The Golgi complex has a central role in the secretory pathway of all higher organisms. To explain the synthesis of its unique stacked structure in mammalian cells, two major models have been proposed. One suggests that it is synthesized de novo from the endoplasmic reticulum. The second model postulates a pre-existing Golgi template that serves as a scaffold for its biogenesis. To test these hypotheses directly, we have developed an approach in which we deplete the Golgi complex from living cells by laser nanosurgery, and subsequently analyze the ‘Golgi-depleted’ karyoplast using time-lapse and electron microscopy. We show that biosynthetic transport is blocked after Golgi depletion, but is restored 12 hours later. This recovery of secretory transport coincides with an ordered assembly of stacked Golgi structures, and we also observe the appearance of matrix proteins before that of Golgi enzymes. Functional experiments using RNA interference-mediated knockdown of GM130 further demonstrate the importance of the matrix during Golgi biogenesis. By contrast, the centrosome, which can also be removed by laser nanosurgery and is not reformed within the considered time frame, is not required for this process. Altogether, our data provide evidence that de novo Golgi biogenesis can occur in mammalian cells.
The Golgi complex is a central organelle in the secretory pathway of all eukaryotes. It receives newly synthesized material from the endoplasmic reticulum (ER), which is processed and sorted at the trans-Golgi network for its final destination in the cell. Live-cell studies monitoring the dynamics of GFP-tagged Golgi proteins have demonstrated that this organelle is highly dynamic, raising the fundamental question of how it acquires and maintains its steady state architecture and function (Lippincott-Schwartz et al., 2000). Indeed, the molecular mechanisms of Golgi biogenesis and maintenance are presently only poorly understood (Lowe and Barr, 2007). In mammalian cells, Golgi growth occurs during interphase and relies on ER-to-Golgi transport (Guo et al., 2006), but whether Golgi biogenesis occurs from a pre-existing template or de novo remains to be resolved. Treatment of mammalian cells with the drug brefeldin A (BFA) disassembles the Golgi complex and most Golgi components redistribute to the ER. Interestingly, after drug washout, a fully intact Golgi complex can reassemble (Lippincott-Schwartz et al., 1989; Puri and Linstedt, 2003; Kasap et al., 2004), which is consistent with a de novo Golgi biogenesis scenario. However, to what extent this drug treatment represents a bona fide Golgi biogenesis model is not clear, because it has been shown that following BFA treatment, Golgi remnants remain throughout the cytoplasm (Orci et al., 1993; Seeman et al., 2000). It is therefore possible that these remnants act as a template for Golgi biogenesis after drug washout. More recent experiments, in which the Golgi complex was chemically inactivated in living cells, have shown that a new Golgi complex growing adjacent to the inactivated one lacks essential Golgi proteins and functionalities, including normal trafficking activity (Jollivet et al., 2007). This led to the conclusion that, although de novo biogenesis of Golgi-like structures can occur in mammalian cells, pre-existing Golgi templates are necessary for their organization into a morphologically and functionally intact organelle. Common to all of these approaches is the inability to remove the Golgi complex or its remnants from intact cells, and therefore, the fundamental question of whether a functional Golgi complex can be synthesized directly from ER membranes cannot be addressed. Alternative approaches to study Golgi biogenesis using cell microsurgery to generate Golgi- and nucleus-free cytoplasts were unable to synthesize a functional, transport-competent Golgi complex from the ER (Pelletier et al., 2000). Microsurgery has also been used to deplete the centrosome and Golgi from living cells (Maniotis and Schliwa, 1991) and subsequently study their biogenesis. The authors suggested that the Golgi complex, which was at least partially depleted from cells together with the centrosome, could be resynthesized to a normal extent within 20 hours of the surgery. However, as the focus of this study was the biogenesis of centrosomes, the extent of initial Golgi depletion by cell surgery was not addressed in detail. Therefore, despite the extensive amount of elegant work to characterize Golgi biogenesis, a clear description of how this occurs is currently lacking, largely because of technical limitations of the methods used.
Here, we describe a new method to study Golgi biogenesis by first displacing the Golgi complex from its juxtanuclear position, followed by its depletion from living mammalian cells using laser nanosurgery. Subsequently, we have followed the karyoplasts containing an intact nucleus using time-lapse microscopy, and investigated whether a functional Golgi complex can re-form under these conditions. In summary, our data not only demonstrate the feasibility of this approach, but also provide strong support for the hypothesis that the Golgi complex is capable of forming de novo in mammalian cells.
Depleting the Golgi complex from living cells by laser nanosurgery
A BSC1 cell line, stably expressing GFP2-tagged galactosyltransferase (GalT–GFP) as a live-cell marker to visualize the Golgi complex, was used. Microdissection of cells was performed using a diffraction-limited laser nanosurgery device, which allows precise intracellular cutting of structures such as microtubules and actin stress fibers (Colombelli et al., 2004; Colombelli et al., 2005). Severing actin stress fibers leads to their retraction, and as a consequence, to the retraction of the plasma membrane and finally, dissection of the cell into two pieces (supplementary material Movie 1). To remove the Golgi complex in an efficient manner, we exploited our recent observations that BSC1 cells grown on 6-μm-wide fibronectin lines show a displacement of the Golgi complex and associated centrosome from the nucleus by distances of up to several micrometers (Pouthas et al., 2008) (Fig. 1A, supplementary material Figs S1 and S2). In experiments using cells growing on non-coated cover-glass surfaces, where the Golgi complex is often close to, or even overlapping, the cell nucleus, substantial depletion of the Golgi complex by laser nanosurgery was difficult and mostly incomplete, as judged by fluorescence-microscopy-based visual inspection using GalT–GFP as a Golgi marker (data not shown). By contrast, the procedure was significantly facilitated when cells were grown on fibronectin lines, resulting in a marked displacement of the Golgi complex from the cell nucleus. In this case, the entire GalT–GFP signal could be eliminated from the karyoplast by laser nanosurgery, as judged at the light microscopy level (Fig. 1A).
Electron microscopy analyses (Colombelli et al., 2008) of serial sections from karyoplasts fixed either immediately (n=3) or 4 hours after (n=4) laser nanosurgery confirmed the light microscopy observations. We could not detect any structures either morphologically similar to a stacked Golgi, Golgi mini-stacks or centrioles (Fig. 1C). Other membrane-bound organelles and structures of the early secretory pathway, including ER-exit sites, could however, be clearly identified in these karyoplasts (Fig. 1C). By contrast, Golgi stacks and associated centrosomes were only detected in the Golgiplasts (Fig. 1C). Consistent with these observations, presence of the Golgi markers GM130 (Golgin subfamily A member 2), giantin, TGN46 and the centrosomal marker γ-tubulin could not be detected above background staining in the karyoplasts up to 4 hours after laser nanosurgery (Fig. 1B,D; Table 1). From a total of 40 cells analyzed by immunofluorescence and by light microscopy at these early time-points after laser nanosurgery, 38 showed no Golgi-specific or centrosome-specific labeling, and only two karyoplasts showed a weak Golgi-like juxtanuclear presence of the Golgi markers (Table 1). By contrast, the Golgiplasts located next to the karyoplasts contained Golgi and centrosomal marker labeling that was similar to labeling in untreated neighboring control cells (Fig. 1D).
Monitoring Golgi biogenesis by correlative light-electron microscopy
Time-lapse analyses of karyoplasts showed that approximately 1 hour after laser nanosurgery, the GalT–GFP-specific fluorescence started to increase steadily in the karyoplasts (Fig. 2A and supplementary material Movie 2) indicating a recovery of essential cell functions such as gene transcription and protein translation. At 6 hours and later, discrete GalT–GFP-specific structures appeared throughout the karyoplasts (Fig. 2A and supplementary material Movie 2). In karyoplasts 48±14% of these discrete GalT–GFP structures (n=267) colocalized with specific markers of the early secretory pathway vesicular coat, COPII and COPI (Fig. 2B). This was similar to the colocalization measured in unperturbed control cells (50±14% colocalization; n=112). Therefore, the discrete GalT–GFP-positive structures are probably transport carriers of the early biosynthetic transport route, as described in previous studies (Presley et al., 1997; Scales et al., 1997). At later time points (~8 hours or later after nanosurgery), these structures clustered in a juxtanuclear area where they colocalized with the endogenous Golgi markers GM130, giantin and TGN46 (Fig. 2C and Table 1). Interestingly, at all time points tested (up to 24 hours after laser nanosurgery) centrosomal γ-tubulin labeling was never observed in the karyoplasts (Fig. 2D). Electron microscopy analyses of serial sections from karyoplasts fixed 12 hours or later after nanosurgery revealed in 7 out of 11 karyoplasts analyzed, the presence of either several Golgi mini-stacks or normal-sized Golgi stacks, ranging from 500 nm to 2.2 μm in length (Fig. 3A), with the latter ones appearing highly similar to the single stacks observed in unperturbed control cells. Furthermore, the new Golgi stacks were always found adjacent to the nucleus. Consistent with the immunofluorescence analyses (Fig. 2D), we were unable to detect the presence of a centrosome or centrosome-like structures in any of these karyoplast serial sections.
Correlation of biosynthetic and retrograde transport activity with Golgi biogenesis
To investigate whether the newly synthesized Golgi complex was functionally intact, we tested the efficiency of biosynthetic transport in the karyoplasts. To this end, we expressed a GPI-anchored CFP (GPI–CFP) (Keller et al., 2001) in cells and followed its transport from the ER to the plasma membrane in karyoplasts at different time points after laser nanosurgery (Fig. 3B). In ‘Golgi-depleted’ karyoplasts, the anterograde transport of GPI-anchored CFP to the plasma membrane was not detectable within the first 6 hours after laser nanosurgery (Fig. 3B, n=5), although the karyoplasts expressed the marker (data not shown). In cells cut intentionally in such a manner that parts of the Golgi complex remained in the karyoplast (‘cut control’), transport of GPI–CFP to the plasma membrane occurred in a similar manner to that in unperturbed neighboring control cells (Fig. 3B, n=4). This suggests that in our assay, even parts of a functional Golgi complex are sufficient to allow GPI–CFP plasma membrane transport. In support of this, in one of the karyoplasts analyzed 6 hours after laser nanosurgery, and in which discrete juxtanuclear GM130-positive Golgi-like structures were observed, transport of GPI–CFP appeared to be restored to similar levels to those observed in control cells (Fig. 3C). This result is consistent with published data showing that a single mini-stack or Golgi fragment is sufficient to support transport from the ER to the plasma membrane (Pelletier et al., 2000). In karyoplasts analyzed 16 hours after laser nanosurgery, levels of GPI-anchored CFP at the plasma membrane were indistinguishable from those found in unperturbed control cells (Fig. 3B, n=4). Similar results demonstrating the transport competence of the newly synthesized Golgi complex were obtained in experiments following the Golgi-dependent transport of cholera toxin (CTX) (Fig. 4). Previous work has shown that the retrograde transport of this toxin from the plasma membrane to the ER requires an intact Golgi complex (Orlandi et al., 1993; Feng et al., 2004). In 70±5% of unperturbed control cells, fluorescently labeled CTX could be seen in internal membranes, including Golgi-like structures, the ER and the nuclear envelope (Fig. 4A). By contrast, we were unable to detect arrival of the toxin in the ER in any of the karyoplasts where the Golgi complex had been depleted by laser nanosurgery (n=6), 6 hours after laser nanosurgery (Fig. 4B). In these karyoplasts CTX-specific plasma membrane staining was predominantly observed, and very few internal membrane structures were labeled. In parallel experiments where cells were cut such that parts or the Golgi complex remained in the karyoplast, transport of the toxin to the ER occurred in a manner indistinguishable from that observed in unperturbed control cells (Fig. 4C, n=9). This result suggests that the laser nanosurgery procedure itself, or indeed removal of part of the Golgi complex, does not block trafficking of CTX through the Golgi complex to the ER. When karyoplasts were labeled with CTX 16 hours after laser nanosurgery, a time-point where Golgi mini-stacks were detectable in the karyoplasts (Fig. 3A), transport of CTX to the ER was restored to levels that were comparable with those seen in control cells (Fig. 4D, n=5).
Altogether, our data are consistent with the hypothesis that the Golgi complex can be substantially depleted from cells using laser nanosurgery and that this organelle is subsequently resynthesized within a period of 12 hours.
The appearance of endogenous Golgi matrix proteins and enzymes during Golgi biogenesis
To commence investigation of the mechanism by which the observed Golgi reassembly could occur, we quantified the appearance of endogenous Golgi matrix proteins (GM130, giantin, GRASP65) and enzymes (Mannosidase II, GalNacT1 and GalT; Fig. 5 and supplementary material Table S1) at different time points after nanosurgery. Consistently, all the matrix proteins tested began to show a juxtanuclear distribution coincident with a similar distribution of the ectopically expressed GFP-tagged Golgi marker (Fig. 5). Quantification of the Golgi-like signals for the matrix proteins showed that they appeared with similar efficiencies after nanosurgery (Fig. 5 and supplementary material Table S1). By contrast, endogenous Golgi enzymes only acquired a Golgi-like appearance at later time points (Fig. 5 and supplementary material Table S1). For example, at 12 hours after nanosurgery, when the newly formed Golgi complex was visible as a stacked structure (Fig. 3A) and was transport competent (Fig. 3B and Fig. 4), we could detect little Golgi-like labeling of endogenous Golgi enzymes in the majority of the cells analyzed (Fig. 5, supplementary material Table S1). These results indicate that a stacked Golgi complex can assemble and regain functionality despite lacking its full complement of resident enzymes.
Golgi biogenesis in GM130-downregulated cells
Based on these results, one possibility is that matrix proteins have a role in the early steps of Golgi assembly. To further test this hypothesis we used an RNA interference approach to specifically assess the role of GM130 in reassembly of the Golgi following nanosurgery. Although the small interfering RNA (siRNA) sequences used for the experiments was designed to target the human gene encoding GM130, a strong knockdown efficiency of up to 90% was also achieved in the monkey BSC1 cell line used in our experiments (see supplementary material Fig. S3A). At the immunofluorescence level, a significant reduction of GM130-specific staining, but not the Golgi marker giantin, was observed in most of the cells 72 hours after siRNA transfection (supplementary material Fig. S3B,D). Immunofluorescence and electron microscopy analyses showed a significant fragmentation of the Golgi complex in cells transfected with the siRNA targeting GM130 (supplementary material Fig. S3): a phenotype that is consistent with earlier work in which this protein was depleted from cells by RNA interference (Puthenveedu et al., 2006; Marra et al., 2007). No effect on GM130 levels or Golgi complex morphology was observed in cells transfected with non-silencing control siRNAs (supplementary material Fig. S3). Also consistent with earlier work (Puthenveedu et al., 2006; Marra et al., 2007), knockdown of GM130 did not interfere with the transport kinetics of a temperature-senitive variant of the vesicular stomatitis G-protein (VSV-G ts045) to the plasma membrane (supplementary material Fig. S4). The arrival of VSV-G at the Golgi complex following its export from the ER was also not significantly inhibited under the experimental conditions used here (supplementary material Fig. S4).
When we analyzed Golgi biogenesis in GM130-knockdown cells 12 hours after nanosurgery, giantin labeling was largely absent and GalT–GFP distribution was confined to the ER and discrete structures throughout the cytoplasm (Fig. 6A and not shown). By contrast, in cells transfected with control siRNAs, a substantial amount of giantin staining and GalT–GFP fluorescence could be observed in juxtanuclear Golgi-like structures (Fig. 6). At later time points, giantin staining could be detected in discrete structures in GM130-knockdown cells and 24 hours after laser nanosurgery, its cellular distribution was similar to that in neighboring control cells not subjected to nanosurgery (Fig. 6). Ultrastructural analyses by electron microscopy confirmed these observations. In GM130-knockdown cells 12 hours after nanosurgery, we could not identify any stacked Golgi-like structures, but rather only carrier-like membranes that were occasionally associated with microtubules or vesicular clusters in the juxtanuclear area (Fig. 7A). A stacked Golgi structure, although largely fragmented, was only observed in GM130-knockdown cells 24 hours after nanosurgery (Fig. 7B).
Here, we describe and apply a new approach to deplete the Golgi complex from living cells and subsequently follow its biogenesis using correlative time-lapse and electron microscopy. Our results show that ‘Golgi-depleted’ karyoplasts are capable of synthesizing a stacked and transport-competent Golgi complex within 12 hours, which is consistent with models for de novo Golgi biogenesis. Surprisingly, Golgi biogenesis apparently occurs in the absence of a centrosome, although it requires the Golgi matrix protein GM130. This result is also consistent with our observation that during Golgi biogenesis, matrix proteins display a juxtanuclear Golgi-like appearance when Golgi enzymes are still largely absent from this organelle.
The phenomenon of Golgi biogenesis has been extensively studied by a number of laboratories using various approaches, in turn resulting in a variety of results and interpretations (for a review, see Lowe and Barr, 2007). Our results appear to be in contrast to earlier work in mammalian cells (Pelletier et al., 2000; Jollivet et al., 2007), which suggested that de novo biogenesis of a stacked and transport-competent Golgi complex is not possible, and that Golgi biogenesis more likely involves templated growth. One reason why Jollivet and colleagues (Jollivet et al., 2007) might not have observed de novo biogenesis of a stacked and transport-competent Golgi complex could be that their time of observation after chemical inactivation of the existing Golgi complex (max. 4 hours) was shorter than the time actually required for Golgi biogenesis in karyoplasts as we observed here (12 hours). Pelletier and colleagues (Pelletier et al., 2000) clearly showed that cytoplasts generated by microsurgery fail to synthesize a Golgi or Golgi-like structures from the ER and are not capable of supporting biosynthetic transport to the plasma membrane. One limitation of this earlier study, compared with the approach that we have used, is that the cytoplasts did not contain a nucleus, and therefore transcription to ultimately drive new protein synthesis could most likely not occur. These cellular events might be essential for de novo Golgi biogenesis. However, our data are consistent with earlier studies that have used BFA in combination with the kinase inhibitor H89, which inhibits COPII recruitment and cargo exit from the ER (Puri and Linstedt, 2003), to induce Golgi complex disassembly. Because Golgi components are only redistributed rather than actually depleted in these approaches, a Golgi complex could therefore rapidly reassemble within 30 minutes after washout of the drugs. The time required for the biogenesis of a morphologically and functionally intact Golgi complex, as we have observed here, strongly argues that a number of factors first need to be synthesized by the cells before Golgi assembly can occur. Our data are supported by analyses using light microscopy and electron microscopy that show a substantial, if not complete, depletion of components of the Golgi complex immediately after nanosurgery. Our observation that RNA-interference-mediated knockdown of GM130 can delay Golgi reformation in karyoplasts for several hours further strengthens the evidence that matrix proteins such as GM130 had been substantially depleted in the karyoplast by the nanosurgery. Although a definitive demonstration of the complete absence of any remaining ‘Golgi seed(s)’ after laser nanosurgery is formally impossible, our results strongly suggest the existence of de novo Golgi biogenesis in mammalian cells.
A close association and functional interplay between the Golgi complex, the centrosome and associated microtubules has been established in recent years (for a review, see Sütterlin and Colanzi, 2010). Surprisingly, our data here suggest that the biogenesis of a stacked and transport-competent Golgi complex can occur in the absence of a centrosome. This raises the immediate question of how juxtanuclear positioning of the newly synthesized Golgi complex, as we have observed here, is achieved. One possible explanation might be through the use of a self-organization process in which microtubules nucleating from the membranes of the nuclear envelope and/or ER-derived Golgi precursors is sufficient to facilitate clustering of the precursors into larger pre-Golgi structures that are then able to mature into a complete Golgi complex. In addition to such a self-organizing mechanism, a number of factors associated with Golgi precursors that would need to be resynthesized after Golgi depletion, and thereby help to form a functionally and structurally normal Golgi complex, should also exist. Likely candidates are proteins proposed to be part of a Golgi matrix, such as GM130 (Shorter and Warren, 2002; Lowe and Barr, 2007). The approach described here should help to further characterize the role of these proteins in relation to Golgi biogenesis. In line with the hypothesis for a crucial role of Golgi matrix proteins in maintaining and organizing Golgi structure and function, the knockdown of GM130 in our experiments delayed Golgi biogenesis in karyoplasts for several hours. Moreover, observations that knockdown of GM130 under our experimental conditions had no apparent effect on secretory transport, excludes the possibility that a lack of GM130 causes a general inhibition of ER export and therefore a subsequent delay in the arrival at the Golgi of material such as lipids and proteins that are necessary to build a new Golgi complex. Therefore, these data are consistent with a structural role of GM130 in Golgi biogenesis at the level of vesicle tethering or lateral cisternal fusion, for example, as suggested previously (Lowe and Barr, 2007; Puthenveedu et al., 2006; Marra et al., 2007).
Altogether, our data strongly support the hypothesis that Golgi biogenesis can occur de novo in mammalian cells, even in the absence of a centrosome. The Golgi matrix protein GM130 appears to have a crucial role in the efficiency with which this biogenesis occurs. With the approach described here, it will now be possible to investigate the role of other factors in Golgi biogenesis more systematically.
Materials and Methods
Reagents were purchased as follows: fibronectin (from bovine plasma, Sigma), Alexa-Fluor-647-conjugated fibrinogen (Molecular Probes), Sylgard 184 (Dow Corning). Suberic acid bis (3-sulfo-N-hydroxysuccinimide ester) sodium salt (BS3, Sigma), 3-aminopropyltriethoxysilane (APTS, Sigma), Poly(L)-lysine hydrobromide (Sigma), mPEG-SPA-2000 or mPEG-SPA-5000 (Shearwater), silicon-free grease (Glisseal N, Borer Chemie), Enhanced GFP-Golgi mammalian expression vector (Clontech), Monosialoganglioside GM1 from bovine brain (Sigma), Cholera toxin (CTX, from Vibrio cholerae, Sigma), Neomycin (G418, Calbiochem) and GFP2 expression vector (Perkin Elmer). Anti-GM130 (Silencer pre-designed siRNA ID 145093) and Scramble (Silencer Negative Control siRNA 1) siRNA were purchased from Ambion. All the cell culture reagents were purchased from Gibco, unless otherwise noted.
Cell lines, cell culture and transfection
BSC1 (ATCC CCL-26) and BSC1-GalT–GFP2 cells were cultured in modified Eagle's medium (MEM, Life Technologies) supplemented with 10% fetal calf serum (FCS), penicillin and streptomycin (100 U/ml and 100 μg/ml, respectively) at 37°C in 5% CO2. Cells were seeded on patterned surfaces or on live-cell dishes (MatTek) at a density of 15,000 cells/ml. Experiments were performed 24 hours or 48 hours after seeding of the cells. One hour before nanosurgery, the medium was exchanged to CO2-independent medium.
Stable cell lines were generated by transfecting BSC1 cells using Fugene6 (Invitrogen) according to the manufacturer's protocol with a modified version of the EGFP–Golgi vector containing a neomycin-resistance cassette (Clontech). It encodes a fusion protein consisting of modified green fluorescent protein (GFP2) and a sequence encoding the N-terminal 81 amino acids of human β1,4-galactosyltransferase (Gal-T). This region contains the membrane-anchoring signal peptide that targets the fusion protein to the trans-medial region of the Golgi complex. Addition of neomycin or G418 (a neomycin analog) selected for cells that had incorporated the vector. Resistant clones were screened for GFP2–Golgi expression levels, growth rates and Golgi morphology, and subject to a second round of selection by cell sorting.
YT2 cells [BSC1 stably transfected with the GalNAcT2-YFP (Axelsson and Warren, 2004)] were a gift from Graham Warren (Max Perutz Laboratories, Vienna, Austria) and were cultured in D-MEM (Dulbecco's modified Eagle medium, GIBCO) supplemented with 10% FCS and Geneticin (800 μg/ml) at 37°C in 5% CO2.
For transient transfection of siRNAs, Oligofectamine (Invitrogen) was used in accordance to the manufacturer's instructions. For the assessment of the effects of GM130 knockdown (supplementary material Fig. S3B–E) cells were analyzed 72 hours after transfection, whereas laser nanosurgery was performed 12–24 hours after transfection, because the removal of the Golgi-complex-associated GM130 in the presence of the siRNA (and thus of new protein translation) allowed us to have a very efficient knockdown efficiency immediately after transfection, without waiting for degradation of the pre-existing protein.
Antibodies and immunofluorescence
Antibodies were used as follows: monoclonal antibodies GM130 (BD Biosciences), Golgin97 (Molecular Probes), GalNAcT1 and GalT (CellMAb), α-tubulin (clone GTU-88, Sigma); polyclonal antibodies Giantin (Abcam), Mannosidase II (AbD Serotec) and Sec31 (BD Biosciences). Rabbit antibodies against recombinant GFP and beta-prime COPI were raised by standard protocols. Sheep antibodies against TGN46 (Biozol) and GRASP65 (a gift from Martin Lowe, University of Manchester, UK) were used. All fluorescently conjugated secondary antibodies were purchased from either Molecular Probes or Amersham.
Cells were fixed with either −20°C methanol for 4 minutes or 4% paraformaldehyde (PFA) for 15 minutes. Cells were subsequently treated with 0.1 M glycine in 1× PBS for 5 minutes and then in 0.1% Triton X-100 in 1× PBS for a further 5 minutes. Fixed and permeabilized cells were incubated with primary antibodies in 1× PBS for 30 minutes. Samples were washed three times for 5 minutes in 1× PBS, and incubated for 30 minutes in fluorescently conjugated secondary antibodies, before being washed three times for 5 minutes in 1× PBS. Where required, cells were permeabilized with 0.1% saponin. Coverslips were finally mounted on glass slides with Mowiol (Calbiochem).
The patterning was performed as previously described (Pouthas et al., 2008). Briefly, the layout design of the stamp was done using CleWin Software (WieWeb) and translated into 5 inch chromium photolithography mask by Delta Mask V.O.F. (Enschede) then used to produce a structured silicon master (3 μm depth) using a positive tone resist process. Poly(dimethylsiloxane) stamps (PDMS stamps) were fabricated by curing Sylgard 184 (Dow Corning) at 60°C for 24 hours. After curing, the elastomer was peeled off from the master. The PDMS stamps were incubated with 100 μl aliquot of 1:2 mixture of Alexa-Fluor-647-conjugated fibrinogen and fibronectin solution, or only fibronectin solution, at a final protein concentration of 50 μg/ml in 1× PBS for 30 minutes. The stamps were then washed in deionized water and blow-dried with a stream of nitrogen (FN-stamp). Silanization of the slides was performed by immersion in 5% 3-aminopropyltriethoxysilane (APTS), acetone solution for 30 seconds followed by several washing steps in 1× PBS. The slides were then incubated in bis(sulfosuccinimidyl)suberate (BS3; approx. 0.5 mg/ml in 1× PBS) for 10 minutes. Blow-dried FN stamps were then placed in contact with silanized glass surfaces for 5 minutes. After removal of the stamp, the slides were incubated in a 1 mg/ml poly(L-lysine)-g-poly(ethylene glycol) (PLL-g-PEG) for 30 minutes for back-fill. All these processes were done at room temperature. Finally, the patterned glass slides were attached to drilled 10 mm holes at the bottom of 35 mm plastic Petri dish using silicon-free grease (Glisseal N, Borer Chemie).
Generation of karyoplasts devoid of Golgi by laser nanosurgery
Cell nanosurgery by low intensity and shockwave-free plasma ablation was achieved with a frequency tripled Nd:YAG-pulsed laser (JDSUniphase), at a wavelength of 355 nm and theoretical pulse duration of less than 500 pseconds on a Zeiss Axiovert 200M using a Zeiss heating stage insert. All irradiations were performed along a manually oriented laser line-target, defined interactively on the image with a graphical mouse interface. Pulse density along the line was set such that the optical intensity was approximately constant with a linear density of about 5 pulses per μm at a speed of 100–1000 pulses per second. In this study, actin severing was successful with typical energy of 200 nJ per pulse. Cell dissection was also successful with lower energies down to about 100 nJ and with repeated irradiations in several axial positions to insure fast and efficient separation of karyoplast and Golgiplast. Alternatively, an Olympus Fluoview 1000 system coupled with a second scanning unit (355 nm pulsed laser) was used for the nanosurgery.
After cell dissection, the positions of successfully cut cells were inscribed inside the glass coverslip about 10 μm deep with the same laser at higher power (Colombelli et al., 2008). By doing so, a square box could be placed around the recently cut cell and a number could be written to clearly identify it during the following steps.
Cells and karyoplasts were imaged using either a multi-position time-lapse microscope: CellR, or CellM (Olympus) with the following objectives: 20× (UPlanSApo, NA 0.75), 40× (UPlanSApo, NA 0.9). Alternatively, cells were imaged using a DeltaVision RT system (Applied Precision) with a 20× (NA 0.4) objective. After immunofluorescence, images were captured using a Zeiss Axiovert 200 microscope, Zeiss LSM 710, or Leica SP2 AOBS confocal microscope (Leica Microsystems).
Cholera toxin (CTX) was labeled with Cy3 dye (Amersham) according to the manufacturer's instructions. Karyoplasts were followed by time-lapse microscopy for 2 hours or 12 hours after laser nanosurgery, incubated with GM1 receptor (final solution 1 μM in MEM without FCS) on ice for 20 minutes, then with CTX (1:750) for an additional 20 minutes followed by 5 minutes at 37°C (toxin release). Unbound toxin was removed and CTX uptake was followed for 4 hours by time-lapse microscopy, before fixation with 4% PFA.
GPI–CFP adenovirus was a gift from Kai Simons (Max Planck Institute of Cell Biology and Genetics, Dresden, Germany) and prepared as described previously (Keller et al., 2001), diluted 1:20 in medium (MEM with 10% FCS) and incubated on cells (100 μl/dish, 1 hour, 37°C). Cells were washed once and left for 10 minutes at 37°C. The laser nanosurgery was performed within 1 hour to avoid any transport of the GPI to the plasma membrane before karyoplast generation. The karyoplasts were followed for 6 hours before fixation with 4% PFA. Alternatively, the karyoplasts were followed for 12 hours or longer. In all cases, the GPI–CFP plasma membrane signal was enhanced using a polyclonal antibody against GFP, without membrane permeabilization.
For the VSV-G ts045 transport assay, cells were transfected with GM130 or scramble siRNA and 55 hours thereafter they were infected with an adenovirus vector for the expression of ts045–ECFP (Keller et al., 2001). After incubation at 40°C for 17 hours (which results in the accumulation of the protein in the ER), cells were either fixed immediately in PFA, or the temperature was shifted to 32°C to release the ER-exit block and then cells were fixed at different time points after the temperature shift.
Cells were then immunolabeled for VSV-G without permeabilization to identify only the surface fraction and the nuclei were stained with Hoechst 33342. 16 images were randomly acquired for each time point with an automated Olympus widefield microscope (operated with the software ScanR) and the analysis was done as described above.
Correlative light and electron microscopy
Correlative light and electron microscopy was performed as previously described (Colombelli et al., 2008). Briefly, cell locations were marked with the laser inside the glass coverslip and further fixed for flat embedding with 2.5% glutaraldehyde in Cacodylade buffer. Samples were washed with CaCo buffer, and kept cold on ice during incubation with 2% OsO4 in CaCo buffer for 40 minutes. The sample was rinsed with water, before uranyl acetate (0.5% in water, 30 minutes) was added to increase the contrast. After several washes with water, the samples were moved back to the laser nanosurgery microscope for laser surface etching. Two lines perpendicular to the fibronectin pattern, one on each side of the cell of interest, were etched at the surface of the coverslip with a depth of a few microns. Additional washing with PBS was done to remove glass debris from the top of the cells, which were then dehydrated with ethanol. The coverslips were embedded in Epon and left to polymerize at 60°C (24 hours). After polymerization and removal of the coverslip, the etched marks were clearly visible on the Epon surface, so that the samples could be trimmed only around the cell of interest. Ultra thin (50–60 nm) sections were made [on average 58±21 (s.d.), n=18] serial sections covered the complete cytoplasm of the cell) and transferred to copper-palladium slot grids (2×1 mm, Plano, Wetzlar) freshly coated with Formvar. The sections were post stained with uranyl acetate and lead citrate. The pictures were taken using a BioTwin CM120 electron microscope (FEI Company).
Golgi complex displacement
Analysis of a complete Golgi complex displacement was done by measuring the immunofluorescence intensity of different Golgi markers in the Golgi area compared with areas adjacent to the nucleus. Fixed control cells grown on fibronectin lines were selected for a displaced Golgi complex in the GalT–GFP channel and confocal images were acquired of the complete cell volume. From the maximum intensity projections in ImageJ, overlay pictures were made from the different channels. RGB pictures were used to create plot profiles, by placing a 70-μm-long line along the cell, through the centre of the nucleus and the Golgi complex.
Colocalization analysis between GalT–GFP and COPI structures was performed as follows. Images were acquired with a Zeiss Axiovert 200 microscope, 63× (NA 1.4) or a Leica SP2 AOBS confocal microscope, 63× (NA 1.4) and loaded in ImageJ (version 1.39p, NIH). Each channel image (GalT–GFP, COPI and COPII) was separately analyzed and fluorescently positive structures were marked, using the ‘point selection’ function. The three channel pictures were thereafter overlaid, and the colocalization in each category was measured (GalT–GFP, GalT–GFP/COPI, GalT–GFP/COPII and GalT–GFP/COPI/COPII), and used for calculation of the mean ± s.d.
Appearance of endogenous proteins in the new Golgi complex
The sequential order of appearance of the different Golgi proteins in the de novo Golgi complex was analyzed as follows. Immunofluorescence images of GM130 and an additional marker of interest were acquired on a Zeiss LSM 710 confocal microscope as a maximum projection of Z stacks across the whole cells. Images were low-pass filtered to remove image noise and background signals were subtracted. Corrected images were then thresholded such that the entire Golgi-marker-specific signal in non-treated control cells was detected. The total fluorescence in each thresholded Golgi structure was then determined and the ratio between the karyoplast under view and the average of the neighboring control cells was calculated, to assess the recovery of the Golgi complex in the karyoplast. GM130 endogenous levels were used as a reference in all cases, except for the GalT measurement shown in Fig. 5, where Giantin was used. The quantification of these experiments is shown in supplementary material Table S1.
The analysis of the Giantin signal in GM130-knockdown or control cells 12 or 24 hours after nanosurgery was carried out as described above, and GM130 signal was in this case used to check the efficiency of the protein knockdown. The ratio of total Giantin signal in the karyoplast and the average in control cells was used here as a measure of the Golgi reformation process and is shown in Fig. 6.
Automated analysis of VSV-G transport assay
Automated image analysis procedures were implemented as modules in a MATLAB-based version of the open-source software CellProfiler (Lamprecht et al., 2007). The quantification of VSV-G surface transport assay was performed as described previously (Simpson et al., 2007). Briefly, from the identified cell nuclei, the cytoplasmic extension of a cell (cell mask) was approximated by step-wise dilation of the nuclear area until neighboring areas touch or the dilation exceeds a user defined maximal radius, which was chosen to match the typical size of a cell. After subtracting background intensities from each image, the integrated optical density within each cell mask was computed for both VSVG-CFP (Total ts045) and anti-VSVG–Alexa-Fluor-647 (Surface ts045). The ts045 surface transport efficiency of a cell was assessed by dividing Surface ts045 by Total ts045. The histogram in supplementary material Fig. S4D shows the mean transport efficiency calculated on more than 100 cells for each experimental condition.
For automated detection of Golgi structures in fluorescence images (GalNAcT2–YFP), we developed a MATLAB-based image-processing module for the segmentation of locally bright objects of arbitrary shape. In brief, we separated locally bright Golgi foreground signals from background signals by a morphological top-hat filter with a circular structural element, whose radius was chosen to match the typical size of Golgi objects. The resulting foreground image was then subjected to a local adaptive thresholding algorithm, neighboring thresholded pixels were combined, and connected components with a minimal number of 4 pixels were assigned as Golgi objects. CFP–ts045 images were subjected to a morphological top-hat filter with a circular structural element, whose radius was chosen to keep locally bright CFP–ts045 signals such as that due to accumulation within the Golgi, but to subtract ‘background’ intensities such as CFP-ts045 residing in the ER. Next, for each cell we computed the integrated optical density of the background-corrected CFP–ts045 images within Golgi masks and this value was divided by the Total-ts045 value to obtain the ts045 Golgi fraction.
SDS-PAGE and western blotting
BSC1 cells were lysed in sodium dodecyl sulphate (SDS) sample buffer, boiled for 10 minutes and resolved in NuPage 4–12% Bis-Tris polyacrylamide gradient gel (Sigma). Proteins were transferred onto PVDF membranes (BioRad) using a Mini Trans-Blot cell (BioRad) at 100 V for 80 minutes. Membranes were blocked in Tris-buffered saline, 0.1% Tween 20, 2% milk powder, probed using antibodies against GM130 (BD Biosciences) and α-tubulin (NeoMarkers) and specific binding was detected via an Alexa-Fluor-680-conjugated secondary antibody (Molecular Probes) with an Odyssey Infrared Imaging System (LI-Cor Biosystems).
We are grateful to F. Pouthas and P. Girard for providing expertise in micropatterning and T. Wolf for providing wafers. We thank C. Tischer for help in the automated image analysis, the ALMF team at EMBL Heidelberg for their continuous support, and C. Schuberth and E. San Pietro in the Pepperkok team for critical reading of the manuscript. P.R. is a recipient of a von Humboldt exchange fellowship.
↵* These authors contributed equally to this work
↵‡ Present address: Institute for Research in Biomedicine (IRB) Parc Científic de Barcelona, C/ Baldiri Reixac 10, 08028 Barcelona, Spain
↵§ Present address: School of Biology and Environmental Science, University College Dublin (UCD), Belfield, Dublin 4, Ireland
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.079640/-/DC1
- Accepted November 17, 2010.
- © 2011.