is the assembly of multiple silencing modules (promotor+sRNA+terminator) on a single phagemid vector. As the main clinical problem with bacteria is multi or even pan-resistance this modularity provide an interesting strategy to explore as it allows simultaneous downregulation of multiple target genes.

   Building a library of sRNA : As our silencing strategy rely on base pair complementarity, a main concern from the clinical point of view would be the fast emergence of resistances to silencing through single point mutations. Indeed even if the majority of mutations that would decrease the efficiency of the silencing would also lead to an impaired or non functionnal protective protein, some neutral mutations will still be made possible by the redundancy of the genetic code. To tackle this issue an interesting solution would be to build a library of small RNA containing single point mutation on those particular neutral positions and load it on our phagemids. By being one move ahead of evolution we expect that such an anticipating strategy would considerably slow down the emergence of resistance to silencing.