Brain pyrimidine nucleotide synthesis and Alzheimer disease

Many patients suffering late-onset Alzheimer disease show a deficit in respiratory complex IV activity. The de novo pyrimidine biosynthesis pathway connects with the mitochondrial respiratory chain upstream from respiratory complex IV. We hypothesized that these patients would have decreased pyrimidine nucleotide levels. Then, different cell processes for which these compounds are essential, such as neuronal membrane generation and maintenance and synapses production, would be compromised. Using a cell model, we show that inhibiting oxidative phosphorylation function reduces neuronal differentiation. Linking these processes to pyrimidine nucleotides, uridine treatment recovers neuronal differentiation. To unmask the importance of these pathways in Alzheimer disease, we firstly confirm the existence of the de novo pyrimidine biosynthesis pathway in adult human brain. Then, we report altered mRNA levels for genes from both de novo pyrimidine biosynthesis and pyrimidine salvage pathways in brain from patients with Alzheimer disease. Thus, uridine supplementation might be used as a therapy for those Alzheimer disease patients with low respiratory complex IV activity.

OXPHOS is much more than an energy-generating cellular pathway [4]. OXPHOS dysfunction can affect many biochemical routes, among them the de novo pyrimidine biosynthesis. This process requires three proteins: CAD, named after its three enzyme activities, carbamoyl-phosphate synthetase, aspartate transcarbamylase and dihydroorotase; DHODH, dihydroorotate dehydrogenase; and UMPS, which enzyme activities are orotate phosphoribosyltransferase and orotidine-5′-phosphate decarboxylase ( Figure 1). DHODH is located in the mitochondrial inner membrane and directly transfers electrons to the OXPHOS electron transport chain (ETC) via coenzyme Q10 [5]. A reduction of electron transport downstream coenzyme Q 10 would slow down DHODH activity and the de novo pyrimidine biosynthesis.
Pyrimidine nucleotides are required for the synthesis of nucleic acids, carbohydrates, and many membrane OM, IMS, IM, and M code for mitochondrial outer membrane, intermembrane space, mitochondrial inner membrane, and mitochondrial matrix, respectively; ETC, electron transport chain; CI, CIII, CIV, CV and Cyt c code for respiratory complexes I, III, IV, ATP synthase and cytochrome c, respectively; Q, coenzyme Q 10 ; NADH and NAD + , reduced and oxidized forms of nicotinamide adenine dinucleotide; H + , protons; e -, electrons; ADP, ATP and Pi, adenosine diphosphate, adenosine triphosphate, and inorganic phosphate; H 2  components, such as glycoproteins, glycolipids, and phospholipids. As proliferating cells need to duplicate their genomes and other cell components like biological membranes, they depend on high concentrations of pyrimidine nucleotides. Then, the de novo pyrimidine biosynthesis pathway is indispensable for these cells. However, differentiated cells do not divide themselves or replicate their genomes. Hence, it is generally considered that the activity of de novo route is low in these cells and they satisfy their need for pyrimidine nucleotides through the salvage pathway [5]. In this process, the uridine-cytidine kinase 2 (UCK2) phosphorylates uridine and cytidine nucleosides to produce UMP and CMP [6].
Neurons are post-mitotic, differentiated cells. Neuronal differentiation includes the formation of axons and dendrites and the maintenance of the neuron's vastly expanded surface. These events require a continuous membrane synthesis, even in adult's post-mitotic neurons [7]. Moreover, it has been observed that highly elaborate axonal arborization of neurons greatly increases their baseline energy demands, rendering neurons more vulnerable to perturbations of mitochondrial function pathways [8]. We had previously hypothesized that an OXPHOS dysfunction in AD patients can secondarily affect de novo pyrimidine biosynthesis and the plasma membrane remodeling [9]. This might explain the alterations in composition and structure of neuronal membranes linked to loss of synapsis that precede neuronal loss in AD patients [10]. Therefore, pyrimidine nucleoside supplementation could be used as a therapeutic agent in AD.

Expression of enzymes from pathways for pyrimidine nucleotide synthesis in neurons differentiated from neuroblastoma SH-SY5Y cells
To check a potential effect of the OXPHOS dysfunction on neuronal pyrimidine nucleotide synthesis, we used the human neuroblastoma SH-SY5Y cell line as a model. This cell line has frequently been used to study AD-related issues. In fact, considering 'SH-SY5Y' and 'Alzheimer's' terms, more than 1,400 publications appear in PubMed database. These cells can be differentiated into dopaminergic neurons. They express βIII-tubulin (TUBB3), a neuronal marker, and tyrosine hydroxylase (TH), a dopaminergic marker (Figure 2A-2I). Neuronal differentiation is also accompanied by an increase in oxygen consumption ( Figure 2J).
Proliferating cells express both DHODH and UCK2 mRNAs ( Figure 2K). Levels are very similar to those reported in the human protein atlas webpage (https://www.proteinatlas.org/). DHODH mRNA levels were approximately five times lower than those for UCK2 mRNA. Neuronal differentiation did not modify mRNA values. The UCK2 protein was not detected in proliferating cells, although CAD and DHODH proteins were found ( Figure 2L-2N). As UCK2 was previously reported in proliferating SH-SY5Y cells [11], we ruled out a general lack of pyrimidine salvage pathway in these cells. Perhaps the anti-UCK2 antibody was not sensitive enough. In fact, we needed 10 μg of a commercial enzyme to obtain a faint band in the Western blot. Additionally, we confirmed the presence of CAD and DHODH, but no UCK2, in these postmitotic cells, i.e. neuron-differentiated SH-SY5Y cells ( Figure 2L-2N).

Neuronal differentiation after DHODH inhibition
As previously commented upon, de novo pyrimidine biosynthesis is relevant for proliferating cells. Orally administered leflunomide is almost completely converted into its active metabolite teriflunomide and, inhibiting DHODH, it decreases pyrimidine nucleotide availability and cell proliferation. Because of this, it has been used for treatment of rheumatoid arthritis and it is also a potent anticancer drug [12]. During patient treatment, teriflunomide steady-state plasma concentrations of 200 μM can be reached [13]. We observed that leflunomide 100 μM largely reduces cell proliferation of human neuroblastoma SH-SY5Y cells ( Figure 3A).
According to our hypothesis, de novo pyrimidine biosynthesis pathway is also important for neurons. To revel its role in these cells, we studied the leflunomide effect on neuronal differentiation. Leflunomide 100 μM prevents an increase in TUBB3 and TH levels ( Figure  3B). It has been shown that uridine can completely overcome growth inhibition by leflunomide [14]. However, uridine 200 or 800 μM were not able to prevent the effect of leflunomide on neuronal differentiation ( Figure 3B).
The additional inhibition of other cell functions by leflunomide, besides its effect on the de novo pyrimidine biosynthesis pathway, might be the reason why uridine has no effect in the maintaining neuronal differentiation [12]. In human cells, leflunomide reduces the mitochondrial inner membrane potential [15]. Mitochondrial inner membrane potential is generated by the electron flow through the respiratory chain, and this electron flow is associated with oxygen consumption. To determine the leflunomide effect on oxygen consumption of neuroblastoma SH-SY5Y cells, we tested different leflunomide concentrations. Leflunomide 100 μM reduces oxygen consumption to a residual 33.5 % ( Figure 3C). Supporting our results, it has been reported that leflunomide, or its metabolite teriflunomide, decreased oxygen consumption in mouse and human cells [13,[16][17][18][19][20]. These results suggest that an important part of oxygen consumption and energy production in proliferating cells is due to pyrimidine biosynthesis.
DHODH knockdown diminished mitochondrial ATP production and inner membrane potential in mouse and human cells [21,22]. Leflunomide causes mitochondrial proliferation in human cells [15], suggesting a compensatory response. Moreover, externally added uridine to the culture media did not improve the mitochondrial inner membrane potential and did not prevent mitochondrial proliferation, despite that uridine normalizes the cell cycle progression [15]. Uridine 200 μM do not recover oxygen consumption of leflunomide-treated cells (Supplementary Figure 1). Similarly, it was reported that uridine had no protective effect against tamoxifeninduced impairment to mitochondrial respiration [23].

Neuronal differentiation after OXPHOS inhibition
The OXPHOS dysfunction due to leflunomide might be responsible for the reduced neuronal differentiation of neuroblastoma SH-SY5Y cells. Because leflunomide also acts on other cell targets [12], we selected the CIV inhibitor potassium cyanide (KCN) that presumably does not act on these other leflunomide targets. KCN 10 μM or 25 μM decrease oxygen consumption in neuroblastoma SH-SY5Y cells to a residual 78 and 62 %, respectively ( Figure 4A). It had been previously shown that KCN abolished DHO-induced oxygen consumption in mouse cells and mitochondria from different rat tissues [16,24], confirming that OXPHOS CIV inhibition affected the de novo pyrimidine biosynthesis pathway. Moreover, in the human neuroblastoma SH-SY5Y cell line, KCN 10 μM and 25 μM do not affect cell proliferation ( Figure 4B).
Neuroblastoma SH-SY5Y cells were cultured with KCN 10 μM to analyze the effect of OXPHOS dysfunction on neuronal differentiation. These cells do not show increased TUBB3 and TH levels ( Figure 4C). Supporting our results, it was previously reported that KCN significantly inhibited the dopaminergic neuronal differentiation of neural progenitor cells, derived from human induced pluripotent stem cells [22]. Hence, this compound, by reducing respiratory chain activity, prevents dopaminergic neuronal differentiation.
In MDA231 human breast cancer cells, it has been shown that de novo pyrimidine synthesis pathway is depressed under low energy conditions, supposedly to reduce ATP-consumption [25]. To determine if that occurs in differentiated cells, we measured DHODH and UCK2 mRNA levels in KCN-treated SH-SY5Y cells after their differentiation to neurons. Although the de novo pathway requires more ATP than the salvage one, DHODH mRNA expression level did not change, but UCK2 mRNA levels significantly decreased ( Figure 4D). It is worth noting that a high percentage of OXPHOS oxygen consumption is due to electrons supplied to the ETC through the DHODH enzyme. Then, de novo pyrimidine biosynthesis is required not only for pyrimidine synthesis and energy production but also for other OXPHOS-related cell functions, such as the adjustment of the levels of signaling molecules (ATP, calcium, NAD + and reactive oxygen species) to modulate different cell pathways [4]. This fact might explain why the de novo pathway is not down-regulated in neurondifferentiated cells under low energy conditions, whereas the salvage pathway only involved in pyrimidine synthesis, is reduced.

Uridine recovers neuronal differentiation of OXPHOS dysfunctional cells
It was previously reported that uridine protected AD fibroblasts and normal human neuronal progenitor cells against azide toxicity [26,27], another CIV inhibitor. Supplementation of uridine to aged rats increased brain neurofilament levels [28]. Besides, a uridine effect on neuronal differentiation was reported. Thus, in rat PC12 pheochromocytoma cells and mouse N2a and human LAN-5 and SH-SY5Y neuroblastoma cells, exogenous uridine enhanced cell differentiation as evidenced by increased number of neurite-bearing cells, neurites per cell, neurite branching, neurite length, and neurite neurofilament levels [27,[29][30][31].
We observed that uridine 200 μM was able to restore UCK2 mRNA levels in KCN-treated neuroblastoma SH-SY5Y cells ( Figure 4D). Perhaps an excess of the UCK2 substrate uridine acts as a positive regulator for the expression of UCK2 mRNA. Uridine completely recovered TUBB3 and TH protein levels in these cells ( Figure 4C). Uridine increases the synthesis of UDP-N-Acetylglucosamine (UDP-GlcNAc) [32]. This compound transfers GlcNAc to different proteins, such as transcription factors or histones, which implicates this modification in transcriptional regulation [33,34]. Maybe, this is the way uridine recovers TUBB3 and TH expression.
All these results indicate that OXPHOS dysfunction affects de novo pyrimidine nucleotide biosynthesis pathway and has negative consequences on neurons generated from neuroblastoma SH-SY5Y cells. As previously commented upon, it is considered that the activity of the de novo pathway is low in fully differentiated cells, such as mature brain neurons [5], but the cell's capacity to salvage uridine is limited, and no cell can tolerate complete DHODH inhibition [35]. Moreover, some observations suggest that de novo pyrimidine biosynthesis pathway is also important in mature neurons from adult human brain [36].

DHODH mRNA
According to RNA-Seq studies reported in the Expression Atlas webpage (https://www.ebi.ac.uk/gxa/ home; accessed April 23, 2018), CAD, DHODH, and UMPS mRNAs are found in brain of primates, rodents, other placental and marsupial mammals, birds, and amphibians (Supplementary Table 1). There are no large differences in the number of transcripts per million for genes from the de novo pyrimidine biosynthesis pathway among species. By comparison, UCK2 mRNA levels, from the pyrimidine salvage pathway, are also very much alike. The presence of these mRNAs in adult rat brain was confirmed by in situ hybridization analysis [37]. Neocortex, cerebellar cortex, and hippocampus, which have high neuronal densities, showed high signal intensities.
Since CAD, DHODH, and UMPS mRNAs are found in brain from different species, expression of these genes may also be in the human brain. We tuned up a RT-qPCR to quantify DHODH, normalized by GAPDH, mRNA levels in four different regions of adult human brain from clinical and histopathological controls (CHPC) (Supplementary Table 2 and Figure 5A). We also determined UCK2 mRNA levels (Supplementary Table 2 and Figure 5B). Depending on the brain region, the DHODH mRNA amount was 4.3 -9.7 times lower than that of UCK2 mRNA. DHODH and UCK2 mRNA levels in the locus ceruleus (LC) were significantly higher than those from the other brain regions. Dentate gyrus (DG) UCK2 mRNA levels were also significantly different from those of other brain regions ( Figure 5A, 5B). A review of the Expression Atlas Webpage (accessed April 23, 2018) showed that DHODH and UCK2, but also CAD and UMPS, mRNAs were found in different regions of adult human brain (Supplementary Table 3). Similar mRNA levels were found for different genes, in different brain regions, and in various studies.

DHODH protein
By using a non-commercial anti-DHODH polyclonal antibody, the immunoblotting of protein from adult rat brain revealed the presence of DHODH in different regions, such as cortex, hippocampus, striatum, cerebellum, brain stem, and spinal cord [38].
To confirm the occurrence of enzymes from the de novo pyrimidine biosynthesis pathway in adult human brain, we first performed a western blot for CAD. The anti-CAD monoclonal antibody showed a unique band of the expected size ( Figure 6A). However, the anti-DHODH polyclonal antibody showed multiple bands ( Figure  6B). One of them was located in the expected position for a protein which size is the same of DHODH (43 kDa). Additionally, this signal coincided with the band for a commercial DHODH lacking its first 31 amino acids (ab128451, Abcam, Cambridge, UK), with approximately 40 kDa expected size. The analysis of different brain regions, such as hippocampus, entorhinal cortex (EC), and putamen suggested the presence of DHODH in all of them, but again, multiple bands were observed ( Figure 6C). An anti-DHODH monoclonal antibody resulted in a cleaner lane, but still with several bands ( Figure 6D). One of them was located in the expected position, coinciding with the site for the commercial DHODH. To produce this monoclonal antibody, a DHODH fragment (amino acid positions 32 -141) was used (ab54621, Abcam, Cambridge, UK). This antibody also recognized the same DHODH fragment generated by us ( Figure 6E).
In an attempt to further confirm the presence of this protein in adult human brain, we carried out proteomic analyses of the western blot candidate band. In parallel to the brain sample, we also loaded a sample from neuroblastoma SH-SY5Y cell line, because proliferating cells require a large amount of pyrimidine and are dependent on DHODH, as we have previously shown ( Figure 6F). Peptide mass fingerprinting did not show the protein in these samples. However, a parallel reaction monitoring (PRM) proteomics analysis showed 3 DHODH peptides in the western blot band corresponding to neuroblastoma SH-SY5Y cell homogenate ( Supplementary Figures 2A and 3), but DHODH could not be confirmed in the band corresponding to brain homogenate, probably due to a high background. Peptide mass fingerprinting and PRM proteomics analyses suggested, but did not confirm, the presence of DHODH in one of 6 spots from a twodimensional electrophoresis gel (Supplementary Figures  2B and 4). It may be that the brain DHODH quantity was too low. However, a western blot of a brain sample using two different concentrations of a commercial DHODH allowed us to estimate its brain concentration as close to 2.2 ng DHODH/mg brain ( Figure 6G). By ELISA, a mean concentration of 128.3 pg DHODH/mg brain was estimated in hippocampus from five control individuals. Surprisingly, the UCK2 mean concentration was much lower, 0.6 pg UCK2/mg brain. These results are in line with those SH-SY5Y cells, in which higher UCK2 mRNA levels but much lower UCK2 protein concentrations than those for DHODH were observed.

DHODH activity
DHODH histochemistry analysis in adult rat brain showed high staining in different regions, with the hippocampus presenting a characteristic pattern [38]. We could not get optimally-cryopreserved adult human brain samples appropriate for histochemical studies. However, we detected high DHODH activity in adult mouse brain homogenate and Brequinar, a DHODH inhibitor, largely reduced the staining ( Figure  7A-7C). Moreover, fresh samples from adult mouse brain consume significantly more oxygen when dihydroorotate was used as electron donor ( Figure  7D). Leflunomide reduces oxygen consumption ( Figure 7E). Similar to proliferating SH-SY5Y cells, in the absence of dihydroorotate, leflunomide was able to reduce oxygen consumption to a residual 60 %. This result highlights the importance of de novo pyrimidine biosynthesis pathway in adult brain, not only for pyrimidine production, but also for OXPHOS function.
Supporting our results on adult brain DHODH activity, in the 1960s and 1970s, different publications showed the presence of the de novo pyrimidine biosynthesis pathway in adult mouse, rat, and cat brain. Thus, the intracranial injection of radioactive intermediates for this pathway resulted in an early appearance of radioactive pyrimidine nucleotides in the brain [39][40][41]. Moreover, administration of these radioactive compounds to adult rat brain sections or homogenates also produced radioactive pyrimidine nucleotides [40,[42][43][44][45][46][47][48]. Finally, high DHODH activity was found in adult rat brain homogenate [49].

DHODH immunohistochemistry
Previous results strongly suggested the presence of the de novo pyrimidine biosynthesis pathway in adult human brain. However, these results did not indicate in which cell type this route was expressed. Immunocytochemical staining with anti-DHODH antibody confirmed DHODH in frontal cortex, hippocampus, striatum, and substantia nigra pars reticulate in the adult rat brain. Immunoreactivity was found in neuronal bodies and absent in neuroglia [38]. Moreover, neuronal incorporation of radioactive orotic acid intraventricularly injected into adult rat brain was higher than into glial cells [50]. Immunohistochemistry staining of adult human samples with anti-CAD and anti-DHODH antibodies confirmed the presence of both enzymes in different brain regions, such as amygdala, cerebellum, DG, EC, frontal cortex, putamen, and striatum ( Figure 8). Stronger staining was observed in zones with high neuronal density whereas the immunoreactivity was hardly seen in white matter. Furthermore, no marking is noted in cytoplasm of glial cells (Figure 8). In neurons, staining is located in perikarya and dendrites (Figure 8 and Supplementary Figure 5). Supporting our results, the Human Protein Atlas Webpage (https://www.proteinatlas.org/, accessed April 26, 2018) shows that CAD and DHODH proteins appear, at low levels, in different adult brain regions, such as cerebral cortex, hippocampus, caudate, and cerebellum. Neurons and glial cells showed mediumlow and low or no staining, respectively.
Although CAD is a cytosolic enzyme and DHODH is a mitochondrial one, the pattern of staining is very similar for both of them. These enzymes together with UMPS, another cytosolic enzyme, belong to the same metabolic pathway, and it has been shown that CAD and UMPS are placed around and outside mitochondria, closely associated with their outer membranes [5,51]. Then, the enzymes for the complete biosynthetic pathway are physically associated in the cell.
All these results confirm the presence of the de novo pyrimidine biosynthesis in mature neurons from adult human brain. We wondered if this biochemical pathway would be affected in Alzheimer disease patients.

UCK2 and DHODH mRNA levels in different Alzheimer disease stages
To study the biochemical pathways for pyrimidine nucleotide synthesis in AD patients, we analyzed UCK2 and DHODH mRNA levels from different brain regions, such as LC, EC, hippocampal cornus ammon 1 (CA1), and DG. Neurofibrillary pathology is found in LC and EC in AD Braak and Braak stages I/II, CA1 in AD stages III/IV, but not in DG, even in AD stages V/VI [52].
AD stages were significantly associated with UCK2 mRNA levels in all these brain areas ( Figure 9A-9D). However, the pattern was different in CA1, where higher levels were found in more severe AD stages, than in the other regions, where UCK2 mRNAs tended to decrease as neuropathology progressed. A potential explanation, already pointed out to justify lower CA1 mitochondrial DNA (mtDNA) deletion levels [53], would be that a selective and progressive apoptosis of CA1 neurons with decreasing UCK2 mRNA levels may shift the pattern towards an apparent increase of 'healthy' cells. In contrast, DHODH mRNA levels were not associated with AD stages (Figure 9E-9H). These results resemble those previously found in KCN-treated SH-SY5Y cells, in which energy stress provoked a decrease in UCK2 mRNA levels but did not modify DHODH mRNA levels ( Figure 4D), and the same explanation might be advanced. Despite this, and except for LC, positive correlations between DHODH and UCK2 mRNA levels were observed in these regions ( Figure 9I-9L), thus suggesting that both pathways are essential.
Associations between these mRNA levels and AD stages are not due to unbalanced sex or postmortem delay distributions. No association was found between AD stages and RNA integrity number (RIN) values. AD stages and age are significantly associated [54]. Although there were no significant correlations between age and DHODH mRNA levels, age and UCK2 mRNA levels were significantly correlated (Supplementary Figure 6A-6H). These correlations were negative in LC, EC, and DG, and positive for CA1. To evaluate the relevance of AD stages versus age on UCK2 mRNA levels, we divided individuals according to their UCK2 mRNA levels (higher and lower half) for each brain region and AD stage, and compared their ages (Supplementary Figure 6I-6L). Except for DG AD I/II stages, there were no significant differences. This result suggests that variation in UCK2 mRNA levels is more dependent on AD stage than on age. If lower UCK2 mRNA levels in different brain regions correspond with a decrease in pyrimidine synthesis pathway, and, as it has been recently reported, AD subjects show lower uridine concentrations in plasma and cerebrospinal fluid [55][56][57][58][59], then AD patients probably have a cerebral deficiency in many cellular functions related to the metabolism of uridine. Therefore, it can be suggested that uridine therapeutic administration would improve biochemical pathways related to brain function. In this line, uridine treatment attenuates the impairments in learning and memory observed in gerbil, stressed rat, and AD mouse models [27,[60][61][62][63]. Nutrient combinations including uridine supplementation have shown promising results in AD therapy [64][65][66][67][68][69]. As previously proposed, reduction in UCK2 mRNA levels might be secondary to a drop in energy production. OXPHOS is the main energy provider to power neuronal activity [70]. Thus, OXPHOS dysfunction in AD patients might explain the brain reduction in UCK2 mRNA levels.

Reduction in MT-CO1 and COX4I1 mRNA levels is associated with an increase in DHODH/UCK2 ratio in entorhinal cortex from Alzheimer disease patients
We used mtDNA-encoded MT-CO1 and nuclear DNA (nDNA)-encoded COX4I1 mRNA levels as surrogates of OXPHOS function. A comparison of mRNA levels and AD stages in different brain regions showed a significant association in EC ( Figure 10A, 10B). In this region, AD stages V/VI patients had lower MT-CO1 and COX4I1 mRNA values than controls. The COX4I1 mRNA levels were also associated to AD stages in DG and LC. We noted above that AD stages and age are significantly associated. Moreover, it was previously reported that MT-CO1 mRNA levels were significantly lower in cerebral hemispheres from 28 than 9 months old rats [71]; and the mRNA levels for the three mtDNA-encoded CIV subunits were lower in cerebral cortex from 24-than 18-month-old mice [72]. We have also observed a negative significant correlation between MT-CO1 and COX4I1 mRNA levels and age, but only in EC (Figure 10C, 10D). To evaluate the importance of AD stages versus age on EC MT-CO1 and COX4I1 mRNA levels, we divided individuals according to their MT-CO1 and COX4I1 mRNA levels (higher and lower half) for each AD stage, and compared their ages ( Figure 10E, 10F). Except for EC AD V/VI stages, there were no significant differences. However, contrary to what would be expected, and what is observed in the correlation, the youngest individuals had lower MT-CO1 mRNA levels. This result suggests that variation in MT-CO1 and COX4I1 mRNA levels is more dependent on AD stage than on age.
Interestingly, in EC, MT-CO1 and COX4I1 mRNA levels negatively correlate with DHODH/UCK2 ratio ( Figure 10G, 10H). In EC, this ratio was associated with AD stages (Figure 10I), and was significantly higher in AD stages V/VI than that in I/II and III/IV. The ratio did not correlate with age. The higher DHODH/UCK2 ratio in EC of AD stages V/VI is perhaps an attempt to compensate for a reduction in energy production by a CIV defect. As down-regulation of the de novo pathway might affect many vital reactions, because of its importance for OXPHOS function, a reduction in the salvage pathway, to avoid dispensable ATP consumption, would cause a higher DHODH/UCK2 ratio.
As a conclusion, our results confirm that an OXPHOS dysfunction, throughout an altered de novo pyrimidine biosynthesis pathway, can affect variables of postmitotic cells, such as neuronal differentiation; that the de novo pyrimidine biosynthesis pathway is present in adult human brain; and that both de novo and salvage pyrimidine nucleotide synthesis pathways appear to be unbalanced in AD brain. However, more work is required to link these pathways for pyrimidine nucleotide synthesis to an OXPHOS defect in AD brain. In particular, the determination of pyrimidine nucleotide levels and DHODH and UCK2 activities in large numbers of AD brains could confirm our RNA results.
Differentiation was induced following optimized protocols [109], but instead of 25 mM glucose, we used 5 mM galactose media to enhance OXPHOS function. Cells were maintained at 37 °C in a humidified atmosphere of 5 % CO2.
Three growth curves were performed for each condition, and each time point (from day 0 to day 12) was counted in triplicate.

Human and mouse brains
Brain samples from three healthy adult individuals were obtained from the Institute of Neuropathology HUB-ICO-IDIBELL Biobank following the guidelines of Spanish legislation on this matter and the approval of the local ethics committee. Mouse brain was obtained from C57BL/6J adult individuals (2 to 8 months) after the approval of the local ethics committee. Animals were killed by decapitation and the brain was immediately removed. For histochemistry analysis, brains were covered with optimal cutting temperature (OCT) compound and frozen at -80 °C. For oxygen consumption studies, brains were used immediately after dissection.

RNA analysis
Brain RNA samples were obtained from the Institute of Neuropathology HUB-ICO-IDIBELL Biobank. The post-mortem interval between death and tissue processing was between 2 and 22 h 15 min. Processing of brain tissue has been detailed elsewhere [110]. Neuropathological diagnosis and staging of AD were based on the classification of Braak [52]. CHPC were middle-aged individuals (n = 20, 11 men, 9 women; age, 50.7 ± 7.2 years) with no neurological or mental disorders and without lesions on the neuropathological study. Four brain regions were examined: CA1, DG, EC, and LC. A summary of all analyzed individuals is provided (Supplementary Table 2). Not all regions were assessed in every case because of the limited availability of tissues.

RNA purification
RNA from frozen cells and tissue was extracted following the instructions of the supplier (NucleoSpin RNAII, Macherey-Nagelsupplier, Düren, Germany and RNeasy Mini Kit, Qiagen, Hilden, Germany). RIN and 28S/18S ratios were determined with the Agilent Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA). Samples were digested with DNase, and RNA concentration was evaluated using NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA).

qPCR
The mRNA levels were measured with the RT-qPCR method using an Applied Biosystems StepOne TM Real-Time PCR System Thermal Cycling Block and a ViiA 7 Real-Time PCR system. DHODH, UCK2 and MT-CO1 mRNA levels were determined in triplicate with RT-qPCR using TaqPath TM 1-Step Multiplex Master Mix (DHODH and UCK2) and TaqMan ® RNA-to-C T TM 1- Step Kit (MT-CO1). The expression levels were normalized using GAPDH mRNA or 18S rRNA. Serial dilutions of recombinant plasmid DNA were included in each experiment to generate a standard curve for DHODH, UCK2, and GAPDH.
The DHODH peptide corresponding to amino acids 32-141 was heterologously expressed in Escherichia coli strain C41 (DE3) as a recombinant peptide using pET28a expression vector.
The quantitative determination of DHODH and UCK2 proteins was performed using commercially available Human Dihydroorotate Dehydrogenase (DHODH) ELISA kit and Human Uridine-Cytidine Kinase 2 (UCK2) ELISA kit (Abbexa, Cambridge, UK). Protein extracts concentrations were measured by Bradford assay. Samples were collected and DHODH and UCK2 levels were immediately measured, following the manufacturer's instructions.

Histochemistry
DHODH histochemistry was performed using the nitroblue tetrazolium/formazan technique on 20 µm cryostat sections of adult mouse brain. The oxidation of its substrate dihydroorotate (DHO) (10 mM) can be determined by trapping the electrons with tetrazolium (1 mM) in the presence of cyanide (10 mM), to prevent their further flow along the chain to oxygen, in phosphate buffer pH 8, 37 °C, 60 min. The insoluble blue product formazan can be revealed by light microscopy [38].

Immunohistochemistry
This protocol was performed at the Anatomic Pathology Core Unit of the Health Sciences Institute of Aragon (Zaragoza, Spain). Once paraffin blocks were made, 2.5 µm sections were cut with a rotation microtome (Leica RM2255) and paraffin sections were taken on superfrost glass slides. Slides were air dried at 37 °C overnight and, for immunohistochemistry stain, they were deparaffinized in xylene for 10 min, rehydrated in a grades series of ethanol (100, 100, 96, 70, 5 min each) and distilled water for 5 min. After rehydration, antigen retrieval was performed by means of the PT Link (Dako) at 95 °C for 20 min in a low/high pH buffer (Dako antigen retrieval, low/high pH). After retrieval, automated immunostaining was performed with a previously optimized protocol. For this process endogenous peroxidase was first blocked (EnVision FLEX Peroxidase-Blocking) followed by 5 min 2 washing steps (Dako wash buffer); mouse monoclonal anti-DHODH 1/50 and rabbit polyclonal anti-CAD 1/200 (ab54621, ab99312, Abcam, Cambridge, UK) primary antibodies were used with 40 min incubation AGING time. After 2 wash steps a dual rabbit/mouse HRP conjugated visualization system for signal amplification was used (Envision Flex HRP, Dako). After 3 wash steps (Dako wash buffer, 5 min each), 3,3'-Diaminobenzidine (DAB) was used for signal development after precipitation at primary antibody binding sites.

Flow cytometry
Cell suspensions were first fixed with 4 % paraformaldehyde for 15 min at 4 °C and permeabilized using a commercial buffer (Thermo Fisher Scientific). Samples were incubated overnight with the primary antibodies rabbit anti-TUBB3 (1:1,000) from Abcam and rabbit anti-TH (1:200) from Sigma, and then washed with PBS (Thermo Fisher Scientific). Cells were incubated for 1 h with appropriate secondary antibodies Alexa 488 (Thermo Fisher Scientific), washed, and then analyzed on a BD FACScan System (Becton-Dickinson, San Jose, CA, USA). 10,000 cells were studied. The results were analyzed using Weasel software.

Oxygen consumption
Neuroblastoma SH-SY5Y cell line: Cells were cultured in 5 mM galactose media. The cells were then collected by trypsinization, washed, counted, and resuspended at 1.5 x 10 6 cells/ml. Endogenous and inhibited respiration (with leflunomide or cyanide) analyses were performed. Each condition was analyzed three times.

Mouse brain
Brain was immediately obtained after animal death, dissected on ice, and weighed on an analytical balance. The dissected brain was directly transferred into ice-cold mitochondrial respiration medium (MIR05: 110 mM sucrose, 60 mM K + -lactobionate, 0,5 mM EGTA, 3 mM MgCl 2 , 20 mM taurine, 10 mM KH 2 PO 4 , 20 mM HEPES and 1 g/l BSA, pH 7.1). Tissues were homogenized in the same medium with a pre-cooled glass potter at 1,000 rpm, 16 strokes. Resulting homogenates containing 10 mg tissue wet weight were suspended in 1 ml of ice-cold MIR05. To measure the oxygen consumption in the presence of the substrate, MIR05 medium contained 6 mM of DHO. Endogenous and inhibited respiration (with leflunomide) analyses were performed. Each condition was analyzed three times.
Oxygen consumption of cells and mouse brain was analyzed using the high-resolution oxygraph OROBOROS ® .

Mass spectrometry analysis of protein spots
Protein identification by peptide mass fingerprint Spots were excised from gels, reduced, alkylated and digested with trypsin sequencing grade (Roche Molecular Biochemicals) [111]. Produced peptides were analyzed in a 4800 Plus Proteomics Analyzer MALDI-TOF/TOF mass spectrometer (Applied Biosystems, MDS Sciex, Toronto, Canada) at the Proteomics Unit of Complutense University of Madrid. The MS data was searched against SwissProt Data Base with taxonomy restriction to human (553,231 sequences) using MASCOT 2.3 (http://www.matrixscience.com) search engine through Global Protein Server v 3.6 software (ABSciex).
The search parameters were carbamidomethyl cysteine as fixed modification and oxidized methionine as variable modification. Peptide mass tolerance was 50 ppm and up to 1 missed trypsin cleavage site allowed. All identified protein outperformed the probability scores fixed by mascot as significant with a p-value minor than 0.05.

Protein identification by PRM
Desalted peptides from digested protein were analyzed by targeted proteomics to identify only the protein of interest in this study (PYRD or DHODH). The software Skyline 4.1 was used to build and optimize the PRM method for detection of unique peptides from DHODH [112]. An inclusion list consisted of m/z of resultant target candidates for each peptide was exported to Xcalibur 4.0 (Thermo Scientific) acquisition software of Q-Exactive HF mass spectrometer. The final PRM methods included one protein, 49 peptides and 112 precursors. The peptides for PRM were analyzed by nanoflow liquid chromatography-tandem mass spectrometry in an EASY-nLC 1000 System coupled to the Q-Exactive HF mass spectrometer through the Nano-Easy spray source (all from Thermo Scientific, Bremen, Germany). MS Data acquisition was performed in Q-Exactive HF using PRM method. Then Data were analyzed in Skyline software and identified with Mascot search engine thorough Proteome discoverer 2.2 software (Thermo Scientific). Peptides identifications based on MSMS data were used by skyline to generate a spectral library. To confirm DHODH detection, data were processed against the PRM-library on Skyline and manually inspected to ensure consistency between the transitions detected and the sequences of peptide searched.

Statistical analysis
The statistical package StatView 6.0 and GraphPad Prism 5.0 were used to perform all the statistical analyses. The normality of distribution was analyzed with Kolmogorov-Smirnov test. Mann-Whitney or Kruskal-Wallis non-parametric tests were used when values did not follow a normal distribution, while the ANOVA test was used for normal variables. To compare more than two groups post-hoc tests were also performed. All data were expressed as mean ± standard deviation and significance levels were set at p < 0.05 and the levels indicated by the post-hoc tests. Statistical values in cell lines were obtained after three or more independent experiments. In each independent experiment, several technical replicates were performed.

ACKNOWLEDGMENTS
We wish to thank Dr. Javier Godino (Servicio de Separación Celular y Citometría, Instituto de Aragonés de Ciencias de la Salud, Zaragoza, Spain) for his assistance with flow cytometry analysis, Santiago Morales (Universidad de Zaragoza, Zaragoza, Spain) for his help with the figures, the Servicio General de Apoyo a la Investigación-SAI (Universidad de Zaragoza, Zaragoza, Spain) and T. Yohannan for editorial help.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

FUNDING
This work was supported by the Instituto de Salud Carlos III (grant numbers PI17/00021 and PI17/00166); the Gobierno de Aragón (grant number LMP135_18 and Grupos de Referencia B33_17R) and FEDER 2014-2020 "Construyendo Europa desde Aragón". We also thank CIBERER and CIBERNED, Institute of Health Carlos III (ISCIII) for cooperation support. AGING

Supplementary Figures
Supplementary Figure 1 showing the selected spots for peptide mass fingerprinting (blue and red arrows) and parallel reaction monitoring (PRM) (red arrow) proteomics analyses.