Background and Rationale
Every ME/CFS study struggles with the same fundamental confound: is an observed biological abnormality a cause of the illness, a consequence of it, or a genetic predisposition shared by both? Cross-sectional case-control designs cannot distinguish these possibilities. Longitudinal studies capture temporality but not causation. Randomised trials test interventions but not underlying biology.
The monozygotic (MZ) twin discordant design eliminates the largest confound in one stroke. MZ twins share 100% of their germline DNA sequence, their early rearing environment, their parental socioeconomic status, and their childhood diet. When one twin develops ME/CFS and the other does not, the unaffected co-twin is the optimal control — matched on every genetic and early-environmental variable that plagues conventional studies. Any difference between the twins is attributable to the illness or its trigger, not to shared predisposition (Buchwald et al. 2001) (Koelle et al. 2002).
This design has already been used productively in ME/CFS: a 2-day CPET case report in identical twins discordant for ME/CFS showed a 13% VO₂peak decline only in the affected twin (Giloteaux, Hanson, and Keller 2016), providing within-subject replication of the Day-2 decline. A platelet mitochondrial proteomics study in two twin pairs found differentially expressed proteins in the affected twins (Ciregia et al. 2016). A 22-pair HSV serology study found no difference in herpesvirus antibodies between affected and unaffected co-twins (Koelle et al. 2002) — a null result that constrains the viral persistence hypothesis in ME/CFS.
But these prior twin studies used small samples (n=1–22 pairs) and measured blood-based biomarkers. No twin study has combined the three domains where DecodeME’s genetic findings now demand investigation: striatal dopaminergic imaging, microbiome composition, and herpesvirus lytic-to-structural antibody ratios.
The Genetic Ratchet: Why Twin Design Now
DecodeME (n=15,579 cases, n=259,909 controls) identified eight genome-wide significant loci and a brain-wide tissue enrichment pattern that changes the strategic value of the twin design (DecodeME Consortium, Ponting, et al. 2025).
Brain tissue enrichment. MAGMA gene-tissue analysis found significant enrichment in all 13 brain tissues examined, with glutamatergic synapse genes (SHISA6, UNC13C) among the implicated loci (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025).
Cell-type convergence on striatal MSNs. The Maccallini 2026 meta-GWAS (n=19,470) found medium spiny neurons (MSNs) in the striatum as the most specific cell-type hit using the Human Brain Atlas pipeline, replicated in Dropviz (7/13 significant cell types are striatal neurons) (Maccallini 2026). Immune cell enrichment (ImmGen) was null.
Rare variant convergence. An independent rare-variant analysis (Snyder 2025) identified synaptic genes (NLGN2, SYNGAP1) — two independent methodological paths converging on neuronal biology (Snyder, Zhao, et al. 2025).
Heritability estimates. Twin studies yield h² ≈ 0.51 (95% CI: 0.37–0.65) for ME/CFS, while DecodeME’s SNP-based heritability is h²_SNP = 0.095 — the 5.4-fold gap means ~81% of the genetic liability is “missing” (Buchwald et al. 2001) (DecodeME Consortium, Ponting, et al. 2025). Shared environment, rare variants, epistasis, and gene-environment interactions fill the gap.
These findings create a specific prediction: if ME/CFS genetic risk operates through striatal circuits, and if environmental triggers (infection, stress) activate this vulnerability, then discordant MZ twins should differ in striatal imaging markers despite identical genomes. The unaffected co-twin carries the same genetic vulnerability — but lacks the illness, meaning the difference between twins isolates the acquired pathology.
Three Convergent Measurement Domains
Three domains cross-validated by the DecodeME findings, plus the LSR connecting to a validated biomarker hypothesis:
Striatal VMAT2 PET. Liu et al. (2026) found 16–20% reduction in VMAT2 binding across ventral striatum, dorsal putamen, and dorsal caudate in long COVID versus controls (P = 4×10⁻⁵), with reductions comparable to mild-moderate Parkinson’s disease (Liu et al. 2026). The same cohort showed striatal TSPO PET (microglial activation) (Braga et al. 2023) and MAO-B PET (astrogliosis) (Braga et al. 2025). COVID-recovered controls had normal VMAT2, confirming specificity to persistent illness. Zero VMAT2 PET data exist in ME/CFS. The DecodeME MSN enrichment provides a genetic rationale for why the striatum might be the circuit where vulnerability resides and pathology manifests.
DAT SPECT. DaTSCAN (ioflupane I-123) binds the dopamine transporter (DAT) on presynaptic dopaminergic terminals — a distinct presynaptic marker from VMAT2 (vesicular monoamine transporter 2). DAT SPECT is widely available in clinical nuclear medicine (unlike VMAT2 PET tracers) and provides a second, independent measure of presynaptic terminal integrity. DAT availability in ventral striatum predicts willingness to expend effort for reward (Treadway et al. 2012). The clinical phenotype of ME/CFS (effort intolerance, motor slowing, apathy) maps onto striatal dopaminergic function. Together, VMAT2 PET and DAT SPECT provide complementary presynaptic information: VMAT2 measures vesicular packaging capacity, while DAT measures transporter-mediated reuptake at the membrane.
Gut microbiome. If striatal differences exist, where does the signal originate? The gut-brain axis is one candidate: microbiome composition predicts fatigue severity in ME/CFS (Nagy-Szakal et al. 2017), and the striatum is a primary target of gut-derived inflammatory signals. Measuring microbiome in the same twin pairs connects peripheral trigger to central outcome.
Lytic-to-Structural IgG Ratio (LSR). The LSR (anti-BZLF1 IgG / anti-VCA-p18 IgG) is proposed as a biomarker discriminating ALR-driven antibody abnormality from LLPC-maintained background serology Lytic-to-Structural IgG Ratio (LSR) as a Diagnostic Biomarker. Measuring LSR in twin pairs tests whether LSR elevation is a trait (present in both twins — genetic predisposition to herpesvirus dysregulation) or a state (present only in the affected twin — acquired dysregulation). This directly tests the mechanism behind the LSR hypothesis.
Connection to the Identical Twin Matcher Software
The Identical Twin Matcher is a patient-facing software concept that finds the K most similar patients in a de-identified registry and reports their trajectories. The present proposal provides the research-grade infrastructure that the Matcher requires: a validated twin registry with deep phenotyping. The “identical twin” in the software concept is metaphorical — a nearest-neighbor lookup into a database. The present proposal deploys literal twins as the ultimate genetic control, generating data that could seed the Matcher’s database with uniquely informative entries.
Hypothesis
In monozygotic twin pairs discordant for ME/CFS:
Primary (striatal imaging): The affected twin will show reduced striatal VMAT2 binding (≥10% reduction on (+)-11C-DTBZ VMAT2 PET or 18F-AV-133 VMAT2 PET versus the unaffected co-twin) in at least one striatal subregion (ventral striatum, dorsal putamen, dorsal caudate), with effect size δ ≥ 0.5 SD of the within-pair difference distribution. The unaffected co-twin will show VMAT2 binding in the normal range (within 1 SD of population control mean). DAT availability (DaTSCAN SPECT) will differ between affected and unaffected twins in a pattern that discriminates vesicular pathology (VMAT2↓ + DAT normal) from structural terminal loss (VMAT2↓ + DAT↓) or isolated transporter pathology (VMAT2 normal + DAT↓).
Secondary (microbiome): The affected twin will show reduced gut microbiome α-diversity (Shannon index, Chao1) and altered β-diversity (Bray-Curtis dissimilarity) relative to the unaffected co-twin, with differentially abundant taxa enriched for butyrate-producing species (Faecalibacterium, Roseburia) in the unaffected twin. Microbiome diversity metrics will correlate with striatal VMAT2 binding across all twins (r ≥ 0.3, P ≤ 0.05), consistent with a gut-striatal axis.
Tertiary (LSR): LSR (anti-BZLF1 IgG / anti-VCA-p18 IgG) will be elevated in the affected twin versus the unaffected co-twin if the LSR reflects acquired antibody dysregulation. If LSR is elevated in both twins versus population controls but not different between twins, the LSR reflects a genetic predisposition to herpesvirus antibody production rather than an illness-specific process.
Falsifiability. The design is falsified at multiple levels:
- Null at all tiers: no significant within-pair difference in VMAT2 binding, DAT availability, microbiome α-diversity, or LSR (all P ≥ 0.05, uncorrected). Demonstrates that the long COVID VMAT2 finding does not generalise to ME/CFS, and that the DecodeME brain enrichment reflects genetic vulnerability without acquired striatal pathology.
- Discordant pattern: VMAT2 reduction in the affected twin but no microbiome difference. Isolates the CNS as the primary locus of pathology and suggests microbiome changes are downstream of illness behaviour (diet, inactivity) or unrelated.
- Discordant pattern: microbiome dysbiosis present but VMAT2 normal. Suggests the microbiome-immune axis as the primary pathology, with striatal changes secondary or absent — consistent with the peripheral immune abnormalities documented in ME/CFS.
- LSR trait pattern: LSR elevated equally in both twins versus controls. Refutes the state-marker interpretation of the LSR hypothesis and establishes LSR as a genetic trait — still diagnostically useful but mechanistically different.
- Severity gradient: within-pair VMAT2 difference magnitude correlates with between-pair illness severity difference (r ≥ 0.4, P ≤ 0.05). If no gradient exists, observed differences may reflect measurement noise rather than pathology.
- Correlation structure: if VMAT2-microbiome correlation is absent despite significant individual-group differences, the gut-striatal axis hypothesis is falsified for this pathway.
Design Type
Cross-sectional within-subject (within-pair) comparison of MZ twin pairs discordant for ME/CFS, with a nested comparison to healthy MZ twin pairs (concordant for health) as an imaging reference group.
Core design: Each twin pair provides its own control. The unaffected co-twin is matched on 100% of germline DNA, early-life environment, parental SES, and childhood diet. The within-pair comparison isolates the nongenetic component of ME/CFS — what changed between two genetically identical individuals.
Reference group: 20 healthy MZ twin pairs (both twins healthy, age/sex-matched to the ME/CFS pairs) provide normative VMAT2 and DAT ranges. This addresses a critical gap: normative VMAT2 PET values in healthy young-to-middle-aged adults are sparsely documented (Liu et al. 2026).
Zygosity confirmation: Microsatellite or SNP-based zygosity testing (≥12 markers) on all pairs. Self-reported zygosity has 2–5% error rate, unacceptable in a design where genetic identity is the foundational premise.
Cross-sectional, not longitudinal: The design captures one timepoint. A longitudinal extension (re-imaging after treatment, natural recovery, or worsening) would dramatically increase power but also cost and participant burden. The longitudinal extension is a future direction.
Recruitment Strategy and Sample Size
Recruitment sources:
- Existing ME/CFS twin registries and cohorts (DecodeME database includes twin status query; Swedish Twin Registry; UK Twins Registry)
- ME/CFS patient organisations (Solve M.E., MEAction, ME Association, Open Medicine Foundation networks)
- Social media outreach targeting ME/CFS patient communities
- Clinical referral networks (ME/CFS specialist clinics with existing patient databases)
Primary sample: 30 MZ twin pairs discordant for ME/CFS (n=60 individuals).
Rationale for n=30 pairs: A feasibility-constrained maximum, not a power-driven optimum. MZ twins discordant for ME/CFS are rare. The disease affects ~0.2–0.4% of the population, and MZ twins constitute ~0.3% of births. The expected yield from comprehensive recruitment across all available registries is 20–40 pairs.
Power analysis at n=30 pairs:
- Paired t-test (within-pair comparison of affected vs unaffected twin): 80% power to detect δ = 0.53 SD at α = 0.05 (two-tailed). The Liu 2026 long COVID VMAT2 effect size was δ ≈ 1.4 – 1.6 SD for ventral striatum and dorsal putamen (Liu et al. 2026), so n=30 pairs has adequate power to detect effects ≥half the magnitude reported in long COVID.
- Correlation (VMAT2 × microbiome diversity): For correlation analyses across all individuals (n=100), 80% power to detect r = 0.28 at α = 0.05 (two-tailed). The more conservative paired analysis (treating each twin pair as the unit, n=30 pairs) has 80% power to detect r = 0.49.
- ANCOVA (severity gradient): 80% power to detect f² = 0.25 (medium effect) with 1 predictor.
Sensitivity analysis: If only n=20 pairs are recruited, power drops to 80% for δ = 0.66 SD and r = 0.59. The study is powered for effects ≥0.7 SD at n=20 — still adequate to detect the Liu 2026 effect magnitude but underpowered for moderate effects.
Reference group: 20 healthy MZ twin pairs (n=40), age/sex-matched to the ME/CFS pairs. These provide normative VMAT2 and DAT reference ranges. Healthy co-twins in the ME/CFS pairs already serve as within-pair controls; the healthy pairs establish whether the unaffected co-twin is truly “normal” or shows subclinical imaging abnormalities (e.g., intermediate VMAT2 binding suggesting incomplete penetrance of genetic vulnerability).
Healthy reference pairs receive all measures (VMAT2 PET, DAT SPECT, microbiome, LSR, blood biomarkers, clinical phenotyping) on the identical protocol. Their microbiome and LSR data provide population reference ranges for within-pair comparison interpretation. A secondary analysis compares unaffected co-twins in the ME/CFS pairs to healthy twins in the reference pairs — this tests whether the unaffected co-twin is truly normal or carries subclinical abnormalities.
Inclusion criteria for affected twin:
- Meets IOM 2015 diagnostic criteria for ME/CFS
- Post-infectious onset preferred (≥60% of sample; mixed onset retained for generalisability)
- Illness duration ≥2 years (avoid early-phase instability)
- Willing to discontinue CNS-active medications for 5 half-lives before imaging (exceptions: stable SSRIs ≥3 months, documented and unchanged)
- Able to travel to imaging site (this selects against severe/very severe patients — a limitation addressed in Section Radiation Exposure Cumulation)
Inclusion criteria for unaffected co-twin:
- Does NOT meet any ME/CFS diagnostic criteria (IOM, CCC, ICC)
- No chronic fatigue (fatigue duration \(<\) 1 month)
- SF-36 physical function score ≥85
Exclusion criteria (both twins):
- Contraindication to PET or SPECT (pregnancy, breastfeeding, radiation exposure concern)
- History of traumatic brain injury with loss of consciousness \(>\) 30 minutes
- Current substance use disorder (excludes dopamine pathway confound)
- Antipsychotic use within 6 months (antipsychotics alter striatal dopaminergic signalling, a potential confound)
- Parkinson’s disease, Huntington’s disease, or other known striatal pathology
- Antibiotic use within 3 months (microbiome confound)
- Proton pump inhibitor use within 1 month (microbiome confound)
- Probiotic or prebiotic supplementation within 2 weeks
Study Visit Schedule
Single study visit per participant, spread over 2–3 days (both twins simultaneously).
-Day 1 (arrival): Consent, clinical phenotyping questionnaires, cognitive testing, motor testing, autonomic testing, blood draw (serum, plasma, PBMC, PAXgene), dietary/lifestyle interview. Stool sample kit provided (return within 48 hours of collection at home or on-site).
-Day 2 (PET imaging): Morning VMAT2 PET (60–90 min dynamic scan, tracer injection at t=0). Afternoon rest. No strenuous activity for 24 hours pre-scan (to avoid exercise-induced dopamine release confound).
-Day 3 (SPECT imaging, optional return): If DAT SPECT is performed on a separate day (radiopharmaceutical scheduling). DaTSCAN injection → 3–4 hour uptake → 30–45 min SPECT acquisition.
-Post-visit (1 week): Actigraphy device returned. Stool sample shipped to microbiome core. Food diary and any outstanding questionnaires completed remotely.
Between-pair timing: Pairs are scheduled such that both twins are imaged within 7 days of each other to minimise temporal confounds (seasonal microbiome variation, time-of-year effects on mood/activity). The healthy reference pairs follow the identical protocol.
PEM management: No 2-day CPET or physical challenge is included because of the risk of triggering PEM that would confound the imaging measures (neuroinflammation, microbiome changes post-exertion). The study is an at-rest assessment of baseline striatal function, not a stress test. If a PEM substudy is added in a future protocol (re-imaging 24–48 hours after controlled exercise in a subset), that would require a separate visit and separate funding.
Measures
Primary PET Imaging: Striatal VMAT2 and DAT
-Tracer: (+)-11C-DTBZ, or alternatively 18F-AV-133 if a longer-half-life tracer is logistically necessary (multicentre design). The Liu 2026 long COVID study used (+)-11C-DTBZ, providing a direct comparison dataset (Liu et al. 2026).
-Scanner: High-resolution PET/CT or PET/MR. MRI-based attenuation correction preferred for striatal subregion delineation.
-Quantification: Non-displaceable binding potential (BP_ND) using the simplified reference tissue model with occipital cortex as reference region. Striatal subregions: ventral striatum (nucleus accumbens + ventral caudate + ventral putamen), dorsal putamen (pre-commissural and post-commissural), dorsal caudate (head, body). Brainstem (substantia nigra, ventral tegmental area, locus coeruleus) as exploratory regions.
-Analytical approach: Paired t-test on BP_ND within each subregion (affected vs unaffected co-twin). False Discovery Rate (FDR) correction across 5 subregions (q < 0.05). Effect size reported as Cohen’s d_z (within-pair standardised mean difference) with 95% CI.
-DAT SPECT: Standard clinical DaTSCAN (ioflupane I-123) protocol. DaTSCAN binds the dopamine transporter (DAT) on presynaptic dopaminergic terminals — a distinct presynaptic marker from VMAT2. This radioligand is FDA-approved and available in most nuclear medicine departments. Quantify specific binding ratio (SBR) in the same striatal subregions. Together, VMAT2 and DAT provide complementary presynaptic information: VMAT2 measures vesicular monoamine packaging capacity, while DAT measures transporter-mediated dopamine reuptake at the terminal membrane.
-Discordance pattern interpretation (both markers are presynaptic):
- VMAT2↓ + DAT↓: concordant reduction on both presynaptic markers. Indicates structural terminal loss or severe presynaptic dysfunction. Most consistent with dopaminergic terminal degeneration.
- VMAT2↓ + DAT normal: vesicular packaging deficit without structural terminal loss. Suggests a functional impairment in vesicular loading (e.g., ATP-dependent VMAT2 proton gradient dysfunction) rather than terminal loss.
- VMAT2 normal + DAT↓: transporter-specific pathology. DAT downregulation or internalization without vesicular deficit — could reflect compensatory adaptation to elevated synaptic dopamine or a primary DAT regulatory abnormality.
- Both normal: striatal dopaminergic system is NOT the locus of pathology in ME/CFS. Falsifies the primary hypothesis.
-TSPO PET (optional, funding-dependent): Subset of n=10 ME/CFS pairs receive TSPO PET (e.g., 11C-PBR28) to quantify microglial activation. Genotyping for TSPO rs6971 polymorphism required (high-affinity binders only). Determines whether VMAT2 reduction is accompanied by neuroinflammation as in long COVID (Braga et al. 2023), or whether the pathology is non-inflammatory.
Secondary Measures
—–Gut microbiome.
- Stool sample collection: OMNIGene-GUT or DNA/RNA Shield collection kit. Single sample per participant (morning void, home collection, shipped at ambient temperature within 48 hours).
- Sequencing: Shotgun metagenomics (Illumina NovaSeq, ≥10M paired-end reads per sample). 16S rRNA gene sequencing (V3–V4 region) as backup/validation.
- Bioinformatic pipeline: KneadData (quality control, host decontamination) → MetaPhlAn 4 (taxonomic profiling) → HUMAnN 3 (functional pathway profiling).
- Primary metrics: α-diversity (Shannon index, Chao1, observed species), β-diversity (Bray-Curtis dissimilarity, weighted/unweighted UniFrac), differentially abundant species (ALDEx2, MaAsLin 2 with within-pair blocking).
- Functional analysis: KEGG modules, CAZy (carbohydrate-active enzymes), short-chain fatty acid synthesis pathways (butyrate, propionate, acetate).
—–Lytic-to-Structural IgG Ratio (LSR).
- Primary ratio: anti-BZLF1 IgG ÷ anti-VCA-p18 IgG, quantified by ELISA with standard curves (not seroprevalence titres). See Lytic-to-Structural IgG Ratio (LSR) Diagnostic Biomarker Validation for full assay methodology.
- Secondary: anti-EA-D IgG, anti-EBNA-1 IgG, avidity (urea 8M wash ELISA for anti-BZLF1 and anti-VCA-p18). Avidity index \(<\) 0.4 = low (recent primary infection or recent ALR with low-affinity SLPB output); 0.4–0.6 = intermediate; \(>\) 0.6 = high (LLPC-derived, remote infection).
- Poly-herpesvirus extension (optional): anti-HHV-6 U45 dUTPase IgG, anti-CMV gB IgG, anti-VZV gE IgG.
- Twin-pair analysis: within-pair LSR difference (affected – unaffected). One-sample paired t-test comparing mean within-pair difference to zero. Also compare: LSR in affected twins vs healthy reference pairs; LSR in unaffected twins vs healthy reference pairs (tests whether unaffected co-twins carry subclinical LSR elevation).
—–Clinical phenotyping (both twins, identical protocol).
- ME/CFS diagnostic: IOM 2015 criteria questionnaire, DSQ-PEM (PEM severity and frequency), DePaul Symptom Questionnaire
- Fatigue severity: Fatigue Severity Scale (FSS), Multidimensional Fatigue Inventory (MFI-20)
- Physical function: SF-36 physical function subscale, 6-minute walk test (if tolerated)
- Cognitive function: Symbol Digit Modalities Test, Trail Making Test A and B
- Apathy and motivation: Apathy Evaluation Scale (AES), Effort Expenditure for Rewards Task (EEfRT) — maps to ventral striatal dopaminergic function (DAT and VMAT2) (Treadway et al. 2012)
- Motor function: Finger Tapping Test (maps to dorsal putamen VMAT2)
- Sleep: Pittsburgh Sleep Quality Index (PSQI), actigraphy (7 days pre-imaging)
- Pain: Brief Pain Inventory (BPI)
- Autonomic: COMPASS-31, 10-minute stand test (heart rate and blood pressure at supine, 1, 2, 5, 10 min)
—–Blood biomarkers (exploratory).
- Inflammatory: hs-CRP, IL-6, TNF-α, IL-1β
- Neurofilament light chain (NfL, serum) — marker of axonal injury; elevated in long COVID with cognitive impairment
- Epstein-Barr virus qPCR (plasma) — correlation with LSR and with VMAT2 binding (tests viral reactivation → striatal pathology pathway)
- TSPO rs6971 genotyping (if TSPO PET substudy is performed)
- Metabotropic: tryptophan, kynurenine, kynurenic acid, quinolinic acid (kynurenine pathway — glial activation marker, links microbiome to striatum)
—–Diet and lifestyle covariates (both twins).
- 3-day food diary with nutritional analysis
- International Physical Activity Questionnaire (IPAQ)
- Current medications and supplements (full documentation)
- Smoking, alcohol, and substance use history
- Menstrual cycle phase at time of imaging (female participants)
Expected Outcomes and Implications
If the Primary Hypothesis Is Confirmed
VMAT2 reduced in affected vs unaffected co-twin. The first direct evidence of striatal dopaminergic terminal pathology in ME/CFS. The twin design isolates this as an acquired pathology — not a genetic predisposition shared by both twins. The DecodeME MSN enrichment transitions from a statistical association to a functionally validated disease locus. The finding:
- Establish VMAT2 PET as a quantitative biomarker for ME/CFS pathophysiology
- Provide a circuit-level target for dopaminergic augmentation trials (L-DOPA, MAO-B inhibitors, dopamine agonists — see Dopaminergic Augmentation for Apathy-Predominant and Motor-Slowing Phenotypes in Post-Infectious ME/CFS)
- Link ME/CFS to the established long COVID VMAT2 finding within a unified post-infectious striatal pathology framework
- Create a stratification tool: VMAT2-low vs VMAT2-normal ME/CFS subgroups, identifiable by imaging, corresponding to apathy/motor-slowing vs non-motor phenotypes
DAT concordance or dissociation. If the affected twin shows concordant VMAT2↓ + DAT↓, structural terminal loss is confirmed — presynaptic replacement strategies (L-DOPA, dopamine agonists) face a reduced terminal substrate. If VMAT2↓ with preserved DAT, the pathology is vesicular rather than structural — dopaminergic augmentation strategies that bypass vesicular packaging (e.g., MAO-B inhibitors to slow synaptic dopamine degradation) may be more rational than L-DOPA. If VMAT2 normal but DAT↓, a primary DAT regulatory abnormality is implicated — a novel mechanism not previously considered in ME/CFS.
Microbiome-gut-striatal axis confirmed. If microbiome diversity correlates with VMAT2 binding, provides the first evidence of a gut-to-striatal pathway in ME/CFS. Links the microbiome literature — which has produced inconsistent findings partly because of uncontrolled confounds — to a specific CNS endpoint. The twin design controls for diet, genetics, and environment, isolating the microbiome-striatum association.
LSR elevated only in affected twin. Establishes LSR as a state marker of ME/CFS — not a genetic trait. Validates the hypothesis that antibody dysregulation is acquired and illness-linked, supporting the ALR-mechanism interpretation of the LSR cascade (Rethinking Herpesvirus Serology in ME/CFS). Accelerates the LSR biomarker validation study proposed in Lytic-to-Structural IgG Ratio (LSR) Diagnostic Biomarker Validation by providing within-subject genetic control evidence before the larger cross-sectional study completes.
If the Hypothesis Is Refuted
Scenario A: All primary measures null (VMAT2, DAT, microbiome, LSR show no within-pair difference).
The most informative null result in the history of ME/CFS biomarker research. The twin design controls every confound that plagues conventional studies. If no difference exists between affected and unaffected MZ co-twins across these four domains, one of two conclusions follows:
- ME/CFS pathology is not detectable by the methods employed (correct biology, wrong tools — e.g., blood-based metabolites, not PET-visible terminal density, or functional connectivity rather than receptor availability).
- ME/CFS pathology is not located in the striatum (DecodeME enrichment reflects developmental vulnerability, not ongoing pathology, and the long COVID VMAT2 finding is COVID-specific, not generalisable to other post-infectious syndromes).
Both conclusions are scientifically valuable. The null twin study is the most powerful negative evidence possible — it eliminates confounds that leave all other null studies ambiguous. A null result here would redirect the field away from striatal dopaminergic imaging and toward alternative circuits (brainstem noradrenergic, cortical glutamatergic, thalamic) or alternative measurement modalities (fMRI connectivity, MR spectroscopy).
Scenario B: Discordant results across domains (e.g., VMAT2↓ but microbiome normal).
This isolates the locus of pathology. VMAT2 reduction without microbiome dysbiosis suggests the striatum is the primary site and microbiome changes in prior studies were downstream of illness behaviour. Conversely, microbiome dysbiosis without VMAT2 reduction suggests the gut-immune axis is primary and the DecodeME brain enrichment reflects developmental vulnerability that does not manifest as ongoing striatal pathology.
Scenario C: LSR elevated in both twins.
Refutes the LSR-as-state-marker hypothesis. The LSR is a genetic (or shared-environment) trait — both twins carry elevated lytic herpesvirus antibodies, reflecting a common early-life infection or a shared immune response genotype. The LSR remains useful as a diagnostic (if elevated in both twins vs population controls) but loses its proposed mechanism (ALR-driven acquired dysregulation). Consistent with the Koelle 2002 finding that HSV antibody levels did not differ between discordant twins (Koelle et al. 2002).
Scenario D: Unaffected co-twin shows intermediate VMAT2 reduction.
If the unaffected co-twin has VMAT2 binding intermediate between the affected twin and healthy reference pairs, the genetic vulnerability hypothesis is supported: both twins carry the striatal vulnerability genotype, but only the affected twin crossed the clinical threshold. The first direct evidence of incomplete penetrance of a striatal endophenotype in ME/CFS — the unaffected twin “looks like ME/CFS” on imaging but has no symptoms. Longitudinal follow-up of these intermediate co-twins is essential to determine whether they are at elevated risk of developing ME/CFS.
Scenario E: Severity gradient absent.
If the affected twin shows VMAT2 reduction but the magnitude does not correlate with illness severity, the imaging finding may be a categorical biomarker (present/absent) rather than a continuous severity marker. Useful for diagnosis but not for monitoring disease progression or treatment response.
Budget and Timeline
Budget Summary
Total estimated cost: ~$3.8M over 4 years.
| VMAT2 PET imaging (30 ME/CFS pairs + 20 healthy pairs = 100 scans) |
$1,500,000 |
~$15,000/scan (tracer synthesis, scanner time, radiologist read, MRI for attenuation correction) |
| DAT SPECT (100 scans) |
$500,000 |
~$5,000/scan (DaTSCAN is FDA-approved; widely available at lower cost than research PET tracers) |
| TSPO PET substudy (20 scans, optional) |
$300,000 |
~$15,000/scan; only if funding permits |
| Microbiome (shotgun metagenomics, 100 samples) |
$150,000 |
~$1,500/sample (DNA extraction, library prep, sequencing, bioinformatics) |
| Serology (LSR panel, 100 samples) |
$80,000 |
ELISA panels (anti-BZLF1, anti-VCA-p18, anti-EA-D, anti-EBNA-1, avidity) |
| Blood biomarkers (NfL, cytokines, kynurenine pathway, 100 samples) |
$60,000 |
~$600/sample |
| Twin registry recruitment and coordination |
$200,000 |
2-year FT study coordinator + outreach + travel reimbursement for participants |
| Participant travel and accommodation |
$300,000 |
~$3,000/pair (travel, 3 nights accommodation, meals for both twins; imaging site may be distant from home) |
| Clinical phenotyping (cognitive, motor, autonomic testing) |
$100,000 |
~$1,000/pair (psychometrist time, equipment) |
| Data management, biostatistics, and analysis |
$200,000 |
1-year FT biostatistician + data infrastructure |
| Zygosity testing (60 pairs) |
$30,000 |
~$500/pair (SNP panel) |
| Actigraphy devices (100 units) |
$15,000 |
~$150/unit (wrist-worn, 7-day) |
| Subtotal |
$3,435,000 |
|
| Contingency (10%) |
$343,500 |
|
| Total |
~$3,778,500 |
|
Cost-saving measures:
- If DAT SPECT is deferred to a Phase 2 extension, subtract ~$500,000
- If TSPO PET is excluded, subtract ~$300,000
- If a single imaging site with high-volume throughput is used, VMAT2 PET cost may reduce to ~$10,000/scan (total saving: ~$500,000)
- Minimum viable study (VMAT2 PET + LSR + microbiome, no SPECT, no TSPO): ~$2.1M
- Minimum viable with DAT SPECT added: ~$2.6M
- Minimum viable with DAT SPECT + TSPO subset: ~$2.9M
Timeline
| Phase 0: Preparation |
Months 1–6 |
IRB approval, tracer synthesis validation, zygosity assay setup, microbiome collection kit validation, LSR ELISA pilot (n=20 healthy controls for reference range), site agreements, database build |
| Phase 1: Recruitment |
Months 7–18 |
Twin registry activation, patient organisation outreach, social media campaign, clinical referral network engagement, eligibility screening, zygosity confirmation |
| Phase 2: Data Collection |
Months 13–36 |
Rolling twin-pair visits (1–2 pairs/month). Clinical phenotyping → microbiome collection → PET → SPECT → actigraphy return. Healthy reference pairs interleaved with ME/CFS pairs. The 6-month overlap between Phase 1 and Phase 2 reflects rolling enrollment — imaging begins for the first enrolled pairs while recruitment continues for remaining slots. |
| Phase 3: Analysis |
Months 30–42 |
Primary analysis (months 30–36): within-pair VMAT2, DAT, microbiome, LSR. Secondary analysis (months 36–42): correlational analyses, severity gradients, VMAT2-microbiome axis, exploratory blood biomarkers. |
| Phase 4: Dissemination |
Months 42–48 |
Manuscript preparation, preprint, peer review, conference presentations, data sharing through dbGaP or equivalent repository. |
| Total |
4 years |
|
Recruitment feasibility note: The rate-limiting step is not recruitment volume (there are likely hundreds of discordant MZ pairs globally) but recruitment within travel distance of the imaging site. A single-site design limits the catchment area. A two-site design (e.g., one North American, one European site) would double recruitment capacity and reduce participant travel burden but increase coordination complexity and inter-site scanner variability. The budget above assumes a single-site design; a two-site design would add ~$500,000 for scanner harmonisation and second-site staffing.
Limitations
Discordant MZ twins with ME/CFS are rare. The disease prevalence (~0.2–0.4%) multiplied by the MZ twin birth rate (~0.3% of births) yields approximately 0.6–1.2 affected MZ twins per 100,000 individuals. Of these, only the subset whose co-twin remains healthy AND who can travel to the imaging site is eligible. The yield from all available registries and outreach is uncertain — the study may recruit fewer than 30 pairs. At n=15, power is ~54% for δ = 0.53 SD, and only large effects (δ > 1.0) would be detectable. The study’s feasibility depends on comprehensive, multi-national recruitment with the willingness to fund travel.
Mitigation: If recruitment lags behind schedule at month 18 (fewer than 15 pairs enrolled), the study pivots from a powered primary analysis to a descriptive/pilot design with effect size estimation and Bayesian analysis (reporting posterior distributions rather than hypothesis tests). The data would still be uniquely informative — even n=10 pairs with imaging is the largest MZ twin imaging study in ME/CFS by an order of magnitude.
The requirement to travel to an imaging site and undergo two days of testing selects for mild-to-moderate ME/CFS patients who can tolerate travel, sustained upright posture, and cognitive testing. Severe and very severe patients (~25% of the ME/CFS population) are systematically excluded. This matters because the striatal pathology might be more pronounced in severe patients — the observed effect size could underestimate the true population effect. If the null hypothesis is not rejected, it could be because VMAT2 reduction exists only in the severe subgroup that was not sampled.
Mitigation: Include home-visit options for clinical phenotyping in patients who cannot travel. Consider a mobile imaging unit (MRI truck with PET capability) in a future extension. Acknowledge that generalisability to severe/very severe patients is unknown.
The cross-sectional design captures one snapshot. It cannot determine whether VMAT2 reduction preceded illness onset (pre-existing vulnerability that lowered the threshold for post-infectious ME/CFS) or developed after onset (acquired pathology). The twin design eliminates genetic confounding, but it does not establish temporality between the trigger event and the imaging findings.
Mitigation: Include questions about pre-illness apathy, motor slowing, and reward sensitivity (retrospective report — limited but informative). The longitudinal extension (re-imaging after recovery or worsening) would resolve temporality but requires a separate study.
(+)-11C-DTBZ requires an on-site cyclotron (carbon-11 half-life = 20.4 minutes). This restricts the study to imaging centres with cyclotron capability — typically academic medical centres. 18F-AV-133 (fluorine-18, half-life = 109.8 minutes) can be distributed regionally but is less established for VMAT2 quantification. The Liu 2026 long COVID data used (+)-11C-DTBZ, so switching to 18F-AV-133 would require tracer-specific reference ranges and limits direct comparison.
Mitigation: Partner with a PET centre that has existing VMAT2 imaging programmes (e.g., Toronto group behind the Liu 2026 study). If 18F-AV-133 is used, include a cross-validation substudy (n=5 healthy volunteers scanned with both tracers).
A single stool sample per participant captures point-prevalence microbiome composition but misses temporal variability. The microbiome fluctuates with diet, sleep, stress, and medication — a snapshot may misclassify participants. ME/CFS patients also have altered gastrointestinal transit time, which affects stool consistency and microbial composition independent of microbial ecology (Nagy-Szakal et al. 2017).
Mitigation: Include dietary and medication covariates in all analyses. Collect stool consistency data (Bristol Stool Scale). A longitudinal microbiome substudy (weekly samples for 4 weeks in n=10 pairs) would quantify within-subject variability and inform interpretation of the single-sample design.
The LSR has never been measured in any disease population — there is no reference range for healthy adults, let alone for discordant twin pairs. The interpretation of within-pair LSR differences depends on the assay’s test-retest reliability, which is unknown. If technical variability exceeds the expected biological effect size, the twin design cannot distinguish signal from noise.
Mitigation: The pilot phase (Phase 0) includes LSR measurements in n=20 healthy controls and n=5 healthy twin pairs to establish reference ranges and within-pair variability. If within-pair variability in healthy twins is high (coefficient of variation \(>\) 20%), the LSR component is demoted to exploratory.
Many ME/CFS patients take medications that affect dopaminergic signalling: SSRIs/SNRIs (indirect DA modulation via serotonin-dopamine interactions), stimulants (methylphenidate, modafinil — DAT blockade directly confounds DAT SPECT), low-dose naltrexone (opioid modulation of DA), antipsychotics (D2 blockade), antiemetics (D2 blockade). DAT blockade by stimulants is a direct confound for the DAT SPECT measure (not just an indirect pathway confound). The washout period (5 half-lives) is feasible for most medications but may be clinically unacceptable for patients stabilised on SSRIs.
Mitigation: Allow stable SSRIs (≥3 months, unchanged dose) as an exception, with sensitivity analysis excluding SSRI users. Document all medications and perform post-hoc analysis stratified by medication class. An observational compromise — the alternative (mandatory washout of all CNS-active medications) would make recruitment impossible.
VMAT2 PET + DAT SPECT + (optional) TSPO PET exposes each participant to ~15–25 mSv total effective dose, depending on tracers and administered activities. Within research ethics guidelines for studies with direct participant benefit potential; precludes repeated imaging (longitudinal design requires separate ethics justification) and may deter participation.
Mitigation: Explicit informed consent with personalised radiation risk counselling. Participant information materials include plain-language radiation dose comparison (e.g., equivalent to 5–8 years of background radiation). The single-site design avoids cumulative exposure from repeated visits.
Analysis Plan
Primary analysis (VMAT2 PET): Within-pair paired t-test of BP_ND in each of 5 striatal subregions (ventral striatum, dorsal putamen pre-commissural, dorsal putamen post-commissural, dorsal caudate head, dorsal caudate body), FDR-corrected across subregions (q < 0.05). Secondary: linear mixed-effects model with pair as random effect and affection status as fixed effect, adjusting for age, sex, and scan order (affected vs unaffected twin scanned first).
Multiplicity correction across primary outcome domains (VMAT2, DAT, microbiome α-diversity): the three primary comparisons are hierarchically ordered (tier 1: VMAT2; tier 2: DAT; tier 3: microbiome α-diversity). Tier 2 and 3 are tested only if tier 1 is significant, providing strong control of familywise error at α=0.05. If the hierarchical structure is not invoked, Benjamini-Hochberg FDR correction (q < 0.05) is applied across the three primary p-values.
DAT SPECT analysis: Identical to VMAT2 analysis within each subregion. Primary readout: within-pair DAT SBR difference. Secondary: VMAT2 × DAT interaction — does the discordance pattern (concordant vs dissociated presynaptic markers) discriminate terminal loss from functional vesicular impairment?
Microbiome analysis: Paired tests: α-diversity (Shannon index) compared within-pair using paired Wilcoxon signed-rank test. β-diversity: PERMANOVA on Bray-Curtis distance with pair as strata. Differentially abundant species: MaAsLin 2 with within-pair blocking (pair as random effect). Correlation analysis: Spearman’s ρ between Shannon index and VMAT2 BP_ND across all twins (n=100).
LSR analysis: Within-pair paired t-test of LSR. Secondary: LSR in affected twins vs healthy reference pairs (independent t-test); LSR in unaffected twins vs healthy reference pairs. Avidity index distribution: histogram with mixture model testing for bimodality.
Correlation structure across domains: Canonical correlation analysis (CCA) or regularised CCA (rCCA) relating the VMAT2 vector (5 subregions), microbiome vector (genus-level relative abundances), and LSR within each participant. This tests the multivariate hypothesis that striatal imaging, microbiome composition, and herpesvirus serology co-vary in a coordinated pattern.
Severity gradient analysis: Within-pair VMAT2 difference regressed on within-pair FSS difference, SF-36 difference, and DSQ-PEM difference. Mixed-effects models with pair-level random intercepts.
Sensitivity analyses: (a) Exclude SSRI/SNRI users; (b) exclude participants with antibiotic use ≤6 months; (c) exclude participants with prebiotic/probiotic use ≤2 weeks; (d) restrict to post-infectious onset pairs; (e) restrict to pairs where affected twin has apathy score above the median.
Data Sharing and Reproducibility
All de-identified data (VMAT2 BP_ND values, DAT SBR values, microbiome OTU tables, LSR values, clinical phenotypes) will be deposited in a public repository (e.g., dbGaP, Synapse, or Zenodo) at the time of publication. Imaging data (DICOM) will be shared through an appropriate neuroimaging repository (e.g., OpenNeuro). Twin data require additional privacy protections: within-pair identifiers will be coded such that co-twin linkage is preserved for analysis but individual identification is impossible. All analysis code will be published as version-controlled repositories with containerised environments (Docker) to ensure computational reproducibility.
Funding and Implementation
Funding sources: NIH (NINDS, NIAID, or Common Fund High-Risk High-Reward programme), Wellcome Trust, Open Medicine Foundation, or European Research Council. The study qualifies as high-risk/high-reward: the recruitment target is ambitious, the imaging modalities have never been combined in ME/CFS, and the null result is as informative as a positive one — a characteristic of well-designed high-risk studies.
Implementation consortium: A multi-institutional collaboration is required. At minimum: (1) a PET centre with VMAT2 imaging capability and an on-site cyclotron (or regional 18F-AV-133 supply chain); (2) a nuclear medicine department with DaTSCAN SPECT capability; (3) a microbiome sequencing core with shotgun metagenomics pipelines; (4) a clinical ME/CFS centre with diagnostic expertise and patient networks; (5) a twin registry with discordant-pair identification infrastructure; (6) a biostatistics group with expertise in paired designs and high-dimensional correlation structures.
Neuro-COVID bridge funding: The Toronto group (Liu, Braga, et al.) has existing VMAT2, TSPO, and MAO-B PET data in long COVID. A direct collaboration would enable cross-syndrome comparison of striatal pathology between ME/CFS and long COVID in genetically controlled (twin) and uncontrolled designs. This bridge would be a unique contribution — no study has compared striatal PET findings across post-infectious syndromes.
Relationship to Existing Work
- Identical Twin Matcher: The present study generates the deep-phenotype twin registry that the Matcher requires. The Matcher is a patient-facing nearest-neighbor lookup; the present study provides the infrastructure-grade entries.
- DecodeME ((DecodeME Consortium, Ponting, et al. 2025)): The MSN enrichment finding provides the genetic rationale for targeting the striatum. The present study tests whether this genetic signal translates to functional pathology.
- Long COVID VMAT2 PET ((Liu et al. 2026)): The present study extends the VMAT2 finding to ME/CFS, testing generalisability across post-infectious syndromes.
- LSR Biomarker Validation (Lytic-to-Structural IgG Ratio (LSR) Diagnostic Biomarker Validation): The present study provides within-subject genetic control evidence that complements the larger cross-sectional validation design.
- Striatal Dopaminergic Terminal Loss Speculation: The present study provides the empirical test of this speculation.
- Chaudhuri & Behan Striatal-Thalamic-Frontal Hypothesis (Striatal Symptom Signalling: Heterogeneous Pathology Converging on a Common Endpoint): The present study provides the first direct striatal imaging test of this 25-year-old hypothesis with genetic control.
References
Braga, Joeffre, Elad J. Y. Kuik, Mariel Lepra, Pablo M. Rusjan, Stephen J. Kish, and Jeffrey H. Meyer. 2025.
“Astrogliosis Marker [11C]SL25.1188 After COVID-19 with Ongoing Depressive and Cognitive Symptoms.” Biological Psychiatry 97 (8): 816–24.
https://doi.org/10.1016/j.biopsych.2024.09.027.
Braga, Joeffre, Mariel Lepra, Stephen J. Kish, Pablo M. Rusjan, Zahra Nasser, Nicolaas Verhoeff, Neil Vasdev, et al. 2023.
“Neuroinflammation After COVID-19 with Persistent Depressive and Cognitive Symptoms.” JAMA Psychiatry 80 (8): 787–95.
https://doi.org/10.1001/jamapsychiatry.2023.1321.
Buchwald, Dedra, Richard Herrell, Suzanne Ashton, Michael Belcourt, Karen Schmaling, Patrick Sullivan, Michael Neale, and Jack Goldberg. 2001.
“A Twin Study of Chronic Fatigue.” Psychosomatic Medicine 63 (6): 936–43.
https://doi.org/10.1097/00006842-200111000-00012.
Ciregia, Federica, Laxmikanth Kollipara, Laura Giusti, René P Zahedi, et al. 2016.
“Bottom-up Proteomics Suggests an Association Between Differential Expression of Mitochondrial Proteins and Chronic Fatigue Syndrome.” Translational Psychiatry 6: e904.
https://doi.org/10.1038/tp.2016.184.
DecodeME Consortium, Chris P Ponting, et al. 2025.
“Initial Findings from the DecodeME Genome-Wide Association Study of Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” medRxiv.
https://doi.org/10.1101/2025.08.06.25333109v1.
Giloteaux, Ludovic, Maureen R. Hanson, and Betsy A. Keller. 2016.
“A Pair of Identical Twins Discordant for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome Differ in Physiological Parameters and Gut Microbiome Composition.” American Journal of Case Reports 17: 720–29.
https://doi.org/10.12659/ajcr.900314.
Koelle, David M, Serge Barcy, Meei-Li Huang, Rhoda L Ashley, Lawrence Corey, Judy Zeh, Suzanne Ashton, and Dedra Buchwald. 2002.
“Markers of Viral Infection in Monozygotic Twins Discordant for Chronic Fatigue Syndrome.” Clinical Infectious Diseases 35 (5): 518–25.
https://doi.org/10.1086/341774.
Liu, Yuhan Karida, Devina Persaud, Erica L. Vieira, Joeffre Braga, Pablo Rusjan, Laura Miler, Jennifer S. Rabin, et al. 2026.
“Loss of Vesicular Monoamine Transporter 2 in Striatum of Long COVID and Relationship to Neuropsychiatric Symptoms.” eBioMedicine.
https://doi.org/10.1016/j.ebiom.2026.106339.
Maccallini, P. 2026.
“Biological Insights from Genome-Wide Association Studies and Whole Genome Sequencing of Myalgic Encephalomyelitis/ Chronic Fatigue Syndrome.” Research Square [Preprint], June.
https://doi.org/10.21203/rs.3.rs-9702020/v1.
ME/CFS Science. 2025.
“DecodeME: The Biggest ME/CFS Study Ever.” 2025.
https://mecfsscience.org/decodeme-the-biggest-me-cfs-study-ever/.
Nagy-Szakal, Dorottya, Brent L. Williams, Nischay Mishra, Xiaoyu Che, Bohyun Lee, Lucinda Bateman, Nancy G. Klimas, et al. 2017.
“Fecal Metagenomic Profiles in Subgroups of Patients with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Microbiome 5: 44.
https://doi.org/10.1186/s40168-017-0261-y.
Snyder, Mark P, Hongjuan Zhao, et al. 2025.
“Rare Coding Variants in ME/CFS Implicate Neuronal Genes and Synaptic Function.” medRxiv.
https://doi.org/10.1101/2025.04.15.25325899v2.
Treadway, Michael T, Joshua W Buckholtz, Robin L Cowan, Neil D Woodward, Rui Li, M Salah Ansari, Ronald M Baldwin, Ariel N Schwartzman, Ronald M Kessler, and David H Zald. 2012.
“Dopaminergic Mechanisms of Individual Differences in Human Effort-Based Decision-Making.” Journal of Neuroscience 32 (18): 6170–76.
https://doi.org/10.1523/JNEUROSCI.6459-11.2012.