Current Challenges of Gene Therapy for Prostate Cancer

June 1, 1997

Gene therapy for prostate cancer faces hurdles similar to those being encountered for other cancers and nonmalignant processes. The greatest obstacle is the identification of efficient delivery systems, since numerous animal models and cell culture systems have shown potential efficacy when most cells express the introduced genetic material. Early prostate cancers are easily accessible to gene vector introduction, and the predictable metastatic patterns of this cancer may offer additional advantages for gene therapy. This article reviews gene vectors and gene products, as well as ongoing trials of gene therapy that have recently begun in prostate cancer. [ONCOLOGY 11(6):845-856, 1997]

ABSTRACT: Gene therapy for prostate cancer faces hurdles similar to those being encountered for other cancers and nonmalignant processes. The greatest obstacle is the identification of efficient delivery systems, since numerous animal models and cell culture systems have shown potential efficacy when most cells express the introduced genetic material. Early prostate cancers are easily accessible to gene vector introduction, and the predictable metastatic patterns of this cancer may offer additional advantages for gene therapy. This article reviews gene vectors and gene products, as well as ongoing trials of gene therapy that have recently begun in prostate cancer. [ONCOLOGY 11(6):845-856, 1997]


Prostate cancer is the most rapidly increasing cancer diagnosis in theUnited States, owing both to its prevalence in an aging "baby boom"population and to increased use of screening with improved technologies,such as the prostate-specific antigen (PSA) test.[1,2] Although early detectionoffers the hope of improved survival, evidence for this outcome is lacking.Furthermore, current therapeutic strategies for prostate cancer, even whenlocalized, are controversial because of unwanted side effects and disagreementover their efficacy.[3,4] Neoadjuvant hormone therapy has shown some promisein reducing the incidence of positive surgical resection margins and biochemicalrelapse after radiation treatment, but as yet no data indicate improvedoverall survival.[5-9]

Although advanced prostate cancer is well-palliated for 2 to 4 yearsby hormonal ablation of testosterone, it inevitably progresses. Furthertreatment with cytotoxic chemotherapy, radiopharmaceuticals, or external-beamradiation to areas of bone pain achieve transient responses but have nodefinite impact on survival. Thus, increasing attention has been givento the development of alternative treatments for prostate cancer of allstages, including gene therapy.

Definition of Gene Therapy

The term "gene therapy" is broadly applied to encompass allapproaches involving the introduction of genetic material into patients'cells. This may involve replacement of an absent or defective gene, suchas adenosine deaminase (ADA) in ADA deficiency, the low-density lipoproteinin familial hypercholesterolemia, and the chloride-transporter gene incystic fibrosis.

Alternatively, gene therapy may involve the introduction of anticanceror antiviral genes, such as those encoding tumor-suppressor proteins (leadingto tumor regression), ribozymes (catalytic RNAs that cleave targeted mRNAs),antisense RNA (interfering with normal RNA processing), and dominant mutationsthat knock out functions essential for viral activity or oncogenesis.

Finally, many of the ongoing gene therapy clinical trials involve avaccination approach with gene-modified tumor cells. In this paradigm,tumor cells from the patient are removed and genetically engineered tosecrete a cytokine or growth factor, such as interleukin-2 (IL-2) or granulocyte-macrophagecolony-stimulating factor (GM-CSF), or to express a foreign protein, suchas HLA-B7. Upon reinfusion of the cells, expression of the introduced geneshould stimulate the patient's own immune system to reject the tumor. Amodification of this approach involves the injection of a DNA vector directlyinto tumor deposits, with the intention of triggering local and systemicimmune responses.

Gene TherapyChallenges Common to All Diseases

The challenges of gene therapy for prostate cancer include those whichare common to gene therapy approaches for all diseases and those whichare unique for this disease. The common issues are grouped into three broadcategories: gene transfer, gene regulation, and safety. These have beenextensively reviewed recently.[10-12] Thus, we will provide only a briefsummary here, with the emphasis on recent clinical outcomes.

Gene Transfer

Efficient delivery of foreign genes into the appropriate target cellsis the major hurdle facing all gene therapy approaches today. Deliverymethods can be divided into viral (eg, retrovirus, adenovirus, and adeno-associatedvirus [AAV]) and nonviral (plasmid DNA delivered either as naked DNA orby physical means, such as cationic liposomes and particle-mediated bombardment).

Two major considerations in determining the optimal delivery methodare: (1) the target cell and its accessibility to the vector used to deliverthe DNA, and (2) whether or not the treatment requires long-term expressionof the gene. The first consideration also involves a decision to deliverthe DNA either in vivo or ex vivo, as well as a determination of the specificroute for in vivo delivery (ie, systemic administration or localized injection).The length of expression required varies with the nature of the defectbeing corrected, as well as the delivery system. For example, approachesthat involve physically mediated, localized, in vivo delivery of DNA canbe administered repeatedly. In contrast, methods that involve ex vivo deliveryof DNA into hematopoietic stem cells derived from bone marrow, followedby reinfusion of these cells, are limited in terms of the number of repetitionsdue to the invasiveness of the procedures.

For most approaches, long-term sustained expression of the therapeuticgene is desirable. With current technologies, most nonviral methods ofgene delivery cannot produce sustained expression. Of the viral vectors,two (retrovirus and AAV) are at least theoretically capable of producinglong-term expression; in fact, this has been demonstrated in ongoing clinicaltrials using retroviral vectors.

Specific features of the different delivery systems have been reviewedrecently[12,13] and are summarized in Table1.

Viral Vectors--All viral vectors share the feature of being disabled,so that the viruses may deliver DNA into the target cell but cannot undergoreplication once inside the target cell (ie, they are replication-incompetent).Since all of the viruses used are known to infect human cells, this isone of the major safety concerns. (For further discussion of this concern,see "Safety" below.) Features that distinguish the differentviral vectors include the size of the gene insert accepted, whether ornot the virus infects nondividing cells, the duration of expression, andpossible host immune response to viral proteins.

Of the approved clinical trials, over 85% have used viral vectors; themajority of these have been retroviral vectors, which have the longestusage history (5 years) in gene therapy studies. Bone marrow hematopoieticcells that were "marked" with retroviral vectors could stillbe detected in patients for up to 3 years following autologous transplantation;in these vectors, the polymerase chain reaction (PCR) technique was usedto amplify retroviral sequences.[14] Retrovirally marked tumor cells weredetected in leukemia and neuroblastoma patients who relapsed followingautologous transplantation with unpurged bone marrow cells.[15] This unequivocallydemonstrated that the graft had residual tumor cells, and pointed to theneed for purging.

In the first gene replacement study, conducted in patients with ADAdeficiency, ADA sequences retrovirally introduced into patients' T-cellscould still be detected 2 years following infusion.[16,17] The ADA genehas also been introduced into hematopoietic cells derived from cord blood,and ADA sequences have been detected 18 months following infusion.[18]

Together, these studies point to the feasibility of using retroviralvectors to deliver genes for long-term expression. However, with currentmethodologies this method is still quite inefficient.

Adenoviral vectors offer the most efficient method for gene transfer,and thus, have been the most frequently used vectors for localized in vivogene delivery, such as delivery of the CFTR gene for cystic fibrosis.[19]These vectors can infect nondividing cells, which is advantageous for applicationsthat target such cells. However, in the case of tumors, retroviral vectors(which do not infect nondividing cells) offer the possibility of selectivetargeting to the dividing tumor cell, whereas adenoviral vectors targetdividing and nondividing cells indiscriminately.

Another disadvantage of adenoviral vectors is that they generally resultin short-term expression, since the viral DNA is not integrated into thehost genome. In the case of cystic fibrosis, the longest-lasting expressionwas 4 to 9 days. Thus, this approach requires repeated administrations.

Adeno-associated virus vectors are being investigated as an alternativeto adenoviral vectors for delivery of CFTR into the lungs of patients withcystic fibrosis. These vectors can infect nondividing cells, albeit ata reduced efficiency compared with dividing cells. Development of producersystems to reproducibly generate high yields of AAV has been problematicand has slowed the clinical application of this vector system.[20]

Nonviral Vectors--The nonviral delivery methods offer severaladvantages over viral vectors with respect to safety and ease of production.Rigorous tests do not need to be applied to validate the absence of replication-competentviruses, which saves both time and money. Another advantage is that nonviralvectors can deliver larger pieces of DNA than can viral vectors.

Such an approach, utilizing liposomes to deliver HLA-B7 along with beta-2-microglobulingenes directly into tumor cells in vivo, has been applied in the treatmentof metastatic melanoma, hepatic metastases of colon cancer, and other tumors.[21,22]This approach falls into the immunotherapy category, in which the foreignantigen is recognized and serves to boost the local immune response. Insome patients treated with this approach, tumor shrinkage has been noted.

Although gene transfer with nonviral vectors is promising, its clinicalapplication has lagged behind viral vectors. This is due primarily to therelatively short-term expression of nonviral vectors. However, short-termexpression may be acceptable or even preferable for some applications,such as expression of a potent toxin or conditionally lethal gene in alocalized cancer. Thus, this strategy for gene delivery has some potentialapplications for prostate cancer, as discussed below.

Gene Regulation

Diseases that have been targeted for initial gene therapy studies donot require stringent regulation of the transferred gene. There are tworelated sides to this issue: On the one hand, expression in a nontargetedcell should not be toxic; on the other hand, even a low level of expressionin the target cell should be effective, since the expression of foreigngenes tends to be low and decreases with time. The ADA gene met these requirements,and thus, ADA deficiency was an ideal candidate for gene replacement eventhough it is a rare disease.

A related consideration is whether to use constitutive regulatory sequences(promoters) that are active in all cell types, or to use a tissue-specificpromoter in order to target expression to a particular cell type. Again,this depends on the gene and the disease. For example, delivery of a geneencoding a lethal product that induces the death of tumor cells would ideallyrequire a tumor-specific promoter so that its expression would be limitedto tumor cells. Unfortunately, such "magic bullets" that directexpression of the lethal gene exclusively to tumor cells have not yet beenidentified.

Promoters that are much more active in the target cell type than inother cell types have been used in preclinical studies, however. In reality,these promoters tend to be weaker than the constitutive promoters, andconsequently, their in vivo use is hampered by the problem of low expression.Of course, many gene therapy approaches skirt the issue of obtaining tissue-specificexpression by using delivery methods that preferentially target the appropriatecells. This is accomplished either by localized injection or by takingadvantage of vectors that preferentially target dividing cells.


Currently, the major safety concerns about gene therapy relate to: (1)generation of replication-competent virus with altered phenotype and/orhost range; (2) activation of an oncogene or inactivation of a tumor-suppressorgene due to insertional mutagenesis; and (3) potential immunogenicity resultingfrom expression of foreign proteins. The first concern relates exclusivelyto viral vectors; the second also relates mainly to viral vectors sincemost nonviral vectors are not integrated into the host genome.

Generation of Replication-Competent Viruses--All viruses in currentuse for clinical studies are replication-incompetent. Viral proteins areprovided in "trans," either in a stable packaging cell line ora transient expression system. "Trans" means that the virus thatis packaged in these cells contains all the necessary viral proteins (suppliedby engineered packaging cells) but lacks the genes to produce more viralproteins. Instead, the viral genome encodes the therapeutic gene underappropriate regulatory sequences. The resulting virus can thus undergoone round of infection, but, once inside the target cell, it is unableto produce more virus, since no viral proteins are made.

Concern about the generation of replication-competent virus stems fromone study in which a retroviral stock known to contain replication-competentvirus induced T-cell lymphoma in primates.[23] Thus, rigorous testing ofviral stocks is required, as well as monitoring of patients at specifiedperiods following infusion to ensure the absence of virus. Replication-competentvirus has not been detected in any of the more than 500 patients treatedwith viral vectors to date.

Insertional Mutagenesis--Theoretically, there is always the possibilitythat insertion of material into the genome may occur at a site, which wouldhave a deleterious effect. However, this has not occurred in any of theclinical trials to date.

Immunogenicity--In contrast, immunogenicity has proven to bea problem in clinical studies involving adenoviral vectors. In early cysticfibrosis trials, an immunogenic response to adenoviral proteins encodedin the vector genome led to decreased dosages and attempts to engineerthe vector to make it less immunogenic. It will be important to resolvethis problem, since the immunogenic response can lead not only to localizedinflammation at the site of delivery but also to rapid elimination of transducedcells.

There have been no deleterious, long-lasting effects attributed to geneticintervention in any of the patients treated in over 60 clinical protocolsto date.

Unique Features ofProstate Cancer

Several features of prostate cancer provide unique opportunities forgene therapy that are not necessarily shared by other tumors or conditions.

Long Latency

Prostate cancer has a very long latency period, with clinical manifestationsor diagnosis of disease occurring at a much lower rate than autopsy incidence.[24-26]This provides a prolonged window of opportunity for modification of geneticdamage via gene replacement or the addition of genes with tumor-suppressingactivity. Long latency also implies that even inefficient gene deliverymodalities would eventually reach most affected cells through repetitiveapplication of the vectors.

Unique Protein Products

Toxin genes targeting unique protein products produced by prostate epithelialtissue would be permissible because the human organism can live withoutthe affected organ; in this respect, prostate epithelium is similar tobreast or thyroid epithelial tissue but unlike lung or colon epithelium.Our group and others have generated data illustrating the "proof ofprinciple" by transducing toxin genes strictly regulated by DNA elementscontrolling unique proteins produced, for example, in the lens of the eye.Despite the presence of the toxin gene in all tissues, transgenic animalswere unaffected anywhere but in the eye.[27] Thus, the PSA, (prostate-specificmembrane antigen (PSM), or probasin gene regulatory sequences could allpotentially be used to direct gene expression to tissues of "prostate"origin anywhere in the body without significant concern about collateraldamage to the normal prostate.

Anatomy and Metastatic Pattern

Finally, anatomic considerations of prostate cancer present unique possibilities.The entire prostate is easily accessible by needles or other instrumentsthrough transperineal, transurethral, or transrectal approaches. Just asbrachytherapy (eg, radioactive seed implants) is commonly used only forcervical, breast, or prostatic neoplasia, locally administered gene vectorscould be easily used if they were available. Furthermore, the relativelyrestricted metastatic pattern of prostate cancer to lymph nodes or bonemay eventually offer unique delivery opportunities.

Candidate Genes

The selection of candidate genes for the treatment of prostate cancerdepends, in part, on whether the desired effect is local or systemic. Systemiceffects are of obvious importance in patients with pathologic stage C orD tumors, who have a high likelihood of or known metastases. Local geneticeffects, in turn, could target aberrant tumor-suppressor genes, amplificationof oncogenes, or growth factors thought to play a role in local tumor proliferation.

Local Gene Therapy

Tumor-Suppressor Genes--In some malignancies, the frequent lossof heterozygosity in expression of single tumor-suppressor genes, suchas p53, RB, p16, or p21, suggests that restoration of a single copy ofthe normal functioning gene and its subsequent production of protein willreestablish proliferative homeostasis. In vitro evidence to support thishypothesis is convincing. Clinical trials based on restoration of p53 inlung cancer are underway.[28]

In prostate cancer, no tumor-suppressor genes have yet been identifiedfor which deletions are universal or even frequent in early-stage disease.However, the replacement of whole portions of chromosomal material hasbeen shown to inhibit prostate cancer cell line DU 145 growth,[29] andsuppression of PC3 cell growth has been shown in experiments using theC-CAM1 tumor-suppressor gene delivered via adenovirus-mediated gene transfer.[30]

Local Growth Factors--Attacking local growth factors may offera second alternative to local gene therapy. Numerous candidate factorshave been identified that are required to maintain proliferation of prostatecancer in different models. These include epidermal growth factor (EGF),transforming growth factor-alpha, transforming growth factor-beta, basicfibroblast growth factor (bFGF), and insulin-like growth factors (IGFs).[31]The localized production of PSA itself may offer a target, since it hasbeen shown that one of the enzymatic targets of PSA beyond the seminalcoagulum is the IGF-binding protein family.[32] To abrogate expressionof these or other growth factors would require localized gene expressionof neutralizing molecules, such as antisense, ribozymes, or single-chainmonoclonal antibodies. In prostate cancer, like other tumors, angiogenesisor vascular endothelium itself offers another intriguing possibility forgene therapy.

Direct Toxins--The third category of gene products that may havelocal efficacy are those producing direct toxic effects on cells. Cellularsynthesis of toxins, such as the diphtheria toxin A-chain and Pseudomonasexotoxin, can produce lethal effects to all virally transduced or physicallytransfected cells, and regulation of expression could further restrictthese lethal effects only to those cells synthesizing PSA, PSM, probasinor other prostate epithelial-specific proteins, as mentioned above.

Conditionally toxic genes would produce effects only when coadministeredwith an additional agent. Examples of such binary systems include herpes-virus thymidine kinase (HSV TK)/ganciclovir (Cytovene) and cytosine deaminase/5-fluorocytosine(flucytosine [Ancobon]).[33] One pleasant surprise of the HSV TK systemhas been the discovery of the "bystander effect," which producescollateral toxicity to cells within the vicinity of single cells expressingTK after ganciclovir administration.[34]

Systemic Gene Therapy

Cytokine Genes--For metastatic disease, it is difficult to contemplatedelivery systems that are efficient enough to deliver therapeutic genesto all cells. In this situation, therefore, investigators have begun toexplore the possibility that cytokine gene expression by local tumors couldproduce distant effects mediated by effector cells "educated"at the primary tumor site. Virtually all of the known cytokine genes havebeen studied in various animal systems, and an excellent review of thesestudies has recently been published.[35]

Table 2 lists many of the cytokinegenes that have been tested in different animal models. At present, themost promising of these appear to be GM-CSF, IL-2, and, possibly, HLA-mismatchedantigens. Clinical trials of some of these cytokine genes are currentlyin progress in prostate cancer (see below).

Clinical Approaches

To date, six gene therapy protocols for the treatment of prostate cancerhave been approved in the United States. Table3 lists the protocols and their designs.

As indicated in the table, gene therapy for prostate cancer is in itsinfancy, with fewer than 10 patients treated to date. It is likely thatnone of the trials underway will show any therapeutic benefit; most aredesigned as phase I safety trials in patients with metastatic disease.As results of such trials are reported, it will be important to keep theirpurpose in mind rather than becoming discouraged. Particularly in thosetrials relying on the generation of an immunologic response, data fromanimal experiments indicate the necessity of beginning treatment beforethe detection of metastases in almost all cases.

The future of gene therapy for prostate and other cancers will dependon further development of vector systems at the basic science level, aswell as a better understanding of the genes involved in tumor inductionand proliferation. At present, it seems prudent to allow focused clinicaltrials to develop information of use in modifying our preclinical approachesand models, especially since patients are often willing to participatein such efforts and may have few alternatives. The greatest disserviceto such patients would be either to oversell expectations or categoricallyeliminate such clinical trials while we await further developments frombasic investigations.


1. Parker SL, Tong T, Bolden S, et al: Cancer statistics, 1996. CA CancerJ Clin 46:5-27, 1996.

2. Kramer BS, Brown ML, Prorok PC, et al: Prostate cancer screening:What we know and what we need to know [see comments]. Ann Intern Med 119:914-923,1993.

3. Fleming C, Wasson JH, Albertsen PC, et al: A decision analysis ofalternative treatment strategies for clinically localized prostate cancer:Prostate Patient Outcomes Research Team [see comments]. JAMA 269:2650-2658,1993.

4. Hanks GE: External beam radiation treatment for prostate cancer:Still the gold standard [see comments]. Oncology 6:79-86, 89; 90-94 [discussion],1992.

5. Cher ML, Shinohara K, Brelin S, et al: High failure rate associatedwith long-term follow-up of neoadjuvant androgen deprivation followed byradical prostatectomy for stage C prostatic cancer. Br J Urol 75:771-777,1995.

6. Soloway MS, Sharifi R, Wajsman Z, et al: Randomized prospective studycomparing radical prostatectomy alone versus radical prostatectomy precededby androgen blockade in clinical stage B2 (T2bNxMO) prostate cancer: TheLupron Depot Neoadjuvant Prostate Cancer Study Group. J Urol 154:424-428,1995.

7. Fair WR, Aprikian AG, Cohen D, et al: Use of neoadjuvant androgendeprivation therapy in clinically localized prostate cancer. Clin InvestMed 16:516-522, 1993.

8. Isreali RS, Wang Y, Fair WR, et al: Neoadjuvant androgen-deprivationtherapy (ADT) prior to radical prostatectomy results in a significant decreasein residual micrometastases as detected by nested RT-PCR with PSM primers(abstract). Proc Ann Meet Am Assoc Cancer Res 36:A3823, 1995.

9. Witjes WP, Oosterhof GO, Schaafsma HE, et al: Current status of neoadjuvanttherapy in localized prostate cancer. Prostate 27:297-303, 1995.

10. Anderson WF: Gene therapy. Sci Am 273:124-128, 1995.

11. Marshall E: Gene therapy's growing pains. Science 269:1050-1055,1995.

12. Ross G, Erickson R, Knorr D, et al: Gene therapy in the United States:A five-year status report. Human Gene Ther 7:1781-1790, 1996.

13. Blaese M, Blankenstein T, Brenner M, et al: Vectors in cancer therapy:How will they deliver? Cancer Gene Ther 2:291-297, 1995.

14. Dunbar CE, Cottler-Fox M, O'Shaughnessy JA, et al: Retrovirallymarked CD34-enriched peripheral blood and bone marrow cells contributeto long-term engraftment after autologous transplantation. Blood 85:3048-3057,1995.

15. Rill DR, Santana VM, Roberts WM, et al: Direct demonstration thatautologous bone marrow transplantation for solid tumors can return a multiplicityof tumorigenic cells. Blood 84:380-383, 1994.

16. Parkman R, Barrantes B, Weinberg KI, et al: Analysis of T lymphocytesfrom patients with adenosine deaminase (ADA) deficiency transplanted withADA transduced cord blood cells (abstract). J Cell Biochem 21A(suppl):369,1995.

17. Mullen CA, Snitzer K, Culver KW, et al: Molecular analysis of Tlymphocyte-directed gene therapy for adenosine deaminase deficiency: Long-termexpression in vivo of genes introduced with a retroviral vector. HumanGene Ther 7:1123-1129, 1996.

18. Kohn DB, Weinberg KI, Nolta JA, et al: Engraftment of gene-modifiedumbilical cord blood cells in neonates with adenosine deaminase deficiency.Nature Medicine 1:1017-1023, 1995.

19. Crystal RG, McElvaney NG, Rosenfeld MA, et al: Administration ofan adenovirus containing the human CFTR cDNA to the respiratory tract ofindividuals with cystic fibrosis. Nature Genet 8:42-51, 1994.

20. McKeon C, Samulski RJ: NIDDK workshop on AAV vectors: Gene transferinto quiescent cells. Human Gene Ther 7:1615-1619, 1996.

21. Nabel GJ, Chang AE, Nabel EG, et al: Immunotherapy for cancer bydirect gene transfer into tumors. Human Gene Ther 5:57-77, 1994.

22. Hersh EM, Akporiaye E, Harris D, et al: Phase I study of immunotherapyof malignant melanoma by direct gene transfer. Human Gene Ther 5:1371-1384,1994.

23. Donahue RE, Kessler SW, Bodine D, et al: Helper virus induced Tcell lymphoma in nonhuman primates after retroviral mediated gene transfer.J Exp Med 176:1125-1135, 1992.

24. Dhom G: Epidemiologic aspects of latent and clinically manifestcarcinoma of the prostate. J Cancer Res Clin Oncol 106:210-218, 1983.

25. Breslow B, Chan CW, Dhom G, et al: Latent carcinoma of prostateat autopsy in seven areas: Collaborative study organized by the internationalagency for research on cancer, Lyons, France. Int J Cancer 20:680-688,1977.

26. Holund B: Latent prostatic cancer in a consecutive autopsy series.Scand J Urol Nephrol 14:29, 1980.

27. Breitman ML, Clapoff S, Rossant J, et al: Genetic ablation: Targetedexpression of a toxin gene causes microphthalmia in transgenic mice. Science238:1563-1565, 1987.

28. Zhang WW, Roth JA: Anti-oncogene and tumor suppressor gene therapy--examplesfrom a lung cancer animal model. In Vivo 8:755-769, 1994.

29. Berube NG, Speevak MD, Chevrette M: Suppression of tumorigenicityof human prostate cancer cells by introduction of human chromosome del(12)(q13).Cancer Res 54:3077-3081, 1994.

30. Kleinerman DI, Zhang WW, Lin SH, et al: Application of a tumor suppressor(C-CAM1)-expressing recombinant adenovirus in androgen-independent humanprostate cancer therapy: A preclinical study. Cancer Res 55:2831-2836,1995.

31. Ware JL: Growth factors and their receptors as determinants in theproliferation and metastasis of human prostate cancer. Cancer Met Rev 12:287-301,1993.

32. Fielder PJ, Rosenfeld RG, Graves HC, et al: Biochemical analysisof prostate specific antigen-proteolyzed insulin-like growth factor bindingprotein-3. Growth Regulation 4:164-172, 1994.

33. Trinh QT, Austin EA, Murray DM, et al: Enzyme/prodrug gene therapy:Comparison of cytosine deaminase/5-fluorocytosine versus thymidine kinase/ganciclovirenzyme/prodrug systems in a human colorectal carcinoma cell line. CancerRes 55:4808-4812, 1995.

34. Freeman SM, Abboud CN, Whartenby KA, et al: The "bystandereffect": Tumor regression when a fraction of the tumor mass is geneticallymodified. Cancer Res 53:5274-5283, 1993.

35. Vieweg J, Gilboa E: Considerations for the use of cytokine-secretingtumor cell preparations for cancer treatment. Cancer Invest 13:193-201,1995.