JNK2 controls fragmentation of the Golgi complex and the G2/M transition through phosphorylation of GRASP65

The Golgi complex is well known for its central role in the processing and transport of proteins and lipids. In mammalian cells, the Golgi complex is organised in the form of stacks that are connected by tubular membranes, thus forming a continuous membranous structure that is referred to as the ‘Golgi ribbon’ (Shorter and Warren, 2002). During cell division the Golgi complex undergoes a multistep disassembly process that facilitates its distribution into the daughter cells (Lowe and Barr, 2007). The preparatory step for this segregation is the G2-specific cleavage of the tubular membranes connecting the stacks (Golgi ‘unlinking’) (Colanzi et al., 2007; Feinstein and Linstedt, 2007). Then, during mitosis, the stacks are fragmented into dispersed vesicles and tubulo-vesicular clusters, which are then partitioned between the daughter cells (Wei and Seemann, 2009).

The G2-specific Golgi unlinking has great functional significance as its inhibition results in cell cycle arrest in G2 (Sutterlin et al., 2002; Carcedo et al., 2004; Yoshimura et al., 2005) or in a delay of G2/M transition (Feinstein and Linstedt, 2007), depending on the experimental approach (Cervigni et al., 2011). Therefore, the partitioning of the Golgi complex is a necessary event for the cell to enter into mitosis, indicating that a ‘Golgi checkpoint’ can sense the integrity of the Golgi complex. A block of Golgi partitioning impairs G2/M transition through the inhibition of centrosome recruitment and activation of the mitotic kinase Aurora-A (Persico et al., 2010). Moreover, MEK-dependent inhibition of the Golgi-localised mitotic kinase Myt1 promotes early entry into mitosis (Villeneuve et al., 2013). Thus, these findings have highlighted the regulatory interplay between organelle inheritance and the signalling pathways that regulate cell division.

The unlinking of the Golgi ribbon in G2 is controlled by the protein CtBP1 (also known as, and hereafter referred to as BARS) (Colanzi et al., 2007) and by the Golgi stacking factors GRASP55 (also known as GORASP2) and GRASP65 (also known as GORASP1) (Lee et al., 2014). In this context, BARS stimulates the fission-mediated cleavage of the tubules that connect the stacks of the Golgi ribbon (Carcedo et al., 2004; Corda et al., 2006; Colanzi et al., 2007), and GRASP55 and GRASP65 are involved in the maintenance of the ribbon structure (Puthenveedu and Linstedt, 2004; Sutterlin et al., 2005; Tang et al., 2012; Jarvela and Linstedt, 2014). GRASP55 and GRASP65 act as membrane tethers through their conserved N-terminal halves, each of which contains two PDZ domains (the ‘GRASP domain’). The more N-terminal PDZ domain of GRASP55 and GRASP65 can homodimerise in trans with the second PDZ domain of another molecule of GRASP55 and GRASP65, respectively. This homotypic trans-oligomerisation is required for the formation of the Golgi stacks, and also for the ‘lateral’ connections between homotypic cisternae to promote the formation of the Golgi ribbon. The trans-oligomerisation is inhibited by phosphorylation events at the C-terminal halves of these proteins (for a review, see Vinke et al., 2011).

GRASP55-dependent Golgi-linking is under the control of the RAF–MEK–ERK kinase module in a protein kinase D (PKD)-dependent manner (Feinstein and Linstedt, 2007,, 2008; Duran et al., 2008; Truschel et al., 2012; Kienzle et al., 2013). However, although depletion of PKD1 and PKD2 leads to the accumulation of cells in the G2 phase of the cell cycle (Kienzle et al., 2013), chemical inhibition of MEK induces only a kinetic delay of G2/M transition (Feinstein and Linstedt, 2007), indicating that additional signalling pathways could control the Golgi unlinking process.

Less is known about the regulation of the tethering function of GRASP65 in the ribbon formation and/or maintenance during G2. During mitosis GRASP65 is phosphorylated by CDK1 at a total of four serine/threonine residues (Ser216 or Ser217, Thr220, Ser277 and Ser376), which leads to a complete disassembly of the Golgi stacks (Barr et al., 1997; Lin et al., 2000; Wang et al., 2003; Preisinger et al., 2005; Yoshimura et al., 2005). In addition, GRASP65 is also phosphorylated by PLK1 at Ser189 (Preisinger et al., 2005; Sengupta and Linstedt, 2010; Truschel et al., 2012). However, this PLK1 action requires the priming phosphorylation of GRASP65 by a yet-to-be-identified kinase (Sengupta and Linstedt, 2010).

Among these phosphorylated residues, Ser277 phosphorylation is strongly increased during the G2 phase of the cell cycle (Tang et al., 2012). Thus, we have focused on the hypothesis that phosphorylation of GRASP65-Ser277 is a key regulator of Golgi unlinking and that this phosphorylation is carried out through a signalling pathway that does not involve ERK1/2 (also known as MAPK3 and MAPK1, respectively) and CDK1. Indeed, our data show that during G2, GRASP65-Ser277 is exclusively phosphorylated by JNK2. Furthermore, this JNK2-mediated phosphorylation has a crucial role in Golgi unlinking in G2, and hence in the entry of cells into mitosis as JNK inhibitors causes a potent and persistent cell cycle block in G2. Collectively, our findings reveal a new role for this JNK2–GRASP65 axis in coupling Golgi unlinking and G2/M transition. An understanding of how Golgi unlinking is regulated is a key step towards the unravelling of novel regulatory mechanisms in cell cycle progression.

To study the role of GRASP65-Ser277 phosphorylation during G2, we generated Flag-tagged vectors for the overexpression of the regulatory domain of GRASP65 (i.e. the SPR domain, denoted GRASP65Δ200) (Sutterlin et al., 2002) and a mutant where the Ser277 of GRASP65 was replaced by alanine (Ser277Ala-GRASP65Δ200). HeLa cells were seeded on glass coverslips and synchronised in S-phase using a double-thymidine block (Colanzi et al., 2007). To induce an ‘acute’ functional block of GRASP65, the cells were transfected with Flag–GRASP65Δ200 and Flag–Ser277Ala-GRASP65Δ200 1 h after the release of the S-phase block by thymidine washout, and they were processed for immunofluorescence and western blotting at the mitotic peak (12 h after S-phase release) (Fig. 1A). Confocal microscopy analysis of the non-mitotic cells, which are mostly in G2 (Colanzi et al., 2007), showed that the structure of the Golgi complex was significantly more compact in cells overexpressing GRASP65Δ200 with respect to both the cells overexpressing Ser277Ala-GRASP65Δ200 and the control non-transfected cells on the same coverslip (Fig. 1B,C). This indicates that the effect of Flag–GRASP65Δ200 on Golgi morphology is crucially dependent on the phosphorylation of GRASP65-Ser277 (Fig. 1B,C). Moreover, western blotting showed that a fraction of Flag–GRASP65Δ200 had a reduced electrophoretic mobility compared to the non-phosphorylatable mutant (Fig. 1D), in line with the phosphorylation of GRASP65-Ser277 (Wang et al., 2003; Yoshimura et al., 2005).

Next, we determined whether the expression of these constructs affected cell entry into mitosis. HeLa cells were treated as above and, 3 h prior to fixing the cells, some of the samples were treated with either brefeldin A (BFA) or nocodazole, to induce the artificial break-up of the Golgi complex, and thus to bypass the Golgi checkpoint (Persico et al., 2010) (Fig. 1E). Transfection of Flag–GRASP65Δ200 resulted in ∼60% inhibition of the mitotic index compared to non-transfected cells, whereas the transfection of Flag–Ser277Ala-GRASP65Δ200 did not result in any alteration in the mitotic index normalised to the non-transfected cells (Fig. 1E). As evidence that Flag–GRASP65Δ200 blocks entry into mitosis as a consequence of inhibition of Golgi fragmentation, the overexpression of Flag–GRASP65Δ200 did not affect entry into mitosis in cells where the Golgi had been artificially disrupted by the BFA or nocodazole treatments (Carcedo et al., 2004) (Fig. 1E). Thus, these data show that Flag-GRASP65Δ200 impairs entry into mitosis only when Ser277 can be phosphorylated, which reveals an important role for this residue in regulating the G2/M transition.

Next, to identify the kinase involved in the phosphorylation of GRASP65-Ser277, we tested a set of candidates based on the surrounding sequence of GRASP65-Ser277, Px(S/T)P, which matches the consensus sequence for the mitogen-activated protein kinases ERK1/2, the p38 family kinases (hereafter denoted p38) and JNK family kinases (hereafter denoted JNK), and for the mitotic kinase CDK1. HeLa cells were subjected to the double-thymidine cell cycle synchronisation protocol. At 10 h after S-phase block release, the cells were treated for 2 h with dimethylsulphoxide (DMSO; as control), or with an inhibitor of each of JNK (SP600125), MEK1 (also known as MAP2K1; U0126), p38 (SB203580) and CDK1 (RO3306). The cells were then fixed and processed for immunofluorescence. The samples were stained with a well-characterised phosphorylation-specific antibody against phosphorylated GRASP65-Ser277 (Fig. 2A) (Yoshimura et al., 2005). Cells in late G2 were identified as having duplicated and separated centrosomes and uncondensed DNA (Persico et al., 2010) (Fig. 2B). Importantly, more than 90% of this G2 cell population showed clear staining of phosphorylated GRASP65-Ser277 at the Golgi complex in control cells (Fig. 2B,C, Ctrl). The treatment of the cells with the inhibitors of MEK1, p38 and CDK1 did not modify the fraction of cells positive for phosphorylated GRASP65-Ser277. By contrast, the treatment with the JNK inhibitor caused a more than 80% reduction in the fraction of cells with detectable levels of phosphorylated GRASP65-Ser277 (Fig. 2B,C), which revealed that JNK has a crucial role in this phosphorylation.

To confirm this finding using an independent approach we used two additional structurally unrelated JNK inhibitors (JNK inhibitors VIII and III; Gao et al., 2013) to monitor GRASP65-Ser277 phosphorylation in NRK cells. These cells were arrested in S phase using an aphidicolin block. At 6 h after S-phase block release, the cells were either treated with DMSO (as control) or treated for 2 h with inhibitor VIII, inhibitor III or SP600125. Western blotting of cell lysates showed that treatment with each of these three JNK inhibitors resulted in a strong decrease in the phosphorylation of GRASP65-Ser277 (Fig. 3A, upper panel). Importantly, JNK activity did not show an increased activation during G2/M (supplementary material Fig. S1A,B), but treatment with the three JNK inhibitors caused a complete inhibition of the phosphorylation of c-Jun, which is an established reporter of JNK activity (Fig. 3A, lower panel). Thus, collectively, these data show that the phosphorylation of GRASP65-Ser277 during G2 is mediated by JNK signalling. In addition, in cells arrested in mitosis by nocodazole treatment, inhibiting either JNK or CDK1 could also reduce the phosphorylation of GRASP65-Ser277 (supplementary material Fig. S1C).

Having established that JNK mediates phosphorylation of GRASP65 during G2, we investigated whether inhibition of JNK affects the Golgi morphology. NRK cells were synchronised as described above and, 2 h before fixation, the cells were treated with DMSO (as control) or with each of the three JNK inhibitors. The samples were scored by confocal analysis according to three phenotypes that were based on the structure of the Golgi complex: compact, partially fragmented, and fragmented, as previously described (Colanzi et al., 2007) (Fig. 3B,C). A compact Golgi appears rounded and in a perinuclear location. A fragmented Golgi appears to be distributed over a wider area and shows breaks that are the consequence of the loss of the connections between the Golgi stacks. A partially fragmented phenotype describes an intermediate condition that cannot be attributed to either of the other two phenotypes. Importantly, all of the three JNK inhibitors induced strong increases in the fraction of cells with a compact Golgi complex: from 23.6% in the control, to 68.9%, 79.4% and 74.7% in cells treated with JNK inhibitors VIII, III and SP600125, respectively (Fig. 3C). These data show that JNK has a key role in the regulation of the Golgi organisation.

We also tested whether the JNK inhibitors are able to induce the formation of a compact Golgi morphology in G2-arrested cells. NRK cells were arrested in G2 using a 20-h treatment with the CDK1 inhibitor RO3306 (Vassilev, 2006). The G2-arrested cells were also treated for 1 h and 4 h prior to fixing, with DMSO (control), with a JNK inhibitor (SP600125) and with inhibitors of MEK1 (U0126) and PLK1 (BI2536), which are two kinases involved in the regulation of Golgi structure (Lin et al., 2000; Feinstein and Linstedt, 2007; Villeneuve et al., 2013). Finally, the cells were fixed and processed for immunofluorescence microscopy. In the control, ∼70% of the cells had an unlinked Golgi complex, with only a minor fraction with a compact Golgi morphology (Fig. 3D,E). Importantly, JNK inhibition induced a significant increase in the compact Golgi cell population (from 6.5% to 56%), with a consequent decrease in the proportion of cells with fragmented Golgi (from 72% to 24%) (Fig. 3E). Interestingly, PLK1 inhibition also induced a detectable increase in the proportion of the cells with compact Golgi morphology, although to a lesser extent (from 6.5% to 22%), along with an increase in the partially fragmented form (from 21.5% to 43.5%). However, MEK1 inhibition induced a modest increase in the proportion of cells with compact Golgi morphology (Fig. 3E). In line with these findings, live imaging of G2/M-enriched cells showed that inhibition of JNK can induce the Golgi membranes to assume a more compact organisation (supplementary material Movie 1). As a control, in G2-blocked cells the Golgi elements became highly dynamic and underwent several aggregation and dissociation events (supplementary material Movie 2).

Next, we asked whether Golgi fragmentation can be reversed in cells that have a fragmented Golgi and that, for this feature, could bypass the Golgi checkpoint. Thus, we examined the effect of JNK inhibition in MCF7 cells, which is a tumour cell line known to have a fragmented Golgi (Kellokumpu et al., 2002) and that requires JNK signalling for its proliferation (Mingo-Sion et al., 2004), as confirmed by their higher JNK activity compared to HeLa cells (supplementary material Fig. S2A). By confocal analysis, the Golgi complex in MCF7 is fragmented into scattered elements (supplementary material Fig. S2B, DMSO). Importantly, the addition of the JNK inhibitor induced reformation of a ribbon-like Golgi complex (supplementary material Fig. S2B, SP600125).

Finally, to determine whether the compact Golgi morphology corresponds to a continuous membranous system, we investigated the Golgi ribbon integrity through analysis of the diffusion-mediated process of fluorescence recovery after photobleaching (FRAP) of Golgi-resident enzymes (Cole et al., 1996; Lippincott-Schwartz and Patterson, 2003). Golgi enzymes diffuse along the length of the Golgi ribbon, as shown previously by their fast FRAP (Cole et al., 1996). However, during G2, when the tubular connections between adjacent Golgi stacks are severed, there is reduced diffusion of these enzymes between the Golgi stacks, as reported previously (Colanzi et al., 2007).

HeLa cells were transfected with galactosyl transferase fused to GFP (GalT–GFP) and subjected to the double-thymidine synchronisation protocol. The GalT–GFP FRAP was then evaluated in a G2-enriched cell population that was represented by the non-mitotic cells 9 h after thymidine washout (Colanzi et al., 2007). Prior to the FRAP, the cells were treated for 2 h with DMSO (control) or the JNK inhibitor VIII. Then, a region corresponding to approximately half of the Golgi complex was bleached by laser illumination, and the FRAP of GalT–GFP was examined (Fig. 4A). In the control G2 cells, the FRAP of GalT–GFP (normalised to the unbleached areas) was partial and slow, which is consistent with a broken Golgi ribbon (Colanzi et al., 2007). In contrast, in cells treated with the JNK inhibitor, the GalT–GFP FRAP was greater and faster (Fig. 4B,C), which indicated that inhibition of JNK led to the maintenance of a continuous Golgi ribbon in these G2 cells.

We next examined whether JNK-dependent Golgi fragmentation is mediated by phosphorylation of GRASP65-Ser277. HeLa cells were transfected with a control non-targeting small-interfering (si)RNA or with a GRASP65-specific siRNA, for 72 h. Western blotting revealed that the GRASP65 levels were reduced by ∼70% (Fig. 5A). The cells were then synchronised using the double-thymidine protocol, and endogenous GRASP65 was replaced with Myc-tagged and siRNA-resistant wild-type (wt) or phospho-depleted (Ser277Ala-GRASP65) constructs. Finally, the cells were stained with anti-GM130 and anti-Myc antibodies. The numbers of cells with the fragmented Golgi phenotype were quantified in the absence and presence of the JNK inhibitor SP600125 (Fig. 5B). As most of the cells were in G2, ∼60% of the control non-targeting cells showed a fragmented Golgi, and this was slightly greater with the GRASP65 knockdown (∼70%). As expected, addition of the JNK inhibitor SP600125 reduced the fraction of cells with fragmented Golgi in control cells (Fig. 5C, NT+SP), but in striking contrast, this failed to restore the Golgi integrity in cells knocked down for GRASP65 (Fig. 5C, siGR65+SP). This suggested that GRASP65 is a major effector of the JNK-mediated effects on the Golgi structure. In line with this hypothesis, transfection with the siRNA-resistant wt-GRASP65 restored the induction of compaction of the Golgi complex by SP600125 in GRASP65 knocked-down cells (Fig. 5D; compare siGR65+WT with siGR65+WT+SP). Finally, the overexpression of phospho-depleted Ser277Ala-GRASP65 induced compaction of the Golgi complex in both control and GRASP65-interfered cells, and this effect was not further increased by treatment with SP600125 (Fig. 5D, compare NT+S277A with NT+S277A+SP, and siGR65+S277A with siGR65+S277A+SP). Collectively, these data indicate that prevention of GRASP65 phosphorylation recapitulates the effects of JNK inhibition and that GRASP65 is a major effector of JNK in the regulation of the structure of the Golgi complex.

Next, we asked whether JNK-dependent inhibition of Golgi fragmentation results in a block of G2/M transition. NRK cells were synchronised with aphidicolin to induce a cell cycle block at the S-phase boundary. At 4 h after aphidicolin washout, the cells were treated with each of the three JNK inhibitors (VIII, III, SP600125), and the mitotic index was analysed at several times after the S-phase block release, as previously described (Feinstein and Linstedt, 2007). Importantly, all of these three JNK inhibitors resulted in strong and persistent reduction in the mitotic index (Fig. 6A).

To determine whether inhibition of JNK blocks the cell cycle because of the activation of the Golgi checkpoint, we investigated whether the artificial fragmentation of the Golgi complex by treatment with BFA can rescue mitotic progression. Thus, synchronised NRK cells were treated with SP600125, as previously described, in the absence and presence of BFA for 4 h. The cells were then fixed and processed for fluorescence microscopy, to determine the mitotic index through the staining patterns of histone H3 phosphorylated at Ser10 (pH3) (Persico et al., 2010) (Fig. 6B). In addition, the inhibition of JNK reduced the mitotic index compared to the control cells, whereas in cells treated with BFA, the JNK inhibitor failed to promote cell cycle block (Fig. 6C).

To confirm this finding using an independent approach, we forced the Golgi complex into the form of isolated stacks by performing siRNA-mediated GRASP65 depletion (Puthenveedu et al., 2006). HeLa cells were transfected with control non-targeting siRNA or GRASP65-specific siRNA and synchronised using the double-thymidine protocol. Finally, the cells were incubated in the absence and presence of the JNK inhibitor SP600125 for 2 h prior to fixing them at the mitotic peak. Depletion of GRASP65 did not affect the overall mitotic index of these synchronised cells in comparison to the control. Importantly, the addition of SP600125 blocked the entry into mitosis for control cells, but not for the cells with an already fragmented Golgi (GRASP65-depleted) (Fig. 6D). As shown above, in MCF7 cells the Golgi is fragmented, but the treatment with JNK inhibitors induces the reformation of an intact Golgi ribbon. Thus, we tested the effect of SP600125 on the G2/M transition rate of MCF7 cells. Our results indicate that inhibition of JNK induced a remarkable inhibition of entry into mitosis (supplementary material Fig. S2C), indicating that the Golgi checkpoint can be activated also in tumour cells with a constitutively fragmented Golgi complex.

Overall, these findings establish a causal link between JNK-mediated fragmentation of the Golgi complex and cell cycle progression.

The JNK MAPK family includes the JNK1 and JNK2 subfamilies, which are ubiquitous, and the JNK3 subfamily (for a review, see Waetzig and Herdegen, 2005). As this last isoform was not detectable by western blotting and real-time quantitative PCR in HeLa cells, we excluded it from our investigation.

HeLa cells were depleted of JNK1 and JNK2 isoforms by siRNA treatments (Fig. 7A), and subjected to the double-thymidine synchronisation protocol. JNK1 depletion caused a minor reduction of the phosphorylation of GRASP65-Ser277. By contrast, JNK2 depletion strongly reduced this phosphorylation (Fig. 7B), suggesting that the JNK2 isoform is the one preferentially involved in the phosphorylation of GRASP65-Ser277. To test whether JNK2 can directly phosphorylate GRASP65-Ser277, recombinant JNK2 was incubated with the recombinant GRASP65Δ200 and the Ser277Ala-GRASP65Δ200 mutant. JNK2 phosphorylated Ser277 of GRASP65Δ200, but not the mutant (Fig. 7C). This finding agrees with the concept that the various JNK isoforms have clear substrate specificities (Bogoyevitch and Kobe, 2006), and it is also in line with previous data reporting that JNK1 cannot phosphorylate GRASP65 (Yoshimura et al., 2005).

The roles of the two JNK isoforms were then analysed in terms of the regulation of the Golgi structure. For this, HeLa cells depleted of JNK1 and JNK2 by siRNA treatments (see above) were processed for immunofluorescence using an anti-GM130 antibody to reveal the organisation of the Golgi membranes (Fig. 7D). Note that here the cells were scored according to three phenotypes, as above: for fragmented, partially fragmented, and compact Golgi structures (see Fig. 3B). This confirmed a major role for JNK2 in Golgi fragmentation, as in cells silenced for JNK2, the Golgi complex showed a more compact organisation compared to control cells (Fig. 7E). In addition, there was a 12-fold increase in the population of cells with mitotic Golgi clusters during metaphase for JNK2-depleted cells compared to non-targeting-siRNA-treated cells (Fig. 7F,G), indicating that in JNK2-depleted cells, the mitotic Golgi fragmentation was not complete. These data are supported by the observation that inhibition of JNK caused an increase in the oligomerisation state of GRASP65 (supplementary material Fig. S3).

Finally, to define the role of JNK depletion in cell cycle progression, we first checked whether a JNK isoform was localised at the Golgi complex. Interestingly, JNK2, but not JNK1, was localised at the Golgi complex (supplementary material, Fig. S4A). Then, HeLa cells that were depleted of JNK1 or JNK2 by siRNA treatments (see above) were fixed at 9 h or 13 h after the second thymidine washout (i.e. at around the mitotic peak, and after it, respectively) and processed for fluorescence-activated cell sorting (FACS) analysis (supplementary material Fig. S4B,C). In all of the samples fixed at 9 h, progression into the G2/M phase of cell cycle was not altered (supplementary material Fig. S4B). However, in the samples fixed at 13 h, although the control cells and the JNK1-depleted cells had exited mitosis, as expected, the JNK2-depleted cells had accumulated in G2/M (supplementary material Fig. S4C, bottom panel). As shown in supplementary material Fig. S4D, the depletion of JNK2 resulted in significant reduction in the mitotic index (control, 16.9%; JNK1 depletion, 13.9%; JNK2 depletion, 2.6%), which indicated that the accumulation in G2/M revealed by FACS analysis was mainly caused by G2 block of the cell cycle. Taken together, these data show that JNK2 is required for G2-specific phosphorylation of GRASP65 on Ser277, and that this phosphorylation is in turn necessary for Golgi ribbon unlinking and entry of cells into mitosis.

The main finding of this study is that a JNK2–GRASP65 signalling axis is the major regulator of Golgi ribbon organisation during G2, and of the G2/M transition of the cell cycle. Using a panel of biochemical and imaging-based assays, we have shown that JNK2 is required for phosphorylation of GRASP65-Ser277, and that this phosphorylation is necessary for Golgi fragmentation during the late G2 phase of the cell cycle, and for cell entry into mitosis.

The importance of the JNK2–GRASP65 signalling axis in the Golgi mitotic checkpoint is highlighted by the inhibition of JNK2 using three structurally unrelated specific inhibitors and the depletion of JNK2 by RNA interference, which resulted in inhibition of Golgi fragmentation and G2 arrest of cell cycle progression. Artificial fragmentation of the Golgi membranes by BFA or GRASP65 depletion abrogated the requirement for JNK2, which indicates that JNK2 exerts its effects on cell cycle progression through the regulation of the cleavage of the Golgi ribbon. Our data also suggest that GRASP65 is the major JNK2 target at the Golgi membranes, as the single point mutation of Ser277 to Ala can recapitulate the effects on the Golgi structure of the inhibition of JNK2. Overall, our data show that JNK-mediated phosphorylation of GRASP65-Ser277 has a major regulatory role in Golgi organisation during G2.

Our identification of a role for JNK2 in the Golgi mitotic checkpoint is supported by the Golgi-localisation of JNK2 itself and of possible upstream activators, such as Mekk4 (also known as MAP3K4) and MLK3 (also known as MAP3K11) (Du et al., 2005; Millarte and Farhan, 2012). In addition, in a functional screening of kinases and phosphatases, JNK2 was categorised into a group of kinases that are necessary for the regulation of cell migration in a Golgi-dependent manner (Farhan et al., 2010). Finally, several reports have shown that the JNK pathway has an important role during G2/M transition through the phosphorylation of key mitotic regulators, such as the phosphatase CDC25C and the histone H3 (MacCorkle-Chosnek et al., 2001; Lee and Song, 2008; Oktay et al., 2008; Gutierrez et al., 2010a,,b). However, our results show that the JNK2–GRASP65 signalling axis has a prominent role in the regulation of G2/M transition, because even in the presence of JNK inhibitors, artificial fragmentation of the Golgi complex restores entry into mitosis.

Our findings also support the proposal that GRASP65 has a key role in maintaining the continuity of the ribbon through its capability of forming large oligomers. Although mouse embryonic fibroblasts derived from mice knocked down for GRASP65 show an apparent normal organisation of the Golgi ribbon (Veenendaal et al., 2014), probably as a result of adaptive mechanisms during embryonic development, a large body of experimental evidence has clearly proven the key role of GRASP65 in the regulation of the Golgi structure. Indeed, many experimental approaches, including depletion of GRASP65 by siRNA treatment (Puthenveedu and Linstedt, 2004; Sutterlin et al., 2005; Tang et al., 2012), or the acute depletion of GRASP65 (Jarvela and Linstedt, 2014), have shown that functional disruption of GRASP65 results in the unlinking of the Golgi ribbon. Importantly, although the breakdown of the Golgi ribbon caused by depletion of a maintenance factor, such as GRASP65, is irreversible (Fig. 5B), the unlinked Golgi complex organisation that can be seen in tumour cell lines with overactive JNK can be fully reversed by treatment with JNK inhibitors (Fig. 3D,E; supplementary material Fig. S2B).

An additional finding of the present study is that different signalling pathways can phosphorylate GRASP65-Ser277, depending on the physiological context: by ERK1/2 upon growth factor stimulation and wound induction (Yoshimura et al., 2005; Gaietta et al., 2006; Bisel et al., 2008) and by JNK2 during G2. Interestingly, we also find that that both JNK and CDK1 appear to be involved in the phosphorylation of Ser277 GRASP65 during mitosis. However, this interplay between JNK and CDK1 during mitosis requires further investigation. The common theme of this structural reorganisation is the breakup of the Golgi ribbon, which can be transient, as in the case of Golgi reorientation during cell polarisation induced by wound healing (Bisel et al., 2008), or more extensive, as during G2 in preparation for cell entry into mitosis (our data). Thus, the extent of Golgi reorganisation might be a function of the level of phosphorylation of GRASP65-Ser277, and of the combination with additional modifications, such as the phosphorylation in mitosis that involves at least four distinct sites (Preisinger et al., 2005), which would lead to Golgi unstacking, and thus complete Golgi fragmentation.

Thus, an important question is how the activity of these kinases is regulated. The MAP kinase signalling module is composed of three-tiered sequentially acting kinases: a MAPK, a MAPK kinase (MAPKK), and a MAPK kinase kinase (MAPKKK). Specificity of the signalling is ensured by the temporary and local formation of a signalling module of specific composition (directed by a ‘scaffold’). The identification of further elements of the module will help in understanding how JNK2 is directed to induce Golgi disassembly even in the absence of a global rise of JNK activity.

In conclusion, the data presented here have several implications that are related to both the mechanisms of Golgi ribbon formation and the interplay between Golgi membrane structure and/or function and cell cycle regulation, as the definition of the molecular mechanisms that regulate Golgi unlinking during G2 should also help in the identification of additional targets for blocking cell cycle progression in anti-cancer therapies.

Normal rat kidney (NRK) and HeLa cells were from the American Type Culture Collection (Manassas, VA, USA) and were cultured in Dulbecco's modified Eagle's medium and minimal essential medium, respectively (Invitrogen, Carlsbad, CA), supplemented with 10% fetal calf serum (Biochrom, Berlin, Germany), 100 μM minimal essential medium non-essential amino-acid solution, 2 mM L-glutamine, 1 U/ml penicillin, and 50 μg/ml streptomycin (all Invitrogen). MCF7 cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, 1 U/ml penicillin and 50 μg/ml streptomycin. All of the cell lines were grown under a controlled atmosphere in the presence of 5% CO₂ at 37°C.

Aphidicolin, thymidine, nocodazole, BFA and SP600125 were from Sigma-Aldrich. DMSO was from Carlo Erba. RO-3306, JNK inhibitors III and VIII, and Mowiol were from Calbiochem. SB203580 was from Tocris Bioscience. U0126 was from Promega. Hoechst 33342 was from Invitrogen. BI2536 was from Axon Medicine BV (Groning, The Netherlands).

The antibodies were from the following sources: anti-phospho-histone H3 (Ser10) from Millipore (Billerica, MA); anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and anti-GM130 from BD Biosciences (San Jose, CA); anti-GRASP65 (pS277) from N. Nakamura, Kanazawa University Kakuma, Japan; anti-Flag-M2 from Sigma-Aldrich (Milan, Italy); anti-pericentrin and anti-GRASP65 from Abcam; anti-Myc tag, anti-phosphorylated ERK1/2 (pERK1/2), anti-phosphorylated Jun (pJun), anti-c-Jun, anti-JNK1 and anti-JNK2 from Cell Signaling Technology; and anti-Plk1 from BioLegend. Alexa-Fluor-488-, Alexa-Fluor-633- and Alexa-Fluor-546-conjugated secondary antibodies were from Invitrogen.

For construction of N-terminally Flag-tagged GRASP65Δ200, subcloning of R. norvegicus GRASP65 cDNA was carried out, with amplification by PCR with the following oligonucleotides, which contain EcoR1 and EcoR5 sites, followed by insertion into the pcDNA3-3XFLAG plasmid (Sigma): forward, 5′-CGAGAATTCGGATCCCAACGCAGCCC-3′; reverse, 5′-CGCGATATCTCACAACCCAGGCTCTGG-3′. A point mutation of GRASP65-Ser277 into Ala was inserted by digestion with DnpI endonuclease of the PCR amplification with the following primers: forward, 5′-CTCCCTGGGCCTGGGGCTCCTGGCCATGGCACT-3′; reverse, 5′-GAGGGACCCGGACCCCGAGGACCGGTACCGTGA-3′.

The cDNA of rabbit full-length GFP-GRASP65 was from Dr Yanzhuang Wang (University of Michigan, Ann Arbor, USA). The cDNA of human full-length Myc-GRASP65 wt was from Dr Christine Sütterlin (University of California, Irvine, USA). The phospho-depleted mutant Ser277Ala was obtained through mutagenesis with the following primers: forward, 5′-CTTCCTGGGCCTGGGGCTCCCAGCCACAGTGCT-3′; reverse, 5′-GAAGGACCCGGACCCCGAGGGTCGGTGTCACGA-3′. The phospho-mimetic Ser277Asp was obtained through mutagenesis with the following primers: forward, 5′-CTTCCTGGGCCTGGGGATCCCAGCCACAGTGCT-3′; reverse, 5′-GAAGGACCCGGACCCCTAGGGTCGGTGTCACGA-3′. The siRNA-resistant forms of the original construct were obtained with the following primers: forward, 5′-GAATGACACCCTGAAAGCACTGCTGAAAGCAAATGTGGAGAAGCCCGTGA-3′; reverse, 5′-CTTACTGTGGGACTTTCGTGACGACTTTCGTTTACACCTCTTCGGGCACT-3′. The cDNA of GalT-GFP was from Dr Lippincott-Schwartz (NIH, Bethesda, MD, USA).

HeLa cells were transfected with the TansIT-LT1 Transfection Reagent (Mirus), according to the manufacturer instructions.

The siRNA duplexes were transfected using Lipofectamine 2000 (Invitrogen), according to the manufacturer instructions. GRASP65 was targeted using siRNA duplexes directed against the sequence 5′-CTGAAGGCACTACTGAAAGCCAAT-3′ (Sigma). JNK1, JNK2 and JNK3 were targeted using siRNA duplexes directed against the sequences 5′-GTGGAAAGAATTGATATATAA-3′, 5′-AAGAGAGCTTATCGTGAACTT-3′ and 5′-CCGCATGTGTCTGTATTCATA-3′, respectively. After transfection, the intracellular protein contents were assessed by SDS polyacrylamide gel electrophoresis followed by western blotting, and the cells were further processed according to the experimental design.

HeLa cells were grown on 10 μg/ml fibronectin-coated glass coverslips (Sigma-Aldrich) and treated as described above. They were then fixed with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) for 10 min at room temperature. The blocking reagent (0.5% bovine serum albumin, 0.1% saponin, 50 mM NH₄Cl) was then added to the cells for 20 min, followed by a 1-h or overnight incubation with the primary antibody in the blocking reagent. The cells were then washed with phosphate-buffered saline (PBS) and incubated with the secondary antibodies (1:400) with 2 μg/ml Hoechst 33342.

For morphological and mitotic index analysis of synchronised cells, HeLa cells were treated as previously described (Colanzi et al., 2007). For the G2 cell synchronisation, the cells were maintained for 20 h in the presence of 9 μM RO3306. The mitotic index analysis of the time-course experiments was performed as described previously (Feinstein and Linstedt, 2007). FRAP experiments were performed as previously described (Colanzi et al., 2007) with the difference that calls treated with JNK inhibitor VIII for 2 h, until the G2/M peak. The half-lives (t_½) of the control and inhibitor-VIII-treated cells were assessed, to compare the relative recovery rates between the samples.

Trypsinised cells were pelleted, washed in cold PBS, and resuspended in ice-cold ethanol while vortexing. The cells were incubated overnight at 4°C. The next day, the ethanol was removed by centrifugation and the cells were washed in cold PBS and incubated with 50 μg/ml propidium iodide (Invitrogen) for 30 min in the presence of RNase (Sigma). The cells were then analysed using a FacsCanto instrument (BD). The data were plotted with Diva software, with 30,000 events analysed for each sample.

A reaction was performed using 1 μg JNK2 kinase incubated with 5 μg substrate in 25 μl kinase buffer (5 mM MOPS, pH 7.2, 5 mM MgCl₂, 2.5 mM β-glycerophosphate, 1 mM EGTA, 400 μM EDTA, 0.05 mM dithiothreitol) at 30°C for 15 min in the presence of 300 μM ATP (Amersham Biosciences). The reactions were terminated by addition of SDS-PAGE sample buffer, and the products were resolved by SDS-PAGE and stained with Ponceau Red to check the loading amounts, and blotted with the antibody against phosphorylated GRASP65-Ser277. GST–JNK2 was purchased from SignalChem. To prepare GST–GRASP65Δ200 and Ser277Ala-GRASP65Δ200, the rat GRASP65Δ200 and the Ser277Ala-GRASP65Δ200 cDNAs were cloned into pGEX-4T2 (GE Healthcare Life Sciences); correct insertion was confirmed by DNA sequencing. The proteins were expressed in BL21 bacteria and purified on glutathione–Sepharose 4B (GE Healthcare Life Sciences), following the manufacturer's protocol.

The authors would like to thank Raman Parashuraman and Alberto Luini for helpful discussion and for providing us with materials. Dr Franck Perez is thanked for the generous gift of the R3G3 antibody. We would also like to thank Chris Berrie for critical reading and editorial revision of the manuscript.