chemotherapy. Additionally, the tremendous heterogeneity of these tumors results in chemotherapy 
resistance in many of the cells. Attempts have been made to improve delivery of chemotherapy into 
brain tumors by giving the drugs directly into the carotid artery, by temporarily opening the blood 
brain barrier with hypertonic agents such as Mannitol [16], and by implantation of chemotherapy 
containing polymers directly into the brain tumors [17]. There have been significant complications of 
some of these methods, without consistent improvement in outcome. High dose chemotherapy with bone 
marrow rescue has also been utilized, but again the toxicity has been high and the results poor [18]. 
In summary, despite a large body of clinical research in the treatment of malignant gliomas, there has 
been little overall improvement in the outcome for these patients. 
II. B Preclinical Studies 
II.B.1 The Use of HSV TK with Ganciclovir In Cancer Models 
The failure of conventional brain tumor therapies stems from their inability to functionally distinguish 
neoplastic from normal cells. Despite considerable efforts, tumor specific molecular attributes that are 
suitable therapeutic targets have not been identified. Creation of "artificial" differences in biochemical 
function is an attractive option [19]. Insertion of the herpes simplex thymidine kinase (HSV-TK) gene 
into malignant cells in conjunction with the systemic administration of gancyclovir (GCV) has become a 
prototypic gene therapy system for the selective destruction of cancer cells [19], This section provides 
a general overview of the background experiments that have used HSV-TK as a "suicide gene" to 
selectively target cancer cells in experimental systems. 
Many investigators have shown that the expression of the HSV-TK gene confers a negative selectable 
phenotype to cancer cells in vitro. Moolton [19] demonstrated acquired GCV sensitivity in a murine 
sarcoma cell line transduced with a retroviral vector that produces HSV-TK. The transduced sarcoma 
tumor cells were 200-1000 times more sensitive to GCV than control tumor cells. This finding has 
been reproduced in several rodent and human cancer model systems including lung cancer [20], 
mesothelioma [21, 22], hepatocellular carcinoma [23] , leukemia[24], melanoma [25] and CNS 
tumors [26-32] models. The efficacy of this approach varies significantly [24] and may be due to a 
variety of factors including promoter function, target cells studied, and efficiency of transduction. 
Retroviral vectors were employed for much of the early experiments with HSV-TK. Several 
modifications have been introduced to overcome the difficulty of producing high titer retroviruses. 
Takamiya et. al. [31] demonstrated that rat glioma cells co-infected with a retroviral TK vector and a 
wild-type (replication competent) vector were 300 fold more sensitive to the the toxic effect of GCV, 
than those cells infected with TK vector alone. The co-infection permits the continued production of TK 
bearing virus and thus infection of neighboring tumor cells not transfected by the initial inoculation. 
This process in effect creates a retrovirus packaging cell line within the tumor. Takamiya et. al. [31] 
subsequently improved upon this system by introducing a murine retrovirus packaging cell line 
directly into the tumor[30]. This approach has been studied in other CNS tumor models [29], in 
experimental hepatic metastases [23], and has become the basis of an ongoing clinical trials for the 
treatment of brain tumors. 
More recently, adenovirus vectors have been used for gene therapy of brain tumors. Chen et. al. [32] 
demonstrated regression of experimental gliomas following in vivo adenovirus-mediated gene transfer 
and GCV treatment. The tumor deposits were not completely eliminated by this treatment, however. 
Tumor cells close to the injection site were more readily transduced than those distant, as judged by 
parallel marker gene transfer experiments. Furthermore, these more distant cells escaped GCV toxicity 
because of a diminished bystander effect attributed to a paucity of gap junctions in the rodent brain 
tumor cell line employed [32]. This can be potentially over come in the clinical setting by more precise 
stereotactic treatment planning (aided by MRI and PET studies), and by multiple tumor injections. 
Recombinant DNA Research, Volume 20 
[355] 
