Spatial heterogeneity of malignant molecular markers in glioma revealed by stereotactic biopsy

Abstract

Purpose

To investigate the spatial distribution of key malignant molecular markers in glioma using MRI-guided stereotactic biopsies for tissue validation.

Materials and methods

Thirty-seven patients with newly suspected gliomas underwent preoperative magnetic resonance imaging (MRI) followed by MRI-guided stereotactic biopsy and subsequent open craniotomy. Using preoperative MRI (T2-weighted imaging [T2WI]/fluid-attenuated inversion recovery [FLAIR] and contrast-enhanced T1WI), each tumor was radiologically subdivided into core, middle, and peripheral regions. Representative tissues from these regions were obtained via stereotactic biopsy. Immunohistochemistry (IHC) was performed to detect the expression of Ki67, ZEB1, NOTCH1, and HIF-1α.

Results

The expression levels of Ki67, ZEB1, NOTCH1, and HIF-1α were positively correlated with increasing glioma grade (Kruskal–Wallis test, all P < 0.05).A significant spatial gradient was observed for Ki67 (P < 0.001), ZEB1 (P < 0.001), HIF-1α (P < 0.001), and NOTCH1 (P = 0.001) expression across the core, middle, and peripheral regions (Friedman test). Post-hoc pairwise comparisons confirmed the most pronounced difference was between the core and periphery for all markers (all P < 0.001).

Conclusions

The expression of Ki67, ZEB1, NOTCH1, and HIF-1α is closely associated with the malignant grade of gliomas and exhibits a conserved “core-to-periphery” decreasing gradient. The tumor core appears to be the molecular epicenter of aggressive phenotypes. These findings underscore the importance of targeting the radiologic core for diagnostic biopsy to avoid under-grading and suggest that spatial heterogeneity should be considered in therapeutic planning.

Introduction

Glioma is the most common and aggressive primary malignant tumor of the central nervous system. The 2021 World Health Organization (WHO) classification integrates histopathological and molecular features to categorize adult-type diffuse gliomas into grades 2 to 4, with glioblastoma (GBM, grade 4) representing the most lethal subtype [1, 2]. Despite multimodal treatment involving maximal safe resection, radiotherapy, and temozolomide chemotherapy, the prognosis for high-grade gliomas remains dismal, with a median survival of 14–16 months for GBM [3, 4]. Emerging therapies, including immunotherapy and targeted agents, have shown limited success, highlighting the urgent need for a deeper understanding of glioma biology to develop more effective strategies [5]. A major obstacle to treatment efficacy is significant intratumoral heterogeneity, which is not fully captured by conventional diagnostic approaches.

Molecular markers associated with core malignant behaviors—such as proliferation, invasion, stemness, and treatment resistance—provide critical insights into tumor aggressiveness. Ki67 is a well-established proliferation index whose expression correlates with glioma grade and poor prognosis [6–9]. ZEB1 is a master regulator of epithelial–mesenchymal transition (EMT), promoting invasion, chemoresistance, and a stem-cell-like phenotype [10–13]. Hypoxia-inducible factor 1-alpha (HIF-1α) drives adaptive responses to hypoxia, contributing to angiogenesis, stemness maintenance, and therapy resistance [14–17]. NOTCH1 signaling is involved in cell fate decisions and is often upregulated in aggressive tumors, potentially sustaining glioma stem cell populations [18–21].

Most previous studies have analyzed these markers in bulk tumor samples, overlooking their potential spatial distribution within the tumor mass. This oversight may compromise the accuracy of prognostic assessments and limit the efficacy of spatially targeted therapies. However, correlative studies using precisely mapped clinical biopsy specimens to define this molecular topography remain scarce.

Therefore, to elucidate the spatial biology of gliomas and inform clinical practice, we investigated the expression patterns of Ki67, ZEB1, HIF-1α, and NOTCH1 across distinct radiologically defined regions within untreated gliomas. Using MRI-guided stereotactic biopsies, we systematically sampled the tumor core, transitional zone, and periphery. We hypothesized that the expression of these malignant markers would not only increase with tumor grade but also exhibit a spatially organized gradient, with the highest concentration in the tumor core. Mapping this molecular landscape aims to provide critical guidance for optimizing biopsy targeting, surgical planning, and the development of spatially-informed therapeutic strategies.

Materials and methods

Study population and ethics

This prospective study was conducted at Shandong Provincial Hospital Affiliated to Shandong First Medical University from August 2019 to April 2022. Patients with a suspected solitary intracranial glioma based on external hospital reports were referred for further evaluation. All patients underwent a standardized preoperative MRI protocol. Exclusion criteria were: (1) refusal to undergo MRI or surgical intervention; (2) final histopathological diagnosis other than glioma; (3) biopsy specimens deemed inadequate for analysis (e.g., minimal tissue, extensive necrosis); (4) pediatric patients excluded. Glioma diagnosis and grading strictly followed the 2021 WHO Classification of Tumors of the Central Nervous System (5th edition). Written informed consent was obtained from all participants. The study protocol was approved by the Hospital Ethics Committee and conducted in accordance with the Declaration of Helsinki.

Preoperative MRI and tumor region definition

Preoperative MRI was performed on a 3.0 T scanner (Philips Ingenia) using a standardized protocol including: axial T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1WI (0.1 mmol/kg gadobutrol). Slice thickness was 5 mm with no gap.

Tumor regions were defined based on a pre-specified decision tree applied by consensus between an experienced neurosurgeon and a neuroradiologist, using fused T2WI/FLAIR and post-contrast T1WI sequences. Peritumoral edema, characterized by FLAIR hyperintensity without corresponding T2WI changes and mass effect distinct from the solid tumor, was excluded from region delineation.

For contrast-enhancing tumors:

• Core Region: The area demonstrating vivid contrast enhancement on T1WI.

• Middle Region: The peri-enhancing zone on T1WI or the intermediate signal intensity area surrounding the core on T2WI/FLAIR, which may show indistinct or patchy enhancement.

• Peripheral Region: The outermost area of the tumor, showing mild T2WI/FLAIR hyperintensity blending with normal brain parenchyma and demonstrating no contrast enhancement.

For non-enhancing tumors:

• Core Region: The region with the highest and most homogeneous signal intensity on T2WI/FLAIR.

• Middle Region: The zone of intermediate T2/FLAIR signal intensity immediately surrounding the core.

• Peripheral Region: The area of faint T2/FLAIR hyperintensity at the tumor-brain interface.

Stereotactic biopsy procedure

Using the BrainLab VectorVision2 navigation system, 3 to 6 biopsy trajectories were pre-planned to target the defined core, middle, and peripheral regions. Areas of obvious liquefactive necrosis, hemorrhage, or large vessels were avoided. During surgery, prior to any resection, stereotactic biopsies were obtained using a side-cutting needle, yielding cylindrical specimens (approximately 1 mm in diameter × 10 mm in length). Each biopsy site was meticulously recorded according to its pre-defined region. For subsequent analysis, the patient was the unit of analysis. If multiple biopsy specimens were obtained from the same predefined region (core, middle, or periphery) within a patient, the immunohistochemistry scores from those specimens were averaged to yield a single, representative value for that region per patient. The procedure’s safety was monitored; no symptomatic hemorrhages or permanent neurological deficits related to the biopsy were recorded in this cohort.

Immunohistochemistry (IHC)

Formalin-fixed, paraffin-embedded tissue Sect.  (5 μm thick) were used. After deparaffinization and rehydration, antigen retrieval was performed using citrate buffer (pH 6.0) under high pressure. Endogenous peroxidase activity was blocked with 3% H₂O₂. Sections were incubated overnight at 4 °C with the following primary antibodies: Ki67 (Proteintech, Cat# 27309-1-AP, rabbit polyclonal, dilution 1:100), ZEB1 (Proteintech, Cat# 21544-1-AP, rabbit polyclonal, dilution 1:200), HIF-1α (Abcam, Cat# ab216842, mouse monoclonal, dilution 1:300), and NOTCH1 (Abcam, Cat# ab52627, rabbit monoclonal, dilution 1:150). This antibody targets the activated intracellular domain of NOTCH1; therefore, nuclear staining was interpreted as indicative of active NOTCH1 signaling and scored as positive. Cytoplasmic or membranous staining was rarely observed and was not considered positive for the purpose of this study. After washing, sections were incubated with the appropriate secondary antibody from the PV-9000 kit. Positive control tissues (e.g., tonsil for Ki67, breast carcinoma for HIF-1α) and negative controls (omission of the primary antibody) were included in each staining batch. All regional samples from the same patient were stained in the same batch to minimize variability. Detection was performed using the PV-9000 Polymer Detection System (ORIGENE), followed by counterstaining with hematoxylin.

Digital analysis and semi-quantitative scoring

Whole slides were digitized using an Olympus VS200 slide scanner at 20× magnification, yielding a resolution of 0.322 μm/pixel. Two experienced neuropathologists, blinded to the tumor region and grade, independently scored the slides. Field Selection and Scoring : For each marker, three high-power fields (HPF, 400×) per specimen were selected from areas with the highest staining density (“hotspots”) according to the following predefined criteria: (1) the field contained a minimum of 100 viable tumor cells; (2) areas of necrosis, hemorrhage, or significant artifacts comprised less than 10% of the field; (3) fields at the extreme edge of the tissue section were avoided. The average score of the three HPFs was used as the final expression value for that marker in that region. The percentage of positive tumor cells per HPF was estimated. A semi-quantitative score was assigned: 0 (< 5%), 1 (6–15%), 2 (16–40%), 3 (> 40%). Nuclear staining was considered positive for Ki67 and NOTCH1. For ZEB1, nuclear with or without cytoplasmic staining was considered positive. For HIF-1α, nuclear and/or cytoplasmic staining was considered positive. Discrepancies in scores were resolved by joint review. The average score of the two reviewers for each specimen was used for statistical analysis. Inter-rater reliability was high (Intraclass Correlation Coefficient > 0.85 for all markers).

Statistical analysis

Statistical analysis was performed using GraphPad Prism (v9.0) and R software (v4.2.0). Continuous data are presented as mean ± standard deviation. The patient was the unit of analysis. The Kruskal–Wallis test was applied separately for each molecular marker (Ki67, ZEB1, HIF-1α, NOTCH1) to compare expression scores across WHO grades (2, 3, and 4). Dunn‘s post-hoc test was used for pairwise grade comparisons, with p-values adjusted using the Holm method to control for multiple testing across the four markers.

For the analysis of spatial heterogeneity, differences in marker expression across the Core, Middle, and Peripheral regions (within-patient comparison) were assessed for each marker separately using the Friedman test. When the Friedman test was significant, Dunn’s post-hoc test was employed for pairwise regional comparisons. To account for multiple comparisons across the four markers and three region pairs, the false discovery rate (FDR) was controlled using the Benjamini–Hochberg procedure. Effect sizes for significant post-hoc comparisons are reported as median differences. A two-sided p-value < 0.05 was considered statistically significant.

Results

Patient cohort and tumor classification

A total of 37 patients with treatment-naïve gliomas were included in this study. All patients underwent preoperative MRI, followed by MRI-guided stereotactic biopsy sampling from predefined tumor regions and subsequent open resection. According to the 2021 WHO classification, the cohort consisted of 11 patients with grade 2 gliomas (IDH-mutant astrocytoma or oligodendroglioma with 1p/19q co-deletion), 8 patients with grade 3 gliomas (IDH-mutant astrocytoma or oligodendroglioma), and 18 patients with IDH-wildtype glioblastoma (grade 4). The clinical and radiological characteristics of the patients are summarized in Table 1.

Table 1.

Baseline characteristics of the study cohort (n = 37)

Characteristic	Value
Age (years), mean ± SD	52.6 ± 12.4
Sex, n (%)
Male	21 (56.8)
Female	16 (43.2)
Tumor location, n (%)
Frontal	15 (40.5)
Temporal	11 (29.7)
Parietal	7 (18.9)
Occipital	4 (10.8)
WHO grade, n (%)
Grade 2	11 (29.7)
Grade 3	8 (21.6)
Grade 4	18 (48.6)
Preoperative tumor diameter (cm), mean ± SD	4.1 ± 1.3

Expression of molecular markers increases with glioma grade

The expression levels of Ki67, ZEB1, HIF-1α, and NOTCH1 were positively correlated with increasing histological grade (Kruskal–Wallis test: Ki67 P = 0.018, ZEB1 P = 0.032, HIF-1α P = 0.041, NOTCH1 P = 0.039).The distribution of semi-quantitative immunohistochemistry (IHC) scores across different grades is detailed in Table 2. Representative photomicrographs illustrating the expression patterns of all four markers in low-grade (grade 2) and high-grade (grade 4) gliomas are presented in Fig. 1.

Table 2.

Distribution of semi-quantitative IHC scores for Ki67, ZEB1, HIF-1α, and NOTCH1 across WHO glioma grades (Number of patients)

Grade	Score	Ki67	ZEB1	HIF-1α	NOTCH1
Grade 2 (n = 11)	0 (< 5%)	0	1	2	4
	1 (6–15%)	6	5	6	2
	2 (16–40%)	5	5	3	5
	3 (> 40%)	0	0	0	0
Grade 3 (n = 8)	0 (< 5%)	0	1	1	2
	1 (6–15%)	2	4	4	1
	2 (16–40%)	3	2	3	3
	3 (> 40%)	3	1	0	2
Grade 4 (n = 18)	0 (< 5%)	1	2	2	5
	1 (6–15%)	4	6	8	3
	2 (16–40%)	5	5	10	9
	3 (> 40%)	8	5	6	9

Fig. 1.

Immunohistochemical analysis of malignant molecular markers in human glioma biopsy specimens.Representative images of immunohistochemical staining for Ki-67 (A1, A2), ZEB1 (B1, B2), HIF-1α (C1, C2), and NOTCH1 (D1, D2) in low-grade glioma (LGG, left panels) and high-grade glioma (HGG, right panels). Nuclear staining for Ki-67 and NOTCH1, nuclear and/or cytoplasmic staining for ZEB1, and nuclear and cytoplasmic staining for HIF-1α (indicative of hypoxic tumor regions) were observed. Scale bars = 50 μm. Corresponding semi-quantitative scoring of Ki-67 (A3), ZEB1 (B3), HIF-1α (C3), and NOTCH1 (D3) expression in LGG and HGG tissues is summarized in Table 2 (*P < 0.05, **P < 0.01)

Spatial heterogeneity reveals a core-to-periphery expression gradient

Analysis of region-specific biopsies revealed a significant spatial gradient in expression from the tumor core to the periphery for Ki67 (Friedman test, P < 0.001), ZEB1 (P < 0.001), HIF-1α (P < 0.001), and NOTCH1 (P = 0.001). Post-hoc pairwise comparisons with FDR adjustment confirmed that for all four markers, expression in the Core was significantly higher than in the Periphery (all adjusted P < 0.001). Detailed results of all pairwise comparisons are available upon request.The mean IHC scores consistently decreased from the Core to the Middle to the Periphery for all markers (Table 3; Fig. 2). Representative IHC staining from the three regions of the same tumor are shown in Figs. 3, 4, 5 and 6.

Table 3.

Semi-quantitative IHC scores (mean ± SD) for Ki67, ZEB1, HIF-1α, and NOTCH1 in Core, Middle, and Periphery tumor regions (n = 37 patients)

Marker	Core	Middle	Periphery	P-value (Friedman Test)
Ki67	1.95 ± 0.80	1.19 ± 0.63	0.51 ± 0.50	< 0.001
ZEB1	1.97 ± 0.91	1.16 ± 0.65	0.62 ± 0.62	< 0.001
HIF-1α	1.54 ± 0.82	0.92 ± 0.67	0.41 ± 0.53	< 0.001
NOTCH1	1.65 ± 1.04	1.11 ± 0.91	0.65 ± 0.90	0.001

Fig. 2.

Consolidated spatial gradient of molecular marker expression. Line graph illustrating the mean semi-quantitative IHC score (± standard deviation) for Ki67, ZEB1, HIF-1α, and NOTCH1 across the Core, Middle, and Periphery tumor regions (n = 37). The consistent declining trend from the core to the periphery is visually summarized

Fig. 3.

Spatial distribution of Ki67 expression. Representative IHC staining of Ki67 in the Core (A), Middle (B), and Periphery (C) regions from the same glioblastoma case. Violin plots with superimposed boxplots showing the distribution of semi-quantitative Ki67 scores across the three tumor regions(D). Statistical significance was determined by the Friedman test with Dunn’s post-hoc test (**P < 0.01, ****P < 0.0001). Scale bars in A = 50 μm

Fig. 4.

Spatial distribution of ZEB1 expression. Representative IHC staining of ZEB1 in the Core (A), Middle (B), and Periphery (C) regions from the same glioblastoma case. Violin plots with superimposed boxplots showing the distribution of semi-quantitative ZEB1 scores across the three tumor regions(D). Statistical significance was determined by the Friedman test with Dunn’s post-hoc test (*P < 0.05, ***P < 0.001, ****P < 0.0001). Scale bars in A = 50 μm

Fig. 5.

Spatial distribution of HIF-1α expression.Representative IHC staining of HIF-1α in the Core (A), Middle (B), and Periphery (C) regions from the same glioblastoma case. Violin plots with superimposed boxplots showing the distribution of semi-quantitative HIF-1α scores across the three tumor regions(D). Statistical significance was determined by the Friedman test with Dunn’s post-hoc test (**P < 0.01, ***P < 0.001, ****P < 0.0001). Scale bars in A = 50 μm

Fig. 6.

Spatial distribution of NOTCH1 expression. Representative IHC staining of NOTCH1 in the Core (A), Middle (B), and Periphery (C) regions from the same glioblastoma case. Violin plots with superimposed boxplots showing the distribution of semi-quantitative NOTCH1 scores across the three tumor regions(D). Statistical significance was determined by the Friedman test with Dunn’s post-hoc test (**P < 0.01). Scale bars in A = 50 μm

Ki67 expression: a gradient of cellular proliferation

Ki67, a canonical marker of active cell cycle progression, exhibited a striking spatial compartmentalization. Ki67-positive nuclei were densely packed within the tumor Core, reflecting a region of high proliferative activity (Fig. 3A). Quantitatively, the expression score demonstrated a significant and stepwise decline from the Core to the Middle, and further to the Periphery (Core vs. Middle: P < 0.01; Core vs. Periphery: P < 0.0001; Middle vs. Periphery: P < 0.01; Dunn’s post-hoc test) (Fig. 3B). This clear proliferative gradient underscores the Core as the primary epicenter of tumor growth and expansion.

ZEB1 expression: mapping the invasive and stem-like niche

The expression pattern of ZEB1, a master regulator of epithelial-mesenchymal transition (EMT) and stemness, mirrored the malignant topography. ZEB1 immunoreactivity, predominantly nuclear, was most intense in the Core region (Fig. 4A). Statistical analysis confirmed a robust decreasing gradient from the Core outwards (Core vs. Middle: P < 0.001; Core vs. Periphery: P < 0.0001; Middle vs. Periphery: P < 0.05; Dunn’s post-hoc test) (Fig. 4B). This spatial restriction suggests that the molecular drivers of invasion, chemoresistance, and a stem-cell phenotype are concentrated within the tumor’s central compartment.

HIF-1α expression: a hypoxic core driving adaptation

Consistent with the typically hypoxic tumor core, HIF-1α protein stabilization was markedly elevated in this region, with staining observed in both the cytoplasm and nucleus (Fig. 5A). The expression scores formed a steep gradient, with the Core showing significantly higher levels than both the Middle and Periphery (Core vs. Middle: P < 0.01; Core vs. Periphery: P < 0.0001; Dunn’s post-hoc test). The Middle region itself retained higher expression than the Periphery (P < 0.01) (Fig. 5B). This gradient visually maps the hypoxic stress within the tumor, which is known to fuel angiogenesis, metabolic reprogramming, and therapy resistance.

NOTCH1 expression: a stemness signal with concentrated activity

NOTCH1 signaling, implicated in maintaining glioma stem cells, displayed a trend of decreasing expression from the Core to the Periphery (Fig. 6A). While a decreasing trend was observed across all three regions, post-hoc analysis identified the most pronounced and statistically significant difference specifically between the Core and the distant Periphery (P < 0.001; Dunn’s post-hoc test) (Fig. 6B). This pattern indicates that active NOTCH1 signaling, potentially supporting a stem-like and therapy-resistant subpopulation, is preferentially localized to the central tumor mass.

Discussion

Our study provides a detailed spatial map of key malignant markers in human gliomas, revealing a consistent “core-to-periphery” decreasing gradient for Ki67, ZEB1, HIF-1α, and NOTCH1 expression, which is intrinsically linked to tumor grade. This finding underscores the profound molecular heterogeneity within individual gliomas and has direct implications for clinical management.

The high expression of these markers in the tumor core aligns with its pathophysiological role as the epicenter of proliferation (Ki67), hypoxic stress (HIF-1α), and processes associated with stemness and invasion (NOTCH1, ZEB1) [22–24]. The observed gradient suggests that the core maintains a microenvironment conducive to aggressive phenotypes. For instance, core hypoxia may stabilize HIF-1α, which can upregulate NOTCH1, potentially reinforcing a stem-like state and promoting EMT via ZEB1, thereby creating a spatially organized axis of malignancy [21, 25]. However, our cross-sectional data cannot establish causality or temporal sequence; they describe a stable spatial pattern consistent with this model.

The clinical implications are significant. First, for diagnostic biopsy, sampling solely the peripheral, lower-grade appearing region risks under-grading and missing the most aggressive molecular profile. Our data strongly advocate for targeting the radiologic core (enhancing region or T2/FLAIR hyperintense core) to obtain a representative diagnostic sample. Second, in surgical planning, the concept of “maximal safe resection” may need to incorporate molecular boundaries. While complete resection of the infiltrative periphery is often impossible, aggressive removal of the molecularly “active” core could be prioritized. Techniques like 5-aminolevulinic acid (5-ALA) fluorescence, which often highlights core regions, align with this principle [26]. Third, for therapy development, our findings suggest that treatments targeting proliferation, hypoxia, or stemness pathways (e.g., NOTCH inhibitors, anti-angiogenics) may need to be optimized for delivery to the relatively less accessible tumor core, whereas strategies aimed at preventing invasion might focus on the peripheral niche.

Our study has limitations. The sample size, though sufficient to detect the primary spatial pattern, is moderate. IHC provides protein expression data but not functional activation. While our semi-quantitative scoring was rigorous and blinded, the hotspot-based approach, although standard for assessing regional peak activity, may influence the magnitude of observed gradients. We applied FDR correction for regional spatial comparisons, but the exploratory analysis of multiple markers warrants validation in independent cohorts. The definition of the tumor core in non-enhancing tumors, while based on clear radiological criteria (highest T2/FLAIR signal), may not be biologically uniform across all subtypes. Finally, the cross-sectional design cannot inform temporal evolution.

Conclusion

In conclusion, our study provides direct histological evidence of a conserved spatial “malignancy gradient” in gliomas, characterized by the core-to-periphery decline of Ki67, ZEB1, HIF-1α, and NOTCH1. This fundamental spatial organization challenges the view of glioma as a uniform entity and compels a paradigm shift towards spatially-informed management. Our findings support anchoring diagnostic biopsy to molecular epicenters and suggest that surgical and radiotherapeutic planning, as well as targeted therapies, could be optimized by considering this regional molecular heterogeneity. Future research integrating spatial omics and longitudinal sampling is needed to validate these gradients, link them to clinical outcomes, and propel the development of personalized, spatially-aware glioma therapeutics.

Author contributions

C.Q., K.W., and S.Z. conceptualized and designed the study. C.Q., K.W., C.W., C.W., and Y.M. performed the experiments and data acquisition. L.P., H.H., and Y.M. conducted the formal analysis and data curation. C.Q. and K.W. wrote the original draft of the manuscript. S.Z., C.W., and H.H. reviewed and edited the manuscript. S.Z. supervised the project, acquired funding, and administered the project. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China [No. 82072780]; the Science and Technology Department of Sichuan Province [No. 2022YFS0630]; and Luzhou Municipal People’s Government [No. 2020-SYF-28].

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Ethics Committee of Shandong Provincial Hospital Affiliated to Shandong First Medical University. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments. Informed consent was obtained from all individual participants included in the study.

Consent for publication

Participants consented to the publication of anonymized data and findings derived from their samples and clinical information.

Competing interests

The authors declare no competing interests.