The induction of a nucleoplasmic reticulum by prelamin A accumulation requires CTP:phosphocholine cytidylyltransferase-α

The nuclear lamina is a proteinaceous mesh that underlies the nuclear envelope. It is involved in a diverse range of functions such as maintaining the structural integrity of the nucleus, regulating DNA replication, controlling gene expression and organising the three dimensional space of the genome (Goldman et al., 2002; Malhas et al., 2007; Malhas et al., 2009; Verstraeten et al., 2007). The lamina is composed of type V intermediate filament proteins called lamins encoded by two gene classes: A-type lamins and B-type lamins. Lamin A and C are encoded by the LMNA gene and are translated from distinct transcripts generated by differential splicing (Fisher et al., 1986; Furukawa et al., 1994; Machiels et al., 1996; McKeon et al., 1986). Lamin A contains a C-terminal CAAX motif that is processed to generate prelamin A, a 74 kDa precursor containing a terminal farnesylcysteine (Firmbach-Kraft and Stick, 1993; Hofemeister et al., 2000; Holtz et al., 1989; Kitten and Nigg, 1991). This precursor undergoes a final proteolytic maturation that ultimately removes the terminal 15 amino acids, including the farnesylated cysteine, producing the 72 kDa mature lamin A (Weber et al., 1989).

Mutations in genes encoding components of the nuclear lamina lead to a group of diseases called the laminopathies (Worman et al., 2010). One of the best characterised is the LMNA gene, which has many mutations associated with it and a diverse range of signs and symptoms that are exhibited depending on where the mutation resides. A single base mutation towards the C-terminus in LMNA causes the activation of a cryptic splice site, generating a protein called progerin, which lacks fifty internal amino acids (Eriksson et al., 2003). The string of amino acids lost includes the cleavage site for the zinc metalloendopeptidase FACE1 (ZMPSTE24 in mice) resulting in the internally truncated protein retaining the farnesylated C-terminus that is normally removed by this enzyme. This leads to the rare accelerated ageing disorder called Hutchinson–Gilford progeria syndrome (HGPS), with the affected individual suffering from lipodystrophy, cardiovascular disease and osteoporosis before death occurs during the teenage years (Merideth et al., 2008).

The consequences of accumulation of farnesylated prelamin A, whether as a result of drug treatment or mutation, has sparked much interest in several areas of cell biology including HIV therapy (Capeau et al., 2006; Capeau et al., 2005; Caron et al., 2007; Clarke, 2007; Coffinier et al., 2007; Coffinier et al., 2008; Goulbourne and Vaux, 2010; Hudon et al., 2008; Rudich et al., 2005; Saillan-Barreau et al., 2008) and the normal ageing process (Ragnauth et al., 2010; Scaffidi et al., 2005; Scaffidi and Misteli, 2006; Scaffidi and Misteli, 2008). After farnesylated prelamin A accumulates the nucleus becomes dysmorphic with a highly convoluted nuclear envelope with many nuclear invaginations. However, the relationship between these morphological changes and the increasingly well-recognised nuclear invaginations observed in different cell types and during various stages of cellular development remains obscure (Fricker et al., 1997a; Fricker et al., 1997b; Malhas et al., 2011). Such structures are believed to increase the interface between nucleoplasm and cytoplasm to improve communication between the two environments as well as add structural support. It has also been reported that this nucleoplasmic reticulum (NR) regulates calcium signals in localised sub nuclear regions, enabling control of gene transcription and cell growth (Chamero et al., 2008; Collado-Hilly et al., 2010; Nalaskowski et al., 2011). NR development seems dependent on the enzyme CTP:phosphocholine-cytidylyltransferase-α (CCT-α), which is responsible for phosphatidylcholine synthesis for the biosynthesis and curvature of membranes (Gehrig et al., 2008; Gehrig et al., 2009; Lagace and Ridgway, 2005). The retention of the farnesylated tail in prelamin A is believed to enhance these highly dysmorphic shapes in cell nuclei by increasing stress on the nuclear membrane (McClintock et al., 2006; Roblek et al., 2010).

The study reported here investigated the effects of acute farnesylated prelamin A accumulation on cells, by assessing the morphology and organisation of the nucleus. It was found that farnesylated prelamin A accumulated in cells treated for 48 hours with Zmpste24 short interfering RNA (siRNA), resulting in dysmorphic nuclei containing a complex NR. Prelamin A, lamin B1 and calreticulin colocalise in the NR, which can form in interphase cells without an intervening mitosis. Treatment of cells accumulating farnesylated prelamin A with an inhibitor of CCT-α, leads to a reduction in NR although the nuclear phenotype is still dysmorphic compared with controls.

Further assessment of the morphology of treated nuclei at the electron microscope (EM) level revealed that the NR in these cells consisted of both double membrane invaginations of the nuclear envelope and invaginations of just the inner nuclear membrane (INM). Furthermore, these structures contained regular fenestrations in mouse embryonic fibroblasts. EM tomography revealed that these fenestrations had dimensions similar to nuclear pore complexes (NPC). Immunolabelling and examination by light and electron microscopy confirmed the presence of Nup153 and monoclonal antibody (mAb)414 staining, which recognises the conserved domain FXFG, at these fenestrations. After prelamin A accumulation, there was a reduction in NPC number on the nuclear surface. Taken together these results show that accumulation of prelamin A results in two forms of NR developing, with many nuclear pores associated with them, and that development of the NR occurs during interphase and requires both CCT-α and prelamin A.

Upon transfection of MEFs with Zmpste24 siRNA the abundance of ZMPSTE24 enzyme was reduced at 48 hours (Fig. 1a). At earlier times of 12 and 24 hours post-transfection there was no noticeable reduction in abundance. The control siRNA produced no alteration of ZMPSTE24 protein levels over the three time points (Fig. 1a). Additionally, using immunofluorescence prelamin A could only be detected after 48 hours transfection (Fig. 1b; see also supplementary material Fig. S1 for similar results with a second independent ZmpSte24 siRNA) and this corresponded to the development of an NR that could also be visualised at the EM level (Fig. 1c). Additional experiments examined in more detail the situation at the 48 hour time point, and found that ZMPSTE24 enzyme levels were reduced, prelamin A was accumulating and cells exhibited a dysmorphic nuclear morphology. Farnesylated prelamin A is the only known substrate for the enzyme (Bergo et al., 2002; Levy et al., 2005; Pendas et al., 2002) so reduced expression of the ZMPSTE24 endoprotease will result in an increased abundance of farnesylated prelamin A accumulation.

On western blots, samples taken at 48 hours post-transfection had a band with a molecular mass that was higher than lamin A, which is indicative of the accumulation of the unprocessed precursor (Fig. 2a). This was further supported by detection of a signal using a prelamin-A-specific antibody, which was absent in control-siRNA-treated cell lysates (Fig. 2b). However, the prelamin-A-specific antibody detects amino acids 645–664 and so does not distinguish between farnesylated and non-farnesylated forms (Mattioli et al., 2011). In order to establish this, cells were incubated with a farnesol analogue, 8-anilinogeraniol, which is incorporated into anilinogeranyl diphosphate (AGPP). AGPP is used as a substrate by protein farnesyltransferase for the modification of CAAX-containing proteins, which includes prelamin A (Troutman et al., 2005). An antibody raised against the analogue detected prelamin A after Zmpste24 siRNA transfection, but this band was absent in control-transfected cells and reduced after Zmpste24 siRNA transfection in combination with farnesyl transferase inhibitor (FTI-277) treatment (Fig. 2c). The prelamin-A-specific antibody detected a band in both the Zmpste24-transfected and Zmpste24-transfected with FTI-277 cells (Fig. 2c).

The Zmpste24^−/− MEFs were shown to accumulate only prelamin A using a Lamin A/C antibody and no mature lamin A was detected (Fig. 2d). Co-transfection of Zmpste24 siRNA with Ccta (also known as Pcyt1a) siRNA into wild-type MEFs did not affect the quantity of accumulating prelamin A or lamin A levels (Fig. 2a). Transfection with Ccta siRNA alone did not cause unprocessed protein formation and did not affect the levels of lamin A (Fig. 2a). We confirmed that transfection of Ccta siRNA substantially reduced CCT-α protein, as indicated by an anti-CCT-α antibody western blot (Fig. 2e).

Farnesylated prelamin A appears to cause a dysmorphic nuclear phenotype with blebs and invaginations. Two hundred DAPI stained nuclei were analysed by thresholding and binarising the images, before determining individual nuclear perimeters and areas. These parameters were used in the contour ratio (CR) formula (4π×area/perimeter²) to establish how dysmorphic nuclei had become, with lower CR values denoting increased dysmorphia (Fig. 2f). Nuclei, after control siRNA transfection, had a CR of 0.79±0.04 (mean ± s.e.m.), and were indistinguishable from control untreated nuclei (0.78±0.05). However, Zmpste24 siRNA transfection lead to a significantly reduced CR of 0.63±0.14 and 0.65±0.12 for two separate ZmpSte24 siRNAs. Zmpste24^−/− nuclei were significantly more dysmorphic than the cells accumulating large amounts of farnesylated prelamin A with a CR of 0.52±0.14. This more extreme phenotype could be the result of long-term accumulation of farnesylated prelamin A in Zmpste24^−/− cells, which would have lacked all ZMPSTE24 expression for many generations. Transfection with Ccta siRNA did not result in a dysmorphic nuclear appearance, a result confirmed by a CR of 0.76±0.08, which is indistinguishable from that of the control. Co-transfection of this siRNA together with Zmpste24 siRNA generated a CR of 0.61±0.27, which was not significantly different to the other treatments that led to the accumulation of farnesylated prelamin A. This would suggest that the perimeter of the nucleus becomes highly convoluted in the presence of farnesylated prelamin A, and that this effect does not depend upon CCT-α. To confirm the role of farnesylated prelamin A, cells were transfected with Zmpste24 siRNA and additionally treated with an FTI. This caused a significant (P<0.01) increase in CR towards control levels. The CR after Zmpste24 siRNA plus DMSO vehicle control was unchanged compared with that after Zmpste24 siRNA transfection alone (Fig. 2g). The experiment was repeated using FTI treatment of ZmpSte24^−/− cells; once again the FTI caused a significant (P<0.01) reversion of the abnormal nuclear morphology towards the control level (Fig. 2h). This is presumably because the FTI removed the farnesylated tail from the accumulating prelamin A and therefore reducing its effect on the nuclear periphery, as has been previously reported (Fong et al., 2006; Toth et al., 2005).

Immunofluorescence microscopy was used to examine the organisation of the nuclear architecture in cells accumulating prelamin A (Fig. 3). The proteins stained were prelamin A, lamin B1 and calreticulin. Lamin B1 is a farnesylated protein, which has a high affinity for the inner nuclear membrane because of its farnesylated tail and serves as a good marker for the outline of the inner nuclear membrane. Calreticulin, an ER luminal protein that is also found within the perinuclear space, served as a good marker of nuclear membrane invagination (Echevarria et al., 2003).

There was little staining of prelamin A in untransfected nuclei, whereas lamin B1 often had a very strong association with the nuclear periphery with very little intranuclear signal (Fig. 3a). Calreticulin labelled ER throughout the cytoplasm and around the nuclear membrane with very little signal detected inside the nucleus (Fig. 3b). Zmpste24 siRNA transfection led to strong prelamin A staining at the periphery and within intranuclear structures that colocalised with lamin B1 and calreticulin (Fig. 3b). Zmpste24^−/− nuclei had a very similar appearance to the siRNA knockdown nuclei, containing an extensive NR that colocalised with prelamin A and lamin B1 (Fig. 3a,b). The staining observed could be as a result of the INM folding in, with just an ER luminal core or both the INM and ONM folding in with a cytoplasmic core surrounded by a ring of ER lumen. The fact that prelamin A also colocalises with these two proteins further supports the possibility that it is the farnesylated form of prelamin A that is accumulating, as the unfarnesylated form has been reported to form isolated intranuclear foci due to reduced affinity with the INM rather than NR (Toth et al., 2005).

Quantification of invaginations was achieved by acquiring z-stacks of nuclei and counting the number of calreticulin-positive intranuclear structures that were continuous with the nuclear membrane (Fig. 3c,d). It was found that in control-siRNA-transfected nuclei there was an average of two invaginations per nucleus. In Zmpste24-siRNA-transfected and Zmpste24^−/− nuclei there were significantly more (an average per nucleus of nine and twelve, respectively (P<0.01; Fig. 3d). Co-transfection with Zmpste24 and Ccta siRNA significantly reduced intranuclear staining of both lamins and calreticulin, a marker for NR, to three calreticulin-positive intranuclear signals per nucleus. However, even though the invagination count was reduced the nuclear morphology still had a dysmorphic phenotype, as supported by the CR (Fig. 2f).

Cells transfected with Zmpste24 siRNA were stained for prelamin A, CCT-α and concanavalin A (ConA; Fig. 4a). Con A is a lectin that binds to high mannose forms of N-linked oligosaccharide side chains on glycoproteins in the ER, inter-membrane space and NR interior (Fricker et al., 1997b). ConA revealed invaginations in the nucleus that colocalised with the prelamin A. Overall pixel intensity for CCT-α labelling within the nuclei of Zmpste24-siRNA-transfected cells revealed a shift to a greater abundance of higher intensity pixels (Fig. 4b). CCT-α appeared as intense foci along the invaginations, as has previously been observed (Gehrig et al., 2008). Quantification of these foci revealed that the abundance of CCT-α foci increased when nuclei acutely accumulated prelamin A (Fig. 4c).

Transmission electron microscopy (TEM) was used to assess the morphology of nuclei developing NR. Adherent monolayer cultures were extensively fixed and crosslinked in situ to ensure that nuclear morphology was preserved. Cells were released from the substrate without scraping, using propylene oxide to limit mechanical damage, and then pelleted before embedding in order to sample all orientations. This approach revealed two types of membranous structure that formed within the nucleus when farnesylated prelamin A accumulated. The nuclei of cells that were untransfected or transfected with a control siRNA were often ovoid with very few nuclear aberrations or intranuclear structures (Fig. 5a,b). However, after Zmpste24 siRNA transfection and in Zmpste24^−/− cells it became apparent that the nuclei contained many membrane-bound, flattened cisternal structures containing membrane-lined fenestrations (Fig. 5d,f). These structures were often perpendicular to the nuclear membrane and varied from several hundred nanometres in length up to the diameter of the nucleus. An electron dense signal was often found in close association with the membrane, which could be heterochromatin or the lamina scaffold. These cisternae were also frequently found to terminate close to nucleoli. Co-transfection with Zmpste24 and Ccta siRNA lead to a convoluted perimeter but no intranuclear membranous structures could be observed (Fig. 5e). The morphology of nuclei transfected with just the Ccta siRNA alone was similar to that observed in untreated and control siRNA transfected nuclei (Fig. 5c).

The fenestrations were regularly distributed along the length of these structures, and upon further examination, using dual-axis EM tomography (EMT) on 250 nm sections, it could be seen that these fenestrations were actually pores that had very similar dimensions to NPCs but had a nucleoplasmic to nucleoplasmic orientation (Fig. 6a). A three-dimensional model generated using IMOD (http://bio3d.colorado.edu/imod/) software revealed pores with a diameter of approximately 100 nm, which correlates well with current estimates of NPC diameters (Fig. 6a; supplementary material Movie 1). Serial section dual-axis electron tomography revealed that these structures are sheet-like and have many pore structures along them (Fig. 6b). EMT also revealed that these structures were in fact invaginations of the inner nuclear membrane (Fig. 6c), which explains the presence of an intranuclear calreticulin signal observed during immunofluorescence analysis. The region of continuity between the NE and NR can be very short, measuring only a few tens of nanometres.

Another type of invagination observed in nuclei accumulating prelamin A was an infolding of both the INM and ONM that contained a cytoplasmic core (Fig. 5d,f, blue boxes). These structures could either be single or branching and they protruded into the nucleus for anywhere from several micrometres up to its entire length. In cross sections these structures appeared to be double-walled vesicles within the nucleoplasm, with NPCs clearly visible bridging the nuclear membrane between the cytoplasmic core and nucleoplasm. There appeared to be two types of invaginations that develop after prelamin A accumulation, which are represented in the schematic in Fig. 5g.

In order to assess the relationship between these fenestrations and NPC, nuclei of MEFs were labelled for nucleoporin153 (Nup153), which is found within the nuclear pore basket, and with a monoclonal antibody mAb414 that identifies phenylalanine–glycine (FG) regions in the core nucleoporins (Fig. 7). A recent study investigated the effect on the distribution of Nup153 in cells expressing nuclear envelopathy-causing lamin A mutants and found it to be highly disorganised (Busch et al., 2009). The Nup153 protein is located in the basket of NPC, but reported to be a dynamic nucleoporin (Daigle et al., 2001). Thus, Nup153 alone is insufficient to define the presence of NPC. However, mAb414 recognises the conserved domain FXFG in the core nucleoporins and is stably associated with the NPC, making it an unambiguous marker of NPC structures.

It was found that after Zmpste24 siRNA transfection there was an intranuclear distribution of Nup153 and mAb414 labelling (Fig. 7a,b). The number of NPCs per 25 μm² of nuclear surface was determined in grazing sections of Nup153 and mAb414 labelled nuclei (Fig. 7c–e). It was clear that there was a significant (P<0.01) reduction in the number of NPCs on the surface of prelamin-A-accumulating nuclei compared with control nuclei. Additionally immunogold labelling revealed association of the Nup153 antibody (Busch et al., 2009) and mAb414 with the NR, supporting the notion that they contain at least certain key components of an NPC (Fig. 7d).

The mechanism by which NR is produced remains unknown. One model requires a mitosis accompanied by disordered NE assembly in the daughter nuclei; the presence of excess prelamin A might effect such a disorganisation. In order to investigate this question we used a cell cycle inhibitor and time-lapse phase-contrast imaging. MEFs were treated with hydroxyurea (HU) for 24 hours, which arrests the cells in the G0–G1 phase of the cell cycle (Fig. 8a). They were then either transfected with Zmpste24 siRNA or control siRNA before being fixed and labelled with an anti-lamin B1 antibody. The antibody was chosen as it was shown to colocalise with the NR marker calreticulin and was a reliable indicator of developing NR. To determine that cells were in the G1–G0 phase of the cell cycle DNA was stained and a cell cycle application in Metamorph software was used. It was found that the majority of the cells treated with HU did remain in the G0–G1 phase (Fig. 8a). The number of invaginations was determined by counting the number of intranuclear lamin B1 signals, with greater than six lamin-B1-positive labels being counted as a nucleus with invaginations (Fig. 8b). It was clear that the HU treatment had no effect on the number of invaginations (Fig. 8b,c).

To further confirm this observation, synchronised Zmpste24-siRNA-transfected MEFs were cultured on a gridded Mattek dish and were time-lapse imaged for 48 hours, with an image acquired every 15 minutes. After this time the cells were fixed and labelled with lamin B1. It was clear that a complex NR formed both in cells that divided and those that did not divide, confirming that NR formation is independent of mitosis (Fig. 8d).

The development of NR in a variety of different cell types is being increasingly reported and recognised (Malhas et al., 2011). The presence of NR has been documented under normal cellular conditions (Fricker et al., 1997b) and pathological states (Bussolati et al., 2008). However, the exact function of such structures remains speculative, and the current study is the first to thoroughly investigate these structures in the context of farnesylated prelamin A accumulation and determine a potential mechanism for their formation.

The observation of an irregular nuclear boundary developing after the accumulation of unprocessed lamin A has been frequently reported, particularly in the context of laminopathies (McClintock et al., 2006; Roblek et al., 2010), ageing (Scaffidi and Misteli, 2006) and HIV therapy regimens (Coffinier et al., 2007). The invagination of the nucleus along the periphery has been suggested to be due to prelamin A putting additional strain on the nuclear envelope (Prufert et al., 2004; Ralle et al., 2004). However, in healthy cells, CCT-α has been implicated in the production of NR by binding to the INM, leading to the generation of positive curvature and an infolding of the nuclear membrane (Gehrig et al., 2008; Gehrig et al., 2009; Gehrig and Ridgway, 2011).

A consequence of reduced expression of Zmpste24 is that there is an accumulation of prelamin A (Figs 1, 2, 3), which has retained 15 C-terminal amino acids including an isoprenylated terminal cysteine. The effects on the nucleus must be highly dominant as even a small quantity of the prelamin A is able to elicit this effect. The presence of this isoprenylated tail has been well documented to increase the affinity of the protein for the INM (Toth et al., 2005); furthermore, prevention of farnesylated prelamin A accumulation, using an FTI, displaces prelamin A to the nuclear interior as foci and restores the dysmorphic nuclear shape to normal (Fong et al., 2006; Toth et al., 2005). The CR of nuclei from cells co-transfected with Zmpste24 and Ccta siRNA revealed that the presence of farnesylated prelamin A and a reduced amount of CCT-α was enough to produce a convoluted nuclear boundary but not an NR (Figs 2, 3). The appearance of a convoluted nuclear boundary has also been reported in HeLa cells that were transfected with FACE1 siRNA (Gruber et al., 2005). That prelamin A was responsible for this morphological alteration was substantiated when lamin-A-deficient HeLa cells or cells that had lamin A acutely knocked down with siRNA, were subsequently transfected with FACE1 siRNA. This resulted in the number of nuclear aberrations being far lower than in wild-type cells transfected with FACE1 siRNA. Treatment with an FTI in our experiments supports our hypothesis that the farnesylated state of a prelamin is the causative factor behind the development of a more dysmorphic nuclear CR and NR.

However, immunofluorescence analysis revealed that the co-transfection of Zmpste24 and Ccta siRNAs led to a substantially reduced amount of NR, further supporting the notion that CCT-α is necessary for its formation (Fig. 3). The fact that the frequency of invaginations is substantially lower in cells that are not accumulating prelamin A suggests that the farnesylated prelamin A is acting in concert with CCT-α to generate the NR, and that the reduced presence of the CCT-α enzyme means the improperly processed lamin accumulates at the periphery and causes the membrane to become dysmorphic without the generation of NR.

The presence of the uncleaved isoprenylated C-terminal fragment is clearly important for initiating NR development, because of the lack of this structure in the untransfected cell nuclei. However, the fact that NR is not present after co-transfection with Zmpste24 and Ccta siRNAs would suggest that the isoprenylated tail is sufficient to generate a convoluted nuclear envelope but is insufficient for NR production. The farnesylated tail of prelamin A might act as a nucleation site for CCT-α activity. The initial positive curvature induced by the farnesylated tail interacting with the INM could attract CCT-α to this site.

CCT-α was observed to form foci along the NR that colocalised with prelamin A and there was an increased abundance of the nucleoplasmic foci after prelamin A accumulation (Fig. 4). These foci are produced when the inactive form of the CCT-α enzyme becoming activated and localises at the nuclear membrane where it might synthesise PC and extend the envelope into the nucleoplasm to form the NR (Gehrig et al., 2008; Gehrig et al., 2009; Lagace and Ridgway, 2005). However, it has recently been revealed that the catalytic activity of CCT-α is not necessarily required for the development of the NR (Gehrig and Ridgway, 2011; Lagace and Ridgway, 2005) and the M domain of CCT-α could act in a similar manner to the amphipathic ENTH domain found in epsin (Itoh and De Camilli, 2006). The epsin ENTH domain causes membrane deformation by inserting itself into the membrane by directly binding to phosphoinositide and pushing the phospholipid head groups apart, so reducing the energy available to curve the membrane, which is supported and maintained by a clathrin lattice (Ford et al., 2002; Itoh and De Camilli, 2006). CCT-α might act in a similar way, but instead of the clathrin lattice supporting invagination it is supported by the nuclear lamina, which is perhaps why the NR is ablated by lamin siRNA (Gehrig et al., 2008).

During live-cell imaging and in fixed cells it was apparent that, unlike prelamin A, CCT-α did not fully colocalise with the NR but instead was found along sections of it. This might be because the enzyme localises at points where the NR is expanding and further invagination is occurring. Cells treated with a fatty acid substrate develop CCT-α foci, which colocalise with the NR whose formation is dependent on the presence of the nuclear lamins (Gehrig et al., 2008). Recently it was revealed that cells expressing GFP–progerin also develop nuclear invaginations that colocalise with CCT-α (Gehrig and Ridgway, 2011), which suggests that a similar mechanism of nuclear membrane deformation occurs in HGPS patients. The common feature between the data presented here and those of Gehrig and Ridgway is the retention of the farnesylated tail on improperly processed lamin A. It will be interesting to investigate, in the context of progerin accumulation in cell nuclei, whether Ccta siRNA alters the distribution of CCT-α and the frequency of NR development.

Incomplete reformation of the NE after mitosis could occur if there is imperfect synchrony between chromosome decongestion and fusion of recruited NE fragments. This would result in chromatin-bound ER and/or NE cisternae becoming trapped in the interstitial space giving the impression of NR development (Gupton et al., 2006). However, it has been shown that NR can develop without mitosis (Fischer et al., 2003), persist throughout interphase (Fricker et al., 1997b; Olins and Olins, 2009) and specific cell types display heritable patterns (Fricker et al., 1997b), which strongly suggests physiological regulation. The data from the HU-treated cells indicated that an NR can develop when the cell is arrested at the G0–G1 phase of the cell cycle (Fig 8a) and transfected with Zmpste24 siRNA. This had very little effect on the number of invaginations compared with cells that were not treated with HU (Fig 8b,c). Further confirmation came from live-cell imaging of MEFs transfected with the Zmpste24 siRNA (Fig 8d), which revealed that NR develops within 24 hours in nuclei that have not been through mitosis. It could also be argued that NR is simply an artefact of cell culture in a two-dimensional environment and that they only provide structural support to a nucleus under compression and serve no functional role. However, the fact that NR has been reported in nuclei from cells in tissue samples (Fricker et al., 1997b; Langevin et al., 2010; Storch et al., 2007) and in three-dimensional cell culture (Chandramouly et al., 2007) confirms that they are a naturally occurring structures and not a side effect of culture conditions.

The NR that forms during prelamin A accumulation had two principal morphologies (Fig. 5). The first consists of an infolding of both the INM and ONM, with a cytoplasmic core that often contains nuclear pore complexes. The purpose of the double membrane invagination has been studied and it is proposed to act as a conduit for localised calcium release within the nucleus to regulate transcription of genes controlled by calcium (Bkaily et al., 2009; Bootman et al., 2009; Chamero et al., 2008; Collado-Hilly et al., 2010; Guatimosim et al., 2008; Lui et al., 1998; Marius et al., 2006; Nalaskowski et al., 2011). This might help the cell to adapt during the pathological accumulation of prelamin A. Because the cells were released from their growth substrate before embedding for TEM, the only information we have about orientation of NR in relation to the substrate comes from the light microscopy experiments and other studies (Malhas et al., 2011).

By shortening the distance between intranuclear expressing genes and the cytoplasm it could also facilitate message export and thus selectively increase translation to aid with the cellular shift to a pathological state. The fact that these invaginations contain nuclear pore complexes and have been reported to contain translational machinery (Paytubi et al., 2009) further supports the notion of increased mRNA export and generation of proteins from active genes deep within the nucleus. Treatment with a histone deacetylase inhibitor actually increases the frequency of invaginations, which might be due to the increased abundance of transcriptionally active genes (Galiova et al., 2008). NR invaginations are also frequently reported as terminating at nucleoli (Fricker et al., 1997b; Schneider et al., 2010), which again might serve as tight spatial coupling between a transcriptionally active compartment and a potential transport channel.

If the positive curvature is being applied to just the INM by CCT-α then why do double membrane invaginations form? One possible reason is that the linker of nucleoskeleton and cytoskeleton (LINC) complexes (Crisp et al., 2006; Houben et al., 2007), bridging the INM and ONM, hold the two membranes together, so when the INM invaginates it brings the ONM with it. In the case of just the INM invaginating, as was frequently observed, the accumulation of the farnesylated prelamin A could disrupt nuclear organisation at the INM and, therefore, affect the distribution of LINC complexes. Consequently, the uniform distance between the two nuclear membranes breaks down and infolding of just the INM occurs. This has been observed in siRNA targeting of nesprin isoforms, leading to the nuclear membranes separating and the INM invaginating (Zhang et al., 2007).

The presence of NPCs in the INM-only invaginations presents a topological conundrum, especially as these structures can form in an intact interphase nucleus. The structures observed cannot represent fragments of complete nuclear envelope trapped in the nucleoplasm after mitosis. Perhaps these features are not a complete pore but simply the nucleoplasmic half of the NPC, which has been torn apart during invagination of the INM. However, tomography reveals that these pores do have the symmetry of electron dense material either side of the pore, but whether they are complete complexes is yet to be determined. They certainly cannot be normally functioning pores as both their sides face the nucleoplasm. The fact that there are many fewer Nup153- and mAb414-positive structures at the nuclear surface and an increased abundance of these nucleoporins localised inside the nucleus (Busch et al., 2009), in association with the NR, suggests that they are being drawn in by the invaginations. Their function could be a transcriptional one, as components of the NPC are being increasingly implicated as necessary components for initiating transcription (Capelson et al., 2010; Ikegami and Lieb, 2010; Kalverda et al., 2010; Vaquerizas et al., 2010).

This study illustrates, for the first time, the connection between the acute accumulation of farnesylated prelamin A and involvement of CCT-α in generating an NR. The exact function of the NR in this pathological state remains speculative but it is likely to be an adaptive response of the cell to the toxic build up of farnesylated prelamin A.

Wild-type, Zmpste24^−/− MEFs (obtained from Stephen Young, UCLA) were cultured in DMEM supplemented with non-essential amino acids, 10% fetal calf serum, penicillin and streptomycin.

MEFs were grown to 80% confluency on glass bottomed 35 mm dishes (MatTek) and transfected with Zmpste24 siRNA (Qiagen, SI00241633, or Invitrogen Zmpste24 Stealth RNAi™ siRNA catalogue: MSS214049), control siRNA (Ambion, AM17012) or Ccta siRNA (Santa Cruz, sc-40395) using Lipofectamine 2000 (Invitrogen).

Wild-type MEFs were grown on 25 mm diameter glass coverslips before transfection. After 24 hours the medium was changed and cells were transfected again before being incubated for another 24 hours and then fixed in 4% paraformaldehyde. Cells were treated with 20 mM glycine in PBS, permeabilised with 0.4% Triton X-100 and blocked with 0.4% fish skin gelatin in PBS. Cells were then incubated with a prelamin-A-specific antibody (Santa Cruz Biotechnology Inc, sc-6214), anti-lamin B1 antibody (Maske et al., 2003), anti-calreticulin antibody (Abcam, ab-2907), anti-CCT-α antibody (R. Cornell) or an anti-Nup153 antibody (Abcam, ab-24700). ConA was used at a concentration of 100 μg/ml. The cells were then incubated with Alexa Fluor 488 donkey anti-mouse IgG (Invitrogen, A-11029), Alexa Fluor 647 donkey anti-goat IgG (Invitrogen, A-21447) and Cy3-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch, 711-166-152). Finally the cells were mounted in Mowiol supplemented with DAPI. Immunofluorescence images were collected using a Plan Apo 63× 1.4 NA oil immersion lens and a Plan Neofluar 10× 0.3 NA lens on a Zeiss Axioplan 2 imaging microscope with fibre optic illumination (Sutter Instrument Company, Novato, CA, USA) and a Cool SNAP HQ 12 bit CCD camera (Photometrics, Marlow, UK), which was controlled with Metamorph software (Molecular Devices, Downingtown, PA). The immunofluorescence was analysed using a quad band dichroic reflector (Chroma Technology Corp, Rockingham, VT) for DAPI, to detect DNA (360 nm exciter with 40 nm bandwidth and 457 nm emitter with 50 nm bandwidth), FITC to detect lamin B1 and CCT-α (490 nm excitation with 20 nm bandwidth and 528 nm emitter with 35 nm bandwidth), Cy3 to detect calreticulin (555 nm excitation with 28 nm bandwidth and 617 nm emission with 73 nm bandwidth) and Cy5 to detect prelamin A and ConA (635 nm excitation with 20 nm bandwidth and 685 nm emission with 40 nm bandwidth).

Images of DAPI-stained nuclei were thresholded and binarised using Metamorph (version 7.1.3). The area and perimeter was measured for each nucleus using the integrated morphometry functions in Metamorph. Contour ratio (CR) was calculated for each nucleus using the formula CR=4π×area/perimeter². Values shown are means ± s.d. for 200 nuclei.

For cell cycle tracking in live cells, a Zeiss inverted laser scanning microscope was used in a non-confocal transmission phase-contrast mode at 633 nm, with a Plan Neofluar 10× 0.3 NA objective. The Zeiss environmental control (PeCon GmbH, Erbach, Germany) was maintained at 37°C and 5% CO₂ while imaging the cells undergoing mitosis. Live-cell imaging took place for 24 hours at 15-minute intervals using excitation at 555 nm and emission at 617 nm with the environmental conditions described above.

Cells were washed twice in PBS, trypsinized and centrifuged at 1000 g for 5 minutes at 4°C. The cells were then resuspended in 100 μl 2× SDS running buffer and 100 μl water before being incubated at 90°C for 10 minutes. Samples were then subjected to a 10% SDS-PAGE at 150 V for 1.5 hours before being transferred to a nitrocellulose membrane and bathed in lamin A antibody at 1:200 (Abcam, ab26300), β-actin antibody at 1:5000 (Abcam, 8226), prelamin A antibody at 1:100 (Santa Cruz, sc-6214) or CCT-α antibody at 1:200 (R. Cornell, affiliation) overnight. The secondary antibodies used were peroxidase AffiniPure goat anti-mouse IgG (Sigma, A4416), peroxidase Affinipure goat anti-rabbit IgG (Sigma, A6154) and peroxidase Affinipure donkey anti-goat IgG (Jackson ImmunoResearch Europe Ltd, 705-035-147) at a dilution of 1:5000, 1:6000 and 1:20,000 respectively, for 45 minutes. Proteins were detected using the Immobilin Western Chemiluminescent HRP substrate (Millipore, P90719).

Pixel intensity of nuclei stained for CCT-α was quantified using Metamorph software to generate a histogram of intensities. The intensities of ten nuclei per transfection were averaged to produce the mean intensity of CCT-α inside a nucleus. CCT-α foci were counted and averaged in fifty separate nuclei.

Cells were treated with 1.5 mM HU for 24 hours before being transfected. Images of DAPI-stained nuclei were acquired using a SNAP HQ 12 bit CCD camera with a Plan Neofluar 10× 0.3 NA lens. The phase of the cell cycle was assessed using the cell cycle application in Metamorph software.

Images of grazing sections of anti-Nup153- and mAb414-labelled nuclei were acquired and processed in Metamorph software. Processing involved thresholding the image, binarising the thresholded features, median filtering the binary image (3×3 kernel) to remove single pixels and ultimate eroding of the resulting binary image such that each NPC is represented by a single pixel. Over a 25 μm² area the number of NPCs was counted for ten different nuclei in each treatment.

The farnesyl transferase inhibitor FTI-277 (Calbiochem, Beeston, UK) was dissolved in DMSO and used at 10 μM for 48 hours. To assess protein farnesylation, cells were incubated with the farnesol analogue 8-anilinogeraniol (provided by Hans Peter Spielmann, University of Kentucky, USA), which was dissolved in DMSO and used at a final concentration of 50 μM for 48 hours. The farnesyl analogue (also provided by H. Spielmann) was then detected by western blotting using a mouse monoclonal antibody against the analogue (Troutman et al., 2005).

Cells were sequentially fixed in a mixture of 2.5% glutaraldehyde, 2% paraformaldehyde and 0.2% picric acid, all in 100 mM cacodylate buffer (pH 7.4) with 2 mM MgCl₂, followed by 1% osmium tetroxide in 100 mM cacodylate buffer (pH 7.4) and, after extensive washing, in 2% aqueous uranyl acetate. After this extensive crosslinking to stabilise nuclear morphology in situ, the cells were dehydrated with ethanol, released from the plastic culture dish using propylene oxide, centrifuged and embedded in Epon resin. For routine EM, 50 nm sections were cut, stained and examined in a FEI Tecnai 12 electron microscope (FEI, Eindhoven, The Netherlands), operating at 80 kV.

For tomography, 250-nm thick sections were collected on Formvar-coated copper slot grids. Following staining, 10 nm colloidal gold particles were applied to both surfaces of the grid to serve as fiducial markers for subsequent image analysis. Dual-axis tilt series from −65°to +65°at 1°intervals were obtained using a computerized tilt stage in a Tecnai 12 electron microscope, operating at 120 kV. Tomographic reconstruction and modelling was performed using the IMOD software package (Mastronarde, 1997).

For immunogold EM the cells were fixed in 4% PFA (~10 minutes) followed by 8% PFA, both in 250 mM HEPES buffer (pH 7.4) for 2 hours. The cells were washed in 250 mM HEPES buffer containing 50 mM glycine for 1 hour at room temperature to quench free aldehydes, and released from the plastic by scraping. The cells were pelleted in 3% gelatine, cryo-protected with 2.0 M sucrose, and cryosectioned using a Reichert Ultracut E with the FCS attachment. Cryosections of 95 nm nominal thickness were collected onto Formvar-coated nickel grids using a 1:1 mixture of 2.3 M sucrose and 2% methylcellulose, and immunolabelled with anti-NuP153 (1:30 dilution) followed by goat anti-mouse IgG conjugated to 10 nm gold (British Biocell, UK). Sections were examined in a Tecnai 12 electron microscope.

We thank Stephen Young (Department of Cardiology, UCLA) for Zmpste24^−/− and wild-type MEFs, and Loren Fong (Department of Cardiology, UCLA) for the anti-ZMPSTE24 antibody. We thank Mike Shaw for TEM sample preparation and acquisition of dual-axis tomography data, and Nick White for advice and instruction in confocal laser scanning and widefield light microscopy. Keith Gull, Sylvain Lacomble, Bill Wickstead and Catarina Gadhela (Sir William Dunn School of Pathology, Oxford) were invaluable in the analysis of the dual-axis tomography data set using the Etomo IMOD software. We thank Neale Ridgway (Dalhouise University, Canada) and Rosemary Cornell (Simon Fraser University, Canada) for the CCT-α antibody. We also thank Peter Cook for critical comments on the manuscript.