**4. Discussion**

Mitochondrial dysfunction is the most common type of metabolic disorder and can be caused by either mitochondrial or nuclear gene mutations. Here, we applied proteome and metabolome profiling to reveal the molecular consequences of gene mutations in *NDUFS1* and *MT-ND5*, which respectively encode for the two core subunits in the hydrophilic and transmembrane arms of CI.

#### *4.1. Specific Disassembly of the N-Module and the Entire Respirasome in Mutated NDUFS1*

Our proteome screening indicated a specific loss of the entire N-module for the *NDUFS1* patient (Figure 3C). For validation, we applied BN-PAGE in combination with western blot to reveal a disrupted assembly for CI. An in-gel activity assay showed a missing band only in this patient, which may either derive from a diminished stability of CI or by the lack of FMN in the protein (Figure 5F). The CI N-module consisted of three core subunits, NDUFS1, NDUFV1, and NDUFV2, which were encoded by nuclear genes and accommodate the FMN prosthetic group, as well as the iron–sulfur clusters N1a, N3, N1b, N4, and N5 [55,56]. The N-module in the patient with mutated *NDUFS1* hence disintegrated easily, resulting in strongly reduced amounts of fully assembled CI, causative for to the observed enzymatic dysfunction. The homozygous mutation p.Asp252Gly in NDUFS1 alone was shown to cause the disassembly of CI in a patient with mild cavitating leukoencephalopathy [30]. Recent studies support a model that functional modules of CI are first assembled independently and then gradually form a mature CI, in which the N-module joins in the final step [57–59]. Over the past two decades, several possible formations of structures for respiratory chain supercomplexes have been identified and resolved at high resolution [60–66]. It has been reported that about 80–90% of the CI population is indeed bound to other OXPHOS complexes in stoichiometry to form supercomplexes (SC), which is named the mitochondrial respirasome and is composed of CI, CIII and CIV [67–69]. The formation of SC was severely reduced in the *NDUFS1* patient compared with the controls (Figure 5A–D). This was further confirmed by the detection of large amounts of individual CIII, which could not be assembled into a mature SC in the *NDUFS1* patient (Figure 5D). Whether the formation of respirasomes indeed enhances the e fficiency of electron transfer and minimizes the electron leakage and thus ROS production is still controversially debated [63,64,70,71]. It can be concluded that the N-module disintegrated easily due to the substitutions of two amino acids in the *NDUFS1* patient, which severely a ffected the maturation and structural stability of CI and, hence, the formation of SC.

#### *4.2. Disruption of The Electron Flow in Mutated NDUFS1*

In the patient with *NDUFS1* mutations, the valine at position 228 was changed to alanine. This valine was located between the iron–sulfur cluster N4 and N5 (Figure 4). A previous study modeled the e ffect of this Val232Gly substitution in bacterial CI and showed that the "Y"-shaped side chain of valine is crucial as a bridge for electron transfer between N4 and N5. The replacement of valine to glycine caused a decrease in the electron transfer e fficiency to about one-thousandth in bacterial CI [37]. Our *in silico* modeling of the ovine enzyme with the specific substitution p.Val205Ala again confirmed the dramatically reduced electron transfer, which was expected in the *NDUFS1* patient. The decreased rate of electron transfer between the N4 and N5 clusters should a ffect the overall rate of electron transfer from NADH via FMN and iron–sulfur clusters to Co-Q10 in CI, with an elevated level of FMN in its reduced state and a consequential increase in the level of ROS production by the enzyme [72–75]. Therefore, both defects, the Val228Ala substitution disrupting electron transfer between the N4 and N5 clusters and the partial disintegration of the N-module due to the Asp252Gly substitution [30], might be the cause of the elevated ROS production, indicated by the decreased GSH/GSSG ratio in the patient (Figure 2). Elevated ROS has been reported in cells with CI assembly defects previously [76]. The reported case, carrying only the p.Asp252Gly substitution, presented a very mild phenotype [30], indicating that the additional interruption of the electron flow between N4 and N5 in our case significantly contributed to the severity of the phenotype.

Regarding tunneling calculations, however, several important points should be mentioned. The change of electron transfer coupling between N4 and N5 was quite significant in all enzymes that we have examined. However, given all the uncertainties in the structure, and di fficulties of theoretical modeling of FeS clusters, these results should be regarded only as qualitative trends. It was also recognized that a slower electron transfer rate will only be important if this is the rate-determining step. Furthermore, we cannot exclude that additional water molecules will occupy the mutated site, possibly changing the rates of electron transfer. Generally, it should be recognized that the accurate

quantitative predictions on human enzymes is still a significant challenge. However, the calculated dramatic disruption of electronic coupling reported here, and in the emerging picture, appears to be in agreemen<sup>t</sup> with overall experimental evidence collected in this work on the *NDUFS1* mutant.

#### *4.3. The Stalling of Proton Translocation in Mutated ND5 Is Assumed to Stop Electron Flow Without Any Consequences for Respirasome Formation*

In contrast, the other patient with the *MT-ND5* mutation, harboring a heteroplasmy level of 70%, had a fully assembled CI (Figure 5). The amino acid substitution in the ND5 subunit, which was located on the distal end of the transmembrane arm, thus had no e ffect on the assembly of CI. The crystal structure of CI suggests that a unique, out-of-the-membrane quinone-reaction chamber enables redox energy to drive concerted long-range conformational changes, resulting in the translocation of four protons upon oxidation of one NADH molecule [77–79]. A study in *Escherichia coli* showed that amino acid substitutions close to the proton translocation channel indeed reduced the functionality of CI [80]. Thus, we expect that the proton pumping activity in the *MT-ND5* patient to be impaired in a similar way. In silico modeling of this e ffect by the online tool STRUM, a structure-based prediction of protein stability changes upon single-point mutation [38], indeed showed increased stability for the p.Phe124Leu mutation in subunit ND5. Hence, the stabilizing e ffect of this mutation for the proton channel might hamper its functionality by losing its flexibility. However, the details of the conformational coupling to electron flow remain unknown. It is worth mentioning that the mutation in proton pumping regions hindering electron transfer far away is one of the marvels of CI, and this has been demonstrated by earlier experiments on isolated proteins. One can argue that if the *MT-ND5* mutation indeed results in disrupted conformational coupling and a ffects the electron transfer chain, the immediate consequence of this may be an elevated production of ROS, either by reduced FMN or by a reversed electron transfer mechanism [81–84].

#### *4.4. A Similar Pattern of Regulated Metabolites Was Identified in Both Patients, Mainly for ROS Defense and TCA Cycle Metabolites*

Interestingly, an almost identical set of significantly regulated metabolites was identified in both patients (Figure 1). The shortage of NAD+ may reflect the deficiency of CI, one of the major consumers that oxidizes NADH and generates NAD<sup>+</sup>. We further want to mention that an imbalance of the NADH/NAD<sup>+</sup> ratio in itself may a ffect all aspects of impaired metabolism [85,86]. A stable NADH/NAD+ ratio is critical toward maintaining the homeostasis of metabolic process in both the cytoplasm and mitochondria [85,87]. An increased NADH/NAD+ ratio might thus a ffect the TCA cycle, since previous studies have shown that the inhibition of CI increased succinate oxidation rates [33,88]. The TCA cycle is primarily regulated by product feedback inhibition by NADH and by ADP/ATP and NAD+/NADH ratios [89]. The lack of NAD<sup>+</sup>, required for the conversion of malate to oxaloacetate, resulted in the elevated malate levels. Secondary metabolic alterations caused by CI deficiency were found previously in urine and may play an important role in the pathogenesis of CI deficiency [52]. Fumaric, malic, and also lactic acid, were found to be dramatically increased in some patients' urine, matching our cell culture results [53]. We therefore believe that NADH cannot efficiently transfer electrons to CI, either because of disassembly or because of dramatic disruption of the electron tunneling flow, and thus a jam of unused NADH was created, leading to dysregulation of these TCA cycle metabolites.

The N-module of CI binds and oxidizes NADH and generates two electrons that are transferred through FMN and seven iron–sulfur clusters to ubiquinone in the Q-module [24]. The fundamental role of FMN for the enzymatic functionality of CI has been demonstrated for specific mutations within the FMN docking side carrying subunit NDUFV1. Mutants lacking FMN were fully assembled, but enzymatically inactive [90]. FMN was diminished threefold in both of our cases and previously in rotenone inhibited cells [33], indicating that the electron transfer to quinone was interrupted. It has been reported for both prokaryotes and eukaryotes that the non-covalently bonded prosthetic group

FMN dissociated reversibly from CI when the later one was reduced by NADH and no suitable electron acceptor was available [91,92]. The dissociation of FMN was proposed as a protective mechanism to decrease ROS production [92], as FMN was shown to be a direct site for superoxide radical formation in the CI N-module [73,74,93,94].

Mitochondria are a major source for ROS [81], which can be eliminated by ROS scavengers such as glutathione (GSH) [95]. Apparently, the more than 35-fold decrease in the GSH/GSSG ratio observed in both patients (Figure 2) was a strong indicator of elevated ROS levels because of a stalled electron flow. In addition, N-acetylputerescine, the inactive form of puterescine, was significantly increased in both patients. Puterescine is known to be a main ROS scavenger as well [50,51].

#### *4.5. CI Deficiency Leads to a Glycolytic Phenotype*

The cell respiration assay confirmed that both patients were indeed more glycolytic in their bioenergetics profiles, compared with the two controls (Figure 6B). The oxygen consumption rate (OCR) linked with ATP synthesis showed severe decreases in both patients' fibroblasts (Figure 6D), which confirmed the dysfunction of the mitochondrial OXPHOS caused by CI mutations. This was in agreemen<sup>t</sup> with the clinical data from patient muscle biopsies, in which the CI activities were found to be below the reference range and matched the elevated lactate levels in the fibroblasts and plasma samples of the patients (Table 1). Furthermore, the maximal and spare respiration capacities were significantly lower in both patients (Figure 6E,F). In order to compensate for the mitochondrial shortage of ATP, the patients' cells exhibited an increased rate of glycolysis, as indicated by the glycolysis-contributed proton e fflux rate measurement (glycoPER, Figure 6G).

#### *4.6. Accumulation of Structural Proteins in Patients*

An increase of proteins involved in the cytoskeleton and the extracellular matrix (ECM) was found in both patients, in agreemen<sup>t</sup> with the diagnosis of ventricular hypertrophic cardiomyopathy in the *MT-ND5* patient. CI deficiency frequently resulted in remodeling of the extracellular matrix, causing cardiomyopathy [96–98]. A relationship between a compromised respiratory chain and alterations in structural proteins has been shown previously [99], and OXPHOS deficiencies have also been linked to the development of hypertrophic cardiomyopathy [100]. Thus, our proteome survey indicated that an insu fficient cellular bioenergetic status led to an increase in the ECM and cytoskeletal mass, but further studies are necessary to provide mechanistic links.
