Introduction The field of gene therapy for cancer now has the luxury (or some may even say the curse) of clinical trial data to assess the efficacy of both the genes and the vectors which have dominated the field through its infancy.1,2,3,4,5,6 On balance, these data suggest that the problems which gene therapy for cancer will take into the new millennium focus far less on the choice of the therapeutic gene(s) to be used than on the means of delivering them. There is already a battery of genes that we know are very effective in killing cells, if they can but be expressed at the right site and at appropriate levels. None the less, until the perfectly targeted vector is developed,7 the choice of gene will remain crucially important in order to compensate for the deficiencies of the vectors which we currently have available (Figure 1).8 GenesThere is still no clear consensus on which tumour-clearing approach should be adopted using gene transfer. Intuitively, it would seem that simple eradication of cells is likely to be the best, and safest. The self-renewing nature of malignant disease dictates that tumour cells should be cleared as efficiently as possible rather than genetically corrected. Thus, the most frequently used genes have been those designed to kill cells directly – such as suicide genes9 – or to induce immune-mediated destruction – such as immunogenic antigens or cytokines.10 None the less, several groups have had success using genetic therapies to turn the aberrant biology of tumour cells against themselves. This has been the basis for the design of replicating vectors targeting genetic mutations specific to tumour cells (see below),11 as well as for the delivery of genes which stand as sentinels over control of the cell cycle such as p53.5 Given the genetic redundancy in human tumours, especially at late stages of evolution, genetic therapies for cancer may run the risk of being short-circuited by the recruitment of other signals driving proliferation of the tumour cells. However, if the genetic targeting approach uses pathways which are central to the continued survival of the cell it may be possible to induce tumour killing by triggering apoptotic effector mechanisms directly12,13 or indirectly.14 In some cases, it is evident that tumour cells may actually have a particular vulnerability to the induction of programmed cell death because of the loss of survival signals, for instance through changes in their cell adhesion properties.15 Whatever its mechanisms, no single gene can be a serious contender unless it has a demonstrable bystander effect. The requirement for such a bystander effect stems directly from the poor delivery efficiency provided by current vectors (see below).7 This bystander effect can come in one of two guises – either local9 or immune mediated16,17 – and a combination of the two is preferable. The prototypical bystander killing involves the transfer of toxic metabolites locally between cells9 but there may be other mechanisms including suppression of angiogenesis18,19 and a ‘kiss of death’ delivered by contact with dying cells.20 The next few years should see a wave of papers describing ways in which this local bystander killing can be enhanced, for example, by co-administration of pharmaceutical or genetic agents which enhance cell–cell communications.21 In addition, immune stimulation through cell killing can also enhance local tumour killing22 and can help to generate systemic immunity to other tumour deposits.23 These effects can be active either over a short period of time through activation of non-specific immune effector mechanisms22,24 or can set the foundations of long-lived, T cell-mediated immunity to tumour cells.17,25 The most effective way to stimulate T cell-mediated immunity to tumours is to kill cells in such a way as to convince the immune system that they are the source of an aggressive, infection-like situation.26 This mimics a pathological challenge against which the immune system is highly evolved to react. This contrasts with the scenario in which the chronic growth of tumour cells in a patient has, more likely than not, managed to ‘tolerise’ the immune system to the very cells which we would like it to attack.27 Thus, exploitation of mechanisms that combine more powerful direct local killing with local and systemic immune activation will be a major priority. As well as the isolation of more potent versions of pre-existing genes such as HSVtk,28 significant advances will come from the discovery of new genes which both kill more cells locally and which do so with as much immunological aggression (‘Danger’) as possible (Figure 1).23,26 VectorsIf the jury is still out on the choice of therapeutic gene, there seems to have been a much more definitive selection of vector type over the past few years. To date, cancer gene therapy trials have variously used the three most common vectors (plasmid, retrovirus and adenovirus). However, except for the situation where tumour/ immune cells are manipulated ex vivo, there will be a clear preference in the coming years for the use of adenoviral vectors for in vivo delivery to tumours.29 Dominant in this selection process is the high titre of adenoviruses (>1011 p.f.u./ml) compared with other vectors. Given the requirement to kill, rather than correct, target cells, there is generally little need for integration (ie long-term gene expression as provided by C-type and lentiviral vectors). The initial rationale of the use of C-type retroviral vectors to target exclusively dividing tumour cells on the background of a quiescent tissue is being gradually superseded by the realisation that human tumours generally cycle much more slowly than the rodent cell lines on which this strategy was based.2 Hence, the trade-off between the total numbers of cells that can be productively infected by an adenovirus, compared with the loss of a potential targeting advantage using C-type retroviruses, clearly favours the more efficient adenoviral system. In addition, the immunogenicity associated with adenoviral vectors probably offers an added ‘adjuvant’ bonus in the context of most cancer protocols. As a result, the number of direct in vivo delivery protocols will continue the escalation of the use of adenoviruses. Of course, the first vector system that comes through with a targeted particle that works in a systemic application may be able to win the initiative back again.7 However, even the highest titre system is clearly not high enough yet to cure even local tumours. Therefore, there is a clear need to explore, and exploit, a range of alternative options. Other systems, such as AAV and HSV, are already well developed for use in other gene therapy contexts and may be valuable in certain conditions within the cancer arena.30 But three areas of intense activity will soon be (1) the development of replicating viruses, (2) the combination of components of different vectors into hybrids with the beneficial properties of different systems and (3) the investigation of novel viral and non-viral delivery systems which have not been explored to their best potential (Figure 1). The development of replication-competent vectors for cancer gene therapy is the inevitable consequence of data from the early clinical trials. So far, a substantial therapeutic gap still exists between the overlap of the efficacy provided by, on the one hand, the potency of the therapeutic gene(s) and, on the other, the efficiency of gene delivery provided by the vector (Figure 1). Only when these two ‘therapeutic domains’ approach each other will clinical efficacy result. Therefore, a natural solution to closing this gap is to use viruses that can replicate in tumour cells to enhance gene spread.31 The trailblazer of recombinant replication-competent viruses has been the ONYX-015 virus that preferentially replicates in cells lacking functional p53 due to a deletion in the adenovirus E1B gene.6,11 Phase I and II clinical trials have proved safe enough to allow the virus to go forward for phase III clinical trials – despite some uncertainties as to what exactly is the dependence on p53 status for viral replication.32 ONYX-015 has no therapeutic transgene itself and relies on the lytic ability of replicating adenovirus to kill cells directly. However, the therapeutic indications from the early trials have again shown that more potency is required over and above that produced by viral lysis of tumour cells and the future development of this system will depend on the incorporation of therapeutic genes into the replicating framework.33 As well as other viruses that use intrinsic properties of the transformed phenotype as the basis for tumour-selective replication of a virus,34 other replication-competent viruses will enter clinical trials based on the use of tumour/tissue-specific transcriptional regulatory elements to drive expression of the critical viral genes required for replication.35 With a strongly established literature on the use of tissue/tumour-specific promoters within recombinant vectors36,37 this is a field which is set to proliferate rapidly within the coming years. An intermediate step on the way to fully replication-competent viruses is the development of hybrid vectors which ‘mix and match’ elements of established vector systems. Thus, adenoviruses have been used to convert target cells in vivo into retroviral producer cells.38 Similarly, a hybrid between adenovirus and EBV has been described which allows the high titre of adenoviruses to be combined with long-term persistence without integration through the maintenance of a stable EBV episome.39 In this way, factors such as persistence of expression, titre and immunogenicity of vectors may be controlled more closely than the properties of individual vectors alone can allow.37 Finally, the increasing awareness of the potential of gene therapy means that a variety of different viral systems, previously not thought of in the context of vector development, will undoubtedly be developed.7 Of the large number of viruses that are already well characterised virologically, many have potential for exploitation as vectors. For any given virus, this may either be in its entirety as a novel vector or it may involve cannibalisation of specific components that can be incorporated into other (hybrid) vectors. For example, the VP22 protein of HSV can increase protein distribution between non-transduced cells – thereby enhancing the potential bystander effects.40 The potential for both well-characterised, as well as lower profile, viral systems to contribute to the recombinant, hybrid and replicating vector pool of the new millennium is, therefore, very great. TargetingA genuine ability to target delivery systems to tumour cells distributed widely throughout the body of a patient would simultaneously increase real titres and efficacy, and decrease potential toxicity. In truth, no such systemically targeted vectors exist yet. Injection of vectors into the bloodstream for the treatment of cancer requires not only that the vectors be targeted (to infect only tumour cells) but also that they be protected (from degradation, sequestration or immune attack) for long periods of time so that they can reach the appropriate sites for infection. Moreover, having reached such sites, the vectors must be able to penetrate into the tumour from the bloodstream before carrying out their targeted infection. Progress in vector targeting has been dramatic in the last few years. Surface targeting would be optimal to prevent non-productive binding and sequestration of vectors before they reach their target cells. It is now possible to activate infection through retroviral envelope binding only in tissues which express, for example, tumour-associated proteases41 and surface targeting is now also possible for adenoviral vectors.42 The advent of in vivo selection of peptide ‘addressin’ sequences to target tumour cells or vasculature will add greatly to the technology required to target delivery at the level of cell binding.43 The challenge will be to show that such addressin peptides can be efficiently and functionally incorporated into vector systems such as viral envelopes. Transcriptional targeting is perhaps more established as a method of limiting gene expression to target cells.44 However, this approach complements more the decreased toxicity of gene therapy rather than contributing to its increased efficiency: a transcriptionally targeted vector still has no means of preventing its sequestration by the mass of non-target cells/tissues which it is likely to encounter before it finds its real target. So, despite impressive advances in both promoter design and envelope modification, it still remains to be shown that any of these systems can be used for genuine systemic delivery in which vectors last long enough, and arrive in high enough quantities, at the tumour sites to be effective. Therefore, the design of most new in vivo trials will necessarily have to be based around local delivery, where targeting comes principally through the location of the catheter or syringe needle. As the refinements of vector targeting become more sophisticated, vectors will become available for testing in loco-regional protocols, escalating up through organ/limb perfusion and, eventually, to genuine systemic delivery. However, given the reluctance of the human immune system to permit free circulation of viruses/vectors for any period of time, achieving this latter goal promises to be one of the most difficult of all the challenges to solve. Anti-angiogenic gene therapy – going for the jugularGiven the difficulties in generating truly targeted vectors for systemic delivery, the alternative is to target those biological properties of tumours which set them apart from all, or most, normal tissues. One of the most notable distinguishing features of tumour growth is the absolute requirement for the tumour to provide itself with an expanding blood supply through the process of angiogenesis, and there is a wealth of targets at the interface between the malignant population and the supporting stroma that could be exploited by approaches including gene therapy.45 For instance, the migration of tumour endothelium can be inhibited by interfering with protease enzyme function46 and it may even be possible to design molecules with both antiangiogenic activity and tumour-homing properties.43 The identification of naturally circulating factors such as angiostatin and endostatin, which appear to be capable of suppressing angiogenesis has sparked an explosion in efforts to deliver and express such recombinant molecules19,47,48,49,50 and more candidates are being reported all the time. Immuno-gene therapyIn principle, the immune system provides exactly what we would like the ideal vector system inherently to provide: (1) an amplification of the therapeutic potential following relatively low level gene delivery, and (2) high level specificity of body-wide target cell killing once correctly activated. The possibility of harnessing these two features to fight metastatic cancer is the reason why the majority of cancer gene therapy protocols have been aimed at immune stimulation (although the cynic might also point out that in the absence of better targeted vectors there has been little other choice!). Many of the first clinical protocols for cancer gene therapy involved the ex vivo modification of freshly isolated tumour cells with cytokines.10 These trials grew out of safety considerations and a reluctance to deliver viruses directly into tumours in situ. However, in the presence of the appropriate in vivo controls, it became apparent that in many cases, cytokine modification may be little better than more conventional adjuvant-based cancer cell therapies with no gene transfer component.51 In addition, the recovery of patient tumour cells and maintenance in culture for long enough to allow transduction with cytokine genes has proved to be laborious and expensive and may significantly alter the phenotype of the cells. None the less, once clear advantages had been shown in animal systems for some cytokines,52,53 clinical trials have shown encouraging signs that cytokine-modified vaccines can generate significant immune responses in patients with minimal toxicity.3 However, given the effort and money involved in the autologous cell gene modification approach it is unclear whether these sorts of approaches will ever offer a useful return on the investment. Perhaps the two most significant areas in which immuno-gene therapy is likely to progress are in the the molecular identification of tumour-associated antigens and exploitation of the central significance of the dendritic cell in generating anti-tumour immune responses. One of the most spectacular advances during the evolution of gene therapy for cancer has been the cloning of tumour-associated antigens from human tumour (usually melanoma) cells which are recognised by either CD8+,54 or more recently, CD4+ T cells.55,56 This has added molecular credibility to the long-held presumption that tumours can indeed express antigens against which T cell-mediated responses can be raised. The tumour vaccination field can now move from the relatively crude level of whole cell vaccines into the molecular arena with defined targets with which to immunise. One cautionary note should be sounded – tumours are highly heterogeneous and unlikely to express only one dominant antigen on all of the cells. Therefore, the trend towards molecular vaccination with defined antigens will undoubtedly have to employ ‘cocktail’ approaches where multiple cDNAs are used in the vaccination protocol. With such antigens in hand – even if only in a limited number of tumour types at the moment, the question remains of how to break tolerance to these antigens which are often not even mutated forms of self antigens. Several key studies have shown that tolerance to tumours can be broken as long as tumour antigens – whether clearly defined or not57,58,59 – are delivered into suitably activated dendritic cells.60,61 Indeed, many of the large numbers of gene transfer experiments which have successfully raised anti-tumour immunity may be attributable simply to the induction of tumour cell killing in vivo, leading to transfer of antigens into professional antigen presenting cells (APC).27,61,62 The efficacy of the resultant anti-tumour responses is likely to be influenced heavily by (1) the efficiency with which DC can be attracted to tumour sites, (2) the availability of tumour antigens to be taken up by DC, and (3) their subsequent maturation/activation and ability to traffic to the lymph nodes to present these antigens.27 All three of these factors will be encouraged if the tumour cell killing occurs in the presence of pro-inflammatory signals which persuade the immune system to associate a pathological event with the killing.26 Thus, tumour cells which are killed at low levels by purely apoptotic means are unlikely to (1) attract DC, (2) load the DC or (3) activate the DC. In contrast, cells which die by non-apoptotic means,23 similar to pathogenic infections, or which die such that the levels of apoptotic death overwhelm the local phagocytic capability to clear them,59 will stimulate all three of the above. Therefore, we now have the ability to identify the key molecules that can serve as targets for immune responses and to isolate and genetically manipulate the central APC involved in antigen presentation. Coupling these two together means that dendritic cell modification is likely to supplant tumour cell manipulations in the coming years. It must always be remembered, however, that where immuno-gene therapy for cancer hopes to end up is exactly the point from which gene therapies for autoimmune disease are starting out (Figure 2). Oncologists using immuno-gene therapies seek to induce autoimmunity to a particular class of cells within the patient – their tumour cells. The induction of autoimmunity to some of these tumour-associated antigens has already been shown and correlates well with the generation of anti-tumour immune responses.63,64,65,66 However, the same or related antigens may be displayed on other cells or tissues within the body, differing only in relative expression level, and hence may also represent a target for recognition and destruction. Such side-effects can be tolerated for the treatment of tumours where the normal tissue type is not crucial to survival of the patient (such as melanocytes). However, as immunotherapies for the common cancers of more ‘important’ tissues develop, the successful induction of autoimmunity to tumour antigens may be accompanied by the potentially disastrous destruction of uninvolved self tissues. The first goal must be the attainment of anti-tumour immune responses; the second will be to learn how to restrict them to tumour, rather than normal tissues of the same type. Gene therapy in the clinic – knowing its placeA key factor in ensuring the success of gene therapy will be to develop a clear understanding of how it can best play its role in the clinic. For example, immuno-gene therapies are only ever likely to be effective in clinical situations where patients are at, or have been returned to, a state of low tumour burden and still have effective, functioning immune systems.67 A consequence of this is that as the early phase I/II trials move ahead in to phase III and IV, it will take some considerable time, and large numbers of patients, to demonstrate their true efficacy. Moreover, gene therapy is likely to be very effective in combination with pre-existing clinical regimens, such as chemotherapy and radiotherapy.6 A large number of studies are now showing great potential for collaboration between gene therapy and pharmaceutical, immunological and radiotherapeutic disciplines to kill cells more effectively and in greater numbers. It is unlikely, therefore, that gene therapy alone will play a curative role in cancer treatments for some years. Its importance as a supporting player is much more likely to establish its sense of worth in the minds of clinicians over the coming years, thereby setting the scene for its ultimate use as a frontline therapy in its own right. Gene therapy and the free market – putting it all togetherGene therapists aspire to creating the Perfect Vector which, for cancer gene therapy, is likely to be rather more multi-component than in other disease situations. It will probably consist of a coat with molecules that allow tumour cell-specific binding; mechanisms to permit transfer of the genes to the nucleus following penetration of the cell membrane; promoter/enhancer elements which target high levels of expression only in the appropriate cells and, of course, the gene(s) that will ultimately lead to the death of the tumour cells directly and/or indirectly through immune stimulation (Figure 3a). Each of these components should be perfected to give the best possible combinations and therapy for the patient. However, even if it were clear what should be put into this perfect vector, will it be commercially possible to assemble it? Few, if any, laboratories can hope to optimise each component from their own endeavours. Therefore, each of the constituent elements (envelopes, promoters, genes, viral systems) will inevitably be the intellectual property of different companies or institutions (Figure 3b). What are determined conceptually as the optimal combinations may, in the harsh reality of the free market, at best take long periods of time to co-ordinate and, at worst, never be possible to bring together into one final product. In addition, as ownership of intellectual property begins to dominate what companies are, or are not, prepared to support, finding the funding (and desire) to put together the best possible finished product is going to become increasingly difficult (Figure 3b). It is highly probable that intellectual property priorities are already preventing the initiation of projects which might be the optimal therapeutic priorities, because ownership of certain components of the target vector are not commercially compatible. It will be important to see if the commercialisation of gene therapy can truly complement, rather than corrupt, its therapeutic promise in the new millennium.
|