Team:Newcastle/Autonomous linear DNA Clock


iGEM Homepage Newcastle University BacillaFilla Homepage Image Map


Linear DNA clock: Richard

Initial idea 12 February 2010

Builds on idea about protecting against mutations. Rather than wait for a mutation to 
happen, could guarantee that genetically-engineered BS cell will not survive after a certain 
time. Idea is to linearise prokaryotic dna and insert two genes: one a repressor protein at 
a variable short distance from the cleaved site, and 2) a gene that codes for a cell-
destroying protein of some sort which the repressor protein usually inhibits. 
Remember telomeres? Idea is that every time the BS divides, its dna will become progressively 
shorter from either end. Thus the gene for the repressor protein will get eaten away. After 
time x (can be set by experimenter) the repressor gene will be rendered non-functional and 
no repressor protein is produced. This means that the second gene we inserted, which codes 
for the cell-destroying protein is free to kill the cell. No mutation, no residual problem 
of removing BS after treatment by idea 3.
At some point in evolution in the move from prokaryotic to eukaryotic genomes, the dna became 
linearised, so shouldn’t be such a hard problem to solve if we want the dna to stay this way. 
Could insert a piece of linear dna containing linear transcription machinery on it as well.

Ongoing references

Baker et al. 2007 A Novel Linear Plasmid Mediates Flagellar Variation in Salmonella Typhi.Took linear plasmid from S.Typhi and transformed into E.coli [1].

Baker et al. 2007 A linear plasmid truncation induces unidirectional flagellar phase change in H:z66 positive Salmonella Typhi. More info on this linear plasmid [2].

Cui et al 2007 Escherichia coli with a linear genome. Linearises genome of E.coli [3].


Literature research and plan [4] Presentation delivered at informal meeting on 26 Feb [5]

Linear DNA in crisis -- 02 April 2010

Please help! I have attached an attempt to model the linear dna system [6]. If my assumptions are correct, then the clock can no longer be a guaranteed kill switch.

The problem is that the clock isn't synchronous. Basically, template stands are immutable once created, will persist, and will not shorten. I have attached a doc which explains. For example in the excel printscreen, follow the green or yellow dna molecule through the replications and see that they don't shorten!

The results are interesting for telomere biology but either we find a way to create blunt ends, accept a one-cell signal, use single-stranded dna, or go back to the drawing board.

I do hope I'm wrong,

Another problem with linear DNA -- 05 April 2010

Even though the previous problem could be solved by having a synchronous population clock with weak extracellular inhibitor signals and strong intracelllar inhibitors, there is still, much to my dissapointment, a greater problem.

The idea is that the linear dna molecule becomes shorter with each round of cell division. This is flawed because after 20 cell divisions, the bacterial population wil be over 1 millon in number afterwhich it cannot divide any further. If one cell division takes 20 mins, and the loss in bp is about 100, and a gene is about 1000, then it takes ten cell divisions to eat through a single gene. Secondly, if it takes 10 x 20 minutes to eat through a single gene, then each timer is really only limited to controlling one gene at a time over a long time period.

The only solution seems to be to have the linear dna plasmids copy themselves without cell division (as Anil suggested was possible). But would this create a useful component or a cell swamped by a hundred thousand linear plasmids?

It is extremely hard now to see how this could ever be applied to repairing cracks in concrete.

There was a mechanism in eukaryotic cells that could be exploited to speed up the shortening but it requires endonucleases and dna modifications.

Probably a good idea to draft alternative ideas now unless this can be salvaged

Renewed hope for linear DNA -- 06 April 2010

I have come up with a potential solution to both problems with the linear DNA idea that I highlighted over the Easter break. (I also apologise for the 3' and 5's being the wrong way round in my diagrams and code).

Please see first slide of attached PowerPoint [7]. The idea requires four things which I could research into but thought you would know about straight away. It uses the assumptions 1) that bacteria protect their dna from nuclease activity through various mechanisms (i.e. but viral dna entering the cell is not). 2) If the dna is not protected it is eaten away by endo/exo nucleases. 3) that these protective mechanisms can be disrupted using the telomere-shortening mechanism during cell division 4) genes are not protected.

It overcomes the immortal strand hypothesis problem because exonucleases would eat into these. It overcomes the time/cap of divisions problem because whole genes are removed at each step and a unit time delay is encoded by a single protective structure. (the length of the protective structure tails would be set to align with the number of bases removed from the telomeres at each step).

A gap/ER2 model with unit shortening and a new mechanism -- 12 April 2010

Created a linear plasmid model for the gap/ER2 mechanism where the ends of the linear plasmid shorten by 1 unit according to the telomere repair option. The .ppt supports the gap/ER2 mechanism, outlines how it could be applied to a synchronous population clock and sketches a new idea in which exonuclease shortening is coupled directly to the DNA-replication machinery [8].

A new design -- 23 April 2010

Today I thought of a new design to add - Design 5: the alternating protein complex model. In outline, the ER2 (exonuclease-resistant secondary structure) in this model is a protein complexed to the DNA. The model is like the ER2 only model except that there is no specificity for exonuclease enzymes to ER2s. The pattern of activity seen in the ER2 only model is instead achieved by having two kinds of ER2 (protein complexes) in alternating pattern along the linear DNA. A signal (not necessarily cellular division) dictates which type of ER2 is disrupted, possibly by binding of the signal to the complexed protein and therefore causing the dissociation of the protein from the linear DNA, or, if the binding affinity of the protein to the DNA is weak, by transcriptional/translational inhibition of the protein by the signal. The linear DNA is then eaten away from the end by general exonucleases, through disrupted ER2s and whole genes, until the next ER2 of the second type is reached.

It has not escaped my notice that such an ER2-specific signal could be used to induce the linear DNA shortening in the first place and another signal to temporarily halt the shortening (if this is desired as well). A combination of this design with Design 4: exonuclease model is looking promising. If the rate of exonuclease attack is, or can be made slow enough (using specific sequences), then multiple genes can lie between these ER2s (with or without a 'junk' DNA biobrick interspersed between the genes). This design requires neither telomere repair structures nor genes to maintain them on the linear plasmid. It could be the case, however, that my ER2 (protein complex) is the modification of an existing telomere structure because these are already known to show resistance to exonucleases.

Another idea, is to have an event to perturb the balance of a chemical in the cell such that the whole DNA unfolds (but remains double-stranded). This means all ER2s will 'unfold' as well and, if they are positioned at the end of the DNA, will become susceptible to exonuclease attack . If the chemical balance was gradually restored (e.g. by the efflux of the perturbing molecule to the outside of the cell), then exonuclease attack would halt. Not sure this should be Design 6 because I dont want to model this in Java!

A deterministic oneRepressor model -- 28 April 2010

This is the simplest system I could conceive. Included in the .zip file is a PowerPoint presentation which describes the model [9].

The only experience I have of modelling has been during the Stochastic Systems Biology module. I have used the R code from this as a template for the Java program. Despite my inexperience, it appears that the part of the simulation before G1 becomes disrupted is showing the same dynamics in both R and Java. The full Java program compiles and the output appears correct.

I had trouble satisfying myself with drawing graphs in Java so have decided it is best to copy the formatted output over into the Excel spreadsheet which is in the .zip folder. To display the results of a simulation under a different parameter set, the Java output can be pasted over the top of the existing Excel data. This should directly transform the graph. If it does not then either a text-to-columns operation needs to be performed or the graph needs to be inserted from scratch.

I have fiddled with the parameters to get a 'good result'. I knew intuitively that the quantity of G2 needed to be near zero at the beginning so that no P2 was produced. I achieved this by making the level of repression high. Whilst bearing this in mind, I also tried to make the level of P2 synthesis high so that after G1 had been disrupted by the shortening linear DNA, the level of P2 increased quickly. I am however aware of the possibility that there exists good maths to find the optimal solution to this problem. The search continues!

Project proposal and presentation -- 10 May 2010

Outline proposal [10] and short presentation [11].

Exonuclease only model update -- 24 May 2010

Attached is a tidy model simulating the shortening of a single linear DNA molecule by exonuclease enzymes. The potential for up to four repressor modules is included (one repressor module is shown in slide 1 of the .ppt). The CSV output option writes the results of the simulation to a comma-delimited file located on the desktop which can be loaded into R for high-quality graphical output. I decided to switch to R because Excel was proving very inadequate for medium-sized data sets and, indeed, the non-linear x-axis scale was extremely misleading. It was however necessary to increase the maximum vector size through R's properties menu so that it could handle deterministic data of 2 million time points. Supporting R code is provided [12].

There are currently however two things missing from this model which I personally feel should be included. The first is to have multiple cells. The second is to account for the estimated 100-300 bp gap created at unrepaired telomeres following replication of the DNA during cell division. My gut feeling is that the rate of exonuclease attack will be too great for this model to be useful over time periods lasting hours. Therefore, my personal priority is to decide how (fully- or partially-) resistant ER2s can be included in the model to slow this rate down. It will not be a small challenge although, if computer memory permits, I am looking forward to it.

Since the linear DNA biobrick has been voted as top priority for the iGEM project, there is strong pressure now to finish the exonuclease model, search for parts, and design the biobrick itself. I have attached my Bug's eye view flowchart with side 2 showing a preliminary indication of where the linear DNA clock might be useful [13].