Polymer Nanocarriers for Targeted Local Delivery of Agents in Treating Brain Tumors

Abstract

Glioblastoma (GBM), the deadliest brain cancer, presents a multitude of challenges to the development of new therapies. The standard of care has only changed marginally in the past 17 years, and few new chemotherapies have emerged to supplant or effectively combine with temozolomide. Concurrently, new technologies and techniques are being investigated to overcome the pharmacokinetic challenges associated with brain delivery, such as the blood brain barrier (BBB), tissue penetration, diffusion, and clearance in order to allow for potent agents to successful engage in tumor killing. Alternative delivery modalities such as focused ultrasound and convection enhanced delivery (CED) allow for the local disruption of the blood brain barrier, and the latter in particular has shown promise in achieving broad distribution of agents in the brain. Furthermore, the development of polymeric nanocarriers to encapsulate a variety of cargo, including small molecules, proteins, and nucleic acids, have allowed for formulations that protect and control the release of said cargo to extend its half-life. The combination of local delivery and nanocarriers presents an exciting opportunity to address the limitations of current chemotherapies for GBM toward the goal of improving safety and efficacy of treatment. However, much work remains to establish standard criteria for selection and implementation of these modalities before they can be widely implemented in the clinic. Ultimately, engineering principles and nanotechnology have opened the door to a new wave of research that may soon advance the stagnant state of GBM treatment development.

Glioblastoma Prognosis, Characterization, and Overview

Glioblastoma (GBM), formerly known as Glioblastoma Multiforme, is the most highly incident and deadliest form of brain cancer^[1]. GBM presents more commonly amongst adults older than 50 years of age, though a distinction exists between those with Isocitrate Dehydrogenase (IDH)-wildtype and -mutant phenotypes^[2]. The former is representative of a large majority of GBM tumors and is presumed as the focus for the rest of this review, while the latter tumors present a unique set of challenges, and thus the prognosis is not fully addressed here. Further classification of GBM tumors is complex and based on a combination of histologic presentation and molecular phenotypes. Classification by morphological features has, in the past decade, given way to a greater focus on mutations in particular molecular markers such as EGFR, PDGFRA, NF1 and MGMT (promoter status)^[3].

Progress on the development of new treatments has been slow, as no alternative to the standard of care has emerged in nearly fifteen years. The standard of care—involving surgical resection, radiation therapy (RT), and chemotherapy—is rigorous, invasive, and leads to only limited survival extension in most patients^[4]. Tumor recurrence is a common and difficult hurdle; the development of chemotherapy resistance in the recurrent tumor due to the failure to eliminate a small number of elusive cells highlights the importance of expanding the options available for treating these tumors^[5]. Greater understanding of the interaction between DNA damaging agents (i.e. chemotherapy and RT) and the multiple pathways by which tumor cells repair such damage can lead to the development of new therapeutic methods based on synergistic drug interactions. Advances in drug delivery modalities are needed to allow for the use of drug combinations that would normally demonstrate deleterious toxicity when simply administered systemically. This review describes the rationale for those delivery systems, and recent progress in achieving better delivery by using nanotechnology approaches.

Glioblastoma Classification

Glioblastoma has traditionally been classified as a Grade IV astrocytoma by the World Health Organization based on a set of histopathological criteria, including: “nuclear atypia, mitotic activity, microvascular proliferation, and necrosis”^[6]. In recent years, GBM tumors have further been categorized into four main subtypes (classical, mesenchymal, proneural, and neural) associated primarily, though not universally, with mutations in particular genes. Work based on the catalogue of tumor sequencing compiled by The Cancer Genome Atlas (TCGA) has associated the classical subtype with amplification of the Epidermal Growth Factor Receptor (EGFR) gene, which is commonly mutated, along with lack of mutations in TP54^[7]. The mesenchymal subtype is most closely associated with deletion, mutation, or under-expression of the NF1 gene^[7]. The proneural subtype is associated with mutations in IDH1 and PDGFRA^[7]. The neural subtype was associated with a set of markers common in neurons, but more recent analyses have suggested that this subtype’s distinction in previous analyses was only due to tumor samples being contaminated with healthy cells^[8]. Thus, the classical-mesenchymal-proneural schema appears to be the most appropriate based on our current understanding.

Standard of Care and Alternative Treatments for Glioblastoma

The standard of care (SOC) for most GBM patients consists of maximal surgical resection of the tumor followed by radiotherapy concurrent with chemotherapy, most typically temozolomide (Fig. 1)^[9],[10]. The addition of temozolomide (TMZ) to the standard of care improved 5-year survival from 1.9% to 9.8%, though since the advent of TMZ in 2005, there has not been any major update to the SOC leading to greater survival benefit though NovoTTF (tumor treatment fields) has emerged as a seemingly safer alternative to chemotherapy^[10],[11]. Furthermore, treatment for patients above the age of 70 remains even more challenging as they respond less favorably to the SOC^[12]. TMZ has become the chemotherapy drug of choice due to its relatively low toxicity, ease of administration (orally), and blood brain barrier penetration^[13].

Figure 1:

A representation of the typical standard of care for glioblastoma patients including maximal surgical resection (A) cycles of radiation and temozolomide (B) and ultimately maintenance chemotherapy to reduce the impact of recurrence (C). Created with Biorender.com.

Other agents, such as lomustine and bevacizumab, have more recently emerged as options for continued treatment after the failure of the SOC. Lomustine is promising as an addition to TMZ for patients with methylated MGMT promoters due to having both TMZ-esque alkylating functionality as well as the ability to cross-link DNA strands and modify certain amino acid residues^[14]. Bevacizumab is a monoclonal antibody that targets Vascular Endothelial Growth Factor (VEGF) and has demonstrated some promise as an adjuvant treatment for recurrent GBM after the failure of the SOC and Lomustine^[15]. However, these additional therapies have not been properly evaluated in newly diagnosed patients. A variety of experimental therapies have arisen in recent years that might shift the paradigm of GBM treatment including immunotherapy, small molecule inhibitors, antibodies or antibody-drug conjugates, and chemo/radio-sensitizers^[16]. These treatments have been designed to exploit synergies between RT or temozolomide in order to overcome resistance to the treatment. Poly-(ADP-Ribose)-DNA Polymerase (PARP) inhibitors, DNA-dependent Protein Kinase (DNA-PK) inhibitors, and inhibitors of the Ataxia Telangiectasia and Rad-3 Related (ATR) pathway involve influencing aspects of DNA repair mechanisms to enhance the efficacy front-line therapies^{[17],[18],[19]}. Alongside bevacizumab, other antibody-based therapies have also been developed. Trials with the EFGR-targeted monoclonal antibodies Cetuximab and Nimotuzumab have shown no survival benefit compared to the SOC as a single agent or in combination therapies^{[20],[21],[22]}. Ultimately, many of these investigational therapies have demonstrated pharmacokinetic challenges—often low penetration into the brain tumor microenvironment—preventing wider clinical use.

Challenges of Drug Delivery to the Brain

The development of drugs targeting GBM and other diseases of the CNS is complicated by a set of pharmacokinetic challenges. Traditional routes of chemotherapy administration, such as oral or intravenous, have been hampered by difficulty penetrating the blood brain barrier (BBB), high and toxic doses required to achieve bioavailability, diffusion through the dense brain parenchyma, metabolic degradation, and clearance. Within the traditional drug design paradigm, it has been viewed as incredibly difficult to achieve safety and efficacy alongside a system conducive to patient compliance for CNS targeted therapies, and thus for many years few pharmaceutical companies focused on developing such treatments^[23]. Furthermore, as more CNS diseases have been elucidated, the investment and development of new treatments have not kept pace due to the high risk of failed clinical trials^[24]. Attempts to address these challenges are diverse; they range from chemical modification of drugs to bypass the BBB, biodegradable implants to ensure sustained release, nanocarriers to protect their molecular cargo, surgical techniques that bypass the BBB, chemical or biomolecular or ultrasonic methods to disrupt the BBB, and alternative routes of delivery such as intranasal. The intersection of pharmacokinetic modeling and theranostic imaging has been crucial for pushing the envelope on our understanding the factors contributing to successful brain delivery. Furthermore, the GBM microenvironment presents additional challenges and opportunities. GBM tumors demonstrate differences in tissue diffusivity due to necrosis and cell density, hypoxic and acidic conditions, have a stiffer extracellular matrix, increased vascularization, and BBB-disruptive functions^[25]. Ultimately, these conditions alongside the general brain delivery concerns present a complex design problem, within which many of these factors must be accounted to achieve better therapy. Especially, as significant progress has stalled toward developing new chemotherapeutic agents for GBM in the past couple decades, more creative solutions must be incorporated to best utilize the molecules that have shown efficacy in models but not within the clinic.

The Blood Brain Barrier and Systemic Drug Delivery

The BBB is one of the most significant and immediate challenges to successful delivery of drugs to the brain through systemic (oral or intravenous) routes. The BBB most simply is the convergence of the endothelial cells (ECs) of the microvasculature surrounding the CNS and their supporting cells such as pericytes, smooth muscle cells, macrophages, and microglial cells (Fig. 2)^[26]. Astrocytes, in particular, have been shown to be important mediators of BBB integrity, partnering with ECs to regulate the ‘leakiness’ of the barrier^[27]. The BBB serves multiple functions, including: managing ion balance between the brain and surrounding tissue in conjunction with the choroid plexus^[28]; dampening the influence of neurotransmitters released in the peripheral nervous system (PNS) on the CNS^[29]; and preventing damaging enzymes, macromolecules, and neurotoxins from entering the brain^[30]. This complexly regulated system helps to ensure the stability of the brain microenvironment, allow for proper functioning, and protect the sensitive CNS from systemically circulating agents^[31]. The primary mechanism by which the BBB regulates and excludes passage of molecules into the brain parenchyma is the maintenance of tight junctions. Tight junctions are formed from a set of ‘interlocking’ proteins including claudins, occludins, and junctional adhesion molecules, which create a lock-and-dam mechanism between ECs, preventing the passage of molecules larger than ions^[32]. Extracellular calcium ions play a pivotal role in maintaining the active state of subunits of tight junction proteins; calcium depletion can lead to BBB disruption^[33]. Certain disease states have been shown to impact tight junction permeability such as hypoxic conditions and inflammatory edema, and presence of tumors. Furthermore, reduced expression of crucial tight junction components has been observed surrounding brain tumors^[34],[35].

Figure 2:

A depiction of the intricacies associated with drug transport at the blood brain barrier. Tight junctions composed of claudins, occludins, and junctional adhesion molecules prevent paracellular transport of macromolecules. The basement membrane provides a scaffold for the interactions between endothelial cells, astrocytes, and pericytes to regulate the permeability and expression of surface receptors. A variety of transporters, efflux pumps, and channels allow for the regulated passage of molecules into the brain interstitial space though remove most drugs efficiently. Created with Biorender.com.

The ability of a molecule to traverse the BBB has been associated with certain properties, though the possession of such characteristics does not guarantee successful transport. Higher lipophilicity and lower polarity, hydrogen bond formation, molecular weight below 450 Da, and low plasma protein binding are associated with greater transport across the BBB^[31],[36]. Furthermore, it has been demonstrated that positively-charged and basic molecules have a greater tendency to pass through the BBB than negatively-charged molecules, which is attributed to their easier permeation through the negatively-charged glycocalyx and phosphoplipid heads on the EC membranes^[37],[31].

Due to the BBB’s relative impermeability to most molecules, the ECs rely on a series of membrane transporters to allow access for some crucial molecules, as well, as preferentially remove others from the brain space. Some of the primary transporters that mediate uptake of agents into the brain include organic anion transporting polypeptides (OATPs) and monocarboxylate transporter 1 (MCT-1), which are respectively known to transport the drugs methotrexate and γ-hydroxybutyrate^[38]. Alternatively, efflux transporters account for the opposite phenomenon, such as the ATP-binding cassette (ABC) family of proteins including P-glycoprotein (PGP), multi-drug resistance associated proteins (MRPs), and breast cancer resistance protein (BCRP)^[39], Thus, even many lipophilic molecules with high BBB penetrability and molecules transportable by the ‘influx’ transporters still achieve only low tissue concentrations due to their rapid efflux, which complicates the drug development process.

Drug Diffusion and Tissue Penetration

After overcoming the initial hurdle of the BBB, GBM therapies must be able to ensure adequate tumor coverage to optimally address the infiltrative nature of tumors. TMZ, for example, fails to reach and fully eliminate all clinically relevant populations of tumor cells not only due to molecular resistance, but also due to pharmacodynamics^[40]. Aside from BBB considerations, an understanding of drug movement through the brain parenchyma itself is crucial toward optimizing methods of delivery for new therapeutic molecules. The brain extracellular space (ECS) is compressed and tortuous due to the density of cells in the tissue, creating an environment where molecules may be less likely to reach certain areas through diffusion alone due to tissue anisotropy (Fig. 3A)^[41],[42]. This heterogeneity in distribution effects molecules over 3000 Da in size most significantly; however, studies focused on small molecules such as the EGFR inhibitor erlotinib and tetrodotoxin have demonstrated the limitations of diffusion alone for small molecules in the brain ECS^[43],^[44],[45]. Interestingly, the ECS takes up a larger volume of GBM tissue compared to that of healthy tissue, yet the diffusion of small molecules through tumor tissue has been shown to be slowed – an effect attributed to increased production of ECM creating a greater barrier to diffusion^[46]. Furthermore, evaluations of another alternative delivery method, the delivery of carmustine from the implantable Gliadel^® wafer, have shown limited penetration of the drug into the surrounding tissue despite the highly concentrated depot and rapid degradation of the implant (Fig. 3B)^[47]. Pharmacokinetic models have further shown that low water solubility decreases the ability of a drug to penetrate brain tissue, which is an added complexity as many leading GBM targeted drugs, such as TMZ, are relatively hydrophobic^[48]. Thus, it is clear that methods to enhance drug permeability through the brain parenchyma itself such as direct infusion techniques, alongside understanding the mechanisms of drug action, are of paramount importance in developing new GBM treatments (Fig. 3C).

Figure 3:

Depictions of primary means of drug delivery to the brain. Systemic drug delivery leads to anisotropic distribution of drug within the brain and difficulties in reaching deep tumors (A). Implantable wafers allow for controlled delivery in resected tumors cavities but have diffusion limited penetration deeper into the parenchyma (B). Direct local infusion methods allow for targeted delivery and good tumor coverage though present difficulties with regard to drug clearance and continued dosing (C). Created with Biorender.com.

Clearance and Metabolism Pharmacokinetics

Alongside barriers to entry and penetration into the brain, drugs face myriad mechanisms of elimination through bulk fluid flow, efflux, chemical degradation, and enzymatic metabolism (Fig. 4A). The elimination rate of a drug is crucial for achieving a therapeutic concentration in the brain – as even a drug with high BBB permeability and brain diffusivity may not achieve a therapeutic effect if its elimination rate is high. Models have suggested that in many cases clearance rates and brain distribution are more determinative of efficacy than BBB permeability on its own^[49]. The most prominent means of elimination are interstitial fluid transport and drug metabolism, the extent of which vary between different drugs.

Figure 4:

An overview of mechanisms of drug clearance from the brain including efflux through transporters across the BBB, diffusion, cell internalization and associated metabolism, enzymatic degradation in the interstitial space, transport through the choroid plexus into the subarachnoid and ultimately venous circulation (A). A qualitative representation of drug pharmacokinetics based on mechanism of delivery. Sustained delivery methods tend to maintain drug concentrations compared to single dose administration (B). Created with Biorender.com.

Fluid flow in the brain progresses primarily in one direction – pressure gradients at the BBB interface drive bulk flow of fluid from the vasculature into the brain interstitial space, then along perivascular spaces and along axons, until reaching the choroid plexus, which regulates filtration of the cerebrospinal fluid (CSF)^[50]. Ultimately, the purified CSF circulates through the subarachnoid space and ventricles, providing mechanical and immune support to the brain, and drains in the venous circulation and the so-called ‘glymphatic’ system (the CNS component of the lymphatic system). The choroid plexus is thought to be ‘leakier’ than the BBB, and less accessible for drug delivery from systemic methods due to its relationship to venous rather than arterial circulation^[51]. These complex relationships between fluid spaces in the brain have created challenges in measuring drug concentrations in the brain. The most widely adopted technique, microdialysis, is often utilized to sample CSF and interstitial fluid (ISF), allowing measurement of concentrations of drug^[52],[53]. Studies have examined the rate of elimination of molecules from animal and human brains, and have suggested half-lives from 4 to over 25 hours for molecules of different sizes – the rate of efflux being dependent not only on route of sampling but also on anesthetization, as elimination half-life has been shown to differ depending on type of anesthesia^[54]. Microdialysis has been successfully used to sample interstitial levels of TMZ in GBM patients, which was demonstrated to reach a peak concentration of approximately 0.6 ug/mL at 2 hrs post oral administration of a 150 mg/m² oral dose^[55]. The relatively rapid elimination of TMZ suggests a need for high systemic doses of the drug to achieve therapeutic concentrations in the brain. Thus, an important area of development for newer therapies are methods to improve drug retention in the brain tissue.

Aside from bulk fluid movement, a drug’s efficacy may be hampered by numerous mechanisms of degradation and metabolism by enzymes in the brain interstitium and at the fluid interfaces. Brain metabolic enzymes tend to be characterized in the broad categories of Phase I, which generally functionalize lipophilic drug molecules, and Phase II, which conjugate functional groups to target drugs for further degradation^[56]. It is important to consider the metabolic expression profile of a GBM when selecting and designing agents for treatment, as it is possible that enzymes with activity toward a particular drug might be expressed differently in the tumor compared to surrounding healthy tissue.

Off Target Toxicity of GBM Chemotherapies and Combinations

Another major challenge associated with chemotherapies and treatment combinations are off target toxicities that limit the maximal dose and are dependent on the method of drug administration. Toxicities may arise as acute symptoms, which include deficiencies in blood cell counts, nausea and vomiting, and liver enzyme elevation. In recent years, evidence of neurological toxicities has emerged, as some long-term survivors of the extensive SOC have developed more complex side effects long after treatment. Deficiencies in white blood cells and platelets are extremely common for concomitant TMZ with RT, perhaps more common than the relatively low incidence reported in the initial trials^[57],[58]. Some patients demonstrate prolonged cytopenia, suggesting intense toxicity in the bone marrow, and hence suggesting further investigation is necessary to elucidate why certain patient cohorts exhibit greater toxicity. Furthermore, Phase II trials of TMZ with the MGMT inhibitor O⁶BG in TMZ-resistant GBM demonstrated a high incidence of serious hematological toxicity^[59]. Thus, when designing future combination therapies based on the synergy of DNA damaging agents with other molecules, it is crucial to evaluate such deleterious impacts. Trials with other well-characterized alternative chemotherapies including carmustine, cisplatin, and lomustine (both as single agents and in combination with TMZ) elicited grade VI hematological adverse events in at least 15% of patients^{[60],[61],[62],[63]}. Of note, the most serious events were associated with TMZ combinations leading, in some cases, to halting of treatment. These incidents of toxicity are most well recognized in elderly patients, who receive an altered course of TMZ administration due to the risk these events pose^[64]. Avoiding systemic toxicities is one of the motivations for direct delivery to the brain, by CED or implants. But the carmustine-releasing Gliadel^® implant has been shown to possess its own set of toxicities at low rates, primarily in the CNS, including cerebral edema, infection, and seizures^[65]. It is thus suggestive that alternative routes of drug administration may help to avoid systemic toxicities while rational design is needed to minimize local toxicities.

Methods to Disrupt or Bypass the Blood Brain Barrier

Methods to overcome the BBB tend to fall into the categories of disruption, exploiting biological mechanisms, or bypassing it through surgical techniques. One of the earliest methods of BBB penetration involved disrupting the BBB endothelial junctions using hyperosmotic solutions, which is very effective within a short time of induction but has never been established as safe due to high neurotoxicity^[66],[67]. A more recent method of disruption is focused ultrasound with microbubbles, which is non-invasive and can be specifically targeted when coupled with imaging techniques^[68]. This technique has been well-evaluated and is thought to be safe though variable in efficacy, subject to time constraints that may not cooperate with the bioavailability of particular drugs, and does not overcome the issue of rapid drug efflux through transporters^[69]. The inhibition of drug efflux transporters has also been a strategy to increase BBB permeability. The most promising candidates for ABC transporter inhibition thus far, Tariquidar and Elacridar, have suffered from low potency, difficult pharmacokinetics, and the lack of stable oral formulations^[70]. One of the major foci of BBB penetration research is exploiting endocytotic mechanisms of endothelial cells by conjugating molecular moieties known to cross the BBB to either nanocarriers or a drug itself. A form of paclitaxel conjugated to the peptide Angiopep-2, has demonstrated relative safety and the ability to penetrate the BBB in patients, though it has not achieved greater efficacy than the SOC treatments^[71]. Alternatively, glutathione-conjugated liposomes have been used to deliver doxorubicin to patients with recurrent glioma and achieved much greater brain penetration than other formulations^[72]. These agents however have yet to be fully evaluated for efficacy. The most well-established method to circumvent the BBB has been the implantation of Gliadel^® wafers into tumor resection cavities, which still suffers from lack of tissue penetration, as described previously^[73]. Ultimately, the one of the most promising methods to bypass the BBB is a direct surgical infusion of agents into the brain in an approach called convection-enhanced delivery (CED). Initially established by Bobo et al. who delivered radiolabeled transferrin throughout the brain, CED utilizes fluid pressure gradients over a slow, extended delivery period to infuse agents over a broader volume^[74]. This technique has become one of the best characterized ways to surpass the BBB and has been extensively evaluated in the clinic for safety and to establish the appropriate parameters governing optimal delivery.

A Clinical History of CED

CED has been utilized to deliver a variety of cargo in a clinical setting for the treatment of high-grade gliomas. Clinical development has focused on parameters such as catheter placement, fluid flow rate, volume of distribution of infusate, as well as safety and efficacy. Many trials have demonstrated the safety of CED with various agents, but none have demonstrated increased therapeutic efficacy compared to the current SOC (Table 1). Despite some small success, most trials demonstrated challenges associated with pharmacokinetics, surgical complications, and toxicities that have prevented widespread use.

Table 1:

Summary of Results from CED Studies and Trials

Delivered Agent	Year	Species (n)	Disease Target	Formulation Details	Outcomes	Other Notes	Citation
Paclitaxel	2004	Human (n=15)	Recurrent high-grade gliomas (mostly GBM)	Dissolved in Saline with cremophor as vehicle	Most patients died within 7 months after treatment	Chemical meningitis and seizures were common complications	[75]
Topotecan	2011	Human (n=16)	Supratentorial, recurrent high-grade gliomas	Not specified	Median survival of 60 weeks, 75% 6-month survival	<20% of patients exhibited high grade adverse events, undetectable systemic drug and associated toxicities after administration	[76]
Nimustine Hydrochloride (ACNU)	2020	Human (n=16)	Recurrent brainstem glioma (GBM and DIPG)	Aqueous solution	Only two cases were considered responders which most demonstrating either stable or progressive disease; Failed to show improvement over Gliadel	ACNU given in conjunction with systemic TMZ; Only 2 patients demonstrated non-transient treatment associated toxicities	[77]
Irinotecan (Onivyde)	2022	Human (n=18)	Recurrent High-grade glioma	Nanoliposomal	No reported results on efficacy	Observed full tumor volume coverage and a single serious adverse event; Real-time MRI guiding of catheter placement	[80]
8H9 Monoclonal Antibody (Radiolabeled)	2018	Human (n=25)	DIPG	Saline	Median survival of 15.3 months post infusion	Low incidence of Grade 3 and 4 adverse events; Difficulty establishing best dose for escalation	[81]
Cintredekin Besudotox	2007	Human (n=51)	Mostly recurrent GBM	Not specified	Median survival 42.7 weeks (10–11 months); survival greater with 2 or more catheters placed	Performed alongside tumor resection; some high-grade adverse events at highest dose	[82]
	2010	Human (n=296, 192 in CED group)	Recurrent GBM	Not specified	Median survival of 11.3 months with CB group compared to 10 months; survival difference not significant	Performed alongside tumor resection and compared to implanted Gliadel wafers; somewhat higher incidence of pulmonary embolism in CB group however non-significant	[83]
	2018	Human (n=5)	DIPG	Not specified	Mean survival of 20.3 weeks post infusion	Low incidence of grade 3 or higher adverse events though one patient lost verbalization after procedure; single catheter placement	[84]
CPG Oligodeoxynucle-otide	2006	Human (n=24)	Recurrent GBM	0.9% Sodium Chloride (aqueous)	Median survival of 7.2 months post infusion	1 incident of dose limiting toxicity though most other adverse events were mild	[85]
	2010	Human (n=31)	Recurrent GBM	0.9% Sodium Chloride (aqueous)	28-week median overall survival; 4 long-term survivors (>24 months); Did not meet ultimate goal for progression free survival	Nearly 50% of patients experienced grade 3 lymphopenia but was reversed in most cases	[86]
	2017	Human (n=81, 39 in CED group)	Newly Diagnosed GBM	0.9% Sodium Chloride (aqueous)	9-month median progression free survival in CED-CPG group vs 8.5 months SOC and increased 2-year survival; Difference was not statistically significant	Compared to SOC (radiotherapy + TMZ)	[87]
Recombinant Poliovirus	2018	Human (n=61)	Recurrent GBM	Not specified	Median overall survivor of 12.5 months in treatment group compared to 11.3 months from historical controls; 21% overall survival at 36 months	One dose limiting toxicity case (intracranial hemorrhage) occurred though the patient survived for 57.5 months; 19% of patients experience grade 3 or higher events	[88]
HSV-1-tk	2003	Human (n=8)	Non-diffuse GBM	Liposomal	Median overall survival of 28.1 weeks after infusion	Given in conjunction with systemic Ganciclovir; Transient worsening of neurological symptoms	[89]
Trabedersen	2011	Human (n=145, 89 in CED groups)	GBM and astrocytoma	Not specified	Median survival of lower dose 39.1 months compared to 21.7 months SOC in astrocytoma; Survival metrics not significantly different for the GBM groups	Compared to SOC chemotherapies (temozolomide or procarbazine/lomustine/vincristine); lower incidence of adverse events compared to SOC	[90]

Clinical CED of chemotherapeutics has demonstrated a range of different outcomes, though most have only comprised Phase 1 single-center studies. Trials with paclitaxel demonstrated robust initial tumor response for recurrent GBM, which did not appear to produce major changes in long term survival while also demonstrating unexpected toxicity in some patients due to drug leakage into the CSF^[75]. More recent studies, however, have shown more promise. A Phase I trial of CED of topotecan and a Phase 1 trial of CED of nimustine hydrochloride (ACNU) in young DIPG patients have recently demonstrated some success in achieving high dosing without such severe toxicities. However, efficacy remains to be truly evaluated and appeared heavily influenced by patient-to-patient variation within these small scale trials^[76],[77]. Other small molecule studies remain in the early stages and await further evaluation in broader efficacy studies^{[78],[79],[80]}.

Many early phase clinical studies have focused on delivering immunotoxins and monoclonal antibodies. A Phase I trial for DIPG used a unique approach, incorporating a radioactive isotope into a tumor-targeted antibody for localized radioimmune therapy^[81]. This trial demonstrated a much cleaner safety profile for both the therapy and surgical procedure as compared to most CED trials, with relatively few adverse events. Coupled with PET imaging, the study coordinators were also able to establish that the infusions covered a majority of the estimated tumor volume with drug. A series of trials into the chimeric IL13-conjugated Pseudomonas exotoxin in both GBM and DIPG demonstrated safety of the CED procedure though identified complexities associated with catheter placement^{[82],[83],[84]}. A Phase III study comparing this immunotoxin to Gliadel^® demonstrated no increased survival benefit despite demonstrating a similar safety profile^[83].

Nucleic acid CED has also been extensively studied for glioma treatment. A series of trials into CpG-based oligonucleotides for GBM demonstrated high tolerable doses with low treatment and low incidence of procedure-related adverse events^{[85],[86],[87]}. However, upon reaching Phase II, CED of CpG, alongside the standard of care treatments, failed to demonstrate survival benefit. A Phase I study of recombinant poliovirus further demonstrated safety of nucleic acid delivered while using a more controversial viral vector^[88]. Of particular note for nucleic acid delivery is an older Phase I study that established the safety of delivering HSV-1-tk gene in a cationic liposomal vector, one of the few examples of using CED and nanocarriers in conjunction in the clinic^[89]. Furthermore, a Phase II study of Trabedersen, an anti-sense oligonucleotide inhibitor, demonstrated both safety of the use of CED for the drug and a possible survival benefit when compared to various commonly administered chemotherapies^[90]. The fate of the planned Phase III trial following, however, remains unclear as there is no publicly available information.

Thus, despite some remaining challenges related to catheter placement, CED has been established as a safe mechanism for delivery of anti-tumor agents, including a variety of nucleic acid formulations. Efficacy, however, remains to be definitively shown amongst even the most advanced trials. One consistent limitation of all trials is the lack of evaluation of the retention half-life of the administered agents. These studies demonstrate the potential of CED as a promising method for delivering therapeutics but also make clear that work remains to understand and improve upon the associated pharmacokinetic challenges.

Parameters for Evaluating CED and Associated Challenges

Early investigations into the use of CED examined the fluid dynamics of the technique and how to optimize the parameters of the infusion to achieve the greatest possible volume of distribution while limiting backflow. These investigations demonstrated that infusion rates of 1 uL/min and higher, or using cannulae of larger than 32 gauge, contributed to excessive backflow compared to lower infusion rates and smaller gauge cannulae^[91]. The ability of infusate to distribute throughout the brain is highly dependent on flow rate, molecule size, porosity of the brain ECS, extent of convection, and volume of infusion (V_i)^[92]. Furthermore, through a comparison of CED with liposomes possessing different surface characteristics, it has been shown that greater tissue binding and lipophilicity can lead to smaller volume of distribution (V_D) of larger molecules^[93]. Hydrophilic, PEGylated liposomes distributed more widely than cationic liposomes or hydrophobic free doxorubicin alone, suggesting that the infusate itself must be optimized to ensure the greatest extent of distribution (this will be discussed in more depth in a later section). An investigation using models based on trial data has suggested that convective velocity in healthy brain tissue decreases radially from the infusion site while within tumors the opposite is true^[94]. This is attributed to a decrease in interstitial fluid pressure surrounding the tumor creating a gradient not present in healthy tissue. Of particular note, and something that has not been fully explored, is the influence of tumor necrosis on volume of distribution. The previously described model suggests that necrosis actually aids CED, as fluid cannot be reabsorbed by absent vasculature and thus the convective velocity is maintained at greater radial distances from the infusion site.

To achieve the optimal V_D, much of the investigative work related to CED has focused on catheter and cannula design and placement. As discussed earlier, small diameter catheters and cannulas tend to reduce backflow though other aspects of design are also important. Flexible catheter tubing has been shown to more reproducibly achieve a broader distribution of infusate as compared to rigid tubing, possibly due to the ability of the flexible tubing to adjust as convective forces cause shifting in the surrounding tissue^[95]. Similarly, a cannula design in which a flexible fused silica tubing was inserted into a 27-gauge needle was used to demonstrate that flow rates above 5 uL/min lead to tissue damage in both rat and primate brains as well as reduced V_D^[96]. Further, this design demonstrated less reflux of infusate compared to the metal needles alone. A more complex implanted catheter system was developed in order to achieve repeatable 5 uL/min infusions in pig brains and was able to remain patent for over 160 days in some animals^[97]. These catheters still experienced moderate incidence of reflux. Furthermore, catheter placement and number are important parameters that have been investigated in order to improve tumor distribution of infusate, diminish CSF leakage, and reduce side effects^{[98],[99],[100],[101]}.

Still, even after successful CED infusion, free drug is susceptible to the rapid clearance, metabolism, and efflux mechanisms described earlier. One study focused on radiolabeled analogs of the PDGFRA inhibitor dasatinib demonstrated that even with chemical modifications intended to improve stability, the free analogs were eliminated from the brain ECS with half-lives of 0.5–2 hours^[102]. However, they also demonstrated that a dasatinib-nanofiber conjugate (with dimensions of approximately 100 nm × 5 nm) was extremely slow to clear after CED, with an estimated half-life of 57 days. Ultimately, these results and many others illustrate the need to develop infusate formulations that allow for a more extended pharmacokinetic profile to treat GBM and other long-term brain diseases without requiring repeat procedures to maintain effective levels of the drug. Ultimately, the advent of nanoparticles and other nanocarriers has paved the way for expanding the toolbox of CED, by creating vehicles to encapsulate and deliver drugs for extended periods of time (Fig. 5C).

Figure 5:

A depiction of a rodent positioned in a stereotactic frame prior to intracranial infusion or CED (A). Representation of a coronal slice of a mouse brain bearing a tumor during CED, in which the catheter tip is placed near the tumor site and the infusate spreads relatively radially (B). A close-up representation of the distribution of nanoparticle infusate during CED through the interstitial space, populated by tumor cells (salmon), astrocytes (purple), oligodendrocytes (green), neurons (pink) and microglia (yellow) (C). Created with Biorender.com.

Nanoparticles as a Means to Improve Drug Pharmacokinetics

Polymeric nanoparticles (NPs) have been a focus of academic advancements in drug delivery for ~50 years^[103]. Nanoparticles can be customized as delivery vehicles for particular conditions; they can be designed to encapsulate myriad cargo of varying size, be biodegradable while lasting for an extended period, have tunable release kinetics, be reactive toward biological microenvironments with pH and temperature sensitivity, be used for theranostic applications, and more^{[104],[105],[106]}. The versatility of the NP paradigm has allowed the expansion of our ability to target specific tissues, reduce off target effects and release over a longer time for disease states where repeated dosing is challenging.

Certain biodegradable polymers are the most common and well-characterized for the purposes of NP preparation including variations of poly(lactic acid) (PLA), poly(lactic-co-glycolic acid) (PLGA), and poly(caprolactone) (PCL). Furthermore, poly(ethylene glycol) (PEG) has emerged as a common additional block to modify these aforementioned polymers to create more amphiphilic and stealthy NPs (Fig. 6B)^[107]. In recent years, the investigation into alternative polymers with distinct surface chemistry and structure have yielded a wealth of opportunities to design NPs for particular purposes.

Figure 6:

Representation of single-layer micellar nanoparticles with hydrophobic core and hydrophilic surface (A). Solid nanoparticle with a more complex interweave of polymer in the core (B). A polyplex of oppositely charged polymer and macromolecular cargo (C). A basic liposome with a lipid bilayer structure (D). A more complex lipid nanoparticle with a mixture of different lipids, cholesterol and containing DNA (E). A polymeric dendrimer with fractal-like branching (F). Gel-like nanoparticles with cross-links such as alginate or chitosan (G). Polymeric nanoparticle with hyperbranched outer coating rather than straight-chain PEG (H). Nanoparticle made of aggregated albumin, which can lead to high loading of cargo that easily interacts with proteins (I). Created with Biorender.com.

The road toward the widespread use of polymeric NPs in research began in the late 1970s with the seminal work by Robert Langer and Judah Folkman describing the release of proteins from a poly(ethylene vinyl acetate) matrix^[108]. Though smaller molecules had previously been incorporated in and released from polymeric systems, such as silicone, this was the first example of larger molecules exhibiting this behavior. This demonstrated that the phenomenon of controlled release was not limited to a small subset of drug agents. Another crucial finding was the work by Iwai et al. that described the “enhanced permeation and retention” (EPR) effect, which led to the accumulation of nano-sized polymer-drug conjugates formulated with an oily medium within implanted liver tumors^[109]. Despite more recent work suggesting that the EPR effect might be limited in scope, and not as important in human tumors, this work inspired further investigation into nanoscale materials to take advantage of the EPR effect in treating various cancers. Another important development was the introduction of nanoscale PEGylated polymeric micelles incorporating the drug adriamycin, which were demonstrated to preferentially accumulate in tumors (Fig. 6A)^[110]. As interest in nucleic acid delivery grew, cationic polymers were developed that could form nanoparticles to protect and enhance nucleic acids, increasing uptake and endosomal escape of nucleic acids while protecting them from degradation by endonucleases^[111]. As alternatives to cationic polymers poly(ethyleneimine) and poly(beta-amino ester) (PBAE), lipid nanoparticles (LNPs) (Fig. 6E) were developed for nucleic acid delivery^[112],[113]. In recent years, LNPs been used to deliver siRNA drugs and the Moderna and Pfizer/BioNtech mRNA COVID-19 vaccines intramuscularly in humans^[114],[115]. However, all of these materials have demonstrated pronounced cytotoxicity^[116] and efforts to create less toxic materials, such as poly(amine-co-ester) (PACE) polymers, may lead to safer options^{[117],[118],[119],[120]}.

Polymeric NPs have become an important tool in the development of new therapies targeting cancers in the CNS, particularly GBM. NPs have been designed with controlled size to preferentially accumulate within tumors based on the EPR effect^[121], with conjugated ligands to target highly expressed tumor receptors^[122], to shed different polymer layers at different stages of delivery in response to changes in pH and ionic equilibrium^[123], and to be responsive to tumor targeting through ultrasound guiding^[124]. Thus, polymeric NPs represent a versatile platform to enhance delivery of therapeutics that can be intentionally designed to preferentially target brain tumors and their microenvironments.

An Overview of Polymeric Nanoparticle Formulation Methods

A variety of nanoparticle formulation techniques have emerged that allow for modulation of the NPs’ physical properties to optimize their size, drug encapsulation, structure, and surface characteristics. The most prevalent methods of nanoparticle preparation include the emulsion-evaporation, double-emulsion, and nanoprecipitation, though many others also exist. They each rely on fundamental concepts surrounding the miscibility of organic solvent and aqueous phases, may incorporate surfactants to stabilize the interface between the phases, and induce the spontaneous formation of NPs at the interface. The emulsion-evaporation technique involves the dissolution of the chosen polymer and drug in the organic phase, which is then added dropwise to the aqueous phase within which NPs spontaneously form. The aqueous phase in this context often contains a surfactant, such as poly(vinyl alcohol) (PVA), depending on the solvent conditions and the amphiphilicity of the primary polymer. The two phases in this case are typically immiscible at room temperature. The formation of NPs occurs due to the tendency to reach a more stable energy state by aggregation of hydrophobic moieties and interaction of hydrophilic moieties with the surrounding aqueous phase^[125]. Thus, a hydrophobic drug is more likely to be encapsulated within the hydrophobic matrix or core of the NP than remain free in aqueous solution. Additionally, sonication applies sound waves to induce cavitation in order to break up larger emulsion droplets into smaller droplets^[126]. The end result is a colloidal suspension in which nanoparticle surfaces form an interface with the aqueous solution, while the likely toxic organic phase is removed through a process of evaporation or dialysis^[127].

An extension of single emulsions, double emulsions most commonly employ a two-step water-in-oil-in-water (w/o/w) emulsion that is used to encapsulate proteins or relatively hydrophilic drugs that would tend to partition into the aqueous phase^[128],[129]. This process creates a nano or microparticle surrounding aqueous droplets containing the cargo. Because of the complexity of stabilizing these double emulsions, this technique produces NPs with greater characteristic variability compared to other methods.

The solvent diffusion or displacement method is a variation in which two or more solvents are incorporated into the organic phase, one of which is immiscible with water and the other of which is water-miscible (e.g. chloroform and ethanol). The water miscible solvent diffuses into the aqueous phase upon mixing, which rapidly reduces the size of the emulsifying droplets^[130]. This method is preferred by some due to its ability to achieve smaller particle sizes more reliably, but the polymer used must be carefully screened for compatibility with particular solvent combinations.

The nanoprecipitation method is another well-characterized technique that does not require surfactants though has a more limited scope. It relies on incorporating water miscible solvents, such as DMSO, ACN, or acetone into the organic phase and thus limits capturable drugs based on solvent solubility^[131]. In this case, amphiphilic polymers are necessary—such as block copolymers of PLA and PEG—because they effectively act as their own surfactant. Furthermore, hydrophilic molecules are difficult to encapsulate with this technique due to the rapid mixing of the phases and their tendency to diffuse into the aqueous external medium. Nanoprecipitation, however, is one of the more reproducible ways to produce nano-sized particles, the size of which is directly tunable dependent on polymer concentration^[132].

Less common techniques include ‘salting out’, emulsion-polymerization, gelation, and complexation. Salting out is similar to nanoprecipitation, as it incorporates water miscible solvents, but also requires dissolving salts (often calcium or magnesium chloride though sucrose can sometimes be used) in the aqueous phase to create an artificial phase separation and cause NPs to precipitate as the salt concentration changes^[133]. Emulsion-polymerization is a family of techniques that use monomers in the aqueous phase and initiators in the organic phase to simultaneously polymerize and form NPs^[134]. Alternatively, multiple polymer systems, such as the natural polymers chitosan and alginate, can form into gel-like NPs by crosslinking when introduced to ions of opposite charge^[135],[136]. This technique is particularly effective for water soluble proteins or nucleic acids and does not require any organic solvents. A related method involves the complexation of charged polymers of opposite charges and is useful for the formation of ‘polyplexes’ between polycationic polymers and negatively charged nucleic acids (Fig. 6C)^[137]. Ultimately, the large toolkit of NP formulation techniques allows for a breadth of options to address formulation challenges.

Polymer Design and Selection

Determining which polymer to use for NP-based applications requires a detailed understanding of the physical and chemical properties of the polymer itself as well as its behavior in NP form. The hydrophobicity of the overall polymer as well as its blocks (in the case of block copolymers) determines its ability to form and maintain stable NPs, as well as its efficiency of encapsulating drugs. Surface charge and modifications are crucial for modulating uptake by target cells, immunogenicity, and blood half-life^[138]. The size of a polymer chain is related to the size of the NPs it produces. Furthermore, polymers with higher degradation rates lead to increased bulk drug release due to the rapid destabilization of the NP while polymers with lower degradation rates lead to more diffusive release of drug^[139]. Also, NP degradation can proceed through either the surface or bulk erosion, depending on a polymer’s ability to form water permeable, porous NPs or more densely packed, hydrophobic NPs^[140]. Consequently, the diversity of polymers capable of NP formation as well as the broad design criteria requires careful planning before choosing an appropriate polymer system for a particular biological application (Table 2).

Table 2:

Summary of Polymeric Nanocarriers for Brain Delivery and Associated Preclinical Models

Polymer	Abbreviation	Formulation Methods	Cargo, Delivery Route, and Animal Model
Poly(lactide-co-glycolide)	PLGA	Double Emulsion[235]; Single Emulsion-Evaporation[233],[232]	CED of carboplatin in rat and pig GBM models[235]; CED of paclitaxel in U87 rat model[233]; CED of camptothecin in 9L rat model[232]
Poly(lactide-co-glycolide)-block-Poly(ethylene glycol)	PLGA-PEG	Double Emulsion-Evaporation[247]; Nanoprecipitation[248]; Single Emulsion-Evaporation[249]	IV delivery of urocortin in rat Parkinson’s model[247]; IV delivery of loperamide for CNS infections[248]; IV administration of paclitaxel to flank and intracranial mouse and rat glioma xenograft models[249]
Poly(lactic acid)	PLA	Solvent Diffusion[250]	CED of carmustine in C6 rat glioma model[250]
Poly(lactic acid)-block-Poly(ethylene glycol)	PLA-PEG	Reverse Emulsion-Evaporation[251]; Single Emulsion-Evaporation[252]; Nanoprecipitation[243]	IV administration of paclitaxel in mouse C6 glioma model[251]; IV delivery of quisinostat in GL261 mouse model[252]; CED of ATR inhibitors for radiosensitization in RG2 rat model[243]
Poly(lactic acid)-block-Polyglycerol	PLA-HPG	Single Emulsion-Evaporation[245]	CED of anti-miR-21 PNAs in U87 rat model[245]
Poly(caprolactone)	PCL	Double Emulsion-Evaporation[253]; Nanoprecipitation[254]	Intranasal delivery of carboplatin and in vitro evaluation in LN229 cells[253]; CED of camptothecin in 9L rat model[254]
Poly(caprolactone)-block-Poly(ethylene glycol)	PCL-PEG	Single Emulsion-Evaporation[255],[256]	Intravenous administration of dye-loaded NPs and paclitaxel in a mouse U87 model[255]; Intranasal delivery of bexarotene towards Alzheimer’s applications[256]
Poly(ethylene glycol)-block-Poly(decalactone-co-dioxanone)	PEG-PDL-co-DO	Single Emulsion-Evaporation[244]	CED of ATR inhibitors (Berzosertib) in combination with radiotherapy in RG2 rat model[244]
Alginate based nanocarriers	Alginate	Ionic Gelation and Cross-linking[257]	Intranasal delivery of venlafaxine to the brain to treat depression in rats[257]
Chitosan based nanocarriers	Chitosan	Ionic Gelation[222],[258]; Complexation[259]	IV administration of lomustine in an U87 mouse model[222]; Intranasal delivery of venlafaxine to the brain to treat depression in rats[258]; CED of anti-EGFR and anti-Galectin-1 siRNAs in U87 mouse model[259]
Poly(amine-co-ester)	PACE, PBAE	Complexation[245],[260]	CED of anti-mIR-21 in U87 rat model[245]; CED of pTK in 9L and F98 rat models[260]
Poly(ethyleneimine) and associated conjugates	PEI	Complexation[261],[262]	IV administration of pORF-hTRAIL DNA in a U87 mouse model[261]; CED of fluorescently labeled NPs in rat brains and evaluation of uptake in 9L cells[262]
Poly(L-lysine)	Poly-Lysine	Complexation[263]	CED of DNA in healthy rat model[263]
Poly(L-aspartic acid)	Poly-aspartic acid	Complexation[236]	CED of cisplatin in F98 rat tumor model[236]
Poly(amidoamine) Dendrimer	PAMAM	Chemical Synthesis[264]	IV administration of DNA in healthy mouse model[264]

One category of commonly used polymers for nanoencapsulation are natural polymers such as albumin, chitosan, and alginate. Albumin is an important protein component in blood that stabilizes the solubility of other molecules such as fatty acids and other proteins, binds many drugs and ions, and is a major stabilizer of the blood’s colloid-like nature (Fig. 6I)^[141]. Albumin nanoparticles have been developed, shown to encapsulate the cancer drug paclitaxel, elicit higher concentrations of the drug in multiple murine tumor models, and improve efficacy compared to other paclitaxel formulations^[142]. Human serum albumin also possesses the advantage of low immunogenicity due to its ubiquity in human blood. Chitosan is a positively charged polymer derived from the exoskeletons of shellfish and has been used to encapsulate and deliver many proteins and nucleic acids in nanoparticles due to its ability to complex with these macromolecules and high cell uptake^[143]. Alginate is an anionic polymer produced by brown algae and some bacteria and, similar to chitosan, can complex with macromolecules through a cation-based gelation process and has been extensively used to create structured gel ‘nanospheres’ and ‘nanocapsules’ for various applications including chemotherapeutic delivery (Fig. 6G)^[144].

Though natural polymers have had some success, they lack the customizability of synthetic polymers, which can be designed and functionalized more readily for specific applications. The most commonly used such polymers for NP preparation include PLGA, PLA, PCL and associated block copolymers. All of these are biodegradable, an important quality to prevent accumulation of non-degradable materials in tissues thereby inducing damage to cells and a long-term immune response^[145],[146]. PLGA, PLA, and poly(glycolic acid) (PGA) are a family of poly(α-esters) based on lactic acid, which occurs naturally as a byproduct of metabolic fermentation, and glycolic acid, a synthetic acid used commonly in skin care products and biomedical products requiring relatively rapid degradation such as sutures^[147]. PLGA is the copolymer of lactic and glycolic acids, and all three are components of delivery systems approved by the FDA. Lactic acid interestingly can take the form of two different isomers, D and L, the latter of which is the naturally produced form, though typically synthetically produced lactic acid contains a mix of both isomers. The D-isomer in itself is of some concern in biological applications due to the body’s inability to process it^[148]. PLA is typically produced either through a direct polycondensation reaction of lactic acid or a ring opening polymerization of the cyclic lactide dimer^[149]. Typically, the L-isomer is utilized for polymerization, however, during the process there is some interconversion into the D-isomer^[148]. PGA can be synthesized by analogous processes, of which ring opening polymerization of cyclic glycolide is most efficient^[150]. PLGA is typically synthesized using a similar ring opening polymerization utilized both cyclic dimers and can be finely tuned by varying the ratio of the two monomers. All three polymers degrade via simple hydrolysis through cleavage of the ester bonds. Degradation of these polymers is both chain length and pH dependent, with higher rates of degradation observed for smaller molecular weight chains and at higher pH^[151],[152]. PLA typically degrades at a slower rate compared to PGA, and thus the degradation rate of PLGA is highly dependent on the ratio of the two^[151],[153]. PLA and PLGA-based NPs have been extensively investigated though PGA NPs are less ideal due to their overly rapid rate of degradation^[154],[155]. The degradation of these polymers produce their acidic monomers creating an acidic microenvironment within nano and microparticles, which could be of concern for certain drug and macromolecular cargos^[156]. A common alternative to the poly(α-esters) is PCL, one of the earliest synthetic biodegradable polymers, which is synthesized through ring opening polymerization of ε-caprolactone or 2-methylene-1–3—dioxepane, and exhibits a much slower degradation profile compared to the poly(α-esters)^[157]. Due to its long hydrolytic degradation time in vivo, PCL has been primarily used for scaffolds and implants in tissue engineering applications though has often been used as a block amongst copolymers for NP preparation^[158].

Another major piece of the polymeric NP toolkit is poly(ethylene glycol) (PEG), a hydrophilic polymer that is often conjugated to the more hydrophobic polymers described above and has been well-characterized to bestow ‘stealth’ properties when present on the surface of NPs^[159]. The relative lack of immune reactivity of PEG was first demonstrated through the administration of PEG conjugated to bovine serum albumin (BSA) in rabbits in comparison to non-PEGylated BSA^[160]. Furthermore, PEGylation has been demonstrated to increase circulation time and reduce tissue uptake of conjugates and nanocarriers in a variety of animal models and patients^[159]. Thus, PEG conjugates became an important area of development and several PEGylated enzymes have been approved for clinical treatment of leukemia and Hepatitis C^[161]. PEG has since been incorporated as a component of liposomes and polymeric micelles(Fig. 6A,D,E)^[162],[163]. PEG has become ubiquitous as the outer coating of many NPs as a block copolymer with PLA, PLGA, PCL and other polymers, such as the cationic poly(ethyleneimine) (PEI) and poly-lysine, which are common nucleic acid delivery vehicles^[164]. Though PEG remains the most common way to extend the half-life of NPs and other nanocarriers, some alternatives have emerged, such as linear polyglycerol or hyperbranched polyglycerol (HPG)^[165] (Fig. 6H). HPG-coated NPs have demonstrated even longer blood circulation, greater stability, and greater efficacy in animal models than PEGylated NPs though it has not yet been as widely adopted^[165]. The polymers discussed in this section are not an exhaustive list and many more complex multi-block polymers have emerged that add further advantages and challenges to the milieu of NP-based delivery^[166],[167].

Nanoparticle Characterization Methods

Nanoparticle characterization typically involves measuring size, polydispersity, surface charge, stability, degradation, drug loading, release kinetics, and occasionally mechanically properties related to flow and shear stress. All of these characteristics are important for most delivery applications though are especially crucial to optimize for the brain tumor microenvironment.

Nanoparticle size is unsurprisingly the most fundamental parameter for successful delivery to target cells as the method of cellular uptake is highly dependent on the size of material to be taken up as well as shape and surface chemistry^[168]. NP is most often characterized by a combination of dynamic light scattering (DLS) and electron microscopy, including the transmission (TEM), scanning (SEM), and cryo (cryo-EM) varieties. DLS involves the detection of the scattering fluctuations of a laser caused by random Brownian motion of sub-micron particles in solution^[169]. The autocorrelation function of the signal is then determined and used to calculate the diffusion coefficient, which in turn can be used to estimate the hydrodynamic diameter of the NPs in suspension via the Stokes-Einstein equation. DLS is a fairly convenient method to use for size characterization though is known to be inconsistent with the more determinative EM techniques, likely due to effects of hydration shells that are present in hydrated samples during DLS ^[170]. DLS also allows an estimate of the polydispersity of the NP sample, which is meant to represent the breadth of the NP distribution. SEM is a powerful technique for observing the general shape and surface of materials by utilizing scattered electrons from the NP sample to produce an image of the NP surfaces^[170]. But SEM is sometimes difficult to interpret as sample preparation requires coating the NPs with molten carbon or gold, often altering or even destroying the NPs’ structure. TEM is perhaps more reliable as it does not alter the NP as easily as it requires a less harsh ‘staining’ methodology involving deposition of heavy metal-based agents (such as uranium or tungsten) onto the NPs’ surfaces^[171]. Cryo-EM presents the most recent advance in imaging of NPs and involves the preservation of NPs in a frozen/hydrated state, allowing greater visualization of an unaltered sample^[172]. An additional technique is Atomic Force Microscopy (AFM), which uses a cantilever probe to measure the repulsion force of the surface of a NP sample to generate a ‘roughness’ profile for the sample^[173]. AFM can provide valuable information related to size and surface characteristics of NPs but can be easily misinterpreted in circumstances of unevenly distributed and poorly prepared samples.

Surface charge is often represented by measuring the ζ-potential, a voltage typically determined through electrophoresis and related to the movement of colloidal particles within the electrophoretic cell^[174]. NPs with ζ-potentials within the range of +/−10 mV are considered to have approximately neutral surface charge while those with more extreme potentials are considered to be cationic or anionic (and typically dependent on the pH of the solution)^[175]. The ζ-potential of a NP solution is also considered to be related to the stability of the NP suspension, as NPs with greater surface charge are more likely to electrostatically repel each other and thus not aggregate as easily. Nanoparticle stability can also be assessed through long-term incubation and repeated size measurements utilizing DLS or other techniques.

More complex techniques such as X-ray photoelectron spectroscopy (XPS), Fourier Transform Infrared (FTIR) spectroscopy, and Raman spectroscopy can be used to better understand NPs’ structure and composition. XPS can be used to determine the surface composition of NPs by detection of electrons that escape from the surface of the NP after irradiation^[176]. FTIR relies on low polarizability of light and large dipole moments while Raman utilizes high polarizability of scattered light by the NP sample in order to elucidate details of the chemical composition^[177].

Drug loading and release kinetics are typically determined by measuring drug concentrations with High Performance Liquid Chromatography (HPLC), a technique that utilizes high pressure and solvent gradients to elute to a drug through a chromatography column. This technique can precisely determine the concentration of a drug in a solution when compared against standards within a linear range. Alternatively, differential scanning calorimetry (DSC) is often utilized to understand the relationship between a polymeric NP matrix and the state of the drug within it. DSC measures temperature-dependent state transitions of a polymer, such as the glass transition temperature (T_g) and can be used to understand how drug-polymer interactions alter the structure of a NP when compared to unloaded NPs and simple mixtures of the two agents^[178].

Drug Release from Polymeric Nanoparticles

The extent and rate of drug release from NPs are amongst the most crucial properties for determining their efficacy as well as their applicability to particular disease states. One important goal for using polymeric delivery systems is to control the release of drug over time, while the use of delivery approaches such as CED aims to control of drug distribution^[179]. Drug release is dependent on NP structure, degradation properties, polymer-drug interactions, and solubility in the release medium. The most fundamental mechanisms of release involve simple diffusion through pores in the polymer structure that tend to grow as the polymer absorbs water, simple diffusion through the polymer matrix itself, and convective transport due to fluid movement along osmotic gradients^[180]. Nanoparticle degradation also influences the methods of transport available to an encapsulated drug. Surface erosion involves degradation primarily along the exterior of the NP, which is often ideal for controlling the rate of release and achieving first order kinetics. Bulk erosion occurs with more porous or hydrophilic polymers that allow water to permeate the NP more easily and thus is not as reproducible^[180].

Drug release data is most often collected by one of two sets of methods: dialysis and ultracentrifugation/filtration^[181]. Dialysis drug release methods typically involve collecting samples of dialysis buffer surrounding or adjacent to a dialysis membrane containing the release sample (Fig. 7A). The ultracentrifugation method typically involves centrifugation the NP suspension to separate free drug in solution from the NPs (Fig. 7B). This can be achieved by simple centrifugation methods in the case of larger NPs but requires centrifugal filtration for smaller NPs that do not pellet as easily^[182]. The supernatant or filtrate is typically collected at different time points and drug concentration measured with relevant analytical methods such as HPLC. Both dialysis and ultrafiltration methods are imperfect and especially complicated when used for hydrophobic drugs, which do not quicky solvate and diffuse from the NPs in aqueous solutions (Fig. 7C). Thus, hydrophobic drug release often requires a sink condition, which could be thought of a simulation of drug clearance or binding from an in vivo medium, in order to force the equilibrium to continue the release of the drug^[183]. The use of sink conditions, though a useful technical tool, leads to difficulty in interpreting release kinetics relevance to in vivo conditions as the sink conditions are not typically equivalent to what will be encountered in vivo and can lead to an artificial saturation of release kinetics^[184]. One further limitation of these techniques is the possibility of drug loss through aggregation or adsorption to either the centrifuge filter or the dialysis membrane^[185]. More advanced techniques have been developed in more recent years that do not have these same challenges, such as the use of DSC to examine drug release relative to NP polymer structure as well as using electrochemical cells to measure drug release in real-time based on voltage changes between different compartments^[186],[187]. Although these techniques require more advanced equipment, they offer more direct methodologies for understanding release. But they are not as well-standardized and require greater investigation before wider implementation.

Figure 7:

Free drug diffusion from nanoparticles into a dialysis bag and ultimately through the membrane during a release assay (A). An ultracentrifuge tube with a filter, in which drug loaded nanoparticles remain, while supernatant containing free drug has collected in the bottom of the tube (B). Unpublished data comparing drug release profiles of the same drug-loaded polymeric nanoparticle sample collected using different means with β-cyclodextrin as an artificial sink condition (C). Created with Biorender.com.

Consequently, a variety of mathematical models have been proposed to explain NP release kinetics, many of which have been applied to or were derived based on release from other polymeric systems, such as implants. The zero-order drug release model (Eq. 1) represents polymeric systems that release drugs slowly at a constant rate and can be used to represent polymeric systems with slow degradation and low porosity as well as highly-controlled osmotic systems:

$$C_t=C_{\text{0}}+k_{\text{0}}t$$ (1)

Here, C_t represents the amount of drug in the release medium at a particular time point (t), C₀ represents in the initial amount of drug in the release medium, and k₀ represents the rate of drug release^[188]. The first-order release model (Eq. 2) is more widely applicable though also empirical and has been used for some liposomes and reversible binding systems:

$$C_t=C_0{e}^{-kt}$$ (2)

Where C_t represents the drug released at the specified timepoint (t), C₀ represents the initial drug concentration, and k represents the drug release rate constant. The Higuchi model is a more complex way to represent a multi-phase release system, which is common amongst some NP systems with different degradation phases. The Higuchi model (Eq. 3) was initially developed for polymer matrices with planar geometry but has since been adapted to other systems^[189]. The generalized and simplified Higuchi equation is:

$$C_t=K_H\sqrt{t}$$ (3)

where K_H is the Higuchi constant that accounts for Fick’s Law of Diffusion^[190]. The complexity and assumptions involved in the Higuchi model make it less applicable to NP systems than other types of drug carriers. The Hixson-Crowell model (Eq. 4) takes into account geometry and does not assume constant size of the drug delivery system; it has primarily been used for tablet-based formulations but has been investigated in NP applications as well:

$$\sqrt[3]{C_0}=\sqrt[3]{C_t}+\kappa t$$ (4)

where C₀ represents the initial amount of drug loaded in the delivery system, C_t represents the remaining drug, and κ is a constant representing the relationship between surface area and volume^[191]. The Korsmeyer-Peppas model (Eq. 5) was developed to represent release from porous cylindrical polymeric systems:

$$C_t=K{t}^n$$ (5)

where C_t represents the drug released at particular timepoint (t), K represents the release rate constant, and n is an index that represents different release kinetics, where different values of n can approximate some of the other models previously described^[192]. Finally, the Weibull model (Eq. 6) is a commonly used empirical model and has been used extensively to model biphasic NP release:

$$C_t=C_0(1-e^{-k(t-T)})$$ (6)

where C_t represents the drug release at a timepoint (t), C₀ represents the maximal drug release, T represents a time delay due to drug dissolution, and k is rate constant^[188]. The Weibull model has been shown to yield the best fit with least error for various NP formulations^[193],[194]. Polymeric nanoparticles often exhibit a tri or biphasic release pattern. In both cases there is an initial burst release of drug adsorbed to or near the surface of the NP, which is problematic for applications where well-controlled release or long-term sustained release is necessary^[195]. In the triphasic release profile, the second phase is often related to diffusion-based drug release while the polymer structure remains stable while the third phase of release is due to a more rapid degradation phase^[196]. The biphasic release patterns tend to occur due to either very low degradation rates and hence a very long sustained release phase or relatively rapid degradation that overcomes simple diffusion. However, most release patterns lie somewhere in between and are more complex due to differences in NP porosity, tortuosity, drug-polymer interactions, NP size, and other physicochemical properties of the polymer and drug. Thus, it is important to consider multiple models, such as those described above, when interpreting drug release kinetics (Fig. 8).

Figure 8:

Simulated implementations of the specified drug release models, demonstrating the range of possible release kinetic profiles for different materials and applications.

Impact of Protein Corona Formation on Nanoparticle Pharmacokinetics

The formation of protein coronas surrounding NPs in biological environments has become a crucial focus of recent investigations, particularly with regard to its implant on pharmacokinetics, in vivo stability, and off target toxicity. The formation of a protein corona has been shown to alter the properties of surface-functionalized NPs, such as reducing the binding affinity of transferrin-coated NPs to associated receptors^[197]. The presence of NP-adsorbed proteins has been detected at various stages of cellular internalization, suggesting that the corona composition is important for regulating NP endocytosis and intracellular trafficking^[198]. Protein coronas are thought to form quickly on the surface of most NPs during blood circulation, and are dependent on size and surface charge^[199]. Alternatively, protein corona composition has been determined to be potentially beneficial with regard to reducing the toxicity associated with some NP materials such as metals, and reducing NP aggregation^[200]. Ultimately, optimization of the protein corona composition based on NP properties has become an important route toward improving NP pharmacokinetics and safety^[201].

Nanoparticles for Glioblastoma Treatment and Optimization for Brain Delivery

Polymeric NP systems are at the forefront of exploration towards improving drug delivery to the brain and in particular to improve chemotherapy delivery for GBM. A subset of these efforts focus primarily on using special surface modifications to allow for NP passage through the BBB by taking advantage of endocytotic mechanisms^[202]. A series of studies into poly(butylcyanoacrylate) (PBCA) loaded with the peptide dalargin demonstrated analgesic efficacy in mice and uptake by endothelial cells at the BBB^[203]. Furthermore, it was determined that the most important aspect of the NPs that allowed such BBB penetration was the use of polysorbate 80 (or Tween 80) as a surfactant in the NP formulation process^[204]. These studies have suggested that Tween 80 interacts with the ECs of the BBB to improve the efficiency of uptake and transport into the brain parenchyma. A similar brain penetration effect was observed with PLGA NPs and Tween 80 though with even greater tissue penetration when poloxamer 188 (F68) was used instead^[205]. Tween 80 has also been shown to inhibit the drug efflux transporter P-glycoprotein as well as increase the adsorption of apolipoprotein E (ApoE), an important factor involved in trafficking of LDL in the brain, on NP surfaces^[206]. Surface modification of NPs has also been effective at increasing NP penetration through the BBB, such as with the trans-activating transcriptor (TAT) peptide on PLA NPs or glutathione on PLGA NPs^[207],[208]. Furthermore, transferrin-conjugated PEG-PCL NPs have shown enhanced delivery of doxorubicin across the BBB^[209]. Thus, surface modification has proven to be a valuable strategy toward allowing some NP formulations to pass through the BBB, though many of these studies have demonstrated only a limited portion of the initial drug dose reached the brain and have also demonstrated neurotoxicity concerns^{[210],[211],[212]}. Since only a small fraction of the drug-carrying NPs enter the brain, there is concern about side effects due to the larger fraction of NPs that remain extracranial. An alternative strategy has also been to use cationic polymers, such as chitosan or PEI, to deliver proteins or nucleic acids across the BBB^[213],[214].

Many NP systems have been designed specifically toward GBM treatment. Various surface modifications have been used to specifically target glioma cells in vitro and in vivo, such as the LDL receptor targeted peptide 22 or the urokinase plasminogen activator (uPA) responsive cell penetrating peptide that have been conjugated to paclitaxel loaded PLA-PEG NPs^[215],[216]. Both of these surface modifications target receptors that are overexpressed in glioma cells and both examples demonstrated robust NP distribution with tumors in mouse models as well as somewhat greater efficacy than unmodified NPs and free drug. NPs with encapsulated TMZ and modified with chlorotoxin, a tumor targeting peptide, have demonstrated greatly increased TMZ half-life compared to free drug as well as tumor targeted delivery^[217]. Numerous other chemotherapeutics, such as docetaxel, camptothecin, doxorubicin, and lomustine have been loaded into NPs composed of primarily surface modified PLGA, PLGA-PEG, PLA-PEG, PCL-PEG, chitosan, and lipids, demonstrating greater efficacy than free drug in rodent GBM models^{[218],[219],[220],[221],[222]}. Alternatively, iron oxide and other metallic NPs have utilized to deliver the EGFR targeted Cetuximab to rodent intracranial tumors and, in some cases, for the separate, investigational approach of inducing magnetic hyperthermia to ablate the tumors^[223],[224]z.

Due to the variation in design features and approaches toward GBM targeting, there not yet a standard profile for NP design. However, there are some guiding principles and features that are important to consider. Certain properties are more crucial for blood brain barrier penetrance, such as lipophilicity, while properties such as size are important for tumor and tissue distribution, cell uptake, and clearance rate. Furthermore, NP surface chemistry governs multiple facets of GBM targeting, including tumor cell specificity and uptake. Furthermore, the interactions between the polymer and cargo are important to consider to optimize the release rate and half-life of the system; an expanded analysis of these design criteria is presented in Table 3. Thus, the plethora of combinations of NPs and drugs have clearly demonstrated the promise of the paradigm though relatively few have been tested clinically for GBM. Some of the biggest hurdles toward clinical implementation of NP-based drug delivery for GBM are the lack of investigation into brain penetration of these NP systems in larger mammalian brains as well as off-target toxicity when administered systemically^[225].

Table 3:

Summary of Design Criteria for Nanoparticle Delivery to Glioblastoma

Design Criterion	Optimal Parameters	Purpose	Examples/References
Hydrophobicity	Amphiphilic	Pass through BBB/cell membranes	Fully hydrophilic NPs flatten during membrane transport leading to more difficult passage while hydrophobic NPs elongate leading to easier diffusion; Amphiphilic NPs retain integrity while also passing through membrane relatively easily[265]
Size	>50 nm, <200 nm	Ensure penetration through ECS and cell uptake; Reduce NP clearance from Brain ECS	PLGA Brain Penetrating NPs[233], Polystyrene NPs of 200 nm and less demonstrate increased cell uptake and membrane transport[266], TPGS coated Polystyrene NPs demonstrate best uptake/transport at 100 nm compared to other sizes[267]
Surface Charge	Cationic	Uptake in Glioblastoma Stem Cells	Polyurethane-PEI NPs loaded with miR-145[268], PAMAM conjugated to PLGA-PEG NPs with paclitaxel[269]
	Neutral/Neutral	Distribution through Brain ECS	PLGA Brain Penetrating NPs[233], Neutral and Anionic NPs less toxic than Cationic NPs[270]
PEGylation	Dense Coating	Brain ECS Distribution/Reduce Clearance	Paclitaxel Polystyrene NPs coated with Dense PEG[234]
Surface Chemistry	Tumor-specific Targeting	Enhance tumor uptake/reduce off target cell uptake, BBB penetration	Transferrin-conjugated[271], Apolipoprotein E[245], Angiopep-2[272], RGD peptide[273], Antibody-conjugated (Cetuximab)[274]
Internal Chemistry	Hydrophobic; Biodegradation	Control release rate of cargo; Nanoparticle Stability	Poorly soluble drugs encapsulated better in hydrophobic NP core[275]
Surfactant Presence/Type	Surfactant Dependent	Prevent off target toxicities/Enhance cell uptake/Other Properties	Polysorbate-80-coating of PBCA NPs enhances uptake and BBB penetration[276], Poloxamer 188 increases PLGA NP size and Polyvinyl Alcohol decreases it[277]
Nanoparticle Structure	Solid Nanoparticle	Prevent disruption of NP during transport and different media	Core-shell Spherical PLGA-PEG NP structure prevents destabilization by salt solution compared to PLGA NPs[278]

Convection Enhanced Delivery of Nanocarriers to the Brain

The leading edge of investigational NP administration to the CNS has involved a coupling with CED to improve the spatial distribution and tissue penetration compared to systemic administration. Various studies have examined the ideal NP characteristics to achieve maximal tissue distribution and have focused primarily on size and surface characteristics; Early studies with liposomes demonstrated that small NPs (<50 nm) with PEGylated or neutral surfaces distributed more widely than larger and positively charged NPs^{[226],[227],[228]}. Furthermore, studies of CED with polymeric microspheres loaded with a variety of agents have demonstrated modest efficacy in rodent tumor models though their efficacy is likely limited by the large size of the microspheres relative to the pores in the brain ECS^{[202],[229],[230],[231]}. Thus, nanoparticles and other nanocarriers have emerged as promising vehicles for CED to brain tumors. PLGA and PLGA-PEG have been used for the delivery of myriad drugs, including camptothecin, dithiazinine iodide, paclitaxel, and carboplatin, demonstrating survival benefit when compared to free drug administered in the same manner^{[232],[233],[234],[235]}. Debates have ‘raged’ within the literature regarding the influence of size or the density of PEGylation on NP surface as the primary factor to achieve highly penetrative NPs, though it appears optimizing both has achieved similar success in rodent models suggesting that PEGylation may not always be necessary^{[233],[234],[236]}. Furthermore, studies comparing NP distribution in tumor bearing rodent brains to healthy brains demonstrated greater heterogeneity and incomplete tumor coverage^[237],[238]. However, surface functionalization of the NPs appeared to selectively favor uptake by tumor cells when compared to healthy cells, suggesting a combination of stealthy and tumor-targeted surface chemistry can assist in the balance between distribution and uptake^[238]. Other strategies have emerged to improve NP distribution after CED, such as osmolar dilation of the extracellular space and degradation of the extracellular matrix, which both expand the volume of distribution but could be of concern as their potential toxicities have not been fully investigated^[239],[240]. Alternatively, taking advantage of NP transport utilizing perivascular pathways has also been suggested as a means to preferentially deliver NPs to vascularized tumors, though there is some risk of NPs that are too large accumulating within these spaces, thereby limiting distribution^[241]. CED of polymeric NPs including MRI contrast agents has been validated in some larger animal glioma models, such as porcine (pig) and canine (dog), however as seen in rodents, complete tumor coverage was not universally achieved^[233],[242]. Ultimately, continued development of CED with NPs should incorporate the best catheter technologies and state-of-the-art experience gained through human trials of free-drug CED.

In recent years, investigations into NP- based CED for GBM have focused on synergistic approaches, in which a dose-limited therapy combination, when delivered systemically, can be utilized more effectively when one agent is sequestered within a NP formulation in the brain. As an example, the ATR inhibitor VE822 (now berzosertib) has been encapsulated in both PLA-PEG and PEG-poly(ω--pentadecalactone-co-p-dioxanone) (PEG-PDL-co-DO) and delivered through CED to rodent GBM models as a radiosensitizer^[243],[244]. Both of these studies demonstrated survival benefit of the combination therapy compared to either therapy alone, which may allow for better targeting of radiotherapy toward the tumors to reduce off target toxicity. Also, other NP-encapsulated agents, such as anti-miR-21 and the DNA intercalators oxaliplatin and 56MESS, have elicited survival benefit in combination with TMZ and reduced tumor growth when delivered through CED in rodent models^[245],[246]. Thus, CED of NP-encapsulated drugs has shown robust promise in expanding the library of drug combinations that are possible to use for GBM treatment by eliminating some of the safety concerns of systemic administration.

Conclusion

Progress towards treating GBM has been slow but technologies such as polymeric nanocarriers and modalities such as convection enhanced delivery have created hope toward overcoming the difficulties associated with brain delivery. As the library of polymers and therapeutic agents has expanded, the challenge of how to select the best possible combination of carrier, treatment paradigm, and delivery modality for a particular tumor phenotype has emerged. Alongside technological development, the pursuit of a more comprehensive understanding of material nanoparticle properties, pharmacokinetics, and compatibility with delivery devices will provide a foundation for translating these systems into the clinic. As clinicians have gained more experience in implementing CED with free drug formulations and with a renewed, broader adoption of nanocarrier technologies since the COVID-19 pandemic, it is only a matter of time before the clinical manifestation of the CED of nanoparticles for GBM patients. Ultimately, combination therapies have begun take center stage, as opposed to single agents, to address the heterogeneity and resistance of the disease. Nanoparticles and similar drug delivery modalities will be crucial for controlling the dosing and safety profile of such combinations.