The Retina within Sight of RNAi Therapeutics

 From the highly prevalent diseases such as the age-related macular degenerations (dry and wet) to the rarer, but more numerous diseases such as the retinitis pigmentosa, the eye is the subject of a significant and growing unmet medical need.  From a genetic point of view, the application of RNAi Therapeutics to these diseases is very attractive.  However- yes, you guessed it- delivery challenges have made it difficult to exploit its potential.  In 2013, I have come across at least two pieces of evidence, one in synthetic and one related to DNA-directed RNAi, which make me believe that the time for ocular RNAi Therapeutics is nigh.

A bit of history

Following an early rush into ocular diseases that saw 3 wet AMD RNAi candidates, one each by Opko Health, Quark Pharmaceuticals (partnered with Pfizer), and Sirna Therapeutics/Merck (partnered with Allergan), speed into phase II and III studies, considerable doubts about their scientific foundations killed off the enthusiasm.  

Firstly, it was thought that the preclinical validations of the wet AMD candidates rested on TLR3-activation artifacts.  While I have not seen this claim more widely supported and chemical modifications and short double-strandedness should readily get around this issue, the more vexing question to me has been how were these either unmodified or simply 2’-O-methyl modified, unformulated RNAi triggers supposed to enter their target cells?

Moreover, for ddRNAi where Genable Technologies is readying an AAV-based candidate for retinitis pigmentosa to commence clinical trials, vector distribution following needle administration represents serious safety and efficacy issues.  Because the AAV and lentiviral gene therapy workhorses do not diffuse on their own from intravitreal injection sites to the back of the eye, ocular gene therapy has largely involved subretinal injections which a) is a relatively dangerous procedure that can seriously harm retinal architecture and integrity, and b) limits vector diffusion and therefore therapeutic activity to a small area surrounding the subretinal administration site.

Smartly evolved AAV can penetrate from the vitreous into the retina

In order to develop viral vectors capable of overcoming the physical barriers between the vitreous and retina, Dalkara and colleaguespublished earlier this year a study on the selection of AAV2 variants capable of doing just that. 

The selection process of this now popular method of directing AAV to various tissues and cell types starts out with large libraries of more or less randomly mutated AAV variants.  In each selection round, the viral DNA is extracted from the target cells and re-amplified for the next round of selection such that the AAV variants that most efficiently transduce the target cells eventually become highly enriched.

 

The 7m8 AAV that they identified was thus capable of broadly transducing retinal cells following intravitreal administration in mice, from the inner ganglion cells to the outer retinal pigment epithelial cells.  In non-human primates, 7m8 also greatly enhanced retinal penetration from the vitreous compared to old gold standard AAV serotypes.  However, retinal cells were not transduced as broadly as in mice suggesting that either additional candidates from mouse evolution experiments need to be characterized in non-human primates or that the selection itself should be performed in non-human primates.

I view such progress exceptionally promising for retinal degenerative diseases in which rescue of only a fraction of cells should be therapeutic (e.g. by getting rid of a mutant mRNA that leads to the death of the cell expressing it), but potentially also for the more challenging disease settings where RNAi action would probably have to occur in the majority of cells in order to be therapeutic.

 

Along these lines, after deciding on the best AAV variant, an important task will be to determine the more precise percentage of cells of each cell type that can be transduced to then match them up with potential indications.

Sticky self-delivering RNAi triggers penetrate deeply into the retina

In the field of synthetic RNAi Therapeutics, a peer-reviewed paper came out by RXi Pharmaceuticals (Byrne et al.) which confirmed earlier claims that self-delivering RNAi triggers (sd-rxRNAs as they call their own versions) were able to fully penetrate the retinal cell layers in both the mouse and rabbit following intravitreal injections.  The rabbit is an important model because their eye volume, believe it or not, is close to ours (~1.5ml vs ~5.0ml).  Formal proof of RNAi-mediated gene silencing was then confirmed in the mouse.

The self-delivering RNAi triggers had guide strands of ~19 nucleotides and passenger strands of less than 15 nucleotides.  An extended phosphorothioate (PS) single-stranded tail on the guide and additional hydrophobic modifications such as sterol groups were supposed to enhance tissue penetration and functional cell uptake.  The 6-8nt PS tail obviously has been borrowed from standard antisense chemistry.  Therefore, a battery of tests was performed to exclude visual abnormalities resulting from these extensive modifications.  No notable tox findings were made.

Despite the similarities in chemistry, sd-rxRNAs appear to be superior to ISIS-type PS-ASOs (RNaseH mechanism).  This is because in the Byrne et al. study, ~3 microgram RNAi trigger was required for 50% gene knockdown, whereas 50 micrograms phosphorotioate antisense oligos were required for a 50% knockdown of MALAT as presented by ISIS at OTS 2013 in Naples (report still available).  To wit­, MALAT as a nuclear RNA can be expected to be far more susceptible to RNaseH ASOs than your average mRNA, so the potency difference could be even bigger.  Also, it would be interesting to determine the safety implications of administering 3 micrograms versus 50 micrograms or whatever the equivalent dosages in humans.

The potency difference would have been much bigger still if RXi were not so married to their less than 15bp double-strandedness, a misguided decision driven by old IP considerations (when the Kreutzer-Limmers were still a concern).  I therefore expect other RNAi companies with more active R&D to thank RXi for the validation and come up with more optimal RNAi reagents for ocular applications.

Ocular RNAi Therapeutics- keep an eye on it in 2014 and beyond.

RXi Pharmaceuticals Continues String of Sensible RNAi Delivery Collaborations

After deciding to focus on the use of so-called ‘self-delivering’ siRNAs (‘sd-rxRNA’) for dermatology and ophthalmology applications, RXi Pharmaceuticals has since entered into a number of relationships with companies to facilitate the uptake of sd-rxRNA in various organs.

The latest such technology partnership with EyeGate Pharmaceuticals for ocular direct RNAi applications follows another direct RNAi delivery partnership in dermatology with TransDerm and a collaboration with Philips Electronics for developing image-guided ultrasound-mediated siRNA delivery technologies. While the long-awaited and much-rumored deals that should bring in non-dilutive funding are still lacking, these steps promise the efficient development of RXi’s RNAi Therapeutics platform by tapping into what appear to be companies with complementary know-how and technologies. If successful, the technologies developed as a result of these partnerships may well pave the way for these Big Pharma partnerships that RXi and its investors so crave.

A common theme of these collaborations is that the partner provides technologies that, through largely physical methods (electrical currents, ultrasound, microneedle injections etc), should allow for the co-localization of the sd-rxRNA with the target cells, after which RXi’s chemistry kicks in to overcome the formidable cellular lipid membrane barrier and induce RNAi gene knockdown- all without the need for nanoparticle formulation.

The relationship with EyeGate also addresses a problem generally facing the development of ocular drugs, particularly those for back-of-the-eye disorders tied to severe vision loss: patient-friendly modes of drug administrations. While the rapidly increasing genetic insights into eye diseases has yielded a number of excellent gene targets, there is considerable doubt whether drugs such as the monoclonal antibody Lucentis for wet age-related macular degeneration are viable in the long run as they need to be injected into the eye by needle every 1-3 months with each administration carrying with it an incremental risk of serious adverse events such as retinal detachments.

One way of addressing this problem is through the development of sustained-release formulations where a depot of drug is placed in the eye which releases drug over time. Another promising approach are gene therapies which may need to be administered only once or twice in a life-time, and may even offer the prospect for real cures. A third approach, the one taken by EyeGate, is to use iontophoresis whereby low electrical currents are applied to generate ions that repulse like-charged drug molecules, appropriately formulated, and push them across the outer sclera to be distributed throughout the eye. Because this technique does not involve needle injections and instead works by placing electrodes on top of the eye, it should be relatively patient friendly. EyeGate is currently in late-stage development of a drug candidate delivered by its proprietary iontophoresis technology.

It is, of course, worth keeping in mind that such collaborations will take time before we will see resulting technologies move into the clinic. Nevertheless, one thing is clear: RXi Pharmaceuticals is highly competitive in attracting qualified partners in the area of conjugate/self-delivering siRNA Therapeutics development. Probably the result of focusing their efforts on one segment of RNAi Therapeutics technology instead of trying to dominate them all. It is now time, however, to leverage the sd-rxRNA platform for that non-dilutive funding.

mdRNA and RXi Coming Out with their RNAi Triggers (Part II)

In the second part of my review of mdRNA’s and RXi Pharmaceutical’s new trigger designs (part I: specificity of mdRNA's usiRNAs), I will try and somewhat de-mystify RXi’s ‘self-delivering rxRNAs’, the industry’s best-kept secret. First insights into the structure and chemistry of sd-rxRNAs were recently provided at the ARVO meeting in Florida and the publication of a related patent application (International Publication number WO 2010/033247 A2).

Shown above is a typical example of an sd-rxRNA. Sd-rxRNAs are based on a so-called ‘asymmetric siRNA’ structure in which the double-stranded region of the RNAi trigger is relatively small (less than 15bp) whereas the length of the guide strand is maintained around 19-23 nucleotides to maintain silencing efficacy. Although on average these RNAi trigger designs are significantly less potent than traditional Tuschl-type siRNAs, extensive screening in some cases allows for quite potent asymmetric siRNAs to be found with picomolar activities.

The double-stranded region is kept short because it is thought that structurally rigid double-stranded nucleic acids don’t wiggle well through the plasma membrane. Single-strand nucleic acids, however, as practiced in the antisense field can enter the cells more readily possibly because of the increased flexibility and an exposed nucleobase that is relatively less charged.Another benefit of keeping the dsRNA region below 15bp is that it avoids conflicting with competing IP estate such as Alnylam's Kreutzer-Limmer series, although in this case the scientific rationale, as just pointed out, should be strong enough to stand on its own feet. Having said that, I can remember conference presentations by a Korean group that reported on similar asymmetric siRNAs, and there was also a publication on 'asymmetric siRNAs' in Nature Biotech2 years ago, although in that case the dsRNA region was more centrally placed, whereas in the case of RXi’s sd-rxRNAs the shorter passenger strand base pairs with the 5’ end of the guide strand.

The guide strand carries a synthetic 5’ phosphate group. This is probably intended to address the less efficient phosphorylation of highly modified, conjugated siRNAs of unusual structure compared to traditional RNAi triggers that undergo rapid phosphorylation following introduction into the cell, a step that is necessary for activating the silencing potential of an siRNA.

In addition to asymmetric structure, there are a number of chemical bells and whistles that render the RNAi trigger more lipophilic (‘fat-loving’) and therefore membrane permeable. One important strategy here is to replace a number of the phophodiester bonds in the RNA backbone with phosphorothioate linkages, especially in the long 3’ overhang of the guide strand. Phosphorothioate backbones are well known in the oligonucleotide therapeutics field and e.g. widely applied to RNAseH-type antisense molecules and have the property of being 'sticky' and contributing to favorable pharmacologies. It is possible that these phosphorothioylated 3' overhangs may not only enhance the membrane permeability of the molecules, but harness specific oligonucleotide uptake receptors thought to play a role in RNaseH antisense delivery.

Based on my understanding of the effect of phosphorothioylation on RNAi performance, and also evident from a number of datasets in the patent application, such modifications should not be used too extensively as this often will compromise knockdown efficacy (best limited to the 3' overhang). Phosphorothioylation, like the extensive 2’-O-methyl and 2’-F modifications of the nucleotides, has the added benefit of stabilizing sd-rxRNAs which in many applications will be directly exposed to body fluids (note: such modifications may also be used to prevent innate immune stimulation and to enhance guide-strand specificity).

In addition to phosphorothioylation, the conjugation of a lipophilic group such as a cholesterol to the 3’ end of the passenger strand is intended to also enhance membrane permeability and cellular uptake. This, of course, is very similar to the cholesterol-conjugated siRNA delivery approach taken by Alnylam and first published 5 years ago (Soutschek et al., 2005). For the same reason, lipophilic groups may also be added to within the RNAi trigger structure itself.

The real question, of course, is how all this translates to silencing in vivo. Unfortunately, the presentations contained only very little in vivo efficacy data. The abundance of tissue culture experiments, however, indicate that compared to lipid-mediated transfection, much higher amounts of sd-rxRNAs are required to achieve similar silencing (50-1000 fold higher). There is, however, one dataset that shows that ~50mg/kg sd-rxRNAs can knock down a gene expressed in a mouse liver (compare this to the 1000-fold lower dosages required to the latest SNALP liposomes), but further improvements in potencies are required before sd-rxRNAs, without further formulation become therapeutically relevant.

Compared to the 2005 Nature study by Alnylam, I agree that sd-rxRNAs may be a little bit more potent on average. Consequently, the patent application is littered with references about the superiority of sd-rxRNA over Alnylam’s siRNA-cholesterol conjugate technology. This, however, is a little bit unfair in my opinion given that it appears to be Alnylam that was really the innovator in this field and that its conjugate technology should have progressed since the initial publication.

Maybe a little bit more surprising, or then again maybe not, were also regular references about the technical superiority compared to Dharmacon’s Accell self-delivering siRNAs. To be clear, references of technical superiority per se are nothing unusual in the patent literature. It becomes noticeable, however, when such references are quite frequent and apparently politically motivated. As discussed on this blog before, I have speculated that Accell is likely to use lipophilic chemistries, too, not least because there are patent applications by Dharmacon covering lipophilic siRNA conjugates. (e.g. WO/2008/036825). I don’t want to speculate any further here, but RXi Pharmaceutical would probably do well to take potential conflicts of interests serious and take the necessary steps to address them before yet another IP drama erupts.

In summary, sd-rxRNAs are an example of how one would aim to render siRNAs more membrane-permeable without sacrificing too much silencing efficacy. The concept and the means employed may not be that revolutionary, but credit has to be given for extensively testing which structures and chemistries could work. For now, sd-rxRNAs are probably most promising in direct RNAi applications such as for the skin (see recently announced TransDerm collaboration), eye, and lung (by inhalation) where the ability to administer large amounts of unformulated siRNAs to the target organ may outweigh the potency disadvantages compared to traditional siRNAs formulated in polyconjugate or nanoparticle formulations.

mdRNA and RXi Coming Out with their RNAi Triggers (Part I)

In conference presentations this week, both mdRNA and RXi Pharmaceuticals finally lifted the veil of mystery shrouding the nature and performance of their claimed proprietary RNAi trigger designs. Independent of what they may mean in terms of IP, they are certainly worthy contributions to the science of RNAi triggers: the use of unlocked nucleic acids to increase the specificity of RNAi gene silencing (mdRNA), and the convergence of delivery and RNAi triggers through nucleic acid chemistry (RXi).

At the ‘RNAi and miRNA World Congress’, mdRNA presented data illustrating how selectively spiking siRNAs of the traditional Tuschl structure with a number of unlocked nucleic acid residues (usiRNAs) can reduce the level of off-targeting commonly provoked by ‘naked’ siRNAs. While similar strategies have been reported before and are broadly attempted in the industry, what differentiates mdRNA’s approach is that this particular nucleic acid analogue does not involve the addition of modifying chemical groups. Instead, in an unlocked nucleic acid the ribose ring is simply broken without making the residue any bulkier, but this is enough to significantly affect the interactions of the siRNA with its target mRNA and the Argonaute protein, the effector protein of the RNAi silencing pathway.

The simplest, and therefore widely practiced first step in reducing off-targeting is to modify the 5’ end of the passenger strand with a variety of chemistries. Because the 5’ end is important for the Argonaute protein to recognize and bind an RNA as the guide strand, half the off-targeting can be eliminated based on the sensitivity of this interaction to chemical alterations. It so happens that a UNA at the 5’ end of the passenger strand is similarly not tolerated…as has in effect recently also been shown in a publication by Sirna Therapeutics/Merck (Kenski et al., 2010). In the example provided by mdRNA at the conference, this meant a reduction in the number of genes that were changed in their expression by a factor of more than 2-fold (generally 2-3 fold) from ~390 to ~ 180, while on-target gene silencing activity was maintained at around 95% (data from RNA microarray experiments).

I should add, however, that additional usiRNA residues in the 3’ overhangs of both passenger and guide strands somewhat complicate determining the exact contribution of the passenger strand 5’ modification in reducing off-targeting. This is because it is known from structural studies that the position of the 3’ end of the guide strand relative to the Argonaute protein changes upon the transition from the microRNA-off-target (in the so called ‘PAZ pocket’) to the on-target cleavage (dislodged from the ‘PAZ pocket’) confirmations. 3’ ends that do not fit into the PAZ pocket may therefore change the ease with which the enzyme adopts either conformation and therefore affect the propensity of microRNA-like off-targeting. Structural studies with UNA residues at the 3’ end may be instructive here.

More interesting is mdRNA’s observation that when, on top of the above modifications, one normal base is replaced with its corresponding UNA analogue in the seed (= the first 2-8 nucleotides), the off-target activity is further reduced to about 35 genes that are changed by at least 2-fold, while again maintaining on-target efficacy. This was theorized to be due to the changes in guide RNA-target mRNA hybridization energies within the seed: the usiRNA lowered this energy sufficiently such that a microRNA-like interaction that is reliant on the seed becomes prohibitively weakened. However, the extended length of the guide RNA-target mRNA base-pairing compensates for the lost energy thus still allowing for Argonaute conformational change and on-target cleavage to ensue.

Previously described modifications have similarly been thought to function in reducing microRNA-like off-targeting by modulating seed energies. usiRNAs, however, could be a particularly attractive approach, because its small size should make it less likely to be detrimental to on-target activity. This, of course, would have to be shown with more examples, preferably in a peer-reviewed publication, and the Sirna Therapeutics paper has shown at least some limitations in where the UNA can be placed in the seed. Similarly, it remains to be demonstrated how much of the reduction in the off-target activity was due to eliminating microRNA-like interactions instead of a reduction in the innate immunostimulatory potential of these siRNAs as UNAs were also shown by mdRNA to mitigate such responses.

I believe this story shows how RNAi trigger design has matured to a point that RNAi Therapeutics truly lives up to its original promise of specificity which had temporarily (2003-6 period) been tarnished by the discovery of wide-spread microRNA-like off-targeting and the innate immunostimulatory potential of exogenously delivered synthetic siRNAs. It also demonstrates that at least on the level of RNAi trigger design mdRNA could certainly be a valuable partner to most in Big Pharma.

In my next entry (link here), I will provide my thoughts about RXi Pharmaceutical’s ‘self-delivering rxRNAs’ for which, until now to the best of my knowledge, the company has created so much expectations without actually disclosing the chemical nature thereof.