*Review* **Viral Vectors as Gene Therapy Agents for Treatment of Glioblastoma**

**Oleg Mozhei 1,\* , Anja G. Teschemacher <sup>2</sup> and Sergey Kasparov 1,2,\***


Received: 21 October 2020; Accepted: 7 December 2020; Published: 11 December 2020

**Simple Summary:** Glioblastoma is the most malignant cancer of the brain and current therapeutic strategies are clearly inadequate. In addition to surgical intervention, conventional drugs and ratio-therapy, scientists are looking at approaches based on gene therapy with genetically modified viruses. In this review we give a snapshot of the current state of play in this field of research and the available information about the clinical trials. We make some suggestions as to what opportunities could be explored further and hope that this review will stimulate discussion and conception of new life saving strategies.

**Abstract:** In this review, we scrutinize the idea of using viral vectors either as cytotoxic agents or gene delivery tools for treatment of glioblastoma multiforme (GBM) in light of the experience that our laboratory has accumulated over ~20 years when using similar vectors in experimental neuroscience. We review molecular strategies and current clinical trials and argue that approaches which are based on targeting a specific biochemical pathway or a characteristic mutation are inherently prone to failure because of the high genomic instability and clonal selection characteristics of GBM. For the same reasons, attempts to develop a viral system which selectively transduces only GBM cells are also unlikely to be universally successful. One of the common gene therapy approaches is to use cytotoxic viruses which replicate and cause preferential lysis of the GBM cells. This strategy, in addition to its reliance on the specific biochemical makeup of the GBM cells, bears a risk of necrotic cell death accompanied by release of large quantities of pro-inflammatory molecules. On the other hand, engaging the immune system in the anti-GBM response seems to be a potential avenue to explore further. We suggest that a plausible strategy is to focus on viral vectors which efficiently transduce brain cells via a non-selective, ubiquitous mechanism and which target (ideally irreversibly) processes that are critical only for dividing tumor cells and are dispensable for quiescent brain cells.

**Keywords:** gene therapy; glioblastoma; glioma; viral vectors

## **1. Introduction**

Glioblastoma multiforme (GBM) is a highly malignant primary brain cancer of predominantly astrocytic origin [1]. The main features of GBM that lead to malignancy and high mortality are its high resistance to DNA-damaging drugs, including the only Food and Drug Administration FDA-approved alkylating agent temozolomide (TMZ), which is achieved by O6-methylguanine-DNA methyltransferase overexpression, moderate response to radiation, genomic instability and powerful clonal selection. A particularly grave feature of GBM is its high invasiveness.

New insights into the genomic landscape of GBM revealed typical mutations in an array of genes, including *TERT*, *PTEN*, *IDH1*, *IDH2*, *TP53*, *ATRX*, *PIK3CA*, *PIK3R1*, *NF1*, *H3F3A*, *CDKN2A*, *EGFR*,

*PDGFRA*, *MET*, *CDK4*, *CDK6*, *MDM2*, *MDM4* [2]. Traditionally, based largely on neuroanatomical considerations, gliomas were subdivided into four grades. Glioblastoma is the most malignant (grade IV) glioma [3].

The introduction by the World Health Organisation (WHO), in 2016, of the "integrated" classification based on histology and genetics was developed in the hope of improving diagnostic accuracy, patient management and prognosis of the response to treatments [4]. However, as of today, most of the treatment algorithms are not based on molecular histological characteristics and are essentially universal, consisting of maximal surgical resection, followed by radiotherapy and chemotherapy with TMZ, followed by TMZ, known as the "Stupp protocol" [5,6].

Unfortunately, even this aggressive treatment has low efficiency, with survival rates remaining between 12 and 15 months and the 3-year survival rate only at about 15%. Despite introduction of newer treatments, such as Carmustine wafers, the monoclonal antibody bevacizumab and cyclin-dependent kinases (CDK) inhibitors, GBM is still an essentially incurable disease, resulting in a patient death rate of more than 95% within five years of diagnosis.

Even though classic metastases are exceedingly rare in GBM, its cells have a tendency to migrate into the parenchyma and eventually spread extensively throughout the brain. For this reason, already upon primary diagnosis, some patients have infiltration in more than one part of the brain, with tumor cells moving across the corpus callosum or through the walls of the ventricles. In cases such as those, surgery may be performed only for the sake of decompression but has little effect on the overall progression of the disease. The only feasible option to pursue, then, is systemic pharmacotherapy and radiotherapy. However, GBM presents formidable challenges for traditional drug design. Movement of drugs across the blood–brain barrier (BBB) is a significant problem because it depends on too many factors (charge, molecular weight and conformation, hydrophobicity, presence of specific transporters, vascularization of the tumor, etc.). Moreover, the relationship between these factors and drug transfer across the BBB is non-linear. It is estimated that less than 2% of small-molecule drugs and no large-molecule drugs or nucleic acid-based constructs can reach the brain because of the BBB [7]. Insufficient saturation of brain tissue with anti-cancer drugs allows GBM cells to benefit from the selection of the most aggressive and drug-resistant subclones. In addition, tumors engage various efflux transport systems (for instance, ATP-binding cassette sub-family B member 1 (*ABCB1*) gene, which extrude drugs from cancer cells) [8]. The other well-known mechanism of tumor defense is expression of high levels of the DNA repair enzyme O6-methylguanine-DNA methyltransferase, mentioned above [9].

However, upon initial diagnosis, GBM tumors frequently appear relatively well-localized and surgically accessible. Nevertheless, due to the infiltration, tumors almost inevitably reoccur after resection, typically originating from sites adjacent to the surgical cavity. Surgeons are limited in their actions because GBM often grows near critical regions of the brain (major nerve tracts, essential centers and large blood vessels). Damage to those areas is too risky and may cause severe disabilities or even be lethal. In cases of well-localized and relatively superficial primary GBM, the key task is, therefore, the prevention of infiltration around the surgical cavity. Here is the scope for locally delivered therapies, such as slow-release formulations of anti-cancer drugs [10], photodynamic therapy [11] or viral vectors, which are the topic of this review.

#### **2. Molecular Strategies for Viral Gene Therapy of the GBM**

For patients with well-localized primary GBM, one could envisage a strategy where after the de-bulking surgery, the adjacent parenchyma is infiltrated by viral gene therapy vectors which selectively destroy the GBM cells. In a more dramatic scenario, a viral gene therapy tool could be injected systemically, selectively affecting tumor cells in the whole of the CNS and eliminate them. Attempts to develop gene therapy with the aid of viral vectors have been under development for some time, and below, we summarize some of the main strategies and their outcomes.

1. Oncolytic viruses which destroy tumor cells were amongst the first vectors which were tested in patients. The rationale for this approach was based on pre-clinical data demonstrating that some strains of various viruses replicate well only in tumor cell lines. It was then suggested that it is possible to selectively destroy cancer cells in situ, with minimal impact on normal cells. In clinical studies, either wild-type or genetically engineered viruses were used; the specificity of the latter was enhanced by targeted changes in their genomes. It needs to be stressed that oncolytic viruses are able to destroy any cells which they invade and, unless tightly controlled by an additional mechanism, might cause excessive tumor necrosis and dangerous brain oedema [12,13]. While several viral progenitors have been used (see Section 2.1 below), the first oncolytic viruses were wild-type viruses, followed by second generations of genetically modified viruses and third-generation vectors equipped with transgenes to further induce therapeutic effects [12].

2. Suicide gene therapy is based on heterologous expression of *Escherichia coli* or yeast cytosine deaminase or *Herpes simplex* virus thymidine kinase in the cancer cells [14]. Cytosine deaminase converts the prodrug 5-Fluorocytosine (5-FC) to a toxic 5-Fluorouracil (5-FU) metabolite, whereas thymidine kinase (HSV-tk) converts ganciclovir to ganciclovir monophosphate, which, in turn, is converted to toxic ganciclovir triphosphate by tumor cells' enzymes. This leads to damage and lysis of transgene-expressing cells and those surrounding them (so-called bystander effect).

3. Immunomodulatory vectors aim to engage a strong immune response against the GBM cells. This can be achieved by expression of strong antigens on tumor cells' surface or by the production of factors which stimulate and attract the immune cells.

4. Introduction of anti-oncogenes and tumor suppressors in cancer cells aims to decrease proliferation, stimulate differentiation or induce apoptosis by a dominant gain-of-function effect.

To achieve maximum efficiency, some approaches can be combined. For example, an oncolytic effect may accompany release of immunomodulatory proteins expressed by genes delivered with a viral vector.

## *2.1. Viral Vector Types Proposed for Gene Therapy of GBM*

The effectiveness of gene therapy tools is a function of virus biology, mechanism of action, specificity and replication competency. If the viral genome is partially deleted to prevent replication, this clears room for the delivery of the therapeutic genes. If, however, the virus is allowed to replicate, it will cause cytopathic effects, lysis and new virions will proceed to infect other cells. There are currently over 20 viral vectors that have been used in clinical trials for gene therapy of GBM, as summarized in Table 1. Figure 1 describes the selection criteria.


**Table 1.** Comparison of key features of viral vectors proposed for treatment of GBM.


#### **Table 1.** *Cont*.

In relation to the ability to replicate, + denotes replication competent vectors, − stands for replication incompetent ones and ± for conditionally replication competent vectors. CAR-chimeric antigen receptor; CMV–cytomegalovirus; RSV-rous sarcoma virus; PKR-protein kinase R; HUJ-Hebrew University, Jerusalem; CD-cytosine deaminase.

*Cancers* **2020**, *12*, x 4 of 25

**Figure 1.** Selection and inclusion criteria for review of glioblastoma multiforme (GBM)-targeting viral vector trials. **Figure 1.** Selection and inclusion criteria for review of glioblastoma multiforme (GBM)-targeting viral vector trials.

#### **Table 1.** Comparison of key features of viral vectors proposed for treatment of GBM. *2.2. Adenovirus-Based Vectors*

**Name Structure of Vector Mechanism of Action Specificity Replication Competent**  DNX2401 Ad5 Lytic viral cycle in targeted cells Replicate in cells defective in the Rb/p16 tumor suppressor pathway and expressing integrins αvβ3 and αvβ5 ± DNX2440 Ad5 Lytic viral cycle in targeted cells and immunomodulatory effect Replicate in cells defective in the Rb/p16 tumor suppressor pathway and expressing integrins αvβ3 and αvβ5 ± Lytic viral cycle in Replicate in tumor Adenovirus (Ad) is a double-stranded DNA virus (Baltimore Classification class I [15]) without an envelope [16]. There are at least 57 serotypes of human Ad, Ad1–Ad57, in seven species, A–G [17]. The human Ad genome contains five early transcription units (E1A, E1B, E2, E3 and E4), four intermediate and one late transcription unit [17]. Main modification of Ad genome are shown in Figure 2. Viral entry is coxsackie-adenovirus receptor (CAR)-dependent. One of the crucial steps in the adenoviral replication cycle is interaction of the *E1A* gene product with E2F-Rb or E2F-DP1 transcription complexes to force the infected cell into the S phase since it is helps the virus to use the cellular DNA replication machinery to replicate its own genome [18]. These processes can be altered to achieve increased selectivity towards GBM and will be discussed later. Most Ad vectors originate from Ad5 (Species C). Non-replicating Ads are widely used as experimental gene delivery tools, while replicating Ads have been engineered to be tumor-specific agents. The conventional strategy to achieve replication deficiency is to delete *E1* and *E3* genes. The genomes of such vectors, after entering the target cell nucleus, remain as additional DNA elements not integrated into the chromosomes (i.e., episomal). This has major implications for their fate in the cancer, as well as in any other dividing cells, because after a few divisions, episomes which do not replicate are diluted and expression drops rapidly.

ONYX-015 chimeric Ad2 and Ad5 targeted cells cells with altered p53 pathway ± Converts harmless Transduce CARexpressing cells. The strategies for targeting Ad vectors to GBM include (1) use of tumor-specific promoters; (2) deletion of critical viral genes which are supplied by tumor cells in trans; (3) modification of the viral capsid to enable selective entry into GBM cells.

Ad-hCMV-TK Ad5 ganciclovir to toxic product in transduced cells CMV-dependent expression mechanism − ONYX-015 was the first oncolytic Ad vector to be described [19]. This is a recombinant selectively replication-competent chimeric Ad2 and Ad5 vector [17]. ONYX-015 lacks the *E1B* gene. The normal function of the protein encoded by *E1B* is to inactivate p53 protein in infected cells. Thus, ONYX-015

[23–25].

[33].

was expected to replicate only in p53-deficient cells [20], but later, it was found that ONYX replication is not, in fact, p53-dependent [21,22]. *Cancers* **2020**, *12*, x 7 of 25

**Figure 2.** Schematic of the genome structures of adenovirus type 5 (Ad5) and Ad5-based vectors. (**a**) Wild-type Ad5 virus. Arrows indicate transcriptional units. ITR—Inverted terminal repeat. (**b**) In the ONYX-015 adenoviral vector, the *E1B* gene is deleted. (**c**) DNX-2401 adenoviral vector structure. Δ24 bp indicates 24 base pairs' deletion in the Rb-binding domain of the E1A gene; RGD ins indicates an insertion of an additional peptide sequence in the Ad fiber-encoding part of the genome. (**d**) Adenoviral vectors, often referred as AVVs in the literature, are replication-incompetent viral particles produced by deleting *E1* and *E3* genes and inserting a desired transgene. Such vectors are widely used in experimental neuroscience for gene delivery by various groups, including ourselves **Figure 2.** Schematic of the genome structures of adenovirus type 5 (Ad5) and Ad5-based vectors. (**a**) Wild-type Ad5 virus. Arrows indicate transcriptional units. ITR—Inverted terminal repeat. (**b**) In the ONYX-015 adenoviral vector, the *E1B* gene is deleted. (**c**) DNX-2401 adenoviral vector structure. ∆24 bp indicates 24 base pairs' deletion in the Rb-binding domain of the E1A gene; RGD ins indicates an insertion of an additional peptide sequence in the Ad fiber-encoding part of the genome. (**d**) Adenoviral vectors, often referred as AVVs in the literature, are replication-incompetent viral particles produced by deleting *E1* and *E3* genes and inserting a desired transgene. Such vectors are widely used in experimental neuroscience for gene delivery by various groups, including ourselves [23–25].

DNX2401 (Delta-24-RGD**)** is a recombinant serotype 5 strain Ad [26]. This oncolytic vector has two modifications in its genome that make it selectively replication-competent in cells defective in the Rb/p16 tumor suppressor pathway. The first modification is the 24-bp deletion (bp 923–946) in the Rb-binding domain of the *E1A* gene [26]. Under normal circumstances, viral E1A proteins promote cells towards a mitotic state by releasing E2F transcriptional factors from the block by Rb proteins. The unstable version of the *E1A* gene in DNX2401 cannot bind to E2F-Rb or E2F-DP1 transcription complexes and release E1A. This prevents replication in cells with a normal Rb/p16 tumor suppressor pathway. GBM often have defective Rb/p16 tumor suppressor pathways, which makes it possible for viruses to replicate selectively in GBM cells because cells are free from the Rb/p16 block anyway. Most cancer cells lack, or poorly express, CAR receptors required for adenovirus binding and internalization. To circumvent this problem, the second modification, an additional RGD peptide sequence in the HI loop of the Ad fiber, allows the virus to bind to cells DNX2401 (Delta-24-RGD) is a recombinant serotype 5 strain Ad [26]. This oncolytic vector has two modifications in its genome that make it selectively replication-competent in cells defective in the Rb/p16 tumor suppressor pathway. The first modification is the 24-bp deletion (bp 923–946) in the Rb-binding domain of the *E1A* gene [26]. Under normal circumstances, viral E1A proteins promote cells towards a mitotic state by releasing E2F transcriptional factors from the block by Rb proteins. The unstable version of the *E1A* gene in DNX2401 cannot bind to E2F-Rb or E2F-DP1 transcription complexes and release E1A. This prevents replication in cells with a normal Rb/p16 tumor suppressor pathway. GBM often have defective Rb/p16 tumor suppressor pathways, which makes it possible for viruses to replicate selectively in GBM cells because cells are free from the Rb/p16 block anyway. Most cancer cells lack, or poorly express, CAR receptors required for adenovirus binding and internalization. To circumvent this problem, the second modification, an additional RGD peptide sequence in the HI loop of the Ad fiber, allows the virus to bind to cells expressing integrins αvβ3 and αvβ5 which are found on the surface of most cancer cells, including glioma and GBM [26,27].

expressing integrins αvβ3 and αvβ5 which are found on the surface of most cancer cells, including glioma and GBM [26,27]. DNX-2440 (Delta-24-RGDOX) is an immunomodulatory recombinant selectively replicationcompetent serotype 5 strain Ad-encoding OX40 ligand (OX40L) driven by the cytomegalovirus DNX-2440 (Delta-24-RGDOX) is an immunomodulatory recombinant selectively replicationcompetent serotype 5 strain Ad-encoding OX40 ligand (OX40L) driven by the cytomegalovirus (CMV) promoter. The protein is able to activate T cells via interaction with its receptor on the surface of T lymphocytes [28,29].

(CMV) promoter. The protein is able to activate T cells via interaction with its receptor on the surface of T lymphocytes [28,29]. AVV-CMV-HSV-tk (Ad-hCMV-TK) uses the suicide gene strategy and is a recombinant replication-defective serotype 5 Ad with *Herpes simplex virus thymidine kinase* (*HSV-tk*) gene under the transcriptional control of the CMV promoter [30]. CMV is often referred to as ubiquitously and constitutively active. However, experimental neuroscience demonstrated that this is, in fact, not the case, since CMV-bearing viral vectors effectively drive expression only in some cell types in the AVV-CMV-HSV-tk (Ad-hCMV-TK) uses the suicide gene strategy and is a recombinant replicationdefective serotype 5 Ad with *Herpes simplex virus thymidine kinase*(*HSV-tk*) gene under the transcriptional control of the CMV promoter [30]. CMV is often referred to as ubiquitously and constitutively active. However, experimental neuroscience demonstrated that this is, in fact, not the case, since CMV-bearing viral vectors effectively drive expression only in some cell types in the normal rodent brain and expression may be transient [31]. It follows that the brain cells have mechanisms to silence CMV and this may very well apply to the clones within GBM.

normal rodent brain and expression may be transient [31]. It follows that the brain cells have mechanisms to silence CMV and this may very well apply to the clones within GBM. AVV-RSV-HSV-tk (ADV/HSV-tk) is a similar suicide gene virus but expresses *HSV-tk* under control of Rous sarcoma virus long-terminal-repeat promoter (RSV) [32]. The RSV promoter is considered a strong constitutive promoter, similar to CMV. RSV, in comparison with CMV, exhibits AVV-RSV-HSV-tk (ADV/HSV-tk) is a similar suicide gene virus but expresses *HSV-tk* under control of Rous sarcoma virus long-terminal-repeat promoter (RSV) [32]. The RSV promoter is considered a strong constitutive promoter, similar to CMV. RSV, in comparison with CMV, exhibits a lag phase prior to the onset of viral DNA replication and has a somewhat different profile of tissue-specific expression, although it is not entirely clear whether this confers an advantage in this case [33].

Ad-hCMV-Flt3L is a recombinant replication-deficient serotype 5 Ad for CMV promoter-driven expression of human fms-like tyrosine kinase 3 ligand (Flt3L). Flt3L is a hematopoietic growth factor

a lag phase prior to the onset of viral DNA replication and has a somewhat different profile of tissue-

Ad-hCMV-Flt3L is a recombinant replication-deficient serotype 5 Ad for CMV promoter-driven expression of human fms-like tyrosine kinase 3 ligand (Flt3L). Flt3L is a hematopoietic growth factor and ligand for the Flt3 tyrosine kinase receptor, which is expressed on the surface of dendritic cells (DCs). The transgene provides an immunomodulatory effect by stimulating both the proliferation of dendritic cells and their migration to the tumor site. The vector is usually used with other conventional drugs for eliciting a stronger response to GBM via release of Flt3L from destroyed cells [34]. *Cancers* **2020**, *12*, x 8 of 25 and ligand for the Flt3 tyrosine kinase receptor, which is expressed on the surface of dendritic cells (DCs). The transgene provides an immunomodulatory effect by stimulating both the proliferation of dendritic cells and their migration to the tumor site. The vector is usually used with other conventional drugs for eliciting a stronger response to GBM via release of Flt3L from destroyed cells [34].

Ad-RTS-hIL12 also aims at immunomodulation. It is a recombinant replication-deficient serotype 5 Ad-encoding human pro-inflammatory interleukin-12 (IL-12: hIL indicates human origin of the gene) gene under control of RheoSwitch Therapeutic System (RTS) promoter. RTS is an artificial veledimex-inducible promoter that leads to uniform and long-term release of interleukin-12 in the tumor area after a single vector injection. This system is based on recruiting transcription factor to a synthetic promoter via Gal4–Gal4-binding site interactions [35]. The cassette consists of Gal4-EcR fusion protein sequence, internal ribosome entry site (IRES) linker and *VP16-RXR fusion protein* gene and is driven by *human ubiquitin C* gene promoter (Figure 3). Upstream, there is a customizable promoter with Gal4 binding sites to which these fusion proteins are recruited and the target gene is transcribed [35]. IL-12 activates the immune system, which may result in immune-mediated tumor cell lysis and inhibition of cancer cell proliferation [36]. Ad-RTS-hIL12 also aims at immunomodulation. It is a recombinant replication-deficient serotype 5 Ad-encoding human pro-inflammatory interleukin-12 (IL-12: hIL indicates human origin of the gene) gene under control of RheoSwitch Therapeutic System (RTS) promoter. RTS is an artificial veledimex-inducible promoter that leads to uniform and long-term release of interleukin-12 in the tumor area after a single vector injection. This system is based on recruiting transcription factor to a synthetic promoter via Gal4–Gal4-binding site interactions [35]. The cassette consists of Gal4-EcR fusion protein sequence, internal ribosome entry site (IRES) linker and *VP16-RXR fusion protein* gene and is driven by *human ubiquitin C* gene promoter (Figure 3). Upstream, there is a customizable promoter with Gal4 binding sites to which these fusion proteins are recruited and the target gene is transcribed [35]. IL-12 activates the immune system, which may result in immune-mediated tumor cell lysis and inhibition of cancer cell proliferation [36].

**Figure 3.** Schematic representation of *RTS* gene switch cassettes. Upon administration of veledimex, RXR-VP16 and GAL4-EcR proteins dimerize and activate transgene expression. The GAL4 domain recognizes unique specific binding sites (GAL4-BS) while VP16 acts as a powerful activation of transcription in mammalian cells. The protein 3D structure was adopted from Yoon et al. [37]. **Figure 3.** Schematic representation of *RTS* gene switch cassettes. Upon administration of veledimex, RXR-VP16 and GAL4-EcR proteins dimerize and activate transgene expression. The GAL4 domain recognizes unique specific binding sites (GAL4-BS) while VP16 acts as a powerful activation of transcription in mammalian cells. The protein 3D structure was adopted from Yoon et al. [37].

Ad.hIFN-β is another immunomodulating replication-defective serotype 5 Ad-encoding human *Interferon-β* (*IFN-β*) gene under control of CMV promoter [38]. Interferon-β (IFN-β) is a pleiotropic cytokine with anti-tumor activity which demonstrated promising outcomes in some clinical trials [39]. However, overall efficacy was limited and transient mainly because of high-dose toxicity (myelosuppression, transaminitis, neurotoxicity, including seizures, etc.) [38]. To overcome this limitation, Ad.hIFN-β was developed to drive synthesis of Interferon-β in cancer cells. A schematic representation of the genome is shown in Figure 2. Ad.hIFN-β is another immunomodulating replication-defective serotype 5 Ad-encoding human *Interferon-*β (*IFN-*β) gene under control of CMV promoter [38]. Interferon-β (IFN-β) is a pleiotropic cytokine with anti-tumor activity which demonstrated promising outcomes in some clinical trials [39]. However, overall efficacy was limited and transient mainly because of high-dose toxicity (myelosuppression, transaminitis, neurotoxicity, including seizures, etc.) [38]. To overcome this limitation, Ad.hIFN-β was developed to drive synthesis of Interferon-β in cancer cells. A schematic representation of the genome is shown in Figure 2.

VB-111 is recombinant replication-defective serotype 5 Ad-encoding Fas-TNFR-1 gene under control of pre-proendothelin-1 promoter. The promoter was chosen with the aim of achieving VB-111 is recombinant replication-defective serotype 5 Ad-encoding Fas-TNFR-1 gene under control of pre-proendothelin-1 promoter. The promoter was chosen with the aim of achieving selectivity to endothelial cells undergoing angiogenesis. Cell apoptosis is induced when circulating TNF-α interacts with the Fas-TNFR-1 receptor [40]. The expected outcome is the prevention of vascularization and, therefore, metabolic insult to the tumor. *Cancers* **2020**, *12*, x 9 of 25 TNF-α interacts with the Fas-TNFR-1 receptor [40]. The expected outcome is the prevention of vascularization and, therefore, metabolic insult to the tumor.

As mentioned above, replication-incompetent Ad vectors stay episomal in the transduced cells and are not propagated when the cell divides. This leads to a rapid dilution of the viral genomes in any dividing cells, such as GBM. In this respect, replication-competent viruses, such as ONYX-015, are different because they replicate in the affected cells. The downside of this strategy is the lack of control over the spread of the virus and infection of the healthy cells, which then, inevitably, become targets for 5FC. In addition, release of the activated, toxic products of pro-drugs non-selectively kills adjacent cells (the "bystander effect"). As mentioned above, replication-incompetent Ad vectors stay episomal in the transduced cells and are not propagated when the cell divides. This leads to a rapid dilution of the viral genomes in any dividing cells, such as GBM. In this respect, replication-competent viruses, such as ONYX-015, are different because they replicate in the affected cells. The downside of this strategy is the lack of control over the spread of the virus and infection of the healthy cells, which then, inevitably, become targets for 5FC. In addition, release of the activated, toxic products of pro-drugs non-selectively kills adjacent cells (the "bystander effect").

#### *2.3. Herpes Simplex Virus-Based Vectors 2.3. Herpes Simplex Virus-Based Vectors*

*Herpes simplex* virus (HSV) is an enveloped double-stranded DNA virus (Class I according to the Baltimore classification [15]). HSV can target both dividing and non-dividing cells and has broad tropism but predominantly infects neurons. Herpes viruses are classified into subfamilies, and for gene therapy applications, HSV-1 is used. The genome of HSV-1 is ~150 kbp long and can, therefore, potentially carry a substantial payload (Figure 4). During the viral life cycle, HSV-1 remains episomal as a circular DNA molecule [41]. *Herpes simplex* virus (HSV) is an enveloped double-stranded DNA virus (Class I according to the Baltimore classification [15]). HSV can target both dividing and non-dividing cells and has broad tropism but predominantly infects neurons. Herpes viruses are classified into subfamilies, and for gene therapy applications, HSV-1 is used. The genome of HSV-1 is ~150 kbp long and can, therefore, potentially carry a substantial payload (Figure 4). During the viral life cycle, HSV-1 remains episomal as a circular DNA molecule [41].

**Figure 4.** Schematic of the *Herpes simplex* virus 1 (HSV-1)-based vectors. The genome of wild-type HSV-1 can be divided into six regions which contain specific genes. Information about specific vectors is provided in the text. **Figure 4.** Schematic of the *Herpes simplex* virus 1 (HSV-1)-based vectors. The genome of wild-type HSV-1 can be divided into six regions which contain specific genes. Information about specific vectors is provided in the text.

The RL1 gene (also known as γ34.5), one of the essential genes for replication, can be used to

The RL1 gene (also known as γ34.5), one of the essential genes for replication, can be used to modulate specificity. During viral replication, the host cellular defense system typically responds with translational arrest and reduction in the global synthesis of viral and cellular proteins [42]. This process is facilitated by phosphorylation of the translation initiation factor eIF2α by protein kinase R (PKR). *RL1* gene encodes The Infected Cell Protein 34.5 (ICP34.5), also known as Neurovirulence factor ICP34.5. This multifunctional protein binds and retargets the host phosphatase PP1α to eIF2α, thus reversing the phosphorylation and the shutdown of the protein synthesis [43]. Mutated ICP34.5 is unable to counteract PKR action, which, theoretically, should protect healthy cells. Since in tumors, the PKR pathway is often inhibited, lack of ICP34.5 function does not limit viral replication and should result in selective replication of this mutated HSV-1 in such cancer cells.

The other important HSV-1 gene is UL39, which encodes the large subunit of ribonucleotide reductase, also known as ICP6. The ribonucleotide reductase complex converts ribonucleotides to deoxyribonucleotides needed for viral DNA replication. The host ribonucleotide reductase enzyme is highly active only in mitotic cells. Thus, UL39-defective HSV-1 UL39 cannot replicate efficiently in non-dividing cells [44]. Specific examples are given below.

HSV 1716 is an oncolytic recombinant replication-competent HSV-1. Deletions in both copies of the *RL1* gene (see above) were made with the aim to permit replication only in PKR-defective tumor cells [45].

C134 is an oncolytic HSV-1. In this virus, *RL1* genes are deleted and human cytomegalovirus (HCMV) *IRS1* gene was inserted between *UL3* and *UL4* genes [46]. The *IRS1* gene enhances replication in fast dividing tumor cells [46]. The exact molecular mechanism of action of IRS1 protein is still not known.

G207 is an oncolytic recombinant replication-competent HSV-1 which has two modifications to increase specificity towards GBM cells: deletions in both copies of the *RL1* gene to target PKR-defective cancer cells and disruption of *UL39* gene to eliminate the possibility to replicate in non-dividing normal cells. During the lytic phase, the vector causes direct cytopathic effect and indirect T cell-mediated cell death [47].

rQNestin34.5v.2 is a recombinant HSV-1 also devoid of *UL39* and all *RL1* genes. Lack of *RL1* gene should limit replication in normal cells via the mechanism explained above. Instead, this vector carries one copy of *RL1* gene under transcriptional control of the nestin promoter, which is frequently upregulated in gliomas [48]. Thus, nestin promoter is expected to drive expression of functional ICP34.5 selectively in glioma cells, resulting in a cytopathic effect. It is worth noting that the selectivity of this promoter is not widely known and that nestin is also expressed in normal brain cells [49].

M032-HSV-1 is a combined (oncolytic and immunomodulatory) replication-competent HSV-1. The virus has deletions of both copies of the *R1* (γ*34.5*) gene and inserted *interleukin-12* (*IL-12*) gene [50]. Deletions limit replication to PKR-defective tumor cells. In addition, interleukin-12 promotes an immune response against surviving tumor cells and decreases angiogenesis.

#### *2.4. Vectors Based on other Viral Backgrounds*

Pelareorep (Reolysin) is a human wild-type reovirus [51,52]. Reovirus is a non-enveloped double-stranded RNA virus (Class III according to the Baltimore Classification [15]). It causes mild infections in humans—for instance, gastroenteritis. Reoviruses can be used as oncolytic agents because they replicate predominantly in cells where the Ras pathway is highly active, as is typical for many cancers [53]. Specific examples are provided in Tables 1 and 2.

Newcastle disease virus (NDV) is a single-stranded enveloped RNA virus whose natural host is poultry. It has been shown that the virus can induce apoptosis in melanoma cultures overexpressing a protein called Livin, encoded by the BIRC7 gene. This protein belongs to a family of anti-apoptotic proteins which are commonly overexpressed by tumors and it has been demonstrated that melanoma tumor cells that do not express Livin are relatively resistant to the virus [54]. Attempts have been made to use it against GBM [54]. NDV-HUJ is a wild-type oncolytic HUJ strain of Newcastle disease virus.

ParvOryx, or H-1PV, is an oncolytic wild-type parvovirus, a small single-stranded DNA virus (Class II according to the Baltimore classification [15]) without an envelope. In nature, this is a rodent virus, but H-1PV is able to infect cells of other species, including humans. Replication of H-1PV greatly depends on the activity of the host enzymes expressed during the S-phase, making it selectively replication-competent in fast dividing cancer cells [55].

PVSRIPO is a poliovirus type 1 (Sabin type) viral vector with its cognate internal ribosome entry site (IRES) replaced with that of human rhinovirus type 2. The vector binds to CD155 (poliovirus receptor, PVR or NECL5), internalizes and eventually causes tumor cell lysis [56]. The exchange of the IRES should, in theory, restrict replication in cells of neuronal origin [56].

Toca 511 is a replicating gamma-retrovirus which carries a *yeast cytosine deaminase (CD*) gene. Administration of 5-FC leads to generation of toxic 5-FU by CD [57]. As a result, tumor cells infected by this virus should die and release 5-FU, which can cause the bystander effect [58]. The vector has specificity for replicating cells, and replication in non-malignant cells in vivo is reportedly insignificant [59].

TG6002—recombinant vaccinia viral vector, also encoding the suicide gene *CD* [60]. Vaccinia virus is a 190-kbp dsDNA-enveloped virus which causes small pox [61]. To increase safety and specificity to fast dividing cells, the *J2R* gene (encoding thymidine kinase) and the *I4L* gene (encoding the large subunit of the ribonucleotide reductase) were deleted [61].

MV-CEA is a recombinant Edmonston strain of measles virus, expressing a soluble extracellular *N*-terminal domain of human carcinoembryonic antigen (CEA) [62]. Internalization is mediated through CD46 binding, leading to formation of syncytium and cell lysis [62]. The expressed CEA is expected to stimulate the immune system to recognize and destroy targeted cells.

#### *2.5. Evaluating Vector E*ffi*cacy*

The main goal of patient treatment is to increase life expectancy and improve the quality of life. Unfortunately, GBMs are a very heterogeneous group of diseases. Even morphologically-similar tumors can have different driver mutations and responses to treatment, which makes it impossible to directly compare the results of clinical trials. It should be noted that regional features of healthcare systems and even personal experiences of the attending physician can introduce bias. Moreover, previous treatment changes the tumor makeup due to clonal selection, which must be taken into account.

For the purpose of this review, we have stratified studies into three types.

1. Dose-escalating studies to assess the maximum acceptable dosage of the gene therapy vector. In accordance with the possible side effects of the administration of viral vectors, these studies are not carried out on healthy volunteers.

2. Comparison of the new therapy with existing ones when used in patients with recurrent or progressive GBM. In such patients, the prior therapy has led to the emergence of resistance and more aggressive clones, thereby diminishing the potential benefit of TMZ and justifying the application of a new therapeutic regime.

3. Comparison of the new treatment with standard treatment in patients with newly diagnosed GBM. If a therapy has shown effectiveness against TMZ-resistant GBM, it is advisable to study it in new cases as an alternative (or even replacement) to standard treatment.

We also deliberately include the date on which the study record was first available on ClinicalTrials.gov [63]. This makes it possible to identify viral vectors which have been discontinued for various reasons (including insufficient efficacy) from those that are still in ongoing trials but without published results yet (Table 2).


#### **Table 2.** Clinical trials using viral vectors.


**Table 2.** *Cont*.





GBM—glioblastoma multiforme; TMZ—temozolomide; IFN—interferon; SoC—standard of care; RTS—RheoSwitch Therapeutic System; VPs—vector particles; HGG—high grade glioma; pfu—plaque forming unit; WHO—world health organization.

#### **3. Discussion**

The search for a gene therapy solution is driven by the abysmal prognosis currently typical for GBM. As of today, many different ideas have been proposed and tested, some of which are summarized above. However, so far, no obvious breakthrough is evident.

*Cancers* **2020**, *12*, 3724

Of the many studies listed in Table 2 and other parts of this review, we have selected two, both using Ad, which have led to interesting results and were published recently. They pursue different strategies and are interesting to compare.

Lang et al. reported the outcome of the trial of DNX 2401 (Delta-24-RGD) on 25 patients without surgical resection and 12 patients where the vector was first injected into the tumor via an implanted catheter, which was followed by surgical removal of the tumor 14 days later and multiple intramural injections of DNX 2401 [64]. Viral loads varied between cases between 10<sup>7</sup> and 3 <sup>×</sup> <sup>10</sup><sup>10</sup> viral particles (vp) in 1 mL volumes. The paper mentions that 3 <sup>×</sup> <sup>10</sup><sup>10</sup> vp in 1 mL was the highest concentration of Ad which could be manufactured, which is close to the experience of our laboratory. In the group treated with a single intratumoral injection of DNX 2401 (no surgery), tumor reduction occurred in 18 of 25 patients (72%). The median survival time in that group was only 9.5 months, regardless of the vector dose, which does not look to be a major success; however, five patients (20%) from this group survived for more than 3 years, which is rather striking given that they were all initially enrolled as recurrent cases with previous history of drug treatments and resistance. Obviously, all patients also received therapies other than DNX 2401. Some limited spread of the vector outside of the brain was detected and anti-Ad5 antibodies appeared in a significant number of patients in both cohorts. In histological specimens, various signs of immune response and inflammatory infiltration as well as viral cell death were evident. The incidence of side effects was very high—for example, 68% experienced headaches, 32% experienced hemiparesis, and 24% convulsions—but the authors argue that they were mainly disease- and not treatment-related. Overall, the paper shows clearly that DNX 2401 can induce an oncolytic effect accompanied by an immune response. This study can, perhaps, be seen as one of the fairly successful preliminary trials which relies on the concept of conditionally replicating oncolytic viruses. From the available information, it seems that the control provided by the requirement for the defective Rb/p16 pathway, as characteristic for many tumors, is sufficiently tight, and the spread of the virus was obviously not too fast and was limited to the locality of injection, rather than becoming generalized encephalitis, which is encouraging news. It is a pity that the integrity of Rb/p16 was not assessed in the patients' biopsies—perhaps that could help to predict the efficacy of the treatment. It would also be important to confirm directly that DNX 2401 is still able to infect the GBM cells after the tumor is given time to undergo clonal selection as it typically happens with GBM. Can GBM cells escape by downregulating the binding sites for the RGD motif, incorporated in this gene therapy agent? It will be very interesting to watch further developments in this dimension.

Recently, the results of NCT02026271 (ClinicalTrials.gov Identifier), which uses Ad–RTS–hIL-12, were published [71]. It is interesting to analyze the approach used in that study in more detail since it highlights many problems facing the field. As mentioned above, Ad–RTS–hIL-12 is a replicationincompetent AVV with a promoter, controllable by a small-molecule drug veledimex (VDX), allowing drug-induced production of interleukin-12 (IL-12) by the cells where AVV genomes are active. The study mainly focused on the demonstration of the ability to induce IL-12 production by VDX and the safety of this treatment. Patients enrolled were all already previously treated with various regimes and, obviously, represented a really tough challenge. After surgical resection of the bulk, AVVs were injected into one spot in white matter as a single injection of 50 <sup>µ</sup>L containing <sup>2</sup> <sup>×</sup> <sup>10</sup><sup>11</sup> viral particles, which corresponds to the titer of 10<sup>13</sup> vp/mL, which our laboratory was never able to achieve and seems to be an extremely concentrated AVV stock administered in a very small volume (compare to the previously mentioned paper [70]). The drug treatment lasted for 14 days. During that period, the drug clearly induced production of IL-12, which spilled over into the systemic circulation, and various signs of inflammatory response were visible in the patients; luckily, they were easily reversible by VDX discontinuation. Interestingly, patients treated with 20 mg VDX seemed to survive better than both those treated with lower and higher doses, the latter probably being a sign of a negative effect of excessive immunostimulation. Over the 30-month observation period, 30 of 31 enrolled patients died, which can hardly be considered a therapeutic success. Nevertheless, the authors successfully demonstrated infiltration of the tumor by the immune cells, indicating that, at least mechanistically, they achieved the

expected result. Considering the results of this study, as reasoned above, non-replicating AVV genomes are inevitably diluted in dividing tumor cells. Since the whole protocol lasted for 14 days, this could be the only period when there was enough active transgene in the remaining GBM cells. Unfortunately, in the paper, there is no information on the presence of the viral genomes in the post-mortem samples. This issue, i.e., survival of the transcriptionally active adenoviral genomes in the GBM, is both interesting and important but we do not have the answer yet. It would be very interesting to know whether VDX could effectively trigger a wave of IL-12 production 3–5 months after the transduction. The other question is whether the cells producing IL-12 were mainly the GBM cells or other cells in the vicinity of the injection track. Overall, this strategy is in progress and seems to critically depend on the ability to quickly destroy the infiltrating GBM cells while the AVV are still functional.

What are the limitations, and can they be overcome, at least theoretically? The first point to consider is that of infection or transduction efficiency and stability of transgene expression. Viral vectors must be able to very efficiently enter the target cells and introduce any transgene cargo into their nuclei. Viral vectors have been extensively used in biomedical research and neuroscience for the last 20 years and there is a wealth of information about many of the vectors, similar to those used in human trials. For example, the internalization mechanism of species C adenoviruses is based on their interaction with CAR and Integrin αvβ5 proteins on the surface of the target cells [90], PVSRIPO requires CD155 [56], MV-CEA cell entry is based on interaction with CD46 [62], and so forth. We argue that this makes strategies involving adenoviral and similar vectors, which require specific GBM surface proteins for entry, vulnerable to the common mechanism of tumor defense based on downregulation of the relevant proteins and consecutive clonal selection and expansion. Ad has been used in vitro by many groups, including ourselves, in experimental neuroscience for transgene expression in both neurons and glia [91,92]. In vivo, however, these vectors clearly prefer astrocytes over all other cell types in the brain [25,92], and thus, unmodified Ad cannot be seen as a universally efficient delivery tool, irrespective of the putative origin of the GBM. In some Ad-derived gene therapy vectors, such as DNX 2410, a specific modification of the fiber H-loop should enable them to bind to specific integrins expressed by many tumor cells, but this mechanism is vulnerable to downregulation of the target integrins. The obvious differences in transductional tropism between adeno- and lentiviral vectors in rodent CNS were demonstrated long ago [93]. It was noted that vesicular stomatitis virus G-protein (VSVG)-pseudotyped lentiviruses which do not utilize a specific receptor-dependent entry pathway have a much wider transduction potential. In our laboratory, VSVG-pseudotyped HIV-derived lentivirus was used to transduce six patient-derived GBM cell lines with an apparent 100% success rate (unpublished observations). We suggest that the requirement for a specific interaction partner protein on the target cells is a limitation of vectors used for gene therapy of GBM because these can be easily eliminated by selection, making tumor cells resistant. Could lentivirus be a route to explore? Another fundamental issue is the possible silencing of exogenous expression cassettes. In experimental neuroscience, this was noted a long time ago for a commonly used promoter CMV, which is incorporated in several viral vectors listed here [31,94]. The mechanisms of CMV-mediated transgene silencing are not well understood but could be based on RNA interference or methylation of the viral promoters by cell defense machinery [95,96]. Additionally, as mentioned above, replication-incompetent vectors which stay episomal fail to propagate to the progeny of the cells they invade, which means that unless the infected GBM cells die immediately, they will eliminate viral genomes by dilution after a few divisions.

The next important point is the mechanism of action of viral gene therapy. Oncolytic viruses use the natural feature of viruses to multiply and destroy cells. Obviously, such processes, if uncontrolled, will be lethal, as exemplified above by Reolysin or C134. Various mechanisms of transcriptional control are used to enable replication predominantly in fast dividing cells. However, if this strategy is really successful and, thus, leads to a powerful cytopathic effect, rapid destruction of GBM in clinical settings can cause brain edema with subsequent impairment of vital functions and even death. Specificity of viral gene therapy is a fundamental problem. For cytopathic viruses, this solely relies on the dependence

of their replication on factors highly expressed by tumor cells. However, GBM cells, even within the same tumor, are heterogenous [97]. Is it even possible to find a ubiquitous driver/controller of viral replication in the pool of diverse GBM cells? At this point, such a possibility remains to be demonstrated. So far, the selectivity of the published vectors is obviously not sufficient to fully prevent destruction of normal brain cells. With some vectors, such damage can be inflicted by the conversion of pro-drugs into toxic specimens which are then released—the so-called bystander effect. This problem is particularly relevant to the brain, where elimination, dilution and biodegradation of these harmful molecules might be slower than in the periphery. An added problem introduced by replicating vectors is the release of viral particles into the bloodstream, leading to an inevitable immune response.

The success of viral gene therapy critically depends on the physical access of the virus to the GBM cells. Shall they be injected into the brain at the time of surgery or administered using some other means? It would be ideal to inject the virus into the bloodstream because it could reach all GBM cells which are spread within the parenchyma, but can this be done? Outside of the field of neuro-oncology, the best current example of an attempt to achieve generalized expression in the human brain with an i.v.-injected viral vector is Zolgensma (AVXS-101), an adeno-associated viral vector carrying the SMN1 transgene [98]. However, in humans, this virus has to be delivered before 2 years of age, when the blood–brain barrier is still not completely mature, and large doses are used, requiring administration of steroids to prevent a severe immune response [99]. This is in stark contrast with multiple studies in mice where a brain-wide expression has been achieved with some strains of adeno-associated virus injected i.v. [100]. Adeno-associated viruses are extremely small and definitely have the best chances of reaching the CNS when their concentration in the bloodstream is high enough, but they do not seem to have any tropism to GBM in addition to the fact that the adult human BBB is probably completely impermeable to them. Moreover, after a single application into the bloodstream, a strong antibody response is inevitable, making this a "single shot only" strategy. It is therefore unlikely that we will see successful targeting of disseminated GBM with any type of currently available viral vector applied via the bloodstream.

To summarize, the attempts to develop an efficient gene therapy for GBM with viral vectors face the following fundamental problems.

(a) Vectors relying on a specific mechanism of internalization are unlikely to be successful because of the extreme instability of GBM genomes, the multitude of clones in the same tumor and the ease of clonal selection of resistant cells to which the virus will have no access. It follows that using less specific mechanisms of viral entry might be a winning strategy.

(b) GBM cells divide, and some do it at a very high pace. In such cells, non-integrating viral genomes will be rapidly diluted and probably become inefficient, unless they cause immediate death of the cell. The ability to silence transgenes adds to this problem. The only way to ensure downwards transmission of the transgene is the use of integrating vectors, such as lentiviruses.

(c) Specificity of the effect is one of the key requirements and we have listed, above, some of the strategies used to limit the impact to GBM cells vs. the rest of the brain. So far, many of these strategies have been demonstrated to work in vitro and sometimes even in GBM-bearing mice in vivo. Whether a sufficiently reliable and universal strategy can be found for clinical application remains to be seen. We hypothesize that one avenue to explore is to try to suppress the mitotic apparatus, since healthy cells in the postnatal human brain rarely or never divide.

(d) Injection in the bloodstream is unlikely to be successful. We are therefore left with a necessity to infiltrate with viral gene therapeutics the areas of the putative GBM growth during the debulking surgery or, possibly, by stereotaxis at a later stage.

We hope that this review will allow readers to get a feel for the current options for the viral gene therapy of GBM and initiate a discussion about its future directions. We suggest that a more plausible strategy might be to focus on viruses which enter via a non-selective, ubiquitous mechanism. We hypothesize that it might be possible to irreversibly block processes critical for dividing tumor cells which are dispensable for quiescent healthy brain cells. Mitosis is a highly specialized stage

of a cell's life and depends on a range of proteins which are expressed in non-dividing cells at low levels. This idea may be illustrated by the current attempts to target, for example, cyclin-dependent kinases with inhibitors. The key difference is that the peripheral cells—for example, in the bone marrow—should not be affected and inhibited by a virus which is delivered into the brain parenchyma. Hence, the issue of systemic toxicity could become less critical.

As stated in the beginning, this review reflects the view of the experimentalist neuroscientists and, hopefully, might stimulate a discussion leading to new discoveries in the field of neuro-oncology.

#### **4. Conclusions**

Viral gene therapy of GBM is a promising field but several major hurdles need to be overcome for it to become an accepted part of the currently available portfolio of therapeutic interventions. As yet, some potentially encouraging results have been obtained with a conditionally replicating oncolytic Ad, but the fundamental challenge of tumor resistance via downregulation of the proteins, critical for viral proliferation remains to be overcome. Obviously not all the options have been yet explored and we hope to see new types of vectors entering clinical trials in years to come.

**Author Contributions:** All authors have read and agreed to the published version of the manuscript.

**Funding:** O.M. was funded by the 5/100 Programme from the Russian Government to Baltic Federal University, Kaliningrad, Russian Federation and The Fellowship of the President of the Russian Federation. A.G.T. was funded by the British Heart Foundation (PG/18/8/33540). S.K. and A.G.T. were funded by the British Heart Foundation (RG/14/4/39736).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Nanomedicine: A Useful Tool against Glioma Stem Cells**

**Elia Bozzato <sup>1</sup> , Chiara Bastiancich <sup>2</sup> and Véronique Préat 1,\***

2

	- Institute Neurophysiopathol, INP, CNRS, Aix-Marseille University, 13005 Marseille, France; chiara.bastiancich@univ-amu.fr

**Simple Summary:** Glioblastoma is one of the deadliest brain cancers, and despite the efforts made in the last few years, the life expectancy of patients is still low. In most cases, even with the best treatments available, the tumor will eventually return. One of the main causes of this appears to be a fraction of cancer cells that are known as glioma stem cells. They have different characteristics than normal cancer cells, and some drugs can eliminate them. However, using such drugs is not always safe or effective, and nanomedicine can have improved effects as well as additional benefits. This review focuses on the nanomedicine strategies that have been employed in the last 5 years and their relative advantages, which make nanomedicine a promising approach for the eradication of glioma stem cells.

**Abstract:** The standard of care therapy of glioblastoma (GBM) includes invasive surgical resection, followed by radiotherapy and concomitant chemotherapy. However, this therapy has limited success, and the prognosis for GBM patients is very poor. Although many factors may contribute to the failure of current treatments, one of the main causes of GBM recurrences are glioma stem cells (GSCs). This review focuses on nanomedicine strategies that have been developed to eliminate GSCs and the benefits that they have brought to the fight against cancer. The first section describes the characteristics of GSCs and the chemotherapeutic strategies that have been used to selectively kill them. The second section outlines the nano-based delivery systems that have been developed to act against GSCs by dividing them into nontargeted and targeted nanocarriers. We also highlight the advantages of nanomedicine compared to conventional chemotherapy and examine the different targeting strategies that have been employed. The results achieved thus far are encouraging for the pursuit of effective strategies for the eradication of GSCs.

**Keywords:** glioblastoma; brain tumor; nanomedicine; cancer stem cell; targeted therapy

## **1. Introduction**

Glioblastoma (GBM) is a grade IV astrocytoma, and the prognosis for GBM patients is very poor. Currently, the standard of care therapy includes surgical resection of the main tumor mass, followed by radiotherapy and concomitant chemotherapy with oral temozolomide (TMZ) [1]. However, this therapy has limited success due to the intrinsic characteristics of the tumor, such as the tumor heterogenicity, development of chemoresistance, and presence of glioma stem cells (GSCs). These factors lead to tumor recurrences. Recently, the overall survival of GBM patients has slightly increased from 16.0 months to 20.9 months with the additional application of tumor-treating fields to the standard of care therapy [2]. Nevertheless, despite this significant improvement, GBM still remains an unmet medical need, and successful long-term therapies urgently need to be found.

GBM is characterized by resistance to treatment and high intertumor and intratumor phenotypic and genetic heterogeneity [3]. Many advances have been made in the past decade to uncover the genetic diversity of GBM and the clone-specific functional profile,

**Citation:** Bozzato, E.; Bastiancich, C.; Préat, V. Nanomedicine: A Useful Tool against Glioma Stem Cells. *Cancers* **2021**, *13*, 9. https://dx.doi.org/ 10.3390/cancers13010009

Received: 6 November 2020 Accepted: 18 December 2020 Published: 22 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/).

showing that even within the same tumor, the combination of various molecular subclasses could be found (e.g., [4–6]). This diversity also indicates the presence of GSCs, which are defined as a quiescent subpopulation of cancer cells with high self-renewing abilities that are able to recreate a tumor after transplantation [7]. Even though the precise cell of origin of GBM is still a controversial issue, as some experts contend that it arises from a subpopulation of neural stem cells, while others argue that it arises from the transformation of more differentiated astrocytes [8], it is now recognized that presence of GSCs and crosstalk with their supportive niche contributes to tumor malignancy [9]. Moreover, they are responsible for the onset of tumor recurrence, and therefore, are a promising therapeutic target to prevent GBM relapse. Several publications have recently highlighted how GSC location at the invasive margins, heterogeneity, and dynamism (transcriptional, epigenetic, and metabolic) can play an important role in the response to surgery, radiotherapy, and chemotherapy (e.g., [10,11]). A review from Liu et al. [12] evaluates the potential involvement of brain tumor stem cells in postoperative stem cell niches and their role in tumor relapse, and their input should be considered for the development of adapted nanomedicines. Indeed, while it is true that most nanomedicines are intended for a post-surgical application, most studies report their efficacy on preclinical models designed to treat established GBM. This overlooks the fact that surgical resection of brain tumors can create an environment that can stimulate the proliferation of residual tumor cells (GSCs, tumor microtubes, and infiltrating GBM cells), leading to tumor recurrences. Here, we would like to highlight how nanomedicines can be used to overcome some of the limitations of conventional chemotherapies targeting GSCs, thus representing a promising approach for GBM therapy.

#### **2. Glioma Stem Cells**

Due to their dormant state, GSCs are intrinsically resistant to conventional chemotherapeutics that act on rapidly proliferating cancer cells, such as alkylating agents, antimetabolites, and mitotic inhibitors. Furthermore, they can actively resist chemo- or radiotherapy by the activation of checkpoint mechanisms, in order to recover efficiently from the genotoxicity induced by the therapy. Another mechanism of resistance for GSCs is the expression of drug efflux mechanisms (ABC transporters) to protect the cells from xenogeneic molecules [13]. Autophagy, which is required for stemness maintenance, not only in normal tissue stem cells but also in GSCs, has been shown to contribute to therapy resistance [14]. Moreover, the Notch signaling pathway is involved in the resistance of GSCs to radiotherapy. The inhibition of this pathway through γ-secretase inhibitors is able to induce radiosensitivity by targeting the subpopulation of cells that bears the GSC marker CD133 [15].

GSCs are also characterized by specific pathways that are implied in the conservation of stemness characteristics or in tumor formation. The Notch pathway can inhibit cell differentiation and therefore maintain the stem-like properties of GSCs [16]. In patient-derived GSCs taken from the periphery of the tumor, Hu and collaborators demonstrated that Notch promotes self-renewal and inhibits differentiation [17]. In recurrent GBM samples, CD133, Notch, and VEGF expression was higher after radiotherapy and chemotherapy, and after a second surgery and treatment with bevacizumab, the overall survival was significantly longer for Notch-negative patients [18]. Furthermore, cells from the interface region are CD133+/Notch1+, and there is a positive-feedback loop between NOTCH1 and SOX2 [19]. The aberrant activation of Wnt signaling causes the transcription of c-Myc and other target genes leading to tumor formation [20]. It also participates in the maintenance of stemness characteristics by regulating the expression of PLAGL2 (pleiomorphic adenoma gene-like 2) that is able to suppress the differentiation of GSCs [21]. Finally, the Sonic Hedgehog (Shh) pathway is essential for cell survival and sustained growth of the tumor. In fact, it regulates the expression of stemness genes in glioma GSCs [22].

GSCs can be isolated from cancer cells and tissue stem cells using specific intracellular or extracellular markers (Figure 1), although functional validation should also be

employed to assess the stem cell characteristics (self-renewal and tumor formation) [8]. The most common marker is CD133 or Prominin-1, a transmembrane glycoprotein that is also expressed by human neural stem cells [23]. However, evidence also suggests the existence of CD133- GSCs [24], and therefore, a single marker cannot automatically identify GSCs. Other common markers are A2B5, a glycolipid found on the cell surface of oligodendrocyte progenitors; stage-specific embryonic antigen-1 (SSEA-1, also known as CD15) an embryonic antigen with a carbohydrate structure; and Nestin, a filament protein that is also expressed by neural progenitor cells [25]. Additionally, high ALDH-1 (aldehyde dehydrogenase 1) activity and the high extrusion of xenobiotics through ABC transporters are two functional markers that have been associated with GSCs [25].

**Figure 1.** Intracellular and extracellular glioma stem cell (GSC) markers. Adapted from [26].

The metabolism of GSCs is very plastic. In fact, the dependence on oxidative or nonoxidative metabolism is heterogeneous throughout the tumor. Fast-dividing cells rely more on anaerobic glycolysis [27], creating the Warburg effect as an adaptation metabolism for their rapid growth. In an acidic environment, GSCs can undergo mesenchymal differentiation, resulting in an increase of therapy resistance [28]. On the other hand, slowly proliferating cells are more dependent on oxidative phosphorylation (OXPHOS) and lipid oxidation, and GSCs in particular can metabolize various substrates, making it difficult to find a pharmacological target [10]. GSCs have been reported to have lower glucose consumption than normal GBM cells [10]. However, depending on their microenvironment, they are able to adapt to nutrient and stress conditions by increasing their glycolytic activity [10]. In fact, GSCs can also upregulate high-affinity transporters, such as GLUT3, to obtain sufficient nutrients and support their rapid metabolism [10].

GSCs can adapt and are able to interact with different niches. For example, GSCs that are located at the perivascular niche are in contact with the endothelium that secretes ligands that bind to the transmembrane Notch receptor on GSCs, leading to the activation of the Notch pathway and supporting GSC self-renewal. In exchange, GSCs can transdifferentiate into pericytes to contribute to the vascular structure, thus promoting tumor growth [26]. GSCs can also interact with immune cells through their metabolism. They can regulate the microenvironment and generate stress for immune cells, thus creating a globally suppressive tumor microenvironment that allows for immune escape and tumor progression [10]. In return, macrophages, which are the most represented type of tumor-infiltrating cell, participate to the regulation of GSC metabolism by increasing their fatty acids synthesis and trafficking, thus promoting lipid oxidation, which is one of the main metabolic pathway of GSCs [10]. Moreover, through the secretion of interleukin 10 (IL-10) and transforming growth factor beta (TGF-β), GSCs are able to suppress the tumor-associated microglia, generating an M2 immunosuppressive phenotype [26]. Furthermore, GSCs are able to regulate immune cells directly, causing the activation of regulatory T cells, the inhibition of cytotoxic T cell proliferation, and the induction of cytotoxic T cell apoptosis [29,30].

GSCs however are not a static, discreet cell subpopulation; their stemness is rather a dynamic and reversible state. There is considerable evidence that EMT (epithelial to mesenchymal transition) is involved in the dynamism of GSCs [31], and that various factors can stimulate or revert this transition [32–34]. Furthermore, based on their location in the tumor, they can have different characteristics and exert different functions: while GSCs in the core hypoxic regions support proliferation and therapy resistance, GSCs from the outer invasive region are enriched for their invasive potential and promote tumor recurrence after resection [11].

#### **3. Chemotherapy against GSCs**

Despite the high number of researchers and clinicians investigating GBM, treatment options for this tumor have remained nearly unchanged for the last 15 years [35]. Some progress has been made in the field of personalized therapy, thanks to the ChemoID assay, which consists of a viability test on GSCs and bulk tumor cells from freshly resected samples, in order to identify the most effective drug or combination of drugs. Patients were therefore treated with the selected drugs, and 12 out of 14 cases had complete or at least partial response to the therapy [36]. In order to better relate to intra-tumor heterogeneity, this same approach could be used on samples obtained from different tumor regions from each patient. After the viability assay on GSCs from each sample, the patient could be treated with the combination of drugs that demonstrated cytotoxicity in the different regions. However, the study from Ranjan et al. [36] suggests that, along with chemotherapy directed against GBM cells, combination therapies also targeting GSCs could be necessary. The possible approaches that can be adopted in order to eliminate GSCs are represented in Figure 2.

**Figure 2.** Anti-GSC molecules and their mechanisms of action.

One of the strategies that has been explored to attack the GSC population is to inhibit specific GSC pathways, such as Notch, Wnt, and Shh. For example, the inhibition of Notch activation through γ-secretase inhibitors is reported to reduce the CD133-positive GBM cell population in vitro and to reduce tumorigenicity of pretreated brain tumor cells subcutaneously injected in athymic mice [37]. Cyclopamine, a Shh inhibitor, was able to reduce neurosphere formation and block the tumor formation of intracranially injected GSC cells [38]. Resveratrol can modulate the Wnt pathway and decrease the proliferation and mobility of GSCs [39]. Metformin can inhibit AKT signaling, which is involved in the response to stress conditions to promote GSC growth and survival [40]. Its analog Phenformin is also able to inhibit the self-renewal of GSCs, thus reducing the growth of xenograft tumors and prolonging mice survival [41]. Napabucasin, a STAT3 inhibitor, can inhibit the expression of stemness-associated genes and the growth of GBM spheroids in vitro [42], and has led to the loss of GSCs associated genes, induction of apoptosis, and inhibition of in vivo tumor growth of GSCs derived from recurrent GBM [43]. This drug has also been used in a phase I/II clinical trial in combination with TMZ [44]. Glasdegib and RO4929097, a Shh pathway inhibitor and a γ-secretase inhibitor, respectively, are also being used in combination with TMZ in two different ongoing clinical studies [45,46].

GSCs are also implied in therapy resistance, and they can actively participate to this process though mechanisms like DNA repair, pro-surviving signaling, and most importantly, drug efflux [47]. Therefore, another approach is to employ P-gp (permeability glycoprotein) or to induce the differentiation in normal GBM cells, in order to sensitize them to conventional chemotherapy. It has been demonstrated that CD133 contributes to the regulation of MDR1 through the phosphoinositide 3-kinase (PI3K)- or Akt–NF-κB signal pathway [48]. Moreover, the invasive margin of GBM displays an increased expression of ABCG2 [49], which is another efflux pump belonging to the ABC transporters superfamily. It has been shown that reduction in ABCG2 expression can decrease the cell migration and invasion of GSCs [50]. An example of P-gp inhibitor is epigallocatechin gallate, which was able to reduce the P-gp expression and neurosphere formation of GSCs obtained from the U87 cell line, and increase the sensitivity of these cells to TMZ [51]. The differentiating agent transretinoic acid was able to deplete GSC markers and reduce the formation of neurospheres, and the effect on cell migration was improved in combination with rapamycin [52]. Resveratrol can induce the degradation of Nanog, which is essential for stemness maintenance, thus leading to the loss of GSC markers and decreased tumorigenicity [53]. Curcumin was demonstrated to activate autophagy, thus triggering the differentiation cascade of GSCs and causing a decrease in its self-renewal and clonogenic abilities [54]. Finally, bone morphogenetic protein 4 (BMP4) is commonly used to reduce the number of GSCs by inducing their differentiation, and therefore increasing the response to conventional therapies [55]. BMP4 is also currently being administered through convection-enhanced delivery (CED) in a phase I clinical trial [56].

Additionally, tackling the tumor microenvironment through antiangiogenic or antivasculogenic molecules can also decrease the number of GSCs. The treatment with bevacizumab was able to reduce the number of CD133+/Nestin<sup>+</sup> cells, along with reducing the microvasculature density and tumor growth in U87 glioma xenografts [57]. Moreover, the administration of antibodies against a proangiogenic factor like IL-6 could delay the growth of tumors obtained by the injection of GSCs in a xenograft model [58]. Another antivasculogenic molecule, the biciclame compound plerixafor (AMD3100), was able to inhibit irradiation-induced vasculogenesis in vivo by preventing the binding of the chemokine stromal cell-derived factor 1 (SDF-1, involved in the migratory process of GBM) to its receptor C-X-C chemokine receptor type 4 (CXCR4) [59].

Targeting the DNA methylation of GSCs through histone deacetylase inhibitors (HDAC) inhibitors is another strategy that has been described in the literature. In fact, suberanilohydroxamic acid (SAHA) is able to induce autophagy in GSCs, thus leading to decreased cell viability in vitro and reduced tumor growth in vivo [60].

Finally, salinomycin has been used on GBM cells in combination with HDAC inhibitors, such as valproate and vorinostat [61], and it has also shown anti-CSC activity in other cancer types [62]. Even though its mechanism of action needs to be elucidated, it has been reported that it can induce ROS production in GSCs, thus leading to endoplasmic reticulum stress and cell death via regulated necrosis [63]. Additionally, verteporfin can target the mitochondria of GSCs and inhibit OXPHOS without any toxicity to normal cells [64].

In many cases, the elimination or impairment of GSCs has led to decreased tumor growth and increased survival in preclinical in vivo models, highlighting once again the importance of tackling GSCs in the treatment of GBM. However, only a few of the abovementioned molecules are being tested in clinical trials (mostly GSC pathway inhibitors), and the results are not yet available.

#### **4. Nanomedicine against GSCs**

#### *4.1. Nanomedicine for GBM Treatment*

The intrinsic limits of chemotherapy are the lack of specificity, harmful side effects, low therapeutic index, and transport limitations [65]. Indeed, many drugs, including those cited in the previous chapter, have poor solubility, high toxicity due to the uncontrolled drug biodistribution, or poor stability in the physiological environment. Moreover, when administered systemically, they need to cross the blood–brain barrier (BBB) to reach the GBM tumor site at therapeutic concentrations, often leading to severe, dose-related systemic side effects. Some drugs are not stable in biological fluids and have a very short half-life; therefore, multiple administrations are required to achieve the therapeutic concentration at the tumor site, reducing patient compliance.

Nanomedicine can help provide a solution for these problems. The encapsulation of drugs in nanosized carriers can protect them from degradation, increase the amount of drug reaching the tumor site, and decrease the intensity of the side effects, thus increasing the safety of the treatment. The maintenance of a correct therapeutic level can be facilitated by the controlled release of the drug over time. Moreover, the surface of the nanocarrier can be suitably modified with targeting moieties in order to actively and specifically recognize GBM cells and GSCs, or to cross the BBB more easily. This can further increase the uptake of the nanoparticles (NPs) by GSCs and enhance their residence time in the tumor.

The BBB is a natural barrier that protects the central nervous system from exogenous compounds or macromolecules. Even though in GBM the patients' BBB parts are disrupted and leaky [66,67], the crossing of the BBB still represents a challenge for GBM treatment, due to the poor blood perfusion and the high interstitial pressure. The BBB can be bypassed by administering drugs locally, through implants or CED. A local delivery has the advantage of increasing the drug concentration in its site of action while minimizing the side effects. However, systemic delivery is still the preferred strategy for inoperable tumors, and thanks to its being less invasive, also allows for the administration of multiple doses.

Herein, we review the nanomedicine approaches that have been developed in the last 5 years against GSCs, dividing them by nontargeted and targeted systems (Tables 1 and 2, respectively).

**Table 1.** Nontargeted nanosystems for the treatment of preclinical glioblastoma (GBM).



**Table 1.** *Cont*.

Legend: \* free drug. Abbreviations: HOTAIR: HOX transcript antisense RNA; TMZ: Temozolomide; NPs: nanoparticles; SPION: superparamagnetic iron oxide NPs; GSCs: glioma stem cells; CSCs: cancer stem cells; i.p.: intraperitoneal; i.v.: intravenous.




#### **Table 2.** *Cont*.


**Table 2.** *Cont*.

Legend: \* free drug. Abbreviations: CK2α: protein kinase CK2 catalytic α subunit; EGFR: epidermal growth factor receptor; TMZ: Temozolomide; NPs: nanoparticles; TfR: transferrin receptor; RGD: Arginyl-glycyl-aspartic acid peptide; GSCs: glioma stem cells; CSCs: cancer stem cells; i.p.: intraperitoneal; i.v.: intravenous.

#### *4.2. Non-Targeted Nanomedicines*

NPs can exploit the enhanced permeation and retention (EPR) effect to accumulate and increase their residency time at the tumor site [92,93]. The EPR effect consists of the preferential accumulation of NPs in the tumor site caused by two components: (i) due to their rapid growth, blood vessels in the tumor present a leaky and less organized structure than normal blood vessels; and (ii) inefficient lymphatic drainage. However, in the past few years, due to its intratumor and intertumor variability, together with the differences between animal models and patients, the EPR effect has been questioned [94,95]. Despite this controversial topic, in order to eliminate GSCs, nanomedicine can still offer many advantages when compared to conventional chemotherapy (Table 1, Figure 3).

**Figure 3.** Potential advantages of nanomedicine against GSCs.

One of the advantages of using a drug delivery system is the increase in safety compared to the free drug. For example, paclitaxel-loaded chitosan NPs covered with 1,3β-glucan were demonstrated to have a lower half maximal inhibitory concentration (IC50) value than the free drug on C6-derived stem-like cells, and significantly lower hemolytic activity than the drug suspension [96], thus showing an increased safety profile. Cytarabine-loaded liposomes showed an increased safety profile compared to the free drug [97]. This formulation is currently being examined in a phase I/II clinical trial [98], and is reported to tackle the subventricular zone, which is one of the proposed sites of origin for GSCs [99].

Another advantage of nanomedicine is the increased stability. The encapsulated molecule can be protected from degradation processes, such as hydrolysis, enzymatic degradation, or metabolism. This is usually the case for nucleic acids, such as miRNAs and siRNAs, as their blood half-life is very low. Various types of nucleic acids have been encapsulated in polymeric NPs [69,74], lipid–polymer NPs [73], superparamagnetic iron oxide NPs [71,76], and gold NPs [72,100]. These formulations were able to increase the internalization of the nucleic acid by passive targeting, inducing an efficient silencing of GSC-related genes, reducing GSC proliferation and invasion, and prolonging animal survival in vivo.

Moreover, encapsulation in a drug delivery system can also reduce the efflux of the drug. Unlike free drugs, which enter the cells through diffusion and locate near the efflux pumps, nanomedicines enter the cells through endocytosis and are transported into the cell via endo-lysosomal trafficking, preventing them from being a substrate for drug efflux pumps [101]. Etoposide, which is an efflux pump substrate, was loaded in layered doublehydroxide nanocomposites, thus prolonging its retention time in the cells and increasing its accumulation in the tumor site. This brought about the elimination of GSCs in vitro and decreased tumor growth in the xenograft mouse model [68].

Nanomedicine can also improve the bioavailability of molecules like curcumin. Curcumin was formulated in liposomes in combination with epicatechin gallate and resveratrol, and after intraperitoneal injection, it obtained an almost constant plasma concentration, which led to increased mouse survival in the in vivo experiment. Furthermore, this liposomal formulation was able to decrease the GSC subpopulation of GL261 cells [70].

Additionally, even though this advantage is less common than others, drug delivery systems can in some cases increase the activity of the drugs. Atorvastatine-loaded polymeric micelles were indeed able to inhibit the growth of CSC spheroids compared to the single drug [102]. In the case of zinc-doped copper oxide nanocomposites, the NPs have an intrinsic inhibitory effect, decreasing the colony formation of TMZ-resistant GSCs, but at the same time exerting lower toxicity on normal cells [75].

#### *4.3. Targeted Nanomedicines*

The design of nanosystems can be implemented by the addition of a targeting agent, usually an antibody or a ligand, that selectively recognizes cell surface markers overexpressed in a certain population. This has the aim of making the carrier interact with the cell surface, and thanks to the interaction, induce its cellular uptake by endocytosis, ultimately acting as a Trojan horse and releasing its cargo directly inside the cell. Therefore, targeted nanomedicines have the advantage of increasing the amount of cytotoxic agent inside the target cell, reducing the proportion of drug that is delivered to healthy tissues.

Different strategies have been employed to specifically target GSCs (Table 2, Figure 4), and the most common and straightforward is the use of antibodies against CD133, which is the most described GSC marker in the literature. The conjugation of anti-CD133 antibodies to polymeric dendrimers loaded with mercaptoundecahydrododecaborate, a substance employed in boron neutron capture therapy, has led to significantly increased drug uptake and the decreased clonogenic survival of CD133+ cells after neutron radiation. This also produced significantly prolonged mouse survival in an orthotopic xenograft model [88]. Anti-CD133 antibodies were also used as carriers and targeting agents at the same time. IR700, an agent employed in near-infrared photoimmunotherapy, was conjugated to the antibody with a theranostic application. The authors successfully detected CD133+ cells

following intravenous administration and laser irradiation in mice bearing orthotopic brain tumors initiated from patient-derived GSCs, and at the same time observed extended overall survival [84].

**Figure 4.** Targeting strategies employed to reach GSCs.

Another common strategy that has been adopted is the conjugation of anti-transferrin receptor (anti-TfR) antibodies. Resveratrol-loaded targeted liposomes are capable of reducing the growth of glioma neurospheres. Moreover, the targeted formulation has shown a significantly increased association with glioma neurospheres compared to the nontargeted liposomes [103]. In addition, targeted polymeric NPs were conjugated to antisense oligonucleotides against laminin-411, which is correlated to GSC marker expression. This nanosystem was able to reduce the protein expression and prolong the survival of mice intracranially transplanted with LN229 and U87 MG cells [78].

Another approach that has been applied is the use of the anti-EGFR antibody. Cetuximab was bound to iron NPs, and showed enhanced uptake by EGFR- and EGFRvIIIexpressing GSCs and neurospheres, as well as a significantly increased animal survival in vivo [87].

One of the main obstacles that nanomedicine encounters in the treatment of GBM is the crossing of the BBB, whose natural function is to prevent exogenous structures from reaching the brain. Consequently, nanocarriers for GBM must be designed to cross the BBB and reach the tumor site in higher amounts. The cyclic RDG peptide was linked to micelles loaded with an antisense nucleotide against TUG1, a gene participating in Notch signaling. The formulation in a targeted micellar delivery system allowed the crossing of the BBB and the accumulation in the tumor site, thus enhancing TUG1 silencing in a mouse xenograft model [91].

Several authors developed multifunctional nanocarriers by combining the targeting of GSCs and the crossing of the BBB. TMZ-loaded liposomes were conjugated with an anti-CD133 antibody for targeting GSCs and angiopep-2 for BBB crossing. Angiopep-2 can bind to the low-density lipoprotein (LDL) receptor-related protein, which is highly expressed on the endothelium of the BBB. This system was able to bind to GSCs more efficiently than the nontargeted system, and showed an increased permeability of the BBB in vitro. Moreover, the dual-targeted liposomes were able to decrease the tumor size and prolong the mice survival in the orthotopic, in vivo GSC model [90]. Paclitaxel and surviving siRNA-loaded liposomes were also conjugated with an anti-CD133 aptamer for targeting GSCs and Angiopep-2 for crossing the BBB. Targeted liposomes had an improved uptake in cancer stem cells compared to the nontargeted ones. Moreover, while Taxol

1

and nontargeted liposomes had almost the same effect, targeted liposomes produced a significant decrease in the cell viability of CD133+ cells. The formulation was also able to significantly reduce tumor growth and prolong mouse survival in vivo [83].

Finally, the same targeting moiety can be employed for both targeting GSCs and crossing the BBB. A mannose derivative, p-aminophenyl-α-d-mannopyranoside, was used to functionalize curcumin- and quinacrine-loaded liposomes. Compared to the nontargeted one, this nanocarrier was able to cross a BBB in vitro model more efficiently and significantly increase the uptake in GSCs. Moreover, the targeted liposomes could increase the median survival and inhibit the tumor growth of tumor-bearing mice [89]. Surprisingly, the anti-TfR antibody has also been demonstrated to exert both functions in p53-loaded liposomes. This formulation was also capable of crossing the BBB and targeting GSCs in vivo. Moreover, the delivery of a p53-encoding plasmid was able to decrease the expression of O<sup>6</sup> -methylguanine-DNA methyltransferase (MGMT), thus increasing the sensitivity of the cells to TMZ. Due to the promising preclinical results, this formulation is currently under investigation in a phase II clinical study; however, no results have been released yet [80,81,104].

#### **5. Conclusions**

Despite extensive research, the need for an efficient, long-term treatment against GBM remains high. As GSCs play a major role in GBM recurrence and resistance to treatment, it is important to take them into account and include anti-GSC molecules in combination regimens to increase their therapeutic benefit. In this review, we have examined the nanosystems that have been developed and used against GSCs in the past 5 years (Tables 1 and 2), trying to highlight their advantages compared to conventional chemotherapeutic treatments. Surprisingly, most of the delivery systems reported in the literature have been developed for systemic administration, while the use of local delivery systems, which have the advantage of bypassing the BBB and delivering high drug concentrations at the tumor site, are poorly represented. In our opinion, a suitable delivery system should be adaptable to the resection cavity to ensure adhesion to the brain tissue, thus delivering the drug(s) in the regions where recurrence is more probable. In fact, most of the recurrences arise nearby the resection cavity [105]. Moreover, this delivery system should include multiple drugs, at least one directed against normal GBM cells and at least one directed against GSCs, as the combination therapy approach is considered promising and is being tested in various clinical trials [106]. Finally, the drug(s) should preferentially be released from the delivery system in a sustained way, in order to maintain a therapeutic drug concentration at least until the beginning of the conventional radio- and chemotherapy (or even beyond, provided that none of the drugs interact with TMZ in an antagonistic manner). However, only a few of the nanomedicine systems included in this review have reached the clinical stage up to now, and therefore, there is still considerable research to be performed in order to explore new potential routes or consolidate established nanomedicine strategies. However, nanomedicine can be a promising strategy for adjuvant GBM therapies, in order to eliminate the GSC population and eradicate these deadly tumors.

**Author Contributions:** Literature search, manuscript writing, figures, and tables, E.B.; critical revision of manuscript and supervision, C.B.; supervision and critical revision of manuscript, V.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors thank the EuroNanoMed III found Gliogel, the Fondation ARC pour la Recherce sur le Cancer, the Fondation contre le cancer (Belgium: FAC-C:2016/830), and the Fonds De La Recherche Scientifique (FNRS: Research Credit 33669945) for their financial support.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**

