The integrity of a lamin-B1-dependent nucleoskeleton is a fundamental determinant of RNA synthesis in human cells

Over the past 20 years, nuclear organisation in higher eukaryotes has been discussed at length (Jackson, 1991; Spector, 1993; Berezney et al., 1996; Cook, 1999; Misteli, 2004), and it is generally agreed that nuclear structure and function are inextricably linked. If this is true, it is also true that remarkably little is known about the molecular principles by which the links are made. A key focus of this debate has been the possible role of a unifying framework – the nucleoskeleton or nuclear matrix – that spreads throughout the nucleus. To date, however, although compelling evidence for such a structure continues to accumulate, a convincing molecular characterisation has remained elusive.

In mammalian cells, the nuclear lamin proteins are major candidates to perform any structural roles; they form a structural nuclear framework that is composed of a peripheral nuclear lamina and a diffuse internal network (Goldman et al., 2002; Gruenbaum et al., 2003; Broers et al., 2006), with associated actin and myosin (Goldman et al., 2002; de Lanerolle et al., 2005). Two classes of lamin proteins, the A- and B-type lamins, are expressed from three genes – LMNA, LMNB1 and LMNB2. Somatic cells express at least one B-type lamin, and A-type lamins are expressed in differentiated cells, with LMNA expressing two major A-type lamins – A and C. Genetic alterations in human laminopathies (Worman and Courvalin, 2005; Broers et al., 2006) and transgenic mice with different patterns of lamin gene expression (Mounkes et al., 2003; Gruenbaum et al., 2003) demonstrate that a major function of the lamin proteins is to provide nuclear strength and integrity. In addition, lamin filaments are known to interact with many proteins, and defects in these interactions contribute to a variety of disease phenotypes (Worman and Courvalin, 2005; Broers et al., 2006).

It is widely reported that lamin proteins contribute to the normal regulation of DNA and RNA synthesis, and it has been suggested that `lamins form a scaffold required for proper organisation of both replication and transcription' (Goldman et al., 2002). Indeed, key published studies supporting this view have been performed using pseudo-nuclei assembled in vitro using Xenopus egg extracts (reviewed by Broers et al., 2006). To date, crucial roles for the nucleoskeleton in DNA and RNA synthesis of mammalian cells are based largely on circumstantial evidence. For example, one preliminary report suggests that disrupting the lamin network in BHK 21 cells will inhibit pre-mRNA synthesis (Spann et al., 2002), although a detailed characterisation of the transcription phenotype was not performed.

Other observations suggest that assembled lamin proteins are not of fundamental importance in RNA synthesis. Studies of transgenic mice that are null for lamin A/C (Sullivan et al., 1999) or express a mutated form of lamin B1 (Vergnes et al., 2004) appear to be decisive. The lamin A/C knockout mice display complex phenotypes and although viable die prematurely. Mice expressing the mutant lamin B1 die at birth from respiratory failure. Yet the ability to generate embryonic fibroblasts from these lamin-modified mice implies that neither lamin A/C nor B1 are essential for cell growth and proliferation. This conclusion does conflict, however, with studies using RNA interference (RNAi) to deplete lamin expression, which show that B-type but not A-type lamins are required for cell proliferation (Harborth et al., 2001).

With particular emphasis on understanding the interplay between nuclear structure and function, we have performed a detailed analysis of the consequences of lamin protein depletion. Using RNAi to reduce the expression of the nuclear lamin proteins, we show that depletion of the major B-type lamin – lamin B1 – results in a specific inhibition of RNA synthesis. Loss of functional capacity is accompanied by profound changes in nuclear organisation, with decreased RNA synthesis correlating with collapse of active chromatin and expansion of the inter-chromatin compartment. As a consequence of this structural reorganisation, chromosome territories (CTs) accumulate at the nuclear periphery. This suggests that the organisation of CTs is maintained by the association of chromatin with active centres of RNA synthesis and that the activity of these centres is dependent on a nuclear structure that incorporates lamin B1.

RNAi was used to reduce expression of lamin proteins in human cells using both vector-based systems and direct delivery of short interfering RNAs (siRNA). Initially, RNAi was performed using a vector-based protocol with co-transfection of vectors expressing a suitable reporter gene to identify individual RNAi-treated cells. Following RNAi, quantitative image analysis showed that selected RNAi targets reduced expression of the appropriate lamin protein to ∼20% of the endogenous levels at 36-48 hours after transfection (Fig. 1). Further reduction was seen between 48 and 72 hours.

We first evaluated how altered lamin gene expression influenced RNA synthesis. Sites of pre-mRNA synthesis were visualised using an antibody (H5) that binds to the active form of RNA polymerase II and hence marks active transcription factories (Fig. 2). Cells were co-transfected with RNAi vector and vector expressing YFP-tubulin. After different times, cells were fixed and transcription sites visualised by indirect immunofluorescence (Fig. 2A). Quantitative image analysis (Fig. 2C) showed that the amount of active RNA polymerase II fell to ∼20% of control levels by 48 hours after transfection, with a progressive loss of active RNA polymerase II seen between 12 and 48 hours following transfection (supplementary material Fig. S1). Cells with reduced expression of lamin A/C (Fig. 2A) showed no reduction in active RNA polymerase II. In parallel samples, RNAi was also performed with co-transfection of a vector that expresses an RNAi immune variant of human lamin B1 gene (Materials and Methods), to counter the possibility of protocol-dependent artefacts such as off-target effects. Cells expressing this modified LMNB1 showed no loss of active RNA polymerase II (Fig. 2A).

To confirm reduced transcription, we also visualised nascent RNA directly after incorporating BrUTP (Fig. 2B). Following RNAi treatment, cells were permeabilised and nascent centres pulse-labelled with BrUTP in vitro. Analysis of cells with reduced lamin B1 showed that the activity of RNA polymerase II was particularly sensitive to levels of lamin B1 expression, with pre-mRNA synthesis falling to as little as 10% of control levels in cells, with <20% of lamin B1 found in controls (Fig. 2B,E). Cells with reduced lamin B1 expression showed a clear loss in pre-mRNA synthesis, despite the broad scatter in control cells. This cell-to-cell variability in RNA synthesis and polymerised lamin B1 result from natural variabilities in HeLa cells, such as variable DNA content and changes during the cell cycle – as the number of active genes doubles as the genome is replicated during S phase. The relationship between assembled lamin B1 and pre-mRNA synthesis suggests a direct link between transcription and the integrity of lamin-B1-containing nuclear structures.

Cells with lamin B1 expression of 10-20% of controls also showed declining levels of pre-rRNA synthesis (Fig. 2B), which eventually resulted in the collapse of nascent pre-ribosomal transcripts into a small number of large nucleolar foci (supplementary material Fig. S2). These progressive changes are emphasised by the relative activities of the two classes of nascent transcript (Fig. 2D). As expected, the ratio of pre-rRNA/pre-mRNA synthesis in control cells was ∼1.2; ∼55% of nascent synthesis is nucleolar in HeLa cells. As lamin B1 expression fell, an initial fall of pre-mRNA synthesis and subsequent decay of pre-rRNA synthesis resulted in a ∼fourfold increase in the synthetic ratio. As predicted from the presence of active RNA polymerase II (Fig. 2A), depletion of laminA/C had no significant effect on the extent of mRNA synthesis (Fig. 2F and supplementary material Fig. S3).

With vector-based RNAi, reporter gene expression provides an unambiguous assignment of treated cells, and untreated cells in the sample provide perfect experimental controls. However, because such mixed cell populations are not suitable for biochemical analysis, we also performed RNAi using direct transfection of siRNA under conditions in which target gene expression is reduced in >90% of cells (supplementary material Fig. S4). Importantly, this confirms that the average depletion of lamin proteins is very similar when quantitative immunofluorescence and immunoblotting are used (Figs 1, 2 and supplementary material Fig. S2); slight differences reflect the natural cell-to-cell variability of the vector-based approach. Using siRNA, a clear and progressive decrease in the active form of RNA polymerase II, in transcription factories, was seen in cells with reduced lamin B1 expression, whereas depletion of lamin A/C had no obvious effect.

We next analysed how conditions of lamin B1 depletion, which compromise RNA synthesis, influenced markers of global nuclear organisation, using well-characterised structural features of nucleoli (Raska et al., 2006) and nuclear speckles (Lamond and Spector, 2003). The specific purpose of these experiments was to map the progressive changes in nuclear organisation that result from depletion of lamin protein expression. Analysis of the timing of specific changes would then give information about their temporal relationship, and hopefully reveal how different defects were causally related.

In these experiments, chromatin organisation was monitored by co-expression of DsRed-histone H2B during lamin depletion – note that in these experiments the DsRed-histone is expressed and assembled into chromatin before cellular defects arise. During interphase (Fig. 3A), the majority (∼80%) of HeLa cells contained 1-3 roughly spherical nucleoli, usually with diameters of 2-4 μm. Typically, each nucleolus had 10-20 active centres of ribosomal RNA synthesis, which are rich in the pre-rRNA processing protein fibrillarin. Fibrillarin staining is used to monitor nucleolar activity as the complexity of the fibrillarin-stained active centres reflects the extent of rRNA synthesis – cells with active rRNA transcription have multiple nucleoli with numerous small active sites, whereas fewer much larger sites are seen when transcription falls. About 15% of proliferating cells had extended, ovoid-shaped nucleoli, which often touched the nuclear periphery. Very few cells contained large peri-nucleolar aggregates of fibrillarin or fibrillarin dispersed throughout the cytoplasm (Fig. 3A). Depletion of lamin B1 correlated with a >2.5fold fall in the most prominent interphase organisation and concomitant increase in the normally minor organisational patterns (Fig. 3A, histogram). These changes correspond with structural changes seen in sites of pre-rRNA labelling when RNA polymerase I activity is inhibited as a result of >90% reduction in expression of lamin B1 (Fig. 2 and supplementary material Fig. S2).

Nuclear speckles are nuclear domains of about 1 μm diameter that are rich in proteins involved in pre-mRNA metabolism, such as RNA splicing. The proteins within speckles are known to be highly dynamic, and this is reflected in the texture of the sites, which is irregular and dispersed in transcriptionally active cells and spherical and condensed when transcription is suppressed (Lamond and Spector, 2003). The majority (∼80%) of interphase HeLa cells contained 10-20 irregular nuclear speckles, which were intensely stained by indirect imunolabelling using antibodies to the splicing protein SC-35 (Fig. 3B). Almost all the other interphase cells contained speckles with rounded and condensed appearance. Depletion of lamin B1 correlated with a >threefold fall in the prominent speckle pattern and increase in previously minor patterns, with the condensed speckle morphology increased from ∼20 to ∼60% and a >fivefold increase in cells with diffuse nuclear and cytoplasmic labelling (Fig. 3B). These structural changes in the morphology of the nuclear speckle compartment are consistent first with a loss of the normal speckle structure, which mimics that seen when pre-mRNA synthesis is inhibited, and second with a general loss of the nuclear targets that normally define the spatial organisation of nuclear speckles in these cells.

During the course of these experiments we noticed that at late time points (48-72 hours) progressively more cells developed atypical patterns of chromatin distribution (Fig. 3C). In nuclei of control cells, expression of DsRed-histone H2B defined the classical amorphous chromatin distribution, with regions of dense and more dispersed chromatin surrounding chromatin-depleted nuclear regions, which contain nucleoli and the interchromatin constituents, such as nuclear speckles. This fenestrated appearance was maintained following depletion of lamin A/C. However, in some cells loss of lamin B1 correlated with a profound alteration in nuclear organisation, with clear condensation of the dispersed chromatin and expansion of the inter-chromatin compartment (Fig. 3C).

Structural changes seen in lamin-B1-depleted cells at late time points have some features in common with nuclear changes that occur during programmed cell death, although we also note that some changes seen in nuclear organisation, particularly within the nucleolar compartment, are never seen during apoptosis. In view of this, we next measured apoptosis in lamin-depleted cells using Annexin V binding (Fig. 4). Treating HeLa cells for only 4 hours with staurosporin provided a positive control, with 100% of cells showing robust Annexin V binding (Fig. 4A). In negative controls and cells treated to deplete lamin A/C ∼2% of cells were Annexin-V-positive at 48 hours post-transfection and ∼10% at 96 hours post-transfection (Fig. 4B). Notably, cells transfected with the lamin B1 RNAi vector showed larger populations of Annexin-V-positive cells – with ∼5 and 50% positive cells at 48 and 96 hours post-transfection, respectively (Fig. 4B). Although this shows that depletion of lamin B1 effects cell viability, consistent with earlier observations (Harborth et al., 2001), the temporal changes in nuclear organisation show that apoptosis is a consequence of events that result from depletion of lamin B1 and is inconsistent with the suggestion that effects seen in lamin-B1-depleted cells arise because of some indirect mechanism that leads to apoptosis.

The status of p53 in HeLa cells is partially compromised by expression of the HPV E6/7 proteins. To rule out any affects this might have on apoptosis, we also performed lamin-depletion experiments in primary lung fibroblast cells (MRC5) and colorectal cells (HTC116) with normal or null p53 status. In all cases, analysis of nascent RNA showed that these three cell types had the same depletion phenotypes as HeLa cells when siRNAs were used to inhibit lamin protein expression (supplementary material Figs S5 and S6). Additionally, the same results were seen when depletion experiments were performed in cells treated with the apoptosis inhibitor Z-VAD-FMK (supplementary material Fig. S6).

Changes in chromatin organisation seen during lamin B1 depletion (Fig. 3C) prompted us to evaluate the organisation of CTs in these cells. We first looked at a specific case using fluorescence in situ hybridisation (FISH) to chromosome 19, which is normally located within the nuclear interior (Croft et al., 1999). Intriguingly, in lamin-B1-depleted cells this CT was seen to be localised at the nuclear periphery. Notably, the nuclei of lamin-B1-depleted cells were frequently disfigured during the FISH protocol so that the resulting signals were more dispersed, and apparently weaker, than in controls – consistent with depletion of lamin B1 compromising global nuclear structure.

As the texture of DNA foci within CTs is lost during FISH, we next imaged individual territories after labelling with BrdU. In this experiment, cell populations were grown in medium supplemented with BrdU so that all DNA was labelled and then grown for ∼5 days to reveal individual CTs. Using high-resolution 3D imaging, BrdU-labelled territories displayed a complex structure of finely dispersed DNA foci – a typical example is shown in Fig. 5B (control). Following depletion of lamin B1, a general loss of the surface texture of CTs was seen, with individual DNA foci appearing to become more condensed and tightly packed (Fig. 5B). At the same time, this loss of structure correlated with an increased tendency to associate with the nuclear periphery. Finally, we confirmed these observations in living cells, using CTs labelled with Alexa Fluor 488 dUTP (Fig. 5C). Strikingly, in the most extreme cases, the Alexa-labelled CTs appeared to be smeared against the nuclear rim. In more typical examples, the general shape of CTs was preserved but an expanded interchromatin domain correlated with CTs concentrated towards the nuclear periphery (Fig. 5D). These observations suggest that interactions that define the natural distribution of CTs are degraded when the expression of lamin B1 is reduced.

As lamin depletion leaves a residual lamina with compromised structural properties (Fig. 5), we next evaluated the stability of the nucleoskeleton in cells with depleted lamin expression. In proliferating mammalian cells, soluble and lamina-associated subunits have t1/2 for exchange of >2 hours (Moir et al., 2000). Under growth conditions used here, the t1/2 for recovery of GFP-lamin A in HeLa was ∼60 minutes (Fig. 6). When lamin A and B1 were depleted, the t1/2 for recovery of GFP-lamin B1 or A of the residual lamin fell to ∼20 and 10 minutes, respectively. These data confirm that although the lamina is a stable structure, the natural stability is dependent on a balance of lamin proteins that contribute to the structural network of the intact lamina.

These changes in the behaviour of the residual lamina raise obvious questions about its structure and any complementation effects that are seen in lamin-depleted cells. In agreement with the analysis of Lammerding et al. (Lammerding et al., 2006), we found that cells with depleted lamin protein expression showed very limited upregulation of other lamin proteins. For example, using vector-based depletion of lamin A/C the distribution of lamin B1 was clearly normal using indirect immunofluorescence (supplementary material Fig. S2). Quantitative analysis, using depletion with siRNAs, showed that in all cases expression of the unaffected lamin proteins was between 100 and 120% of the normal levels (supplementary material Figs S4 and S5).

Lamin-dependent nuclear structures are known to play a key role in nuclear mechanics and shape determination (Broers et al., 2006; Lammerding et al., 2006). In addition, the nucleoskeleton has also been implicated in the organisation of chromatin loops (Jackson, 1991). With this in mind, we next evaluated the structure of DNA loops in nucleoids prepared from cells with reduced lamin gene expression (Fig. 7). Lamin gene expression was reduced using vector-based RNAi and co-expression of reporter genes that mark the nuclear rim (Fig. 7) and nucleoids treated with ethidium bromide to monitor the structure of expanded nuclear halos (Fig. 7). Notably, the DNA halos from lamin-A/C-depleted nuclei were indistinguishable from untreated cells, whereas cells with reduced expression of lamin B1 showed general loss of structure of the residual nucleoid cage and corresponding deformities in the structure of the DNA halo (Fig. 7A). Specifically, in lamin-B1-depleted cells, DNA halos were more expanded than in controls but also much more irregular; in control as well as lamin-A/C-depleted cells halos were generally spherical, whereas halos from lamin-B1-depleted cells were asymmetrical, with huge DNA bundles spreading many microns away from the nucleoid cage. This loss of general loop organisation was particularly evident during time-lapse analysis, where the majority of lamin-B1-depleted cells showed profound loss of nucleoid structure, often with huge plumes of DNA escaping from nucleoids under ultraviolet illumination (Fig. 7B). These observations confirm that the integrity of the lamin-B1-containing nucleoskeleton is a major determinant of the organisation of chromatin domains in HeLa cells.

In human cells, nuclear lamin proteins provide an intermediate filament network that maintains nuclear structure (Broers et al., 2006). This network may also support other organisational roles. For example, it is possible that by organising nuclear actin and myosin the lamin network will orchestrate spatial dynamics within the nucleus (Chuang et al., 2006). Nuclear lamin proteins have also been reported to influence different aspects of nuclear function (reviewed by Broers et al., 2006; Gruenbaum et al., 2003). However, if lamin proteins do regulate fundamental functions, it is surprising that viable cells can be isolated from mice with deficient LMNA (Sullivan et al., 1999) or LMNB1 (Vergnes et al., 2004) expression.

In the mouse models, deletion of lamin A/C expression – as a null phenotype – has similar properties to some classes of human laminopathies, with developmental defects particularly in muscle, heart and peripheral nerves, and the lamin-B1-defective mice die at birth with severe developmental phenotypes (Vergnes et al., 2004). Even so, the ability to establish cell lines (MEFs) from embryos of LMNB1 mutant mice implies that loss of normal lamin B1 function is not inevitably cell lethal. However, these observations on murine cells do contradict RNAi experiments with human cells, which suggest that B-type but not A-type lamins are required for cell proliferation, with prolonged reduction in expression of B-type lamins resulting in apoptosis (Harborth et al., 2001).

To explain this conflict, it might be argued that specific functions played by nuclear lamins are cell-context-specific. Even so, it is notable that the LMNB1 mouse was generated using a recombination strategy that left the promoter intact to express a fusion protein in which the 3′ half of the lamin B1 was replaced by the βgeo reporter (Vergnes et al., 2004). Despite the fact that the nuclear localisation signal was deleted in this construct, the altered lamin B1 in MEFs localises to the nucleus and concentrates at the nuclear periphery. This construction and the presence of a mutant form of lamin B1 in the residual lamin complicates any conclusions that might be made about the normal function of lamin B1. Clearly, LMNB1 MEFs have lamina with altered structural properties and compromised differentiation potential (Vergnes et al., 2004), which might also explain the observation that MEFs with altered lamin B1 expression display abnormalities in CT distribution (Malhas et al., 2007). However, these observations do not exclude the possibility that some lamin B1 functionality might be at least partially retained.

Adding to this controversy, a preliminary study from Goldman and colleagues (Spann et al., 2002) provides compelling data to link the organisation of the lamin network to RNA synthesis in somatic mammalian cells. Here, a dominant-negative knockout strategy – using N-terminally-deleted lamin A – was used to disrupt the lamin filaments. The loss of lamin A filaments was reported to correlate, qualitatively, with a decay in transcription by RNA polymerase II. However, the interpretation of this observation is complicated because the disruption also removes B-type lamins from the nuclear periphery to nuclear aggregates that sequester essential transcriptional factors, giving potentially profound bystander effects.

We wanted to confirm possible roles for nuclear lamin proteins in linking nuclear structure and function. To do this we used RNAi strategies that deplete individual lamin proteins without significantly altering the concentration of the remaining lamin proteins in the residual lamina. We describe a detailed analysis of structure-function relationships in four human cell lines. In HeLa cells, depletion of lamin A/C expression had no obvious phenotype, whereas depletion of the major B-type lamin – lamin B1 – correlated with a pronounced loss of transcriptional activity. The same transcription defects were seen in primary lung fibroblasts and colorectal cells with normal and null p53 status.

During lamin B1 depletion, levels of pre-mRNA synthesis fell concomitantly with loss of lamin B1 (Fig. 2); active RNA polymerase was lost (Fig. 2), although overall expression of RNA polymerase II changed little (supplementary material Fig. S5). The consequence of this was seen in global changes in nuclear organisation (Fig. 3). During the early stage of lamin B1 depletion, pre-rRNA synthesis in nucleoli was unaffected. However, at later stages, pre-rRNA synthesis was inhibited and characteristic changes in nucleolar morphology seen (Figs 2, 3 and supplementary material Fig. S2). This is consistent with a lamin-B1-dependent nuclear structure being required for assembly of active sites of RNA synthesis. By contrast, reducing the level of lamin A/C had no obvious phenotype under the routine knockdown conditions used. This dominant behaviour of lamin B1 is surprising, as embryonic mouse fibroblasts with no lamin B1 are viable in culture (Vergnes et al., 2004). A possible explanation for this is that the structural and functional properties of the nucleoskeleton are defined by its overall composition, so that differential expression of the B-type lamins in different cells and tissues will allow some functional complementation [see Lammerding et al. (Lammerding et al., 2006) for discussion of effects of complementation in cells from knockout mice]. Even so, in experiments performed here, altered expression of the remaining lamins was no more than 20% of the normal levels and therefore unlikely to significantly compensate for any defects seen. Additionally, in all four cells lines used, depletion of lamin B1 but not lamin B2 resulted in transcriptional inhibition.

Conditions that activate programmed cell death in human cells are extremely complex. In the nucleus, checkpoint pathways monitor DNA integrity and activate apoptosis if DNA damage accumulates (Roos and Kaina, 2006). Activation of apoptosis correlates with a number of diagnostic nuclear changes (Robertson et al., 2000), with many nuclear proteins, including both A- and B-type lamins, targeted for caspase-dependent degradation. Well-characterised nuclear changes then follow.

Depleting lamin B1 in HeLa cells leads to a progressive loss of RNA synthesis and eventually results in cell death (Fig. 4). However, this is a late phenotype, and we stress that all functional and structural observations (Figs 2, 3, 4, 5) detailed here were made using cells that were morphologically normal using bright-field microscopy and chromatin organisation. Changes in RNA synthesis were first evident within 12 hours of inducing lamin B1 depletion, and a robust and predictable programme of changes was seen between 18 and 48 hours (Figs 2, 3), at least 2 days before a significant proportion of depleted cells were positive for Annexin V binding (Fig. 4). The timing of onset of apoptosis is consistent with cell death arising as a result of defects in gene expression, so that cells become starved of components that are required to sustain essential cellular processes as levels of key transcripts fall.

This conclusion is supported by analysis of the effect of lamin B1 depletion on DNA synthesis. Although our detailed analysis of DNA synthesis (C.W.T., A.M.-M. and D.A.J., unpublished) falls outside the scope of this study, two key observations are worthy of comment. First, as for transcription, depletion of lamin B1 – but not lamin A/C or B2 – gives rise to a replication phenotype (48-60 hours post-RNAi), which results in a ∼50% decline of cells within S phase but increase in the proportion of mid/late S-phase cells. Second, time-lapse analysis shows that cells with reduced lamin B1 expression have frequent defects in the replication programme, including extended S-phase duration and abortion of synthesis before S phase is complete. It appears that S-phase-dependent defects in cells with compromised lamin B1 expression will be sufficient to activate replication checkpoints and induce apoptosis.

Mammalian nuclei are structured so that nuclear functions are performed within dedicated nuclear sites (Cook, 1999). Chromatin that interacts with these sites is locally structured, to give spatially defined CTs (Cremer and Cremer, 2001; Misteli, 2004), which occupy preferred nuclear positions that reflect transcriptional activity and differentiation status. However, as global molecular mechanisms that regulate the location of CTs have not been defined, the extent to which the position of a CT might influence any aspect of function remains a matter for debate. Even so, recent studies have suggested that the organisation of CTs could allow genes that interact with similar regulatory transcription factors to be expressed within common transcription factories (Chakalova et al., 2005; Spilianakis and Flavell, 2006), perhaps as independently regulated gene networks.

Our analysis shows that nuclear function, and specifically RNA transcription, is a key determinant of chromosome position. Depletion of lamin B1 and corresponding loss of RNA synthesis correlates with redistribution of CTs towards the nuclear periphery (Fig. 5), mimicking changes seen in cells treated with inhibitors of RNA synthesis (Croft et al., 1999). In addition, changes described herein are reminiscent of structural changes described by Albiez et al. (Albiez et al., 2006) in cells grown under conditions of differing osmolarity. In this study, transferring cells to hypertonic media induced a dramatic but reversible collapse of chromatin and corresponding expansion of the inter-chromatin domain. Hence, the correlation between chromatin organisation and function suggests that chromosome position will normally reflect the balance of interactions at the nuclear lamina and internal nucleoskeleton; a balance that is altered in lamin-B1-depleted cells.

Our observation that loss of lamin-B1-dependent structures in HeLa leads to reduced RNA synthesis implies that lamin-based structures function at transcription sites throughout the nuclear interior. However, although there is some evidence for an internal nucleoskeleton containing B-type lamins (Hozak et al., 1995; Barboro et al., 2002), the significance of this is difficult to assess as so little is known about the structure of the peripheral and internal lamin networks (Gruenbaum et al., 2003; Broers et al., 2006).

Following lamin depletion, the dynamic behaviour of the residual proteins might hint at the structure of the network. Proteins of the nuclear lamina are known to exchange slowly. In proliferating cells, ectopically expressed fluorescent A- or B-type lamins exchanged with a t1/2 in the range 1.5-3 hours (Moir et al., 2000; Gilchrist et al., 2004), although mutants with substantially increased mobility have been described (Gilchrist et al., 2004). In our hands, GFP-lamin A/C expressed ectopically in HeLa cells incorporated into the endogenous lamina and exchanged with t1/2 of ∼60 minutes (Fig. 6). A clear increase in the rate of exchange was seen when either A- or B-type lamins were depleted from the endogenous lamina, implying that both A- and B-type lamins contribute to the stability of the endogenous network.

Alteration in the structure of the lamin network was evident in nucleoids prepared from lamin-B1-depleted cells, where loss of stable DNA loops suggests that lamin-B1-dependent interactions are a key determinant of chromatin loop organisation. Once again, these experiments emphasise a clear difference in the roles played by A- and B-type lamins. While both classes of lamin protein appear to contribute to the structure of an integrated lamin protein network, the major roles of the two classes are distinct. Most notably, lamin B1 is essential to maintain natural levels of gene expression, whereas loss of lamin A/C has no obvious functional implications. The major role of A-type lamins, by contrast, appears to be linked to the structural maintenance of the nuclear compartment and regulation of nuclear mechanics (Lammerding et al., 2006).

In conclusion, we show here that lamin B1 is an essential component of a nuclear structure that is required to maintain the natural activity of RNA synthesis in human cells. When lamin B1 is depleted, the decay of RNA synthesis also correlates with structural alterations in major nuclear compartments, including CTs. We propose that the molecular network that connects a lamin-B1-dependent nucleoskeleton, nucleoskeleton-associated transcription factories and chromatin templates, which interact with the active sites, provides a general molecular mechanism to explain how nuclear structure and function are linked in higher eukaryotes.

HeLa cells (S3) and p53 normal and null human colorectal cells (HCT116) were grown in DMEM supplemented with antibiotics and 10% FBS. Primary human lung fibroblast cells (MRC5) were grown in MEM supplemented with antibiotics and 10% FBS.

For each target gene, three expressed target sequences were selected (Harborth et al., 2001) (www.ambion.com). RNAi was performed using double-stranded RNA (Ambion) or short hairpin RNAs expressed from pSuper. RNAi by duplex siRNAs was validated by western blotting and immunolabelling and the most effective targets cloned in to the RNAi pSilencer™ expression vectors, pSECpuro and pSEChygro (Ambion).

The following target sequences were chosen for detailed study: Lamin A/C, CTGGACTTCCAGAAGAACA; Lamin B1, AGAGTCTAGAGCATGTTTG; Lamin B2, GCTGAGCTCTGACCAGAAC; B1-Scrambled, GTCGATAACGTGCTACTAT.

RNAi was performed on HeLa cells at 75% confluency in six-well plates. For each well, 1.5 μg plasmid DNA in 100 μl Optim medium without serum was mixed with 12 μl Polyfect (Qiagen) and incubated for 5 minutes (20°C), after which time 0.6 ml fresh medium was added. Transfection mix was added to wells with 2 ml fresh medium (37°C). After 24 hours, cells were seeded on to 13 mm coverslips coated with poly-L-lysine. RNAi vectors were co-expressed with appropriate reporter genes, for identification of the transfected cells. The reporters used were: YFP-tubulin (Clontech); DsRed- and GFP-lamin A/C and B1, DsRed-histone H2B. Plasmids with LMNA and LMNB1 were kind gifts from Chris Hutchison (Durham, UK) and Jan Ellenberg (EMBL, Germany), respectively. Cells were monitored for up to 96 hours following transfection and changes analysed by quantitative microscopy. Typically, 100-200 cells were analysed for each sample.

Rescue of lamin B1 knockdown was performed using point mutations generated by site-specific mutagenesis by overlap extension. The third nucleotides of each codon in the target sequence of lamin B1 cDNA were altered, to preserve the natural protein sequence. Changes were made using a three-step PCR protocol with Roche Expand High Fidelity PCR Kit and a 3′ FLAG tag incorporated. Modified cDNA was cloned into PCEP4 vector (Invitrogen) to give PCEP4-lamin B1/RNAi^–, with the modified lamin B1 cDNA expressed from the CMV promoter. Following transfection, expression of the RNAi immune lamin B1 protein was monitored by immunoblotting, using both anti-lamin B1 (Zymed) and anti-FLAG (Sigma). Lamin A/C siRNA (sc-35776) and Lamin B1 siRNA (sc-29386) were purchased from Santa Cruz. Lamin B2 siRNA and Scrambled siRNA were from Sigma-Genosys.

Transcription factories were labelled by indirect immunofluorescence of the active form of RNA polymerase II. Nascent transcripts were visualised following incorporation of BrUTP in vitro (Jackson et al., 1993). Cells on coverslips were permeabilised using Triton X-100 (0.1% for 1 minute at 20°C) and transcription performed in buffer supplemented with 500 μM BrUTP, GTP, CTP and 2 mM ATP (4 minutes at 37°C).

For live imaging of DNA foci, cells were labelled with either Cy3-dUTP or Alexa Fluor 488 dUTP for visualisation of DNA foci and CTs in living cells. Cells were seeded on to 13 mm coverslips coated with poly-L-lysine and transfection performed at 50% confluency. For each coverslip, 3 μl Fugene6 was mixed with 1 μl dUTP analogue (10 minutes; 0°C) and 17 μl PBS (further 5 minutes; 0°C) before applying to cells (10 minutes; 0°C). Coverslips were rinsed and returned to fresh medium and seeded as required to visualise individual labelled CTs 4-7 days later.

Live cell imaging was performed using a Zeiss LSM 510 META microscope. Fluorescence recovery after photobleaching (FRAP) was performed as follows. Cells expressing fluorescent lamin proteins were selected using laser scanning at low laser power; RNAi-treated and control cells with similar, low level, expression were selected for comparison. A pre-bleach image was recorded (1-2% 488 laser line; two scans for 1 μm confocal section) and a selected area then bleached using high laser power (full power for 488 and 543 lines). FRAP was recorded using time-lapse images of the original section and quantitative image analysis of the bleach zone performed using the Zeiss software.

For live imaging of CTs, labelled cells were grown on 35 mm Petri dishes with 13 mm glass inserts (MatTek Corporation) in medium without Phenol Red. Using a 1.4NA 100× Plan-Apo DIC objective, cells were selected under bright-field illumination and then 3D image stacks taken under live scanning mode using minimum laser power.

Immunofluorescence was performed under standard conditions (Pombo et al., 1999). Briefly, samples were rinsed in PBS, fixed with 4% paraformaldehyde (Electron Microscopy Sciences; 15 minutes on ice), blocked and immunolabelled using the reagents below. Slides were then washed, DNA stained with 5 μg/ml Hoechst 33258 (Sigma) for 10 minutes, and mounted with Vectashield (Vector Laboratories). The lamin antibodies used throughout this study were from Santa Cruz (Lamin A/C sc-7292, Lamin B2 sc-56147) and Abcam (lamin B1 ab6048).

Active RNA polymerase II in transcription factories – mouse anti-human RNA polymerase II antibody (Covance; H5 antibody – MMS-129R; 1/1000; 15 hours; 4°C) and Cy3-donkey anti-mouse IgG (1/1000; 1 hour; 20°C).

BrdU detection in nascent RNA – sheep anti-BrdU antibody (Biodesign; M20105S; 1/1000; 1 hour; 20°C) followed by Cy3- or Alexa-Fluor-488-conjugated donkey anti-sheep IgG (1/1000; 30 minutes; 20°C).

BrdU detection in DNA – after fixation, slides were rinsed twice in ddH₂O and then incubated with HCl (2.5 M; 1 hour; 20°C) to denature DNA. Slides were then washed three times in PBS, blocked and cells immunolabelled using sheep anti-BrdU antibody (Biodesign; M20105S; 1/1000; 1 hour; 20°C) and Cy3-donkey anti-sheep IgG (1/1000; 1 hour; 20°C).

Nucleoli – mouse monoclonal antibody to fibrillarin (Cytoskeleton – AFB01; 1/200; 15 hours; 4°C) and FITC-conjugated donkey anti-mouse IgG (1:1000; 1 hour; 20°C).

Nuclear speckles – mouse monoclonal antibody to SC-35 (Sigma – S4045; 1/500; 15 hours; 4°C) and FITC-conjugated donkey anti-mouse IgG (1:1000 dilution; 1 hour; 20°C).

Cells undergoing apoptosis were monitored by Annexin V binding. At appropriate times after transfection (typically 48 and 96 hours) cells were rinsed in PBS, fixed and stained with Annexin V conjugated directly with Alexa 568 (Roche – diluted 1:50 in incubation buffer) as detailed by the manufacturer. Samples were washed extensively in PBS and nuclei counterstained with Hoechst 33258.

Caspase inhibition: 50 μM of Z-VAD-FMK (MBL International Corporation) was added into the medium the day before transfection.

Control and RNAi-treated HeLa cells were grown on poly-L-lysine-coated slides and fixed with 4% paraformaldehyde (10 minutes at 20°C). The distribution of chromosome 19 was determined using WCP 19 labelled with SpectrumGreen™ (Vysis). FISH was performed as recommended by the manufacturer, with hybridisation for 18 hours at 37°C.

To prepare nucleoids, HeLa cells were suspended in ice-cold PBS without Ca²⁺ and Mg²⁺ and aliquots of 50 μl containing 3.5×10⁵ cells were gently mixed with 150 μl lysis solution containing 2.6 M NaCl, 1.3 mM EDTA, 2.6 mM Tris, 0.6% Triton X-100, pH 8.0. After 20 min at 4°C, the mixture was overlaid on sucrose step gradients containing 0.2 ml 30% sucrose under 0.6 ml 15% sucrose. Both sucrose layers contained 2.0 M NaCl, 1.0 mM EDTA, 10 mM Tris, pH 8.0. The gradients were spun at 4°C in a microfuge for 4 minutes at 9000 g. Nucleoids (in ∼250 μl buffer) were recovered, 50 μl stained with ethidium bromide to final concentration of 5 μg/ml and examined by fluorescence microscopy. DNA halos expand continuously when ethidium-bromide-stained nucleoids are exposed to ultraviolet light. To monitor the structure of the halos, it is therefore essential to use identical exposure times; in Fig. 7, imaging was performed at 20 seconds.

Whole cell extracts (10 ng whole cell protein/lane) were used in all western blots. Actin anti-body was from Sigma-Aldrich (A1978). Goat anti-mouse HRP (172-1011) and goat anti-rabbit HRP (172-1019) were from Bio-Rad. Anti-RNA polymerase II, clone CTD4H8 was from Upstate. ECL Western Blot Analysis System (RPN2109) was from GE Healthcare.

3D images were generated from confocal Z series using Imaris software (Bitaplane, Switzerland). Centre of mass for chromosome territories (Fig. 5) was determined and assigned to specific nuclear sectors as described by Croft et al. (Croft et al., 1999). Statistical analysis was performed by linear regression and Student's t test using Microsoft Excel.

We thank Chris Hutchison and Jan Ellenberg for providing reagents that were used in this study. This work was undertaken with financial support from BBSRC (C.W.T., A.M.-M., C.M., S.C., S.W., D.A.J.) and European Commission (K.Z., D.F., D.A.J.).