Stem cells and RNA interference offer the latest and, to some, the greatest hope for novel therapies to fight neurodegenerative diseases. The prospect of replacing lost or dysfunctional neurons with stem cells, or knocking down pathogenic protein accumulation with RNAi, or even fixing disease-causing mutations by gene replacement drives the development of new technologies specifically for neurological diseases. A batch of recent studies dealing with all these possibilities provides a broad look at the promise of new treatments, and a reminder that the road to the clinic will not be a short one.

Stem cells to the rescue
Stem cells, often looked to as replacement parts for lost or faulty neurons, could play a completely different role in treating neurodegenerative diseases, according to a new study in the July 26 Journal of Neuroscience. The work, from Richard Sidman and colleagues at Harvard Medical School and Evan Snyder at the Burnham Institute for Medical Research, La Jolla, California, shows that rather than replace ailing neurons, implanted stem cells provide support for endogenous, at-risk neurons and preserve their health and function by modulating the tissue environment.

A major obstacle to using stem cells for neurodegenerative diseases has been the problem of how new neurons would be incorporated into functional circuits. The new results suggest that stem cells could ultimately find their use in preserving existing neurons and their connections rather than the considerably more challenging task of rebuilding networks after neurons have been destroyed. The stem cell results were obtained in a mouse model of inherited Purkinje neuron degeneration, but could be applicable to many other diseases, if treatment can be initiated early.

To study the effects of early stem cell transplantation, the researchers chose a well-defined model of neuron degeneration, the nervous mouse. Previously, first author Jianxue Li showed that in these mice, Purkinje cell death in the fourth to fifth postnatal week is triggered by a 10-fold increase in expression of tissue plasminogen activator (tPA), an important mediator of cerebellar development (Li et al., 2006). In the new study, Li and colleagues injected a neuron stem cell (NSC) line into the cerebellum during the first postnatal week, and found that Purkinje neurons are maintained in normal numbers and with normal morphology and connections right into adulthood. Cell transplants resulted in functional recovery as well—mice that received the transplants showed better motor performance on the rotarod, and this correlated with the number of rescued Purkinje neurons.

Somewhat unexpectedly, the bolstered neuronal complement did not derive from transplanted stem cells. The neurons did not express the β-galactosidase marker of the NSCs, and many of the neurons in female mice had detectable Barr bodies (an inactivated X chromosome), indicating that they derived from the host and not the donor line, which was male. The injected NSCs did differentiate, gaining neuronal cell makers and losing stem cell markers, but none stained with the Purkinje cell marker calbindin.

Consistent with the idea that NSCs provided a supportive function, transplantation reduced high brain levels of tPA protein and mRNA to normal. This normalization rectified downstream alterations in channel function, synaptic structure, and neurotrophic factor levels, all of which contributed to normalizing cell survival and function.

Just how NSCs can depress abnormal gene expression is not clear, but cell-cell contact appears to be important. The researchers showed that NSCs could rescue Purkinje cells in culture, but that neuron survival required direct contact between the cells. This was consistent with in vivo results showing that neuron rescue required NSCs to move into the parenchyma: Surviving Purkinje neuron number was positively correlated with integrated NSC number, but not with the number of NSCs in the superficial meninges.

“The present study strengthens the concept of stem cells serving an underappreciated therapeutic role as ‘chaperone cells’,” the authors write, which “may represent a more tractable therapeutic strategy for this neurogenetic disorder than the more conventionally considered goal of cell replacement. In most neural systems, as in the cerebellum, the prospect of reconstructing the complexity of connections and feedback loops beyond the periods of normal neurogenesis seems quite staggering. Preserving extant circuitry clearly would be more approachable.”

The strategy of using stem cells to restore the neuronal environment and rescue host cells may also apply to other neurodegenerative disorders. Of course, treatment would need to start early before cells are totally lost, but the approach could be useful for slowly progressing diseases like Alzheimer disease, once the early stages can be readily identified.

A hit and a miss for RNAi
For many neurodegenerative diseases where the accumulation of misfolded, toxic proteins causes cell death, using RNA interference and antisense approaches to knock down the expression of specific proteins is a promising approach. That is exactly the path taken by Don Cleveland and colleagues at the University of California at San Diego, who report that delivery of synthetic antisense oligonucleotides directly into the CSF reduces levels of disease-causing SOD1 mutant mRNA and protein in the brain and spinal cord in a rat model of ALS. While the treatment did not delay onset of motor neuron disease, it did slow progression. The studies are a prelude to clinical trials in humans, which the researchers hope to start within a year, according to a press release on the study. The work, which was published online July 27 in the Journal of Clinical Investigation, offers a therapeutic strategy to downregulate almost any CNS protein.

One barrier to successful use of antisense oligos in the CNS is the problem of delivery. The researchers, led by joint first authors Richard Smith and Timothy Miller, tried delivering the oligos by osmotic pump into the cerebral ventricles. This method is already used in humans, they note, to deliver pain medication. Studies suggest that from the cerebral ventricles, the oligos would circulate in the cerebrospinal fluid, which bathes all regions of the CNS. Indeed, the authors showed that after a 14-day infusion in rats or rhesus monkeys, micromolar concentrations of oligonucleotides turned up in the brain parenchyma, as well as both the upper and lower regions of the spinal cord. They found oligonucleotides in both lumbar motor neurons and in non-neuronal cells.

Smith and colleagues then tested the ability of antisense SOD1 oligos to reduce protein levels in transgenic rats expressing the human G93A SOD1 mutant. These transgenic animals are a widely used ALS model. Even though the animals express very high levels of mutant SOD1 (5-10 times higher than endogenous SOD1), they effected a 40-60 percent reduction in mRNA and an approximately 25 percent reduction in protein when they administered a human-specific oligonucleotide. When they started infusion treatment in 65-day-old rats (~30 days before disease onset), they observed no change in onset, but the treatment did seem to slow progression, delaying the emergence of severe symptoms from day 122 to day 134, a 37 percent extension compared to the normal course of the disease.

Previous studies of viral delivery of antisense SOD1 in animals also showed promising results (Raoul et al., 2005; Ralph et al., 2005; Miller et al., 2005), but intraventricular delivery of oligonucleotides presents some advantages. First, the treatment can be easily regulated, with doses readily increased, reduced, or stopped. Also, in contrast to viral delivery schemes targeted only to neurons, oligonucleotides in the CNS get into other cells as well. This feature may be particularly important for ALS, where the expression of SOD1 in both neurons and surrounding glia is important in disease onset and progression (see ARF related news story and Boillee et al., 2006)

What of the application to other diseases? When the researchers infused antisense oligonucleotides to two Alzheimer disease targets, presenilin-1 and GSK3β, they saw a reduction in mRNA for the two proteins in the right frontal/temporal cortex after 14 days. While this does not ensure that protein levels would be reduced, the results suggest that the widespread delivery of oligos throughout the CNS could open opportunities to treat a number of diseases.

Animal toxicity studies are now under way in preparation for a planned phase I of the SOD antisense oligo in humans. If the treatment proves safe, it could open up opportunities to knock down many more targets. The authors cite, for example, proteins such as the amyloid-β precursor protein, β-secretase, tau, and presenilin-1 for AD, or huntingtin for Huntington disease, as targets of interest for future studies.

Any future trials will have to consider the specter of side effects. For RNA interference therapies, the possibility of off-target effects raises safety concerns. In particular, a recent report showed that the administration of viral vectors that produce short hairpin RNAs (shRNAs, the precursors to small interfering RNAs) can be fatal to mice (Grimm et al., 2006): By overwhelming the cellular machinery that produces endogenous microRNAs, the shRNAs perturb gene expression generally, which leads to cell death.

Now, another paper, this one from Bernardo Sabatini and colleagues at Harvard Medical School, shows an additional off-target effect of shRNAs that is specific to neurons. In work published in the Journal of Neuroscience on July 26, Sabatini, Veronica Alvarez, and Dennis Ridenour show that expression of shRNAs in neurons interferes with dendritic spine structure and function, and can result in decreases in synapse number. (For a complete description of off-target effects of interfering RNAs, see the comment below from Zuoshang Xu.)

The effects are independent of which mRNA is targeted, and even of the generation of siRNAs, but do depend on the sequence of the shRNA. In particular, shRNAs that induced an antiviral-type response, as measured by activation of an interferon target gene, affected neuron morphology and function, while shRNAs to the same protein that did not induce the interferon gene, did not. For research, the results demonstrate the need to use caution in using RNA knockdown data to probe protein roles in synaptic function and remodeling. More careful sequence selection and use of stringent controls must be the norm—the use of scrambled sequences is not sufficient, the authors say, and they stress the need for protein rescue experiments to ultimately prove the specificity of any observed effects.

For therapeutic RNAs, the same cautions will apply. However, synthetic antisense DNA oligonucleotides may have an advantage over shRNAs because, as Cleveland and colleagues argue, they do not require further processing and do not trigger an antiviral response.

Corrective measures
Finally, what about gene therapy—a permanent fix for those genetic errors that give rise to many neurodegenerative diseases? For Alzheimer, Parkinson, and Huntington diseases, and others whose familial forms spring from dominant gain-of-function mutations, the strategy of adding a good copy of the gene will not work. Instead, the goal of gene therapy must be to actually repair the mutations in the gene in situ. While novel methods have been devised for therapeutic recombination in cells (see ARF related news story), the ultimate goal is to find techniques that work in vivo.

A new approach that uses adeno-associated viral vectors comes closest yet to this elusive prize. In a report published online in Nature Biotechnology, researchers from David Russell’s lab at the University of Washington in Seattle used the vector to repair a model mutation in a β-gal reporter gene in mouse liver in vivo. The group, led by first author Daniel Miller, also repaired a disease-causing mutation in the GusB gene encoding the liver enzyme β-glucuronidase, although the frequency of repair (one to two cells per 10,000) was too low to demonstrate a therapeutic effect. For that enzyme, they estimate a 10-fold increase in the efficiency of recombination would achieve a clinical effect. Though gene replacement in the CNS will be many times harder than correcting liver enzymes, the work provides a tantalizing promise of what may be possible, one day.—Pat McCaffrey


  1. Alvarez et al. show that some shRNAs, when expressed in cortical neurons, can cause RNAi-target-independent effects, including shortened dendrites, reduced spine density and length, and decreased synaptic transmission. Based on these results, they caution that studies using RNAi should include proper controls. Particularly important is a rescue experiment using the shRNA-resistant target construct.

    This paper comes on the heels of several others raising a serious issue in designing RNAi experiments and therapeutic strategies. In general, there are three types of off-target effects in RNAi: off-target silencing; the interferon response; and the interference with the endogenous miRNA production.

    Off-target silencing is the silencing of the unintended genes (Jackson and Linsley, 2004). An important contributor to this is homology in the 5’ half of the siRNA (called the seed sequence) with the sequences of the non-targeted mRNAs (Jackson and Linsley, 2004).

    The interferon response is the expression and activation of antiviral genes (which include PKR as described by Alvarez et al.) by the cell after it is exposed to the double-stranded RNA. Many studies have contributed to the understanding of the mechanism whereby siRNA or shRNA triggers interferon response in mammalian cells. Among the contributing factors are the use of siRNA or shRNA synthesized by T7 polymerase (Kim et al., 2004; Sledz et al., 2003), the specific sequence motifs in the siRNA (Hornung et al., 2005; Judge et al., 2005), some characteristic of the shRNA expression vector (Bridge et al., 2003; Pebernard and Iggo, 2004), the inclusion or exclusion of the two 3’-overhang nucleotides (Marques et al., 2006), the length of the siRNA (Marques et al., 2006), the methods of siRNA or shRNA delivery (Heidel et al., 2004; Judge et al., 2005; Robbins et al., 2006), and even the purity of the chemically synthesized siRNAs (Robbins et al., 2006). Activation of the interferon response can trigger complex consequences that include apoptosis. Alvarez et al. showed that the off-target effects that they observed on neurons depend on the activity of PKR, suggesting that these effects are the consequence of the interferon response to neurons.

    The interference with endogenous miRNA is caused by introducing a large amount of shRNA or siRNA into the cells. This can saturate the endogenous miRNA processing and effecting machinery, leading to reduced function of the endogenous miRNAs (Grimm et al., 2006). One manifestation of a compromised endogenous miRNA function is the reduction in the levels of endogenous miRNAs (Grimm et al., 2006). Another manifestation is the increased expression of target genes of the endogenous miRNA. Perhaps this explains why a large number of genes were reported to increase when high doses of siRNAs were transfected into cells (Persengiev et al., 2004). Based on studies on Dicer knockout mice and other studies on miRNA function in developing vertebrates (Song and Tuan, 2006), it is clear that the miRNA function is essential for the survival of mammals. Thus, interference of the endogenous miRNA function can cause serious cellular dysfunction and even death (Grimm et al., 2006).

    With these multiple off-target effects, it is extremely important to carefully consider controls in reverse genetics experiments. Most useful in addressing these off-target effects are using multiple siRNA or shRNA to silence the target and conducting rescue experiments using the siRNA- or shRNA-resistant target constructs. A common character of these off-target effects is that they are siRNA or shRNA dose-dependent. Higher doses tend to trigger more severe off-target effects and can cause cell death. Therefore, the lowest effective silencing dose should always be used in reverse genetic experiments. Other helpful controls include measuring the expression levels of interferon response genes, checking the levels of endogenous miRNAs, and gene profiling of the RNAi treated cells.

    As far as the therapeutic RNAi is concerned, this report raises new challenges. Therapeutic siRNA or shRNA should be screened and the ones with the lowest effective silencing dose and highest toxic dose should be selected and thoroughly tested for their off-target effects. Ideally, the dose of the therapeutic siRNA or shRNA should be controlled so that it will not exceed the non-toxic range.


    . Induction of an interferon response by RNAi vectors in mammalian cells. Nat Genet. 2003 Jul;34(3):263-4. PubMed.

    . Fatality in mice due to oversaturation of cellular microRNA/short hairpin RNA pathways. Nature. 2006 May 25;441(7092):537-41. PubMed.

    . Lack of interferon response in animals to naked siRNAs. Nat Biotechnol. 2004 Dec;22(12):1579-82. PubMed.

    . Sequence-specific potent induction of IFN-alpha by short interfering RNA in plasmacytoid dendritic cells through TLR7. Nat Med. 2005 Mar;11(3):263-70. PubMed.

    . Noise amidst the silence: off-target effects of siRNAs?. Trends Genet. 2004 Nov;20(11):521-4. PubMed.

    . Sequence-dependent stimulation of the mammalian innate immune response by synthetic siRNA. Nat Biotechnol. 2005 Apr;23(4):457-62. PubMed.

    . Interferon induction by siRNAs and ssRNAs synthesized by phage polymerase. Nat Biotechnol. 2004 Mar;22(3):321-5. PubMed.

    . A structural basis for discriminating between self and nonself double-stranded RNAs in mammalian cells. Nat Biotechnol. 2006 May;24(5):559-65. PubMed.

    . Determinants of interferon-stimulated gene induction by RNAi vectors. Differentiation. 2004 Mar;72(2-3):103-11. PubMed.

    . Nonspecific, concentration-dependent stimulation and repression of mammalian gene expression by small interfering RNAs (siRNAs). RNA. 2004 Jan;10(1):12-8. PubMed.

    . Stable expression of shRNAs in human CD34+ progenitor cells can avoid induction of interferon responses to siRNAs in vitro. Nat Biotechnol. 2006 May;24(5):566-71. PubMed.

    . Activation of the interferon system by short-interfering RNAs. Nat Cell Biol. 2003 Sep;5(9):834-9. PubMed.

    . MicroRNAs and cell differentiation in mammalian development. Birth Defects Res C Embryo Today. 2006 Jun;78(2):140-9. PubMed.

  2. This is an interesting paper, which casts a new light on functional restoration of brain functions using neural stem cells. Unlike ES cells, which can theoretically give rise to any neurons in the brain, neural stem cells have a limited potential of differentiation repertories. In this paper, neural stem cell graft did not replace neurons but helped normalize the function and signals of host neurons with a genetic problem. This sort of supporting function has received some expectations for cell graft therapy for spinal cord injury, for which the injection of cells may somehow contribute to rejuvenation of the tissue environment and create more favorable conditions for reinnervation.

    It is worth testing whether a similar strategy is applicable to human neurodegenerative diseases.

  3. Smith et al. have shown the preclinical potential for antisense phosphorothioate oligonucleotides as therapeutic agents for neurodegenerative diseases, focusing on anti-SOD1 for ALS. Delivered as a saline solution via Alzet miniosmotic pump directly into the CNS, the antisense oligos are detectable throughout the CNS at micromolar (pharmacologically relevant) concentrations, most notably in the lumbar spinal cord region that is the earliest affected area in ALS rodent models. Similar results are obtained from infusion of an irrelevant tracking oligo directly into the brains of rhesus monkeys.

    Dose-dependent reduction of SOD1 mRNA is shown from transfection studies using cultured cells, but similar dose-dependent effects data are not shown from in vivo pump delivery of antisense SOD1 into animals. That’s too bad: it’s not clear what determines an “optimum dose” for rat or monkey CNS, and whether reduction in SOD1 was the only readout for determining the optimum dose, and the authors don’t report whether any toxicity effects or tolerability issues were observed at any of the tested doses.

    Using a SOD1 transgenic rat model as a test for therapeutic efficacy, the authors show that a month of pump-delivered antisense does nothing to delay onset or early weight loss of the treated animals—similar to saline. The authors do report a temporary shift in the survival curves (50 percent of treated rats are alive at a date when only 10 percent of control rats are still alive), but these curves converge at the end stage of disease, with steep losses in the treated group in the last 2 weeks, resulting in no clear extension of life. These data are consistent with the inability of the antisense to delay onset or weight loss stages of disease.

    The story as a whole is interesting and potentially promising. The Alzet pump delivery methods, requiring surgery for implantations and changes, are problematic and could limit the usefulness of such studies. And, why is there an apparent limit to reducing SOD1 to 40 percent of baseline? Is this an absorption issue, or other limit of the technology, or does this say something else about the expression of SOD1 gene and its mRNA? Either way, it’ll be awhile before this type of experimental therapy is applicable to Alzheimer disease, in which the time course of disease development is more gradual and over greater periods of time (even in animal models), as compared to ALS.

Make a Comment

To make a comment you must login or register.


News Citations

  1. ALS—Is It the Neurons or the Glia?
  2. A Lucky Break for Gene Therapy: Designer Nuclease Boosts Repair

Paper Citations

  1. . Purkinje neuron degeneration in nervous (nr) mutant mice is mediated by a metabolic pathway involving excess tissue plasminogen activator. Proc Natl Acad Sci U S A. 2006 May 16;103(20):7847-52. PubMed.
  2. . Lentiviral-mediated silencing of SOD1 through RNA interference retards disease onset and progression in a mouse model of ALS. Nat Med. 2005 Apr;11(4):423-8. PubMed.
  3. . Silencing mutant SOD1 using RNAi protects against neurodegeneration and extends survival in an ALS model. Nat Med. 2005 Apr;11(4):429-33. PubMed.
  4. . Virus-delivered small RNA silencing sustains strength in amyotrophic lateral sclerosis. Ann Neurol. 2005 May;57(5):773-6. PubMed.
  5. . Onset and progression in inherited ALS determined by motor neurons and microglia. Science. 2006 Jun 2;312(5778):1389-92. PubMed.
  6. . Fatality in mice due to oversaturation of cellular microRNA/short hairpin RNA pathways. Nature. 2006 May 25;441(7092):537-41. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Neural stem cells rescue nervous purkinje neurons by restoring molecular homeostasis of tissue plasminogen activator and downstream targets. J Neurosci. 2006 Jul 26;26(30):7839-48. PubMed.
  2. . Antisense oligonucleotide therapy for neurodegenerative disease. J Clin Invest. 2006 Aug;116(8):2290-6. PubMed.
  3. . Retraction of synapses and dendritic spines induced by off-target effects of RNA interference. J Neurosci. 2006 Jul 26;26(30):7820-5. PubMed.
  4. . Gene targeting in vivo by adeno-associated virus vectors. Nat Biotechnol. 2006 Aug;24(8):1022-6. PubMed.