The Golgi complex in one of the largest cellular organelles, which is increasingly viewed as the main coordinator of different trafficking pathways and molecular events that play a grand role in cellular homeostasis [1
]. For years, researchers have obsessed over the molecular mechanisms that contribute to Golgi function and unique ribbon-like architecture. The emerging consensus is that the sequential distribution of enzymes in the interconnected stacks of flat cisternae provides the appropriate modifications to the cargo [3
]. This is the main paradigm of secretory transport through the Golgi, regardless of whether it is postulated by the cisternal or vesicular model [5
]. Therefore, the intact and perinuclear position of Golgi is ideal for the complete processing of proteins. However, in response to different intracellular perturbations, including inhibition of endoplasmic reticulum (ER) to Golgi trafficking and subsequent ER stress, as well as the treatment with or consumption of various chemicals and alcohol, Golgi undergoes disorganization, characterized by unstacking its membranes and abnormal glycosylation [7
]. Nevertheless, Golgi has a phenomenal self-organizing mechanism [14
], and, under drug- or stress-free conditions, Golgi is able to return to its classical ribbon-like structure and perinuclear positioning [16
]. This fact is of particular note because, despite over 100 years of Golgi research, scientists are still at a loss to explain the mechanisms of Golgi’s ability to maintain and recover harmony in such complicated structure.
It has now been more than thirty years since fungal metabolite Brefeldin A (BFA) was employed as a model for the effective blockade of ER-to-Golgi transportation and rapid Golgi disorganization [20
]. BFA-induced Golgi disassembly is reversed upon drug washout (WO) [16
], and BFA acts as an uncompetitive inhibitor that binds to the Arf1–GDP–GEF complex, thus blocking the activation of ARF1 and subsequent COPI coat assembly [21
]. Indeed, loss of COPI vesicles from the Golgi is one of the earliest responses to BFA treatment [24
]. However, knockdown (KD) of COPI coatomer proteins, including β-COP, despite Golgi disorganization, does not mimic the crucial effect of BFA, i.e., collapse of Golgi membranes and their absorption into the ER [25
]. This suggests that BFA is the drug having the potential to perform multiple tasks, especially in that it may have an effect on the structure of Golgi scaffold proteins.
The integrity of the Golgi requires structural cooperation of its matrix (golgins) and residential proteins [30
]; in turn, Golgi compact organization is likely to be a paramount condition for the successful trafficking of some critical enzymes to the Golgi [33
]. While enzymes are responsible for the processing of proteins, golgins serve as the guardians of Golgi’s monolithic architecture and docking sites for many Golgi targeting vesicles [35
]. The function of golgins is coordinated by Golgi reassembly stacking proteins (GRASPs) [40
]. We recently identified that Golgi glycosyltransferases employ two different recruiting sites at the Golgi: giantin and the complex of GM130 and GRASP65 [43
]. Giantin is the largest (376 kDa) Golgi matrix protein. It has a long (≥350 kDa) cytoplasmic N-terminal region followed by one-pass trans-membrane domain and short Golgi lumenal C-terminal domain which stabilizes dimeric giantin via a disulfide bond [44
]. It has been suggested that giantin could be essential in the cross-bridge structures that maintain Golgi morphology and it is required for efficient SNARE-mediated fusion [44
]; however, this has never been directly tested. A strong argument in defense of this hypothesis came from a study showing that, during apoptosis, giantin is more stably associated with Golgi fragments than other golgins [47
]. In addition, the Warren group, using a microsurgery approach, developed Golgi-free cytoplasts from large African green monkey cells [49
], and these cells failed in de novo Golgi biogenesis. However, when cells were treated with BFA, followed by microsurgery, even a small amount of giantin in the cytoplast was sufficient to restore Golgi membranes. In addition, it was recently shown that depletion of giantin resulted in more dispersed Golgi stacks after treatment with Nocodazole, a microtubule depolymerization agent [50
]. Thus, it appears that giantin is the core Golgi protein that drives Golgi biogenesis, but the underlying mechanism remained enigmatic.
Recently, we revealed that ethanol (EtOH) treatment blocks activation of SAR1A GTPase, thus preventing COPII vesicle-mediated Golgi targeting of protein disulfide isomerase A3 (PDIA3), the chaperone that catalyzes dimerization of giantin [13
]. In EtOH-treated hepatocytes, Golgi is consequently disorganized and Golgi targeting of mannosyl (α-1,3-)-glycoprotein beta-1,2-N-acetylglucosaminyltransferase (MGAT1), the key enzyme of N-glycosylation, is altered [51
]. In addition, we previously showed that giantin determines Golgi localization for core 2 O-glycosylation enzymes (both N-acetylglucosaminyltransferase 2/M and 1/L, C2GnT-M and C2GnT-L) [34
], and glycogen synthase kinase β (GSK3β) [52
]. Moreover, we recently found that EtOH-induced Golgi disorganization in prostate cancer (PCa) cells is accompanied by mislocalization of the N-glycosylation enzyme, N-acetylglucosaminyltransferase-III (MGAT3), again implying its sensitivity to giantin [53
]. Thus, different classes of Golgi residential enzymes seem giantin-dependent: recent observation in the giantin knockout model reveals another Golgi protein, N-acetylgalactosaminyltransferase 3 (GALNT3), which is down-regulated in cells lacking giantin [54
]. In addition, the siRNA-mediated KD of giantin results in an abnormal rate and glycosylation of some plasma membrane (PM)-associated anterograde cargo [50
]. Importantly, we also recently reported that giantin is required for post-alcohol recovery of Golgi in liver cells and PM-directed trafficking of important hepatic proteins [55
Here, using different approaches, we found that BFA-induced Golgi disorganization is associated with the monomerization of giantin. In cells recovering from BFA, giantin, but not other golgins and GRASPs, is the key protein responsible for the restoration of juxtanuclear and flattened Golgi. The fusion of the nascent Golgi membranes is assisted by Rab6a GTPase. The ability to assess a Golgi 3D-structure using high-resolution microscopy enabled us to discover the sequential Golgi targeting of residential enzymes: giantin-independent proteins filled Golgi membranes during recovery, while giantin-sensitive did so only after a complete restoration of stacks. Finally, we found that giantin is involved in the formation of “long” intercisternal communications.
2. Material and Methods
2.1. Antibodies and Reagents
The primary antibodies used were: (a) rabbit polyclonal—giantin (Novus Biologicals, Centennial, CO, USA, NBP2-22321), giantin (Abcam, Cambridge, UK, ab24586 and ab93281), NMIIA (Abcam, ab75590), and GST (Abcam, ab19256); (b) rabbit monoclonal—SAR1B (Abcam, ab155278), GAPDH (Cell Signaling Technology, Danvers, MA, USA, 14C10), Man-I (Abcam, ab140613), MGAT1 (Abcam, ab180578), GM130 (Abcam, ab52649), GRASP65 (Abcam, ab174834); (c) mouse monoclonal—HSP70 (Abcam, ab2787), NMIIB (Abcam, ab684), GRASP65 (Santa Cruz Biotechnology, Dallas, TX, USA, sc365434), β-actin (Sigma-Aldrich, St. Louis, MO, USA, A2228), giantin (Abcam, ab37266), SAR1A (Abcam, ab77029), Rab6a (Santa Cruz Biotechnology, sc-81913); and (d) mouse polyclonal—MGAT1 (Abcam, ab167365), GM130 (Abcam, ab169276). The secondary antibodies (Jackson ImmunoResearch, West Grove, PA, USA) were: (a) HRP-conjugated donkey anti-rabbit (711-035-152) and donkey anti-mouse (715-035-151) for W-B; and (b) donkey anti-mouse Alexa Fluor 488 (715-545-150) and anti-rabbit Alexa Fluor 594 (711-585-152) for immunofluorescence. Brefeldin A (EMD Chemicals, Gibbstown, NJ, USA) and MG-132 (Frontier Scientific, Logan, UT, USA) were dissolved in DMSO immediately before use. Brefeldin A was added to the cultured cells at a final concentration of 36 μM, which was followed by incubation at 37 °C for the 1 h. To study Golgi recovery after Brefeldin A treatment, cells were rinsed at least 3 times with pre-warmed drug-free medium followed by incubation under regular culture conditions for various durations as described in each experiment. NEM (N-ethylmaleimide) was obtained from Thermo Scientific (Waltham, MA, USA), dissolved in water and applied to the cell pellet in at 2 mM prior to the cell lysis buffer. All other chemicals and reagents including methanol and DMSO were of MS-grade/analytical grade and purchased from Sigma.
2.2. Immunoprecipitation (IP), Plasmid Constructions and Transfection
For identification of proteins in the complexes pulled down by IP, confluent cells grown in a T75 flask were washed three times with 6 mL PBS each, harvested by trypsinization, and neutralized with soybean trypsin inhibitor at a 2× weight of trypsin. IP steps were performed using Pierce Co-Immunoprecipitation Kit (Thermo Scientific) according to manufacturer instructions. Mouse and rabbit non-specific IgG was used as non-specific controls. All cell lysate samples for IP experiments were normalized by appropriate proteins. To determine whether the target protein was loaded evenly, input samples were preliminarily run on a separate gel with different dilutions of control samples vs. treated, and then probed with anti-target protein Abs. The intensity of obtained bands was analyzed by ImageJ software, and samples with identical intensity were subjected to IP. MYH9 (myosin, heavy polypeptide 9, non-muscle, NMIIA), MYH10 (myosin, heavy polypeptide 10, non-muscle, NMIIB), SAR1A, SAR1B, GOLGB1 (giantin), GOLGA2 (GM130), GORASP1 (GRASP65), Rab6a, and scrambled on-targetplus smartpool siRNAs were purchased from Santa Cruz Biotechnology. All products consisted of pools of three target-specific siRNAs. Cells were transfected with 100 nM siRNAs using Lipofectamine RNAi MAX reagent (Life Technologies, Carlsbad, CA, USA). PCMV-intron myc Rab6 T27N was a gift from Terry Hebert (Addgene, Cambridge, MA, USA, plasmid # 46782) [56
]. Transient transfection of cells was carried out using Lipofectamine 3000 (Life Science technologies) following manufacturer protocol. Rab6a–pCMV3–C–OFPSpark (RFP tag) was ordered from Sino Biological. GOLGB1 (giantin)–pCMV6–AC–GFP plasmids were obtained from OriGene.
The Cys3254Ser substitution in giantin protein was performed using the standard cloning and PCR-mediated site-directed mutagenesis (SDM) procedures. Given that the original plasmid GOLGB1 (giantin)–pCMV6–AC–GFP is too big for amplification (16 kb), we generated a smaller (8 kb) substrate for SDM as follows: 4 kb C-terminal fragment of the GOLGB1 from the GOLGB1 (giantin)–pCMV6–AC–GFP was subcloned into pET28b vector within EcoRV and NotI restriction sites. Then resulting plasmid pET28b-GOLGB1-C-terminus was amplified with mutagenic overlapping primers: Ser3254F (5′-GCTCATTCTGTCTTTTACGGGCCATCTAACGCGTACGCGG) and Ser3254R (5′-CCGTAAAAGACAGAATGAGCAGGACATGAATCATTAGAAAGTAGATGGCTGC). For SDM PCR we used Phusion High-Fidelity DNA Polymerase (Thermo Scientific) and cycler program: 95 °C 2′ + 15 × [95 °C 30″ + 60 °C 1′ + 72 °C (12′ + 6″)] + 72 °C 6′. After PCR completion, the PCR reaction mixture was treated with DpnI restriction enzyme (to digest methylated template) and then used to transform E. coli TOP10 strain. A positive clone was confirmed by restriction analysis and Sanger sequencing. Then, mutated plasmid was digested with EcoRV, NotI, and PvuI restriction enzymes. PvuI was used to cut pET28b backbone which has same (4 kb) size as subcloned C-terminus of the GOLGB1. A 4 kb EcoRV NotI fragment of the pET28b-GOLGB1-C-terminus-MUT was ligated with 12 kb EcoRV NotI fragment of the GOLGB1 (giantin)–pCMV6–AC–GFP. Positive clones were selected by restriction analysis and sequencing.
2.3. In Vitro Crosslinking
The protocol of crosslinking was followed according to the manufacturer’s (Thermo Scientific) instructions. Briefly, PBS-washed (three times) microsomal fraction of cells were exposed to 0.2 mM dithiobis (sulfosuccinimidylpropionate) (DTSSP) in water for 30 min at room temperature. Cross-linked protein was analyzed by SDS-PAGE under non-reducing conditions since the DTSSP cross-linker is thiol-cleavable.
2.4. Confocal Immunofluorescence Microscopy
The staining of cells was performed by methods described previously [29
]. Slides were examined under a Zeiss 510 Meta Confocal Laser Scanning Microscope and LSM 800 Zeiss Airyscan Microscope (Carl Zeiss Microscopy, Jena, Germany) performed at the Advanced Microscopy Core Facility of the University of Nebraska Medical Center. Fluorescence was detected with fixed exposure time, using an emission filter of a 505–550 nm band pass for green, and a 575–615 nm band pass for red. Images were analyzed using ZEN 2.3 SP1 software. For some figures, image analysis was performed using Adobe Photoshop and ImageJ. Statistical analysis of colocalization was performed by ImageJ, calculating the Pearson correlation coefficient [57
2.5. Three-Dimensional Structured Illumination (3D-SIM) Microscopy and Image Analysis
SIM imaging of Golgi ribbons was performed on a Zeiss ELYRA PS.1 super-resolution scope (Carl Zeiss Microscopy) using a PCO.Edge 5.5 camera equipped with a Plan-Apochromat 63 × 1.4 oil objective. Optimal grid sizes for each wavelength were chosen according to manufacturer recommendations. For 3D-SIM, stacks with a step size of 110 nm were acquired sequentially for each fluorophore, and each fluorescent channel was imaged with three pattern rotations with three translational shifts. The final SIM image was created using modules built into the Zen Black software suite accompanying the imaging setup. Analyses were undertaken on 3D-SIM datasets in 3D using IMARIS versions 7.2.2–7.6.0 (Bitplane AG, Zurich, Switzerland). The calculation of intercisternal distances was based on nearest neighbor distances to consider the Nyquist limited resolution, which in our case was around ~94 nm [58
]. The 3D mask was obtained by applying a Gaussian filter to merged channels, thresholding to remove low-intensity signals, and converting the obtained stack into a binary file that mapped all voxels of interest for coefficient calculation. For colocalization studies, IMARIS “Colocalization Module” was used. To avoid subjectivity, all thresholds were automatically determined using algorithms based on the exclusion of intensity pairs that exhibit no correlation [59
]. Colocalization was determined by Pearson’s coefficient, which represents a correlation of channels inside colocalized regions. After calculation, colocalization pixels were displayed as white. 3D animation was generated using IMARIS “Animation Module”.
2.6. AFM Imaging and Image Analysis
Giantin-GFP was isolated from DMSO and BFA-treated cells using GFP-Trap Magnetic Agarose (ChromoTek, Planegg, Germany) according to the manufacturer’s recommendations. Eluted IP samples were isolated using Millipore UFC500324 Amicon Ultra Centrifugal Filters and then dissolved in PBS for pH neutralization. Next, about 10 μL samples were treated with 2% of β-mercaptoethanol and deposited onto a piece of freshly cleaved mica. After 2 min incubation samples were rinsed briefly with several drops of deionized water and dried with a gentle flow of argon. Images were collected with MultiMode Nanoscope IV system (Bruker Instruments, Santa Barbara, CA, USA) in Tapping Mode at ambient conditions. Silicon probes RTESPA-300 (Bruker Nano Inc., Billerica, MA, USA) with a resonance frequency of ~300 kHz and a spring constant of ~40 N/m were used for imaging at scanning rate for about 2.0 Hz. Images were processed using the FemtoScanOnline software package (Advanced Technologies Center, Moscow, Russia).
2.7. In Situ Proximity Ligation Assay
The assay was performed using the Duolink kit (Sigma-Aldrich) according to the manufacturer’s protocol. The mouse monoclonal anti-giantin Ab, which recognizes the cytoplasmic N-terminus (3–91 aa) (Sigma, WH0002804M1) was used in combinations with: (a) rabbit polyclonal, which is raised against the central region of giantin (1781–1907 aa) (Sigma, HPA011008); and (b) rabbit polyclonal to giantin (region within N terminal 108–157 aa, Abcam ab93281). After incubation with primary Abs, cells were incubated with oligonucleotide-conjugated anti-mouse minus and anti-rabbit plus proximity ligation assay secondary probes. Subsequent PLA signal was detected by confocal microscopy and the corrected integrated fluorescence intensity was calculated using the ImageJ software.
2.8. Isolation of Golgi Membrane Fractions by Sucrose Gradient Centrifugation
Golgi membrane fractions were isolated using methods described previously [29
]. Cells from ten-to-twelve 75 cm2
cell culture flasks were harvested with PBS containing 0.5× protease inhibitors (1.2 mL per flask). Then, after centrifugation for 5 min at 1000 rpm and 4 °C, the pellet was resuspended in 3 mL of homogenization buffer (0.25 M sucrose, 3 mM imidazole, 1 mM Tris-HCl; pH7.4, 1 mM EDTA). Cells were homogenized by drawing ~ 30 times through a 25-gauge needle until the ratio between unbroken cells and free nuclei became 20%:80%. The postnuclear supernatant was obtained by centrifugation at 2500 rpm and 4 °C for 3 min, and then, the supernatant was adjusted to 1.4 M sucrose by addition of ice-cold 2.3 M sucrose in 10 mM Tris-HCl (pH 7.4). Next, 1.2 mL of 2.3 M sucrose at the bottom of the tube was overlaid with 1.2 mL of the supernatant adjusted to 1.4 M sucrose followed by sequential overlay with 1.2 mL of 1.2 M and 0.5 mL of 0.8 M sucrose (10 mM Tris-HCl, pH 7.4). Gradients were centrifuged for 3 h at 38,000 rpm (4 °C) in an SW40 rotor (Beckman Coulter, Brea, CA, USA). The turbid band at the 0.8 M/1.2 M sucrose interface containing Golgi membranes was harvested in ~500 µL aliquot by syringe puncture. The fraction at a concentration of ~1.0–1.4 mg protein/mL was used for the experiments mentioned in the Section 3
2.9. Zonal Sedimentation Analysis on Sucrose Gradients
Dimeric and monomeric giantin was analyzed by ultracentrifugation on layered 5–25% (wt/vol) sucrose gradients (9 mL) with a 2-mL 60% sucrose cushion prepared in 50 mM Tris-HCl (pH 7.5) and 100 mM NaCl. Isolated Golgi fractions were loaded on the top of separate sucrose gradients and centrifuged in an SW41 rotor (Beckman Coulter Inc., Fullerton, CA, USA) at 35,000 rpm for 16 h at 4 °C. Fractions of 0.25 mL were collected from the top of the tube and analyzed by immunoblotting with giantin Ab. As high molecular weight standard protein, we used an 800 kDa protein Nesprin, for which the specific band on SDS-PAGE was used for consideration of the size of giantin dimer. Densitometric analysis of the obtained band was performed using the ImageJ software.
2.10. Isolation of Microsomal Fraction
Isolation of microsomes was performed using Endoplasmic Reticulum Isolation Kit (Sigma) according to the manufacture’s protocol. Briefly, cells were suspended in Hypotonic Extraction Buffer and incubated for 20 min at 4 °C to allow the cells to swell. Then, after centrifugation at 600× g for 5 min and homogenization, the homogenate was centrifuged at 1000× g for 10 min (4 °C). The obtained postnuclear supernatant was centrifuged at 12,000× g for 15 min (4 °C), and supernatant (post mitochondrial fraction) was further subjected to isolation of rough endoplasmic reticulum (RER) using precipitation by 8 mM CaCl2.
2.11. Statistical Analysis
Measurements for the giantin KD do not follow a normal distribution as detected via the Lilliefors test at p
≤ 0.05 [60
]; hence, hypothesis testing for different medians was performed via the non-parametric Wilcoxon rank sum test [61
] for control versus giantin KD for the three parameters: cisternal length, intercisternal distance, and number of intercisternal connections. Data are expressed as mean ± SD. The rest of the analysis was performed using the two-sided t-test. A value of p
< 0.05 was considered statistically significant.
Protein concentrations were determined with the Coomassie Plus Protein Assay (Pierce Chemical Co., Rockford, IL, USA) using BSA as the standard. Densitometric analysis of band intensity was performed using ImageJ.
The ability of the Golgi apparatus to recover after severe attacks is unique and could play a significant role in cellular homeostasis. Here, we describe the role of the largest golgin giantin in the maintenance of Golgi stability and the dynamics of its membranes. We show that BFA-induced Golgi disorganization is associated with the loss of giantin dimeric structure. At first glance, it seems hard to imagine how a long dimer could unwind quickly. However, we are not the first to describe such kind of phenomenon. For example, several coiled-coil proteins, including golgin GCC185, may unwind in several domains [116
]. In addition, some dimeric myosin motor proteins may exist in their monomeric form [118
]. In addition, the induction of ER stress by the Arf1 mutant results in the relocation of GRASP55 dimer to the ER and its subsequent monomerization [111
]. Notably, in different cells and organs, EtOH-induced Golgi disorganization was associated with giantin de-dimerization [13
]. In our preliminary screening, treatment of prostate cells with the ER stress inducer, Thapsigargin, results in Golgi fragmentation, which again was accompanied by the loss of giantin dimer (manuscript in preparation). Highly aggressive PCa cells, PC-3 and DU145, which are experiencing ER stress under normal conditions, demonstrate fragmented Golgi phenotype and lack of giantin dimer [34
]. Thus, it is attempting to speculate that link between giantin monomerization and Golgi disorganization can be attributed to not only the BFA-induced cellular perturbations but also to the other clinically relevant cases of ER stress. This assumption still requires rigorous investigations.
We have shown that giantin C-terminal is critical not only for its Golgi retention but also for its dimerization and Golgi reassembly. We would not exclude the potential contribution of the coiled-coil domain of giantin to the fusion of Golgi membranes. In our preliminary screening, we observed that giantin (C3254S) construct with additional potentially critical point mutation in the coiled-coil domain (Glu1978Lys) still able to reside in the perinuclear space and recover Golgi after BFA (data not presented). However, the precise role of coiled-coil domains in the stabilization of giantin dimer still remains uncovered. Our results are echoing previous observation of Warren’s group [104
]; however, they do not fit completely with the publication of Misumi et al. [64
], who showed that truncated giantin mutant lacking C-terminal transmembrane domain (2619–3162aa) is not only able to reach the Golgi in HeLa cells but also does not block post-BFA Golgi reassembly. Since this giantin peptide lacks a disulfide-bonded lumenal domain, it seems conflicting with the data presented in our story. However, these authors used low dosage (7 µM) and short time (30 min) treatment of BFA. Using different combinations of BFA dosage and timing, we noticed that treatment of HeLa cells with 36 µM BFA for 1 h was necessary to induce complete redistribution of giantin IF signal to the cytoplasm, when cells already exhibit no perinuclear Golgi elements, but does not undergo apoptosis yet. Thus, we prefer the model of Golgi collapse, as described by Perez group, when giantin is “transported back to the ER in the presence of BFA as the Golgi ‘blinks out’” [119
]. Quite contrarily, in their figures, Misumi et al. presented BFA-treated cells that still demonstrate predominant disposition of giantin punctae in the perinuclear area.
Our results clearly indicate that Golgi biogenesis requires giantin, and this finding harmonizes well with recently published data by Stevenson et al. that the recovery of Golgi after BFA was abolished in the giantin KO model [54
]. While we observed the reduction of cisternal length in cells lacking giantin, these authors could not detect significant changes in this Golgi parameter. Moreover, another publication by Satoh et al. [120
] observed that loss of giantin elongates Golgi cisternae, contradicting their own previous observation [50
]. There are several explanations for this discrepancy. First, given that individual Golgi cisternae are complex structures that are extensively interconnected, we used 3D reconstructed images of Golgi, while authors of these two publications analyzed the mean length of cisternae from separate Z-stacks obtained by electron microscopy (EM). Second, the conclusion of these groups was based on the calculations for all Golgi cisternae, including trans
-Golgi and elongated membranes of the trans
-Golgi network. Instead, we are focused on the cis-medial
-Golgi, where GRASP65 and giantin mostly reside. Finally, the difference between our data and results from Satoh et al. can be partially ascribed to the variety of sizes found in HeLa cells. Indeed, using the EM technique, two different groups presented contradicting results regarding the effect of double knockdown of GRASP65/55 on unstacking of Golgi in HeLa cells [109
Here, we found that the reformation of Golgi membranes also depends on the activity Rab6a GTPases. The appearance of Rab6a in the Golgi coincides with giantin dimerization and Golgi reconstitution. At this point, we may assume the existence of at least two events that require the virtual involvement of giantin. First, during clustering to the perinuclear region, giantin monomers from the rims of opposite cisternae form the dimer via disulfide bond in the luminal domain. Then, giantin’s and Rab6a’s N-terminus tether to each other, thus forming dual giantin-Rab6a dimeric complexes (Figure 10
); given the ability of Rab6a to dimerize [95
], this scenario seems realistic. This, in turn, stabilizes giantin dimer via the formation of a coiled-coil structure. We believe that this is the critical step in the fusion of Golgi membranes because cells lacking Rab6a or its GTPase activity are unable to recollect Golgi membranes in the perinuclear space. Second, giantin seems indispensable for the formation of a communication between Golgi stacks. Our results also indicate that, in addition to the dimeric form, giantin may exist as an oligomer (Figure 2
). The nature of this phenomenon remains unknown and could be subject to future studies that could also address the question of whether giantin oligomeric complexes are essential for the formation of large cisterna-cisterna communications. While the function of these internal Golgi “bridges” remains elusive, it was suggested that they could play a role in intra-Golgi trafficking and maintenance of the compact shape of the Golgi [113
]. The alternative “short” intercisternal communications provided by GRASP65 oligomers seems efficient for Golgi compaction and positioning, but not sufficient for proper processing, because giantin depletion cardinally changes rate and quality of glycosylation, at least in HeLa cells [50
]. It is important to note here that, in normal cells, giantin depletion itself does not lead to the loss of Golgi centralization, because the gradual decline in giantin expression can be compensated by the oligomerization of GRASP65. However, giantin appears to be critical for the recovered cells that attempt to restore Golgi architecture after different stresses.
It is known that cells pretreated with NMIIA siRNA or its inhibitor Blebbistatin demonstrate a significant delay in BFA-induced Golgi disorganization [65
]. We have also shown that NMIIA is tethered to Golgi membranes under different drug and stress conditions, such as treatment with EtOH, heat shock, or inhibition of heat shock proteins (HSPs), and depletion of the beta-COP subunit of COPI [8
]. Thus, it seems that NMIIA is a master regulator of ER stress mediated Golgi disorganization. However, we recently found that post-alcohol recovery of compact Golgi in hepatocytes is blocked in cells depleted from non-muscle Myosin IIB (NMIIB), the isoform of NMIIA [55
]. Our preliminary data (data not shown) indicate that NMIIB is also required for post-BFA Golgi recovery, implying that intercisternal fusion requires a force that can be created by the action of NMIIB. Although there exists no direct evidence as to how NMIIB is tethered to the Golgi, our data make this a likely scenario.
The relationship between the matrix and resident Golgi proteins is increasingly viewed as a mutual alliance, where the function of some may influence the behavior of others [30
]. On the one hand, golgins and GRASPs are able to form a Golgi matrix framework regardless of the presence of Golgi enzymes [71
]. On the other hand, Golgi appears as a disassembled structure when some glycosyltransferases fail to localize to the Golgi [126
]. For example, a mutation in the membrane-spanning domain of an N-acetylglucosaminyltransferase I caused a dramatic effect on Golgi morphology [129
]. In CHO cells lacking N-acetylglucosaminyltransferase V (MGAT5), the Golgi volume density was significantly reduced [130
]; however, this was not ascribed to the lack of enzymatic function, given that in the absence (or inhibition) of its precursors, MGAT1 and Mann-II, the Golgi seems unaffected. However, we do not rule out that this possibility may be limited only to the glycosyltransferases that form complexes with cytoskeletal proteins [131
]. It is important to note, however, that the cytoplasmic tail of Golgi enzymes plays an essential role not only in their Golgi retention but that it also mediates tethering of NMIIA to the Golgi membranes, thus providing the force for alcohol- or BFA-induced Golgi disorganization [29
]. Here, we provide the evidence which indicates that the cytoplasmic domain of MGAT1 is also required for the Golgi targeting. Since the N-terminal of giantin is projecting into the cytoplasm, it is clear that the direct interaction between MGAT1 and giantin occurs outside of the Golgi lumen. However, we still do not know whether giantin, in addition to the docking function, can also serve as a Golgi retention partner for MGAT1. This possibility requires further investigation.
Our results confirm that the response of giantin to rapid ER stress is distinct from that of GM130 and GRASP65: while some golgins (including Golgin-45 and giantin) and GRASP55 can be partially detected in the ER, both GM130 and GRASP65 are still retained in the Golgi remnants [110
]. Moreover, we observed that BFA treatment does not impair the oligomerization of these proteins and provided evidence that the function of the GM130-GRASP65 complex is not critical for the Golgi recovery. This may explain why GM130-GRASP65-dependent enzymes are able to quickly refill emerging Golgi membranes. Our model suggests that giantin dimerization is a necessity for successful Golgi biogenesis. Since the formation of giantin-dimer requires functional COPII [51
] and, broadly speaking, the dynamic of Golgi membranes is linked to the integrity of ER exit site [36
], we believe that any substantial disturbance of ER function would inevitably result in Golgi dysfunction and its subsequent disorganization.
We show here that in cells lacking giantin, GRASP65 oligomerization may take a leading role in Golgi remodeling and maintenance of its positioning. However, GRASP65 consistently fails to form oligomers in different PCa cells with fragmented Golgi phenotype (our manuscript, in press). We would not exclude the possibility that cancer-specific Golgi fragmentation [136
] is influenced by the dysfunction of ER, because multiple studies observed the virtual link between ER stress and cancer [141
]. Therefore, prolonged but sub-lethal ER stress may lead to not only impairment of giantin dimerization, but also to alteration of posttranslational modification of GRASPs and other golgins [142
]. In this case, the consequences of Golgi disintegration will appear more severe than one induced by the deficiency in giantin dimerization after treatment with BFA. Lastly, most golgins and GRASPs have multiple potential N- or O-glycosylation sites, and in the case of giantin, this was preliminary confirmed [34
], suggesting their posttranslational modification may require the assistance of Golgi enzymes. Thus, it is reasonable to conclude that full maturation of Golgi matrix proteins occurs after the appearance of all resident enzymes at the Golgi membranes and complete Golgi recovery. For example, in parasitic protozoan Giardia lamblia
, Golgi was not identified in nonencysting cells; nevertheless, the induction of Golgi enzyme activities coincides with the stacking of Golgi during differentiation to cysts [143
]. Therefore, our model requires that some of the proteins (which Golgi targeting likely requires giantin) prefer to hold the occupation of membranes until complete recovery of Golgi (Figure 5
F). The appearance of MGAT1 in the reformed Golgi membranes coincides with giantin dimerization, which prompts us to speculate that the binding site of giantin for MGAT1 becomes available only after giantin dimerization. This assumption requires further experimental support. In some aspects, this conception does not fit with the idea that enzymes refill membranes during Golgi biogenesis according to their intra-Golgi (from trans
) location [80
]. However, given that early forming structures in Golgi assembly are inactive in cargo transport [81
], the holdup of giantin-sensitive enzymes seem reasonable, because, in this scenario, the accomplishment of Golgi recovery would coincide with the commencement of protein glycosylation.
If one assumes that, in mammalian cells, compact and perinuclear Golgi is essential for complete glycosylation, does it follow that any disturbance in Golgi morphology will result in the abnormal glycan processing of cargo? The answer lies in the comparison of BFA with other Golgi disruptive agents, such as Cytochalasin D or Nocodazole. Indeed, destabilization of actin by Cytochalasin D [144
] or alteration of microtubules after Nocodazole [145
] results in extensive Golgi fragmentation [29
]. Nevertheless, in these cells, the giantin-dependent enzyme, for instance, C2GnT-M, still localizes in the Golgi [29
], suggesting that these chemicals have no significant effect on the structure or function of giantin. This could explain the abnormal glycosylation in BFA-treated cells [146
] that is not found in cells incubated with either Cytochalasin D or Nocodazole [30
]. Therefore, while Golgi is anchored in juxtanuclear space inter alia by cooperation with cytoskeleton proteins, successful glycosylation is determined by the intra-Golgi location of residential enzymes rather than by the positioning of Golgi [149
]. The important question of how these separated “mini-Golgi”, induced by either Nocodazole or Cytochalasin D, communicate to maintain glycosylation of cargo remains to be answered. At first glance, clues may come from the vesicular Golgi model rather than a cisternal maturation conception; however, we are still far from a complete understanding of the nature of the many COPI- and COPII-independent vesicular and tubular complexes that have been detected in the Golgi [43
]. One of the promising arsenals for these studies is high-resolution microscopy, without which accomplishment of the current story would have been impossible.