Genetic Modifiers of the HD Phenotype in Mice and Men

Reported by: Lisa J. Bain

Abstract:

Variation in age of onset of Huntington’s disease is highly correlated with CAG repeat length. However, when variation in repeat length is controlled for, variation in onset is still highly heritable, suggesting that other genetic factors modify the disease process. At a workshop sponsored by The Hereditary Disease Foundation in October 2002, 19 scientists met to discuss research aimed at identifying genetic modifiers in animal models and humans. Assays to screen for modifiers were described in Drosophila melanogaster, C. elegans, and mouse models. In C. elegans, one genetic modifier, dubbed pqe-1 (polyglutamine enhancer-1) has been identified in the laboratory of Anne Hart. Studies are underway to further characterize this low abundance but strongly enhancing protein; and to understand how it enhances polyglutamine toxicity. Searching for modifiers in mice is complicated by the variability in both background strains and the transgene itself, i.e., whether it is full length or truncated, the length of the CAG repeat, whether it is a knock-in or transgene, and which promoter is used. Two possible genetic modifiers were discussed: caspase-1 and Msh-2. In humans, investigators must determine which disease characteristics they wish to modify: age of onset, progression of disease, individual phenotypic features, or all of these characteristics. The next step will be to decide which approaches are most likely to be successful in assessing changes in these characteristics. Katrina Dipple described searching for genetic modifiers by looking for SNPs in candidate genes identified by understanding the basic biology of another rare disease, glycerol kinase deficiency. The workshop ended with more questions than answers, and with some clarification of what lies ahead.

            In Huntington’s disease (HD), the gene for the protein huntingtin mutates by expanding the number of CAG repeats, resulting in a protein with an expanded glutamine tract. The length of the expanded CAG repeat correlates inversely with the age of onset of disease. Individuals with CAG repeats below a certain threshold (about 35 repeats) may live to old age with no symptoms, while those above 40 repeats almost certainly will manifest symptoms at some point. Between 35 and 39 repeats there is variable penetrance; a phenomenon once thought not to occur in HD families. More than 60 repeats results in juvenile onset HD.

            However, when the variation in repeat length is controlled for, variation in onset is still highly heritable, suggesting that other genetic factors modify the disease process. Moreover, genetic factors may also modify the phenotypic expression of the disease gene, resulting in variable symptomatology. On October 15, 2002, The Hereditary Disease Foundation (HDF) convened a workshop in Baltimore, Maryland, to discuss research aimed at identifying genetic modifiers of the HD mutation. Nineteen scientists working in flies, mice, worms, and humans shared their research findings. The goals of the workshop, as set forth by David Housman and Carl Johnson, were to identify genes that modify the HD phenotype and develop hypotheses worth testing in terms of translating modifier genes into therapeutic targets.

Fly models 

            Juan Botas started the discussion by describing the work his lab has been conducting in Drosophila melanogaster. The HD models he uses are htt exon1 transgenic flies with either wild-type or a mutant number of glutamine repeats (Htt-16Q and Htt-128Q). The exon 1 transgene is silent in these flies unless they are crossed to another strain with a driver. Strains with different drivers can elicit transgene expression in different cell types or even in specific neurons. He further modulates expression of the transgene by manipulating temperature within the normal culture range. With temperature variation, moderate to very strong phenotypes can be seen in the same strain. The capacity to keep the expanded htt transgene silent is important because it allows him to keep high-expressing transgenic lines healthy and with constant phenotypes through many generations.  Botas said he is also working on an inducible full-length fly model, which is not yet ready.

Three assays are used to screen for modifiers: eye degeneration, survival, and climbing (a locomotor/behavioral assay that assesses the ability of flies to climb the walls of a glass vial). Drosophila melanogaster expressing the mutant htt gene in the eye retina develop a neurodegenerative phenotype that has been well characterized, and can be easily scored by visual examination of the eye in living flies (i.e., no need to do histology or immunocytochemistry). For the survival assay, Botas is expressing the transgene at high levels with a GAL4 driver inserted into a pan-neural gene. This drives expression in every neuron in both the central and peripheral nervous systems.

The neurons of these flies have big aggregates, with obvious disintegration of nuclei and projections, said Botas. “We don’t know why these flies are dying,” he said. “For the purpose of establishing a survival assay, we don’t want something that kills, very early, 100% of the flies. We’re looking for an expression level that leads to, say, 90% lethality so we can tune it.”  Botas said they are able to modulate the survival phenotype either pharmacologically (e.g., SAHA) or genetically (e.g., chaperones).

The wall-climbing assay is very quantifiable, said Botas. At birth, HD flies are indistinguishable from wild-type flies in terms of wall climbing. Climbing ability begins to decline in wild-type flies at about day 35 or 40; however, in HD flies it begins to decline at day 22. Botas stressed the importance of having several independent assays for validating the effects of modifiers. Any given assay is going to produce some false positives that reflect the shortcomings of that particular assay. Testing modifiers in other independent assays is an efficient way to filter out these false positives. In the Botas lab, hits are confirmed histologically by looking for an improvement in neuronal integrity, shown by a reduction in nuclear inclusions and an increase in projections of neurons to which expression has been directed. The primary screen varies depending on the purpose of the test, said Botas. For drug screening, the primary screen would be survival; while for genetic modifiers the primary screen would be eye examination.

Two types of modifier screens can be carried out in Drosophila: 1) screens for genes that modify the HD phenotype when their activity is reduced; and 2) screens for genes that modify the HD phenotype when their activity is increased. In searching for genetic modifiers that work by reduction of gene activity, they use heterozygous mutations and thus are looking for modifier genes that alter the HD phenotype when activity is reduced up to 50%. This is very important because most modifier genes may not be recovered under conditions of severe lack of function if they are required for normal development or viability. For the most part, they are using EP- and P-element mutagenesis in primary screens, but have also done screens fro candidate genes using all kinds of mutations.

Botas’ lab has run similar screens to look for genes that modify ataxin-1-induced neurodegeneration[i]. Spinocerebellar ataxia type 1 (SCA1) is a human neurodegenerative disease that results from a CAG expansion in the ataxin-1 gene. To conduct these screens, 3000 lines with P- and EP-element insertions that have been previously characterized are obtained from a stock center. These flies are crossed to SCA1 flies and screened for the eye phenotype. Hits are screened further with the climbing assay; those with no phenotype are discarded.  About one-fourth of the modifiers obtained suppressed neurodegeneration in the SCA1 flies; the other three-fourths enhanced neurodegeneration. The study identified modifiers involved in protein folding, protein clearance, RNA processing, transcriptional regulation, and cellular detoxification. Several of these same genetic modifications may be relevant in studies of HD and other polyglutamine diseases.

Enhanced toxicity modifier found in worms

            Anne Hart next described the studies her lab is doing with C. elegans. They have expressed exon 1 of the htt gene with either Q23, Q95 or Q150 in the neurons of C. elegans.  They use an endogenous promoter that drives expression in four classes of neurons. The only transgene that affects normal worms is Htn-Q150, which causes neurodegeneration in 15% of the neurons at 8 days of age. Given the low penetrance, they looked for enhancing modifiers.

            Surprisingly, they got only one enhancer, but it is a strong enhancer. They identified seven independent alleles of this gene, which they named PQE-1 (polyglutamine enhancer-1)[ii]. It enhances Htn-Q95 but not Htn-Q2; and more dramatically enhances ataxin-1 with expanded polyglutamine. It also modestly enhances ataxin-1 Q2. 

            PQE-1 is a low abundance, but large protein, about 1600 amino acids. It is alternatively spliced, with three domains in the longest protein. One domain is a long QP rich stretch; another a highly charged domain with a nuclear localization signal; and a third, an exonuclease domain near the C-terminus. More detailed studies of the alternative splice forms suggest that the QP domain matters. It does not appear to interact with other proteins shown to be important in HD pathology: CREB, CBP, or HDAC.

            According to Hart, the level of PQE-1 is a dramatic modifier of polyglutamine toxicity. Even if half of the dose is removed, toxicity is modestly enhanced. “We’re guessing it’s a required gene,” she said. This would explain why no null mutant has been seen. However, the partial loss of function alleles on hand don’t seem to affect the lifespan of worms and in animals with no transgenes, the behavior is normal.

            “We’re still trying to figure out why it’s protecting, why it’s modifying,” said Hart. They are also doing suppressor screens. So far they have screened 27,000 animals and have 7 suppressor strains that vary from 30% to 100% survival.

            Nancy Wexler asked what would be involved in taking Hart’s system to flies. Hart said they would have to clone the cDNA behind the Gal4 promoter into one of those constructs; then they would have to make transgenic fly lines. Carl Johnson suggested that, rather than trying to express this huge protein in other systems, it might help to break down the sequence and figure out what is the critical set of amino acids that are having the effect.

            Another way to address this problem, said Hart, would be to do RNAi (RNA interfering) screens. The biggest problem she said is that neurons are somewhat refractory to RNAi.  RNAi is easy to study in worms, because it can be put inside of bacteria that the worms will eat. Each cell in a microtiter plate could contain bacteria with a different RNAi, which the worms would eat. The plates could then be easily screened to look for an effect.  In contrast, transgenes would be needed to do RNAi screens in flies, said Juan Botas. Marcy MacDonald mentioned that human siRNA expression libraries are already becoming available.

            Carl Johnson mentioned that Michael Sherman has looked for expressors and expanders in yeast. He has found that deletion of the RNQ1 prion gene suppressed both aggregation and toxicity. The major lesson from yeast, said Johnson, is that the protein can flip between a prion and non-prion configuration.

            “It hasn’t escaped our attention that this could be a prion,” said Anne Hart.

Mouse Models for Modifier Studies

            Numerous mouse models have been developed for studying HD, raising the question of which strains would be most useful for genetic modifier studies. Significant variability in phenotype arises from variations in background strains as well as the transgene itself, i.e. whether it is full length or truncated, the length of the CAG repeat, whether it is a knock-in or transgene, and which promoter is used.

            The question is, are any of these mice good models of the human condition, asked Chris Ross. “All of these mice so far have interesting phenotypes. We don’t know where they come from; we don’t know why the mice die; we don’t know which parts of the brain are responsible for the behavioral phenotype; and really almost none of these mice have very much in the way of neuronal degeneration.”

            David Housman added a further complication: clinicians take extraordinary measures to keep people with HD alive during the late stages of their illness, yet mice die for different reasons, such as that they can not reach their food or water. No one has ever done studies in mice to see what happens if they are kept alive with feeding tubes and other measures.

 What is known, however, is that humans with relatively subtle symptoms still have massive neuronal loss in the striatum. And functional measures that are done in human patients for motor coordination, rigidity, bradykinesia, etc. do correlate with cell death, said Ross. “So I do think cell death is important from a clinical standpoint, and therefore it’s important to try and make mouse models that can reproduce neuronal cell loss.”

Housman asked “What if there is more or less a fixed amount of time, rather than age, before a neuron dies?”  Dysfunction could persist for a year and half before the neuron dies, and by that time a mouse would have aged itself out of the study. 

Marcy MacDonald and Vanessa Wheeler described their work with HD knock-in models. “We’ve approached knock-ins like they were people,” said MacDonald. All lines are on an outbred background. Phenotypes are identified on an outbred background and then the alleles are put on an inbred background. They did a meta-analysis of phenotype versus CAG repeat length -- including biochemical and mRNA phenotypes, n-terminal fragment accumulation, neurodegeneration at the level of the medium spiny neuron, reactive gliosis, neuronal morphology – and showed a “nice repeat length dependence.[iii]”

Wheeler said that in mice with 111 repeats they saw striatal accumulation of protein in the nucleus at about 6-10 weeks, inclusions at 10-12 months, and neuropil aggregates at about 17 months; but nothing degenerative or behavioral until 18 months, although MacDonald acknowledged that they haven’t had the money or expertise to behaviorally phenotype these animals in the way that it needs to be done. Darren Monckton mentioned that Peggy Shelbourne sees a subtle phenotype at 3 months in her knock-in mice.

MacDonald suggested thinking of knock-ins as resembling juvenile HD in the first three years of life, before obvious symptoms have appeared. That would mean phenotyping for things that are striatal specific, rather than cerebellar specific such as rotarod. “The tests aren’t really out there,” she concluded.

Carl Johnson emphasized the importance of this. “From the point of view of chemical screens, one would like to use the weakest allele you possibly could; the allele which is closest in many senses to the human phenotype.”  For chemical screens, the R6/2 may have too strong a phenotype if it is used alone with no other models, raising the likelihood of getting false negatives. Screening for genetic suppressors is very different than screening for chemical suppressors, continued Johnson. With genetic mutations, strong mutants will generally be seen first; then you have to work to get weaker mutants that give partial loss of function.

Jim Gusella added that if you detect an effect with a weak allele, your strong allele could be a secondary assay. “The fact that it doesn’t work in the strong allele would mean had you used the strong allele in the primary assay you never would have seen it. But it doesn’t mean you should throw it away. In humans, you’re not trying to cure a strong allele, you’re trying to cure a weak allele [except in juvenile cases.]”

Genetic modifiers that have been investigated to some degree include caspase-1 and Msh-2.  Marcy MacDonald’s lab has tested the dominant negative caspase-1 allele in  knock-in mice, and found no early effects. Anne Messer’s lab was the first to report that mismatch repair enzymes are required for somatic repeat expansion; and she is continuing to study this phenomenon using multiple transgenics and knockouts on inbred backgrounds. Darren Monckton said that at least two HD mouse models have been crossed onto Msh-2 knock-outs leading to suppression of somatic mosaicism.  Pms2 and Msh3 have also been shown to suppress somatic mosaicism, while Msh6 appears to increase the rate of somatic expansion.  Given that longer repeats cause a more severe disease and the repeat continues to expand throughout life, it is logical to suggest that suppressing somatic expansion would lead to a reduced phenotype. Consistent with this, MacDonald revealed that the rate of aggregate formation in HD knock-in/Msh2 deficient mice was slower than in the HD knock-ins alone.

The point is, said David Housman, that there are many different ways to modulate the pathogenic process. Finding a system in which they can be studied will be necessary to sort out all the different pieces and the pathways that are affected.

Modifiers in humans

            What do you want to modify? The final session of this one-day workshop was devoted to discussion of modifiers in humans. “What do you want to modify?” asked Housman. Age of onset? Rate of progression? Or something else?  Nelson Freimer suggested looking more broadly at how variable the phenotype is: Do different phenotypic characteristics progress at different rates?

            Darren Monckton raised two questions: Is variation in the rate of progression heritable, and is it distinct from variation in age of onset? Or are they linked? In other words, if a person has an earlier onset than predicted by their repeat length, would you predict they would also have a more rapid rate of progression than is average for the HD population? How do you measure rate of progression?  Richard Myers suggested that duration (the time between onset and death) is not the same as rate of progression, because people may live an extra 10 years with the best terminal care available. But Jim Gusella said that in one study on HD confirmed brains from the brain bank, CAG repeat length correlated similarly well with both age of onset and age of death. So if you look at the interval between onset and death (duration), it doesn’t correlate well with CAG length. Myers argued that everyone with HD, both juvenile-onset and later-onset, lives approximately 20 years after onset. Those with onset at age 50, however, face many other age-related problems that could hasten their death than would those with childhood onset.

            What measures to use. Housman asked how one would assess the effectiveness of genetic modifiers if death is not used as an endpoint? Gusella said you must define what it is you are trying to treat. “In Huntington’s disease, you’re not trying to prevent death, you’re trying to prevent dysfunction much earlier on. So you’ve got to measure that.”

            Russ Margolis suggested breaking the disease into different components: chorea, voluntary motor function, dementia, etc. Presumably each of these components could be susceptible to different modifiers.  A problem with assessing these features clinically is that they fluctuate considerably from day to day. The way around this in a clinical study is to have a very large sample size, long follow-up with lots of data points, and good controls.  Richard Myers noted that this type of analysis might be easier to do in Venezuela because people see the doctor on the schedule imposed by the researchers, rather than when they don’t feel good. In contrast, the Huntington’s Study Group (HSG) data is collected through “ad-lib” clinical follow-up, creating data with lots of variance.

            Nancy Wexler said that they have collected 22 years of data in Venezuela, where they follow people prospectively and try to examine about 1,000 people per year. “So we have almost total ascertainment,” she said. She asked what kind of measures should be done that would help identify modifiers.

            Nelson Freimer added that, assuming there are modifiers to be identified, they probably do not exist to be modifiers of HD.  These are things that you might expect to see segregate with families independent of HD. For that reason, it would be interesting to collect data from non-gene carriers in a family. Wexler said that some data from neurological exams have been collected from unaffected siblings in HD families. These family members would likely share similar environmental factors including diet with their affected siblings.

            Housman joked, “In flies, worms and mice, we’d call them controls.”

            MacDonald chimed in, “Littermate controls!”

            Katrina Dipple asked whether psychiatric testing is done in these Venezuelan families, or development testing in younger members of the family. Wexler noted that cultural differences make it hard to find appropriate measures.  Dipple noted that modifier genes are likely to affect subtle characteristics such as learning disabilities that may be hard to pinpoint, she said.  

            Richard Myers suggested using weight loss as a phenotypic measure. Weight is readily measured and shows up in both humans and mice with HD. Other measures that could prove useful in identifying modifiers include chorea, voluntary movement, results from the Stroop and Trails tests (measures of executive function,) retinal scans and MRI scans. Biomarkers in blood samples could also prove useful. Although biomarkers might not correlate with all these other measures, the robustness of the test may yield valuable information, said Andrew Schectel. Darren Monckton said that DNA stability can be measured in blood and might be a useful measure.

            Human population studies.               Several different groups have reported similar results in heritability studies. Richard Myers said in his groups’ study, after adjusting for CAG repeat, age of onset was still about 70% heritable.

Michael Andresen said that in Venezuela, 72% of the variance in age of onset is explained by CAG repeat length, and heritability of the remaining variance is about 78%. Andresen commented that one might expect heritability to be higher in Venezuela because of a higher shared environmental component.  Venezuela has the added advantage that sibships can be compared. “You would expect first degree relatives to have twice the heritability of second degree relatives if it’s purely genetic; and that in fact is not the case,” he said. “You see a higher than expected heritability for second degree relatives compared to first degree relatives, suggesting that there is a significant component of that heritability which is due to shared environment and not to genes.”

Housman added that the heritability of excess variants is better if you eliminate the higher end variants. With greater than 50 repeats, there is much less variation in age of onset. After 60 repeats, onset will be 20 years or longer.

How to look for modifier genes. Several different approaches for finding modifier genes were discussed. The only published, confirmed modifier gene in HD is the GRIK2 gene, which codes for the kainate GluR6 receptor subunit[iv][v], according to Jim Gusella. This gene appears to have a big effect in a very small number of people, and therefore doesn’t account for a lot of the variance.

Katrina Dipple suggested doing a candidate gene approach using what has been learned in Drosophila and mice. Using patient samples, one could look for SNPs or mutations in regions related to some aspect of disease progression, such as apoptosis. Michael Andresen said the Venezuela group had constructed a list of 200 highly probable candidate genes and is doing a genome scan of patients.           

             Darren Monckton offered an approach for prioritizing genes within such large candidate lists. By looking at the natural variation in a gene, one might be able to find polymorphisms that stand a good chance of affecting the function of the gene (e.g., coding variants) and, hence, increase the likelihood that they may function as modifiers in human populations.

            Dipple has been using the candidate approach to find genetic modifiers that affect glycerol kinase deficiency (GKD), a rare inborn error of metabolism. Using this approach, she has identified three potential genes that are known to be involved in fat and/or carbohydrate metabolism and predisposition to non-insulin dependent diabetes. She looked at other genes in which there are known polymorphisms or mutations that cause or predispose to diabetes. She found that those proteins (transcription factors and enzymes in the related metabolic pathways) may also be important for glycerol kinase expression and function. “The more you can learn about the basic biology of the gene product, the more you can target where to look for modifiers,” she concluded.  

            Jim Gusella pointed out a strong reason for using a candidate approach. “You may detect a strong effect in a small subset of people, but more importantly, it really is to me the only way of a priori deciding whether the particular biochemical mechanisms, pathways, or whatever, that you’ve been studying in model systems are really relevant to these diseases.” For example, he said, people have talked about excitotoxicity being important in HD for 20 years, but there was no way to really determine what role it played. Finding modifiers of excitotoxicity that have an impact in HD models might clarify the role of excitotoxicity and point to possible interventions.

 What about non-genetic factors?

            At the conclusion of the workshop, other questions arose. As Darren Monckton said, “The idea was to talk about genetic variation, but maybe we need to understand environmental variation.” Nancy Wexler said the Venezuela group has been trying to study environmental effects by doing an urbanization rating, which takes into account pollution in lakes, smoking, etc. Another variable that affects the data is interrater reliability. 

            Scientists left the meeting with many more questions than answers, yet with some clarification about the direction that genetic modifier studies may take.  Carl Johnson and Nancy Wexler expressed the strong desire to facilitate cooperative research in whatever way the Foundation can.