Inhibition of DGK and PLD increases retrograde transport to the Golgi
To modulate the DAG and PA levels in the cells, we have used the DGK inhibitors R59022 (RI) and R59949 (RII) and the PLD inhibitors CAY10593 (CAY93) and CAY10594 (CAY94) (Fig. 1a). These inhibitors have been reported to have different selectivities for DGK and PLD isoforms. Both DGK inhibitors strongly inhibit the DGKα isoform; RI moderately inhibits DGKε and DGKθ, whereas RII strongly inhibits DGKγ and moderately inhibits DGKδ and DGKκ . CAY93 primarily inhibits PLD1, whereas CAY94 primarily inhibits PLD2 . To determine the effect of DAG and PA modulation on transport to the Golgi apparatus, we have taken advantage of the sulfation process that occurs specifically in the trans-Golgi network (TGN). Using modified protein toxins containing sulfation sites that can be labeled with radioactive sulfate, we can monitor retrograde transport to the Golgi. Treatment with both DGK and PLD inhibitors strongly increased sulfation of ricin-sulf1 (Fig. 1b) in HEp-2 cells, suggesting that retrograde transport is increased when DAG and PA levels are altered. Combination of the DGK and PLD inhibitors gave an additive effect, indicating that DGK and PLD inhibition may increase transport by different mechanisms. The inhibitors did not affect the sulfation process per se, as total protein sulfation was not affected. Since RI and CAY94 had a tendency for higher effect on ricin sulfation compared to RII and CAY93, respectively, we decided to focus on the DGK inhibitor RI and the PLD inhibitor CAY94 for the rest of the study.
Next, by measuring ricin sulfation at different time points ranging from 30 to 90 min, we investigated how the inhibitors affect the kinetics of the retrograde transport of ricin. We found that ricin was more rapidly transported to the Golgi after both DGK and PLD inhibition, and detectable levels of sulfated toxin could be seen already after 45 min in inhibitor-treated cells, whereas only faint bands were visible after 60 min in control-treated cells (Fig. 1c). Stronger band intensities also indicate that more ricin was able to reach the Golgi after DGK and PLD inhibition. The increase in retrograde transport of ricin is not restricted to HEp-2 cells, as we also saw a similar, although lower, increase in ricin sulfation after DGK and PLD inhibition in Caco-2, PC-3 and U-2 OS cells (Fig. 1d).
To visualize ricin transport to the Golgi, cells were treated with a 30 min ricin-sulf1-pulse and chased for 60 min in the presence of 1 mM lactose to prevent reuptake of recycled toxin. We then looked at ricin colocalization with the TGN marker Golgin-97 by immunofluorescence confocal microscopy. As the Golgi morphology is highly variable in HEp-2 cells, we used U2-OS cells for this assay. After ricin-sulf1 pulse-chase, we could see a clear perinuclear staining partially overlapping with the TGN marker (Fig. 1e), and quantification of the ricin intensity/area in the TGN (Golgin-97 positive structures) showed an increase after inhibitor treatment, in agreement with the sulfation data.
To further support the notion that retrograde transport of ricin is increased after DGK and PLD inhibition, we measured ricin transport to the ER and into the cytosol. ER transport was measured using a modified ricin molecule containing both sulfation and glycosylation sites (ricin-sulf2). N-linked glycosylation occurs in the ER and controls proper folding of proteins, and in the presence of radioactive mannose, ricin-sulf2 will be radioactively labeled when it reaches the ER. As shown in Fig. S1a, ricin mannosylation was increased after treatment with DGK and PLD inhibitors, corroborating the idea that retrograde transport is increased. Since ricin-sulf2 also contains a sulfation site, this molecule can be used to investigate transport from the Golgi to the ER when incubated in the presence of radioactive sulfate. Ricin-sulf2 will then be radioactively labeled in the Golgi and upon reaching the ER, glycosylation will increase the size of ricin-sulf2 which can be visualized by autoradiography. Comparison of the intensity of the two ricin-sulf2 bands shows that DGK and PLD inhibition has no effect on transport between the Golgi and the ER (Fig. S1b). Ricin exerts its toxic action by removing an adenine from the 60S ribosomal subunit, thereby preventing protein synthesis. We found that in the presence of DGK and PLD inhibitors, less ricin is needed to inhibit protein synthesis, resulting in a two-fold sensitization towards ricin (Fig. S1c,d). Overall, these experiments clearly show that DGK and PLD inhibition significantly increase the retrograde transport of ricin to the Golgi, the ER and the cytosol.
The increase in retrograde transport can be caused by a change in endosomal sorting or by increased internalization, and we therefore measured the endocytic uptake after DGK and PLD inhibition using 125I-labelled ricin. As shown in Fig. 2a, equal amounts of ricin were internalized, suggesting that DGK and PLD inhibition increases retrograde transport by altering endosomal sorting. We next investigated if sorting into the degradative and recycling pathways was also affected by DGK and PLD inhibition. We treated cells with a 20 min pulse of 125I-labelled ricin, followed by a 2 h chase in medium containing lactose to prevent reuptake of ricin and determined the amount of cell-associated, recycled and degraded ricin. There were no significant changes in ricin degradation or recycling after DGK or PLD inhibition (Fig. 2b). Thus, DGK and PLD inhibition seems to increase the endosomal sorting of ricin into the Golgi without affecting the recycling or degradation of the toxin. We have previously shown that transport of ricin to the Golgi apparatus is dependent on Vps34 and formation of PI3P . Thus, we tested whether the Vps34 inhibitor SAR405 could prevent RI- and CAY94-induced increase in ricin transport (Fig. S2). Indeed, SAR405 gave a significant reduction in ricin sulfation when added in combination with RI or CAY94, indicating that the inhibitor-mediated increase in ricin sulfation is dependent on Vps34 activity.
Next, we investigated if the increase in retrograde transport is specific for ricin or if other retrograde cargos are similarly affected. To this end, we studied the Golgi transport of the bacterial toxin Shiga toxin. Shiga toxin binds to the glycosphingolipid Gb3 on the cell surface and is also transported retrogradely from endosomes to the Golgi and the ER [11, 13]. We used a modified Shiga toxin molecule containing sulfation sites (StxB-sulf2) to investigate transport to the Golgi and measured sulfation in a similar manner as described above for ricin. We found that treatment with RI gave a clear increase in StxB-sulf2 sulfation, whereas treatment with CAY94 had a variable effect. In most experiments, CAY94 increased StxB-sulf2 sulfation, but generally had less effect than the DGK inhibitor (Fig. S3a). Colocalization studies between a non-toxic version of Shiga toxin (Stx1m) and the TGN marker Golgin-97 showed a similar trend of increased transport, without reaching statistical significance (Fig. S3b,c). These data suggest that the increase in retrograde transport mediated by DGK and PLD inhibitors is not restricted to ricin, but that not all pathways are changed to the same extent.
Changes in the lipidome following inhibition of DGK and PLD
Inhibition of DGK is expected to increase DAG levels while decreasing PA levels, whereas PLD inhibition can be expected to decrease PA and perhaps also DAG (Fig. 1a). To test whether this was indeed the case, we treated cells with inhibitors for 1 or 3 h and measured the changes in the cellular lipidome by MS lipidomics. As expected, after both 1 and 3 h, the DGK inhibitor RI gave a weak, but consistent, increase in DAG levels without affecting any of the other lipid classes measured (Fig. 3 and Supplementary material 2). The PLD inhibitor CAY94, however, gave a surprising transient increase in PA that was reversed after 3 h and a persistent increase in DAG and PG (Fig. 3 and Supplementary material 2). The inhibitors increased the levels of several DAG, PA and PG lipid species, with the most prominent effect on 16:0–18:1 (Fig. 4). DAG 16:0–16:0 was also increased after 1 h treatment with the PLD inhibitor and PA was increased after 1 h treatment with the PLD inhibitor. The DAG, PA and PG species that were most affected by inhibitor treatment share the same fatty acyl composition as the most abundant PC species. There were no major changes in the 18:0–20:4 species of DAG, PA and PG, which is the most abundant PI species, suggesting that PC is the source of the increased DAG, PA and PG levels (Fig. 4).
It should be noted that after 3 h of inhibitor treatment, one of the major DAG species, 16:0–16:0, is completely lost from control and RI-treated samples. During inhibitor treatment, cells are serum-starved. Our preliminary data show that incubations in the absence of serum reduced the cellular DAG levels in a time-dependent manner and that 1 h and 3 h in the absence of serum gave similar DAG levels as found in the cells treated with DMSO for 1 and 3 h (Fig. 3 and Supplementary material 2 and 3). Interestingly, treatment with the PLD inhibitor CAY94 gave similar DAG levels as in cells grown in complete medium. The DAG species most affected by removal of serum were 16:0–16:0 and 16:0–18:1 (Supplementary material 3), the same species that were most changed by CAY94 treatment (Fig. 4).
The effects of different PLD inhibitors on PLD activity and ricin transport
Given the surprising effects of the PLD inhibitor on the cellular PA level, we wanted to verify that the inhibitor actually inhibits PLD activity. To this end, we used a modified version of the recently described IMPACT (Imaging Phospholipase D Activity with Clickable alcohols via Transphosphatidylation) method to measure PLD activity in living cells . Basal PLD activity depends both on PLD1 and PLD2, while phorbol ester PMA stimulates mainly PLD1 . We tested the inhibitor efficiency on both basal and PMA-stimulated PLD activity and found that at a concentration of 10 µM, CAY94 inhibited both the basal and PMA-stimulated PLD activity, indicating that at this concentration, CAY94 inhibits both PLD1 and PLD2 (Fig. S4). Thus, the increase in PA seen in the lipidomics assay is not due to inefficient PLD inhibition. Based on our sulfation data, both CAY93 and CAY94 increased ricin transport (Fig. 1b). The IC50 values of CAY93 in cells are 11 nM for PLD1 and 1.8 µM for PLD2, and for CAY94, the IC50 values are 1 µM for PLD1 and 110 nM for PLD2 . The 10 µM concentration of CAY93 and CAY94 should therefore inhibit both PLD1 and PLD2, while at 1 µM they should be isoform specific. In agreement with the reported IC50 values, 1 µM CAY93 efficiently blocked PMA-induced PLD activity, but had only a slight reduction in basal PLD activity, whereas 1 µM CAY94 efficiently reduced basal PLD activity and had no effect on PMA-induced PLD activity (Fig. S4). At 1 µM concentrations, CAY93 and CAY94 gave only a slight increase (15–40%) in ricin-sulf1 sulfation, suggesting that both PLD1 and PLD2 need to be inhibited to strongly stimulate ricin transport to the Golgi (Fig. S5). To investigate whether PLD1 and PLD2 need to be inhibited, we also tested the dual PLD1 and PLD2 inhibitor FIPI. At 1 µM concentration, FIPI strongly inhibited both basal and PMA-stimulated PLD activity (Fig. S4), in agreement with published IC50 values in a nanomolar range for both PLD1 and PLD2 . In ricin-sulf1 sulfation assays, FIPI gave similar results as CAY93 and CAY94 when used at 1 or 10 µM (Fig. S5). The CAY inhibitors and FIPI inhibits PLD via binding to its HKD domain. We also tested a new PLD inhibitor VU 0364739 (VU036), which in addition to binding to the HKD domain, also interacts with an allosteric site of PLD and has IC50 values 1.5 µM for PLD1 and 20 nM for PLD2 in cells . At 1 µM concentration, VU036 effectively inhibited basal PLD activity but had no effect on PMA-induced PLD activity, while at 10 µM concentration, VU036 blocked both basal and PMA-induced PLD activity (Fig. S6a,b) which is in agreement with the published IC50 values . When tested in the ricin sulfation assay, VU036 gave very similar results as CAY94, with no or very little effect at 1 µM concentration and a high increase in ricin sulfation at 10 µM (Fig. S6c).
Several PLD and DGK isoforms are involved in regulating ricin transport
Based on The Human Protein Atlas, HeLa cells, which have a similar karyotype as HEp-2 cells , express four isoforms of PLD (PLD1, PLD2, PLD3 and PLD6), and seven isoforms of DGK (α, β, δ, ε, ζ, η and θ) [36, 37]. Therefore, we chose to test the expression levels of three PLD isoforms, PLD1, PLD2 and PLD3 (we did not include the mitochondrial PLD6), and six DGK isoforms, α, δ, ε, ζ, η and θ (we did not include β, which is enhanced in brain tissue  and has very low expression in HeLa cells ). Gene expression was analyzed by qPCR and we found that HEp-2 cells have high expression of PLD3 and lower expression of PLD1 and PLD2 (Fig S7), which is similar to the expression pattern of PLDs in HeLa cells. It should be mentioned that the primer efficiency for PLD2 was lower than for the other primers, leading to higher Cp values and underestimation of the expression level. For the DGK expression in HEp-2 cells, the highest mRNA level was observed for DGKδ (Fig S7), which is similar to HeLa cells. However, the expression of some DGK isoforms seems to differ between HEp-2 and HeLa cells: DGKθ is one of the abundant DGK isoforms in HeLa cells, while it was least expressed of all of the analyzed DGKs in HEp-2 cells (Fig. S7). On the contrary, we found DGKε to be the second most abundant of the six DGK isoforms tested in HEp-2 cells, while in HeLa cells, it has lowest expression of the six .
Based on the expression pattern of PLD and DGK isoforms in HEp-2 cells, we chose to knock down PLD1, PLD2, PLD3, DGKα, DGKδ, DGKε, DGKζ and DGKη by siRNA for 48 h and then analyzed ricin transport in these cells. Knockdown of PLD1 or PLD3 did not have a significant effect on ricin sulfation, while knockdown of PLD2 led to a significant reduction in the sulfation of ricin (Fig. 5a), without affecting its binding or uptake (Fig S8). The knockdown efficiency was more than 90% for all three siRNAs, and the downregulation of one isoform also affected the mRNA levels of other isoforms but to a lower extent than the targeted isoform (Fig. 5b). Since our inhibitor data indicated that more than one PLD needs to be inhibited to increase ricin transport to the Golgi, we also tested double and triple knockdown of the PLDs. Similar to knockdown of PLD2 only, combined knockdown of PLD1 and PLD2 gave a significant reduction in ricin transport, while combined knockdown of PLD1 and PLD3 led to varying increase in ricin sulfation (Fig. 5c). Importantly, the increase in ricin sulfation after the double knockdown of PLD1 and PLD3 correlates with the knockdown efficiency of PLD1 and PLD3: higher knockdown efficiency of PLD1 and PLD3 led to higher increase in ricin sulfation (Fig. 5d,e), which, together with the lack of effect on ricin sulfation after single knockdown of PLDs, shows that both of these enzymes need to be efficiently knocked down to increase ricin transport to the Golgi.
The single knockdown of all tested DGK isoforms had a significant effect on ricin sulfation: the knockdown of DGKα, DGKδ and DGKε gave an increase in ricin sulfation, while the knockdown of DGKζ and DGKη led to a reduction in ricin sulfation (Fig. 6a). The knockdown of DGKα also gave a significant reduction in total protein sulfation (Fig. 6a) and also led to increased expression of all other isoforms of the DGKs (Fig. 6b), making it difficult to say whether DGKα has a direct effect on ricin transport. Since DGKα and DGKε are the main targeted isoforms by RI , we also tested whether combined knockdown of the two could lead to an even higher increase in ricin transport to the Golgi. However, the combined knockdown of DGKα and DGKε did not give significantly higher increase in ricin sulfation than the knockdown of DGKε alone (Fig. 6c).
The DAG effectors PKC and PKD are not involved in upregulating retrograde transport after DGK and PLD inhibition
Both the DGK and the PLD inhibitor resulted in increased DAG levels, and thus we next investigated whether membrane recruitment of DAG-binding proteins could play a role in increasing sorting into the retrograde pathway. PKC is the most characterized DAG-binding protein and it has previously been shown that downregulation of PKCδ strongly reduced the retrograde transport of Shiga toxin . PKD, another DAG-binding protein, has been shown to be essential in the secretory pathway, where it activates PI4K in the Golgi apparatus and contributes to the generation of cell surface specific transport carriers [2, 40]. Activation of PKD was tested by Western blotting using an antibody that detects proteins containing phosphorylated Ser/Thr at the PKD consensus sequence. Treatment with the DGK inhibitor gave an increase in phosphorylated proteins (Fig. 7a), suggesting an increase in PKD membrane translocation and activity, in agreement with the increase in DAG seen in the lipidomics assay. Although treatment with the PLD inhibitor gave a stronger relative increase in DAG levels, it gave only a small increase in phosphorylated PKD substrates (Fig. 7a). PKC can act as an upstream activator of PKD  and treatment with the PKC inhibitors bisindolylmaleimide I (BIM) and sotrastaurin (Sotra) strongly inhibited the RI-induced phosphorylation of PKD substrates (Fig. 7a). To check whether PKC activity was important for the increased retrograde transport of ricin after DGK or PLD inhibition, we measured ricin-sulf1 sulfation after treatment with RI or CAY94 in the presence or absence of PKC inhibitors. The PKC inhibitors did not affect the increase in ricin-sulf1 sulfation after DGK or PLD inhibition (Fig. 7b,c), indicating that it is not caused by increased PKC activity. Since the PKC inhibitors also prevented RI-mediated PKD activation, we reason that the DAG effectors PKC and PKD are dispensable for the up-regulation of ricin-sulf1-sulfation after DGK- and PLD inhibition.
DGK and PLD inhibitors alter the morphology of endosomes
By altering the DAG levels, the inhibitors may also change the biophysical properties of the membrane, and a local increase in the concentration of DAG has been shown to affect both fission- and fusion processes [2, 42]. We therefore studied how treatment with the DGK and PLD inhibitors affect endosomal morphology. HEp-2 cells were treated with inhibitors and stained with antibodies against the endosomal marker EEA1 and investigated by immunofluorescence confocal microscopy. Both RI and CAY94 gave a significant increase in the size of EEA1-positive structures, and we also noticed a more irregular shape of the endosomes (Fig. 8a,b).
To investigate whether inhibitors affect biophysical properties of endosomal membranes, we used an environment-sensitive probe NR12S, which has been proven useful in analyzing plasma membrane order in living cells [43, 44]. Recently, the NR12S probe has also been employed to probe the membrane packing in the endocytic recycling compartment, as it is taken up by non-selective endocytosis and delivered to the endocytic recycling compartment . Thus we employed NR12S to analyze whether PLD or DGK inhibition affects membrane packing in the endosomes. HEp-2 cells were incubated with Alexa647-labeled transferrin to mark early and recycling endosomes and then stained with the NR12S probe and imaged using a Zeiss LSM 780 confocal microscope (Fig. S9a). The mean generalized polarization (GP) value was quantified within the plasma membrane and endosome (transferrin-positive) mask as described in the Supplementary material 1. The GP value in the transferrin-positive endosomes was higher than the GP value in the plasma membrane (Fig. S9b), indicating higher lipid packing in the endosome membrane, which is in agreement with results in . Since the NR12S probe has a slow flipping rate across the bilayer , it should mainly be localized at the outer leaflet of the plasma membrane and the luminal leaflet of the endosomes in the time frame of the experiment. As shown, inhibitor treatment did not affect the GP value at the plasma membrane (at least at the outer leaflet, which is probed by NR12S), but there seemed to be an increase in the GP value at the endosomes after PLD inhibition (Fig. S9b) possibly mediated via interaction of the two leaflets of the membrane and/or flipping of lipids, such as DAG, from the cytosolic to the luminal side of the endosomal membrane. The absolute value of GP depends on the imaging settings and cannot be compared between individual experiments without calibration; therefore we calculated the difference between the GP value at the endosomes and the GP value at the plasma membrane for each cell (delta GP). The deltaGP values represent the difference in membrane packing between the plasma membrane and the endosomes, and can be used to combine data from individual experiments. Using this approach, we saw a significant increase in deltaGP in the PLD inhibitor treated cells, but no significant change after DGK inhibitor treatment (Fig. S9c).
To study the endosomal morphology in more detail, we used an RPE1 cell line stably expressing GFP-WDFY2. WDFY2, which interacts with VAMP3, was recently described as an endosomal protein localizing to endosomal tubules and found to regulate endosomal sorting . Under conditions of mild overexpression, WDFY2 also labels the limiting membrane of the endosome, allowing us to study endosomal morphology. After treatment with RI and CAY94, we saw an increase in endosome size and clustering (Fig. 8c). To quantify this increase, Ilastik software  was used for automated pixel and object classification, allowing us to quantify the number of small, medium, large/clustered and giant/clustered endosomes, as shown in Fig. 8d. The number of small endosomes was similar in control- and CAY94-treated cells, but was reduced after RI treatment (Fig. 8e), whereas the number of endosomes with visible lumens was strongly increased after both RI and CAY94 treatment (Fig. 8e). The distribution of medium, large and giant endosomes within the subpopulation of endosomes with visible lumens was also changed after RI and CAY94 treatment. CAY94-treatment gave a larger proportion of giant endosomes than RI and control treatment, whereas both RI and CAY94-treatment increased the proportion of large endosomes (Fig. 8f). Interestingly, cells treated with the combination of RI and CAY94 seem to have the combination of the morphological changes induced by the inhibitors individually: many large endosomes that are more clustered than in cells treatment with CAY94 alone (Fig. 8c).
To test whether there is a possible link between changes in the endosome morphology and the increase in ricin transport, we analyzed WDFY2-positive endosome size in the cells treated with two different concentrations (1 µM and 10 µM) of CAY94 and FIPI. As shown in Fig. S10, at 1 µM concentration, PLD inhibitors did neither affect the number of endosomes with detected lumen, nor the endosomes size distribution, while at 10 µM concentration, both CAY94 and FIPI gave a clear increase in the number of endosomes with lumen as well as an increase in their size. It seems that the changes in the endosomal size correlate with the increase in ricin sulfation, as 10 µM is required to give a strong increase in the ricin transport (Fig. S10).
Retrograde cargo is sorted towards the Golgi from tubular endosomal structures, therefore we also investigated whether the inhibitors increased endosomal tubulation. To this end, we looked at tubule dynamics by live-cell imaging of RPE GFP-WDFY2 cells. Tubulation events were more frequent in RI and CAY94-treated cells than in control-treated cells and we could observe long tubular carriers being released from endosomes travelling through the cytosol (Fig. 9a and Supplementary material 4). For quantification, we fixed cells and performed 3D scans of the cells to capture all visible tubules (Fig. 9b). First, we determined the proportion of cells with visible tubular structures by manual inspection. After CAY94 treatment, all cells were positive for tubular structures, and these cells also had a higher number of long tubular structures (> 0.7 µm) and a higher mean length of the tubules, while treatment with RI increased only the number of long tubular structures per cell without affecting the mean length (Fig. 9c). The combined treatment of RI and CAY94 resembled treatment with CAY94 alone, but induced more clustered endosomes and thus slightly shorter tubules (Fig. 9c). Treatment with DGK and PLD inhibitors increases endosome size and tubulation, but to a different extent and lead to different morphology, indicating that these inhibitors affect the endosome morphology via different pathways.
Retrograde cargo destined to the Golgi can be segregated into tubular structures by retromer and the SNX-BAR complexes . To analyze whether ricin is transported via WDFY2-positive endosomes and/or tubules, we treated RPE-GFP-WDFY2 cells with the inhibitors and then added ricin-Alexa555 prior to cell fixation. The cells were immunolabeled for the retromer component Vps35 and the SNX-BAR component SNX2 and imaged using fluorescence super-resolution imaging. Indeed, we clearly saw ricin inside the WDFY2-positive endosomes in control and in inhibitor treated cells. In addition, some of the WDFY2-positive tubules were also positive for ricin, although the signal for ricin was always strongest within the endosomal lumen (Fig. 10). Interestingly, CAY94 treatment often led to endosomes with thick tubular tails positive for WDFY2. Based on the 3D reconstruction, these structures look as multiple tubular structures extending from the endosome. Worth noticing, ricin was often accumulated at the base of such thick tubular structures (Fig. 10). Finally, both Vps35 and SNX2 decorated WDFY2-positive endosomes and tubules and they were often found located at the same position or close to the ricin signal within the endosome (Fig. 10).