Ocular Gene Therapy: Quo Vadis?

The question of where gene therapy is going has almost as many answers as there are practitioners in the field. Narrowing our focus to diseases of the eye, this diversity of opinions does not contract proportionately. Therefore, rather than forecasting the future, it is more useful to assess the field today, including a consideration of where problems remain before clinical trials would be generally feasible. Progress is clearly most advanced in developing therapies for retinal degenerative diseases, principally because the genetics of many forms of retinitis pigmentosa (RP) are relatively well understood, and many natural and transgenic animal models exist. However, given the surgically simpler and less traumatic access afforded by any part of the eye other than photoreceptors and retinal pigment epithelium (RPE), the primary cellular targets of RP, it is not necessarily true that gene therapy for outer retinal diseases will be the first or most successful clinically. The brief opinion presented here will therefore attempt to cover the status of gene therapy for all parts of the eye in as balanced a manner as possible. 

Ideally there are four basic prerequisites that should be met for any genetic therapy targeted to an ocular disease. First, a gene delivery technique must be available that is efficient and nontoxic. Second, the genetic basis of the disease, or minimally its biochemical basis, should be well characterized so that an appropriately matched therapeutic approach can be selected. Third, expression of the therapeutic gene needs to be properly controlled, both insofar as levels of the gene product are concerned as well as which tissues or cell types support or do not support expression. Finally, having an experimental animal model of the disease available for preclinical testing of the therapy is clearly key in demonstrating proof of principle. The field of ocular gene therapy has been dealing with these issues for about 5 years, but as yet no clear consensus has emerged for any single approach, ¹ and, given the diversity of tissues in the eye and the range of pathology they can experience, this should not be surprising. On the other hand, over the past 2 years, there have been major advances toward satisfying one or more of the four prerequisites for several ocular diseases. It is this progress that will be reviewed and then placed in the larger context of ocular gene therapy, quo vadis? To logically review the eye as a gene therapy target, issues regarding individual ocular tissues and conditions will be discussed generally in the order in which they occur, from the back of the eye forward. 

Outer Retina/RPE

Three classes of therapy for photoreceptor/RPE diseases have seen recent developments worth highlighting, each exhibiting photoreceptor rescue in rodent models of RP. These include ribozyme therapy targeted against mRNA from individual dominant negative alleles, neurotrophin/growth factor gene therapy against a variety of genetic forms of RP, and, more indirectly, therapy aimed at promoting avoidance of photoreceptor apoptosis, the apparent common final cell death pathway in RP. 

Ribozyme Therapy

Morphologic and electrophysiological preservation of photoreceptors was seen for 3 months in the P23H transgenic rat using recombinant adeno-associated virus (rAAV)-vectored ribozymes. ² This experiment has been recently repeated, assessing the longevity of rAAV-mediated ribozyme therapy (M. LaVail, A. Lewin, and W. Hauswirth, unpublished observations, 2000). After 8 months, ribozyme-treated eyes in 12 rats retained on average 4 to 5 rows of photoreceptor nuclei, whereas contralateral, phosphate-buffered saline (PBS)–injected eyes now had only approximately 1 row. Rod electroretinographic (ERG) amplitudes reflected this difference. Another key question is whether residual photoreceptors in a retina with advanced RP are still responsive to rescue by gene therapy. This situation was modeled by delaying rAAV-ribozyme injections in the P23H rat until 4 rows of photoreceptor nuclei had been lost. Ninety days later, treated right eyes retained an average of 5 rows of photoreceptor nuclei compared with 2 rows in the control partner eyes. It, therefore, appears that even intervention after photoreceptor loss is well under way by AAV-vectored genes can delay the loss of the remaining photoreceptors. 

Neurotrophin/Growth Factor Genes

Neurotrophin gene therapy using recombinant adenovirus carrying a CNTF (ciliary neurotrophic factor) cDNA has led to structural rescue of photoreceptors for several months in the rd ³ and rds ⁴ mouse models of RP. In the latter, scotopic ERG amplitudes were also enhanced, suggesting functional rescue as well. However, the level of rescue was only modest and of limited duration. Using an AAV vector, CNTF genes delivered to photoreceptors in the P23H opsin transgenic rat model of RP resulted in morphologic and electrophysiological preservation for at least several months (W. Peterson, unpublished observations, 1999). Various other forms of neurotrophins in AAV vectors are currently under testing, and, given the more persistent expression of AAV passenger genes in photoreceptors, ^{5 6} this should be a critical test of its value in the retina. On the downside, CNTF as an injected protein clearly has toxic effects in the retina and brain. Phase I trials of Regeneron Corporation’s Axokine in the central nervous system for weight loss were halted after occasional herpes reactivations were reported. Regeneron also halted its transgenic P347L opsin pig trial due to a lack of photoreceptor rescue, occasional cataracts, and CNTF-related ERG abnormalities. Whether secreted CNTF from virally vectored genes will have similar side effects remains to be established. This is an important contrast because there are likely to be significant differences between CNTF as an ocular bolus of injected protein and CNTF as a gene leading to lower but more sustained local levels of the protein. 

Basic fibroblast growth factor (bFGF, FGF2) has been of long-standing therapeutic interest in the eye. When the bFGF gene is delivered by an adenovirus vector, photoreceptor survival is prolonged for at least 2 months in the Royal College of Surgeons (RCS) rat. ⁷ This is significantly longer than the rescue duration reported for intraocular administration of the protein, highlighting the relative effectiveness of prolonged survival factor production locally through transgene expression. However, the potentially toxic angiogenic properties of bFGF have yet to be carefully evaluated in vector-bFGF–treated eyes. There is currently a growing list of molecules within the FGF family, many of which have little or no angiogenic potential, and these appear to be prime candidates for testing by gene delivery in animal models of RP (J. Flannery, unpublished observations, 2000). 

Anti-Apoptosis Therapy

Because apoptosis appears to be the common death pathway for photoreceptors in RP animal models, its manipulation by genetic means seems a logical therapeutic option. However, results of crosses between RP mice and transgenic mice either overexpressing the anti-apoptotic genes bcl-2 or c-fos, or a knockout of the pro-apoptotic p53 gene, have been mixed at best. More recently, the photoreceptor survival effect of bcl-2 gene therapy was significantly enhanced in the rd/rd mouse if a wild-type copy of the defective βPDE (β-subunit of cGMP-phosphodiesterase) gene was also delivered, ⁸ suggesting that anti-apoptotic therapy can be combined with gene augmentation for an additive effect. In contrast, AAV-vectored bcl-2 expression in ganglion cells leads to increased glutamate-mediated cell damage. ⁹ However, such toxicity should be avoidable by limiting the cell type expressing the passenger gene through precise local injection and tight, cell-specific promoters. Two recent reports portend other gene therapy approaches. ^{10 11} In the S334ter transgenic rat, activation of caspase 3, a primary mediator of apoptosis, accompanied photoreceptor cell loss. ¹⁰ This opens a variety of anti-caspase gene strategies to evaluation. A major new inhibitor of apoptosis (IAP), X-linked IAP, was described that inhibits caspase 3, ¹¹ thus its testing in animal models of RP seems logical. 

Gene Augmentation Therapy

Delivery of a normal gene to compensate for lack-of-function mutations such as null mutations in the βPDE gene has shown structural and biochemical rescue in the rd mouse. Rescue was mediated either by recombinant adenovirus, ¹² lentivirus, ¹³ or encapsidated adenovirus minichromosomes ¹⁴ (EAMs, “gutless adenovirus”), and in each case treatment preserved photoreceptor morphology to a modest degree for 1 to 2 months. Why neither longer nor more complete rescue was seen remains unresolved. Both problems could be vector-related or model-related because the rd mouse loses photoreceptors very quickly over the first 3 to 4 weeks of life, perhaps too quickly for effective rescue by any vector. 

Studies in Large Animals

AAV-vectored green fluorescent protein (GFP) under control of an opsin promoter was stably expressed in rats for more than 3 years (J. Flannery and W. Hauswirth, unpublished observations, 2000) and with a cytomegalovirus promoter in the rhesus monkey for nearly 2 years (J. Bennett, unpublished observations, 2000) without obvious toxicity in either species. Importantly, in the monkey, after transduction of one retina and seroconversion to modest anti-AAV titers, the contralateral retina could be efficiently transduced by the same vector. ⁵ Thus, the fact that 60% to 80% of humans are seropositive for AAV should not limit this vector’s utility in the retina. Because all current retinal gene therapy reports have involved rodents, it is vital to determine whether the analogous therapy can be equally effective in a human-sized eye. Retinal gene therapy experiments are currently under way in the P347S transgenic pig using ribozymes and in the rdc1 dog using βPDE gene augmentation to answer this key question. In neither species did subretinal injection of high titer AAV vector result in noticeable inflammation. Therapeutic results have been slow to come, however, because rates of retinal degeneration are significantly slower than in rodents, yet still faster than in human RP. 

Viral Vector Toxicity

Several recent reports have refocused concern on the safety of viral vectors, particularly the adenoviruses. In the retina, immune reaction to adenoviral vectors appears to limit its effectiveness to 3 to 8 weeks in mice, suggesting that vector antigens are recognized and responded to relatively normally in the subretinal space. Consistent with this, coexpression of immune-modulators enhances the persistence of adenoviral-vectored retinal gene expression. ¹⁵ Without doubt, the gene therapy outcome that has generated the most attention over the past year was the fatality during an adenovirus-OTC (ornithine transcarbamylase) human trial. The patient’s immune reaction to the large systemic dose of vector administered (4 × 10¹³ particles) seems to have been the initial pathogenic trigger. It is important to note, however, that at current viral titers well less than 10% of this amount could be injected into a human subretinal space. Nevertheless, this unfortunate phase I outcome will certainly place a much greater safety burden not only on all adenoviral vector protocols but on other vector systems as well. In this regard it should be noted that AAV (unrelated to adenovirus), although able to elicit a modest immune response, has shown no toxicity at similar systemic or subretinal doses in several large animal species, including monkeys. 

Three other developments this past year highlight the growing opportunities for new retinal gene therapy approaches. Enhanced survival of photoreceptors was noted in several mouse RP models after ocular administration of GDNF (glial-derived neurotrophic factor), ¹⁶ PEDF (pigmented epithelium–derived factor), ¹⁷ or a calcium-channel blocker, ¹⁸ suggesting alternative, perhaps complimentary, retinal gene therapy strategies. 

Retinal Ganglion Cells

Glaucoma is a leading cause of blindness in the world, and several mechanisms of retinal ganglion (RGC) cell death have been proposed. The two predominant hypotheses currently involve neuronal damage due to either long-term elevation in intraocular pressure or the presence of genetic factors predisposing a patient to RGC loss with age without significantly elevated intraocular pressure. Proximal causes of RGC death appear to include glutamate toxicity, induction of nitric oxide synthase (NOS-2) leading to reactive oxygen species damage, and/or loss of trophic factor support. ¹⁹ As a result, apoptotic cell death ensues. Because the genetic basis of glaucoma is not well understood, gene therapy must currently either target these biochemical manifestations of glaucoma or, at a more downstream stage, help RGCs avoid apoptosis. An additional problem has been the lack of well-defined animal models. Currently, the most useful models are produced by the induction of RGC death in rodents by axotomy, cautery of episcleral tissue, or episcleral vein occlusion, with the latter two procedures leading to increases in intraocular pressure before RGC loss. In these systems, essentially any neuroprotective paradigm may be useful, including neurotrophin/cytokine gene therapy or anti-apoptotic approaches, ²⁰ and a number of pharmacological agents have shown promise. For gene-based therapy, the observation that optic nerve transection leads to activation of the tumor-suppressor protein p53 and, hence, to upregulation of the proapoptotic gene bax prompted the evaluation of bax antisense oligonucleotides. Vitreal inoculation of antisense oligonucleotides reduced axotomy-induced RGC death, ²¹ suggesting the potential utility of a broad range of anti-apoptotic genes. RGC loss due to optic nerve axotomy can also be slowed by adenoviral vector delivery of the BDNF (brain-derived neurotrophic factor) gene. ²² Interestingly, Müller cells appeared to be the primary cell type transduced on vitreal injection of the vector. This suggests that transduction of supporting Müller cells with genes for secretable trophic factor genes or with genes that promote their production may be another way to protect adjacent RGCs. Thus, a variety of death-avoidance strategies may be useful in gene-based glaucoma treatment. 

For pressure-induced glaucoma in particular, an obvious target tissue is the trabecular meshwork (TM), where intraocular outflow is regulated. Modulation of the TM physiology by gene transfer could be a way to achieve long duration pressure reduction. Cells of the TM appear to be good targets for gene transduction, as demonstrated by liposome-mediated delivery of reporter genes or oligonucleotides into rat or primate TM. ²³ Recently, good levels of transduction have been reported using adenovirus vectors as well. ²⁴ Viral vector transduction was achieved without affecting outflow facility, damaging neither tissue architecture nor TM morphology. ²⁵ Thus, TM manipulation by vectored gene expression seems a viable alternative for some forms of glaucoma. 

Optic Nerve

Recurrent and acute optic neuritises are vision-threatening sequelae of several local and systemic maladies, including Leber’s hereditary optic neuropathy and multiple sclerosis. Recently, viral vectors achieved efficient gene transfer to the optic nerve after intravitreal injection near the optic nerve head in the mouse and guinea pig. AAV-mediated transfer of the catalase gene, a reactive oxygen species scavenger, suppressed demyelination and blood–barrier disruption at the foci of experimental autoimmune encephalomyelitis in the mouse optic nerve. ²⁶ The long-term (>1 year) reporter gene expression with AAV-vectored gene transfer observed in nerve fibers, glial cells, and associated blood vessels suggests the feasibility of this strategy for chronic or recurrent optic neuropathies as well. ²⁷ The use of a replication-deficient adenovirus rather than an AAV vector allowed more rapid transduction with the catalase gene in the same model system, ²⁸ a modification that could be helpful in the treatment of more acute forms of optic neuritis. 

Retinoblastoma

The most common primary ocular malignancy of childhood is retinoblastoma. Large retinoblastomas are currently treated by enucleation, often with long-term problems (visual loss and severe cosmetic deformity secondary to enucleation and/or irradiation of the orbital region). Radiation treatment often eliminates any residual tumor, but an increased risk of second cancers in the radiation field is a serious side effect. Because small tumors might be surgically treated, thus avoiding enucleation, control of tumor burden may be an important therapeutic goal. Introduction of the herpes simplex virus thymidine kinase gene (HSV-tk) followed by ganciclovir treatment to create a toxic nucleotide analog that can be incorporated specifically into the DNA of dividing cells and thereby kill them has proven useful in a variety of systems. This strategy was tested as tumor-targeted suicide gene therapy in Y79Rb cell line–induced ocular tumors in mice ²⁹ and was shown to be effective in reducing the ocular tumor burden. A similar result was obtained on induced subcutaneous tumors in mice when the Rb cell line had been transduced ex vivo with HSV-tk in a retrovirus vector. ³⁰ An alternative strategy involved HSV mutants defective in the virus-encoded ribonucleotide reductase gene. This virus can replicate and kill dividing cells well, but it is ineffective against nondividing cells. Such mutant viruses indeed inhibited the growth of subcutaneous Y79–induced tumors in mice. ³¹ Thus, in appropriate cases, suicide gene therapy mediated either by viral vectors and local ganciclovir treatment or by a suicide HSV virus infection could be useful in avoiding the need for enucleation and radiotherapy, thus allowing more conservative treatment (cryotherapy or laser photocoagulation) as used currently for small ocular tumors. 

Retinal and Choroidal Vasculature

Choroidal neovascularization (CNV) is the major cause of severe loss of vision in patients with age-related macular degeneration (AMD), and retinal neovascularization is the prime cause of vision loss in diabetic retinopathy and in retinopathy of prematurity. Together, AMD and diabetic retinopathy are by far the most frequent causes of untreatable blindness in the world. Thus, the development of gene strategies that counteract neovascularization stimuli would be a major therapeutic advance. Because vascular endothelial growth factor (VEGF) appears to be a major regulator of the ocular vasculature, its control may be key. Unfortunately, antisense oligonucleotide therapy against VEGF failed to prevent experimental CNV in a laser-induced mouse model. ³² However, because VEGF is constitutively expressed in the vasculature of normal eyes, ³³ it is possible that significant inhibition of VEGF activity may have adverse consequences. 

AMD and diabetes are chronic diseases; therefore, long-term expression of therapeutic genes is likely to be required. Toward this end, retrovirus-mediated reporter gene transfer after intravitreal or subretinal injection was tested in a laser photocoagulation–induced CNV mouse model. Gene transfer and expression were seen selectively near sites of CNV. ³⁴ Because retroviral vectors require proliferating cells for effective transduction, such vector homing to sites of neovascularization could be very useful in gene-based CNV treatment. Clearly, however, we are at a very early stage in the development of gene therapy for AMD-related CNV, and other viral vectors (AAV, adenovirus) as well as obvious candidate therapeutic genes (e.g., angiostatin and PEDF) need to be tested. Perhaps most importantly for AMD, a better animal model of CNV is likely to be needed before significant headway will be made. 

In the diabetic patient, fibronectin synthesis associated with basement membrane thickening is a key indicator of diabetic microangiopathy. Fibronectin antisense oligonucleotides delivered into normal rat retinal capillary cells after intravitreal injection reduced fibronectin expression 47% to 87% over the first week after treatment. ³⁵ Although potentially useful for acute inhibition of retinal neovascularization, longer-term treatments are likely to be needed, and the vector and therapeutic gene considerations noted above for the choroidal vasculature would seem to apply to the retinal vasculature as well. Significantly, the development of an opsin promoter–regulated VEGF mouse that exhibits retinal neovascularization ³⁶ provides an important new model for evaluating ocular anti-neovascularization gene therapy. 

Retinal Detachment

Proliferative vitreoretinopathy (PVR) is the most common cause of failure in treating rhegmatogenous retinal detachment. Removal of inappropriately dividing cells, such as proliferating RPE cells, using suicide gene strategies is therefore worth testing. In a rabbit with vitreally implanted fibroblasts that had been transduced ex vivo with HSV-tk, the severity of experimental PVR was reduced after ganciclovir treatment. ³⁷ Similar selective reductions in proliferating RPE cells or in fibroblasts in culture have been reported for infection with a ribonucleotide reductase mutant HSV ³⁸ or an E2F decoy oligonucleotide delivered by a liposome, ³⁹ respectively. An additional problem associated with retinal detachments is the local loss of photoreceptor function, perhaps due to an apoptotic mechanism related to a partial loss of trophic support from the RPE. BDNF protein injected during experimental detachment in cats led to diminished cell proliferation locally and to better photoreceptor morphology, ⁴⁰ suggesting that BDNF or other neurotrophin gene delivery may be useful at the site of detachment as well. However, in this instance as well as for PVR, any gene therapy strategy should be designed for limited duration because once proliferating cells in the vitreous have been eliminated or the detachment resolved, the therapeutic gene’s continued presence is unnecessary and may be undesirable. 

Lens

Posterior capsule opacification (PCO) is a common late complication of cataract surgery. The current treatment for PCO, laser capsulotomy, can cause serious complications such as macular edema or retinal tears or detachment. The benefit of surgical techniques such as cortical cleanup or removal of lens epithelial cells, the source of inappropriately mitotic cells, has yet to be demonstrated. Similarly, antimetabolites or toxins linked to antibodies against lens epithelial cells have also been unsuccessfully tested. The removal of dividing cells by suicide gene transfer as described above is therefore a possibility for PCO as well. ⁴¹ In a rabbit model, intracameral injection of HSV-tk adenovirus followed by ganciclovir treatment prevented experimental PCO. Thus, targeted destruction of pathogenic mitotic cells seems a viable strategy for a variety of ocular disorders. 

Conjunctival and Corneal Epithelia

The conjunctival and corneal epithelia play important roles in the pathogenesis of various eye diseases. Their external position offers several alternative gene delivery paradigms not available to deeper tissue such as the retina. In theory, transfer of genes expressing antiinflammatory proteins or growth factors could be a valuable way to prevent an inflammatory cascade or to supply components for healthy epithelial turnover. This could be of particular use in the wound-healing process. Gene delivery to a predetermined area of a rabbit cornea was reported without evidence of corneal damage using gene gun technology (high velocity gold beads coated with plasmid DNA). ⁴² Adenovirus vectors have also been used for gene transfer to conjunctival and corneal epithelia. ^{43 44 45} Because the corneal epithelium seems less sensitive to transduction by adenovirus than the conjunctiva, topical application of low titer vectors may be suitable for selective conjunctival gene transfer without affecting the corneal epithelium. However, because of the rapid turnover of corneal and conjunctival epithelia, in vivo expression with either technology was sustained for only a few days after treatment. This could be an advantage if a limited therapy is required. However, if prolonged expression is needed, gene transfer to corneal and conjunctival stem cells, localized at the limbus and fornix, respectively, will be necessary. Because both nonviral plasmid DNA and adenoviral vectors lead to relatively short-term transgene expression in most tissue, extended-term gene therapy in the cornea has been approached recently using a retroviral vector for an ex vivo transduction of corneal epithelial progenitor cells. ⁴⁶ Retrovirus vectors stably integrate into the host genome of mitotic cells, providing the potential for long-term passenger gene expression. In this study, keratolimbal autografts transduced with retroviruses continued to produce progeny cells that expressed the transgene for at least 6 months in vivo. Because stem cells located in the limbus appear to be responsible for the continuous repopulation of the corneal epithelium, this could provide a long-term therapy for corneal disorders either by direct in vivo transduction of stem cells or by ex vivo corneal stem cell transduction and implantation into the limbal area. ⁴⁷  

A novel corneal gene delivery technology is low-voltage electroporation for transferring genes as naked DNA to a relatively circumscribed area of the corneal endothelium. ⁴⁸ With circular plasmid DNA, gene expression persisted for at least 21 days. Such transient expression could be an alternative or adjuvant to drug therapy in wound healing or in acute diseases of the cornea. More recently, plasmid DNA-encoding tissue plasminogen activator (tPA) was transferred to a portion of the corneal endothelium with a similar low-voltage electroporation strategy. ⁴⁹ Intracameral fibrin formation after laser-induced bleeding and subsequent corneal cloudiness were inhibited significantly. Thus, given the special external access afforded by the cornea, physical gene delivery techniques have clear potential for therapeutic application in the cornea. 

Concluding Remarks

Although there remain many untested genes that have therapeutic potential for ocular diseases, current proof-of-principle successes, primarily in rodent models of RP, suggest that the next experimental steps toward human trials are justified. There are two key goals remaining for all ocular gene therapies at the preclinical level. ¹ The given therapy must show therapeutic results in a species with a human-sized eye. For many of the ocular diseases and conditions discussed above, this will be a daunting challenge for two reasons. First, animal models with the relevant ocular condition may not exist and/or the technology for their creation may be unknown or prohibitively expensive. Second, the question of ocular dimensions and therapeutic coverage (i.e., number and topography of target cells that are required to be transduced to yield a measurable vision-preserving outcome) may not be trivial, and the issue can only be resolved by experimental verification. In the process, noninvasive clinical end-point criteria over longer time periods, more like those expected in human trials, need to be developed as well. ² A Federal Drug Administration–approved phase I/II ocular gene therapy protocol will require detailed animal toxicity and biodistribution studies. Potential for toxicity will be intimately tied to the vector used. Current herpes virus vectors retain toxic cellular properties that have yet to be fully evaluated in the eye. As noted, adenovirus vectors have immune response and persistence problems that have been well documented in ocular tissue. AAV vectors, in contrast, appear to be relatively nontoxic at high doses in the eye and elsewhere and, with the appropriate promoter, lead to persistent gene expression. Lentiviral vectors appear to have many of the same favorable properties as AAV vectors, at least in photoreceptors, ⁵⁰ but they have received only limited attention to date in ocular tissue. Important issues for lentivirus vector preparations that remain to be fully resolved include the possibility of low-level infectious wild-type (HIV) contaminants, relatively low vector titers, and expression shut off/mutagenesis upon chromosomal integration. At the moment, therefore, AAV vectors are best poised for initial clinical trials. Finally, after any vector administration, systematic toxicity analysis of ocular and adjacent tissues would be best carried out on the same animals used for efficacy studies. In this way any unintended vector dissemination and associated pathology could be accurately balanced against any therapeutic effects. 

In conclusion, the vision community stands on the brink of a revolution in therapies for blinding and acuity-altering conditions. It is an exciting time, because ocular gene therapy seems well poised to be among the earliest successful applications of this new approach to disease and pathology. On a more sobering note, the field is well-advised to build strong preclinical cases for safety and efficacy before proceeding to the clinic for any ocular applications of gene therapy. To launch prematurely into a poorly supported trial could delay this promising therapy for years. As always, the message is that careful science, particularly when building a case for human trials, is never a mistake. 

 Submitted for publication February 10, 2000; revised April 8, 2000; accepted April 11, 2000.

 Commercial relationships policy: N.