Advanced Age in Humans and Mouse Models of Glioblastoma Show Decreased Survival from Extratumoral Influence

Abstract

Purpose:

Glioblastoma (GBM) is the most common aggressive primary malignant brain tumor in adults with a median age of onset of 68 to 70 years old. Although advanced age is often associated with poorer GBM patient survival, the predominant source(s) of maladaptive aging effects remains to be established. Here, we studied intratumoral and extratumoral relationships between adult patients with GBM and mice with brain tumors across the lifespan.

Experimental Design:

Electronic health records at Northwestern Medicine and the NCI SEER databases were evaluated for GBM patient age and overall survival. The commercial Tempus and Caris databases, as well as The Cancer Genome Atlas were profiled for gene expression, DNA methylation, and mutational changes with varying GBM patient age. In addition, gene expression analysis was performed on the extratumoral brain of younger and older adult mice with or without a brain tumor. The survival of young and old wild-type or transgenic (INK-ATTAC) mice with a brain tumor was evaluated after treatment with or without senolytics and/or immunotherapy.

Results:

Human patients with GBM ≥65 years of age had a significantly decreased survival compared with their younger counterparts. While the intra-GBM molecular profiles were similar between younger and older patients with GBM, non-tumor brain tissue had a significantly different gene expression profile between young and old mice with a brain tumor and the eradication of senescent cells improved immunotherapy-dependent survival of old but not young mice.

Conclusions:

This work suggests a potential benefit for combining senolytics with immunotherapy in older patients with GBM.

Translational Relevance.

Patients with glioblastoma (GBM) ≥65 years of age represent more than 70% of GBM diagnoses and have poorer survival outcomes compared with similarly treated younger counterparts. The biological mechanism(s) contributing to the significantly worse survival outcomes of older patients with GBM is largely unexplored. The objective of this work was to determine whether the primary source of maladaptive aging effects arise from within the GBM (i.e., intratumorally) or if those effects primarily originate from outside of the tumor (i.e., extratumorally). The collective clinical and preclinical observations suggest that the negative effects of aging unexpectedly arise from host tissue outside of the bulk tumor mass, and, that the effects of senescence are selectively maladaptive in older adults with a primary brain tumor. This work provides rationale for future clinical trials that combine senolytics with immunotherapy that may selectively benefit older adults with GBM.

Introduction

Advanced age is one of the most important risk factors for developing cancer (1), and in-turn, cancer is the leading cause of death for adults 60 to 79 years of age (2, 3). The NCI's Surveillance, Epidemiology, and End Results (SEER) database shows that while the overall cancer mortality rate declined between 1975 and 2015, the mortality rate for patients with primary brain and other central nervous system cancer increased (2). Glioblastoma [GBM; wild-type (WT) isocitrate dehydrogenase 1 or 2 (IDHwt)] is the most common malignant primary brain tumor in adults with a median age of onset between 68 to 70 years old (4). The incidence rate for GBM increases with age and is maximal among patients who are 75 to 84 years old (2, 5, 6). Older adult patients with GBM who are ≥65 years of age experience worse survival outcomes after standard-of-care and/or immunotherapeutic therapies as compared with similarly treated younger patients with GBM (2, 4, 5). The underlying biological determinants that contribute to treatment resistance and/or lack of therapeutic tolerability have not been comprehensively explored for older adults with GBM (6).

Established biomarkers of GBM include the favorable O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status, as well as EGFR amplification, PTEN, and TP53 mutations. Although previous studies have reported a different prevalence of tumor cell–specific biomarkers between younger and older patients with GBM, an established relationship between intratumoral biomarker prevalence and overall survival (OS) has yet to be identified (7–9). Interestingly, despite the increased HR among older adults with GBM, previous analyses suggested that intratumoral gene expression does not vary between GBM isolated from younger versus older patients (10). In contrast, expression for select factors including immunosuppressive indoleamine 2,3 dioxygenase 1 (IDO1) expression is increased in the older adult extratumoral brain as compared with the younger adult extratumoral brain (11, 12). Furthermore, the increased extratumoral IDO1 expression is functionally maladaptive because its systemic genetic deletion increases the survival of older adult mice with a brain tumor after treatment with immunotherapy (5). The data collectively support the hypothesis that the predominant source of maladaptive age-dependent factors are primarily derived from an extratumoral source rather than the bulk GBM itself.

Cellular senescence is a hallmark of biological aging and is induced by age-dependent events such as DNA damage or oncogenic signaling (13, 14). Senescent cells remain in a state of permanent cell-cycle growth arrest and modify their microenvironment through the secretion of interleukins, chemokines, growth factors, proteases, and extracellular matrix components. The full composition of those senescent cell–derived factors are referred to as the senescence-associated secretory phenotype (SASP; ref. 15). As senescent cells accumulate with normal aging or at sites of age-related disease, SASP factor levels increase and contribute to chronic inflammation (14, 16). This results in the compensatory enhancement of immune suppression and/or immunosenescence, the latter of which is characterized by increased T-cell exhaustion, a reduced capacity for clonal expansion, and a diminished capacity for mounting immune responses to new antigens (17–20). Increased levels of immunosuppression and immunosenescence in older adults may also represent key mechanisms of resistance that arise in response to standard-of-care and/or experimental therapies (6, 20). Although chronic inflammation plays a role in the initiation and progression of age-related disease, a causal link between senescence, disease progression, and age-dependent mechanisms of treatment resistance in patients with GBM remains to be established (14, 16, 17, 21, 22).

The unexpected discovery that intratumoral gene expression of IDHm and IDHwt grade IV glioma does not vary with host age (10) led to our study objective for validating and extending that initial observation in a cohort of IDHwt-exclusive GBM tumors based on World Health Organization's (WHO) reclassification in late 2021. Using two commercial platforms in addition to The Cancer Genome Atlas (TCGA), we compared tumors isolated from younger patients with GBM who were <65 years of age as well as older adult patients with GBM who were ≥65 years of age. The analysis of all three databases concordantly showed negligible differences for intratumoral gene expression and tumor mutational burden (TMB) when comparing resected GBM from younger and older patients. This supports the hypothesis that the biological determinants of unfavorable survival outcomes in older adults with GBM predominantly arise from outside of the bulk tumor mass. We therefore investigated whether potential age-related differences occur in the extratumoral non-bulk tumor brain parenchyma. Given the limited availability of pretreated extratumoral human brain tissue from patients with GBM, which requires autopsy, we instead used murine brain tumor models of varying ages. As aging has been associated with global mRNA changes in healthy murine brains (23, 24, 25), we performed a global RNA sequencing (RNA-seq) analysis indicating that the presence of a brain tumor causes dramatic changes to gene expression in the extratumoral older adult mouse brain as compared with the young mouse brain. Brain tumor progression, as well as brain tumor therapy with radiation and anti–programmed cell death protein 1 (PD-1) mAb both contributed to an increased level of senescence in the older adult extratumoral mouse brain. The eradication of systemic host p16^INK4A-expressing cells with senolytic treatments or through the inducible ablation of host p16^INK4A-expressing cells in transgenic INK-ATTAC mice led to long-term survivorship of older adult mice with a brain tumor after co-treatment with radiation, anti–PD-1 mAb, and an IDO1 enzyme inhibitor.

Materials and Methods

SEER database analysis

OS for patients with GBM was defined by the time from tissue diagnosis to date of death or last known time alive. Individualized demographic and clinical variables including age, survival status, survival time, and the cause of death from the SEER database were accessed through SEER*Stat (Version 8.3.9.2). Further statistical analyses were conducted in R version 4.0.2. Kaplan–Meier survival curves were estimated for different age groups and compared using the Log-rank test.

Northwestern Medicine Enterprise Data Warehouse database analysis

Medical records for patients diagnosed with GBM were accessed through the Northwestern Medicine Enterprise Data Warehouse (NMEDW). Data were de-identified before being released for the Institutional Review Board–approved study entitled, ‘Survival Analysis of Glioblastoma Patients Treated with Psychosocial Modifiers’ (#STU00214383). Accessed de-identified variables included the date of diagnosis, age at diagnosis, gender, date of last follow-up, date of death, IDH mutation status, and MGMT methylation status. OS was defined as the time from tissue diagnosis to death or last known time alive.

TCGA database analysis

Clinical and molecular data were obtained from TCGA manuscripts, Genomic Data Portal, and the Broad Institute (23, 26, 27). Analyses were adjusted for gender and limited to IDHwt cases (identified using clinical and molecular data), which resulted in the exclusion of 26 tumors. Associations between OS and progression-free survival among age groups were identified using survfit and coxph functions from the survival R package (25). Differentially expressed genes (DEG) were identified using limma after filtering for genes that had more than two counts per million in at least 5 samples. Differential analysis involved voom transformation of counts, model fitting, and empirical Bayes shrinkage (28). Differential methylation was similarly tested using limma with the exception that Beta values were transformed to M-values for the analysis (29). Epigenetic age was calculated using the online DNAm age calculator (https://dnamage.genetics.ucla.edu/home; ref. 30). This predicts age from the DNAm coefficients of 353 CpG sites. Estimations that correlated with the internal gold standard less than 0.8 were discarded, as recommended by the software, from downstream analyses. DNAm age acceleration was calculated by subtracting a subject's chronologic age from DNAm age. MC3 mutation calls were filtered for genes frequently mutated in public exomes and differential analyses employed Fisher exact tests. Nonstop, nonsense, splice site, and frame-shift mutations were collapsed for truncation. Heat maps for mutations were generated using maftools (31). Statistical analyses were performed on R (v 4.0.3). Unless otherwise mentioned, Wilcoxon Rank Sum test (as implemented in wilcox.test) was used to compare groups and statistical significance was considered as P < 0.05. Additional statistical details are denoted in the figure legends.

Commercial database analyses

For Caris (herein denoted as database A) and Tempus (herein denoted as database B) analyses, patient demographics and clinical characteristics, along with molecular and sequencing data were compared for patients <65 and ≥65 years of age at diagnosis by Pearson χ² tests/Fisher exact tests or Wilcoxon rank-sum tests, as appropriate. The goal was not to make direct comparisons between these platforms, but rather, to ensure that results were consistent across both clinical datasets. Tumors diagnosed as GBM, but were IDH 1/2 mutated, were excluded per WHO 2021 criteria because these are now classified as astrocytoma IDH mutant WHO grade 4 (26). The vast majority of patient samples were from pretreated tumor specimens at the time of diagnosis. Next-generation sequencing (NGS) metrics were compared between age groups employing the same appropriate tests with FDR correction for multiple testing. Analogously, for the subset of samples with RNA-seq results, databases A and B used transcripts per million (TPM) and log₁₀ of the normalized RNA read counts, respectively, for an a priori list of genes with false-discovery correction. All categorical fields were described as N (%) and continuous fields as median (interquartile range). Analyses were two-sided with statistical significance evaluated at the 0.05 alpha level or 0.05 q-value level in the event of false-discovery correction. Analyses were performed using R 4.0.4 and took place in October to November 2021, which included patient records sequenced from November 2017 to September 2021.

Analysis of the clinical database A

NGS was performed as previously described (28). Briefly, genomic DNA isolated from formalin-fixed, paraffin-embedded (FFPE) tumor samples using the NextSeq or NovaSeq 6000 platforms (Illumina, Inc.). For NextSeq sequenced tumors, a custom-designed SureSelect XT assay was used to enrich 592 whole-gene targets (Agilent Technologies). For NovaSeq sequenced tumors, a panel of more than 700 clinically relevant genes at high coverage and high read-depth was used along with another panel designed to enrich for an additional >20,000 genes at lower depth. All variants were detected with >99% confidence based on allele frequency and amplicon coverage, with an average sequencing depth of coverage of >500 and an analytic sensitivity of 5%. Prior to molecular testing, tumor enrichment was achieved by harvesting targeted tissue using manual microdissection techniques. Genetic variants identified were interpreted by board-certified molecular geneticists and categorized as ‘pathogenic’, ‘likely pathogenic’, ‘variant of unknown significance’, ‘likely benign’, or ‘benign’, according to the American College of Medical Genetics and Genomics standards. When assessing mutation frequencies of individual genes, ’pathogenic’ and ‘likely pathogenic’, variants were counted as mutations. The copy-number alteration of each exon is determined by calculating the average depth of the sample along with the sequencing depth of each exon and comparing the calculated result to a pre-calibrated value. TMB was measured by counting all non-synonymous missense, nonsense, in-frame insertions/deletions, and/or frameshift mutations found per tumor that had not been previously described as germline alterations in the dbSNP151 database, Genome Aggregation Database, or benign variants identified by Caris geneticists. A cut-off point of ≥10 mutations per MB as “high TMB” was used on the basis of the KEYNOTE-158 pembrolizumab trial (29). Detection of gene fusions was performed on mRNA isolated from a FFPE tumor sample using the Illumina NovaSeq platform (Illumina, Inc.) and Agilent SureSelect Human All Exon V7 bait panel (Agilent Technologies). FFPE specimens underwent pathology review to diagnose percent tumor content and tumor size. A minimum of 10% tumor content in the area for microdissection was required to enable enrichment and extraction of tumor-specific RNA. A Qiagen RNA FFPE tissue extraction kit was used, with RNA quality and quantity determined by the Agilent TapeStation. Biotinylated RNA baits were hybridized to the synthesized and purified cDNA targets and the bait-target complexes were amplified in a post-capture PCR reaction. The resultant libraries were quantified, normalized, denatured, diluted, and sequenced. The reference genome used was GRCh37/hg19 and analytical validation of this test demonstrated ≥97% positive percent agreement, ≥99% negative percent agreement, and ≥99% overall percent agreement with a validated comparator method. For gene expression, the whole transcriptome from patients was used to sequence to whole transcriptome to an average of 60M reads. Raw data was de-multiplexed by Illumina Dragen BioIT accelerator, trimmed, counted, PCR-duplicates removed, and aligned to human reference genome hg19 by STAR aligner. For transcript counting, TPM molecules was generated using the Salmon expression pipeline (32). Immune cell fraction was calculated by quanTIseq (31). A combination of multiple test platforms was used to determine the MSI or MMR status of the tumors profiled including fragment analysis (FA; Promega), IHC [MLH1, M1 antibody; MSH2, G2191129 antibody; MSH6, 44 antibody; and PMS2, EPR3947 antibody (Ventana Medical Systems, Inc.)], and NGS (2,800 target microsatellite loci were examined and compared with the reference genomes from the University of California, Santa Cruz). The three platforms generated highly concordant results as previously reported (30) and in the rare cases of discordant results, the MSI or MMR status of the tumor was determined in the order of IHC, FA, and NGS. MGMT promoter methylation was evaluated by pyrosequencing. DNA extraction from paraffin-embedded tumor samples was performed for subsequent pyrosequencer-based analysis of 5 CpG sites (CpGs 74–78). All DNA samples underwent a bisulfite treatment and were PCR amplified with primers specific for exon 1 of MGMT (GRCh37/hgl9 – chr10: 131,265,448- 131,265,560). Methylation status of PCR amplified products was determined using the PyroMark system. Samples with ≥7% and <9% methylation were considered equivocal.

Analysis of the clinical database B

The Tempus xT assay is a laboratory developed test that detects single-nucleotide variants, indels, and copy-number variants, as well as chromosomal rearrangements in 22 genes with high sensitivity and specificity (33). DNA sequencing of 595 to 648 genes and full-transcriptome RNA-seq were performed as previously described (33). Variant detection, visualization, and reporting were performed as previously described (33). The immune infiltration algorithm estimates the relative proportion of immune subtypes using a support vector regression (SVR) model that includes an L2 regularizer and an epsilon insensitive loss function similar to that of CIBERSORT (http://cibersort.stanford.edu). The SVR was implemented in Python using the nuSVR function in the SVM library of scikit-learn (0.18) with the LM22 reference matrix downloaded from the supplement of Newman and colleagues (34).

Mouse strains, orthotopic injections, and in vivo treatments

Male C57BL/6 WT mice were obtained from either Jackson Laboratories (Catalog no. 000664) or from the National Institute on Aging (NIA). WT mice from Jackson Laboratories or the NIA were obtained and experiments were initiated at either 1.5 to 4 months or at 18 to 26 months of age. Animal procedures were carried out in accordance with Northwestern University's Institutional Animal Care and Use Committee (IACUC). Transgenic INK-ATTAC mice on the C57BL/6 background mice were a generous gift from Dr. Sheila Stewart, PhD, at Washington University School of Medicine in St. Louis. The generation and characterization of the INK-ATTAC mouse line and their backcrossing onto the C57BL/6 genetic background was previously described (27, 35). INK-ATTAC mice were maintained in the Northwestern University Center for Comparative Medicine. Unmodified GL261 cells were acquired from the NIH/NCI at Frederick (Frederick, MD) and unmodified CT-2A cells were provided by Dr. Thomas Seyfried (Boston College, Boston, MA). Cell lines were authenticated prior to experimental use and cultured at 37°C in DMEM supplemented with 10% FBS, penicillin (100 μg/mL), and streptomycin (100 mg/mL). All cell culture reagents were from Gibco Invitrogen. For mouse studies, OS was defined as the time from intracranial injection until endpoint criteria/death. Median survival and its confidence intervals were estimated on the basis of the Kaplan–Meier estimates and survival curves plotted using the Kaplan–Meier method and compared by Log-rank test or Cox proportional hazard regression models. Mice were intracranially engrafted at 2 to 4 months or at 18 to 26 months of age with 1 × 10² to 5 × 10⁴ syngeneic (to C57BL/6 background) murine glioma cells. IACUC guidelines on the care and use of laboratory animals were followed for all surgical procedures. Before surgery, mice were subcutaneously injected with meloxicam (2 mg/kg) and at both 24 and 48 hours post-surgery for pain management. The surgical site was shaved and swabbed with iodine followed by the application of 70% ethanol. After making a 1-cm midline incision, a parietal burr hole was drilled 2 mm posterior to the coronal suture and 2 mm lateral to the sagittal suture. Injections were performed with a stereotactic frame and a 2.5-μL volume of glioma cells reconstituted in PBS was intracranially injected at a depth of 3 mm with a 22-gauge Hamilton syringe. After needle removal, the skin was stapled. Treatments for each experiment are indicated in the figure legends. Whole brain radiotherapy was administered once daily for 5 days after mouse placement in a lead box with head-only exposure and subsequent irradiation with 2 Gy via cesium-137. Prior to radiotherapy, mice were anesthetized with an intraperitoneal injection of a 0.15-mL solution containing ketamine HCl (90 mg/kg) and xylazine (10 mg/kg). Anti–PD-1 mAb (clone J43; BioXcell BP0033–2) treatments were administered via intraperitoneal injection as a 500-μg loading dose followed by three 200-μg maintenance doses every 3 days. Temozolomide (TMZ) treatments (33 mg/kg; Sigma) were administered via intraperitoneal injection once per day for 5 consecutive days. The IDO1 enzyme inhibitor, BGB-5777 (100 mg/kg; BeiGene) or BGB-7204 (100 mg/kg; Beigene), and/or the senolytics, dasatinib (5 mg/kg; LC Laboratories), and quercetin (50 mg/kg; MP Biomedicals), were suspended in ORAplus (Perrigo) and administered via oral gavage once (for senolytics) or twice (for IDO1 enzyme inhibitor) per day, Monday through Friday, for up to 4 weeks. AP20187 was solubilized in ethanol at a concentration of 62.5 mg/mL and stored at −20°C. Ten mg/kg AP20187 treatments were administered via intraperitoneal injection 3 times per week on Monday, Wednesday, and Friday, in a solution of 4% ethanol, 10% PEG-400, and 2% Tween in H₂O.

Flow cytometry for p^16INK4A+ or beta-galactosidase⁺ cells

To determine the proportion of GFP⁺ cells in WT versus INK-ATTAC mice, the stromal vascular fraction was isolated from inguinal adipose tissue as previously described (34, 35). The percentage of p^16INK4A+GFP⁺ cells in each group was assessed with flow cytometry as previously described (12). To determine the proportion of beta-galactosidase⁺ cells in WT mice, extra-tumoral tissue was isolated from mice at 20 days after engraftment with GL261 cells. Single-cell suspensions were made of extratumoral tissue using the Adult Brain Dissociation Kit (Miltenyi Biotec) according to the manufacturer's protocol. Flow cytometry was performed as previously described (12, 36), and oligodendrocyte cells were identified by staining for oligodendrocyte marker O1 (Clone O1, eBioscience) without co-expression of GFAP (Clone GA-5, Novus Bio) and TMEM119 (Clone V3RT1GOsz, eBioscience). Senescent cells were detected via β-galactosidase hydrolysis using the CellEvent TM Senescence Green Flow Cytometry kit (Invitrogen) according to the manufacturer's instructions.

RNA isolation and real-time quantitative PCR

Murine brain tissue was carefully dissected to isolate extratumoral tissue from within the contralateral brain. Total RNA was isolated with Trizol reagent and a PureLink RNA Mini Kit (Ambion 12183025). One microgram of total RNA was reverse transcribed into cDNA with the iScript cDNA Synthesis Kit (Bio-Rad). Quantitative real-time PCR (RT-PCR) was performed on the CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with the SYBR Green Supermix (Bio-Rad 1725274). Primers used for RT-PCR are shown in Supplementary Table S1. The 2^−ΔΔCT method was used to calculate the relative quantity of gene expression and the sample target threshold cycle (C_T) values were normalized to the internal housekeeping gene (GAPDH). mRNA expression of target genes relative to GAPDH was summarized using mean and SEM and compared between age and treatment groups. Statistical analyses were conducted using R Studio and R v. 3.6.1 statistical software, Cox models were fitted using the survival package in R, and emmeans package was used to calculate adjusted P values. A P value of ≤0.05 was considered significant. Figures were created using GraphPad Prism 9.0 software.

RNA-seq of murine extratumoral brain

Total RNA was isolated from murine extratumoral brain and sent to Novogene for bulk RNA-seq. Messenger RNA was purified from total RNA using poly-T oligo-attached magnetic beads. After fragmentation, the first strand cDNA was synthesized using random hexamer primers, followed by the second strand cDNA synthesis using either dUTP for a directional library or dTTP for a nondirectional library. For quality control, raw reads of FASTQ format were first processed through in-house perl scripts. Clean reads were obtained by removing low quality reads and reads containing adapters from raw data and the Q20, Q30, and GC content of the clean data were calculated to gauge quality. All the downstream analyses were based on the clean data with high quality. Hisat2 v2.0.5 was used to build the reference genome index and align the clean reads to the reference genome. FeatureCounts v1.5.0-p3 was used to count the number of reads mapped to each gene and the fragments per kilobase million of each gene was calculated on the basis of the gene length and read counts. Prior to differential gene expression analysis, the read counts were adjusted with the edgeR program package via one normalized factor. Differential expression analysis of two conditions was performed using the edgeR R package (3.22.5). To prevent statistical errors (false positives), P values were adjusted using the Benjamini and Hochberg method. This is standard practice when analyzing genomic data. Corrected P value of 0.05 and absolute fold-change of 2 were set as the threshold for significantly different expression. Gene ontology (GO) enrichment analysis of DEGs was performed with the clusterProfiler R package (Supplementary Table S2). GO terms with a corrected P value less than 0.05 were considered significantly enriched by DEGs. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed with the clusterProfiler R package to assess the statistical enrichment of differential expression genes in KEGG pathways. Pathways with a corrected P value less than 0.05 were considered significantly enriched by DEGs. All plots used to analyze RNA-seq results were created using Partek Flow software version 10.0.22.0121.

Single-cell RNA-seq

Single-cell suspensions were prepared from fresh tumor tissue using the Adult Brain Dissociation Kit (Miltenyi Biotec) according to the manufacturer's protocol. Tumor infiltrating leukocytes and peripheral immune cells were further isolated using biotinylated CD45 antibodies coupled with magnetic streptavidin microbeads (Miltenyi Biotec) form the extratumoral single-cell suspensions. Cell number and viability were analyzed using Nexcelom Cellometer Auto2000 with AOPI fluorescent staining method. Sixteen thousand cells were loaded into the Chromium Controller (10X Genomics, PN-120223) on a Chromium Next GEM Chip G (10X Genomics, PN-1000120), and processed to generate single-cell gel beads in the emulsion (GEM) according to the manufacturer's protocol. The cDNA and library were generated using the Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 (10X Genomics, PN-1000286) and Dual Index Kit TT Set A (10X Genomics, PN-1000215) according to the manufacturer's manual. Quality control for constructed library was performed by Agilent Bioanalyzer High Sensitivity DNA kit (Agilent Technologies, 5067–4626) and Qubit DNA HS assay kit for qualitative and quantitative analysis, respectively. The multiplexed libraries were pooled and sequenced on Illumina HiSeq 4000 sequencer with 2 × 50 paired-end kits using the following read length: 28bp Read1 for cell barcode and UMI, and 90 bp Read2 for transcript. The demultiplexing, barcoded processing, gene counting, and aggregation were made using the Cell Ranger software. To visualize the single-cell RNA-seq (scRNA-seq) results, the normalized gene barcode matrix was used to compute a neighborhood graph of cells, then Uniform Manifold Approximation and Projection (UMAP) was performed with default parameters. The whole pipeline was implemented using SCANPY 6. Cell type annotation was performed using R package SingleR 7.

Cell viability assay

To assess the potential for cytotoxicity attributable to AP20187, in vitro–cultured GL261 cells were seeded in a 96-well plate at a density of 1 × 10⁴/well. AP20187 was dissolved in 100% ethanol. Twenty-four hours later, cells were rinsed with 1 × PBS and treated with AP20187 at 10 nmol/L, 100 nmol/L, or 1 μmol/L. Controls included no treatment, 1 μmol/L staurosporine, and vehicle. At 48 hours posttreatment, the media was changed and an MTT assay was performed with the R&D Systems TACS MTT Cell Proliferation Assay (Catalog no. 4890–025-K) according to the manufacturer's instructions. The absorbances at 570 nm were used to assess cell viability.

Statistical analysis

Data are represented as the mean ±SEM. Lines were fit to enzyme activity plots using linear regression. The statistical significance of the differences in mRNA expression and tumor-infiltrating T-cell response between the two groups was determined by a Student t test or Wilcoxon rank sum test as appropriate. Differences among multiple groups were assessed using ANOVA with post hoc Tukey test, or Kruskal–Wallis test followed by Bonferroni correction as appropriate. OS was defined as the day of tumor cell engraftment until reaching endpoint criteria and/or death. Survival curves were generated using the Kaplan–Meier method and compared by Log-rank test and Cox Regression analysis. Data were analyzed using Prism 7.0 software (GraphPad Software) and R 3.6.1. A P < 0.05 was considered significant.

Data availability statement

The RNA-seq data have been deposited in Gene Expression Omnibus and the accession number is GSE214294. The data generated in this study are available within the article and its Supplementary Data files.

Results

OS decreases in older adults with GBM and is independent of other prognostic factors

SEER data were analyzed across all age groups (n = 46,998) for OS and estimated median OS (mOS). Older patients with GBM have decreased OS and mOS as compared with their younger counterparts. Patients with GBM who are 18 to 44 (n = 3,793), 45 to 54 (n = 7,297), 55 to 64 (n = 12,510), 65 to 74 (n = 12,434), and 75+ (n = 10,964) years of age have a mOS of 1.58, 1.17, 0.83, 0.50, and 0.25 years, respectively (Fig. 1A; Supplementary Fig. S1; P < 0.0001 for all pairwise comparisons). To address potential interference from non–GBM-related mortalities, a competing risks analysis was performed on the same patient data analyzed in Fig. 1A and B). Patients with GBM are more likely to succumb from GBM (solid lines) compared with non–GBM-related causes (dashed lines) across all age groups.

Figure 1.

The relationship between MGMT promoter methylation status, tumor cell infiltration, OS, and GBM patient age. A, Life expectancy from the SEER database for patients with GBM. Subjects were stratified by 18 to 44 (blue), 45 to 54 (orange), 55 to 64 (purple), 65 to 74 (green), or 75+ (red) years of age (YOA). B, Competing risk analysis of patients with GBM to account for non–GBM-related mortality. Subjects were stratified by 18 to 44 (red), 45 to 54 (blue), 55 to 64 (green), 65 to 74 (purple), or 75+ (orange) YOA. The NMEDW was accessed to assess GBM patient survival based on (C) IDHwt status and (D) MGMT promoter methylation status. Subjects were stratified by (C) <65 (blue) or ≥65 (red) YOA or by (D) <65 YOA and MGMT unmethylated (purple), ≥65 YOA and MGMT-unmethylated (red), <65 YOA and MGMT methylated (blue), ≥65 YOA and MGMT methylated (green), respectively. Levels of GBM cell brainstem infiltration (E) and associated OS (F) for human patients that were stratified by age. Groups include <65 YOA and no brainstem infiltration (yellow, n = 1), ≥65 YOA and no brainstem infiltration (red, n = 1), <65 YOA and microscopic brainstem infiltration (green, n = 4), ≥65 YOA and microscopic brainstem infiltration (purple, n = 5), <65 YOA and extensive brainstem infiltration (blue, n = 20), or ≥65 YOA and extensive brainstem infiltration (pink, n = 2). *, P < 0.05; ****, P < 0.0001.

We next examined the relationships between IDH status, MGMT promoter methylation status, and age at the time of diagnosis by extracting electronic health information from the NMEDW for patients with GBM (Fig. 1C and D). To enhance the rigor of the survival analysis, patients with unknown IDH or MGMT status were excluded. As compared with younger patients with GBM <65 years of age (n = 245) with a mOS of 1.60 years, patients with GBM ≥65 years of age (n = 163) have a decreased mOS of 0.92 years (Fig. 1C; P < 0.0001). Similarly, patients with GBM <65 years of age with methylated or unmethylated MGMT promoter have a mOS of 4.63 and 1.27 years, respectively, which is increased as compared with patients with GBM ≥65 years of age who have a mOS of 1.30 and 1.05 years, respectively (P < 0.001; Fig. 1D). Collectively, these data indicate that increased age is a negative prognostic factor for patients with GBM even after accounting for MGMT promoter methylation status.

We next analyzed age-associated relationships from a former study that discovered a negative relationship between extent of brainstem tumor cell infiltration and GBM patient OS (37). Thirty-three patients with GBM ranging between 26 and 74 years of age were stratified by <65 years of age, ≥65 years of age, and extent of brainstem infiltration upon postmortem autopsy (Fig. 1E and F). Brainstem infiltration was categorized as either none, microscopic, or extensive. Microscopic tumor cell infiltration indicates the presence of sparse tumor cells and no tissue damage whereas extensive tumor cell infiltration is defined as the presence of numerous tumor cells and tissue damage. While the majority of patients with GBM <65 years of age presented with extensive infiltration (blue), the majority of patients ≥65 years old presented with microscopic infiltration (purple; Fig. 1E). Of the 20 patients <65 years of age that presented with extensive brainstem tumor cell infiltration, 75% exhibited premortem brainstem symptoms. In contrast, of the 5 patients <65 years of age with microscopic or no brainstem infiltration, only 20% exhibited premortem brainstem symptoms. The situation in older adults ≥65 years of age was similar with the 2 patients presenting with extensive brainstem tumor cell infiltration also demonstrating premortem brainstem symptoms whereas only 1 of 6 older adult patients with microscopic or no brainstem infiltration exhibited premortem brainstem symptoms. Generally, older patients with GBM tended to have less brainstem involvement as compared with younger patients. Overall, there was concordance between brainstem infiltration and premortem brainstem symptoms among both younger and older adults with GBM. However, there was a substantially increased frequency of younger patients with GBM with extensive tumor cell infiltration of the brainstem as compared with older adults (Fig. 1F; Supplementary Table S3).

Intra-GBM gene expression, tumor methylation status, and TMB are largely unchanged with varying patient age

The GBM dataset from TCGA was analyzed to explore differences between intra-GBM gene expression, TMB, and DNA methylation (DNAm) for older as compared with younger patients. The mOS of patients with GBM ≥65 years of age (n = 194) was significantly decreased compared with their younger counterparts (n = 363; Fig. 2A). A Cox proportional hazards regression model demonstrated that the HR for patients with GBM ≥65 years of age is 2.1 (P < 0.001; Fig. 2B) indicating that the prognosis for individuals <65 years of age is significantly improved compared with those ≥65 years of age. In contrast, the analysis of sex for the same patient population failed to demonstrate a significant change in mOS (HR = 1.2; P = 0.077).

Figure 2.

Intratumoral gene expression, TMB, and DNAm does not substantially change in younger as compared with older patients with GBM. A, OS of patients with GBM from TCGA as stratified by <65 (blue) or ≥65 (red) years of age (YOA). B, Cox proportional hazards model for GBM patient age and sex. C, Differentially methylated genes in patients ≥65 YOA (blue; n = 92) as compared with patients <65 YOA (red; n = 170). D, RNA expression in patients with GBM ≥65 YOA (red; n = 57) as compared with patients <65 YOA (n = 96). E, TMB in patients with GBM <65 (blue) and ≥65 YOA (red) stratified by mutation type. F, DNAm age acceleration in patients with GBM <65 (blue) and ≥65 YOA (red).

Of the ∼20,000 genes in the human genome, only the methylation of PCOLCE2 and SLC10A4 was increased in patient-resected GBM in older adults compared with younger counterparts (Fig. 2C). However, a comparison of intra-GBM RNA gene counts found no significant age-dependent differences in gene expression for PCOLCE2 and SLC10A4 (Fig. 2D). In addition, TMB and neoantigen burden is similar between younger and older adult patients except for a small increase in missense mutations in older adults than in the younger counterparts (P = 0.0058; Fig. 2E; Supplementary Fig. S2). We obtained sample-specific annotations for prior exposure to radiotherapy. Although there was a lack of information for 40% (n = 106) of total samples, 49% of GBM (n = 130) were confirmed to have been irradiated prior to sample collection and only 10% (n = 26) were radiation-naïve. We controlled for this and for gender in differential analyses with the resulting identification of a single differentially methylated gene (PCOLCE2; Supplementary Fig. S2). The difference between epigenetic age and chronologic age is referred to as DNAm age acceleration and was previously defined (38). DNAm age acceleration is higher in patient-resected GBM from younger adults as compared with those from older adults (P = 0.0022; Fig. 2F), which is consistent with the premise that normal patterns of age-dependent methylation are altered in malignant tissue (39). Overall, as our analysis of the TCGA database for GBM suggests only minor and/or negligible intratumoral differences associated with subject age, tumor-centric analyses do not fully explain the survival differences of younger versus older patients with GBM.

A comparison of multiple clinical databases generally indicates similar levels of intra-GBM prognostic and immunologic markers between young and old patients

To validate the conclusion(s) of the TCGA database analysis, an additional analysis of IDHwt GBM from two commercial repositories was stratified by age for common immune-associated gene expression, markers for immune cells, and gene amplification/mutations (Tables 1–3). Both commercial databases defined young and old patients with GBM as <65 and ≥65 years of age, respectively. Intra-GBM transcript expression levels for biomarkers which included Lag3, PDCD1, CD274, CD3ε, TNFRSF18, CD40, CD8α, TNFRSF4, IDO1, CTLA4, HAVCR2, TNFSF9, MGMT, and CDKN2A, did not change among younger and older adults across both clinical databases (Table 1). Similarly, intra-GBM RNA expression profiling for immune cells representing macrophages (MQ), NK cells, CD4⁺ T cells, CD8⁺ T cells, and B cells did not differ between younger and older adults across both clinical databases (Table 2). Finally, there were no age-dependent intra-GBM changes for programmed death-ligand 1 (PD-L1) IHC, common DNA amplifications including CDK6 and EGFR, as well as mutations associated with EGFRvIII, MET fusion, PTEN, TP53, or NF1 (Table 3). There was a small but significant increase for TERT promoter mutations in GBM resected from older adults compared with younger adults across both databases. Differences between the two databases are attributable to classification of pathogenic versus likely pathogenic TERT promoter mutations. MGMT-promoter methylation showed enrichment from older adults compared with younger adults in one of the databases, but not both databases (Table 3). Overall, there were minimal age-dependent intra-GBM differences that could be validated by multiple clinical databases.

Table 1.

Commercial database analyses for gene expression of select immunologic markers in tumor-resected GBM from younger and older adult human patients.

Gene	A <65, N = 902	A ≥65, N = 530	A P value	B <65, N = 648 (log10)	B ≥65, N = 367 (log10)	B P value	Significant datasets
LAG3	0.38	0.41	0.544	1.50	1.44	< 0.0001	1
PDCD1	0.30	0.33	0.144	1.62	1.62	0.935	0
CD274	3.74	3.61	0.369	1.87	1.9	0.444	0
CD3E	0.65	0.59	0.098	1.27	1.24	0.922	0
TNFRSF18	0.26	0.25	0.724	1.41	1.40	0.251	0
CD40	2.14	2.10	0.291	1.95	1.93	0.099	0
CD8A	0.69	0.61	0.226	1.15	1.11	0.690	0
TNFRSF4	0.46	0.43	0.278	1.84	1.80	0.120	0
IDO1	0.31	0.23	0.002	0.90	0.89	0.939	1
CTLA4	0.30	0.29	0.076	1.17	1.18	0.840	0
HAVCR2	32.44	31.37	0.637	2.83	2.85	0.061	0
TNFSF9	0.22	0.20	0.116	0.98	0.96	0.817	0
CDKN2A	1.97	2.03	0.945	1.84	1.75	0.044	0

Data from Caris Life Sciences (Database A) and Tempus (Database B) for gene expression of common immunologic markers as stratified by <65 or ≥65 years of age. Significance, or lack thereof, for each data point across both datasets is included in the final column.

Table 2.

Commercial database analyses for gene expression of select immune cells in tumor-resected GBM from younger and older adult human patients.

Immune cell	A <65, N = 902	A ≥65, N = 530	A P value	B <65, N = 648	B ≥65, N = 367	B P value	Significant datasets
MQ	0.010259	0.020132	0.008	0.71	0.71	0.853	1
NK cell	0.059850	0.060562	0.576	0.12	0.11	0.838	0
CD4+ T cell	0.00	0.00	0.939	0.05	0.06	0.700	0
CD8+ T cell	0.00	0.00	0.454	0.000	0.000	0.850	0
B cell	0.064778	0.064242	0.249	0.06	0.06	0.969	0

Data from Caris Life Sciences (Database A) and Tempus (Database B) for gene expression of common immunologic markers as stratified by <65 or ≥65 years of age. Significance, or lack thereof, for each data point across both datasets is included in the final column.

Table 3.

Commercial database analyses of select tumor cell–associated markers in resected GBM from younger and older adult human patients.

Feature	A Positive (Age <65)	A Negative (Age <65)	A Positive (Age ≥65)	A Negative (Age ≥65)	A P value	B Positive (Age <65)	B Negative (Age <65)	B Positive (Age ≥65)	B Negative (Age ≥65)	B P value	Significant datasets
MGMT-Me	283 (34.13%)	546 (65.87%)	277 (49.73%)	280 (50.27%)	6.63E-09	N/A	N/A	N/A	N/A	N/A	1
IHC PD-L1	176 (21.54%)	641 (78.46%)	98 (17.82%)	452 (82.18%)	0.092	57 (20.0%)	230 (80.0%)	41 (21.0%)	152 (79.0%)	0.712	0
dMMR/MSI-H	7 (0.78%)	895 (99.22%)	8 (1.51%)	522 (98.49%)	0.188	11 (0.9%)	882 (99.1%)	1 (0.2%)	483 (99.8%)	0.067	0
CDK6 amplification	8 (0.89%)	890 (99.11%)	9 (1.70%)	519 (98.30%)	0.172	11 (1.2%)	904 (98.8%)	1 (0.2%)	494 (99.8%)	0.067	0
EGFR amplification	324 (36.04%)	575 (63.96%)	182 (34.47%)	346 (65.53%)	0.549	267 (29.0%)	648 (71.0%)	151 (31.0%)	344 (69.0%)	0.603	0
NGS-EGFR	167 (18.56%)	733 (81.44%)	81 (15.31%)	448 (84.69%)	0.118	110 (12.0%)	805 (88.0%)	65 (13.0%)	430 (87.0%)	0.546	0
EGFRvIII mutations	197 (21.86%)	704 (78.14%)	104 (19.62%)	426 (80.38%)	0.315 (RNA-seq)	87 (9.5%)	828 (90.0%)	44 (8.9%)	451 (91.0%)	0.702 (DNAseq)	0
EGFR Fusion	11 (1.24%)	874 (98.76%)	4 (0.76%)	520 (99.23%)	0.397	45 (4.9%)	870 (95.0%)	23 (4.6%)	472 (95.0%)	0.820	0
MET Fusion	11 (1.22%)	890 (98.78%)	8 (1.51%)	520 (98.48%)	0.639	0 (0.0%)	915 (100%)	0 (0.0%	495 (100%)	N/A	0
NGS-TERT	662 (77.07%)	197 (22.93%)	473 (82.55%)	100 (17.45%)	0.012	452 (49.0%)	463 (51.0%)	288 (58.0%)	207 (42.0%)	0.002	2
NGS-PTEN	252 (30.39%)	577 (69.61%)	198 (35.23%)	364 (64.77%)	0.059	261 (29.0%	654 (71.0%)	144 (29.0%)	351 (71.0%)	0.823	0
NGS-TP53	258 (28.67%)	642 (71.33%)	170 (32.14%)	359 (67.86%)	0.167	170 (19.0%)	745 (81.0%)	77 (16.0%)	418 (84.0%)	0.154	0
NGS-NF1	131 (14.57%)	768 (85.43%)	79 (14.96%)	449 (85.04%	0.841	99 (11.0%)	816 (89.0%)	59 (12.0%)	436 (88.0%)	0.532	0

Data from Caris Life Sciences (Database A) and Tempus (Database B) for DNA expression, amplification, mutations, fusion events, and PD-L1 protein detection as stratified by <65 or ≥65 years of age. Significance, or lack thereof, for each data point across both datasets is included in the final column. NGS was used to detect mutations.

RNA-seq reveals a unique gene expression profile induced by a brain tumor in the older adult brain

The negligible intratumoral changes across patient-resected GBM from younger as compared with older adults are unlikely to explain the substantial differences in survival between younger and older patients with GBM. We therefore hypothesized that age-dependent changes contributing to worse survival outcomes in older adults occur outside of the bulk tumor mass. Bulk RNA-seq was performed on tissue isolated from the extratumoral brain from 2-month- and 23-month-old male C57BL/6 mice with or without an intracranial syngeneic GL261 cell-based brain tumor (Fig. 3A; Supplementary Fig. S3). Principal component analysis (PCA) of DEGs reveals a distinct gene expression profile of extratumoral tissue from an older adult mouse with a brain tumor as compared with all other groups (Fig. 3B). Heat map comparison of DEGs confirms the relatively distinct gene expression profile of extratumoral tissue that was isolated from an older adult mouse brain with a contralateral brain tumor as compared with all other groups (Fig. 3C).

Figure 3.

Gene expression profiling of the extratumoral brain for young and older adult mice with or without a brain tumor. A, Schematic representation of how the extratumoral (left brain hemisphere) and intratumoral (right brain hemisphere when indicated) compartments are defined within the same brain. B, PCA of the naïve or extratumoral brain from C57BL/6 mice at: 2 months of age without a brain tumor (green), 2 months of age with a brain tumor (yellow), 23 months of age without a brain tumor (red), and 23 months of age with a brain tumor (blue). C, Global RNA-seq heat map of the samples shown in B. D, Gene expression associated with p53, senescence, and SASP pathways/products. E, The brain from 2-month-old and 23-month-old mice with intracranial GL261 was isolated from naïve mice or at 7 (n = 5), 14 (n = 5), or 21 (n = 5) days post-intracranial injection (i.c.) for RT-PCR analysis of the intra- and extratumoral tissues for CDKN2A. F, Gene expression associated with immune-related pathways/products. **, P < 0.01; ***, P < 0.001. (A, Created with BioRender.com.)

Because senescence is a process that tends to progressively increase with age, and because senescence promotes cancer relapse in a non-brain tumor setting (40), we next hypothesized that the brain tumor would enhance extratumoral gene expression for markers of senescence and/or the SASP. In accordance with this notion, CDKN2A and several genes of the p53 pathway were upregulated in the older adult extratumoral brain in response to brain tumor burden (blue) as compared with all other groups (green, yellow, red; Fig. 3D).

Confirmatory analysis of extratumoral tissue showed a progressive age-dependent upregulation of CDKN2A levels that selectively occurred in older adult mice with a brain tumor but not in younger counterparts (Fig. 3E). In contrast, there was no age-dependent difference for intratumoral CDKN2A levels between young and older adult mice. Pathway analysis of the extratumoral mouse brain showed additional age and brain tumor co-dependent changes including the adaptive immune response, cytokine production, T-cell–mediated immunity and exhaustion, IFNγ production, and immune checkpoint blockade molecules (Fig. 3F; Supplementary Table S2).

Senescent cells, senolytic effects, and/or therapeutic treatment effects in older adult mice with a brain tumor

UMAP reconstruction of scRNA-seq analysis shows the major representative cellular populations that were included in the analysis of extratumoral brain from young and old mice with a brain tumor including astrocytes (Astro), B cells, choroid plexus epithelial cells, endothelial cells (EC), ependymocytes, immature neurons (ImmN), macrophages (MAC), microglia (MG), neutrophils, oligodendrocytes (OLG), oligodendrocyte precursor cells (OPC), T cells, and mature neurons (mNeur; Fig. 4A). The analysis confirms that a low 1% to 5% level of Astro, EC, ImmN, MAC, MG, T cells, and mNeur express p16^INK4A and are present in both the young and old extratumoral brain of mice with a brain tumor. There is also a preferential presence of B cells and OLG that express p16^INK4A in the older adult, but not the younger extratumoral brain. Finally, the number of OPC and T cells that express p16^INK4A is doubled in the older adult, as compared with the young extratumoral brain of mice with a brain tumor. The proportionality of the cellular populations analyzed by scRNA-seq is roughly similar with an exception for ImmN and MAC that were substantially decreased in the older adult, as compared with the younger extratumoral brain with a brain tumor (Fig. 4B). Senolytic treatment with dasatinib and quercetin decreases the number of extratumoral cells that are beta-galactosidase positive in the older adult brain with a brain tumor (Fig. 4C). The data collectively suggest that the extratumoral cellular lineage with the highest frequency of senescent cells in the older adult brain with a brain tumor is the OPCs and that they can be eradicated with dasatinib and quercetin treatment.

Figure 4.

Cellular identification in the brain of young and older adult mice with an intracranial tumor and after treatment with radiotherapy (RT) and PD-1 mAb or dasatinib and quercetin. A, UMAP analysis for scRNA-seq analysis of the extratumoral brain for young 7-week-old and old 85-week-old C57BL/6 mice with intracranial CT-2A to visualize populations of cells including astrocytes (Astro), B cells, choroid plexus epithelial cells (CPC), EC, ependymocytes (EPC), ImmN, MAC, MG, neutrophils (Neut), OLG, OPCs, T cells, and mNeur. Among these cell populations, the relative frequency of p16^INK4A+ expression is shown. B, Of the total number of cells from the identified cellular populations, each one is demonstrated as a proportion of the total for the extratumoral brain from young and old mice with a brain tumor. scRNA-seq data represent the pooled analysis of 2 mice/group. C, 21- to 28-month-old C57BL/6 mice with intracranial GL261 were treated with dasatinib (5 mg/kg) and quercetin (50 mg/kg), Monday–Friday, for 2 weeks beginning on Day 7 post i.c. Extratumoral brain isolation occurred on Day 20 after tumor cell injection and flow cytometric analysis for beta-galactosidase in cells positive for oligodendrocyte marker O1 (Olig) and negative for GFAP and TMEM119 are shown. A flow cytometry contour plot is shown for one mouse that represents the outcomes for each group (n = 3 mice/group). D, 8-week-old (blue) and 90-week-old (red) C57BL/6 mice with intracranial GL261 were treated with 2 Gy radiation (RT) x 5 days total and one 500-μg loading dose followed by three 100-μg maintenance doses of anti–PD-1 mAb beginning at day 14 after tumor cell injection. Extratumoral brain and brain tumor samples were collected on days 15 and 25 posttreatment for RT-PCR analysis of CDKN2A, p16^INK4A, and p19^ARF. E, 8-week-old (blue) and 90-week-old (red) C57BL/6 mice were intracranially injected with GL261 cells and treated with or without dasatinib (D; 5 mg/kg) and quercetin (Q; 50 mg/kg) on days 9, 10, 11, 14, 15, and 16 post-engraftment. Extratumoral brain and brain tumor samples were collected for quantitative gene expression analysis of CDKN2A. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant

The mouse CDKN2A locus encodes two structurally unrelated tumor suppressor proteins including p16^INK4A and p19^ARF in mice (41). Activation of the CDKN2A locus results in the expression of p16^INK4A and p19^ARF that is observed in most senescent cells and may play a causal role in their growth arrest (42). Because increased senescent cell burden has previously been shown to promote tumor outgrowth in non-primary brain tumor models (40), we next evaluated CDKN2A, p16^INK4A and p19^ARF levels in the extratumoral brain of young and older adult mice with a brain tumor. We coupled these investigations with radiotherapy and an anti–PD-1 mAb, which was predicated on the unexpectedly poor survival outcome of patients with newly diagnosed MGMT unmethylated GBM treated with radiotherapy and anti–PD-1 mAb (43) combined with our interest in the phase I trial by our group combining radiotherapy, anti–PD-1 mAb, and an IDO1 enzyme inhibitor (NCT04047706). Two-month- and 23-month-old WT mice with intracranial GL261 were treated with radiotherapy and PD-1 mAb and analyzed at 1 day after treatment initiation (day 15 post-GL261 injection) and 11 days after treatment initiation (day 25 post-GL261 injection; Fig. 4D). For both time points, CDKN2A (inclusive of detection for both p16^INK4A and p19^ARF), p16^INK4A, and p19^ARF expression levels were increased in the older adult extratumoral brain as compared with younger counterparts (P < 0.05). In addition, the expression levels of CDKN2A, p16^INK4A, and p19^ARF increased during brain tumor progression and posttreatment in the older adult brain. There was no increase for CDKN2A, p16^INK4A, and p19^ARF levels in the young adult extratumoral brain. There was also no change in transcript expression for CDKN2A, p16^INK4A, and p19^ARF when comparing brain tumors isolated from young and old mice. In a follow-up experiment, 2-month- and 23-month-old mice were intracranially injected with GL261 and treated with vehicle or the blood–brain barrier–penetrating senolytics combination of dasatinib and quercetin (Fig. 4E; ref. 44). CDKN2A levels were increased in the extratumoral brain of older adult mice with a brain tumor as compared with young counterparts. Dasatinib and quercetin treatment decreased extratumoral CDKN2A levels suggesting that the senolytics successfully eradicated senescent cells in the older adult brain. Collectively, these data show that senescent cell markers selectively increase in the aged brain parenchyma after brain tumor treatment and that those markers are decreased after treatment with the senolytics, dasatinib, and quercetin.

Treatment with combination immunotherapy and senolytics improves survival in older adult mice with a brain tumor

Previous work demonstrated that the simultaneous treatment with radiotherapy, anti–PD-1 mAb, and an IDO1 enzyme inhibitor produces a robust and long-term survival (LTS) benefit in young mice with a well-established syngeneic brain tumor (12). However, the LTS arising from that therapeutic approach was subsequently shown to be ablated in older adult mice with a brain tumor, indicating an age-dependent treatment benefit that was preferentially experienced during youth (5). On the basis of the ability of radiotherapy and PD-1 mAb treatment to induce markers of senescence in the extratumoral brain, we next evaluated the potential therapeutic benefit(s) of senescent cell eradication in young 2-month-old (Fig. 5A) and older adult 18-month-old (Fig. 5B) mice with intracranial GL261 and treated with: (i) vehicle control and IgG antibodies; (ii) dasatinib and quercetin (D/Q); (iii) immunotherapy with radiotherapy, anti–PD-1 mAb, and an IDO1 enzyme inhibitor; or (iv) immunotherapy combined with D/Q (Supplementary Fig. S4). Regardless of age, there was no survival difference between mice treated with vehicle + IgG antibodies versus those animals treated with D/Q. As was previously demonstrated, a majority of young mice (60%) experienced improved LTS after treatment with radiotherapy, anti–PD-1 mAb, and IDO1 enzyme inhibitor treatment (Fig. 5A). The addition of senolytics to immunotherapy did not further improve LTS in young mice as compared with young mice treated with immunotherapy alone. In further agreement with previous observations (5), the combination of radiotherapy, anti–PD-1 mAb, and IDO1 enzyme inhibitor treatment failed to produce LTS in older adult mice with a brain tumor (Fig. 5B). In contrast, LTS was experienced by 15% of older adult mice that were simultaneously treated with senolytics and immunotherapy. There was no LTS experienced by older adult mice with a brain tumor treated with radiotherapy and TMZ treatment regardless of co-treatment with senolytics (Fig. 5C).

Figure 5.

Evaluation of LTS after treatment with immunotherapy and/or senescent cell eradication strategies in young or older adult mice with a brain tumor. A, 2- and (B) 18-month-old C57BL/6 WT mice were intracranially engrafted with 50,000 GL261 cells and treated with either (i) IgG control antibodies and empty OraPlus (red circle), (ii) 5 mg/kg dasatinib (D) and 50 mg/kg quercetin (Q), (iii) 2 Gy x 5 days radiotherapy (RT), one 500-μg loading dose followed by three 100-μg maintenance doses of anti–PD-1 mAb (clone J43), and 100 mg/kg IDO1 enzyme inhibitor (IDO1i; BGB-7204) x 5 days/week for up to 4 weeks total, or (iv) the combination of D, Q, RT, anti–PD-1 mAB and IDOi, beginning at day 14 post-5×10⁴ GL261 cell injection per the dosing schedule described in Supplementary Fig. S4. C, 20- to 22-month-old WT mice were intracranially engrafted with 5×10⁴ GL261 cells and treated with either (i) vehicle control, (ii) D and Q, (iii) 2 Gy x 5 days RT and 33 mg/kg TMZ, or the combination of D, Q, RT, and TMZ beginning at day 14 post-5×10⁴ GL261 cell injection. D, 3- to 5- and 19- to 23-month-old WT and INK-ATTAC (C57BL/6 background) mice began treatment with 10 mg/kg AP20187×2 to 3 days/week at day -21 prior to intracranial engraftment with 5×10³ GL261 cells. Mice were then further treated with 2 Gy x 5 days RT, one 500-μg loading dose followed by three 100-μg maintenance doses of anti–PD-1 mAb (clone J43), and 100 mg/kg IDO1 enzyme inhibitor IDOi beginning at day 14 post–tumor cell injection. E, Schematic representation of the findings reported in this manuscript. *, P < 0.05; ***, P < 0.001. (E, Created with BioRender.com.)

To validate the age-dependent observations above with a different in vivo modeling system, INK-ATTAC mice on the C57BL/6 background were studied. The INK-ATTAC mouse model was previously shown to facilitate the inducible eradication of senescent cells after treatment with the pharmacologic compound, AP20187 (27). Importantly, AP20187 is not toxic to in vitro cultured GL261 cells (Supplementary Fig. S5) and decreases p16^INK4A-expressing cells in treated INK-ATTAC mice as compared with WT mice (Supplementary Fig. S6A). Young (3- to 5-month-old) and older adult (19- to 23-month-old) WT littermates and INK-ATTAC mice began treatment with AP20187 beginning at 21 days prior to intracranial engraftment with 5 × 10³ GL261-luciferase expressing cells (Fig. 5D; Supplementary Fig. S4). At 14 days post intracranial engraftment, mice began treatment with radiotherapy, anti–PD-1 mAb and an IDO1 enzyme inhibitor. Similar to the outcome arising from young WT mice co-treated with immunotherapy and senolytics (Fig. 5A), young WT littermates treated with radiotherapy, anti–PD-1 mAb, and IDO1 enzyme inhibitor experienced a slightly higher LTS as compared with young INK-ATTAC mice during co-treatment for senescent cell eradication (Fig. 5D). In agreement with outcomes observed in older adult WT mice co-treated with immunotherapy and senolytics (Fig. 5B), older adult WT littermates experienced a markedly reduced LTS as compared with older adult AP-treated INK-ATTAC mice during co-treatment for senescent cell eradication (Fig. 5D). The conclusions found by our investigation are presented in Fig. 5E.

Discussion

Advanced age is a strong prognostic factor for poorer survival outcomes in patients with GBM (45). Historically, the presumed reason(s) for inferior survival outcomes in older adult patients were primarily attributed to less aggressive treatment choices out of the concern for potential treatment-related toxicities, a reduced physiologic reserve, and/or subject frailty (6, 46). However, even among optimal conditions that require meeting strict eligibility criteria and include highly controlled treatment regimens such as a prospective randomized phase III clinical trial, older patients with GBM have worse outcomes as compared with younger counterparts (47). To investigate the mechanistic basis for the worse survival outcomes in older adults with GBM, the current study evaluated human GBM patient data from the SEER program database, the NMEDW electronic health record, TCGA database, and the commercial Tempus Labs and Caris Life Sciences clinical databases. The collective conclusion of the clinical analysis is that, while older adult human patients with GBM have decreased survival as compared with their younger counterparts, the biological underpinnings for this age-dependent survival difference does not primarily arise from within the brain tumor itself. This was confirmed by a general lack of change for intratumoral gene expression, DNAm, TMB/MMR, and immune-related profiles when comparing untreated resected GBM isolated from younger adults <65 years of age to untreated resected GBM from older adults ≥65 years of age. Instead, and in accordance with the preclinical analysis of young and old mice with an intracranial brain tumor, we discovered marked age-dependent changes that occurred in the extratumoral brain parenchyma of older adult mice, and importantly, these changes were associated with a functionally worse survival outcome after treatment with immunotherapy. Specifically, extratumoral markers for senescence were increased in response to tumor progression and/or the treatment with radiotherapy and PD-1 mAb. Moreover, the treatment with senolytics in WT mice, or the inducible eradication of host (non-tumor) p16^INK4A-expressing cells in INK-ATTAC mice, were further combined with radiotherapy, PD-1 mAb, and IDO1 enzyme inhibitor in older adult mice with a brain tumor, a significant improvement in OS was observed.

The rationale for choosing radiotherapy and PD-1 mAb treatment for evaluating markers of senescence in the extratumoral brain of young and old mice was based on several recent clinical observations including a recently completed negative phase II trial (NCT04195139) in older patients with GBM treated with hypofractioned radiotherapy and concurrent TMZ, followed by adjuvant TMZ with or without the addition of nivolumab (PD-1 mAb; ref. 48). Similarly, phase III clinical trial evaluation of patients with newly diagnosed MGMT promoter unmethylated GBM treated with radiotherapy and nivolumab did not show an improved OS as compared with those individuals treated with radiotherapy and TMZ when analyzed holistically by 2 separate groups (NCT02617589; ref. 44). However, the data became more interesting when analyzing the trial among patients with GBM across different age groups. The younger patients with GBM who were <65 years of age and treated with radiotherapy and nivolumab experienced an improved HR [1.33 (1.07–1.67)] as compared with similar ages of patients with GBM treated with radiotherapy and TMZ standard of care. The HR improved further to 1.65 (1.09–2.51) among even younger patients with GBM who were <50 years of age and treated with radiotherapy and nivolumab as compared with similarly aged patients with GBM treated with standard of care. Strikingly and concordant with the phase III trial outcome described above (NCT04195139), the HR was not significantly different for patients with GBM ≥65 years of age who were treated with radiotherapy and nivolumab as compared with those treated with radiotherapy and TMZ. Because these outcomes collectively suggest that the dual treatment combination with radiotherapy and PD-1 mAb produces a modest but significant survival benefit in younger adults but not older adults with GBM, and because radiotherapy has previously been demonstrated to increase senescence in normal tissues (15, 49), we hypothesized that the poorer survival outcome of radiotherapy and PD-1 mAb treatment may be associated with an enhanced level of senescence in the extratumoral brain of older adults. In support of this hypothesis, we found an age-dependent increase for the senescence markers, CDKN2A, p16^INK4A, and p19^ARF in the older adult as compared with young extratumoral mouse brain.

To understand whether the treatment- and age-dependent enhancement of senescence is therapeutically relevant, young and old WT mice with a brain tumor were evaluated with a derivative immunotherapeutic approach whereby radiotherapy and PD-1 mAb was further combined with an IDO1 enzyme inhibitor. The rationale for evaluating the triple-agent rather than dual-agent combination was 4-fold including: (i) the overarching negative clinical trial result of Checkmate 498 suggesting that radiotherapy and PD-1 mAb fails to improve survival as compared with standard-of-care treatment for patients with MGMT promoter unmethylated GBM (44); (ii) the synergistic survival benefit that is demonstrated after combining radiotherapy and PD-1 mAb with an IDO1 enzyme inhibitor in a preclinical mouse brain tumor model (12); (iii) the striking age-dependent disparity whereby young mice with a brain tumor experience an impressive survival benefit after treatment with radiotherapy, PD-1 mAb, and IDO1 enzyme inhibitor as compared with the poor survival outcome of older adult mice with a brain tumor that are treated similarly (5); and (iv) our ongoing phase I clinical evaluation of patients with GBM treated with standard radiotherapy, nivolumab, and BMS-986205 (IDO1 enzyme inhibitor; NCT04047706) with a preliminary indication that younger adults survive longer than older adults posttreatment (50).

Despite the comprehensive nature of our study, there are limitations to be acknowledged. First, the diagnostic methodologies that were used across the Tempus, Caris, and TCGA databases were sufficiently distinct to prevent our capability for analyzing the data in aggregate and directly side-by-side. Next, younger patients with GBM <65 years of age were overrepresented in our analyses at a frequency of 60.0%, 64.9%, and 65.2% in commercial database A, commercial database B, and the TCGA, respectively, which is significantly greater than the normal distribution for patients with GBM of that age group that represents ≤35% of younger individuals according to the NCI's SEER database. Third, refining and testing optimal treatment approaches in older adult mice is not trivial, as cost(s) quickly becomes a major consideration. The final limitation of our study is related to feasibility. As of March 24, 2023, a 90-week-old C57BL/6 mouse that is age-equivalent to a 65-year-old human, is $533.59/mouse at Jackson Laboratories, as opposed to a 6-week-old C57BL/6 mouse that is age-equivalent to a 12-year-old human, and is $32.07/mouse (5). Because animal costs contribute to the considerations of experimental capability, the ability to access aged mice is a final and critical limitation of our investigation.

It's notable that advanced age plays a role in many neurologic disorders including Alzheimer's disease, Parkinson's disease, and stroke, with senescence being confirmed to play a negative role in each of those diseases (44, 51, 52). We hypothesize that the similar median age of onset for GBM, Alzheimer's disease, Parkinson's disease, and stroke at ∼70 years of age is not a coincidence. As such, it's tempting to speculate that the mechanisms giving rise to an increased incidence of those age-related neurologic diseases, as well as the mechanisms that contribute to the lack of therapeutic response, may possess overlap in biological origin(s). The landmark clinical trial that standardized radiotherapy and TMZ treatment for patients with GBM limited enrollment to human subjects ≤70 years of age. There have been subsequent randomized prospective studies focused specifically on older patients with GBM that aim to discover an optimal therapeutic approach for this subgroup cohort (48). The current study suggests that there may be clinical benefit of targeting senescence with senolytics in older adults with GBM. What is not yet understood is whether IDO1, which is another biological determinant responsible for decreasing immunotherapy-dependent survival during advanced age (5), plays a redundant role with senescence in the therapeutic resistance of older adults with GBM (5, 6). To address that question convincingly and comprehensively, IDO1KO mice must be crossed with INK-ATTAC mice, followed by aging out to ∼80 to 100 weeks of age, followed by intracranial injection of syngeneic glioma cells and treatment with immunotherapy, and then compared with similarly aged and treated WT, IDO1KO, and INK-ATTAC controls. Unfortunately, that experiment will take time. Ultimately, this study supports the continued exploration of identifying and therapeutically validating age-dependent biological factors that selectively reduce survival in the largest cohort of patients with GBM: older adults.

Supplementary Material

Supplementary Data F1

Supplementary Figures

Supplementary Data T1

Supplementary Tables

Authors' Disclosures

J. Xiu reports other support from Caris Life Sciences during the conduct of the study. O. Elemento reports other support from Owkin, Freenome, Exai, and OneThree; personal fees from Champions Oncology; and personal fees and other support from Volastra during the conduct of the study. D.J. Baker reports to be coinventor on patents held by Mayo Clinic, patent applications licensed to or filed by Unity Biotechnology, and a Unity Biotechnology shareholder. T.L. Walunas reports grants from Gilead Sciences outside the submitted work. R.V. Lukas reports personal fees from Merck, Novocure, Cardinal Health, AstraZeneca, Elsevier, Medlink Neurology, and EBSCO, as well as nonfinancial support from BMS outside the submitted work. No disclosures were reported by the other authors.

Authors' Contributions

M. Johnson: Conceptualization, resources. A. Bell: Conceptualization, investigation, methodology, writing–original draft, writing–review and editing. K.L. Lauing: Conceptualization, resources, investigation. E. Ladomersky: Conceptualization, investigation. L. Zhai: Conceptualization, resources, investigation. M. Penco-Campillo: Formal analysis, investigation. Y. Shah: Investigation. E. Mauer: Investigation. J. Xiu: Investigation. T. Nicolaides: Investigation. M. Drumm: Investigation. K. McCortney: Investigation. O. Elemento: Investigation. M. Kim: Investigation. P. Bommi: Investigation. J.T. Low: Investigation. R. Memon: Investigation. J. Wu: Conceptualization, resources, formal analysis. J. Zhao: Formal analysis. X. Mi: Formal analysis. M.J. Glantz: Investigation. S. Sengupta: Investigation. B. Castro: Investigation. B. Yamini: Investigation. C. Horbinski: Investigation. D.J. Baker: Investigation. T.L. Walunas: Investigation. G.E. Schiltz: Visualization. R.V. Lukas: Conceptualization, investigation. D.A. Wainwright: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.