Start here: what Huntington's disease is
Most diseases in this collection are caused by something that happens to you. Huntington's disease is caused by something written down before you were born.
In 1993, after a decade-long hunt, the cause was found: "a novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington's disease chromosomes" [1]. The gene is HTT, and near its start sits a stretch of the three letters CAG, repeated. CAG codes for glutamine, so a long CAG run produces a long polyglutamine tract in the huntingtin protein. Everyone has this repeat; the disease is what happens when it is too long.
The expanded protein misfolds and aggregates, forming intranuclear neuronal inclusions — a finding demonstrated in both human brain and transgenic mice, and shown to underlie the neurological dysfunction rather than merely accompany it [2] [3] [4]. The neurons that die first and worst are the medium spiny neurons of the striatum, the deep brain structure that gates movement, motivation and habit; selective striatal loss is reproduced in mouse models carrying the human mutation [5] [6] [7]. Astrocytes contribute too, with expanded huntingtin in striatal astrocytes impairing glutamate transport [8], and excitotoxicity is part of the final common path [9] [10].
What that produces clinically is a triad that unfolds over 15 to 20 years: a movement disorder — most characteristically chorea, involuntary flowing movements — together with cognitive decline and psychiatric symptoms, the last often arriving first [11] [10].
Who gets it. Huntington's is autosomal dominant: one expanded copy is enough, so each child of an affected parent has a 50% chance of inheriting it. That single fact shapes everything about how the disease is experienced — it is a family diagnosis, not an individual one, and it arrives with a decision about whether to be tested. Prevalence differs markedly between populations, and HTT haplotype structure explains part of the difference between Europe and East Asia [12] [13] [14].
Why it is in the news. For thirty years after the gene was found there was no therapy that touched the disease itself. That is beginning to change: huntingtin-lowering approaches have moved from concept to patients, with dose-dependent reduction of mutant huntingtin demonstrated in cerebrospinal fluid [15], and AAV-delivered microRNA against HTT has advanced from mouse models into the clinic [16] [17]. The Progress section takes this up, including what has gone wrong along the way.
Three pillars follow — measurements, management, and progress — with the disease's defining quantitative relationship in between.
Pillar 1: measurements and diagnosis
The diagnosis is a number
Huntington's is unusual, perhaps unique among common neurodegenerative diseases, in that the definitive test is a count. A blood sample, PCR across the HTT CAG tract, and the answer is an integer. The interpretation runs in bands:
- Under ~27 repeats — normal.
- ~27–35 — an intermediate allele. The carrier will not develop Huntington's, but the repeat is unstable in transmission and can expand into the disease range in a child. This is how the disease appears in families with no prior history.
- 36–39 — reduced penetrance. Some carriers develop Huntington's; some do not. Ten paternal transmissions of 37–39 repeat alleles were followed in collateral branches of families with de novo HD, and all ten descendants — many elderly — were without symptoms, supporting genuine reduced penetrance for repeat sizes that overlap the clinical range [18]. These alleles are not rare: direct assessment of 7,315 individuals from three population cohorts found roughly 1 in 400 people (0.246%) carry an allele of 36 or more [19] [20] [21].
- 40 or more — fully penetrant. The person will develop Huntington's if they live long enough.
That 36–39 window is a genuine grey zone, and it is a good illustration of why a genotype is not automatically a diagnosis [22].
The repeat is unstable between generations, and asymmetrically so. Analysis of 36 families found CAG repeats unstable on parent-to-offspring transmission, with instability "more frequent and stronger upon transmission from a male than from a female, with a clear tendency towards increased size"; all three juvenile-onset cases analysed had an expansion of paternal origin [23] [24]. That paternal bias is the mechanism behind anticipation — the disease appearing earlier in successive generations.
The ethics that come attached
Because the test is definitive and predictive, and because there is as yet no treatment that changes the outcome, a person at 50% risk can learn their fate decades before symptoms. Predictive testing is therefore delivered inside a structured genetic-counselling protocol rather than as a routine blood test, and the field has studied its consequences carefully: psychological functioning one year after the predictive test has been formally assessed [25], a UK consortium has reported 22 years of experience with the programme [26], and people at risk report real concerns about genetic discrimination [27]. Most people at risk choose not to be tested. That is a rational choice, not a failure of understanding.
Rating the disease, and seeing it before it starts
The clinical instrument is the Unified Huntington's Disease Rating Scale (UHDRS), which scores motor features (including chorea), cognition, behaviour and functional capacity, and which serves as the endpoint in the treatment trials below [28] [29].
But the disease starts long before it can be diagnosed. Predict-HD studied 438 gene-mutation-positive participants who did not yet meet diagnostic criteria and had no functional decline, modelling cognitive, motor, psychiatric and imaging measures against estimated time to diagnosis based on CAG length and current age. Detectable changes commenced "one to two decades prior to the predicted time of clinical diagnosis," a pattern robust across markedly different marker types [30] [31] [32] [33].
Imaging shows it as atrophy, and the striatum goes first: longitudinal MRI has charted the onset and progression of pathological atrophy [34], tensor-based morphometry detects structural change in preclinical carriers [35], caudate volume tracks with plasma markers [36], and PET shows progressive striatal and cortical dopamine-receptor loss [37].
Fluid biomarkers are the newest layer, and they matter most for trials. Measuring mutant huntingtin and neurofilament light (NfL) in CSF and blood alongside clinical and MRI measures, investigators found NfL correlated with all non-biofluid measures more closely than CSF mutant huntingtin did; CSF mutant huntingtin distinguished carriers from controls, while NfL — in both CSF and plasma — separated premanifest from manifest disease, and both were highly stable within individuals over weeks [38] [39] [40] [41] [42] [43].
Centerpiece: repeat length and age at onset
Huntington's offers something almost no other disease does: a single integer, measurable at birth, that predicts when a person will fall ill. It is the cleanest quantitative genotype-to-phenotype relationship in common medicine — and examining exactly how clean it is turns out to be the most instructive thing about it.
The relationship. Longer repeat, earlier onset. The correlation was established immediately after the gene was found: a "significant inverse correlation (p = 0.0001) between the age of onset and the CAG repeat length" [23]. It is not linear. The decline is steep at the low end — the difference between 40 and 45 repeats is decades — and flattens at the high end, where onset is already in early adulthood. A convenient way to write that shape is an exponential approach to an asymptote:
A(n) = A_∞ + (A_40 − A_∞) · e^(−k(n − 40))
with A(n) the mean predicted age at motor onset for repeat length n.
Grounding, and what is illustrative. The inverse relationship and its steepening are grounded [23]. The parametric form above and its three constants are illustrative and flagged — the published parametric onset model is not retrievable in this review's substrate, so no coefficients are quoted from it and none should be inferred from the figure.
What the illustration is not free to misrepresent is how much the relationship actually explains, and that number is grounded precisely: "CAG repeat length explains around half of the variation in age at onset, but genetic variation elsewhere in the genome accounts for a significant proportion of the remainder" [44]. The script therefore sets the residual spread so that CAG length explains exactly half the variance, and asserts it. The consequence is the wide band in the figure — and it reproduces the caveat the original 1994 report attached to its own correlation: "The observed scatter would, however, not allow an accurate individual prediction of age of onset" [23].
What the model explains. Three things.
First, why a genetic test cannot tell someone when they will get ill. A person with 42 repeats can be told a mean and a range spanning decades. For someone deciding whether to have children, change career or be tested at all, that distinction between "we know your genotype" and "we know your future" is the entire clinical conversation.
Second, where the other half of the variance lives — and it is not noise. Genome-wide association studies have identified variants associated with onset and progression [45] [46] [47] [48], and the strongest signals converge on DNA repair. Which leads to the third point.
Third, and most consequentially: the repeat keeps expanding inside neurons after birth. Somatic instability means the CAG tract in a striatal neuron at 40 grows far beyond the length in the blood sample taken at 20. The modifier genes are the machinery of mismatch repair and related pathways — MSH3 variants associate with a reduced rate of somatic expansion, delayed onset and slower progression in both Huntington's and myotonic dystrophy type 1 [49] [50]; FAN1 is protective, with increased expression associated with delayed onset and slower progression, and FAN1 binding to the expanded repeat reducing expansion in patient-derived medium spiny neurons [44] [51]; and Msh2, Mlh1/Mlh3, Pms2 and base-excision-repair stoichiometry all modify instability in models [52] [53] [54] [55]. The length of the uninterrupted CAG tract, independent of the polyglutamine it encodes, drives somatic instability and hastens onset [56] [57], and expansions are cell-type-specific within human brain [58].
That reframes the disease. The inherited number sets the starting conditions; a somatic process running for decades inside neurons determines the rest. And a process is something a drug can target — which is the last section.
Limits, honestly. The curve describes a mean, and the residual band is modelled as constant across repeat lengths, which it is not. Onset itself is a judgement call — motor onset is conventional, but psychiatric and cognitive changes precede it by years [11] [30]. Disease duration also varies with age at onset [59], and late-onset disease has its own character [60].
Pillar 2: management
There is no approved disease-modifying therapy. Everything in this section treats symptoms. Saying so plainly matters, because the gap between a definitive genetic diagnosis and an absence of disease-modifying treatment is the defining experience of this illness.
Chorea
The one target with well-established drug therapy is the involuntary movement. Tetrabenazine "selectively depletes central monoamines by reversibly binding to the type 2 vesicular monoamine transporter (VMAT2)" — reducing dopamine available for release and so damping chorea. In a randomised trial of 84 ambulatory patients, tetrabenazine reduced UHDRS chorea severity by 5.0 units versus 1.5 on placebo (adjusted effect −3.5 UHDRS units, 95% CI −5.2 to −1.9, p < 0.0001), with benefit also on global clinical impression [28].
Deutetrabenazine is the same molecule with deuterium substituted at metabolically vulnerable positions, which "attenuates CYP2D6 metabolism and increases active metabolite half-lives," giving steadier exposure. Ninety adults with manifest disease and a baseline total maximal chorea score of 8 or higher were randomised to deutetrabenazine or placebo across 34 Huntington Study Group sites [61], with subsequent work on conversion from tetrabenazine [62] and indirect tolerability comparison between the two [63] [64] [65].
Antipsychotics blocking the D2 dopamine receptor are the other option, and they carry a real trade-off: antidopaminergic treatment is associated with reduced chorea and irritability but impaired cognition [29]. Practice guidance and reviews weigh these choices [66] [67].
An honest caveat runs through this whole section: chorea is often not what disables people most. Cognitive and functional decline usually matter more, and nothing here touches them.
Everything else is multidisciplinary
Psychiatric care — for depression, irritability, obsessionality and apathy — physiotherapy, speech and swallowing therapy, nutrition, and family and caregiver support form the substance of care. Genetic counselling is central and continuing, not a one-off event at diagnosis [26] [25]. Registries such as Enroll-HD exist partly to make this population trial-ready [33].
Pillar 3: progress
The strategic logic is simple: if a single toxic protein causes the disease, reduce it. The execution has been hard.
Antisense oligonucleotides (ASOs) are short synthetic nucleic acids that bind HTT messenger RNA and trigger its degradation, lowering huntingtin production [68] [69] [70]. The first-in-human study delivered an ASO intrathecally to patients with early Huntington's and reported dose-dependent reductions in the concentration of mutant huntingtin in CSF, without serious adverse events [15]. That was a genuine milestone: a drug measurably engaging the causal molecule in the human brain. The subsequent large efficacy programme did not confirm clinical benefit and was halted — a reminder that target engagement and clinical benefit are different claims, and that huntingtin-lowering strategy remains under active redesign rather than settled [71] [72].
Gene therapy takes a one-time approach. An adeno-associated virus carries a microRNA against *HTT into the brain, where transduced cells produce the silencing RNA continuously. AAV5-miHTT delivered by bilateral striatal injection was evaluated in a severe homozygous mouse model using behaviour, MRI, magnetic resonance spectroscopy and striatal RNA sequencing, and improved brain health [16]. That construct is the basis of the clinical programme now reporting the first disease-slowing signals in people. Those clinical results are the reason this disease is in the news, but they are not* in this review's substrate, so no efficacy figures are quoted here; what is grounded is the preclinical evidence, the delivery strategy, and the surrounding field [17] [73] [74] [75] [76].
Other lowering modalities are in earlier development: RNA interference [77] [78], and zinc-finger transcriptional repressors that shut the gene off rather than degrade its message, with neuronal and astrocytic contributions dissected separately [79].
And somatic instability has become a target in its own right — arguably the most important conceptual shift in the field. If the repeat expands in neurons over decades, and if slowing that expansion delays onset, then the mismatch-repair machinery is druggable ground. The human genetics already provides the proof of principle: naturally occurring MSH3 and FAN1 variation shifts onset and progression in the direction that stabilising the repeat would predict [49] [44] [51] [56].
Biomarkers are what will make those trials readable. CSF mutant huntingtin gives direct pharmacodynamic readout of whether a lowering drug is working [39] [38], NfL gives a measure of ongoing neurodegeneration [40] [41], and the premanifest cohorts show that measurable change begins one to two decades before diagnosis [30] — which is both the opportunity and the difficulty. The people most likely to benefit from a disease-slowing drug are those who are not yet ill, and demonstrating benefit in them requires either very long trials or biomarkers trusted enough to substitute for outcomes.
Older mechanistic lines — enhancing autophagic clearance of mutant huntingtin [80] [81], histone deacetylase inhibition [82], targeting ER stress and excitotoxicity [83] [9] — have not yet produced a therapy but built the understanding the current programmes rest on [10].
Dig deeper in lmmol
Related reviews:
- Alzheimer disease — the other neurodegeneration review here, and a useful contrast: both are protein-aggregation disorders, but Alzheimer's cause is multifactorial while Huntington's is a single fully penetrant repeat, which is why the therapeutic logic differs so sharply.
- Sickle cell disease — the closest structural sibling in this collection: another monogenic disease with a clean genotype-to-phenotype relationship, and the one where gene therapy has already reached approval, which is the road Huntington's is now on.
- The health reviews index collects the rest of the series.
The molecules this review turns on:
- Synaptic vesicular amine transporter (VMAT2) — the tetrabenazine and deutetrabenazine target; depleting vesicular monoamines is what damps chorea [28] [61].
- D(2) dopamine receptor — the antipsychotic target used as the alternative for chorea, with its cognitive trade-off [29].
- Tetrabenazine — the compound record for the first drug approved for Huntington's chorea.
- Huntingtin itself does not yet have a static page in this graph; use the graph index, all proteins, or all diseases rather than guessing an entity URL.