Stress Granule Protein DZIP3 Tracks a Cytoskeletal Program in Glioblastoma

Glioblastoma remains one of the most formidable challenges in modern oncology. Representing roughly 45 to 50 percent of all gliomas, this aggressive brain tumor carries a median overall survival of just 12 to 15 months, extending only to about 18 months even with the current standard-of-care STUPP protocol of surgical resection followed by concomitant radiotherapy and chemotherapy. Now, a team of researchers from Mohammed VI University of Sciences and Health and collaborating institutions in Morocco has turned its attention to an unlikely molecular player in this disease: DZIP3, or DAZ-interacting zinc finger 3, an RNA-binding protein that resides within stress granules, the membraneless cytoplasmic structures that cells assemble when under duress. Their new bioinformatic investigation, published in Frontiers in Bioinformatics, reveals that DZIP3 expression in glioblastoma is most strongly and independently associated with a cytoskeletal dynamics and intracellular transport gene program, along with four mitochondrial modules, an RNA granule and neurodegeneration signature, and a nucleocytoplasmic transport gene set.

The study began with a systematic literature review of stress granule-associated proteins to identify candidates lacking mechanistic characterization in glioblastoma. Stress granules themselves are fascinating structures: they coalesce from heterogeneous elements including RNA-binding proteins, translation initiation factors, non-RNA-binding proteins, ribosomal subunits, and stalled messenger RNAs. Under acute stress, they act as a temporary protective mechanism by blocking mRNA translation, thereby regulating protein synthesis. Once the stress is relieved, they dissolve. But under chronic stress, they can adopt pathological features. In cancer, stress granules inhibit apoptosis by negatively regulating the SAPK apoptotic pathway, remodel translational processes through the PI3K/mTOR pathway to regulate mitochondrial proteins, and even control key components of the antitumor immune response by sequestering the messenger RNA of immune checkpoints such as PD-1. Despite these findings, how stress granules participate in gliomagenesis remains largely unexplored, making them an underexploited reservoir of candidate regulators.

DZIP3, also known as hRUL138, is an E3 ubiquitin ligase responsible for protein degradation via the proteasome, yet it remains understudied in cancer, particularly with respect to its impact on tumor microenvironment remodeling. Previous literature indicates that DZIP3 can act as both a repressor and an activator of transcription. It has been found to promote tumorigenesis in gastric, breast, lung, colon, liver, and pancreatic cancers, while exhibiting antitumoral properties in low-grade glioma, where its high expression is negatively correlated with angiogenesis. It has also been characterized as a transcriptional repressor by monoubiquitinating histone H2A, but it can act as a transcriptional activator by binding CARM-1 and enhancing transcription of the estrogen receptor alpha. This duality of function made DZIP3 an intriguing candidate for investigation in the context of glioblastoma, where cellular heterogeneity, including stem-like tumor cell states such as NPC-like and OPC-like phenotypes, contributes to treatment resistance and tumor recurrence.

To evaluate the prognostic value of DZIP3, the researchers downloaded data from The Cancer Genome Atlas TCGA-GBM project, retrieving 287 glioblastoma samples after matching RNA-sequencing data with clinical information and removing duplicate patients. They converted raw counts to Transcripts Per Million, log2-transformed the values, and stratified patients into high- and low-DZIP3 groups based on the median expression. Initially, high DZIP3 expression was associated with better survival in the TCGA cohort, with a Kaplan-Meier analysis showing prolonged overall survival compared with patients with low DZIP3 expression. However, when the team performed a multivariate Cox proportional hazards regression adjusting for age, gender, transcription subtypes, IDH mutation status, MGMT promoter methylation status, and Karnofsky performance score, the prognostic utility of DZIP3 did not persist. The hazard ratio for DZIP3 expression was 1.08 with a p-value of 0.66, indicating no statistically significant independent association with survival. Similar results were observed in the Chinese Glioma Genome Atlas CGGA cohort, where DZIP3 showed no prognostic correlation after adjustment for clinical and molecular covariates.

Having established that DZIP3 is not an independent prognostic biomarker, the researchers turned to the transcriptional correlates of its expression. They performed differential expression analysis using the Limma package with empirical Bayes moderation, identifying 4028 upregulated and 364 downregulated genes in the high-DZIP3 group. The top upregulated genes were associated with neurodevelopmental properties, RNA processing, tumor suppressors, and biomarker roles, while the top downregulated genes were linked to oxidative phosphorylation, extracellular matrix composition, inflammation, and cancer remodeling. Gene Ontology enrichment analysis across biological processes, molecular functions, and cellular components revealed terms related to nucleosome organization, spliceosomal snRNP assembly, glial and oligodendrocyte differentiation, cytoskeletal transport, and metal ion homeostasis. Reactome pathway analysis identified 190 significantly enriched pathways, while KEGG-based Gene Set Enrichment Analysis uncovered 224 significantly enriched pathways, and Hallmark GSEA identified 30 significantly enriched gene sets associated with cell cycle, mitotic programs, metabolic signaling, and developmental pathways such as NOTCH and WNT beta-catenin signaling.

But here is where the study becomes methodologically remarkable. The researchers recognized that bulk tumor transcriptomics is prone to confounding by cell proliferation, immune infiltration, and technical variation in library complexity. To separate DZIP3-specific transcriptional associations from potential artifacts, they conducted a robustness assessment of 30 candidate gene sets. They evaluated the correlation between module scores for each candidate gene set and DZIP3 expression using Spearman correlation, followed by partial Spearman correlation controlling simultaneously for a proliferation marker score comprising MKI67, PCNA, TOP2A, MCM2, and CCNB1; total immune cell infiltration estimated by Cibersort; and the number of genes detected per sample representing library complexity. They retained gene sets as independent signals only if both statistical significance after Benjamini-Hochberg FDR correction and a minimum effect size threshold of partial rho of at least 0.30 in absolute value were met. Of the 30 gene sets evaluated, 23 were poorly or not significantly associated with DZIP3 expression after comprehensive confounder control. These included P-body elements, chromatin-modifying enzymes, glial-OPC-like and NPC-like malignant cell states, NOTCH and WNT signaling pathways, the mitotic spindle gene set, and the panel of 49 inhibitory immune checkpoints.

Only seven gene sets retained independent association with DZIP3 expression. The strongest correlate was the cytoskeletal dynamics and intracellular transport gene set, with a raw correlation of 0.83 and a partial correlation of 0.51 after controlling for all confounders. This 17-gene set encompasses kinesin family motor proteins including KIF1A, KIF1B, KIF2A, KIF3A, KIF5A, KIF5B, KIF5C, KIF21A, and KIF21B; the cytoplasmic dynein heavy chain DYNC1H1; microtubule-associated proteins MAP1A, MAP1B, MAP2, and MAPT; stathmins STMN2 and STMN4; and the dynein regulator PAFAH1B1. Together, these genes represent the core molecular machinery for microtubule-based intracellular transport and cytoskeletal remodeling. Glioblastoma is often described as a tumor with a diffuse infiltrative growth pattern, and this transcriptional program is safely associated with invasiveness and cellular motility. Previous studies have highlighted the role of the kinesin family in glioma progression, with KIF1AB, KIF1B, KIF2A, KIF3A, and KIF21B associated with worse prognosis in glioma. KIF3C enhances invasion, proliferation, and migration of glioblastoma cells by stimulating epithelial-mesenchymal transition and activating the PI3K/Akt signaling pathway, while KIF4A promotes glioblastoma progression by modulating the HIF1A/VEGFA signaling pathway and resistance to temozolomide.

The cytoskeletal signal was replicated in the CGGA cohort with a correlation of 0.81, corroborating the results from the TCGA cohort. It was also validated at single-cell resolution using the Core GBmap harmonized single-cell reference atlas for glioblastoma, which contains 338564 GBM tumor cells. The researchers filtered the cells to obtain only malignant cells, 127521 in total, and determined the Spearman correlation between DZIP3 expression and the cytoskeletal dynamics gene set. The correlation was 0.19 at single-cell resolution and remained significant even after controlling for per-cell sequencing depth, with a partial correlation of 0.11. Although the magnitude of the correlation was attenuated relative to bulk data, the result was directionally coherent and statistically significant, corroborating that the association between DZIP3 expression and the cytoskeletal gene set is not an artifact of bulk tissue composition but is evident at the cellular level. The other six correlates, including four mitochondrial modules related to branched-chain amino acid metabolism, NAD biosynthesis, translation factors, and pyruvate metabolism, as well as the RNA granule and neurodegeneration gene set and the nucleocytoplasmic transport gene set, were all replicated in the CGGA cohort with correlations ranging from 0.58 to 0.76.

The mitochondrial modules deserve particular attention. Using the MitoCarta 3.0 database, which contains 1136 mitochondrial genes, the researchers grouped genes into 51 functional modules and established per-sample scores for each module containing five or more detected genes. Ten modules were strongly associated with DZIP3 expression, with raw correlations ranging from 0.73 to 0.79. After the robustness assessment, four modules retained independent associations with partial correlations between 0.36 and 0.44. The branched-chain amino acid metabolism module exhibited the strongest independent correlation, indicating its position as the second most robust transcriptional correlate after the cytoskeletal gene set. This finding is consistent with existing literature reporting that branched-chain amino acid catabolism promotes glioblastoma aggressiveness and that BCAT1 activity is higher in IDH-wild-type gliomas. Increased mitochondrial pyruvate carrier activity and mitochondrial mitotic activity have been associated with more aggressive glioblastoma in patients treated with chemotherapy, while mitochondrial translation factors are upregulated in glioblastoma upon glucose starvation, promoting tumor cell growth. Targeting the mitochondrial translation machinery represents an interesting therapeutic approach.

Finally, the researchers constructed a miRNA-mRNA network to propose a candidate regulatory mechanism linking DZIP3 to the cytoskeletal program. Using the multiMiR package with experimentally validated interactions from miRTarBase, TarBase, and miRecords, they cross-referenced differentially expressed miRNAs against validated targets of the 17 genes in the cytoskeletal gene set. From 51 candidate miRNAs, nine targeted at least 13 of the 17 genes and were identified as core hub regulators. Seven of these hub miRNAs directly targeted DZIP3, suggesting a possible post-transcriptional regulatory loop controlling DZIP3 expression and its linked cytoskeletal transcriptional program. Six of the nine hub miRNAs are members of the miR-17-92 cluster and its paralogs, including miR-18a-5p, miR-20b-5p, miR-106b-5p, miR-17-5p, and miR-20a-5p, while the remaining three are miR-27a-3p, miR-185-5p, and miR-15a-5p. These clusters represent the most defined oncogenic miRNA clusters in brain tumors. Members of the miR-17-92 cluster have been found to disrupt cellular homeostasis in glioblastoma, increase tumor cell proliferation, and reduce apoptotic capacity. Individual members promote glioblastoma invasion by activating the WNT signaling pathway, and miR-25 promotes glioblastoma proliferation, migration, and invasion by activating the mTOR signaling pathway. The researchers stress that this is a correlative, speculative observation: whether DZIP3 is mechanistically implicated in this regulatory circuit or co-regulated in parallel cannot be answered by transcriptomic correlation data alone. Functional validation through knockdown or overexpression of DZIP3 and assessment of microtubule reorganization in glioblastoma cell lines would be required to reveal the mechanistic effect of DZIP3 beyond the transcriptomic results reported here. Nevertheless, the study characterizes DZIP3 as a transcriptional associate of harmonized intracellular transport and biosynthetic activity, demonstrating that clinical prognostic value is distinct from transcriptional reproducibility, a finding with relevance far beyond DZIP3 itself.

Subject of Research: DZIP3 is associated with the cytoskeletal dynamics in glioblastoma and mitochondrial machinery regulation

Article Title: DZIP3 is associated with the cytoskeletal dynamics in glioblastoma and mitochondrial machinery regulation

Article References: Bichri, K., El Badisy, I., Rghioui, M., Chihab-Eddine, M., Sqalli Houssaini, M., & Ghazi, B. (2026). DZIP3 is associated with the cytoskeletal dynamics in glioblastoma and mitochondrial machinery regulation. Frontiers in Bioinformatics, 6, Article 1878616. https://doi.org/10.3389/fbinf.2026.1878616

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1878616

Keywords: DZIP3, associated, cytoskeletal, dynamics, glioblastoma, mitochondrial, machinery, regulation, scientific research