INTRODUCTION

TOP ABSTRACT INTRODUCTION GENE TRANSFER TO THE... CONCLUSION REFERENCES

Corneal Grafting

Corneal transplantation is the only form of treatment available to reverse the loss of vision caused by damage to the cornea (1). Such damage can be caused by infection, for example by herpes simplex, keratoconus (a condition in which the cornea becomes conical in shape), or by failure of the endothelial layer of the cornea. In addition, there are congenital defects that are seen in children. Corneal transplantation is the commonest form of solid organ transplantation, with about 40,000 a year performed in the United States alone (2).

Corneal transplantation has always been considered to be successful, and is typically carried out with no systemic immunosuppression or HLA matching. The reason for this success is that the cornea and anterior chamber of the eye form an immunologically privileged site, in other words the immune system shows a reduced ability to react against foreign antigens in the cornea or anterior chamber (3). There are several mechanisms that are responsible for this privilege. These can be divided into those that lead to an ignorance and/or isolation of the foreign antigen and those that actively suppress or deviate the immunological response. The former mechanism was recognized by Medawar, who demonstrated that skin transplanted into the cornea was poorly rejected (4). These data suggest that it was the anatomical location of the cornea that was responsible for the immunological privilege. Vascularization of the cornea reduces the extent of the privilege, leading to the concept that privilege is due in part to the immune system failing to see the antigen. This leads to a blockage in the afferent arm of the immune response. Factors that lead to this ignorance and isolation include the lack of a blood supply, poor lymphatic drainage, and the paucity of dendritic cells in the central cornea, as reviewed in Niederkorn (5).

Active suppression and deviation of the immune response in the cornea is caused by a number of mechanisms. These include the expression of Fas ligand (CD95L) on cells in the anterior chamber, which interact with Fas (CD95)-expressing activated lymphocytes, inducing apoptosis (6). There is also active suppression of the immune response in the eye, termed anterior chamber-associated immune deviation (ACAID) (5). ACAID describes the experimental finding that if an antigen has been injected into the anterior chamber of the eye, then the animal develops a systemic reduction in cellular response on subsequent challenge with the same antigen. Many of the mechanisms responsible for ACAID have been elucidated; the key components seem to include the exposure of antigen-presenting cells in the anterior chamber to cytokines and hormones such as transforming growth factor ₂ (TGF-₂) and  melanocyte-stimulating hormone (-MSH).

Rejection of Corneal Grafts

It should be noted that immune privilege should be a relative term, not an absolute term (9). It is possible to produce immunological responses in the cornea and anterior chamber of eye and, while corneal grafting has been highly successful, there is a significant failure rate. The main reason for such failure is immunological rejection of the graft. At present graft survival is about 75% at 4-5 yr, a figure not dissimilar to survival of cardiac and renal grafts (10). The survival figures for corneal grafts have not changed much over the last few decades, while of course the figures for other grafts have improved greatly. So the widespread assumption among immunologists that corneal transplantation "is solved" needs revising.

In addition, there are high-risk patients, whose grafts have a much higher risk of rejection. These include recipients of previous grafts, patients with vascularized or inflamed graft beds, and children. These groups can show poor graft survival, for example, in one study of corneal grafting in infants only 66% of first grafts survived 1 yr, and the figure for subsequent grafts was worse (R. M. Comer, S. Daya, and M. O'Keefe, unpublished data). These patients provide a strong impetus for the development of novel strategies aimed at preventing graft rejection. It should be noted that the existence of a group of patients with poor prognosis provides an ethical opportunity for the testing of new forms of treatment. Equivalent groups of patients are unlikely at present to receive cardiac or renal grafts because of the shortage of organs.

The rejection of corneal grafts can involve one of three major mechanisms; epithelial, stromal, or endothelial rejection (1). Epithelial rejection involves the outer layer of the cornea (see below for anatomical description). In stromal rejection leukocytes migrate through the stroma of the cornea. In endothelial rejection the leukocytes pass through the anterior chamber and adhere to the posterior (endothelial) surface of the cornea, leading to destruction of the endothelium. Endothelial rejection is normally considered the most serious type of rejection, as the endothelial cells in the human are essentially nonreplicative, and so any damage sustained is irreversible. In addition, as is described below, they have an essential function in maintaining the clarity of the cornea. The epithelial layer is capable of regeneration (and will eventually in any case be replaced by recipient cells), and so rejection at this site is clinically less important. Stromal rejection is less common, is often associated with minor symptoms, and is not normally accompanied by aqueous inflammation (1).

One feature of corneal rejection that is of interest is the nature of the allogeneic recognition. There are two ways in which allospecific T cells can recognize allogeneic cells: the direct and indirect pathways (11). Direct recognition involves the T cells recognizing the donor-derived allogeneic MHC molecule directly (Figure 1A). In humans and animals there is a high frequency of T cells that can participate in direct allorecognition, and so this pathway is responsible for the strength of a primary alloresponse (as seen, e.g., in a mixed lymphocyte reaction). Indirect recognition is essentially the same as the recognition of any foreign antigen by specific T cells; donor-derived allogeneic MHC molecules or other polymorphic proteins are presented by recipient antigen-presenting cells in the context of recipient autologous MHC molecules (Figure 1B). Indirect pathway responses are thought to be important in the chronic rejection of allografts. In addition, they will dominate in settings where there are not enough donor-derived dendritic cells to initiate the direct pathway alloresponse. This includes corneal transplantation. The importance of the indirect pathway, together with the reduced expression of MHC antigens on corneal grafts, may explain why in rodent models minor histocompatibility antigens are more significant than major histocompatibility antigens in determining graft survival (12).

View larger version (24K): [in this window] [in a new window]   View larger version (20K): [in this window] [in a new window]   Figure 1.   Direct and indirect pathway of allorecognition. In the direct pathway (A) of allorecognition the T cell receptor (TCR) on the recipient T alloreactive T cell (shown in black) directly recognizes peptide antigen presented in the context of donor allogeneic MHC molecules (shown in gray), on donor-derived antigen-presenting cells (APCs) (or other donor cells). In indirect presentation (B) allogeneic antigens (including MHC molecules) are released from the donor cells (e.g., by death or shedding), these are then taken up by recipient APCs, processed, and presented in the context of recipient MHC molecules. The resulting peptide-MHC complex consists of allogeneic peptide (gray) in autologous MHC (black). This is then recognized by T cells showing indirect allospecificity.

Corneal Anatomy

The cornea is composed of three major layers: the epithelial, stromal, and endothelial layers (13) (Figure 2). The epithelium forms the anterior surface of the cornea, and is six or seven layers thick. It is continually regenerating, with stem cells located in the limbal region (in penetrating keratoplasty the limbus is not included in the donor button, although limbal transplants are indicated in some conditions). In humans the epithelium lies on top of the Bowman's layer, which acts as a basement membrane. The stromal layer makes up approximately 90% of the cornea and contains collagen fibers that are precisely arrayed so as to be transparent. This layer contains some scattered keratocytes.

View larger version (96K): [in this window] [in a new window]   Figure 2.   Cross-section of cornea. This photograph shows a cross-section of a rat cornea with the epithelium on the top (anterior surface), the stroma (containing ketatocytes) in the middle, and a monolayer of endothelial cells on the bottom (posterior surface), lying on top of Descemet's membrane.

The endothelium is a monolayer, which lies on Descemet's membrane (Figure 3). The main physiological function of the endothelial cells is to pump water out of the stromal surface by use of Na⁺, K⁺-ATPase pumps. If this function is not maintained the stroma swells because of the accumulation of water, and the precise orientation of the collagen fibers is disrupted, resulting in opacity. In humans the corneal endothelial cells are arrested in the G₁ stage of the cell cycle, and are essentially nonreplicative (14). Soon after birth the total number of endothelial cells is fixed (~ 3,000-4,000 cells/mm²) and it slowly reduces so that by middle age it is 2,500/mm² and by old age 2,000/mm² (13). If the number of endothelial cells falls below ~ 800/mm² then the pump function of the layer cannot be maintained and there is swelling and consequent opacity of the cornea (13). It should be noted that the corneal endothelium is distinct from vascular endothelium, being of neural crest origin.

View larger version (108K): [in this window] [in a new window]   Figure 3.   Corneal endothelial cells. Specular microscopy of human corneal endothelial cells in situ, from a healthy volunteer. The endothelial cells form a roughly hexagonal array.

	GENE TRANSFER TO THE CORNEAL ENDOTHELIUM

TOP ABSTRACT INTRODUCTION GENE TRANSFER TO THE... CONCLUSION REFERENCES

The main approach that we have investigated for modulating allograft rejection is gene therapy. The cornea makes an excellent model for gene transfer as it has a relatively simple anatomy, the effects of the gene transfer can be monitored in vivo, and perhaps most importantly, the cornea can be maintained in culture for long periods. Thus, in eye banks, corneas can be routinely stored for up to 1 mo before transplantation. This simplifies the experimental procedure as it allows assessment of the effects of gene transfer in vitro.

Adenoviral Vectors

In our initial experiments we concentrated on using adenoviral vectors to mediate gene transfer. Adenoviruses are double-stranded DNA viruses that are capable of infecting nondividing cells. Most adenoviral vectors are based on adenovirus serotypes 2 and 5 (Ad2 and Ad5).

Adenoviruses infect cells in a multistage process. First, the fiber protein interacts with receptors on the cell surface (coxsackie-adenovirus receptor [CAR]) (15, 16). The penton base protein then binds to _v integrins present on the cell, promoting internalization into endosomes (17). Acidification of the endosome then promotes virus escape from the endosome into the cytosol, and subsequent transport to the nucleus.

A variety of vectors have been based on the adenovirus. The development of these vectors has been a process of first removing genes essential for viral replication (hence disabling the virus and rendering them safe), and then suppressing production of viral proteins (thus reducing the immunological response and any other undesired effects of viral gene expression). The latest vectors (often termed "gutless adenovirus") contain only the minimal cis-acting DNA sequences needed for replication and packaging of the vector DNA (18). As a result there is no viral gene expression in infected cells. It is likely that these cells will be essentially nonimmunogenic. However, all the work that we have performed concentrates on earlier generations of adenoviral vectors that contain deletions in the E1A, E1B, and E3 regions of the viral genome.

Our first set of experiments used the marker gene lacZ, which encodes the bacterial enzyme -galactosidase. This allows ready identification of cells expressing the protein. In initial experiments we incubated rabbit corneas with adenovirus containing lacZ under control of the cytomegalovirus (CMV) promoter (19). High levels of -galactosidase expression were seen in the corneal endothelium, with essentially no expression in the stroma or epithelium (some expression was seen at the cut edge of the stroma).

The expression was transient, peaking on Day 1-3 and falling off to undetectable levels by Day 21 (Figure 4). The gene transfer had no detectable effect on the behavior of the cornea in vivo, as indicated by no increase in the rejection rate of gene-modified corneas when compared with unmodified corneas and a normal deturgescence of the cornea after transplant (19). The latter could be determined by ultrasonic pachymetry, in which the thickness of the cornea is monitored; the cornea swells during storage and then thins out after grafting as the endothelial cells pump water out of the stroma. It should be noted that these in vivo experiments were all carried out in a low-risk setting, in which a cornea from an outbred New Zealand White line of rabbits was transplanted into an avascular corneal bed from the same line. Supraoptimal concentrations of adenovirus can cause toxicity (data not shown).

View larger version (14K): [in this window] [in a new window]   Figure 4.   Time course of expression of -galactosidase in rabbit corneas. Rabbit corneas were incubated with adenovirus containing the lacZ gene. They were then incubated for various times, and the amount of -galactosidase enzyme was determined by a colorimetric assay. Maximal expression is seen on day 3, and falls to undetectable levels by day 21. (Data taken from Reference 19 [Larkin, D. F. P., H. B. Oral, C. J. A. Ring, N. R. Lemoine, and A. J. T. George. Transplantation 1996;61:363-370], where further details of the experimental procedure can be found.)

Subsequently we have shown that adenoviral gene delivery is effective in human and rat corneas (20, 21). In the rat we found that there was rapid overgrowth of the epithelium (within 24 h), which would grow over the cut edge of the stroma and then enclose the endothelium by 72 h (20). This would block gene expression, and so we routinely remove the epithelium from rat corneas for in vitro applications. In all species tested the kinetics of expression of -galactosidase were broadly similar. Similar data concerning the feasibility of adenoviral gene delivery to the cornea and the anterior chamber have been obtained by other groups (22).

There are several unanswered questions that arise from these preliminary studies. The first concerns the location of gene expression. As described above, expression is seen only in the endothelial layer, with occasional low-level expression near the cut edge of the stromal layer. It is perhaps not surprising that no expression is seen in the keratocytes within the body of the stroma: there will be little access of virus to this site. However, in our gene transfer studies no attempt was made to prevent access of the virus to the epithelial layer. Wild-type adenovirus is capable of infecting the epithelium; adenoviral infections of the epithelium can be a serious clinical condition (25). In addition, we have shown that the epithelial cells express the _v intregrins needed for viral internalization (26). It is possible that there is a physical barrier to epithelial infection by the virus, or that the epithelial cells are not capable of expressing the construct. Work is being done to determine how to make the epithelium permissive for expression from these constructs.

The second question concerns the transient nature of gene expression. Adenoviral DNA does not integrate into the host genome, but remains episomal. In most systems studied loss of expression over time has been attributed to either an immunological response against the virus (or expressed protein) or to dilution of the nonreplicating adenovirus in dividing cells. However, in our situation the expression is monitored ex vivo, in the absence of any immune response. Furthermore, human endothelial cells do not replicate (and rabbit endothelial cells are virtually nonreplicative). In the human we have looked for persistence of virus-derived DNA (both lacZ gene and E2A segment of the viral genome) by polymerase chain reaction (PCR) and expression of -galactosidase by reverse transcription (RT)-PCR. As expected, the transcription of the lacZ gene was transient, following the levels of protein. However, virus-derived DNA persisted up to Day 56 (the last time point studied), long after protein expression was lost (21). These data have been obtained using both the CMV and Rous sarcoma virus (RSV) promoters to drive -galactosidase expression. These data indicate that loss of expression is not due to loss of DNA, although it is possible that they are due to sequestration or modification of the DNA in some way. It should also be noted that we see longer term expression (up to 30 d, although considerably less than the peak levels of secretion) of genes that encode soluble proteins (such as CTLA4-Ig [cytotoxic T lymphocyte-associated antigen 4-Fc fusion protein] and TNFR-Ig [tumor necrosis factor receptor-Fc fusion protein]) (21) (and our unpublished observations). This might suggest that the nature of the expressed protein can affect the time course of expression; however, differences in the turnover of different proteins and sensitivity and nature of the assay systems used make it difficult to draw any firm conclusions from these data.

Other Viral Vectors

The major drawback of the adenoviral vectors that we have used is their immunogenic and inflammatory nature (23), although this will be less of a problem with the gutless vectors that are under development. As described above, we and others observe only transient expression with this vector system. We have therefore looked at a number of other viral systems, including adeno-associated virus and herpesvirus.

Adeno-associated virus is a single-stranded DNA virus that can transduce both dividing and nondividing cells (27). The virus genome is flanked by terminal repeats, which are necessary and sufficient for encapsulation and integration of the viral genome. As a result no viral genes need to be included in the vector. Production of the virus is dependent on the presence of a helper virus, normally adenovirus. On infection adeno-associated virus integrates into the human genome at a specific site on chromosome 19 (28). Such site-specific integration would be advantageous for gene delivery; however, the current adeno-associated viral vectors do not retain this targeting capacity, showing random integration. We are able to obtain long-term expression with adeno-associated viral vectors (29). Indeed, this is the only vector system by which we are able to obtain long-term expression. However, it has proved difficult to produce high yields of virus, and so the overall level of gene expression is low. Improved production techniques are now enabling the production of high titers of adeno-associated virus, and this vector probably merits further investigation in the context of corneal transplantation.

Herpesvirus vectors are attractive as the virus is capable of long-term latent infection of cells. However, herpesvirus is cytopathic. To that end constructs have been created in which genes that are responsible for the toxic effect are removed. We have viruses containing such deletions. However, although they appear to be highly efficient at transducing corneal endothelial cells, they kill the cells rapidly (29). Use of this vector system in the context of corneal endothelium will have to await the development of improved herpes vectors.

Nonviral Vectors

The alternative is to use nonviral vectors. They have several theoretical advantages, including a relatively low immunogenicity, the ability to target expression to certain cell types, and reduced safety issues in the production and administration of the therapeutic agent (30). They will probably be more amenable to large-scale industrial production than viral vectors. In addition, they have the considerable advantage for experimental work that it is not necessary to generate appropriate recombinant viruses, which can be a lengthy and laborious process. Instead, the gene of interest need only be cloned into a simple plasmid vector, allowing many different constructs to be rapidly evaluated. Their main drawback at present is that they are considerably less efficient at mediating gene transfer than their viral counterparts.

There are several classes of nonviral vector, including naked DNA (with or without particle bombardment to increase uptake into the cell), liposomes, cationic polymers, and various forms of receptor-mediated gene transfer. In the context of corneal endothelium, we have experience with the last three forms of gene transfer. In addition, we initiated studies in this area by developing lipoadenofection as a "half-way house" between viral and nonviral gene transfer.

Liposomes and lipoadenofection. Liposomes work as vectors by forming electrostatic interactions between the plasmid DNA and cationic groups on the lipid molecules. The resulting complex forms electrostatic interactions with the cell surface, is then internalized, and enters the cell cytosol. There are several stages in the process that are inefficient, and a number of groups have shown that complexes of liposomes and adenovirus can enhance gene delivery of plasmid, when compared with liposomes alone (31). The viruses may be promoting binding to the cell surface through the appropriate viral receptors, allowing internalization of the virus (e.g., through interactions between the penton base protein and _v integrins in the case of adenovirus), facilitating escape from the endosomes, and promoting transport of the DNA to the nucleus.

We have shown that liposomes are inefficient at gene delivery (32). However, mixtures of liposome and adenovirus mediate more efficient gene transfer (32). The distribution and kinetics of gene expression are similar to those seen with adenovirus alone. In addition, we demonstrated that it is possible to deliver a construct in which the gene is under the control of an inducible promoter. We used the E-selectin promoter, which can be activated by proinflammatory cytokines such as TNF, and showed that gene expression was seen only in the presence of TNF (32). TNF is present at high levels during corneal graft rejection (33), and so this approach offers the possibility of restricting gene expression to when it is needed. Obviously a similar strategy could be developed using a drug-inducible promoter system.

Clearly lipoadenofection is a form of "virus-enhanced" gene transfer, and is not truly nonviral. However, it shows some of the advantages of nonviral gene transfer, including the ability to use plasmid DNA rather than cloning the gene of interest into viral vectors. The immunogenicity of the virus might be a concern; however, UV irradiation of the virus, sufficient to prevent expression of viral genes, does not abolish the ability of the virus to assist gene transfer (Figure 5).

View larger version (29K): [in this window] [in a new window]   Figure 5.   Effect of UV irradiation on lipoadenofection. Expression of viral genes will greatly contribute to the potential immunogenicity of the virus-liposome-DNA complex used in lipoadenofection. To determine whether the viral DNA could be destroyed, while preserving the ability of the virus to assist in gene delivery, lipoadenofection of an endothelial cell line was performed essentially as described (32), except that the Rad35 virus containing the lacZ gene was used (19). In addition, the virus was irradiated with UV light (0 [control] to 3,980 cGy). The DNA that was delivered contained the gene encoding chloramphenicol acetyltransferase (32). After adenolipofection the amount of -galactosidase and CAT enzymes was determined. Data is expressed as a percentage of control levels. As can be seen, UV irradiation of the virus was effective at abolishing expression of -galactosidase. Some loss of CAT expression was seen when irradiated virus was used; however, when compared with the -galactosidase, it can be seen that the use of irradiated virus would permit gene transfer by lipoadenofection, without the expression of virus-derived proteins.

Cationic polymers. Cationic polymers work in a manner similar to liposomes, forming complexes between the negatively charged DNA and the positively charged polymer. These bind to the cell surface and are internalized. These molecules probably encounter the same bottlenecks as liposomes, involving binding to the cell surface, efficient internalization, endosomal escape, and transport to the nucleus. We have investigated the use of dendrimers for gene transfer.

Dendrimers are spherical, hyperbranched macromolecules possessing a large number of terminal groups (34). Dendrimers are synthesized via a repetitive and controlled three-dimensional procedure, which yields highly branched macromolecules. As a result of this control, dendrimers can be considered to be completely monodispersed spherical polymers. Many possible and important applications have been postulated, including catalysis, imaging, drug delivery, and molecular machines, and some of these applications are beginning to be realized. In the context of gene delivery dendrimers have been shown to be capable of mediating gene transfer to a variety of cell types (35). This is mediated by the electrostatic interaction between the positively charged amine groups on the surface of the dendrimers with DNA (39), resulting in complexes that can be visualised by electron microscopy (39, 40). These bind to the cell surface, through electrostatic interactions, and are then internalized. The dendrimers buffer the endosomes, thus protecting the DNA (36). A further advantage of dendrimers for in vivo applications is that they protect the DNA from the action of DNase found in serum (39).

There are several types of dendrimer. These can vary by virtue of the chemical groups used to make the branched structure, the nature of the branching (symmetrical or asymmetric), the terminal groups displayed by the molecule, and the number of generations (i.e., the number of branches). We have used a polyamidoamine dendrimer that has been "activated" by heat treatment in a solvent to trim the structure. Such treatment improves the flexibility of the molecule and increases the efficiency of gene transfer (38). This dendrimer was provided by a commercial collaborator (Qiagen, Hilden, Germany), which is now marketing the product (termed Superfect) for in vitro gene transfer. We have shown that it is possible to use these activated dendrimers to mediate gene transfer to the corneal endothelium (41). We have not seen any toxic effects of the gene transfer at the optimal doses of reagents. In addition, transfer is possible in conventional medium containing fetal calf serum. Similar polyamidoamine dendrimers have also been used for in vivo gene transfer by other groups to cardiac allografts (42).

Receptor-mediated gene delivery. One of the drawbacks with both the liposome and the polymer approach to gene delivery is that targeting to the cell surface relies on electrostatic interactions. This means that delivery is essentially nonspecific (although the nature of the liposome or polymer may have an influence on the delivery), relatively inefficient, and also targeted at cell surface structures that may not be efficiently internalized into endosomes. For this reason receptor-mediated gene delivery is highly attractive. In this approach the DNA- vector complex is targeted to a particular target molecule on the cell surface. This has the potential for specific delivery to particular cells, and also delivery to molecules that are optimal for gene delivery. We have investigated two targeting approaches to receptor-mediated gene delivery, using peptides to deliver genes to integrins and ligands (transferrin) to delivery the genes to the appropriate receptor (transferrin receptor).

For targeting to integrins we have used peptides, developed by Hart and colleagues (43), that contain 16 lysine residues (to bind to DNA) fused to cyclic peptides that bind to integrin molecules. The integrin molecules are good targets for this approach as they are involved in the entry of a number of viruses into cells during infection. When the peptides are mixed with plasmid DNA they form complexes (44). These have been shown to be inefficient at mediating gene transfer unless the cells are treated with chloroquine (45). This suggests that the limiting step is endosomal escape. When liposomes are included in the complex the transfection efficiency increases. When applied to corneas these complexes formed between the peptides, liposomes, and plasmid DNA can mediate gene delivery to the corneal endothelium (44, 45).

Another target molecule that is attractive for gene delivery is the transferrin receptor. This rapidly recycles through acidic compartments. Transferrin has been used in combination with liposomes to improve gene delivery (46). We have shown that this approach can be used to obtain gene expression in the corneal endothelium (49). The levels of expression seen are higher than with the activated dendrimer.

	CONCLUSION

TOP ABSTRACT INTRODUCTION GENE TRANSFER TO THE... CONCLUSION REFERENCES

In this article we have concentrated on looking at the various methods we have employed to achieve gene transfer to the corneal endothelium. We have not discussed our studies relating to the nature of the rejection response, or to the types of genes that we are using in an attempt to modulate this rejection. These include molecules designed to induce tolerance (CTLA4-Ig [21]), deviate the immune response (vIL-10 [49]), frustrate effector mechanisms (TNFR-Ig [33]), or protect corneal cells from damage (catalase [50]). The cornea offers not only an important target for gene delivery in its own right, but also can be used as a tool for the development of new vector systems, as a result of its accessibility and the ease of maintaining the organ ex vivo.

The two types of vectors used, viral and nonviral, have complementary advantages and disadvantages. In general viral vectors can be considered to be efficient, but toxic and immunogenic. Nonviral vectors are relatively nontoxic and nonimmunogenic, but are inefficient. There are two broad approaches to this problem: one is to take the viral vectors and strip out of them everything that is not needed for gene delivery. The second is to start to add into nonviral vectors features that will increase their efficiency (such as molecules that target them to particular cell surface receptors). In many cases these features will be borrowed from viruses! Thus, in several years' time we may well find ourselves using hybrid vectors, uncertain whether to term them viral or nonviral, that encapsulate the best features of both systems (Figure 6).

View larger version (18K): [in this window] [in a new window]   Figure 6.   The future of gene delivery vectors. The future "perfect vectors" will combine many of the best features of current viral and nonviral vectors. They will allow controlled expression of therapeutic proteins, and may be targetable to specific cells or tissues.