Lipid Droplet‐Localized Spindle Apparatus Coiled‐Coil Protein 1 Regulates Lipid Droplet Distribution
Honggang Su, Huimin Pan, Yaqiang Liu, Wei Wang, Yang Jiao, Yumo Hang, Xiahe Huang, Yong E. Zhang, Yuhang Chen, Mei Ding, Xun Huang
TL;DR
This study identifies a protein that helps organize lipid droplets in cells by linking them to the microtubule network, influencing their distribution and cell division.
Contribution
The paper discovers SPDL1-L as a novel lipid droplet-associated protein that regulates their transport and clustering via dynein.
Findings
SPDL1-L localizes to lipid droplets through hydrophobic and basic residues.
SPDL1-L promotes lipid droplet clustering at the microtubule-organizing center.
SPDL1-L uses dynein to mediate minus end transport of lipid droplets.
Abstract
Lipid droplets (LDs) are central to cellular energy homeostasis and lipid metabolism. While LD‐associated proteins are known to regulate LD dynamics and function, their roles in LD transport and their broader cellular significance remain poorly understood. In this study, we identify lipid droplet‐localized spindle apparatus coiled‐coil protein 1 (SPDL1‐L) as a key regulator of intracellular LD transport. SPDL1‐L localizes to LDs through the synergistic action of its hydrophobic region and nearby basic residues. SPDL1‐L expression promotes LD clustering at the microtubule‐organizing center (MTOC) and the formation of donut‐shaped (toroidal) nuclei during cell division. SPDL1‐L‐mediated LD clustering is likely through dynein‐mediated minus end transport. LD clustering activity is diminished when the LD targeting or dynein binding function of SPDL1‐L is impaired. Together, our findings…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6- —Key Technologies Research and Development Program10.13039/501100012165
- —National Natural Science Foundation of China10.13039/501100001809
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsLipid metabolism and biosynthesis · Microtubule and mitosis dynamics · Protein Degradation and Inhibitors
Introduction
1
A neutral lipid core, a phospholipid monolayer and various associated proteins constitute a lipid droplet (LD) [1]. With their central role in neutral lipid storage and energy metabolism, LDs are dynamically formed, transported and degraded in response to metabolic demands and environmental conditions. Imbalances in LD dynamics are closely associated with a range of diseases, including diabetes, lipodystrophy, liver disease, neurodegeneration, and more [2].
In general, LDs exhibit two distinct distribution states, clustered and dispersed, in cells. The distribution is quite dynamic across different phases of the cell cycle and under different nutrient statuses. In NIH 3T3 cells, LDs show increased dispersion during the S phase [3]. In HuH7 human hepatoma cells, LDs are positioned outside the spindle during metaphase [4]. During nutrient starvation, LDs translocate to the cell periphery in a microtubule‐dependent manner, adopting a dispersed distribution in Vero cells. This promotes the interaction between LDs and mitochondria, thereby facilitating mitochondrial β‐oxidation [5]. In Drosophila embryos, Klar regulates LD transport by modulating the relative activities of kinesin‐1 and dynein. Loss of Klar alters this balance, misdirecting LDs to the yolk cell instead of the peripheral epithelium [6]. Except for the molecular motors and the cytoskeleton, the mechanisms regulating LD distribution in different cells have not been well explored. Few molecular linkers between LDs and the cytoskeleton have been reported. Plin2 plays a crucial role in regulating the distribution of LDs by interacting with microtubules through dynein and kinesin motors. In the presence of Plin2, LDs are evenly distributed. However, when Plin2 is displaced by core protein loading in HCV‐infected HuH7 cells, the motor balance is disrupted, causing LDs to move toward the microtubule‐organizing center (MTOC) and the nuclear periphery [7]. Plin3 interacts with Dync1i1 to facilitate dynein‐dependent lipid export from the ER, suggesting a role for Plin3 as an adaptor for neutral lipid transport during lipogenesis in hepatocytes [8]. A human long non‐coding RNA, Lipid‐Droplet Transporter (LIPTER), binds to phosphatidic acid and phosphatidylinositol 4‐phosphate on the surface of LDs. LIPTER interacts with MYH10 to link LDs to actin, thereby promoting their transport and playing a key role in lipid metabolism in the human heart [9]. Uncovering new molecular adaptors between LDs and the cytoskeleton may offer new insights into LD dynamics.
Besides the well‐studied LD‐resident Perilipin (PLIN) proteins, canonical LD‐associated proteins are usually involved in lipid metabolism. For example, CCT1, GPAT4 and DGAT2 function in lipid synthesis; ATGL, HSL and CGI58 control lipid breakdown; and Rabs and ARF1 regulate lipid trafficking [10]. In addition, many LD‐associated proteins which do not function in lipid metabolism have also been identified. This greatly expands the cellular and physiological functions of LDs. The discovery of isoforms localized to LDs in proteins previously not linked to LDs further broadens the catalog of known LD‐associated proteins. For example, the M1 isoform of Spastin, a classic microtubule‐severing protein, is responsible for LD dispersion [11, 12]. Spindle apparatus coiled‐coil protein 1 (SPDL1), commonly known as Spindly, is a dynein adaptor crucial for accurate chromosome segregation during mitosis [13]. In human cells, the short isoform (Spindly or SPDL1‐S) localizes predominantly to the nucleus during interphase and to kinetochores (KTs) during mitosis [14]. Our recent study revealed that a long isoform of chicken SPDL1 (SPDL1‐L) localizes to LDs and regulates their positioning in chicken cone cells [15]. However, the conservation and the functions of SPDL1‐L in other organisms, as well as the LD targeting mechanism of SPDL1‐L remain completely unexplored.
In this study, we investigated the localization and function of human SPDL1‐L in cultured cells. Our results demonstrate that the targeting of the SPDL1‐L to LDs is governed by both its hydrophobic region and nearby basic residues. Moreover, we found that SPDL1‐L regulates LD transport in a dynein‐dependent manner toward the minus end of microtubule. Functional impairment of SPDL1‐L reduces LD clustering. Additionally, overexpression of SPDL1‐L leads to the formation of donut‐shaped nuclei (also called toroidal nuclei) during mitosis. Together, our findings reveal that LD‐localized SPDL1 regulates LD distribution.
Results
2
The Alternative Polyadenylation Generated SPDL1‐L is Conserved
2.1
To identify SPDL1‐L‐like proteins in different organisms, we performed a BLASTP search of the NCBI database using the 67 amino acids specific to the long isoform of chicken SPDL1 (aa 560–626 of chicken SPDL1‐L). A total of 185 candidates were selected based on the following criteria: 1) they were the longest isoform from each species, and 2) they displayed strong homology, defined by an E‐value of < e‐5. Among them, 116 are from mammals, 47 from birds, 21 from reptiles, and 1 from an amphibian. The phylogenetic tree, constructed based on the protein sequences from 27 representative species, shows that closely related species tend to cluster together in the same branch (Figure 1A; Table S1).
SPDL1‐L, an isoform derived via alternative polyadenylation, is conserved. (A) Phylogenetic tree of representative SPDL1‐L‐like proteins. The numbers in square brackets indicate the number of SPDL1‐L‐like proteins within each class. (B) Generation of two isoforms of human SPDL1 via alternative polyadenylation. Top: Schematic of the human SPDL1 gene, showing the location of the two polyadenylation (pA) sites, one within the intron between exons 11 and 12, and the other in exon 12. Bottom: Use of the distal (exon 12) pA site generates SPDL1‐S, while use of the intronic pA site generates SPDL1‐L. E11a refers to an additional part of exon 11 when the intronic pA site is used. (C) The AlphaFold 3 model of the human SPDL1‐S and SPDL1‐L. Coiled‐coil regions are rendered as green helices, and selected residues are labeled for orientation. A boxed region denotes the C‐terminal segment (residues 558–622) subjected to detailed analysis. Selected residues are annotated by physicochemical classification: polar (magenta), hydrophobic (blue), arginine (red), and prolines (yellow). (D) Secondary structure predictions of the C‐terminal of SPDL1‐L were made using Phobius (https://phobius.sbc.su.se/). (E) A summary of XL‐MS data reporting intramolecular crosslinks of SPDL1‐S and SPDL1‐L is provided. For clarity, only crosslinks between residues that are at least 20 amino acids apart are shown. A comprehensive list of all crosslinks can be found in Table S2.
We analyzed the structure of the SPDL1 locus in the human genome. Structural analysis of SPDL1 transcripts from the NCBI and PolyASites databases (https://polyasite.unibas.ch/) reveals that human SPDL1‐S and SPDL1‐L are generated through alternative polyadenylation (APA). The use of polyadenylation sites in the last intron (between exons 11 and 12) and at the distal end of the gene (in exon 12) results in the production of SPDL1‐L and SPDL1‐S, respectively (Figure 1B). SPDL1‐S encodes a 605 aa protein (SPDL1‐S) and SPDL1‐L encodes a 622 aa protein (SPDL1‐L). They share the same aa 1–557 region. Notably, SPDL1‐L is present in humans and in several rodent species, but is absent in mice (Figure 1A; Table S1). In the mouse Spdl1 locus, a 25× tandem repeat of a 31‐bp sequence of unknown origin is inserted into the putative Spdl1‐L‐specific coding sequence, resulting a premature stop codon (TGA) within exon 11, which may explain the absence of SPDL1‐L in mice (Figure S1). The evolutionary conservation of SPDL1‐L suggests that it may play a conserved role in a wide range of vertebrates, including human.
SPDL1‐S has long dynein‐dynactin binding coiled‐coils regions at the N‐terminus and a farnesylated C‐terminal region which binds to the ROD‐Zwilch‐ZW10 (RZZ) complex [16]. We used AlphaFold 3 to compare the structural features of SPDL1‐S and SPDL1‐L. Both SPDL1‐S and SPDL1‐L adopt a structure with four long helices (Figure 1C). The C‐terminal region of SPDL1‐L contains two helices separated by two proline residues (P597 and P602). The first helix is predominantly hydrophobic, containing hydrophobic amino acids such as phenylalanine (F), leucine (L), and isoleucine (I) (Figure 1C,D). That said, the possibility of an amphipathic character cannot be ruled out, as hydrophilic residues (e.g., serine (S), threonine (T)) are positioned on the same side of the helix, suggesting potential interactions with aqueous environments.
Crosslinking mass spectrometry (XL‐MS) revealed that SPDL1‐S adopts a compact intramolecularly folded conformation, characterized by multiple long‐range contacts between distinct coiled‐coil segments [16]. We compared the intramolecular interaction patterns of SPDL1‐S and SPDL1‐L by XL‐MS. For both SPDL1‐S and SPDL1‐L, we detected crosslinks between CC1a and CC1b, extensive interactions of CC1a with CC2 and CC3, and additional contacts between CC2 and CC3 (Figure 1E). In addition, there are extensive crosslinks between the sequence spanning residues 558–572 of SPDL1‐S with CC1a and CC3, in line with the requirement of RZZ engagement to relieve autoinhibition and enable SPDL1‐S activation. Similar intramolecular interaction was not found in SPDL1‐L, likely because this region differs substantially between the two isoforms. Together, these findings imply that SPDL1‐L may share the structural basis for dynein adaptor function, although its regulatory mechanism likely differs from that of SPDL1‐S.
SPDL1‐L Targets LDs in a Partially Seipin‐Dependent Manner
2.2
Given the conserved structural features of SPDL1‐L, we next determined its localization. We expressed N‐terminal mEGFP‐tagged SPDL1‐L in HeLa cells. Under normal conditions, SPDL1‐L colocalized with the ER marker BFP‐Sec61β as well as LipidTOX‐stained LDs. After treatment with oleic acid (OA), SPDL1‐L was almost exclusively localized to LDs (Figure 2A). To identify the LD targeting mechanisms of SPDL1‐L, we expressed the SPDL1‐L‐specific region (aa 558–622) with an N‐terminal mScarlet tag and found that it is localized to LDs (Figure 2B). To further delineate the LD‐targeting region, we constructed a series of mScarlet‐tagged truncations of the SPDL1‐L‐specific region. We identified the LD‐targeting sequence as amino acids 573–614 of SPDL1‐L, which we refer to as the lipid droplet targeting motif (LDTM) (Figure 2B,C). The LDTM covers the whole hydrophobic region (Figure 1C,D). Deletion of five amino acids from the N‐terminus (LDTM^△N^) or four amino acids from the C‐terminus (LDTM^△C^) of the LDTM abolished its LD targeting as shown by fluorescence imaging as well as the LD‐binding index (Figure 2B–D). The LD‐binding index is determined based on the fluorescence intensity ratio between LDs and other cellular regions [ 17 ]. Similar to full‐length SPDL1‐L, the N‐terminal mEGFP‐tagged LDTM is predominantly targeted to the ER and LDs under normal conditions. Following OA treatment, it is primarily localized to LDs in both COS‐7 (Figure 2E) and HeLa cells (Figure S2A). These findings indicate that lipids regulate the localization of SPDL1‐L, and the LDTM determines the membrane targeting of SPDL1‐L.
SPDL1‐L translocates from the ER to LDs in a manner partially dependent on Seipin. (A) Localization of mEGFP‐SPDL1‐L. HeLa cells were transfected with mEGFP‐SPDL1‐L and the ER marker BFP‐SEC61β, with or without 100 µM OA treatment for 16 h prior to fixation. LDs were stained with LipidTOX (red). Scale bars represent 10 µm (2 µm for enlarged images). (B) Schematic showing the truncations of the SPDL1‐L‐specific region used to identify the domain required for LD targeting (hereafter called the LDTM). The fragments were tagged at the N‐terminus with mScarlet. (C) HeLa cells were transfected with the indicated mScarlet‐LDTM truncations from panel B and treated with 100 µM OA for 16 h prior to fixation. LDs were stained with BODIPY (green), and DAPI was used to label the nuclei (blue). Scale bars represent 10 µm (2 µm for enlarged images). (D) Quantification of the fluorescent signal from the indicated proteins on LDs (n ≥ 20). The binding index of the LDTM was normalized to 1. Data are presented as mean ± SD. Asterisks indicate statistical significance (one‐way ANOVA followed by Tukey's test): **** p < 0.0001. (E) Representative images of COS‐7 cells expressing mEGFP‐tagged LDTM of SPDL1‐L (mEGFP‐LDTM) and BFP‐SEC61β under normal and OA treatment conditions. Cells were treated with 100 µM OA for 16 h prior to fixation. LDs were stained with LipidTOX (red). Scale bars represent 10 µm (2 µm for enlarged images). (F) Confocal images of HeLa and BSCL2‐KO cells expressing mScarlet‐LDTM or mScarlet‐ACSL3. Cells were treated with 300 µM OA for 6 h before fixation. Scale bars represent 10 µm (2 µm for enlarged images). (G) Quantifications of fluorescent protein signal on LDs (n ≥ 20) in panel F. Data are presented as mean ± SD, with asterisks indicating statistical significance as determined by one‐way ANOVA followed by Tukey's test: **** p < 0.0001.
LD proteins with hydrophobic regions usually move from the ER to LD through ER‐LD contacts marked by Seipin or ER exit site proteins [18, 19]. To assess the role of Seipin in the LD targeting of SPDL1‐L, we evaluated the LD targeting efficacy of LDTM in WT HeLa cells and BSCL2‐KO cells. In addition, we used ACSL3 as a positive control. Seipin stabilizes ER‐LD contact sites and facilitates the correct localization of ACSL3 [19]. We found that LD targeting of both the LDTM and ACSL3 was significantly impaired in BSCL2‐KO cells at the 6‐h, 300 µM OA treatment timepoint (Figure 2F,G). Notably, LDTM and ACSL3 preferentially accumulated at putative LD‐LD or LD‐ER contact sites in BSCL2‐KO cells. Therefore, Seipin facilitates the LD targeting of SPDL1‐L.
The Hydrophobic Region and Conserved Basic Residues Nearby Jointly Contribute to SPDL1‐L LD Targeting
2.3
Based on AlphaFold 3 predictions [20] and second structure analysis, the LDTM contains two helices separated by two proline residues (P597 and P602) and the first helix is hydrophobic (Figure 1C,D and Figure 3A). Moreover, several basic residues near the hydrophobic region are also highly conserved (colored red in Figure 3A,B).
*The hydrophobic region and basic residues collectively mediate the targeting of the LDTM to LDs. (A) Amino acid sequence of the LDTM, with several mutations were designed to disrupt the structure of the LDTM. (B) Conservation analysis of amino acid residues in the LDTM of SPDL1‐L‐like proteins, based on sequences from 185 different species. (C) Confocal images showing the localization of proline mutants within the LDTM. HeLa cells were transfected with mScarlet‐tagged constructs and incubated with 100 µM OA for 16 h before fixation. LDs were stained with BODIPY (green). Scale bars represent 10 µm (2 µm for enlarged images). (D) Quantifications of fluorescent protein signal on LDs (n = 21) in panel C. The binding index of the LDTM was normalized to 1. Data are presented as mean ± SD, with asterisks indicating statistical significance as determined by one‐way ANOVA followed by Tukey's test: ***p < 0.001. (E) Confocal images showing the localization of LDTM mutants with a deletion in the segment (601‐610). Cells were treated with 100 µM OA for 16 h before fixation. Scale bars represent 10 µm (2 µm for enlarged images). (F) Quantifications of fluorescent protein signal on LDs (n = 21) in panel E. Data are presented as mean ± SD, with asterisks indicating statistical significance as determined by one‐way ANOVA followed by Tukey's test: ****p < 0.0001. (G,H) Confocal images showing the localization of LDTM mutants in which the blue‐labeled residues in panel A, representing highly conserved hydrophilic amino acids, were mutated to the weakly hydrophobic alanine (A) in panel G or the hydrophilic glutamic acid (E) in panel H. Scale bars represent 10 µm (2 µm for enlarged images). (I) Quantifications of fluorescent protein signal on LDs in panel G (n = 20) and panel H (n = 20). Data are presented as mean ± SD, with asterisks indicating statistical significance as determined by one‐way ANOVA followed by Tukey's test: ****p < 0.0001. (J) Confocal images showing the localization of LDTM mutants where red‐labeled positively charged residues, arginine (R) and lysine (K), were mutated to uncharged alanine (A). Scale bars represent 10 µm (2 µm for enlarged images). (K) Quantifications of fluorescent protein signal on LDs (n = 20) in panel J. Data are presented as mean ± SD, with asterisks indicating statistical significance as determined by one‐way ANOVA followed by Tukey's test: ***p < 0.0001. (L) A cartoon model depicting the localization of the LDTM to LDs and the ER. Under normal conditions, the LDTM primarily localizes to the ER. Upon OA induction, the LDTM predominantly localizes to LDs. The LDTM contains a hydrophobic region and basic residues. The hydrophobic region may interact with LDs through hydrophobic interactions, and the basic residues may interact with the negatively charged heads of phospholipids on the surface of the LDs. Both hydrophobic and electrostatic interactions are likely to contribute to targeting of the LDTM to LDs. In contrast, the localization of the LDTM to the ER seems to rely primarily on hydrophobic interactions. We can't rule out the possibility that the LDTM interacts indirectly with LDs by binding to other LD‐associated proteins, such as Seipin, rather than directly binding to the LD surface. PC, phosphatidylcholine; PE, phosphatidylethanolamine; PA, phosphatidic acid; PI, phosphatidylinositol; PS, phosphatidylserine. The yellow and pink heads of the phospholipids indicate PC/PE and PA/PI/PS, respectively.
We then used site mutants to assess the LD targeting contribution of different residues. Replacing the prolines with valines (Figure 3C,D) or deleting the segment containing the second proline (Figure 3E,F) slightly reduces the protein's LD targeting. Mutating the two prolines to valines results in a more continuous α‐helix without the kink, as evidenced by the smoother helical structure (Figure S2B). This structural alteration suggests that the proline‐induced kink has a moderate effect on the LDTM's localization to LDs (Figure 3C,D; Figure S2B). Therefore, it is possible that the hydrophobic region targets SPDL1‐L to LDs.
To test this, we created a series of missense mutations designed to disrupt the hydrophobic region in the LDTM. We substituted the conserved hydrophobic residues phenylalanine (F), isoleucine (I), leucine (L), cysteine (C), and methionine (M) within hydrophobic region with neutral alanine (A) or hydrophilic glutamic acid (E), which would weaken or abolish its hydrophobicity, respectively (Figure 3A). We expressed mScarlet‐tagged LDTM mutants in HeLa cells and assessed their LD targeting. Consistent with the prediction, the mutants with lower hydrophobicity (LLCM→4A, FFFL→4A, and IIF→3A) showed significantly reduced LD targeting (Figure 3G,I). When the hydrophobic residues within the hydrophobic region were reversed to hydrophilic glutamic acid, binding to LDs was completely abolished (LLCM→4E, FFFL→4E, and IIF→3E) (Figure 3H,I). Therefore, these results indicate that the hydrophobic region is required for SPDL1‐L LD targeting.
Above, we showed that deletion of the C‐terminal part of the LDTM abolished the LD localization of SPDL1‐L (Figure 2B–D), indicating that the hydrophobic region alone is insufficient for LD targeting. Notably, both the N‐ and C‐termini of the LDTM contain several highly conserved basic arginine (R) and lysine (K) residues (R576, R608, R611, K612, K614) (Figure 3A,B). Typically, basic residues electrostatically interact with negatively charged phospholipids in membranes [21, 22, 23, 24]. Basic residues are enriched in LD‐associated proteins and may serve as sorting signals for recruitment to the LD from the ER [25]. To evaluate the role of electrostatic interactions in SPDL1‐L LD targeting, we substituted these conserved charged residues (R and/or K) with the uncharged residue A. LD targeting of the LDTM was almost completely abolished in these mutants (Figure 3J,K). Similar mutants of full‐length SPDL1‐L (FFFL→4E and 2K/3R→5A) almost completely lost LD localization and instead accumulated predominantly in the nucleus (Figure S2C), likely due to impaired LD association and yielded a pattern reminiscent of the short isoform SPDL1‐S, which is predominantly nuclear during interphase.
We also examined whether the hydrophobic region and basic residues are crucial for the ER localization of SPDL1‐L under normal conditions. We co‐expressed mScarlet‐tagged LDTM mutants with EGFP‐SEC61β. Disruptive point mutations of the hydrophobic region almost completely abolished the ER localization of the LDTM, causing it to diffuse throughout the entire cell (Figure S3A,C). In contrast, when the basic residues (R and/or K) were mutated to A, the tagged proteins still mostly localized to the ER (Figure S3B,D). Consistently, full‐length SPDL1‐L mutants carrying the same hydrophobic substitutions (LLCM→4E, FFFL→4E, and IIF→3E) also lost ER localization and failed to colocalize with SEC61β (Figure S3E). Therefore, the hydrophobic region, but not the basic residues, is critical for the ER localization of SPDL1‐L. Putting together, it is likely that the hydrophobic region helps SPDL1‐L to target to the ER membrane, while under OA‐loaded conditions, LD membranes with more negatively charged phospholipids which provide additional binding affinity for SPDL1‐L LD targeting through Seipin‐involved ER‐LD contacts (Figure 3L). After defining the determinants of ER and LD localization, we next characterized the cellular function of SPDL1‐L in culture cells by overexpression and loss‐of‐function analyses.
SPDL1‐L Promotes LD Clustering in a Dynein‐Dependent Manner
2.4
As shown previously (Figure 2A), expression of mEGFP‐SPDL1‐L in HeLa cells leads to a highly penetrant LD clustering phenotype. Most LDs are clustered together and many cells have only one cluster. Each cluster contains 50 to 80 LDs and they do not fuse. This may be due to the low expression of LD fusion proteins, such as CIDE family members, because expression of both mScarlet‐SPDL1‐L and CIDEA results in the formation of large LDs (Figure S4A). In addition, the degree of LD clustering was positively correlated with the fluorescence intensity of mEGFP‐SPDL1‐L (Figure S4B).
The formation of LD clusters could be due to increased LD‐LD contact and/or LD transport to the same subcellular location. In chicken cone cells, SPDL1‐L regulates LD positioning near the MTOC [15]. Similarly, in Hela cells, SPDL1‐L‐decorated LDs clustered around the centrosome marker pericentrin, which marks the MTOC (Figure 4A). The cytoskeleton and motor proteins are responsible for transporting organelles and other cargos, influencing their distribution within the cell. We treated HeLa cells expressing mScarlet‐SPDL1‐L with the microtubule‐depolymerizing compound nocodazole for 1 h prior to fixation. Microtubule depolymerization resulted in reduced LD clustering, demonstrating the requirement for microtubules to maintain LD clustering (Figure 4B). During mitosis, microtubules are reorganized to form the spindle, with their minus ends concentrated at the two poles, and organelles undergo a redistribution process [26, 27]. Similarly, SPDL1‐L‐decorated LDs were enriched at mitotic spindle poles at metaphase and telophase during cell division compared to the dispersed distribution of LDs in wild‐type cells (Figure 4C).
SPDL1‐L‐mediated LD clustering is microtubule‐ and dynein‐dependent. (A) Immunofluorescence staining of pericentrin in HeLa cells expressing mScarlet‐SPDL1‐L following treatment with 100 µM OA. Scale bar represents 10 µm (2 µm for enlarged image). (B) Fluorescence images of HeLa cells transfected with empty vector (EV) or vector expressing full‐length mScarlet‐SPDL1‐L, with or without nocodazole treatment (5 µg/mL), applied 1 h prior to fixation. 100 µM OA was added to induce LD formation. Scale bars represent 10 µm. (C) Fluorescence images of HeLa cells with or without mEGFP‐SPDL1‐L expression during mitosis upon 100 µM OA treatment. Tubulin Tracker Deep Red was used to label microtubules (red) in living cells. Scale bars represent 10 µm. (D) Fluorescence images of HeLa cells expressing full‐length (FL), N‐terminal truncation (∆N: aa 361–622), and point mutants of mScarlet‐SPDL1‐L (∆CCS: A24V, Y60A and F258A) upon 100 µM OA treatment. EV: empty vector control. Scale bars represent 10 µm. (E) Fluorescence microscopy images of HeLa cells co‐expressing mScarlet‐tagged SPDL1‐L and 3×Myc‐tagged p50‐dynamitin, or treated with siRNA targeting dynein heavy chain 1 (DHC). The Myc channel shows the localization of Myc‐tagged p50‐dynamitin (green). Scale bars represent 10 µm. (F–H) Statistical analysis of LD clustering in panels B and D (n ≥ 20). Each dot represents a cell. DAPI labels the nuclei, and LDs are labeled with BODIPY. Data are shown as mean ± SD, with asterisks indicating statistical significance as assessed by one‐way ANOVA and subsequent Tukey's test: ** p < 0.01, *** p < 0.001, **** p < 0.0001; ns: not significant.
Minus‐end‐directed cytoplasmic dynein motors and plus‐end‐directed kinesin motors contribute to cargo transport along microtubules [28]. SPDL1‐S binds to dynein through its N‐terminal domain and interacts with the kinetochore through its C‐terminus [16]. The recruitment of dynein was nearly abolished in cells expressing SPDL1‐S^∆CCS(A24V, Y60A, F258A)^, in which three core motifs for dynein binding were mutated [13]. Since SPDL1‐L shares the same N‐terminal domain as SPDL1‐S, we speculated that SPDL1‐L links dynein to LDs, promoting the minus end movement of LDs. To test that, we expressed two different mutant forms of SPDL1‐L: an N‐terminal truncation called SPDL1‐L^∆N^, which lacks the dynein‐interacting domain while retaining aa 361–622, and SPDL1‐L^∆CCS^, which has mutations in the three core dynein binding motifs. Compared to wild‐type SPDL1‐L, both the SPDL1‐L^∆N^ and SPDL1‐L^∆CCS^ mutants greatly reduced LD clustering (Figure 4D). A similar phenotype was observed in cells where dynein‐heavy chain 1 (DHC) was depleted by RNAi or in cells expressing p50‐dynamitin, a protein known to disrupt the dynein/dynactin complex (Figure 4E,C) [29]. To further examine the interaction between SPDL1‐L and dynein, we assessed the interaction between DHC and wild‐type SPDL1‐L or its LDTM mutants. As shown in Figure S4D, DHC interacted with both SPDL1‐L and its LDTM mutants, and the interaction strength between DHC and the LDTM mutants was comparable to that observed with wild‐type SPDL1‐L. These findings suggest that LD binding may not be essential for activation of the SPDL1‐L‐dynein interaction.
We further quantified the LD clustering phenotype. A cluster was defined as an aggregation of more than five LDs in 2D images. The “cluster number” refers to the total number of clusters in each cell, while the “cluster percentage” represents the proportion of clustered LDs relative to the total number of LDs in each cell. The “cluster size” indicates the number of LDs within each cluster. Compared to control HeLa cells, nearly all LDs in cells expressing full‐length SPDL1‐L cluster at 1–2 positions (Figure 4F,G). In contrast, over 80% of LDs cluster at 3–4 positions in cells expressing mScarlet‐SPDL1‐L^∆N^ and mScarlet‐SPDL1‐L^∆CCS^, with significantly smaller cluster sizes (Figure 4F–H). The SPDL1‐L‐induced LD clustering phenotype was significantly diminished when microtubules were disrupted by nocodazole, particularly in terms of the reduced cluster percentage (Figure 4F–H). Together, these results indicate that LD‐localized SPDL1‐L promotes LD transport to the MTOCs/spindle poles in a dynein‐dependent manner.
SPDL1‐L Overexpression Causes Donut‐Shaped Nuclei (DSNs)
2.5
Interestingly, we found irregular nuclear morphologies in HeLa cells overexpressing mScarlet‐SPDL1‐L. For detailed phenotyping, we also generated HeLa cells lines with stable overexpression (OE) of SPDL1‐L and characterized one such cell line, OE‐12. DAPI staining revealed that the aberrant nuclei included both donut‐shaped nuclei (DSNs) and bean‐shaped nuclei (BSNs) (Figure 5A). A DSN has a central hole without chromatin and the internal rim of the DSN is marked by Lamin B1, a nuclear envelope marker. We observed that SPDL1‐L‐decorated LD clusters were localized at the hole of DSNs and at the indent of BSNs. Under normal culture conditions, 13% of OE‐12 cells have DSNs, and 21% have BSNs. OA treatment significantly increased the incidence of both DSNs and BSNs, with around 30% of cells exhibiting each phenotype (Figure 5B,C). 3D reconstruction based on confocal images of DAPI/BODIPY staining and mScarlet fluorescence confirmed that the central hole in DSNs penetrates the entire nucleus and is continuous with the cytoplasm (Figure 5B; Video S1). Cytosolic subcellular structures such as the ER (marked by EGFP‐SEC61β), mitochondria (stained with anti‐Hsp60 antibody), microtubules (stained with anti‐alpha‐tubulin antibody), and centrosomes (marked by EGFP‐CETN2) were found in the hole of DSNs (Figure 5D). SPDL1‐L overexpression also induced DSNs in several other cell lines, including COS7, SUM159, LN229, and U251 cells following OA treatment, although the frequency varied among cell types (6% in COS7, 16% in SUM159, 19% in LN229, and 35% in U251) (Figure S5A,B).
Overexpression of SPDL1‐L results in the formation of donut‐shaped nuclei (DSNs) at mitotic exit. (A) Representative images of DSNs and BSNs under normal conditions. HeLa cells with stable overexpression of mScarlet‐SPDL1‐L (OE‐12 cells) were immunostained for lamin B1. Scale bars represent 10 µm. (B) 3D reconstruction of a DSN and LDs labeled with BODIPY in an OE‐12 cell under OA treatment conditions. Scale bar represents 10 µm (2 µm for BODIPY and mScarlet‐SPDL1‐L panels). (C) The percentage of DSNs and BSNs in HeLa and OE‐12 cells under normal (Con) and 100 µM OA treatment conditions (n > 150). (D) The distribution of cellular organelles in OE‐12 cells with DSNs under normal conditions. The cells were labeled the following organelle markers: including plasmid‐transfected EGFP‐Sec61β for the ER, Hsp60 antibody staining for mitochondria, alpha‐tubulin antibody staining for microtubules, and plasmid‐transfected mEGFP‐CETN2 for centrosomes. Scale bars represent 10 µm. (E) Live imaging of OE‐12 cells expressing H2B‐mEGFP to monitor DSN formation during cell division after treatment with 100 µM OA for 12 h. Images were captured every 8 min, with the upper panels showing the time‐lapse series. 3D reconstructions of the micrographs are displayed below. Scale bar represents 10 µm. (F) The proportion of DSNs in HeLa cells transfected with various forms of mScarlet‐SPDL1‐L, including ∆N (n = 207) and ∆CCS (n = 232). (G) The proportion of DSNs in HeLa cells transfected with mScarlet‐SPDL1‐L (n = 424) and mScarlet‐BICD2‐N‐SPDL1‐L‐C (n = 374). The N‐terminal region (1‐400) of BICD2 binds to dynein, and the C‐terminal region (558‐622) of SPDL1‐L directs targeting to LDs. (H) Model for DSN formation. Overexpression of SPDL1‐L leads to the accumulation of LDs at the spindle poles during mitosis. At the end of mitosis, the spindle poles are positioned closer to the cell interior relative to the DNA plate. The LDs act as a hindrance, causing the chromatin to disperse around the centrosome, ultimately resulting in the formation of a DSN in interphase.
Although there are only a few previous studies on the DSN phenotype [30, 31], they linked DSNs to microtubule abnormalities and aberrant centrosome positioning. Similarly, we observed abnormal centrosome movement during DSN formation in dividing OE‐12 cells (Figure 5E; Video S2). Notably, the SPDL1‐L‐coated LD clusters were observed to shift from the outside of the DNA plate toward its interior from metaphase to telophase, while the segregating chromosomes appeared to move toward these LD clusters. This suggests that DSN formation is associated with the spatial convergence of SPDL1‐L‐coated LDs and chromosomes during mitosis (Figure 5E; Video S2).
Next, we examined whether the formation of DSNs is dependent on dynein and LD clustering. Compared to full‐length SPDL1‐L, cells expressing mScarlet‐SPDL1‐L^∆N^ and mScarlet‐SPDL1‐L^∆CCS^ failed to form DSNs (Figure 5F), indicating that dynein binding is required for SPDL1‐L‐mediated DSN formation. We also treated OE‐12 cells with the dynein ATPase inhibitor Ciliobrevin D. Notably, while 20 µM Ciliobrevin D treatment had no effect on LD clustering (Figure S5C), it markedly reduced DSN formation (Figure S5D), highlighting the critical role of the dynein‐driven force in DSN formation. We then substituted the N‐terminal region (1‐557) of SPDL1‐L with the N‐terminal region (1‐400) of BICD2, one of the most well‐known dynein adaptors [32]. Interestingly, we found that DSNs were not observed in HeLa cells expressing BICD2‐N‐SPDL1‐C (558‐622), even though LDs still clustered (Figure 5G,E). Put together, these results indicate that LD clustering is insufficient to induce DSN formation. It is possible that the accumulation of SPDL1‐L‐coated LD clusters at spindle poles, together with dynein‐driven forces, could contribute to local chromatin displacement during mitosis, which might predispose cells to abnormal nuclear morphologies (Figure 5H).
Additionally, we conducted a co‐culture experiment to evaluate the growth of OE‐12 cells. The results indicated that under co‐culture conditions, OE‐12 displayed a growth disadvantage compared to wild‐type (WT) cells (Figure S6A). Interestingly, despite a significant increase in the proportion of DSNs following OA treatment, the growth disadvantage did not markedly worsen upon OA treatment, suggesting that DSN production has a limited impact on the growth of OE‐12 cells (Figure S6A). Since changes in nuclear morphology may affect the invasive properties of cells, we performed an invasion assay on OE‐12 cells. We found that the capacity of OE‐12 cells for invasion was significantly reduced compared to WT cells (Figure S6B,C). It is possible that structural alterations in the DSN cells could reduce the flexibility and deformability of the nucleus, thereby affecting the cells' invasive potential [33].
SPDL1‐L Regulates LD Clustering in SH‐SY5Y Cells
2.6
Because the aforementioned results are from SPDL1‐L‐overexpressing cells, we next investigated the function of SPDL1‐L on LD clustering in wild‐type cultured cells. Since cellular LD clustering is related to the expression level of SPDL1‐L (Figure S4B), we hypothesized that LD clustering may occur in cells with a high level of SPDL1‐L expression. We used a SPDL1‐L‐specific antibody to detect the level of SPDL1‐L protein by western blot in various cell lines. The results showed that the anti‐SPDL1‐L antibody successfully recognized overexpressed mScarlet‐SPDL1‐L. However, endogenous SPDL1‐L was undetectable in all tested cell lines, likely due to the sensitivity limitations of the antibody (Figure S7A). We then measured the levels of SPDL1‐L mRNA. We found that SPDL1‐L was expressed at a relatively higher level in SH‐SY5Y, a human neuroblastoma cell line (Figure 6A). In addition, the relative level of SPDL1‐L transcription compared to SPDL1‐S transcription was highest in SH‐SY5Y cells (Figure S7B). In line with this observation, RNA‐seq data from the GTEx database (https://gtexportal.org/home/) show that SPDL1‐L expression is relatively high in human brain tissue, moderate in the heart and testes, and almost undetectable in adipose tissue (Figure S7C). Furthermore, immunostaining with the SPDL1‐L‐specific antibody showed that endogenous SPDL1‐L is localized to LDs in SH‐SY5Y cells (Figure 6B).
Loss of SPDL1‐L results in dispersed distribution of LDs in SH‐SY5Y cells. (A) The mRNA expression levels of SPDL1‐L in diverse cell types, determined by qRT‐PCR using B2M (β2‐microglobulin) as an internal control. (B) Immunofluorescence staining of endogenous SPDL1‐L in SH‐SY5Y cells. Scale bars represent 10 µm (2 µm for lower enlarged images). (C) Representative fluorescence images of different cell types treated with varying concentrations of OA for 24 h to achieve comparable LD sizes. LDs were labeled with Nile Red. Scale bars represent 10 µm. (D) Quantification of the cluster percentage in panel C (n ≥ 39), with each dot representing a single cell. (E) The Spearman correlation between SPDL1‐L mRNA levels and cluster percentage (r = 0.9030, p = 0.0008) was calculated for the cell lines indicated by pink circles. The outliers HEK293 and HepG2 (blue circles) were excluded from the analysis. (F) Comparison of LD distribution among SH‐SY5Y cells, SPDL1‐ knockdown SH‐SY5Y cells and SH‐S2F cells. Scale bars represent 10 µm. (G) Quantification of the cluster percentage in panel F (SH‐SY5Y and SH‐S2F cells: n ≥ 20; shSPDL1 SH‐SY5Y cells: n = 20), with each dot corresponding to an individual cell. Data are presented as mean ± SD. Asterisks indicate statistical significance (one‐way ANOVA followed by Tukey's test): ** p < 0.01, **** p < 0.0001.
We also examined the distribution of LDs in different cell lines. Cells were treated with various concentrations of OA to achieve similar LD sizes. Consistent with the high expression of SPDL1‐L mRNA, multiple clusters of LDs were found in SH‐SY5Y cells compared to other cell lines (Figure 6C). Quantitative analysis revealed that the cluster percentage in SH‐SY5Y cells was significantly higher than that in other cell types (Figure 6D). Additionally, except for HepG2 and HEK293 cells, SPDL1‐L mRNA levels were positively correlated with LD cluster percentage (Figure 6E).
Next, we examined the LD distribution in SPDL1‐L loss‐of‐function conditions. Because we were unable to obtain an effective SPDL1‐L specific shRNA, we knockdown both SPDL1‐S and SPDL1‐L first. As SPDL1‐L transcription is far more abundant than SPDL1‐S in SH‐SY5Y cells, the effect may reflect SPDL1‐L depletion. Knockdown of SPDL1 impairs LD clustering, as evidenced by the reduction in LD percentage (Figure 6F,G; Figure S7D). During the generation of a SPDL1‐L FLAG‐tag knock‐in SH‐SY5Y cell line, we accidentally generated a cell line, SH‐S2F, with a 51‐bp insertion in the LDTM of SPDL1‐L (Figure S7E). This creates a SPDL1‐L variant with an insertion of an extra 18 amino acids at aa 598 and a deletion of F598 (Figure S7E). Compared to WT SH‐SY5Y cells, SH‐S2F cells exhibited more dispersed LDs, as evidenced by a reduced cluster percentage (Figure 6F,G). Lastly, since SH‐S2F cells exhibited reduced LD clustering, we examined the LD targeting efficiency of mScarlet‐tagged wild‐type LDTM and mScarlet‐tagged LDTM^insertion^ in HeLa cells. LD^insertion^ represents the LDTM with the 51‐bp insertion in SH‐S2F cells. Compared to mScarlet‐LDTM, mScarlet‐LDTM^insertion^ has reduced LD targeting efficiency (Figure S7F,G). It is possible that the physical proximity and proper spatial orientation between the hydrophobic region and the basic residues are important for effective protein targeting to LDs. We next examined LD distribution in two other neural cell lines, LN229 and U251. Similar to SH‐SY5Y cells, LDs formed clusters in both cell lines, and SPDL1 knockdown reduced LD clustering, indicating a conserved role of SPDL1‐L in LD organization (Figure S7H,I). Together with the overexpression results, these findings support that SPDL1‐L regulates LD distribution.
Discussion
3
LD‐associated proteins play important roles in regulating LD dynamics and LD functions. Previous studies have highlighted the crucial role of SPDL1 in mitosis [34, 35]. Here, we showed that SPDL1‐L, the long isoform of SPDL1, is a bona fide LD‐associated protein that directly regulates LD distribution through a dynein‐dependent mechanism. Besides causing LD clustering, SPDL1‐L overexpression affects nuclear morphology, resulting in a DSN phenotype.
LDs are not stationary organelles. They constantly move and contact with other organelles. Similar to most other organelles, the directional movement of LDs along the cytoskeleton, mostly along microtubules, relies on their interaction with the cytoskeleton tracks. Transport toward the plus and minus ends of microtubules depends on kinesin and dynein motors, respectively, yet the specific adaptors linking LDs to microtubule motors remain largely unknown [36]. In mammalian cells, glucose deprivation induces the relocation of LDs from the perinuclear region to the cell periphery in a microtubule‐dependent manner. The absence of Spastin impairs this redistribution process, which is dependent on Spastin's microtubule‐binding domain [37]. In Drosophila embryos, the dynein adaptor Bicaudal D (BicD) plays a dynamic role in regulating LD transport, which occurs in both directions. Instead of just being a static tether or solely involved in dynein recruitment, BicD helps balance the activity of plus‐end and minus‐end motors, controlling the overall direction of transport. NudE and Lis1 regulate LD motion by modulating dynein motor activity in COS1 cells. They enhance dynein attachment to LDs and promote persistent, directional movement along microtubules [38]. It has been reported that kinesin motors are recruited to LDs in liver cells by ARF1, a key activator of lipolysis, and drive LD transport to the periphery of hepatocytes [39]. Additionally, by binding to the dynein subunit Dync1i1, Plin3 may serve as an adaptor for microtubule‐mediated LD transport in AML12 cells [8].
Our findings expand this limited repertoire by identifying SPDL1‐L as a novel dynein adaptor that couples LDs to the microtubule network. Unlike other LD‐associated adaptors, SPDL1‐L originates from a canonical mitotic dynein adaptor and has been functionally repurposed to regulate LD motility. Through its unique C‐terminal hydrophobic region and adjacent basic residues, SPDL1‐L anchors to the LD surface and recruits dynein‐dynactin complexes, driving LD movement toward the microtubule minus end and promoting perinuclear clustering. Loss of dynein activity or disruption of the N‐terminal dynein‐binding region of SPDL1‐L abolishes this clustering phenotype, confirming the dynein dependence of the process. Endogenous SPDL1‐L is highly expressed in neuronal cell lines such as SH‐SY5Y, where it is required for maintaining LD clustering, suggesting that this regulatory mechanism may be particularly important in cells with polarized morphology and high metabolic demand.
Overexpression of SPDL1‐L results in the formation of DSNs during mitosis. DSNs have been observed in various contexts, including normal granulocyte developmental processes, and under pathological conditions in mammalian cells and tissues [40, 41, 42, 43, 44]. Altered nuclear mechanics, such as increased Lamin A/C levels or dysregulation of Lamin‐associated proteins, can promote nuclear deformation and DSN formation [45, 46, 47]. The mechanism of DSN formation under in vivo conditions has not been revealed. In vitro, previous studies have linked DSNs to mitotic errors, particularly due to defects in mitotic progression as well as aberrant microtubules and centrosomes [30, 31, 48]. Treatment with protein farnesyltransferase inhibitor (FTI) induces DSN formation, which is linked to defects of centrosome separation [31]. Similar nuclear abnormalities have also been observed under conditions affecting lysosomal or motor protein function [30, 49, 50]. When present in high numbers or forming clusters, LDs‐although inherently fluid organelles‐can exert substantial mechanical stress on the nucleus, leading to compression, deformation of the nuclear envelope, and in some cases, structural rupture, despite the nucleus's intrinsic rigidity [51]. In line with the contribution of centrosomes and microtubules, although LD clustering alone is not sufficient for DSN formation, our results suggest that LD clusters may act as physical hindrances to proper nuclear envelope reformation.
The mechanisms which determine the targeting of proteins to LDs have been intensively studied. Although several distinct protein domains or protein modifications in LD‐associated proteins for LD targeting have been reported [52], one major question remains largely unanswered: how are proteins specifically targeted to LD membranes versus other cellular membranes? Amphipathic helices, hydrophobic region including short hairpins, and post‐translational lipid modifications of proteins all favor membrane binding. How do these domains differentiate the hydrophobicity of the LD membrane from other cellular membranes? Previous studies suggested the membrane packing defects of LDs may expose more hydrophobic areas than other cellular membranes [17, 53]. Furthermore, the neutral lipid core may provide higher and more persistent hydrophobicity than phospholipid tails [54].
Here, based on our finding that both the hydrophobic region and nearby basic residues are required for targeting of SPDL1‐L to LDs, we propose a coincident detection model for LD targeting of proteins. It is possible that the hydrophobic region‐membrane interaction is relatively weak. However, when combined with other binding forces, such as electrostatic interactions between positively charged amino acids and negatively charged phospholipids, the hydrophobic region is able to achieve a much stronger and more specific interaction. We found that OA loading promotes the LD localization of SPDL1‐L. It is possible that LD membranes have relatively more negatively charged phospholipids, such as PS and PA [55, 56], potentially supporting the electrostatic interactions required for SPDL1‐L localization to LDs. Our results show that basic residues play a key role in enhancing the affinity of SPDL1‐L for LDs rather than the ER. Similarly, the central hydrophobic domain of caveolins aids its attachment to the ER, while the basic residues direct caveolins to LDs [25]. Positively charged residues located near the hinge region of membrane‐embedded hairpin motifs play a critical role in targeting proteins from the ER to LDs. These residues, such as R179 in LiveDrop and R105 in ALG14, undergo snorkeling to interact with the LD monolayer interface, stabilizing protein localization [57]. The LD targeting domain of Klar also contains several conserved basic amino acids located at its C‐terminus, although the precise role of these basic amino acids in LD targeting remains unclear [58]. Recent studies have highlighted that the arginine‐rich segment of the adipocyte‐specific protein CLSTN3B is essential for enhancing the interaction between the ER and LDs, facilitating the transfer of phospholipids from the ER to LDs during LD expansion [59]. The coincident detection model has been widely reported in phospholipid signaling and the localization of phospholipid‐binding proteins. Coincident detection may be a common mechanism to achieve high compartmental specificity of non‐covalent protein‐lipid or protein‐membrane interactions.
In summary, our work identifies SPDL1‐L as a new LD‐dynein adaptor that connects LDs to the microtubule network, thereby driving their perinuclear accumulation. The specific LD‐localized variant of SPDL1 expands the function of the SPDL1 gene beyond mitosis into lipid organelle regulation. Our study offers new insights into the functional dynamics of LDs. It will be interesting to study the in vivo function and regulation of SPDL1‐L in the future.
While this study establishes SPDL1‐L as a dynein adaptor that regulates LD organization, several limitations should be acknowledged. Our analyses are primarily based on cultured human neural cell lines, and thus the physiological relevance of SPDL1‐L‐mediated LD clustering remains to be verified in vivo. Whether LDs in mouse cells exhibit tight perinuclear clustering under physiological or stress conditions remains unclear‐an open question warranting future investigation. The correlation between SPDL1‐L mRNA expression and LD clustering across cell types is modest, likely reflecting differences in protein abundance or activity. Moreover, we cannot exclude the possibility that redundant or compensatory mechanisms contribute to LD positioning in other cellular contexts.
Experimental Section
4
Plasmids
4.1
mEGFP‐SPDL1‐L and mScarlet‐SPDL1‐L were generated by cloning human SPDL1‐L (GenBank accession number NM_001329639.2) into the pEGFP‐C1 and pmScarlet‐C1 vectors, respectively. mScarlet‐SPDL1‐L^∆N^ (aa 361–622), mScarlet‐LDTM (aa 573–614), mScarlet‐LDTM^∆N^ (aa 578–614), and mScarlet‐LDTM^∆C^ (aa 578–610) were created by homologous recombination based on the mScarlet‐SPDL1‐L construct. mScarlet‐SPDL1‐L^∆CCS^ (A24V, Y60A, and F258A) was generated via PCR‐based mutagenesis from mScarlet‐SPDL1‐L. A series of LDTM point mutants were also created by PCR‐based mutagenesis using mScarlet‐LDTM as the template. mScarlet‐BICD2‐N‐SPDL1‐L‐C was constructed by inserting the N‐terminal region of human BICD2 (aa 1–400) and the C‐terminal region of human SPDL1‐L (aa 558–622) into the pmScarlet‐C1 vector. Human CETN2 cDNA was cloned into the pEGFP‐C1 vector. For the cloning of 3xMyc‐p50‐dynamitin, the cDNA was inserted into the pcDNA3.1 vector, encoding an N‐terminal 3xMyc tag. BFP‐SEC61β, mEGFP‐SEC61β, and H2B‐mEGFP constructs were generously provided by Dr. Kangmin He (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences).
Cell Culture and Plasmid Transfection
4.2
Caki‐1 (ATCC, RRID: CVCL_0234), COS‐7 (ATCC, RRID: CVCL_0224), HCT116 (ATCC, RRID: CVCL_0291), HeLa (ATCC, RRID: CVCL_0030), HepG2 (ATCC, RRID: CVCL_D5I8), HEK293 (ATCC, RRID: CVCL_0045), hHEF (Haoyi Wang lab, IOZ, CAS; generated from WIBR3 hESCs, RRID: CVCL_9767), LN229 (ATCC, RRID: CVCL_0393), U2OS (ATCC, RRID: CVCL_0042), U251 (ATCC, RRID: CVCL_0021), and SUM159 (ATCC, RRID: CVCL_5423) cells were cultured in DMEM (Gibco, Cat# C11995500BT), SH‐SY5Y (ATCC, RRID: CVCL_0019) cells in DMEM/F12 (Hyclone, Cat# SH30023.01), and HT29 (ATCC, RRID: CVCL_A8EZ) cells in McCoy's 5A (Gibco, Cat# 16600082), all supplemented with 10% fetal bovine serum (Biological Industries, Cat# 04‐001‐1A) and 100 U/mL penicillin‐streptomycin (Biological Industries, Cat# 03‐031‐1B). Cells were maintained at 37 °C in a humidified incubator with 5% CO_2_. WIBR3 hESCs (Haoyi Wang lab, IOZ, CAS; RRID: CVCL_9767) were cultured following the method described by An et al. [60]. All cells were confirmed to be free of mycoplasma contamination.
For plasmid transfection, HeLa, COS‐7, LN229, U251, and SH‐SY5Y cells were seeded onto coverslips in 6‐well plates at a density of 3 × 10⁵ cells per well. The cells were allowed to adhere overnight, reaching 50–70% confluence by the time of transfection. Plasmids (2 µg) were transfected into the cells using Lipofectamine 2000 (Invitrogen, Cat# 11668019) for HeLa, COS‐7, LN229, and U251 cells and Xfect transfection reagent (Clontech, Cat# 631317) for SH‐SY5Y cells according to the manufacturer's instructions. After 4–6 h, the medium was replaced with 2 mL of fresh complete DMEM. Cells were harvested for fluorescence observation 24–48 h post‐transfection. Unless otherwise specified, all LD induction experiments were performed with Oleic acid (Sigma‐Aldrich, Cat# 364525, 100 µM) for 16 h before harvest.
Immunofluorescence Staining and LD Staining
4.3
Where indicated, cells were treated with Ciliobrevin D (Aladdin, Cat# C421408, 20 µM) or Nocodazole (Sigma‐Aldrich, Cat# M1404, 5 µg/mL) for the specified durations before fixation, as described in the figure legends. After treatment, live cells in 6‐well plates were washed three times with phosphate‐buffered saline (PBS), and, when applicable, microtubules were labeled with Tubulin Tracker Deep Red (Invitrogen, Cat# T34077) according to the manufacturer's instructions. Cells were then fixed with 4% formaldehyde for 20 min, permeabilized with 0.1% Triton X‐100 in PBS for 20 min, and non‐specific binding was blocked by incubating the cells in 3% BSA in PBS for 30 min at room temperature. Cells were incubated with primary antibodies diluted in 3% BSA in PBS overnight at 4°C, including anti‐SPDL1‐L (this paper, 1:200), anti‐Lamin B1 (EASYBIO, Cat# BE3191, 1:500), anti‐Hsp60 (Selleck, Cat# F0482, 1:500), anti‐α‐tubulin (EASYBIO, Cat# BE0031, 1:500), and Pericentrin polyclonal antibody (Proteintech, Cat# 27084‐1‐AP, 1:200). After washing with PBS, cells were incubated with the appropriate Alexa Fluor‐conjugated secondary antibodies for 1 h at room temperature in the dark, including donkey anti‐mouse IgG Alexa Fluor 488 (Thermo Fisher Scientific, Cat# A21202, 1:500) and donkey anti‐rabbit IgG Alexa Fluor 488 (Thermo Fisher Scientific, Cat# A21206, 1:500). LDs were stained with BODIPY 493/503 (ThermoFisher Scientific, Cat# D3922), Nile Red (Sigma‐Aldrich, Cat# 7385‐67‐3), or LipidTOX Deep Red (Invitrogen, Cat# H34477) according to the manufacturer's instructions. Finally, nuclei were stained with DAPI (Invitrogen, Cat# 62247, 1 µg/mL) or Hoechst 33342 (Solarbio, Cat# 14533, 1 µg/mL) for 10 min at room temperature, followed by washing with PBS.
Confocal Imaging and Live Cell Imaging
4.4
Fixed cells were imaged using Leica SP8 and Zeiss LSM 980 confocal systems, with images captured using a 63× oil immersion objective.
For live‐cell imaging of DSN formation, OE‐12 cells were seeded onto glass‐bottomed 35‐mm dishes (NEST, Cat# 801002), specifically designed for live‐cell imaging, and cultured in complete DMEM for 16 h to achieve approximately 50%–70% confluence. H2B‐mEGFP was then transfected to label DNA, and the cells were incubated for an additional 12 h. RO‐3306 (Selleck, Cat# S7747, 10 µM) was subsequently added to arrest the cells at the G2/M boundary, while OA was added simultaneously at a final concentration of 100 µM to induce LD formation. After 12 h of treatment, cells were washed with PBS and replaced with complete DMEM to allow them to enter mitosis. Two hours after release, cells were subjected to live‐cell imaging using a Zeiss LSM 980 confocal system.
To determine the fate of DSN‐containing OE‐12 cells, the cells were seeded onto 6‐well plates and cultured in complete DMEM for 16 h to reach approximately 50%–70% confluence. H2B‐mEGFP was then transfected to label DNA. After 24 h of transfection, cell growth was monitored using the Incucyte S3 Live‐Cell Analysis System (Sartorius). Images were captured every hour for 24 h using the system's integrated software.
Real‐Time Quantitative PCR
4.5
Total RNA was extracted from cells samples using TRIzol reagent. Subsequently, 2 µg of total RNA was reverse transcribed into cDNA using the EasyScript One‐Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech). qPCR was performed on a 7500 Real‐Time PCR System (Applied Biosystems) using SYBR Green SuperMix (TransGen Biotech). The primer sequences for qRT‐PCR are listed in Table S3.
Antibody Preparation and Specificity
4.6
The monoclonal antibody specific to SPDL1‐L was generated using hybridoma technology. Mice were immunized with a C‐terminal peptide from human SPDL1‐L (aa 606–622, FSRTLRKTKLIAKGKDS) to elicit an immune response. Spleen cells from the immunized mouse, which contained the antibody‐producing B cells, were then fused with myeloma cells to create hybridomas. The hybridomas were cultured in a selective medium, and those capable of producing antibodies against SPDL1‐L were selected. The specificity of the resulting antibodies was validated through Western blotting. Among the generated clones, the 15F3 antibody, with a higher affinity for SPDL1‐L, was selected for all subsequent experiments.
Co‐Culture Assay
4.7
mEGFP‐tagged HeLa cells and mScarlet‐tagged OE‐12 cells were generated through lentiviral transduction. Equal numbers of both cell types were sorted by flow cytometry and co‐cultured in a 6‐well plate. The cells were then subjected to two different treatments: one group was left untreated, while the other was treated with 50 µM OA. After culturing for 4, 7, and 10 days, flow cytometry was used to assess the proportions of each cell type in the co‐culture based on mEGFP and mScarlet fluorescence.
Transwell Invasion Assay
4.8
HeLa cells and OE‐12 cells were cultured in complete medium supplemented with 100 µM OA to induce DSN formation. The cells were then suspended in serum‐free medium at a concentration of 2×10^6 cells/mL. For the invasion assay, 200 µL of the cell suspension was added to the upper chamber of a Matrigel‐coated transwell insert (8‐µm pore size; Corning, Cat# 3422) pre‐coated with Matrigel (Corning, Cat# 356234). The lower compartment contained 600 µL of 3T3‐L1 cell culture supernatant with 20% FBS, acting as a chemoattractant. Additionally, 50 µM OA was added to both the upper and lower chambers to maintain the proportion of DSNs. The cells were incubated for 48 h to allow migration through the Matrigel. After the incubation period, cells in the chamber were fixed with 4% paraformaldehyde for 20 min, followed by staining with 0.5% crystal violet for 20 min. Non‐invasive cells were gently removed by wiping, and the cells that had migrated through the Matrigel to the underside of the membrane were observed under a light microscope fitted with a 40× oil immersion objective for quantification.
Generation of a Stable HeLa Cell Line Overexpressing mScarlet‐SPDL1‐L
4.9
Stable HeLa cell lines overexpressing mScarlet‐SPDL1‐L were generated by employing a viral‐based method involving lentiviral transduction. A cDNA encoding mScarlet‐SPDL1‐L was inserted into a lentiviral expression vector and co‐transfected with packaging plasmids into HEK293 cells using Lipofectamine 2000, following the manufacturer's protocol. After 48 h, the supernatant containing lentivirus was collected and filtered. HeLa cells were then infected with the lentivirus in the presence of polybrene (8 µg/mL) to enhance infection efficiency. After 16 h, the medium was replaced with fresh DMEM containing 10% FBS. Stable transduced cells expressing the fluorescent reporter were selected, and single cells were sorted by FACS. These single cells were plated in a 96‐well plate to form individual clones, which were visualized under a fluorescence microscope. The clones were further analyzed by Western blotting or fluorescence microscopy. The stable HeLa cell lines overexpressing mScarlet‐SPDL1‐L were then used for subsequent experiments.
shRNA‐Mediated Knockdown of SPDL1 in SH‐SY5Y Cells
4.10
We employed an shRNA‐mediated gene knockdown approach to suppress the expression of SPDL1. The shRNA sequence, listed in Table S3, was cloned into a pLKO.1 vector containing an mCherry fluorescent marker, and lentiviral particles were generated to infect SH‐SY5Y cells. Infected cells were subsequently analyzed by confocal microscopy and qRT‐PCR to evaluate knockdown efficiency.
siRNA‐Mediated Knockdown of Dynein Heavy Chain (DHC) in HeLa Cells
4.11
siRNA‐mediated knockdown of Dynein Heavy Chain (DHC) in HeLa cells was performed using Lipofectamine 2000. HeLa cells were cultured to 50%–70% confluence and transfected with DHC‐specific or negative control siRNA (Suzhou Jima Biotechnology Co., Ltd., Suzhou, China). After 48 h of incubation, knockdown efficiency was validated by qPCR to assess DHC mRNA levels. Subsequently, mScarlet‐SPDL1‐L was transfected into these cells to assess LD distribution.
Co‐Immunoprecipitation Analysis of SPDL1‐L and DHC Interaction
4.12
HEK293 cells were transfected with the indicated plasmids and lysed 36 h later in RIPA buffer. Lysates were incubated with anti‐HA magnetic beads (M132‐10, Medical & Biological Laboratories Co., Ltd., Nagoya, Japan) overnight at 4°C, washed, and eluted in SDS sample buffer. 5% of the lysate was used as input control. Proteins were analyzed by immunoblotting with anti‐HA (M180‐3, Medical & Biological Laboratories, 1:5000) and anti‐mCherry (ab167453, Abcam, 1:3000) antibodies.
Generation of SPDL1‐L‐FLAG Knock‐in SH‐SY5Y Cells Using the CRISPR/Cas9 System
4.13
To generate SPDL1‐L‐FLAG knock‐in SH‐SY5Y cells, a 2x FLAG tag followed by a stop codon (TGA) was inserted into the LDTM‐coding region of the genomic SPDL1‐L locus using the CRISPR/Cas9 system. Guide RNAs (gRNAs) were designed to target sequences flanking the desired insertion site within the LDTM. The single guide RNAs (sgRNAs) were designed using the online platform (https://www.crispick.org). A donor DNA template containing the FLAG‐coding sequence flanked by homology arms (700–800 bp) matching the regions surrounding the cut site was cloned into the PUC19 plasmid. The gRNAs were synthesized and inserted into the Cas9‐containing PX458 construct, which also expressed EGFP for selection. PX458 and the donor template were introduced into SH‐SY5Y cells via Lipofectamine‐based transfection. EGFP selection was applied post‐transfection, followed by FACS sorting to isolate single cells into a 96‐well plate. The cells were cultured for several days, and single‐cell clones were screened for successful insertion by PCR amplification of the target locus. Positive clones were expanded and further validated by PCR to confirm the insertion.
3D Reconstruction
4.14
3D imaging in Figure 5B,E was performed using a Zeiss LSM 980 confocal microscope with a 63× and 40× oil immersion objective, respectively. Nuclei were labeled with DAPI or H2B‐mEGFP. LDs were labeled with BODIPY 493/503. mScarlet‐tagged SPDL1‐L was expressed from a construct. Z‐stack images were acquired with a step size of 0.3 µm for Figure 5B and 2 µm for Figure 5E, and processed using Imaris software (Bitplane) for 3D reconstruction. Surface rendering and volume measurements were performed to analyze the spatial relationships between the nuclei, LDs, and SPDL1‐L.
LD Clustering Analysis
4.15
LD clustering was quantified using CellProfiler. Images were processed to segment the LDs and measure their centroid positions and numbers. These measurements were then used to assess the number of clusters, cluster percentage, and the number of LDs within each cluster. Data analysis was performed using R scripts in combination with Microsoft Excel.
Crosslinking‐Mass Spectrometry (XL‐MS)
4.16
The SPDL1‐L and SPDL1‐S genes were cloned into a pFastBac vector with GFP and twin‐strep tags, along with a PreScission protease recognition site. Baculovirus was generated using the Bac‐to‐Bac system, transfected into Sf9 cells to produce primary (P1) and amplified (P2) virus stocks. Sf9 cells were infected with the P2 virus and incubated for 48–60 h to express the protein. The cells were collected by centrifugation and lysed in buffer A, followed by protein purification using Streptactin beads and elution with desthiobiotin. The GFP tag was removed using PreScission protease, and the protein was concentrated. Gel filtration on Superdex 200 and Superose 6 columns further purified the SPDL1‐S and SPDL1‐L proteins, respectively
For DSBU crosslinking [16], the protein was diluted to 5 µM in Spindly buffer, and DSBU was added at 3 mM concentration. The reaction was incubated for 1 h at room temperature, and the proteins were precipitated with cold acetone at −20°C overnight. The precipitated proteins were resuspended in 50 mM NH_4_HCO_3_, reduced with DTT, alkylated with chloroacetamide, and digested with trypsin overnight. Peptides were desalted using StageTips [61] and dried using a SpeedVac concentrator.
For LC‐MS/MS analysis, the peptides were resuspended in 0.1% formic acid and analyzed using an LTQ Orbitrap Elite mass spectrometer coupled to an Easy‐nLC 1000 in data‐dependent mode. Precursor ions were measured at 120 000 resolution, and the top 50 most intense ions from each scan were selected for MS2 analysis. Finally, the raw data was analyzed using Thermo Proteome Discoverer to identify proteins, and MaxQuant was used for crosslink site analysis by searching against the SPDL1‐L and SPDL1‐S sequence database with the appropriate parameters.
Statistical Analysis
4.17
All assays were conducted with three independent experiments. Cells or images were randomly selected for analysis. Statistical analysis was performed using Microsoft Excel, GraphPad Prism 8, and ImageJ. Results are presented as mean ± SD, with p‐value calculation methods indicated in the figure legends.
Author Contributions
H.S., M.D., and X.H. designed the experiments. H.S., H.P., W.W., Y.J., and Y.H. performed the experiments. H.S., Y.L., Y.Z., Y.C., and X.H. analyzed the data. H.S. and X.H. wrote the manuscript.
Funding
Key Technologies Research and Development Program 2024YFA1306101; National Natural Science Foundation of China, 92354301, 32230044 and 32321004.
Conflicts of Interest
The authors declare no conflict of interest.
Supporting information
Supporting File 1: advs73589‐sup‐0001‐SuppMat.zip;
Supporting File 2: advs73589‐sup‐0002‐FigureS1‐S7.zip;
Supporting File 3: advs73589‐sup‐0003‐TableS1‐S3.zip;
Supporting File 4: advs73589‐sup‐0004‐VideoS1‐S2.zip.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1A. R. Thiam , R. V. Farese Jr , and T. C. Walther , “The Biophysics and Cell Biology of Lipid Droplets,” Nature Reviews Molecular Cell Biology 14 (2013): 775–786.24220094 10.1038/nrm 3699 PMC 4526153 · doi ↗ · pubmed ↗
- 2A. Zadoorian , X. Du , and H. Yang , “Lipid Droplet Biogenesis and Functions in Health and Disease,” Nature Reviews Endocrinology 19 (2023): 443–459.10.1038/s 41574-023-00845-0PMC 1020469537221402 · doi ↗ · pubmed ↗
- 3A. L. S. Cruz , N. Carrossini , L. K. Teixeira , L. F. Ribeiro‐Pinto , P. T. Bozza , and J. P. B. Viola , “Cell Cycle Progression Regulates Biogenesis and Cellular Localization of Lipid Droplets,” Molecular and Cellular Biology 39 (2019): e 00374.30782775 10.1128/MCB.00374-18PMC 6469922 · doi ↗ · pubmed ↗
- 4T. Makiyama , T. Obama , Y. Watanabe , et al., “Behavior of Intracellular Lipid Droplets during Cell Division in Hu H 7 Hepatoma Cells,” Experimental Cell Research 433 (2023): 113855.37995922 10.1016/j.yexcr.2023.113855 · doi ↗ · pubmed ↗
- 5A. Herms , M. Bosch , B. J. N. Reddy , et al., “AMPK Activation Promotes Lipid Droplet Dispersion on Detyrosinated Microtubules to Increase Mitochondrial Fatty Acid Oxidation,” Nature Communications 6 (2015): 7176.10.1038/ncomms 8176 PMC 444679626013497 · doi ↗ · pubmed ↗
- 6M. D. Kilwein , T. K. Dao , and M. A. Welte , “Drosophila Embryos Allocate Lipid Droplets to Specific Lineages to Ensure Punctual Development and Redox Homeostasis,” Public Library of Science Genetics 19 (2023): 1010875.10.1371/journal.pgen.1010875 PMC 1044916437578970 · doi ↗ · pubmed ↗
- 7S. Boulant , M. W. Douglas , L. Moody , A. Budkowska , P. Targett‐Adams , and J. Mc Lauchlan , “Hepatitis C Virus Core Protein Induces Lipid Droplet Redistribution in a Microtubule‐ and Dynein‐Dependent Manner,” Traffic 9 (2008): 1268–1282.18489704 10.1111/j.1600-0854.2008.00767.x · doi ↗ · pubmed ↗
- 8Y. Gu , Y. Yang , X. Cao , et al., “Plin 3 protects against Alcoholic Liver Injury by Facilitating Lipid Export from the Endoplasmic Reticulum,” Journal of Cellular Biochemistry 120 (2019): 16075–16087.31119787 10.1002/jcb.28889 · doi ↗ · pubmed ↗