Differential Effects of Yeast NADH Dehydrogenase (Ndi1) Expression on Mitochondrial Function and Inclusion Formation in a Cell Culture Model of Sporadic Parkinson’s Disease

Parkinson’s disease (PD) is a neurodegenerative disorder that exhibits aberrant protein aggregation and mitochondrial dysfunction. Ndi1, the yeast mitochondrial NADH dehydrogenase (complex I) enzyme, is a single subunit, internal matrix-facing protein. Previous studies have shown that Ndi1 expression leads to improved mitochondrial function in models of complex I-mediated mitochondrial dysfunction. The trans-mitochondrial cybrid cell model of PD was created by fusing mitochondrial DNA-depleted SH-SY5Y cells with platelets from a sporadic PD patient. PD cybrid cells reproduce the mitochondrial dysfunction observed in a patient’s brain and periphery and form intracellular, cybrid Lewy bodies comparable to Lewy bodies in PD brain. To improve mitochondrial function and alter the formation of protein aggregates, Ndi1 was expressed in PD cybrid cells and parent SH-SY5Y cells. We observed a dramatic increase in mitochondrial respiration, increased mitochondrial gene expression, and increased PGC-1α gene expression in PD cybrid cells expressing Ndi1. Total cellular aggregated protein content was decreased but Ndi1 expression was insufficient to prevent cybrid Lewy body formation. Ndi1 expression leads to improved mitochondrial function and biogenesis signaling, both processes that could improve neuron survival during disease. However, other aspects of PD pathology such as cybrid Lewy body formation were not reduced. Consequently, resolution of mitochondrial dysfunction alone may not be sufficient to overcome other aspects of PD-related cellular pathology.


Introduction
Parkinson's disease (PD) is a neurodegenerative movement disorder characterized clinically by resting tremor, bradykinesia, rigidity, and postural instability [1]. Motor symptoms of PD correlate with the progressive loss of axonal terminals in the striatum and degeneration of dopaminergic neurons in the substantia nigra pars compacta [2]. PD patients also experience non-motor symptoms

Generation of PD Cybrids and Cell Culture
As described by Swerdlow et al. [12], rho0 SH-SY5Y cells were fused with platelets isolated from aseptically-drawn PD patient blood samples by brief incubation with polyethylene glycol (J.T. Baker, Thermo Fisher Scientific, Waltham, MA, USA). Residual rho0 cells and cybrid cells with insufficient platelet mitochondria repopulation were eliminated by growth in cybrid selection medium containing high glucose Dulbecco's Modified Eagle Medium (DMEM, Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA), dialyzed fetal calf serum (Hyclone/Thermo Fisher Scientific, Waltham, MA, USA), antibiotic/antimycotic, and lacking pyruvate and uridine for 5-6 weeks [12,50]. Cybrid cell lines that tested negative for mycoplasma were aliquoted and stored in liquid nitrogen.
All cell lines were cultured in T75 cm 2 flasks (Greiner Bio-One, Monroe, NC, USA) with growth medium consisting of DMEM supplemented with 10% fetal bovine serum (FBS, Hyclone/Thermo Fisher, Waltham, MA, USA), pyruvate/uridine, and antibiotic/antimycotic as previously described [60]. Ndi1 selection media was made with glucose-free DMEM, supplemented with 10% FBS, pyruvate/uridine, antibiotic/antimycotic, 5 mM galactose, and 30 nM rotenone [60]. Rotenone was prepared in advance as a stock and added directly to the media during the selection period. Ndi1-transfected cell lines were provided by Yagi Lab at passage 4-5. Cells used for the studies reported here were at passage 12-20. Ndi1-expressing cell lines were maintained in growth medium but were returned to selection media every 5 days for 48 h to maintain a stable Ndi1 expression level. To minimize the chance of mycoplasma infection, all cell lines were discarded after 2 months and replaced by fresh cells from frozen aliquots.

Viral Transfection
Gene delivery by adeno-associated virus (AAV) is safe, effective and elicits long-lasting expression [44,45]. Cells were transfected with the rAAV-NDI1 (serotype 2), as previously described [33]. Briefly, PD61 and SH-SY5Y cells were plated for 48 h in 6-well plates before addition of the virus. Viral transfection used 5-8 × 10 12 virus particles/mL of growth medium to ensure efficient transfection [37]. Cells were incubated with the virus for 5 days, then returned to regular growth medium and allowed to proliferate. Transfected cell populations were then exposed to selection media for at least 2 weeks. Immunocytochemistry with a Ndi1-specific antibody was used to verify that the selection process was complete [26]. Ndi1-expressing cell lines were removed from selection media before preparation of frozen stocks.

Immunoblots
Cells were grown in T175 cm 2 flasks (Greiner Bio-One, Monroe, NC, USA) to 80-90% confluency and harvested in 1X radioimmunoprecipitation assay buffer with protease inhibitors and phenylmethanesulfonylfluoride as previously described [59]. The soluble fraction was isolated by centrifugation. Protein quantity was measured using the Micro BCA kit (Pierce/Thermo Fisher Scientific, Waltham, MA, USA). Equal amounts of protein were loaded on to Bis-Tris gels (Bio-Rad, Hercules, CA, USA) and run using the Bio-Rad Criterion system. Proteins were transferred onto nitrocellulose membranes using the iBlot transfer system (Thermo Fisher Scientific, Waltham, MA, USA). Membranes were blocked using Li-Cor blocking buffer [26] and stained with primary antibodies to Ndi1 (1:2000, provided by T. Yagi), MitoProfile Total OXPHOS cocktail, (1:400, Mitosciences/abcam, Cambridge, MA, USA) or αSYN (sc-7011R, 1:100, Santa Cruz Biotechnology, Dallas, TX, USA) at room temperature. Membranes were then washed and stained at room temperature with appropriate secondary antibodies labeled with an infrared dye (1:4000, Li-Cor, Lincoln, NE, USA). Membranes were imaged using the Odyssey scanner (Li-Cor, Lincoln, NE, USA). Band densities were calculated as integrated intensities using the Odyssey software. Integrated intensities were normalized to porin (MSA03, 1:2000, Mitosciences/Thermo Fisher Scientific, Waltham, MA, USA) or actin (A2103, 1:2000, Millipore-Sigma, Burlington, MA, USA). Student's t-test was used to compare normalized integrated intensities between non-transfected and Ndi1 transfected cells, with Welch's correction for unequal variances when necessary (Prism, GraphPad, San Diego, CA, USA).

Quantitative Real-Time PCR (qRT-PCR)
Pellets of approximately 10 × 10 6 cells were collected from sub-confluent T175 cm 2 flasks. Total RNA and DNA were isolated using an RNA/DNA isolation kit (Qiagen, Germantown, MD, USA) and measured using a NanoDrop (Thermo Fisher). From RNA, complementary DNA (cDNA) was made using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). Quantitative real-time PCR was run for single genes (EvaGreen Power mix, Bio-Rad, Hercules, CA, USA) or in a multiplex set (iQ Multiplex Power mix, Bio-Rad) in a CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). Bio-Rad CFX Manager software calculated starting quantities for samples, based on a standard curve with known starting quantities. All values were normalized to endogenous reference genes based on geNorm (BioGazelle, Technologiepark 3 Zwijnaarde, 9052, Belgium) analysis to find the genes with the highest expression stability. For gene expression, we used the endogenous reference genes for glyceraldehyde 3-phosphate dehydrogenase and 14-3-3-zeta. We used 14-3-3-zeta and beta-2-microglobulin as the endogenous reference genes for DNA copy number studies. Student's t-tests were used to analyze data (Prism, GraphPad, San Diego, CA, USA).

Cellular Respiration (OXPHOS)
Cellular respiration was measured using the Seahorse Extracellular Flux Analyzer XF24 (Seahorse Bioscience/Agilent, Santa Clara, CA, USA) as previously described [60]. Cells were grown to a confluent monolayer in XF24 culture plates for 24 h prior to the assay. Mitochondrial inhibitors used were oligomycin (1 µM) to inhibit ATP synthase, carbonyl cyanide 4-(trifluoromethyoxy) phenhydrazone (FCCP, 300 nM) to dissipate the inner mitochondrial membrane proton gradient, rotenone (100 nM) to inhibit complex I, and antimycin A (10 µM) to inhibit complex III, all at a pH of 7.4. All measurements were acquired after a 3-min mix and 2-min waiting period to re-oxygenate the media. Oxygen concentration was then measured in sequential 3-min windows in order to calculate an oxygen consumption rate (OCR). Prism software was used to determine basal, maximum capacity and non-mitochondrial OCR as well as to calculate leak, ATP-linked OCR and spare capacity [61]. The basal extracellular acidification rate (ECAR) was also determined. Following each experiment, individual OCR and ECAR values were normalized to protein content (Micro BCA Kit, Pierce/Thermo Fisher Scientific, Waltham, MA, USA). The contribution of Ndi1 expression to the OCR measurements for each cell line was measured by inhibiting endogenous complex I (30 nM rotenone in the running media starting 1 h prior to initiating the XF24 experiment-Pre-RTN). Statistics were calculated using two-way ANOVA with Bonferroni multiple comparison post-hoc tests. (Prism, GraphPad, San Diego, CA, USA).

Aggregated Protein and CLB Measurements
Ndi1 transfected and non-transfected cell lines were stained with 100 mM Congo red for 24 h to visualize beta-pleated sheet aggregated proteins and CLB as previously described [60]. The density of Congo red aggregates (Congo red positive pixels) was quantified using MetaMorph image analysis software (Molecular Devices, San Jose, CA, USA) and then normalized to cell number per each image. CLB, defined as spherical Congo red positive inclusions over 2 µm in diameter, were visualized using a FV1000 laser scanning confocal microscope, counted and measured using Fluoview software (Olympus America, Center Valley, PA, USA). Experiments were repeated with cells from a different passage. Outcomes from transfected and non-transfected cell lines were compared using Student's t-test (Prism, Graph Pad, San Diego, CA, USA).

Mitochondrial Movement
Transfected and non-transfected cell lines were differentiated in 35mm polyethyleneimine-coated dishes as previously described [60,62]. Undifferentiated cells were grown for 24-48 h in regular growth media before differentiation. Differentiation media consisted of Neurobasal medium (Thermo Fisher Scientific, Waltham, MA, USA), B27 supplement (Thermo Fisher Scientific, Waltham, MA, USA), antibiotic/antimycotic, pyruvate/uridine and freshly diluted staurosporine (4-6 nM, Millipore-Sigma, Burlington, MA, USA) as previously described [60]. Transfected and non-transfected cell lines were maintained in differentiation media for 12 days. To image mitochondrial movement in differentiated neurites, transfected and non-transfected neuronal cells were incubated with MitoTracker CMXRos (MitoTrackerRed, Molecular Probes, Thermo Fisher Scientific, Waltham, MA, USA) as previously described [60,63]. Time-lapse recordings of fluorescently labeled mitochondria were collected using MetaMorph Imaging software (Molecular Devices, San Jose, CA, USA) every 3 s for 2 min using an Olympus IX70 inverted microscope equipped with epifluorescence and Nomarski optics, a Lambda 10-2 filter wheel, a Photometrics CoolSnap HQ progressive scan CCD camera, and a heater/controller to maintain cybrid cells at 37 • C during image collection (World Precision Instruments, Inc., Sarasota, FL, USA). All the mitochondria were individually tracked in each movie frame in randomly selected processes (PD61 n = 3 cultures, 51 processes; PD61 Ndi1 n = 4 cultures, 49 processes; SH-SY5Y n = 3 cultures, 53 processes; SH-SY5Y Ndi1 n = 3 cultures, 49 processes). Mitochondrial velocities were calculated using MetaMorph Imaging Software and analyzed using the Students t-test.

Ndi1 Expression in PD61 Cybrid and SH-SY5Y Cell Lines
Successful transfection and selection of PD61 and SH-SY5Y cell lines for expression of rAAV-NDI1 (PD61 Ndi1 and SH-SY5Y Ndi1 ) was established using immunocytochemistry, qRT-PCR and immunoblots ( Figure 1). Levels of Ndi1 protein as well as gene expression were undetectable in non-transfected PD61 and SH-SY5Y cell lines (data not shown). The incorporation of Ndi1 protein into the mitochondria was confirmed by co-localization of Ndi1 immunostaining with immunostaining for complex Vα ( Figure 1A,B). Ndi1 protein level in SH-SY5Y Ndi1 cells was significantly increased ( Figure 1C,D). In contrast, Ndi1 protein expression levels were six-fold higher in PD61 Ndi1 than in SH-SY5Y Ndi1 ( Figure 1C). Similarly, using qRT-PCR, NDI1 gene expression was three-fold higher in PD61 Ndi1 than in SH-SY5Y Ndi1 ( Figure 1D).  Figure 2A shows a typical bioenergetics profile for PD61 and PD61 Ndi1 . Basal OCR (net sum of all oxygen consuming processes at baseline) in PD61 was low (2,082 pmolO2/min/mg total protein) while basal OCR of SH-SY5Y cells was 20% higher at 11,796 pmolO2/min/mg total protein (compare Figures 2C and 3C). Values for ATP-linked, leak, maximum, spare capacity, and non-mitochondrial OCR were also minimal in PD61 ( Figure 2C). The presence of Ndi1 protein in PD61 had a robust effect on respiration, resulting in a 10-fold increase in basal OCR levels (Figures 2A and 2C). As shown in Figure 2C, PD61 Ndi1 also exhibited a considerable and significant increase in ATP-linked, maximum capacity and non-mitochondrial OCR (not shown) as well as spare capacity and leak. OCR was also measured for four control cybrid cell lines generated from disease-free, age-appropriate donor platelets. The basal OCR for PD61 Ndi1 was 2.1X higher than the mean basal OCR for the four control (CNT) lines (21,955 pmolO2/min/mg total protein for PD61 Ndi1 as compared to 10,311 pmolO2/min/mg total protein for CNT, p < 0.001, 1-way ANOVA). Similarly, the max capacity OCR in PD61 Ndi1 also Expression of Ndi1 was greater in PD61 Ndi1 than in SH-SY5Y Ndi1 based on immunofluorescence, Western blot analysis and qRT-PCR. (A,B) Mitochondria in PD61 Ndi1 and SH-SY5Y Ndi1 cells stained with Ndi1 (green), complex Vα (red) and nuclei labeled with DAPI (blue). Box outline designates an enlarged area of merged image (high mag). Scale bars = 10 µm; high mag scale bar = 2 µm. (C) Western blot visualization of Ndi1 protein levels in SH-SY5Y Ndi1 and PD61 Ndi1 . Values shown are integrated intensities normalized to actin. (D) Quantitative real-time (qRT-PCR) analysis of NDI1 gene expression levels in PD61 Ndi1 and SH-SY5Y Ndi1 and shown as fold change from SH-SY5Y Ndi1 levels. There was an approximately three-fold increase in PD61 Ndi1 . Students t-test, n = 3, # p < 0.005.

Effects of Ndi1 Expression on Oxygen Consumption Rates (OCR) and Extracellular Acidification Rates (ECAR)
Oxygen utilization in non-transfected and Ndi1-expressing cell lines was assessed using the Seahorse XF24 (Seahorse Bioscience/Agilent) [61,64]. The OCR and ECAR (a surrogate measure of glycolysis) were calculated using XF24 Analyzer software (see Methods). Sequential injection of inhibitors permits the measurement of specific components of the ETC such as basal, ATP-linked and maximum capacity, non-mitochondrial OCR, and calculation of spare capacity and leak as shown in Figures 2A and 3A (see [60,61]). ATP-linked OCR was determined after the addition of oligomycin to inhibit ATP synthesis (Figures 2A and 3A). The difference between basal and oligomycin OCR is the oxygen utilization that is ATP-linked. Leak is the OCR remaining after oligomycin (adjusted for non-mitochondrial oxygen utilization, see Figures 2A and 3A). Maximum OCR is determined after the addition of FCCP (Figures 2A and 3A) and spare or reserve capacity is calculated as the difference between maximum and basal OCR.  Basal ECAR was higher in PD61 compared to SH-SY5Y (compare Figures 2B and 3B). This is consistent with other studies showing that cells with compromised oxygen utilization maintain ATP levels by increased glycolysis [65]. PD61 Ndi1 exhibited a significant increase in basal OCR that was for SH-SY5Y Ndi1 . Previous studies using transfected PC12 cells showed that Ndi1 expression did not adversely affect the morphology of neuronal cells that were already differentiated [37].   Figure 2A shows a typical bioenergetics profile for PD61 and PD61 Ndi1 . Basal OCR (net sum of all oxygen consuming processes at baseline) in PD61 was low (2082 pmolO 2 /min/mg total protein) while basal OCR of SH-SY5Y cells was 20% higher at 11,796 pmolO 2 /min/mg total protein (compare Figures 2C and 3C). Values for ATP-linked, leak, maximum, spare capacity, and non-mitochondrial OCR were also minimal in PD61 ( Figure 2C). The presence of Ndi1 protein in PD61 had a robust effect on respiration, resulting in a 10-fold increase in basal OCR levels (Figure 2A,C). As shown in Figure 2C, PD61 Ndi1 also exhibited a considerable and significant increase in ATP-linked, maximum capacity and non-mitochondrial OCR (not shown) as well as spare capacity and leak. OCR was also measured for four control cybrid cell lines generated from disease-free, age-appropriate donor platelets. The basal OCR for PD61 Ndi1 was 2.1X higher than the mean basal OCR for the four control (CNT) lines (21,955 pmolO 2 /min/mg total protein for PD61 Ndi1 as compared to 10,311 pmolO 2 /min/mg total protein for CNT, p < 0.001, 1-way ANOVA). Similarly, the max capacity OCR in PD61 Ndi1 also exceeded the max capacity for the four control lines by 2.4X (30,768 vs. 13,028 pmolO 2 /min/mg total protein, p < 0.0001, 1-way ANOVA).

PD61/PD61 Ndi1
Basal ECAR was higher in PD61 compared to SH-SY5Y (compare Figures 2B and 3B). This is consistent with other studies showing that cells with compromised oxygen utilization maintain ATP levels by increased glycolysis [65]. PD61 Ndi1 exhibited a significant increase in basal OCR that was accompanied by a significant 50% decrease in ECAR in PD61 Ndi1 ( Figure 2B) to become more reliant on OXPHOS than on glycolysis.
To confirm that Ndi1 was functionally contributing to the changes seen in OXPHOS in PD61 Ndi1 ( Figure 2E), we measured OCR in the presence of 30 nM rotenone (rotenone pretreatment or Pre-RTN, Figure 2E). Under these conditions, rotenone will inhibit endogenous complex I but will not affect the activity of rotenone-resistant Ndi1. In spite of the poor oxygen utilization in PD61, pretreatment with rotenone significantly decreased endogenous basal, ATP-linked, maximum, and spare capacity OCR in PD61 ( Figure 2D). Rotenone pretreatment of PD61 Ndi1 ( Figure 2E) significantly reduced but did not eliminate oxygen utilization. Since Ndi1 is not inhibited by rotenone, the respiration remaining after pretreatment with rotenone is due to the expression of Ndi1 in PD61 mitochondria. The reductions in PD61 Ndi1 in basal, leak, maximum and spare capacity OCR were all significant when pre-treated with rotenone ( Figure 2E). However, levels of basal, leak, maximum, and spare capacity OCR remained high even after rotenone pre-treatment ( Figure 2E). Figure 3A shows the bioenergetic profile of SH-SY5Y and SH-SY5Y Ndi1 . Ndi1 expression in SH-SY5Y cells had no significant effect on basal, ATP-linked or leak OCR ( Figure 3C). However maximum and spare capacity OCR levels were significantly increased in SH-SY5Y Ndi1 ( Figure 3C). Even though basal OCR levels were unchanged in SH-SY5Y Ndi1 , there was a significant 55% decrease in ECAR in SH-SY5Y Ndi1 compared to SH-SY5Y ( Figure 3B) indicating a shift toward greater reliance on OXPHOS rather than glycolysis.

SH-SY5Y/SH-SY5Y Ndi1
Pretreatment with rotenone ( Figure 3D) to eliminate endogenous complex I activity resulted in significant reductions in basal, ATP-linked and maximum capacity OCR in non-transfected SH-SY5Y cells that virtually eliminated oxygen utilization. In comparison, rotenone pre-treatment in SH-SY5Y Ndi1 ( Figure 2E) had no effect on basal, ATP-linked or leak OCR. Rotenone pretreatment of SH-SY5Y Ndi1 resulted in a significant decline only in maximum and spare capacity OCR.

Effects of Ndi1 Expression on Mitochondrial Movement
Previous studies have shown reduced velocities of mitochondrial movement in PD cybrid neurites compared to SH-SY5Y [60,63]. Mitochondrial movement was measured in Ndi1-expressing and non-transfected, differentiated PD61 and SH-SY5Y neurites. The velocities of all the mitochondria in each neurite were averaged and then graphed to show the range of mitochondrial velocities ( Figure 4A,B). The average velocity of mitochondria in all PD61 Ndi1 neurites was not significantly different from the average mitochondrial velocity in PD61. However, a histogram analysis to visualize average mitochondrial velocity per neurite revealed that Ndi1 expression resulted in a shift from slow (0.04 to 0.20 microns/second) to more rapid mitochondrial movement (0.21 to 0.44 microns/second) in PD61 Ndi1 ( Figure 4B). Furthermore, the percentage of mitochondria classified as "not moving" (velocity ≤ 0.05 microns/second) was significantly reduced in (p = 0.03, matched t-test) in PD61 Ndi1 neurites compared to PD61 neurites (5.7% ± 1.7 SEM for PD61 Ndi1 vs. 13.2% ± 2.2 SEM for PD61). Also, the percentage of mitochondria that moved for less than 9 consecutive seconds was significantly reduced (p = 0.025 matched t-test) in PD61 Ndi1 (23.3% ± 5.3 SEM) compared to PD61 (36.8% ± 3.5 SEM).

Effects of Ndi1 Expression on Endogenous Mitochondrial Gene Expression
We measured gene expression and copy number of four genes encoded by mtDNA-ND2 and ND4 (complex 1 subunits), the complex IV subunit (cytochrome oxidase) COX3, and the 12s ribosomal RNA (rRNA). The copy number of all four genes in PD61 was significantly below the SH-SY5Y level (Dashed line at 1.0 in Figure 5A). Mitochondrial gene expression for ND2, ND4, and COX3 Ndi1 Mean mitochondrial velocities for SH-SY5Y and SH-SY5Y Ndi1 per neurite were not significantly different based on a similar histogram analysis as described above. There were no significant changes in the percentage of non-moving mitochondria or in the percentage of mitochondria that moved for less than nine consecutive seconds in SH-SY5Y vs. SH-SY5Y Ndi1 (data not shown).
PD61 and PD61 Ndi1 cells successfully differentiated into neurons with multiple long processes that resemble the axons (long, unbranched, narrow caliber) and dendrites (tapering and branched) similar to those formed by SH-SY5Y neurons (not shown). In contrast, many SH-SY5Y Ndi1 cells produced shorter processes in response to the low dose of staurosporine differentiation. Further study is needed to determine if the altered appearance of neurites in SH-SY5Y Ndi1 is a consequence of Ndi1 expression in this neuroblastoma cell line or if the differentiation protocol requires modification for SH-SY5Y Ndi1 . Previous studies using transfected PC12 cells showed that Ndi1 expression did not adversely affect the morphology of neuronal cells that were already differentiated [37].

Effects of Ndi1 Expression on Endogenous Mitochondrial Gene Expression
We measured gene expression and copy number of four genes encoded by mtDNA-ND2 and ND4 (complex 1 subunits), the complex IV subunit (cytochrome oxidase) COX3, and the 12s ribosomal RNA (rRNA). The copy number of all four genes in PD61 was significantly below the SH-SY5Y level (Dashed line at 1.0 in Figure 5A). Mitochondrial gene expression for ND2, ND4, and COX3 but not 12s rRNA was significantly increased in PD61 Ndi1 compared to PD61 ( Figure 5A). In fact, Ndi1 bypass of complex I dysfunction in PD61 Ndi1 resulted in mtDNA gene copy numbers that were significantly increased to levels above SH-SY5Y gene expression levels (Dashed line at 1.0 in Figure 5C,D). Gene expression for ND2, ND4, COX3 and 12s rRNA was also increased in SH-SY5Y Ndi1 , however, these changes did not achieve significance ( Figure 5B). mtDNA copy number also did not change significantly in SH-SY5Y Ndi1 compared to SH-SY5Y ( Figure 5D).

Effects of Ndi1 Expression on Mitochondrial Biogenesis
Peroxisome proliferator-activator receptor gamma co-activator 1-alpha (PCG-1α) is a transcription factor that serves as the master regulator for mitochondrial biogenesis [66]. In both cell lines expressing Ndi1, we observed a significant increase in PGC-1α expression compared to the nontransfected cell lines ( Figure 6A). PD61 Ndi1 had the highest relative expression of PGC-1α, which was not surprising given the robust changes in mitochondrial function and mtDNA gene expression. Nuclear respiratory factor 1 (NRF1) is a DNA binding protein that regulates nuclear encoded ETC subunits as well as mitochondrial transcription factor B1 and mitochondrial transcription factor A (TFAM) which both bind to mtDNA, initiate mitochondrial-encoded gene transcription, and also participate in mitochondrial biogenesis [67][68][69]. Neither PD61 Ndi1 or SH-SY5Y Ndi1 exhibited significant changes in these mitochondrial transcription factors (not shown).

Effects of Ndi1 Expression on Mitochondrial Biogenesis
Peroxisome proliferator-activator receptor gamma co-activator 1-alpha (PCG-1α) is a transcription factor that serves as the master regulator for mitochondrial biogenesis [66]. In both cell lines expressing Ndi1, we observed a significant increase in PGC-1α expression compared to the non-transfected cell lines ( Figure 6A). PD61 Ndi1 had the highest relative expression of PGC-1α, which was not surprising given the robust changes in mitochondrial function and mtDNA gene expression. Nuclear respiratory factor 1 (NRF1) is a DNA binding protein that regulates nuclear encoded ETC subunits as well as mitochondrial transcription factor B1 and mitochondrial transcription factor A (TFAM) which both bind to mtDNA, initiate mitochondrial-encoded gene transcription, and also participate in mitochondrial biogenesis [67][68][69]. Neither PD61 Ndi1 or SH-SY5Y Ndi1 exhibited significant changes in these mitochondrial transcription factors (not shown).

Effects of Ndi1 Expression on ETC Assembly
In light of the increases in mitochondrial gene expression and biogenesis signaling, we also measured the assembly of the individual ETC complexes using an antibody cocktail (MitoProfile Total OXPHOS human WB antibody cocktail, abcam, Cambridge, MA, USA) that recognizes subunits that are labile when each ETC complex is incorrectly assembled [70]. The assembly of complex III and complex V in PD61 Ndi1 was significantly increased compared to PD61 ( Figure 6C). Levels of assembled complexes I and IV in PD61 Ndi1 were not significantly different from PD61 ( Figure 6C) despite the observation that mtDNA gene expression of subunits for complexes I and IV was significantly increased ( Figure 5A). In SH-SY5Y Ndi1 , levels of assembled complexes I, II, III, and IV were all increased significantly compared to SH-SY5Y ( Figure 6D). In comparison, mtDNA gene expression of complex I and IV subunits were not significantly changed by Ndi1 expression in SH-SY5Y ( Figure  5B).

Effects of Ndi1 Expression on ETC Assembly
In light of the increases in mitochondrial gene expression and biogenesis signaling, we also measured the assembly of the individual ETC complexes using an antibody cocktail (MitoProfile Total OXPHOS human WB antibody cocktail, abcam, Cambridge, MA, USA) that recognizes subunits that are labile when each ETC complex is incorrectly assembled [70]. The assembly of complex III and complex V in PD61 Ndi1 was significantly increased compared to PD61 ( Figure 6C). Levels of assembled complexes I and IV in PD61 Ndi1 were not significantly different from PD61 ( Figure 6C) despite the observation that mtDNA gene expression of subunits for complexes I and IV was significantly increased ( Figure 5A). In SH-SY5Y Ndi1 , levels of assembled complexes I, II, III, and IV were all increased significantly compared to SH-SY5Y ( Figure 6D). In comparison, mtDNA gene expression of complex I and IV subunits were not significantly changed by Ndi1 expression in SH-SY5Y ( Figure 5B).

Effects on Ndi1 Expression on αSYN
We measured αSYN protein levels using immunoblots because αSYN has been shown to permeabilize mitochondrial membranes and to play a role in the formation of LB [71][72][73]. There was no detectable difference in the quantity of soluble αSYN in Ndi1 expressing cell lines compared to non-transfected lines (data not shown). We also assessed αSYN mRNA levels using qRT-PCR. As shown in Figure 7B, αSYN gene expression was increased almost two-fold in PD61 Ndi1 versus PD61 but this change did not achieve statistical significance (p = 0.055). There was also no significant difference in αSYN gene expression between SH-SY5Y Ndi1 and SH-SY5Y ( Figure 6B).
shown in Figure 7B, αSYN gene expression was increased almost two-fold in PD61 Ndi1 versus PD61 but this change did not achieve statistical significance (p = 0.055). There was also no significant difference in αSYN gene expression between SH-SY5Y Ndi1 and SH-SY5Y ( Figure 6B).

Effects of Ndi1 Expression on Levels of Protein Aggregates and CLB
We determined the total cellular level of aggregated proteins, as well as the frequency of CLB expression (see Methods) using live-cell staining with Congo red, a histochemical dye that binds to proteins containing beta-pleated sheet configurations as well as to the N-terminal aggregation prone region of αSYN [60,74,75]. In PD61 Ndi1 , total cellular aggregated protein content (Congo red positive pixels/cell) was significantly decreased, compared with PD61 ( Figure 7A). A similar change was observed in SH-SY5Y Ndi1 compared to SH-SY5Y ( Figure 7A).
Expression levels of Congo red-labeled CLB (round inclusions greater than 2μm in diameter) were also unchanged in PD61 Ndi1 (1-2% in both PD61 and PD61 Ndi1 , Figure 8B) despite a significant improvement in OXPHOS function and a reduction in the levels of Congo red positive protein aggregates. There was also no change in size of CLB in PD61 Ndi1 compared to PD61 (4.336+/− 0.5902 for PD61 vs. 4.217 +/− 0.938 for PD61 Ndi1 ). As shown in Figure 8C, D, CLB continued to be expressed in transfected cells with robust Ndi1 expression. Furthermore, Ndi1 immunostaining was heterogeneously incorporated into CLB. In some cases, CLB contained Ndi1 immunostaining that colocalized with staining for the mitochondrial outer membrane protein porin ( Figure 8D). counted and normalized to cell number as a measure of cellular aggregated protein content. Both SH-SY5Y Ndi1 and PD61 Ndi1 showed a decrease in total aggregated protein content after Ndi1 expression. Student's t-test, n = 3. * p < 0.05, ** p < 0.01 (B) qRT-PCR for αSYN showed no significant change in αSYN expression after Ndi1 expression in SH-SY5Y Ndi1 . PD61 Ndi1 shows a trend towards increased αSYN gene expression (p = 0.055). Student's t-test, n = 3.

Effects of Ndi1 Expression on Levels of Protein Aggregates and CLB
We determined the total cellular level of aggregated proteins, as well as the frequency of CLB expression (see Methods) using live-cell staining with Congo red, a histochemical dye that binds to proteins containing beta-pleated sheet configurations as well as to the N-terminal aggregation prone region of αSYN [60,74,75]. In PD61 Ndi1 , total cellular aggregated protein content (Congo red positive pixels/cell) was significantly decreased, compared with PD61 ( Figure 7A). A similar change was observed in SH-SY5Y Ndi1 compared to SH-SY5Y ( Figure 7A).
Expression levels of Congo red-labeled CLB (round inclusions greater than 2µm in diameter) were also unchanged in PD61 Ndi1 (1-2% in both PD61 and PD61 Ndi1 , Figure 8B) despite a significant improvement in OXPHOS function and a reduction in the levels of Congo red positive protein aggregates. There was also no change in size of CLB in PD61 Ndi1 compared to PD61 (4.336 ± 0.5902 for PD61 vs. 4.217 ± 0.938 for PD61 Ndi1 ). As shown in Figure 8C,D, CLB continued to be expressed in transfected cells with robust Ndi1 expression. Furthermore, Ndi1 immunostaining was heterogeneously incorporated into CLB. In some cases, CLB contained Ndi1 immunostaining that co-localized with staining for the mitochondrial outer membrane protein porin ( Figure 8D).

Discussion
The yeast mitochondrial NADH dehydrogenase, Ndi1, has been widely used to bypass complex I dysfunction in vitro and in vivo. We expanded on existing literature to investigate Ndi1 expression in a human cell culture model of sporadic PD that exhibits critical aspects of PD pathology including compromised mitochondrial oxygen utilization, the expression of aggregated proteins, and Lewy body-like CLB. Similar to previous studies, we found that trans-species expression of a nuclearencoded yeast gene NDI1 in our human, cell culture model of sporadic PD compensated for complex I dysfunction, restored NADH oxidase activity, and enabled more efficient operation of the mitochondrial ETC (e.g., References [44,49,76]). While the presence of Ndi1 in PD61 resulted in

Discussion
The yeast mitochondrial NADH dehydrogenase, Ndi1, has been widely used to bypass complex I dysfunction in vitro and in vivo. We expanded on existing literature to investigate Ndi1 expression in a human cell culture model of sporadic PD that exhibits critical aspects of PD pathology including compromised mitochondrial oxygen utilization, the expression of aggregated proteins, and Lewy body-like CLB. Similar to previous studies, we found that trans-species expression of a nuclear-encoded yeast gene NDI1 in our human, cell culture model of sporadic PD compensated for complex I dysfunction, restored NADH oxidase activity, and enabled more efficient operation of the mitochondrial ETC (e.g., References [44,49,76]). While the presence of Ndi1 in PD61 resulted in reduced levels of Congo red-positive aggregated proteins, CLB expression levels and αSYN gene expression were not significantly changed. This suggests that bypass of complex I dysfunction by Ndi1 expression in a model of sporadic PD was sufficient to improve mitochondrial function but was not sufficient to reduce other cellular dysfunctions relevant to the formation and maintenance of CLB.
One of the compelling reasons to explore an alternate NADH oxidase like Ndi1 for PD is that it can overcome OXPHOS dysfunction without regard to the cause of the complex I dysfunction. For example, NDI1 gene expression has been successfully used in a range of models from Caenorhabditis elegans to vertebrates [77,78], primary and tumor cell cultures where complex I dysfunction is due to neurotoxicity [17,36,38,44,46], nuclear gene mutation [41,49], or mitochondrial gene mutation [32,40]. Whether functional compromise is the result of complex I inhibition or altered structure, NDI1 gene expression improved mitochondrial function. Furthermore, Ndi1, like other alternate oxidases from lower organisms, appears to be inactive when cells function normally and only becomes active under conditions when the ETC is dysfunctional [79]. Rustin, Jacobs and the Alternatives Consortium [79] suggest that alternative respiratory enzymes like Ndi1 confer "significant flexibility" to the ETC allowing it to "overcome potential constraints".

Cellular Consequences of Ndi1 Expression in PD61
We selected PD61 as a test case because the complex I activity and ETC dysfunction were comparable to PD brain [59,80]. Platelets for the creation of the cybrid came from a 65-year-old male with a Hoehn and Yahr score of 2 and disease duration of 15 years. The donor was haplogroup L2e1a (sub-Saharan) with diagnostic mutations as well as six mutations that were unique to the individual relative to the revised Cambridge Reference sequence [58]. Examination of the substantial changes in cellular functions induced by Ndi1 expression including improved oxygen utilization, axonal transport of mitochondria, mitochondrial gene expression, and copy number as well as increased mitochondrial biogenesis signaling by PD61 Ndi1 could provide insights for future PD therapy development.
Previous studies have shown that oxygen utilization in PD61 is crippled by ETC damage [59,63]. Our measurements of oxygen utilization using the XF24 suggest that increased NADH oxidation due to Ndi1 expression was enough to robustly alter the respiratory profile of PD61 Ndi1 . Although we expected some functional improvements, we were surprised by the magnitude of the respiratory improvements in PD61 Ndi1 . Immunoblot analysis confirmed that porin expression was unchanged by the presence of Ndi1 ( Figure 6B) so this change was not due to increased mitochondrial mass. The increased basal, ATP-linked, maximum, and spare capacity OCR in PD61 Ndi1 is consistent with a significant increase in the volume of electrons being transferred along the ETC.
The ATP-linked OCR of PD61 Ndi1 was significantly improved compared to PD61 but a greater increase was seen in maximum capacity. Desler et al. [81] suggested that complex IV, the final electron acceptor of the ETC, balances the activity of the ETC in response to cell needs. In support of this proposal, PD61 Ndi1 had increased expression and gene copy number of COX3, a mtDNA encoded complex IV subunit ( Figure 5). Non-mitochondrial OCR was also significantly increased in PD61 Ndi1 compared to PD61 (not shown, p < 0.05). Increased availability of NADH due to Ndi1 expression was available for use by the ETC as well as non-mitochondrial NADPH oxidases accounting for the increase in non-mitochondrial oxygen utilization [82,83].
Since the presence of Ndi1 increases electron transfer along the ETC, electron leak also increased. Electron leak can result in increased free radical generation at complexes I and III [82]. Qualitative assessment of ROS generation by PD61 Ndi1 using DCFDA (2 ,7 -Dichlorodihydrofluorescein diacetate) was not appreciably different from PD61 (data not shown). Lack of increased free radical generation after Ndi1 expression is, however, consistent with Ndi1 studies in other systems [36]. The increased expression of PGC-1α seen in PD61 Ndi1 could also contribute to suppression of ROS through its role in ROS detoxification [84].
Spare capacity is critically important to synaptic function and neuronal survival since decreases in spare capacity increase neuronal vulnerability and can lead to an energy crisis when ATP demands are increased [81,85]. PD61 Ndi1 exhibited a significant increase in spare capacity compared to PD61 suggesting that PD61 Ndi1 cells have increased reserves and are better equipped to handle respiratory stress like that induced by MPTP or rotenone.
The significant increase in OCR we observed in PD61 Ndi1 was matched by a compensatory 56% decrease in ECAR (a surrogate for glycolysis). Mitochondrial oxygen utilization (OCR) can be significantly reduced by high levels of glucose in culture media that are needed to support dysfunctional PD cybrid cells [86]. SH-SY5Y neuroblastoma and cybrid cell lines exhibited the Warburg effect (low OCR and high ECAR) [87]. Ndi1 expression in PD cybrid cells also reduced their reliance on glycolysis and improved oxygen utilization and facilitated ATP production.
To determine how increased respiratory function after Ndi1 expression alters mtDNA genes, we measured changes in mtDNA gene expression and copy number. Reduced expression of mitochondrial genes has been observed in previous studies of PD human brain, peripheral cells, and cybrids [59,60,80,88]. In keeping with the changes in OCR, we also detected increased levels of mitochondrial gene expression and copy number for subunits of complex I, IV and 12s rRNA in PD61 Ndi1 compared to PD61.
Increased copy number for complex I genes did not result in an increase in assembled complex I based on immunoblots ( Figure 6C). We did not expect complex I assembly to increase since Ndi1 expression only bypasses dysfunctional endogenous complex I and does not restore the production of normal constituent subunits, assembly or normal function. Increased assembly of complexes III and V in PD61 Ndi1 compared to PD61 is a reasonable outcome since Ndi1 feeds electrons into complex III driving increased ATP-linked respiration. However, increases in mtDNA-encoded complex I and IV subunits ( Figure 5) did not translate into increased assembly of ETC complexes ( Figure 6). A recent study found subunits that are not immediately incorporated into assembled complexes may be quickly degraded [89]. This could explain why an increase in mtDNA gene expression was observed without a consistent effect on ETC complex assembly.
In PD61 Ndi1 , the observed changes in mitochondrial gene expression and gene copy number could be due to increased mitochondrial biogenesis signaling, driven by increased expression of PGC-1α which suggests that retrograde signaling from mitochondria back to the nucleus is increased [90]. Changes in mitochondrial NADH/NAD+ ratios as a result of Ndi1 could underlie the increase in PGC-1α. PGC-1α increases mitochondrial biogenesis as well as enzymes that detoxify ROS and has a global effect on mitochondrial function as reviewed by Austin and St. Pierre [91]. The finding that Ndi1 expression can lead to increased PGC-1α expression is a significant finding worthy of further investigation considering the evidence of reduced biogenesis signaling in PD patients [88]. Therapies that increase PGC-1α have shown some promise in improving mitochondrial function [92,93].
In this study, we anticipated that CLB expression would decline due to increased availability of ATP to fuel autophagic processes in Ndi1-expressing cells. We found significantly reduced levels of Congo red-labeled small protein aggregates in PD61 Ndi1 (Figure 7), however, the size and number of CLB expressed by PD61 Ndi1 was not significantly different from PD61. Lack of change in the expression of CLB despite improvements in OXPHOS and reduction in small aggregate density suggests that Ndi1 expression was not sufficient to-(1) prevent CLB formation or (2) induce clearance of CLB. The life cycle of a CLB is not well understood. Puncta of αSYN and Congo red-labeled aggregated proteins are present in nigral neurons in postmortem brain and in PD cybrids [50,94]. The presence of Ndi1 protein in CLB suggests that at least some CLBs were formed after Ndi1 transfection and expression ( Figure 8D). This finding indicates that improving ETC function alone is insufficient to prevent CLB formation. Inclusion of protein and gene expression changes as well as autophagy markers in future studies would be informative (see Nashine et al. [95]).

Cellular Consequences of Ndi1 Expression in SH-SY5Y
Unlike the robust changes in PD61 Ndi1 OCR, OCR in SH-SY5Y Ndi1 exhibited only a few altered features. This is consistent with other studies of alternate oxidases like Ndi1. Cannino et al. [36] reported that NDI1 gene expression did not detectably alter cell physiology or mitochondrial function in the absence of ETC dysfunction. Significant changes in SH-SY5Y Ndi1 were detected in maximum and spare capacity. Pretreatment of SH-SY5Y cells with rotenone, reduced basal OCR to levels comparable to PD61 (Figures 2C and 3C). In contrast, rotenone pretreatment of SH-SY5Y Ndi1 only significantly altered maximum and spare capacity ( Figure 3E) confirming that Ndi1 was functionally incorporated into SH-SY5Y mitochondria.
ECAR was significantly decreased ( Figure 3B) even though basal OCR in SH-SY5Y was not significantly improved by Ndi1 expression and mitochondrial mass was unaltered in SH-SY5Y Ndi1 ( Figure 3B). This outcome suggests that SH-SY5Y Ndi1 has become more dependent on OCR than SH-SY5Y.
The changes in cell function induced by Ndi1 in SH-SY5Y cells were not as dramatic as the changes seen in PD61. In fact, previous studies have shown that expressing Ndi1 in differentiated neurons did not affect their viability [96]. There was however increased biogenesis signaling via PGC-1α in SH-SY5Y Ndi1 . At the protein level, the ETC complex assembly was increased in SH-SY5Y Ndi1 compared to SH-SY5Y for complexes I, II, III, and IV. This finding was unexpected considering there was no upregulation of mtDNA gene expression or copy number. Previous studies have shown that undifferentiated SH-SY5Y cells have greater dependence on glycolysis than differentiated SH-SY5Y cells [97]. Interestingly, as mentioned above, we observed an increase in maximum capacity in SH-SY5Y Ndi1 . These data together suggest that SH-SY5Y Ndi1 cells could have a higher capacity for OXPHOS compared to the SH-SY5Y cells ( Figure 3E), which indicates a shift towards greater dependence on OXPHOS, resulting in the higher levels of ETC complex assembly in SH-SY5Y Ndi1 .
One drawback to these findings is that we only used two transfected lines in this study, a PD cybrid cell line with typical mitochondrial dysfunction, abnormal mitochondrial morphology, and αSYN aggregation and the parent SH-SY5Y cell line [50]. This preliminary study was designed to test if Ndi1 would be effective at improving mitochondrial function in a sporadic PD model with mitochondrial dysfunction. In light of these findings, this study should be expanded to include cybrid cell lines from PD and other neurodegenerative diseases with mitochondrial dysfunction as well as cybrid lines made from disease-free age-matched control platelets. We expect there will be heterogeneity in the impact of Ndi1 on individual PD cybrid lines, due to the differences in mitochondrial dysfunction among these cell lines [58][59][60].

Conclusions
In conclusion, the expression of Ndi1 in a trans-mitochondrial cybrid cell model of sporadic PD led to improved mitochondrial function, morphology, and transport. Mitochondrial gene expression, copy number, and biogenesis were also all increased in our PD cybrid cell line after the expression of Ndi1, suggesting that Ndi1 warrants further investigation as a source of insight into potential therapies for PD and other mitochondrial diseases. Although we saw a significant improvement in mitochondrial OCR that exceeded levels in control cybrid cell lines and a reduction in cellular aggregated protein content, Ndi1 expression did not prevent the formation of CLB in the PD cybrid cell line. This finding suggests that supplementing complex I with alternate oxidases like Ndi1 is sufficient to improve mitochondrial function in this PD model, but other additional interventions may be necessary to address the multisystem nature of PD pathogenesis.