Depletion of the human N-terminal acetyltransferase hNaa30 disrupts Golgi integrity and ARFRP1 localization

The organization of the Golgi apparatus (GA) is tightly regulated. Golgi stack scattering is observed in cellular processes such as apoptosis and mitosis, and has also been associated with disruption of cellular lipid metabolism and neurodegenerative diseases. Our studies show that depletion of the human N-α-acetyltransferase 30 (hNaa30) induces fragmentation of the Golgi stack in HeLa and CAL-62 cell lines. The GA associated GTPase ADP ribosylation factor related protein 1 (ARFRP1) was previously shown to require N-terminal acetylation for membrane association and based on its N-terminal sequence, it is likely to be a substrate of hNaa30. ARFRP1 is involved in endosome-to-trans-Golgi network (TGN) traffic. We observed that ARFRP1 shifted from a predominantly cis-Golgi and TGN localization to localizing both Golgi and non-Golgi vesicular structures in hNaa30-depleted cells. However, we did not observe loss of membrane association of ARFRP1. We conclude that hNaa30 depletion induces Golgi scattering and induces aberrant ARFRP1 Golgi localization.


Introduction
The structure and functionality of the Golgi apparatus (GA) is maintained by at least four systems: microtubule-associated transport proteins, the actin-associated cytoskeleton, Golgi matrix proteins and proteins involved in targeting and fusion of vesicles such as GTPases and SNARE proteins [1]. Disruption of any of these systems can lead to changes in GA organization, such as for example GA collapse or disassembly and fragmentation [2][3][4][5]. GA fragmentation is found in physiological processes such as mitosis, apoptosis and organelle trafficking [1,6,7], as well as disruption of cellular lipid metabolism, amyotrophic lateral sclerosis, Alzheimer's disease and Creutzfeldt-Jacob's disease [1,[8][9][10].
The small GTPases of the ADP ribosylation factor (Arf) family are regulators of membrane traffic [11]. The Arfs bind to their target membranes through an N-terminal membrane anchor and an N-terminal amphipathic helix. Typically, the Arfs contain a glycine in the second position and are thus subjected to N-myristoylation [12,13]. Interestingly, a subset of the Arfs are not myristoylated, but rather N-terminally acetylated (Nt-acetylated) [14].
Arf-related protein 1 (ARFRP1) is an Arf that binds GA-and trans-Golgi network (TGN) membranes in its GTP-bound state [15,16]. The yeast homologue of ARFRP1, Arl3, requires protein N-α-terminal acetylation (Nt-acetylation) for correctly targeting the GA [17,18]. ARFRP1 has been suggested to function in TGN-to-plasma membrane transport and endosome-to-TGN transport and is needed for recruitment of human ADP ribosylation like factor 1 (hArl1) to Golgi compartments [2,16,19]. hArl1 recruits various organelle-organizing factors such as arfaptins and golgins to GA and TGN membranes [20,21]. Depletion of ARFRP1 in HeLa cells induces dislocation of TGN protein Syntaxin6, but it does not lead to disturbance in cis-Golgi markers Golgin subfamily A member 2 (130-kDa cis-Golgi matrix protein (GM130), GOLGA2) and Giantin [2]. The physiological and cellular importance of ARFRP1 are underscored by the observation that Arfrp1 −/− mice die at the embryonic stage [22].
The NatC complex is one of several Nt-acetyltransferases (NATs) that perform Nt-acetylation in eukaryotes. Nt-acetylation or protein N-α-terminal acetylation, is the addition of an acetyl group on the Nα-amino group of proteins. It is one of the most abundant protein modifications in eukaryotes and displays a wide array of biological functions [23,24].
In the present study, we show that depletion of the hNatC catalytic subunit hNaa30 leads to disassembly of the GA and TGN. We further show that ARFRP1 shifts from a GA and TGN localization in control cells, to localize to smaller vesicle-like membranous compartments in hNaa30-depleted cells. On the basis of these findings, we conclude that hNaa30 is required for GA and TGN integrity and normal ARFRP1 distribution.

Results
hNaa30 knockdown leads to scattering of GA and TGN in HeLa and CAL-62 cell lines hNaa30 was depleted by siRNA-mediated knockdown of hNAA30 gene expression. In order to ensure that knockdown phenotypes are specific for hNaa30 depletion and not a result of si-hNAA30-independent effects, two hNAA30-specific siRNAs that target different regions of the hNAA30 transcripts were used for all the experiments. Western blot analysis was routinely performed to confirm knockdown efficiency of the siRNA constructs on hNaa30 protein levels ( Figure 1). siRNA knockdown will not give a complete depletion of protein levels, and all results need to be evaluated as consequences of protein reduction rather than complete loss of protein.
To investigate a potential effect of hNaa30 depletion on the GA and TGN, we compared the localization of cis-Golgi proteins GM130 and Giantin and the TGN-protein Syntaxin6 in sihNAA30 and siCONTROL treated (siCTR) HeLa cells using immunofluorescence microscopy. In siCTR-cells, GM130, Giantin and Syntaxin6 took a perinuclear stack-like localization, resembling the classical Golgi ribbon (Figure 2A,C,E). A minor fraction of the cells displayed a dispersed, but still compartmentalized localization of GM130, Giantin and Syntaxin6. In hNaa30-depleted cells, a significantly larger fraction of cells displayed the dispersed localization of GM130, Giantin and Syntaxin6 than was observed in the control (Figure 2A-F). Dispersion was not observed as a discrete phenotype, rather cells displayed various degrees of fragmentation. The cutoff for scoring was set to cells having one continuous Golgi ribbon or not.
As the GA disperses during mitosis and apoptosis, visible mitotic or apoptotic cells were not counted. To investigate whether the observed cis-Golgi fragmentation is specific for HeLa cells or whether it might represent a more general role of hNaa30, we depleted hNaa30 in anaplastic thyroid carcinoma CAL-62 cells and investigated the localization of GM130. Also in CAL-62 cells, we observed an increase in GM130-fragmentation after hNaa30 depletion (Supplementary Figure S1). CAL-62 cells contain a non-functional form of p53 [34]. In cells with functional p53, hNaa30  depletion leads to p53-dependent apoptosis [27]. Therefore, the observation of GM130 dispersion after hNaa30 depletion in CAL-62 cells supports that the observed cis-Golgi dispersion is not a result of apoptotic events. We conclude that depletion of hNaa30 leads to fragmentation of GA and TGN, independent of mitotic or apoptotic GA dispersion.
GA integrity is dependent on ER-Golgi transport, microtubule integrity and microtubule-associated motor proteins, proteins facilitating vesicular transport and actin cytoskeleton. ER morphology, β-tubulin or actin architecture is not affected by hNaa30 depletion [31]. Disruption of ER-to-Golgi transport has been previously shown to give an ER-like localization of Golgi markers [35]. The scattered appearance of GM130 and Giantin after hNaa30 depletion did not resemble an ER-like localization and costaining of hNaa30-depleted cells with ER-marker Calnexin and cis-Golgi marker Giantin revealed that there was no relocalization of Giantin to the ER (Supplementary Figure S2). It is therefore not likely that hNaa30 induces cis-Golgi dispersion through disruption of ER-to-Golgi traffic.

Depletion of hNaa30 leads to a shift in ARFRP1 localization to non-GA compartments and loss of TGN localization
To further elaborate the role of hNatC catalytic subunit hNaa30 for GA and TGN functions, we investigated endogeneous ARFRP1 localization after hNaa30 depletion. We confirmed that ARFRP1 at least partly colocalized with cis-Golgi markers GM130 ( Figure 3A) and Giantin ( Figure 3B) and TGN marker Syntaxin6 ( Figure 3C), as has been shown previously [16,18]. In hNaa30-depleted cells, a reduction in ARFRP1-GM130 colocalization was observed, as compared with the control ( Figure 3A). The degree of colocalization in hNaa30-depleted cells varied from partial colocalization to a total loss of colocalization. ARFRP1 still colocalized with Giantin in sihNAA30-depleted cells, but in addition, ARFRP1 localized to compartments that did not contain Giantin ( Figure 3B). ARFRP1 lost colocalization with TGN-marker Syntaxin6 in hNaa30-depeted cells ( Figure 3C). In addition, ARFRP1 localized to smaller, vesicle-like compartments in siCTR cells. In sihNAA30-treated cells, this localization pattern was more pronounced (Figure 3A-C). We did not observe a loss of membrane association of ARFRP1 in hNaa30-depleted cells by immunofluorescence microscopy.
To further clarify the role of Naa30 for membrane attachment of ARFRP1, hNaa30-depleted cells were lysed and organelles were sedimented by centrifugation. ARFRP1 cosedimented with the organelle-enriched pellet both in control-and sihNAA30-treated cells ( Figure 3D). Nt-acetylation can function as a degradation signal in a ubiquitin-dependent manner [36]. However, we did not observe any consistent reduction in ARFRP1 protein level in hNaa30-depleted cells ( Figure 3D).

ARFRP1-Y2P-GFP but not ARFRP1 depletion induces GA scattering
We then hypothesized that loss of ARFRP1 from the GA might be responsible for the fragmented GA phenotype in hNaa30-depleted cells. We used a pool of four different siRNA oligos to deplete ARFRP1 ( Figure 4A). GA morphology was then visualized by GM130 staining ( Figure 4B) and cells with scattered GA morphology were quantified. Knockdown of ARFRP1 did not affect GA morphology ( Figure 4C). Thus, the GA-dispersion effect of hNaa30 depletion is not due to loss of ARFRP1 per se.
To investigate whether Nt-acetylation affected ARFRP1 localization and function, we generated an ARFRP1-Y2P-GFP mutant. N-termini holding a proline in the second position are not Nt-acetylated [37,38]. ARFRP1-GFP displayed a somewhat less punctuated localization than what was observed for the endogeneous protein. This might be an artefact from the overexpression or tagging of the protein. However, GA localization was clear. ARFRP1-Y2P-GFP displayed a predominantly cytoplasmic localization in HeLa cells ( Figure 4D), suggesting that the acetylated N-terminus is important for localization. Of note, cells overexpressing ARFRP1-Y2P-GFP displayed a scattered GM130 localization ( Figure 4D).

ARFRP1-Q79L-FLAG does not lose GA localization in hNaa30-depleted cells
ARFRP1 binds to target membranes in its GTP-bound state [16]. To investigate whether Nt-acetylation affects ARFRP1 localization dependent on GTP-or GDP-bound state, we transfected sihNAA30-treated cells with ARFRP1-FLAG or the constitutively GTP-bound mutant ARFRP1-Q79L-FLAG and co-immunolabelled cells with antibodies directed towards the FLAG epitope and GM130. ARFRP1-FLAG colocalized with GM130 in siCTR cells and this colocalization was partly lost in hNaa30-depleted cells ( Figure 5A,C). ARFRP1-Q79L-FLAG colocalized with GM130 to a higher degree than the wild-type ( Figure 5B,C) and maintained this colocalization in sihNAA30-treated cells. A small reduction in colocalization of ARFRP1-QL-FLAG was seen for one of the hNAA30targeting oligos, but this was minor compared with the changes seen in the wild-type. Thus, in its active form, ARFRP1 is not dependent on hNaa30 for GA localization.

Discussion
hNaa30 is the catalytic subunit of the human NatC complex. Several studies have pointed towards a possible role of NatC in organelle traffic. In the present study, we show that depletion of hNaa30 leads to scattering of the cis-Golgi and TGN in HeLa and the cis-Golgi in CAL-62 cells. GA disassembly is observed as a normal part of apoptosis and mitosis [1]. Since hNatC depletion does not lead to accumulation of mitotic cells [27], it is not very likely that the observed GA fragmentation is due to an accumulation of mitotic cells after hNaa30 depletion. Since the GA takes a fragmented appearance during apoptosis, no cells with fragmented nuclei were counted. Further, hNaa30 depletion mediated apoptosis is p53 dependent and is therefore not expected in CAL-62 cells. From this, we conclude that the observed increase in GA and TGN fragmentation is independent of mitosis-or apoptosis-specific GA fragmentation.
Also another human NAT, Naa60, was recently found to be important for GA integrity [39,40]. Although Naa30 and Naa60 display partially overlapping substrate specificities in vitro and in a yeast model [39,41], Naa60 Nt-acetylates a distinct group of transmembrane substrates in vivo [40]. Thus, the Naa30-KD and Naa60-KD GA phenotypes are likely to be mediated via different substrates and different pathways.
In a recent work, we showed that while hNaa30 depletion had profound effects on mitochondrial architecture, it had no effect on ER, endosome, peroxiosome, β-tubulin or actin architecture [31]. hNaa30 depletion does therefore not induce general changes in cellular or organellar architecture and the effects are thus specific for GA and mitochondrial compartments. Whether these phenotypes are functionally connected remains to be determined. Disruption of ER-to-GA anterograde transport will lead to fusion of the GA with the ER [35]. As we did not see merging of the cis-Golgi marker Giantin with the ER-marker Calnexin in hNaa30-depleted cells, it is not likely that hNaa30 depletion scatters GA through disturbing the ER-to-Golgi traffic. The dispersion of syntaxin6 could be due to disturbances in retrograde endosome-to-TGN traffic. As mentioned above, hNaa30 depletion did not have any effect on endosomal morphology, but further studies are needed to determine the role of hNaa30 in retrograde traffic in more detail.
ARFRP1 is a GTPase involved in endosome-to-TGN trafficking. ARFRP1 mutants display disruption of the TGN. ARFRP1 was earlier suggested to be an Naa30-substrate, where acetylation is needed for ARFRP1 membrane targeting through recruitment to hSys1 [17,18,33]. We show that ARFRP1 localization is indeed altered after hNaa30 depletion, from a predominant GA and TGN localization, to also being localized to smaller non-Golgi compartments and losing TGN localization. hNaa30 depletion might disturb retrograde endosome-to-TGN traffic and lead to endosomal accumulation of ARFRP1. This needs to be investigated further in future studies.
We did attempt to immunoprecipitate ARFRP1 in hNaa30-depleted cells for analysis of Nt-acetylation by mass spectrometry, but unfortunately we were not able to obtain a satisfactory pulldown for analysis. To circumvent this, we took advantage of the observation that a proline in the second position blocks Nt-acetylation. A non-acetylatable ARFRP1-Y2P-GFP mutant did not localize the GA compartments. This suggests that hNaa30 is the NAT responsible  for Nt-acetylation of ARFRP1. Our previous observations that hNaa30 could functionally replace yeast Naa30 suggest that this is an evolutionarily conserved function [41].
According to our observations, it is the localization of ARFRP1 within the endomembranous system rather than the membrane attachment per se that is affected by hNaa30 depletion. Two hypotheses have been made about the role of Nt-acetylation for membrane-associated GTPases: Nt-acetylation as a lipid anchor and Nt-acetylation mediates protein-protein interactions with membrane recruiting factors [14,17,18]. Our observations are in accordance with the latter hypothesis. This may also give an indication to why a group of Arfs is Nt-acetylated instead of N-myristoylated: rather then binding to the membrane, they localize to their target membranes through protein-protein interactions. But as NatC is not a general determinant of substrate cellular localization, this is likely to be specific for each substrate [42].
Behnia et al. [18] previously showed that that the GDP-locked ARFRP1-T31N mutant of ARFRP1 was still recruited to target membranes by hSys1 in COS cells. Shin et al. [16] showed that the GTP-locked mutant ARFRP1-Q79L-FLAG predominantly localized to GA compartments, while the GDP locked mutant protein ARFRP1-T31N-FLAG predominantly localized to the cytoplasm in HeLa cells. We find that ARFRP1-Q79L-FLAG GA localization is unaffected by hNaa30 depletion.
We further wanted to clarify whether the siNAA30-induced GA dispersion was due to loss of ARFRP1 from target membranes. Surprisingly, ARFRP1 depletion did not induce GA dispersion, while cells overexpressing the non-acetylatable Y2P mutant displayed a scattered GA.
Based on this, we suggest a model where GDP-bound but not GTP-bound ARFRP1 requires Nt-acetylation for targeting GA membranes. Loss of Nt-acetylation could lead to accumulation of ARFRP1 in non-GA compartments, as observed in hNaa30-depleted cells. In this model, the GA dispersion observed after ARFRP1-Y2P-GFP overexpression but not ARFRP1 depletion could be due to a dominant-negative effect of the Y2P mutant. However, care must be taken when comparing overexpression systems directly with siRNA experiments.
In the present study, we have shown that hNaa30 indeed affects both GA and TGN integrity and subcellular distribution of the Golgi-associated GTPase ARFRP1. This has potentially large implications for cellular membrane traffic, secretion, lipid biogenesis and protein maturation. Taken together with our recent findings that hNaa30 is important for mitochondrial integrity, the present work shows the importance of hNaa30 and NatC in intracellular organization.

Cell culture and transfection
HeLa cells (epithelial cervix adenocarcinoma; A.T.C.C. no. CRL-1573) and CAL-62 cells (anaplastic carcinoma, 8305C DSMZ number ACC 448) were cultured and transfected as described previously [27]. Cells were harvested 72 h post siRNA transfection. siRNA-mediated knockdown was performed using Dharmafect 1 transfection reagent (Dharmacon) according to the instruction manual. Gene-specific siRNAs were purchased from Dharmacon and used at a final concentration of 50-100 nM to silence hNAA30. Two different siRNAs targeting hNAA30 were used to ensure that phenotypes were specific for hNAA30 depletion: sihNAA30-1 Dharmacon cat. number D-009961-01 and sihNAA30-2 cat. number D-009961-05. For ARFRP1 knockdown, a pool of four different siRNAs were used (Dharmacon cat. number L-019250-00-0010). Non-targeting siRNApool cat. number D-001810-10 was used as a negative control. Knockdown efficiency was inspected routinely by Western blotting for all experiments. Plasmid transfection was performed using Roche X-tremeGENE9 transfection reagent. In experiments where cells were subjected to both siRNA and plasmid transfection, 20 μM carbobenzoxy-VAD (O-methyl)-fluoromethylketone (z-VAD-fmk) pan-Caspase inhibitor (R&D Systems Europe Ltd.) was added to avoid cell death.

Cell lysis and organelle sedimentation
Cells were harvested by scraping and sedimented at 2000 × g for 5 min. Total cell lysates were prepared by resuspending cell pellets in total lysis buffer (50 mM Tris/HCl, pH 8, 50 mM NaCl, 0.5% Nonidet P40, 5 mM EDTA, 1 mM Na 3 VO 4 and 1 mM Pefabloc (Roche)) and incubated for 5 min on ice. Cell membranes were removed by centrifugation at 15700 × g for 1 min and the cell lysate was transferred to a new tube. Organelle sedimentation was performed using a modification of previously described methods [43]. Cells were harvested and resuspended in a hypotonous KSHM-buffer (100 mM potassium acetate, 85 mM sucrose, 20 mM Hepes-KOH, pH 7.4, 1 mM Mg acetate, Phosphatase Inhibitor Cocktails 2 & 3 (Sigma-Aldrich) and Complete Protease Inhibitor (Roche)). Resuspended cells were snap frozen in liquid N 2 and centrifuged at 1500 × g for 5 min. Supernatant was collected, pellet was resuspended in KSHM buffer, snap freezing and centrifugation was repeated. Supernatants were pooled after snap freezing and centrifuged at 25000 × g for 30 min. The supernatant (S) was analysed by immunoblotting ( Figure 3D), while pellet was pooled with organellar pellet. The pellet was re-suspended in KSHM-buffer after snap freezing and centrifuged at 25000 × g for 30 min. The resulting pellet was enriched for organelles. Pellets were resuspended in total lysate buffer as described above to remove membranes and remaining nuclei.

Immunofluorescence
HeLa or CAL-62 cells grown on cover slips were washed in PBS, fixed in paraformaldehyde or methanol, permeabilized in 0.1% Triton X-100 and blocked in 10% BSA. Proteins of interest were labelled with primary antibodies as indicated in the figures. Secondary antibodies were Alexa Fluor 488-, Alexa Fluor 594-or Alexa Fluor 555-conjugated IgGs (Invitrogen). Blue Hoechst 33342 or DAPI staining was used to stain the nuclei. Images were acquired using a Leica DMI 6000b microscope or a Zeiss 510 Meta confocal microscope, as indicated. Where indicated, microscopic recordings were processed by de-convolution (Leica 4000 software). Z-stacks and Z-stack projections were handled using the Fiji Image Processing Software. The data for quantification of GM130, Giantin and Syntaxin6 localization after hNAA30 knockdown are shown as the mean of at least 100 cells counted in three independent samples.

Statistics
Data for quantification of phenotypes from micrographs were analysed using SPSS statistical software packaging. Student's t test was used to compare the mean percent of defined phenotypes between hNAA30-targeting siRNAs and the non-targeting control siRNA. Significance level was set at 5% (95% confidence intervals) for all analyses. For colocalization analysis, Mander's colocalization coefficient was calculated for at least 75 individual cells per condition using the Fiji Image Processing Software [44].