Our work shows that heterochromatin, long recognized as a key factor in the developmental programming of gene expression, also plays an integral role in the timing of the early embryonic cell cycles. Satellite sequences successively acquire features of heterochromatin, becoming late replicating by cycle 14, which prolongs S phase. This prolongation of S phase slows the early cell cycles and allows the progression to MBT [16
Our description of the successive introduction of the features of heterochromatic reveals a lack of interdependency of these features. For example, since it occurs earlier, the compaction of the satellite sequences is independent of late replication and of HP1 binding. We also have been able to visualize the events of late replication with unprecedented spatial and temporal resolution that gives insights into the replication of compacted HP1-bound chromatin.
Onset of late replication of satellites prolongs embryonic S phases
Origin spacing could contribute to S phase length (). However, EM studies show that origin spacing changes only slightly from 7.9 kb to 10.6 kb from preblastoderm embryos to cycle 14 [10
]. Since forks are thought to converge at a rate of 3 kb/min, the additional separation would extend S phase by about 1 min [10
], a minor contribution to the change from a 3.4 to a 50 min S phase.
S phase duration would also increase if all of the replicons did not replicate at the same time [1
]. However, asynchrony in replicon firing can happen in two ways, organized and unorganized. By unorganized asynchrony, we mean that origins fire at different times without regard to their position in the genome. In this case early and late firing origins can be juxtaposed. When replication from an early firing origin reaches an adjacent later firing origin just before it fires, one, rather than two, forks replicates the inter-origin distance, doubling replication time. Greater unorganized asynchrony will result in passive replication of later origins and reduce the number of origins. Thus, given the known origin spacing, unorganized asynchrony is unlikely to make a very major contribution to the more than ten-fold increase in S phase length between pre-blastoderm cycles and cycle 14.
If replication asynchrony is organized so that large regions of the genome (replication units) have many similarly behaving replicons, early initiated forks invading a late region from the outside will not have time to replicate a significant portion of the large domain. The insulation resulting from distance can greatly magnify the impact of asynchrony on S phase duration. Organization of genomes into large replication units is widespread but poorly understood. We show that the satellite sequences are replication units, and that embryonic changes in S phase duration result from change in the schedules of their replication. In preblastoderm cycles, satellite sequences replicate early, finishing in synch with general replication (). Subsequently, satellite replication is increasingly delayed in parallel to S phase prolongation. Importantly, when satellite replication is late, it is deferred, not slow. For example, AATAC sequences begin to replicate 18 min into S phase 14 (). Each type of satellite sequence exhibits distinct replication delays. The 359 sequence has almost no delay in S phase 14, while AATAT and AATAACATAG (data not shown) finish replicating after 359 but before AATAC. The stereotyped schedules suggest that each replication unit has a characteristic “lateness” parameter. This lateness parameter appears to be continuously variable in that there are many replication units each with its own schedule of replication.
The Changing Character of Replication in Embryonic S phases
Though its replication is delayed, a unit such as AATAC replicated quickly once initiated (10 min). Although the 359 satellite is more slowly replicating (~15 min), we suggest that it may be composed of separately and asynchronously replicating subdomains that we sometimes resolve (e.g. 2–4). We conclude that the dynamics of replication within a replication unit changes only modestly during the early cycles (from 3.4 to roughly 10 min).
In summary, our results argue that by S phase 14, the genome is replicated as a series of units each of which replicates relatively quickly, but that a temporal program of sequential replication of these units creates a long S phase (concept embodied in the schema shown in ). This replication program resembles a consensus view of replication in slowly replicating cells of mammals and plants. We conclude that progression from coincident replication of all of the replication units in a rapid S phase to sequential replication in a prolonged S phase 14 underlies prolongation of early embryonic S phases in Drosophila.
Developmental regulation of heterochromatin and its role in late replication
In widely divergent species and biological settings, heterochromatin has common characteristics including, compaction, transcriptional quiescence, late replication, “repressive” histone modifications, and association of specific heterochromatin proteins. This intimate association suggests mechanistic coupling of these features. If this were so, the various heterochromatin characteristics would emerge coordinately at the same moment during development. Instead, our observations show temporal uncoupling during early Drosophila embryogenesis.
Although it was suggested that heterochromatin appears at cycle 14 [19
], both the cytological and biochemical manifestations of heterochromatin develop progressively [20
] (Victoria Foe, p.c.). Foci of compacted chromatin that align with satellite sequences appeared in pre-blastoderm embryos prior to, and hence independently of, late replication and HP1 recruitment (, S2, S3
). Furthermore, HP1 binding to satellite sequences occurred late in cycle 14, after the onset of late replication. Since HP1 would have to decorate the satellite sequences at the onset of cycle fourteen if it were required to suppress early replication and promote late replication, we conclude that the late replication of satellite sequences is specified independently of the HP1 binding. Finally, satellite sequences reorganize; the previously independent foci of satellites aggregate into a large coherent HP1 positive region at the very beginning of interphase 15. This intimate association of satellites, which makes the chromocenter more coherent, is downstream of cycle 14 events and the MBT.
Together our findings show that satellite sequences acquire the features of heterochromatin progressively. Compaction is present early, late replication is introduced subsequently, and recruitment of HP1 and then chromocenter maturation follow. Onset of position effect variegation suggests that heterochromatic suppression of transcription begins in G2 of cycle 14 and mounts subsequently [23
]. Thus, heterochromatin does not form in a single step, and it acquires increasing influence during critical developmental events surrounding the MBT and gastrulation.
Replicating compacted sequences
If compaction of chromatin prevents replication, decompaction might accompany or provoke replication. Our real-time observations of PCNA and HP1 in cycle 15 show replication adjacent to, but not overlapping HP1-bright foci of compacted chromatin. A more diffuse HP1 region appears adjacent to bright HP1 foci; PCNA overlies these fainter partner foci. Each partner focus appears and disappears as the PCNA signal rises and declines. We conclude from this that replication does not occur in the compacted domain and that the sequences in the compacted HP1-bright focus unfurl during replication.
The persistence of in situ foci for 359 and AATAC shows that the satellites are not fully decondensed during replication. The size of partner HP1 foci also argues for limited decompaction. If an entire focus of compacted HP1-bright chromatin were to disperse, it would expand in volume, but the partner focus is about the same size as the brighter parent focus. Thus, we suggest that that a partner focus represents decompaction of a portion of the sequences harbored in the adjacent HP1-bright focus.
Following replication heterochromatic sequences rapidly recompact. After an initial expansion, the partner HP1 focus does not grow throughout replication, and it shrinks and disappears as replication declines. When pulsed with fluorescent nucleotides for less than the replication time of the satellite, fluorescence overlies the compacted satellite sequence. Thus, we suggest that DNA is “spooled” out of compacted foci, replicated and returned to compacted foci shortly after replication. We roughly estimate the duration of replication-associated decompaction in embryonic cycle 15 as one min. The dynamics, which are not easily consistent with decompaction of large topological domains, suggest that active replication forks drive local unfolding of chromatin structure, but we cannot exclude the possibility that transient decompaction might promote replication.
The developmental program
We are interested in mechanisms that couple the changing cell cycle behavior with development. Previous work suggested that the gradual prolongation of early cycles is secondary to gradual prolongation of S phase [16
]. A model in which the exponentially increasing amounts of DNA titrates replication components to prolong S phase is attractive, but not presently supported.
Our results show that if a titration mechanism governs S phase duration, it is indirect. Injection of α-amanitin in cycle 13 prevented onset of late replication, accelerated S phase 14, and caused an early synchronous mitosis. Thus, activity of at least one of the DNA dependent RNA polymerases is required to slow S phase, and the replication “hardware” needed for a rapid S phase are not limiting. Accordingly, if a titration mechanism were involved, the titrated component would regulate an upstream process. For example, transcription is restricted prior to cycle 14, and titration of a repressor might derepress transcription in late cycle 13, indirectly triggering onset of late replication.
Three findings suggest an abrupt switch to late replication at the beginning of cycle 14: the dramatic increase in S phase length, the accompanying switch of satellite sequences to delayed replication, and the requirement for transcription in cycle 13 for this transition. However, we see that the late replication program of cycle 14 is anticipated by slight delays in replication of satellite sequences in cycles 12 and 13. These early changes suggest a more progressive process. We propose that early slight changes in replication timing and transcription initiate a positive feedback process that precipitates an abrupt change at the MBT. Rapid cell cycles suppress transcription [7
] and limit the time available to modify newly replicated chromatin, but, once the cycle begins to slow, transcription and heterochromatin modifications would accelerate to create conditions permissive for late replication, which would further slow the cycle.