Question bank — Part D: Spinocerebellar ataxia type 12
Covers D1-D5, with the final section drawn from lab 11 and lab 12.
Convert to Anki: python reference/to_anki.py question-banks/ --out anki-import.tsv
The neuron, the cerebellum and selective vulnerability
Q: A neuron at rest is often described as doing nothing. What is it spending ATP on, and why does that matter for a degeneration? A: The Na/K-ATPase burns ATP to push 3 Na out for every 2 K in, holding a resting potential near -70 mV, with an honest textbook range of -40 to -90 mV across cell types. In the Attwell-Laughlin budget for rodent grey matter, action potentials take 47% of the signalling bill, postsynaptic effects of glutamate 34%, resting-potential maintenance 13% and glutamate recycling 3%. That is a calculated budget for rodent grey matter and the split has since been revised, so use it only for the qualitative point: firing dominates the bill, and a cell that fires continuously runs a permanent ATP deficit it must keep servicing.
Q: The is about a tenth of brain mass. What fraction of the brain's neurons does it hold, and what follows? A: About 80% -- roughly 69 billion of the brain's ~86 billion neurons, in ~154 g of tissue, against ~16 billion (~19%) in the cerebral cortex plus its white matter at ~1,233 g. Almost all of them are granule cells, the smallest neurons in the vertebrate brain. "Cerebellar disease" is therefore not disease of a small corner of the brain. The same study also retired the 10:1 glia-to-neuron ratio: non-neuronal cells number ~85 billion, so the ratio is about 1:1.
Q: How many does a human cerebellum contain? A: Nobody knows to better than a factor of two, and the disagreement is the fact. Fractionator counts of nucleoli give 15.4 x 10^6 (CV 19%, n = 12); optical-disector counts of nuclei give 30.5 x 10^6 (CV 0.13, n = 5). Granule-cell estimates conflict the same way: 101 x 10^9 by stereology against an implied 69 x 10^9 or fewer for all cerebellar neurons by isotropic fractionator. Quote 15-30 million Purkinje cells and roughly 70-100 billion granule cells, and say the spread is a fact about counting methods, not about people.
Q: In what order does a signal travel through the cerebellar cortical circuit, and which cell is its output? A: Mossy fibres from the pontine nuclei (also spinal cord, vestibular nuclei, reticular formation) synapse on granule cells in the granular layer; granule axons ascend into the molecular layer and bifurcate into parallel fibres running transversely across the dendritic trees of many Purkinje cells in series; climbing fibres from the inferior olive make hundreds of contacts on the proximal Purkinje dendrite and produce an all-or-nothing complex spike. Purkinje cells are the sole output of the cerebellar cortex and they are GABAergic, so that entire output is inhibitory, projecting onto the deep cerebellar nuclei (dentate, interposed, fastigial), which are excitatory and carry the signal to thalamus, red nucleus and brainstem.
Q: What is architecturally unusual about parallel-fibre input to a Purkinje cell, and why can you not quote one synapse count? A: A single parallel fibre contacts tens of thousands of Purkinje cells, and each Purkinje cell integrates a very large number of parallel fibres -- massive divergence followed by massive convergence. The count is reconstruction-method-dependent: the same group's 2024 primary paper reports 38,417 +/- 29,210 spines per human Purkinje cell (max observed 97,853) and its 2025 review quotes about 360,000, with total dendritic lengths of 20,167 +/- 15,249 um and 30,000-63,645 um respectively. Write "tens of thousands to a few hundred thousand parallel-fibre synapses". The qualitative claim is not in dispute.
Q: Every textbook draws one climbing fibre per Purkinje cell. Where does that come from, and where does it fail? A: It is a rodent fact. About 80% of human Purkinje cells emit 2-3 primary dendritic trunks against 10.5% in mouse, and multi-trunk human cells appear to receive more than one climbing fibre. Human dendritic length is 7.24x mouse and the dendritic complexity index 27x, at about 2 spines/um in both species. Do not teach the mouse wiring diagram as the human one.
Q: Why is Purkinje pacemaking the hinge of the whole selective-vulnerability argument? A: Because Purkinje cells fire with all synaptic input blocked, driven by a resurgent Na current and matching K conductances. Spontaneous simple spikes average about 40 Hz in vivo with a 0-200 Hz range, the intrinsic regular-firing range is 30-150 Hz, and climbing-fibre complex spikes arrive at about 0.5-1 Hz. A cell firing tens of times a second forever, supplying a vast arbour, has less metabolic headroom than a quiet interneuron.
Q: Name the four candidate explanations for selective neuronal vulnerability and say what evidential state each is in. A: Firing energetics -- plausible, widely argued, not shown sufficient. Calcium handling -- the strongest datum is that calbindin-D28k null mice are ataxic with altered dendritic Ca signalling, and reduced calbindin and parvalbumin immunoreactivity precedes onset in SCA1 transgenic mice. Protein burden -- large, long, highly branched cells synthesise, fold, traffic and degrade more protein per cell. Morphology and transport length. None is established as the mechanism for any ataxia; present them as a toolbox and say which observation would discriminate. On the track's three-tier ladder, calcium handling is Supported -- demonstrated, but in one design class, the mouse mutants -- while firing energetics, protein burden and transport length are Conjectured, argued from adjacent facts rather than measured.
Q: Nearly every neurodegeneration gene is broadly expressed, yet each disease kills a reproducible subset of neurons. State the puzzle for . A: HTT is expressed everywhere and Huntington disease destroys striatal medium spiny neurons; PPP2R2B is expressed across the brain and presents as a cerebellar-plus syndrome. Expression pattern does not pick the dying cell. For SCA12 this is sharper than usual, because the gene is not even cerebellum-enriched -- so the cerebellar is an instance of the selective-vulnerability problem, not a solution to it.
Q: "Neurons never divide, so a neuron cannot dilute a toxic protein." Where is that too strong? A: The consequence is right: a postmitotic cell cannot dilute damage by dividing and cannot be replaced, so an eighty-year-old Purkinje cell is eighty years old and every insult it survived it survived cumulatively. But "postmitotic" is an actively maintained state, not a locked door -- neurons in Alzheimer's disease and other degenerations show cell-cycle re-entry markers, and re-entry is followed by death rather than division. Whether any neurons are born in the adult human hippocampus is genuinely unsettled: two 2018 papers six weeks apart reached opposite conclusions on overlapping markers. Nobody claims neurogenesis in the Purkinje layer, so SCA12 is unaffected -- say so rather than leaning on the slogan.
Q: Ubiquitin-positive inclusions are found in a degenerating neuron. What follows? A: That bulk degradation failed, not what caused it. Deleting Atg5 or Atg7 from the mouse CNS produces progressive motor deficits, neuronal loss and ubiquitin-positive cytoplasmic inclusions, including in cerebellar cortex, with no disease-associated aggregating protein present anywhere in the experiment. Inclusions are a readout of clearance failure, not a diagnosis of cause -- which is exactly why the tau-negative, polyQ-negative, TDP-43-negative inclusions in the SCA12 autopsy series are informative as negatives.
Q: What does the cerebellum compute, and how does the clinical picture fall out of it? A: The framework -- a well-supported theory, not a measurement -- is that the cerebellum learns internal models: forward models predicting the sensory consequences of a motor command, inverse models supplying the command for a desired outcome, with climbing-fibre input as the training error signal. If it predicts and calibrates rather than commands, damage should decalibrate rather than paralyse. Ataxia, dysmetria, , dysarthria, nystagmus and dysdiadochokinesia then stop being a list and become one prediction.
Q: Give the MDS activation-condition vocabulary for tremor, and say which term is the cerebellar one. A: Rest tremor occurs in a body part not voluntarily activated and fully supported against gravity, and damps on movement -- the classic Parkinsonian one. is any tremor produced by voluntary contraction and subdivides into postural (holding a position against gravity), kinetic (during any voluntary movement) and intention tremor, a kinetic tremor whose amplitude increases as the limb approaches the target. Intention tremor is the cerebellar one. The MDS consensus classifies on two axes: Axis 1 clinical features, Axis 2 aetiology.
Q: A patient's tremor is worst at rest and damps on movement. What does that say about the cerebellum? A: That the cerebellum is probably not the culprit -- the activation condition is the diagnostic axis, and a rest-predominant tremor points away from cerebellar disease. This vocabulary is what makes the SCA12 sentence precise: its presenting feature is usually an action tremor of the upper limbs that can be, and often is, mistaken for essential tremor.
Q: What is the loop-oscillation account of an action tremor, and what does it fail to explain in SCA12? A: A limb under feedback control is a loop with delay, and a delayed loop with too much gain oscillates. On the internal-model account the cerebellum supplies the prediction that lets a correction be issued before the error has been felt, so losing it leaves raw delayed feedback: each correction overshoots and is itself corrected, which is why intention tremor grows as the limb closes on the target. Grade that Supported -- it is a well-supported theory of cerebellar function, not a measurement. What it does not explain is SCA12's actual presenting sign, which is postural in 87.7% of a 49-patient cohort against intention tremor in 57.1%; no published work localises the oscillating circuit in SCA12, so the circuit basis of its postural tremor is Conjectured, and saying so is better than borrowing the cerebellar story wholesale.
Q: An ataxic patient has hyperreflexia. What does it add? A: That the pathology extends beyond the cerebellar cortex. Hyperreflexia -- exaggerated stretch reflexes, often with a widened reflex zone and sometimes clonus -- is an upper motor neuron sign, indicating loss of descending corticospinal inhibition of the spinal reflex arc rather than a muscle or peripheral-nerve problem. Pure cerebellar degeneration produces hypotonia and pendular reflexes if anything, so hyperreflexia is a "plus" feature, and it is a recognised part of the SCA12 phenotype.
Q: SARA has interrater ICC 0.98 and no established minimal clinically important difference. Why are both facts load-bearing? A: Because reliability is not interpretability. SARA is 8 items scoring 0 (no ataxia) to 40 (most severe), validated in cohorts of 167 and 119 SCA patients, with test-retest ICC 0.90, Cronbach's alpha 0.94, administration in 14.2 +/- 7.5 min, and correlations of -0.80 with the Barthel Index and -0.89 with UHDRS part IV. But no MCID has been established, so a trial reporting a 1.2-point SARA improvement is reporting a number whose clinical meaning has never been independently anchored.
Q: Why does SARA move more for one deteriorating patient than another? A: Because the items are not equally weighted. Gait contributes 8 of the 40 points and stance 6, so 35% of the maximum comes from two axial items; a patient whose gait worsens while limb function holds moves the score far more than the reverse. Its correlation with disease duration is only r = 0.34, which is another way of saying the scale is not a clock.
Q: The SCA12 trial reported TETRAS and SARA together. What does each measure, and why in that order? A: They measure different things. TETRAS is the tremor instrument: its performance score rates tremor on examination -- the trial's entry criterion was an upper-limb performance component of at least 2 -- and its ADL component rates tremor-related daily function; change in TETRAS PS and TETRAS ADL+PS at weeks 4 and 8 were the primary outcomes. SARA is the ataxia instrument: 8 items, 0 to 40, with gait and stance supplying 35% of the maximum, and it appeared as a secondary outcome alongside SF-36. The order encodes the shape of the disease -- tremor first, ataxia second -- and the interpretability caveat applies on both scales, since SARA has no established minimal clinically important difference and no anchored point-change is quoted here for TETRAS either.
Kinases, phosphatases and PP2A
Q: Why is the level of phosphorylation at a site not a property of the kinase? A: Because a kinase transfers the gamma-phosphate of ATP onto a hydroxyl-bearing side chain and a hydrolyses it off, so the steady-state fraction phosphorylated is set by the ratio of kinase to phosphatase flux at that site. Nothing in the cell is simply "phosphorylated" -- and knocking out a phosphatase raises that fraction exactly as surely as activating a kinase does. That framing is what makes a phosphatase-subunit dosage change a signalling rather than a housekeeping detail.
Q: 518 kinases against about 30 Ser/Thr phosphatase catalytic subunits. What does that arithmetic force? A: The human genome encodes 518 protein kinases (about 1.7% of genes), of which 428 are Ser/Thr and 90 Tyr, opposed by 107 protein tyrosine phosphatases but only about 30 Ser/Thr phosphatase catalytic subunits. Specificity on the Ser/Thr phosphatase side therefore cannot live in the catalytic subunit -- there are not enough of them. It has to be built combinatorially, by targeting subunits, which is exactly what 's B subunits are.
Q: What are the three subunits of a PP2A holoenzyme, and which slot is the variable one? A: A heterotrimer. The scaffolding A subunit (PPP2R1A, PPP2R1B) is 15 HEAT repeats in an L-shaped superhelix -- the elastic spine everything docks onto. The catalytic C subunit (PPP2CA, PPP2CB) is a metal-dependent PPP-family phosphoesterase whose C-terminal tail ends in Leu309. The third slot takes one variable regulatory B subunit drawn from four structurally unrelated families.
Q: Name the four B-subunit families and their folds. A: B/B55/PR55 (PPP2R2A-D: Balpha, Bbeta, Bgamma, Bdelta), a seven-bladed WD40 beta-propeller with an acidic top groove. B'/B56/PR61 (PPP2R5A-E), HEAT-like helical repeats that recognise an LxxIxE short linear motif. B''/PR72/PR130 (PPP2R3A-C), EF-hand and Ca-responsive. B'''/striatin (STRN, STRN3, STRN4), which scaffolds the STRIPAK complex. Two unrelated folds docking onto the same scaffold is the structural fact behind combinatorial specificity.
Q: How many distinct PP2A holoenzymes exist, and what kind of number is that? A: Reviews say "over 60" and "nearly 100", so quote about 60-100 -- and say what kind of number it is: a combinatorial upper bound computed from 15 B-subunit genes with at least 26 transcript and splice variants against 2 A genes and 2 C genes, not a census of complexes observed in a cell. Not every combination assembles, and observed abundances differ by orders of magnitude.
Q: Why is treating "PP2A" as one enzyme fatal to understanding ? A: Because the A-C core has essentially no substrate selectivity on its own; the B subunit supplies substrate recognition, subcellular targeting and regulation. Two holoenzymes sharing a catalytic subunit but carrying different B subunits are functionally different enzymes. PP2A accounts for at least 50% of cellular Ser/Thr dephosphorylation in most cell types and up to about 1% of total cellular protein in some tissues, so "PP2A activity" without a named B subunit is not a claim about anything in particular.
Q: What are 's identifiers, and what is its product called? A: PPP2R2B at 5q32, about 500 kb on the minus strand of GRCh38, with 10 RefSeq transcript variants; NCBI Gene gives the span as chr5:146,580,742-147,081,520 and the GTEx reference API one about 3 kb different, because gene boundaries are -dependent. Entrez 5521, Ensembl ENSG00000156475. The product is B55beta, also written Bbeta or PR55beta, a brain-enriched PP2A regulatory B subunit. Note the convention throughout: PPP2R2B the gene, B55beta the protein.
Q: Which PP2A subunits are brain-restricted, and why does that matter for a phosphatase disease? A: Bbeta and Bgamma protein are detectable only in brain, unlike the A subunit, Balpha and C, which are broadly expressed. Bbeta2 mRNA is dramatically induced postnatally and on neuronal differentiation, while total Bbeta protein decreases after birth as Bgamma rises. A ubiquitous enzyme with a brain-restricted specificity subunit is how a housekeeping phosphatase acquires a neurological .
Q: Bbeta1 versus Bbeta2 -- what differs, and what does not? A: They are alternatively spliced isoforms differing only in the N-terminal tail. That divergent Bbeta2 N-terminus has no effect on phosphatase activity -- it is a pure targeting signal, and the tail alone is sufficient to send GFP to mitochondria. Bbeta1 is cytosolic; Bbeta2 sits in the outer mitochondrial membrane. Bbeta2-containing holoenzymes are about tenfold less abundant than Bbeta1-containing ones, and Bbeta2 is predominantly forebrain.
Q: How does Bbeta2 get stuck in the outer mitochondrial membrane, and how is that gated? A: By a cryptic mitochondrial import signal in roughly residues 1-24 plus a structural arrest domain: the protein engages the import translocase but resists unfolding, so it stalls in the outer membrane. The targeting is itself phospho-regulated -- phosphorylation of Ser20/Ser21/Ser22 neutralises the import signal's +3 charge and holds Bbeta2 in the cytosol, and autodephosphorylation permits mitochondrial accumulation. Regulation all the way down.
Q: What is the DRP1 Ser637 switch, and which enzymes work it in each direction? A: DRP1 is a dynamin-family GTPase recruited to the outer mitochondrial membrane to divide it, opposed by MFN1/MFN2 and OPA1 fusion. Its inhibitory phosphosite is Ser637 in human numbering, Ser656 in rat. PKA, anchored by AKAP1 at that same membrane, phosphorylates it and inhibits fission; calcineurin (PP2B) and PP2A with Bbeta2 remove the phosphate and promote fission. Under PP2A/Bbeta2 action, neuronal mitochondria fragment and depolarise, are depleted from dendrites, and dendritic outgrowth is stunted while synapse number rises. One residue, two opposed enzymes anchored on one membrane.
Q: What is B55beta actually known to do inside a neuron, and what is the negative that belongs beside it? A: Established (more than one system): Bbeta and Bgamma protein are detectable only in brain -- Strack et al. 1998, corroborated across systems. Supported, in one design class (overexpression, one laboratory) -- which is where D2's §5.4 table puts the rest: the Bbeta2 isoform's divergent N-terminal tail acting as a targeting signal that parks PP2A in the outer mitochondrial membrane, gated by phosphorylation of Ser20/Ser21/Ser22; Bbeta2-containing holoenzymes dephosphorylating DRP1 at Ser637 and driving fission; Bbeta2 overexpression accelerating apoptosis on growth-factor withdrawal in a way that requires holoenzyme assembly, since an assembly-defective mutant is inert; and the PP2A-LC3B link routing PINK1-Parkin mitophagy to SCA12. None of those has a patient-derived system behind it, which is what Established would require. The negative: there is no published catalogue of validated B55beta substrates in a neuron -- and the widely quoted Akt Thr308 result belongs to B55alpha (PPP2R2A), not the beta paralogue -- so "B55beta dephosphorylates X" has to be sourced substrate by substrate rather than inherited from the family.
Q: What is PP2A's share of tau phosphatase activity in human brain? A: About 71%, against roughly 11% PP1, 10% PP5 and 7% PP2B. The Km for tau dephosphorylation by PP1, PP2A and PP5 is 8-12 uM, comparable to intraneuronal tau concentration in human brain, while PP2B's is fivefold higher. In Alzheimer's disease brain, total tau-phosphatase and PP2A/PP5 activity fall while PP2B rises, and PP2A activity correlates negatively with tau phosphorylation at most sites.
Q: Why does a Km near the physiological substrate concentration matter for a dosage disease? A: Because an enzyme in that regime is neither saturated nor idle, so its flux tracks its abundance. Change how much active PP2A is present and you change the steady-state phosphorylation of the substrate directly. That is the enzymological reason a dosage change in a PP2A subunit is not automatically buffered away.
Q: "More PP2A means less phospho-tau." What is wrong with that prediction? A: It ignores an opposite-signed indirect route. Akt phosphorylates GSK-3beta at Ser9 to inhibit it; PP2A dephosphorylates Akt Thr308 to inhibit Akt; less Akt means less Ser9 phosphorylation, so more active GSK-3beta and more tau phosphorylation. PP2A therefore affects tau phosphorylation with opposite signs along two routes, so any single-signed prediction is naive. Note too that the Akt result is for B55alpha (PPP2R2A), not B55beta -- do not transfer a result across paralogues.
Q: How is PP2A actually regulated? A: Overwhelmingly by which holoenzyme gets assembled, not by dialling a catalytic rate up and down. LCMT1 methylates the C subunit's C-terminal Leu309 carboxylate, and neutralising that charge permits both B55 and B56 binding; PME-1 removes the methyl ester, disfavouring B binding and protecting the free catalytic subunit from proteasomal degradation; alpha4 (IGBP1) and PTPA are chaperone-like biogenesis factors that sequester and activate nascent C, preventing promiscuous free-C activity. Assembly is the control point, which is precisely why the amount of one B subunit is a credible pathogenic variable.
Q: State the argument that raising one B subunit is dangerous, and name its weak step. A: Step 1: specificity, targeting and regulation reside in the B subunit -- Established, structurally. Step 2: 15 B-subunit genes and at least 26 variants compete for cores built from 2 A and 2 C genes -- Established. Step 3: assembly is a methylation- and chaperone-gated binding equilibrium, not dedicated one-to-one pairing -- Established. Step 4: therefore raising one B subunit forms more of that holoenzyme and, by mass action on a limited core pool, fewer of the others, producing gain of one activity and loss of several unrelated ones simultaneously. Step 4 is the honest weak link and grades Conjectured: it is asserted routinely in reviews and follows from steps 1-3, but no primary study quantifies the free-core pool in a neuron or measures displacement of endogenous B subunits at disease-relevant dosage.
Q: What general lesson does a specificity subunit teach about gene dosage? A: That dosage of a subunit competing for a shared catalytic core is not a smooth dial on one activity -- it is a redistribution of a scarce resource. and triplosensitivity intuitions built on enzymes that act alone do not transfer, and neither do simple "more enzyme, more product" predictions.
Repeat-expansion disorders as a class
Q: Why is a repeat threshold not like the mass of the electron? A: Because it is a clinical-laboratory convention, revised whenever someone reports a shorter in a symptomatic person. The , SCA3 and C9orf72 thresholds all moved within the last decade and are still moving. Any chapter or report that states a threshold must state whose threshold it is and when it was set.
Q: Give the HTT allele classes and the unit warning that goes with them. A: CAG in coding exon 1: 26 or fewer normal, 27-35 intermediate, 36-39 reduced , 40 or more fully penetrant. Prevalence in European-ancestry populations is 9.71:100,000, up to 17:100,000 with multi-source ascertainment, against 0.1-2:100,000 in East Asian and African populations. The units are repeat units, not base pairs -- a 40-repeat allele is 120 bp of tract, and a lab that reports base pairs is not reporting the same number.
Q: What is the least intuitive number in repeat genetics? A: That as many as 1 in 400 people of European ancestry carry an HTT allele of 36 or more CAG, the majority of them reduced-penetrance 36-39 alleles, against a disease prevalence near 1 in 10,000 -- a roughly 25-fold gap, almost entirely alleles that never declare themselves. Penetrance of 36-38 alleles across a typical lifespan is only 0.2%-2%. SBMA repeats the shape: clinical prevalence about 1:300,000 males, but 1:6,887 males carry a pathogenic expansion.
Q: Give the DM1 allele classes, location and mechanism. A: DMPK CTG in the 3' UTR: 5-34 normal, 35-49 (asymptomatic but unstable), over 50 pathogenic, congenital disease usually over 1,000 repeats and sometimes 730-1,000. The mechanism is untranslated CUG RNA forming nuclear foci that sequester muscleblind proteins -- MBNL, MBLL and MBXL co-localise with foci of expanded-repeat transcripts in DM1 and DM2 cells -- and the catastrophe follows.
Q: One repeat, two diseases: give the FMR1 numbers and say what flips. A: CGG in the 5' UTR: about 5-44 normal, 45-54 grey zone, 55-200 premutation, over 200 full , with hypermethylation causing transcriptional silencing. Fragile X syndrome is the full mutation acting by . FXTAS is caused by the premutation, through toxic RNA and RAN-translated FMRpolyG -- and the full mutation cannot cause FXTAS, because it is methylated and silent. Identical motif, identical position, different size, and past about 200 repeats the disease changes rather than merely worsening.
Q: Give the Friedreich row, and say why it is the odd one out. A: FXN GAA in 1: 5-33 normal (most under 12), 34-65 mutable normal with 44-65 borderline, and 66 to about 1,300 pathogenic with most patients at 600-1,200. It is , which is unusual in a table of , and the mechanism is loss of frataxin: the GAA tract forms intramolecular triplex (H-DNA), "sticky DNA" and R-loops that block transcriptional elongation, and repressive chromatin flanks the expansion.
Q: Where is the widest source disagreement in the whole taxonomy? A: C9orf72. GGGGCC in intron 1 or the region: GeneReviews gives 2-24 normal, 25-60 of uncertain significance, and 61 to over 4,000 pathogenic with age-dependent reduced penetrance. STRipy calls 24 or more pathogenic, and diagnostic labs commonly call 30 or more an expansion. The gap is 37 repeats wide and the biology gives no clean step. Do not paper over it.
Q: Give the SCA12 row of the taxonomy table. A: CAG at 5q32, in the 5' region and probable promoter, about 133 nt upstream of the reported start site -- not in a glutamine reading frame in the canonical transcript, so no polyglutamine tract, though the v4 API annotates the same interval "coding: polyserine" against a different transcript while STRchive calls it 5' UTR. Controls 4-32; 43-50 called intermediate and argued pathogenic; 51 or more pathogenic by the classical convention, with 43 argued as the true floor. Autosomal dominant. Grade the whole row contested: there is no current GeneReviews chapter and the field has not converged.
Q: Motif does not determine mechanism. Make the case using CAG alone. A: HTT's CAG sits in the reading frame and is translated into polyQ, giving protein and aggregation. PPP2R2B's CAG sits about 133 nucleotides upstream of the transcription start site and acts as a cis element that up-regulates expression, with no polyQ. ATXN8OS/ATXN8's CTG.CAG sits in one gene's 3' UTR and the overlapping gene's sense ORF, yielding CUG RNA toxicity, polyQ, and RAN polyAla and polySer -- one repeat, three products. Same motif, three different molecular crimes: position decides.
Q: How does secondary structure differ by motif, and what does that predict? A: CAG.CTG forms imperfect hairpins with one mismatch per repeat unit -- a replication and repair problem, a substrate for MutSbeta, expansion-biased. CGG.CCG forms hairpins and G-quadruplexes and is heavily methylatable, so it stalls replication and then builds . GAA.TTC forms triplex, sticky DNA and R-loops, which is a transcription problem and gives loss of function. GGGGCC forms DNA and RNA G-quadruplexes and R-loops, generating repeat-length-dependent abortive transcripts, nucleolin sequestration and nucleolar stress. The structure predicts which process breaks.
Q: SCA6 breaks almost every generalisation in the class. How? A: CACNA1A CAG at the coding 3' end (the alpha1ACT ORF): normal 18 or fewer, 19 uncertain, pathogenic 20-33 with 22 commonest -- a pathogenic range that overlaps the normal range of most other polyQ genes. has not been observed, and GeneReviews states plainly that expansions are not commonly observed in transmission from parent to child. Whatever makes CAG unstable, it is not the CAG alone.
Q: Every threshold is really a threshold on what? A: Uninterrupted repeat length. ATXN1: a 36-44 CAG allele with CAT interruptions is normal, without them mutable or pathogenic, and one CAT-interrupted CAG-44 carrier was asymptomatic at 66 despite an affected father. FMR1: AGG interruptions every 9-10 CGG reduce the risk of expansion to full mutation for alleles under 100 repeats. CNBP: 11-26 CCTG with GCTC/TCTG interruptions is normal. ATXN10: ATCCT-interrupted expansions associate with more epileptic seizures and show anticipation with paradoxical contraction. RFC1 makes the point by substitution instead -- (AAAAG) and (AAAGG) are benign, (AAGGG) at the same locus and length class is pathogenic.
Q: What is the cleanest natural experiment separating repeat length from protein length? A: The HTT CAA result from GeM-HD 2019, in 9,064 European-ancestry subjects. Individuals whose allele had lost the CAACAG had earlier onset than their total glutamine count predicted; individuals with a CAACAG duplication -- that is, more glutamines -- had later onset. Polyglutamine length is not the driver; uninterrupted CAG length is. Add glutamines while shortening the pure CAG tract and onset gets later.
Q: Distinguish , premutation and reduced-penetrance allele. A: An intermediate or mutable-normal allele does not affect its carrier but is meiotically unstable, so the carrier's children are at risk -- HD 27-35, SCA1 36-38 uninterrupted, DM1 35-49, FMR1 45-54, of which about 14% expand into the premutation range on maternal transmission. A premutation causes a different disease by a mechanism the full mutation cannot use: FMR1 55-200 gives FXTAS and FXPOI. A reduced-penetrance allele is the genuine disease allele in someone who may never manifest: HD 36-39. Conflating them produces wrong counselling.
Q: What did the Huntington onset-modifier find, and what may you not quote from it? A: GeM-HD 2015 found a 15 locus with two independent effects -- hastening onset by 6.1 years and delaying it by 1.4 years -- plus a chromosome 8 locus hastening onset by 1.6 years, with association at MLH1 and pathway support for DNA handling and repair. GeM-HD 2019, in 9,064 subjects, put DNA-repair genes at the significant modifier loci: FAN1 (the top locus), MSH3/DHFR, PMS1, MLH1, PMS2, LIG1 and RRM2B/UBR5, several carrying multiple independent acting in opposite directions at the same locus. What you may not quote is a per-locus effect size in years from the 2019 paper without reading Table 1 of the source yourself.
Q: Give the MutSbeta model of expansion, mechanistically. A: MutSbeta is MSH2-MSH3, the loop-recognising arm of . It binds a CAG or CTG hairpin but hydrolyses ATP more slowly when bound to one, so the repair reaction stalls at recognition instead of proceeding to excision -- the loop is protected rather than removed, then fixed into the genome as an expansion. MutSbeta also recruits beta to the nascent hairpin and stimulates its retention, with Polbeta using the hairpin as a primer. Stoichiometry matters: abundant MSH3 out-competes MSH6 for MSH2, shifting the MutSalpha/MutSbeta balance towards expansion -- which is the axis the MSH3 modifier haplotypes sit on. FAN1 is the counterweight, a nuclease that opposes expansion.
Q: There is no in a postmitotic neuron. So what expands the repeat? A: Repair, not replication, by two candidate routes. Oxidative-damage-initiated base-excision repair: OGG1 excises 8-oxoG inside the repeat, the gap is filled by error-prone strand-displacement synthesis that captures a hairpin, and the longer repeat is a bigger oxidation target -- an escalating cycle. And MutSbeta-driven loop repair going wrong. Both are transcription- and repair-coupled, both run in cells that will never divide again, and both are biased towards addition, because the intermediate that gets fixed into the genome is an extra loop rather than a missing one.
Q: State the two-step model of onset and the evidence behind it. A: Step 1: the inherited allele expands somatically in vulnerable neurons, slowly at first. Step 2: past a critical length a length-dependent toxic mechanism fires and that cell dysfunctions and dies. Onset is then set by the rate of somatic expansion, and the inherited length matters chiefly because it sets how far the cell has to travel. The strongest evidence is single-cell work on postmortem HD brain: striatal projection neurons carry frequent somatic expansion of the disease allele while most other cell types show limited variation, and the pathogenic length for those neurons is about 150 CAG, with transcriptionally normal cells below it and loss of identity above. Treat the three- trajectory numbers as rough.
Q: Summarise the parent-of-origin asymmetries across the class. A: Paternal bias in HD (expansions greater than 7 CAG almost exclusively paternal), DRPLA (offspring onset about 26-29 years earlier than affected fathers against 14-15 years earlier than affected mothers), SCA1, SCA2, SCA7 (infantile onset at 200-400 repeats), SCA10, SCA37, and weakly SCA3. DM1 splits: congenital DM1 is most often maternally inherited, yet paternally transmitted expansions increase more (median 425, range 70-2,000) than maternal ones (median 200, range 57-1,400). SCA8 reverses it, expanding maternally and contracting on paternal transmission. SCA6 and DM2 show no anticipation, and C9orf72 is disputed.
Q: What is , and why does it change what "non-coding repeat" means? A: Non-ATG-initiated of an expanded repeat, discovered in SCA8 and DM1, where CAG constructs expressed polyGln, polyAla and polySer with no ATG present. Its famous case is C9orf72: sense poly-GA (dominant), poly-GP and poly-GR from three frames, and antisense (GGCCCC)n transcripts translated into dipeptide repeats too. In FXTAS, 5' UTR RAN translation produces FMRpolyG, toxic in human cell lines and Drosophila neurons. A repeat outside any reading frame can still make protein.
Q: Why should you stop thinking of a repeat as sitting in one gene? A: Because these loci are transcribed from both strands, so one expansion can make two toxic RNAs and RAN products in up to six frames. At the DM1 (CTG.CAG)n, antisense transcription and heterochromatin are constrained by , which restricts the antisense transcript and confines H3K9 methylation to the repeat ; at C9orf72 the antisense transcript is RAN-translated as well. The same bidirectionality is reported at the SCA12 locus.
Q: Which mechanism causes C9orf72 disease? A: GeneReviews lists three side by side -- , RNA toxicity via sequestration, and dipeptide-repeat aggregation from bidirectional RAN translation -- and does not adjudicate between them. That non-adjudication is the current state of the field, not a gap in the article. Most real repeat diseases run more than one of these at once.
SCA12: locus, alleles and mechanism
Q: How was found, and why does the method matter? A: By detection, a ligation-based assay that detects a long CAG tract in genomic DNA without knowing where it is. RED flagged an expansion in a proband from a single large American reported as being of German descent; the group is described as having cloned a 2.5-kb genomic fragment containing 93 uninterrupted CAG units and mapped it to , publishing a two-page letter in Nature Genetics in 1999. The repeat was found first and the gene second, the reverse of SCA1/2/3 where narrowed a locus and the repeat turned up inside a candidate gene. SCA12 is the clean case of an orphan repeat -- twenty-seven years on the association is rock solid and the mechanism is still argued about.
Q: Give the SCA12 repeat coordinates in every assembly you should ever quote. A: GRCh38 chr5:146,878,728-146,878,759; GRCh37/hg19 chr5:146,258,291-146,258,322; chr5:147,414,734-147,414,765. Cytoband 5q32, gene on the minus strand, reference copy number 10.7, region annotated 5' UTR. All three are quoted 1-based inclusive, converted from STRchive's 0-based half-open intervals. Every coordinate names its assembly and its convention, because a coordinate without a build is meaningless -- and these three differ by hundreds of kilobases for the same 32 bp. Note also that STRchive's interval runs 2 bp longer than the ExpansionHunter catalog's definition of the same tract (32 bp against 30 bp): locus boundaries are curator decisions, not measurements.
Q: (CAG)n, (CTG)n, (AGC)n and (GCT)n all appear in the SCA12 literature. How many loci is that? A: One, a single 32-bp reference element. PPP2R2B is on the minus strand, so the tract reads CAG on the sense/mRNA strand and CTG on the plus/reference strand; and a trinucleotide repeat has no canonical , so databases differ by a cyclic rotation -- STRchive normalises to GCT in reference orientation and AGC in gene orientation. Never infer a different locus from a different motif spelling.
Q: Is the SCA12 repeat in the or the 5' UTR? A: Both, depending on which transcript you mean, and that is the honest answer. The 1999 report placed it 133 nucleotides upstream of the then-reported start site, i.e. in the promoter; later work moved the start site further 5', placing the repeat inside the 5' UTR, and STRchive's current calls it 5' UTR. 's HGVS for the locus, NM_181675.3(PPP2R2B):c.27CAG[(7_28)], is written in coding numbering even though ClinVar's own summary text calls the region 5-prime; the v4 API annotates the same 32 bp "coding: polyserine" against a different first exon -- three authoritative databases, two incompatible annotations, and a third placement, the 1999 promoter one, that neither of them uses. The reconciliation is that PPP2R2B has multiple alternative first exons and therefore multiple start sites: relative to one, the repeat is promoter-proximal; relative to another, it is transcribed 5' UTR.
Q: Which PPP2R2B isoform actually carries the repeat, and why is that the crux? A: The repeat lies in the 5' region of the Bbeta1 transcript, whose first exon is the repeat-containing exon -- exon 7 in the original PPP2R2B numbering, with splice variants 7B7D and 7C7D. It is not in the 5' region of the Bbeta2 transcript, which uses a different first exon. Any Bbeta2-based mechanism therefore has to route through trans effects on total PPP2R2B output rather than through the repeat sitting in the Bbeta2 message. That is the weak joint in the entire Bbeta2 story.
Q: What is PPP2R2B-AS1, and what is unfinished about it? A: An antisense transcript across the repeat on the opposite strand, so it carries a CUG repeat. It was mapped as covering a region 44 bp upstream of the CTG repeat and 145 bp downstream of it, is polyadenylated and carries SP1 sites, forms CUG RNA foci in SK-N-MC cells and induces apoptosis, and has been proposed as a therapeutic target. What is unfinished: its 5' end remains uncharacterised despite repeated sequencing attempts, which matters if you want to target it.
Q: Give the SCA12 ranges by source, and say what is actually agreed. A: Normal: 4-32 (retired GeneReviews chapter), 6-32 (STRchive), 4-31 (Srivastava 2017), with ClinVar's benign call anchored at 7-28. Intermediate: 40-49 (STRchive), or 43-50 reclassified as intermediate pathogenic (Srivastava 2017). Pathogenic: 51-78 (STRchive), 51 or more diagnostic (retired GeneReviews), or 43 or more proposed (Srivastava 2017). Agreed: 51 or more with a compatible is diagnostic, and 31 or fewer is normal. Everything from 32 to 50 is under active dispute.
Q: A patient's SCA12 allele is 44 repeats. How should the report read? A: As an allele in a range where pathogenicity has been reported but not established -- never as "negative". The floor has moved downward twice since 2015: 46 reported as probably the shortest pathogenic allele in a Chinese patient presenting with , then 43 proposed on 18 patients from 16 unrelated families carrying 43-50 repeats, then symptomatic patients at 40 and 42 with a consistent clinical phenotype. Against that, CAG-39 carriers within an affected family were unaffected, which is the argument for putting the floor at 43 rather than lower.
Q: What is known about in SCA12? A: Complete penetrance was documented in the index family at 66-78 repeats. Individuals carrying 45-62 repeats have been observed with very-late-onset disease or with no disease at all, so penetrance in the lower band is reduced and apparently age-dependent, and STRchive tags the locus length_affects_penetrance. The earliest direct evidence of non-penetrant carriers is one Indian family with the expansion in 6 affected and 3 unaffected at-risk individuals.
Q: What do the biallelic SCA12 carriers show? A: Two are published: one homozygous CAG-45/45 and one compound heterozygous CAG-42/51. Neither differed from heterozygous CAG-51 carriers in age at onset or severity, so no dosage effect was detected -- in a sample of two. That is a statement about power, not a demonstration that dosage does not matter.
Q: What do Established, Supported and Conjectured mean in this track, and why three tiers rather than two? A: Established means demonstrated in more than one system, including a patient-derived one; Supported means demonstrated, but in one system or one design class; Conjectured means argued from adjacent facts, not measured. The middle rung is the one doing the work, because most of the SCA12 literature sits there -- real experiments, one design class, usually overexpression -- and a two-tier proven/unproven split either promotes the fly and cell-line results to fact or throws them away. The lettered hypotheses A, A', A'', B, C1, C2, C3 and D exist for the same reason: they let a claim be checked against the row it belongs to instead of against "the SCA12 literature" in aggregate.
Q: What are the live mechanistic hypotheses for SCA12, and how confident is the field? A: Five main ones, lettered so they can be argued about one at a time: A, promoter and 5'-UTR-driven PPP2R2B overexpression; B, Bbeta2-driven mitochondrial fission; C1, sense CAG RNA toxicity; C2, antisense CUG RNA toxicity from PPP2R2B-AS1; and C3, in several frames -- with three further rows, A' for a toxic polyserine product of the expanded tract, A'' for a translational-efficiency effect on the repeat-containing message, and D for tau hyperphosphorylation via altered . None is Established. STRchive scores the gene-disease evidence 8.5 out of 18 ("Moderate") and calls the mechanism incompletely established -- and its own one-line summary, polyalanine via RAN translation, rests on a single citation and is considerably more confident than the underlying literature.
Q: What supports the overexpression hypothesis? A: Hypothesis A is the oldest and best-supported in vitro, and it grades Supported -- demonstrated, but within one design class. The repeat acts as a cis element that up-regulates PPP2R2B expression, with CREB1 and SP1 binding upstream and up-regulating and TFAP4 binding downstream and down-regulating. The repeat-containing region has promoter activity that increases with repeat length and depends on cell type and repeat sequence, and PPP2R2B overexpression disrupts neuronal morphology. A 2024 study reports the expansion raising the 7B7D transcript and Bbeta1 protein, and also producing a protein bearing a long polyserine tract that triggers apoptosis.
Q: What is the single most damaging result for the overexpression model? A: That in patient-derived mature neurons most PPP2R2B isoforms were down-regulated relative to controls. That directly contradicts "expansion means more Bbeta means toxicity", and both results are primary and recent. The direction of the expression change may be cell-type- and maturation-stage-dependent; as things stand, whether Bbeta1 is up or down in SCA12 neurons is unresolved.
Q: Hypothesis A'' says the repeat changes rather than transcription. State it, and name the experiment that separates it from A. A: A is a transcript-level claim -- the repeat is a cis element raising PPP2R2B output -- and A' is a product-level one, a toxic polyserine species from the expanded tract. A'' is the third possibility: the expanded tract lies in the 5' region of the Bbeta1 message, so it can change how well that message is translated, and protein output can move while transcript abundance does not. It is the one class of hypothesis that can move protein without moving transcript -- but D4 records it as at best a partial reconciliation, since the patient neurons in question show transcript that does move, and a maturation-stage effect remains the competing account. The discriminating experiment is polysome or profiling in patient and isogenic neurons across a repeat-length series: ask whether ribosome loading per molecule of the repeat-containing transcript shifts, which is a question no measurement of transcript abundance can answer. Nobody has run it, so A'' is Conjectured.
Q: Give the evidence for the Bbeta2 mitochondrial hypothesis, and the objection to it. A: The evidence for Hypothesis B: Bbeta2's divergent N-terminus targets PP2A to mitochondria and accelerates apoptosis on growth-factor withdrawal; stressors drive Bbeta2 to mitochondria where it dephosphorylates Drp1 at S637 to activate fission, and fragmentation is required for the death; silencing Bbeta2 protects hippocampal neurons against free-radical, excitotoxic and ischaemic insults; Drosophila overexpressing ppp2r2b show neurodegeneration, apoptosis, shortened lifespan, mitochondrial fragmentation and raised ROS, rescued by antioxidant treatment and SOD2 expression; and PP2A-mediated dephosphorylation of LC3B links PINK1-Parkin mitophagy to SCA12, with deferiprone rescuing neuronal survival. The objection: every one of those experiments overexpresses Bbeta2 artificially, and the repeat is not in the Bbeta2 transcript's 5' region. Mechanistically rich, causally under-anchored -- Supported in cell, neuron and fly systems, and nowhere near Established in a patient-derived one.
Q: How would you tell the three main hypotheses apart experimentally? A: Ask what each predicts and go and look in patient tissue rather than in an overexpression construct. Overexpression (A) predicts a repeat-length-dependent rise in specific PPP2R2B transcripts and in Bbeta protein in the affected cells -- so measure isoform-resolved expression in patient neurons by maturation stage, which is exactly where the field currently contradicts itself. The Bbeta2 mitochondrial hypothesis (B) predicts raised Bbeta2 specifically, hypophosphorylated DRP1 Ser637, and fragmented, dendritically depleted mitochondria in patient neurons -- and it needs a route from a repeat in the Bbeta1 first exon to a Bbeta2 protein increase. RNA toxicity (C1 and C2) and RAN translation (C3) predict foci and non-canonical peptides in human brain tissue, in a stated frame, above a stated repeat length. Each prediction is checkable and none has been checked in SCA12 brain.
Q: What has been shown about RNA foci in SCA12? A: In patient iPSC-derived neural stem cells, nuclear CAG RNA foci were detected by RNA-FISH and were absent from controls, in lines genotyped 14/59, 10/67 and 17/65 against controls at 10/13 and 09/16. Pull-down against a CAG-54 probe identified 13 proteins binding exclusively to the expanded repeat -- KRT27, HSPA1B, H1-3, DYNLL2, C1QBP, S100A16, CTSA, KIAA0100, PSMA2, FN1, SNRPB, GSN and PSMA1 -- enriched for protein-clearance machinery. The paper reports neither the fraction of cells bearing foci nor a repeat-length threshold for foci formation, and that gap is worth stating: it leaves C1 Supported in one patient-derived system and untested in tissue.
Q: Is RAN translation demonstrated for SCA12 specifically, and do the labs agree? A: Demonstrated, and the reported frames only partly agree, which is itself informative. Both groups report a polyserine product -- the Delhi lab from the AGC frame, the Hopkins lab as an expansion-derived polyserine tract that triggers apoptosis. What is unreconciled is the rest: the Delhi lab reports polyglutamine from the sense CAG frame, with polyQ appearing even when the sole ATG was mutated to GGG; the Hopkins lab reports polyalanine from the antisense CUG/alanine ORF, and explicitly no polyleucine and no polycysteine product, using (CTG)10, (CTG)73 and interrupted (CTG)73 constructs. Both are primary. Nobody has demonstrated a RAN product in SCA12 human brain tissue, so C3 is Supported in cells and iPSC neurons and Conjectured for human brain.
Q: SCA12 is often floated as a tau-related disease because PP2A is the main tau . How good is that? A: It is a twenty-five-year-old speculative list item that has never been tested. A 2001 review named cell-cycle regulation, tau phosphorylation and apoptosis as processes an altered PP2A B subunit could perturb; no study has measured tau phosphorylation in SCA12 patient tissue, SCA12 iPSC neurons or an SCA12 animal model. Against it, the one autopsy series found intranuclear inclusions negative for tau as well as for long polyQ, alpha-synuclein and TDP-43. A separate association between normal-range PPP2R2B CAG length and Alzheimer disease in a Japanese population is a different disease and is not SCA12 evidence. That is row D, and it is Conjectured: never measured in any SCA12 system, and the cheapest untried experiment in the table.
Q: How much human neuropathology is the SCA12 literature actually built on? A: A handful of brains. The paper of record reports enlarged ventricles, marked cerebral cortical atrophy, loss, less prominent cerebellar and pontine atrophy, and neuronal intranuclear ubiquitin-positive inclusions consistent with Marinesco bodies that stained for none of expanded polyglutamine, alpha-synuclein, tau or TDP-43. A second study examined a single brain; a third could not reliably detect the expanded transcript in one formalin-fixed brain. Any statement about "the SCA12 brain" is a statement about single cases.
Q: What did the one somatic-instability study of an SCA12 brain find, and why is it awkward? A: Somatic was detected across brain regions, but the -- a target tissue -- showed the least instability, coupled with increased PPP2R2B methylation and lower expression, alongside raised expression of DNA-maintenance genes and reduced expression of cell-cycle modulators. In Huntington disease the striatum shows the most somatic expansion. SCA12 does not obviously follow the HD pattern, and n = 1.
Q: Does SCA12 show ? A: Treat it as unsupported. The current GeneReviews Hereditary Overview, last revised 2026-07-09, states there is insufficient evidence for anticipation. The SCA12 chapter retired in 2018 said a moderate degree had been observed and that expanded repeats are modestly unstable, with length variations of a few triplets among sibship members. Patient-facing material says repeats typically do not expand between generations. The current source supersedes the retired one, and the field should stop asserting either way.
Q: Is there a parent-of-origin effect in SCA12? A: Effectively unknown. There is one dataset: 21 patients, 7 maternal, 7 paternal, 1 from both parents, no bias detected -- and far too small to detect one. Contrast HD, DM1 and most polyQ SCAs, where a paternal expansion bias is well documented. SCA12 has no such finding and no study has been powered to look for one, which is a different statement from "no effect".
SCA12: population, clinic and therapy
Q: What is the presenting sign of , in numbers? A: of the upper limbs in the overwhelming majority. Tremor was present at presentation in 95.9% of a 49-patient cohort and was the first symptom in 90% of a 21-patient cohort. Bilateral upper-limb tremor 85.7%, postural-type 87.7%, 57.1%, head tremor 55.1% (62% in the smaller cohort), voice tremor 42.8%. was present at presentation in 73.5% -- present, but second. The rest of the picture in the same series: impaired tandem gait 89.8%, dysmetria 75.5%, dysarthria 57.1%, walking aid required 18.2%, urinary disturbance 20.4%, cognitive dysfunction 22.4% and psychiatric disturbance 8.1% -- against 61.76% cognitive impairment in a dedicated non-motor study of 34 patients and 67% psychiatric disorders in a series of 21, because a clinical screening series and a neuropsychological battery are not measuring the same construct.
Q: Name three ways SCA12 differs from the other , and the common cause. A: Tremor comes first and ataxia later, the reverse of SCA1/2/3/6 where the cerebellar syndrome leads. Cortex is more atrophic than on imaging and at autopsy, where most SCAs are the other way round. And there is no polyglutamine tract, so no expanded-polyQ protein and no 1C2-positive inclusion. All three follow from the being non-coding -- there is a whole review titled "Why is SCA12 different from other SCAs?".
Q: The reported age at onset in SCA12 differs by two decades between cohorts. What are the numbers and the candidate explanations? A: The German-American index family looks like a fourth-decade disease at a mean of 34-38 years; Indian cohorts look fifth-to-sixth decade -- 51.33 +/- 8.98 (range 41-69) in one, 46.38 +/- 11.7 in another, and 58.3 (range 35-72) for CAG 43-50 against 54.1 for the CAG-51 group in a third. STRchive gives a typical window of 26-50 with a full range of 8-62. No paper resolves it. Candidates, none demonstrated: ascertainment, since the index family was found because it was severe; repeat-length differences between founder populations; and modifier background.
Q: Does repeat length predict age at onset in SCA12? A: Unresolved, with directly contradictory published results. One Indian cohort of 21 found r = -0.760, P = 0.0001; another of 49 found no significant correlation; the largest series reports r = -0.65, P < 10^-4 in 124 unrelated patients. Both smaller cohorts are Indian and largely Agarwal, and both are small. Quote a cohort, never a consensus.
Q: What does MRI show in SCA12, and what does a normal scan mean? A: In 49 MRIs: cerebellar atrophy only 34.7%, cerebral atrophy only 16.3%, both 34.7%, normal 6.1%, basal ganglia mineralisation 4%, periventricular ischaemic change 4%. Another series found variable cerebro-cerebellar degeneration with no clear correlation with , cortical atrophy exceeding cerebellar, white-matter hyperintensities in the cerebellar peduncles in some patients, and two patients with short disease duration and no atrophy at all. A normal MRI does not exclude SCA12 -- imaging is supportive, never a rule-out.
Q: How often does SCA12 turn up outside India? A: Rarely. Orphanet describes it as a very rare subtype of type I autosomal cerebellar ataxia, and screening yields are low: 1 family among 247 European index cases, with Italian, Thai and Turkish cohorts also screened, and a report that it may be underdiagnosed in Chinese populations. In a study of 515 familial essential-tremor probands, 6 carried a pathogenic STR expansion, of which exactly one was . SCA12 does hide inside essential-tremor cohorts, but rarely.
Q: How common is SCA12 in India? A: In a pan-India series of about 5,600 ataxia referrals over ten years, SCA12 was 8.6% (490 patients) against SCA2 at 8.5% (482), SCA1 4.8% (272) and SCA3 2.0% (113) -- so the honest statement is that SCA12 and SCA2 are jointly the commonest SCAs in Indian referral cohorts, since 8.6 against 8.5 is well inside the noise. The familiar line "second most common SCA in India" comes from a smaller AIIMS series where SCA12 was about 16% (20/124) of autosomal dominant ataxia.
Q: What is the population prevalence of SCA12? A: There is no published estimate, anywhere, including India. STRchive records prevalence as null, and neither GeneReviews nor Orphanet gives a rate. Anyone quoting a per-100,000 figure is extrapolating from referral fractions, and a referral fraction describes who reaches a genetics clinic, not who has the disease.
Q: How was the Indian actually proved? A: By typing 20 Indian SCA12 families plus ethnically matched controls with four novel SNPs and a dinucleotide marker spanning about 137 kb downstream of the CAG repeat. One was significantly associated with the affected (P = 0.000), establishing a common founder, and expanded alleles in that population ran 51-69 CAG. Crucially, the same haplotype was absent from the American 's expanded , so the SCA12 expansion has arisen independently at least twice. A founder effect proved with five markers and no genome.
Q: Is SCA12 confined to the Agarwal community? A: It is founder-enriched, not founder-restricted, and both halves of that are solid. In a 49-patient cohort 79.6% were Agarwal, 10.2% explicitly non-Agarwal and 10.2% undisclosed, with about 70% of patients from north and north-east India and 18% from southern India; the authors concluded SCA12 is not confined to a single ethnicity. In a smaller series 20 of 21 patients were Agarwal. The community is endogamous and originates in Haryana.
Q: What obligation does naming a founder community in a paper create? A: A that individual consent cannot represent: stigma and discrimination risk attach to everyone who shares the label, including people never enrolled and never tested. Combine that with an adult-onset, slowly progressive, untreatable disease and predictive testing becomes a decision about identity and reproduction as much as about medicine. Report the finding as enrichment with its measured limits -- 79.6% in one cohort with 10.2% explicitly outside -- not as "the Agarwal disease".
Q: Name two that sit on the SCA12 differential, and give one discriminator each. A: FXTAS, from the FMR1 CGG of 55-200 -- and note the premutation is the disease allele, since the full mutation silences the locus and cannot cause it. Discriminator: it is with strongly age-dependent in male carriers, 17% at 50-59 rising to 75% at 80 or over, and the motif is CGG at a locus a CAG-oriented ataxia panel does not size, so you have to order the FMR1 assay by name. SCA27B, from a deep-intronic GAA expansion in FGF14 -- 250-300 with reduced penetrance, 300 or more fully penetrant -- found in 61% of French-Canadian index patients in the discovery cohort and now among the commonest causes of late-onset ataxia. Discriminator: position and motif, since a deep-intronic GAA is invisible to and to any panel built around coding CAG tracts.
Q: Why does a standard whole-genome pipeline miss SCA12? A: Because a read must span the repeat plus unique flank on both sides to size it, and a 150-bp read holds at most 50 CAG units with no flank at all. Ordinary short-read genome and exome pipelines therefore report the locus as reference or say nothing, which is why repeat expansions are described as invisible by construction and are one of the named places undiagnosed cases hide after a negative exome. The fix is a targeted assay: , a repeat genotyper run over the alignment, or long reads.
Q: What can repeat-primed PCR tell you, and what can it not? A: RP- uses one locus-specific flanking primer plus one primer that anneals within the repeat at multiple offsets, so products of every possible length ending in the repeat produce a stutter ladder of peaks at 3-bp intervals on capillary electrophoresis. It tells you an expansion is present -- the ladder continues beyond the normal range instead of stopping. It cannot give the exact size of a large expansion, because the ladder decays and merges before the true end of a big allele. The result is "expanded, at least N", not a size.
Q: What is the classic trap in flanking PCR with fragment sizing? A: A homozygous-normal result and a normal-plus-unamplifiable-expansion result look identical -- one peak. Flanking PCR sizes normal-range alleles exactly, but large GC-rich amplicons amplify poorly or not at all, so the expanded allele can simply fail to appear. One peak is a or a dropout and the assay cannot tell you which, which is exactly why RP-PCR exists -- and why Southern blotting, slow and low-throughput and needing micrograms of DNA, remained the reference method for alleles far beyond any PCR, such as FMR1 full mutations above 200 CGG and DMPK expansions in the thousands.
Q: A patient with action tremor and late-onset ataxia has a negative repeat panel. Why is that not a negative answer? A: Because a panel reports the loci it contains, at the motifs it was built for, against the thresholds its laboratory has adopted, and each of the three is a separate way to be missed. Loci: SCA27B and the RFC1 expansions behind CANVAS -- now implicated in 14-22% of adult-onset ataxia -- sat undiscovered through three decades of molecular ataxia genetics because nobody was looking at those motifs in those positions. Method: an ordinary short-read pipeline reports a repeat locus as reference or says nothing, because a 150-bp read holds at most 50 CAG units with no flank on either side. Threshold: a laboratory calling 51 or more will report a 44-repeat PPP2R2B allele as normal, when 44 sits squarely in the disputed 32-50 band. The report should say what was tested, by what assay, at what threshold -- and what remains untested.
Q: How well does the handle a repeat-length variant? A: Awkwardly. It was built for SNVs and , and its population-frequency, computational-prediction and functional criteria do not map onto a length threshold that has a contested floor, a reduced-penetrance band and age-dependent penetrance. For SCA12 there is no consensus threshold to anchor a pathogenic call at all. The honest report states the measured repeat number, the assay that produced it, the published ranges from each source with their dates, and what is and is not established -- which is a different document from a five-tier classification.
Q: What is hard about predictive testing in SCA12 specifically? A: The same structure as Huntington disease -- adult onset, slow progression, no disease-modifying therapy -- plus one extra problem. Pedigree Bayes supplies the prior, but the conditional term needs an age-dependent penetrance function, and for SCA12 that function does not exist: penetrance in the 43-50 band is genuinely uncertain and non-penetrant carriers have been reported at 45-62 repeats. A positive result in that range cannot be converted into a risk figure, and that should be said before the test rather than after it.
Q: If you were designing an ASO for SCA12, what would the mechanism decide? A: The target strand, and therefore the whole programme. If the disease is -driven overexpression of PPP2R2B, you lower the sense transcript with an RNA-degrading ASO. If it is antisense CUG RNA toxicity, the target is PPP2R2B-AS1 instead, and knocking down the sense transcript achieves nothing useful. If it is , you must block the initiating event on whichever strand and frame matters -- and the two labs that have looked report different frames. Not knowing the mechanism is a therapeutic obstacle, not an academic one. Delivery is the shared bottleneck in either case.
Q: What is the state of disease-modifying therapy for SCA12? A: There is none, and the negative is documented: as of August 2026 no published study tests an , siRNA or other RNA-targeting agent against PPP2R2B in a neurological context, and no SCA12 disease-modifying clinical trial has been reported. What exists is stated intention -- one group proposes PPP2R2B-AS1 as a therapeutic target, another proposes repeat-stabilising drugs -- both in discussion sections.
Q: SCA12 has one randomised controlled trial. What did it test and what did it show? A: Extended-release propranolol escalated to 240 mg/day against placebo in 60 genetically confirmed SCA12 patients aged 18-65 with a TETRAS performance-score upper-limb component of at least 2, single-centre and double-blind, with primary outcomes at weeks 4 and 8. It significantly reduced TETRAS performance score, improved ADL, and improved SARA and SF-36, with accelerometry confirming tremor reduction in most positions; fatigue and headache were commonest and there were no discontinuations for safety. It is symptomatic tremor control, not disease modification.
Q: What is an SCA12 patient offered today, and what negative belongs in the same sentence? A: Symptomatic tremor control with a real evidence base: extended-release propranolol escalated to 240 mg/day beat placebo in 60 genetically confirmed patients, reducing TETRAS performance score and improving ADL, SARA and SF-36, with fatigue and headache commonest and no discontinuations for safety; deep brain stimulation exists only at case level, in two published single-patient reports. Nothing disease-modifying exists -- as of August 2026 no published study tests an antisense oligonucleotide, siRNA or other RNA-targeting agent against PPP2R2B in a neurological context, and no disease-modifying trial has been reported. The negative worth stating out loud: there is no maintained SCA12-specific management guideline at all, because the GeneReviews SCA12 chapter was retired on 2018-12-13 and only the general Hereditary Ataxia Overview remains current.
Q: What does SCA12 have, and lack, for trial readiness? A: Has: outcome measures validated in a real RCT (TETRAS PS, TETRAS ADL+PS, SARA, SF-36, accelerometry); longitudinal gait data in 3 patients over 5-6 years, in which step-length, stance-time and step-time variability progressed and may precede clinical gait abnormality; a machine-learning reaching-kinematics classifier separating SCA12 from essential tremor at 83.3% accuracy; plasma Abeta40 decreased with the Abeta42/Abeta40 ratio increased; and five mitochondrial quality-control genes down in patient PBMCs. Lacks: a multi-site prospective natural-history cohort with annualised SARA progression rates, a fluid biomarker replicated in a second cohort, a published animal model with a progressive phenotype, and any target-engagement assay.
Q: What is the honest statement about life expectancy in SCA12? A: "Believed near-normal; no survival study published." A patient organisation states that SCA12 progresses very slowly and lifespan is generally not shortened, which is consistent with the late onset and slow course in every cohort -- but there is no SCA12 equivalent of EUROSCA and no survival analysis exists. The difference between "lifespan is normal" and "no published survival data" is the difference between a claim and an absence of one.
Genotyping repeats and reading expression
Q: Do the read-length arithmetic that makes a made-for-teaching locus. A: The 1000 Genomes 30x data is paired-end 150 bp on NovaSeq 6000 -- verified by measurement, since all 888 reads extracted from the region of HG00096 are exactly 150 bp. A 150-bp read holds at most 150/3 = 50 CAG units with zero flanking sequence, and to be informative about length a read must anchor in unique flank on both sides, so reads fully resolve only comfortably below 50 units. The SCA12 pathogenic threshold of 51 or more sits almost exactly at the failure point.
Q: What does an aligner do with a read consisting entirely of repeat motif? A: It has no way to place it. Such a read matches every position within the repeat equally well, and matches other loci carrying the same motif genome-wide, so it maps with low or zero mapping quality, maps to the wrong repeat locus, or is left unaligned and placed by its mate. That is the premise of repeat genotyping: the reads the pipeline throws away are the signal.
Q: Name the three read classes a repeat genotyper counts, and what each buys. A: Spanning reads contain the whole repeat plus both unique flanks and size the allele exactly, but only for alleles at or below read length. Flanking reads start in unique flank and end inside the repeat, giving a lower bound. lie entirely within repeat sequence and are evidence that the allele exceeds read length; the genotyper counts IRRs whose mate anchors near the locus, plus off-target IRRs for loci declared RareRepeat, to estimate sizes beyond the read. ExpansionHunter's JSON reports all three counts as CountsOfSpanningReads, CountsOfFlankingReads and CountsOfInrepeatReads.
Q: Why does ExpansionHunter report REPCI as well as REPCN, and what should a report never omit? A: Because beyond about read length are estimates with wide confidence intervals, not measurements. REPCI carries that interval alongside the point estimate REPCN, and a clinical report that quotes the point estimate alone has dropped the part that says how much to trust it. That is the same discipline as reporting an effect size with its confidence interval anywhere else in the course.
Q: Give the SCA12 variant-catalog entry and its coordinate conventions. A: The default Illumina hg38 catalog contains 31 loci and PPP2R2B is one of them: LocusId "PPP2R2B", LocusStructure "(GCT)*", ReferenceRegion "chr5:146878727-146878757", VariantType "Repeat". Catalog regions are 0-based half-open, so the same interval in 1-based samtools form is chr5:146,878,728-146,878,757 (GRCh38), and its 30 bp is 10 GCT units in the reference. The v4 API's main_reference_region agrees exactly; STRchive's interval includes 2 extra bp. The catalog motif GCT is the plus-strand reading -- on the minus strand, where the gene is transcribed, the same repeat reads CAG.
Q: What genotype does HG00096 carry at the SCA12 locus, and how confident should you be? A: 10/14 GCT units: GT 0/1, REPCN 10/14, REPCI 10-10/14-14, from 12 and 14 spanning reads with 26 flanking reads and zero in-repeat reads. Both alleles are normal-range and both were called from spanning reads alone, so the confidence intervals are degenerate -- which is what an easy genotype looks like, and is exactly the regime that disappears once an allele exceeds read length.
Q: Counting reads in a remote works without a reference; extracting their sequences does not. Why? A: Because CRAM stores reads as differences from the reference, so decoding bases requires the reference while counting records does not. samtools view -c against the CRAM URL over chr5:146878000-146879500 returns 486 reads in about 5 seconds with no reference at all; asking for sequences fails with "Unable to fetch reference chr5:...", because the file's embedded reference path is the sequencing centre's internal filesystem. Two fixes: point REF_PATH at the EBI CRAM registry, which serves reference slices by MD5 and pulled only chr5 at 173 MB, or download the full 3.26 GB analysis-set FASTA. Run the failure first, then the fix.
Q: Why look at a repeat pileup after genotyping it, and what does the tool for that teach beyond the picture? A: Because a genotype should never be reported without seeing the alignment it came from: REViewer reconstructs local from the genotyper's realigned BAMlet and renders an SVG pileup per locus. Two practical facts from running it are teaching points in their own right: the repository is archived at v0.2.7 (2022-02-02, checked 2026-08), and there is no macOS arm64 build, so on Apple Silicon it runs under Rosetta from the osx-64 package. Tool lifecycles are part of clinical bioinformatics.
Q: What do long reads do to the repeat-sizing problem? A: They dissolve it. A 10-20 kb read spans any clinically observed SCA12 allele -- 78 units is 234 bp -- thousands of times over, so allele size is read straight off the alignment. Measured on HG002 HiFi over this locus: 69 reads in the region, mean read length 14,552 bp, 53 reads individually spanning the repeat, and a modal per-read count of exactly 10 CTG units, matching the short-read genotype with no inference. The honest per-read counts are 5, 6, 7, 9, 9, 9 and then 47 reads at 10 -- per-read noise around an unambiguous consensus. Nanopore ultralong reads span it too: 41 reads over the repeat, counted in 11 seconds by ranged access without downloading the 187.5 GB (size as of 2026-08). A purpose-built genotyper (TRGT) additionally reports consensus allele sequences and repeat methylation from HiFi 5mCpG calls, with Mendelian concordance of 98.38% across 937,122 tandem repeats, but is linux-only in bioconda.
Q: How is somatic of a repeat actually measured, and what does each method give you? A: Bulk hands you one smear whose modal peak is the inherited allele, so mosaicism has to be recovered from the shape around it. Small-pool PCR dilutes genomic DNA until only a few amplifiable molecules enter each reaction, so molecules are sized one at a time and the distribution is built up reaction by reaction. The expansion index instead summarises a bulk fragment trace by how much of its signal sits above the modal allele, giving one number per tissue and per age; how that summary is computed is a protocol convention, and neither wet-lab assay has an entry in this course's verified-facts file, so treat both descriptions as orientation rather than sourced protocol. Long reads replace the summary with the distribution itself, because every read is one molecule: over this locus in GIAB HG002 HiFi, 53 spanning reads give per-read counts of 5, 6, 7, 9, 9, 9 and then 47 reads at exactly 10 -- which is what per-read noise around a stable allele looks like, and is the baseline any claim of mosaicism has to beat. For SCA12 itself the whole evidence base is one brain, in which mosaicism was detected across regions with the least unstable.
Q: Give the PPP2R2B tissue expression profile, and say why the version number matters. A: GTEx v10 medians in : Brain_Frontal_Cortex_BA9 29.71 (the highest tissue), Brain_Cortex 24.80, Brain_Cerebellum 10.02, Brain_Cerebellar_Hemisphere 8.78, Whole_Blood 0.40, Liver 0.05, Muscle_Skeletal 0.04. The v8 figures are close -- 29.90, 24.66, 10.16, 8.76, with a median of 0.50 across 41 non-brain tissues and 0.011 in cultured fibroblasts. Name the release wherever these appear, because the numbers move with every release and the gene identifier version moves with them.
Q: "PPP2R2B is cerebellum-enriched, which is why SCA12 is a cerebellar disease." Why is that wrong? A: Because the data say brain-enriched, not cerebellum-enriched. Cerebellum ranks below every cortical and basal-ganglia region GTEx samples, and frontal cortex is about 3x cerebellum; the Human Protein Atlas calls the gene group-enriched (brain, retina, testis) with low region specificity within brain. Independently, the Bbeta2 isoform is reported as predominantly forebrain. So the cerebellar of SCA12 is not explained by where the gene is transcribed -- it is an instance of the selective-vulnerability problem, and saying so is more honest and more interesting than the tidy story.
Q: What can bulk cerebellar not rule out, and why? A: Purkinje-cell enrichment. Bulk cerebellar RNA is dominated by granule-cell transcripts, because granule cells outnumber by roughly 10^3 to 10^4, so a gene expressed strongly and specifically in Purkinje cells would be diluted into invisibility. What the bulk data rules out is cerebellum-level enrichment, not cell-type-level enrichment. No single-nucleus analysis of PPP2R2B by cerebellar cell type has been published, so the question is open and should be stated as open.
Q: Why must an SCA12 expression analysis be isoform-resolved rather than gene-level? A: Because the repeat sits in one alternative first exon, so "the gene's expression" is a sum over transcripts that do not all carry the repeat, and a gene-level TPM can be flat while the repeat-containing transcript moves. The GTEx transcript API returns 21 annotated PPP2R2B transcripts, with top cerebellar isoforms ENST00000504198.5 at 6.26 TPM, ENST00000394413.7 at 2.97 and ENST00000394411.9 at 2.38. There is no median-transcript flat file in the v10 bucket and the full transcript matrix is 4.34 GB, so the portal API is the laptop-scale route.
Q: Which single-cell resource resolves cerebellar cell types on a laptop, and what are its practical gotchas? A: The Siletti human brain atlas -- about 3 million nuclei, roughly 100 dissections, 3 donors -- on CZ CELLxGENE. The cerebellar vermis dissection is 71,874 cells in a 431 MB h5ad, downloadable without registration; the lateral hemisphere dissection is 28,028 cells at 160 MB and the cerebellar-inhibitory supercluster 14,411 cells at 150 MB -- all file sizes as of 2026-08, and they move with every dataset revision. Gotchas: cellxgene h5ads index var by Ensembl ID, so PPP2R2B is ENSG00000156475, not a symbol; the download URLs are content-addressed and can rotate on dataset revision, so re-resolve via the collection on a 404; and the zero-install fallback is the browser Explorer, with all quantitative work done on the GTEx tables.