Frequently asked questions about DamID

Before you start reading this FAQ...

The intention of this web site is to make DamID accessible to everyone in the scientific community. We all benefit from exchange of information, so please share your experiences and ideas!

How does DamID work?

DamID is based on the creation of a fusion protein consisting of Escherichia coli DNA adenine methyltransferase (Dam) and a chromatin protein or transcription factor of interest. Dam methylates adenines in the sequence GATC. Endogenous methylation of adenines is absent in most eukaryotes. Upon expression of the fusion protein in cultured cells or in an intact organism such as Drosophila, Dam will be targeted to the native binding sites of the chromatin protein. This will then result in local methylation of adenine residues. Hence, the sequences near a binding site of the protein will be marked with a unique methylation tag, which can be detected using Southern blot-, PCR- and microarray based assays that take advantage of restriction enzymes that are methylation sensitive.

Is it really that simple?

Almost, but there's one complication. The Dam-fusion protein is never perfectly targeted to the native binding sites — a fraction of the fusion protein molecules is inevitably diffusing around in the nucleus. This will cause considerable background methylation of non-target sites. To make things worse, some GATCs in the genome are more accessible to Dam than others, and as a result the non-specific background methylation is not homogeneous throughout the genome. Thus, one might easily mistake a non-target GATC in a very "open" chromatin region for a target of the chromatin protein.

Luckily, it turns out that one can correct for this non-targeted background methylation by measuring in a parallel control experiment the methylation levels by unfused Dam. For each GATC the methylation levels obtained with the fusion protein are then compared to the methylation levels obtained with Dam only. In practice, this comparison can be done by calculating the ratio methylation by Dam-fusion protein: methylation by Dam. We found that in this way the variation in chromatin accessibility is normalized for.

The task that remains is to compare the normalized methylation level of your locus of interest with the normalized methylation levels of a few negative control loci, i.e., loci of which you are quite sure that they are non-targets. Preferably you also include a few positive controls in the experiment (if you know any target loci already). Repeat the experiment several times and do a proper statistical test, because there is some noise associated with the whole procedure!

Do I really have to do the Dam-only control experiment?

YES! If you only measure methylation levels after expression of a Dam-fusion protein, you will be unable to tell whether high methylation in a particular locus is due to the "open" chromatin structure, or due to real targeting of Dam by its fusion partner. That's why you need the Dam-only control. For the same reason the methylation levels of each locus have to be determined quantitatively. There's no way around this!

In which organisms can DamID be used? Does it work in mammalian cells?

DamID has been extensively used and validated in Drosophila — both in cultured Kc cells and in intact flies. Proof-of-principle has also been obtained in Arabidopsis thaliana (Germann et al., 2006). And yes, it also works in mammalian cells — we used it to generate detailed binding maps of heterochromatin proteins both in human and mouse cells (Vogel et al., 2006). With the exception of some ciliates, eukaryotes do not appear to have endogenous methylation of adenines in DNA. Hence, DamID may work well in model organisms such as S. cerevisiae, S. pombe, or C. elegans.

What's the trickiest part in the procedure?

Our work in Drosophila (flies and cell lines) suggests that it is very important to keep the expression level of the Dam-fusion protein very low, to avoid saturating methylation levels. We used the Drosophila heat-shock promoter for expression of our Dam fusions, but did the actual experiments in the absence of heat-shock. In human cells we also use an inducible promoter, but again for DamID we only use the leaky expression in the absence of induction. Under these conditions we could not detect the fusion proteins themselves (by western blotting or immunofluorescence microscopy), but there was specific methylation of target sequences. Our interpretation is that only trace amounts of the Dam proteins are present, but that — thanks to the high enzymatic activity of Dam — this level is just right to obtain detectable but non-saturating levels of methylation. After induction the Dam-fusion proteins themselves could be detected easily, but at the same time the targeted methylation levels had reached saturation, and background methylation had become so high that no target sequences could be identified.

Thus, the trick is to keep the expression levels of the Dam (fusion) proteins low. In flies I would therefore suggest to work with the (uninduced) hsp70 promoter or some derivative. In human cells we suggest that you use our vectors, which contain the inducible promoter from the pIND system. This low expression level has the additional advantage that the Dam fusion protein is unlikely to perturb the function(s) of the endogenous chromatin protein or its targets, since it is only present in trace amounts.

Which types of microarrays can be used with DamID?

In Drosophila, DamID works surprisingly well with conventional cDNA arrays; we have now identified large numbers of target loci for several different transcription factors and chromatin proteins using cDNA arrays (van Steensel et al., 2003; Orian et al., 2003; Greil et al., 2003; de Wit et al., 2005; Tolhuis et al., 2006; Pickersgill et al., 2006). With these arrays only binding close to transcribed regions is detected, and cDNA arrays may be less useful when genes are large (e.g., in mammals). In a collaboration with Kevin White's lab at Yale University, we have used genomic tiling path arrays that were made of spotted 1 kb PCR fragments to map binding of 15 different regulatory proteins in intergenic and coding regions in a 2.9 Mb contiguous genomic region (Sun et al., 2003; Moorman et al., 2006). DamID also performs very well with 380 k NimbleGen oligonucleotide arrays (Vogel et al., 2006), and preliminary results suggest that Agilent promoter arrays are also fine for DamID. We have not tested Affymetrix genomic tiling arrays yet.

How does DamID compare to chromatin immunoprecipitation?

Chromatin immunoprecipitation (ChIP) has been successfully used to identify target DNA sequences of various chromatin proteins. At present, it is impossible to tell whether you should use ChIP or DamID for studying a particular protein. Many excellent and exciting ChIP results have been published. However, if you ask around in the community you'll learn that ChIP certainly does not work for every protein. In Drosophila cells, we find that DamID works for > 90% of all proteins (transcription factors, chromatin proteins, etc.), without ever having to adjust the protocol. So, needless to say: we like DamID!

Only a few side-by-side comparisons of ChIP-on-chip and DamID have been done so far. This showed that data were in good agreement in the case of GAGA factor (Moorman et al., 2006; Negre et al., 2006) and Polycomb group proteins (Tolhuis et al., 2006; Negre et al., 2006), many thanks to the lab of Giacomo Cavalli for helping us with these comparisons).

Here are a number of additional practical and theoretical considerations (also reviewed in van Steensel & Henikoff, 2003):

1. ChIP requires an antibody against the protein of interest. Of course this antibody should be absolutely specific for your protein under ChIP conditions. DamID is not dependent on antibodies, and thus there's no need to worry about cross-reactivity.
2. With DamID, one can also study the behaviour of mutant proteins. This can be useful to confirm the specificity of the DamID profile, but also to study protein targeting mechanisms, and to investigate the effects of disease-associated mutations, etc.
3. When doing DamID in cultured cells, transient transfections with high transfection efficiencies are needed. Ideally we aim for 30% or higher, although we have obtained good binding maps from lower efficiency transfections. Depending on the cell line, this may take some optimization. In mammalian cells, we prefer our lentivirus vectors, which can be used in almost any cell type. Stably transfected cell pools also work. Clonal stable lines may be tricky, because clonal cell lines sometimes are aneuploid. Transfections are usually not required for ChIP, unless an antibody against an epitope tag is used for the immunoprecipitations.
4. When working in metazoans such as D. melanogaster, DamID requires the generation of transgenic lines expressing the Dam fusion protein. Of course, once you have the transgenic lines, then it should be possible to generate DamID data from dissected tissues or flow-sorted cell populations.
5. In most cases, ChIP involves the use of formaldehyde to crosslink chromatin proteins to their DNA targets, followed by immunoprecipitation of chromatin fragments using an antibody against the protein of interest. The assumption is that crosslinking by formaldehyde is instantaneous and does not cause artifactual rearrangements. This assumption may be correct in most cases. However, one should keep in mind that electron microscopists who study subnuclear structures usually do not use formaldehyde alone as a fixative, because it yields poor morphology. In addition, when working with multicellular organisms, formaldehyde may not penetrate the tissues quickly enough to preserve the native chromatin structure.
6. For ChIP-on-chip, one needs a relatively large amount of starting material to do reliable immunoprecipitations. DamID (when combined with the PCR-based methylation assay) requires only small numbers of cells — we did DamID in single fruitflies!
7. Wolffe and Leblanc (2000) have suggested that DamID may work well for proteins that interact only transiently with their DNA targets, because targeted Dam could leave a permanent mark even after a brief interaction. Detecting such transient interactions by DamID is an interesting possibility, but at present we do not have any experimental evidence for this.

How precisely can I map a protein binding site?

Targeted methylation 'spreads' in cis from the protein binding site. We don't know exacly how this happens, but it does limit the resolution of DamID. We estimate that this resolution is roughly 1-2 kb. For ChIP, numbers between 0.5 and 2 kb have been reported, but this value may correlate with the modesty of the authors – and note that various definitions of 'resolution' exist. The accuracy of mapping by ChIP may be affected by the degree of crosslinking and shearing of the chromatin. In our side-by-side comparison of ChIP-on-chip and DamID for GAGA factor, we did not see a clear difference in peak widths between the two methods (Moorman et al., 2006).

Where can I find more information? Where can I get DamID protocols?