Chromatin remodelling ATPases play essential roles in organizing the chromatin landscape which regulates eukaryotic gene expression, yet their mechanism of action is poorly understood. ATP-dependent nucleosome remodelling is likely to be a multi-step process that involves binding of the remodeller to the nucleosome and ATP substrates, ATP hydrolysis, movement of the histone octamer relative to DNA and dissociation of the enzyme from the nucleosome. A tight coordination between these activities is required to achieve efficient remodelling.
The simultaneous presence of tPHD and tCHD is a unique characteristic of the ATP-dependent chromatin remodellers CHD3, CHD4 and CHD5 
. However, many chromatin remodelling complexes contain multiple conserved domains which are thought to be involved in chromatin targeting. Although many studies have described the role and properties of these domains in isolation, it remains unclear how these domains influence each other and contribute to the remodelling action and its regulation.
In the present work we show that the two tandem domains of CHD4, tPHD and tCHD, are structurally coupled and modulate each other’s affinity for their respective substrates (a). The tCHD activity is modulated by tPHD domain, which prevents its aggregation into large complexes in presence of DNA. Sequence analysis of tCHD suggests that each individual chromodomain (CHD1 and CHD2) contains a DNA binding motif (Supplementary Fig. 6a
). This is consistent with both individual chromodomains of dMi-2 being able to bind to DNA 
. The presence of two binding sites could effectively crosslink DNA fragments together, thereby creating large aggregates.
Perhaps, given the physical interaction between tPHD and tCHD domains in the stalk region, tPHD-mediated modulation of tCHD–DNA interaction could be explained by tPHD masking one of the DNA binding sites. Sequence analysis of tPHD shows that a long linker (32 a.a.), which is rich in acidic residues (15 a.a), connects PHD1 to PHD2, making it a plausible electrostatic mask that repels DNA from one of the binding sites on tCHD (Supplementary Fig. 6b
Fig. 6 Summary of activities and simple model of ATP hydrolysis cycle. (a) Summary of DNA, core particles and histone H3 binding and ATPase activity of the various CHD4 domain constructs. Nucleosome-recognition by CHD4 is promoted by interdomain interactions (more ...)
The tPHD recognizes both un-methylated as well as K9-methylated N-terminal tails of histone H3. The affinity for the H3 tail is increased by the presence of tCHD which does not bind histones. Thus, the effect is most likely structural. The proposed intimate association between tCHD and tPHD domains within the stalk suggests that tCHD may stabilize an active conformation of tPHD and enable it to bind H3 with high affinity (b). Interestingly, the previously measured affinity of the isolated PHD2 domain for the same peptide (Kd
= 18 μM 
) is very close to the Kd
obtained here for the tPHDtCHD (Kd
= 14–19 μM, ) and is in contrast with the low binding affinity of the tPHD construct in this study. Recently, the isolated PHD1 was also found to bind with similar affinity to H31–12
values ranging from 3.2 ± 0.6 μM to 17 ± 5 μM [30,31]
. It is possible that the long (32 amino acids) charged and likely flexible linker between PHD domains has an inhibitory effect on the intact tPHD and that only the presence of the tCHD domain restores the correct binding surface.
In our study, we show that the ATPase domain enhances the histone-tail binding affinity of the tPHDtCHD moiety. The ATPase domain greatly increases affinities compared to those observed for the isolated tPHDtCHD (, ). To a lesser degree ATP also modulates the histone tails affinity, a feature potentially important for remodelling. This is consistent with the solution structure studies in which only subtle structural changes may occur during ATP binding.
Vice versa, the tPHDtCHD moiety is essential for the DNA-stimulated ATPase activity since only the tPHDtCHD/ATPase shows DNA-stimulated ATPase activity (). This result could be explained by intimate interaction between the ATPase domain and both tCHD and tPHD within the head part. tPHD arm is wrapped around the ATPase and may stabilize an active configuration or prevent the nearby tCHD domain from sterically blocking access to the ATP or DNA binding sites of the ATPase. This latter possibility is further supported by the observation that the tCHD/ATPase construct, which lacks tPHD, displays severely attenuated ATPase activity that is not stimulated by DNA. These results are in line with the structure of the yeast tCHD/ATPase construct from CHD1 
, in which the tCHD is seen occluding the DNA binding site of the ATPase motor.
Thus, the intertwined arrangement of DNA binding domains, histone tail binding domains and the motor domain in CHD4 might enable coordination of nucleosome binding and release with the ATP-dependent nucleosome remodelling activity (b). In the ATP-free state, CHD4 is initially targeted to nucleosomes via binding of the tPHDtCHD module to DNA. The tPHD is able to bind H3 tails through recognition of K4 (unmethylated) and K9 (preferably methylated) but the affinities are very low. It is the influence of the tCHD and ATPase domains and the allosteric regulation by ATP binding that increase the tPHD histone H3 tails affinity to biologically significant values. Nucleosome binding stimulates the ATPase activity and hydrolysis (or any subsequent steps) and leads to translocation and change in the relative position of the DNA-bound CHD4 and the nucleosome. The increased strain on the histone binding site may promote dissociation from the tail. This allows mobility of CHD4 between nucleosomes while translocating without compromising the tPHD domain mediated targeting.
These results suggest that the conserved domains flanking the ATPase motor do not only direct CHD4 to the correct substrate but also participate in the remodelling activity. Since this combination of a motor domain and “chromatin targeting” domains is the unifying feature of all chromatin remodellers, we propose that inter-domain allosteric regulation might be a general feature of this class of enzymes.
NOTE ADDED IN PROOF: Since this work was being revised a paper which describes a similar overall shape and domain interactions of CHD4 has been published