SCD1 and SCD5 modulate PARP-dependent DNA repair via fatty acid desaturation in glioblastoma

SUMMARY

Glioblastoma (GBM) relies on fatty acid metabolism for aggressive growth. This study identifies stearoyl-CoA desaturase-5 (SCD5), a brain-enriched isoform, as a critical driver of glioblastoma stem cell (GSC) maintenance and genomic stability. While SCD1’s role in GBM is well-established, our research reveals that SCD5 plays a non-redundant role by preferentially desaturating C18:0 and uniquely remodeling sphingolipids. Genetic silencing of SCD5 disrupts the cell cycle, impairs DNA repair, and triggers parthanatos—a form of cell death caused by PARP1 hyperactivation. Mechanistically, loss of SCD activity or saturated fatty acid accumulation triggers PARP1 hyperactivation and subsequent degradation, depleting RAD51 to compromise homologous recombination and induce parthanatos. These findings uncover a lipid-mediated vulnerability in GBM, linking fatty acid desaturation to PARP1-dependent genome integrity. Targeting SCD5 may offer a therapeutic strategy to eliminate therapy-resistant GSCs and enhance the efficacy of genotoxic or immunotherapeutic interventions.

Graphical abstract

In brief

The role of SCD5 in cancer metabolism is unclear. Mnatsakanyan et al. show that both SCD1 and SCD5 sustain glioblastoma stem cell survival by regulating lipid homeostasis and DNA repair. Its loss triggers saturated fatty acid accumulation, PARP1 hyperactivation, and parthanatos, exposing a metabolic vulnerability in therapy-resistant cancer stem cells.

INTRODUCTION

Glioblastoma (GBM) remains the most aggressive and therapy-resistant primary brain tumor in adults. Despite standard treatment involving maximal surgical resection followed by radiation therapy (RT) and temozolomide (TMZ), which induce lethal DNA double-strand breaks, nearly all tumors recur due to intrinsic or acquired resistance.¹ This resistance is largely driven by active DNA damage repair (DDR) mechanisms,^2,3 which are particularly enhanced in GBM stem-like cells (GSCs).^4–6 These observations underscore the need to target repair pathways in combination with conventional therapies.

Central to DDR is poly(ADP-ribose) polymerase 1 (PARP1), which is overexpressed in various cancers, including GBM.^7,8 In GSCs, PARP1 plays a vital role in maintaining genomic stability and facilitating DNA repair, thereby enabling resistance to treatments such as RT and TMZ.⁹ PARP1 also interacts with key DNA repair proteins, such as RAD51,^10,11 which is essential for homology-directed repair (HDR), further contributing to GSC resilience.⁶ Beyond its role in DDR, PARP1 is increasingly implicated in metabolic regulation, including lipid metabolism,^12,13 a critical driver of tumorigenesis, progression, and therapy resistance.^14,15 Alterations in lipid metabolism and fatty acid biosynthesis, frequently observed in GBM,¹⁶ support rapid tumor growth and promote survival under therapeutic stress. Notably, upregulation of de novo lipogenesis is associated with poor clinical outcomes, and targeting lipid synthesis pathways seems therapeutically promising in preclinical GBM models.^15–18

Fatty acid desaturation generates unsaturated fatty acids (UFAs) that serve diverse structural and signaling functions within cells.¹⁹ UFAs are incorporated into phospholipids,²⁰ enhancing membrane fluidity, flexibility, and permeability, properties essential for maintaining cell integrity and supporting membrane trafficking and signal transduction.^20,21 Stearoyl-CoA desaturase 1 (SCD1), the most abundant desaturase in humans, catalyzes the conversion of saturated fatty acids (SFAs) into monounsaturated fatty acids (MUFAs),²² thereby promoting tumorigenesis and cancer progression in GBM.^17,23 Prior work from our group and others established that SCD1 is required for GSC self-renewal and tumor initiation, and that SCD1 inhibition sensitizes GBM to DNA-damaging therapies.^17,23

Although SCD1 has been extensively characterized as the dominant desaturase, supporting tumor lipid metabolism, recent evidence points to additional desaturation pathways operating in GBM¹⁸ and in other cancers, such as liver and lung carcinoma.²⁴ Because brain tumors may rely on neural-lineage metabolic enzymes that remain poorly defined, we focused on stearoyl-CoA desaturase 5 (SCD5), a brain-enriched isoform²⁵ whose function in GBM has not been systematically explored. The restricted expression pattern of SCD5, coupled with evidence of alternative desaturation pathways in GBM, raises the possibility that SCD5 may support tumor growth and therapy resistance through mechanisms distinct from or complementary to SCD1.

Given these observations, we hypothesized that SCD5 plays a previously unrecognized and biologically significant role in GBM metabolism. In this study, we demonstrate that both SCD1 and SCD5 are critically required for GBM progression, functioning through overlapping yet complementary mechanisms that regulate fatty acid desaturation, cell-cycle progression, and DDR. These findings reveal an unexpected metabolic-genomic connection and highlight new therapeutic opportunities for targeting lipid desaturation as a vulnerability in GBM.

RESULTS

SCD5 is widely expressed across GBM subpopulations

SCD5 expression is largely restricted to the brain (Figure 1A) and represents the predominant fatty acid desaturase in neural cells, with particularly high expression in glial populations, including oligodendrocytes, oligodendrocyte precursor cells, and astrocytes (Figure S1A). Analysis of single-cell RNA sequencing (scRNA-seq) datasets from patients with GBM^26,27 revealed unexpectedly high SCD5 expression across multiple tumor subpopulations, whereas SCD1 expression was limited to distinct cellular clusters (Figures 1B–1C and S1B). In contrast, bulk RNA-seq data from the Glioma Longitudinal AnalySiS (GLASS) consortium²⁸ showed higher overall expression of SCD1 than SCD5 in GBM (Figure 1D). This apparent discrepancy is likely attributable to tumor heterogeneity, as bulk RNA-seq integrates transcriptional signals from both tumor and non-tumor cell populations. A similar pattern was observed in normal brain tissue. While scRNA-seq data revealed predominant SCD5 expression in neural cells, bulk transcriptomic analyses favored SCD1 expression (Figures 1E and 1F), suggesting that SCD1 is expressed by a broader range of cell types beyond the neural lineage (Figure S1C). Together, these findings indicate that elevated SCD5 expression in GBM reflects inheritance from the neural cell of origin rather than an adaptive oncogenic response.

Figure 1. SCD5 is highly expressed in glioma cells at both transcriptional and translational levels.

(A) Representation of human tissue distribution with the expression of either SCD1 or SCD5 for each respective tissue determined by Human Protein Atlas (HPA) bulk RNA-seq data.

(B and C) scRNA-seq analysis of two independent datasets (GEO: GSM3828673, n = 28 patients; GEO: GSE2285000, n = 39 patients) showing SCD1 and SCD5 expression patterns across cellular subpopulations.

(D) Bulk RNA-seq analysis of glioma tissues from GLASS consortium database comparing SCD1 and SCD5 expression.

(E and F) scRNA-seq (E) and bulk RNA-seq (F) analysis of normal brain tissue from the HPA database.

(G) Immunohistochemical staining of SCD1 and SCD5 in healthy brain tissue and tissue from patients with GBM (n = 3 each; top: overview; and bottom: high-magnification insets). Scale bar, 50 μm.

(H) Immunoblot analysis of SCD1 and SCD5 protein levels in eleven patient-derived GSC lines and two iPSC-derived NSCs and quantification of SCD1 and SCD5 protein levels normalized to GAPDH.

(I) Real-time qPCR analysis of the expression of GSCs markers as well as SCD1 and SCD5 in patient-derived GSCs and iPSCs-derived NSCs.

(J) Immunostaining for SCD1 and SCD5 in GSCs. Scale bar, 50 μm.

(K) Confocal microscopy analysis of SCD1/SCD5 co-localization with Pearson’s correlation coefficient (R) quantification. Scale bar, 25 μm.

Regarding the role of SCD5 in GBM tumorigenesis, analysis of The Cancer Genome Atlas (TCGA) data showed that patients with GBM with high SCD5 expression do not exhibit worse clinical outcomes (Figure S1D). Consistently, data from the GLASS consortium indicated that treatment with alkylating agents or radiation does not significantly alter SCD1 or SCD5 expression in either primary or recurrent GBM (Figures S1E and S1F). These findings were further validated in our patient-derived glioma stem cells (GSCs), in which neither SCD1 nor SCD5 expression changed substantially following 7 days of TMZ treatment or 24 h after radiation exposure (Figures S1G and S1H).

To investigate SCD5 protein expression, we developed and validated a custom antibody. Specificity was confirmed through overexpression and knockdown (KD) experiments in GSCs using multiple short hairpin RNAs (shRNAs) targeting both SCD5 transcript variants (Figures S1I and S1J) and further verified using miRFP670nano-tagged SCD1 and SCD5 constructs (Figure S1K). Immunohistochemical analysis demonstrated SCD5 protein expression in both GBM patient specimens and normal brain tissue (Figure 1G). The examination of eleven patient-derived GSC lines representing diverse molecular subtypes revealed elevated protein expression of both SCD isoforms, particularly SCD5, compared with two induced pluripotent stem cell (iPSC)-derived neural stem cells (NSCs), with distinct expression patterns between the two isoforms, suggesting differential transcriptional or post-translational regulation (Figures 1H and 1I). It is noteworthy that no significant difference in SCD isoform expression was observed across GSCs by subtype classification. Furthermore, co-immunostaining analysis demonstrated that SCD5 expression was uniform across GSCs, while SCD1 showed heterogeneous distribution, consistent with scRNA-seq findings (Figures 1J and S1L).

While SCD5 exists as two transcript variants (SCD5 and SCD5B), analysis of GLASS database samples and our GSC models revealed minimal expression of SCD5B (Figures S1M and S1N), suggesting that this variant has limited biological relevance in GBM. Structural analysis showed that although SCD5 shares 65% amino acid similarity with SCD1 (Figure S1O), it exhibits distinct distributions of PEST motifs (Figure S1P) that are known to regulate the low stability of the SCD1 protein.²⁹ However, protein stability assays following cycloheximide (CHX) treatment demonstrated comparable degradation rates for both isoforms (Figure S1Q). Subcellular localization studies confirmed that both enzymes reside in the endoplasmic reticulum (ER) but with minimal spatial overlap (Figures 1K, S1R, and S1S), suggesting that they occupy distinct functional microdomains. Together, these findings indicate that SCD5 is broadly expressed in GBM and, along with SCD1, likely contributes to tumor pathogenesis through complementary roles in lipid metabolism.

Perturbation of SCD1 and SCD5 expression reveals distinct lipidomic profiles

To characterize the functional roles of SCD1 and SCD5 in lipid metabolism, we performed ¹³C metabolic flux analysis in two GSC lines following isoform-specific KD. Both SCD1 and SCD5 KD resulted in reduced levels of newly synthesized MUFA (C16:1 and C18:1; Figure 2A). The levels of the polyunsaturated fatty acid (PUFA) C18:2 were undetected in the synthesized fraction, as expected (Figure 2A). However, we observed an increase in C18:2 levels in the total fraction, suggesting an increased PUFA uptake.

Figure 2. SCD1 and SCD5 silencing induces distinct lipidomic alterations in GSCs.

(A) Isotope tracing analysis (¹³C incorporation) of FA synthesis pathways in GSCs following SCD1 or SCD5 knockdown (48–72 h). Dot plot displays log₂ fold changes in FA levels versus controls (filled dots: p < 0.05, Student’s t test, n = 4).

(B) Volcano plot of lipid species alterations in SCD5-overexpressing GSCs versus controls (x axis: log₂ fold change; y axis: – log₁₀ p value).

(C) PCA of global lipid profiles following SCD1 or SCD5 knockdown.

(D and E) Venn diagrams showing shared (D) downregulated and (E) upregulated lipid species between SCD1- and SCD5-deficient GSCs.

(F) Differential abundance of lipid classes after 4-day SCD1/SCD5 silencing relative to SCR control (*p < 0.05, **p < 0.01, and ***p < 0.001; Student’s t test; n = 4).

(G) LION enrichment analysis of the top 10 significantly altered biological terms.

To comprehensively characterize the lipidomic changes mediated by SCD1 and SCD5 and elucidate their non-redundant functions, we employed shotgun lipidomics as an unbiased discovery approach. First, we confirmed the desaturase activity of SCD5 in GSCs through overexpression studies that demonstrated a significant increase in MUFA and a corresponding decrease in SFA (Figure 2B). Subsequent lipidomic profiling of SCD1- and SCD5-KD GSCs revealed distinct lipid remodeling patterns. Principal-component analysis (PCA) separated the lipidomic profiles of SCD1-KD and SCD5-KD cells from controls and from each other, indicating isoform-specific metabolic roles (Figures 2C and S2A). While SCD1 KD induced broad alterations across multiple lipid classes, SCD5 KD resulted in more selective changes, with only 15 lipid species showing significant downregulation (Figures 2D and S2B). Intriguingly, both KD conditions led to increased levels of several MUFA-containing lipids, suggesting compensatory activation of the remaining SCD isoform (Figure S2B). A potential exogenous source of these lipids can be ruled out, as the cell culture medium lacked C18:1.

Comparative analysis of affected lipid species revealed that ~74% of upregulated lipids following SCD5 KD were distinct when compared to SCD1 KD (Figures 2E and S2C), supporting non-redundant functions of these enzymes. Lipid classes analysis showed that both KDs decreased free FAs , while differentially affecting other lipid classes (Figure 2F). SCD5 KD increased most glycerophospholipids (Figure S2D) and upregulated sphingolipids (SLs), particularly sphingomyelin (SM) with saturated acyl chains (Figures 2F and S2E). In contrast, SCD1 KD increased diacylglycerols while decreasing cholesterol esters and triacylglycerides (TG), leading to reduced lipid droplet formation (Figures 2F and S2F).

To functionally interpret the lipidomic alterations, we performed Lipid Ontology (LION) enrichment analysis,³⁰ which links changes in lipid species to their biophysical and functional characteristics. In SCD5-KD GSCs, the most significantly upregulated terms were related to membrane characteristics, including increased bilayer thickness, elevated transition temperature, and reduced lateral diffusion. The most downregulated terms were associated with free FA abundance (Figure 2G). In contrast, SCD1 KD primarily enriched terms related to lipid saturation and significantly reduced those linked to lipid storage, lipid droplets, and neutral lipid content, consistent with the observed decrease in lipid droplet formation (Figures S2F). At the molecular level, SCD5 KD was associated with enrichment of ceramide- and phosphatidylethanolamine-related terms, whereas SCD1 KD was linked to an increase in plasmalogens (acylglycerophosphoethanolamines and lipids with vinyl ether bonds) and mitochondrial membrane components. Importantly, several of these lipid remodeling patterns were reversed in SCD5-overexpressing cells (Figure S2G). Notably, LION analysis revealed an increase in SFAs, particularly C18:0, following both SCD1 and SCD5 KD (Figure 2G). However, while SCD1 KD also led to an increase in C16:0, this was not observed in SCD5-deficient cells. Supporting these findings, SCD5 overexpression resulted in a decrease in C18:0 and a corresponding increase in C18:1 level (Figure S2G).

Overall, our findings demonstrate two key functional distinctions: (1) SCD5 appears specialized for C18:0 to C18:1 conversion, while SCD1 desaturates both C16:0 and C18:0, and (2) SCD5 KD uniquely increases SLs, particularly saturated SMs, without affecting neutral lipids, whereas SCD1 KD alters neutral lipid metabolism and lipid droplet formation. These differences, coupled with their distinct ER localization, suggest that SCD1 and SCD5 operate in separate metabolic networks despite sharing desaturase activity.

SCD5 preferentially desaturates C18:0 but not C16:0

Our results suggest that SCD1 and SCD5 exhibit differential substrate specificity. To test this, we analyzed FA composition after SCD1 or SCD5 KD (Figure S2A) or SCD5 overexpression (Figure S2B). Shotgun lipidomics revealed distinct FA profiles between these conditions. Notably, while SCD1 KD significantly reduced the desaturation index (DI) of C18:0, SCD5 KD showed no significant effect on the DI of either C16:0 or C18:0 (Figure 3A), suggesting potential functional compensation by SCD1. Importantly, SCD5 overexpression specifically enhanced the DI of C18:0 without affecting C16:0 DI (Figure 3B), providing further evidence that SCD5 selectively desaturates C18:0.

Figure 3. SCD5 preferentially desaturates C18:0 in GSCs.

(A) Desaturation indices (C16:1/C16:0 and C18:1/C18:0) in GSCs following 4-day SCD1 or SCD5 knockdown (mean ± SD; *p < 0.05, **p < 0.001, and ***p < 0.0001, Student’s t test, n = 4).

(B) Desaturation indices in SCD5-overexpressing GSCs.

(C) Quantification of the mono-unsaturation index (for fatty acids ≥ C18) by lipid class in GSCs following 4-day SCD1 or SCD5 knockdown.

(D and E) Dose-response viability curves of GSCs overexpressing SCD1 or SCD5 treated with (D) C16:0 or (E) C18:0 for 4 days (mean ± SD, n = 4).

(F) Three GSC lines were treated with FASNi ± C16:0/C18:0, 2 days after SCD1/SCD5 knockdown. Cell viability was measured 3 days after treatment and represented as fold change relative to BSA/DMSO-treated control (bar plot: mean ± SD; *p < 0.05, **p < 0.001, and ***p < 0.0001; n = 4).

(G) Immunoblot analysis of SCD1/SCD5 in GSCs treated with C16:0 or C18:0 for 8 or 24 h.

(H) qPCR analysis of SCD1/SCD5 mRNA 3 days post-knockdown in four GSC lines (M12, M76, M83, and M120).

(I) Immunoblot analysis of SCD1/SCD5 protein levels 3 days following knockdown.

In addition to potential compensation by the other isoform, another factor that may explain the unchanged C18:0 DI upon SCD5 KD is the dynamic nature of FA metabolism, as both C18:0 and C18:1 can be elongated by elongases into longer-chain MUFAs or SFAs. To account for this, we calculated the MUFA/SFA ratio across lipid classes for FAs ≥ C18 (Figures 3C and S3C). Consistently, SCD1 KD reduced the MUFA/SFA ratio across most lipid classes. In contrast, SCD5 KD selectively lowered the MUFA/SFA ratio in PIs, TGs, and SLs (Cer d18:1, HexCER, SM). Moreover, SCD5 KD increased the proportion of C18:0 within Cer d18:1, HexCER, and SM (Figure S3D), underscoring its role in SL metabolism.

Functional assays corroborated substrate specificity. The overexpression of SCD1 protected GSCs from lipotoxicity induced by both C16:0 and C18:0, whereas SCD5 overexpression only rescued C18:0 toxicity in two GSC lines (Figures 3D and 3E). To support these findings, we silenced either SCD1 or SCD5, inhibited de novo FA synthesis using fatty acid synthase inhibitors (FASNi), and attempted to rescue GSC viability by supplementing the culture medium with either C16:0 or C18:0 (Figure 3F). In GSCs expressing a non-targeting shRNA (SCR), both C16:0 and C18:0 rescued cell viability after FASN inhibition (Figure 3F). In SCD5 KD cells, adding either C16:0 or C18:0 prevented the loss of cell viability due to FASNi, since SCD1 expression remained unaffected and could desaturate both C16:0 and C18:0 to produce functional lipids. However, after SCD1 KD, adding C16:0 led to a significant increase in cell toxicity, even in GSCs not exposed to FASNi (Figure 3F). Conversely, C18:0 supplementation in SCD1 KD cells treated with FASNi rescued cell viability (Figure 3F), further supporting an exclusive role of SCD5 in C18:0 desaturation.

SFA supplementation increases SCD1 expression.^23,31 Further, we have previously shown that the pharmacological inhibition of SCD results in increased SCD1 protein levels, due to SFA accumulation.²³ Consistently, treatment with both C16:0 and C18:0 upregulated SCD1 expression in both GSC lines (Figure 3G). In contrast, only C18:0 induced SCD5 expression in both lines (Figure 3G). Additionally, SCD1 KD in four GSC lines triggered a transient compensatory upregulation of SCD5, which was only detectable at early time points (Figures 3H and 3I). This asymmetry is likely because SCD1 KD significantly reduces the DI of C18:0 across most of the lipidome, while SCD5 KD does not cause widespread changes beyond the SL class.

In sum, these data demonstrate that SCD5 is a dedicated C18:0 desaturase with a pronounced role in SL metabolism, while SCD1 exhibits broader activity toward both C16:0 and C18:0.

SCD5 is essential for GSC self-renewal, survival, and tumor growth

To explore whether SCD expression is linked to GSC stemness, we analyzed microarray data comparing different GBM cell types (GEO: GDS3885),³² including differentiated GBM cell lines (cultured in serum), GSCs grown as neurospheres, and primary GBM cells (Figure 4A). Correlation analysis revealed a positive correlation between SCD5 and markers of neural stem and progenitor cells (Figure 4B). We observed that cells cultured in serum (differentiation conditions) lost the expression of SCD5 along with stemness markers like SOX2, NES, OLIG1, PROM1 (CD133), and OLIG2. In contrast, non-primary cell lines, regardless of whether they are cultured with serum or serum-free conditions, exhibited increased SCD1 expression. These findings highlight key metabolic shifts, including the loss of SCD5 and the upregulation of SCD1, in established cell lines that do not recapitulate the SCD expression patterns observed in tumors or primary GSCs.

Figure 4. SCD1 and SCD5 are essential for GSC maintenance and tumorigenesis.

(A) Heatmap of SCD5, SCD1, and stemness marker expression across glioma cell lines, neurosphere-cultured GSCs, and primary GSCs (GDS3885 microarray dataset).

(B) Spearman correlation analysis between SCD isoforms and stemness markers using the same transcriptomics data.

(C) Morphological changes in GSCs following 6-day BMP4 differentiation (20 ng/mL). Scale bar, 100 μm.

(D) Heatmap of stemness (NES, OLIG1/2, and SOX2) and differentiation (GFAP) marker expression post-BMP4 treatment.

(E) Immunoblot analysis of SCD1/SCD5 protein levels after BMP4-induced differentiation.

(F) Viability assay of GSCs following 4-day SCD1/SCD5 knockdown (mean ± SD; n = 4; *p < 0.05, **p < 0.001, and ***p < 0.0001; Student’s t test). Lower: mRNA knockdown efficiency heatmap.

(G) ELDA of stem cell frequency in M12/M120 GSCs after SCD1/SCD5 knockdown (14 days).

(H) Cell viability assay of GSCs and non-stem glioma cells (NSGCs) generated by FBS-induced differentiation of GSCs following 4 days of SCD5 knockdown (n = Lower: heatmap showing mRNA knockdown efficiency and expression of stemness and differentiation markers.

(I and J) Drug response assays: (J) SCD1i-resistant GSCs (treated with the SCD1 inhibitor CAY10566, 1 μM) with SCD5 knockdown, and (K) SCD1i-sensitive GSCs overexpressing SCD5 (n = 4).

(K) In vivo tumor growth measured by Fluc bioluminescence imaging of control (SCR) versus SCD5-knockdown GSCs (n = 7 mice/group) at the indicated time points.

(L) Kaplan-Meier survival curves of mice bearing control or SCD5-deficient GSCs (n = 7/group).

Further supporting this, analysis of the DepMap database showed that established GBM cell lines (grown in serum) depend on SCD1 but not SCD5 (Figure S4A). To validate these findings, we differentiated patient-derived GSCs using BMP4 (Figures 4C–4E) and assessed the mRNA and protein levels of stemness and differentiation markers, along with SCD1 and SCD5. Upon differentiation, SCD5 mRNA levels decreased in all GSC lines tested (Figure 4D), while protein levels of both SCD1 and SCD5 were markedly reduced in three of four lines (Figure 4E). These results align with the microarray data, confirming that SCD5 is associated with stemness in GSCs. In contrast, while SCD1 levels declined during BMP4-induced differentiation, its upregulation in long-term serum culture suggests distinct regulatory mechanisms. This likely reflects the necessary role of SCD1 in maintaining basal lipid homeostasis and membrane integrity in rapidly dividing cells while also supporting GSC self-renewal through mechanisms independent of stemness regulation.

To assess whether SCD5, similar to SCD1, is required for GSC self-renewal, we performed shRNA-mediated KD in multiple GSC lines and normal human astrocyte (NHA) cultures. The KD of either SCD1 or SCD5 had minimal effects on NHAs but significantly reduced GSC viability (Figure 4F). Extreme limiting dilution assays (ELDA) and secondary sphere formation assays further demonstrated reduced stem cell frequency and impaired sphere-forming capacity (Figures 4G, S4B, and S4C), confirming that both enzymes are required for GSC self-renewal. Consistent with data from DepMap database, differentiation of GSCs with fetal bovine serum (FBS) rendered them insensitive to SCD5 downregulation (Figure 4H).

To rule out the off-target effects, we engineered a lentiviral vector expressing either a wild-type SCD5 (SCD5wt) or an shSCD5-resistant SCD5 (SCD5^shRes; resistant to shSCD5(2)) fused to miRFP670nano to monitor protein stability post-KD (Figures S4D and S4E). GSCs expressing miRFP670nano-SCD5^shRes maintained fluorescence and viability upon shSCD5(2) transduction, whereas shSCD5(1) reduced both (Figures S4F and S4G).

To determine whether the viability loss was due to the metabolic function of SCD5, we supplemented GSC medium with C18:1 (the direct product of SCD), post-KD. This significantly rescued the viability loss in all four GSC lines after SCD1 or SCD5 KD (Figure S4H). To test functional redundancy, we overexpressed SCD5 in SCD1-KD cells (and vice versa) and found only partial rescue (Figures S4I and S4J), supporting our lipidomics data that SCD1 and SCD5 have distinct biological roles despite shared enzymatic activity.

Resistance to SCD inhibitors (SCDi) has been associated with alternative desaturation pathways; however, the role of SCD5 in this context has not been previously explored.^18,24 Using a panel of SCDi-sensitive and -resistant GSCs we previously characterized,¹⁷ we found that SCD5 KD sensitized cells to pharmacological SCD1 inhibition with CAY10566 (Figure 4I), whereas overexpression of either SCD1 or SCD5 conferred resistance (Figure 4J). These findings suggest that SCD5 functions as an alternative source of MUFAs, compensating for SCD1 inhibition.

Finally, to assess the role of SCD5 in tumorigenesis, we intracranially implanted two independent patient-derived GSC lines (83 and 326) expressing either a scrambled control (SCR) or one of the two shRNAs targeting SCD5 into the brains of nude mice. SCD5 KD significantly suppressed tumor growth and prolonged survival (Figures 4K, 4L, S4K, and S4L), with 7 of 8 mice in the shSCD5(1) group and 4 of 7 mice in the shSCD5(2) group, implanted with 83 GSCs, surviving to 240 days and showing no detectable tumors by bioluminescence imaging (Figure S4M). Together, these findings establish SCD5 as a critical regulator of GSC maintenance and highlight its potential as a therapeutic target in GBM.

SCD activity is necessary for cell cycle progression

To identify gene expression changes and regulatory networks affected by SCD loss, we performed bulk RNA-seq on GSCs following SCD1 or SCD5 KD (Figure 5A). Gene set enrichment analysis revealed that the top five significantly downregulated biological processes were shared between SCD1 and SCD5 depletion, underscoring functional redundancy between these two enzymes (Figure 5B).

Figure 5. SCD1 and SCD5 regulate cell cycle progression in GSCs.

(A) Transcriptomic changes in GSCs following 4-day SCD1 or SCD5 knockdown (volcano plot: x axis = log₂ fold change; y axis = – log₁₀ p value).

(B) Top 5 significantly enriched GO terms from GSE GO analysis. Terms are ranked by lowest adjusted p values (dot size: gene count; color: adjusted p value; filled dots: padj < 0.05, n = 4).

(C) Enrichment map of the 40 most altered GO terms.

(D) Venn diagram of downregulated transcripts shared between SCD1- and SCD5-deficient GSCs.

(E) GO enrichment analysis of commonly downregulated genes in SCD1/SCD5 knockdown conditions.

(F) Cell cycle distribution analysis by flow cytometry 4 days post-SCD1/SCD5 knockdown.

Visualization of the top 30 significantly altered pathways in an enrichment map further demonstrated that clustered terms were predominantly associated with cell cycle progression, mRNA translation, and protein synthesis (Figure 5C). Consistent with this, Gene Ontology (GO) analysis showed that the top 10 most downregulated terms after SCD1/SCD5 KD were linked to cell division (Figure 5E). Conversely, upregulated terms included neurotrophic signaling and responses to metal ions (Figures S5A and S5B), with the former potentially reflecting a compensatory mechanism by which GSCs foster adaptive neuron-cancer cell interactions³³ to survive SCD depletion.

To functionally validate these observations, we analyzed cell cycle distribution using flow cytometry 4 days post-KD. Depletion of either SCD1 or SCD5 resulted in G1 phase accumulation and a concomitant reduction in G2/M phase cells (Figure 5F), mirroring the transcriptional changes. Furthermore, synchronized GSCs (double-thymidine block) released into the cell cycle exhibited delayed progression at 3, 6, and 9 h post-release upon SCD KD, compared to controls (Figure S5C). Collectively, these data demonstrate that SCD activity is essential for efficient cell cycle progression in GSCs.

SCD downregulation triggers PARP hyperactivation and parthanatos

While the mechanistic details remain unclear, our previous work demonstrated that SCD inhibition depletes RAD51, impairs HDR, and increases DNA damage in GSCs.^17,23 RNA-seq analysis revealed significant downregulation of base excision repair (BER) pathways following SCD1 or SCD5 KD (Figure 6A), corroborated by increased DNA damage, as evident through elevated γH2A.X levels (Figures 6B and S6A).

Figure 6. SFA accumulation induces DNA damage, PARP1 hyperactivation, and parthanatos in GSCs.

(A) Reactome pathway analysis of DNA damage repair mechanisms altered by SCD1/SCD5 knockdown (4 days).

(B) γH2AX immunofluorescence (left) and quantification (violin plot, right) in SCD1/SCD5-deficient GSCs (n > 100 cells; *p < 0.05, **p < 0.001, and ***p < 0.0001; Student’s t test). Scale bar, 25 μm.

(C) Immunoblot analysis of PARP1 and PARylation at the PARP1 expected molecular weight post-SCD1/SCD5 knockdown.

(D) PARP1 immunoprecipitation with PAR immunoblot in knockdown conditions.

(E) Time-course immunoblot analysis of PARP1, PAR, and γH2AX in GSCs treated with 150 μM of C18:0.

(F) Immunoblot analysis of PARP1, PAR, and RAD51 expression after treatment with H₂O₂ (0–2 mM, 15 min) and PARG inhibitor (1 μM, 24 h).

(G and H) Subcellular fractionation followed by immunoblot analysis of AIF expression following SCD1/SCD5 knockdown (G) or C18:0 treatment (H). TATA-binding protein (TBP; nuclear) and GAPDH (cytoplasmic) were used as loading controls. Quantification shows mean nuclear AIF/TBP ratio ± SD (n = 3 GSC lines: M12, M76, and M120).

(I) PicoGreen staining of nuclear DNA fragmentation post-SCD1/SCD5 knockdown. Scale bar, 10 μm.

(J) Immunoblot analysis of DNA damage markers in PARP1-GFP-expressing GSCs treated with 150 μM of C18:0 for 24 and 48 h.

(K) Schematic illustration of the experimental setup used to graft CFP-expressing GSCs into cerebral organoids and to treat them with an SCD1 inhibitor (SCD1i) for 3 days.

(L) CFP immunofluorescence of GSCs (MGG6) engrafted into 90-day-old cerebral organoids following 3 days of SCD1i treatment. Scale bars, 1,000 μm.

(M) Quantification of GSC infiltration into the core of cerebral organoids based on CFP staining (n = 14 sections).

(N) CFP and Hoechst staining of GSCs engrafted into 90-day-old cerebral organoids following 3 days of SCD1i treatment. White arrows highlight ring-shaped nuclear condensation (parthanatos), while yellow arrows indicate scattered nuclear condensates (apoptosis). Scale bar, 50 μm.

(O) CFP and AIF immunofluorescence of GSCs engrafted into 90-day-old cerebral organoids after 3 days of SCD1i treatment. Scale bar, 10 μm.

PARP1, the predominant PARP isoform in GBM (Figures S6B), plays a central role in DNA repair, including BER and HDR.³⁴ Prior studies have established that PARP inhibition transcriptionally represses RAD51,³⁵ implicating PARP1 in the regulation of DNA repair machinery. Although PARP1 mRNA levels were unchanged post-SCD KD (Figure S6C), we observed reduced PARP1 protein without caspase-mediated cleavage (no increase in PARP1c; Figure 6C). Strikingly, autoPARylation, detected by PAR bands at PARP1 molecular weight, was markedly enhanced (Figure 6C), confirmed by PARP1 immunoprecipitation (Figure 6D). Conversely, SCD1 or SCD5 overexpression reduced γH2A.X and restored PARP1 levels (Figure S6D), while BMP4-induced differentiation (which lowers SCD expression) mirrored SCD KD effects (Figure S6E). Further, we observed a robust inverse correlation between PARP1 and γH2A.X across GSC lines (Figures S6F and S6G), suggesting a mechanistic link between SCD, PARP1, and γH2A.X.

To determine whether SFA accumulation mediates the observed effects, we supplemented GSCs with C18:0, a direct substrate of SCD. This treatment triggered a time-dependent cascade characterized by PARP1 hyper-PARylation, subsequent depletion of PARP1 protein, and increased DNA damage, all of which peaked at 24 h before partially recovering—likely due to the activation of compensatory DNA repair mechanisms (Figures 6E and S6H). Notably, RAD51 levels declined markedly at 48 h, coinciding with the lowest observed PARP1 levels (Figure S6H). Further analysis revealed that the C18:0-induced DNA damage was mediated by reactive oxygen species (ROS), as co-treatment with the antioxidant N-acetylcysteine (NAC) significantly reduced γH2A.X expression (Figure S6I). These findings suggest that sustained PARP1 auto-PARylation leads to its degradation, which in turn contributes to RAD51 depletion. To directly test this hypothesis, we treated GSCs with hydrogen peroxide (H₂O₂) to induce PARP1 auto-PARylation, in combination with a PARG inhibitor (PARGi) to prevent PAR removal. This dual treatment led to a synergistic reduction in PARP1 protein levels (Figures 6F, S6J, and S6K), confirming that persistent PARylation promotes PARP1 degradation.

We next investigated the degradation mechanisms by treating GSCs with C18:0 for 48 h, followed by administration of either autophagy inhibitors (bafilomycin A or 3-methyladenine) or proteasome inhibitors (MG-132 or bortezomib). Both treatments restored PARP1 protein levels (Figure S6L), indicating that PARP1 is degraded via both autophagy and proteasomal pathways. Consistent with this, we observed increased LC3 cleavage following C18:0 treatment, SCD inhibition, or SCD KD (Figure S6M), demonstrating that SFA accumulation activates autophagy and contributes to PARP1 degradation.

Excessive PARP1 autoPARylation can induce parthanatos, a form of cell death where nuclear PAR translocates to the cytosol, binding apoptosis-inducing factor (AIF) and triggering its nuclear import.³⁶ SCD KD or C18:0 treatment increased nuclear AIF (Figures 6G, 6H, and S6N–S6P) and cytosolic PAR accumulation (Figure S6N). Once in the nucleus, AIF recruits migration inhibitory factor (MIF), which cleaves DNA into large fragments, resulting in chromatin condensation and fragmentation.³⁶ PicoGreen staining revealed chromatin condensation and fragmentation in SCD-depleted GSCs (Figure 6I), hallmarks of parthanatos. Elevated PARP1 levels allow cells to repair DNA without excessive accumulation of PARylated PARP, thereby preventing parthanatos.³⁷ Crucially, PARP1 overexpression mitigated DNA damage, auto-PARylation, and cell death (Figures 6J, S6Q, and S6R), confirming that PARP1 levels determine cell fate under SFA stress. PARP1-overexpressing SCD KD GSCs exhibited a DDR response comparable to that of control cells (Figure S6S). Furthermore, C18:1 supplementation completely rescued GSCs from SCD1i-induced parthanatos, as evidenced by the absence of PARP1 degradation and hyperPARylation (Figure S6T).

To determine whether these effects occur in a more complex microenvironment, we modeled the tumor niche using 90-day cerebral organoids (COs). These COs contain key cellular components of the GBM microenvironment, including neurons, astrocytes, and oligodendrocytes (Figure S7A). We introduced cyan fluorescent protein (CFP)-expressing GSCs into the COs and allowed them to invade for 14 days, followed by a 3-day treatment with an SCD1i (Figures 6K and S7B). Immunostaining confirmed that PARP1 expression in GSCs was significantly higher than in the normal cells present in the COs (Figure S7C). Within the organoids, GSCs were the predominant proliferative cell population, as indicated by Ki67 staining; this proliferation was markedly reduced by SCD1i treatment (Figure S7D). Consistent with this finding, SCD1i treatment significantly decreased GSC invasion (Figures 6L and 6M) and induced hallmarks of parthanatos, including nuclear condensation and the translocation of AIF to the nucleus (Figures 6N and 6O). Notably, analysis of cell death markers in SCD1i-treated organoids revealed evidence of both apoptosis and parthanatos. We observed ring-shaped nuclear condensation in cells undergoing parthanatos (where the nuclear envelope remains intact) and scattered nuclear condensates in apoptotic cells (due to the formation of apoptotic bodies; Figure 6O). This suggests that SCD inhibition triggers multiple cell death pathways within the heterogeneous tumor population.

In conclusion, SCD deficiency or SFA accumulation in GSCs hyperactivates PARP1, leading to its autodegradation via autophagy/proteasomal pathways, impaired DNA repair, and parthanatos.

DISCUSSION

GSCs drive therapeutic resistance and recurrence, yet their metabolic vulnerabilities remain incompletely understood. Here, we identify SCD5, a brain-enriched fatty acid desaturase, as a critical regulator of GSC self-renewal that links lipid metabolism to genomic stability through a novel PARP1-dependent mechanism.

Our study makes several conceptual advances beyond previous work on SCD1. First, we identify SCD5 as functionally non-redundant with SCD1, exhibiting a substrate preference for C18:0 and uniquely regulating SL metabolism. Second, SCD5 dependency is specific to the stem cell state as GSCs require SCD5 while differentiated tumor cells do not, revealing a vulnerability tied to cellular identity rather than oncogenic transformation. Third, we establish that SFA accumulation triggers PARP1 hyperactivation, ATP depletion, and parthanatos, bypassing apoptosis resistance in GSCs. Finally, we demonstrate that SCD inhibition induces dual cell death pathways (apoptosis and parthanatos), potentially reducing resistance development. These findings explain alternative desaturation pathways in GBM¹⁸ and establish lipid metabolism as a direct regulator of DNA repair and cell fate.

While the role of SCD1 in cancer is well established, the function of SCD5 remains poorly understood. Conflicting reports have described SCD5 depletion as having no effect on proliferation,^18,38 promoting necrosis,³⁹ or even enhancing cell growth.⁴⁰ Similarly, ectopic expression of SCD5 has been linked to both pro-tumorigenic effects, such as promoting thymic cancer cell migration⁴¹ and anti-tumorigenic outcomes, including reduced metastasis in breast cancer and melanoma models.^42,43 Our study demonstrates that SCD5 plays a critical role in the GBM, where it regulates lipid metabolism, tumor growth, and cell cycle progression. The observed role of SCD5 in cell cycle regulation aligns with previous reports implicating it in cyclin D1-mediated neuronal proliferation.⁴⁴ Notably, SCD5 expression is markedly reduced in differentiated cells and in serum-cultured GBM cell lines, which predominantly rely on SCD1 for fatty acid desaturation. This context-dependent expression underscores the need for careful model selection when investigating SCD5 function. Importantly, our metabolic and genetic analyses reveal functional redundancy between SCD1 and SCD5. In cells with high SCD1 activity, ectopic expression of SCD5 may disrupt lipid homeostasis by driving excessive MUFA production, potentially leading to lipotoxicity or lethal phospholipid imbalances. This mechanistic insight may help reconcile previously conflicting findings regarding the role of SCD5 in various cancer contexts.

Although both SCD1 and SCD5 exhibit delta-9 desaturation activity, their expression patterns and likely regulatory mechanisms diverge. SCD1 is broadly expressed and known to be regulated by transcription factors, such as SREBP1, LXR, and PPARα, while SCD5 is preferentially expressed in neural tissues and pancreatic cells,^25,45 suggesting cell-type-specific metabolic roles. In GBM, elevated SCD5 expression reflects inheritance from the neural cell of origin rather than adaptive upregulation by core stemness or oncogenic pathways. This contrasts sharply with SCD1, which appears to be more dynamically regulated in cancer contexts.

Despite this constitutive expression pattern, SCD5 is functionally essential for maintaining the stem cell state. GSCs exhibit specific dependency on SCD5, whereas serum-differentiated non-stem glioma cells remain largely insensitive to SCD5 loss. This stem cell-specific requirement, combined with the reciprocal compensatory regulation we observed between SCD1 and SCD5, underscores their interdependence in maintaining lipid homeostasis. Our data now explain a prior observation that some patient-derived GBM lines with low SCD1 expression can still synthesize MUFAs through undetermined mechanisms.¹⁸ SCD5 serves as the compensatory enzyme in these contexts, maintaining desaturation capacity even when SCD1 activity is compromised. This reciprocal compensation also explains why some tumors resist SCD1-specific inhibitors and supports the rationale for developing dual SCD1/SCD5 inhibitors.

The pharmacological specificity of existing SCD inhibitors remains unclear, particularly regarding their ability to target both SCD1 and SCD5. Our findings support the development of dual SCD1/SCD5 inhibitors for GBM therapy based on three key observations. First, scRNA-seq reveals that SCD5 maintains uniform expression across GBM subpopulations, suggesting broad therapeutic relevance. Second, while SCD1 exhibits ubiquitous expression, the brain-enriched distribution of SCD5 offers potential for reduced systemic toxicity, though potential effects on oligodendrocytes and other neural cells require further evaluation. Third, and most critically, SCD5 silencing potently impairs three fundamental GSC properties—self-renewal, tumor initiation, and cell cycle progression—while inducing parthanatos-mediated cell death. This triple vulnerability suggests that SCD5 inhibition could simultaneously target multiple oncogenic processes in GBM.

Our study reveals how lipid desaturation directly regulates DNA repair capacity and cell fate in GBM. SCD inhibition or genetic depletion leads to SFA accumulation, particularly C18:0, which causes DNA damage likely through ROS.^46–48 This triggers PARP1 hyperactivation as cells attempt DNA repair, but the resulting auto-PARylation consumes NAD⁺ and ATP, creating an energetic crisis that arrests cell cycle progression in G1 phase and depletes cellular energy reserves. The coordinated rescue of both ATP levels and cell viability by C18:1 supplementation or PARP1 overexpression establishes this as a causal pathway rather than correlative effects. Prolonged PARP1 hyperactivation ultimately leads to its degradation via autophagy and proteasomal pathways,^48,49 further compromising DNA repair capacity through RAD51 downregulation and initiating parthanatos through PAR-mediated AIF nuclear translocation.

While impaired SCD activity or SFA accumulation can promote caspase activation and apoptosis in GBM,⁵⁰ our data indicate that GSCs that express high levels of PARP1 predominantly undergo parthanatos following SCD inhibition. We propose that intracellular PARP1 levels determine cell death modality: lower PARP1 expression favors apoptosis, while elevated PARP1 levels, characteristic of GSCs and linked to stemness maintenance,⁵¹ shift cell fate toward parthanatos. This differential vulnerability creates a therapeutic advantage, whereby differentiated tumor cells undergo apoptosis, while apoptosis-resistant GSCs⁵² are selectively eliminated through parthanatos, bypassing canonical resistance mechanisms. Indeed, our cerebral organoid experiments revealed evidence of both apoptotic and parthanatos pathways following SCD inhibition, with different cells within the heterogeneous tumor population exhibiting distinct death modalities consistent with varying PARP1 expression levels. The engagement of parallel cell death pathways may reduce the likelihood of resistance development and ensure more complete tumor elimination. Beyond direct cytotoxicity, insufficient C18:1 impairs membrane phospholipid synthesis required for cell division, creating convergent metabolic and biosynthetic blocks to proliferation that explain the profound cell-cycle arrest we observed.

Standard GBM therapies such as radiation and TMZ are limited by the robust DNA damage response, particularly through PARP-dependent repair pathways.⁵³ Fatty acid oxidation has been shown to fuel PARP activity and support HDR.⁵⁴ Additionally, radiotherapy increases levels of UFAs in GBM, potentially promoting resistance.⁵⁵ Our previous work demonstrated that SCD inhibition reduces RAD51 expression and impairs HDR, sensitizing GBM cells to TMZ and radiation,^20,26,52 and PARPi-resistant triple-negative breast cancer cells to PARPi.⁵³ We now extend these findings by showing that SFA-induced PARP1 depletion downregulates RAD51, providing further support for a mechanistic link between lipid desaturation and DNA repair pathways. These effects parallel the findings where PARP1 disruption suppresses RAD51 expression³⁵ and reinforce the bidirectional regulation between lipid metabolism and DNA repair.

We propose a model in which SFA accumulation induces DNA damage and PARP1 hyperactivation, resulting in depletion of NAD⁺ and ATP⁵⁶ and initiating the degradation of PARP1 via lysosomal⁴⁸ and proteasomal pathways.⁴⁹ This energetic crisis renders GBM cells vulnerable to further stress and DNA damage. Beyond direct tumor cell death, SCD inhibition may also enhance anti-tumor immunity. Damage-associated molecular patterns (DAMPs) released during parthanatos can stimulate immune responses,⁵⁷ as we recently demonstrated in an immunocompetent brain metastasis model where SCD inhibition enhanced dendritic cell activation, interferon signaling, and T cell-mediated tumor control.⁵⁸ These results suggest that targeting SCD could yield a dual therapeutic benefit: direct tumor cell killing and immune microenvironment reprogramming to favor anti-tumor immunity.

Limitations of the study

Our lipidomics analyses indicate that SCD1 and SCD5 have distinct substrate specificities, suggesting they play non-overlapping roles in shaping the GBM lipidome. However, several important questions remain. First, while we demonstrate that SCD5 expression reflects neural lineage inheritance, the precise transcriptional mechanisms governing SCD5 in normal neural development and GBM remain unclear. Second, our study focuses on GBM, but the broader role of SCD5 in other CNS pathologies, including different glioma grades, brain metastases, and neuroinflammatory conditions, warrants investigation. Third, direct comparative lipidomics profiling across GSCs, differentiated tumor cells, and normal neural cells would provide deeper insights into the cell state-specific metabolic dependencies we observed functionally.

From a therapeutic perspective, a critical unanswered question is whether compensatory upregulation of either SCD isoform could mediate acquired resistance to SCD-targeted therapies. Our observation of reciprocal regulation between SCD1 and SCD5 suggests this possibility, though its functional significance in vivo remains to be determined. Additionally, while our CO model provides valuable insights, fully immunocompetent animal models will be essential for evaluating how SCD inhibition interacts with anti-tumor immunity and for assessing potential toxicities in normal neural cells, particularly oligodendrocytes with high endogenous SCD5 expression. Finally, developing isoform-specific inhibitors will require detailed structural characterization to exploit differences between SCD1 and SCD5 despite their shared catalytic mechanism.

In conclusion, our study reveals a mechanistic connection between fatty acid desaturation and DNA repair in GBM, identifying SCD5 as a critical metabolic vulnerability in GSCs. SCD inhibition triggers SFA accumulation, PARP1 hyperactivation, and parthanatos, thereby disrupting tumor maintenance. Importantly, these effects may be further amplified in combination with DNA-damaging therapies that exploit compromised repair capacity or immunotherapies that leverage parthanatos-induced immune activation. By revealing how a brain-enriched lipid desaturase controls stem cell fate through PARP-dependent genome maintenance, our findings establish SCD5-targeted therapy as a strategy to selectively eliminate the therapy-resistant stem cell population driving GBM recurrence.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Christian E. Badr (badr.christian@mgh.harvard.edu).

Materials availability

• All unique/stable reagents generated in this study will be made available from the lead contact with a completed materials transfer agreement, in accordance with any additional institute or funding agency policies.

• Plasmids generated in this study will be deposited in Addgene.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

Male and female BCr-nu/nu nude mice (6–8 weeks old) were used for all in vivo experiments. Mice were maintained in a pathogen-free environment at the Massachusetts General Hospital animal care facility under a 12h light/dark cycle with ad libitum access to standard chow and water. All animal procedures were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care (Protocol 2016N000267) and conducted in accordance with NIH guidelines for the care and use of laboratory animals.

Primary patient-derived glioma stem cells and control cell models

Primary glioblastoma stem-like cells (GSCs) were obtained from consented patients with GBM at Massachusetts General Hospital (MGG4, MGG6, MGG8 and MGG23; provided by H. Wakimoto) under the appropriate IRB-approved protocols, or sourced from I. Nakano (326, 157 and 83) and the Brain Tumor PDX National Resource at Mayo Clinic (M76, M12, M39 and M120; provided by J. Sarkaria).

GSCs were maintained as neurospheres in DMEM/F12 (Thermo Fisher) supplemented with B27 without vitamin A (1:50), heparin (2 μg/mL), recombinant human EGF (20 ng/mL), recombinant FGF2 (10 ng/mL), GlutaMAX (1:100), and penicillin/streptomycin (1:100). Normal human astrocytes (NHAs) were cultured as a monolayer in ScienCell Astrocyte Medium, while human induced pluripotent stem cells (iPSCs) were maintained in B8 medium on Geltrex-coated dishes according to established protocols.^59,60 All cell lines were regularly tested for mycoplasma using a DNA-based assay.

METHOD DETAILS

Glioma stem cells differentiation

To induce differentiation, GSCs were dissociated with Accutase and seeded onto tissue culture-treated plates in DMEM/F12 supplemented with B27, GlutaMAX, penicillin/streptomycin, and BMP4 (20 ng/mL) for 6 days, with a media refresh at day 3. Alternatively, differentiation was achieved by culturing cells in 2% FBS for 6 days (media refreshed every 3 days), followed by a 3-day serum-free incubation period to minimize experimental variability associated with FBS components.

Organoids creation

Cerebral organoids were generated from iPSCs as previously described⁶¹ and cultured on an orbital shaker for up to 90 days to facilitate nutrient diffusion. This duration ensured a predominantly postmitotic cell population, as confirmed by Ki-67 staining, prior to Glioma Stem Cell (GSC) implantation. To prevent tissue attachment, organoids were maintained in low-adherence dishes coated with Anti-Adherence Rinsing Solution. For tracking and invasion analysis, GSCs were transduced with a lentivirus expressing cyan fluorescent protein (CFP). At day 90, organoids were transferred to V-bottom 96-well plates in a 1:1 mixture of DMEM/F12 and Neurobasal media supplemented with 2% B27, 1% N2, 1% non-essential amino acids, 0.2% P/S, and 1% Geltrex to assist with engraftment. Approximately 10,000 CFP-labeled GSCs were added to each organoid and maintained in V-bottom wells for 24 h. Subsequently, organoids were transferred to ultra-low attachment dishes for an additional 14 days before starting SCD1i treatment. For histological analysis, organoids were fixed in paraformaldehyde for 1 h, equilibrated in 30% sucrose until they sank, and embedded in OCT compound for cryosectioning at a thickness of 15 μm.

Lentiviral production

HEK293T cells (5 × 10⁶) were seeded in 150 mm dishes. After 24 h, 30 μL chloroquine (25 mM; Sigma-Aldrich) was added, and cells were transfected with 15 μg of plasmid encoding the gene or shRNA of interest, 3.75 μg PMD2.G (Addgene, Cat. No. 12259), and 11.25 μg psPAX2 (Addgene Cat. No. 12260) using PEI (Polysciences; 1:3 DNA:PEI ratio).

At 72 h post-transfection, the medium was centrifuged (500 × g, 10 min) and filtered (0.22 μm PES filter: Corning, Cat. No. 431222). The filtrate was ultracentrifuged at 70,000 × g for 90 min at 4°C. Viral pellets were resuspended in 200 μL of 1% BSA in PBS, aliquoted, and stored at −80°C.

Lentiviral transduction

GSCs were dissociated with Accutase and seeded at 5 × 10⁵ cells/mL in 12-well plates. Polybrene (1 μg/mL; Millipore, Cat. No. TR-1003-50UL) was added to enhance transduction efficiency. Virus was added and incubated for 8–14 h at 37°C. Media were then replaced, and cells were expanded for subsequent experiments.

Radiation therapy

For all γ-irradiation experiments, a ¹³⁷Cs Mark 1 - Model 25 Irradiator (J.L. Shepherd & Associates) equipped with a rotating turntable was used to ensure dose uniformity. GSCs were irradiated with a single total dose of 6 Gy. Following irradiation, cells were maintained in culture for an additional 24 h prior to collection and protein extraction for downstream analysis.

Cell viability assay

Cells were plated in 96-well plates in 100 μL total volume. Viability was assessed using the CellTiter-Glo according to the manufacturer’s instructions. The reagent was diluted 1:3 in PBS, and 25 μL was added per well. After a 10-min incubation, 95 μL from each well was transferred to an opaque 96-well plate, and luminescence was measured using a Flexstation III plate reader (Molecular Devices).

Mouse orthotopic brain tumor models

GSCs (10,000 cells) expressing Firefly luciferase (Fluc) were implanted into the left forebrain of athymic nude mice (1.0 mm anterior, 2.0 mm lateral to bregma, 2.5 mm depth) using a stereotactic frame. Tumor growth was monitored by bioluminescence imaging (Xenogen IVIS 200, PerkinElmer) after intraperitoneal injection of D-luciferin (150 mg/kg). Signal intensity was quantified using Living Image 4.3.1 software.

Immunoblotting

Cells were lysed in 1× RIPA buffer (Millipore, Cat. No. 20–188) for 20 min on ice, followed by one freeze–thaw cycle. Protein concentration was determined using the BCA assay (Thermo Fisher), and equal amounts of protein were loaded onto NuPAGE 4–12% Bis-Tris gels (Thermofisher). After transfer to PVDF membranes, blots were blocked with 5% skimmed milk in TBST (TBS +0.5% Tween 20; Sigma-Aldrich, P1379-500ML) and incubated overnight at 4°C with primary antibodies diluted in 2.5% milk. Membranes were washed and incubated with secondary antibodies for 1 h at room temperature. Detection was performed using SuperSignal West Pico PLUS or West Femto chemiluminescent substrates (Thermofisher). The complete list of antibodies used for immunoblotting is provided in the Key Resources table. A custom rabbit polyclonal antibody against SCD5 was generated using the following peptide antigen: MPGPATDAGKIPFC (SynoBiologics).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
SCD5	ThermoFisher	Cat# PA5-89006; RRID: AB_2805291
NG2	ThermoFisher	Cat# 14-6504-82; RRID: AB_10870987
SCD5 (custom)	SynoBiologics	N/A
SCD1	ThermoFisher	Cat# MA5-27542; RRID: AB_2723611
PARP1	ThermoFisher	Cat# 14-6667-82; RRID: AB_10698017
PAR	Cell Signaling Technologies	Cat# 89190S; RRID: AB_3716623
γH2A.X	Cell Signaling Technologies	Cat# 9718S; RRID: AB_2118009
AIF	Cell Signaling Technologies	Cat# 5318S; RRID: AB_10634755
LC3 A/B	Cell Signaling Technologies	Cat# 12741S; RRID: AB_2617131
GFP	ThermoFisher	Cat# MA5-15256; RRID: AB_10979281
TBP	Proteintech	Cat# 66166-1-Ig; RRID: AB_2881562
β-Tubulin	Cell Signaling Technologies	Cat# 86298S; RRID: AB_2715541
β-Actin	Cell Signaling Technologies	Cat# 3700S; RRID: AB_2242334
GAPDH	Proteintech	Cat# 60004-1-Ig; RRID: AB_2107436
Anti-rabbit IgG, HRP-linked	Cell Signaling Technologies	Cat# 7074V; RRID: AB_2099233
Anti-mouse IgG, HRP-linked	Cell Signaling Technologies	Cat# 7076V; RRID: AB_330924
PAR	Cell Signaling Technologies	Cat# 89190S; RRID: AB_3716623
KI67	Cell Signaling Technologies	Cat# 9449S; RRID: AB_2797703
NFH	Antibodies Inc	Cat# NFH; RRID: AB_2313552
OLIG2	Proteintech	Cat# 82806-7-RR; RRID: AB_3670559
Goat anti-rabbit AF 647	ThermoFisher	Cat# A-21245; RRID: AB_2535813
Goat anti-mouse AF 488	ThermoFisher	Cat# A-11001; RRID: AB_2534069
Goat anti-rabbit Biotin-XX	ThermoFisher	Cat# B-2770; RRID: AB_2536431
Goat anti-mouse Biotin-XX	ThermoFisher	Cat# B-2763; RRID: AB_2536430
Bacterial and virus strains
ElectroMAX Stbl4 Competent Cells	ThermoFisher	Cat# 11635018
Chemicals, peptides, and recombinant proteins
GSK2194069 (FASN inhibitor)	MedChemExpress	Cat# HY-12325
CAY10566 (SCD1 inhibitor)	Cayman Chemical	Cat# 10012562
PDD00017273 (PARG inhibitor)	MedChemExpress	Cat# HY-108360
Cycloheximide	Sigma-Aldrich	Cat# 01810-1G
Hydrogen peroxide	Fisher Scientific	Cat# H325-100
DMSO ThermoFisher	ThermoFisher	Cat# 20688
Triton X-100	Sigma-Aldrich	Cat# X100-500ML
D-Glucose-13C6	Sigma-Aldrich	Cat# 389374
Halt Protease and Phosphatase Inhibitor Cocktail	ThermoFisher	Cat# 78440
Hoechst 34580	ThermoFisher	Cat# H21486
Trypsin 0.25%	ThermoFisher	Cat# 25200056
D-luciferin	Gold Biotechnology	Cat# LUCK-100
DMSO	ThermoFisher	Cat# 20688
RIPA buffer	Millipore	Cat# 20-188
Heparin	Sigma-Aldrich	Cat# 375095-100KU
Thymidine	Sigma-Aldrich	Cat# T1895-1G
RNase A	New England Biolabs	Cat# T3018L
Epidermal Growth Factor (EGF)	Biolegend	Cat# 713008
Fibroblast Growth Factor 2 (FGF2)	Biolegend	Cat#710308
Bone Morphogenetic Protein 4 (BMP4)	Biolegend	Cat# 595201
Critical commercial assays
Quick-RNA Microprep	Zymo Research	Cat# R1051
BCA kit	ThermoFisher	Cat# 23227
NEBNext Ultra II RNA Library Prep Kit for Illumina	New England Biolabs	Cat# E7770S
NEBNext Poly(A) mRNA Magnetic Isolation Module	New England Biolabs	Cat# E7490S
CellTiter-Glo 2.0 Assay	Promega	Cat# G9242
NE-PER Nuclear and Cytoplasmic Extraction Reagents	ThermoFisher	Cat# 78833
Luna Universal One-Step RT-qPCR Kit	New England Biolabs	Cat# E3005S
Luna Universal qPCR Master Mix	New England Biolabs	Cat# M3003E
Deposited data
Raw RNAseq data	GEO	GSE316537
Experimental models: Cell lines
HEK 293T	Massachusetts Institute of Technology	N/A
Mayo12 (M12)	Mayo Clinic	N/A
Mayo76 (M76)	Mayo Clinic	N/A
Mayo120 (M120)	Mayo Clinic	N/A
Mayo39 (M39)	Mayo Clinic	N/A
MGG4	Massachusetts General Hospital	N/A
MGG6	Massachusetts General Hospital	N/A
MGG8	Massachusetts General Hospital	N/A
MGG23	Massachusetts General Hospital	N/A
83 cells	The Ohio State University	N/A
157 cells	The Ohio State University	N/A
326 cells	The Ohio State University	N/A
iPSCs (33114B)	Massachusetts General Hospital	N/A
iPSCs (33113)	Massachusetts General Hospital	N/A
Primary Human Astrocytes (NHA)	ScienCell Research Laboratories	Cat# 1800
U-251	Sigma-Aldrich	Cat# 09063001
Experimental models: Organisms/strains
Nude mice	Charles River NCI	BCr-nu/nu (code: 553)
Software and algorithms
ImageJ FIJI	ImageJ	RRID:SCR_002285; V2.0.0
R	R Foundation	RRID:SCR_001905; V4.30
BioRender	Web tool	RRID:SCR_018361
Other
Accutase	ThermoFisher	Cat# A1110501
Penicillin/Streptomycin	ThermoFisher	Cat# 15140122
GlutaMAX	ThermoFisher	Cat# 35050061
B27 without vitamin A	ThermoFisher	Cat# 12587010
Astrocyte Medium	ScienCell Research	Cat# 1801
DMEM/F12	ThermoFisher	Cat# 07010
Anti-Adherence Rinsing Solution	Stemcell Technologies	Cat# 07010

Immunocytochemistry

Cells were plated on glass coverslips coated with poly-D-lysine (1 h) and laminin (20 μg/mL, overnight; Sigma-Aldrich). After 4 h, cells were fixed in 4% paraformaldehyde for 20 min and blocked with 5% normal goat serum, 0.5% BSA, and 0.1% Triton X-100 in TBS.

Primary antibodies were incubated overnight at 4°C. After washing, secondary antibodies were applied, and Hoechst (1:1000; Thermofisher) was added during the secondary incubation. Slides were mounted using ProLong Diamond (Thermofisher) and imaged on a Keyence BZ-X810 or Nikon W1-SoRa confocal microscope. The complete list of antibodies used for immunofluorescence is provided in the key resources table.

Endoplasmic reticulum and mitochondria imaging

GSCs expressing miRFP670-nano-SCD1 or miRFP670-nano-SCD5 were seeded at 20,000 cells/cm² on laminin-coated (10 μg/mL) glass-bottom 8-well chambers (Ibidi). After 24 h, cells were stained with:ER-Tracker Green (1 μM; Thermofisher) for 30 min at 37°C. MitoView Green (200 nM; Biotium) for 15 min at 37°C. Cells were then imaged using a NIKON W1-SoRa confocal microscope.

Laser micro-irradiation induced DNA damage

GSCs were transduced with lentivirus expressing PARP1-GFP, followed by control shRNA or shRNA targeting SCD1 or SCD5. Cells were cultured for 4 days post-transduction prior to DNA damage assessment. To induce localized DNA damage, cells were imaged using a Nikon W1 SoRa spinning disk confocal microscope mounted on a Nikon Ti2-E inverted base. Targeted photostimulation was performed using the Nikon Ti2-LAPP galvano-scanner module equipped with a 405 nm solid-state laser and delivered through a 60x/1.49 oil-immersion objective.

Defined regions of interest (ROIs) within the nucleus were subjected to a single 600-μs pulse of laser illumination at 70% laser power. To capture the rapid recruitment kinetics of PARP1, time-lapse imaging was initiated immediately prior to the stimulation event, with images acquired every 5 s for a total of 300 s. PARP1-GFP recruitment to damage sites was quantified using ImageJ/Fiji. The mean fluorescence intensity at the stimulated ROI was measured for each time point and normalized to the pre-irradiation baseline.

RNA expression analysis

Total RNA was extracted using the Quick-RNA Miniprep Kit (Zymo) and quantified with a NanoDrop spectrophotometer (Thermo Fisher). cDNA was synthesized from 600 ng of RNA using the LunaScript RT SuperMix Kit (NEB). Quantitative PCR (qPCR) was performed using the Luna Universal qPCR Master Mix (NEB) on a QuantStudio 3 system (Applied Biosystems). Relative gene expression was calculated using the comparative Ct method (ΔΔCt), with normalization to TBP as a housekeeping gene. The list of primers used in this study is provided in Table S3.

Bulk RNA sequencing

Total RNA (500 ng) was extracted with the Quick-RNA Miniprep Kit and assessed for quality on a Bioanalyzer (Agilent). Libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit (NEB), with poly(A) selection, fragmentation, cDNA synthesis, end repair, adapter ligation, and PCR amplification. Libraries were sequenced on an Illumina NextSeq 2000 (paired-end).

Reads were aligned to GRCh38 using the Rsubread package in R. Gene counts were generated using featureCounts, and differential expression was analyzed with DESeq2. Significance was defined as adjusted p < 0.05.

Carbon tracing analysis

GSCs were transduced with control (SCR) or targeted shRNAs for SCD5 and SCD1 knockdown. Following a 48-h incubation, the medium was replaced with glucose-free medium supplemented with a 1:1 ratio of unlabeled D-glucose and ¹³C₆-glucose (2.25 g/L each). At 24 and 48 h post-medium change, 200,000 cells were harvested and resuspended in 3 M methanolic guanidine HCl before immediate transfer to glass tubes for derivatization. FAME internal standards (Nu-Chek Prep) were used for sample quantification. Isotopomer spectral analysis (ISA) was performed via GC-MS (Agilent 7890/5975C) as previously described,⁶² with data normalized to total cell counts.

Shotgun lipidomics

Cells were transferred into extraction tubes containing phosphate-buffered saline (PBS), and lipid extraction was performed using a modified Bligh and Dyer protocol as previously described.⁶³ Samples were spiked with 74 internal standards (Avanti) prior to biphasic extraction. Organic phases were dried via SpeedVac (35°C, 90 min) and resuspended in 1:1 methanol:dichloromethane with 10 mM ammonium acetate. Targeted analysis of 1,450 lipid species was performed by direct infusion on a Sciex 5500 mass spectrometer equipped with a Differential Mobility Device (DMS), with compensation voltages tuned using EquiSPLASH LIPIDOMIX.

As described in Su B. (2021),⁶⁴ raw data were processed using the Shotgun Lipidomics Assistant (SLA). To ensure data quality and handle background noise, Multiple Reaction Monitoring (MRM) transitions were filtered to exclude species with >10% zero values or low signal intensity (raw intensity <100). Quantitative accuracy was maintained by applying a computationally derived isotope correction to mitigate isobaric overlap within lipid classes. Species concentrations (nM) were calculated using response factors derived from class-specific internal standards and subsequently normalized to cell number to account for biological starting material.

Cell cycle synchronization

GSCs were synchronized at the G1/S phase using 2 mM thymidine (Sigma-Aldrich, T1895-1G). Cells were treated for 16 h, released into thymidine-free medium for 8 h, then treated again for 16 h. After the second treatment, cells were washed and placed in fresh medium. This time point was designated as time 0.

Flow cytometry for cell cycle analysis

GSCs were transduced with SCD1/5-targeting shRNA and cultured for 4 days. Cells were fixed in 70% ethanol after PBS washes. For cell cycle profiling, cells were stained with propidium iodide (0.02 mg/mL), RNase A (0.5 mg/mL; NEB), and Triton X-100 (0.1%) in PBS, incubated at 37°C for 30 min, and analyzed using a Cytek Aurora flow cytometer.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses were performed using R v4.3.0. Each experiment included at least four biological replicates, and the specific details for each dataset are indicated in the figure legend. Statistical comparisons were made using two-tailed Student’s t-tests. Data normality was assessed using the Shapiro-Wilk test. Non-parametric tests (Wilcoxon signed-rank test) were used when data were non-normal or heteroscedastic. p < 0.05 was considered statistically significant.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2026.117223.

Highlights.

• SCD5 is essential for glioblastoma stem cell maintenance and tumor initiation

• SCD5 preferentially desaturates C18:0 and uniquely regulates sphingolipids

• Loss of SCD1/5 activity causes PARP1 hyperactivation, autodegradation, and parthanatos

• Targeting SCD5 exposes a metabolic vulnerability in glioblastoma stem cells

Data and code availability

• RNA-seq data have been deposited in GEO: GSE316537 and are publicly available as of the date of publication. The raw shotgun lipidomics data are available in spreadsheet 1.

• Microscopy and immunoblot data reported in this paper will be shared by the lead contact upon request

• This paper does not report any original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.