Successful Use of a Plant Gene in the Treatment of Cancer in Vivo

ML Corte´s1, P de Felipe1, V Martıń1, MA Hughes2 and M Izquierdo1 1Departamento de Bioquı´mica y Biologıá Molecular-Centro de Biologıá Molecular ‘Severo Ochoa’, Universidad Autońoma de Madrid, Facultad de Ciencias, Madrid, Spain; and 2Department of Biochemistry and Genetics, The Medical School, University of Newcastle upon Tyne, UK

A new strategy for cancer gene therapy has been substrate, linamarin, followed by cell death. We show that developed using a plant gene which encodes the enzyme, the system can eradicate very large intracerebral gliomas linamarase, that hydrolyzes the cyanogenic glucoside sub- in vivo helped by a cyanide bystander effect. Animals strate, linamarin, into glucose, acetone and cyanide. Retro- showing a total regression of the tumor by magnetic reson- viral vectors that carry linamarase as a potential killer–suic- ance imaging (MRI), do not show other appreciable toxic ide gene cause a marked sensitization to the innocuous effects.

Keywords: gene therapy; linamarase/linamarin; retroviral vectors; rat glioblastoma

Introduction (Figure 1a).8,9 Acetone cyanohydrin is unstable at pH above 6 and spontaneously breaks down to acetone and The use of a plant gene represents a novel therapy for HCN.10 The cyanogenic glucoside, linamarin, is not cancer. A number of strategies have been developed in hydrolyzed by most mammalian tissues, therefore once the past to sensitize tumor cells to specific prodrugs. One absorbed, in the absence of linamarase, it is excreted of the most successful approaches to date is the herpes 11,12 unchanged in the urine. The LD50 dose of orally 1–3 simplex thymidine kinase (Hstk)/ganciclovir system. administered linamarin is 300 mg linamarin per kilogram HsTK (the corresponding protein products to genes writ- of body weight.11 ten in italic are in capital letters) phosphorylates nuc- The novelty of the system we are describing is the use leoside analogs (such as ganciclovir or acyclovir) that of a plant gene to release cyanide in neoplastic retrovirus- normal eukaryotic cells fail to phosphorylate. The incor- expressing cells, with the potential of causing a large poration of these analogs into the replicating DNA blocks bystander effect among neighboring cells. We hypothes- the process, thereby killing the cell. Malignant brain ize that a substantial bystander killing effect of the cyan- tumors (glioblastoma) have been cured in the rat animal ide would counteract the poor estimates (1–10%) for in model provided the volume of the tumor does not exceed vivo retroviral infection of solid tumors. The present work 3 4 100–150 mm in size. When the same treatment is demonstrates that transduction of mouse, rat and human 4–7 applied to humans, complete remission of the tumor is cells by retroviral vectors expressing the cassava linamar- not achieved, although there is in some cases a measur- ase gene, cas5,9 drastically increases sensitivity of these able size reduction probably responsible for a consider- cells to the cytotoxic effects of linamarin in vitro and in able increase in the expected survival time of these vivo. Neighboring cells can also be affected by the cyan- 4,7 patients. A more aggressive killer–suicide system ide released, thereby amplifying the killing effect. would represent a promising strategy for one of the most frequent (1 per 50 000 people per year) and deadly (mortality close to 100%) brain tumors in human beings. Results We have developed an alternative system to pre-existing ones, based upon the plant linamarase gene from cassava Construction of a retroviral vector carrying the (Manihot esculenta Crantz).8 The enzyme encoded by this linamarase gene, linamarin sensitivity of infected cells ␤ gene is a -glucosidase that hydrolyzes the innocuous and toxicity to sodium cyanide ␤ substrate linamarin (2-H0-isobutyronitrile- -d-gluco- A therapeutic retrovirus was engineered by inserting the pyranoside) to acetone cyanohydrin and glucose linamarase gene (cas5) into the Moloney murine leukemia based vector pBabepuro.13 The resulting recombinant retroviral plasmid carries the cas5 gene expressed under the control of the 5′ LTR (long terminal repeat) promoter, and the selection gene, pac (puromycin N-acetyl Correspondence: M Izquierdo, Departmento de Bioquı´mica y Biologıá transferase), under control of the SV40 promoter (Figure Molecular-Centro de Biologıá Molecular ‘Severo Ochoa’, Universidad Autońoma de Madrid, Facultad de Ciencias, Cantoblanco 28049, Mad- 1b). The estimated size of the retroviral plasmid is 6.9 kb rid, Spain and it was designated pLlinSp (plasmid, LTR, linamarase, Received 27 February 1998; accepted 5 June 1998 SV40 promoter, puromycin N-acetyl transferase). Plant gene in the treatment of cancer ML Corte´s et al 1500

Figure 1 Enzymatic action of linamarase and structure of retrovirus carriers. (a) Schematic representation of linamarin breakdown by linamarase into glucose and acetone cyanohydrin and further spontaneous decomposition of this product releasing free cyanide and acetone. (b) Recombinant retroviral plasmid pLlinSp. Linamarase expression is under the control of the left LTR (long terminal repeat); the SV40 promoter directs the puromycin N-acetyl transferase (pac) synthesis.

Amphotropic retroviral packaging cells ⌿CRIP were hybridization with cas5 conﬁrmed the results of the grown in increasing amounts of linamarin to determine enzyme assays (data not shown). the sensitivity of these cells to the compound. Values up The toxicity of sodium cyanide upon the cells was also to 2 mg/ml of linamarin did not markedly decrease cell evaluated. The glioblastoma tumor cell lines (C6, L9 or survival values. No cyanide was detected in any type of U373 MG), as well as one glioblastoma explant from a cell line tested (C6, ⌿CRIP and 3T3) even after treatment human patient, were 10 to 500 times more resistant to with very high concentrations of linamarin (5 mg/ml, the toxic effect of exogenous sodium cyanide than ⌿CRIP 48 h), showing the stability of the substrate and the cells. This ﬁnding could be related to the observation that absence of linamarase activity in these mammalian cells. malignant tumor cells (C6) have a low oxidative respir- The ecotropic murine packaging cell line ⌿CRE was ation rate and a decreased number of mitochondria, mak- transfected by the retroviral plasmid vector pLlinSp and ing them more resistant to metabolic poisons, such as supernatants from ⌿CRElin lawns, were collected to KCN,14,15 a property that can be partly reversed in vivo infect ⌿CRIP cells producing titers of 3–6 × 105 c.f.u. by the low pH conditions maintained within the tumor. (colony forming units)/ml. Some clones showed a Neoplastic cells forming a tumor are exposed to a sub- marked toxicity to the presence of linamarin, with esti- stantial and consistent decrease in extracellular pH from mated concentrations that resulted in 50% cytotoxicity 7.4 to 6.6 in the case of glioblastoma, or lower for other ± ␮ 16,17 (IC50), of 32 25 g linamarin/ml (Table 1 and Figure 2a). types of tumors resulting in higher sensitivity (50% In a similar way, rat glioblastoma cells C6 were or greater) to the inhibitory effects of sodium cyanide. infected with ⌿CRElin supernatant and about 20 clones Nutrient deprivation due to poor vascularization within

resistant to puromycin were selected. In general, the IC50 solid tumors is probably the major cause of this acidity, of the C6lin clones was 14 times higher than ⌿CRIPlin with limited oxygen diffusion causing induction of individual clones (Table 1), while comparison of best anerobic glycolysis and lactic acid acumulation. clone values gave an IC50 of 59 times higher for C6lin than In vitro, cultures are normally maintained closer to pH ⌿CRIPlin (Table 1 and Figure 2a and b). The packaging 7.4 than 6.6. Lowering the culture pH to 6.6 reduced ⌿ cell line FlyA13 (human fibrosarcoma in origin) has an slightly the IC50 of the best CRIPlin clone to linamarin ␮ ␮ ␮ intermedian IC50 value of 205 g linamarin/ml at pH 7.4 from 7.2 g/ml to 4.5 g/ml (Figure 2a). Although the (Figure 2c). Other cell lines whose IC50 were estimated only significant difference between values in this graph is ⌿ ␮ showed values closer to C6lin than to CRIPlin (Table 1). at 10 g linamarin/ml; the IC50 of C6lin cells to linamarin ␮ ␮ We can not fully explain the extraordinary low IC50 drops from 430 g/ml at pH 7.4 to 240 g/ml at pH 6.6, value estimated for ⌿CRIPlin but it could be related to a being significant for all concentrations but the highest

very high linamarin permeability of these particular cells. (Figure 2b). In a similar way, the IC50 of FLYA13 cells to Cell extracts from C6lin cells incubated with linamarin linamarin drops from 205 ␮g/ml to 77 ␮g/ml at pH 6.6. produced higher amounts of cyanide (purple colour) The lower interstitial pH in tumors compared with nor- than equivalent extracts from ⌿CRIPlin cells, suggesting mal tissues would result in a greater and speciﬁc sensi- a very efﬁcient expression of the linamarase gene in the tivity to the toxic effect of cyanide on tumor cells. This C6lin cells (Figure 3). Isolation of mRNA and Northern is very convenient because it will selectively increase the Plant gene in the treatment of cancer ML Corte´s et al 1501 Table 1 Sensitivity of different cell lines to linamarin

Cell line % of IC50 survival (␮g linamarin/ml) of cells with individual 1 mg puromycin-selected linamarin/ml clones

Best clone Mean valuea

C6 100b 430 452 ± 40 Rat glioma

U-373 MG 80 145 289 ± 145 Human glioblastoma

Hs 683 80 230 323 ± 115 Human glioma

Sk-N-MC 80 183 333 ± 125 Human neuroblastoma

FLYA13 92b 205 385 ± 126 Human amphotropic packaging cells

⌿CRIP 85 7.2 32 ± 25 Murine amphotropic packaging cells

NIH 3T3 90b ND ND Murine ﬁbroblasts

Ag 9181 89b ND ND Human ﬁbroblasts

ND, not determined. aMean value ± s.d. determined with at least four different clones. bIn these cell lines the concentration of linamarin was increased to 5 mg/ml without a substantial drop in survival values. In the other cell lines the effect of 5 mg linamarin/ml was not determined.

specificity of the therapeutic action towards inner tumor cells rather than normal surrounding cells. Figure 2 Sensitivity to linamarin as a percentage of cell survival. The The cyanide bystander effect initial number of cells was 105 in all cases. Following 24 h incubation The type of bystander effect expected in the present sys- increasing linamarin concentrations were added and 2 days later, the per- tem is rather different from the more commonly centage of surviving cells was determined to evaluate drug toxicity. Clonal lines were assayed for vector titer, using NIH 3T3 cells. Best titers were described killing of nearby unmodified cells which is 3–6 × 105 c.f.u./ml. (a) ⌿CRIP sensitivity to linamarin compared with dependent on connexin expression and cell communi- one clone of retroviral LlinSp infected ⌿CRIP, at increasing linamarin 18,19 − ␮ cation via gap junctions. The cyanide ion (CN ) can concentrations and two pH values. The estimated IC50 drops from 7.2 g diffuse freely and HCN is a gas, therefore toxicity to the linamarin/ml at pH 7.4 to 4.5 ␮g linamarin/ml at pH 6.6. (b) C6 sensi- neighbors of a linamarase-containing cell is due to cyan- tivity to linamarin compared with one clone of retroviral LlinSp infected C6, at increasing linamarin concentrations and two pH values. The esti- ide release, and cyanide action does not require cell-to- ␮ mated IC50 for this C6lin clone is 430 g linamarin/ml at pH 7.4 and cell contact or gap junctions. 240 ␮g linamarin/ml at pH 6.6. (c) FLYA13 sensitivity to linamarin com- In order to investigate the bystander effect in this sys- pared with one clone of retroviral LlinSp infected FLYA13 at increasing tem, we performed some experiments in which linamar- linamarin concentrations and two pH values. The estimated IC50 for this ase retroviral producer cells were mixed in a 50:50 initial FLYA13lin clone is 205 ␮g linamarin/ml at pH 7.4 and 77 ␮g ± cell ratio with non-producer cells. One of the two popu- linamarin/ml at pH 6.6. Values are means s.d. Asterisk indicates a sig- lation cells was labeled in red in order to follow its fate. nificant difference between pH 7.4 and 6.6 at that linamarin concentration point (P Ͻ 0.05). Cells were seeded in flat-sided tubes provided with screw caps to prevent HCN release. One of the mixtures was the packaging cell line ⌿CRIP infected with the pLlinSp retrovirus (⌿CRIPlin) plus normal C6 labeled red with DiI (1,1′-dioctedecyl-3,3,3′,3′-tetramethyl-indocar- Plant gene in the treatment of cancer ML Corte´s et al 1502

Figure 3 Cyanide presence in homogenates is proportional to cristal violet produced (purple color) as estimated by the Lambert method.31 (a) ⌿CRI- Plin; (b) C6; (c) C6lin; (d) normal rat brain; (e) normal human brain; (f) human glioblastoma. Samples a to c are cell homogenates and d to e, tissue homogenates.

bocyanine perchlorate).20 In the presence of 2 mg/ml of linamarin, both types of cells (⌿CRIPlin and C6-labeled) died; while 86% of C6 alone survived at linamarin concentrations of 10 mg/ml. During the 48-h experimental time, some C6 cells are expected to be infected by the retrovirus rather than die as a consequence of the bystander effect. Mixtures of ⌿CRIPlin with either ⌿CRIP or 3T3 labeled in red (⌿CRIPlin/⌿CRIP red or ⌿CRIPlin/3T3 red) were devastated at a low concentration of linamarin (100 ␮g/ml) due to the greater sensitivity of ⌿CRIP or 3T3 cells to cyanide released by producer cells. In the mixtures ⌿CRIPlin/⌿CRIP red, no infections can take place and all the observed effect is due to bystander mediated by cyanide. These results Figure 4 Gadolinium enhanced magnetic resonance views in sagittal (a1,b1) and axial (a2,b2) projections of a rat brain, before (a1,a2) and after indicate that the system can be very efﬁcient in neighbor- 3 linamarin treatment (b1,b2). A brain tumor (70 mm volume size, a1 a2 ing cells compensating for the poor infection estimates arrow) of glioblastoma-C6lin transfected cells, has been eradicated com- for retroviral vectors in solid tumors.7 pletely (b1 b2 arrow) after 1 week treatment with constant local release of 12.5 ␮g of linamarin per hour. Curability of rat glioblastoma in vivo Based on the encouraging in vitro data, we proceeded to

evaluate the sensitivity of the system in vivo using an toma with internal hemosiderin (Figure 4b1,b2). The ani- intracerebral tumor model. Two types of experiments mal was killed 7 months later and the brain fixed and were performed: in the first type, C6 cells, already trans- processed for microscopic analysis, no tumor cells were fected with pLlinSp and selected for puromycin resistance found, only a chronic periventricular gliosis. One of the (C6lin), were inoculated into the animal brain and animals having a similar size tumor, was treated with allowed to form a tumor. The glioblastoma would be sub- only a total of 1 mg of linamarin, this animal finished the sequently treated with the prodrug, linamarin. In the treatment but finally died as a consequence of the tumor, second type of experiment, brain tumors were induced indicating a lower dose limit dosage for the prodrug. by non-transfected C6 cells and subsequently ⌿CRIPlin Local, rather than intraperitoneal, administration of the retrovirus-producing cells were inoculated into the grow- prodrug was chosen, because we have estimated and ing tumor, before treatment with linamarin. experimentally confirmed in other systems (tk/ganci- A total of 106 C6lin transfected cells were inoculated clovir or tk/5–2-bromovinyl-2′-deoxyuridine), that the (in the first type of experiment), into the white matter of amount of prodrug needed to eradicate 1 mm3 of tumor the right cerebral hemisphere (4 mm to the right of the is about 1000 times lower if the prodrug is administered bregma and 4.5 mm deep from the skull) of three Wistar locally rather than intraperitoneally (Table 2), thus rats. A few weeks later (2 to 4), an intracranial tumor was decreasing substantially the cost of the treatment. No clearly visible by magnetic resonance imaging (around 70 3 mm volume size) (Figure 4a1 and a2). A day later, linamarin was administered by a direct brain tumor infusion Table 2 Prodrug dose in micrograms needed to eradicate 1 mm3 cannula connected by a catheter tube to a mini-osmotic of experimental rat tumor with the system Herpes simplex thy- pump. A total of 2.4 mg of linamarin was delivered midine kinase/ganciclovir (n = 3) throughout the 8-day period. Release was at a constant rate of 1 ␮l per h (12.5 ␮g of linamarin/␮l). The magnetic Prodrug Intraperitoneal delivery Intracranial delivery resonance image taken 3 weeks after the treatment showed a total absence of the tumoral mass and in its Ganciclovir 3032 ± 2228 20 ± 11 place a residual hypointense image at the site of injection, interpreted as a fibrous scar or as the remains of hema- Plant gene in the treatment of cancer ML Corte´s et al 1503 apparent toxicity was observed by local administration increase the potential access of retrovirus to all neoplastic of up to 9 mg of linamarin into normal rat brains using cells. A few days later (3 for the largest tumor and 5 for the same conditions as in the experimental animals. One the others), 6 mg of linamarin in 200 ␮l were adminis- of the animals treated with 6 mg of linamarin was killed tered at a constant rate of 1 ␮l per h (osmotic pump and the brain fixed and processed for microscopic analy- delivery). A total eradication of the tumor was achieved sis. No lesion could be detected. for the largest (Figure 5b1,b2 and b3) and the smallest In the second type of experiment, a total of 106 C6 cells glioblastomas; the medium-sized tumor animal, on the were inoculated into the right hemisphere of 10 Wistar other hand, also finished the treatment but died a month rats and allowed to develop into tumors. The three ani- later. Examination of the dead animal brain showed a mals with the largest tumors, approximate volumes of hemorrhage within the tumor remains, being probably 3 3 3 400 mm , 600 mm and 1000 mm , (Figure 5a1,a2 and a3), the cause of death. were selected for further inoculation with ⌿CRIPlin Control animals (three) with tumor volume sizes (200 retrovirus-producing cells (7 × 106 cells for the two small- mm3, 400 mm3 and 250 mm3), treated only with the lina- est and 107 cells for the largest) and linamarin local treat- marin (6 mg) prodrug (no linamarase-producing ment. The murine retroviral-producer cells (⌿CRIPlin) retrovirus) in the same conditions as described earlier, had a titer of 5 × 105 c.f.u./ml and were inoculated in a died: 3, 0 and 2 days before finishing the treatment. single place (the same used for C6 delivery) in all but the The linamarase/prodrug treated animals showed no biggest size tumor where an additional small craniotomy apparent neurological damage for a period up to 6 was made at a distal site within the tumor, in order to months. We have observed recurrence in very large tumors (over 500 mm3 in size) about 6 months after finishing the treatment in approximately 50% of the animals. In all recurrent cases the tumoral mass had crossed the interhemispheric fissure and reached the ventricles. At this point tumoral cells could be distributed via cer- ebrospinal fluid to parts of the brain quite distant from the original inoculum. Indeed some of the recurrent tumors were multiple and very far from the original site. Toxicity to normal brain tissue was estimated in three animals by direct inoculation of ⌿CRIPlin producer cells, followed by linamarin treatment (2.5, 4 and 6 mg) in the absence of tumor cells. The magnetic resonance of the treated animals (Figure 6) does not show any large damage but a small scar similar to the one normally present in brains after the tumor has been eradicated. Two types

of relaxation time (in Tesla, T) images were analyzed: T1 and T2 in order to visualize lesions other than tumoral. In general, T1-weighted images primarily provide anatomic information, T2-weighted images are more sensitive to early pathologic changes in most tissues. We have not seen any lesion related to neoplasm, edema, ischemia, infection, demyelination or trauma; animals only exhib- ited small MRI abnormalities and no gross lesions. We interpret these results as a proof that the treatment itself does not produce appreciable damage to the brain of the animals.

Discussion The use of a retrovirus in brain cancer gene therapy has the advantage of targeting only dividing cells in a quiesc- ent background of neurons. The transduction rate of tumor cells by retroviral vectors in vivo is nevertheless very low, being usually in the range of 1–10% of the tumor cell population. An efficient bystander effect is hypothesized to explain the total remission of tumors in animal models by the Hstk/ganciclovir system, and transfer of phosphorylated ganciclovir between cells, via Figure 5 Gadolinium enhanced magnetic resonance views in axial (a1,b1), sagittal (a2,b2) and coronal (a3,b3) projections of a rat brain, before (a1, gap junctions, appears as one of the best documented ⌿ 18 a2 a3) and after mouse retroviral producer cell, CRIPlin, inoculation and theories to explain the mechanism. Some tumor cells, linamarin treatment (b1,b2,b3). A very large brain tumor (approximately like rat glioblastoma C6 for example, have lost the 3 1000 mm volume size, a1,a2,a3 arrows) of rat glioblastoma C6 cells, has capacity to synthesize some of the connexin proteins and been eradicated completely as shown in the three magnetic resonances (b1, therefore the benefit from this important effect will be b2 and b3 arrows). These images show the appearance of the living brain after murine therapeutic retroviral producer cell (⌿CRIPlin) inoculation severely limited. The linamarase/linamarin enzyme/ into the tumor and 1 week linamarin treatment (with constant local release prodrug combination does not require cell-to-cell contact of 30 ␮g/h) started 3 days after the inoculation. for its bystander effect, since evidence of cyanide in the Plant gene in the treatment of cancer ML Corte´s et al 1504 glycosides, such as linamarin, are not hydrolyzed by cultured mammalian cells: rat brain tumor or normal tissue (Figure 3), and human brain tumor or normal tissue (Figure 3). Nevertheless gut microbial flora would be able to catabolize some cyanogenic glycosides, such as amygdalin or linamarin, to a certain extent by bacterial ␤-glucosidases in the intestinal lumen. This would result in the release and absorption of hydrogen cyanide. Female hamsters given an oral dose of 108 mg of linamarin/kg body weight show signs of cyanide poisoning causing death in 18% of the animals.11 The toxicity is much lower in animals receiving similar doses intravenously and is practically absent when administered intraperitoneally.22 Local administration of linamarin is as safe or probably safer than intraperitoneal delivery since it avoids gastrointestinal interference and it is clear from the in vitro and in vivo tests reported here that we are well below dangerous concentrations of the prodrug. One alternative method for the oral administration of cyanogenic gluco- sides is to pretreat the animals (or patients) with especially extended-spectrum antibiotics to reduce gastrointestinal flora and therefore make them less sus- ceptible to cyanogenic glucoside toxicity.23 In many tropical communities a potential health haz- ard is recognized by the consumption of leaves and roots of the cassava plant. When mechanical damage of the tissues of this plant occurs, the cyanogenic glucoside (linamarin) is exposed to the action of the endogenous plant linamarase and ␣-hydroxynitrile lyase and hydro-

Figure 6 Axial T2 (a1,b1) and sagittal T1 (a2,b2) magnetic resonance cyanic acid is released. Cooking/boiling, drying, soaking views of two control animals treated in the right hemisphere with 5 × 106 and fermentation are the usual processing techniques to ⌿ mouse non-malignant CRIPlin-retroviral producer cells and then with inactivate linamarase or breakdown linamarin before 4 (a) and 6 mg of linamarin (b). The left hemisphere image can be con- consumption. The occasional poisoning reported in sidered as a control of untreated brain tissue. humans is generally due to inadequately processed leaves or roots.24 Cyanide doses of 0.5 to 3.5 mg/kg body cell culture medium indicates that it can cross cell mem- weight are predicted to be lethal to adults.25 This would branes. Furthermore, the bystander effect expected in vivo be equivalent to 3 mg/kg of sodium or potassium cyan- should be greater than we estimate in vitro since the vol- ide and to 500 to 1000 mg of linamarin. The dose of lina- atile character of cyanide makes it rather difficult to esti- marin potentially needed to eradicate a human tumor of mate the killing effect for surrounding cells in vitro about 4 cm diameter, is below dangerous concentrations because the compound dissipates into the medium and of the prodrug and would be administered over the per- air faster than to neighboring cells. On the other hand, in iod of a week. vivo within a tightly packaged tumor, the cells nearest to The second concern is related to the damage that cyan- the producers will be the first and main target of the ide release could cause in healthy parts of the brain. cyanide released. Cyanide released within a tumor would Cyanide can rapidly enter the plasmatic membrane and also be more effective against neighboring cells by the slowly accumulate proportional to the concentration, in increased responsive effect of the low extracellular pH.16 mitochondria and membrane elements of the neuronal Cells outside the tumor, having extracellular pH values cell.26 It is an advantage of our system that the local extra- close to 7.4, would be rather resistant to the cyanide cellular pH created within tumors is lower and the sus- bystander effect. ceptibility to cyanide is increased in such conditions. On One of the main concerns in the present study is the the other hand, infiltrating glioblastoma cells causing fre- putative toxicity of the system. Two aspects are parti- quent recurrence of the tumor, will probably not benefit cularly delicate and should be further analyzed: one from this pH effect. related to the possibility of linamarin breakdown in other In one of the in vivo experiments we have performed, parts of the body by alternative enzymes or situations; a large majority (if not all) of the tumoral cells had incor- and the other, to the putative damage of healthy brain porated the linamarase gene and therefore could process tissue surrounding the tumor. linamarin to glucose, acetone and cyanide (C6 transfected With reference to the first issue, there is one ␤-gluco- with pLlinSp). Nevertheless magnetic resonance images sidase in mammals of broad specificity and obscure show a comparable fibrous scar to the one obtained pre- physiological function able to hydrolyze toxic plant gly- viously with the Hstk/ganciclovir system3,4 or to the C6 cosides found in the diet of man. In general the richest infected in vivo by the retrovirus LlinSp, indicating source of this cytosolic ␤-glucosidase in most vertebrates absence of a devastating effect in the brain. The control appears to be the liver and the small intestine, except in animals treated with producer cells and linamarin in the the mouse and rat, where the kidney is the richest source absence of tumor (Figure 6) also provide evidence against of the enzyme.21,22 We have confirmed that cyanogenic brain damage with the full treatment. All of the surviving Plant gene in the treatment of cancer ML Corte´s et al 1505 animals are healthy and well, over 6 months after the A plausible model for the tumor is given by the volume treatment. We have seen tumor recurrence about 6 generated placing sections homothetic to C0 at each pos- months after finishing the treatment, in half of the tumors ition x with a ratio S(x)/d0. 3 2 larger than 500 mm volume size, all of which had orig- Therefore the area ratio is [S(x)/d0] inally crossed the interhemispheric fissure and gained C L N L 2 access to the lateral ventricles. No recurrence was V Х 0 ͸ ͩSͩj ͪͪ observed in smaller tumors confined to one brain hemi- d2 N N 0 j=1 sphere. Glioblastoma is a particularly nasty tumor because of where = the infiltrating characteristic of its cellular constituents: C0 area of the given coronal section = recurrence tumors are generally not further than 2 cm d0 diameter of that coronal section in the z (sagittal) from the primary site but it may be difficult to eliminate direction = infiltrated neoplastic cells without damaging some sur- S(x) length of the sagittal section in the z direction at rounding brain cells. We have been unable in the past, position x to cure rat glioblastomas larger than 150 mm3 in volume N will be chosen taking into account the shape of the using the Hstk/ganciclovir system and this indicates that sagittal picture the new linamarase/linamarin approach is more efficient. It is possible to obtain similar formulae replacing sagittal When the Hstk/ganciclovir system was used in a small 4,7,27 sections by axial. However, taking into consideration the number of patients, a genuine tumor size reduction standard geometries of these tumors, the axial section in about 30% of the treated patients was generally seems to be less significant than the sagittal one. In gen- acknowledged, but no total disappearance of the residual eral, we expect to have an estimate from below using the tumor has been achieved despite several consecutive axial-coronal approach. treatments. Survival time was increased in many cases, but recurrence finally killed the patient. The method we Construction of lin-recombinant plasmid describe here could certainly be more aggressive than The plasmid pcas 59 was first cut with the restriction previous ones and the animal results can be considered enzyme SacII and filled in with Klenow, it was sub- highly encouraging. sequently cut with SalI to generate a blunt-ended SalI lin- To our knowledge this is the first successful use of a amarase coding fragment. This fragment was inserted in plant gene in the treatment of cancer in vivo. The possi- pBabe Puro,13 cut with SnaBI plus SalI restriction bility of applying this procedure to humans remains open enzymes (SnaBI gives a blunt end, so the insert was ori- and promising. ented by the Sal enzyme). The recombinant is called pLlinSp and can be selected by its resistance to puromycin Materials and methods (1–2.5 ␮g/ml puromycin) under control of the promoter of SV40; linamarase is under control of the LTR promoter. Volume estimation The prokaryotic origin of replication allows the recom- If [0,L] are the parameters of the tumor interval in the binant plasmid to grow in bacteria (E. coli) before trans- р р sagittal section and C (x), 0 x L denotes the area of duction of mammalian cells. the corresponding coronal section, then we have the exact formula for the volume of the tumor: Cell lines and culture L = V ͵ C(x)dx, Cells: C6 and L9 (rat glioblastoma), U373MG (human o glioblastoma), Hs683 (human glioma), NIH-3T3 (murine which yields the approximation fibroblasts), Ag9181 (human fibroblasts), ⌿CRE ⌿ L N L (ecotropic packaging line), CRIP (anphotropic packag- V = ͸ Cͩj ͪ. ing line) and FlYA13 (amphotropic packaging line) were app N N j=1 grown in Dulbecco’s modified Eagle’s medium (DMEM) In practice, however, we are forced to estimate the vol- supplemented with 10% heat inactivated fetal calf serum ume with the knowledge of only one section of each type. (C6, L9, U373MG, Hs683, Ag9181, FLYA13) and calf ° It is not a well-constructed problem. Nevertheless, these serum (the rest). Cells were cultured at 37 C, 7% CO2 and tumors are usually adapted to the rat’s head geometry: 97% relative humidity. they tend to be rather symmetric (elliptical) in the coronal The pH of DMEM was artificially lowered to pH 6.6 sections, while they have long and irregular sagittal in some experiments by decreasing the NaHCO3 concen- shapes. tration from 3.7 g/l to 0.7 g/l. Given two ‘well-taken’ sections, let us introduce coordinates: Transfections and infections The retroviral plasmid pLlinSp was introduced to the ecotropic ⌿CRE packaging cells using the adenovirus enhanced transferrinfection technique (AVET) for DNA entrance.28–30 Viral supernatants from lawns, collected 48 h after transfection, were used to infect amphotropic ⌿CRIP packaging cells (⌿CRIPlin). After selection with puromycin (1.5 ␮g/ml) colonies were isolated using clon- ing rings. Clonal lines were assayed for vector titer, using NIH 3T3 cells.6 Best titers were 3–6 × 105 c.f.u./ml. Plant gene in the treatment of cancer ML Corte´s et al 1506 Cell lines showed in Table 1 were infected with ⌿CRI- Wistar rat brain inoculation of C6lin transfected cells Plin supernatants (U373MG, Hs683, Ag9181, Sk-N-MC and linamarin administration and FLYA13) or ⌿CRElin supernatants (C6 and 3T3), to We have essentially followed procedures described pre- obtain linamarase producing (lin+) cell lines. After selec- viously.3–4 Wistar male rats weighing 250–300 g were tion with puromycin, colonies were isolated and linama- anesthetized with a mixture of ketamine (50 mg/ml), val- rin sensitivity was estimated. ium (5 mg/ml) and atropine (1mg/ml) in a 5:4:1 ratio by In Figure 4, the inoculated C6 cells were transfected volume at a dose of 0.3 ml/100 g of body weight, before with the retroviral plasmid pLlinSp using the AVET tech- placing them in a stereotactic apparatus. C6 glioma cells nique for DNA entrance.28–30 or the same cells transfected with the pLlinSp plasmid and selected by puromycin resistance were injected at a Labeling of cells concentration of 105 cells/␮l in complete PBS (with cal- Cells were labeled essentially as described previously.20 cium and magnesium) supplemented with 0.1% glucose. Two milligrams of DiI (1,1′-dioctadecyl-3,3,3′,3′-tetrame- With the aid of the manipulating arm of the stereotactic thyl-indocarbocyanine perchlorate) were dissolved in apparatus, a total of 10 ␮l was introduced, over a 5-min 1 ml of ethanol, and 50 ␮l of the solution was added to interval, into the fronto-parietal lobe of the right cerebral 5 ml of cells in a 6-cm culture dish. After 1-h incubation hemisphere (4 mm to the right from the bregma and 4.5 at 37°C, the cells were trypsinized, washed twice with mm deep from the skull) using a 10 ␮l Hamilton syringe PBS, and resuspended in DMEM at a concentration of connected to a 26-gauge needle; the needle was kept in 5 × 104 cells/␮l. place 3 min before and after injection. ⌿CRIPlin producer cells were inoculated at the same stereotactic coordinates Linamarin sensitivity and cyanide toxicity assays in as the C6 cells except in a very large tumor in which an culture additional tumor site was also used, cells were delivered Cells were seeded in a flat-sided tube (provided with a at different heights along the tumor. All rats received screw cap to prevent HCN release), at a density of 105 tetracycline in the drinking water (approximately 75 cells per tube. Following 24 h incubation, increasing lina- mg/kg) and dexamethasone (1 mg/500 ml) for 3 days marin concentrations were added. Two days later, the before and 1 week after surgery. The linamarin treatment percentage of surviving cells was determined to evaluate was started after magnetic resonance visualization of the drug toxicity. Cells were incubated for 2.5 h with the tumor and size estimation. The prodrug was adminis- mitochondrial substrate MTT (3-[4,5-dimethylthiazo-2- tered by a brain infusion cannula placed at the site of yl]-2,5-diphenyltetrazolium bromide, 200 ␮g/ml) which tumor inoculation and fixed to the skull using dental can be transformed to formazan only by living cells. cement; a small stainless steel screw acts as anchor to The medium was removed and 3 ml of DMSO secure the cannula and the dental cement. The brain (dimethylsulfoxide) were added to dissolve the forma- infusion cannula was then connected subcutaneously zan. Samples were measured spectrophotometrically at (s.c.) to an osmotic pump (Alzet; model 2001; Alza Phar- 540 nm after 10 min. maceuticals, Palo Alto, CA, USA) which delivers 1 ␮l/h Cyanide toxicity was assayed as described above for 7 or 8 days. The osmotic pump was filled with the except by the addition of increasing amounts of NaCN appropriate amount of linamarin in PBS. to the cultures. Magnetic resonance imaging Measurements of linamarase activity and cyanide In this study axial, sagittal and coronal views of the skull presence were made with a Cpflex small coil around the animal. Linamarase activity and cyanide presence is determined T1 sagittal and axial projection images were obtained by the Lambert method31 with the following modifi- using spin eco, with TR 650, TE 17.0/1 and acquisitions. cations: 5 ml of the succinimide/N-chlorosuccinimide Slice thickness was 3 mm; field of view 50 × 100 and high reagent was added to the flat tubes, containing the cells resolution matrix of 256 × 512. Sequence post-enhance- with differents amounts of linamarin (48 h later), cul- ment was achieved with gadolinium (0.8 ml/kg) and tured in 1 ml of DMEM without phenol red. Then 1 ml magnetization transfer (MT). of the barbituric acid/pyridine reagent was added, mixed and the absorbtion at 592 nm was read after 10 min Acknowledgements (calibration curves were made with NaCN). We gratefully acknowledge the special collaboration of Tissue homogenates the neuroradiologist Arantxa Royo with the magnetic res- The tissues were minced and homogenized in five vol- onance imaging; the mathematician Antonio Co´rdoba for umes of a medium containing sucrose (0.25 m), ethylene- the volume estimation, the anatomopathologist Javier diamine tetraacetate (1 mm) and Tris HCl (0.01 m, pH 7.2) Figols for microscopic analysis of brain tissue sections buffer supplemented with the protease inhibitor phenyl- and to Lorenzo Seguido for other technical aspects of this methylsulfonyl fluoride (PMSF) (1 mm) all procedures work. We thank Antonia Ruiz for expert technical assist- were carried out at 1–4°C.32 The homogenate was centri- ance. We are also indebted to Jean-Michael Heard for the fuged (600 g, 10 min) and the supernatant fraction was packaging cell lines ⌿CRIP and ⌿CRE and to Hartmut stored at −70°C. The protein content was measured by Land for the plasmid pBabepuro. The FLYA13 cell line BCA protein assay. was kindly provided by Yasuhiro Takeuchi. We are very The experiment in Figure 3 was carried out at 37°C, grateful to all the members of the animal house, 1mm linamarin, 0.2 mg protein, pH 7.4, in a total volume especially to Javier Palacıń Urquijo for continuous advice. of 1 ml. Cyanide presence was determined by the Lam- The work was supported by the Plan Nacional de Salud bert method31 after 4 h incubation. (PNICyD, grant number 96–0037) from the Spanish Plant gene in the treatment of cancer ML Corte´s et al 1507 government and by the local government: comunidad de ular synthesis and tumor cell division. Biochem Pharmacol 1988; Madrid (Acciones coordinadas, grant number 57–1996). 37: 2259–2266. The Centro de Biologıá Molecular is the recipient of an 17 Gerweck LE, Seetharaman K. Cellular pH gradient in tumor ver- institutional grant from the Fundacioń Ramoń Areces. sus normal tissue: potential exploitation for the treatment of cancer. Cancer Res 1996; 56: 1194–1198. 18 Wygoda MR et al. Protection of herpes simplex virus thymidine kinase-transduced cells from ganciclovir-mediated cytotoxicity by bystander cells: the good samaritan effect. Cancer Res 1997; References 57: 1699–1703. 19 Vrionis FD et al. The bystander effect exerted by tumor cells 1 Culver KW et al. In vivo gene transfer with retroviral vector- expressing the herpes simplex virus thymidine kinase (HSVtk) producer cells for treatment of experimental brain tumors. gene is dependent on connexin expression and cell communi- Science 1992; 256: 1550–1552. cation via gap junctions. Gene Therapy 1997; 4: 577–585. 2 Ram Z et al. In situ retrovirus-mediated gene transfer for the 20 Tamura M et al. Targeted killing of migrating glioma cells by treatment of brain tumors in rats. Cancer Res 1993; 53: 83–88. injection of HTK-modified glioma cells. Hum Gene Ther 1997; 8: 3 Izquierdo M et al. Long-term rat survival after malignant brain 381–391. tumor regression by retroviral gene therapy. Gene Therapy 1995; 21 Glew RH, Gopalan V, Forsyth GW, VanderJagt DJ. The mam- 2: 66–69. malian cytosolic broad specificity ␤-glucosidase. In: Zsen A (ed). 4 Izquierdo M et al. Implications of the size of glioblastoma on its ␤-Glucosidases: Biochemistry and Molecular Biology. Am Chem Soc curability. Acta Neurochirurgica 1997; 68 (Suppl.): 111–117. Symp series 533: Washington, 1993, pp 83–112. 5 Culver K et al. Gene therapy for the treatment of malignant brain 22 Lima Go´rniak S, Haraguchi M, de Souza Spinosa H, Nobre D. tumors with in vivo tumor transduction with herpes simplex Experimental intoxication in rats from a HCN-free extract of thymidine kinase gene/ganciclovir system. Hum Gene Ther 1994; Holocalyx glaziovii Taub: probable participation of the cyano- 5: 343–379. genic glycoside. J Ethnopharmacol 1993; 38: 85–88. 6 Izquierdo M et al. Human malignant brain tumor response to 23 Newton GW et al. Amygdalin toxicity studies in rats predict herpes simplex thimidine kinase (HSVtk)/ganciclovir gene ther- chronic cyanide poisoning in humans. West J Med 1981; 134: apy. Gene Therapy 1996; 3: 491–495. 97–103. 7 Ram Z et al. Therapy of malignant brain tumors by intratumoral 24 Tylleska¨rTet al. Epidemiological evidence from Zaire for a diet- implantation of retroviral vector-producing cells. Nature Med ary etiology of konzo, an upper motor neuron disease. Bull W 1997; 3: 1354–1361. HO1991; 69: 581–589. 8 Mkpong OE, Yan H, Chism G, Sayre R. Purification, characteriz- 25 Solomonson LP. Cyanide as a metabolic inhibitor. In: Vennes- ation and localization of linamarase in cassava. Plant Physiol land B et al (ed). Cyanide in Biology. Academic Press: London, 1990; 93: 176–181. 1981, pp 12–26. 9 Hughes MA et al. A molecular and biochemical analysis of the 26 Borowitz JL et al. Accumulation of labeled cyanide in neuronal ␤ structure of the cyanogenic -glucosidase (linamarase) from cas- tissue. Toxicol Appl Pharmacol 1994; 129: 80–85. sava (Manihot esculenta Cranz). Arch Biochem Biophys 1992; 295: 27 Roth JA, Cristiano RJ. Gene therapy for cancer: what have we 273–279. done and where are we going? J Natl Cancer Inst 1997; 89: 21–39. 10 Selmar D, Carvalho FJP, Conn EE. A colorimetric assay for ␣- 28 Cotten M et al. High-efficiency receptor-mediated delivery of hydroxynitrile lyase. Analyt Biochem 1987; 166: 208–211. small and large (48 kilobase gene constructs using the endo- 11 Frakes RA, Sharma RP. Comparative metabolism of linamarin some-disruption activity of defective or chemically inactivated and amygdalin in hamsters. Fd Chem Toxic 1986; 24: 417–420. adenovirus particles. Proc Natl Acad Sci USA 1992; 89: 6094–6098. 12 Hernandez T et al. Fat in humans of dietary intake of cyanogenic 29 Wagner E et al. Coupling of adenovirus to transferrin- glycosides from roots of sweet cassava consumed in Cuba. Natu- polylysine/DNA complexes greatly enhances receptor- ral Toxins 1995; 3: 114–117. mediated gene delivery and expression of transfected genes. 13 Morgenstern JP, Land H. Advanced mammalian gene transfer: Proc Natl Acad Sci USA 1992; 89: 6099–6103. high titre retroviral vectors with multiple drug selection mark- 30 von Ru¨ den T et al. Generation of high-titer retroviral vectors ers and a complementary helper-free packaging cell line. Nucleic following receptor-mediated, adenovirus-augmented transfec- Acids Res 1990; 18: 3587–3596. tion. BioTechniques 1995; 18: 484–489. 14 Lichtor T, Dohrmann GJ. Respiratory patterns in human brain 31 Lambert JL, Ramasamy J, Paukstelis JV. Stable reagents for the tumors. Neurosurgery 1986; 19: 896–899. colorimetric determination of cyanide by modified Ko¨nig reac- 15 Maduh EU, Borowitz JL, Isom GE. Cyanide-induced alteration tions. Analyt Chem 1975; 47: 916–918. of the adenylate energy pool in rat neurosecretory cell line. J 32 Glew RH, Peters SP, Christopher AR. Isolation and characteriz- Appl Toxicol 1991; 11: 97–101. ation of ␤-glucosidase from the cytosol of rat kidney cortex. 16 Hu JJ et al. pH-related effects of sodium cyanate on macromolec- Biochem Biophys Acta 1976; 422: 179–199.