Sortilin emerges as a two-faced switch driving glioblastoma stemness and drug resistance in 3D spheroids

Glioblastoma remains one of the most lethal human cancers, and its resistance to temozolomide and other therapies has long been attributed to a stubborn subpopulation of tumour cells that behave remarkably like stem cells. Now, a team of Canadian researchers has identified sortilin (SORT1), a Vps10p-domain sorting receptor best known for its role in lipoprotein handling and receptor trafficking, as a molecular node that appears to weld together three previously separate tumour programs: adipogenic plasticity, stemness, and drug resistance. The findings, published in BMC Cancer, suggest that SORT1 operates through two distinct arms, one that can be pharmacologically blocked and one that requires the receptor’s physical presence.

The study, led by Abdallah Fallah, Alain Zgheib, Bogdan Alexandru Danalache, Nicoletta Eliopoulos and Borhane Annabi at the Université du Québec à Montréal and McGill University, used focused PCR arrays to compare U87 glioblastoma cells grown as conventional two-dimensional monolayers or as three-dimensional spheroids. The researchers then perturbed SORT1 in two ways: genetically, using SORT1-targeted siRNA (siSORT1), and pharmacologically, using AF38469, a Vps10p-domain antagonist that blocks the receptor’s ligand-binding arm. A non-targeting scrambled siRNA served as the genetic control.

The results reveal that when U87 cells are grown as 3D spheroids, they activate a coordinated transcriptional program that is largely absent in 2D monolayers. This program includes upregulation of angiogenic markers such as VEGFA and HIF1A, adipogenic and lipogenic regulators including PPARG, DDIT3, ADIPOQ, CEBPD and FOXO1, and a battery of glioblastoma stemness and drug-resistance markers including PROM1/CD133, SOX2, NANOG, CD44, ABCG2, ABCB5, TAZ and KITLG. Together, these changes indicate that 3D growth triggers a lipid-remodeling program with cancer stem cell features that are directly relevant to chemoresistance.

When the researchers silenced SORT1 or blocked it with AF38469, both interventions reduced spheroid size and repressed the adipogenic module, supporting a role for SORT1 in regulating adipogenesis and lipid-synthesis programs in glioblastoma spheroids. However, the two perturbations did not behave identically across the full transcriptional landscape. The cancer stem cell module proved far more sensitive to genetic SORT1 silencing, with 18 of 27 genes responding, than to pharmacological AF38469 treatment, which affected only 10 of 27 genes.

Among the transcripts that emerged as siSORT1-dependent but AF38469-insensitive were CXCL8, CD38, ZEB2, SOX2, NANOG and CD44. This bifurcation is significant because it suggests that SORT1 functions segregate into two arms: a ligand-dependent arm that AF38469 can target, and a ligand-independent, scaffold-like arm that requires SORT1 expression itself. The pharmacological inhibitor reproduces the genetic silencing effect nearly perfectly on the lipid and metabolic module but fails to do so on the developmental backbone of the cancer stem cell program.

Principal component analysis confirmed this dual-module organization. When siSORT1 and AF38469 were applied together, they jointly displaced the adipogenesis profile, whereas only siSORT1 clearly separated the cancer stem cell profile. This systems-level observation supports the idea that the bifurcation operates on entire transcriptional modules rather than on individual genes. The lipid and metabolism machinery plus drug-resistance elements form a co-regulated, ligand-dependent arm, while stemness, plasticity and the developmental backbone form a scaffold-dependent persistence arm.

The SORT1-dependent transcriptional state is organised into six functionally coherent modules, providing what the researchers describe as a modular blueprint for combinatorial therapeutic targeting. An AF38469 phenocopy index, a quantitative measure of how well the pharmacological inhibitor mimics genetic SORT1 silencing, provides strong support for this bifurcation model. The index shows that AF38469 reproduces the siSORT1 effect on the lipid and metabolic module but fails on the developmental backbone.

These findings are consistent with a model in which SORT1 behaves as a state-persistence node, associating adipogenic and lipogenic reprogramming with cancer stem cell-associated and drug-resistance-associated transcriptional programs. The researchers caution that this model remains to be tested with direct functional assays in 3D U87 spheroids, and the current study is primarily transcriptional in scope. Nevertheless, the modular organization offers a framework for designing combination strategies that could circumvent the acquired SORT1-regulated cancer stem cell and drug-resistance phenotype.

The broader implications are considerable. By creating hydrophobic compartments that sequester lipophilic chemotherapeutics, increased adipogenic reprogramming and lipid synthesis correlate with poor treatment response and increased drug tolerance in glioblastoma. SORT1 has been implicated in cancer stem cell biology and chemoresistance, and this study now provides a systems-level dissection of how the receptor links lipid remodeling to a chemoresistance-associated molecular signature. If the two-arm model holds up under direct functional testing, it could open new avenues for therapeutic intervention, combining AF38469-like ligand-dependent antagonists with strategies that target the scaffold-dependent arm of SORT1 biology.

The choice of a three-dimensional spheroid model deserves particular attention when weighing these results. Standard two-dimensional monolayer culture has long been criticized for stripping glioblastoma cells of the hypoxic, nutrient-gradient and cell–cell contact conditions that define the tumour microenvironment in patients. Hanging-drop spheroid systems restore some of these gradients, and it is precisely under these conditions that the adipogenic and stemness programs described here emerge. The fact that markers such as CD133 and SOX2 rise in spheroids relative to monolayers echoes a broader literature showing that 3D architecture alone, without any genetic manipulation, can push glioma cells toward a stem-like, therapy-tolerant state. This makes the spheroid platform a useful intermediate between simplified monolayers and animal models for probing drug-resistance biology.

The lipid dimension of the study connects to a growing appreciation of metabolic reprogramming in aggressive cancers. Lipid droplets act as hydrophobic sinks that can partition weakly lipophilic drugs away from their intracellular targets, effectively lowering the free drug concentration reaching the nucleus or microtubules. Enzymes of fatty acid synthesis and adipogenic transcription factors such as PPARG and CEBPD have each been independently linked to poor prognosis in high-grade gliomas. What the present work adds is a plausible coordinator: a single membrane receptor whose perturbation collapses much of this lipid program, suggesting that lipid remodeling in glioblastoma spheroids is not a diffuse set of parallel changes but a regulated module with an identifiable upstream node.

Sortilin itself has an unusual profile among trafficking receptors. It shuttles between the trans-Golgi network and endosomes, sorting cargo proteins involved in metabolism, inflammation and growth factor signaling. Its involvement in lipoprotein handling made it an attractive candidate when the authors sought a molecular bridge between lipid biology and cancer stem cell maintenance. The observation that a ligand-binding antagonist and a genetic knockdown diverge in their transcriptional consequences is consistent with a receptor that does more than transmit signals through ligand occupancy; it may also act as a physical scaffold, organizing protein complexes or cargo trafficking routes that persist as long as the protein is expressed, regardless of whether its ligand-binding pocket is occupied.

The specific transcripts that resist pharmacological blockade are informative. CXCL8, the gene encoding interleukin-8, is a chemokine associated with inflammatory recruitment, angiogenesis and stem cell maintenance in multiple tumour types. ZEB2 is a transcriptional repressor best known in epithelial–mesenchymal transition, but it also participates in developmental gene-regulatory networks in neural lineages. CD38, SOX2, NANOG and CD44 round out a set of markers tied to self-renewal and immune interaction. That these genes respond to loss of SORT1 expression but not to ligand-pocket antagonism implies that the scaffold-dependent arm of the receptor touches developmental and inflammatory programs that the ligand-dependent arm does not.

The drug-resistance markers in the spheroid signature also merit comment. ABCG2 and ABCB5 encode ATP-binding cassette transporters that actively pump structurally diverse chemotherapeutics out of cells, and both have been detected on glioblastoma stem-like populations. TAZ, a co-activator in the Hippo pathway, links mechanical cues from the 3D environment to transcriptional programs supporting proliferation and survival. The co-regulation of transporter genes with adipogenic regulators in a single module raises the possibility that lipid accumulation and drug efflux are not merely coincident but co-selected features of the same cellular state, which would explain why lipid-rich tumour cells so often prove chemoresistant.

Several caveats frame the interpretation. The work relies on a single cell line, U87-MG, which, although widely used, does not capture the molecular heterogeneity of patient glioblastomas, and the analysis is transcriptional rather than functional: no direct assays of spheroid viability under temozolomide challenge, stem cell frequency or lipid droplet burden are reported. Supplementary data on fibronectin distribution between cytosolic and nuclear fractions following SORT1 perturbation hint at downstream mechanistic threads, but these remain to be pulled. Validation across patient-derived glioblastoma stem cell cultures, organoids and in vivo models would be the natural next step.

If the two-arm model is confirmed, the therapeutic logic is straightforward but demanding. Blocking the ligand-dependent arm with AF38469-like molecules could deflate the lipid and metabolic machinery that shelters tumour cells from chemotherapy, while the scaffold-dependent arm would require entirely different approaches, such as degrading the receptor protein itself or disrupting its intracellular trafficking partners. The modular blueprint offered by this study, with its quantitative phenocopy index, gives medicinal chemists and translational researchers a measurable target profile for such combination strategies, and the open-access availability of the underlying datasets should facilitate independent evaluation of this provocative model.

Subject of Research: Sortilin-associated transcriptional programs in 3D glioblastoma spheroids

Article Title: A SORT1-associated transcriptional signature links adipogenic plasticity, stemness and drug-resistance programs in 3D glioblastoma spheroids: dissection of ligand-dependent and ligand-independent arms

Article References: Fallah, A., Zgheib, A., Danalache, B. A., Eliopoulos, N., & Annabi, B. (2026). A SORT1-associated transcriptional signature links adipogenic plasticity, stemness and drug-resistance programs in 3D glioblastoma spheroids: dissection of ligand-dependent and ligand-independent arms. BMC Cancer. https://doi.org/10.1186/s12885-026-16856-9

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16856-9

Keywords: glioblastoma, SORT1, sortilin, cancer stem cells, 3D spheroids, chemoresistance, adipogenesis, AF38469, drug resistance, transcriptional modules, SORT1-associated, transcriptional