Abstract
Background
Tertiary lymphoid structures (TLSs) correlate with improved survival in various cancers, but their prognostic significance and imaging correlations in glioblastoma (GBM) remain unclear. This study investigated the relationship between TLSs, survival outcomes, and imaging features in GBM patients.
Methods
A retrospective analysis of 190 newly diagnosed GBM patients was conducted. TLSs were identified via hematoxylin-eosin staining, with maturity assessed by multiplex immunofluorescence (CD20/CD3/CD21/CD23) into mature (mTLSs) and immature (imTLSs) subgroups. Survival analysis used Kaplan–Meier method, and TLS predictive models were developed via logistic regression. Clinicopathological and VASARI imaging features were compared.
Results
Of 190 cases, 85 (44.73%) were TLSs+, comprising 47 imTLSs and 38 mTLSs. Mean overall survival (OS) was 11.82 months. Significant OS differences were observed between TLSs+ (14.67 months) and TLSs− (9.51 months) groups, and between mTLSs (19.36 months) and imTLSs (10.87 months) groups (all P < .05). VASARI features f2, f8, f17, and f22 differed significantly between TLSs+/TLSs− groups, while f6, f16, and f18 differed between mTLSs/imTLSs groups (all P < .05). These features were independent predictors of TLS presence and maturity.
Conclusion
TLSs exist at varying maturation stages in GBM and represent favorable prognostic biomarkers. Their presence and maturity significantly correlate with OS and specific VASARI features. Preoperative imaging prediction of TLSs may facilitate risk stratification and individualized treatment for GBM patients.
Keywords: glioblastoma, survival analysis, tertiary lymphoid structure, VASARI features
Key Points.
Tertiary Lymphoid Structures (TLSs) are favorable prognostic biomarkers in GBM. The presence and maturity of TLSs are significantly associated with improved overall survival in glioblastoma patients.
Specific imaging features can non-invasively predict TLS status. Preoperative VASARI MRI features are significantly correlated with the presence and maturity of TLSs, allowing for non-invasive prediction and potential risk stratification.
Importance of the Study
The prognosis of glioblastoma remains poor, with considerable heterogeneity among patients. Tertiary lymphoid structures (TLSs) have emerged as promising immune-related prognostic biomarkers associated with improved survival in various cancers. However, their significance in GBM remains underexplored. Current methods for detecting and quantifying TLSs rely heavily on expensive, complex, and time-consuming histopathological techniques, which limit their clinical applicability. This study demonstrated that TLSs are present in patients with GBM and are independently associated with prolonged survival. Furthermore, for the first time, we established a significant correlation between the presence and maturation of TLSs and specific preoperative VASARI MRI features. These findings support the use of routine MRI-based characteristics as non-invasive, accessible, and reproducible biomarkers for predicting TLS status. Our proposed imaging model could enhance preoperative risk stratification and help guide immunotherapeutic strategies, offering a practical tool for personalized management of patients with GBM.
Glioblastoma (GBM) is the most common primary malignant tumor in the brain, characterized by the blood-brain barrier, tumor angiogenesis, tumor microenvironment heterogeneity, and treatment resistance.1 Most patients with GBM experience recurrence shortly after initial diagnosis and treatment, with a 5-year survival rate of 7.1% and a median survival of only 9 months.2 Therefore, it is necessary to conduct in-depth research into prognosis-related microenvironmental subregions in GBM to identify reliable prognostic factors, provide evidence for immunotherapy strategies, and thereby enable early, precise risk stratification to improve patient outcomes.
Tertiary lymphoid structures (TLSs) are ectopic lymph node-like structures that contain B and T cells.3 Currently, TLSs are regarded as hubs of anti-tumor immunity, as they can promote anti-tumor immune responses by effectively inducing stronger or broader immune reactions.4–6 Studies have confirmed that the presence and maturity of TLSs are associated with improved prognosis in patients with cancer, making them important biomarkers for evaluating the efficacy of immunotherapy and predicting outcomes.7
TLSs in GBM were first reported in 2021, when researchers discovered that an agonistic CD40 antibody (αCD40) could induce and enhance the formation of TLSs in GBM.8 In mouse models of GBM, tumor-associated TLSs have been demonstrated to correlate with better treatment responses and improved prognosis.9,10 Among these, targeting the LTβR signaling axis has shown the greatest promise in promoting TLSs formation, improving survival rates, and enhancing anti-tumor responses in GBM.10 Therefore, identifying and quantitatively analyzing TLSs can pave the way for the application of immune checkpoint inhibitors in GBM, provide a theoretical basis for targeting TLSs, and facilitate the assessment of patient prognosis.11 Although the research value of TLSs in tumor immunology has been well established, their detection continues to face significant challenges.7 TLSs are important biomarkers for predicting treatment efficacy and prognosis, and objective, simple, and novel technologies are needed to characterize them.
Mapping and topological validation of imaging and histopathological features are highly significant for non-invasive studies of the tumor microenvironment. Several studies have confirmed that imaging features and radiomics models can predict the presence of TLSs in various solid tumors, such as hepatocellular carcinoma and breast cancer, offering new insights for the clinical diagnosis and treatment of tumors.12,13 The Visually Accessible Rembrandt Images (VASARI) is a standardized feature set for characterizing MR images of human GBMs; it is considered clinically meaningful, widely available, and relevant to GBM prognosis.14 Han et al.applied a dichotomization approach to VASARI features and identified seven features associated with overall survival (OS) in patients with GBM.15
However, the non-invasive prediction of TLSs in GBM based on preoperative imaging features has not been studied. Therefore, this study evaluated the expression and maturity of TLSs in GBM, analyzed patient survival rates between TLSs-positive and TLSs-negative groups and between mature and immature TLSs groups, examined the relationship between clinical, molecular, and pathological features, VASARI characteristics, and TLSs expression and maturity, and established a connection between macroscopic imaging features and tumor microenvironment characteristics.
Methods
In this study, we retrospectively collected and evaluated tumor tissue specimens from patients. During the research, the participants’ samples were properly preserved. This study was approved by the Medical Ethics Committee of our institution (Ethics Approval No. 2025A-364), and the research protocol adhered to the principles of the Declaration of Helsinki.
Study Design and Patients
We collected clinical, pathological, and radiographic data from patients newly diagnosed with IDH wild-type GBM between January 1, 2019, and August 1, 2024, at the Second Hospital of Lanzhou University. Patients’ imaging was obtained from the Picture Archiving and Communication System (PACS; Syngo, Siemens Healthineers AG, Forchheim, Germany).
The inclusion criteria were as follows: according to the 2021 World Health Organization Classification of Central Nervous System Tumors, the histopathological type was GBM, with molecular testing confirming the wild-type status of IDH. The PACS contains complete clinical diagnostic and treatment information, along with comprehensive molecular pathology data (including Ki-67, P53, MGMT, and TERT). All patients underwent preoperative multisequence MRI scans (T1WI, T2WI, CE-T1WI, FLAIR, and DWI) with clear and complete images. All patients received standard treatment. As per institutional protocol for symptomatic management, all patients routinely received preoperative steroid therapy (primarily dexamethasone) for edema control prior to surgical resection. Surgically resected tumor tissue samples and formalin-fixed, paraffin-embedded specimens were available for collection and analysis. The exclusion criteria were as follows: patients under 18 years of age; patients with missing preoperative multisequence MRI scans, poor image quality, or significant artifacts; and patients whose postoperative tumor tissue samples (H&E staining, IHC staining, mIF staining slides) were unsuitable for analysis. To minimize potential confounders affecting the tumor immune microenvironment, we also excluded patients with any of the following: a known diagnosis of immune-related comorbidities (eg, autoimmune diseases) or other prior cancers; and active infection or other major systemic disorders known to affect immune function. Finally, patients lost to follow-up were excluded.
Patient follow-up: OS was defined as the period from the date of surgery to the date of death or the last follow-up. The follow-up methods in this study included reviewing patient revisit records, analyzing MRI follow-up images in the hospital information system and PACS, and conducting telephone follow-ups to assess survival status. Survival time was calculated accordingly, and the follow-up period for OS in surviving patients was set at a minimum of 6 months post-surgery.
Imaging Review
MRI-VASARI feature analysis
The acquisition parameters of MRI sequences are presented in Supplementary Material, Methods S1. VASARI features for MRI annotation include 25 imaging descriptors and 2 numerical variables based on different MRI modalities. The exact description of all features can be found in the Cancer Imaging Archives of the National Cancer Institute (https://.com/cancerimagingarchive.net/display/public/VASARI+Research+Project).
The VASARI feature set includes: f1-Tumor location, f2-Side of lesion center, f3-Eloquent brain, f4-Enhancement quality, f5-Proportion enhancing, f6-Proportion nCET, f7-Proportion of necrosis, f8-Cystic, f9-Multifocal or multicentric, f10-T1/FLAIR ratio, f11-Thickness of enhancing margin, f12 Definition of enhancing margin, f13-Definition of non-enhancing margin, f14-Proportion of edema, f15-Edema across the midline, f16-Hemorrhage, f17-Diffusion characteristics, f18-Pial Invasion, f19-Ependymal extension, f20-Cortical involvement, f21-Deep white matter invasion, f22-nCET crosses midline, f23-CET crosses midline, f24-Satellites, f25-Calvarial remodeling, f29-transverse diameter, and f30-longitudinal diameter. The MRI scans of all patients were independently evaluated by two radiologists (with experience in neuroimaging diagnosis using MRI) who described the imaging features. Any discrepancies in the retrospective assessment of the imaging features were resolved through discussion to reach a consensus.
Evaluation and quantification of TLSs
Preserved postoperative formalin-fixed paraffin-embedded tissue samples were collected and sectioned. Sectioning was performed using a LEICA RM 2235 microtome, with maximum section dimensions of 3.0 × 2.0 × 0.3 cm and a section thickness of 4 μM. H&E staining was performed. (The steps for H&E staining are detailed in Supplementary Material, Methods S2.) H&E-stained slides were scanned using a 3DHISTECH digital whole-slide scanner (Pannoramic MIDI, Budapest, Hungary) to obtain Whole Slide Images (WSIs). The output formats for the WSIs were SVS, MRXS, TIFF, etc., 2 pathologists (with 6 and 4 years of diagnostic experience, respectively) manually reviewed the WSIs using SlideViewer software (3DHISTECH, Budapest, Hungary), adjusting magnification levels as needed to assess the presence of TLSs. Patients were categorized into TLS-positive (TLSs+) and TLS-negative (TLSs−) groups. Before evaluating the H&E-stained slides, the two pathologists reached a consensus on the definition of TLSs, characterizing them as dense lymphoid aggregates located within tumor cell regions or areas of tumor fibrosis. Multiplex immunofluorescence (mIF) staining (detailed procedure in Supplementary Material, Methods S3) was employed to detect the distribution of B cells (CD20, 1:100 dilution, Kit-0001), T cells (CD3, 1:300 dilution, MAB-0740), follicular dendritic cells (CD21, 1:100 dilution, RMA-0811), and germinal centers (CD23, 1:100 dilution, RMA-0504) within TLSs and to analyze their maturity. TLSs+ were classified into an immature group (imTLSs+, characterized by the presence of only CD20+ and CD3+ cells, or CD20+, CD3+, CD21+, and CD23− cells) and a mature group (mTLSs+, characterized by the presence of CD20+, CD3+, CD21+, and CD23+ cells).
Statistics
All primary source data were processed into Excel datasheets. Data processing and analysis were performed using R version 4.3.3 (2024-02-29), along with Zstats 1.0 (www.zstats.net).
Survival analyses were performed using univariate Cox regression analyses for TLSs+/TLSs− and imTLSs/mTLSs, and results were plotted using the Kaplan–Meier method.
The Kolmogorov–Smirnov test was used to assess the normality of all continuous variables. Measurement data and count data are expressed as mean ± standard deviation (mean ± SD) and percentage, respectively. Group comparisons were performed using the t-test, χ2-test, or Fisher’s exact test. Statistical significance was defined as a 2-sided P ≤ .05. A prediction model for the expression of TLSs in GBM was established based on variables from multiple logistic regression analysis.
Results
Patients and Demographic Data
A total of 190 patients were evaluated, and the mean age at initial diagnosis was 53.69 years; 104 males (54.74%) and 86 females (45.26%). Based on H&E-stained WSIs, 85 cases (positive rate of 44.74%) showed dense lymphoid aggregates in clusters or streaks between tumor cells or perivascular areas. mIF staining revealed 47 cases of immature TLSs, including early TLSs (CD20+ and CD3+) and primary follicle-like TLSs (CD20+, CD3+, CD21+, and CD23−), and 38 cases of mature TLSs: secondary follicle-like TLSs (CD20+, CD3+, CD21+, and CD23+).The mean OS was 14.67 months for the TLSs+ group and 9.51 months for the TLSs− group. Within the TLSs+ group, the mean OS was 19.36 months for the mTLSs subgroup and 10.87 months for the imTLSs subgroup. Survival analysis demonstrated statistically significant differences in survival between the TLSs+/TLSs− groups and between the mTLSs/imTLSs groups (P < .05) (Kaplan–Meier curves are shown in Figure 1).
Figure 1.
(A) Kaplan–Meier survival analysis of TLSs+, and TLSs–. (B) Kaplan–Meier survival analysis of mTLSs+, and imTLSs+.
No significant differences were observed in patient gender or age between the TLSs+ and TLSs− groups, or between the mTLSs and imTLSs groups. Additionally, no significant correlations were observed between pathological parameters, such as the cell proliferation index Ki-67, P53 (positive/negative), MGMT methylation status (positive/negative), TERT mutation status (mutant/wild), and TLSs expression or maturity. (Correlation analysis between TLSs+/TLSs−, mTLSs+/imTLSs−, and clinicopathological features is shown in Supplementary Table S1).
Analysis of Differences in VASARI Features and Logistic Regression Modeling Between TLSs+ and TLSs− Groups
Significant differences were observed between the TLSs+ and TLSs− groups in the following VASARI features: f2-Side of lesion center (χ2 = 8.59; P = .01), f8-Cysts (χ2 = 3.88; P = .04), f17-Diffusion characteristics (P = .03), f22-Non-contrast-enhancing tumor crosses midline (χ2 = 4.07; P = .04) (Table 1). Multivariate logistic regression analysis of the f2-Side of the lesion center showed that f22-nCET crossing the midline was an independent risk factor for predicting TLSs+/TLSs− expression in patients with GBM (Table 2, forest plot as shown in Supplementary Figure S1).
Table 1.
Analysis of differences in VASARI features between TLSs+, and TLSs− groups
| Feature name | Total (N = 190) | TLSs− (N = 105) | TLSs+ (N = 85) | Statistic | P-value |
|---|---|---|---|---|---|
| f1-Tumor location, n (%) | χ² = 6.88 | .23 | |||
| Frontal | 63 (33.16) | 30 (28.57) | 33 (38.82) | ||
| Temporal | 61 (32.11) | 34 (32.38) | 27 (31.76) | ||
| Insular | 14 (7.37) | 10 (9.52) | 4 (4.71) | ||
| Parietal | 28 (14.74) | 20 (19.05) | 8 (9.41) | ||
| Occipital | 19 (10.00) | 9 (8.57) | 10 (11.76) | ||
| Cerebellum | 5 (2.63) | 2 (1.90) | 3 (3.53) | ||
| f2-Side of lesion center, n (%) | χ² = 8.59 | .01 | |||
| Right | 87 (45.79) | 39 (37.14) | 48 (56.47) | ||
| Left | 87 (45.79) | 58 (55.24) | 29 (34.12) | ||
| Center/Bilateral | 16 (8.42) | 8 (7.62) | 8 (9.41) | ||
| f3-Eloquent brain, n (%) | χ² = 4.73 | .32 | |||
| None | 23 (12.11) | 12 (11.43) | 11 (12.94) | ||
| Speech motor | 57 (30.00) | 27 (25.71) | 30 (35.29) | ||
| Speech receptive | 56 (29.47) | 30 (28.57) | 26 (30.59) | ||
| Motor | 27 (14.21) | 19 (18.10) | 8 (9.41) | ||
| Vision | 27 (14.21) | 17 (16.19) | 10 (11.76) | ||
| f4-Enhancement quality, n (%) | – | .06 | |||
| None | 3 (1.58) | 3 (2.86) | 0 (0.00) | ||
| Minimal/Mild | 21 (11.05) | 15 (14.29) | 6 (7.06) | ||
| Marked/avid | 166 (87.37) | 87 (82.86) | 79 (92.94) | ||
| f5-Proportion enhancing, n (%) | – | .69 | |||
| <5% | 6 (3.16) | 4 (3.81) | 2 (2.35) | ||
| 6-33% | 56 (29.47) | 31 (29.52) | 25 (29.41) | ||
| 34-67% | 79 (41.58) | 47 (44.76) | 32 (37.65) | ||
| 68-95% | 40 (21.05) | 19 (18.10) | 21 (24.71) | ||
| >95% | 9 (4.74) | 4 (3.81) | 5 (5.88) | ||
| f6-Proportion nCET, n (%) | χ² = 5.14 | .16 | |||
| <5% | 21 (11.05) | 7 (6.67) | 14 (16.47) | ||
| 6-33% | 40 (21.05) | 23 (21.90) | 17 (20.00) | ||
| 34-67% | 102 (53.68) | 61 (58.10) | 41 (48.24) | ||
| 68-95% | 27 (14.21) | 14 (13.33) | 13 (15.29) | ||
| f7-Proportion of necrosis, n (%) | – | .34 | |||
| None | 6 (3.16) | 3 (2.86) | 3 (3.53) | ||
| <5% | 34 (17.89) | 19 (18.10) | 15 (17.65) | ||
| 6-33% | 68 (35.79) | 34 (32.38) | 34 (40.00) | ||
| 34-67% | 75 (39.47) | 47 (44.76) | 28 (32.94) | ||
| 68-95% | 7 (3.68) | 2 (1.90) | 5 (5.88) | ||
| f8-Cysts | χ² = 3.88 | .04 | |||
| No | 82 (43.16) | 52 (49.52) | 30 (35.29) | ||
| Yes | 108 (56.84) | 53 (50.48) | 55 (64.71) | ||
| f9-Multifocal or multicentric, n (%) | – | .11 | |||
| Indeterminate | 155 (81.58) | 84 (80.00) | 71 (83.53) | ||
| Multifocal | 25 (13.16) | 13 (12.38) | 12 (14.12) | ||
| Multicentric | 9 (4.74) | 8 (7.62) | 1 (1.18) | ||
| Gliomatosis | 1 (0.53) | 0 (0.00) | 1 (1.18) | ||
| f10-T1/FLAIR ratio, n (%) | χ² = 0.85 | .65 | |||
| Expansive | 79 (41.58) | 41 (39.05) | 38 (44.71) | ||
| Mixed | 62 (32.63) | 37 (35.24) | 25 (29.41) | ||
| Infiltrative | 49 (25.79) | 27 (25.71) | 22 (25.88) | ||
| f11-Thickness of enhancing margin, n (%) | – | .05 | |||
| Indeterminate | 3 (1.58) | 3 (2.86) | 0 (0.00) | ||
| Thin | 40 (21.05) | 19 (18.10) | 21 (24.71) | ||
| Thick/solid | 69 (36.32) | 45 (42.86) | 24 (28.24) | ||
| Solid | 78 (41.05) | 48 (45.71) | 30 (35.29) | ||
| f12-Definition of enhancing margin, n (%) | χ² = 3.55 | .06 | |||
| Well-defined | 132 (69.47) | 67 (63.81) | 65 (76.47) | ||
| Poorly-defined | 58 (30.53) | 38 (36.19) | 20 (23.53) | ||
| f13-Definition of non-enhancing margin, n (%) | χ² = 0.03 | .86 | |||
| Well-defined | 97 (51.05) | 53 (50.48) | 44 (51.76) | ||
| Poorly-defined | 93 (48.95) | 52 (49.52) | 41 (48.24) | ||
| f14-Proportion of edema, n (%) | – | .61 | |||
| None | 7 (3.68) | 5 (4.76) | 2 (2.35) | ||
| <5% | 37 (19.47) | 23 (21.90) | 14 (16.47) | ||
| 6-33% | 71 (37.37) | 40 (38.10) | 31 (36.47) | ||
| 34-67% | 72 (37.89) | 36 (34.29) | 36 (42.35) | ||
| 68-95% | 3 (1.58) | 1 (0.95) | 2 (2.35) | ||
| f15-Edema across the midline, n (%) | χ² = 0.89 | .35 | |||
| No | 149 (78.42) | 85 (80.95) | 64 (75.29) | ||
| Yes | 41 (21.58) | 20 (19.05) | 21 (24.71) | ||
| f16-Hemorrhage, n (%) | χ² = 0.52 | .47 | |||
| No | 139 (73.16) | 79 (75.24) | 60 (70.59) | ||
| Yes | 51 (26.84) | 26 (24.76) | 25 (29.41) | ||
| f17-Diffusion characteristics, n (%) | – | .03 | |||
| No image | 54 (28.42) | 32 (30.48) | 22 (25.88) | ||
| Facilitated | 65 (34.21) | 40 (38.10) | 25 (29.41) | ||
| Restricted | 62 (32.63) | 32 (30.48) | 30 (35.29) | ||
| Mixed | 9 (4.74) | 1 (0.95) | 8 (9.41) | ||
| f18-Pial invasion, n (%) | χ²=0.63 | .43 | |||
| No | 82 (43.16) | 48 (45.71) | 34 (40.00) | ||
| Yes | 108 (56.84) | 57 (54.29) | 51 (60.00) | ||
| f19-Ependymal extension, n (%) | χ² = 0.19 | .67 | |||
| No | 75 (39.47) | 40 (38.10) | 35 (41.18) | ||
| Yes | 115 (60.53) | 65 (61.90) | 50 (58.82) | ||
| f20-Cortical involvement, n (%) | χ² = 0.88 | .35 | |||
| No | 37 (19.47) | 23 (21.90) | 14 (16.47) | ||
| Yes | 153 (80.53) | 82 (78.10) | 71 (83.53) | ||
| f21-Deep white matter invasion, n (%) | χ² = 0.00 | .98 | |||
| No | 49 (25.79) | 27 (25.71) | 22 (25.88) | ||
| Yes | 141 (74.21) | 78 (74.29) | 63 (74.12) | ||
| f22-nCET crosses midline, n (%) | χ² = 4.07 | .04 | |||
| No | 157 (82.63) | 92 (87.62) | 65 (76.47) | ||
| Yes | 33 (17.37) | 13 (12.38) | 20 (23.53) | ||
| f23-CET crosses midline, n (%) | χ² = 1.49 | .22 | |||
| No | 163 (85.79) | 93 (88.57) | 70 (82.35) | ||
| Yes | 27 (14.21) | 12 (11.43) | 15 (17.65) | ||
| f24-Satellites, n (%) | χ² = 0.09 | .77 | |||
| No | 145 (76.32) | 81 (77.14) | 64 (75.29) | ||
| Yes | 45 (23.68) | 24 (22.86) | 21 (24.71) | ||
| f25-Calvarial remodeling, n (%) | χ² = 0.11 | .74 | |||
| No | 181 (95.26) | 101 (96.19) | 80 (94.12) | ||
| Yes | 9 (4.74) | 4 (3.81) | 5 (5.88) | ||
| f29-transverse diameter, Mean ± SD | 44.78 ± 14.13 | 44.82 ± 14.41 | 44.73 ± 13.87 | t = 0.04 | .97 |
| f30-longitudinal diameter, Mean ± SD | 40.44 ± 13.19 | 40.85 ± 13.46 | 39.95 ± 12.93 | t = 0.47 | .64 |
Abbreviations: VASARI, visually accessible rembrandt images; TLSs, tertiary lymphoid structures; CET, contrast enhancing tumor; nCET, non-contrast enhancing tumor; t, t-test; χ2, Chi-square test; –, fisher exact. Bold values (P < 0.05) indicate that the corresponding factors show statistically significant differences in distinguishing TLSs+ and TLSs– groups.
Table 2.
Results of univariate and multivariate logistic regression for predicting TLSs+/TLSs– status
| Variables | Univariate |
Multivariate |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | SE | Z | P | OR (95% CI) | β | SE | Z | P | OR (95% CI) | |
| f2-Side of lesion center | ||||||||||
| Right | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Left | −0.90 | 0.31 | −2.87 | .004 | 0.41 (0.22 ∼ 0.75) | −0.86 | 0.34 | −2.51 | .012 | 0.43 (0.22 ∼ 0.83) |
| Center/Bilateral | −0.21 | 0.54 | −0.38 | .703 | 0.81 (0.28 ∼ 2.36) | −0.74 | 0.71 | −1.04 | .298 | 0.48 (0.12 ∼ 1.92) |
| f8-Cysts | ||||||||||
| No | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Yes | 0.59 | 0.30 | 1.96 | .050 | 1.80 (1.01 ∼ 3.23) | 0.49 | 0.32 | 1.52 | .129 | 1.64 (0.87 ∼ 3.09) |
| f17-Diffusion characteristics | ||||||||||
| No mage | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Facilitated | −0.10 | 0.38 | −0.25 | .800 | 0.91 (0.43 ∼ 1.90) | −0.10 | 0.40 | −0.26 | .794 | 0.90 (0.41 ∼ 1.98) |
| Restricted | 0.31 | 0.38 | 0.83 | .409 | 1.36 (0.65 ∼ 2.85) | 0.18 | 0.39 | 0.46 | .644 | 1.20 (0.56 ∼ 2.58) |
| Mixed | 2.45 | 1.10 | 2.24 | .025 | 11.64 (1.36 ∼ 99.75) | 2.13 | 1.11 | 1.91 | .056 | 8.39 (0.94 ∼ 74.46) |
| f22-nCET crosses midline | ||||||||||
| No | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Yes | 0.78 | 0.39 | 1.99 | .047 | 2.18 (1.01 ∼ 4.69) | 1.21 | 0.50 | 2.42 | .016 | 3.35 (1.26 ∼ 8.91) |
Abbreviations: TLSs, tertiary lymphoid structures; nCET, non-contrast enhancing tumor; OR, odds ratio; CI, confidence interval.
Analysis of Differences in VASARI Features and Logistic Regression Modeling Between mTLSs+ and imTLSs+ Groups
Significant differences were observed between the mTLSs and imTLSs groups in the following VASARI features: f6-Proportion nCET (χ2 = 9.76; P = .02), f16- hemorrhage (χ2 = 4.00; P = .04), and f18-Pial invasion (χ2 = 4.57; P = .03) (Table 3). Multivariate logistic regression analysis of F6-proportion nCET showed that it was an independent risk factor for predicting mTLSs+/imTLSs+ in patients with GBM (Table 4, forest plot as shown in Supplementary Figure S2). Figure 2 shows representative images of GBM and examples of TLSs stained with H&E and mIF.
Table 3.
Analysis of differences in VASARI features between mTLSs+, and imTLSs+ groups
| Feature name | Total (N = 85) | imTLSs+ (N = 47) | mTLSs+ (N = 38) | Statistic | P-value |
|---|---|---|---|---|---|
| f1-Tumor location, n (%) | – | .92 | |||
| Frontal | 33 (38.82) | 18 (38.30) | 15 (39.47) | ||
| Temporal | 27 (31.76) | 16 (34.04) | 11 (28.95) | ||
| Insular | 4 (4.71) | 2 (4.26) | 2 (5.26) | ||
| Parietal | 8 (9.41) | 5 (10.64) | 3 (7.89) | ||
| Occipital | 10 (11.76) | 4 (8.51) | 6 (15.79) | ||
| Cerebellum | 3 (3.53) | 2 (4.26) | 1 (2.63) | ||
| f2-Side of lesion center, n (%) | – | .65 | |||
| Right | 48 (56.47) | 27 (57.45) | 21 (55.26) | ||
| Left | 29 (34.12) | 17 (36.17) | 12 (31.58) | ||
| Center/Bilateral | 8 (9.41) | 3 (6.38) | 5 (13.16) | ||
| f3-Eloquent brain, n (%) | – | .84 | |||
| None | 11 (12.94) | 6 (12.77) | 5 (13.16) | ||
| Speech motor | 30 (35.29) | 17 (36.17) | 13 (34.21) | ||
| Speech receptive | 26 (30.59) | 16 (34.04) | 10 (26.32) | ||
| Motor | 8 (9.41) | 4 (8.51) | 4 (10.53) | ||
| Vision | 10 (11.76) | 4 (8.51) | 6 (15.79) | ||
| f4-Enhancement quality, n (%) | χ² = 0.48 | .49 | |||
| None | 6 (7.06) | 2 (4.26) | 4 (10.53) | ||
| Minimal/Mild | 79 (92.94) | 45 (95.74) | 34 (89.47) | ||
| Marked/avid | |||||
| f5-Proportion enhancing, n (%) | – | .26 | |||
| <5% | 2 (2.35) | 0 (0.00) | 2 (5.26) | ||
| 6-33% | 25 (29.41) | 14 (29.79) | 11 (28.95) | ||
| 34-67% | 32 (37.65) | 19 (40.43) | 13 (34.21) | ||
| 68-95% | 21 (24.71) | 13 (27.66) | 8 (21.05) | ||
| >95% | 5 (5.88) | 1 (2.13) | 4 (10.53) | ||
| f6-Proportion nCET, n (%) | χ² = 9.76 | .02 | |||
| <5% | 14 (16.47) | 3 (6.38) | 11 (28.95) | ||
| 6-33% | 17 (20.00) | 13 (27.66) | 4 (10.53) | ||
| 34-67% | 41 (48.24) | 24 (51.06) | 17 (44.74) | ||
| 68-95% | 13 (15.29) | 7 (14.89) | 6 (15.79) | ||
| f7-Proportion of necrosis, n (%) | – | .22 | |||
| None | 3 (3.53) | 1 (2.13) | 2 (5.26) | ||
| <5% | 15 (17.65) | 5 (10.64) | 10 (26.32) | ||
| 6-33% | 34 (40.00) | 19 (40.43) | 15 (39.47) | ||
| 34-67% | 28 (32.94) | 18 (38.30) | 10 (26.32) | ||
| 68-95% | 5 (5.88) | 4 (8.51) | 1 (2.63) | ||
| f8-Cysts | χ² = 0.07 | .79 | |||
| No | 30 (35.29) | 16 (34.04) | 14 (36.84) | ||
| Yes | 55 (64.71) | 31 (65.96) | 24 (63.16) | ||
| f9-Multifocal or multicentric, n (%) | – | .10 | |||
| Indeterminate | 71 (83.53) | 42 (89.36) | 29 (76.32) | ||
| Multifocal | 12 (14.12) | 4 (8.51) | 8 (21.05) | ||
| Multicentric | 1 (1.18) | 1 (2.13) | 0 (0.00) | ||
| Gliomatosis | 1 (1.18) | 0 (0.00) | 1 (2.63) | ||
| f10-T1/FLAIR ratio, n (%) | χ² = 0.84 | .66 | |||
| Expansive | 38 (44.71) | 20 (42.55) | 18 (47.37) | ||
| Mixed | 25 (29.41) | 13 (27.66) | 12 (31.58) | ||
| Infiltrative | 22 (25.88) | 14 (29.79) | 8 (21.05) | ||
| f11-Thickness of enhancing margin, n (%) | χ² = 3.40 | .18 | |||
| Indeterminate | 21 (24.71) | 11 (23.40) | 10 (26.32) | ||
| Thin | 24 (28.24) | 17 (36.17) | 7 (18.42) | ||
| Thick/solid | 40 (47.06) | 19 (40.43) | 21 (55.26) | ||
| Solid | χ² = 2.29 | .13 | |||
| f12-Definition of enhancing margin, n (%) | 65 (76.47) | 33 (70.21) | 32 (84.21) | ||
| Well-defined | 20 (23.53) | 14 (29.79) | 6 (15.79) | ||
| Poorly-defined | χ² = 0.84 | .66 | |||
| f13-Definition of non-enhancing margin, n (%) | χ² = 0.02 | .89 | |||
| Well-defined | 44 (51.76) | 24 (51.06) | 20 (52.63) | ||
| Poorly-defined | 41 (48.24) | 23 (48.94) | 18 (47.37) | ||
| f14-Proportion of edema, n (%) | – | .26 | |||
| None | 2 (2.35) | 0 (0.00) | 2 (5.26) | ||
| <5% | 14 (16.47) | 7 (14.89) | 7 (18.42) | ||
| 6-33% | 31 (36.47) | 19 (40.43) | 12 (31.58) | ||
| 34-67% | 36 (42.35) | 21 (44.68) | 15 (39.47) | ||
| 68-95% | 2 (2.35) | 0 (0.00) | 2 (5.26) | ||
| f15-Edema across the midline, n (%) | χ² = 0.66 | .41 | |||
| No | 64 (75.29) | 37 (78.72) | 27 (71.05) | ||
| Yes | 21 (24.71) | 10 (21.28) | 11 (28.95) | ||
| f16-Hemorrhage, n (%) | χ² = 4.00 | .04 | |||
| No | 60 (70.59) | 29 (61.70) | 31 (81.58) | ||
| Yes | 25 (29.41) | 18 (38.30) | 7 (18.42) | ||
| f17-Diffusion characteristics, n (%) | – | .15 | |||
| No image | 22 (25.88) | 9 (19.15) | 13 (34.21) | ||
| Facilitated | 25 (29.41) | 14 (29.79) | 11 (28.95) | ||
| Restricted | 30 (35.29) | 17 (36.17) | 13 (34.21) | ||
| Mixed | 8 (9.41) | 7 (14.89) | 1 (2.63) | ||
| f18-Pial Invasion, n (%) | χ² = 4.57 | .03 | |||
| No | 34 (40.00) | 14 (29.79) | 20 (52.63) | ||
| Yes | 51 (60.00) | 33 (70.21) | 18 (47.37) | ||
| f19-Ependymal extension, n (%) | χ² = 0.53 | .47 | |||
| No | 35 (41.18) | 21 (44.68) | 14 (36.84) | ||
| Yes | 50 (58.82) | 26 (55.32) | 24 (63.16) | ||
| f20-Cortical involvement, n (%) | χ² = 2.60 | .11 | |||
| No | 14 (16.47) | 5 (10.64) | 9 (23.68) | ||
| Yes | 71 (83.53) | 42 (89.36) | 29 (76.32) | ||
| f21-Deep white matter invasion, n (%) | χ² = 0.01 | .93 | |||
| No | 22 (25.88) | 12 (25.53) | 10 (26.32) | ||
| Yes | 63 (74.12) | 35 (74.47) | 28 (73.68) | ||
| f22-nCET crosses midline, n (%) | χ² = 2.47 | .12 | |||
| No | 65 (76.47) | 39 (82.98) | 26 (68.42) | ||
| Yes | 20 (23.53) | 8 (17.02) | 12 (31.58) | ||
| f23-CET crosses midline, n (%) | χ² = 0.55 | .46 | |||
| No | 70 (82.35) | 40 (85.11) | 30 (78.95) | ||
| Yes | 15 (17.65) | 7 (14.89) | 8 (21.05) | ||
| f24-Satellites, n (%) | χ² = 0.04 | .84 | |||
| No | 64 (75.29) | 35 (74.47) | 29 (76.32) | ||
| Yes | 21 (24.71) | 12 (25.53) | 9 (23.68) | ||
| f25-Calvarial remodeling, n (%) | χ² = 0.46 | .50 | |||
| No | 80 (94.12) | 43 (91.49) | 37 (97.37) | ||
| Yes | 5 (5.88) | 4 (8.51) | 1 (2.63) | ||
| f29-transverse diameter, Mean ± SD | 44.73 ± 13.87 | 45.40 ± 14.84 | 43.89 ± 12.71 | t = 0.50 | .62 |
| f30-longitudinal diameter, Mean ± SD | 39.95 ± 12.93 | 41.93 ± 12.89 | 37.49 ± 12.71 | t = 1.59 | .12 |
Abbreviations: VASARI, Visually Accessible Rembrandt Images; mTLSs, mature tertiary lymphoid structures; imTLSs, immature tertiary lymphoid structures; CET, contrast enhancing tumor; nCET, non-contrast enhancing tumor; t, t-test; χ2, Chi-square test; –, fisher exact. Bold values (P < 0.05) indicate that the corresponding factors show statistically significant differences in distinguishing imTLSs+ and mTLSs+ groups.
Table 4.
Results of univariate and multivariate logistic regression for predicting mTLSs+/imTLSs+ status
| Variables | Univariate |
Multivariate |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | SE | Z | P | OR (95% CI) | β | SE | Z | P | OR (95% CI) | |
| f6-Proportion nCET | ||||||||||
| <5% | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| 6-33% | –2.48 | 0.87 | –2.86 | .004 | 0.08 (0.02 ∼ 0.46) | –2.54 | 0.91 | –2.79 | .005 | 0.08 (0.01 ∼ 0.47) |
| 34-67% | –1.64 | 0.72 | –2.27 | .023 | 0.19 (0.05 ∼ 0.80) | –1.63 | 0.76 | –2.15 | .031 | 0.20 (0.04 ∼ 0.86) |
| 68-95% | –1.45 | 0.86 | –1.70 | .090 | 0.23 (0.04 ∼ 1.25) | –1.80 | 0.91 | –1.98 | .048 | 0.17 (0.03 ∼ 0.98) |
| f16-Hemorrhage | ||||||||||
| No | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Yes | –1.01 | 0.51 | –1.96 | .050 | 0.36 (0.13 ∼ 0.99) | –1.03 | 0.58 | –1.77 | .076 | 0.36 (0.11 ∼ 1.11) |
| f18-Pial Invasion | ||||||||||
| No | 1.00 (Reference) | 1.00 (Reference) | ||||||||
| Yes | –0.96 | 0.46 | –2.11 | .034 | 0.38 (0.16 ∼ 0.93) | –0.64 | 0.51 | –1.24 | .213 | 0.53 (0.19 ∼ 1.44) |
Abbreviations: mTLSs, mature tertiary lymphoid structures; imTLSs, immature tertiary lymphoid structures; nCET, non-contrast enhancing tumor; OR, odds ratio; CI, confidence interval.
Figure 2.
Imaging, H&E staining, and mIF staining of TLSs at different stages in patients with glioblastoma.
Discussion
This study identified and quantitatively analyzed the expression and maturity of TLSs in tumor tissues from patients with GBM. We investigated survival differences between TLSs-positive (TLSs+) and TLSs-negative (TLSs−) groups, as well as between mature (mTLSs) and immature (imTLSs) TLSs subgroups. Our findings demonstrate that approximately 44.74% of GBM tissues contain TLSs at varying stages of maturity. Critically, survival analysis revealed that patients with TLS-positive status or mature TLSs had significantly longer survival times, indicating that TLSs serve as favorable prognostic biomarkers in GBM.
The prognostic significance of TLSs observed in our GBM cohort is strongly supported by a growing body of evidence across various solid tumors. For instance, in intrahepatic cholangiocarcinoma, the presence of intratumoral TLSs is an established positive prognostic factor, and a CT-based radiomics model predicting TLS status can effectively stratify patients’ recurrence-free survival.16 Similarly, in breast cancer, the presence of TLSs is significantly associated with improved disease-free survival and distant metastasis-free survival.13 Perhaps the most compelling evidence comes from a large-scale multicenter study on gastrointestinal cancers, which developed a machine learning model to quantify TLSs on routine H&E-stained images; this study confirmed the independent prognostic value of a quantitative TLS score across six cancer types and demonstrated that patients with high TLS scores had a significantly lower risk of death compared to those with low scores or no TLSs (Hazard Ratio [HR] 0.27).17 Our findings in GBM therefore align with this established oncological principle, reinforcing the role of TLSs as a robust and favorable prognostic biomarker across diverse malignancies.
Currently, the identification, quantification, and visualization of TLSs rely on histopathological techniques such as H&E staining, IHC, and mIF applied to postoperative tumor tissues. These methods primarily reveal the cellular composition of TLSs by identifying specific immune cell types and visualizing their spatial distribution.18 In our evaluation, the positive rate of TLSs based on H&E whole-slide images (WSIs) was 44.74%. And, we classified TLSs+ according to established immunohistological criteria by mIF.19 Briefly, immature TLSs (imTLS) lack a germinal center structure (CD23−) and consist primarily of aggregates of T cells (CD3+) and B cells (CD20+). In contrast, mature TLS (mTLS) are highly organized structures featuring a distinct germinal center (CD23+).
Although these pathological methods are accurate for identifying, quantifying, and visualizing TLSs, their drawbacks, including high cost, operational complexity, and time-consuming procedures, outweigh the benefits. Given the practical limitations of pathological assessment for TLSs detection, the development of non-invasive predictive methods using routinely available preoperative MRI is of considerable clinical interest. Notably, VASARI is a standardized feature set for characterizing MR images of human GBMs. Specific VASARI features show strong potential for predicting OS in patients with glioma.20 Rao et al.1 reported that a combination of three features, namely volume, hemorrhage, and T1/FLAIR ratio, was capable of stratifying survival in patients with GBM. Our team previously utilized VASARI features to predict the infiltration level of CD8+ T cells in GBM s. The most predictive features were f8-Cysts, f16-Hemorrhage, and f19-Ependymal invasion. These findings demonstrate that VASARI-based characteristics can non-invasively predict CD8 expression levels in GBM.22 Additionally, another study by our team explored the value of VASARI features in preoperatively predicting PD-L1 expression levels. The results revealed that f5-Proportion Enhancing differed significantly between the groups. Logistic regression analysis further demonstrated that f5-Proportion Enhancing is an independent risk factor for predicting PD-L1 expression in patients with GBM.23 VASARI features provide novel imaging biomarkers for assessing the tumor immune microenvironment. Therefore, in situations where advanced methods and functional imaging techniques are difficult to apply in routine clinical practice, our approach of preoperatively predicting TLS’s status based on VASARI features is easy to analyze and can be seamlessly integrated into daily clinical workflows.
Our findings indicate significant differences in several imaging features between the TLSs+ and TLSs− groups, including side of the lesion center, cysts, diffusion characteristics, and nCET crossing the midline. However, it is noteworthy that while cysts showed significance in univariate analysis, they were not retained as an independent predictor in the multivariate logistic regression model. Logistic regression analysis ultimately demonstrated that only side of the lesion center and non-contrast-enhancing tumor crossing the midline were independent predictors of TLSs+/TLSs− status. These results suggest that the formation of TLSs is not random but is closely associated with specific macroenvironmental characteristics of the tumor. For instance, the imaging feature “nCET crosses midline” is typically associated with more aggressive tumor biological behavior. However, this study revealed a negative correlation between nCET crosses midline and the presence of TLSs, suggesting that the anti-tumor immune response represented by TLSs may partially restrain invasive tumor expansion. Similarly, the differences observed in diffusion characteristics may reflect structural alterations in the tumor microenvironment caused by dense lymphocyte infiltration in TLS regions. This aligns with previous reports indicating that immune cell infiltration influences apparent diffusion coefficient values derived from imaging.24
Studies in other solid tumors have suggested that a TLS-positive status is associated with marked tumor enhancement. For instance, research by Xu et al.5 in intrahepatic cholangiocarcinoma demonstrated that diffuse hyperenhancement in the arterial phase was an independent predictor of TLS status (OR[95% CI]:3.082[1.987-4.178]). The underlying mechanism may be that the formation and maintenance of TLSs require oxygen and nutrients, which stimulates local neovascularization. Consequently, after contrast agent injection during MRI, tumors exhibit pronounced enhancement. Furthermore, the aggregation of immune cells necessitates adequate blood flow support. Activated immune cells release various inflammatory mediators, leading to local vasodilation, increased blood flow, and enhanced delivery of contrast agent.26 However, in our study, no significant difference was observed in the Enhancement Quality (f4) feature between the TLSs+ and TLSs− groups (P = .06). We postulate that this finding may be attributable to the characteristic intense enhancement widely prevalent in GBM. As the vast majority of GBM cases present with marked enhancement on imaging, the sample size disparity between the “enhancing” and “non-enhancing” subgroups is substantial. This imbalance likely diminishes the specificity of conventional enhancement intensity for predicting TLS status in GBM. Therefore, relying solely on standard enhancement characteristics may be insufficient to effectively distinguish the presence or absence of TLSs in this specific tumor type.
This study further revealed that TLSs at different maturation stages exhibited distinct radiological phenotypes. Significant differences were observed between mTLSs and imTLSs in features such as f6-proportion of nCET, f16-hemorrhage, and f18-pial invasion. proportion of nCET was confirmed to be an independent predictor of TLS’s maturity. A lower proportion of nCET may be associated with more efficient anti-tumor immunity in mTLSs, leading to increased tumor cell destruction and expansion of the central necrotic area, thereby reducing the non-enhancing tumor component. Hemorrhage is typically associated with abnormal tumor angiogenesis and high expression of vascular endothelial growth factor, which can suppress immune cell function and impede the maturation of TLSs. Consequently, the lower incidence of hemorrhage observed in the mTLSs group may reflect a vascular microenvironment that is more favorable for immune responses, potentially supporting the development and maintenance of mature TLSs. Furthermore, the absence of pial invasion was associated with mTLSs. This suggests that a robust anti-tumor immune response may effectively restrict tumor dissemination along the meningeal spaces, which aligns with the improved survival outcomes observed in patients with mTLSs in this study.
These distinct radiological phenotypes not only illuminate TLS biology but also reveal promising therapeutic avenues. The survival benefit linked to mTLS underscores the value of inducing TLS maturation in GBM.27 Preclinical studies highlight several strategies, such as activating the lymphotoxin-beta receptor (LTβR) pathway or administering agonistic CD40 antibody (αCD40), both of which promote TLS formation and improve anti-tumor responses.8,28 Notably, αCD40 acts through B cells, enhancing lymphotoxin-alpha (Lta) expression.8 However, therapeutic TLS induction is not without challenges. While αCD40 triggers TLS development, it may also expand immunosuppressive CD11b+ B cells and impair T-cell function, thereby blunting checkpoint inhibitor efficacy.8 Thus, merely inducing TLS is inadequate; the goal should be generating functional mTLS that sustain antitumor immunity without inducing immunosuppression. Future efforts should therefore prioritize combination approaches that not only initiate TLS assembly—via localized delivery of chemokines like CXCL13, CCL19, or CCL219,28—but also support their functional maturation, potentially by co-targeting regulatory T cells or inhibitory B-cell subsets in the tumor microenvironment. A deeper understanding of the imTLS-to-mTLS transition will be essential to guide such targeted combination therapies in GBM.
This study had certain limitations. First, this was a single-center, retrospective investigation. Second, the patient cohort size was relatively small, which may have limited the robustness of the findings. Third, as the study focused on radiological-pathological correlations, preoperative laboratory values reflecting systemic immune status were not included; these could potentially interact with the local tumor immune microenvironment and represent confounding factors. However, the strengths of this research include the identification and quantitative analysis of TLSs expression, maturity, and spatial distribution in GBM, demonstrating that TLSs are a favorable prognostic factor; the correlation analysis between TLSs and VASARI imaging features; and the utilization of methods that are feasible for routine clinical practice. Future improvements should focus on expanding the cohort size to validate the results and enhancing the stability of imaging biomarkers for predicting TLS expression, thereby further elucidating the intrinsic relationship between imaging features and microenvironment characteristics. Furthermore, future prospective studies that integrate multiparametric MRI, deep immunophenotyping, and concurrent peripheral blood immunomonitoring are warranted to provide a more comprehensive understanding of the anti-tumor immune response in GBM.
This study demonstrated that the presence and maturity of TLSs in GBM are significantly associated with improved patient survival, confirming the potential value of TLSs as a positive prognostic biomarker in GBM. Furthermore, this study systematically revealed for the first time that TLS expression status and maturity are significantly correlated with VASARI imaging features and identified specific imaging indicators capable of independently predicting TLSs. Consequently, a prediction model based on preoperative conventional MRI features holds promise for the non-invasive assessment of TLS status before surgery, providing valuable imaging-based insights for personalized risk stratification and the development of immunotherapy strategies in patients with GBM.
Supplementary Material
Contributor Information
Qing Zhou, Department of Radiology, Lanzhou University Second Hospital, Lanzhou, Gansu, China; Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence, Lanzhou, Gansu, China.
Peng Zhang, Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Department of Pathology, Lanzhou University Second Hospital, Lanzhou, Gansu, China.
Caiqiang Xue, Department of Radiology, The Affiliated Hospital of Qingdao University, Qingdao, Shandong, China.
Bin Zhang, Department of Radiology, Lanzhou University Second Hospital, Lanzhou, Gansu, China; Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence, Lanzhou, Gansu, China.
Xiaoai Ke, Department of Radiology, Lanzhou University Second Hospital, Lanzhou, Gansu, China; Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence, Lanzhou, Gansu, China.
Jiangwei Man, Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Department of Surgical, Lanzhou University Second Hospital, Lanzhou, Gansu, China.
Jia Kang, Department of Radiology, Lanzhou University Second Hospital, Lanzhou, Gansu, China; Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence, Lanzhou, Gansu, China.
Zhiyong Zhao, Department of Surgical, Lanzhou University Second Hospital, Lanzhou, Gansu, China.
Junlin Zhou, Department of Radiology, Lanzhou University Second Hospital, Lanzhou, Gansu, China; Second Clinical School, Lanzhou University, Lanzhou, Gansu, China; Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence, Lanzhou, Gansu, China.
Supplementary Material
Supplementary material is available online at Neuro-Oncology Advances (https://academic.oup.com/noa).
Author Contributions
Q.Z., P.Z., C.Q.X., B.Z., and J.L.Z. conceived, designed and supervised the study; Q.Z., P.Z., C.Q.X., B.Z., X.A.K, J.W.M., J.K., and Z.Y.Z. collected and assembled the data; Q.Z. and P.Z. were responsible for data analysis and interpretation; and Q.Z. wrote the manuscript. All authors approved the manuscript.
Conflict of Interest Statement
The authors declare that they have no competing financial interests or personal relationships that could have influenced the work reported in this study.
Funding
This study is supported by grants from the National Natural Science Foundation of China (grant number: 82371914), the Cuiying Science and Technology Innovation Program of Lanzhou University Second Hospital (grant number: CY2023-YB-A03), the Science and Technology Program of Lanzhou City (grant number: 2024-9-122), and the Science and Technology Program of Gansu Province(grant number: 24JRRA376).
Data Availability
The datasets generated or analyzed during this study are available from the corresponding author upon reasonable request.
References
- 1. Asija S, Chatterjee A, Yadav S, et al. Combinatorial approaches to effective therapy in glioblastoma (GBM): current status and what the future holds. Int Rev Immunol. 2022;41:582-605. [DOI] [PubMed] [Google Scholar]
- 2. Price M, Ballard C, Benedetti J, et al. CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States between 2017-2021. Neuro Oncol. 2024;26:vi1-vi85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Vaccaro A, de Alves PB, van de Walle T, Dimberg A. Tertiary lymphoid structures in central nervous system disorders. Methods Mol Biol. 2025;2864:21-42. [DOI] [PubMed] [Google Scholar]
- 4. Sarti KG, Da SMT. Tertiary lymphoid structures ares the hubs of antitumour immunity. Nat Rev Cancer. 2023;23:803. [DOI] [PubMed] [Google Scholar]
- 5. Schumacher TN, Thommen DS. Tertiary lymphoid structures in cancer. Science. 2022;375:eabf9419. [DOI] [PubMed] [Google Scholar]
- 6. Lin WP, Li H, Sun ZJ. T cell exhaustion initiates tertiary lymphoid structures and the turbochargeds cancer-immuneity cycle. EBioMedicine. 2024;104:105154. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Cui X, Gu X, Li D, et al. Tertiary lymphoid structures as a biomarkers in immunotherapy and beyond: advancing towards clinical applications. Cancer Lett. 2025;613:217491. [DOI] [PubMed] [Google Scholar]
- 8. van Hooren L, Vaccaro A, Ramachandran M, et al. Agonistic CD40 therapy induces tertiary lymphoid structures, but impairs responses to checkpoint blockade in gliomas. Nat Commun. 2021;12:4127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Yang C, Cai YX, Wang ZF, Tian SF, Li ZQ. Tertiary lymphoid structures in the central nervous system. Trends Mol Med. 2025;31:509-521. [DOI] [PubMed] [Google Scholar]
- 10. Ramachandran M, Vaccaro A, van de Walle T, et al. Tailoring the vascular phenotype through AAV therapy promotes anti-tumor immunity in gliomas. Cancer Cell. 2023;41:1134-1151.e10. [DOI] [PubMed] [Google Scholar]
- 11. Chen J, Yang Y, Luan S, Xu W, Gao Y. Tertiary lymphoid structures in gliomas: impact on tumour immunity and progression. J Transl Med. 2025;23:528. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Li P, Liang Y, Zeng B, et al. Preoperative prediction of intra-tumoral tertiary lymphoid structures based on CT in hepatocellular cancer using computed tomography. Eur J Radiol. 2022;151:110309. [DOI] [PubMed] [Google Scholar]
- 13. Li K, Ji J, Li S, et al. Analysis of the correlation and prognostic significance of tertiary lymphoid structures in breast cancer: a radiomics-clinical integration approach. J Magn Reson Imaging. 2024;59:1206-1217. [DOI] [PubMed] [Google Scholar]
- 14. Sacli-Bilmez B, Firat Z, Topcuoglu OM, Yaltirik K, Ture U, Ozturk-Isik E. Identifying overall survival in 98 glioblastomas using VASARI features at 3T. Clin Imaging. 2023;93:86-92. [DOI] [PubMed] [Google Scholar]
- 15. Han Y, Wang YY, Yang Y, et al. Association between dichotomized VASARI features and overall survival in patients with glioblastoma patients: a single-institution propensity score matching analysis. Cancer Imaging. 2024;24:109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Xu Y, Li Z, Yang Y, et al. A CT-based radiomics approach to predict intra-tumoral tertiary lymphoid structures and recurrence of intrahepatic cholangiocarcinoma. Insights Imaging. 2023;14:173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Li Z, Jiang Y, Li B, et al. Development and validation of a machine learning model for detection and classification of tertiary lymphoid structures in gastrointestinal cancers. JAMA Netw Open. 2023;6:e2252553. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Klein C, Devi-Marulkar P, Dieu-Nosjean MC, et al. Advancement of techniques for precise visualization and quantification of tertiary lymphoid structure-associated immune cells in tissue samples. Methods Mol Biol. 2025;2864:181-203. [DOI] [PubMed] [Google Scholar]
- 19. Vanhersecke L, Bougouin A, Crombé A, et al. Standardized pathology screening of mature tertiary lymphoid structures in cancers. Lab Invest. 2023. May;103:100063. [DOI] [PubMed] [Google Scholar]
- 20. Wegscheid ML, Jennings JW. VASARI glioma features: insights into their impact and performance. Radiol Imaging Cancer. 2024;6:e249018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Rao A, Rao G, Gutman DA, et al. A combinatorial radiographic phenotype may stratify patient survival and be associated with the invasiveon and proliferativeon characteristics ofin glioblastomas. J Neurosurg. 2016;124:1008-1017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Xue C, Zhou Q, Zhang B, et al. Vasaria-based features nomogram to predict the tumor-infiltrating CD8+ T cell levels in glioblastoma. Acad Radiol. 2024;31:2050-2060. [DOI] [PubMed] [Google Scholar]
- 23. Zhang B, Zhou Q, Zhang P, et al. MRI histogram and vVisually AcceSAble rembrandt iImages features a nomogram to predict PD-L1 levels in glioblastoma. Neurosurg Rev. 2025;48:371. [DOI] [PubMed] [Google Scholar]
- 24. Zhou Q, Xue C, Man J, et al. Correlation of tumor-associated macrophage infiltration in glioblastoma with magnetic resonance imaging characteristics: a retrospective cross-sectional study. Quant Imaging Med Surg. 2023;13:5958-5973. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Xu Y, Li Z, Yang Y, et al. Association between MRI radiomics and intratumoral tertiary lymphoid structures in intrahepatic cholangiocarcinoma and its prognostic significance. J Magn Reson Imaging. 2024; Aug; 60:715-728. [DOI] [PubMed] [Google Scholar]
- 26. Zhang Y, Xu M, Ren Y, et al. Tertiary lymphoid structural heterogeneity determines tumour immunity and prospects for clinical application. Mol Cancer. 2024; Apr 6; 23:75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Zhou X, Li W, Yang J, et al. Tertiary lymphoid structure stratifies glioma into three distinct tumor subtypes. Aging (Albany NY). 2021; Dec 26; 13:26063-26094. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. van de Walle T, Vaccaro A, Ramachandran M, et al. Tertiary lymphoid structures in the Central nervous system: implications for glioblastoma. Front Immunol. 2021; Sep 1; 12:724739. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated or analyzed during this study are available from the corresponding author upon reasonable request.