Given the role of the DmCdt1 N-terminal IDR in promoting phase separation, we next examined whether other replication initiation factors might possess similar disordered regions of analogous function. Using DISOPRED (Jones and Cozzetto, 2015), we calculated the percent of predicted disordered residues, as well as the longest continuous disordered segment for each Drosophila protein required for initiating replication, and ranked them according to low (0–10% predicted unstructured residues), moderate (10–30%), or high disorder (>30%) (Figure 3A). In addition to Cdt1, Cdc6 and two subunits of ORC (Orc1 and Orc2) were found to possess a high level of disordered content as a proportion of their total polypeptide chain length, and each contained a continuous region of disorder longer than 200 amino acids. Conversely, the Mcm2-7 subunits, though possessing unstructured regions, contain less than 30% predicted disordered sequence overall, with no single disordered region extending beyond 150 amino acids. Similar patterns were seen for other metazoan replication initiation factors, but not S. cerevisiae proteins (Table 1).

(A) Graphical comparison of the disorder for each Drosophila replication initiation factor. Orc1, Orc2, Cdc6, and Cdt1 each contain long IDRs (as denoted by the numbers under each circle) and a high percentage of overall disordered sequence (predicted >30%, depicted by color shading). (B) Analysis of ORC phase separation by the depletion assay. ORC (500 nM) phase separates in a DNA-dependent fashion in the presence and absence of ATP. (C) Cy5-dsDNA (2.5 µM) was imaged alone and as a mixture with ORC (2.5 µM). In the presence of ORC, large phase-separated droplets formed. (D) Two ORC subunits, Orc1 and Orc2, have large N-terminal IDRs. The IDR of Orc1 is longer and is enriched for positively-charged residues. (E) Analysis of ORC∆1IDR (500 nM) and ORC∆2IDR (500 nM) phase separation by depletion assay. Loss of the Orc1 IDR abolishes phase separation but loss of the Orc2 IDR has no effect. (F) Droplets form when Cy5-dsDNA (2.5 µM) is mixed with ORC∆2IDR (2.5 µM) but not when mixed with ORC∆1IDR (2.5 µM). (G) Cdc6 phase separation was assessed by depletion assay at 500 nM and (H) 5 µM concentrations. DNA induced Cdc6 phase separation but this was inhibited in the presence of 1 mM ATP. (I) Fluorescence imaging reveals phase-separated droplets when Cy5-dsDNA (2.5 µM) is mixed with Cdc6 (20 µM). (J) Phase separation analysis for a Cdc6 construct lacking the N-terminal IDR (Cdc6∆IDR). Cdc6∆IDR (500 nM) shows no depletion in the presence of DNA (500 nM). (K) Cdc6∆IDR (20 µM) is unable to induce droplet formation as assessed by fluorescence microscopy with Cy5-dsDNA (2.5 µM). Gel images are representative of three independent experiments.

For proteins classified with a high fraction of disordered content (>30%), we next assessed the location of the disordered regions relative to known protein domains (Figure 3—figure supplement 1A). In all cases, the longest regions of uninterrupted disorder reside N-terminal to the bulk of the folded domain content of the chain (i.e., upstream of the Cdt1 WH domains and of the AAA+ and WH domains of Orc1, Orc2, and Cdc6). Cdc6 possesses the shortest N-terminal IDR (246 amino acids) and Orc1 the longest (362 residues). There also exist shorter regions of disorder that serve as linker sequences between tandem globular domains, such as between the AAA+ and WH domain of Cdc6 (26 amino acid IDR) or the two WH domains of Cdt1 (152 amino acid IDR).

The realization that Cdc6 and two subunits of ORC possess long N-terminal IDRs suggested that these factors, like Cdt1, might also undergo phase separation in a DNA-dependent fashion, and that this property might promote their functional integration within a single condensed phase. This idea was first tested by examining the ability of recombinant DmORC to phase separate using the depletion assay (Figure 3B, lanes 1–3). Similar to the behavior of DmCdt1 when it forms condensates, ORC was found to be depleted from the supernatant in the presence, but not in the absence, of DNA. The inclusion of ATP did not have an appreciable effect on phase separation by ORC (Figure 3B, lanes 4–5), and fluorescence microscopy with Cy5-dsDNA confirmed that the DNA-dependent depletion of ORC from the supernatant was due to phase separation (Figure 3C, no droplets are observed for DNA alone). Assaying ORC phase separation over a range of protein concentrations (50 nM to 800 nM) consistently revealed ORC depletion when DNA was present but not when it was absent (Figure 3—figure supplement 1B). Depletion occurred regardless of oligonucleotide sequence (Figure 3—figure supplement 1C), although like Cdt1, maximal phase separation occurred with oligonucleotides longer than 25 basepairs (Figure 3—figure supplement 1D). Thus, like Cdt1, ORC is able to undergo LLPS at physiological concentrations in the presence of a variety of DNA substrates without any apparent DNA sequence dependence.

We next investigated whether ORC’s ability to undergo phase separation could be targeted to the IDR of a specific subunit, or whether the IDRs of both Orc1 and Orc2 are required for this behavior. In terms of length, the Orc1 IDR, at 362 amino acids, is approximately 100 residues longer than the Orc2 IDR (Figure 3D). Comparisons of sequence composition between Orc1 and Orc2 IDRs – performed by calculating the relative percentage of hydrophobic (A, G, I, L, M, P, V), hydrophilic (C, N, S, T, Q), charged (D, E, H, K, R), and aromatic (F, W, Y) residues – show that the IDRs of Orc1 and Orc2 are highly similar in content, with the values for Orc1 and Orc2 within 3% for each amino acid category. Interestingly, Orc1 and Orc2 IDR amino acids are near equally distributed across hydrophobic (36/35%), hydrophilic (30/29%), and charged (32/33%) classes; this preponderance of hydrophobic residues is somewhat unexpected given the predicted unstructured nature of these sequences. Glycine, which is grouped within the hydrophobic class, is often enriched in protein unstructured regions, yet for both Orc1 and Orc2, glycine content is lower than expected for such a region (Brüne et al., 2018).

Although the Orc1 and Orc2 IDRs are highly similar in terms of amino acid types, they do show a marked difference in their isoelectric points (pI): the Orc1 IDR (pI = 10.1) is enriched for positively-charged residues (20% positive and 12% negative), whereas the Orc2 IDR (pI = 6.0) is weakly enriched for negatively-charged residues (17% negative and 16% positive). Speculating that this divergence in net charge might be important for facilitating LLPS by ORC, we constructed two mutant DmORC complexes, one with the Orc1 IDR deleted (ORC1ΔIDR) and the other lacking the Orc2 IDR (ORC2ΔIDR). Phase separation for both constructs was then assessed by the depletion assay and fluorescence imaging with Cy5-dsDNA. ORC1ΔIDR showed no evidence for phase separation in either assay, whereas the construct lacking the Orc2 IDR exhibited wild-type behavior in both assays (Figure 3E–F). These findings demonstrate that Drosophila ORC phase separates using interactions that require the N-terminal IDR of Orc1 but not Orc2. Given the predicted lack of structure in the Orc1 IDR we were surprised to discover that this region is relatively well-conserved across the Drosophila genus (67% identity, Figure 3—figure supplement 2) and that the composition, length, and pI of this region is conserved across the metazoan phyla (Table 1). Although S. cerevisiae Orc1 possesses an N-terminal IDR, this region is shorter than observed in metazoa and also has an acidic pI (Table 1). Consistent with this observation, budding yeast ORC does not phase separate (Figure 3—figure supplement 1E).

We next asked whether Cdc6 can undergo DNA-dependent LLPS. Similar to budding yeast Cdc6 (Feng et al., 2000), DmCdc6 was found to interact with DNA by dsDNA-coupled agarose bead pull-down assays; the presence of ATP had no significant effect on Cdc6 DNA-binding (Figure 3—figure supplement 1F). Upon testing whether Cdc6 phase separates in the presence of DNA using the depletion assay (Figure 3G), we observed a reproducible DNA-dependent depletion of Cdc6 from the supernatant (Figure 3G lanes 1–3); however, in contrast to ORC, this depletion was inhibited by the presence of ATP (Figure 3G lanes 4–5). A recent report that ATP can function as a hydrotrope to enhance intracellular protein solubility may provide an explanation for its effect on Cdc6 phase separation (Patel et al., 2017). Repeating the assay at higher concentrations of DmCdc6 shows that unlike DmCdt1, which appeared to fully partition into phases at higher concentrations (Figure 2E), Cdc6 was only partially depleted from the supernatant (Figure 3H). When examined over a range of protein concentrations and DNA sequences (Figure 3—figure supplement 1G–H), Cdc6 consistently showed phase separating behavior, although the protein never completely partitioned into the pellet. Notably, Cdc6 underwent complete partitioning into the condensed phase in the presence of plasmid DNA, whereas only partial depletion was observed with short oligonucleotides (Figure 3—figure supplement 1I), indicating that for DmCdc6, LLPS requires longer DNA segments than either DmCdt1 or DmORC. Fluorescence microscopy with Cy5-dsDNA confirmed that Cdc6 forms condensates in these conditions (Figure 3I), while tests with a Cdc6 mutant lacking the N-terminal IDR (Cdc6ΔIDR) (Figure 3J) showed that this construct has no capacity to phase separate (Figure 3J–K). Collectively, these studies confirm that Drosophila Cdc6 can form DNA-dependent condensates in a manner that requires its N-terminal IDR.