Identification of Fkh1 and Fkh2 binding site variants associated with dynamically bound DNA elements including replication origins.

Forkhead Box (Fox) DNA binding proteins control multiple genome activities, including transcription, replication, and repair. These activities are organized spatially and temporally in the nucleus, and Fox proteins Fkh1 and Fkh2 have emerged as regulators of long-range chromosomal interactions involved with these activities, such as the clustering of replication origins programmed for early initiation. Fkh1 and Fkh2 bind a subset of replication origins and are thought to dimerize to mediate long-range chromosomal contacts between these origins. The binding of Fkh1 and/or Fkh2 (Fkh1/2) to replication origins and the recombination enhancer (RE), which is involved in DNA repair required for mating-type switching, is cell cycle-regulated and thus appears to be more dynamic than Fkh1/2 binding at regulated target genes. Here we report the identification of Fkh1/2 binding sequence variants at replication origins and the RE compared with Fkh1/2 binding sequences found at target genes of the CLB2 group. These different binding sequences have previously been characterized as weak and strong, respectively, suggesting that the presence of weak sites contributes to more dynamic interactions at replication origins and RE, possibly facilitated by Fkh1/2 dimerization and cooperative interactions with accessory proteins. We discuss the wealth of regulatory potential imbued in these features of the DNA and its binding proteins.

Introduction

Genome architecture is increasingly recognized as an important contributor to the regulation of fundamental genomic tasks, including DNA replication, repair and transcription. We have recently identified the Forkhead Box (Fox) transcription factors, Fkh1 and Fkh2, as key regulators of replication origin initiation timing in the S. cerevisiae genome. Fkh1 and/or Fkh2 (Fkh1/2) bind a subset of replication origins referred to as Fkh-activated origins to stimulate their early initiation in S phase. Deletion of FKH1 and FKH2 or their binding sites proximal to Fkh-activated origins results in delayed activation of these origins; as a consequence, other origins, referred to as Fkh-repressed origins are increased in activity in the absence of FKH1 and FKH2, likely due to reduced competition from Fkh-activated origins for dose-limiting replication initiation factors. While it remains unclear exactly how Fkh1/2 stimulate origin initiation (or “firing”), Fkh1/2 are required for the spatial clustering of early replication origins in G1 phase, suggesting a role in establishing chromosomal architecture with functional consequence for origin firing. We have proposed that origin clustering enables cooperativity between origins in the recruitment of limiting initiation factors, thus stimulating early firing.

We have recently discovered the presence of a domain-swapping motif in Fkh1 and Fkh2 that allows for homo-dimerization of Fkh1 and thereby provides a plausible mechanism to establish physical contacts between origins for clustering in anticipation of replication. Mutation of specific residues to create “domain-swap minus” (dsm) alleles of FKH1 and FKH2 that cannot dimerize results in loss of Fkh-activated origin clustering and early firing, supporting a causal relationship between origin architecture and activation. In contrast, the dsm mutations appear to have a limited or minimal effect on the functions of FKH1 and FKH2 in gene regulation. For example, unlike fkh1Δ fkh2Δ cells, fkh1-dsm fkh2Δ and fkh1Δ fkh2-dsm cells do not exhibit pseudohyphal growth that is attributable to mis-regulation of Fkh1/2 target genes. Because of the very high similarity of Fkh1 and Fkh2 amino acid sequences in the Forkhead domain, and because the dsm mutations affects the function of both Fkh1 and Fkh2, we infer that both Fkh1 and Fkh2 are capable of homo-dimerization, though our previous analysis centered on Fkh1 as the major player in replication origin control. We also cannot rule out potential Fkh1-Fkh2 hetero-dimerization, however, the finding that deletion of FKH2 alone has no effect on replication origin function indicates that Fkh1-Fkh2 hetero-dimers, should they occur, are not required for origin regulation. Thus, we will limit discussion to the potential function of Fkh1 and/or Fkh2 homo-dimers.

Fkh1/2 binding sequence variants are associated with replication origins

Attempting to explain the differential requirement for domain-swapping in replication versus transcription, we considered the following: Fkh1/2 binding to replication origins is cell cycle-regulated, peaking during G1 and S phases, suggesting a dependence on origin-licensing, which occurs in G1 phase and persists until the origin replicates in S phase. In contrast, Fkh1/2 binding to target genes occurs throughout the cell cycle. To gain further insight, we examined the DNA sequences of putative Fkh1/2 binding sites proximal to replication origins and to Fkh1/2-regulated genes, in particular the “CLB2 cluster” genes that are expressed in S and G2 phases to regulate mitotic functions.

We identified potential Fkh1/2 binding site by searching for sequences matching the core consensus RYMAAYA within 500 bp regions surrounding origins or within 500 bp regions upstream of CLB2 cluster genes. We identified two main variants of the consensus sequence differing at the first position: GYMAAYA and AYMAAYA, which appeared to differ in their relative abundance at origins versus genes (Fig. 1A). Previous analysis of Fkh1/2 binding sites in the SWI5 promoter identified GTAAACA as a “strong” Fkh1/2 binding site and ATAAACA as a “weak” Fkh1/2 binding site. Accordingly, CLB2 cluster genes are significantly enriched for strong binding sites relative to a simulated random distribution whereas Fkh-activated origins are not (Fig. 1A). In comparison, Fkh-repressed origins are significantly depleted of weak sites relative to a random distribution (Fig. 1A). These findings suggest that differences in the density of Fkh1/2 binding sequence variants play a role in controlling the aforementioned binding characteristics of Fkh1/2 to these genetic elements. Indeed, the combination of weak binding sites and proximal licensing proteins at origins may cooperate to regulate loading of Fkh1/2 onto origins only during G1 and S phases. It will be interesting to determine whether altering weak to strong sites will alter Fkh1/2-origin binding dynamics or origin activity.

Figure 1.

Density of Fkh1/2 binding motif variants correlates with differential regulation of Fkh1/2-chromatin binding. (A) For actual results, genomic sequences of 500 bp centered on the indicated features were queried for the presence of AYMAAYA or GYMAAYA motifs. Significance was calculated by comparison with simulation results, entailing searches (10,000 iterations per feature-binding motif combination) of n random 500 bp genomic regions; n is indicated for each feature class. For the Recombination Enhancer, n = 2 because two non-overlapping, 500 bp regions on each side of the center point were used to calculate mean motifs per window. For each feature class, search results for both motifs were recorded from the same simulation. Fkh-activated and –repressed origins were defined in [2], and CLB2 cluster genes were defined in. (B) Model depicting the possible effect of Fkh1/2 dimerization and presence of DNA motif variants on binding site hierarchies. Fkh1-dsm binds CLB2 cluster promoters and Fkh-activated origins similarly to Fkh1, but fails to cluster early origins in G1, potentially leaving Fkh-activated origins with a lower density of Fkh1/2 binding motifs unbound. In G2/M phase (right panel), origins are unbound while CLB2 cluster promoters (enriched for GYMAAYA motifs) remain bound by either Fkh1 or Fkh1-dsm. Intra-chromosomal origin clustering is likely to be much more frequent than inter-chromosomal origin cluster, which is depicted for emphasis. Cell cycle phase and Fkh1 genotype are depicted above.

We extended the analysis of Fkh1/2 binding sequence variants to the recombination enhancer (RE) for donor preference in mating-type switching, which depends on FKH19. The minimal 700bp RE contains numerous Fkh1/2 consensus binding sequences and we find that the AYMAAYA motif is greatly enriched and the GYMAAYA motif is mildly enriched (Fig. 1A). As with replication origins, there is evidence that Fkh1 binding to the RE is cell cycle-regulated, consistent with the idea that the AYMAAYA motifs may facilitate more dynamic binding. Again suggesting a link between the AYMAAYA motif and domain-swapping, we have found that fkh1-dsm strains are defective in donor preference (S.K. Villwock et al., in review).

Hierarchies of Fkh1 and Fkh2 chromatin binding?

Fkh1 and Fkh2 are relatively abundant proteins, however, their consensus DNA binding sequences are in great excess relative to protein. Thus, Fkh1/2 are likely to select binding sites based on sequence variation and context. For example, it is well known that Mcm1 binds DNA cooperatively with Fkh2 (but not Fkh1), distinguishing the in vivo distributions of Fkh1 and Fkh2, and replication origin proteins physically interact with and are required for Fkh1 binding to Fkh-activated origins. Moreover, DNA bases immediately flanking the seven base-pair consensus sequence we have used for analysis also contribute to binding specificity of Forkhead DNA binding domains. We further propose that a hierarchy of Fkh1/2 binding affinities for different sequences indirectly influence binding site selection due to spatial constraints established by robustly bound and dimerized “primary” sites that place some weaker “secondary” sites in favorable locations for binding, such as loci near concentrations of Fkh1/2, and others in less favored locations (Fig. 1B). Such an effect is consistent with and may be responsible for the linear grouping along chromosomes of Fkh-activated origins separately from Fkh-repressed origins. Consistent with the notion of spatial constraints established by dimerized, primary sites, Fkh1-dsm binds certain loci more avidly than Fkh1,5 possibly due to an altered chromosomal architecture re-localizing these sites to more favorable positions.

Much remains to be learned about the regulation of Fkh1/2 binding to replication origins in addition to numerous other genomic elements, including centromeres and tRNA genes, which both exhibit spatial clustering. A very intriguing possibility is that dimerization allosterically regulates binding specificity and/or avidity. In this case, dimerization itself may be regulated, for example, by cell cycle kinases, to restrict dimerization and consequently DNA binding spatially and temporally. A related possibility is that the DNA sequence influences dimerization such that only a subset of Fkh1/2 binding sites are bound to dimerized Fkh1 or Fkh2. For example, DNA sequence can influence DNA shape that may in turn influence the structure of the bound protein through an induced-fit mechanism. Moreover, the Forkhead DNA binding domain induces a bend in DNA such that a change in Fkh1/2 conformation resulting from dimerization might enhance DNA bending to facilitate unwinding, and hence, initiation rate. Clearly, these non-exclusive, hypothetical mechanisms have great potential to regulate genome structure and function. Indeed, dynamic regulation by Fkh1/2 may allow coordination of cell cycle regulation with stress responses to protect against genome instability and replicative aging. Along these lines, a recent study has reported that Fkh1 is degraded during mitosis under stress conditions. Though we have no information on how Fkh1 domain-swapping influences stress responses, one possibility is that dimerization might protect Fkh1 from degradation, thereby protecting replication initiation potential upon cell cycle entry following recovery from stress. On the other hand, degradation of Fkh1 without cell cycle arrest would potentially enable deregulated replication as in fkh1Δ cells, and while this might be expected to lengthen S phase and provoke a replication stress response, the length of S phase is not substantially altered and we have not detected replication stress resulting from FKH1 deletion or overexpression (unpublished and).

Discovery of the domain-swap motif in Fkh1 and Fkh2 in S. cerevisiae was based on its prior identification exclusively in the human FoxP family from amongst the 18 Fox protein subfamilies. The other well conserved Fox protein in yeast, Hcm1, shares the α–helix-breaking proline (i.e., domain-swap minus) found in virtually all other human Fox subfamilies, leading to the remarkable realization that exclusively monomeric as well as dimer-capable Fox proteins are conserved from yeast to humans. Thus, a deeper understanding of Fkh1/2 function in yeast should continue to reveal important insights into the wide array of molecular functions controlled by Fox proteins, particularly the still obscure area of genome architecture.