How We Drugged a Leukemia’s Favorite Transcription Network

Share this post:Tweet about this on TwitterShare on Facebook0Share on LinkedIn12Share on Reddit0

If we “cure cancer” in the woods and no one is there to read about it, will patients reap the rewards? I did computational analyses for a recent paper in Nature that describes a new chemical that kills leukemia cells and how we think it does so. Somehow this has escaped mention by nearly all science press. </sour grapes> Surprisingly, this chemical messes with a process all cells need to do– healthy and cancer, yet it doesn’t kill all kinds of cells. So, not only did we invent a chemical that kills cancer cells, but we also identified a new way to specifically target cancers with drugs.

Our head model for how THZ1 works: CDK7 (red) normally puts phosphate molecules (blue) onto the tail of RNA Pol II, which is part of its “transcribe now” signal. Left unchecked in a leukemia, cells multiply. But THZ1 binds to and prevents addition of the phosphate, preventing RNA Pol II from transcribing and killing the cells.

Our collaborators at the Dana-Farber Cancer Institute (Harvard Med’s cancer wing) created a chemical compound called THZ1, named after its most direct inventor, Tinghu Zhang. This drug binds to and de-activates a protein called CDK7, short for cyclin-dependent kinase 7. Kinases are proteins that stick phosphate molecules onto other proteins, cyclin-dependent ones involve the cell cycle, and CDK7 was the 7th one discovered. THZ1, the drug, binds to CDK7 and prevents it from adding phosphate molecules to its target proteins.

By testing THZ1 across a bunch of cell types, we saw that our chemical does a number on leukemias. It kills leukemia cells in a dish, and it kills them in mice. We’re pretty happy with this, but my job was to figure out what it’s doing to cells that makes them die. In doing so, we saw that THZ1 is a great tool for figuring out how biology works in addition to being a potential drug.

To understand how THZ1 works, we compared the same kind of cells treated with it to those treated with a control chemical. All cells transcribe DNA genes into messenger RNAs (mRNAs). This is largely done by a group of proteins that form a molecular machine called RNA Polymerase II (RNA PolII). PolII is like a molecular photocopier, and phosphate molecules on its long, protein tail are important to its function. CDK7 is responsible for adding some of the phosphate molecules PolII needs to function onto a longish protein tail of PolII.

Like most of my more pride-inducing findings, our first results came under the pressure of and just ahead of a deadline, in this case, the Thursday night before a Friday morning presentation of our work. We had just gotten data from T cell leukemia cells treated with THZ1. Specifically, we got the levels of mRNA transcripts per gene for all genes by performing an expression microarray experiment. This experiment measures how many molecules of mRNA per gene that existed in a pool of cells, so we compared the counts of mRNAs in untreated cells vs. cells treated with THZ1.

We treated cells with THZ1 and measured the levels of all transcripts. Each transcript is a row. The color represents how different this transcript’s level is compared to control. The greener it is, the less mRNA for that gene in the THZ1-treated cells.

After some biologically-informed data normalization, we saw that virtually all genes had lower mRNA levels in the THZ1-treated sample than in the untreated sample. This made a degree of sense– if most genes get transcribed into mRNA by the action of PolII, and, if you mess with CDK7’s ability to add phosphate groups to PolII’s tail, it can’t transcribe. But this also didn’t make complete sense– all cell types need PolII to transcribe, and not all cells died at the same rate when treated with THZ1. Nearing midnight, our collective brains alternated between bursts of insight and longings for bed.

We treated different kinds of cells with THZ1 and counted how many cells lived compared to an untreated control. Each cell type is a line; the first four are leukemias. Each point is the percent (1.0 = 100%) of the cells that survived at that level of drug treatment.

So how could a chemical that inhibits the function of a machine necessary for all cells to stay alive have specific effects on specific cell types? Well, we thought, not all cell types transcribe the same genes, since not all genes are important for all cell types. Only a specific subset (~1/2 to 2/3) are transcribed in a given cell type, and only a small number (~3-10) seem responsible for making/keeping that cell type’s identity. So we followed a hunch that there was something about the T cell acute lymphoblastic leukemia cell identity genes that made these genes and cells susceptible to futzing with a part of general transcription.

Thanks to the work of many leukemia and transcription researchers, we have a pretty good sense of the important genes in this particular kind of leukemia, including the genes responsible for keeping this cancer cancerous: RUNX1, TAL1, and GATA3. These genes encode transcription factor proteins that help control each other’s transcription, i.e. GATA3 protein from the GATA3 gene helps control the transcription of the TAL1 gene, etc. Our microarray experiments showed that the RUNX1 gene encodes a RUNX1 mRNA whose levels go down a lot when you treat cells with THZ1. In fact, it doesn’t take a lot of THZ1 to make the RUNX1 mRNA levels go down 10 times as much, since it is equally sensitive to low dose (1x) and high dose (10x) treatment.

We knew two things that led us to a hypothesis:
1) Not all genes react in the same way to messing with CDK7, but all genes do react.
2) Not all genes are equally important for a given cell to have a certain identity and live.
What if the transcription of all genes going down was related to a regulator gene going down first?

So we did another experiment. We suppressed the level of RUNX1 transcription and used another microarray experiment to ask how all genes responded. We saw that the same genes whose transcription went down the most in response to THZ1 also went down a lot when RUNX1 was suppressed. This makes us think that genes controlled by RUNX1 are sensitive to THZ1 chemical treatment. This also might explain why not all cell types are equally sensitive to THZ1 in the same way– maybe there are genes that are important for some cell type or identity that dictate if that cell type is sensitive to CDK7 inhibition by THZ1 or not.

Yeah, we’re pretty excited about this one. Some of our other collaborators are taking the compound through more tests, hoping it becomes a drug that patients actually get. Because, after all: As for me, this was my second big project in the lab, and I moved abruptly from transcription mechanics into drugging cancer– not bad for a computer geek. Oh, and that Friday meeting? Knocked it out of the park.

This entry was posted in Science 101, Work I've Done. Bookmark the permalink.

Leave a Reply

Your email address will not be published.