Gene therapy and immunotherapy of prostate cancer: Adenoviral-based strategies

Extensive research focusing on viral vectors has allowed for significant advances in the field of gene therapy. Oncolytic adenoviruses may in the near future become important therapeutic agents in the management of organ-confined prostate cancer. An adenovirus, whose replication is controlled by a prostate-specific regulatory sequence, specifically replicates in and lyses cells of prostate epithelial origin, including prostate adenocarcinoma cells. Such a virus offers two appealing properties: it specifically kills normal and neoplastic prostate cells and its replication leads to controlled virus spread and thereby amplification of the therapeutic effect. Intraprostatic injections of oncolytic adenoviruses have shown good safety profiles and promising potential when combined with external radiation therapy. Current development of oncolytic viruses focuses on how to further improve their efficacy by including therapeutic genes or genes that can modulate the immune response mounted against them. Recent discoveries in tumor immunology have renewed hopes that immunotherapy may play a role in the management of advanced prostate cancer. One interesting approach to overcome the immunological tolerance that appears to be firmly established in many advanced tumors is ex vivo expansion and adoptive transfer of prostate antigen-specific cytolytic T lymphocytes. In parallel, extensive research on cancer vaccine development based on dendritic cells, tumor antigens, cytokines, and novel immune adjuvants is currently ongoing. It will be essential to include immunotherapy protocols early on in the course of recurrent prostate cancer in order to extend survival and improve the quality of life.

Prostate cancer

Prostate cancer is a leading cause of illness and death among men in Western countries. Current statistics from the American Cancer Society estimate that 220 900 men will be diagnosed with prostate cancer and that 28 900 will die from this disease in the USA in 2003 [1]. The current standard therapies employed for organ-confined prostate cancer include radical prostatectomy, external beam irradiation, and brachytherapy, under some circumstances incorporating neoadjuvant or adjuvant hormonal therapy [2], [3]. While these therapies are relatively effective in the short term, a significant proportion (30–40%) of patients who initially present with localized disease ultimately relapse [4]. Moreover, these therapies are often associated with significant side effects due to injury to adjacent tissues. For metastatic prostate cancer, the main therapy is androgen ablation. While this provides cytoreduction and palliation, progression to hormone-refractory disease typically occurs within the order of 14–20 months. A large number of clinical studies have been reported in the field of chemotherapy for advanced androgen-independent prostate cancer. Only recently have two trials revealed that chemotherapy marginally improves the overall survival of patients with hormone-refractory disease [5], [6]. Thus, there is a need for improved techniques to control localized prostate cancer and treat advanced hormone-refractory prostate cancer, which can complement the standard techniques.

The tendency for increased prostate cancer screening, typically with digital rectal examination and measurement of serum prostate-specific antigen (PSA), means that prostate cancer can be diagnosed at an earlier stage than was previously possible [7]. In the USA the clinical picture of prostate cancer is very different today from what it was 10–15 years ago, with a higher proportion of men with organ-confined disease and negative surgical margins [8]. In Sweden, where systematic PSA screening has not yet been initiated and where prostate cancer is still most commonly diagnosed during the work-up of symptomatic patients [9], such a clinical shift has not yet been observed. However, recent studies clearly indicate that early detection and early removal of the cancerous cells would be advantageous for prostate cancer patients [10]. Therefore, PSA testing will probably increase in Sweden, reducing the death toll from prostate cancer it is hoped, though at the same time it will cause a dramatic increase in the demand for organ-confined prostate cancer treatments.

Since the prostate gland is an organ not essential for life, proteins that are specifically expressed by neoplastic (and normal) prostate cells constitute suitable targets for prostate cancer immunotherapy. In addition, gene promoters of such prostate-specific proteins are transcriptionally active only in normal and neoplastic prostate cells and may therefore be included in viral vectors to restrictively control expression of therapeutic molecules or control replication of oncolytic viruses. Prostate cancer is particularly appropriate for gene therapy and immunotherapy, since a large number of genes and proteins with specific or preferential expression in prostate and prostate cancer cells have been identified, including prostate specific antigen (PSA) [11], kallikrein 2 [12], prostatic acid phosphatase (PAP) [13], prostate-specific membrane antigen (PSMA) [14], prostate-specific transglutaminase [15], homeobox NKX3.1 [16], prostate stem cell antigen (PSCA) [17], PAGE [18], TARP [19], [20], DD3 [21], STEAP [22], prostase/kallikrein 4 [23], PDEF [24], hPSE [25], PCGEM1 [26], PART-1 [27], PSGR [28], PMEPA1 [29], prostein [30], GDEP [31], Trp-p8 [32], PSDR1 [33], Cten [34], PATE [35], POTE [36], STAMP-1 [37], AIbZIP [38], HPG-1 [39], NGEP [40], and PrLZ [41]. Furthermore, many of the promoters controlling expression of these genes have been cloned and characterized.

Gene therapy

Gene therapy is defined as the introduction of a gene coding for a molecule that can cure or slow down the progression of a disease. It may be intended for gene correction, i.e., to replace a missing or non-functional gene, or to carry out gene-mediated destruction of diseased cells. Gene correction is an attractive approach for the treatment of monogenetic inherited diseases. However, in the case of cancer, where there are several genetic alterations, it is far more complicated to revert the malignant phenotype of a tumor cell through gene corrections. Instead, an attractive approach is to deliver a “therapeutic” gene which encodes a molecule that directly or indirectly kills tumor cells. Both viral and non-viral vectors have been developed for gene transfer to tumor cells. Gene transfer vectors have been extensively reviewed elsewhere [42–44]. This review will focus on the use of adenoviral vectors for gene delivery, and in particular adenoviral vectors for gene delivery to prostate cancer cells.

A general review on clinical gene therapy trials worldwide 1989–2004 was recently published [45] and several reviews focusing on cancer gene therapy trials have been published [46–49]. Furthermore, information about clinical trial protocols for prostate cancer gene therapy can be found at the NIH webpage (http://www.gemcris.od.nih.gov/).

Adenoviral vectors

The adenovirus genome is composed of a linear double-stranded DNA molecule of approximately 36 kb. Adenovirus genes are transcribed in a complex temporal manner which can be divided into an early and a late phase, respectively occurring before and after virus DNA replication [50]. Early viral gene expression starts at the E1 locus and includes gene products that modulate cellular metabolism in order to make the infected cell more susceptible to virus replication. The E1A proteins interfere with the processes of cell division by binding to members of the cellular retinoblastoma protein [pRb] family. Blockage of pRb leads to the release of transcription factors of the E2F family that activate transcription of target genes required for entering the S phase of the cell cycle. To circumvent premature apoptotic cell death due to excessive E1A-induced transcriptional activity, the E1B-55kD protein binds to p53 and inhibits its apoptotic activity while the E1B-19kD protein functions as a viral homologue of the cellular anti-apoptotic protein Bcl-2. E1A also plays a central role in activating the adenoviral gene expression cascade. Therefore, viruses lacking E1A are replication defective. The remaining early regions (E2, E3 and E4) encode proteins involved in viral DNA synthesis (E2), modulation of the host immune response and cell lysis (E3), and the regulation of viral gene expression, mRNA transport, DNA replication, and apoptosis (E4). The late gene (L1–L5) products encode viral structural proteins and proteins involved in virion assembly [50].

Adenoviruses can infect both dividing and non-dividing cells in a wide variety of cell types and tissues. This characteristic, together with their relative ease of preparation and purification, has led to their extensive use as gene transfer vectors. The adenovirus genome does not integrate into the host cell genome. Instead it remains episomal in the nucleus, thereby minimizing the risks for recombination and disruption of critical endogenous genes or formation of undesired viral mutants. The first generation of replication-defective adenoviral vectors have the E1 and E3 genes removed, allowing for the introduction of up to 7 kb of foreign DNA. The E1 proteins are then provided in trans by producer cells such as 293, 911, or PerC6. The E3 region is not necessary for replication of adenovirus in vitro and may therefore be abolished. Later generations of replication-defective adenoviral vectors have E2 and/or E4 gene deletions, allowing for the introduction of larger pieces of foreign DNA [50]. Helper-dependent adenoviral vectors that are completely devoid of all viral protein-coding sequences have also been developed [51].

Replication-competent adenoviruses include, at least, the E1A, E2, and E4 regions together with the late genes. They have the ability to infect cells, replicate therein, form progeny viruses that are released through lysis of the infected cell, and spread to neighboring cells. Since the virus itself exerts oncolytic activity, the use of replicating viruses is also referred to as virotherapy. Replication-competent adenoviruses have an advantage over replication-defective adenoviral vectors in the sense that every infected cell serves as a virus-producing cell, thereby augmenting the spread of therapeutic gene expression. Generally speaking, replication-competent viruses raise serious safety concerns about toxicity, risk of mutagenesis, and spread from one patient to other human subjects. However, the type 5 adenovirus meets the safety criteria as it is already prevalent in the human population, is of low pathogenicity, and is a non-integrating virus.

In order to make replication-competent adenoviruses tumor- or tissue-specific, the E1A gene, which controls adenovirus replication by controlling the viral gene expression cascade, can be placed under transcriptional control of tumor- or tissue-specific promoters. These viruses are replication-competent under certain conditions and are referred to as conditionally replication-competent adenoviruses (CRAds). An alternative way to produce tumor CRAds is to develop genetically modified adenoviruses that preferentially replicate in tumor cells with alterations in certain cell cycle pathways, such as the p53 and pRb pathways. Yet another way to restrict oncolytic viruses to certain tumors and tissues is to retarget virus infection either through genetic modification of the viral surface proteins that mediate viral attachment and cell entry, or by using bi-specific linkers that bridge virus surface proteins to specific cell surface receptors (see Figure 1).

Figure 1. Strategies for adenovirus retargeting. Restricted adenovirus replication can be achieved through tumor- or tissue-specific promoter-controlled E1A gene expression (a), or through genetic modification of E1A and/or E1B to create virus strains with selective replication in tumor cells with deregulated cell cycle pathways (b). Adenovirus infection can be retargeted from its natural binding receptor to a specific cell surface molecule through genetic modification of the virus fiber or capsid proteins which the virus uses for attachment and cell entry (c) or by using bi-specific linkers that bridge virus surface proteins to specific cell surface receptors (d).

Targeting adenoviral replication to tumor cells using tumor- and tissue-specific promoters

A number of therapeutic approaches relying on prostate-specific transcriptional elements have been envisioned, including therapeutic viruses whose replication is limited to prostate cells and viral vectors with therapeutic genes expressed under the control of prostate-specific regulatory sequences. A prostate CRAd, e.g. an adenovirus with E1A expression under transcriptional control of a prostate-specific regulatory sequence, may be injected intraprostatically to treat localized prostate cancer (see Figure 2). The virus will specifically replicate in prostate epithelial cells and prostate cancer cells, leading to lysis of infected cells and viral spread to neighboring cells that are in turn infected and lysed. Alternatively, the prostate CRAd may be injected intravascularly in an effort to treat metastatic prostate cancer.

Figure 2. Oncolytic adenovirus therapy. A conditionally replication-competent adenovirus (CRAd) with E1A gene expression controlled by a prostate-specific promoter specifically replicates in and lyses cells of prostate epithelial origin, including prostate adenocarcinoma cells. The oncolytic prostate CRAd virus may be injected intraprostatically in patients with localized prostate cancer or intravascularly in patients with metastatic prostate cancer.

Henderson and colleagues pioneered the work of prostate CRAds by using promoter and enhancer elements from human kallikrein 2 (hK2), prostate-specific antigen (PSA) and rat probasin (rPB) to control E1A and E1B gene expression [52–54]. Two of these viruses, CV706 (originally named CN706) and CV787 have entered phase I clinical trials for prostate cancer. CV706 contains the E1A gene under control of the PSE sequence (the minimal PSA promoter together with the upstream PSA enhancer), and it is E3-deleted [52]. Combination therapy using oncolytic CV706 and radiation therapy of prostate adenocarcinoma cells, LNCaP, resulted in synergistic cell killing both in vitro and in vivo [55]. The combination therapy was not associated with increased toxicity and it was found that 50-fold less CV706 was needed to achieve complete tumor regression in the animal model when combined with radiation. The phase I clinical trial used intraprostatic injection of CV706 in patients relapsing after radiation therapy, as characterized by increased PSA. The therapy was demonstrated to be safe and serum PSA levels dropped ≥50% in the five patients receiving the highest dose of virus (1×10¹³ particles) [56]. CV787 contains the E1A gene under control of the rPB promoter and the E1B gene under control of PSE, as well as the wild-type E3 region [53]. Insertion of E3, which encodes proteins that assist virus release and play a role in evading or slowing host immune responses to the virus, enabled CV787 to eliminate distant LNCaP prostate cancer xenografts [53]. The phase I clinical trial using intravenous delivery of CV787 has not yet been completed, although clinical safety of CV787 was reported at the 7th Annual American Society of Gene Therapy Meeting in June 2004. Final data from this trail are awaited.

In addition to prostate-specific promoters, the human osteocalcin promoter and the human telomerase reverse transcriptase (hTERT) promoter have been used to control E1A expression in CRAds with demonstrated effect in prostate cancer cells [57], [58]. Osteocalcin expression is restricted to bone-associated tissues, brain, and prostate cancer specimens and osteocalcin promoter-regulated CRAds have great potential for treatment of prostate cancer bone metastases. Telomerase is a unique DNA polymerase whose major function is to catalyze the addition of telomeric repeats to the ends of chromosomes to maintain chromosomal integrity. Most normal adult somatic cells are devoid of telomerase, which leads to a progressive shortening of the telomeres and ultimately results in senescence and cell death. Telomerase reactivation is a critical step for tumorigenesis, allowing cancer cells to proliferate indefinitely. Therefore, hTERT promoter-regulated CRAds have great potential for the treatment of a broad spectrum of tumors, including prostate cancer.

Therapeutic gene expression from tumor- and tissue-specific promoters

Promoters and enhancers

In addition to prostate CRAds, a number of replication-defective adenoviral vectors have been developed with therapeutic gene expression under the control of prostate-specific regulatory elements. Examples are adenoviral vectors with transgene expression based on promoter and enhancer elements from the genes encoding PSA [59–64], kallikrein-2 [65], prostate specific membrane antigen (PSMA) [66], osteocalcin [67], mouse prostate secretory protein of 94 amino acids (PSP94) [68], and the composite rat probasin promoter (ARR₂PB) [69–74]. To increase the strength of prostate promoters while retaining satisfactory prostate specificity, two-step transcriptional activation systems have been developed. Zhang et al. [75] developed a system in which a prostate-specific regulatory sequence, based on PSA promoter and enhancer elements, controls expression of a fusion protein comprising the DNA binding domain of the yeast transcription factor GAL4, and two transactivation domains of the Herpes simplex virus activator VP16. The produced fusion protein binds to a promoter, composed of GAL4 binding sites, which drives transgene expression. This system leads to higher transgene expression than a system where transgene expression is directly controlled by the prostate-specific regulatory sequence. Yoshimura et al. [76] combined the PSA promoter and the bacteriophage P1 Cre-loxP system to generate a new prostate-specific transcription system. PSA-dependent expression of the Cre recombinase triggers the excision of a stop codon flanked by a pair of loxP sequences which releases transgene expression from the strong cytomegalovirus (CMV) promoter.

Since prostate cancer patients are often treated by androgen withdrawal, it may be beneficial to the patient to use a regulatory sequence that results in high therapeutic gene expression even in the absence of testosterone. Therefore, combining different prostate promoters and enhancers with different characteristics may be advantageous to treat both hormone-dependent and hormone-refractory prostate cancer. Lee et al. [66] constructed an artificial prostate-specific enhancing sequence (PSES) comprising a minimal androgen-responsive PSA enhancer element and a minimal PSMA enhancer element (from the third intron) linked to a TATA-box promoter. When inserted into an adenoviral vector the PSES sequence showed prostate specificity and androgen independence both in vitro and in vivo. The activity of PSES in androgen-deprived cells comes from the PSMA enhancer that exerts increased activity in response to testosterone removal. Martiniello-Wilks et al. [77] constructed an adenovirus with an artificial regulatory sequence comprising the rPB promoter and the simian virus 40 (SV-40) enhancer. The artificial regulatory sequence conferred prostate-specific transgene expression with activity even in the absence of testosterone. Furuhata et al. [78] modified the androgen response elements (ARE) in the rPB promoter to construct a regulatory sequence that is induced by retinoids rather than by testosterone. Recently, we cloned the promoter for a prostate-specific transcript from the unrearranged T cell receptor γ-chain (TCRγ) locus, encoding the TCRg alternate reading frame protein (TARP) [79]. An adenoviral vector with a recombinant regulatory sequence comprising the TARP promoter, the PSMA enhancer, and the PSA enhancer (named PPT) was constructed [80]. In order to obtain absolute prostate specificity, a DNA insulator was introduced to shield the PPT sequence from adenoviral backbone sequences that otherwise interfere with transcriptional regulation. The insulator-shielded PPT adenovirus confers high prostate-specific gene expression in both the presence and absence of testosterone. Intravenous administration of the virus in nude mice with an orthotopic prostate tumor model revealed absolute tumor selectivity and a transcriptional activity fully comparable to the CMV promoter [80]. The PPT sequence is currently being used to control E1A expression in an effort to produce highly active prostate-specific CRAds in our laboratory.

Therapeutic genes

The adenoviral E1A gene is a potent therapeutic gene in the sense that when controlled by a tumor- or tissue-specific promoter, it restricts viral replication and oncolysis of cells where the promoter is active. Besides E1A, a number of genes encoding various therapeutic molecules have been utilized for adenoviral-based prostate cancer gene therapy. Examples are adenoviral vectors with prostate-specific expression of prodrug-activating enzymes, also referred to as suicide genes, such as herpes simplex virus thymidine kinase (HSV-TK) [60], [78], [81], Escherichia coli cytosine deaminase (CD) [76], and Escherichia coli nitroreductase [61]. HSV-TK facilitates the conversion of the prodrug ganciclovir (GCV) to the toxic GCV triphosphate, CD catalyzes the conversion of the nontoxic prodrug 5-fluorocytosine (5-FC) to the cytotoxic 5-fluorouracil (5-FU) and nitroreductase activates the prodrug CB1954 to a toxic metabolite. Other examples are adenoviral vectors with prostate-specific expression of apoptotic molecules such as Bax [69], [70], [72], [74], Bad [72], caspase-8 [82], and caspase-9 [65]. Yet other examples include adenoviral vectors with prostate-specific expression of diphtheria toxin A [64] and the sodium iodine symporter (NIS), which pumps iodine into cells and has utility in both imaging and therapy when combined with radioiodine [73]. Other molecules of interest that could be inserted in adenoviral vectors under prostate-specific expressional control include TRAIL, Fas Ligand, Caspase 3, and pseudomonas exotoxin, some of which have been developed in non-viral systems.

In addition to adenoviral vectors with prostate-specific transgene expression, numerous adenoviral vectors have been developed with suicide genes, apoptotic genes, tumor suppressor genes, or immunostimulatory genes under control of strong ubiquitous promoters, such as the CMV promoter. Many such viral vectors have been used for experimental and clinical prostate cancer therapy [83–85].

Restricting adenoviral replication to tumor cells using viral gene mutants

A number of genetically modified adenoviruses that preferentially replicate in tumor cells have been developed. The dl1520 strain (also called ONYX-015) contains alterations in the E1B-55kD gene that result in complete abrogation of E1B-55kD protein expression [86]. In wild-type adenovirus infection the E1B-55kD protein binds to p53 and blocks its apoptotic activity, thereby allowing efficient virus replication. The theory was that dl1520 would not be able to inactivate p53 in normal cells and, as a result, the viral replication cycle would not be completed. In cancer cells lacking normal p53 function, however, replication was predicted to proceed. This means that ONYX-015 should selectively kill cancer cells with mutated p53. However, several reports have indicated that ONYX-015 replication in tumor cells can be independent of the p53 status [87], [88]. The reason for this may be that functional inactivation of p53, caused by alterations in MDM-2 or p14^ARF protein expression, can make the tumor cell permissive to viral replication. Other examples of adenoviruses with genetic modification are the Δ24 strain [89] and the dl922-947 strain [90]. They both have small deletions in conserved region 2 of E1A that mediate interaction with the pRb family of proteins, which regulate the G1- to S-phase cell cycle checkpoint. These viruses confer selective replication in cells with a disrupted pRb pathway. Neither p53 nor pRb is mutated at a high frequency in early invasive prostate carcinoma [91], which may somewhat limit their use in the treatment of localized prostate cancer.

To improve therapeutic efficacy, Freytag and colleagues have constructed a replication-competent dl1520 adenovirus strain (FRG, also called Ad5-CD/Tkrep) that, in addition to the lack of E1B-55kD protein expression, encodes a fusion protein comprising the CD and HSV-TK prodrug converting enzymes [92]. They found that both the CD/5-FC and HSV-TK/GCV systems enhanced the cytopathic effect of the virus. The greatest therapeutic effect was seen in the virus plus double suicide gene therapy group compared with virus alone or with virus plus single suicide gene groups, suggesting synergistic interactions between the two suicide gene/prodrug systems [92], [93]. Moreover, a trimodal approach with Ad5-CD/TKrep, involving cell lysis by the replicating virus, double suicide gene therapy, and radiotherapy, led to significantly enhanced therapeutic efficacy compared with external beam radiotherapy alone in an orthotopic mouse prostate cancer model [94]. Both suicide gene systems can be used to suppress viral replication, thereby providing means to control viral spread, but at the same time dampen the lytic effect of the virus in cancer cells. Therefore, it is important to examine other suicide gene systems in combination with oncolytic adenoviruses that possess primarily anti-tumor and not anti-viral action.

The Ad5-CD/TKrep replicating adenovirus has recently entered phase I clinical trials for localized prostate cancer. It is used in combination with double suicide gene therapy (5-FC and GCV) and conventional-dose three-dimensional conformal radiation therapy [95], [96]. The safety of the trimodal therapy is good with a toxicity profile that does not differ from radiation therapy alone [96]. There are indications that the PSA level is declining much faster in the trimodal therapy group than in the group receiving radiotherapy alone. Furthermore, the number of negative biopsies after trimodal therapy is higher than expected for radiation therapy, suggesting possible interactions between the oncolytic adenovirus therapy, the double-suicide gene therapies and the radiation therapy, as reported at the 7th Annual American Society of Gene Therapy Meeting in June 2004.

Retargeted adenovirus infection

Three distinct sequential steps are required for adenoviral infection. These involve attachment of the adenovirus to a specific receptor on the surface of the target cell, internalization of the virus and transfer of the viral genome to the nucleus where it can be expressed. The primary virus to cell interaction of type 5 adenovirus occurs between the knob domain of the viral fiber and the cell surface Coxsackie/adenovirus receptor (CAR). CAR is strictly a docking site for the adenovirus, providing a high affinity virus to host association. The second interaction occurs between the RGD (Arg-Gly-Asp) amino acid motif of the viral penton base with the cellular integrins αvβ3 and αvβ5. This interaction triggers internalization of the virions through clathrin-mediated endocytosis and acidification of the endosomes, which allows the virions to escape and enter the cytosol. Following escape from the endosomes the partially dismantled virus is transported to the nuclear membrane and the virus genome is released into the nucleoplasm through nuclear pore complexes. Therefore, the capability of an adenovirus type 5 vector to infect a cell is based on CAR and integrin expression, and cells expressing CAR and integrins below a certain threshold level are refractory to adenovirus infection. It has been observed that many primary tumor cells and tumor cell lines are indeed refractory to adenovirus infection due to a deficiency of CAR expression. However, in the case of clinical prostate cancer, CAR expression is relatively well maintained [97], [98], which is advantageous for adenovirus-based prostate cancer gene therapy.

In order to retarget adenovirus infection to specific cells types, two general approaches have been utilized. One approach involves conjugate-based strategies, which require that the adenovirus is complexed with a targeting molecule that recognizes a specific cell surface receptor. A second approach is genetic retargeting, in which the virus fiber or capsid is genetically modified to recognize a specific cell surface receptor.

Conjugate-based adenovirus retargeting

Several strategies have been described to achieve functional linkage between the adenovirus fiber and a cell surface receptor. For example, a bi-specific antibody conjugate was used to target adenovirus to the epidermal growth factor (EGF) receptor [99]. In an alternative approach, a single chain Fv antibody specific for the adenovirus fiber knob was genetically fused to EGF [100]. The recombinant fusion protein was able to retarget adenovirus to the EGF receptor. Prostate cancer cells generally show abundant expression of EGF receptors [101] and adenoviruses retargeted to EGF receptors may be utilized for enhanced adenovirus targeting to prostate cancer cells. So far, no paper has been published where an adenovirus was successfully retargeted to a prostate-specific cell surface molecule, although there are several known candidate molecules expressed on the surface of prostate cancer cells, including PSMA [14], PSCA [17], STEAP [22], PSGR [28], and NGEP [40]. Cell surface molecules with potential use for prostate-specific retargeting of adenoviruses were recently reviewed by Maitland et al. [102].

Genetic modification of the adenovirus fiber or capsid

The adenovirus fiber knob may be genetically reengineered to bind to a specific cell surface molecule. Wickham et al. [103] constructed an adenovirus that contains a heparin-binding domain at the C-terminus of the fiber protein, which targeted the virus to broadly expressed heparan-containing cellular receptors. Dmitriev et al. [104] reported the construction of modified adenoviral vectors containing the RGD (Arg-Gly-Asp)] peptide in the HI loop rather than at the C-terminus of the fiber domain. Attempts to retarget adenoviral vectors have also included genetic modifications of the penton base or hexon capsid proteins, although these proteins have not been exploited to the same extent as fiber-based targeting strategies. Hexon proteins are abundant on the adenovirus capsid and they contain hyper-variable regions that can be engineered to include targeting ligands. Vigne et al. [105] modified the hyper-variable domain of the hexon proteins with a RGD peptide and found that the modified virus significantly increased the transduction of human vascular smooth muscle cells in vitro. Insertions of a targeting ligand into the knob domain have led to the expansion of viral tropism but not necessarily the ablation of CAR binding.

Fiber genes from other adenovirus serotypes can replace the native fiber gene in adenovirus type 2 and 5 vectors. This was first demonstrated in a publication by Gall et al. [106] where removal of the adenovirus type 5 fiber and subsequent substitution with the adenovirus type 7 fiber resulted in altered viral tropism. It has been shown that substitution of the type 5 fiber with the type 35 fiber increases the adenovirus transduction efficiency of cells of hematopoietic origin, such as CD34⁺ stem cells and dendritic cells [107], [108]. Replacement of only the knob domain of the fiber can also alter viral tropism. This was first demonstrated by Krasnykh et al. [109], who replaced the adenovirus type 5 knob with the type 3 knob.

Routes of adenovirus administration and immune responses

Initial clinical trials used intratumoral injections of replication-defective or CRAd adenoviruses in order to maximize safety. More recently systemic administration routes have been utilized, for example intravenous injection of ONYX-015 virus for the treatment of head and neck cancer [110]. Clinical trials using systemic administration of adenovirus appear to be safe and the treatment well tolerated as long as lower doses are injected [111]. However, systemic administration faces problems of systemic immune responses in the form of innate immunity and neutralizing antibodies. The innate response, which occurs within 24 hours, is dose-dependent and independent of viral transcription. It is directed against the viral particle and results in inflammation of transduced tissues. Ongoing clinical trials with single or repeated systemic administration of adenovirus will hopefully enhance our understanding of adenoviral pharmacokinetics and anti-adenoviral immune responses.

Adenoviruses also induce adaptive antiviral immunity, which can either inhibit or enhance the intended therapeutic effect. In the case of CRAd viruses, it is desirable to obtain maximal virus spread within the tumor before the virus is cleared by the host immune system. Therefore, insertion of the adenovirus E3 region into the virus vector, in full or in part, may help to slow down immune responses to the virus. The E3 gp19K protein binds to MHC class I molecules and prevents the export of MHC class I/antigenic peptide complexes to the cell surface, thereby blocking the killing of infected cells by cytolytic T lymphocytes. The E3 RID and E3 14.7K proteins also help to blunt the immune response by inhibiting tumor necrosis factor (TNF)-mediated apoptosis. The E3 11.6K protein named adenovirus death protein ensures efficient release of progeny virus from infected cells. CRAd viruses, which overexpress the adenovirus death protein, have increased cell lysis capacity and this leads to improved adenoviral spread [112]. The intentional tuning of the antiviral immune response in adenoviral-based therapies is delicate. A delayed immune response ensures efficient spreading of CRAd viruses. On the other hand, a strong antiviral immune response may activate cytolytic T lymphocytes against tumor-associated antigens, which may lead to a desirable immunologic bystander effect with enhanced destruction of tumor cells.

Concluding remarks and future directions

Extensive research focusing on adenoviral vectors has allowed for significant advances in the field of cancer gene therapy. The field is rapidly moving from replication-defective to replication-competent adenoviral vectors. CRAd viruses offer two appealing properties as agents for cancer therapy: they specifically kill targeted cells and their replication leads to amplification of the therapeutic effect. Intraprostatic and intravenous injections of oncolytic CRAds have shown good safety profiles and encouraging potential when combined with external radiation therapy. In addition, hundreds of patients have been injected with various oncolytic adenoviruses, Herpes simplex viruses, or Newcastle disease viruses in a large number of clinical trials and without severe toxicity or side effects. Therefore, the focus on CRAd development is moving from primary concerns about safety towards how to make these viruses more efficient by increasing virus replication and virus spread within the tumor. Preclinical data suggest that efficacy may be improved by arming CRAds with therapeutic genes. It may be equally important to include genes encoding molecules for immune system modulation into the CRAds. Genes encoding molecules that suppress the immune response against the adenovirus may increase virus spread and thereby enhance the oncolytic effects. However, the host immune system will eventually clear the CRAds viruses and when this happens it might be advantageous to have genes encoding molecules that direct the immune response to a T helper 1 [T_H1] cell-type response, in order to activate cytolytic T lymphocytes against tumor-associated antigens and thereby obtain an immunologic bystander effect.

Immunotherapy

A central question in tumor immunology is whether recognition of tumor-associated antigens (TAAs) by the immune system leads to activation or tolerance. Evidence that T lymphocytes can help to control tumor growth is provided by the analysis of tumor prevalence in immunodeficient individuals. The incidence of tumors of viral origin is increased markedly in patients with drug-induced or HIV-related immunodeficiency [113]. However, for human tumors of non-viral origin, the incidence is similar in immunodeficient and healthy individuals, suggesting that for most human tumors no significant pressure is exerted by naturally occurring tumor immunity. If a specific immune surveillance system operates at early stages of tumorigenesis, it has most likely been eluded by the tumor at the time of tissue invasion and metastasis. It is therefore plausible that most tumors induce tolerance at an early stage of development and that advanced tumors may be particularly resistant to immunotherapies, as tolerance may be firmly established.

However, a number of important discoveries over the past two decades have led to renewed enthusiasm for cancer immunotherapy. First, the identification of human TAAs that can be recognized by autologous T cells, a finding that was first reported by Boon and colleagues [114], [115]. Second, discoveries were made showing that professional antigen-presenting cells must be activated in order to induce effective T cell immunity and that pattern recognition receptors and danger signals link innate to adoptive immunity via dendritic cells [DCs] [116], [117]. Third, the discovery that regulatory T lymphocytes, which are important for maintenance of immunological self-tolerance, are also involved in the induction of peripheral tolerance to TAAs and that depletion of these regulatory T lymphocytes can abrogate immunological unresponsiveness to tumors in vivo [118–121].

Cancer immunotherapy through T cell activation

Early attempts at cancer immunotherapy focused on interleukin (IL)-2 or other T cell-activating cytokines that were intended to expand the number and potency of T cells reactive with TAAs. Such activated T cells were expected to infiltrate the tumor and lead to the specific destruction of tumor cells. In addition, early attempts were carried out with tumor cell lysate or irradiated tumor cells that were either mixed with adjuvants or genetically modified to increase immunogenicity. Although these attempts to boost anti-tumor immunity have yielded some limited examples of tumor regression, they have failed so far to control advanced-stage disease to a significant extent [122]. As we learned more about antigen presentation, vaccination strategies were developed where peptide antigens or plasmids encoding TAAs, emulsified in immune adjuvants, were injected directly into patients, typically intramuscularly. Clinical trials using such vaccination strategies have only rarely and sporadically led to regressions of established tumors in patients with metastatic disease [123].

The low clinical response rates may partly be due to the fact that the tumor environment lacks the danger signals that link innate to adoptive immunity, which is central for the immune system in eradicating foreign pathogens. The lack of danger may prevent the development of a general inflammatory reaction and may account for the failure of antigen-specific T cells to eliminate antigen-expressing tumor cells in vivo. Pattern recognition receptors such as the toll-like receptors (TLRs) link innate to adaptive immunity. Therefore, these receptors may be targeted to develop cancer immunotherapy strategies that mimic the efficient immune responses raised against foreign pathogens. For example, oligodeoxynucleotides (ODNs) with unmethylated CpG motifs, abundant in viral and prokaryotic genomes, can trigger an immunomodulatory cascade in humans which involves B cells, T cells, natural killer cells and DCs [124]. The immune response triggered by CpG ODNs, via TLR-9 signaling, skews the immune system in favor of a T_H1 cell-type response and pro-inflammatory cytokine production. The inclusion of CpG ODNs as vaccine adjuvants has shown promising results in some animal models [125]. Double-stranded RNA (dsRNA), a major viral signature, activates DCs via TLR-3 or cytosolic dsRNA-binding enzyme protein kinase R. This activation leads to secretion of IL-12 and type-1 interferons which are important cytokines linking innate and adaptive immunity. An alphavirus replicon, which is a major source of dsRNA, or poly(I:C), a synthetic analog of dsRNA, may therefore be used to increase danger in the tumor environment. In addition to the lack of danger signaling, it will be important to overcome the escape mechanisms of tumors, including the loss of antigen expression and antigen presentation by tumor cells, the local presence of immunosuppressive factors or cells in the tumors, and the tumor-induced inability of lymphocytes to home to and infiltrate tumor tissues.

Dendritic cell vaccines

The ultimate effector cell that mediates the rejection of solid tumors in preclinical animal models is the cytolytic T lymphocyte (CTL). CTLs are derived from naïve or memory CD8⁺ T cells. Each CTL expresses a clonotypically unique T cell receptor (TCR) that confers specificity for a particular target antigen. The antigens recognized by TCRs on CTLs consist of peptide fragments bound within the major histocompatibility complex (MHC) class I molecules on the surface of the antigen-presenting cell. The peptides displayed by MHC class I molecules are generally from endogenous proteins expressed by the antigen presenting cell. DCs are specialized professional antigen presenting cells that are very efficient in presenting antigenic peptides to T cells along with co-stimulatory molecules needed for T cell activation. They exist as two distinct phenotypes, immature DCs specialized in antigen uptake and mature DCs specialized in antigen presentation. DCs have the unique capacity to stimulate naïve T cells and thereby initiate primary immune responses [126]. Therefore, in order to initiate anti-tumor CTL responses, peptides derived from TAAs must be presented by MHC class I molecules on activated DCs.

Immature DCs can readily be generated from peripheral blood monocytes with IL-4 and granulocyte macrophage colony stimulating factor (GM-CSF). They may then be matured with tumor necrosis factor (TNF)-a or other maturation stimuli. Immature DCs can be loaded with the full-length TAA in the form of a recombinant protein, in vitro transcribed mRNA encoding the antigen, or a viral vector comprising the cDNA encoding the antigen. DCs are then matured for optimal processing and presentation of immunogenic peptide epitopes. Alternatively, fully mature DCs can be pulsed with synthetic MHC class I-restricted peptides for direct presentation of known immunogenic peptide epitopes. Autologous antigen-modified mature DCs can be infused as a vaccine into cancer patients [127]. Clinical trials using DC vaccines against established tumors show that they are safe, with minimal side effects [128]. The clinical results have been variable with notably long-lasting clinical responses in some studies [128]. However, several of the initial vaccines used immature rather than mature DCs, which may have affected the immunological and clinical outcome. Only mature DCs have the ability to stimulate T cell responses and enhance homing to draining lymph nodes, the site where therapeutic T cell responses are initiated.

Adoptive transfer of tumor antigen-specific T lymphocytes

One interesting approach that may improve therapeutic outcome is to expand CTLs directed against TAAs ex vivo and adoptively transfer them back to the cancer patient. In such a therapeutic setting lymphocyte cultures are repeatedly stimulated with autologous DCs modified to present immunogenic TAA epitopes. Specific CTLs with high avidity for an antigenic epitope can then be identified by peptide/MHC tetramer or enzyme-linked immunosorbent assay (ELISA) technologies. T cell clones can be raised from a single T cell, or a specific T cell population can be isolated by fluorescent-activated cell sorting (FACS) using a peptide/MHC tetramer. Isolated lymphocyte cultures can be rapidly expanded ex vivo, to clinically relevant numbers for adoptive transfer back to the patient (see Figure 3). Alternatively, the molecular characterization of TAA epitopes and the TCRs that recognize them has opened up the possibility to genetically engineer autologous T cells ex vivo to express a full-length TCR or a chimeric single chain Fv receptor that is specific for a TAA epitope [129].

Figure 3. Adoptive transfer of prostate antigen-specific T cells. Monocyte-derived dendritic cells (DCs) are genetically modified with a prostate antigen-encoding adenovirus vector. The modified DCs are mixed with autologous lymphocytes for stimulation of antigen-specific T cells ex vivo. Antigen-specific T cells are selected and rapidly expanded to large numbers for adoptive therapy. The patient receives systemic immunosuppressive chemotherapy prior to the adoptive transfer of T cells.

In contrast with early studies of adoptive therapy that used lymphokine-activated killer (LAK) cells or tumor infiltrating lymphocytes (TILs), the use of antigen-specific T cell clones for adoptive transfer provides effector cells of uniform specificity and phenotype that retain physiological responses to antigen stimulation and low-dose IL-2 [130]. Very large numbers of highly selected cells can be adoptively transferred. One important aspect of adoptive T cell transfer is that the T cells are expanded ex vivo, thereby circumventing the host immune regulatory mechanisms and the potentially immunosuppressive tumor environment. Another important aspect is that the patient can receive systemic immunosuppressive chemotherapy before the T cell transfer. This lymphodepletion creates space and provides homeostatic lymphocyte survival and proliferative signals for the adoptively transferred T cells [131]. Moreover, it has been suggested that the success of adoptive T cell transfer for the treatment of Epstein–Barr virus (EBV)-related lymphoma and for prophylaxis of CMV infection is due, in part, to the immunosuppressed status of the treated patients. Therefore, it may well be that lymphodepletion prior to adoptive transfer of TAA-specific CTLs can help to induce the inflammatory response that is needed for the CTLs to exert their cytolytic action.

It is well established that CD25⁺ CD4⁺ regulatory T cells (T_reg) are important for the maintenance of immunological tolerance. They have also been found to have a profound effect on the inhibition of T cell responses against cancer [118]. Systemic depletion of T_reg cells with high dose anti-CD25 antibody or immunotoxin prior to adoptive transfer of antigen-specific T cells may have implications for cancer immunotherapy. However, activated effector CD8⁺ and CD4⁺ cells also express CD25, and depletion of these cells during the acute phase of antitumor T cell response may limit the use of this approach. Moreover, it creates a greater risk for autoimmunity than a general lymphodepletion, since the potentially autoreactive T cells would not be depleted along with the T_reg cells. It has recently been shown that TAA-specific CD25⁺ CD4⁺ T_reg cells are present at tumor sites [132]. General application of T_reg cell depletion in the clinics will probably await methods to preferentially deplete T_reg cell subsets that regulate antitumor responses.

Prostate cancer immunotherapy

There are two main reasons why prostate cancer is particularly well suited to immunotherapy and adoptive T cell therapy. First, prostate tissues express many unique antigens, which may be exploited for the induction of anti-tumor CTL responses. Prostate cancer cells also express many epithelial cancer-related antigens that may be exploited as well. Moreover, immune responses to prostate-related antigens can be considered tumor-specific in patients with recurrent metastases after radical prostatectomy. Second, detection of serum PSA is available as an evaluation marker of clinical responses.

A fair number of immunogenic peptides derived from the prostate tissue antigens PSA [133–137], PSMA [138–142], PAP [143], [144], PSCA [145–147], kallikrein 4 [hK-4], prostein [148], [149], trp-p8 [150], and TARP [151], [152] have been identified in vitro. They are restricted by HLA-A2, HLA-A3, or HLA-A24, which are all relatively common MHC class I molecules in the general population. MHC class II-binding peptides that can activate CD4⁺ T helper cells have so far been identified from PSA [153], PSMA [153–155], PAP [156], and hK4 [157]. T helper cell activation is essential for a sustained immunological response against tumor cells. In addition, our laboratory has recently shown that potent CTLs directed against the prostate tissue antigen TARP can be generated ex vivo by repeated stimulation of lymphocyte cultures with autologous peptide-pulsed DCs followed by peptide/MHC tetramer-guided sorting of the stimulated T cells and 1000-fold rapid expansion of the sorted T cells [152]. The sorted and expanded T cells retained both specificity and activity, demonstrating that large numbers of prostate tissue antigen-specific CTLs can be generated ex vivo for adoptive transfer to prostate cancer patients [152].

Several phase I and phase II clinical trials have been conducted with autologous DCs preloaded with prostate tissue antigens. Murphy and colleagues were first, using DCs pulsed with HLA-A2-restricted PSMA peptides [141], [158], [159]. Burch et al. [160] and Small et al. [161] used DCs pulsed with a recombinant fusion protein composed of PAP linked to GM-CSF, while Fong et al. [162] used DCs pulsed with murine PAP as a xenoantigen. Heiser et al. [163] used DCs transfected with mRNA encoding PSA while Barrou et al. [164] used DCs pulsed with recombinant PSA. Immunotherapy using antigen-presenting cells (APC8015, Provenge) pulsed with the recombinant PAP/GM-CSF fusion protein has been conducted as phase II [165] and phase III (not yet published) clinical trials for metastatic androgen-independent prostate cancer with sporadic although encouraging results. Generally speaking, the DC vaccine trials for prostate cancer have been safe and well tolerated by the patients. Immunologic responses have been limited but specific T cell responses were observed in certain patients, demonstrating that it is possible to activate CTLs against self-antigens such as PSA, PSMA, and PAP in vivo in a clinical setting.

As mentioned above, telomerase reactivation is a critical step for tumorigenesis. Therefore, peptides derived from hTERT may be used to expand T cells with activity against a broad spectrum of tumors, including prostate cancer. hTERT peptides restricted by HLA-A2 [166], [167], HLA-A3 [168], and HLA-A24 [169] have been identified, as well as various MHC class II-restricted hTERT peptides [170], [171]. Vonderheide et al. [172] recently performed a phase I clinical vaccination trial with autologous dendritic cells pulsed with an HLA-A2-restricted hTERT peptide. hTERT-specific T lymphocytes were induced in four out of seven patients with advanced breast or prostate carcinoma, as assessed by peptide/MHC tetramer, enzyme-linked immunospot, and cytotoxicity assays.

So far, no clinical trial has been conducted with adoptive transfer of prostate antigen-specific T cells to prostate cancer patients.

Concluding remarks and future directions

Immunotherapy of cancer is a field of extensive research activity and recent discoveries in tumor immunology have renewed hopes that immunotherapy may play a role in the treatment of clinical cancers. So far, the clinical results have not fulfilled expectations and only sporadic tumor regressions have been documented. The main problem that remains to be solved is how to overcome the immunological tolerance that appears to be firmly established in most advanced tumors. One interesting approach that is being explored is ex vivo expansion and adoptive transfer of autologous TAA-specific CTLs to cancer patients who have received systemic immunosuppression prior to the adoptive T cell therapy. Growing knowledge about lymphocyte characteristics, other than antigen specificity, that can impact lymphocyte behavior in vivo may be utilized for the design of improved adoptive therapy protocols. For example, the discovery that the maturation state of CD8⁺ T cells determines their in vivo transport and persistence during an immune response [173] opens up new possibilities to improve the behavior of T cells. However, adoptive T cell transfer is a treatment that requires advanced logistics. Therefore, vaccination strategies based on dendritic cells, tumor antigens, cytokines, and novel immune adjuvants such as CpG ODNs will be continuously explored. Considering the fact that a large proportion of prostate cancer is hereditary and that many prostate tissue antigens have been characterized, prophylactic prostate cancer vaccines may be explored as well.

General conclusions

Prostate cancer remains a leading cause of illness and death among men in Western countries. Early detection and aggressive treatment are essential parameters to improve the overall survival of patients, although this will lead to over-treatment for many individuals. In the case of localized prostate cancer, intraprostatic injection of oncolytic prostate CRAds has shown great potential when combined with radiation therapy. Furthermore, it is plausible to administer CRAds concomitant with radical prostatectomy in order to trace tumor cells that may have disseminated from the prostate capsule. In the case of metastatic prostate cancer the situation is far more complicated and no curative treatment exists today. Several studies have demonstrated that TAA-specific CTLs can be activated both in vitro and in vivo, clearly showing that the host immune system can be activated against tumors. As DC-based vaccination and adoptive T cell transfer strategies become more and more sophisticated, it is essential to include them early on in the course of recurrent prostate cancer, preferably along with the standard hormonal treatments. It is hoped thay this will lead to a role for immunotherapy in controlling prostate cancer by extending survival and improving the quality of life.

The author would like to thank Christina Ninalga, Helena Dzojic, Wing-Shing Cheng, Björn Carlsson, and Dr Valeria Giandomenico for critical reading of the paper.