Dr. Kaplitt had just bored a hole about the size of a quarter through the top of Mr. Klein's skull, in preparation for an ambitious experiment: the infusion deep into the brain of 3.5 billion viral particles, each bearing a copy of a human gene meant to help relieve the tremors, shuffling gait and other abnormal movements caused by Parkinson's disease. . .The article cites quite a chorus of skeptics, who think that the approach will either be ineffectual or actually make things worse:
Genes alone cannot get into cells, but viruses can, and in gene-therapy experiments viruses are commonly used to carry genes to their destination. Dr. Kaplitt and Dr. During chose the virus AAV, or adeno-associated virus. It does not cause disease in people, Dr. Kaplitt said, and its genetic material is removed.
In experiments in mice with a disorder that is intended to mimic Parkinson's, the gene therapy helped all the animals somewhat and helped about half of them a great deal, Dr. Kaplitt and Dr. During reported last October in the journal Science. They have also tested the treatment in monkeys but have declined to discuss the results, because they have not yet been published.
"You don't have to take the risk of putting in a virus and you don't have to take the risk that it's uncontrollable," Dr. Olanow [chairman of the department of neurology at Mount Sinai School of Medicine] said. "The danger is that if you inhibit too much you can induce wild, flinging movements which people have been reported to die from." . . .Dr. Kaplitt himself says that there is no evidence for an immune response in his animal models, which makes sense given the relative isolation of the brain from the immune system. On the other hand, not having published primate studies before going ahead with a human clinical trial is a recipe for inviting skepticism.
Another potential danger is that the virus could spread to other areas of the brain, wreaking destruction, said Dr. Inder Verma, a gene therapy researcher at the Salk Institute, in San Diego, and past president of the American Society of Gene Therapy. . .
Even if the virus does not spread in the brain, it could elicit an immune reaction. "You may get a brain inflammation and swelling," Dr. Verma said. "You may lose some neurons."
AAV is a very interesting virus from the gene therapy p.o.v. (you can read more about it at John Kimball's excellent biology site); among its chief attractions is the fact that the viral DNA (in this case, the viral genes have been deleted, and replaced by a transgene of interest) stably integrates into a single site within the human genome -- i.e. from one patient to the next, one cell to the next, the viral integration event should be both predictable and identical.
What makes this important? First, having a gene therapy vehicle that integrates into the human genome is extremely useful. Many so-called vectors are derived from DNA viruses like adenovirus, the genomes of which normally (i.e. in their unmodified form) sit aloof in the nucleus of an infected cell, unwilling or unable to integrate. From the viral point of view, this is no problem, because they can replicate themselves there and, once they're ready to lyse the host cell and move on, their DNA is nicely separate and ready to go.
On the other hand, if one uses an adenovirus (stripped bare of all the genes that normally allow it to replicate and kill host cells) as a gene therapy vector, the potentially curative gene that you hope to introduce will, like the rest of the viral genome, hang out independent of the host cell genome. If the host cell divides, the gene therapy DNA will be diluted away, since the chromosomal replication machinery can't get to it. In addition, most cells don't like to have this sort of non-chromosomal DNA hanging around their nucleus, and over time will either degrade it or modify it so that its genes become permanently inactive. Non-integrating DNA vectors therefore will probably not effect lifelong cures, but rather might serve as treatments that need to be periodically repeated.
The second advantage of AAV, among integrating vectors, is that it goes (in theory) only to a single place in the genome, a location away from other endogenous genes. This distinguishes it from the other major gene therapy vehicle, the retrovirus, which also integrates during infection but does so at relatively unpredictable spots within the genome. When doing so, it can accidentally land near one of your own genes, and this will disrupt the normal transcriptional regulation of that gene.
A useful property of the retrovirus, from the gene therapy perspective, is that almost its entire genome can be removed (to prevent the untoward results of infection) and replaced by your gene of interest; what remains of the original virus is a so-called transcriptional enhancer, which directs high-level expression of the therapeutic gene that you hope to introduce. However, as I discussed in January, this viral enhancer can also interact with nearby host genes, including ones that are potentially dangerous, so-called proto-oncogenes. These genes, when active, promote proliferation, and thus are usually kept silent in resting tissues. If a retrovirus lands nearby, though, the gene will be expressed at high levels, driving proliferation and promoting tumorigenesis.
This explains the disturbing news that prompted my January post, that two children in a retroviral gene therapy trial to cure their immune deficiency had gotten leukemia, due to viral integration next to the same gene, called LMO-2. The odds of this seemed staggeringly low, but it turns out to be less unlikely than it seemed back then, thanks to a fascinating if distressing Science article that came out in June.
The authors of that paper infected human cells, in vitro, with the same sort of retrovirus as used in the SCID gene therapy trials, and then looked at where the viral genome ended up. Was it like throwing darts at a normal dartboard, where they should land at random (assuming that the thrower is as unskilled at darts as myself), or was it more like a dartboard with very strong magnets hidden behind it, biasing the target sites?
The answer was emphatically the latter: the viral genome was found near one or another gene 34% of the time, versus a predicted 22% for a purely random integration, and even more strikingly around 20% of the integration events were very close to the beginning of a gene's transcriptional start site. This means that possibly 20% of viral integration events will be poised to upregulate expression of a nearby gene. The authors note the dire implications for human gene therapy:
In the X-linked severe combined immune deficiency syndrome clinical trials, >5 x 10E6 cells with MLV integrations were injected into each child. Assuming that 20% of integrations are near transcriptional start sites, there will be 1 million integrations distributed among the 18,214 RefSeq genes or an average of 55 integrations into the 5' region of the LMO2 locus per treatment. [my emphasis]Suddenly it seems like good luck that only two of the kids in the trial ended up getting leukemia, given that every one must have gotten multiple cells with hyperactivated LMO2 genes. (The possible explanations for the relative rarity of leukemia, given these numbers, might make an interesting story for another time.)
In the Times story, Dr. Kaplitt is portrayed injecting 3.5 billion viral particles into his patient's brain: with a retrovirus, this would lead to over 1,000 times more potentially dangerous integration events than even the SCID patients experienced. Since AAV genomes should all land in the same place, away from any proto-oncogenes, this worry is considerably diminished.
An interesting side note comes from a recent New Scientist story on a human retrovirus, HHMMTV:
New evidence for a link between a virus and human breast cancer has been revealed in a series of studies by Australian researchers. The virus, dubbed HHMMTV, is very similar to a version known to trigger mammary cancer in mice.This work is still controversial, but very provocative. How exactly might HHMMTV cause breast tumors? Exactly the way that the SCID gene therapy virus caused leukemias: by integrating next to a proto-oncogene.
The researchers stress that they have not proven that the human form causes cancer in people - but if it does, its raises the possibility of developing a vaccine against the deadly disease.
A team at the Prince of Wales Hospital in Sydney published research in March that found the virus in 19 of 45 breast cancer biopsies taken from caucasian Australian women (Clinical Cancer Research, vol 9 , p 1118). In contrast, they identified HHMMTV in less than two per cent of normal breast tissue samples.
A related mouse virus, MMTV (mouse mammary tumor virus), causes tumors in this way, and in this way helped researchers identify some of the first proto-oncogenes: take a mouse, give it MMTV, wait for tumors to appear, and then ask what cellular genes lie adjacent to the viral insertion site. These experiments were part of what netted Harold Varmus his share of the Nobel Prize in 1989. This strategy, termed insertional mutagenesis, is still used widely today in the mammary gland and other mouse tissues; perhaps we should be less surprised at the fact that retroviral gene therapy can have tumorigenic side effects than at the fact that such effects are so rare!