|Home | About | Journals | Submit | Contact Us | Français|
Brain tumors—particularly glioblastoma multiforme (GBM)—pose an important public health problem in the US. Despite surgical and medical advances, the prognosis for patients with malignant gliomas remains grim: current therapy for is insufficient with nearly universal recurrence. A major reason for this failure is the difficulty of delivering therapeutic agents to the brain: better delivery approaches are needed to improve treatment. In this article, we summarize recent progress in drug delivery to the brain, with an emphasis on convection-enhanced delivery of nanocarriers. We examine the potential of new delivery methods to permit novel drug- and gene-based therapies that target brain cancer stem cells (BCSCs) and discuss the use of nanomaterials for imaging of tumors and drug delivery.
Improved methods for treating glioblastoma multiforme (GBM) and other malignant tumors in the brain are desperately needed. Malignant gliomas are the most common primary malignant brain tumors in adults. More than 15,000 new cases are diagnosed in the United States each year (1, 2). Despite surgical and medical advances, the five-year survival rate for GBM, the most common malignant glioma, has been a dismal 4% for the past few decades (3, 4). Current treatments for GBM are insufficient with a nearly universal recurrence after surgery, radiation therapy, and chemotherapy (5).
A major problem in therapy for GBM and other CNS diseases is getting drugs into the brain. The blood-brain barrier (BBB) prevents the transport of most systemically-delivered molecules into the brain. Recent research has created better methods for drug delivery to the brain. Here, we review the state-of-the-art in the development of new approaches for drug delivery to the brain, highlighting those methods that are most relevant to GBM.
The BBB limits the entry of systemically administered drugs to the brain, making delivery behind the BBB an appealing strategy. One method for accomplishing delivery behind the BBB is implantation of a drug-eluting material into the brain tissue. Controlled release systems for direct delivery of chemotherapy to brain tumors were first approved by the FDA in 1996 (6, 7). The rationale for this approach is that biocompatible materials can be introduced directly into the brain, perhaps during surgical resection of the tumor. If the materials are loaded with drugs and are engineered such that the drug is slowly released after implantation, then their implantation into the brain can provide long-term chemotherapy at the tumor site, circumventing the need for BBB penetration. In this approach, drug penetrates into the tissue near the implant by diffusion, producing the highest drug concentrations in the region most in need of treatment and eliminating side effects caused by delivery of drug to other tissues (Figure 1a).
The most successful controlled release systems are based on biocompatible polymers, particularly polymers that are also biodegradable. The first clinical trials of this approach used a 1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU)-loaded polyanhydride delivery system (8–11): BCNU is an alkylating agent, which was the most commonly used chemotherapy drug for GBM in the 1990s. An early clinical study demonstrated that the procedure was safe (12) and a subsequent placebo-controlled clinical trial showed that implantation of BCNU-loaded polymers was more effective than standard therapies in patients with recurrent GBM (13). This new method for brain tumor therapy was rapidly adopted (the clinical product is called Gliadel®), leading to dozens of clinical studies by investigators around the world, and improved lifespan for many brain tumor patients. Experimental controlled delivery systems have been produced with a variety of other compounds, including methotrexate (14), paclitaxel (10, 15), steroids (16), and camptothecin (CPT) (17), but so far none have been tested in humans.
The safety and effectiveness of this new approach for treating GBM has been established, but the increases in survival observed in clinical trials with Gliadel® are modest. However, diffusion-based implant systems—such as Gliadel®--can produce substantial increases in the duration of survival of animals, and optimal formulations can eliminate tumors in animal models. Implantation of a single polymer implant in the brain of an animal produces unprecedented concentrations of active BCNU within the tissue near the implant (10, 11). The contrast between these observations in animal models and clinical studies suggests that one of the limitations of Gliadel® is the small volume treated by diffusion-based delivery systems.
Convection-enhanced delivery (CED) is a promising approach to overcome the limited distribution volume associated with diffusion-based delivery systems. For diffusion-based delivery, drug molecules move passively from regions of high concentration to regions of lower concentration. As a result, large molecules such as antibodies diffuse no more than 1 mm in 3 days, and small drugs that may have better diffusion are often quickly eliminated by capillary clearance or metabolism (18). In contrast, for dispersion using convection, agents are delivered to the brain via flow through a cannula under constant pressure. In this scenario, the dispersion of agents is powered by bulk flow kinetics or gradients of pressure, in addition to gradient of concentrations. As a result, it is possible to distribute agents widely in the brain (Figure 1b). CED technology and factors affecting drug distribution by CED have been extensively reviewed (18, 19) and thus will not be discussed in detail here.
CED to the brain was first reported in the early 1990s (20). Since then, CED has been used in clinical trials, but this experience has revealed some limitations. Conventional CED of drug solutions results in an increased depth of penetration, but these results are transient. Free drugs are subject to high rates of elimination (i.e. they are transported into cerebrospinal fluid (CSF) or blood) or have short half-lives in the brain; therefore, they disappear soon after the infusion stops (18). Interstitial fluid flow might particularly impact CED of free agents to brain tumors. Mathematical models—and some experimental evidence—suggest that tumors maintain an elevated interstitial fluid pressure in their cores (21), which produces an outwardly-directed flow of extracellular fluid at their periphery (22). These limitations could explain the failure of the recent PRECISE trial, in which a potent targeted toxin, CB (cintredekin besudotox, IL13-PE38QQR) in aqueous suspension, was delivered to brain tumors via CED but failed to show advantage when compared to diffusion-based Gliadel® wafers (23).
To improve CED, agents can be loaded into nanocarriers, such as liposomes, micelles, dendrimers, or polymeric nanoparticles, which are small enough (typically 100–200 nm in diameter) to allow for infusion and penetration in brain. These carriers can protect therapeutic agents from loss and control their release for long periods after infusion (Figure 2). Nanoparticle delivery systems for drugs have been available for many years (24). Some research groups focus on the use of nanoparticles introduced systemically, with the hopes that some of these particles will enter the brain through the BBB. This approach appears to work in some cases, but the percentage of intravenously administered particles that enter the brain is very low (25–27). It is not yet clear whether sufficient quantities of drug can be delivered by systemically-administered nanoparticles to make this a useful method for treating tumors in the brain (although there is some evidence that systemically-administered nanoparticles may be useful for diagnostic purposes, such as iron oxide-containing nanoparticles that facilitate imaging of brain tumors (28)). An alternate approach is to deliver the nanoparticles directly into the brain, perhaps using CED to facilitate the distribution of the nanoparticles throughout the volume of the brain that needs therapy (Figure 1c).
CED of lipid-based nanoparticle systems, such as liposomes, has been well studied. Liposomes are artificial phospholipid vesicles that form a “core-shell” structure and can be readily loaded with therapeutic agents. CED of liposomes has been evaluated in rat brain tumor xenografts (29–32) and canine spontaneous brain tumors (33); CED has also been tested in the normal brain of non-human primates (34, 35). Compared to CED of free drugs, CED of liposomes increases the distribution of delivered agents (36), reduces toxicity, and extends half-life (31, 33, 37). CED of liposomes has recently been tested in human GBM patients: it appears to be safe and to provide some therapeutic benefit (38, 39).
As an alternate to liposomes, polymer nanoparticles can be delivered via CED. In a recent study, CED of camptothecin-loaded nanoparticles led to longer survival in animals with intracranial tumors than CED of camptothecin alone (40). Previous efforts to deliver polymer nanoparticles via CED have achieved a limited volume of distribution, due to an inability to fabricate polymer nanoparticles of small enough size to allow for unhindered interstitial convective transport (40, 41). The pore size of the normal brain extracellular space is 38–64 nm (42) versus ~70–100 nm within a tumor in the brain (43), which suggests that nanoparticles for CED to intracranial tumors should be 60–80 nm in diameter to allow for access to tumorous tissue while sparing the normal brain. We recently developed methods for fabricating non-aggregrating PLGA nanoparticles of this size (44). Our data suggest that these nanoparticles are able to penetrate brain tissue with a mean volume of distribution that is at least 10-fold larger than previously published formulations (Zhou J. unpublished data).
The most recent advance in drug therapy of GBM was the FDA approval in 2005 of temozolomide (TMZ), which has since replaced BCNU as standard therapy. Both TMZ and BCNU cross the BBB, but TMZ produces fewer side effects and is bioavailable after oral administration (BCNU is administered by infusion or through implantation of Gliadel®) (45–47). Even with these drugs, the median survival for GBM patients is less than 15 months: drug resistance and tumor recurrence remain major challenges.
In the past, drugs needed to cross the BBB to be candidates for use in treating GBM. New delivery technologies—such as implantable controlled release systems, CED, and nanoparticles—make it possible to use drugs that do not cross the BBB into the brain. These technologies create a new opportunity: drugs can be selected based solely on their potential for effectiveness against tumors, without regard for their ability to cross the BBB. For example, it is now clear that paclitaxel—which does not cross the BBB—is an effective agent for treating brain tumors when administered via paclitaxel-loaded implants or CED of paclitaxel nanoparticles or microparticles (15, 48). Likewise, CED permitted the clinical testing of immunotoxins, which do not cross the BBB, in the PRECISE trial (23).
Many experimental drugs are currently being examined for treatment of GBM (Table 1): these agents target a wide variety of molecular pathways important in glioma cell survival. But few of these agents will cross the BBB at doses sufficient to treat tumors in the brain without systemic side effects. Therefore, using conventional methods of drug delivery (infusion or oral administration), these novel agents will likely suffer from the same problem that has plagued drug therapy of brain tumors for decades: it will be difficult to deliver these agents to human tumors at a sufficient dose for effectiveness without also creating serious toxicity. New delivery systems may enhance the clinical value of some of these agents, by allowing sustained local delivery that concentrates the drug at the target.
Further, it is becoming increasingly clear that certain cells within tumors play a major role in tumor progression, and that these cells are not sensitive to current drugs (49, 50). A small fraction of glioma cells has been identified as brain cancer stem cells (BCSCs), of which many are CD133+ (51–58). These cells have features similar to primitive neural stem cells, but they also have tumor-initiating functions (59). BCSCs appear to arise from deranged neural stem cell or glial cell progenitors (50–54): these cells have the ability to drive tumor formation, promote angiogenesis, and influence tumor cell migration (50, 55, 60). Current chemotherapeutic agents do not eliminate BCSCs effectively (49, 61, 62): BCSCs in culture are resistant to standard chemotherapy drugs, including temozolomide, carboplatin, cisplatin, paclitaxel, doxorubicin, vincristine, methotrexate, and etoposide (63–66). BCSCs are also resistant to standard radiotherapy (55). These observations suggest that current therapeutic regimens may produce short-term remissions, but are unlikely to cure GBM: long-term remissions require eradication of BCSCs (49, 50, 61, 67).
One promising approach to identifying novel therapies for BCSCs is to target key signaling pathways governing BCSC proliferation and self-renewal, such as the PTEN/PI3K/Akt pathway (68). Inhibition of the Akt pathway induces BCSC differentiation and inhibits self-renewal and tumorigenicity in vivo (69, 70). Developmental pathways, including the Notch (71, 72) and BMP pathways (73), are important for BCSC self-renewal and differentiation and can be targeted to preferentially eliminate BCSCs. In addition, several other signaling pathways, including EGFR (74), TGF-® (75) and HIF (76), are critical for BCSC proliferation, survival, and self-renewal. An alternate approach is to take advantage of recent advances in genomics and informatics to identify therapeutic agents via high-throughput screening. For example, an RNAi screen against a kinome library successfully identified TRRAP as a regulator of BCSC differentiation (77). More recently, eight small molecule drugs that preferentially inhibit BCSCs from GBM over tumor-matched non-BCSC GBM cells were identified from a library of over 30,000 (78). These early successes suggest that high-throughput screening is a promising approach for identifying novel BCSC-targeted therapeutics. Some of the most promising drugs now being examined for GBM therapy (Table 1) were identified through their activity on BCSCs.
Delivery remains a major obstacle for use of any agent that targets BCSC. The new delivery technologies described earlier will likely prove necessary for translating these agents to clinical practice.
The promise of gene therapy in GBM has long been known, but the difficulty carrying it out is equally well recognized. An early gene therapy approach involved introduction of genes at the tumor site that would activate systemically administered prodrugs (79). One such system, which employed herpes simplex virus thymidine kinase and ganciclovir showed success in mouse models (80, 81), but in a phase III trial post-operative delivery failed to show benefit compared to post-operative radiotherapy (82). Other approaches have since emerged, particularly since the recognition of the role of microRNA (miRNA) in tumors and the power of RNA interference for silencing genes.
A variety of gene therapy approaches may be useful in GBM. Deletion of known tumor suppressors, such as PTEN, is a common mechanism of gliomagenesis (83). Rescued expression of PTEN in U87 glioma cells suppresses tumorigenicity in vivo and promotes entry into G1 phase of the cell cycle (84). Genes in the tumor-suppressing p53 pathway (85, 86), such as TP53, MDM2/4 and p14-ARF, and the retinoblastoma (Rb) pathway (85, 87), such as Rb, CDK4/6 and p16-INK4A, are also promising targets for treatment of GBM. Alternately, brain tumor growth can potentially be inhibited by blocking expression of genes, such as EGFR (88). High-throughput sequencing efforts and the compilation of data into The Cancer Genome Atlas, as well as the development of the Cancer Bioinformatics Grid have led to the identification of a number of highly novel chromosomal mutations, amplifications, and deletions that are important in GBM (85). Chief among them is mutation to isocitrate dehydrogenase (IDH) 1 and 2 (89). Mutated IDH1 leads to induction of the HIF pathway (90), which leads to an increased fraction of BCSCs (91).
MicroRNAs play important roles in gliomagenesis and can be targets for treatment. Reduced expression of a number of miRNAs has been noted in GBM (92). miR-7 is capable of repressing EGFR signaling (93), so rescue or ectopic expression of miR-7 in tumor cells may limit GBM progression. Expression of miR-221, an antiapoptotic factor, is markedly elevated in human GBM tissues (94), and inhibition of miR-221 decreases tumor growth in tumor xenograft (95). miR-10b, which inhibits cell cycle arrest and apoptosis while promoting growth through p16-INK4A and p21 targeting, is another miRNA known to be elevated in GBM tissue samples and cell lines (96). Several other miRNAs are important for brain tumor cell survival and represent potential targets for therapy (97, 98).
Delivery remains the major obstacle for gene-based therapies. Viral vectors—including retrovirus (82), adenovirus (99) and adeno-associated virus (100, 101)—are often used in evaluations in animals or humans, due to their high transduction efficiency. In addition to safety concerns, free viral particles in suspension may not reach disseminated tumor cells (102, 103). Synthetic non-viral vectors, such as nanoparticle-based gene delivery vehicles, might be a better approach due to their limited immunogenicity, ability to accommodate and deliver large pieces of genetic material, and potential for modification of their surface structures to allow targeting. Among non-viral vectors, cationic lipid-based systems, including liposome and solid lipid nanoparticles, are the most extensively studied (104, 105). Liposome-based gene delivery vehicles have been used for gene therapy to brain tumors in animals (106, 107) and have shown some promise in clinical trials (38, 39, 108). Cationic polymers, such as dendrimer- (109, 110) and polyethyleneimine-based nanocarriers (111), also have potential for gene delivery to brain tumors. But these highly charged vectors are usually toxic; because of excess positive charge at their surface, these nanocarriers inhibit normal cellular processes, such as clathrin-mediated endocytosis (112) and cell survival signaling (113), leading to substantial toxicity (112, 114, 115).
Most lipid and polymer gene delivery systems rely exclusively on cation density to form complexes with DNA. But it has recently been discovered that the ability of polymer nanocarriers to deliver genes depends on factors other than charge density, particularly polymer molecular weight and hydrophobicity. High molecular weight and increased hydrophobicity can compensate for low charge density to provide efficient gene delivery with minimal toxicity (116, 117). For example, poly(amine-co-ester) terpolymers, which have significantly lower charge density than traditional polycationic polymers, are able to deliver genes in vitro and in vivo with efficiency superior to existing commercially available gene delivery systems and with limited toxicity (118).
Another approach for developing safe and efficient polymer-based vehicles is to start with polymers that are known to be safe for clinical use and to modify them to enhance their ability to deliver genes or oligonucleotides. Poly(lactide-co-glycolide) (PLGA) was approved by FDA in 1969 and has been in continuous clinical use since that time. Octa-functional nanoparticles were produced by modification of PLGA nanoparticles through the use of chemical conjugation and physical formulation techniques. These nanoparticles exhibit limited toxicity but are able to efficiently deliver DNA or siRNA to tumor cells in vitro and in vivo (119), making them suitable for brain tumor gene therapy.
Accurate measurement of the distribution of a delivered agent is needed to assess drug or gene delivery in GBM. Magnetic resonance imaging (MRI) allows for monitoring of delivery in certain cases. Recent studies have used diffusion-weighted imaging (DWI) as a means to monitor the administration of fluid into the brain; it is assumed that the location of changes in DWI correlates to the area occupied by infused drug. These studies find an area of decreased diffusion immediately after infusion that evolves into an area of increased fluid movement (120, 121). In the setting of CED, where many factors can influence the distribution of the infusion, this finding is particularly useful: optimizing fluid distribution is needed for effective treatment (121). Significantly, DWI is frequently used in the clinical setting and does not require the administration of contrast agents. But DWI must be used with caution: it relies upon water movement to produce signal changes, effectively only measuring the movement of the solvent, and measuring it indirectly. There is also debate as to the mechanisms producing the observed intensity changes, and whether they represent the infusate or reactive fluid movement (122).
Drug distribution can be quantified by MRI using gadolinium-based contrast agents. Contrast provided by gadolinium agents, such as gadolinium-diethylene triamine pentaacetic acid (Gd-DTPA), can be used to monitor disruption of the blood-brain barrier (123). When co-administered with a therapeutic agent or other macromolecule (Figure 3), Gd-DTPA can be used to assess the volume of drug delivery (123, 124). Gadolinium complexes can be conjugated to or encapsulated in nanocarriers, such as polymer nanoparticles (125, 126) or liposomes (29, 34) in order to assess the dispersion of drug-loaded particles. This approach is appealing, but it has limitations. Gadolinium contrast agents are smaller than many delivery vehicles. Thus, when used as surrogates, they can diffuse over a larger volume than the actual agent (although, in practice, this effect may be small) (123, 124, 127). Some groups have addressed this issue by conjugating therapies to materials that can be imaged directly, such as iron-oxide nanoparticles (128). But this approach can also present problems, because the addition of an imaging moiety can alter the agent in unforeseen ways, such that distribution behavior does not precisely mirror the natural, unmodified agent. Contrast agent-derived signals are more easily related to drug movement than diffuse signaling changes, such as those seen on DWI, but they have limitations.
Nuclear medicine imaging can also be used to monitor drug distribution. In single-photon emission computed tomography (SPECT), a gamma-emitting tracer allows for three-dimensional visualization of the distribution area of an agent (123, 129). The radioisotope can be either co-administered with the drug of interest or directly bound to the biologically active molecule (such as siRNA) to allow for direct assessment of its volume of distribution (130). SPECT imaging can be coupled with CT or MRI for anatomic correlation (130). This approach is direct, and uses equipment that is significantly less expensive than those used in other nuclear medicine imaging techniques (131). The main drawback of conventional SPECT is resolution, although recent advances in hardware (such as pinhole-SPECT) allow for resolution at the millimeter level (132, 133).
Positron-emission-tomography (PET) is another promising modality for imaging drug delivery to tumors. As with gadolinium and SPECT contrast agents, PET tracers can be infused concurrently with drug or bound to the delivery system, such as nanoparticles (134–136). When PET is coupled with CT, molecular movement can be correlated with anatomy, allowing for measurement of the anatomical area of diffusion of tracer or tracer-laden substance (134, 135). Further, through the use of tracers that are derivatives of amino acids—such as O-(2-[18F]fluoroethyl)-L-tyrosine (FET)—PET imaging can estimate the borders of a tumor, allowing for more accurate assessment of drug distribution relative to tumor volume than MRI (134, 137, 138). Challenges for PET imaging include radiation exposure (which is also a challenge in SPECT), the high cost of the studies, and the short-lived nature of typical PET tracers (139). Another limitation is similar to the one mentioned for gadolinium agents in MRI: unless the tracer is directly coupled to the delivery agent, then the measurement of the area of diffusion is indirect. Direct radiolabeling of nanoparticles shows that it is possible to overcome this limitation (140).
New drug delivery strategies are already impacting treatment of GBM, and it seems clear that delivery systems will be needed for therapies of the future, which will be targeted to particular cells and depend on intracellular delivery of agents that do not readily cross the BBB or enter cells. Polymer implants, CED techniques, and degradable nanoparticles are powerful platform technologies for creation of new methods to treat GBM.
Our work on delivery systems for treating GBM is funded by the National Institutes of Health (Grant CA149128). We thank Nha Duong for preparation of Figure 1 and excellent editorial assistance.
Sources of support that require acknowledgment: National Institutes of Health (Grant CA149128 to WMS)
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
The authors have no conflicts of interest.