SEARCH

SEARCH BY CITATION

Keywords:

  • chromatin;
  • live-cell imaging;
  • nuclear organization;
  • stochastic gene expression;
  • transcription

Abstract

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

The Dictyostelium model has a set of features uniquely well-suited to developing our understanding of transcriptional control. The complete Dictyostelium discoideum genome sequence has revealed that many of the molecular components regulating transcription in larger eukaryotes are conserved in Dictyostelium, from transcription factors and chromatin components to the enzymes and signals that regulate them. In addition, the system permits visualization of single gene firing events in living cells, which provides a more detailed view of transcription and its relationships to cell and developmental processes. This review will bring together the available knowledge of the structure and dynamics of the Dictyostelium nucleus and discuss recent transcription imaging studies and their implications for stability and accuracy of cell decisions.

The last few years have a seen a resurgence of interest in gene regulation in Dictyostelium. This new interest has been stimulated by the conservation of nuclear components apparent from the Dictyostelium discoideum genome sequence, the consequent utility of reagents such as histone modification-specific antibodies designed for other organisms, and the development and improvement of methods for imaging nuclear processes in living cells. The purpose of the review is to summarize recent developments in our understanding of Dictyostelium nuclear biology, with a focus on chromatin organization and transcription imaging, and to integrate these developments into broader themes of transcriptional mechanism and cell specification during development. However, it is first appropriate to introduce the context of the biology we will discuss, by outlining what is known of the structure of the Dictyostelium nucleus.

Nuclear organization in Dictyostelium

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

Dictyostelium discoideum amoebae have nuclei of approximately 3 microns in diameter, and contain a haploid genome of 34 Mb (the yeast genome is three times smaller and human 100 times larger). The nuclear morphology is somewhat round, but heterogeneous, with distortions perhaps caused by the highly dynamic cytoskeleton. The genome is organized on six chromosomes, with the exception of around 100 rDNA-containing minichromosomes of 88 kb (Sucgang et al. 2003). Nuclei undergo a fenestrated mitosis (Moens 1976), where the spindle microtubules operate through a partially disrupted nuclear envelope. The components of the nuclear envelope are sparsely characterized, although they contain conserved components such as the inner nuclear membrane protein, Sun1 (Xiong et al. 2008). Mitosis in Dictyostelium takes around 5 min (Muramoto & Chubb 2008). Nucleoli reside at the periphery of nuclei, and there are usually one to three lobes (Balbo & Bozzaro 2006).

Chromosome organization has been revealed by the genome project (Eichinger et al. 2005). The gene density is high, with 62% of the genome predicted to encode protein. Each chromosome carries a cluster of repeats rich in Dictyostelium intermediate repeat sequence (DIRS) transposable elements near one end. It is likely that these tracts contain the centromeres, as they cluster adjacent to the centrosome at the nuclear edge during interphase, while dispersing at mitosis. The repeats are enriched in histone H3 methylated at lysine 9 (Chubb et al. 2006a), and recruit HP1 (Kaller et al. 2006a) and the centromeric histone H3 variant, H3v1 (DdCenH3) (Dubin et al. 2010). No “magic” centromere sequence has been associated with these repeats (Glockner & Heidel 2009), unlike the defined sequences of the yeast models. The ends of the chromosomes consist of some rDNA sequence (Eichinger et al. 2005). It is presently unclear if repeats exist more distally. Sequences at the telomeres of the rDNA minichromsomes have been identified, comprising four near perfect 29bp tandem repeats and a more distal CnT repeat (where n is 1-8) (Emery & Weiner 1981). The Dictyostelium genome encodes a telomerase homologue (DDB_G0293918).

Nuclear DNA is replicated in a short S-phase (around 40 min per 8 h average cell cycle) and DNA replication begins immediately after completion of mitosis (Muramoto & Chubb 2008). There are distinct phases to DNA replication, with most of the nuclear DNA replicating in the first half of S-phase and the heterochromatin-like sequences (transposons/centromeres) replicating in the second half. Using the DNA-replication marker, GFP-PCNA (proliferating cell nuclear antigen), the early part of S-phase is characterized by a diffuse labeling overlaid by faint spots (Muramoto & Chubb 2008). These spots may be equivalent to mammalian replication factories (Gillespie & Blow 2010). In late S-phase, the heterochromatin replicates at a single bright fluorescent spot (visualized by GFP-PCNA or BrdU) at the nuclear edge. Late replication of heterochromatin is a feature shared by mammalian nuclei, contrasting the early replication of silent chromatin in budding and fission yeast.

Nucleosomal components and modifications

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

The genome complement of histones in Dictyostelium reveals greater diversity than budding yeast, yet retains enough simplicity for genetic approaches to be very useful. The organism lacks the large repetitive tracts of histone genes found in many animal species but most of the histone subtypes have a number of divergent variants, permitting diversity of function. Histone molecular genetics are sufficiently advanced in Dictyostelium that point mutations can be targeted into endogenous histone loci (Muramoto et al. 2010).

The major histone types in the core Dictyostelium nucleosome can be revealed by mass spectrometry of acid extracted nucleosomes (Fig. 1A), and comprise the standard core nucleosomal histones H3, H4, H2A and H2B. In addition, a linker histone, designated histone H1, is encoded by a single gene. H4 is encoded by two genes (at different genomic locations) predicted to encode identical polypeptides. The other histone types have additional variants. The H3 family comprises five genes. Three variants, H3a, H3b and H3c have a high level of conservation with mammalian H3 genes. They are H3.3-like histones, containing substitutions that make them distinct from H3.1 and H3.2 classes (Elsaesser et al. 2010). H3a has AAIG in the histone core at residues 91–95 (typical of H3.3), H3b has AAIQ and H3c has AAIE, compared to SAVM in H3.1/H3.2. H3.3 is capable of replication-independent deposition (Ahmad & Henikoff 2002), unlike H3.1, which is incorporated into chromatin only during DNA replication. An apparent paradox is that H3b mRNA contains a clear SL (stem loop) motif (Davila Lopez & Samuelsson 2008), normally a feature of replication-dependent histones. Unlike replication-independent histones, replication-dependent histone mRNAs are not polyadenylated, and contain a 3′ SL motif associated with a partially independent RNA processing pathway involving U7 snRNP and the stem loop binding protein SLBP (DDB_G0288225 in Dictyostelium). The H3b and H3c genes are adjacent in the genome. H3c expressed sequence tags (ESTs) cannot be detected in EST databases (Urushihara et al. 2006), and we found no convincing evidence for H3c protein. Two more H3 variants, H3v1 and H3v2 have also been defined in the genome sequence. H3v1 is likely to be a centromeric H3 variant (Dubin et al. 2010). There is also a Dictybase gene model for an additional distant H3 variant (DDB_G0278587).

image

Figure 1.  Post-translational modification of core histones in Dictyostelium. (A) Crude nuclei were purified as described (Charlesworth & Parish 1975) with minor modifications. Histone extraction was performed by incubation with 0.4 N sulfuric acid for 1 h on ice and then the sample was precipitated with trichloroacetic acid (TCA). (B) Individual histones were digested with Trypsin, Chymotrypsin, Lys-C, or Arg-C, to cover various amino acid regions. The digested samples were analyzed by nano-HPLC electrospray ionization multistage tandem mass spectrometry (nLC-ESI-MS/MS). The resulting data were submitted to a MASCOT program (Perkins et al. 1999) for searching against Dictyostelium and histone protein databases. ac, acetylation; me, methylation; P, phosphorylation. Numbers of dots under methylated lysines represent mono-, di-, or tri-methylation.

Download figure to PowerPoint

Dictyostelium has five H2A genes. The major nucleosomal variant is H2AX, although peptides encoded by the gene annotated H2AZ can also be identified. Three other divergent variants (H2Av1-3) are encoded in the genome. Dictyostelium has three H2B variants. The major nucleosomal H2B is H2Bv3, and the genome has a hybrid gene, H2Bv1, which consists of a H2B histone domain followed by a divergent H2A sequence. The placement of an H2A-H2B dimer in the same polypeptide may have interesting structural implications. A further variant, H2Bv2, consists of N-terminal H2B sequence, with a long C-terminal domain with no identifiable sequence homologies.

Whilst commercial antibodies against modified histone residues react with Dictyostelium histones (Kaller et al. 2006b), small changes in epitopes may render some antibodies ineffective. We therefore surveyed, using mass spectrometry (Martin et al. 2000; Aebersold & Mann 2003; Garcia et al. 2007), the modifications of the major Dictyostelium histone variants from asynchronously growing cells. Post-transcriptional modifications on histones H2AX, H2Bv3, H3a, H3b and H4 in Dictyostelium are shown in Figure 1. We observed patterns of methyl and acetyl marks on histone H3 and H4 with many characteristics of mammalian histones. We detected many modifications usually associated with transcriptional activation on H3 such as methylated K4, K36 and K79 (using standard H3 numbering after alignment with human H3) and acetylated K9, K14, K18, K23, and K27. In addition, we identified both H3K9Me2 and H3K9Me3, modifications absent from yeast, in agreement with earlier antibody data (Chubb et al. 2006a; Dubin et al. 2010). The mitosis-associated phosphorylation of H3S10 was also detected, although was rare, perhaps because mitosis is around 1% of the cell cycle (Muramoto & Chubb 2008). Although H3K27Me is found in many complex eukaryotic systems (Kouzarides 2007), we found no evidence of this modification in Dictyostelium. H3K27 methylation is linked to polycomb function in many metazoans, so the absence of H3K27Me in Dictyostelium may coincide with the apparent absence of polycomb complex components in the organism. By preparing the sample in the DNA damaging agents bleomycin and cisplatin, we were able to detect phosphorylation of S150 on Dictyostelium H2AX, as previously detected by immunoblot (Hudson et al. 2005). Evidence for modification of H2AZ was not detected, although we identified peptides spanning more than 70% of the protein.

Transcription dynamics

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

It is now straightforward to visualize nascent RNA at single endogenous genes in live Dictyostelium cells. Standard methods to measure transcription such as northern blotting, microarrays, reverse transcription–polymerase chain reaction (RT–PCR) and RNAseq are ensemble measurements of disrupted cells. Although useful, these techniques create a population average and individual cells cannot be followed through time. In many systems, single cell transcriptional activity has been inferred by measuring fluctuations in fluorescent proteins (Elowitz et al. 2002; Ozbudak et al. 2002; Bar-Even et al. 2006; Sigal et al. 2006), or enzymes (Rutter et al. 1995) at the single cell level. Although extremely useful, these techniques are influenced by protein folding as well as RNA and protein turnover, and depending upon these parameters, can potentially miss fluctuations occurring over shorter time scales (Dong & Mcmillen 2008).

Single cell and in some cases single molecule analyses of transcription have been possible in fixed cells for a number of years, using RNA fluorescence in situ hybridization (FISH) (Femino et al. 1998; Raj et al. 2008). These techniques were instrumental in revealing the noise inherent to transcription (Larson et al. 2009), have illustrated how network architecture can minimize developmental error (Raj et al. 2010), highlighted the lack of transcriptional order in cells over short integration times (Gandhi et al. 2011), and have provided evidence for important models of transcriptional mechanism (Raj et al. 2006; Zenklusen et al. 2008). However, the cells are dead, therefore dynamic information is lost. Live analysis of multicopy transgene expression has been possible in recent years in human tissue culture cells (Janicki et al. 2004; Darzacq et al. 2007), in addition to more indirect methods for looking at transcription, such as the appearance of regulators at heat shock loci of polytene chromosomes (Yao et al. 2006). More recently, single allele transcriptional events were measured in HEK-293 cells using Flp recombinase-integrated reporters (Yunger et al. 2010), and mice bearing a transcriptional reporter in the locus encoding β-actin have also been generated (Lionnet et al. 2011).

Detecting transcription in living cells takes advantage of a RNA hairpin from the genome of the RNA bacteriophage MS2 (Fig. 2) (Bertrand et al. 1998). The hairpin has a high affinity, sequence-specific interaction with the MS2 phage coat protein. If GFP is fused to the coat protein, the fluorescence is directed to the RNA in living cells. Dictyostelium allows rapid generation of strains with insertions at endogenous loci, so hairpins can be introduced as tandem arrays into genes of interest. The hairpins are incorporated into nascent RNA at the site of transcription. In the recombinant cells, the MS2-protein-GFP fusion is co-expressed. At the site of transcription, the MS2-GFP binds the hairpins, causing an accumulation of GFP at the site of transcription, which can be revealed, using standard laboratory microscopes, as a fluorescent spot.

image

Figure 2.  (A) The MS2 protein (grey) binds with high affinity and specificity to the MS2 RNA stem loop. The MS2 protein is fused to green fluorescent protein (GFP) (green circle). (B) Multiple MS2 repeats (24 repeats; 1.3 kb) are integrated into the 5′ region of the gene of interest by homologous recombination. A drug resistance cassette is used for selection of recombinants. (C) Upon transcription by RNA polymerase (grey oval), the MS2 repeats are incorporated into the newly synthesized RNA and form stem-loops, creating a binding site for the MS2-GFP fusion protein, which is constitutively expressed in cells. (D) Discontinuous transcription of developmentally induced gene in Dictyostelium cells. An example of single cell transcription during development is shown. Arrows indicate the transcription site in Dictyostelium cells. Timing is in minutes.

Download figure to PowerPoint

By minimizing photodamage using imaging conditions of low light together with sensitive cameras, the fluorescent RNA spots were observed to appear and disappear at irregular intervals, a behavior we called pulsing (Fig. 2). Pulsing was first observed for the early developmental gene dscA (Chubb et al. 2006b), but has subsequently been observed in Dictyostelium for two housekeeping genes (act5 and scd) (Muramoto et al. 2010) and a second developmental gene, ecmA which can be induced in the lab by the addition of specific signals (Stevense et al. 2010). Discontinuity of transcription is not something apparent in standard ensemble measures of transcription, although it is perhaps not especially surprising. For dscA, the pulse lengths and intervals were highly variable, with shorter pulses (5 min or less) the most common, but longer pulses, of over 20 min also detected (Chubb et al. 2006b). The dscA gene was observed in several thousand cells at different 30 min time windows during early development, which revealed several features of the pulsing phenomenon. First, cells expressing early during capture, showed a greater tendency to express later in the window than a cell that had not previously expressed. This was interpreted as an observation of a “transcriptional memory”, which may relate to earlier observations that dsc genes are expressed in only a subset of cells (Clarke & Gomer 1995). Second, there was variability in the number of cells expressing during development- with a strong surge in expressers during the first half hour, dropping over the next few hours, followed by a strong surge in the number of cells expressing just before aggregation (over 40%). Third, the individual cell responses, measured as pulse length, interval, intensity and frequency, showed only small changes over the whole of early development, evidence of a binary transcriptional response, which we return to later.

From other data in the literature, it appears discontinuous transcription is a conserved phenomenon, from bacteria to mammals (Raj & van Oudenaarden 2008; Chubb & Liverpool 2010). Using live cell RNA counting, irregular build-up of transcripts was observed in Escherichia coli (Golding et al. 2005). In yeast, flies and mammalian cells, snapshots of transcriptional activity inferred by single molecule FISH revealed broad distributions of RNA number per cell, implying significant proportions of the population in active and inactive transcriptional states (Raj et al. 2006; Zenklusen et al. 2008; Pare et al. 2009). Put together, these data have been suggested to imply that transcription is a bursting process, where genes exist in two states, one inactive, and one with a certain probability of transcription, with a slow fluctuation between these states. Bursting is not likely to be a property of all genes, as a study in yeast found strong evidence for a more simple transcription mechanism (Zenklusen et al. 2008). The variance for transcript number was low, suggesting a simple probability describing the likelihood of productive transcription. One potential confusion arising from bursting models is that extrinsic noise could distort RNA number without the need for bursting transcription, for example via inaccuracies in segregation of cellular components at cell division (Huh & Paulsson 2011) or heterogeneity in the cell cycle (Brooks 1981). Does the noise arise from transcriptional mechanism or is it simply channeled through it?

The question “what does pulsing mean for the cell?” is perhaps not one with a satisfying answer. How could most genes do otherwise? Protein complexes are dynamic rather than static. Transcriptional complexes, potentially unstable, may fall apart, and transcription would cease until a new complex forms. A nucleosome might hinder an initiating or elongating polymerase, causing a block, and a delay in transcription (Voliotis et al. 2008). The appropriate chromatin environment, and transcription factor concentration may only be found when a locus arrives at a particular nuclear location. The signals that regulate the gene are also unlikely to be continuous in many cases. Pulsing may be advantageous, in situations where flexibility is beneficial. The analogy with the thermostat has been made (Larson et al. 2009), whereby pulsing allows finer and more rapid control over transcriptional responses than a single burst. Noise in gene expression, potentially arising from noisy transcription, provides the source of initial variability in many models of cell fate specification (Losick & Desplan 2008). Lateral inhibition is thought to operate by the amplification of initial differences between cells. This type of amplified noise has also been predicted for changes in ES cell fate (Kalmar et al. 2009). As biology evolved out of noise, we must also remember that just because there is some residual noise left in the system, it is not necessarily important.

Noisy transcription may be an obstacle to cells as they struggle to exert control over gene expression (Raser & O’shea 2005). During development, cells are exposed to signaling factors at many different concentrations. In a number of developmental biology models, cells assess signal strength with precision and respond appropriately, within limited timescales. Cells may be expected to adopt different fates with <15% change in signal (Lander 2007), a problem if gene expression is too approximate. The potential heterogeneity of a transcriptional response to signal was revealed by a second study in Dictyostelium using the MS2 reporter integrated into the ecmA gene (Stevense et al. 2010). The ecmA gene is a stalk cell marker and induced by the extracellular signals, DIF and cAMP, rapidly and by a signaling mechanism involving no intermediate gene expression (Verkerke-Van Wijk et al. 1998). A dose-response analysis revealed the pulse length, frequency and intensity as highly variable between cells. Some spots were continuous for nearly an hour after induction, whereas other cells pulsed only briefly, or not at all. High variability was apparent at all signal doses, even at doses normally high enough to convert all cells into stalk cells in a standard overnight exposure. With this variability in response to strong uniform signal, how could developmental signaling be accurate?

In situations where the cell has several hours to integrate all its transcriptional data, this may not be a problem. The slow responders obviously catch up in the case of stalk cell differentiation. In addition, robustness in cell signaling networks may provide stability in cell function, despite noise in individual genes, although one can envisage that noise could also be amplified by a network, if its source were events early in signal perception.

In other developmental contexts, such as the rapid laying down of pattern in flies and amphibians, there may not be enough time for “signal integration”, particularly with gradients as shallow as 10% change in signal concentration per cell row (Gregor et al. 2007). In these cases one observes strategies that might reduce noise: 1. Restricting transcription to keep noisy fluctuations below “threshold”. 2. Synchronizing cell cycles (this is an interesting idea – in Dictyostelium, cell cycles are highly asynchronous, as they are in many mammalian cell types, so at a constant transcriptional rate of a stable RNA, daughter progeny will have considerable diversity of transcript amounts. By synchronizing divisions, this diversity could be restricted, although noise arising from uneven segregation might persist). 3. Minimizing cell movement. Dictyostelium are very motile, and sampling many environments will favor variable transcriptional responses, unlike many embryo situations, where cells are near relatives. The resultant “salt-and-pepper” pattern of differentiation is reminiscent of differentiation in the early mouse embryo, although heterogeneity here may have other sources.

An additional feature of the ecmA stimulation experiments was that average cell responses were not greatly changed over a 20-fold range of signal concentration. Effects were detected, but these were mild, and parameters such as pulse frequency and pulse intensity had similar magnitudes at high signal strengths as they did in the few cells expressing without added signal. This comes back to the “all-or-none” effect suggested by the dscA experiments, although for dscA the independent variable was developmental time, not signal strength. All-or-none effects have been observed in analogous experiments in bacteria (Vilar et al. 2003), in synthetic (Becskei et al. 2001) and viral (Ko et al. 1990; Walters et al. 1995) systems and in the yeast mating type response (Malleshaiah et al. 2010). Cooperativity is suggested as a cause, perhaps involving a signaling scaffold, chromatin modification, or simply that individual signaling factors might have the opportunity to activate multiple “downstream” components before being turned off, generating amplification. A key question is to what extent one can observe “graded” effects in a native system.

Inheritance of transcriptional states

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

Another problem posed during development, after the specification of cell state has been initiated, is how cells maintain gene expression programs. To what extent do cells maintain transcriptional states down cell lineages? Cell division and growth results in dilution and turnover of initial materials – how is this manifest in change or stability of transcriptional behavior?

To gain insight into the extent to which cells maintain transcriptional states through complete cell cycles, we used the MS2 system in Dictyostelium to compare the transcriptional frequency of mother and daughter cells. By imaging strongly expressed housekeeping genes over complete cell cycles, we found the transcriptional frequency of actin (act5) and fatty acid hydrolase (scd) genes is similar in related cells (Muramoto et al. 2010). Differences in frequency between daughter cells were compared to differences between randomized cell pairs. Daughter cells were significantly more similar in their transcriptional frequencies than unrelated cell pairs. This effect was also detectable in the second generation (mean 8 h later), although considerably weaker.

Does a cell need a strategy for maintenance of state? Many cells retain their neighborhoods during development, so their signaling environments are also likely to be retained (Stathopoulos et al. 2002; Zeitlinger et al. 2007). Obviously, this may not apply so often for motile cells, such as neural crest, germ cells, blood cells, or Dictyostelium, which traverse different environments over their lifetimes. Indeed when transcription was measured with all the related and randomized cells coming from the same imaging field, the persistence of transcriptional frequency relative to random comparisons was still fully apparent for act5, implying no stabilizing effect of the environment, at least for this gene (Muramoto et al. 2010). Simply removing signals is not sufficient to dedifferentiate the cell in many situations, otherwise there would be no barrier to reprogramming, and explants would not retain commitment. Furthermore, tissues are of mixed composition. Cells are likely to receive signals from other cells with different fates, although they may be programmed to receive and act upon these signals in certain ways, which could stabilize alternative fates despite shared signals.

What strategies can cells use to maintain transcriptional programs? The divided cells share, at least temporarily, cytoplasmic and nuclear components. Without looking for specific regulators, gene regulatory networks themselves are predicted to provide robustness (Huang 2009). Positive and/or negative feedback between transcription factors provide stability of transcription (Acar et al. 2005) and stability can have strong cytoplasmic effectors (Zacharioudakis et al. 2007). Chromatin states have been implicated in the transmission of active transcriptional states through mitosis (Brickner et al. 2007; Kundu et al. 2007; Laine et al. 2009; Tan-Wong et al. 2009). We found that mutations in components of the pathway required for histone H3K4 methylation were required for transcriptional stability (Muramoto et al. 2010), most simply illustrated by observations that 95% of daughter pairs have <1.5 fold difference in transcriptional frequency in wild type, whilst 35% of daughter pairs had greater than two fold difference in frequency in H3K4 methylase mutants. This instability of transcriptional state is paralleled by instability of cell state, with cells lacking H3K4Me having a strong tendency to differentiate precociously. Transcriptional stability was also perturbed in cells with an alanine into K4 position of genomic histone H3a, suggesting stability mediated by chromatin. There was no detectable effect on transcriptional stability in mutants of the H3K36 methyltransferase Set2 (DDB_G0268132) and the Dnmt2 methyltransferase DnmA (Kuhlmann et al. 2005; Katoh et al. 2006), which appears to methylate DNA in Dictyostelium. It is presently unclear whether the effects of H3K4Me are direct or indirect. An obvious indirect hypothesis to test is whether the loss of transcriptional stability is a consequence of the loss of cell state stability (precocious development). Do mutations affecting developmental speed without perturbing H3K4Me phenocopy the transcriptional effects of loss of H3K4Me? Consistent with a direct role for H3K4Me, both H3K4Me2 and H3K4Me3 are associated with the 5′ ends of the act5 and scd genes. However, these marks are very abundant in eukaryotic genomes, so this alone is insufficient evidence of directness.

One hypothesis is that the effects of H3K4 methylation to provide stability may be mediated by chromatin remodeling factors. Chromatin remodeling has been implicated in the minimization of gene expression noise (Raser & O’shea 2004). H3K4 methylation can recruit the chromatin remodeling factors, ISWI and NURF, in addition to histone deacetylases and components of the core transcription complex TFIID (Santos-Rosa et al. 2003; Li et al. 2006; Wysocka et al. 2006; Vermeulen et al. 2007; Kim & Buratowski 2009). How might defective chromatin remodeling cause a noisy or unstable transcription phenotype? A recent study detected a conversion from a unimodal to bimodal expression distribution when a nucleosome was artificially positioned over a transcription factor binding site (Bai et al. 2010). Is this increase in variance equivalent to a loss of transcriptional stability? While the association of H3K4Me3 with components of the TFIID complex (Vermeulen et al. 2007) implies a model whereby H3K4Me3 is a mark for positive feedback, enhancing the probability of subsequent initiation, we found no consistent changes in mean transcriptional frequency in the mutations affecting H3K4Me3, so although TFIID may be important, the effects may not be intuitive. One obvious feature of the pulsing patterns in the H3K4Me mutants was a tendency for irregular long strong bursts. One can envisage how this might be a problem caused by the sluggish or inappropriate deposition of a nucleosome.

The H3K4Me dependent stability may have some similarity with effects of Trithorax mutations in flies, which cause failure to maintain patterns of homeotic gene expression (Ringrose & Paro 2004). Trithorax is also an H3K4 methylase. H3K4Me is a barrier to reprogramming in transplanted frog nuclei (Ng & Gurdon 2008) and the nematode germline (Katz et al. 2009), again linking these modifications to stability of gene expression state. Fluctuations in transcriptional states took around a cell generation in Dictyostelium (Muramoto et al. 2010). Fluctuations in fluorescent protein levels took around two cell generations in Human H1299 lung carcinoma cells (Sigal et al. 2006). The sensitivity of Hela and MCF10A cells to inducers of apoptosis (time to death) was also more similar between siblings from a cell division, an effect detectable for up to two cell generations (Spencer et al. 2009). These data, from transformed human cells and Dictyostelium, are not so different in magnitude that the processes at work, presumably dilution and turnover of cell components, may be similar in the different contexts. This type of transient stability may be important in situations where variability and fluctuations in gene expression are required. Transcriptional fluctuations are a feature of the pluripotent state (Chambers et al. 2007; Chang et al. 2008). Fluctuations in gene expression may also contribute to the development of small populations of cells with transient increased tolerance to anti-cancer agents in tumors (Sharma et al. 2010) and generate sub-populations of slow cycling melanoma cells with enhanced self-renewing capacity (Roesch et al. 2010). In both cases, for resistance and self-renewal, the generation of these pools of cells requires Jarid 1 H3K4Me demethylases, again pointing to a requirement for H3K4Me restricting fluctuations in cell state.

Surges and decays in transcriptional behavior

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

All the live cell studies of transcription in Dictyostelium suggested one additional feature, for four different genes, despite their different protein products and expression characteristics, and despite the different types of experiment carried out. After a period of relative transcriptional silence, there was an initial surge in the transcriptional response, before a relaxation back to a lower level. For dscA, this surge came in the first hour after the induction of starvation. For ecmA, this surge in the number of cells responding peaked at around 15 min of stimulation. Perhaps related to this, for act5 and scd, the transcriptional pulses were much longer in the first hour after mitosis (we have never observed a nascent RNA spot during mitosis) and more cells could be observed to express during this window. What could explain these surges, and their subsequent relaxation?

An initial hypothesis, particularly for the mitotic experiments, which involved long-term imaging, was that the apparent transcription decay occurs because of phototoxicity. However, the rather normal cycle times and re-elevated spot frequency after subsequent mitoses implied that cells were still healthy (Muramoto et al. 2010). One explanation is that there is just more transcription during these phases and the initial signaling may take time to decay. In the case of ecmA, where we controlled the extracellular signal, use of a non-degradable cAMP analogue did not greatly alter the trajectory of the response, arguing degradation of cAMP was not causing the decay of transcription, although DIF degradation could not be ruled out. Downregulation could also be an intracellular phenomenon. An alternative hypothesis is that the gene is not initially sufficiently primed, and that initial sluggish polymerase could generate a dense backlog of polymerases behind, generating slow transcribing trains, but longer or brighter pulses, with greater prominence under the microscope. The mechanics of such a process might involve nucleosomes in the way of the first polymerases. Perhaps the first polymerases destabilize nucleosomes (perhaps by the use of histone modifications they bring with them, perhaps by just forcing them out of the way). The initial sluggishness of polymerases would also be expected just after mitosis, when chromatin would be more compact. In addition, S-phase commences in Dictyostelium just after mitosis (Weijer et al. 1984; Muramoto & Chubb 2008). Additional obstructions might be replication forks, or simply the temporary coercion of nuclear structure and chemistry for a non-transcriptional role.

The strategy to distinguish between the two models of higher activity vs. slow polymerase traffic is to photobleach the spots. Rapid recovery would indicate elevated activity during initial firing. Slower recovery during the initial surge would indicate slow polymerases. One caveat here is that the bleached MS2-GFP might be replaced on the same RNA with the fresh MS2-GFP. Earlier FRAP studies on the MS2 system suggest the RNA-protein interaction is stable (Boireau et al. 2007), implying a bleaching approach would be appropriate.

Perspectives

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

The next issues to be resolved are clearly issues for the single cell field as much as they are for Dictyostelium, evidenced by the developments in live cell transcription imaging in other systems. A better understanding of transcriptional mechanics will require the detection of transcriptional events in living cells with greater sensitivity, and the ability to count single RNAs appearing at the site of transcription will be desirable, although non-trivial. This may require improvements along many fronts, from improved camera technology, to improved RNA detection technology. Improving signal to noise levels may also be facilitated by the selection of cell lines with reduced autofluorescence, which may or may not be possible, and the selection of fluorescent proteins with enhanced folding efficiency. Although there is a popular assumption in the field that heterogeneity in gene expression arises from transcriptional mechanism, evidence that transcription is the source rather than a channel for the noise is lacking. Other sources of variability, such as uneven mitotic segregation or heterogeneity of cell cycle timing may also prove important. How variability becomes solidified into choice and stability of cell fate is a problem requiring a combination of imaging approaches with molecular genetics, something Dictyostelium can certainly handle.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References

This work was supported by a JSPS Postdoctoral Fellowship for Research Abroad to TM and a Wellcome Trust Senior Research Fellowship to JRC. We would like to thank the College of Life Sciences Fingerprints Proteomics Facility for assistance with identification of Dictyostelium histone modifications.

References

  1. Top of page
  2. Abstract
  3. Nuclear organization in Dictyostelium
  4. Nucleosomal components and modifications
  5. Transcription dynamics
  6. Inheritance of transcriptional states
  7. Surges and decays in transcriptional behavior
  8. Perspectives
  9. Acknowledgments
  10. References