RNA polymerase II (RNAPII) transcription has been proposed to occur at transcription factories; nuclear focal accumulations of the active, phosphorylated forms of RNAPII. The low ratio of transcription factories to active genes and transcription units suggests that genes must share factories. Our previous analyses using light microscopy have indicated that multiple genes could share the same factory. Furthermore, we found that a small number of specialized transcription factories containing high levels of the erythroid-specific transcription factor KLF1 preferentially transcribed a network of KLF1-regulated genes. Here we used correlative light microscopy in combination with energy filtering transmission electron microscopy (EFTEM) and electron microscopy in situ hybridization (EMISH) to analyse transcription factories, transcribing genes, and their nuclear environments at the ultrastructural level in ex vivo mouse foetal liver erythroblasts. We show that transcription factories in this tissue can be recognized as large nitrogen-rich structures with a mean diameter of 130 nm, which is considerably larger than that previously seen in transformed cultured cell lines. We show that KLF1-specialized factories are significantly larger, with the majority of measured factories occupying the upper 25th percentile of this distribution with an average diameter of 174 nm. In addition, we show that very highly transcribed genes associated with erythroid differentiation tend to occupy and share the largest factories with an average diameter of 198 nm. Our results suggest that individual factories are dynamically organized and able to respond to the increased transcriptional load imposed by multiple highly transcribed genes by significantly increasing in size.
Introduction
The mode of transcription in which RNA polymerase II (RNAPII) scans the genome for available promoters, followed by binding, transcription initiation, and tracking along the template during elongation has been challenged by recent studies examining the location of active transcription units in the nucleus. Nascent transcript labelling studies indicate that RNAPII transcription occurs primarily, if not exclusively, at transcription factories, which are focal accumulations of phosphorylated RNAPII (Iborra et al., 1996; Pombo et al., 1999). Several RNA immunoFISH (fluorescence in situ hybridization combined with immunofluorescence) studies examining nascent transcript signals for over a dozen genes have shown that they invariably overlap with RNAPII transcription factories (Mitchell and Fraser, 2008; Osborne et al., 2004; Osborne et al., 2007; Ragoczy et al., 2006; Schoenfelder et al., 2010), which supports the proposal that RNAPII transcription occurs exclusively in factories. Estimations of the number of serine-5-phosphorylated RNAPII (RNAPII-PS5) transcription factories in several mouse tissues using immunofluorescence light microscopy have indicated that each nucleus contains only several hundreds of factories. Genome-wide gene analyses have shown that these tissues actively express greater than 6000 genes (>12,000 alleles) (Schoenfelder et al., 2010). This suggests that actively transcribed genes share these sites, and indeed nascent RNA FISH signals from genes separated by tens of millions of bases in cis, or from different chromosomes (in trans), frequently overlap at shared transcription factories (Mitchell and Fraser, 2008; Osborne et al., 2004; Osborne et al., 2007; Papantonis et al., 2010; Schoenfelder et al., 2010). In each of these cases, close spatial proximity of colocalizing genes has been confirmed by chromosome conformation capture (3C) assays.
The high frequencies of observed factory sharing coupled with the failure to observe transcription outside factories, suggests that genes either move to factories immediately after RNAPII binding and transcript initiation, or that RNAPII contact and initiation of transcription occur after relocation to a factory. Induced genes rapidly associate with factories and colocalize with constitutively transcribed genes both intra- and interchromosomally, suggesting that many alleles move to preassembled factories when they become active, rather than nucleate their own factories (Dhar and Wong-Riley, 2010; Osborne et al., 2007; Papantonis et al., 2010). 3C analyses confirmed that intra- and interchromosomal contacts are induced coincident with transcriptional activation, and 3D DNA FISH inter-gene measurements suggested that long-range associations at factories occur through movement of the induced gene toward a preassembled factory (Osborne et al., 2007). This dynamic model is supported by the finding that long-range intra- and interchromosomal interactions between genes are lost and constitutively active genes dissociate from transcription factories when transcription initiation is blocked by heat shock, whereas RNAPII-PS5 foci are still observable in the absence of transcription (Mitchell and Fraser, 2008). Shortly after return to 37°C, active genes were once again transcribed at factories, which indicates that gene association is dynamic and transcription-dependent. Detailed analyses of the long-range contacts between two distal induced genes revealed dynamic contacts between the two genes that followed the progress of the polymerase complex through the transcription unit, consistent with a mechanism in which the two transcription units slid past each other during transcription, as if being reeled through a shared factory (Papantonis et al., 2010).
Transcription factories appear to specialize. Episomal constructs driven by various promoters resulted in the clustering of episomes at a subset of factories dependent upon the similarities of their promoters (Xu and Cook, 2008). Constructs with the same promoter driving different reporters tended to colocalize during transcription, whereas those with heterologous promoters tended to localize to different sites. Similarly, networks of endogenous genes tend to cluster. A genome-wide enhanced 4C screen (e4C; a variant of 3C) for genes that share factories with the mouse α-(Hba) and β-globin (Hbb) genes revealed intra- and interchromosomal associations with hundreds of other gene loci from nearly all chromosomes in erythroblast nuclei (Schoenfelder et al., 2010). These studies revealed spatial networks of preferred transcription partners that were overrepresented in genes regulated by the erythroid-specific Kruppel-like transcription factor, KLF1. Immunolocalization of KLF1 in erythroblast nuclei revealed 30–40 foci, the vast majority of which overlapped with RNAPII factories. Investigation of the position of active alleles of KLF1-regulated genes suggested that they had a higher probability of being active if associated with a KLF1 factory and, when there, were clustered with several other KLF1-regulated genes. Collectively, these data demonstrate the compartmentalization of the transcriptional machinery in mammalian nuclei and indicate that the genome is non-randomly organized to take advantage of specialized sites that might be optimized for efficient transcription of co-regulated gene networks.
Previous studies using energy filtering transmission electron microscopy (EFTEM) have revealed that transcription factories in HeLa nuclei are nitrogen-rich proteinacious structures with a mass of approximately 12 MDa and a mean diameter of 87 nm (Eskiw et al., 2008). Here, we used correlative microscopy on ultrathin physical sections in combination with EFTEM to show the size and mass distribution of transcription factories in mouse foetal liver erythroblasts. We showed that factories are located in regions of low chromatin density with an average of 545 factories per nucleus. EFTEM demonstrated that these foci are protein-rich domains with a mean diameter of 130 nm, containing on average 24 MDa of protein. Using this method, we also compared the size and mass distribution of factories associated with the transcription factor KLF1. We also investigated factories containing the highly transcribed Hbb and Hba gene loci using a novel electron microscopy in situ hybridization (EMISH) technique. Together, these data characterize transcription factories in primary mouse erythroblasts tissues at the ultrastructural level and begin to describe the physical diversity of the nuclear structures involved in transcribing different subsets of genes.
Results
Transcription factories in mouse erythroblasts
We used immunolabelling of active, RNAPII-PS5 in embryonic day 14.5 (e14.5) mouse foetal liver cells to determine the distribution of transcription factories (Fig. 1). In whole cells (Fig. 1A–C), many factories are discernable as foci; however, the numbers of these are difficult to assess due to the high amount of signal present in other focal planes and the poor z-resolution of light microscopy. To improve image resolution, immunostained cells were embedded in epoxy resin and 90 nm ultrathin sections prepared. Imaging of factories in ultrathin physical sections greatly improves z-resolution to the thickness of the section, through removal of out-of-focus light. Imaging by this method demonstrates that RNAPII-PS5 is located in foci primarily in regions of low chromatin density (Fig. 1D–F). The numbers of transcription factories were counted in 34 medial, physical sections from different nuclei (supplementary material Table S1; raw counts per section). Performing stereological calculations to account for the Holmes effect [see Materials and Methods, and (Eskiw et al., 2008)], we determined that there were an average of 545±275 factories per nucleus with a range of 120–1044 (Fig. 1G). Previous studies have shown there are a minimum of 6000 active genes (potentially 12,000 active alleles) at this stage of erythroid development (Schoenfelder et al., 2010). Even if only half of the potentially active alleles are transcribing at any given moment in a single cell, these results suggest that the average transcription factory contains approximately ten transcription units.
We then performed correlative microscopy in combination with EFTEM to quantify the size and mass distribution of transcription factories at the ultrastructural level. Briefly, after immunolabelling of e14.5 liver cells, embedding and ultrathin sectioning, fluorescence images of transcription factories in physical sections were collected (Fig. 2A–C) and the positions of these cells marked. EFTEM was then performed on the transcription factories identified within these nuclei. EFTEM detects the energy loss of electrons that have interacted with specific atoms within the specimen and provides quantitative elemental information on the composition of structures (Eskiw et al., 2008). For example, electrons that interact with phosphorus (P) atoms within the specimen lose 132 e-V of energy, whereas electrons that interact with nitrogen (N) atoms lose 409 e-V of energy. Images, or ‘maps’, of P and N are then generated to demonstrate the position of these atoms within the specimen. First, a low magnification mass-sensitive (Fig. 2D) image was collected and overlaid with the fluorescence micrographs to indicate the location of transcription factories within the specimen (Fig. 2E). High magnification P maps (Fig. 2G) and N maps (Fig. 2H) of the regions corresponding to factories were then collected and false coloured red (P) and green (N). Merged images identify structures that are phosphorous-rich, such as chromatin or ribonucleoprotein complexes (RNP) or nitrogen-rich proteinacious structures. We identified transcription factories by correlating the RNAPII-PS5 fluorescence signal overlaid on the high magnification P and N maps (Fig. 2I,J). Previous studies (Eskiw et al., 2008) have shown that transcription factories are N-rich structures, composed primarily of protein, that make contact with the surrounding chromatin (Fig. 2J). Consistent with previous work, we found large N-rich structures underlying the RNAPII-PS5 fluorescence signals and defined the boundaries of transcription factories as the points at which the N:P ratio changed dramatically (supplementary material Fig. S1). We analysed 30 transcription factories identified by this method. Few P atoms were detected within the transcription factory structure, consistent with structural RNA components or the heavily phosphorylated C-terminal domain (CTD) of the large subunit of RNAPII. However, the level of P within the structure was not consistent with the presence of chromatin within the factory, in agreement with previous studies indicating that transcription occurs at the surface of factories (Eskiw et al., 2008). P-rich structures present near the surface of factories are consistent with RNP particles and chromatin. We defined transcription factories based on the continuity of the N-rich structures underlying the fluorescence signals. Factories had a normal distribution of diameters with a mean of 130 nm, after correction for the Holmes effect (Fig. 3). This mean diameter is significantly larger than the 87 nm mean diameter for transcription factories in HeLa cell nuclei (Eskiw et al., 2008).
EFTEM provides quantitative information on elemental composition
We used the EFTEM images to calculate approximate protein and nucleic acid content of transcription factories through comparison with structures of known composition. Nucleosomes within the image are recognizable as N- and P-rich structures with a diameter of 11 nm and a N to P signal ratio of 1:1, consistent with previous EFTEM analyses of physical sections of human cells and purified nucleosomes (Bazett-Jones et al., 1999; Hendzel et al., 1999). Using nucleosomes as internal structures of known composition, we calculated the mass of protein within each factory. Assuming that most nucleosomes contain two of each canonical core histone (H2A, H2B, H3 and H4) (supplementary material Table S2A) wrapped by 146 bp of DNA with a GC content of 42%, we determined that there are 1980 N atoms and 292 P atoms per nucleosome (supplementary material Table S2B). Of these, 1438 N atoms would be contributed from the histone octamer and 542 from DNA. These values were used to determine the arbitrary intensity units per atom within each map (see Materials and Methods). A minimum of five nucleosomes per image were measured from random points to eliminate statistical anomalies and variations between nucleosomes. For factory masses, we determined the nucleosomal equivalents of nitrogen atoms and extrapolated the mass of protein contained within the structures. The elemental composition of transcription factories in mouse erythroblast nuclei was similar to that of transcription factories identified in HeLa cells; the structure being primarily protein-based with chromatin contacts on the surface and P-rich granules within the core. Erythroblast factories have an average protein mass of 26 MDa, considerably greater than HeLa cells.
KLF1-specialized transcription factories are larger than the population average
Many erythroid-specific genes are regulated by the erythroid kruppel-like transcription factor, KLF1. Previous studies have demonstrated that KLF1-regulated genes preferentially cluster at a subset of specialized, KLF1-enriched transcription factories (Schoenfelder et al., 2010). We immunolabelled e14.5 mouse erythroblasts for RNAPII-PS5 and KLF1, embedded them in epoxy resin and generated ultrathin sections. Fluorescence micrographs of RNAPII-PS5 factories (Fig. 4A), KLF1- (Fig. 4B) and H33342-counterstained chromatin (Fig. 4C) were merged (Fig. 4D). As before, RNAPII-PS5 foci were found in regions of low chromatin density. KFL1 was also in foci in regions of low chromatin density, with some KLF1 foci colocalized with transcription factories. We found an average of 60±40 (range 8–132) KLF1-specialized transcription factories per cell after correction for the Holmes effect. Chromatin immunoprecipitation sequencing data predicts that there are between 950 and 1380 KLF1-regulated genes (Tallack et al., 2010), which provides further evidence that these genes share the limited number of KLF1-associated transcription factories.
We then used EFTEM to examine transcription factories enriched in KLF1. Fluorescence images of RNAPII and KLF1 were overlaid with low magnification mass-sensitive images (Fig. 4E) to show the areas of interest, which were then analysed at higher resolution. P (Fig. 4F) and N (Fig. 4G) maps were false coloured red and green, respectively, and merged (Fig. 4H). Line scans through KLF1-associated transcription factories demonstrated that these structures are protein-based and nucleic acid-poor (Fig. 4K), similar to other factories. However, they were significantly larger (P=0.0021; two-way Mann-Whitney test) with a mean diameter of 174 nm (supplementary material Fig. S2). KLF1-containing transcription factories also had significantly more mass of protein (36 MDa).
Factories transcribing the globin genes are larger than the population mean
To examine factories associated with the highly transcribed Hbb and Hba gene loci, we developed a variation of the RNA immunoFISH protocol, called electron microscopy in situ hybridization (EMISH), which allows for the detection of nascent transcripts while preserving nuclear architecture at the ultrastructural level (supplementary material Fig. S3). Erythroblasts were incubated with hapten-containing oligonucleotide probes, followed by antibody detection of nascent transcripts and transcription factories. Following detection, cells were again embedded in resin, thin sectioned, and subjected to fluorescence microscopy to identify the nascent RNA and RNAPII-PS5 signals (Fig. 5). Both Hba and Hbb-b1 nascent RNA signals were detectable in physical sections, and >97% were associated with transcription factories, which was consistent with previous studies (Osborne et al., 2004; Schoenfelder et al., 2010). EFTEM was then performed on transcription factories associated with the globin nascent transcripts. P and N maps were again collected and aligned with the mass-sensitive and fluorescence images (Fig. 5F). High resolution images demonstrated a protein-rich structure, corresponding to the factory fluorescence, adjacent to RNP particles (Fig. 5G). These RNP particles corresponded to the fluorescence signal identified as Hbb-b1 nascent transcripts. Analysis of transcription factories adjacent to Hbb-b1 or Hba transcripts were indistinguishable from one another and together had a mean diameter of 194 nm, which was significantly larger (P=0.0028; two-way Mann-Whitney test) from the mean diameter of the population (supplementary material Fig. S2). EFTEM comparison of the ultrastructure of cells treated by EMISH and control cells demonstrated no significant differences in nuclear morphology or chromatin architecture, which indicated that this technique is suitable for analysis of transcription sites (supplementary material Fig. S3). Quantification of the mass from these factories demonstrated that they contained ~38 MDa of protein.
Genes share transcription factories at the ultrastructural level
Next, we used EMISH to investigate gene sharing at transcription factories. The results of several RNA immunoFISH studies using light microscopy led to the proposal that active genes can share factories. Although 3C-like techniques support these conclusions, others have suggested that light microscopy does not have the resolution to detect colocalization at factories (Brown et al., 2008; Lawrence and Clemson, 2008). We performed EMISH with probes for Hba and Hbb nascent transcripts (Fig. 6). We located nuclei in the ultrathin sections in which the nascent RNA signals overlapped and again performed EFTEM. Fig. 6 clearly shows two overlapping nascent RNA signals for Hbb and Hba with a single underlying N-rich structure indicative of a transcription factory, which strongly supports the previous conclusion that genes share factories. Line scans indicate that this structure is N-based and low in P content (Fig. 6J), similar to factories identified previously. We confirmed that the size and composition of factories transcribing these gene pairs were indistinguishable from those identified from single-gene EMISH analyses.
Discussion
Here, we have shown that transcription factories in murine erythroblast nuclei are significantly larger than those reported in HeLa cell nuclei. To investigate the ultrastructure of sites of nascent transcript synthesis of specific genes, we developed a new and relatively simple EMISH method. We used EMSIH and correlative EM to show that specialized factories enriched in KLF1, as well as those factories transcribing the Hba and Hbb genes, are larger than the population mean. Similarly to HeLa cells (Eskiw et al., 2008), murine erythroblasts contain transcription factories that are primarily protein-based structures that contain very little phosphorus-rich material within the core of the structure. These protein-based factories make intimate contact with the surrounding chromatin environment with adjacent RNP.
We were successful in identifying sites of specific nascent transcript synthesis, through the development of EMISH, to examine factories responsible for transcribing specific genes. From these analyses, we demonstrated that transcription factories specializing in the transcription of genes involved in erythroid differentiation form a distinct subset of larger transcription factories.
To demonstrate sites of nascent transcript synthesis of specific genes we developed a new and relatively simple method that allows for ultrastructural analysis. This method is based on chemical fixation in formaldehyde of ex vivo foetal liver cells to preserve structure and the addition of single-stranded DNA (ssDNA) probes against intronic sequences to localize specific nascent transcripts. This method, in combination with EFTEM, provides the first high-resolution images of the environment in which specific genes are transcribed and provides quantitative information on the transcription factories responsible for their synthesis.
Previous studies had identified between 30–40 KLF1-enriched transcription factories in e14.5 erythroid cells (Schoenfelder et al., 2010). This was derived using confocal imaging of whole cells and performing deconvolution of the images. Here, we used physical sections to estimate the number of KLF1-enriched factories and determined that there are ~60 per nucleus. Given the number of predicted KFL1-dependent genes (Tallack et al., 2010), we estimate that 15–25 genes are transcribing in a single factory, assuming that not all KFL1-dependent alleles are transcribing simultaneously. This, in combination with the loss and regaining of interactions seen from heat shock studies (Mitchell and Fraser, 2008), is compelling evidence for transcription factories mediating gene co-associations.
Analysis of the volume indicates that transcription factory structures might differ from cell type to cell type. Comparisons of the mean factory volume demonstrated that transcription factories in cells from the erythroid lineage (1.05×10−3 μm3) are three times that of HeLa cell factories (0.35×10−3 μm3), and that KLF1-associated factories are eight times larger (2.76×10−3 μm3). The spatial distribution of transcription factories within the nucleoplasm was approximately 7.4 factories/μm3, consistent with the values predicted by Faro-Trindade and Cook, (2006) (between 6.8 and 14 factories/μm3). This average value, however, is slightly misleading because the distribution of factories is not uniform throughout the nucleoplasm, tending to be primarily in regions of low chromatin density. The amount of nucleoplasm with low chromatin density does vary dramatically between cell types and therefore the likelihood of observing transcription factories is highly dependent on the location examined. Regardless, this density does allow rough prediction of the total number of transcription factories present within a nucleus of given volume.
Our transcription factory density calculation is reassuring. The value demonstrates that we detected a significant proportion of total transcription sites and that our calculations are accurate. We were unable to label structures with gold particles, as it is notoriously difficult to label nuclear proteins with gold-conjugated antibodies. Attempts were made to post-section label or to label thick-cut cryo-sections; however, this was unsuccessful probably due to insignificant epitope presentation. Due to these technical limitations, we used correlative microscopy to successfully identify transcription factories. Correlative microscopy allowed us to identify the structures of interest without the use of gold particles. This is fortuitous because gold particles, and other electron-dense materials, mask the underlying ultrastructural and elemental information and so we were able gather unobstructed images of the entire physical structure of transcription factories.
One caveat of our method is the use of antibody labelling to identify structures of interest because antibodies contain N atoms that add mass to our predicted values. This is, however, a requirement and the minimal amount of antibodies were used to perform the labelling. For example, we performed direct labelling of primary antibodies to increase signal. We chose to directly label ssDNA probes for the purpose of EMISH, which proved to be adequate for the detection of nascent transcripts in whole cells; however, upon physical sectioning not enough fluorescently labelled material was present to be detected. We are therefore confident that we have limited the amount of material added to provide reliable detection of the structures of interest.
The quantitative elemental information collected using EFTEM allowed us to predict the mass of transcription factories. This was based on using nucleosomes as internal standards for elemental composition. Nucleosomes are the only unit of chromatin with known dimensions and compositions: 11 nm nucleosomal core consisting of eight histones and ~146 bp of DNA. Although we have made assumptions that the histones are canonical and that the DNA is made up of 42% GC content, nucleosomes containing a histone variant or alterations in GC compositions would not significantly change the predicted 1980 N atoms and 292 P atoms. Using these values, we predicted that the mean mass of protein within a transcription factory from e14.5 erythroid cells was 26 MDa. The protein mass is the equivalent to 52 RNAPII holoenzyme complexes, based on the amino acid sequences of proteins in the RNAPII holoenzyme complex. However, it is likely that other accessory proteins required for initiation, elongation and termination are present, as well as a host of proteins ranging from histone-modifying enzymes to specific transcription factors and adaptor proteins. This mass and the co-association of genes at a single factory strongly support the view that these large complexes contain more than a single polymerase or single gene. This becomes more apparent when examining the average mass of protein associated with transcription factories enriched in KLF1 (36 MDa) or those factories transcribing the Hbb and Hba genes (38 MDa).
Studies have shown that transcriptional machinery can accumulate at artificially inserted sites within the nucleus (Janicki et al., 2004; Kumaran and Spector, 2008). These artificially induced arrays appear to accumulate large amounts of RNAPII, building ‘super factories’ that might reflect the high rate of transcription and the demand of high amounts of transcriptional machinery. This might be parallel to the puffs in Drosophila polytene chromosomes. These puffs are several hundred copies of transcribing heat shock genes with a large accumulation of transcriptional machinery; essentially a large super factory (Yao et al., 2007). In both the artificially induced systems and polytene chromosomes it appears that factories grow in size; however, it is unclear whether there was a smaller factory at these sites prior to transcriptional activation, which would indicate that factories grow in size as the transcriptional load increases. This could account for the increased size in KLF1-enriched factories as well as in those factories that are transcribing highly expressed genes of the erythroid lineage. These genes have a higher transcriptional load and, therefore, require more polymerase and hence larger factories. This increase in transcription factory size could represent both the accumulation of RNAPII components as well as accessory proteins (such as splicing machinery) needed for efficient transcription to occur.
Materials and Methods
Cell preparation
Plugged blb/c or blk6 female mice were terminated at e14.5 by cervical dislocation. Foetuses were removed and placed in ice-cold PBS. Foetal livers were removed, placed in 1 ml ice-cold PBS and gently homogenized. 60 μl of cell suspension was placed on polylysine-coated slides (Sigma) and the cells allowed to attach for 2 minutes. Slides were then placed in PBS containing 4% formaldehyde for 5 minutes at room temperature. Slides were then washed for 5 minutes in PBS at room temperature and permeabilized for 10 minutes in PBS containing 0.5% Triton-X 100. Slides were rinsed in PBS for 5 minutes at room temperature and then placed in 70% ethanol at room temperature for 10 minutes. Slides were stored in 70% ethanol at −20°C. The ultrastructure appeared consistent with previous reports, showing regions of both heterochromatin and euchromatin, as well as other nuclear structures such as nuclear pore complexes that are indicative of good nuclear preservation. All animal experiments were performed according to the relevant regulatory standards.
Stereological calculation of factories per nucleus
The number of foci was determined by acquiring fluorescence images of single sections, containing RNAPII-PS5-labelled e14.5 foetal liver cells, at constant exposure times. Images were processed identically and thresholds set. Masks of images were then created and the number of the factories within each physical slice of nucleus determined. We used the diameter of factories as determined by electron microscopy, i.e. 130 nm, to calculate the number of factories per cell. Given the thickness of the section and the diameter of nucleus, we extrapolated the total number of factories for the entire nuclear volume. This number was then corrected for the Holmes effect, which states that because the structure is larger than the section thickness, objects will be represented multiple times. We therefore derived the formula S/(D+S) to calculate the number of factories per nucleus, where D is the diameter of factories and S is the section thickness (90 nm in these experiments). This formula demonstrates that only 41% of the factories calculated for the entire nuclear volume can be considered because they will be represented in at least two of the physical sections.
Probes
Probes for Hba and Hbb-b1 transcripts (Eurogentec, Southampton, UK) contained three modifications of either dinitrophenyl (DNP) or digoxigenin (DIG) at the 5′-end of the oligonucleotide.
EMISH
Slides with attached cells were removed from −20°C 70% ethanol, and incubated at room temperature in 70% ethanol. Cells were rehydrated in PBS at room temperature for 5 minutes. Slides were equilibrated in 50% formamide, 2× saline–sodium citrate buffer for 15 minutes at room temperature. Then, 100 μl of hybridization mix (Schoenfelder et al., 2010) containing the probe was placed on the slide and incubated at 37°C for 1 hour. Probes for Hbbb1 and Hba were used at concentrations described previously (Schoenfelder et al., 2010). Following hybridization, slides were washed for 5 minutes in Tris–saline solution (TS) at room temperature and then in TS containing 0.1% Triton-X 100 for 5 minutes. Slides were blocked in blocking agent (Roche) and the probes detected by antibody labelling. DNP-containing probes were detected using rat anti-DNP antibody conjugates to A488 dye (1:50, Roche) and then donkey anti-rat FITC (1:200, Jackson Scientific). DIG-labelled probes were detected using sheep anti-DIG (1:1000, Roche) conjugated to A555 dye and donkey anti-goat Cy3 conjugates (1:200, Jackson Scientific). RNAPII was detected using a 1:200 dilution of a rabbit antibody against the serine 5 phosphorylation of the C-terminal domain of Rpb1 (Abcam, AB5131) directly conjugated to A555 dye (Invitrogen).
EFTEM
Cells were prepared for EFTEM after EMISH or immunofluorescence by first post-fixing slides overnight at 4°C in 8% formaldehyde in PBS. After quenching in 155 mM glycine for 30 minutes at room temperature, cells were dehydrated in a graded series of ethanol washes (50, 70, 90 and 100%) at room temperature. Cells were then infiltrated with Quetol 651 (Agar Scientific) epoxy resin mix (with nonenyl succinic anhydride, NSA, and methyl-5-norbornene-2,3-dicarboxylic anhydride, NMA) for 2 hours with two changes, and then infiltrated with Quetol resin with hardener (2% final concentration) for 2 hours before polymerization overnight at 65°C. Physical sections were cut on a Lieca ultracut microtome, and fluorescence images collected on an Olympus X61 upright epi-fluorescence microscope. EFTEM images were collected on a Tecnia 20 (FEI) TEM with a 200 Ke-V field emission gun fitted with a post-column Gatan imaging filter.
Diameter calculation
To determine the diameters of the factories, the structures identified by correlative microscopy were measured by EFTEM and the area occupied by the N-rich factory was calculated. Since the aspect ratio (length of longest axis divided by the length of the perpendicular axis) of these structures was <1.5, we could treat these structures as circles while introducing <3% error in the calculation (Weibel, 1979). Using , where r is radius and A is area, we determined the average radius of a factory. Because the average diameter of the factories is larger than the section thickness (~90–100 nm sections) we needed to correct for the Holmes effect using Fullman's formula (Weibel, 1979). The corrected value for the diameter of transcription factories was calculated to be ~130 nm (129.6 nm).
Mass calculation
Mass calculation for the protein and nucleic acid content was done using the same formula previously described (Eskiw et al., 2008). Mass estimates were calculated as nucleosome equivalents and converted to RNAPII core enzyme (murine Rpb1-12) equivalents. The area of a nucleosome (94.99 nm or 78 pixels) was multiplied by the average intensity to give the total intensity units for N or P. This number was then divided by the total number of atoms within the nucleosome to give the average intensity units per atom. Each nucleosome on average will contain 292 P atoms, which is equal to 146 bp of DNA. Given an average GC content, 542 N atoms per nucleosome will be present from DNA, and using the canonical histone sequences 1438 N atoms will be present from the nucleosomal core proteins for a total of 1980 N atoms per nucleosome. The total integrated intensities (area multiplied by average intensity units per pixel) for N and P were divided by the average intensity per atom derived from nucleosomal calculations, and the total number of N and P atoms determined per factory by dividing the total intensity of N or P within the identified area by the intensity of units per atom. Correction for the Holmes effect was performed as previously described (Pombo and Cook, 1996).
Funding
This work was supported by the Medical Research Council and the Biotechnology and Biological Sciences Research Council, UK [grant number BB/E017460/1]. P.J.F. is a Senior Fellow of the MRC [grant number G117/530]. Deposited in PMC for release after 6 months.