15/ These are truly specific focal amplifications, occurring in an otherwise relatively stable genome.

14/ A major limitation in studying ecDNA-driven cancers is the lack of autochtonous and immunocompetent preclinical models. In the final set of experiments, we combined the ecMdm2 mouse and Myc to induce the formation of primary liver tumors harboring massive Mdm2 amplification.

13/ The Mdm2 ecDNAs also accumulate in primary MEFs, driving their immortalization and cooperating with HRAS in transforming them.

12/ We next generated two mouse strains harboring Cre-inducible Mdm2- and Myc-containing ecDNAs. Using primary cells from the ecMyc mice we show that the engineered ecDNAs accumulate very rapidly and without any extrinsic selective pressure following Cre-mediated recombination.

11/ A useful feature of this strategy is that we can take advantage of the hygromycin resistance transgene encoded by the ecDNAs to force their accumulation. As we increase hygromycin concentration we see a concomitant increase in GFP intensity and ecDNA abundance.

10/ Crucially, when we sequenced the genome of these cells we saw a focal amplification whose border perfectly matched the location of the two circularization cassettes (light blue shaded area, invMDM2 is a negative control).

9/ This is exactly what we expected: ecDNAs lack centromeres and therefore segregate randomly at mitosis. In fact, we can find the engineered MDM2-containing ecDNAs in the GFP+ population, and their number correlates nicely with GFP intensity.

8/ After delivering Cre, ~20-30% of the cells undergo circularization and, start expressing GFP and mScarlet.

7/
To test this strategy, we engineered the formation of a 1.5Mbp ecDNA encompassing the MDM2 locus (as well as a bunch of other genes) in HCT116 cells, which are otherwise diploid and chromosomally stable.

6/ ...this allows us to to visualize cells that have undergone circularization (they become green and red), infer ecDNA copy number (more ecDNAs = more GFP), and follow their dynamics over time.

4/ The underlying idea is to exploits Cre recombinase to induce circularization of any genomic regions flanked by 2 loxP sites in the same orientation. We use CRISPR to insert the two loxP sites at the desired circularization breakpoints.

2/ What is an ecDNA?
ecDNAs had been known since the 60s (they were called "Double Minutes"). They are large circular DNA fragments found in some of the most aggressive forms of cancers. They amplify oncogenes, boost gene expression, facilitate tumor evolution, and increase resistance to therapy.

I had a ton of fun discussing this famous paper with a group of very enthusiastic and bright Weill-Cornell Medical College students. In my opinion it is one of the most elegant studies ever published. Any suggestion for what we should read next? The class is about logical thinking. #science