In 1984, we described two patients with "Mitochondrial Myopathy, Encephalopathy, Lactic Acidosis, and Stroke-like Episodes".1 We proposed the acronym MELAS for this newly described distinctive clinical entity, and speculated about maternal non-Mendelian inheritance, and a mitochondrial DNA (mtDNA) mutation disturbing synthesis of proteins embedded in the respiratory chain. Six years later, Goto and colleagues identified a point mutation in mtDNA (m.3243A>G) of MELAS patients.2 This mutation now accounts for 80% of MELAS cases (commonly referred to as MELAS/3243). For two decades, our team has conducted a long-term longitudinal study of MELAS/3243 patients, establishing a strong foundation in terms of natural history, outcome measures, and biomarkers. However, since there are no clearly validated biomarkers for predicting the risk of conversion, prodromal and mildly symptomatic family relatives continue to live with the uncertainty of converting to the severe MELAS phenotype. This reality emphasizes the need to expand our nascent observations about the natural history of MELAS and the predictive value of brain biomarkers. Our strategies are based on (Specific Aim #1) the need to replicate, using 1H MRSI, our highly promising, preliminary observations of abnormal levels of lactate, NAA, tCr and tCho, in 100 mutation carriers and 30 group-matched healthy control subjects, first at baseline and then again at 2-year follow-up;(Specific Aim #2) the need to measure, using 31P MRSI in synchrony with 1H MRSI, brain levels of (a) phosphocreatine (PCr) to complement and corroborate tCr levels, measured by 1H MRSI, as a marker of cell energetics;b) ATP, to complement and corroborate the 1H MRSI measures of NAA and lactate as indices of mitochondrial dysfunction;c) phosphomonoesters (PME) and phosphodiesters (PDE), to complement and corroborate tCho levels, measured by 1H MRSI, as indices of membrane biosynthesis and turnover, and (d) inorganic phosphate (Pi) as an index of intracellular pH;and (Specific Aim #3) the need to measure temporally concordant levels of metabolite biomarkers in the plasma and urine samples collected from all 130 participants. These three Aims will strengthen our earlier findings of predictive neuroimaging biomarkers, inform us of brain mechanisms underlying metabolic and clinical disturbances, and provide complementary plasma and urine metabolites that, if correlated with the brain biomarkers, will serve as less expensive and more accessible biomarkers predicting risk of conversion to the severe MELAS phenotype.