More Recent Comments

Friday, July 13, 2012

Slip Slidin' Along - How DNA Binding Proteins Find Their Target

Many proteins bind to double-stranded DNA and most of them bind specifically to a particular target site. The lac repressor, for example, binds to a specific DNA site that blocks transcription of the genes required for lactose uptake and utilization. The lac repressor protein is a dimer of two identical subunits and each one binds a short segment with the core sequence ATTGT.1

If you look closely at the structure shown above, you can see how parts of the protein lie in the grove of double-stranded DNA where they can detect the sequence by "reading" the chemical groups on the edges of the base pairs. It's important to realize that DNA binding proteins interact with the DNA double helix and not with unwound DNA where the individual bases are exposed.

How does a DNA binding protein like lac repressor find its specific site in the genome? The most obvious explanation is that the protein binds non-specifically to any piece of DNA and checks to see if it's a specific binding site. If it is, the protein binds very tightly and doesn't fall off. If it isn't, the interaction is much weaker and the protein falls off quickly so it can check out another potential site.

It's been known for a long time that all specific binding proteins also bind non-specifically. This method of searching for the target site depends on the number of proteins, the size of the genome, and the on-off rates for specific and non-specific binding. The overall rate is limited by diffusion or the rate of collisions between the DNA binding protein and the DNA.

It's also been known for a long time that when the search for the target site is measured in a test tube with purified repressor and purified DNA, the rate of finding the target is FASTER than simple diffusion would predict! What actually happens is that the repressor (and other similar DNA binding proteins) sticks non-specifically to DNA then slides along the DNA molecule looking for a target sequence. It can scan about 50 bp of DNA in a single interaction and this form of "facilitated diffusion" leads to a much quicker search for the target sequence. I described this a few years ago, giving the values for the various interactions and rates: DNA Binding Proteins.

Just because this sliding works in vitro, in a test tube, doesn't mean that it works that way in vivo, inside a living cell. The DNA in a cell is covered with lots of other proteins that could block sliding and the concentrations of salts (ionic strength) in the cell are different than the concentrations used in the test tube.

Recently Hammar et al. (2012) developed a method to detect facilitated diffusion in bacterial cells. They did this by labeling the lac repressor with a yellow flourescent protein and following the protein under a flourescent light microscope. You can see two photos from the paper on the left. In the top picture, the lac genes are repressed and there's a lac repressor molecule sitting on every target site (operator). Since the DNA of the genome doesn't move around very much, the bound molecule appears as a stable, yellow, spot. In the bottom picture the lac genes have been "induced" and the repressor isn't bound specifically to DNA. There are about four or five repressors in each cell but they are moving around so much that the yellow flouresence is smeared out and diffuse.

Hammar et al. induced the lac genes and then allowed the lac repressor to re-bind to its target site. They measured the appearance of "fixed" flourescent spots as a function of time. The "fixed" spots represent specific binding to the target sequence.

The results are shown in the figure on the right. It measures the "kinetics" (rate) of binding to the target site. As you can see, most of the target sites are occupied after about four minutes and the binding curve has the normal shape you would expect.

The rate is faster than a diffusion limited reaction. The data can be fitted to a curve that predicts facilitead diffusion with each search covering about 45 bp of DNA. There were several control experiments and additional experiments showing that the presence of other cellular proteins on the DNA didn't have much of an effect.

The amount of slip slidin' measured in this experiment is close to the value determined in the test tube so it looks like sliding is used inside the cell as well as in vitro. I think we can safely conclude that most specific DNA binding proteins find their target sites by binding non-specifically to DNA then sliding along the double-stranded helix until they find their specific binding site.


1. The two sites are inverted relative to each other so that each subunit is, in theory, binding to exactly the same piece of double-stranded DNA. In practice, the binding sites (operators) differ slightly in their sequences. Furthermore the active repressor inside the cell is actually a tetramer that binds a different piece of DNA are either end (see Repression of the lac Operon).

Hammar, P., Leroy, P., Mahmutovic, A., Marklund, E.G., Berg, O.G., and Elf, J. (2012) The lac repressor displays facilitated diffusion in living cells. Science 336:1595-8. [DOI: 10.1126/science.1221648]

Musical reference: Paul Simon & Art Garfunkel - Slip Slidin' Away

9 comments :

Georgi Marinov said...

I think we can safely conclude that most specific DNA binding proteins find their target sites by binding non-specifically to DNA then sliding along the double-stranded helix until they find their specific binding site.

Does the above statement include eukaryotic transcription factors? It is much more difficult to see how that works when you have nucleosomes.

DK said...

What actually happens is that the repressor (and other similar DNA binding proteins) sticks non-specifically to DNA then slides along the DNA molecule looking for a target sequence. It can scan about 50 bp of DNA in a single interaction and this form of "facilitated diffusion" leads to a much quicker search for the target sequence.

In really simple terms, 3D diffusion is replaced by 1D diffusion: for any connected points A and B, 1D will always be much, much faster. 1D diffusion is still a diffusion - there is nothing "facilitated" about it.

Anonymous said...

Involves sliding, detaching, attaching again, sliding ...

Anonymous said...

Actually the explanation is that diffusion of the protein to the binding site is facilitated by its sliding along the DNA strand, not that it slides "until it finds its binding site." The process, as should be expected, is not perfect. For one, the continuity of the sliding process depends on how long the protein can stay semi-attached to the DNA molecule. For another, there are many other proteins attaching, detaching, and or sliding on the DNA.

Anonymous said...

Exactly.

DAK said...

3D diffusion (to use your terminology) is not replaced by 1D diffusion. It's two separate processes - regular diffusion (to land somewhere very, very near the operator) plus the slip'n slide. Both together are faster than plain old regular diffusion on its own; hence the "facilitated". It's fundamentally different from the way in which some small substrate may diffuse to the active site of an enzyme.

DK said...

3D diffusion (to use your terminology) is not replaced by 1D diffusion.

Of course it is. I haven't bothered to read the paper or its abstract when I posted the message above but now that I did, I am pleased (but not at all surprised) to find that that's exactly the terminology they use:

"a combination of 3D diffusion in the cytoplasm and 1D diffusion (sliding) along the DNA." (first paragraph).

I don't know what you imagine that happens (it's not at all clear from your post) but here is how it works in the physical world we live in:

1. DNA-binding protein reaches a piece of DNA by regular 3D diffusion and binds to it ("non-specifically" - which isn't really non-specific; just a lot less specific).
2. Because the affinity is relatively low, it at some point (usually soon) detaches and diffuses some distance. That distance is very small because the affinity is still high enough for the the protein to quickly bind back to the DNA - but now at a slightly different position than the first time. So the observer who only pays attention to the fact that the protein remains associated with DNA most of the time can conclude that the protein "slides" alone DNA.
3. It's not a true sliding (although it can be modeled as such) and it most certainly is not directional - like any random diffusional process, be it in 3D (cytoplasm), 2D (membrane), or 1D (linear polymers like DNA). This is important to stress because there are certainly examples of true directional sliding (powered by motors). Overall, the simplest formalism to describe what happens is to say that the protein undergoes 1D diffusion along DNA.
4. The important thing that follows is that there exists an optimal strength of the non-specific binding that enables the fastest target search: bind too tightly and the "sliding" will be too slow (spend too much time stuck to the same point on DNA); bind too weakly and the "sliding" will also be too slow (too high a chance to get lost far into surrounding solution after unbinding from DNA).

Anonymous said...

As a commentator up-thread noted, any slip-and-slide model of sequence-specific DNA binding activity by transcription factors fails the sniff test: how is the activator (or repressor) able to effectively scan the nucleotide side-groups to achive site-specificity when the latter are coated with histones (in most eukaryotes) and with other attendant DNA-binding molecules (in all organisms). The notion that the chromosomal DNA molecule exists in all of its double-helical beauty for all proteins to probe seems rather tired and readily debunked to my mind. I've been a hesitant skeptic of the "histone code" as anything other than correlative observations, but given the ubiquitous habit of histone compaction of large chromosomal segments, some portions of which obviously remain accessible to transcription factors, it seems clear to me that we're missing some vital pieces of the puzzle.

Unknown said...
This comment has been removed by the author.