Complex I (NADH:ubiquinone oxidoreductase), a major electron entry point to the mitochondrial respiratory chain, couples electron transfer from NADH to ubiquinone to proton pumping across the mitochondrial inner membrane, and generates the proton motive force that drives ATP synthesis and transport processes. The ~1 MDa mammalian complex contains 45 subunits, and pathological mutations in both its mitochondrial and nuclear encoded subunits result in diverse neuro-muscular disorders. Recent high-resolution mammalian complex I structures have been solved by single-particle cryo-electron microscopy (cryo-EM) in characterised biological states and provide mechanistic insights. However, the molecular bases of genetically-determined complex I dysfunctions remain unclear. Here, two mouse models of complex I-linked mitochondrial disease were analysed structurally by cryo-EM to understand the mechanisms of their pathogenesis. The first part of this thesis explores complex I from the ND6-P25 mouse model, which contains a mitochondrial-DNA point mutation leading to a proline to leucine substitution at position 25 in the ND6 subunit of complex I. The cryo-EM structure of ND6-P25L complex I showed a subtle local structural change resulting in rapid global conversion to a deactive state of the enzyme. Furthermore, the mutant enzyme was unable to catalyse reactive oxygen species production by reverse electron transfer, and the mutant mouse heart is protected against ischemia-reperfusion injury, substantiating a direct link between the two effects. The second part of this thesis describes a structural study of complex I from the ndufs4 knockout mouse model. Although the variant complex I is highly unstable, following sample optimisation its structure was obtained at 2.9 . resolution by cryo-EM. The variant complex I lacking the NDUFS4 subunit is in the active state and, unusually, contains a density resembling ubiquinone in its active site. Absence of NDUFS4 allows motion of the NADH dehydrogenase domain and loss of the NDUFA12 subunit, explaining the instability of the variant complex. Finally, investigations aimed at improving cryo-EM grid preparations for complex I and tackling the problems of limited sample concentration and preferred orientation are described. Grids were modified with graphene, graphene-oxide, polylysine and thiol-PEG; improved numbers of particles could be observed using very low protein concentrations, although with preferred orientation and partial loss of enzyme integrity.