Introduction

Among primary brain tumors, gliomas are the most common with 6.5 per 100,000 individuals diagnosed annually in the United States [96]. The prognosis varies widely depending on glioma subtype and grade, ranging from a few months to several decades, with each subtype requiring different treatment strategies [81]. Brain metastases are the most prevalent brain tumors, affecting 10–40% of patients with solid cancers, primarily those with lung (19–40%) or breast cancer (6–22%), or melanoma (6–15%) [13, 18, 62, 93]. The improvements in systemic disease control and better imaging technologies have led to a significant increase of brain metastases [62]. Brain metastasis diagnosis frequently relies on clinical information and imaging combined with prior histopathological information from the primary tumor or other metastases. However, increasing knowledge of (molecular) heterogeneity and resistance mechanisms in brain metastasis as compared to primary tumor and systemic metastases has emphasized the limitations of relying solely on prior histopathology of the primary tumor, particularly in the context of targeted therapies [84]. Accurate and up-to-date molecular information of the primary tumor and its metastases, is therefore, essential to guide and monitor treatment.

The current monitoring strategies for both primary and secondary brain tumors rely heavily on magnetic resonance (MR) imaging, supplemented by clinical evaluation and, if indicated, tissue histopathological analysis. The higher-grade gliomas warrant more frequent imaging, typically every 3–6 months, burdening patients, scanner capacity, and healthcare funds [81]. Imaging alone does not provide information on the biology underpinning tumor response and treatment resistance. Notoriously challenging is the increase of contrast-enhancing lesions on MR imaging following (chemo-)radiation. This can be due to true tumor progression or may be treatment-induced, such as pseudoprogression or radionecrosis. These phenomena occur in approximately 30% of glioblastoma patients and in up to 30% of patients with brain metastasis [3, 134]. These treatment-induced imaging effects, relevant in both glioma and brain metastasis care, complicate clinical decision making as they may be misinterpreted as tumor progression [125, 131]. Advanced MR imaging modalities, such as spectroscopy, perfusion, and diffusion-weighted imaging may aid the discrimination from true tumor progression but are far from perfect [34, 77]. Needle biopsy or even repeat resection and histopathological analysis may not yield a conclusive answer, partially debit to tumor heterogeneity, and the lack of pathological definition of radionecrosis versus tumor progression. These factors emphasize the need for additional minimally invasive, accurate and affordable methods for the monitoring of brain tumors.

Over the recent decades, liquid biopsies have emerged as a promising candidate, including circulating nucleic acids, extracellular vesicles, platelet RNAs, and circulating tumor cells. These can be collected from whole blood and its derivates, but also from urine and cerebrospinal fluid (CSF). In the current era of personalized medicine, which has significantly improved the survival of cancer patients, increasing attention is being directed toward the use of liquid biopsies for disease monitoring. A decade ago, the potential of liquid biopsies has been extensively discussed in this journal, viz. Best et al. [22], which primarily focused on diagnostic capabilities of blood biomarkers for glioma diagnostics. As treatment monitoring appears to be the most relevant application of liquid biopsies in a neuro-oncological setting, we here present an initial overview of the current liquid biopsy platforms available for monitoring gliomas and brain metastases in adults. The value of liquid biopsies for molecular diagnosis and monitoring treatment response in pediatric brain tumors has been described previously [44, 124]. We provide perspective on the use of liquid biopsies for brain tumor monitoring and considerations for effective biobanking in clinical studies. The studies from this literature search are discussed per biomolecule for both diffuse gliomas and brain metastases. We discuss monitoring markers for diffuse glioma patients, which mainly focus on the distinction of tumor progression from pseudoprogression. In contrast, brain metastasis monitoring is primarily focused on differentiating radionecrosis from intracranial tumor progression. As a first step towards achieving this, a large number of studies have identified markers that differentiate between systemic and intracranial disease.

We expect liquid biopsies to complement histopathological, molecular and imaging diagnostics in the multidisciplinary treatment of neuro-oncological patients.

Circulating nucleic acids

Circulating nucleic acids encompass cell-free DNA (cfDNA) and RNA. CfDNA has a half-life of < 1.5 h, theoretically allowing for real-time depiction of tumor activity [127]. Circulating tumor DNA (ctDNA) specifically refers to the tumor-derived subset of cfDNA, which is mainly shed from tumor cells during necrosis and apoptosis. In blood, ctDNA typically constitutes < 1% of total cfDNA [32, 50]. Since it is not always possible to differentiate between cfDNA derived from normal cells and ctDNA, we will from here on use the term ‘cfDNA’.

For over a decade, cfDNA and RNA have been investigated as potential cancer biomarkers [36]. This was primarily initiated by the seminal paper of Bettegowda et al. (2014), describing the detection of tumor-derived DNA molecules in blood from 15 different tumor types, with detection rates increasing as tumor stages advance [23]. Potential purposes range from diagnosis and identification of treatment targets to monitoring for minimal residual disease [83]. As tumor-derived cfDNA detection in blood has been difficult in central nervous system (CNS) tumors, it has been suggested that the blood–brain barrier (BBB) hampers release of cfDNA in the circulation. In brain metastases, a low cerebral tumor load compared to systemic metastases may also contribute to this reduced detection efficiency.

Circulating nucleic acids in glioma

DNA

Starting with detection rates below 10% in 2014, the detectability of glioma cfDNA in blood plasma has improved over time [23]. More recent studies have detected glioblastoma somatic alterations in cfDNA in up to 83% of patients [10, 24, 37, 86, 89, 99, 114, 143]. Novel methods to detect cfDNA keep emerging rapidly, such as individual tumor-guided sequencing. Using an assay based on each patient’s individual tumor characteristics, Moulière et al. detected tumor-derived cfDNA in a cohort of IDH-wild type glioblastoma patients in CSF of 7/8 (88%) patients, in plasma of 10/12 (83%) patients, and in urine of 6/8 (75%) patients [86]. Tumor-derived cfDNA was more fragmented than non-mutant DNA in these biofluids. Selection of specifically fragmented DNA molecules may thus enhance detection sensitivity, as well as the presence of somatic mutations. Muralidharan et al. showed that in TERT-promoter mutant glioma (81% IDH-wildtype), this mutation can be detected in 63% of cases in plasma cfDNA employing digital droplet PCR (ddPCR) [89]. Moreover, in longitudinal monitoring of five patients, the TERT mutant allele frequency (MAF), i.e., the proportion of tumor-derived DNA among total detected cfDNA, decreased following resection and chemoradiation and increased upon tumor progression. Although TERT MAF did not correlate with MRI tumor volume, contrast-enhancing tumors had increased TERT MAF-values, indicating more tumor DNA leakage from a more disrupted and thereby more permeable BBB. Although not directly applicable to non-mutant TERT promoter gliomas (i.e., 40% of gliomas), this study delivers an important proof-of-concept regarding cfDNA as a blood-based monitoring marker in glioma [60]. In blood plasma of eight patients with glioblastoma, Iorgulescu et al. detected tumor cfDNA with 88% sensitivity and 99% specificity using their proprietary MAESTRO-Pool (minor allele enriched sequencing through recognition oligonucleotides) assay. Additionally, they suggest that it may help distinguish true progression from pseudoprogression as defined by follow-up MR imaging and histopathology, although longitudinal samples were collected in only seven patients [48]. Furthermore, two pilot studies noted that total cfDNA concentration in plasma might provide prognostic and monitoring value in glioblastoma regardless of whether somatic alterations were detected [10, 95]. Lower-grade gliomas are more difficult to detect as MAF decreases with glioma grade [99].

Epigenetics

CfDNA analysis also encompasses epigenetic alterations, such as methylation patterns. In patients with glioma, Nassiri et al. analyzed plasma-based DNA methylation profiles with cell-free methylated DNA immunoprecipitation and high-throughput sequencing (cfMeDIP-seq) and employed machine learning to distinguish gliomas from other brain tumor types, extracranial cancer, and healthy individuals with high accuracy (AUC 0.99; 95% confidence interval (CI): 0.96–1.00) [92]. The detection of specific types of glioma within a subgroup of common intracranial tumors was more challenging (IDH-wild type versus others: AUC 0.71; 95% CI 0.53–0.90; IDH-mutant versus others: AUC 0.82; 95% CI 0.66–0.98). Sabedot et al. developed a serum-based glioma-epigenetic liquid biopsy (GeLB) score that allowed for discrimination between 149 glioma patients and patients with various other brain tumor types with 98% accuracy [112]. Additionally, they were able to distinguish tumor progression from pseudoprogression in two patients with astrocytoma grade 2 and one patient with glioblastoma.

RNA

RNA is increasingly studied as a liquid biomarker source. Although unbound mRNA is unstable in blood, RNA that forms protein complexes, such as microRNA (miRNA), are more resistant to RNAses [9, 104]. Other RNA, such as long non-coding RNA and circular RNA (circRNA), are intrinsically more resistant to degradation and thus detectable in blood [46, 128].

MiRNA consists of small, non-coding fragments of RNA that serve as protein expression regulators. Distinct tissue miRNA profiles have been found in various cancer types, including brain tumors [72]. In contrast to cfDNA, miRNA seems to be less dependent on BBB disruption to enter the circulation or is released by ‘bystander’ cells (e.g., endothelial or immune cells) [85].

One study screened the expression of 739 miRNAs in serum obtained from patients with diffuse astrocytoma, anaplastic astrocytoma, or glioblastoma (according to the 2007 WHO classification) and age- and sex-matched healthy controls and found a panel of 108 differentially expressed miRNAs [142]. After subsequent validation in 90 astrocytoma patients and 110 healthy controls, the resulting 9-miRNA panel reached an AUC of 0.97 for the identification of astrocytoma cases, with a significant decrease of these miRNAs following tumor surgery in 73 patients. Another study aimed to identify a miRNA serum biomarker to monitor lower-grade glioma and glioblastoma patients post-treatment [85]. This group also first identified a 9-miRNA panel, demonstrating near-perfect accuracy (AUC 0.998) distinguishing between glioma patients and healthy controls. The authors analyzed 11 patients with longitudinal blood samples and found a close correlation between miR-223 and lower-grade glioma volume, and miR-320e and glioblastoma volume as measured on MR imaging. In two cases of pseudoprogression, miRNA levels did not increase. These studies suggest the potential of miRNA as a biomarker in glioma monitoring and warrant a prospective validation study.

In summary, circulating nucleic acids show promise in distinguishing between glioma subtypes, differentiating gliomas from other brain tumors, and may allow treatment monitoring. However, many studies have small sample sizes, are retrospective, and lack external validation. Furthermore, the detection of glioma-derived cfDNA in blood is challenging due to its low abundance and short half-life. Advancements in detection techniques as well as multimodal testing approaches could improve glioma monitoring. As miRNA is more stable than cfDNA and easier to detect in the blood, they show promise in glioma detection. Clues towards their monitoring potential have been found but require further investigation.

Circulating nucleic acids in brain metastasis

DNA

Brain metastasis-derived circulating nucleic acids are difficult to identify and to distinguish from those originating from systemic metastases. Several studies have explored the role of cfDNA for diagnosis and monitoring of brain metastasis in patients with metastatic solid tumors. Liang et al. detected cfDNA in blood of 28% (2/7) of cancer patients with brain metastases, with genetic alterations matching those in brain tumor tissue [67]. However, it was not specified whether these patients also had systemic metastases, which may have positively affected cfDNA detectability in the blood.

Several studies have shown that cfDNA detection is particularly difficult in isolated progressive intracranial disease. First, Aldea et al. identified plasma cfDNA in 52% (n = 28/54) of NSCLC patients with intracranial progression only (iCNS), compared to 84% (n = 83/99) with extracranial (eCNS) and 92% (n = 86/94) with concurrent extra- and intracranial progression (cCNS) [5]. Driver and resistance genetic alterations were also lower in the iCNS group compared to the eCNS and cCNS group, respectively 37% versus 77% and 73% for driver alterations and 6% versus 45% and 44% for resistance alterations. Second, Alder et al. assessed genomic alterations in serum cfDNA of 253 patients with brain metastases of various solid cancers, primarily breast cancer (12%) and NSCLC (76.4%) [6]. The proportion of detected MAFs ≥ 1% was 65.5%, 80.6%, and 73.4% (p = 0.40) for the iCNS (n = 29), cCNS (n = 160), and eCNS (n = 64) group, respectively. Additionally, the median MAF percentage per patient identified in any gene was also higher in the cCNS group (median 4.75) compared to the iCNS (median 1.6) and eCNS group (median 2.55, p = 0.003). Although the identified MAFs were lower in the iCNS patients, the researchers successfully identified unique genomic alternations in their cfDNA. Furthermore, they demonstrated a comparable ability to detect actionable mutations in cfDNA of patients with iCNS and those with extracranial progression. These findings suggest that cfDNA might find an application in characterizing molecular profiles of brain-metastases, thereby informing and optimizing treatment strategies. Kim et al. showed plasma cfDNA’s clinical utility. Among 164 patients with positive tissue or plasma EGFR mutation tests, 34 (20.7%) were detected only in plasma, leading to first-line EGFR TKI treatment in 85.3% (29/34) of patients [54]. Furthermore, they found a significant association between EGFR mutation status in plasma cfDNA and the presence of brain metastasis in 311 treatment-naive stage IV non-small cell lung cancer (NSCLC) patients, with an adjusted odds ratio of 2.73 (95% CI 1.39–5.36; p = 0.003).

Two additional studies analyzed plasma cfDNA in (brain) metastatic melanoma patients before and during anti-PD1 therapy. Seremet et al. detected BRAFV600E/K mutations in 47% (22/47) and NRASQ61/G12/G13 mutations in 38% (6/16) baseline samples but found no detectable cfDNA in 36 samples collected at baseline and subsequent assessment during therapy from ten patients with progressive intracranial disease only [115]. Lee et al. detected tumor-derived cfDNA in plasma of 53% (40/76) patients with active brain metastasis at baseline and in 24% during therapy [64]. Both studies linked undetectable cfDNA levels at baseline and follow-up with better survival. However, both studies failed to detect cfDNA in patients with isolated brain metastases.

In conclusion, previous studies illustrate the challenges of detecting isolated progressive brain metastases and monitoring brain metastasis response to therapy using cfDNA analysis in blood. Despite these difficulties, the unique genomic alterations can still be identified in cfDNA from patients with isolated intracerebral progression, offering potential insights for treatment decisions.

Epigenetics

Methylation patterns could also aid in early diagnosis in patients at risk for brain metastases and help identify an unknown primary tumor. Barciszewska found that differences in 5-methylcytocine (m5C) content between DNA of matched brain metastasis tissue and peripheral blood samples were significantly associated with primary tumor type and negatively correlated with histopathological tumor grade (from G1: highly differentiated, least malignant to G3: low differentiated, most malignant) [12]. Furthermore, m5C content in DNA of matched brain metastasis tissue and peripheral blood samples was strongly correlated. Pangeni et al. analyzed cfDNA in plasma from brain metastatic breast cancer patients and found concordant methylation patterns in cfDNA and brain metastasis tissue [97]. Zuccato et al. evaluated DNA methylation patterns of lung cancer patients with and without brain metastasis [144]. Plasma methylome-based cfMeDIP-seq enabled them to identify brain metastasis in patients with lung cancer with high accuracy (AUC = 0.80, 95%-CI 0.68–0.93). Additionally, the plasma methylome signals of patients with brain metastasis correlated well with matched brain tumor tissue methylation values. Thus, the methylation patterns in plasma may harbor potential as a diagnostic and prognostic tool for brain metastasis.

In general, it is known that cancer inflicts damage to the surrounding organ tissue. Lubotzky et al. demonstrated that cancer-induced cell death in organs is reflected in tissue-specific cfDNA methylation patterns [73]. Although levels of brain-derived cfDNA in plasma were low compared to cfDNA derived from other organs, they were measurable in most patients with brain metastases (neuron-derived in 27/29 patients, oligodendrocyte-derived in 25/29 patients, and astrocyte-derived cfDNA in all 29 patients), as opposed to patients without brain metastases or healthy controls. These findings suggest that brain cell type-specific cfDNA methylation markers may enable detection of healthy brain tissue damage resulting from brain metastasis.

RNA

Several non-coding RNAs are involved in brain metastases in breast cancer patients. Fu et al. identified circBCBM1 as a proliferation and migration-promoting circRNA in a preclinical breast cancer model and noted that it was markedly upregulated in plasma samples of brain metastatic patients [38]. Another study comparing serum samples from advanced breast cancer patients with and without brain metastasis identified miR-4428 and miR-4480, which detected presence of brain metastasis with an AUC of 0.78 [113]. Additionally, serum miR-330-3p levels were also found to be significantly higher in NSCLC patients with brain metastasis than in those without brain metastasis [129]. These studies indicate that non-coding RNAs may serve as diagnostic biomarkers for brain metastases in breast cancer. No studies have been performed on the value of non-coding RNAs in treatment response of brain metastasis.

Regarding the previously mentioned circulating nucleic acids, thus far only somatic mutation analysis of plasma cfDNA may provide value for diagnosing brain metastases and, potentially, for monitoring treatment response. Brain tissue specific cfDNA methylation patterns in blood appear to be promising, as these are directly linked to brain metastasis and not to systemic metastasis.

Extracellular vesicles

Extracellular vesicles (EVs) are defined as particles that are actively released from cells, delimited by a lipid bilayer and cannot replicate on their own [132]. They have been found in most biofluids, including blood, saliva, urine, ascites, and CSF, and are released by all cell types, although the majority originates from circulating blood platelets [33, 63]. The EVs are essential for intercellular signaling in (patho)physiological processes, including cancer. EVs are heterogeneous and have distinct biological functions based on their molecular composition, structural characteristics and size, ranging from small EVs of 30 nm to large EVs with a size up to 10 µm [51]. Tumor-derived EVs contribute to different stages of cancer development by promoting and regulating both tumor cells and the tumor microenvironment [78].

EVs are investigated as tumor biomarkers as they bear various useful properties. Due to their envelope, the content of EVs is relatively resistant to degradation. Additionally, they may be able to cross an intact BBB as shown in a mouse model with xenografted human glioma stem cells [41]. As EVs are shed by all cell types, they are believed to reflect the full molecular properties of heterogeneous tumors such as gliomas and their cell of origin [76]. They also contain a plethora of biomolecules, such as DNA, RNA, proteins, and lipids, allowing for multimodal biomarker testing. Here, we cover the brain tumor monitoring potential of EVs.

EVs in glioma

The release of EVs by gliomas was first reported in 2008 by Skog et al., who found that EVs contain tumor-specific RNAs [118]. Normal cells were shown to sequester these EVs, which could also be isolated from serum. Even without investigating the contents of EVs, their numbers can be informative. A 5.5-fold increase in total plasma EV concentration was observed in 101 glioblastoma patients compared to 29 age-matched healthy controls [105]. Within the glioblastoma group, increased EV plasma levels were significantly associated with shorter survival. EV levels correlated with MR imaging FLAIR hyperintensity volume, but not with T1 contrast-enhancing volume. FLAIR hyperintensity in glioma signifies diffuse infiltration and edema, suggesting increased permeability of the BBB and consequently elevated EV levels. Alternatively, the peritumoral cells may release additional EVs due to tumor-induced brain edema. Importantly, EV concentration did not correlate with platelet counts, indicating that the mechanism causing increased EVs is distinct from disease-associated thrombocytosis.

EV plasma concentrations may also reflect treatment response. In a subset of 34 patients with glioblastoma, EV levels decreased more in patients that underwent gross total resection compared to subtotal tumor resection. During the stable disease phase, EV concentrations remained low, whereas levels rose upon tumor progression. In two of 11 follow-up patients, the EV levels even increased 3–6 months before MR imaging revealed tumor progression, highlighting the potential of this biomarker for glioblastoma monitoring.

Batool et al. have detected several somatic alterations in EV RNA of patients with glioma [15]. They developed a novel ddPCR assay for the detection of EGFRvIII—a glioma specific deletion mutant—in blood plasma, reaching a sensitivity of 73% and a specificity of 98% in a cohort of 40 glioblastoma patients and 14 age-matched healthy controls. The plasma samples were drawn from four patients after glioblastoma treatment, in which EGFRvIII-mutant copies seemed to mirror clinical status. In CSF of three additional patients with recurrent glioblastoma treated with chimeric antigen receptor (CAR) T-cells, the number of mutant copies was concordant with response based on RANO criteria [30]. EGFRvIII was undetectable at time of CAR-T infusion, but increased dramatically in the following weeks, potentially due to treatment-initiated cell death. Subsequently, EGFRvIII became undetectable after treatment. Plasma taken from a single patient before and after infusion showed a decrease to undetectable levels. No plasma or CSF was collected at time of recurrent disease, but given that immunohistochemical analysis of post-treatment tumor tissue was negative for EGFRvIII, increased levels of EGFRvIII-mutant copies in plasma or CSF at tumor recurrence are unlikely. The loss of EGFRvIII through treatment-induced tumor evolution suggests that the detection of a single somatic alteration is not optimal for the monitoring of targeted (immune) therapy. A combination of biomarkers may be preferred.

Analogous to the EGFRvIII assay, the group of Batool et al. developed an assay for the IDH1.R132H mutation [14]. A sensitivity of 75% and specificity of 89% was reached in a cohort of 124 glioma patients (n = 80 IDH-mutant) and 9 age-matched healthy controls. Multiple follow-up samples were taken from eight patients with IDH1-mutant glioma. The patients in the disease progression group (n = 3; grades 2–4 astrocytoma) and the treatment response group (n = 3; subtypes not mentioned) displayed EV RNA-based MAF concordant with disease status. Notably, in the treatment response group, MAF increased at the start of treatment in all three patients, presumably due to increased tumor shedding. In an additional patient, the same initial spike in IDH-mutant MAF was observed, which later returned to baseline. This drop in MAF co-occurred with suspected pseudoprogression, indicating its potential utility in tumor treatment monitoring.

While EV liquid biopsies in their current state are not yet ready for adoption, EVs have promise for monitoring of glioma due to their rich EV cargo, stability and abundance in the blood. The biomarker has undergone significant development, yet a standardized protocol for EV identification is essential before it can be considered ready for large-scale prospective validation studies. As the isolation of EVs is relatively time-consuming, efforts to improve labor intensity are needed to progress its utility as a biomarker.

EVs in brain metastasis

While studies on EV-associated biomolecules specifically for monitoring brain metastasis are still scarce, growing evidence supports their diagnostic and prognostic utility, with most research focusing on miRNAs and proteins present in EVs.

The screening for brain-metastasis specific miRNAs, Wei et al. conducted RNA-sequencing from plasma-derived EVs and identified 22 differentially expressed miRNAs in plasma of lung cancer patients with and without brain metastasis, of which miR-550a-3-5p was significantly enriched in EVs from patients with brain metastasis [130]. Accordingly, Ruan et al. sequenced RNA of plasma of 42 stage IV breast cancer patients, of which 21 had brain metastasis and found that high levels of miR-199b-5p in EVs are associated with brain metastasis [111].

Additional studies have focused on size and protein cargo of circulating EVs. Carretero-González et al. found that patients with brain metastasis had lower levels of plasma EVs and higher protein concentration in small EVs (sEVs) compared to patients without brain metastasis and healthy controls [28]. Melanoma patients with brain metastasis had decreased STAT3 activation and increased PD-L1 levels in sEVs as compared to patients without brain metastasis, possibly because of systemic immunosuppression in melanoma brain metastasis patients. Rodrigues et al. demonstrated high expression of the cell migration-inducing and hyaluronan-binding protein (CEMIP) in brain metastasis tissue and their secreted EVs in plasma in contrast to tissue from systemic metastasis [110]. Li et al. compared EV-associated proteins in 42 metastatic lung cancer patients, 25 locally advanced lung cancer patients, and 5 healthy controls and identified 120 differentially expressed EV-associated proteins in 28 brain metastatic lung cancer patients, of which MUC5B and SELL could be used as diagnostic biomarkers (AUC 0.774 and 0.720, respectively) [66]. Both miRNA and proteins in plasma EVs are therefore promising biomarkers for brain metastasis.

In addition to aiding in brain metastasis diagnosis, plasma EV content has been linked to disease progression and survival. Chen et al. found that elevated EV-associated integrin β3 levels in 75 lung carcinoma patients who received whole brain radiotherapy for brain metastasis were associated with poorer intracranial control (HR: 1.22 per 1 ng/mL increase; 95% CI 1.012–1.46; p = 0.037) and reduced overall survival (HR: 1.15 per 1 ng/mL increase; 95% CI 1.01–1.32; p = 0.04) [29]. These results indicate that proteins in EVs, in particular integrin β3, may serve as prognostic biomarkers for brain metastasis.

In conclusion, the diagnostic and prognostic potential of plasma EVs has been demonstrated in brain metastasis patients. Further research is needed to confirm the diagnostic and prognostic role of EVs and to establish whether they can be used for treatment monitoring.

Platelet RNA

The blood platelets play a significant role in the progression of systemic cancer. Direct and indirect contact between platelets and tumor cells of colon and breast cancer facilitates transition towards a more mesenchymal phenotype, enabling tumors to invade surrounding tissues and metastasize [59]. The platelets even help circulating tumor cells to evade the immune system by forming a physical shield around them [70, 101]. Despite the recognized role of platelets in cancer pathogenesis, platelet counts have inconsistently shown utility as biomarkers in glioma patients [7]. Consequently, research efforts focus on analyzing platelet content rather than platelet counts. Platelets and tumor cells have demonstrated ‘cross-talk’. The transcriptome and proteome of platelets incorporate a tumor signature that includes tumor-specific mutant transcripts [55, 91, 94]. Possibly, this occurs via alternative splicing of pre-mRNAs from megakaryocytes and sequestration of (circulating) tumor-derived RNA molecules. As platelets have a life span of ~ 7 to 10 days, they may provide a real-time snapshot of the tumor status. Additionally, they are practical biomarker candidates: platelets are widely abundant, can be isolated from only 4–6 mL of whole blood, and the isolation procedure is simple. Before processing, whole blood can be stored for as long as 48 h at room temperature, enabling for sample shipment [19].

Over the last decade, our group has developed a platelet mRNA biomarker platform that enables distinction between cancer, non-neoplastic disease, and healthy controls with high accuracy [19,20,21, 47]. In a cohort of 126 patients with one or multiple brain metastases primarily from NSCLC (n = 85), and 89 patients with glioblastoma, we demonstrated that platelet RNA profiles can be used to discriminate brain metastases from glioblastoma with an AUC of 0.84 (95% CI 0.76–0.92; p < 0.001) [119]. A subsequent study investigating patients with 18 different tumor types showed that RNA profiles of 93 patients with brain metastasis were distinct to profiles of 299 patients with a similar primary tumor without brain metastasis. Furthermore, platelet RNA profiles of brain metastasis patients had similarities to those of patients with gliomas, suggesting that the platelet transcriptome is influenced by both the primary tumor and the metastatic site [47].

The platelets may also be of value for glioblastoma treatment monitoring. Distinguishing true progression from pseudoprogression, a platelet RNA-based glioblastoma-specific classifier has demonstrated an AUC of 0.86 (95% CI 0.70–1.00; p < 0.012) [119]. In a follow-up cohort of 48 glioblastoma patients, the classifier score seemed to mirror disease course, although the correlation varied across patients. In some patients, the score correctly indicated tumor progression before radiological progression occurred. These properties render platelet RNA potentially valuable in therapy monitoring, which is currently being evaluated in the multicenter, prospective PREDICT-study.

A disadvantage of the use of platelet RNA as a biomarker source is that the mechanism of platelet ‘education’ is not completely identified. Recent findings by Karp et al. suggest that platelets are indirectly influenced by other circulating cells rather than sequestering tumor transcripts [53]. Conversely, previous studies have demonstrated transfer of EGFRvIII RNA from glioblastoma to platelets, and RNA transfer between platelets and other cells [42, 61, 94, 107]. Future studies should aim to elucidate these contradictions, through improved wet-lab protocols as well as computational cleaning methods. Platelet RNA remains a promising biomarker in the monitoring of brain tumors, warranting large-scale prospective validation studies.

Circulating tumor cells

Circulating tumor cells (CTCs) originate from primary or metastatic tumor sites and offer a minimally invasive approach to retrieve information about tumor characteristics [68]. Since most tumors metastasize hematogenously, tumor release of CTCs is a proxy of its metastatic potential. However, CTCs require adaptive mechanisms to survive, resulting in low levels in blood, with only a rare subset capable of initiating brain metastases [58]. Recent technological advances, such as immunomagnetic bead methods, microfluidic technologies, and high-throughput sequencing technology, have significantly improved the detection and characterization of CTCs associated with brain tumors [4, 52, 138].

CTCs in glioma

Compared to cell-free nucleic acids and EVs, the field of CTCs in glioma has advanced less rapidly due to the technical difficulties in enriching CTCs from glioma. As glioma does not express the epithelial cell adhesion molecule (EpCAM) typical for carcinomas, standard EpCAM-based CTC isolation methods cannot be employed. This has resulted in the development of various protocols that employ different markers and isolation methods [43]. Most studies have focused on improving enrichment of glioma CTCs rather than correlating them to clinical status, a purpose for which current methods may not be sufficiently mature. The reported blood detection rates of CTCs in glioma patients range from 20 to 84%, with the number of cells detected in a tube of blood often in single digits [40, 75, 87, 88, 102, 123, 140, 141]. Several studies have investigated the prognostic and monitoring potential of CTCs in glioma, of which we describe here the most recent and the most important ones.

Using a telomerase reverse transcriptase-based assay, Zhang et al. were able to detect CTCs in plasma of 106 glioma patients with 83% sensitivity [141]. The presence of postoperative, but not preoperative CTCs was associated with poor prognosis, and a significant decrease in the number of CTCs following tumor resection was observed. Unfortunately, CTC levels during the chemotherapeutic treatment phase, and thereafter, are not described. In addition, Sullivan et al. report 39% sensitivity in detecting CTCs in plasma of 33 glioblastoma patients [123]. A significant difference in CTC counts was observed between patients with stable disease and patients with disease progression. Notably, all detected CTCs had the mesenchymal subtype, indicating that a molecular transition of glioblastoma cells may be necessary to enter the circulation. Consequently, the blood may only contain a subset of tumor-derived cells, not fully reflecting the heterogeneity of the primary tumor. A further increase in sensitivity might enable CTCs to play a role in glioma disease monitoring.

Although we did not find any studies that formally investigated CTCs as a biomarker in the monitoring of glioma, some studies report a decrease in CTCs after tumor resection and an increase upon progression. More research is needed to determine their monitoring value.

CTCs in brain metastasis

In systemic metastasis, CTCs have shown utility in treatment response assessment and post-treatment surveillance [35, 122]. Unfortunately, no studies have specifically evaluated CTCs for monitoring brain metastasis. Most research on CTCs has focused on cellular and molecular adaptations that allow them to cross the BBB and blood‐CSF barrier, colonize the brain microenvironment and form brain- or leptomeningeal metastases [58].

The translational research in human cell lines and mouse models has identified specific molecular profiles that differentiate CTCs associated with brain metastasis from those associated with systemic metastases from the same primary tumor [17, 25, 26, 57, 58, 82, 98, 103, 106, 126, 135, 139]. Building on these findings, several studies have investigated CTCs in blood of patients with brain metastasis. In a mixed cohort of brain metastatic patients with NSCLC, breast cancer, or melanoma, Loreth et al. found that CTCs in blood mostly expressed CD74 and CD44, unlike matched brain tumor tissue [71]. This suggests plasticity of CD44 and CD74 expression on CTCs that survive in blood and penetrate the BBB. Aljohani et al. reported mutations in KEAP1-NRF2-ARE pathway genes in CTCs in blood of lung cancer patients with brain metastasis [8]. In melanoma patients, the RPL/RPS-gene signature in CTCs has been linked with the onset of brain metastasis [27]. These characteristics across various primary solid tumors may provide leads for future development of a CTC-based brain metastasis monitoring platform.

Several studies have investigated overall CTC counts as a prognostic tool for brain metastasis. In NSCLC patients with oligo-metastatic brain disease and patients with concurrent systemic metastases, Hanssen et al. found that presence of ≥ 2 and ≥ 5 CTCs/7.5 mL predicted poorer survival [45]. Naito et al. found no association between pre- and post-treatment CTC counts and brain metastasis presence in small-cell lung cancer (SCLC) patients before, during and after receiving chemo(radio)therapy, although only seven patients had brain metastasis [90]. In contrast, the LANDSCAPE trial showed that early CTC clearance in HER2-positive breast cancer patients with brain metastases predicted intracranial response to HER2-directed therapy combined with chemotherapy and overall survival [100]. However, both responders and non-responders showed a decline in CTC count after one cycle, and only 15% of patients had brain-only metastatic disease, suggesting that reduced CTC levels in blood may reflect overall disease control, rather than intracranial tumor control.

In conclusion, CTC-based liquid biopsies are increasingly implemented in various cancer types. However, the applicability for brain tumor monitoring is insufficiently studied. Important steps have been made in differentiating brain metastasis-associated CTCs from systemic metastasis-associated CTCs. Despite this, further research is required to develop a clinically applicable brain metastasis-specific CTC signature. This may pave the way to CTC-based monitoring of brain metastases.

Role of myeloid cells/monocytes

Currently, liquid biopsies in brain metastasis patients primarily focus on diagnosis and prognosis. There is a critical need to expand their role in monitoring tumor progression, especially since the incidence of cerebral radiation necrosis is rising, which may be hard to distinguish from tumor progression on MR imaging. A recent study proposed an immunosuppressive marker as a surrogate for differentiation between patients with active and inactive brain metastases [121]. Flow cytometry was used to quantify monocytic myeloid-derived suppressor cells (Mo-MDSC) from peripheral blood in 22 patients with biopsy-proven active brain metastasis or radiation necrosis. They concluded that the HLA-Dr-Vnn2 Index could reliably discriminate recurrent brain metastasis from radiation necrosis. The patients with brain metastasis recurrence showed significantly increased CD14 + HLA-DRneg/low Mo-MDSCs and reduced expression of Vnn2 on circulating CD14 + monocytes compared to those with radiation necrosis. This study has set a foundation for further research finding biomarkers for post-treatment brain metastasis monitoring.

Cerebrospinal fluid as liquid biopsy source

The BBB restricts the release of biomolecules into the bloodstream, making CSF a potentially superior compartment for studying brain-derived biomarkers. However, CSF collection via lumbar puncture is more invasive than venipuncture and not always possible due to the risk of cerebral herniation. While intraventricular devices such as an Ommaya reservoir enables easier repeat sampling, the surgical placement carries risks such as infection and hemorrhage. Despite these challenges, CSF-derived biomarkers for brain cancer monitoring have been studied; we will highlight the most relevant studies.

Glioma

In glioma patients, several CSF-based biomarkers are being investigated. One promising candidate is D-2 hydroxyglutarate, a metabolite of mutant IDH, whose CSF concentration differs between patients with IDH-mutant and IDH-wild-type gliomas, as well as pre- and post-resection of IDH-mutant tumors [39, 109]. Ongoing studies will show whether this relatively new CSF biomarker is suitable for treatment monitoring of IDH-mutant gliomas.

Furthermore, several studies have shown the potential of liquid biopsies using cfDNA from CSF. Iser et al. demonstrated molecular-based classification of 75% (n = 24) of glioblastoma patients and 53% (n = 10) of patients with other types of glioma using targeted next-generation sequencing of cfDNA from CSF [49]. They also included patients with recurrent/progressive glioma (n = 14). Klinsing et al. evaluated the diagnostic ability of somatic copy number alterations from CSF cfDNA in brain tumor patients, of which six (26%) glioma patients [56]. They were able to differentiate tumor recurrence from other potential causes of deterioration, e.g., postoperative infection, during surveillance in one patient with a previous history of glioma. Interestingly, a recent case-report indicated the value of CSF-based cfDNA methylation analysis using nanopore sequencing for the diagnosis of a difficult-to-diagnose intracranial lesion [120]. Additionally, Afflerbach et al. demonstrated the potential of nanopore sequencing on cfDNA from CSF for the classification of brain tumors [1]. Most patients in their study were children and adolescents (n = 91/129) with pediatric brain tumors. Although a minority of included patients were adults, this study highlights the utility of cfDNA Nanopore sequencing for brain tumor monitoring. This approach shows particular promise in the early detection of tumor recurrence and identification of molecular characteristics, as evidenced by two follow-up patients. Several other studies report on the detection of glioma-derived DNA in CSF, showing that it can be used to track tumor evolution, which lays the foundation for CSF-based tumor monitoring in future [56, 80, 108, 116].

Brain metastasis

In brain metastasis patients, CSF is deemed a more reliable source than plasma for detecting tumor-derived cfDNA [31, 74, 136]. Wu et al. found that CSF cfDNA had significantly higher concordance with brain tumor tissue than plasma cfDNA (99% vs. 67%) in patients with single brain metastasis, although performance was similar in patients with multiple brain metastases [136]. Additionally, Li et al. observed that cfDNA changes in CSF correlate with intracranial response, while plasma cfDNA seems to reflect extracranial disease response [65].

In summary, CSF collection is more invasive than blood, making it a theoretically suboptimal compartment for frequent liquid biopsies. Nevertheless, CSF is anatomically closer to brain tumors and possibly reflects intracranial disease better than plasma. If CSF-based liquid biopsies, despite all drawbacks, reach excellent performance in clinical practice, the added value of CSF-derived biomarkers could overcome these drawbacks.

Focused ultrasound-enhanced liquid biopsies

An alternative approach to enhance biomarker detection may be to enrich the repertoire of tumor-derived biomolecules in the blood, for example using focused ultrasound (FUS). Originally designed to temporarily open the BBB to enhance delivery of chemotherapeutic agents, the technique was found to increase the blood concentrations of cfDNA by up to 2.6-fold, neuron-derived EVs by 3.2-fold, and brain-specific protein S100b by 1.4-fold [11, 79, 137]. TERT-mutation cfDNA blood plasma levels rose up to 5.6-fold in a cohort of four patients, indicating that focused ultrasound increased tumor-derived cfDNA [137]. With this proof-of-concept, the LIBERATE-trial, in which low-intensity FUS will be used for cfDNA-based liquid biopsy in glioblastoma, has been initiated. Further research, including this trial, should further elucidate the applicability and efficiency of FUS-enhanced liquid biopsies [2].

Conclusions and considerations for future research and biobanking

The ideal brain tumor biomarker is reliably detectable, stable after collection, and does not require complicated processing (Figs. 1 and 2). Currently, this best describes EVs, miRNA, and platelet RNA (Table 1). Although total cfDNA levels may also prove informative, tumor-derived cfDNA is difficult to detect in blood and repeated CSF collection is less desirable due to its invasive nature. Following cfDNA collection either in blood or CSF tubes, cfDNA has a half-life of less than 1.5 h, making timely processing a challenge. Brain tumor-associated CTC detection in blood is also difficult and time-intensive. In Tables 2 and 3, we provide an overview of the most important studies.

Fig. 1
figure 1

Overview of blood-based biomarkers in glioma and brain metastasis, including ctDNA (1), RNA (2), platelet mRNA (3), extracellular vesicles (4) and circulating tumor cells (5). a Primary tumor with cancer cells which intravasate (CTCs), cells secrete EVs, necrotic and apoptotic cells shed nucleic acids (ctDNA, RNA) in circulation; b Brain metastasis formation by CTCs; c Brain metastasis derived CTCs; d Glioma derived CTCs, EVs, ctDNA, RNA

Full size image
Fig. 2
figure 2

The ideal theoretical liquid biopsy for monitoring treatment response in brain tumors as compared to MR-imaging

Full size image
Table 1 Overview of analytes in brain tumor liquid biopsies and their monitoring potential
Full size table
Table 2 Clinical studies on liquid biopsies in glioma
Full size table
Table 3 Clinical studies on liquid biopsies in brain metastases
Full size table

The molecular information obtained from liquid biopsies could complement current clinical, histopathological, tissue molecular and imaging diagnostics tests. Despite extensive research on liquid biopsy in brain tumors over the past decades, no platform is ready for clinical implementation. Progression monitoring using liquid biopsy in neuro-oncological diseases is still in its infancy, with most studies pursuing diagnostic or prognostic goals only. Studies assessing diffuse gliomas show primarily potential in the higher-grade tumors, whereas lung and breast cancer are the most frequent tumor types assessed in the context of brain metastases. The sampling time points are often limited to pre- and post-resection in glioma and to the first weeks of systemic treatment in brain metastasis. We foresee the main clinical applicability of liquid biopsies in brain tumors to be therapy monitoring, with markers reflecting treatment response, especially when treatment-related effects (pseudoprogression/radionecrosis) are suspected. Treatment-naïve detection of brain tumor-associated biomarkers is a prerequisite for future therapy-monitoring and aligns with numerous studies searching for diagnostic markers. However, the next step—evaluating monitoring potential of liquid biopsies—is a significant challenge. It requires longitudinally collected samples, extensive patient and sample annotation, and sufficient research funds. To demonstrate clinical potential, biomarkers subsequently require validation in large, prospective, preferably multicenter cohorts. The final step towards clinical implementation is an interventional trial to confirm that the proposed biomarker informs clinical decisions including treatment continuation in case of pseudoprogression/radionecrosis or alternative treatment options in case of true tumor progression. MR imaging, with all its modalities, may be complemented by liquid biopsies.

A prospective biomarker validation study should be meticulously planned to ensure proper evaluation of performance in a clinical context. For a fair comparison, time points of the investigational test should be harmonized as much as possible with the gold standard test. For a progression versus pseudoprogression design, this would mean collecting blood at the moment of MR imaging evaluation. Moreover, a well-defined ‘gold standard’ outcome measure should be selected, such as RANO-evaluation of imaging, clinical characteristics, and/or systematically collected tissue biopsies [69, 133]. The clinical feasibility of the investigational test needs to be considered, and its clinical relevance should be defined. When studying tumor molecular evolution, sample collection is preferred when this will have consequences for treatment, including potential inclusion in experimental studies. The results of a biomarker test should be included in the multidisciplinary decision about the patient’s treatment, jointly made by neurologists, neurosurgeons, medical oncologists, radiotherapists, and pathologists.

When collecting samples, the costs for patients and institutions need to be weighed against the potential benefits of isolating multiple markers at various time points. Incorporating a broader range of markers allows for multimodal platforms that combine a panel of biomarkers with the purpose of increasing sensitivity and specificity. As some analytes require rapid processing after collection, the choice of biomarkers should also depend on expected logistics. The measurement of unstable biomolecules can be unfeasible in multicenter research, especially if the samples need to be transported to a central processing facility. The preservatives exist for a subset of biomolecules but amount to an extra processing step and additional costs. Furthermore, following the wide adoption of techniques such as next-generation sequencing, machine learning can play a significant role in interpreting extensive datasets and drawing meaningful conclusions.

Finally, simple, standardized protocols should be used as much as possible for each analyte for the benefit of reusability, comparability and clinical adoption. Ideally, the protocol can be performed in a standard diagnostic laboratory in non-academic hospitals of a medium size. To preserve valuable samples, leftover material should be kept in a biobank for long-term storage and reuse in future studies. We encourage world- or nationwide initiatives in central registration of samples to make maximal use of collected samples and avoid unnecessary spending of resources collecting new samples. An example of such combined efforts is the Brain-Liquid Biopsy Consortium [16, 117].

In conclusion, liquid biopsies have the potential to monitor progression of brain cancer using novel molecular biomarkers, and to inform treatment decisions. To take full advantage of this potential, focus will need to shift to large-scale prospective validation, preferably in a collaborative manner. Well-considered choices need to be made when designing these studies to ensure maximum impact.

Methods

A comprehensive PubMed search was conducted from its inception in 1996 through January 15, 2025. Language was restricted to English. Key search terms included synonyms of “liquid biopsy” and biomarker subtypes, such as “cfDNA”, extracellular vesicles”, and “circulating tumor cells”, as well as “pseudo-progression”, "radionecrosis", “glioma” and “brain metastases”. The references from identified studies were reviewed for any additional relevant research.