DDHD2, whose mutations cause spastic paraplegia type 54, enhances lipophagy via engaging ATG8 family proteins

Abstract

Hereditary spastic paraplegia (HSP) is a group of inherited neurodegenerative disorders characterized by progressive lower limb spasticity and weakness. One subtype of HSP, known as SPG54, is caused by biallelic mutations in the DDHD2 gene. The primary pathological feature observed in patients with SPG54 is the massive accumulation of lipid droplets (LDs) in the brain. However, the precise mechanisms and roles of DDHD2 in regulating lipid homeostasis are not yet fully understood. Through Affinity Purification-Mass Spectroscopy (AP-MS) analysis, we identify that DDHD2 interacts with multiple members of the ATG8 family proteins (LC3, GABARAPs), which play crucial roles in lipophagy. Mutational analysis reveals the presence of two authentic LIR motifs in DDHD2 protein that are essential for its binding to LC3/GABARAPs. We show that DDHD2 deficiency leads to LD accumulation, while enhanced DDHD2 expression reduces LD formation. The LC3/GABARAP-binding capacity of DDHD2 and the canonical autophagy pathway both contribute to its LD-eliminating activity. Moreover, DDHD2 enhances the colocalization between LC3B and LDs to promote lipophagy. LD·ATTEC, a small molecule that tethers LC3 to LDs to enhance their autophagic clearance, effectively counteracts DDHD2 deficiency-induced LD accumulation. These findings provide valuable insights into the regulatory roles of DDHD2 in LD catabolism and offer a potential therapeutic approach for treating SPG54 patients.

Introduction

Hereditary spastic paraplegia (HSP) is a group of inherited neurodegenerative disorders characterized by progressive spasticity and weakness in the lower extremities [1, 2]. HSP can be classified into pure forms, which primarily involve motor neuron degeneration, or complex forms, which present with additional neurological or non-neurological manifestations [3]. Currently, there are 80 identified genes or loci associated with HSP, numbered as spastic paraplegia (SPG) 1–80. The inheritance pattern of HSP subtypes can be autosomal dominant (AD), autosomal recessive (AR), or X-linked [4, 5]. These genes are involved in various cellular processes, such as axonal transport, endoplasmic reticulum shaping and dynamics, membrane trafficking, lipid metabolism, and mitochondrial function [3, 6]. One specific rare form of complex AR HSP is spastic paraplegia type 54 (SPG54), which is caused by biallelic mutations in the DDHD2 gene [7]. SPG54 typically presents as early-onset disease, with symptoms appearing in the first decade of life. Characteristic features of SPG54 include spastic gait, intellectual disability, thin corpus callosum, and the presence of a detectable lipid peak in the brain through magnetic resonance spectroscopy [7, 8]. Most variants in DDHD2 result in loss-of-function of the protein products, including nonsense or frameshift truncating mutations, while a smaller group of variants are missense mutations [7, 8].

The DDHD2 gene encodes the DDHD-domain-containing 2 protein, which belongs to the intracellular phospholipase A1 (PLA1) family, along with DDHD1 and SEC23IP [9]. The initial biochemical studies showed that DDHD1 and DDHD2 have the potential to function as lipases, hydrolyzing various (phospho) lipid substrates in vitro [10]. However, the specific functions of these enzymes and their endogenous substrates are not yet fully understood. Notably, studies on DDHD2^–/– mice have indicated that DDHD2 deficiency leads to a marked elevation of triacylglycerols (TAGs) levels in the brain [11, 12]. This elevation correlates with the accumulation of LDs in neurons and results in cognitive and motor abnormalities resembling SPG54. Biochemical assays have demonstrated DDHD2 as a principal brain TAG lipase, which indicates that the brain possesses a specialized pathway for metabolizing TAGs [11, 12]. Disruption of this pathway leads to the massive LD accumulation in neurons and the development of severe neurodegenerative symptoms.

LDs are specialized structures that store excess neutral lipids, such as TAGs and cholesterol esters (CE), to be used as an energy source when needed [13]. Moreover, the storage of neutral lipids in LDs protects cells from lipotoxicity caused by excess free fatty acids (FFAs) and cholesterol [14]. Lipophagy is a cellular process that selectively degrades LDs through autophagy, playing a crucial role in maintaining lipid homeostasis and preventing the abnormal accumulation of lipids. Lipophagy involves recognizing and engulfing LDs with double-membraned autophagosomes, which then fuse with lysosomes for degradation [15, 16]. To facilitate the targeted degradation of LDs, cells employ specific cargo receptors for lipophagy, such as ORP8 and Spartin [17–19]. These receptors recognize, bind, and deliver LDs into autophagosomes for subsequent degradation by way of LIR motifs that bind to ATG8 family proteins (LC3/GABARAPs). Accumulating evidence has highlighted that dysregulation of lipid metabolism and abnormal LD accumulation are pathological hallmarks in various neurodegenerative and neuroinflammatory disorders, but the regulatory mechanisms of lipophagy in these diseases still remain poorly understood [20].

In this study, we investigated the DDHD2-mediated lipid catabolism pathway by identifying DDHD2-interacting partners. Through AP-MS methods, we revealed a robust interaction between DDHD2 and LC3/GABARAPs. Subsequent functional studies have demonstrated that DDHD2 likely acts as a lipophagy cargo receptor to promote LD elimination through a LC3/GABARAP-binding dependent mechanism, in addition to its previously reported TAG lipase activity. These findings are in concert with the fact that DDHD2 deficiency leads to the pathological LD accumulation in SPG54.

Results

Identification of ATG8 family proteins as DDHD1/2 interactors

To elucidate the unidentified molecular mechanisms of DDHD2 in lipid homeostasis, we performed a comprehensive analysis utilizing the publicly available AP-MS dataset of the autophagy-related protein interactome [21]. Our primary objective was to investigate potential interactions between DDHD2 and autophagy-related proteins. Intriguingly, this analysis revealed that when employing any member of the ATG8 family proteins (LC3/GABARAPs) protein family as bait, dozens of peptides corresponding to DDHD2 were detected within their protein complexes. By contrast, the presence of peptides corresponding to DDHD1 was relatively limited (Supplementary Fig. 1A).

To validate the potential interactions between DDHD2 and LC3/GABARAPs, we employed DDHD2 as bait in AP-MS analysis to identify potential interacting proteins from 293 T cells transiently transfected with FLAG-DDHD2 (Fig. 1A). The MS data revealed enrichment of LC3B, GABARAP, and GABARAPL1 peptides in the purified DDHD2 complexes, which is consistent with the study mentioned above (Fig. 1B, Supplementary Table 1). Co-IP assays were subsequently conducted and the results showed that ectopically overexpressed DDHD2 displayed similar binding capacity with six numbers of ATG8 family proteins (LC3/GABARAPs) (Fig. 1C, D). Furthermore, endogenous co-IP assay results confirmed DDHD2 interactions with LC3/GABARAPs (Fig. 1E). Treatment of cells with DC-LC3in-D5, a small molecule that modifies LC3A/B protein at Lys49 and thus prevents its binding to LIR-containing partners [22], markedly reduced the interaction between DDHD2 and LC3A or DDHD2 and LC3B, but not DDHD2 and LC3C (Fig. 1F). Regarding DDHD1, we observed its interaction with LC3/GABARAPs as well, although the strength of these interactions is relatively weaker compared to the DDHD2-LC3/GABARAP interactions (Fig. 1D, Supplementary Fig. 1B).

Fig. 1. DDHD1 and DDHD2 interact with LC3/GABARAPs in cells.

A Tandem affinity purification of the DDHD2 protein complexes was performed in 293 T cells transfected with FLAG-DDHD2 for 24 h. The associated proteins were separated by SDS-PAGE and visualized by Coomassie blue staining. MW, molecular weight. B Quantifications of peptides identified by mass spectrometry analysis were presented, the total and unique number of peptides associated with the proteins. C, D 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. (E) The WCL from 293 T cells were prepared and then subjected to co-IP with IgG or anti-DDHD2 antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. F 293 T cells were transfected with the indicated plasmids for 24 h, and then the cells were treated with or without DC-LC3in-D5 (50 µM) for 24 h. The WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies.

Collectively, these data indicate that DDHD1 and DDHD2 have the capacity to interact with LC3/GABARAPs in cells.

Identification of two LIR motifs in DDHD2 required for its interaction with LC3/GABARAPs

Since the interactions between DDHD2 and LC3A/B were reduced by treatment with DC-LC3in-D5, it was hypothesized that one or more LIR motifs in DDHD2 mediate their interactions. There are a dozen potential LIR motifs (consensus: [W/F/Y]xx[L/I/V]) present in the DDHD2 protein sequence (Fig. 2A). To identify the authentic LIR motif(s), three truncated mutants of DDHD2 (D1, D2, and D3) were constructed, and their interactions with GABARAP, a representative of the ATG8 family proteins, were examined. However, co-IP assay results showed that three deletion mutants all retained the ability to interact with GABARAP (Fig. 2B). The DDHD2-D1 segment contains six potential LIR motifs. Each LIR motif was mutated, and the interaction capabilities of these mutants with GABARAP were tested (Fig. 2C). As shown in Fig. 2D, mutating the first potential LIR motif is sufficient to abolish the interaction between DDHD2-D1 and GABARAP, while the mutations of the other motifs did not have such effect, indicating that the first one is an authentic LIR motif. Similarly, another authentic LIR motif was identified in the D2 segment (Fig. 2E, F). In the D3 segment, none of the potential LIR motif mutations had any impact on its interaction with GABARAP. Unexpectedly, even simultaneous mutations of all the potential LIR motifs did not reduce the interaction between DDHD2-D3 and GABARAP (Fig. 2G, H). This fact suggests that D3 segment may bind to GABARAP independently of the LIR motifs or that the construction of the D3 segment potentially created cryptic binding sites. To clarify this issue, a full-length DDHD2-mLIR mutant with both LIR motifs mutated was generated. The co-IP assay results showed that DDHD2-mLIR mutations nearly abolished the interaction between DDHD2 and LC3/GABARAPs, indicating that these two conserved authentic LIR motifs function in the full-length DDHD2 (Fig. 2I, J). Thus, we speculated that the generation of deletion mutants may lead to the formation of cryptic binding sites, allowing non-physiological interactions between LC3/GABARAPs and D3 segment. Most of the known SPG54-associated DDHD2 mutants result in loss-of-function, including nonsense or frameshift truncating mutations, while a minor group of variants are missense mutations (Supplementary Fig. 1C) [8]. The co-IP assay result demonstrated that SPG54-associated DDHD2 mutants (W103R, V220F, and D660H) exhibited similar LC3/GABARAP-binding capacity as the wild-type DDHD2 (Supplementary Fig. 1D).

Fig. 2. Identification of two LIR motifs in DDHD2 required for its interaction with LC3/GABARAPs.

A Schematic representation of potential LIR motifs in DDHD2-FL and its deletion/point mutants. B 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. C Schematic representation of potential LIR motifs in DDHD2-D1 segment and its point mutants. D 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. E Schematic representation of potential LIR motifs in DDHD2-D2 segment and its point mutants. F 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. G Schematic representation of potential LIR motifs in DDHD2-D3 segment and its point mutants. H, I 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. J The protein sequences of two LIR motifs were aligned between human DDHD2 and its mouse homolog. Hs, Homo sapiens. Mm, Mus musculus.

Collectively, these data indicate that two authentic LIR motifs in DDHD2 are responsible for its interaction with LC3/GABARAPs.

LC3/GABARAP-binding capacity contributes to DDHD2’s LD-eliminating activity

The robust interaction between DDHD2 and LC3/GABARAPs led us to examine whether the LC3/GABARAP-binding capacity influences DDHD2’s LD-eliminating activity. To test this, we first generated DDHD2-KO HeLa cells through CRISPR/Cas9-mediated genomic editing. The ablation of DDHD2 expression didn’t affect basal or rapamycin-induced macroautophagy, as assessed by protein levels of p62 and LC3B-II (Fig. 3A, Supplementary Figs. 2A, 3A). We then visualized LDs using a widely-used LD-detecting probe BODIPY (BODIPY 493/503) in HeLa cells treated with extracellular oleic acid (OA). DDHD2 deficiency caused by either KO or shRNA-mediated knockdown (KD) resulted in an observable increase in LD numbers and sizes when compared to control cells (Fig. 3B, C, Supplementary Fig. 3B–D). Similar results were observed when flow cytometry analysis was performed to detect lipid content (Fig. 3D, E).

Fig. 3. The LC3/GABARAP-binding capacity contributes to DDHD2’s ability to eliminate LDs.

A DDHD2-KO HeLa cells were generated through CRISPR/Cas9 methods. The WCL from parental and DDHD2-KO HeLa cells were analyzed by WB with the indicated antibodies. B Parental or DDHD2-KO HeLa cells were treated with or without OA (0.5 mM) for 18 h, followed by 3 h of OA withdrawal. LDs were visualized using BODIPY 493/503 fluorescent dye (green). DAPI (Blue). Scale bar, 10 µm. C Quantifications of averaged LD number and total LD area were presented. Mean ± SD, n = 15 fields of view. D Parental or DDHD2-KO HeLa cells were treated with or without OA (0.5 mM) for 18 h, followed by 3 h of OA withdrawal. After fixation and LD staining using BODIPY 493/503, the cells were subjected to FACS analysis. E Mean Fluorescence Intensity (MFI) values of LD content D were calculated using FlowJo software. Mean ± SD, n = 3. F The WCL from control CRISPRa and DDHD2 CRISPRa HeLa cells were analyzed by WB with the indicated antibodies. G Control CRISPRa and DDHD2 CRISPRa HeLa cells were treated with or without OA (0.5 mM) for 18 h, followed by 3 h of OA withdrawal. LDs were visualized using BODIPY 493/503 (green). DAPI (Blue). Scale bar, 10 µm. H Quantifications of averaged LD number and total LD area in (G). Mean ± SD, n = 15 fields of view. I, J Parental or ATG7-KO HeLa cells were transfected with the indicated plasmids for 24 h. Then, the cells were treated with OA (0.5 mM) for 18 h followed by 3 h of OA withdrawal, stained with FLAG (red), BODIPY 493/503 (green), and DAPI (blue). Scale bar, 10 μm. Quantifications of total LD area per cell under different transfection were shown (J). Mean ± SD, n = 15 fields of view. P values are calculated by the Two-way ANOVA test in (C, E, H) and student’s t test (J).

Human Ewing’s Sarcoma SK-N-MC and mouse hippocampal neuron HT22 cells, which are frequently employed as in vitro cell models for investigating neuronal pathogenesis, were also tested. The impacts of DDHD2 deficiency on LDs in these cell lines mirrored that observed in HeLa cells (Supplementary Fig. 2B, C, 3E–J). Reciprocally, we examined whether enhancing DDHD2 expression could decrease LD formation. We utilized a CRISPR activation (CRISPRa) system, which incorporated MS2 and PP7 RNA stem-loop structures to join VP64 and dCas9 proteins. This manipulation substantially upregulated endogenous DDHD2 expression (Fig. 3F), leading to a significant reduction in LD accumulation induced by OA (Fig. 3G, H).

The function of DDHD2 as a brain TAG lipase may explain the significant changes in LD formation when its protein levels are altered. However, we questioned whether the LC3/GABARAP-binding capacity also partially contributes to DDHD2’s LD-eliminating activity. Thus, we generated a catalytically inactive serine to alanine mutant (S351A) of DDHD2. The LDs were largely eliminated in cells transfected with wild-type DDHD2. In comparison, LD-eliminating activity was considerably compromised but not entirely abolished in those transfected with the DDHD2-S351A mutant. Additionally, the DDHD2-mLIR mutant, which is unable to bind LC3/GABARAPs, showed reduced LD-eliminating activity when compared to DDHD2-WT. Any ability to eliminate LDs was completely lost when there were simultaneous mutations of the catalytic site and the LIR motifs (Fig. 3I, J). Moreover, we performed similar transfection in ATG7-KO HeLa cells, in which the canonical macroautophagy was abolished (Supplementary Fig. 3K). We observed that the DDHD2-S351A mutant totally lost the capacity to eliminate LDs in ATG7-KO cells (Fig. 3I, J). Also, DC-LC3in-D5 treatment decreased DDHD2’s LD-eliminating activity (Supplementary Fig. 3L, M). For comparison, we also assessed DDHD1 and found that while DDHD1 had some capacity to eliminate LDs, it was not as strong as DDHD2 (Supplementary Fig. 3L, M).

To further clarify whether the enzymatic activity and LC3/GABARAP-binding capacity of DDHD2 are interdependent, we conducted several assays. Firstly, the results from the in vitro TAG lipase activity assays confirmed that the mutation of catalytic serine (S351A) abolished the TAG lipase activity of DDHD2. In contrast, the DDHD2-mLIR mutant exhibited comparable TAG lipase activity to the wild-type DDHD2 (Supplementary Fig. 4A). These results indicated that mutations affecting the LC3/GABARAP-binding capacity of DDHD2 do not directly impact its TAG lipase activity in vitro. Secondly, the co-IP assay results demonstrated that the DDHD2-S351A mutant exhibited similar ATG8-binding capacity as the wild-type DDHD2, indicating that mutagenesis of the catalytic serine did not affect the binding of DDHD2 to LC3/GABARAPs (Supplementary Fig. 1D).

KLH45 is a potent and selective DDHD2 inhibitor, we observed a higher LD accumulation in DDHD2-KO HeLa cells compared to KLH45-treated HeLa cells (Supplementary Fig. 4B, C). While KLH45 treatment partially reversed the elimination of LDs induced by DDHD2 overexpression in DDHD2 Ca cells, it did not completely restore the levels to that of control (Supplementary Fig. 4D, E). These results indicate that the LD-eliminating activity of DDHD2 cannot be solely attributed to its TAG lipase activity. Adipose triglyceride lipase (ATGL) is a major cytosolic TAG lipase and regulator of LDs in most tissues. A recent study has indicated the cooperative role of ATGL and DDHD2 in the catabolism of neuronal TAGs [23]. We observed that NG-497 (a potent and selective ATGL inhibitor) treatment further exacerbated the DDHD2 deficiency-caused LD accumulation in HeLa cells (Supplementary Fig. 4F, G), indicating that DDHD2 and ATGL act in parallel to regulate cellular LD catabolism.

Collectively, these data suggest that the LC3/GABARAP-binding capacity of DDHD2 and the canonical autophagy pathway both contribute to DDHD2-mediated LD elimination.

DDHD2 enhances the colocalization between LC3B and LDs to promote lipophagy

We hypothesized that the interplay between LC3/GABARAPs and DDHD2 could help the recruitment of LC3/GABARAPs to LDs, thus promoting lipophagy. To validate this, HeLa cells were initially exposed to OA followed by lipid withdrawal to stimulate LD turnover. The treatment led to an elevated interaction between DDHD2 and LC3/GABARAPs (Fig. 4A, B). By overexpressing FLAG-DDHD2 in DDHD2-KO HeLa cells that were treated with OA, it became evident that DDHD2, LC3B, and LDs manifested as colocalized puncta (Fig. 4C, D). Noteworthily, DDHD2 deficiency reduced the colocalization between LC3B and LDs, but total LC3 puncta area was not affected. (Fig. 4E–G). To investigate whether LDs decorated by DDHD2 serve as intermediates of autophagic degradation, we examined their location relative to lysosomes. It was found that a proportion of LDs co-localized with the lysosomal marker LAMP1; however, their association was reduced in DDHD2-KO cells, corresponding with the role of DDHD2 in facilitating the delivery of LDs to the lysosomes (Fig. 5A, B). To further demonstrate the function of DDHD2 in autophagic degradation of LDs, we employed a lipophagy reporter system in which mCherry and GFP were tandem-tagged to livedrop, a specific marker for LDs [17]. In mCherry-GFP-livedrop expressing parental HeLa cells, treatment with OA and subsequent withdrawal stimulated the production of red puncta (mCherry⁺ and GFP⁻). This corresponds to the quenching of acidity-sensitive GFP in lysosomes, indicating enhanced lipophagy flux. In contrast, most of the livedrop puncta remained yellow (mCherry⁺ and GFP⁺) in DDHD2-KO HeLa cells (Fig. 5C, D). Consistent with this observation, FACS analysis of DDHD2-KO HeLa cells showed a significantly lower number of cells in the mCherry > GFP gate (Fig. 5E, F). Conversely, DDDH2 Ca cells exhibited attenuated autophagic degradation of LDs than that in control cells (Supplementary Fig. 4H, I).

Fig. 4. DDHD2 enhances the colocalization of LC3B and LDs.

A, B 293 T cells were transfected with the indicated plasmids for 24 h. Then, the WCL were prepared and subjected to co-IP with anti-FLAG antibody. The WCL and immunoprecipitates were analyzed by WB with the indicated antibodies. C DDHD2-KO HeLa cells were transfected with FLAG-DDHD2 for 24 h. Then, the cells were treated with OA (0.5 mM) for 18 h followed by 3 h of OA withdrawal, stained with BODIPY 493/503 (green), FLAG (red), LC3 (magenta), and DAPI (blue). Scale bars,10 μm. D Fluorescent line intensity plot was used to assess the co-localization of DDHD2, LDs, and LC3B in (C). The profiles display the fluorescence distribution across the cell along a line indicated by the white dotted line (shown only for the merged insets). Red, green, and magenta fluorescence signals were independent and displayed relatively. E Parental and DDHD2-KO HeLa cells were treated with OA (0.5 mM) for 18 h followed by 3 h of OA withdrawal, stained with BODIPY 493/503 (green), LC3B (red), and DAPI (blue). Scale bars, 10 μm. F The overlaps between LDs and LC3B shown in (E) were quantified by Pearson’s coefficient analysis using Fiji software. Mean ± SD, n = 15. G The areas of LC3B puncta per cell shown in (E) were quantified using Fiji software. Mean ± SD, n = 15. P values are calculated by the One-way ANOVA test in (F, G).

Fig. 5. DDHD2 KO decreases lipophagy.

A Parental or DDHD2-KO HeLa cells s were treated with OA (0.5 mM) for 18 h followed by 3 h of OA withdrawal, stained with BODIPY 493/503 (green), LAMP1 (red), and DAPI (blue). Scale bars, 10 μm. B The overlaps between LDs and LAMP1 per cell shown in (A) were quantified by Pearson’s coefficient analysis using Fiji software. Mean ± SD, n = 15. C Representative confocal images of parental and DDHD2-KO Hela cells transfected with livedrop reporter plasmids. After transfection with livedrop reporter plasmids for 24 h, the cells were then seeded in 35 mm glass-bottom dishes and incubated with OA for 16 h followed by 6 h of OA withdrawal. Subsequently, the cells were stained with Hoechst (1 μg/mL) and underwent confocal analysis. Hoechst (Blue). Scale bar, 10 μm. D The Pearson’s coefficients of GFP overlap with mCherry shown in (C) were quantified using Fiji software. Mean ± SD, n = 15. E FACS scatter plots depicted the levels of GFP and mCherry fluorescence in GFP-mCherry livedrop expressing parental or DDHD2-KO HeLa cells. After transfection with livedrop reporter plasmids for 24 h, cells were treated with OA (0.5 mM) for 18 h followed by 6 h OA withdrawal, then the cells were subjected with flow cytometry analysis for the changes of acidified livedrop. F Quantifications of acidified livedrop populations shown in (E) were calculated using Flowjo software and visualized using GraphPad Prism 9 software. Mean ± SD, n = 3. P values are calculated by the One-way ANOVA test in (B, D, F).

Collectively, these data indicate that DDHD2 functions as a cargo receptor to boost lipophagy.

LD·ATTEC reverses DDHD2 deficiency-caused LD accumulation

Currently, SPG54 is recognized as an incurable neurodegenerative disorder, for which no targeted therapy has been reported. The most prominent pathological feature seen in affected patients is the excessive accumulation of LDs in the brain. Recently, we have successfully developed a group of small molecule compounds called Lipid Droplets·AuTophagy TEthering Compounds (LD·ATTECs) [24]. These compounds have demonstrated the remarkable capability to stimulate autophagic clearance of LDs in various cultured cell lines and genetically modified mouse models [24, 25]. Therefore, we conducted tests to determine whether LD·ATTEC can reverse DDHD2 deficiency-induced LD accumulation. As shown in Fig. 6A, B, the LDs induced by OA were more prominent in DDHD2-KO HeLa cells. However, LD·ATTEC treatment efficiently eliminated the LDs in both OA-treated parental or DDHD2-KO HeLa cells. Similar results were observed in neuronal HT22 cells (Fig. 6C, D). Additionally, LD·ATTEC efficiently eliminated LDs in parental or DDHD2-KO HT22 cells cultured with an added lipid mixture for a week (Fig. 6E, F).

Fig. 6. LD·ATTEC reverses DDHD2 deficiency-caused LD accumulation.

A Parental or DDHD2-KO HeLa cells were treated with BSA, OA (0.5 mM), or OA (0.5 mM) + LD·ATTEC (5 μM) for 24 h, then the cells were stained with BODIPY 493/503 (green) and DAPI (blue). Scale bars,10 μm. B Quantifications of averaged LD number and total LD area presented in (A). Mean ± SD, n = 15 fields of view. C Parental or DDHD2-KO HT22 cells were treated with BSA, OA (0.5 mM), or OA (0.5 mM) + LD·ATTEC (5 μM) for 24 h, then the cells were stained with BODIPY 493/503 (green) and DAPI (blue). Scale bars,10 μm. D Quantifications of averaged LD number and total LD area presented in (C). Mean ± SD, n = 15 fields of view. E Parental or DDHD2-KO HT22 cells were cultured in normal media (NM) or lipid-rich media (LM) for one week. After that, the cells were treated with or without LD·ATTEC (5 μM) for 24 h and then stained with BODIPY 493/503 (green) and DAPI (blue). Scale bars,10 μm. F Quantifications of averaged LD number and total LD area presented in (E). Mean ± SD, n = 15 fields of view. P values are calculated by the Two-way ANOVA test in (B, D, F).

Collectively, these data indicate that LD·ATTEC can effectively eliminate accumulated LDs caused by DDHD2 deficiency in cell culture models.

Discussion

Based on our finding, we propose that DDHD2 is a key mediator of lipophagy by engaging LC3/GABARAPs and its deficiency causes SPG54 via delaying LD clearance (Fig. 7). Inefficient or dysfunctional autophagy can lead to the accumulation and aggregation of misfolded proteins, which are common pathological characteristics observed in various neurodegenerative diseases [26]. Furthermore, the accumulation of LDs has also been reported in Alzheimer’s disease, Parkinson’s disease, HTT disease, and Amyotrophic lateral sclerosis, raising the possibility that the pathology of these disorders is not solely a proteinopathy but also a lipidopathy [27]. However, these disorders usually show defects in the general autophagy machinery, the direct evidence for lipophagy abnormality as a critical step in the disease pathology is currently lacking. In most tissues, ATGL performs the initial and rate-limiting step of lipolysis. Mice with ATGL ablation have a severe lipolytic defect leading to systemic TAG accumulation in various tissues. However, no LD accumulation was observed in the neurons of these mice [28]. In contrast, DDHD2 serves as a key TAG lipase in the mammalian brain, where its dysregulation leads to vast LD buildup in neurons, but other tissues are generally unaffected [11]. Our study has revealed that DDHD2 deficiency selectively affects lipophagy, without impacting general macroautophagy. Therefore, dysregulation of DDHD2 caused by SPG54-associated mutations results in defects in both lipolysis and lipophagy, thus providing an explanation for the massive LD accumulation in the brains of SPG54 patients. There is significant crosstalk between these two pathways, as evidenced by the fact that ATGL overexpression enhances LC3 colocalization with LDs in primary hepatocytes and promotes lipophagy [29]. Another example is SPG20 (Troyer syndrome), which is caused by mutations in the SPART (spartin) gene. Recent research has shown that spartin functions as a lipophagy cargo receptor, and disruption of spartin leads to LD accumulation in cultured human and murine neurons [18]. Additionally, defects in the biogenesis, maturation, turnover, movement, and organelle contacts of LD have been observed linked to various proteins implicated in HSP [30]. This suggests that dysregulation of LD biology may be a common feature in the pathogenesis of HSP.

Fig. 7. Schematic diagram illustrating that DDHD2 enhances lipophagy via engaging ATG8 family proteins.

DDHD2 plays a crucial role in promoting lipophagy through its interaction with LC3/GABARAPs, and its absence may lead to SPG54 by delaying the clearance of LDs. LD·ATTEC treatment effectively eliminates abnormally accumulated LDs in DDHD2-deficient cells.

Despite the growing recognition of the active roles of LDs in regulating cell fitness in the brain, there have been limited attempts to assess whether eliminating LDs could alleviate adverse neurological phenotypes, mainly due to the lack of pharmacological intervention tools. The early onset and monogenic nature of SPG54 make it an ideal model for testing the benefits of LD elimination. This is because LD accumulation is a predominant pathological feature in this disease whereas the general macroautophagy machinery remains intact, distinguishing it from other well-known neurodegenerative disorders. We have recently developed a degrading strategy called LD·ATTEC, which tethers LDs to autophagosomes through their direct binding to LC3, enabling their degradation [24]. The LD-eliminating activity of LD·ATTEC has been verified in genetic mouse models of obesity, non-alcoholic steatohepatitis, and age-related macular degeneration [24, 25]. In this study, we demonstrate that LD·ATTEC can effectively eliminate accumulated LDs on DDHD2-deficient cells. Further investigation is needed to assess the in vivo effects of LD·ATTEC in DDHD2^-/- mice and evaluate its potential to improve cognitive and motor abnormalities. Another concern is about drug delivery. Most brain-targeting drugs currently in clinical practice are soluble small molecules with molecular weights under 400 Da. This presents a challenge for the LD·ATTEC molecule (772.95 Da) as it may not efficiently cross the blood-brain barrier (BBB) [31]. However, intracerebroventricular (ICV) injection could offer a solution. This method bypasses the BBB by injecting the drug directly into the brain’s ventricular system, allowing it to reach the cerebrospinal fluid [32]. Using ICV injections would allow us to examine the in vivo effects of LD-ATTEC on DDHD2^-/- mice, providing valuable data for future studies.

Materials and methods

Cell line, cell culture, transfection, and lentiviral infection

The HeLa, 293 T, HT22, and SK-N-MC cell lines were obtained from the American Type Culture Collection and maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37 °C with a 5% CO₂. Transient transfection was performed using EZ Trans reagent (Shanghai Life-iLab Biotech). Lentiviral transfection was carried out by co-transfecting pLKO.1-based shRNA KD or pCDH-based overexpression plasmids with virus-packing constructs into 293 T cells. After 48 h, the viral supernatant was collected for further use. The cells were infected with the viral supernatant in the presence of polybrene (8 µg/mL) and subsequently selected and enriched through treatment with puromycin (1 μg/mL) or fluorescence-activated cell sorting (FACS).

Plasmid construction

The plasmids used for transient overexpression were constructed using the pCMV-FLAG/Myc vector (Clontech). Point and deletion mutants were engineered utilizing the KOD-Plus-Mutagenesis Kit (TOYOBO) following the manufacturer’s instructions. Single-guide RNAs (sgRNAs) targeting human or mouse DDHD2 (http://crispr.mit.edu) were subcloned into the pSpCas9(BB)-2A-Puro (PX459) vector for gene knockout (KO). Short hairpin RNAs (shRNAs) targeting DDHD2 were subcloned into the pLKO.1 puro vector (Addgene) for gene KD. The sequences of gene-specific sgRNAs and shRNAs are listed in supplementary table 2.

Antibodies, chemicals, and kits

The information of antibodies, chemicals, and kits used in this study is listed in Supplementary Table 3, 4.

Generation of gene KO cell lines

PX459 plasmids carrying a gene-specific sgRNA were transfected into the indicated cells. 24 h after transfection, puromycin (1 μg/mL) was added for positive selection. After four days, surviving cells were isolated using FACS and then seeded into 96-well plates for single clone cultures. Ten days later, genomic DNAs from the single cell clones were extracted, followed by PCR of the target region and Sanger sequencing. Single cell clones containing frameshift mutations were chosen, and the protein levels of DDHD2 or ATG7 were further confirmed through Western blot (WB) analysis.

CRISPRa/Cas9-mediated endogenous DDHD2 overexpression

SgRNAs targeting the DDHD2 promoter region were designed using the CRISPICK platform (https://portals.broadinstitute.org/gppx/crispick/), based on sequence data obtained from Ensembl (https://useast.ensembl.org/). Cells were infected with lenti-dCas9-VP64-Blast (Addgene), followed by a selection period using blasticidin (10 µg/mL) for 5 days. Subsequently, dCas9-VP64-infected cells were subjected to a secondary infection using pXPR502 (Addgene), containing either negative control or DDHD2-targeting CRISPR activation sgRNAs. These cells underwent selection using puromycin (1 µg/mL) for another four days. The detailed sgRNA sequences are listed in Supplementary Table 2.

Protein complexes purification and mass spectrometry analysis

293 T cells were transfected with pCMV-FLAG-DDHD2 constructs for 24 h. The Whole cell lysates (WCL) were prepared using NP40 lysis buffer [20 mM tris-Cl (pH 7.4), 100 mM NaCl, 0.2 mM EDTA, 0.5% NP-40, and 1× protease inhibitor cocktail], and then incubated with anti-FLAG antibody-conjugated M2 agarose beads (Smart-Lifesciences) at 4 °C overnight with rotary shaking. The beads were washed five times with BC100 buffer containing 0.2% Triton X-100 and eluted with 3×FLAG peptide. The eluted proteins were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) on a gradient gel (Bio-Rad). The protein bands were excised and subjected to mass spectrometry-based sequencing analysis. The proteomic data of mass spectrometry have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD044430.

In vitro TAG hydrolase activity assays

The TAG hydrolysis activity of DDHD2 was assessed utilizing the Lipase (LPS) Activity Assay Kit (Solarbio) following the manufacturer’s instructions. LPS catalyzes the conversion of triglycerides into fatty acids. Utilizing the copper soap method to measure the rate of fatty acid production enables the calculation of LPS activity. Briefly, 293 T cells were transiently transfected EV, DDHD2-WT, or its mutants (S351A or mLIR) for 24 h. The WCL were collected using buffer 1 (1 mL), followed by sonication for preparation. A standard curve was established by diluting standard solutions to specific concentrations. Based on the absorbance value at 710 nm, a linear regression equation was derived. Subsequently, the concentration of fatty acids in the sample can be determined by subtracting the absorbance value of the blank control at 710 nm from the absorbance value of the sample at 710 nm and then substituting this value into the linear regression equation.

RNA isolation and quantitative RT-PCR

Total RNA from cells were extracted by using TRIzol reagent (TIANGEN), followed by reverse transcription into cDNA using the HiScript III First Strand cDNA Synthesis Kit (Vazyme). The synthesized cDNAs were then subjected to PCR amplification using ChamQ SYBR qPCR Master Mix (Vazyme) in CFX Real-Time PCR system (Bio-Rad). The relative mRNA levels of DDHD2 were quantified using the 2^−ΔΔCT method with normalization to GAPDH. The primer sequences are listed in the Supplementary Table 2.

OA-induced LD formation and LD detection

A 10 mM stock solution of OA was prepared using 3 mM fatty-acid-free BSA (Millipore) in PBS. The mixture was incubated in a shaking incubator at 37 °C for 1 h to ensure complete OA dissolution. The obtained stock solution was filtered using a 0.22 μm filter and stored at −20 °C. To induce LDs, an OA-BSA solution (500 μM OA) or Lipid Mixture (Millipore) was used, with 10% BSA as the control. Fresh solutions were prepared for each assay. After 12–18 h induction and 3 h withdrawal, cells were washed with ice-cold 1×PBS. Cells were fixed with 4% paraformaldehyde (PFA, Sangon Biotech) for 10 min at room temperature. After fixation, the cells were washed thrice with 1×PBS for 5 min. For LD staining, BODIPY 493/503 dye (Thermo) was resolubilized in anhydrous DMSO to create a 1 mg/mL stock solution and used for LD staining at a dilution ratio of 1:1000. Staining was carried out for 30 min at room temperature, shielded from light. The cells were washed thrice with 1×PBS and mounted on microscope slides using Flouroshield containing DAPI (Abcam). For immunostaining, the cells were permeabilized with 0.2% Triton X-100 in 1×PBS for 10 min at room temperature and then blocked with 5% BSA for 1 h. The cells were incubated with primary antibodies at 4 °C overnight, followed by secondary antibodies and BODIPY 493/503. The number and size of LDs were analyzed using Fiji software.

Confocal microscopy and analysis

Cells adhered to slides were prepared following the protocol and subsequently imaged using a laser scanning confocal microscope (FV3000 or Zeiss 880). For live-cell imaging, cells with specific manipulations were plated in 35 mm glass bottom dishes (NEST Biotechnology) and incubated with OA or drug treatment. Following treatment, the cells were stained with Hoechst (1 μg/mL, Thermo) for nuclei approximately 10 min before scanning. Three-dimensional eight-bit image stacks of the various fluorescent signals were sequentially captured using a 63× (1.4 NA) objective, with an interval of 0.35 μm between each image. These image stacks were then projected into single stacks for two-dimensional analysis using either a maximum intensity projection macro in the Zeiss Zen software or a similar macro in Fiji software.

The images were imported into Fiji software for further analysis of the fluorescent signals in each cell. The Coloc 2 function of Fiji was used to assess the level of colocalization between different signals. For quantification of the number and area of LDs, the cell outlines were manually delineated based on the BODIPY channel. After applying a defined threshold above the background and autofluorescence, all relevant BODIPY 493/503 signals per cell were identified. The values for LD number and total LD area per cell were calculated using the “analyze particles” tool. At least 15 cells from each condition were included in this statistical analysis. Data analysis and graphical representation were performed using Excel or GraphPad Prism 9.

Flow Cytometry Analysis

The cells subjected to different treatments were treated with trypsin (Gibco Trypsin with 0.25% EDTA) and then washed three times with 1xPBS in 1.5 mL Eppendorf tubes. The resulting cell pellets were resuspended and fixed in 4% PFA, followed by three washes with 1×PBS. Subsequently, the cells were incubated with BODIPY 493/503 (diluted 1:1000 from a 1 mg/mL stock solution in DMSO) in PBS for 30 min at room temperature. After three washes, the cells were resuspended in 500 μl of 1xPBS and filtered through 45 µm strainers prior to FACS analysis. All flow cytometry analyses were carried out using a Fortessa flow cytometer (BD Bioscience). The subsequent data analysis was conducted using FlowJo software.

Statistical analysis

Statistical tests were selected based on the number of groups being compared. Paired or unpaired two-tailed Student’s t-test was employed for comparisons between two groups, while one-way ANOVA or two-way analysis of variance ANOVA were utilized for analyses involving multiple groups, unless otherwise specified. P values were calculated using GraphPad Prism 9 to ensure consistent and reliable statistical analysis.

Author contributions

C.W. conceived the study. F.J. and X.W. performed the experiments and data analyses. C.W., B.L., S.Z. and Y.F. analyzed and interpreted the data. B.L. provided LD·ATTEC compound. C.W. wrote the manuscript.

Funding

This work was in part supported by the National Natural Science Foundation of China (No. 92357301, 32370726, 91957125, 81972396 to C.W. 31821002, 31930062 to S.Z., 92049301, 82050008 to B.L., 31970748 to Y.F), the State Key Development Programs of China (No. 2022YFA1104200 to C.W.; 2018YFA0800300 to S.Z), the Natural Science Foundation of Shanghai (No. 22ZR1406600 to C.W), Science and Technology Research Program of Shanghai (No. 9DZ2282100). Science and Technology Commission of Shanghai Municipality (22S11900100 to Y.F).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41418-024-01261-1.