H3K27-altered diffuse midline glioma: a paradigm shifting opportunity in direct delivery of targeted therapeutics

Abstract

Introduction:

Despite much progress, the prognosis for H3K27-altered diffuse midline glioma (DMG), previously known as diffuse intrinsic pontine glioma when located in the brainstem, remains dark and dismal.

Areas covered:

A wealth of research over the past decade has revolutionized our understanding of the molecular basis of DMG, revealing potential targetable vulnerabilities for treatment of this lethal childhood cancer. However, obstacles to successful clinical implementation of novel therapies remain, including effective delivery across the blood-brain barrier (BBB) to the tumor site. Here, we review relevant literature and clinical trials and discuss direct drug delivery via convection-enhanced delivery (CED) as a promising treatment modality for DMG. We outline a comprehensive molecular, pharmacological, and procedural approach that may offer hope for afflicted patients and their families.

Expert opinion:

Challenges remain in successful drug delivery to DMG. While CED and other techniques offer a chance to bypass the BBB, the variables influencing successful intratumoral targeting are numerous and complex. We discuss these variables and potential solutions that could lead to the successful clinical implementation of preclinically promising therapeutic agents.

1. Introduction

Among pediatric high-grade gliomas, H3K27-altered diffuse midline glioma (DMG), formerly known as diffuse intrinsic pontine glioma (DIPG), is associated with an abysmal prognosis and 5-year survival rate (<2%) [1, 2]. About 300 children are diagnosed with brainstem DMG per year in the United States, comprising 10% of all primary pediatric brain tumors and 80% of brainstem tumors in this age group, most commonly in middle childhood (5–10 years old) [1, 3, 4]. Reflective of the anatomical localization and diffusely infiltrative growth pattern (Figure 1a), patients with DMG may present with a wide range of symptoms including headache, nausea, and the triad of cranial nerve dysfunction (i.e. diplopia and facial asymmetry), cerebellar signs (i.e. ataxia and dysmetria), and long-tract signs (i.e. hyperreflexia and paresis) [5, 6]. Obstructive hydrocephalus is a common late manifestation of disease [5, 7].

Figure 1. Location and implicated anatomy of H3K27-altered diffuse midline glioma (DMG).

(a) DMG tumors were traditionally believed to be pontine, as indicated on the (Ai) sagittal T1 and (Aii) axial T2 MRIs. However, improvements in imaging, biopsy and molecular biology revealed these tumors throughout the midline structures including the thalamus, midbrain, pons, medulla, cerebellar peduncles, and cerebellum. These regions (black dashed line) can be appreciated on both sagittal (Ai) and axial (Aii) planes. (b) The diffuse nature of DMG brainstem gliomas has the potential to impact important neurological structures. (Bi) Projection fibers carrying information from the spinal cord must pass through the midline structures to reach the cortex and may be disrupted by tumor infiltration. (Bii) Additionally, cranial nerves and other nuclei located as displayed by the cross-section can be disturbed by the tumor causing malfunction. Figure created in BioRender.com.

Previously, diagnosis of brainstem DMG was made based on characteristic imaging features on magnetic resonance imaging (MRI), including diffuse T2-weighted signal with ill-defined margins, occupation of >50% of pontine diameter, and encasement of the basilar artery [5, 8]. More recently, it has been established that DMG has the proclivity to migrate along adjacent midline structures including the thalamus, cerebellar peduncles, and spinal cord [9]. Due to its highly elo-quent location (Figure 1b), any opportunity for meaningful surgical resection of DMG is not possible [10]. The established standard of care for these tumors is radiotherapy, although it is mostly palliative [11, 12]. Many patients ultimately enroll in a clinical trial following radiotherapy, and while there are numerous trials ongoing, perhaps the drug attracting the most attention of late is the DRD2/3 antagonist ONC201, which we will discuss in this Special Report.

Seminal studies over the past decade have revolutionized our understanding of the molecular basis of DMG, revealing potential targets for treatment of this lethal childhood cancer. However, obstacles to successful clinical implementation of novel therapies remain. In this invited review, we discuss direct drug delivery via convection-enhanced delivery (CED) as a promising treatment modality for DMG and outline a comprehensive molecular, pharmacological, and procedural approach that may offer hope for afflicted patients and their families.

2. Molecular features and therapeutic targets

2.1. Molecular features of H3K27-altered DMG

Historically, surgeons and oncologists alike have been reluctant to obtain DMG tissue due to concerns over procedure-associated morbidity [13]. Over the last two decades, multiple neurosurgical groups have demonstrated the safety of image-guided stereotactic biopsy of these tumors, increasing availability of previously scarce tumor specimen for further study [14-17]. Pioneering research of available biopsy and autopsy samples has revealed somatic oncogenic mutations affecting chromatin regulation and has greatly improved our understanding of the unique biology of these tumors. Most DMGs harbor gain-of-function mutations in gene coding for select histone subunit 3 (H3) isoforms, specifically H3F3A and HIST1H3B, encoding histone H3.3 and H3.1, respectively. Most mutations result in a lysine residue substitution for methionine at position 27 of histone H3 (H3K27M) [18-20].

An array of characteristic genetic and epigenetic alterations in pediatric high-grade gliomas including DMG has subsequently been identified in recent years [21-23]. To reflect the practical and conceptual importance of these groundbreaking findings, a new family of tumor types, entitled ‘pediatric-type diffuse high-grade gliomas,’ was added to the 2021 WHO Classification of CNS Tumors, and DMGs are now designated as ‘diffuse midline glioma, H3K27-altered,’ recognizing that non-K27M mutations in these genes exist in a small subset of these tumors [24].

2.2. Epigenetic modifiers

Genome-wide loss of H3K27me3 is a hallmark of H3K27-altered DMG and is believed to result in de-repression of developmental transcriptional programs controlled by bivalent enhancers [25-30]. Preclinical studies using the histone demethylase inhibitor GSKJ4 have demonstrated promising antitumor activity specific to H3K27M-mutant tumors [31, 32]. Histone deacetylase inhibitors such as vorinostat and panobinostat were shown to be effective against multiple DMG cell lines in vitro and at present have moved into clinical trials as both single (NCT03566199) and combination agents (NCT02420613, NCT01189266, NCT04341311, and NCT03632317) [33-35]. While activation of bivalent enhancers is believed to contribute to DMG tumorigenesis, these tumors likewise display a dependency on polycomb repressive complex 2 (PRC2) [36, 37]. A variety of small-molecule inhibitors targeting transcriptional regulators, including the PRC2 methylase EZH2 and bromodomain family proteins, are currently being exploited to identify other actionable epigenetic vulnerabilities [38-40].

2.3. Tyrosine-kinase inhibitors

Additional genetic alterations are common in DMG, including those affecting tumor protein p53 (TP53), platelet-derived growth factor receptor alpha (PDGFRA), activin receptor type 1 (ACVR1), Myc proto-oncogene family members (MYC, MYCN), and others [22, 23, 40]. Drug screens targeting some pathways impacted by these mutations have paved the way for multiple clinical trials, notably involving multi-tyrosine kinase inhibitors (e.g. dasatinib [NCT02233049, NCT01644773, NCT00996723], ribociclib [NCT03355794, NCT02607124]) and mTOR inhibitors (e.g. everolimus [NCT03355794, NCT05476939, NCT02233049, NCT03632317], temsirolimus [NCT02420613]).

2.3.1. Cell-based therapies

Tumor-specific surface molecules are also being harnessed to transport a cytotoxic payload or cellular therapy to DMG cells while minimizing off-target side effects. B7-H3 has been shown to be overexpressed in the majority of high-grade gliomas, including DMG [41]. In a phase-I clinical trial of the radiolabeled antibody ¹²⁴I-8H9, which targets B7-H3, the therapy was safe with a promising volume of distribution (V_d) in the targeted area (NCT01502917) [42, 43]. To determine the efficacy and safety of ¹³¹I-omburtamab, a B7-H3 antibody conjugated to a different isotope of iodine, Souweidane and colleagues are in the final stages of initiating phase-I/II study for children with DMG that have not progressed following radiation (NCT05063357). Since the early 1990s, the non-canonical IL-13 receptor subunit alpha 2 (IL-13Rα2) has been used in a similar fashion to transport exotoxin chimeric molecules to pediatric and adult high-grade gliomas [44-46]. Heiss et al. recently completed the first phase-I clinical trial of IL13-PE38QQR in DMG with no patient experiencing a dose-limiting toxicity (NCT00088061) [47]. The disialoganglioside GD2 is highly expressed in tumors of neuroectodermal origin, and an anti-GD2 CAR-T study portrayed positive results in DMG in vitro and in vivo models [48]. Recruitment for a first-in-human phase-I clinical trial (NCT04196413) is ongoing, but the authors have reported promising results from the first four patients treated with GD2-directed CAR-T cells [49].

2.3.2. ONC201

Amid a plethora of failed trials, the DRD2 antagonist, ONC201 has coveted much attention as of late. In 2019, a case report detailing tumor regression and facial palsy reversal in a 10 year old patient with brainstem DMG [50]. This sparked a number of clinical trials with ONC201 alone or in combination with radiotherapy and/or additional drugs. Later, in 2019, the same group published the results of the first 14 patients with recurrent H3K27M DMG treated with ONC201 (NCT02525692) where they reported a median progression-free survival of 14 weeks and median overall survival of 17 weeks [51]. Three adult patients remained on therapy with a progression-free follow-up of 49.6 weeks, which might indicate that the adult form of this tumor is more targetable by ONC201 [51]. In a different study, 28 pediatric patients with confirmed H3K27M tumors with treated with a German-sourced ONC201 showed a median overall survival of 18 months and this was increased to 22 months in those who underwent reirradiation [52]. More recently, an open label, multisite trial (NCT03416530) was published and the five patients who had begun ONC201 following radiation and prior to occurrence were still alive 24 months post-diagnosis [53].

There are a number of ongoing ONC201 trials including one study for adults with H3K27M tumors (NCT03295396). There is a large phase III study underway across many institutions for ONC201 following radiotherapy (NCT05580562) as well as two studies looking at ONC201 in combination with everolimus (NCT05476939) and paxalisib (NCT05009992). These studies will hopefully provide insight into the long-term potential of ONC201 as a therapy for H3K27M tumors.

3. Pharmacokinetic determinants

The molecular heterogeneity and anatomical location of DMG present major obstacles to successful translation of preclinical activity of therapeutic candidates into effective anti-tumor therapies in patients, requiring tumor-specific targeting and innovative drug delivery strategies.

Heterogenous therapeutic vulnerabilities among DMG tumors may have contributed to the inadequate efficacy of previous clinical trials [47, 54]. Consistent with this idea, the IL-13Rα2 immunotoxin GB-13 extended survival in a manner strongly associated with target receptor expression in adult glioblastoma and DMG animal models [55] and the results seen in the original ONC201 trial [51]. Furthermore, intratumoral heterogeneity has been found to be a major contributor to drug resistance and differing therapeutic responses [56, 57]. These observations call into question the value of locally restricted fine-needle biopsy sampling and highlight the need for more rigorous biological-based trial inclusion criteria.

A critical determinant of drug efficacy is achieving adequate exposure of an active agent at its site of action (i.e. effective concentrations of drug in its active, unbound form present at the tumor site for a long enough period of time). Restricted systemic drug delivery to the central nervous system (CNS) is most frequently attributed to the blood–brain barrier (BBB). This complex neurovascular unit excludes nearly all macromolecules and most small molecules from extravagating into the brain parenchyma [58, 59]. In the case of DMG, the absence of contrast-enhancement is indicative of a largely intact BBB [60]. Preclinical evidence further suggests that the brainstem may have a lower density of capillaries than cortical regions and is home to an even more robust BBB, additionally increasing the level of difficulty in delivering therapeutic payloads systemically [61, 62]. Application of CNS pharmacokinetics is therefore critical to the design of new drug formulations and mechanisms of drug delivery, which must be specifically engineered with the aim of optimizing therapeutic exposure at the tumor site in order to impart therapeutic effect.

Attempts to manipulate the BBB and optimize drug delivery to their sites of action have led to the development of noninvasive and invasive administration techniques [63-65]. Osmotic disruption with mannitol, microbubble-mediated focused ultrasound, and pharmacologic manipulation by bradykinin and its agonists enable transport of a wide range of therapeutics to the CNS, but their effects are transient and nonspecific [66-68]. Invasive strategies such as biodegradable polymers, osmotic pumps, and intrathecal or intraventricular injection facilitate administration directly into the cerebrospinal fluid or brain parenchyma, thereby minimizing toxicities associated with systemic exposure [69, 70]. However, drug distribution and dispersion are limited by the principles of diffusion [59, 71]. Direct loco-regional perfusion of brain parenchyma using a small hydrostatic pressure gradient relies primarily on bulk flow to distribute infusates (i.e. a pressure gradient rather than a concentration gradient) [72]. This technology has been termed convection-enhanced delivery (CED) and is becoming more widely used in research models and clinical trials of DMG.

4. Convection-enhanced delivery

CED is a surgical technique that has gained traction as a promising method that addresses the key problem of low drug penetration and residence time in DMG [73]. CED involves placement of one or multiple catheters directly into the tumor, which may be implanted at the time of biopsy, thus bypassing the BBB and allowing for prolonged administration of infusates into the tumor mass [10, 74] (Figure 2a). The technical capabilities and limitations of CED have been explored by multiple groups and are well characterized [74-77]. Pharmacokinetic studies in both small- and large-animal models have demonstrated that nano-scale particles with a diameter of less than 100 nm are the ideal size for achieving a large V_d per volume of infusion but are readily cleared by efflux mechanisms [78–81]. Hydrophobic and positively charged molecules tend to have a poor V_d [80, 82, 83], while surface modification of drugs has been shown to improve V_d and facilitate tumor specificity [42, 80, 84]. CED-paired nanoparticles allow hydrophobic drugs to traverse the extracellular space, facilitate controlled release of drug over time, and prevent premature drug degradation [85, 86]. Two clinical trials of panobinostat in patients with newly diagnosed DMG are currently underway using such a formulation (MTX110; NCT03566199, NCT04264143).

Figure 2. Implications for direct intratumoral targeting of H3K27-altered diffuse midline glioma.

(a) The delicate location within the brainstem and intact blood–brain-barrier (BBB) eliminates surgical intervention and limits systemic therapeutic efficacy. Delivering therapy directly to the brainstem tumor via an implanted catheter has the potential to enhance therapeutic benefit. (b) The variable impacting the potential success of intratumoral targeting are complex and multifold. First, targeted therapies with significant therapeutic potential must be identified by in vitro drug screens and dose-response analysis. Furthermore, intratumoral targeting is dependent on inherent physicochemical properties of the drug itself such as solubility, lipophilicity, molecular and/or the physicochemical properties of the drug packaging, like a nanoparticle. Finally, successful delivery will require optimization of the delivery technique itself. This might include such things as infusion rate, catheter design, or trajectory planning. Figure created in BioRender.com.

CED has been proven safe in multiple clinical studies of malignant brain tumors including DMG [42, 43, 47, 75, 77, 87, 88]; however, a myriad of variables ought to be considered in developing an effective CED platform. Catheter configuration, design, and positioning; the number of catheters employed; as well as the infusion duration, flow rate, and volume remain areas of ongoing investigation [69, 89].

While traditional end-port cannulae have been most commonly used, improved infusion profiles have been observed with recessed step catheters and those with a porous tip [90-93]. In the largest clinical trial of CED for high-grade glioma, post-hoc MRI analysis revealed that only half of the catheters were adequately placed in the targeted area [54, 88]. Software algorithms to model and estimate infusion distribution based on the planned catheter trajectory, infusion parameters, and patient/tumor-specific anatomy are being developed in an effort to address this shortcoming. For example, Wembacher-Schroeder et al. compared the distribution of a radiolabeled antibody as determined by positron emission tomography (PET) to the distribution estimated by the iPlan Flow simulation algorithm, and found acceptable similarity in 8 out of 10 patients [94]. Use of robotic guidance to improve the accuracy of catheter placement was shown to be feasible to stereotactically implant a CED catheter in a 5-year-old patient with a large brainstem DMG [95].

Barua et al. developed a bone-anchored port system that can be connected to multiple-catheters (up to four catheters) and facilitates repeated CED administration to the brain without requiring multiple surgeries [96, 97]. Chronic, continuous CED via an implantable subcutaneous pump has been tested extensively in small and large animal models for up to 32 days of uninterrupted infusion [98-101]. This system has recently been used in DMG patients to safely deliver two 48-hour-infusions of MTX110 7 days apart [102].

Most studies to date have relied on gadolinium-based contrast agents as a surrogate tracer to track infusate distribution over time. Gadolinium (Gd) formulations are available at a large range of molecular weights and have been shown to accurately track the distribution of IL13-PE38QQR at certain concentrations [91, 103, 104]. However, the hydrophilic nature of most Gd formulations is likely to lead to a different distribution pattern than most small-molecule drugs and chemotherapeutic agents used with CED [59, 82]. Furthermore, drug-specific physicochemical properties, which may cause metabolic breakdown or tissue binding depending on the duration of infusion, differ from inert tracers and, consequently, impact surrogate imaging accuracy over time [104].

5. Expert opinion

A wealth of research over the past decade has revealed multiple targetable vulnerabilities in DMG, but effective drug delivery to the tumor remains a central obstacle to improving patient outcomes. Numerous drug screens in patient-derived cell lines and patient-derived xenografts have identified molecular dependencies and promising treatments, yet considerable systemic toxicity and a low BBB penetrance in the brainstem limit translation of these findings to the clinic [59]. Direct drug delivery via CED is a promising method to address these limitations but is constrained by intrinsic physicochemical properties of the infusate and the brainstem microenvironment, which impact tumor targeting, distribution volume, and drug elimination from the brain. Finding the right drug, getting it to the tumor, and then keeping it there long enough is likely what it is going to take (Figure 2b).

While CED allows for small molecules to cover larger tumor volumes in animal models, they are more rapidly cleared from the brainstem [81, 101]. Conversely, larger therapeutics have been shown to be limited in terms of their interstitial distribution, due to spatial and steric hindrance and interaction with cell surface/extracellular molecules, but may reside at the tumor site for longer periods following loco-regional infusion [78]. Future studies must address these pharmacokinetic principles when considering translatability of potential therapeutics.

Drugs that have known efficacy in drug screens and that possess chemical handles allowing for simple chemical modification can be used to modify their structure to create drug ‘depots’ to allow sustained delivery following direct delivery [105]. Further development and evaluation of the safety, distribution, and clearance effects of different molecular weight polyethylene glycol and boron-dextran complexes when directly delivered to the brainstem will assess the prospects of this approach. Furthermore, the effects of blocking efflux pumps on drug clearance may be analyzed for DMG [106, 107].

An alternative approach to optimize tumor-coverage and achieve longer sustained drug delivery when administered directly to the brain is to modify nanoparticle formulations for CED. Various nanoparticle-based drug delivery systems have caught the attention of researchers worldwide, encouraging the field to rapidly develop improved ways for effective drug delivery across the BBB [108]. Nanoparticles can be loaded with drugs and tuned to protect and release drug over extended periods of time [109, 110]. Additionally, co-loading with imaging agents like Gd can be used to visualize in real-time the volume of distribution when directly administered to the brain [111, 112]. Optimization of nanoparticle loading and release with several preclinical drug hits and contrast agent as well as in vivo pharmacokinetic studies will be required to evaluate direct nanoparticle delivery and clearance from the brainstem.

In sum, extensive research efforts have revealed a central role for epigenetic dysregulation in the pathogenesis of DMG. Clinically, standard fractionated radiotherapy to a dose of 54–59 Gy over 30 fractions remains the mainstay of treatment for newly diagnosed DMG, which alleviates some symptoms for a short period of time without providing a significant improvement in overall survival [113]. Over 100 clinical trials with therapeutics have yet to find benefit [11].

Moving forward, the next generation of therapeutic approaches should implement recent groundbreaking findings in the fields of DMG biology and CNS pharmacokinetics into clinically efficacious treatments for this intractable disease.

Article highlights.

1. H3K27-altered diffuse midline glioma (DMG) is a rare but deadly brain tumor predominately found in children.

2. Despite decades of research, clinicians have not yet been able to overcome the complex obstacles for successful clinical implementation of novel therapies.

3. While some therapies may be limited by intratumoral heterogeneity or lack of sufficient exposure time, the crux in almost all therapies is the blood–brain-barrier (BBB).

4. To circumvent the BBB, convection-enhanced delivery (CED) has emerged as a safe technique to deliver drugs directly into the tumor with additional optimization currently underway in preclinical and clinical studies.

5. A successful therapy for DMG tumors will likely require that we are able to get a cytotoxic drug to the tumor and keep it there for long enough to elicit its effects.