Mammalian Homologue NME3 of DYNAMO1 Regulates Peroxisome Division

Peroxisomes proliferate by sequential processes comprising elongation, constriction, and scission of peroxisomal membrane. It is known that the constriction step is mediated by a GTPase named dynamin-like protein 1 (DLP1) upon efficient loading of GTP. However, mechanism of fuelling GTP to DLP1 remains unknown in mammals. We earlier show that nucleoside diphosphate (NDP) kinase-like protein, termed dynamin-based ring motive-force organizer 1 (DYNAMO1), generates GTP for DLP1 in a red alga, Cyanidioschyzon merolae. In the present study, we identified that nucleoside diphosphate kinase 3 (NME3), a mammalian homologue of DYNAMO1, localizes to peroxisomes. Elongated peroxisomes were observed in cells with suppressed expression of NME3 and fibroblasts from a patient lacking NME3 due to the homozygous mutation at the initiation codon of NME3. Peroxisomes proliferated by elevation of NME3 upon silencing the expression of ATPase family AAA domain containing 1, ATAD1. In the wild-type cells expressing catalytically-inactive NME3, peroxisomes were elongated. These results suggest that NME3 plays an important role in peroxisome division in a manner dependent on its NDP kinase activity. Moreover, the impairment of peroxisome division reduces the level of ether-linked glycerophospholipids, ethanolamine plasmalogens, implying the physiological importance of regulation of peroxisome morphology.


Introduction
Peroxisomes proliferate by division involving elongation, constriction, and fission [1][2][3]. Peroxisome division is mediated by several factors including Pex11β, dynamin-like protein 1 (DLP1) [4,5], mitochondrial fission factor (Mff) [6,7], and Fission 1 (Fis1) [8,9] in mammals [10]. Except for Pex11β, these proteins are originally identified as fission factors of mitochondria, of NME3. Interestingly, slightly elongated peroxisomes were frequently observed in the patient's fibroblasts by immunofluorescence microscopic analysis using antibodies to peroxisomal matrix proteins including catalase as well as Pex14 (Figure 2A-C), which is ascertained by measuring the length of each peroxisome ( Figure 2E), although the elongated peroxisomes are reported to be more evident in fibroblasts from patients with a homozygous mutation in the genes encoding DLP1 and Mff [21,37,38]. The elongated peroxisomes were also observed in HeLa cells transfected with three different dsRNAs against NME3 ( Figure 3A) as quantified the elongated peroxisomes ( Figure 3C), where the level of NME3 mRNA were significantly reduced as compared with those of untreated HeLa cells ( Figure 3B). Taken together, these results suggest that NME3 is involved in the morphogenesis of peroxisomes as well. observed in the patient's fibroblasts by immunofluorescence microscopic analysis using antibodies to peroxisomal matrix proteins including catalase as well as Pex14 (Figures 2A-C), which is ascertained by measuring the length of each peroxisome ( Figure 2E), although the elongated peroxisomes are reported to be more evident in fibroblasts from patients with a homozygous mutation in the genes encoding DLP1 and Mff [21,37,38]. The elongated peroxisomes were also observed in HeLa cells transfected with three different dsRNAs against NME3 ( Figure 3A) as quantified the elongated peroxisomes ( Figure 3C), where the level of NME3 mRNA were significantly reduced as compared with those of untreated HeLa cells ( Figure 3B). Taken together, these results suggest that NME3 is involved in the morphogenesis of peroxisomes as well.  [39]. Amino-acid identity was also indicated between NME1, 2, 3, and 4 by taking NME3 as 100%. Proteins GTP-fueled by respective NME proteins are shown in the right column. Gray and solid boxes are N-terminal hydrophobic segment of NME3 and mitochondrial targeting signal of NME4, respectively. (B) Amino acid alignment of DYNAMO1 and NME3. C. merolae DYNAMO1 and human NME3 were aligned by using a Clustal W program. Identical amino acids are represented by asterisks; hydrophobic segment in the N-terminal region of NME3 is marked by an underline. The conserved NDP kinase active site is designated by a box. Amino-acid sequence used for generation of antibody is indicated by a broken-underline.  [39]. Amino-acid identity was also indicated between NME1, 2, 3, and 4 by taking NME3 as 100%. Proteins GTP-fueled by respective NME proteins are shown in the right column. Gray and solid boxes are N-terminal hydrophobic segment of NME3 and mitochondrial targeting signal of NME4, respectively. (B) Amino acid alignment of DYNAMO1 and NME3. C. merolae DYNAMO1 and human NME3 were aligned by using a Clustal W program. Identical amino acids are represented by asterisks; hydrophobic segment in the N-terminal region of NME3 is marked by an underline. The conserved NDP kinase active site is designated by a box. Amino-acid sequence used for generation of antibody is indicated by a broken-underline.

Intracellular Localization of NME3
From the finding that a reduced level of expression and the absence of NME3 cause the elongation of peroxisomes (Figures 2 and 3), we suspected that NME3 localizes to peroxisomes and mediates their fission. Mitochondrial localization of exogenously expressed NME3-HA 2 was observed in HeLa cells ( Figure 4A, lower row), as shown by expressing NME3-GFP or FLAG-tagged NME3 (FLAG-NME3) [28,29]. In addition, NME3-HA 2 -positive punctate structures were stained with Pex14, indicative of peroxisomes, where mitochondria were absent as judged by the signal of Tom20 ( Figure 4A, upper row). However, NME3-HA 2 signal was not discernible in most of peroxisomes ( Figure 4A, upper row), which might be due to a lower targeting-efficiency of NME3-HA 2 to peroxisomes and/or a high turnover of NME3-HA 2 on peroxisomes despite of the detectable expression of NME3-HA 2 as judged by immunoblotting with anti-NME3 antibody against the cytosolic catalytic domain of NME3 ( Figure 4B). Endogenous NME3 in HeLa cells was not readily detectable with this anti-NME3 antibody, apparently due to a very low level of NME3 expression [29] ( Figure 4B, lane 4).
Next, we assessed the complex formation of NME3-HA 2 with the cytosolic membrane protein receptor, Pex19, in the cytosol, which is well-characterized step during the transport of peroxisomal membrane proteins prior to targeting to peroxisomes [40][41][42]. We followed the previously established method to monitor the amount of peroxisomal membrane protein in the cytosol by the co-expression of Pex19 [41,42]. Upon expressing HA-tagged Pex26, a peroxisomal C-tail anchored protein, with FLAG-Pex19, Pex26-HA 2 was recovered more in the cytosolic fractions than that expressed alone, suggesting that the complexes of FLAG-Pex19 with Pex26-HA 2 were formed in the cytosol ( Figure 5A,C). Similarly, amount of NME3-HA 2 in the cytosolic fraction was elevated by co-expression with FLAG-Pex19 ( Figure 5B,D). Together, these results suggest that NME3-HA 2 forms a complex with Pex19 in the cytosol and targeted/transported to peroxisomes via the Pex19-Pex3 pathway.
We further assessed the intracellular localization of NME3 by expressing non-tagged NME3 in HeLa cells. As anticipated, mitochondrial localization of NME3 was discernible by the antibody raised to DYNAMO1, presumably due to a high degree of the similarity in the primary sequences between DYNAMO1 and NME3 (see Figure 1B), where the tubular network of mitochondria was not discernible ( Figure 6). In the same cells, punctate immunofluorescence signals for NME3 merged with Pex14 but not Tom20 were discernible ( Figure 6).

Intracellular Localization of NME3
From the finding that a reduced level of expression and the absence of NME3 cause the elongation of peroxisomes (Figures 2 and 3), we suspected that NME3 localizes to peroxisomes and mediates their fission. Mitochondrial localization of exogenously expressed NME3-HA2 was observed in HeLa cells ( Figure 4A, lower row), as shown by expressing NME3-GFP or FLAG-tagged NME3 (FLAG-NME3) [28,29]. In addition, NME3-HA2-positive punctate structures were stained with Pex14, indicative of peroxisomes, where mitochondria were absent as judged by the signal of Tom20 ( Figure 4A, upper row). However, NME3-HA2 signal was not discernible in most of peroxisomes ( Figure 4A, upper row), which might be due to a lower targeting-efficiency of NME3-HA2 to peroxisomes and/or a high turnover of NME3-HA2 on peroxisomes despite of the detectable expression of NME3-HA2 as judged by immunoblotting with anti-NME3 antibody against the cytosolic catalytic domain of NME3 ( Figure 4B). Endogenous NME3 in HeLa cells was not readily detectable with this anti-NME3 antibody, apparently due to a very low level of NME3 expression [29] ( Figure 4B, lane 4).
Next, we assessed the complex formation of NME3-HA2 with the cytosolic membrane protein receptor, Pex19, in the cytosol, which is well-characterized step during the transport of peroxisomal membrane proteins prior to targeting to peroxisomes [40][41][42]. We followed the previously established method to monitor the amount of peroxisomal membrane protein in the cytosol by the co-expression of Pex19 [41,42]. Upon expressing HA-tagged Pex26, a peroxisomal C-tail anchored protein, with FLAG-Pex19, Pex26-HA2 was recovered more in the cytosolic fractions than that expressed alone, suggesting that the complexes of FLAG-Pex19 with Pex26-HA2 were formed in the cytosol ( Figure  5A,C). Similarly, amount of NME3-HA2 in the cytosolic fraction was elevated by co-expression with FLAG-Pex19 ( Figures 5B,D). Together, these results suggest that NME3-HA2 forms a complex with Pex19 in the cytosol and targeted/transported to peroxisomes via the Pex19-Pex3 pathway.
We further assessed the intracellular localization of NME3 by expressing non-tagged NME3 in HeLa cells. As anticipated, mitochondrial localization of NME3 was discernible by the antibody raised to DYNAMO1, presumably due to a high degree of the similarity in the primary sequences between DYNAMO1 and NME3 (see Figure 1B), where the tubular network of mitochondria was not discernible ( Figure 6). In the same cells, punctate immunofluorescence signals for NME3 merged with Pex14 but not Tom20 were discernible ( Figure 6).      immunoblotting with antibodies to HA, FLAG, lactate dehydrogenase A (LDHA), a cytosolic protein; Tom20, a mitochondrial outer membrane protein; Pex3, a PMP. (B) NME3-HA2 was likewise expressed and assessed for its subcellular distribution as in (A). Two bands (solid and open arrowheads) were detected as in Figure 4B. (C,D), relative Pex26-HA2 and NME3-HA2 signals in cytosol fractions (S) were represented as a percentage of the total signal (S + P) (n = 3). ** p < 0.01, by Student's t-test. Figure 6. Intracellular localization of non-tagged NME3 in HeLa cells. Non-tagged NME3 was expressed in HeLa cells and stained with the antibody raised to DYNAMO1 (green). Peroxisomes and mitochondria were visualized with guinea pig anti-Pex14 (upper panels) and mouse anti-Tom20 (lower panels) antibodies, respectively. Peroxisomes are shown by a pseudo-color image. Scale bar, 10 µm. Higher magnification images of the boxed regions were shown (Inset). Scale bar, 5 µm. Arrowheads and arrows indicate peroxisomal and mitochondrial localization of NME3, respectively.

NME3 Is Elevated by Knockdown of ATAD1
NME3 and NME3-HA2 were partly localized to peroxisomes ( Figures 4A and 6), whereas fission proteins such as Mff and Fis1 are widely localized to peroxisomes [8,9,11]. Based on these results, we suspected that C-terminal tagging suppresses peroxisomal targeting of NME3-HA2 and/or the protein level of NME3 is post-translationally regulated on peroxisomes. We examined a possibility whether ATPase family AAA domain-containing protein 1 (ATAD1) regulates the turnover of NME3, because ATAD1 is known to play a role in the elimination of membrane protein mitochondrially mislocalized C-tail anchored proteins such as GOS28, peroxisomal Pex26, and Pex15 [43], maybe including other peroxisomal membrane proteins. To our surprise, the protein level, but not at the transcription level, of endogenous NME3 was elevated and peroxisomal localization of NME3 became readily discernible by knocking down ATAD1, but not in control HeLa cells ( Figure 7A,C,E-G), where the protein level of Pex14 was not altered ( Figure 7H), implying that ATAD1 is involved in regulating the expression level of NME3. The elevation of NME3 by ATAD1 knockdown was confirmed by ectopic expression of NME3 in HeLa cells, where about 60% decrease in ATAD1 mRNA and ~1.4-fold increase in NME3-1 were detectable ( Figure 7J), consistent with the profile of endogenous NME3 ( Figures 7E-G,I). The immunofluorescence signal of NME3 merged with Tom20 was discernible in Figure 6. Intracellular localization of non-tagged NME3 in HeLa cells. Non-tagged NME3 was expressed in HeLa cells and stained with the antibody raised to DYNAMO1 (green). Peroxisomes and mitochondria were visualized with guinea pig anti-Pex14 (upper panels) and mouse anti-Tom20 (lower panels) antibodies, respectively. Peroxisomes are shown by a pseudo-color image. Scale bar, 10 µm. Higher magnification images of the boxed regions were shown (Inset). Scale bar, 5 µm. Arrowheads and arrows indicate peroxisomal and mitochondrial localization of NME3, respectively.

NME3 Is Elevated by Knockdown of ATAD1
NME3 and NME3-HA 2 were partly localized to peroxisomes ( Figures 4A and 6), whereas fission proteins such as Mff and Fis1 are widely localized to peroxisomes [8,9,11]. Based on these results, we suspected that C-terminal tagging suppresses peroxisomal targeting of NME3-HA 2 and/or the protein level of NME3 is post-translationally regulated on peroxisomes. We examined a possibility whether ATPase family AAA domain-containing protein 1 (ATAD1) regulates the turnover of NME3, because ATAD1 is known to play a role in the elimination of membrane protein mitochondrially mislocalized C-tail anchored proteins such as GOS28, peroxisomal Pex26, and Pex15 [43], maybe including other peroxisomal membrane proteins. To our surprise, the protein level, but not at the transcription level, of endogenous NME3 was elevated and peroxisomal localization of NME3 became readily discernible by knocking down ATAD1, but not in control HeLa cells ( Figure 7A,C,E-G), where the protein level of Pex14 was not altered ( Figure 7H), implying that ATAD1 is involved in regulating the expression level of NME3. The elevation of NME3 by ATAD1 knockdown was confirmed by ectopic expression of NME3 in HeLa cells, where about 60% decrease in ATAD1 mRNA and~1.4-fold increase in NME3-1 were detectable ( Figure 7J), consistent with the profile of endogenous NME3 ( Figure 7E-G,I). The immunofluorescence signal of NME3 merged with Tom20 was discernible in HeLa cells with anti-DYNAMO1 antibody, indicative of mitochondrial localization, which was then diminished by the knockdown of NME3 ( Figure 7A,B). Peroxisomal localization of NME3 was observed in HeLa cells that had been treated for knocking down ATAD1, but not control HeLa cells ( Figure 7A,C). Taken together, these results suggest that NME3 is stably localized to peroxisomes in ATAD1-suppressed cells.

Peroxisomes Are Increased in Number upon Knocking down ATAD1
Peroxisomes are slightly elongated in F741 fibroblasts and HeLa cells suppressed in the expression of NME3 (Figures 2 and 3). Furthermore, in PEX11β-knocked out mouse embryonic fibroblasts (MEF) [12], ectopically expressed NME3 is observed at an apparent constriction site where DLP1 was accumulated but signal of Pex14 is weak, and the both sides adjacent to the constriction site in an elongated peroxisome ( Figure 7K). These results resemble the Pex11β-enriched constriction site formed along the elongated peroxisome devoid of Pex14 [7], suggesting that NME3 is involved in a fission step of peroxisomes. Noteworthily, the elongated peroxisomes are more evident in fibroblasts from the patients defective in DLP1 and Mff [21,37,38]. Interestingly, overexpression of Pex11β but not DLP1 or Mff promotes peroxisome division, leading to an increase in the number of peroxisomes [4,11,44]. Given these findings and together with the enhanced expression of NME3 by the ATAD1 knockdown (Figure 7), we analyzed the abundance of peroxisomes in HeLa cells. In ATAD1-knocked down HeLa cells, the number of peroxisomes was increased by~40% as compared to that in mock-treated HeLa cells, which was returned to the control level by co-suppression of NME3 ( Figure 7D,I), where peroxisomes were often elongated ( Figure 7D). These results suggest that the elevated protein level of NME3 in peroxisomes is responsible for the increase in the number of peroxisomes in ATAD1-knocked down HeLa cells. HeLa cells with anti-DYNAMO1 antibody, indicative of mitochondrial localization, which was then diminished by the knockdown of NME3 (Figures 7A,B). Peroxisomal localization of NME3 was observed in HeLa cells that had been treated for knocking down ATAD1, but not control HeLa cells ( Figures 7A,C). Taken together, these results suggest that NME3 is stably localized to peroxisomes in ATAD1-suppressed cells.   Note that NME3 was detected in the limited area of an elongated peroxisome (arrowhead). Panel (e), signal intensity of NME3, DLP1, and Pex14 in the elongated peroxisome indicated with its length in the merged view was analyzed by line scanning and represented. Note that NME3 was localized at the DLP1-accumulated potential constriction site (arrowhead in (a-d)) where signal of Pex14 is weak and the both sides adjacent to this region the constriction site.

Peroxisomes Are Elongated by Expression of Catalytically Inactive NME3
Elevation of NME3 protein level facilitated peroxisome division (Figure 7). We next examined whether NDP kinase activity of NME3 is required for the division of peroxisomes. To this end, we expressed a catalytically inactive mutant NME3, termed NME3H135N, where the catalytic site histidine at amino-acid position 135 had been mutated to asparagine (NME3H135N). Mutation of the catalytic site histidine in NME3 and DYNAMO1 abolishes their NDP kinase activity [27,29]. In wild-type NME3-expressing HeLa cells, double protein-bands were detected with anti-NME3C antibody raised against the C-terminal 18 amino-acid sequence of NME3 ( Figure 8A, lane 2). However, upon tagging of a tandem HA peptide at the C-terminus of NME3, the NME3-HA 2 protein was no longer detectable with the NME3C antibody, implying that the NME3C antibody specifically recognizes the amino-acid sequence harboring the free carboxyl group of the peptide derived from NME3 ( Figure 8A, lane 3).
The expressed level of NME3H135N was about a half of the wild-type NME3, where the NME3H135N band appeared to be with the same migration as the band with "a lower mobility" of NME3-1 ( Figure 8A, lane 4). Under this condition, peroxisomes were frequently elongated in HeLa cells expressing NME3H135N ( Figure 8B). We interpret this result to mean that NME3H135N affected division of peroxisomes by interfering with the endogenous NME3, hence suggesting that NDP kinase activity of NME3 is required for the division of peroxisomes.
site in an elongated peroxisome ( Figure 7K). These results resemble the Pex11β-enriched constriction site formed along the elongated peroxisome devoid of Pex14 [7], suggesting that NME3 is involved in a fission step of peroxisomes. Noteworthily, the elongated peroxisomes are more evident in fibroblasts from the patients defective in DLP1 and Mff [21,37,38]. Interestingly, overexpression of Pex11β but not DLP1 or Mff promotes peroxisome division, leading to an increase in the number of peroxisomes [4,11,44]. Given these findings and together with the enhanced expression of NME3 by the ATAD1 knockdown (Figure 7), we analyzed the abundance of peroxisomes in HeLa cells. In ATAD1-knocked down HeLa cells, the number of peroxisomes was increased by ~40% as compared to that in mock-treated HeLa cells, which was returned to the control level by co-suppression of NME3 ( Figure 7D,I), where peroxisomes were often elongated ( Figure 7D). These results suggest that the elevated protein level of NME3 in peroxisomes is responsible for the increase in the number of peroxisomes in ATAD1-knocked down HeLa cells.

Peroxisomes Are Elongated by Expression of Catalytically Inactive NME3
Elevation of NME3 protein level facilitated peroxisome division (Figure 7). We next examined whether NDP kinase activity of NME3 is required for the division of peroxisomes. To this end, we expressed a catalytically inactive mutant NME3, termed NME3H135N, where the catalytic site histidine at amino-acid position 135 had been mutated to asparagine (NME3H135N). Mutation of the catalytic site histidine in NME3 and DYNAMO1 abolishes their NDP kinase activity [27,29]. In wildtype NME3-expressing HeLa cells, double protein-bands were detected with anti-NME3C antibody raised against the C-terminal 18 amino-acid sequence of NME3 ( Figure 8A, lane 2). However, upon tagging of a tandem HA peptide at the C-terminus of NME3, the NME3-HA2 protein was no longer detectable with the NME3C antibody, implying that the NME3C antibody specifically recognizes the amino-acid sequence harboring the free carboxyl group of the peptide derived from NME3 ( Figure  8A, lane 3).   We further characterized NME3 by addressing the difference between the two forms of NME3, NME3-1 and NME3-2. We investigated intracellular distribution of NME3 by subcellular fractionation assay ( Figure 8C). Tom20 was mostly detected in a heavy mitochondria (HM) fraction, whereas Pex3 was mainly in a post-HM membrane fraction ( Figure 8C,D). In contrast to the differential distribution of mitochondrial and peroxisomal marker proteins, Tom20 and Pex3, respectively, NME3-1 was recovered in both HM fraction and post-HM fraction containing light mitochondria (LM) and microsome (Ms) fractions, and a small amount of NME3-1 was detected in the cytosol fraction ( Figure 8C,D). Cell-free synthesized NME3 migrated at the same position as the NME3-1 band ( Figure 8F). However, NME3-2 with a higher mobility was mostly recovered in the post-HM membrane fraction, similar to Pex3 (Figure 8C,D). Contrary to this, NME3H135N was recovered mostly in HM and post-HM fractions, similar to NME3-1, and partly in the cytosol fraction ( Figure S1A,B). Interestingly, two additional NME3H135N bands at 20 kDa, termed NME3H135N-a and NME3H135N-b, respectively, were detected in PNS fraction ( Figure S1A,B). Both proteins were mainly recovered in a post-HM fraction, suggesting that NME3H135N is likely post-translationally modified on peroxisomes ( Figure S1A,B). NME3H135N level was increased like NME3 by ATAD1 knockdown (Figure S1C), where NME3H135N and NME3H135N-b were detected. Furthermore, both of NME3-1 and NME3-2 were largely resistant to the sodium carbonate extraction, similar to Pex3, an integral membrane protein residing on peroxisomes ( Figure 8E) [45]. NME3H135N was likewise recovered in the membrane fraction upon the alkaline extraction, while both NME3H135N-a and NME3H135N-b were recovered in both membrane and soluble fractions (Figure S1D), postulating a less membrane association of the modified forms of NME3H135N. Together, these results suggest that NME3 is localized to both mitochondria and peroxisomes as an integral membrane protein, where the NME3-2 seems to be post-translationally modified on peroxisomes in a manner dependent on histidine at amino-acid position 135, thereby detectable with an apparently higher mass in SDS-PAGE.

Decreased Level of Plasmalogens in F741 Patient Fibroblasts
Biosynthesis of plasmalogens is initiated in peroxisomes [46]. The level of ethanolamine plasmalogens (PlsEtn) is severely reduced in peroxisome-deficient fibroblasts from patients with Zellweger spectrum disorders [47,48]. To investigate whether the milder morphological alteration of peroxisomes in F741 fibroblasts affects peroxisomal lipid metabolism, PlsEtn level in control and F741 fibroblasts were determined by liquid chromatography connected to tandem mass spectrometry (LC-MS/MS). Total amount of PlsEtn in F741 fibroblasts was reduced to about 50% of that in fibroblasts from a healthy control (Figure 9), suggesting that biosynthesis of PlsEtn was attenuated by the morphological alternation of peroxisomes.

Discussion
In the present study, we show that NME3 localizes to peroxisomes as well as mitochondria. Silencing of NME3 expression in HeLa cells causes the elongation of peroxisomes, which is also observed in the fibroblasts from NME3-deficient patient devoid of the expression of NME3. An

Discussion
In the present study, we show that NME3 localizes to peroxisomes as well as mitochondria. Silencing of NME3 expression in HeLa cells causes the elongation of peroxisomes, which is also observed in the fibroblasts from NME3-deficient patient devoid of the expression of NME3. An elevated protein level of NME3 by the knockdown of ATAD1 in HeLa cells leads to an increase in peroxisome number. Moreover, synthesis of plasmalogens was affected in the patient's fibroblasts. These results indicate that NME3 is involved in the morphogenesis and functions of peroxisomes.
The findings of the localization of NME3 to peroxisomes and mitochondria (Figures 4, 6 and 7) and the NDP kinase activity of recombinant NME3 [29] suggest that NME3 provides GTP to the GTP-requiring proteins localized in peroxisomes and mitochondria, as in the case of DYNAMO1 in C. merolae [27]. In mitochondria, NME3 interacts with mitofusin (MFN), a GTPase essential for mitochondria fusion [29]. However, the catalytically inactive NME3 appears to restore the impaired MFN-dependent mitochondrial elongation, suggesting that other GTPases are potential acceptor(s) for GTP provided by NME3. Our morphological results showing the elongated peroxisomes in HeLa cells silenced in the NME3 expression and the nme3 patient's fibroblasts suggest that NME3 more likely provides GTP to GTPases required for fission of peroxisomes. Interestingly, the elongated peroxisomes are formed by the expression of catalytically inactive NME3 and the elevation of NME3 expression level increases the number of peroxisomes, similar to the peroxisomal phenotype reported in the earlier studies by exogenous expression of Pex11β in wild-type cells [9,11,44,49]. Noteworthily, Pex11β functions in the formation of constriction sites in a manner dependent on DHA-containing glycerophospholipids [7] and plays as a GTPase-activating protein of DLP1 [24]. Therefore, NME3 most likely provides GTP to DLP1 for the fission of peroxisomes in mammals as well, as for the DYNAMO1 in C. merolae [27], although the interaction between NME3 and DLP1 is not observed by in vivo proximity ligation assay [29]. This is partly due to the limitation of antibodies available for the analysis of in vivo proximity ligation assay, where anti-DLP1 antibody recognizes DLP1 that shuttles between the cytosol and peroxisomes or mitochondrial outer membranes, while NME3 was detected with anti-FLAG antibody recognizing the N-terminus of FLAG-NME3 presumably located in the intermembrane space of mitochondria. We observed the elongated peroxisomes in fibroblasts and HeLa cells devoid of NME3, although the elongated peroxisomes are more evident in the fibroblasts from patients with a homozygous mutation in the genes encoding DLP1 or Mff [21,37,38]. Depletion of both NME1 and NME2 causes the accumulation of invaginated clathrin-coated pits (CCPs) with about 200 nm in length [28], while 300-400 nm-long CCPs are accumulated in dynamin-knockout cells [50,51]. Given these observations, it is conceivable that less tubular phenotype of peroxisomes and CCPs by the NME3-knockdown or -knockout cells are partly due to the utilization of cytosolic GTP. Division of peroxisomes is spatially and temporally well-regulated, hence more detailed analyses would be required for addressing the interaction of NME3 with DLP1 and the effect of NME3 on the GTPase activity of DLP1, leading to the delineation of the molecular mechanisms underlying the division of peroxisomes.
Fission of peroxisome and mitochondrion in C. merolae is highly dependent on DYNAMO1 that locates in the cytosol at G1 phase and sequentially localizes to the division sites of mitochondrion and peroxisome [27]. In mammals, NME3 is localized to peroxisomes and mitochondria as an integral membrane protein ( Figure 8E). Therefore, the recruitment of NME and DYNAMO1 at the division sites is regulated, likely in a different manner. The protein level of NME3 was elevated by the knockdown of ATAD1 (Figure 7). The role of ATAD1 in degradation of the mislocalized C-tail anchored proteins has recently been well defined [43,52,53]. The N-terminal hydrophobic segment of NME3 and digestion of the catalytic domain of NME3 by exogenously added proteinase [29] suggest that NME3 is a type I integral membrane protein localizing to peroxisomes and mitochondrial outer membrane. NME3 appears to be a potential clue, together with ATAD1, to investigate molecular mechanisms underlying the regulation in division of peroxisomes and mitochondria.
LC-MS/MS analysis revealed that the level of PlsEtn was decreased in the F741 fibroblasts ( Figure 9). Impairments of PlsEtn metabolism have also been reported in other cells with abnormal peroxisome morphology. In Pex11β −/− mouse with elongated peroxisomes, PlsEtn level in brain is reduced to about 80% of that in control level [12]. Dlp1-deficient ZP121 CHO mutant showing tubular peroxisomes and dysmorphology of mitochondria shows the impairment of biosynthesis of PlsEtn and phosphatidylethanolamine [5]. In agreement with F741 fibroblasts analyzed in this study, the peroxisomal matrix protein import is normal in both cases. Therefore, dysmorphology of peroxisomes likely affects the PlsEtn biosynthesis despite normal peroxisomal localization of enzymes for the PlsEtn synthesis. Precise mechanism underlying the attenuation of PlsEtn biosynthesis in the cells with peroxisomal dysmorphology remains to be defined. A pathogenic defect, the decreased cerebellar foliation in the NME3 patient remains undefined [29]. We recently reported a similar developmental defect of cerebellum including abnormal foliation in a peroxisome biogenesis disorder mouse model, Pex14 ∆C/∆C mouse, manifesting nearly 50% reduction of the cerebellar PlsEtn level [54,55]. Therefore, the reduced level of PlsEtn is a potential cause of the defects in cerebellar development of the NME3 patient, although the PlsEtn level in the cerebellum of the patient is not determined.

Cell Culture, DNA Transfection, and RNAi
PEX11β −/− MEF [12], HeLa cells and fibroblasts from a healthy control and an NME3-deficient patient were maintained in DMEM (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS (Sigma, St. Louis, MO, USA). All cell lines were cultured at 37 • C under 5% CO 2 . DNA transfection was performed using Lipofectamine 2000 (Invitrogen) for HeLa cells according to the manufacturer's instructions and cells were cultured for the indicated time periods.

Immunoblotting
Immunoblotting was performed as described [61]. In brief, protein samples were separated by SDS-PAGE and electrotransferred to a polyvinylidene fluoride membrane (Bio-Rad Laboratories, Hercules, CA, USA). After blocking in PBS containing 5% nonfat dry milk and 0.1% Tween 20, blots were subjected to immunoblotting with the indicated antibodies. Immunoblots were developed with ECL Western blotting detection reagents (GE Healthcare, Chicago, IL, USA), and scanned with an LAS-4000 Mini luminescent image analyzer (Fujifilm, Tokyo, Japan).

Immunofluorescence Microscopy
Cells on glass coverslips were fixed with 4% paraformaldehyde in PBS for 15 min at RT, permeabilized with 1% Triton X-100 in PBS for 2 min at RT, and blocked with PBS-BSA (PBS containing 1% BSA) for 30 min at RT. Subsequently, cells were incubated with primary antibodies indicated. Antigen-antibody complexes were visualized with Alexa Fluor conjugated secondary antibodies and observed as described [62]. Images were obtained under a laser-scanning confocal microscope (LSM710 with Axio Observer.Z1; Carl Zeiss, Oberkochen, Germany).
Number of peroxisomes per cell was determined in randomly selected cells. Peroxisomal number was calculated using the Particle Analysis package of Image J using a threshold images converted from optical images obtained by confocal fluorescence microscopy [7].

Subcellular Fractionation and Biochemical Analysis
Cells were harvested in buffer H (0.25 M sucrose, 20 mM Hepes-KOH, pH 7.4, 1 mM EDTA, and a protease inhibitor cocktail) and homogenized on ice by passing through a 27-gauge needle (with 1 mL syringe). Homogenates were centrifuged at 1000× g for 5 min to yield a post-nuclear supernatant (PNS) fraction. The PNS fraction was separated into cytosolic and organelle fractions by ultracentrifugation at 100,000× g for 30 min [63]. For subcellular fractionation, PNS fraction was centrifuged at 2500× g for 10 min to separate a heavy mitochondria fraction [11]. Post-HM fraction was further centrifuged at 100,000× g for 30 min to separate a light mitochondria and cytosol fractions.
Alkaline extraction was performed as described [64]. In brief, organelle fractions were treated with 0.1 M Na 2 CO 3 on ice for 30 min, and separated into soluble and membrane fractions by ultracentrifugation at 100,000× g for 30 min.
In vitro transcription/translation reactions in a rabbit reticulocyte-lysate were performed using the TNT T7 Quick Coupled Transcription/Translation System (Promega, Madison, WI, USA) according to the manufacturer's instruction.

Lipid Analysis
Analysis of PlsEtn in fibroblasts was performed as described [65] using 4000 Q-TRAP quadrupole linear ion trap hybrid mass spectrometer (AB Sciex, Foster City, CA, USA) with an ACQUITY UPLC System (Waters).
Supplementary Materials: The following are available online at http://www.mdpi.com/1422-0067/21/21/8040/s1, Figure S1: Intracellular distribution of NME3H135N and its membrane integrity. Funding: This work was supported in part by JSPS Grants-in-Aid for Scientific Research Grant Numbers JP26116007, JP15K14511, JP15K21743, and JP17H03675 (to Y.F.); grants (to Y.F.) from the Takeda Science Foundation, the Naito Foundation, Japan, and the Novartis Foundation (Japan) for the Promotion of Science.