Background

Transcription in eukaryotes is organized in transcription factories (TFs), which arenuclear domains where several genes are grouped to be transcribed together [1, 2][3–5]. The current opinion is that the genes in a TF interact by a loopingmechanism [5–8]. It has been suggested that chromatin looping plays an important role incontrolling gene activity by bringing together promoters and enhancers or TFs [9]. Some studies suggest that promoter-enhancer loops are maintained by theinteraction of proteins associated with these cis-regulatory elements [10]. This interaction precedes chromatin activation, which is required for generelocation to the TF by an unknown mechanism [11]. It has been proposed that TFs are maintained by depletion attraction forces(excluding volume effect) between RNA pol II molecules [12]. However, experimental evidence has shown that genes remain at the factoryeven when active RNA pol II is not present [13]. This makes the excluding volume model very improbable and suggests that thisstructure is not the result of transcription. Instead experimental evidence points tohistone acetylation as being responsible for loop formation [11]. For these reasons we explored the possible contribution of chromatinacetylation in the formation of TFs.

Chromatin at the TF is decondensed [2] and contains active transcription marks like histone acetylation or H3K36me3 [14]. Among all the possible Lysine residues that can be acetylated, H4K16Ac isvery special because it prevents the formation of compacted chromatin by inhibiting theinter-fibre interaction [15–18]. Moreover, H4K16 acetylation is associated with both active chromatin [19] and with the active transcription marker H3K4me3 [19–21].

Results and discussion

To confirm whether H4K16Ac is associated with active chromatin, we analysed thedistribution of H4K16Ac in the nucleus of the TFs of circulating lymphocytes. The TFswere visualised as sites of incorporation of Br-UTP into nascent RNA. TFs appeared asdiscrete foci distributed along the edge of condensed chromatin (Figure 1a) as previously described in other cell types [2, 22, 23]. H4K16Ac was scattered in foci overlapping or very close to these Br-RNAsites (Figure 1a). To study the extent and degree of thehyper-acetylated chromatin in individual transcription units (TUs), we deconstructed thenuclei of these cells by making chromatin spreads. Under these conditions active RNApolymerases and epigenetic modifications of the chromatin are preserved. This treatmentdisassembled nuclei and spread templates over a wide area. The DNA adopts a linearstructure with no visible nucleosomes [22] and 95% of active polymerases remain associated with the DNA [24]. When we stained the chromatin spreads with antibodies against H4K16Ac, theyshowed almost continuous fluorescent tracks along the DNA fibres (Figure 1b). These H4K16Ac tracks corresponded with active chromatin, as demonstratedby co-localisation of H4K16Ac with the nascent transcripts that were labelled eitherin vivo or in vitro by using Bromo-Uridine (BrU) or BrUTPrespectively. The area covered by acetylated histones was larger than that stained bythe nascent transcripts (Figure 1c). This was to be expectedbecause histone acetylation extends over long stretches of genes, whilst only a few RNApol II molecules are ever found on a given gene [22]. Nevertheless, to demonstrate that the distribution of active RNA pol IImolecules is not an artifact of non-natural nucleotide incorporation, we carried outchromatin spreads with cells where transcription was not labelled with BrU. Ourexperiments demonstrated a similar co-localisation of H4K16Ac with P-RNA pol II(hyper-phosphorylated Ser2) (Figure 1c).

Figure 1
figure 1

Transcription on acetylated chromatin. (a) The nascent Br-RNA(green) and H4 K16Ac (red) signals are closely associated. BrUTP incorporation inhuman lymphocyte was carried out for 15 min and after fixation immunolabeledtogether with histone H4 K16Ac (rabbit antibody). (b) The deconstruction ofcell nuclei. After sarkosyl treatment, chromatin was spread and immunolabelledwith H4 K16ac, to show tracks of hyperacetylated chromatin. (c) Thecolocalisation of Br-RNA after BrUTP and H4 K16ac. Br-RNA appears as little spotson tracks of acetylated chromatin, equivalent images were obtained when P-RNA polII (Ser2) antibody (H5) was used. (d) Tracks of acetylatedchromatin appeared in clusters. (e) The distribution of sizes of chromatinacetylated tracks. (f) The distribution of sizes of chromatin betweenconsecutive acetylated tracks. (g) Expression data from FCDP mix cells onmouse chromosome 10. Expressed genes tend to cluster along the chromosome. Forcluster analysis we used a 500 Kb window. When clustering was significant (p>0.95)a blue line is drawn. Bars: a = 2 μm, merge = 200 nm; b, c, d = 10μm.

Full size image

The chromatin spreading technique allowed us to measure the length of H4K16Ac tracks.The distribution of H4K16Ac stretches showed a lognormal distribution with average sizeof ~15 Kb (Figure 1e). H4K16Ac tracks rarely appeared isolated,instead they tended to cluster, spanning several hundreds of Kb (348 ± 90; range235–530 Kb) (Figure 1d). The extension of the gaps betweentwo consecutive H4K16Ac tracks in the cluster showed a lognormal distribution with anaverage distance of ~30 Kb (Figure 1f). The analysis of thepolymerases loaded onto H4K16Ac tracks showed that not all the tracks were stained withBr-RNA or P-RNA pol II. The number of nascent transcripts or P-RNA pol II per track waslow (0.7 ± 1 transcripts/track and 0.8 ± 0.9 P-RNA pol II/track). This was inaccordance with our previous findings, suggesting that most of the TUs contain onemolecule of RNA pol II [22]. The fact that some H4K16Ac tracks of chromatin were not associated to RNApol II or Br-RNA could reflect a temporal discrepancy between the transcription andacetylation processes of chromatin. Indeed, transcription by RNA pol II takes only a fewminutes [25–27] while deacetylation of active chromatin can take several hours [28], providing a molecular memory of recently-transcribed chromatin. On the otherhand, H4K16Ac tracks are not a special feature of lymphocytes as we were able to findthe same chromatin organisation in all the mammalian cell types tested including: Hela,Epstein Barr transformed lymphocytes, human lymphocytes, primary human fibroblasts,primary mouse fibroblasts and murine erythroleukemia cells (both differentiated andundifferentiated).

The clusters in all the different cell types analysed were identical with respect to thenumber of TUs (8 ± 2 TUs/Cluster), suggesting that co-linear active genes expressedat the same time, in agreement with the analysis of expression data using FDCP mix cells [29]. The sliding window analysis (applying a window of 500 Kb) over the entiregenome showed that genes are active in clusters (Figure 1g), inaccordance with our chromatin spreads data. Moreover, our results are consistent withthe co-expression data after a Serial Analysis of Gene Expression where the cluster sizewas <500 Kb [30]. From these data we can conclude that co-linear TUs are active at the sametime in the same cell.

How are these TUs organised in the cell nucleus?

Collinear active TUs are enriched in H4K16Ac which confers stiffness and inhibitsinter-fiber interaction [15–17]. In this way, chromatin appears as a multi-block copolymer with stiff andflexible monomers (rod-coil)n system, where the rod is the stiff activeTU. The multi-block copolymers function as amphiphiles whose components segregateinto domains to avoid unfavourable contact with each other. In these systems,complete phase separation is prevented by the covalent linkage between the components [31]. The rod block does not have the same conformational entropy as the coilblock and this restricts homogeneous packaging. In consequence anisotropicinteractions occur between the stiff blocks ending in a liquid crystalline domainwhere the different TUs are aligned in a high order smectic phase [31] (Figure 2a). Multi-block copolymers can adopt manydifferent structures depending on the relative proportions of the rod and coilphases. For example, when the rod phase is lower than 20% the structures obtained aremicrospheres [32] (Figure 2b). A calculation of the amount of activechromatin in a Hela cell line gives a proportion of rods to coils of ~12%, which isconsistent with active chromatin separated in many microspheres. These spheres forH4K16Ac chromatin were observed in the cell nucleus of human lymphocytes (the shapefactor was 0.93 + 0.05). Microspheres are regularly distributed in artificialpolymers with regular coils and rods resulting from the repulsion forces of coilspushing in all directions. In the cell nucleus TFs are not regularly distributedbecause the sizes of genes and intergenic distances are not as regular as inartificial polymers. Moreover, H4K16Ac foci concentrated at the edge of condensedchromatin (Figure 1a). A possible reason for this discrepancyis that chromatin cohabits with the inter chromatin compartment (ICC), which iscomposed by RNPs and proteins. This results in a biphasic system where the ICC(inelastic phase) segregates from chromatin [33], creating an interphase between both components. Under these conditionsmicrospheres containing H4K16Ac may be pushed by the coil polymers to the interphasebetween chromatin and ICC (Figure 2c), resulting in thelocalisation along the edge of the chromatin as observed.

Figure 2
figure 2

Multi-block copolymer. (a) Chromatin exists as a multi-blockcopolymer with flexible and stiff (coil-rod) chromatin. Stiff blocksself-interact creating an anisotropic crystalline phase (pile of red blocks).(b) These segregated blocks generate a symmetric microphase patternwhere the stiff phase (minor component) self organise in micro-spheres (redcircles) embedded in the major phase of flexible DNA (blue background).(c) The organization of the active chromatin microphase in the cellnucleus. Microspheres remain in a fixed position due to the balance of opposingforces. In the cell nucleus the existence of interphases creates an imbalancebetween the forces acting in opposing directions, with a net force pushing themicrospheres to the interphase. This explains the position of active chromatinat the edge of the condensed chromatin. (d) Model of self-organizationof active chromatin. Collinear active gene chromatin is hyper-acetylated whichresults in its self-interaction with neighbouring TUs. These interactions aredestroyed by the deacetylation of chromatin.

Full size image

A prediction of the multi-block copolymer model is that microphase separation mustpersist as long as H4K16Ac is present in chromatin. In fact, H4K16Ac foci wereunperturbed by treatments like 2 M NaCl extraction, which disrupts chromatin;transcription inhibition by DRB (5,6-Dichlorobenzimidazole1-β-D-ribofuranoside), which reduces RNA pol II transcription by 98%; and heatshock (1h 45°C), which releases RNA pol II from the DNA [26] (Additional file 1: Figure S1). The only way todisrupt these foci was by formamide treatment, which works as a solvent for theelectrostatic self-assembled polymers (Additional file 2:Figure S2).

These experiments contradict the excluding volume model [12] and are in agreement with the multi-block microphase separation hypothesisproposed in this study.

The next question about the genes in a TF is where they come from. Several studieshave shown that genes in cis and in trans are able to interact in the same TF [5–7]. However, the analysis of chromatin spreads showed that collinear genesare active in the same cell at the same time. This guarantees that several H4K16Actracks are in close proximity. Therefore, most of these collinear TUs would probablyaggregate in the same microsphere, as occurs in similar situations with themulti-block copolymers [34]. The experimental evidence from chromosome configuration capture analysissuggests that local chromatin is the primary source of interaction for any genomicloci [8]. Nevertheless, we cannot exclude the possibility that some genes locatedfurther away in the same chromosome or in another chromosome can interact due toproximity or chromatin folding.

Finally, a remarkable feature of TFs is their constant size across species anddifferentiation stages [23]. According to the multi-block copolymer model for chromatin organisation,the way to change the size of H4K16Ac foci (and consequently TFs) is by increasingthe number of active genes in a given region or by unrestricting the mobility of theactive chromatin. The latter has been reported in experiments using plasmids thatrendered larger TFs than the endogenous ones [35, 36].

In summary, we present evidence of the relationship between epigenetic marks and theTF structure. Our model proposes that active chromatin self-organises in the nucleusdue to the special physical properties of H4K16Ac modified chromatin. Therefore, ourmodel implies that chromatin becomes activated (H4K16Ac modified) before joining aTF. This is conceptually very different from current transcription factory model,which proposes that genes are targeted to TFs to “enhance production byconcentrating the relevant machines, resources, and expertise in one place” [37].

Materials and methods

Transcription in vivo and in vitro

For in vivo transcription, cells grown on coverslips were incubated in presence of2.5 mM BrU for several min.

For in vitro transcription, cells grown on coverslips were treated as described [3].

Chromatin spreading

Cells (103 cells in 5 μl) were spotted onto a 22 × 50mm glass slide and 5 μl of lyses buffer were added (Lyses buffer: 1%sarkosyl, 25 U/ml ribonuclease inhibitor, 10 mM EDTA, and 100 mM Tris–HCl (pH7.4)). After 10 min at 20°C, the slide was tilted to allow the drop to run down.Samples were air-dried and fixed in 4% Paraformaldehyde for 10 min. Clusters weredefined as two o more hyper acetylated tracks in less than 100 Kb. For quantificationof clusters of hyper acetylated chromatin between 150 and 200 tracks of hyperacetylated chromatin were analysed.

Immunofluorescence

After blocking for non-specific antibody binding, immunolabelling was carried out asdescribed [3]. For detection of primary transcripts, we used mouse anti-IdU/BrdU (5mg/ml; Caltag Laboratories, Burlingame, CA). For detection of H4K16ac we usedantibodies raised in rabbit and mouse (Serotec, Kidlington, UK, Abcam). RNA pol IIhyperphosphorylated in Ser 2 was detected with H5 antibody (Covance). Secondaryantibodies donkey anti-mouse IgG or IgM tagged with Cy3 (1/200 dilution; JacksonImmunoResearch, Bar Harbor, ME) and donkey anti-rabbit IgG tagged with Alexa 488(1/200; prepared using a Molecular Probes kit, Inc., Eugene, OR). DNA staining wasperformed with 200 nM TOPRO-3 (Molecular Probes) for 5 min. Then coverslips weremounted on slides using Vectashield (Vector laboratories), and images were collectedusing a Radiance 2000 confocal microscope (Bio-Rad Laboratories, Hemel Hempstead,Herts, UK), Distances were measured using EasiVision software (Soft Imaging SystemsGmbH, Münster, Germany) and data exported to Excel (Microsoft) for analysis.

The degree of spreading of the chromatin was measured by hybridising the spreads witha fragment of DNA of 47.26 Kb; the spreading was 3.9 + 0.2 Kb/μm.

Microarrays and sliding window analysis

Mouse FCDP mix cells were used. cRNA synthesis and hybridisation to oligonucleotidearray were performed as described [29].

The sliding window analysis was performed by applying a window of 500 Kb over thechromosomes and moved at 5 Kb steps along a chromosome to know whether the genescontained in that window were more likely to be transcribed together than just bychance.