Six patient-derived glioblastoma (GBM) cell lines (KBT#12137, KBT#10135, KBT#10170, PDM19, PDM22, and PDM123) were kindly provided by Dr. Takuichiro Hide (Department of Neurosurgery, Kitasato University School of Medicine, Kanagawa, Japan). All cell lines were established from primary GBM tissues using protocols approved by the Institutional Ethics Committee, and informed consent was obtained from all patients prior to tissue collection.

Primary GBM tissues obtained by surgical resection were mechanically minced and enzymatically dissociated into single-cell suspensions. The cell suspensions were filtered through a 70-µm cell strainer to remove debris and plated under serum-free conditions to promote sphere formation. Under these conditions, the cells formed free-floating neurospheres with a spherical morphology and well-defined borders, consistent with previously described patient-derived GBM neurosphere cultures.

Cells were maintained as neurospheres in serum-free Dulbecco's modified Eagle's medium/Nutrient Mixture F-12 (DMEM/F12; Gibco, Massachusetts, USA; Cat. No. 21331020) supplemented with 20 ng/ml human epidermal growth factor (EGF; PeproTech, USA; Cat. No. AF-100-15-500UG), 20 ng/ml human basic fibroblast growth factor (bFGF; Oriental Yeast Co., Ltd., Japan; Cat. No. 47079000), 20 ng/ml leukemia inhibitory factor (LIF; Oriental Yeast; Cat. No. 47076000), 0.5x N2 supplement (Thermo Fisher Scientific, USA; Cat. No. 17502048), 0.5x B27 supplement (Thermo Fisher Scientific; Cat. No. 17504044), 1x GlutaMAX (Thermo Fisher Scientific; Cat. No. 35050061), 1x penicillin-streptomycin (Thermo Fisher Scientific; Cat. No. 15140122), 5 µg/ml heparin (Sigma-Aldrich, Japan; Cat. No. H3149-10KU), and 10 ng/ml insulin (Sigma-Aldrich, Japan; Cat. No. I6634-50MG). Cultures were incubated in a humidified atmosphere with 5% CO₂ at 37˚C and passaged every 7-10 days by gentle mechanical dissociation. The 5-aminolevulinic acid (5-ALA) hydrochloride was purchased from Cosmo Oil Co., Ltd. (Tokyo, Japan; Cat. No. AL-00-1).

To assess the cellular accumulation of PpIX following 5-ALA exposure, patient-derived cells were dissociated into single cells and seeded at a density of 5x10⁴ cells/ml. The cells were then incubated with 1 mM of 5-ALA for 4 h at 37˚C under 5% CO₂. After incubation, PpIX fluorescence was analyzed by flow cytometry (FACS Aria II, BD Biosciences, USA). Fluorescence was excited with a 488-nm laser and detected using a 660/20-nm band-pass filter. The data were processed using FlowJo software (v7.6.5; TOMY Digital Biology, Japan).

For microscopic observation, PpIX fluorescence was visualized using a fluorescence microscope (Bio-Revo BZ-9000, Keyence, Japan) equipped with a QD625 filter cube (Olympus, Japan). At least three independent experiments were performed for each condition, and representative results are shown.

Cell motility was evaluated using 24-well Transwell chambers with 8-µm pore polycarbonate membranes (Corning, #3422). Single-cell suspensions (2x10⁵ cells/100 µl) in serum-free DMEM/F12 were placed into the upper chamber, and 600 µl of complete sphere medium was added to the lower chamber as a chemoattractant. After 72 h incubation at 37˚C, non-migratory cells remaining on the upper surface were collected by gentle trypsinization, while migratory cells that had traversed the membrane were recovered from the lower chamber.

To ensure reproducibility, migration assays were performed in triplicate using independent cell preparations. Both fractions were treated with 5-ALA, as described above, to assess the relationship between migratory capacity and PpIX accumulation. The combination of migration status (migratory vs. non-migratory) and fluorescence phenotype (PpIX⁺ vs. PpIX^-) yielded four distinct subpopulations, which were used for downstream transcriptomic analysis.

Total RNA was extracted from the four subpopulations described above (migratory PpIX⁺, migratory PpIX^-, non-migratory PpIX⁺, and non-migratory PpIX^-) using the RNeasy Plus Micro Kit (Qiagen, #74034) according to the manufacturer's protocol. RNA integrity was assessed using a Bioanalyzer (Agilent Technologies), and only samples with an RNA integrity number (RIN) >9.0 were used. For transcriptome profiling, genome-wide cDNA microarray analysis was outsourced to the Chemicals Evaluation and Research Institute (CERI, Tokyo, Japan) and performed using the Agilent Human Whole Genome Oligo Microarray platform (4x44 K, G4112F). Differentially expressed genes (DEGs) were defined as those showing a fold-change greater than 2 and a false discovery rate (FDR) <0.05. Data normalization and statistical analyses were conducted using the GeneSpring GX software (Agilent Technologies).

To elucidate biological pathways associated with PpIX detection resistance and cell motility, DEGs identified in microarray data were subjected to GSEA (31). Analyses were performed using the fgsea R package (v1.26.0; Bioconductor) with Molecular Signatures Database Hallmark and Gene Ontology gene sets (32,33). Genes were ranked according to the log₂ fold change between groups, and enrichment scores were calculated using 10,000 permutations. Significantly enriched pathways were defined by a FDR<0.05. Functional categories related to cytoskeleton regulation, heme metabolism, and membrane transport are highlighted.

Public transcriptomic data of patients with GBM (n=525) were obtained from TCGA (TCGA_GBM; HG-U133A platform) using the GlioVis portal (https://gliovis.bioinfo.cnio.es/) (34).

All data are expressed as the mean ± standard deviation (SD). Comparisons between two groups were performed using unpaired two-tailed Student's t-test, and correlations were analyzed using Pearson's correlation coefficient. Survival differences in TCGA data were evaluated using Kaplan-Meier survival analysis with log-rank and Gehan-Breslow-Wilcoxon tests. A P-value <0.05 was considered to indicate a statistically significant difference.

To assess whether migratory potential influences intraoperative detectability in GBM, we examined the relationship between cell motility and 5-ALA-induced PpIX fluorescence in six patient-derived GBM cell lines (KBT#10135, KBT#12137, KBT#10170, PDM19, PDM22, and PDM123). Flow cytometric analysis of the recurrent GBM cell line PDM123 revealed heterogeneous PpIX fluorescence with a distinct population lacking detectable signals (Fig. 1A). Across the six lines, the proportion of PpIX-negative cells varied substantially, with the primary line KBT#12137 showing the highest PpIX accumulation and the recurrent line PDM123 showing the lowest (approximately 70% PpIX-negative cells) (Fig. 1A and B). Although the recurrent lines tended to exhibit weaker fluorescence than the primary lines, the difference between the two groups was statistically significant (P<0.05; two-tailed t-test, n=3 lines per group). Transwell migration assays demonstrated that recurrent GBM cell lines exhibited significantly higher migratory activity than primary cell lines (P<0.05; two-tailed t-test, n=3 lines per group) (Fig. 1C). When the proportion of PpIX-negative cells was plotted against the migratory index [defined as the migration rate (%)=(lower-/upper-chamber cell counts) x100], positive correlations were observed within both the primary and recurrent groups (Pearson's r=0.8578 and 0.9849, respectively; Fig. 1D). The overall correlation across all six lines was positive (Pearson's r=0.7852). These findings demonstrate that highly motile GBM cells accumulate lower levels of PpIX upon 5-ALA exposure and are thus less detectable during fluorescence-guided surgery. Collectively, these data support functional coupling between intrinsic motility and diagnostic invisibility, which may contribute to incomplete tumor removal and postsurgical recurrence.

To further test whether the migration status directly affects 5-ALA detectability at the subpopulation level, each patient-derived GBM culture was fractionated into migratory and non-migratory populations using Transwell assays, followed by PpIX fluorescence analysis. In the recurrent line PDM123, the migratory (lower chamber) population contained a markedly higher proportion of PpIX-negative cells than the non-migratory (upper chamber) population (Fig. 2A). This trend was consistently observed across all six GBM cell lines, with five lines (all except PDM19) showing a significantly greater proportion of PpIX-negative cells in the migratory fraction (^*P<0.05, ^**P<0.01) (Fig. 2B). The relative abundance of the migratory PpIX-negative population was positively correlated with the overall migratory index across cell lines (r=0.72, P<0.05). Thus, 5-ALA detectability is inversely linked to migratory behavior and a minor subpopulation that combines high motility with low fluorescence likely represents a surgically invisible fraction that contributes to tumor infiltration and recurrence.

To define the molecular programs underlying this dual phenotype, we performed genome-wide transcriptomic profiling of four phenotypically defined subpopulations-migratory PpIX-high, migratory PpIX-low, non-migratory PpIX-high, and non-migratory PpIX-low-sorted from representative GBM lines. Genes specifically upregulated in the migratory PpIX-low subset were defined by combined criteria of selectivity and amplitude (log₂FC_rest >1 and log₂FC_CA, CB, CD >0.5). Under this definition, migratory PpIX-low cells exhibited a focused but distinct transcriptional signature comprising 10 upregulated and 42 downregulated genes, relative to the other three groups. The upregulated genes included SLCO4A1-AS1, HES5, LOC105374643, NMU, and NT5E (CD73), whereas the most strongly downregulated genes were MT1G, GPNMB.1, C1S.1, FOSB, and SPOCK2.1 (Fig. 3A and B). These gene expression patterns suggest a coordinated shift toward a proliferative, stress-resistant, and metabolically adaptive phenotype. For instance, HES5, a Notch-pathway effector, and NT5E, an ecto-5'-nucleotidase that generates adenosine and suppresses immune activation, both contribute to stem-like persistence and microenvironmental tolerance. NMU encodes neuromedin U, a neuropeptide known to enhance GBM motility and angiogenic signaling, whereas SLCO4A1-AS1 is an antisense RNA associated with hypoxia-driven drug transport regulation. Conversely, MT1G and FOSB are oxidative stress-responsive genes typically induced by differentiation or inflammatory cues, and SPOCK2 encodes an extracellular matrix proteoglycan involved in glial maturation; their suppression further indicates an undifferentiated, high-motility cell state. GSEA of the full expression dataset provided a broader functional overview (Fig. 3C) (31), with full Hallmark enrichment results listed in Table SI. Hallmark pathways enriched among the upregulated genes included E2F_TARGETS, MYC_TARGETS_V1, G2M_CHECKPOINT, and OXIDATIVE_PHOSPHORYLATION, reflecting enhanced proliferative and metabolic activation consistent with a cycling, self-renewing phenotype. In contrast, downregulated pathways included TNFA_SIGNALING_VIA_NFKB, INTERFERON_GAMMA_RESPONSE, INFLAMMATORY_RESPONSE, and CHOLESTEROL_HOMEOSTASIS, indicating the suppression of stress-responsive and immunomodulatory transcriptional programs. This pattern suggests that migratory PpIX-low cells adopt a metabolically efficient, yet immunologically quiescent state that may permit tissue dissemination and survival under oxidative or therapeutic stress. Together with the downregulation of heme and porphyrin biosynthetic genes, these data imply that the migratory PpIX-low phenotype is sustained by a coupled transcriptional network integrating cell cycle activation, metabolic adaptation, and reduced immunogenicity, which collectively diminish 5-ALA-derived fluorescence and promote tumor persistence beyond the surgical margin. Collectively, these findings define the migratory PpIX-low state as a transcriptionally integrated mode of high-motility, metabolically adaptive, and diagnostically invisible tumor persistence.

To evaluate the clinical relevance of this program, we constructed a migratory PpIX-low gene signature comprising the top 200 differentially expressed genes (70 upregulated and 130 downregulated; Table SII) and applied it to the TCGA GBM dataset (n=525) using the GlioVis data portal (34), with the clinical survival dataset provided in Table SIII. Kaplan-Meier analysis revealed that patients with high migratory PpIX-low signature scores (n=263) exhibited a significantly shorter overall survival than those with low scores (n=262; P=0.0404, log-rank test) (Fig. 3D). Taken together, these findings establish that the migratory PpIX-low state is defined by a transcriptional program that links motility and metabolic evasion, thereby conferring both high migratory capacity and diagnostic invisibility. Its association with poor patient outcomes underscores its translational significance and suggests that therapeutic modulation of this state, by restoring porphyrin biosynthesis or sensitizing migratory cells to 5-ALA, could improve intraoperative fluorescence detection and surgical efficacy in GBM.