Regional Delivery and Selective Expression of a High-Activity Yeast Cytosine Deaminase in an Intrahepatic Colon Cancer Model¹

Despite the development of aggressive systemic and regional approaches, colorectal cancer metastatic to the liver continues to kill approximately 15,000 patients a year in the United States (1). One approach toward improving this outcome is enzyme/prodrug gene therapy. This strategy involves introducing a gene that produces an enzyme capable of converting a nontoxic prodrug into a chemotherapeutic and/or radiosensitizing agent within the tumor. The most common approach has used viral vectors. A prototype of such systems is the gene for herpes simplex virus, thymidine kinase, introduced in conjunction with the administration of an antiviral drug, ganciclovir. In this strategy, ganciclovir is preferentially phosphorylated by herpes simplex virus thymidine kinase and subsequently by cellular kinases to ganciclovir-triphosphate, which disrupts DNA synthesis (2). Another enzyme-prodrug strategy uses CD⁴ (usually from bacteria) to convert the benign antifungal drug 5FC into the chemotherapeutic and radiosensitizing agent 5FU. Unlike the phosphorylated forms of ganciclovir, which depend at least partially on gap junctional communication (often lacking in cancer cells) to move from the transduced cell to a bystander cell, 5FU can diffuse through cell membranes to achieve significant bystander effect independent of cell contact (3, 4). In addition, 5FU is a radiosensitizer at approximately one-tenth of the concentration required to produce direct cytotoxicity (5).

Our previous studies indicate that yCD, a yeast-derived CD, improves 5FC/5FU conversion compared with bCD (6). Tumor cells expressing yCD are significantly more sensitive to 5FC administration with or without radiation, producing a significant delay in tumor growth (or cure) in a variety of animal models, including intrahepatic colon cancer xenografts and head and neck cancers (6, 7, 8, 9). Bystander effects mediated by 5FC/5FU conversion were also demonstrated to enhance the therapeutic effects in vitro and in vivo (7, 8). We have developed an adenoviral vector containing an RSV-driven yCD gene that has demonstrated promising tumor-growth inhibition (6, 7, 8, 9).

An important potential limitation to improving the outcome of gene therapy concerns the nonselective expression of the yCD gene causing 5FC to 5FU conversion in normal tissues. The strategy of placing genes under the control of tumor-associated regulators has been used to achieve tumor-specific gene expression. CEA is an important marker for tumors of epithelial origin, including cancer of the colon, stomach, breast, and lung (10). Utilization of the CEA promoter in an adenovirus vector for tumor-specific bCD gene expression has been shown to improve the selectivity of 5FC/5FU conversion in CEA-expressing tumors (11, 12). However, the low efficiency of the basal CEA promoter has decreased the enthusiasm for this approach, particularly when combined with the relatively low activity of bacterial CD enzyme in converting 5FC to 5FU.

Our previous study indicated that an enhanced CEA promoter could initiate selective expression of yCD in various colon carcinoma cells (13). In the present study, we developed an adenoviral vector carrying a much more efficient CD enzyme from yeast, driven by an enhanced CEA promoter, CEA-yCD, and tested this vector for yCD gene expression in CEA-expressing human colon carcinoma cells and non-CEA-expressing normal cells. The selectivity of CEA-yCD gene expression was further studied in an intrahepatic model of nude rats bearing colon cancer xenografts with regional intravascular viral administration.

Human colon carcinoma cell lines (LoVo, HT29, and CaCo2) and normal human fibroblasts were obtained from American Type Culture Collection (Manassas, VA). The rat liver epithelium cell line (WB) was a kind gift from Dr. James Trosko (Michigan State University, East Lansing, MI). Cells were maintained in RPMI media containing 10% FBS and antibiotics at 37°C with 5% CO₂.

Recombinant adenoviral vectors of RSV-yCD or CEA-yCD were constructed as described previously (8). An adenoviral vector of CMV-β-gal was provided by the Vector Core Facility of the University of Michigan (Ann Arbor, MI).

For viral infection, cells were seeded on 6-well plates at 3 × 10⁵/well overnight and then infected by a combination of either CMV-β-gal and RSV-yCD, or CMV-β-gal and CEA-yCD (at 3 × 10⁷ pfu for each) in 0.5 ml of RPMI with 2% FBS for 4 h. This was followed by additional incubation in 2 ml of complete medium for 48 h. Cells were washed twice in PBS and lysed in 50 μl of 1× report lysis buffer (RLB; Promega, Madison, WI) for assays described below.

Fifteen μg of cell lysates were fractionated on a 4–20% polyacrylamide gradient gel and electrotransferred onto a nitrocellulose membrane. An anti-yCD rabbit serum (1:1 × 10⁶ dilution; custom made by Berkeley Antibody Company, Richmond, CA) or mouse MAb, against human CEA (1:1000 dilution; NeoMarkers, Fremont, CA) was applied to the membrane, followed by a goat antirabbit or a goat antimouse antibody conjugated to horseradish peroxidase (Southern Biotechnology Associates, Inc., Birmingham, AL). Proteins were visualized by enhanced chemiluminescence (Pierce, Rockford, IL). Signals were measured by densitometry scanning using NIH Image software.

β-gal activity was measured using a commercial assay system (Promega). Twenty-five μg of cell lysate or 100 μg of tissue extracts were applied in 96-well plates, assessed for absorbance at 420 nm, and plotted on a standard enzyme curve following the manufacturer’s recommendation. CD activity was quantified by percentage conversion of ³H-labeled 5FC described previously (13). The percentage of total counts converted to 5FU permitted yCD to be quantified using a standard curve. The yCD activity is expressed as the amount of yCD corrected for infection by individual β-gal activity among different cell lines or separate viral infectants as formulated below:

Animal experiments using nude rats (6 to 8 weeks old; Charles River, Boston, MA) for human colon carcinoma cell implants (LoVo) were conducted using procedures approved by the University of Michigan Committee on Use and Care of Animals. Briefly, LoVo cells were trypsinized, washed in PBS, and resuspended at a concentration of 1 × 10⁸/ml in PBS. Twenty μl of cell suspension containing 2 × 10⁶ cells were then injected directly into the liver visualized by laparotomy.

The presence of tumor was verified 3 weeks after tumor implantation. Immediately afterward, 4 × 10⁹ pfu of CMV-β-gal and of CEA-yCD adenovirus vectors were infused by either hepatic artery or portal vein injection. For hepatic arterial infusion, the hepatic artery/gastroduodenal artery junction was exposed before gastroduodenal artery distal site ligation. The common hepatic branch was also isolated for bleeding control and a V-3 catheter with a flame-thinned tip was then securely placed into the artery. Two ml of mannitol (1 mg/ml) were infused through the artery at 0.5 ml/min before the viral injection, after which the catheter was removed, and the gastroduodenal artery was ligated at a proximal site. For portal vein infusion, the portal vein was isolated with a distal encircling suture for potential bleeding control, and a 30G needle was used for both mannitol administration and viral injection as described for the hepatic arterial infusion procedure. Animals then recovered under a heat lamp for 2 h and were assessed for gene transfer 72 h later. In a separate experiment, animals were given 5FC at 500 mg/kg i.p. for 1 h after 72 h of portal venous viral infusion. Liver and tumor tissues were collected for quantitation of 5FC and 5FU by gas chromatography/mass spectrometry as described below.

Rats were sacrificed 1 h after receiving the 5FC treatment. Liver and tumors were quickly collected, weighed, and then placed in 10 ml or 10 times the tumor volume of ice-cold Tris/EDTA [100 mm Tris-HCl (pH 7.8) and 1 mm EDTA] solution. Tissues were then homogenized using a Teflon/glass homogenizer followed by extraction into ethyl acetate in the presence of chlorouracil as an internal standard. The extracts were derivatized with N,O-bis-(trimenthysilyl) trifluoroacetamide. Quantification of the derivatized products was performed using a Hewlett-Packard 5987A gas chromatography/mass spectrometry in selected ion-monitoring mode.

Cells or tissues were homogenized in the appropriate amount of TRIzol reagent (Life Technologies, Inc., Grand Island, NY) following the manufacturer’s protocol for total RNA extraction. Total RNA from cells (10 μg) or from tissues (30 μg) was fractionated on a 1% agarose gel and then transferred onto a Hybond-N nylon membrane (Amersham Biosciences, Piscataway, NJ). The cDNA probe for yCD was PCR-amplified from the CEA-yCD vector described above, using primers for yCD gene (5′-atggcaagcaagtgggat, 3′-ctactcaccaatatcttc), and was then labeled with [α-³²P]dCTP using a random primer labeling system (redPrime II; Amersham Biosciences) according to the manufacturer’s protocol. The membrane was first prehybridized in QuikHyb solution (Stratagene, Cedar Creek, TX) for 1 h at 68°C in a rotating oven before adding radiolabeled probe for an additional 2-h incubation at 68°C. The membrane was then washed twice in 2× SSC, 0.1% SDS for 30 min at room temperature, followed by a stringent wash in 0.2× SSC, 0.1% SDS at 60°C for 1 h. Signals were detected by autoradiography on film. Total cellular DNA was purified from liver and tumors after viral infusion using a Wizard DNA purification kit (Promega). Viral DNA was measured by PCR amplification of yCD sequence (exclusively in viral DNA) using primers described above from an equal amount of total cellular DNA. These PCR products obtained from 20 amplification cycles were then used as an internal control for viral entry.

Cryo-tissue sections of liver and tumor were fixed in 1% formaldehyde in PBS for 5 min at 4°C, then washed twice in PBS before they were stained in an X-gal staining solution containing 77.4 mm Na₂HPO₄, 22.6 mm NaH₂PO₄, 2 mm MgCl₂, 5 mm K₃Fe(CN)₆, 5 mm K₄Fe(CN)₆, and 1 mg/ml X-gal at 37°C overnight. Sections were counterstained with Eosin-Y before mounting. Images were captured using an Olympus IX70 microscope equipped with a digital camera.

Signals from Western blots and Northern blots were measured by using NIH Image software. Means are expressed as ± the SE. Values were compared by ANOVA and were considered significantly different when P < 0.05.

We used a series of human colon carcinoma cells (LoVo, HT29, and CaCo2) as well as normal rat hepatocytes and human fibroblasts to test whether the CEA promoter could preferentially initiate yCD gene expression in comparison with the nonspecific, active RSV promoter. We first confirmed that these human colon carcinoma cells expressed high levels of CEA by Western blot analysis (20:5:1 ratio between LoVo:HT29:CaCo2 based on densitometry analysis). As expected, CEA expression was not detectable in hepatocytes and fibroblasts (Fig. 1,A). To determine whether the enhanced CEA promoter conferred selective expression of the yCD gene in colon carcinoma cells, adenovirus containing either RSV-yCD or CEA-yCD (3 × 10⁷ pfu) was used in cell infection experiments. We found that CEA-yCD virus consistently transduced higher levels of yCD expression than did RSV-yCD virus within the colon carcinoma cell lines, whereas it induced a much lower or equivalent level of yCD expression within hepatocytes or fibroblasts (Fig. 1,B). A densitometric scanning of the bands for CEA and yCD demonstrated a general agreement between the levels of CEA expression and relative fold changes of CEA-/RSV-induced yCD expression (Fig. 1 C). Thus, our data suggest that CEA promoter is preferentially activated in CEA-expressing cells.

To quantify yCD activity, we performed control experiments using CMV-β-gal adenovirus to measure infectivity in the different cell lines. Cell lines showed differing susceptibilities to infection determined by measurements of β-gal activities in total cellular extracts. LoVo and hepatocytes were the most able to undergo infection (0.29–0.39 milliunit/μg), followed by CaCo2 (0.24–0.28 milliunit/μg), HT29 (0.06–0.07 milliunit/μg), and fibroblasts (0.04–0.05 milliunit/μg). The β-gal activity was similar regardless of whether the β-gal vector was coinfected with the CEA-yCD or the RSV-yCD virus within each cell line (Fig. 2 A).

We were then in a position to determine whether our immunoblot results reflected an actual difference in enzyme activity. We used a highly sensitive yCD assay based on 5FC conversion into 5FU (described above), which can quantify as little as 2 pg of yCD enzyme (13), and we corrected for infectivity as described in “Materials and Methods.” We found that CEA-induced yCD activity was significantly enhanced in all of the cultured colon carcinoma cell lines in comparison with RSV-induced yCD activities [2.4 ± 0.2-, 1.6 ± 0.1-, and 2.1 ± 0.2-fold increase for LoVo, HT29, and CaCo2, respectively (P < 0.05)]. In contrast, CEA-driven yCD expression was minimal in non-CEA-expressing hepatocytes and fibroblasts compared with the RSV promoter counterpart [3.9 ± 0.3-, and 1.9 ± 0.2-fold decrease for hepatocytes and fibroblasts, respectively (P < 0.05)]. Moreover, yCD gene expression controlled by RSV promoter was similar in all of the cell lines (8.3, 5.1, 7.0, 9.0, and 5.7 ng/milliunit of β-gal activity for LoVo, HT29, CaCo2, hepatocytes, and fibroblasts, respectively; Fig. 2 B). This suggests that the CEA promoter can be activated preferentially in CEA-expressing colon carcinoma cells to achieve a high level of yCD expression, whereas the expression remains low in normal cells.

We then wished to assess the potential for CEA-yCD gene therapy in a relevant animal model of human colorectal cancer metastatic to the liver. LoVo cells were chosen for direct tumor implantation because of susceptibility to adenoviral infection and the ability of transgene expression demonstrated in vitro. Because the liver has a dual blood supply, we decided to infuse virus through either the hepatic artery or the portal vein. We found that the administration of CMV-β-gal virus by either hepatic arterial or portal venous infusion, successfully transduced liver and tumors (Fig. 3). It was also noted that CMV promoter-driven β-gal expression levels of liver and tumor were similar, although the normal liver appeared to have moderately higher activities after portal vein infusion compared with hepatic arterial infusion.

To determine whether CEA-yCD adenovirus could improve the tumor specificity of yCD production in vivo, nude rats carrying LoVo tumors were infused with a mixture of CEA-yCD and CMV-β-gal adenovirus through either the hepatic arterial or portal venous route. Normal liver and tumor were collected for Northern blot analysis using yCD cDNA as the probe. Representative data from three rats for each group (Fig. 4 A) demonstrated that yCD mRNAs driven by the CEA promoter were dramatically increased in tumors compared with the levels in liver (∼4-fold increase based on densitometry; P < 0.01) in both portal vein and hepatic arterial infusion groups. No yCD transcripts were detected in control rats.

To evaluate the possibility that the increased yCD transcript expression in tumors was attributable simply to greater viral infection in tumors than in liver, a semiquantitative PCR was carried out. In these experiments, we amplified the yCD sequence directly from viral genomic DNA that was included in an equal amount of total DNA contents from viral-transduced liver and tumors. We found that PCR products for the yCD gene were similar between liver and tumors, regardless of whether the virus was administered by hepatic arterial or portal venous infusion (Fig. 4 B). These findings suggest that adenovirus infects liver and tumor in a similar fashion, and that the yCD gene transcription is enhanced in colon carcinoma cells when CEA promoter is used.

We next wished to determine whether yCD protein expression after viral infusion was increased by the use of CEA-yCD. In contrast to the similar β-gal gene expression driven by CMV promoter in both liver and tumors from either the portal vein or the hepatic arterial infusion group (Fig. 5 A), yCD expression was markedly enhanced in tumors when CEA promoter was used. This was evidenced by an increase in functional yCD levels of 4.2 ± 0.4- and 4.3 ± 0.4-fold (corrected for β-gal activities) in tumors as compared with liver for hepatic arterial infusion and portal vein infusion, respectively (P < 0.01). Importantly, yCD levels remained equally low in normal liver (2.65 ± 0.5 and 4.21 ± 0.7 ng/milliunit for hepatic artery and portal vein, respectively), although above the background level in the control group (0.3 ± 0.2 ng/milliunit). To evaluate the potential of yCD gene therapy to generate 5FU within tumors, we assessed 5FU concentration in tumors, normal liver, and plasma from rats that had received 5FC treatment after a portal venous viral infusion. We found that intratumoral levels significantly exceeded plasma levels (21.1 ± 2.5 μm versus 8.4 ± 0.3 μm, respectively; data not shown), demonstrating that selective expression of yCD in tumors can produce therapeutic 5FU levels. Importantly, the ratio of 5FU concentration in the tumor, compared with plasma, was significantly greater (∼3-fold) than the normal liver:plasma ratio (2.51 ± 0.81 versus 0.85 ± 0.51, respectively; data not shown), suggesting that selective yCD expression in tumors can be achieved by using the CEA promoter.

In this study, we have found that the enhanced CEA promoter initiated high levels of yCD gene expression in colon carcinoma cells in both cell culture and in nude rats bearing intrahepatic xenografts. This specificity of yCD expression was reflected at both mRNA and functional protein levels compared with the expression levels produced by the RSV promoter. In contrast, CEA-driven yCD expression is barely detectable in normal rat hepatocytes and human fibroblasts. Concentrations of 5FU can be generated by this approach that would be anticipated to produce tumor regression if they were maintained.

Strategies using tumor-specific promoters in conjunction with therapeutic genes have been widely attempted in adenoviral gene therapy for cancers of colon, prostate, and liver (11, 12, 13, 14, 15, 16). The efficacy of this approach depends on both the activity and specificity of the promoter. The CEA promoter has been used in conjunction with bCD gene in an adenovirus vector and has been shown to improve the selectivity of 5FC/5FU conversion in CEA-expressing tumors (11, 12). However, the activity of the basal CEA promoter is far less than that of the nonspecific promoter or the enhanced CEA promoter used in this study. Furthermore, although these previous studies suggested that the CD strategy has the potential to be effective in an intrahepatic model, it is important to note that the dose-limiting factor for 5FC, when given as an antifungal agent, is intestinal toxicity resulting from 5FC/5FU conversion in the intestine by enteric bacteria (17, 18). Thus it seems likely that a gene therapy strategy using bCD will generate similar levels of 5FU in the tumor and the intestine, which would be anticipated to produce little or no therapeutic advantage compared with systemic 5FU infusion. In contrast, the K_m for the conversion of 5FC to 5FU for yCD is 22-fold lower than for bCD (6), suggesting that the yCD strategy has a greater potential to generate cytotoxic and radiosensitizing levels of intratumoral 5FU with acceptable intestinal toxicity in a clinical application.

Our present study, using an enhanced CEA promoter in conjunction with yCD in an adenoviral vector, demonstrated substantial activity and specificity of 5FC/5FU conversion in colon carcinoma cells. However, this system has limitations. It is clear that different human cancer cell lines evidence differing amounts of CEA, resulting in a corresponding range of CEA-yCD gene expression. We selected LoVo cells for intrahepatic tumor xenografts in our nude rat model because this cell line is more sensitive to adenovirus infection and is more responsive to CEA-mediated yCD expression, as demonstrated in our in vitro experiments. It should be noted, however, that the therapeutic effects in individual patients may vary significantly as a function of CEA expression within the intrahepatic metastases. Although we failed to detect CEA messenger transcriptions by PCR from normal human liver tissues (data not shown), it remains possible that normal liver or bile duct, which has been reported to produce small amounts of CEA (19, 20), would support CEA-driven yCD expression. We expect that yCD expression from tissues outside of intrahepatic colon carcinoma would be minimal. This is supported by our recent studies in mice, that CEA-yCD adenovirus infusion results in no conversion or a very low level of 5FC conversion in normal colon, bone marrow, spleen, and muscles (13).

Another limitation concerns problems related to low efficiencies in vector delivery, vector infectivity, and the potential toxicity of high vector doses. Despite the use of a regional delivery approach through the portal vein and the hepatic artery, only a small fraction of the tumor cells can be transduced (1–5% based on β-gal staining of tissue section) after a single dose of adenovirus infusion at 4 × 10⁹ pfu/rat. In addition, there was substantial heterogeneity of staining throughout the tumor. Development of methods to improve in vivo vector delivery and infectivity or the use of multiple vector infusion is important for future clinical application. Although the efficiency of tumor cell transduction by adenovirus is relatively low, our system clearly demonstrated that a clinically relevant concentration of 5FU converted from prodrug 5FC in tumor is achievable by a single and tolerable dose of adenovirus containing CEA promoter/enhancer-driven yCD gene in rats.

The liver appears to be an optimal organ for adenoviral gene therapy mainly because of nonselective viral absorption through the sinusoids. We aimed to compare efficiencies of viral targeting into intrahepatic colon cancer xenografts intra-arterially (hepatic artery) and i.v. (portal vein). Although previous studies from our group demonstrated that hepatic arterial delivery of chemotherapeutic agents may result in significant improvement regarding tumor growth inhibition (21), we found no advantage of viral delivery through hepatic artery over portal vein for gene expression within tumors. This is probably explained by the different mechanisms for cell entry used by infectious agents and small molecules. Our observation is consistent with recent findings (22), implying that other approaches are required to achieve selective gene expression in tumors. The approach used in this study involved a tumor-associated marker to drive selective gene expression in tumors. We demonstrated in this study that the CEA promoter selectively initiates yCD gene expression in colon carcinoma cells and produces significant 5FC to 5FU conversion in tumors compared with normal liver (4–5-fold) in vivo. Moreover, 5FU, the chemotherapeutic agent from this conversion, can offer further radiosensitization and bystander effects confirmed by similar studies (5, 23, 24, 25). Thus, CEA-yCD adenovirus gene therapy has potential therapeutic value in multimodality therapy against colon cancer metastases. In addition, the combination of an enhanced CEA promoter and a replication-activated adenovirus (26) appears especially attractive, because the increased expression demonstrated with this approach would be anticipated to be better targeted to tumors.