In vivo mouse models for adult brain tumors: Exploring tumorigenesis and advancing immunotherapy development

Abstract

Brain tumors, particularly glioblastoma (GBM), are devastating and challenging to treat, with a low 5-year survival rate of only 6.6%. Mouse models are established to understand tumorigenesis and develop new therapeutic strategies. Large-scale genomic studies have facilitated the identification of genetic alterations driving human brain tumor development and progression. Genetically engineered mouse models (GEMMs) with clinically relevant genetic alterations are widely used to investigate tumor origin. Additionally, syngeneic implantation models, utilizing cell lines derived from GEMMs or other sources, are popular for their consistent and relatively short latency period, addressing various brain cancer research questions. In recent years, the success of immunotherapy in specific cancer types has led to a surge in cancer immunology-related research which specifically necessitates the utilization of immunocompetent mouse models. In this review, we provide a comprehensive summary of GEMMs and syngeneic mouse models for adult brain tumors, emphasizing key features such as model origin, genetic alteration background, oncogenic mechanisms, and immune-related characteristics. Our review serves as a valuable resource for the brain tumor research community, aiding in the selection of appropriate models to study cancer immunology.

Brain tumor models underpin the investigation of tumor biology and the development of antitumor strategies. Consequently, ideal models should satisfy several requirements (see Figure 1): (I) genetic and; (II) tumor microenvironment (TME) features observed in patients; (III) intratumoral heterogeneity; (IV) reproducibility; and (V) cost-effectiveness and technical feasibility.¹ However, available models remain imperfect due to insufficient recapitulation of the genetic and immune landscape of human tumors, at a reasonable cost and technical feasibility. Despite the advancement of three-dimensional (3D) ex vivo culture techniques, which better reflect tumoral architecture compared to traditional 2D cultures,² in vivo models are indispensable in studying tumor–host immune interactions and developing immunotherapies. Tumor immunogenicity varies among patients with the same molecular cancer subtype and across different cancers, resulting in differential immune responses among the model diaspora.³ Tumor intrinsic and extrinsic factors influence immunotherapy efficacy. Increased tumor mutational burden (TMB) is hypothesized to promote tumor-specific neoantigens, thereby enhancing immune recognition and tumor destruction.⁴ Conversely, a hallmark of cancer is immune evasion often driven by oncogenic signaling.⁵ For instance, intrinsic factors like MYC/beta-catenin signaling promotes immune evasion through impaired dendritic cells (DCs) recruitment and T-cell function.^6,7 Alternatively, loss of tumor suppressor function can impact host immune responses against cancer. For example, TP53 inactivation results in T-cells exclusion, and PTEN inactivation suppresses autophagy within the tumor.⁵ Additionally, extrinsic factors including tissue and tumor-associated microbiomes can influence oncogenesis by altering immune responses. Notably, the human gut microbiome is implicated in clinical responses to immune checkpoint blockade (ICB).^8,9 The microbiome’s salutary effects may stem from its ability to activate innate and adaptive immunity to overcome tumor-mediated immunosuppression. Bacteroides cellulosilyticus was highly abundant in brain tumor patients who responded to ICB.⁹ Emerging evidence suggests that immune recognition of microbiota-derived molecules in the brain promotes T-regulatory cell migration to the CNS. The microbiome affects microglial maturation and function¹⁰and microbiome-derived short-chain fatty acids help regulate CNS homeostasis.¹¹ Given that these factors affect antitumor immunity, it’s imperative for researchers to judiciously choose the appropriate mouse model that aligns with their research objectives.

Figure 1.

Factors that determine the rigor of a brain tumor model. Balancing technical and fiscal considerations, models ideally reflect the pathological features observed in patients such as the tumor microenvironment milieu, intratumoral heterogeneity, and the mutational landscape, in addition to being highly reproducible across research studies. Figure is generated by biorender.

In this comprehensive review, we provide an in-depth summary of genetically engineered mouse models (GEMMs) and syngeneic models specifically for adult brain tumors, with a specific emphasis on glioblastoma (GBM). For GEMMs, we extensively cover the methods of generation (Figure 2), the genetic background, and the potential impact of these oncogenes on the TME (also presented in Table 1). Syngeneic implantation models are given special attention by addressing the cell line origins and their driving genetic alterations. Furthermore, we summarize the immune-related characteristics and immunotherapy responses in syngeneic mouse models (see Table 2) to provide a thorough overview of each model’s immune aspects.

Figure 2.

Normal cells can transform into cancer cells through spontaneous carcinogenesis or exposure to carcinogenic chemicals and other agents. GEMMs can be created by introducing genetic material into normal cells using methods such as transposon-based systems, viral vectors, or electroporation. The temporal and spatial control of these genetic modifications allows for more accurate modeling of tumorigenesis

Table 1.

GEMM for Adult Brain Tumor

	Method	Oncogenic Factors	Tissue Specific	Medium Survival	Tumor Type	Genetic Background	Ref.
SV40 T-Ag	Transgenic	SV40 T-Ag	GFAPR-drive	N/A	Astrocytomas, anaplastic astrocytomas	C57BL/6 X DBA	15
PDGFR	Lentiviral deliver, CreERT2	hPDGF-B, Cdkn2a kd	In situ injection	60–80 days	Proneural GBM	C57BL/6J and 129S1, backcross to C57BL/6J	28
	retroviral deliver	PDGF, ptenf/f, p53f/f	In situ injection	27 days		C57BL/6J	119
	Lentiviral deliver	PDGFA, STOPfl/fl-PDGFRα, TP53fl/fl	Spatiotemporal	70 days		N/A	27
TP53	Transgenic, CKO	p53∆E2–10/∆E2–10 or p53R172H, IDHWT,p53∆E5–6	GFAP-cre	256 days	GBM	C57BL/6 and 129S1/Svj	120
	Transgenic, CKO	TP53–/–, NF1fl/fl	hGFAP-cre	15 weeks	Grades 2–4 astrocytomas	C57BL/6, 129S4/SvJae,backcross to C57BL/6	31,32
		TP53fl/+, NF1fl/fl		22 weeks	GBM
		TP53fl/+, NF1+/fl		30 weeks	GBM
	Transgenic, CKO	TP53–/–, NF1fl/fl; TP53–/+, NF1fl/+	hGFAP-cre	18 weeks	Grades 3–4 astrocytoma	Mixed genetic background of C57/BL6, Sv129, and B6/CBA	35
	Transgenic, inducible CKO	Nf1fl/+;p53fl/fl;Ptenfl/+and Nf1fl/fl;p53fl/fl	Nestin-cre-ERT2	46 weeks	High-grade astrocytoma	129Svj/C57Bl6/B6CBA background	34
	CKO	TP53fl/fl, PTEN fl/+	hGFAP-cre	30 weeks	Anaplastic astrocytoma, GBM	FvB/C57Bl6	33
	CKO	TP53fl/fl, PTEN fl/fl	Olig1-Cre	90 days	GBM	C57BL/6J;129Sv	121
	CKO	TP53fl/fl, PTEN fl/fl	Olig2-tva-Cre	115 days	GBM
	MADR	NF1ko, p53ko, ptenko	In situ electroporation	N/A	Oligodendrogliomas	C57BL/6J	36
	Transgenic	NF1+/–, p53+/–	Globle	7 months	Anaplastic astrocytomas, grade IV astrocytomas	Mixed C57BL/6, 129S4/SvJae, backcrossed to wild-type C57BL/6 (B6)	31
	SB	NRAS-GV12/shTP53/shATRX	In situ injection	86 days	GBM	C57BL/6J	37
		NRAS-GV12/shTP53/Pdgfb		74 days	GBM	C57BL/6J
IDH1 IDH1	CKI	IDH1 R132H	Nestin-CreERT2	6–7 months	GBM precursors	C57BL/6	51
	SB	IDH1R132H, TP53sh, ATRXsh, NRAS G12V	In situ injection	163 days	N/A	C57BL/6	54
		IDH1WT, TP53sh, ATRXsh, NRAS G12V		70 days	N/A
	CKI	IDH1 R132H (PM/Flex)	GFAP-creERT2	N/A	N/A	B6N	53
	SB	NRAS-GV12/shTP53/shATRX/IDH1-R-132H	In situ injection	213 days	N/A	C57BL/6J	37
EGFR	RCAS/tv-a	EGFR, INK4a–ARF deletion	GFAP or nestin promoter	N/A	High-grade glioma	FVB/N, 129, and C57BL6	19
	Transgenic	V12Ha-ras, hEGFRvIII	hGFAP promoter	2-4 weeks	Oligodendrogliomas	ICR	21
	Transgenic	v-erbB(transforming allele of EGFR)	S100β promoter	N/A	Low-grade oligodendroglioma	C57B6/JX DBA/2, back-crossed into FVB/N	24
		v-erbB, p53-/+ or ink4a/arf–/+, orink4a/arf–/–			GBM, high-grade anaplastic oligodendroglioma	p53–/– is FVB/N, ink4a/arf–/– is outbred
	Transgenic, adenoviruses delivered EGFRvIII	hGFAP:V12Ha-Ras, CMV:EGFRvIII	hGAP promoter	N/A	High-grade glioma	hGFAP:V12Ha-Ras is ICR	23
	Adenoviruses delivered Cre	Col1α1-EGFRvIII,EGFRwt, Ink4A/Arf–/–, pten–/–	IC injections	6–8 weeks	GBM	N/A	122
	PiggyBac	CAG-STOPfl/fl-EGFRvIII, ptenfl/+	Nestin-cre	13 weeks	Spinal and brain glioma	Mixed FVB;C57BL/6J; C57BL/6J albino background	22
PTEN	Transgenic, adenoviruses delivered cre	hGFAP:V12Ha-Ras; Ptenfl/fl	IC injection, hGFAP promoter	2.5 weeks	GBM	hGFAP:V12Ha-Ras is ICR	23
	RCAS virus	Kras, Ntv-a PTENf/f	RCAS-Cre, RCAS-Kras	N/A	GBM	Ptenf/f mice were a mixture of 129/Sv, CJ7, and C57BL6/J; tv-a transgenic mice were a mixture of FVB/N, C57BL6, BALB/C, and 129	38
CDKN2A	RCAS virus	INK4a-ARF–/–, KRAS + AKT	NESTIN or GFAP promoter	N/A	GBM	mixture of FVB/N, C57BL6, BALB/C, and 129	42
H-ras	Transgenic	GFAP-HRasV12;p53KI/KI	GFAP-promoter	10 weeks	astrocytoma	mixed 129SvJ/C57Bl6/CD1	56
	Transgenic	V12Ha-ras	hGFAP promoter	4 weeks	Astrocytoma	ICR	55
	Lentivirus IC injection	H-RasV12, AKT,p53-/+	GFAP-cre	100–120 days	GBM	GFAP-cre and TP53–/– were maintained by backgrossing to C57BL/6	57

Abbreviations: CKI: conditional knock in; CKO: conditional knock out; MADR: mosaic analysis with dual recombinase; RCAS: replication-competent avian sarcoma-leukosis virus (ASLV) long-terminal repeat (LTR) with splice acceptor; SB: sleeping beauty.

Table 2.

Responses to Immunotherapy in Syngeneic Brain Tumor Mouse Models

Tumor Model	Genetic Background	PD1	CTLA4	Others
Gl261	C57BL/6J, carcinogen induced59	Anti-PD1 (yes)90,123–125	Anti-CTLA-4 (no)91,123	anti-CD40 (no)73
		Anti-PD1 (yes, minor)86	Anti-CTLA-4 (yes, minor)126	Anti-CD40+ COX-2 inhibitor73(synergistic, minor)
		Anti-PD1 (no)127–129	anti-CTLA-4 (yes)130	anti-OX40 (yes)125
		Anti-PD1 + anti-CTLA-4 (yes)84,126,131	Anti-CTLA-4 + anti-PDL1(synergistic)130	GVAX (yes, minor)125
		Anti-PD1+ anti TIGIT (yes, big)86	Intratumoral IL-12 + anti-CTLA-4 (YES, big)91	IDO inhibitor (no)130
		Anti-PD1+ anti TIGIT (yes, mild)127	Anti-PD1 + anti-CTLA4 + CD40131 stimulation (no synergistic)	STING antonistic93
		Anti-PD1 + TMZ (synergistic)124	Anti-PD1 + anti-CTLA4 + anti-IL6 (synergistic)131	Intratumoral IL-12 (yes, minor)91
		Anti-PD1 + focal RT (synergistic)88	Anti-PD1 + anti-CTLA4 + anti-IL6 + CD40 stimulation (overall survival)131	mImp3 peptide vaccine + PolyIC (yes)132
		Anti-PD1 + GMCI (synergistic)101		NKG2D CAR-T (yes, mild)133
		Anti-PD1 + Zika (synergistic)92		Anti-IL6 + CD40 stimulation (synergistic)131
		Anti-PD1 + GVAX (synergistic)125		Anti-IL6 (minor)131
		Anti-PD1 + anti-OX40 + GVAX (synergistic, dramatic)125		CD40 stimulation (minor)131
		Anti-PD1 + anti-TIM3 + Focal RT87 (overall survival)		Anti-TIM3 (no)87
		Anti-PD1 + anti-CTLA4 + CD40 stimulation (no synergistic)131		Anti-TIM3 + Focal RT (synergistic)87
		anti-PD1 + anti-CTLA4 + anti-IL6 (synergistic)131		Anti-PD1 + anti-TIM3 + focal RT (overall survival)87
		Anti-PD1 + anti-CTLA4 + anti-IL6+ CD40 stimulation (overall survival)131		Anti-TIGIT (no)86,127
		Anti-PD1 + anti-TIM3 + Focal RT (overall survival)87		Anti-PDL1 (yes)132
				Anti-PDL1-LNP/CDK5 inhibitor + RT (synergistic)83
				Anti-CTLA-4 + anti-PDL1 (synergistic)130
CT2A	C57BL/6J. carcinogen induced58	Anti-PD1 (yes, minor)100–102	Anti-PD1 + anti-CTLA-4 + oncolytic virus delivery of IL12 (synergistic)97	STING antonistic94
		Anti-PD1(NO)129	Anti-PD1 + anti-CTLA-4 (no)97	Anti-PDL1 (yes, minor)132
		Anti-PD1 + GMCI (synergistic)101	Anti-CTLA-4 (no)102	Neoantigen vaccine + PolyIC + anti-PDL1 (yes, synergistic)132
		Anti-PD1 + Zika (synergistic)92		Anti-PDL1 + D2C7 (synergistic)102
		Anti-PD1 + anti-CTLA-4+ Oncolytic virus delivery of IL12 (synergistic)97
CT2A	C57BL/6J, carcinogen induced	Anti-PD1 + anti-CTLA-4(NO)97		Anti-PDL1-LNP/CDK5 inhibitor + RT (synergistic)83
		Anti-PD1 + D2C7(synergistic, moderate)102
005GSC	FVB/N,C57BL/6, backcross to C57BL/657	Anti-PD1 (no)98	Anti-CTLA-4 (yes, minor)97	Oncolytic virus delivery of IL12 (yes, minor)97
		Anti-PD1 (yes, minor)97	Anti-PD1 + anti-CTLA-4 (yes)97	CCR2 antagonist (NO)98
		Anti-PD1 + oncolytic virus delivery of IL12 (synergistic)97	Anti-PD1 + anti-CTLA-4 + Oncolytic virus delivery of IL12 (overall long-term survival)97	Anti-PDL1 (yes, minor)97
		Anti-PD1 + CCR2 antagonist (yes, modest)98	Anti-CTLA-4 + oncolytic virus delivery of IL12 (synergistic)97	Anti-PDL1 + oncolytic virus delivery of IL12 (synergistic)97
		Anti-PD1 + anti-CTLA-4(YES)97
SMA-560	VM/Dk, spontaneous model	Anti-PD1 (yes, minor)102	Anti-CTLA-4 (no)91,102	Intratumoral IL-12 (yes, minor)91
		Anti-PD1 + D2C7 (immunotoxin) (synergistic, moderate)102	Anti-CTLA-4 (yes)104	NKG2D CAR-T (yes)133
			Anti-CTLA-4 + D2C7 (synergistic, moderate)102	Anti-PDL1 + D2C7(synergistic)102
			Intratumoral IL-12 + anti- CTLA-4 (YES, big)91
SB28	C57BL/6, SB GEMM(NRasV12, TP53 shRNA, and mPDGF)73	Anti-PD1 + irradiation (no synergistic)95	Anti-PD1 + anti-CTLA-4 (NO)84	FLT3L (yes, minor)95
		Anti-PD1 + anti-CTLA-4 (NO)84		anti-CD40 (no)73
				Anti-CD40 + COX-2 inhibitor (synergistic, minor)73
				Anti-PDL1 (yes, minor)95
KR158B	C57BL/6, 129S4/SvJae, backcross to C57BL/6, NF1 and TP53 KO transgenic mice31	Anti-PD1 (no)90,98		CCR2 antagonist (yes, modest)98
		Anti-PD1 + CCR2 antagonist (synergistic)98
GEM tumor cell orthotopic injection	Pten–/–, Ink4a-Arf–/–, PDGFb	Anti-PD1 + anti-CTLA4 + CD40 stimulation + anti-IL6 (synergistic)131	Anti-PD1 + anti-CTLA4 + CD40 stimulation + anti-IL6 (synergistic)131	Anti-IL6 (minor)131
		Anti-PD1 + anti-CTLA4 + anti-IL6 (same as anti-IL6)131	Anti-PD1 + anti-CTLA4 + anti-IL6(same as anti-IL6)131	CD40 stimulation (no)131
		Anti-PD1 + anti-CTLA4 (no)131	Anti-PD1 + anti-CTLA4 (no)131
QPP4	FvBXC57BL/6, QPP GEMM (Qki,TP53,PTEN triple ko)109	Anti-PD1 (no)110	Anti-CTLA4 (on)110
QPP5		Anti-PD1 (NO)110	Anti-CTLA4 (YES)110
QPP7		Anti-PD1 (yes)110	Anti-CTLA4 (yes)110
QPP8		Anti-PD1 (no)110	Anti-CTLA4 (no)110

Genetically Engineered Mouse Models

SV40 T-Antigen

The first brain tumor Genetically Engineered Mouse Model (GEMM) was established using egg microinjection of SV40-T-antigen DNA, a dominant-acting oncoprotein derived from the polyomavirus SV40.¹² SV40 T-antigen has pleiotropic functions including regulation of viral genomic replication and host cell cycle. Importantly, SV40 T-antigen can induce malignant transformation by disrupting pRb and p53 tumor suppressors.^13,14 The model’s clinical relevance is related to the minority of human brain tumor specimens that harbor DNA sequences and protein of SV40 T-antigen,¹³ suggesting that polyomavirus SV40 infection could stimulate brain tumor formation. First-generation SV40-T tumor models lacked tissue specificity, inducing cancer in various cell types. However, the second-generation models improved tissue specificity by combining CNS-specific regulatory elements with SV40 T-antigen, leading to CNS-specific malignancies.^12,15 Pathologically, SV40-T infection generated diffuse cell proliferation in the periventricular subependymal zone with diffuse parenchymal invasion. Transformed cells exhibited secondary structuring similar to the histopathological features of human astrocytomas.¹⁵

Epidermal Growth Factor Receptor

Epidermal growth factor receptor (EGFR) is an oncogene commonly expressed in adult GBM and is associated with poorer prognosis. EGFR functions as a receptor tyrosine kinase (RTK) that binds epidermal growth factor as its ligand and downstream signaling activates cell proliferation, tumor invasion and inhibition of apoptosis.¹⁶ Genetic alterations of EGFR, including gene amplification, exon deletion, and point mutation, have been found in 30–50% of human GBM, promoting aberrant EGFR activation. Specifically, EGFRvIII, consisting of an intragenic deletion in exons 2 to 7, affects its extracellular domain, causing constitutive EGFR activation in a ligand-independent manner. Clinically, EGFRvIII is commonly expressed in GBM¹⁷ and emerges as a late molecular event after EGFR amplification.¹⁸

Given EGFR’s oncogenic role in brain tumors, models expressing mutant EGFR are desirable to recapitulate human gliomagenesis. A GEMM based on RCAS (replication-competent ASLV long terminal repeat with a splice acceptor) vector system for EGFR-driven glioma has been established.¹⁹ RCAS viruses only infect tumor virus A (TVA) receptor-expressing cells which confers tissue-specificity. Murine transgenic expression of TVA in specific cell types renders these cells susceptible to infection by RCAS viruses with genetic cargos.²⁰ Paradoxically, multiple studies employing constitutively active GFAP-EGFRvIII resulted in failed or inefficient gliomagenesis, with significantly longer tumor latency of up to 60 weeks.^21,22 When EGFR-driven models integrated coexpression of other predisposing mutations, this improved gliomagenesis efficiency. Mutant-activated EGFR expression with tumor suppressor ablation (p16Ink4a, p19Arf, and PTEN) or V(12)Ha-ras potentiated higher-grade glioma formation, but not for wildtype EGFR.^21,23 These molecular alterations are nonspecific to cranial malignancies but also cooperate with spinal cord malignancies. For instance, PTEN loss is a codriver of EGFRvIII in spinal gliomagenesis rather than brain gliomagenesis.²² Related to EGFR, V-erbB is the oncogenic protein in the avian erythroblastosis virus. It acts as an auto-activating oncogene following significant deletion in its extracellular domain, and it shares similar molecular characteristics with EGFR. In tumor models and CNS stem cells, V-erbB expression, driven by the S100β promoter, results in low-grade oligodendroglioma development. Coexpression of V-erbB with INK4a/ARF or TP53 mutations accelerates gliomagenesis.²⁴ Unfortunately, inconsistent tumor formation hinders model development, despite expression of tumor-associated mutations. For example, coexpression of mutant TP53 with constitutively active mutant EGFR failed to induce glioma unless CDK4 is expressed.¹⁹

PDGFR

PDGFR including its 2 isoforms, PDGFRα and PDGFRβ, is a cell surface receptor tyrosine kinase (RTK). Binding of its ligand, PDGF, to PDGFR results in receptor dimerization, transphosphorylation, and activation of intracellular signaling pathways, PI3K/AKT and RAS/MAPK.²⁵

PDGFRα is the second most commonly amplified RTK gene in GBM, with 48% of PDGFRα-positive GBMs showing loss-of-function mutations in TP53.^17,26 A GEMM driven by PDGFRα activation and TP53 knockout was generated using a lentiviral vector with a doxycycline-inducible PDGFRα and a constitutively active Cre. This vector was intracranially injected into transgenic mice expressing conditionally deleted alleles for PDGFRα^fl/fl and TP53^fl/fl, efficiently inducing gliomagenesis. However, activating PDGFRα with wildtype TP53 lacked tumor formation.²⁷ In another model leveraging conditionally deleted Pten and TP53 alleles, retroviral-mediated intracranial PDGF delivery generated proneural gliomas with a median survival of 27 days, underscoring PDGFR’s cooperative role with other mutations. A lentiviral model introducing PDGFRβ and Cdkn2a short-hairpin RNA created GBM in mice, with a median survival of 77 days.²⁸

TP53

TP53 is a transcription factor orchestrating multiple cellular functions as a tumor suppressor, including apoptosis, maintenance of genomic stability, inhibition of angiogenesis, and regulation of cell metabolism and the TME. Clinically, alterations of TP53 are commonly seen in 25–30% of primary GBM and 60–70% of grade IV astrocytomas. TP53 mutations in GBM most frequently occur in the DNA-binding domain, namely within 6 hotspot mutation sites (codons 175, 245, 248, 249, 273, and 282).²⁹ These missense mutations result in the loss of DNA binding specificity and the gain of oncogenic functions. The p53-ARF-MDM2 pathway is dysregulated in 84% of GBM patients.³⁰ In mice, germline loss of TP53 alone favors lymphoma and sarcoma development instead of gliomas. Similarly, loss of NF1 alone is insufficient to induce gliomagenesis, but heterozygous loss of wildtype alleles in TP53 and NF1 led to gliomagenesis in mice, generating the cell line, KR158.³¹ The sequence of TP53 loss relative to NF1 loss is significant for gliomagenesis in mice; losing TP53 before or concurrent to NF1 is necessary to form malignant astrocytomas.³² Loss of other tumor suppressors, like PTEN, with TP53 loss, significantly hastens gliomagenesis, resulting in higher-grade tumors and reduced survival.³³

A neural stem/progenitor-based astrocytoma mouse model was generated by crossbreeding transgenic nestin-creERT2 mice with strains expressing inducibly floxed tumor suppressor alleles (Nf1^fl/+; p53^fl/fl; Pten^fl/+ or Nf1^fl/fl; and p53^fl/fl), resulting in high-grade astrocytomas.³⁴ Additional heterozygous loss of PTEN with TP53 and NF1 mutations accelerates grade III astrocytomas formation, whereas loss of PTEN heterozygosity and Akt activation coincided with grade IV progression.³⁵ It is well-established that neural stem cells (NSCs) contribute to the cancer stem cell (CSC) niche and tumor initiation in malignant brain tumors. Stemness was not sustained outside of the SVZ until QKi was additionally depleted with PTEN and TP53 because QKi acts as a significant regulator of NSC stemness.

Besides transgenic mice, other models utilized somatic tumorigenesis via NF1, TP53, and PTEN knock-out. Dual recombinase-mediated cassette exchange (MADR) is an in situ transgenesis method that permits stable labeling of mutant cells expressing single copy insertion of transgenic elements in precisely defined chromosomal loci. Recipient mice expressing loci with loxP and flp elements for insertion were electroporated in the early postnatal lateral ventricle with 3 plasmids: (1) smBFP2-P2A-SpCas9 for labeling and gene editing, (2) guide RNA targeting NF1, TP53, PTEN, and (3) Flp-Cre expression cassette.³⁶ MADR-mediated tumor suppressor loss increased GBM’s penetrance to 92%.³⁶ MADR confers several advantages by including single locus somatic modification, lineage tracing, dosage response of transgenes, personalized modeling, and adaptability to other strains.³⁶ Tumors generated from the sleeping beauty (SB) transposon system, with NRAS-GV12/shTP53/shATRX (NPA) or NRAS-GV12/shTP53/PDGFb (NPD), have been reported to contain oncostream structures. These contain heterogeneous cellular populations including tumor cells, ACTA2⁺ mesenchymal cells, IBA1⁺ and CD68⁺ tumor-associated microglia/macrophages (TAMs), Nestin⁺ cells, and GFAP⁺ glial-derived cells, and positively correlate with tumor aggressiveness in human.³⁷

PTEN

PTEN, a tyrosine phosphatase and tumor suppressor gene, governs numerous biological processes in cancer, including the maintenance of genomic stability, cell survival, migration, proliferation, and metabolism, and is frequently altered in brain cancer. Homozygous PTEN deletion under the control of the GFAP promoter is lethal in 6-week-old mice but not for heterozygous PTEN deletion. In the CNS, murine brain tissue expressing homozygously deleted PTEN showed significantly increased astrocytic proliferation and brain size but did not stimulate gliomagenesis. Like germline PTEN loss, somatic PTEN deletion through intracranial administration of CMV-cre adenoviral constructs into PTEN^f/f or PTEN^fl/+ mice did not induce gliomagenesis. Similar to other models, loss of PTEN with coexpressed V12Ha-Ras induced gliomas of varying grade, with heterozygously deleted PTEN generating lower grade astrocytomas, but homozygously deleted PTEN producing higher grade astrocytomas.²³ In an RCAS-based system, conditional PTEN knockout with Kras expression in neural progenitors led to GBM formation. However, astrocyte-specific PTEN knockout did not result in tumor formation.³⁸ As described previously, CNS-specific PTEN heterozygosity in an NF1/TP53-deficient astrocytoma model accelerated morbidity, and shortened survival with full penetrance of high-grade astrocytomas. Haploinsufficiency of PTEN accelerated the formation of grade III astrocytomas, while loss of PTEN heterozygosity progressed into grade IV astrocytomas.³⁵ The dependency of PTEN loss on the coexpression of other genetic alterations must be considered when generating brain tumor models studying PTEN mutations.

CDKN2A/B-CDK4/6-RB

CDKN2A locus encodes 2 tumor suppressor proteins, p16INK4a and ARF.³⁹ ARF regulates TP53 stability through interactions with MDM2 and ARF-BP1/Mule ubiquitin ligases.⁴⁰ p16INK4a inhibits CDK4 and CDK6, the D-type cyclin-dependent kinases that initiate the phosphorylation of the tumor suppressor protein, RB. When hypophosphorylated, Rb inhibits genes related to S phase progression, thus promoting G1-phase arrest.⁴¹

About 77% GBM patient specimens harbor RB pathway aberrations, most commonly in deletion of the CDKN2A/CDKN2B locus on chromosome 9p21 (55% and 53%), followed by amplification of the CDK4 locus (14%).²⁶ Leveraging the RCAS/tv-a system in mice, p16INK4a -ARF loss cooperated with Akt and Kras to increase tumor formation in glial progenitor cells. Akt and Kras alone are insufficient to induce gliomagenesis in astrocytes, but p16INKr4a-ARF loss promotes gliomagenesis suggesting that it sensitizes astrocytes to transformation.⁴² Transgenic mice expressing v-erbB with p16INK4a-ARF knockout increased the oligodendroglioma grade and penetrance.²⁴

IDH1

The IDH1 (isocitrate dehydrogenase 1) mutation is characteristically found in >80% of grade IV astrocytomas and low-grade gliomas and is associated with increased overall survival. IDH1 mutational status is included in the pathological classification of glioma specimens given its prognostic value. IDH1 mutations frequently arise before the acquisition of TP53 mutations, loss of 1p/19q, and copy number alterations in PTEN and EGFR.^43,44 IDH1 encodes isocitrate dehydrogenase 1, which catalyzes the oxidative carboxylation of isocitrate to α-ketoglutarate, resulting in the production of nicotinamide adenine dinucleotide phosphate (NADPH). This point mutation R132H is evolutionarily conserved and is localized to the substrate binding site, where it forms hydrophilic interactions with the alpha-carboxylate of isocitrate.^45–47 Despite the loss of its canonical function, it obtained a new ability to catalyze the NADPH-dependent reduction of alpha-ketoglutarate to R(–)-2-hydroxyglutarate (2HG).⁴⁸ However, the mechanism of how IDH1 R132H contributes to carcinogenesis remains unclear. It is thought that R-2-HG antagonizes α-KG by competitively inhibiting various α-KG-dependent dioxygenases, including the JmjC domain-containing histone demethylases (KDMs) and the TET (10–11 translocation) family of DNA hydroxylases that catalyze the sequential oxidation of 5-methlycytosine (5mC) to 5-hydroxymethylcytosine (5hmC), 5-formyl cytosine (5fC), and 5-carboxyl cytosine (5caC), promoting DNA demethylation.⁴⁹ Another report showed 2HG binds to CDC42 and abolishes its association with MLK3, disrupting JNK-mediated apoptosis.⁵⁰ Multiple models have been generated to elucidate the mechanisms related to mutant IDH1.

A conditional IDH1^R132H GEMM was generated by crossing IDH1^fl(R132H)/+ knock-in mice with tamoxifen-inducible nestin-CreERT2-expressing mice, which expresses Cre in the SVZ and the subgranular zone (SGZ) of the hippocampal dentate gyrus. Tamoxifen induction in 5-week-old mice produced phenotypes resembling gliomagenesis.⁵¹ Interestingly, offspring born from the parental crossing of IDH1^fl(R132H)/+ mice with noninducing nestin-Cre mice died perinatally and exhibited brain hemorrhages.⁵² A similar model driving the conditional IDH1^fl(R132H)/+ knock-in utilized GFAP-Cre–ERT2 mice to study this mutation.⁵³ Analogous to other models, IDH1 mutations co-occur with other molecular lesions, including 1p/19q codeletion, inactivating TP53 mutations, and loss-of-function mutations in alpha thalassemia/mental retardation syndrome X-linked gene (ATRX) in glioma patients. To better recapitulate patient-relevant mutations, a nonviral gene delivery system, SB transposon, established a GEMM expressing IDH1^R132H or wildtype IDH1 with TP53 and ATRX knock-down. IDH1^R132H increases median overall survival without treatment and paradoxically elicits tumor radio-resistance.⁵⁴ Clinically, this finding parallels the improved prognosis observed in patients with mutant IDH1-expressing brain tumors. IDH1 mutation status affects TME heterogeneity, with IDH1wt exhibiting greater heterogeneity.³⁷ These models offer promising potential for understanding the favorable prognostic outcomes associated with IDH1 mutations, and mechanisms of coexpressed mutations like TP53, ATRX, etc.

Hras-V12

Though V12Ha-Ras is rarer in human GBMs, mouse models of transgenic GFAP-Harvey Ras (HRas)V12 exhibit MAPK pathway activation at comparable molecular levels to human GBMs,^55,56 with higher doses of V12Ha-Ras promoting tumorigenesis.⁵⁵

To generate a V12Ha-Ras-based model, mice with a GFAP-Cre background received a lentiviral construct harboring Cre-mediated expression of H-RasV12 or Akt cassettes to the hippocampus, SVZ, and cortex. In the hippocampus and SVZ, tumor formation failed or was significantly inefficient with individual or concurrent vector administration, respectively. Upon induction of TP53 mutations, tumor formation was significantly improved with shorter tumor latency. This model resulted in the formation of a cell line, 005GSC, generated from these tumors.⁵⁷ Given this model’s dependency on TP53 alteration, a switchable TP53 functional model was developed that replaces endogenous Tp53 with 1 encoding the TP53ERTAM [estrogen receptor (ER)] fusion protein. p53ERTAM is functional only in the presence of the synthetic steroid ligand 4-hydroxytamoxifen (4-OHT).⁵⁶ This model provides novel insight into the temporal sequence of tumor suppressor inactivation and oncogenic activation in glioma. Additionally, the model’s switchable functionality yields insight into how TP53 disrupts gliomagenesis following gene rescue.

Syngeneic Implantation Model

Syngeneic implantation models have been ubiquitously used in tumor research due to their shorter latency and consistently high tumor incidence.

Carcinogen-Induced and Spontaneous GBM Cell Lines

Carcinogen-induced cell lines most commonly used in neuro-oncology include GL261 and CT-2A. Both were generated by intracranially injecting methylcholanthrene into C57BL/6J mice and were maintained by serial transplantations of tumor pieces into the syngeneic mouse strain.^58,59 The GL261 transplantation tumor model is the most extensively used GBM mouse model. However, it has been observed that GL261 exhibits a bulky growth pattern, forming a nodular mass without considerable parenchymal infiltration. These tumors rarely display necrosis and invasive growth, which are notable characteristics of human GBM. Further, it is molecularly driven by an activating Kras mutation, which is rarely observed in human GBM.^17,60 Monolayer-cultured GL261 exhibits a range of intratumoral pathology, oscillating between mesenchymal-like (MES-like) and astrocyte-like (AC-like). In contrast, GL261-GSCs cultured in neurospheres (NS) and implanted GL261-GSCs shift towards oligodendrocyte-progenitor-like (OPC-like) and neural-progenitor-like (NPC-like) states, which more closely mimic patient samples.⁶¹ Despite the model’s divergent features from the clinical data, these models possess high utility as NSs.⁶² CT-2A NS were generated by serum-free media culturing.⁶³ Unlike monolayer cultures of the parental line, CT-2A neurospheres preserve cancer stemness characteristics,⁶³ providing insights into the glioma stem cell (GSC) niche. CT-2A recapitulates several features of human GBM, including high mitotic index and cell density, nuclear polymorphism, hemorrhage, pseudopalisading necrosis, and microvascular proliferation.⁶⁴ This cell line demonstrates high intratumoral heterogeneity, displaying a differential expression of Sox9, Sox10, and GFAP across different cells within the tumor mass.⁶⁴ However, conflicting results found that cells derived from sphere cultures were less tumorigenic, had similar temozolomide resistance and showed no significant differences in vascularization or immune cell infiltration compared to nonsphere cultures. Despite displaying elevated stemness markers and reduced sensitivity to radiation (RT), these findings challenge the notion that neurospheres are superior to 2D cultures.⁶⁵ Since carcinogens stochastically induce mutations, excessive oncogenic alterations increase the model’s TMB, which is hypothesized to increase tumor immunogenicity and differ from patient samples which are not highly immunogenic. Aside from carcinogen-induced models, SMA-560 is one of the few models that spontaneously arose in VM/Dk mice.⁶⁶ It was established as a cell line through serial transplantation of spontaneous tumor tissue in syngeneic mice.⁶⁷ SMA-560 is highly tumorigenic and possesses astrocytic features but with low levels of S-100 and GFAP. However, the oncogenic driver mutation in this line is not well characterized. Two additional lines, SMA-497 and SMA-540, were also generated from VM/Dk mice, though not widely disseminated. Among these 3 lines, SMA-560 exhibits the most differentiated astrocytic features, whereas SMA-497 is the least differentiated and the least chemosensitive.⁶⁸

Targeted Induced Genetic Alteration

The KR158B cell line is a commonly used mouse GBM cell line derived from a GEMM model expressing germline global NF1^–/+ and TP53^s/+ modifications. Tumor development in vivo results in the loss of NF1 and TP53 wildtype alleles, leading to homozygous NF1 and TP53 deletion.³¹ KR158B tumor tissue demonstrated tumoral and nontumoral cellular heterogeneity. Tumor cells exhibited both fast-cycling and slow-cycling characteristics, with slow-cycling cells (SCCs) displaying features reminiscent of CSCs, similar to human SCCs. Furthermore, tumors arising from gliomasphere cultures of KR158B exhibited more pronounced GBM characteristics, including extensive infiltration and well-defined pseudopalisading necrosis, compared to those from monolayer cultures.⁶⁹ Similar to KR158, mut3 is derived from a GEMM that expresses heterozygous TP53^–/+ and conditional cre-mediated NF1 knockout in the CNS (NF1^f/+) with mut3 tumors losing both alleles.³⁴ While KR158B and mut3 models lose TP53 and NF1, another model, mGB2, is derived from PTEN and TP53 double knock-out GEMM (Tlx-CreERT2/p53-floxed/pten-floxed). The original line, mGB0, is generated by tamoxifen-induced knockout of PTEN and TP53. Fully invasive mGB0 tumors are isolated after 10–24 months and 2 serial transplantations yielded the mGB2 line.⁷⁰ mGB2 recapitulates human GBM characteristics including necrosis, invasion, intratumoral hemorrhages, and platelet aggregation.^70,71 Unlike models that are reliant on an inducible approach, another GEMM model, SB28, is generated via the SB transposon integration system.^72,73 Transfection reagent and plasmids containing SB transposon-flanked NRasV12, TP53 shRNA, and mPDGF were injected into lateral ventricles of mouse neonatal pups. With a note, NrasV12 mutation in human glioma is rarely found and tumor histology features are not well characterized.⁷⁴ Other cell lines and culture conditions have innovated model approaches to better reflect human glioma biology. QPP cell lines are generated from a GEMM harboring triple knockout of Qki, TP53, and PTEN and are cultured specifically in NeuroCult serum-free medium. The QPP mouse model and associated syngeneic cell line mirror human GBM pathology and TME characteristics.⁷⁵ To account for GSC biology, a cell line, 005GSC, was generated from the GFAP-cre TP53^–/+ mouse that received hippocampal lentiviral injections containing H-rasV12 and AKT, resulting in CNS-specific expression. NSC culture conditions were utilized to generate this model since 005GSC tumor cells only survive 2 weeks in a monolayer. The advantages of NSC culture conditions include preservation of CSC features and expression of immature neural progenitor marker, nestin. 005GSC is histologically akin to human GBM, exhibiting tumor heterogeneity, invasiveness, vascularity, and an immunosuppressive TME.⁷⁶ In comparison, bRiTs-G3s are generated in vitro first by retroviral transduction of constitutively active HRasV12 in neural stem/progenitor cells isolated from the SVZ in adult mice with homozygously deleted Ink4a/Arf loci. Transduced cells were cultured as neurospheres and implantation in mice resulted in highly invasive, hypervascular GBM-like tumors.⁷⁷ Lastly, NSCL61 is another NSC-modified GBM cell line that overexpresses H-rasL61 in TP53-deficient NSC, which is highly enriched in GSC.⁷⁸ The diverse methodology of model generation and culture conditions permits a detailed investigation of glioma heterogeneity. It is noteworthy that new in situ transgenesis methods increased the efficiency of GEMM generation. Besides in situ viral delivery of oncogenic factors, in situ electroporation is another option to rapidly create somatically transgenic mice. In utero electroporation (IUE) is a widely used method which when combined with genetic cargos, CRISPR, transposons, and MADR tools, offers multiple advantages including spatial and temporal specificity, rapid generation, flexible gene manipulation, wide applicability, and preservation of an immunocompetent context.³⁶ IUE-mediated PiggyBac overexpression of RasV12 with HA-NFIA stably generated murine oligodendroglioma.⁷⁹ Another model utilizing IUE involved electroporation of Cre-containing CRISPR-Cas9 vector into the frontal SVZ with highly efficient single-guide RNAs that target Trp53, Pten, and EGFR, with over 90% of electroporated mice developing brain tumors.⁸⁰

Immune Therapy Response

GL261

Even though GL261 is a highly immunogenic cell line harboring 4932 nonsynomymous exome mutations, ICB in this syngeneic model shows equivocal efficacy dependent on dosing strategy, tumor inoculation route, or mice strains.⁸¹ Though considered a highly immune-infiltrated tumor, TAMs increased throughout the tumor development.⁶¹ Significant synergistic antitumor responses have been reported when combining PD1 blockade with other immune checkpoint inhibitors, including CTLA-4, OX40, and TIGIT. Concurrent ICB against PD1, OX40 and GVAX demonstrated increased overall survival. Unfortunately, combined ICB with inhibitors against IDO, CTLA-4, TIM3, or TIGIT showed inconsistent or no responses in this model.^82–86 When combining ICB (PD1 or TIM3 blockade) with RT, these modalities produce synergistic therapeutic responses and survival benefits. Additional combinations like focal RT, anti-PD1 with anti-TIM3 or IDO inhibitors significantly increased survival.^82,87,88 While RT may potentiate immunotherapy responses, it can also induce the expression of immune checkpoint molecules such as PD-L1 on TAMs. Given that CDK5 has a critical role in interferon-gamma (IFNγ)-stimulated PD-L1 production, inhibition of CDK5 using PD-L1-targeting lipid nanoparticles (LNPs) that encapsulate a CDK5 inhibitor could abrogate this RT response mechanism. Combining RT with PD-L1-targeting LNPs encapsulated with a CDK5 inhibitor synergistically enhanced survival in GL261 and CT2A⁸³. Other conventional treatment modalities such as RT and chemotherapy synergize with cancer vaccines and ICB in GL261. Vaccines harboring either neoantigen peptide or whole irradiated tumor resulted in significant responses. Cancer vaccine efficacy was enhanced via intratumoral TMZ administration using convection-enhanced delivery in GL261, but not KR158B.⁸⁹ In GL261, standard TMZ dosing induced T-cell exhaustion and attenuated ICB responses.⁹⁰ This effect was not observed when using metronomic doses of TMZ and when combined with PD1 blockade, lower levels of T-cell exhaustion were observed relative to the standard dose.⁹⁰ Standard TMZ dosing may have negatively impacted immunotherapy efficacy due to iatrogenic myelosuppression. However, intratumoral IL-12 administration significantly increased survival in GL261.⁹¹ Other immunotherapeutic combinations have been evaluated in GL261 such as Zika oncolytic virus and gene-mediated cytotoxic immunotherapy, which demonstrated synergistic antitumor efficacy.⁹² Potent immunostimulatory effects have also been observed with the stimulator of interferon genes (STING) DNA sensing pathway, and STING agonists have significantly increased survival in GL261 and CT2A, but this is dependent on NK cell populations.^93,94 In sum, the ubiquitous use of GL261 in neuro-oncology has led to its application in evaluating multiple immunotherapy modalities, sustaining promising avenues of bench-to-bedside research.

SB28

SB28 is a poorly immunogenic, ICB-resistant line, with low MHC-I expression and modest CD8⁺ T-cell infiltration.⁸⁴ Relative to GL261, SB28 has significantly fewer somatic mutations at approximately 108 in total and differs in its sensitivity to ICBs.⁸⁴ When administering ICB, antitumor immune responses are not observed against SB28. PDL1 blockade alone has minor survival benefits, with no additional benefits derived from RT. Other models demonstrated that agonistic CD40 monoclonal antibody (mAb) activates myeloid cells and promotes antitumor immunity, but no response to anti-CD40 is observed in SB28. When CD40 blockade is combined with COX2 inhibition in SB28, a minor survival benefit is produced which has been linked to a reduction in immunosuppressive myeloid cells.⁷³ Aside from the direct effects of immunotherapy, biological sex immensely impacts the immune response against cancer. This is self-evident in SB28 because blocking granulocytic myeloid-derived suppressor cells has shown sex-dependent survival benefit.⁸⁵ Additionally, SB28’s immunogenicity is dependent on the tumor’s anatomic site. Unlike intracranial tumors, significant ICB responses occur against subcutaneously administered SB28. Mechanistically, subcutaneous SB28 has greater infiltration of DCs, T cells and NK cells. In an intracranial model, increased DC abundance mediated by FLT3 administration with RT or ICB failed to significantly increase survival in SB28. These findings epitomize the challenges of mounting effective antitumor immune responses in the brain, likely hindered by compensatory immune regulatory cells like T regulatory cells (T-regs).⁹⁵

005GSC

Murine 005GSC is poorly immunogenic, lacking cell surface expression of MHC-I, CD40, CD80, and CD86. In contrast to SB28, 005GSC expresses increased levels of NK-cell receptor ligands, retinoic acid early inducible-1 (Rae-1) and CD155.⁷⁶ ICB efficacy has been evaluated in 005GSC, but responses are checkpoint molecule dependent. Unlike PD1 blockade, CTLA-4 blockade displayed significant efficacy against orthotopically implanted 005GSC. αCTLA-4-specific responses were mechanistically reliant on CD4⁺ T-cell-secreted IFNγ that sensitized glioma cells to microglia phagocytosis.⁹⁶ This model has also evaluated the impact of IL-12-expressing oncolytic viruses, resulting in macrophage polarization and intratumoral T-reg reduction.^76,97 Modest survival benefits were achieved with combinatorial oncolytic virus and ICB. However, significant improvements were noted in triple combination with PD1 and CTLA-4 blockade, which was verified in CT2A.⁹⁷ Lastly, targeted approaches against immunosuppressive cells were evaluated in 005GSC. Despite inhibiting immunosuppressive myeloid cells, CCR2 antagonism failed to significantly extend survival in 005GSC, but combinatorial approaches with ICB significantly improved survival.⁹⁸ Aside from myeloid cells, other suppressive cells like T-regs limit immunotherapy responses. Significant therapeutic effects were observed in 005GSC by targeting T-regs, which were amplified with ICB.⁹³ Unlike SB28, 005GSC as a model yielded mixed immunotherapy responses and is less immunogenic relative to GL261.

CT2A

CT2A is a less immunogenic line than GL261. Immunophenotyping of the TME revealed that CT2A has the lowest level of activated and resting microglia compared to GL261, 005GSC, and Mut3. However, CT2A has the highest level of TAMs and exhausted CD8 T cells.⁹⁹ Given the high abundance of immunosuppressive cells, this model is highly relevant to immunotherapy evaluation since immunosuppression is a significant barrier to antitumor immune responses. Multiple studies have evaluated ICB using this model. For example, minor therapeutic responses were observed after the PD1 blockade, but no responses were observed after the CTLA-4 blockade.^100–102 This observation is consistent with other models in that ICB responses are often checkpoint molecule-dependent. However, mechanisms underlying these differential responses across models are unknown but may underlie poor clinical responses. Other targeted therapies aimed at antagonizing oncogenic stimuli have been evaluated in CT2A. In CT2A and SMA-560 subcutaneous models, a novel mAb, D2C7, that targets wildtype EGFR and mutant EGFRvIII revealed a significant survival benefit dependent on CD4 and CD8 T-cells. The combination of D2C7 with ICB moderately improved survival, with significant survival benefit seen against PD1. However, combining D2C7 with mAbs against T-cell exhaustion did not augment the survival benefit.¹⁰² As previously discussed, oncolytic viruses have been evaluated with ICB in CT2A. PD1 blockade with oncolytic viral therapy showed significantly improved benefits compared to either monotherapy.^92,97,101 These findings support the observations that combinatorial approaches with different immune modalities improve survival outcomes. Similarly, irradiation may enhance immunotherapy responses through its immune-stimulatory effects.^82,83 Unlike models of 005GSC, the combination of anti-PD1 with anti-GITR, a T-reg targeting antibody, does not augment antitumor responses in CT2A. However, combining these antibodies with irradiation synergized to produce significant antitumor responses.⁹³ Shifting to STING antagonism, significant therapeutic responses with STING antagonist monotherapy were noted in CT2A, similar to SB28 models.⁹³

SMA-560

SMA-560 harbors 2171 nonsynomymous exome mutations, less than 50% of the total number observed in GL261. Additionally, this cell line expresses 2 neoantigen peptides: Odc1Q129L and E2f8K272R.¹⁰³ Inconsistent ICB responses have been observed with modest responses after PD1 blockade and varied responses to CTLA-4 blockade.^91,102,104 Additionally, ICB was combined with IL-12 administration. In SMA-560, intratumoral administration of IL-12 with CTLA-4 blockade dramatically increased survival and tumor eradication, but monotherapy failed to generate robust responses. This therapeutic effect is mediated by CD4⁺ cells, significantly decreasing FoxP3⁺ T-regs and increasing effector T cells.⁹¹

KR158B

Like other models, KR158B is a poorly immunogenic line and fails to generate therapeutic responses to ICB monotherapy.^90,98,105 PD1 blockade administered with hematopoietic stem cell (HSC) transfer and lymphodepletive irradiation overcomes therapeutic resistance to ICB monotherapy.¹⁰⁵ Aside from ICB, adoptive immunotherapies have been evaluated in KR158B, and monotherapy employing adoptively transferred tumor-reactive T-cells did not elicit survival benefits. However, treatment resistance to adoptively transferred T-cells was overcome by combining myeloablation with total body irradiation, HSC transfer, and DC vaccine.^106,107 An important consideration is differential responses observed across models using the same treatment. Unlike in GL261, tumor vaccine combined with CED-TMZ treatment did not augment significant therapeutic effects against KR158B.⁸⁹

Newly Developed Syngeneic model

With a note, freshly isolated tumor spheres from GEMMs were reportedly used in immunotherapy development. RCAS-based GBM neurospheres expressing PDGF-B, Cre, Ink4a-arf^–/–, and PTEN^fl/fl were orthotopically implanted and ICB responses were achieved by targeting TAMs through IL-6 inhibition and CD40 stimulation. To study immune evasion mechanisms, GSCs were generated from novel NSCs engineered to express NF1^–/–, PTEN^–/– and EGFRvIII.¹⁰⁸ Engineered NSCs were transplanted into immunocompetent mice and immune-escaped tumors were generated following serial transplantation.¹⁰⁸ In this study, TAMs mediated tumor immune escape, with macrophage depletion via CSF1R blockade increasing overall survival. Interestingly, GSCs undergo transcriptional reprogramming consistent with myeloid “mimicry” hypothesized to help establish an immunosuppressive TME.¹⁰⁸ Other models also efficiently replicate the human brain TME. As discussed earlier, QPP cell lines are derived from the QPP GEMM (Qki, TP53, and PTEN triple ko) mouse and maintained in neurosphere culture.¹⁰⁹ The TME immunophenotypes of implanted QPP7 and QPP GEMM tumors reflected the immunosuppressive milieu of human gliomas.¹¹⁰ QPP4 and QPP8 resisted ICB, but QPP5 demonstrated responses to CTLA4 blockade. The only line to demonstrate sensitivity to PD1 and CTLA4 blockade is QPP7. Underlying this effect, QPP7 demonstrates higher immunogenicity relative to other lines likely due to the line’s higher neoantigen burden.¹¹⁰

Conclusion and Discussion

Mouse models serve a vital role in understanding tumorigenesis and identifying more efficacious therapeutic strategies. While human-derived tumor lines and immunocompromised mouse models have generated patient-derived xenografts (PDXs), they are hindered in their ability to adequately address critical questions in cancer immunology like the interplay among tumor, treatment and host immunity. As new immunotherapy strategies arise, there is increased demand for mouse models with available molecular, genetic, and immune status information along with sharing the methodology of model generation, the effects of introduced perturbations, and immunogenicity status. Despite the high model diversity of brain tumors, the field consistently relies on a minority of specific mouse models like GL261, which displays divergent immunogenicity and tumor neoantigen burden compared to patients.

Features of models exhibit strengths and weaknesses depending on the research objectives, but they generally facilitate scientific translation from bench to bedside. PDXs capture human intratumoral and intertumoral heterogeneity. However, these models typically require immunodeficient mice, and human–mouse tissue interactions can confound results. Humanized mouse models also recapitulate human tumor heterogeneity and include human immune components, ideally suitable for studying interactions between the immune system and tumors, though human-mouse tissue interactions may still confound results. GEMMs are advantageous for expressing specific mutations, with inducible functionalities. They are immune competent and avoid human–mouse interactions, but they are expensive to create, and inconsistent tumor formation limits their reproducibility. Although syngeneic implantation models surpass other models in time and cost-effectiveness, consistency, repeatability, and in vitro stability of glioma cell lines remain a challenge. Surgical interventions also disrupt the blood–brain barrier, with potential interactions with analgesic medications, tumor engraftment, and immunotherapy responses. Another critical gap remains in recapitulating brain tumor outcomes between older and younger adult patients. Models are frequently derived from adolescent-aged mice, in part, because aged mice are prohibitively expensive. Clinically, youthfulness bias in models harbors significance given that older age is associated with poorer brain tumor outcomes. Tumor-bearing brains are divergent in an age-dependent manner that impacts therapeutic efficacy.¹¹¹ Future model efforts should address therapies in youthful and aged contexts to better translate preclinical immunotherapies to the appropriate patient population.

Mouse model selection should be based on the specific scientific question since no model contains every ideal aspect. Despite immune deficiencies in PDX models, it is still a favorable tool that recapitulates human tumor heterogeneity and genetics. Specifically, preclinical therapeutics like synNotch-CAR T-cells have been evaluated in PDX GBM models, demonstrating efficient tumor killing amid neoantigen heterogeneity.¹¹² Supporting this premise that models promote translation, synNotch-CAR T-cells are being evaluated in a phase I clinical trial (NCT06186401) for patients diagnosed with EGFRvIII⁺ GBM.

In patients diagnosed with GBM, MGMT promoter methylation is associated with favorable prognoses following TMZ administration.¹¹³ Presently, human GBM cell lines and PDXs harboring MGMT promoter methylation have demonstrated increased TMZ sensitivity, but there is a paucity of GEMMs expressing MGMT methylation. Limited evidence suggests that MGMT expression may impact the immune landscape of primary GBM,¹¹⁴ but future efforts should be dedicated to generating models that express MGMT methylation.

Humanized mouse models exhibit features of human hematopoietic and immune cells. These models are advantageous in their unique human tumor-immune cell context and humanized GBM models reflect pathologic and immune-TME features observed in patients.¹¹⁵ PD1 blockade in these models reduced tumor size and immunosuppressive cells in the TME.^116,117 These models have evaluated ICB with oncolytic viruses, in addition to STAT3 inhibition, which increased TILs within the TME.¹¹⁸

Our review serves as a valuable resource for cancer immunologists who seek to select appropriate mouse models for their specific research questions. We have provided a detailed summary of GEMMs and syngeneic mouse models for studying adult brain tumors. This review encompasses model characteristics, including the generation method, genetic background, oncogenic mechanisms, and immune-related features. Within the research community, there is a pressing need for the development of new mouse models that include clinically relevant genetic backgrounds and comprehensive profiling of immune-related characteristics. These advancements would significantly contribute to the progress and translation of cancer immunology studies.

Funding

This work was supported by the National Institute of Neurological Disorders and Stroke (R01NS112315 and R01NS111033 to C.F.).

Conflict of interest statement

C.F. is a founder of iOncologi. Other authors declare no conflicts of interest.

Authorship statement

J.F., D.J., D.C., and L.F. contributed to drafting the manuscript. C.F. and D.J. contributed to revising the manuscript.