Proteomic profiling of the hypothalamus in two mouse models of narcolepsy

Abstract

Narcolepsy is a disabling neurological disorder of sleepiness linked to the loss of neurons producing orexin neuropeptides in the hypothalamus. Two well-characterized phenotypic mouse models of narcolepsy, loss-of-function (orexin-knockout), and progressive loss of orexin (orexin/ataxin-3) exist. The open question is whether the proteomics signatures of the hypothalamus would be different between the two models. To address this gap, we utilized a label-free proteomics approach and conducted a hypothalamic proteome analysis by comparing each disease model to that of wild type. Following data processing and statistical analysis, 14 484 peptides mapping to 2282 nonredundant proteins were identified, of which 39 proteins showed significant differences in protein expression across groups. Altered proteins in both models showed commonalties in pathways for mitochondrial dysfunction and neuronal degeneration, as well as altered proteins related to inflammatory demyelination, insulin resistance, metabolic responses, and the dopaminergic and monoaminergic systems. Model-specific alterations in insulin degraded enzyme (IDE) and synaptosomal-associated protein-25 were unique to orexin-KO and orexin/ataxin-3, respectively. For both models, proteomics not only identified clinically suspected consequences of orexin loss on energy homeostasis and neurotransmitter systems, but also identified commonalities in inflammation and degeneration despite the entirely different genetic basis of the two mouse models.

1. Introduction

Narcolepsy is a neurological disorder characterized by excessive daytime sleepiness and is linked to the loss of neurons producing orexin neuropeptides. Orexin neurons are exclusively localized to the hypothalamus and project to multiple neuronal systems (i.e. monoaminergic and cholinergic neurons) [1]. The orexinergic system (including orexin-producing neurons, orexin neuropeptides, and orexin receptors) has been implicated not only in sleep/wake regulation, but also in metabolism, energy homeostasis, obesity, and type-2 diabetes [2, 3]. To date, the exact cause of the loss of these neurons in narcoleptic patients remains unclear. Genetically engineered animal models of orexin loss have provided a means for a better understanding of the orexinergic system and the pathophysiology of narcolepsy [4]. The two common models are a loss-of-function model (orexin knockout (KO)) and a progressive loss model (orexin/ataxin-3 (Atx)). The KO mouse has a null mutation induced by targeted disruption of the orexin gene, and has intact orexin-producing neurons lacking orexin neuropeptides. The Atx mouse expresses a human ataxin-3 truncated fragment under control of the human prepro-orexin promoter, resulting in progressive degeneration of orexin-containing neurons from birth to an almost complete loss of orexin neurons by 17 weeks of age [5,6]. These two mouse models, however, show a phenotype strikingly similar to humans with narcolepsy. Importantly, an analysis of the hypothalamus proteome would characterize any biological differences or similarities between the two models.

Recent advances in quantitative proteomics have significantly improved protein profiling of complex systems [7–9]. Label-free quantification with LC-MS/MS allows for the identification of hundreds of proteins, and facilitates quantitative comparisons of protein levels between samples and across groups [10–12].

In this study, we utilized a label-free quantitative approach based on ion intensities by peak area, followed by pathway and network analysis to profile each disease model and to pinpoint commonalities and differences of the hypothalamic proteome associated with orexin loss by different genetic mechanisms. Our work identified altered proteins and dysregulated pathways that provide a mechanistic insight into the functional consequences of orexin loss at a cellular level on energy homeostasis and neurotransmitters systems, and identified new commonalities in inflammation and degeneration.

2. Materials and methods

2.1. Animal and genotyping

Three biological groups, KO (n = 6; 14–16 weeks old), Atx (n = 5; 14–16 weeks old), and WT (n = 5; 14–16 weeks old) were studied. At 4 weeks of age, the number of orexin-expressing cells is notably diminished, and by 17 weeks of age, no orexin-positive neurons are found [13]. Prepro-orexin knockout (129S-Hcrttm1Ywa/J) and orexin/ataxin-3 transgenic mice (B6.Cg-Tg (HCRT-MJD) 1Stak/J) were supplied by Dr. Priyattam J. Shiromani (Medical University of South Carolina; Charleston, SC). These mice have been backcrossed for more than 20 generations on C57BL/6J mice [14,15]. WT mice were of the same background strain (C57BL/6J) as the disease models. Mice were housed at 22C on a 12h light/dark cycle with ad-libitum access to food and water. Mice were euthanized by carbon dioxide inhalation between 9:00 AM and 10:30 AM. Hypothalamus tissues were dissected and immediately stored at −80°C. Genotyping analyses were performed on DNA extracted from tail biopsies (Transnetyx Inc., Cordova, TN). Primers 5-CATGAAGGAAGAAGGTCCTGG and 3-CCTTGCACC CAGGAATCTGG were used against the orexin/ataxin-3, and 5-GACGACGGCCTCAGACTTCTTGGG and 3-TCACCCCCTTGGGATAGCCCTTCC against the endogenous prepro-orexin. Primers 5-TAGTTGCCAGCCATCTGTTG and 3-ACTCTCCACCCACAGACAGG were used against orexin-KO. Only homozygous mice identified by genotyping were used in this study, and gene expression of orexin-mRNA in the hypothalamus was present in WT but absent in the orexin-KO (fold change = −13) and orexin/ataxin-3 (fold change =−9) mouse models (data not shown).

All manipulations done to the mice were in accordance to the guidelines established by the National Institutes of Health (USA). All research protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at Case Western Reserve University.

2.2. Sample preparation

Thehypothalamictissuewassuspendedin200Llysisbuffer (50 mM Tris, 3% SDS, and protease inhibitors). The suspension was subjected to four cycles of vortexing, probe sonication, and 20 min of incubation on ice. Lysates were then centrifuged at 12 000 rpm for 20 min at 4°C. Samples were cleaned from detergent using a 10 kDa molecular weight cutoff filter and buffer exchanged with 8 M urea in 50 mM Tris [16]. Proteins were reduced with 10 mM DTT (8 M urea, 50 mM Tris) for 1 h at 37°C, followed by alkylation with 25 mM iodoacetaminde (8 M urea, 50 mM Tris) for 30 min in the dark. The 8 M urea was then adjusted to 4 M using 50 mM Tris. Protein concentration was measured using a Bradford assay (Bio-Rad, Hercules, CA), and the sample amount was normalized to the amount of BSA present in the sample as a reference protein. Ten micrograms of protein was digested with Lysyl-Endopeptidase (ratio of 1:20) for 2 h at room temperature. The urea concentration was then adjusted to 2 M using 50 mM Tris, followed by an overnight trypsin digestion (ratio of 1:20) at 37°C.

2.3. Reverse phase LC-MS/MS analysis

Three hundred nanograms of each sample were analyzed by LC-MS/MS using a LTQ-Orbitrap Velos mass spectrometer (Thermo Scientific, San Jose, CA) equipped with a nanoAcquity™ Ultra high pressure LC system (Waters, Taunton, MA). The injection order on the LC-MS was randomized over all samples and blank injections were run after each sample. Mobile phases were organic phase-A (0.1% formic acid in water) and aqueous phase-B (0.1% formic acid in acetonitrile). Peptides were loaded onto a nano-ACQUITY UPLC^R 2G-V/MC18 desalting trap column at a flow rate of 0.300 L/min. Subsequently, peptides were resolved in a nano-ACQUITY UPLC^R BEH300 C18, followed by a gradient elution of 1–90% of phase B over 240 min. A nano ES ion source at a flow rate of 300 μL/min, 1.5 kV spray voltage, and 270°C capillary temperature was utilized. Full-scan MS spectra (m/z 380–1800) were acquired at a resolution of 60 000 followedby20data-dependentMS/MSscans.MS/MS spectra were obtained by CID of the peptide ions of normalized collision energy of 35% to yield a series of b- and y-ions as major fragments.LC-MS/MS raw data were acquired using the Xcalibur software (version 2.2-SP1, Thermo Scientific, San Jose, CA). To facilitate the label-free quantitative analysis and to ensure that the peak retention time (RT) and the MS instrument performance were maintained within restricted tolerances, an equal amount of internal standard (88321, Thermo Scientific, San Jose, CA) was spiked into each sample [10,12]. The retention time, peak intensity, and mass accuracy of three peptides from the internal standard were monitored and showed less than a 5 min retention time drift, less than 4 ppm mass drift, and consistent peak widths and area under the curve (Supporting Information Table 1). Further, quality assessment of biological replicates revealed: (i) 89% of total peptides in the Atx group have a CV of 30 or less, (ii) 80% of total peptides in the KO group have a CV of 30 or less, and (ii) 88% of total peptides in the WT group have a CV of 30 or less (Supporting Information Fig. 1A–C).

2.4. Protein identification and quantification

To correct for possible systematic variations resulting from sample preparation, protein loads, peptide ionization, LC-MS reproducibility, and other sources of instrument drift, the LC-MS/MS raw files were imported into Rosetta Elucidator™ (Rosetta-Bio-software, version-3.3.0.1.SP.25) [17, 18] (Supporting Information Fig. 2). Retention time alignment, background subtraction and smoothing, peak identification, and peak extraction across the entire chromatographic time window (RT and m/z dimensions) were performed using the PeakTeller™ algorithm. Following accurate alignment of detected LC-MS peaks (retention time and m/z), the ion intensities were measured based on the peak area of the identified peaks. The generated peak files (.dta) were then searched by Mascot (version 2.1, Matrix Science London, UK) against the mouse Uniprot database (538 585 sequences). Search settings were as follow: trypsin enzyme specificity, mass accuracy for precursor ion: 10 ppm, mass accuracy for fragment ions:0.8Da, carbamidomethylation of cysteines as fixed modifications, oxidation of methionine as variable modification, and one missed cleavage. Peptide identification criteria included a mass accuracy of ≤10 ppm, an expectation value of p < 0.05, and an estimated FDR of less than 1%. Peptides were identified by matching accurate mass and elution time peaks to those stored in a reference database of peptides. Intensities of each peptide were then normalized based on the median intensity to adjust the data for differences in the TIC. The values were then log (10) transformed for statistical analysis by one-way ANOVA. One-way ANOVA was utilized to compare protein intensities across groups, and p-values were generated for these comparisons. Quantification of each peptide was generated by averaging the ion intensity of that peptide in all samples within a group. Relative peptide expression (fold change) was calculated using the averaged peptide ion intensities for the three possible pairwise group-based comparisons. Proteins were considered differentially expressed if the following cutoff criteria were met: (1) two or more peptides per protein, (2) each peptide has a p ≤ 0.05, and (3) each peptide has a fold change of ±1.5 or greater (see Section 2.8).

2.5. Pathway and network analysis

Peptides meeting the above cutoff criteria and their corresponding fold changes were imported into Ingenuity Pathways Analysis (www.ingenuity.com). Ingenuity pathways analysis calculated the p-value using a right-tailed Fisher’s exact test, and all pathways with p ≤ 0.05 and 2 or more associated proteins were considered significant. Molecular networks were generated using a combination of protein interaction databases and literature, looking primarily at commonalities between the two narcoleptic mouse models.

2.6. Protein validation using SRM MS

SRM validation of three selected proteins was performed on a total of 18 hypothalamus samples from a new mouse cohort (WT n = 6, KO n = 6, Atx n = 6). The hypothalamus tissues were lysed and then digested (as described above). Six target peptides without missed cleavages or modifications and with clear intense y-ion fragments were selected as follows: three peptides for SIRT2, two peptides for monoamine oxidase enzyme (MAOB), and one peptide for PPP1R1B (Supporting Information Table 2). Synthetic heavy-labeled peptides (C-terminal arginine or lysine residue labeled with ¹³C and ¹⁵N,AQUA-Quant-Pro-grade,with>97%purity,ThermoScientific, San Jose, CA) were spiked into each digest. 600 ng of each digest containing 30 fmole of each synthetic peptide were analyzed by LC/MS using a Waters Nano-ACQUITY UPLC system and a TSQ-Quantum Ultra MS system (Thermo Scientific, San Jose, CA). The digests were desalted on a reversed-phase C18 trapping column washed with 2% acetonitrile with 0.1% formic acid at 15 μL/min for 5 min. The trapping column was switched in-line with a reversed-phase C18-column. The peptides were separated using acetonitrile 10–45% in aqueous 0.1% formic acid over a period of 45 min at 0.3 L/min using the following parameters: ESI-positive mode, ion spray 2500 v, capillary 250°C, skimmer offset 7v, scan width 0.002 m/z, and scan time 0.02 s. Four SRM transitions were monitored and the precursor and product ion pairs were monitored for each peptide (Supporting Information Table 2). Data were processed using Pinpoint software (Thermo Scientific, San Jose, CA). Signal intensities of each endogenous peptide and its synthetic heavy-labeled peptide were obtained, and the data were interpreted using the ratio of endogenous to synthetic peptide signal intensity. In addition, the linearity range for each peptide was determined by generating a standard concentration curve. The linear range was 0.1–50 fmole/μL, the coefficient of determination R² was >0.99, and the lower limit of detection was 50–100 amole, all with a signal to noise ratio >200. The abundance of each peptide was calculated on the basis of the peak area intensity, and the peak areas for the target peptide were normalized to the internal standard.

2.7. Protein validation using Western blotting

Fifteen hypothalamic samples from a new mouse cohort (WT n = 5, KO n = 5, Atx n = 5) were used to validate insulin degrading enzyme (IDE) protein. Twenty micrograms of protein lysate was denatured in a sample buffer and heated at 95°C for 5 min. Protein samples were separated and resolved on 4–12% Bis-Tris gels (Invitrogen, Waltham, MA), and then transferred to nitrocellulose membranes overnight at 10 V. Membranes were blocked using Odyssey blocking buffer (927–40 000, LI-COR Bioscience) for 1 h at room temperature, and incubated with primary antibody against IDE protein (sc-393887, Santa Cruz) for 1 h at room temperature. The membranes were washed in TBST-T three times for 5 min each, and then incubated for 1 h at room temperature with IRDye-680 and IRDye-800 secondary antibody(925–68071 and 926–32210, LI-COR Bioscience). Membranes were scanned and visualized by Odyssey Infrared Imaging System (LI-COR Biotechnology, Lincoln, NE). Quantification was performed using LI-COR Image Studio-Lite-4.0 software. The protein density of each band was normalized to the endogenous β-actin loading control (SC-130656, Santa Cruz) on the same blot (Supporting Information Table 2). All comparisons across groups were performed by Student t-test using Prism 5.01 (GraphPad software, La Jolla, CA). Mann–Whitney correction was applied to comparisons where the variances were significantly different.

2.8. Statistical analysis

This study was performed across three mouse groups (WT, KO, and Atx) with five or six independent biological replicates from each group. We designed the study with sufficient power (95%) to detect 1.5-fold change in most of the measured analytes. Our previous studies have shown that for 75% of analytes, this experimental design typically achieves 80% or more power with a sample size of n = 5–6 independent biological replicates per experimental group [19,20]. Furthermore, with the use of an inbred strain, most environmental variances are eliminated in comparison to outbred mice and humans [21]. For label-free quantification, one-way ANOVA was utilized to compare protein intensities across groups, and p-values were calculated for these comparisons. The statistical significance of the differences across groups was performed using Student t-test (Prism 5.01, GraphPad software, La Jolla, CA), and Mann–Whitney correction was applied to comparisons where the variances were significantly different.

3. Results

3.1. Label-free protein quantification

Hypothalamic samples from disease models were compared to that of age-matched wild-type mice (C57BL/6J) (Fig. 1A). The strategy for sample preparation, protein identification, and relative quantification is illustrated in Fig. 1B. The chromatographic reproducibility in terms of retention time and intensity across groups are highlighted in the Supporting Information Fig. 3. Proteome coverage of over 14 484 quantified peptides mapping to 2282 nonredundant proteins was obtained across the groups. Of those, 921 peptides were statistically significant at p ≤ 0.05 using one-way ANOVA (Supporting Information Table 3). The charge state and p-value for each peptide as well as the total protein % coverage are reported (Supporting Information Table 3). Based on our cutoff criteria (Section 2.4), 39 proteins were identified across the groups with 27 unique proteins that were significantly changing in at least one group (Supporting Information Fig. 4). Figure 1C, visualizes the intersections between the 39 differentially expressed proteins. Twelve and three significant proteins were specifically associated with Atx versus WT (Atx/WT) and KO versus WT (KO/WT), respectively. Interestingly, ten proteins were shared between the Atx/WT and KO/WT comparison groups. Additionally, only two proteins were shared with Atx/WT versus Atx/KO, and none were shared with KO/WT versus Atx/KO or between all groups.

Figure 1.

Label-free quantitative analysis of two mouse models. (A) Hypothalamus tissues were collected from three biological groups, KO (n = 6), Atx (n = 5), and WT (n = 5). (B) Overview of MS-based label-free quantitative proteomics workflow. (C) Venn diagram visualizing the intersections of the 39 differentially expressed proteins across groups.

3.2. Differentially expressed proteins: The orexin-knockout model

The list of the 13 differentially expressed proteins (Supporting Information Table 4) brings attention to several potential targets including: MAOB (monoamine oxidase-B), IDE, SIRT2 (sirtuin-2), and PPP1R1B (protein phosphatase-1 regulatory subunit-1B), as discussed later. Pathway analysis revealed dysregulation in dopamine receptor signaling, melatonin degradation, NRF2-mediated oxidative stress response, mitochondrial dysfunction, and axonal signaling pathways (Supporting Information Fig. 5). Relevant to dopamine signaling, melatonin signaling, and mitochondrial dysfunction, is the MAOB protein. MAOB, a monoamine oxidase-B enzyme located in the outer mitochondrial membrane, was upregulated in KO compared with WT (Supporting Information Table 4). In addition, IDE, an insulin degrading enzyme and abundant protein in the brain, was detected to be downregulated in KO compared with WT. Figure 2A shows the top significant network(p ≤ 0.05)of direct connections between altered proteins including SIRT2, IDE, and PPP1R1B. A complete view of this network is provided in the Supporting Information Fig. 6.

Figure 2.

Top significant networks. Two protein networks highlight target proteins and molecular interactions in KO versus WT comparison (A), and Atx versus WT comparison (B) Complete networks are provided in the Supporting Information Figs. 5 and 7.

3.3. Differentially expressed proteins: The orexin/ataxin-3 model

The list of the 24 differentially expressed proteins (Supporting Information Table 4) highlights several target proteins including PKM (pyruvate kinase), synaptosomal-associated protein-25 (SNAP25), SIRT2, and PPP1R1B proteins. Pathway analysis revealed dysregulation in glycolysis, calcium signaling, mitochondrial dysfunction, and axonal signaling pathways (Supporting Information Fig. 7). Relevant to the glycolysis signaling pathway, PKM, a glycolytic enzyme, was upregulated in Atx compared with WT (Supporting Information Table 4B). SNAP25, a presynaptic plasma membrane protein, was downregulated in Atx compared with WT (Supporting Information Table 4). Figure 2B shows the top significant network that is highly populated with altered proteins including SIRT2 and PPP1R1B. A complete view of this network is provided in the Supporting Information Fig. 8.

3.4. Validation of selected proteins

Our label-free analysis revealed a number of differentially expressed proteins. In this work, protein expression of four target proteins was validated using SRM MS and Western blotting. Figure 3A–D highlights the peptide intensity-based quantification via label-free analysis for these four selected proteins. Using SRM, three, two, and one target peptides were selected to validate SIRT2, MAOB, and PPP1R1B, respectively (Supporting Information Table 2). The SIRT2 level was decreased in KO and Atx compared to WT for all three targeted peptides (Fig. 4A). Consistent with the proteomics analysis, this difference was significant (p ≤ 0.05) between KO and WT, and between Atx and WT. The MAOB level was increased in KO and Atx compared to WT for both targeted peptides (Fig. 4B). This level was more significant between KO and WT (p ≤ 0.01) than between Atx and WT (p ≤ 0.05). The PPP1R1B level was decreased in KO and Atx compared to WT (Fig. 4C). This difference was more significant between Atx and WT (p ≤ 0.05) than between KO and WT. Further, Western blotting analysis against IDE protein showed a decrease in the IDE level in both KO and Atx compared to WT (Fig. 4D). This difference was significant between Atx and WT (p ≤ 0.01). Overall, our SRM and Western blotting revealed consistent findings between the disease and healthy mice compared to label-free analysis in these selections.

Figure 3.

Box-and-whisker plots highlight the peptide intensity-based quantification of the hypothalamic proteome for four selected proteins, MAOB (A), Sirt2 (B), PPP1R1B (C), and IDE (D). The x-axis represents the three biological groups (WT, KO, and Atx). The y-axis represents the mean intensity of peptides that correspond to a specific protein. Differences across groups were assessed by Student t-test using Prism 5.01 (GraphPad software, La Jolla, CA). Mann–Whitney correction was applied to comparisons where the variances were significantly different. [Nonsignificant (ns) = p > 0.05; ^*= p ≤ 0.05; and ^**= p ≤ 0.01].

Figure 4.

Validated targets in both disease models. The abundance of three proteins was assayed by SRM MS: (A) Sirt2, (B) MAOB, and (C) PPP1R1B. Abundance of IDE protein was assayed by Western blotting (D). The bar graph plot illustrates the relative fold change across the three groups: WT (white-bar), KO (gray-bar), and Atx (black-bar). Differences across groups were assessed by Student t-test (Prism 5.01, GraphPad software, La Jolla, CA). Mann–Whitney correction was applied to comparisons where the variances were significantly different. (Nonsignificant (ns) = p > 0.05; ^*= p ≤ 0.05; and ^** = p ≤ 0.01).

4. Discussion

This study represents the first quantitative work on a global hypothalamus proteome in a narcolepsy model, and to our knowledge it provides proteomics data for narcolepsy model in any species. When comparing the wild type and two different genetically determined narcolepsy phenotypes, there were more similarities and few differences in the hypothalamic proteome.

The protein changes in the hypothalamus tissue are downstream of any genetic or epigenetic effects, and therefore indicate potential dysregulation in cellular protein functions. Visualizing the intersections between the 39 differentially expressed proteins suggest similarities in the hypothalamus proteome between the models independent of whether orexin is absent during development or lost after development. From these proteins, SIRT2 and PP1R1B are altered in both disease models compared with wild type. In contrast, there are few changes that are unique to each disease model.

Our study also suggests dysregulation in the monoaminergic and the dopaminergic systems, represented by the upregulation of the MAOB protein and the downregulation of the PPP1R1B protein in both models compared to wild type. Studies have demonstrated that orexins activate monoaminergic neurons of the locus coeruleus area as well as dopaminergic neurons of the ventral tegmental area in the hypothalamus, and thus the loss of orexin in narcolepsy may lead to an imbalance in these systems [3]. MAOB cat-alyzes the oxidative-deamination of monoamine neurotransmitters (including dopamine), leading to the production of reactive oxygen species and mitochondrial dysfunction [22,23]. PPP1R1B, a dopamine- and cAMP-regulated phosphoprotein (DARP-32), is a critical regulator of neuronal signals in the dopaminergic system [24]. As a target for dopamine and a key downstream effector in dopamine signaling, PPP1R1B is implicated in the pathogenesis of several neurodegenerative and neurological disorders [25]. Additionally, elevated levels of MAOB, decreased levels of dopamine, and increased levels of ROS are reported in Parkinson’s disease and Alzheimer’s disease [26,27]. Further, studies have shown that monoaminergic hypoactivity results in narcolepsy in the canine model, and the MAOB inhibitor has been a useful agent for narcolepsy [28].

While monoaminergic systems can be regulated by insulin, the KO model compared to Atx model has shown more dysregulation in insulin resistance and alteration in insulin signaling. In the brain, this may lead to alterations in mitochondrial function, increased levels of monoamine oxidases (such as MAOB), and decreased dopamine levels [29,30]. Insulin receptors are abundant in the hypothalamus, and orexin has been shown to be involved in the maintenance of insulin sensitivity [31]. In narcoleptic patients, orexin deficiency results in insulin resistance and increased obesity and type-2 diabetes [1, 32]. Mice with orexin-deficient neurons also exhibit late-onset obesity despite eating less [6,31]. Consistent with these findings, our data indicates a downregulation of IDE in the KO versus wild-type mice, and downregulation of SIRT2 in both disease models versus wild type.

Insulin resistance increases the circulating levels of insulin. Iqbal et al. reported that insulin promotes cellular metabolism by upregulating PKM expression and decreasing its activity [33]. PKM, which generates glycolytic ATP, is upregulated in the Atx model. PKM was also upregulated in KO; however, it did not meet our cutoff criteria of significance. Further, PKM plays a role in transporting neurotransmitters (including dopamine) into synaptic vesicles through ATP consumption [34]. Taken together, PKM in the Atx model is involved not only in pyruvate metabolism, but also in neurotransmission. Additionally, SNAP25, an essential protein for normal synaptic vesicle release, is downregulated in Atx, indicating a possible neurotransmitter release alteration and neuronal degeneration [35].

The pathway analyses from both disease models show a commonality in pathways for mitochondrial dysfunction, axonal/neuronal degeneration, as well as altered proteins related to inflammatory demyelination, insulin resistance, metabolic responses, and the dopaminergic and monoaminergic systems. Additionally, dysregulation in melatonin signaling (suggested by the upregulation of MAOB) was observed in the KO model, whereas dysregulation in glycolysis signaling (suggested by upregulation of PKM) was observed in the Atx model. We conclude that there are more similarities than differences among models.

Although there are no protein profiling studies related to orexin loss and narcolepsy, a number of gene expression studies have been performed. To our knowledge, there are only three gene expression studies that have been performed on these mouse models using different gene profiling techniques [36–38]. However, none of these studies have compared wild type to both disease models. Some of the reported genes were also identified in our proteomics study (i.e. Scn9a, Ptprn, Adk, Rap2, PDHA1, and hexokinase, Supporting Information Table 4C); however, they did not pass our selection criteria of significance. These studies have focused on genes enriched or colocalized in the Hcrt (orexin) neurons or in the brain, whereas we analyzed the entire hypothalamus region. Therefore, these protein changes are more likely to reflect secondary changes across the entire hypothalamus than within orexin cells. Additionally, differences in the regulatory purposes of RNA messages and proteins, posttranslational regulation of gene and protein expression, as well as sample preparation and profiling techniques, all may contribute to these differences. Therefore, one will need to design experimental comparisons of both mRNA and proteins in a common protocol of genetics and proteomics expression.

Despite the remarkable advances in quantitative proteomics [7–9], many limitations remain. One limitation is the coelution of low abundance peptides with high abundance peptides, which leads to ion suppression and an inability to detect the coeluting low abundance peptides. Another limitation is that of false discovery due to the small sample size and many measurements. In our study, we tried to minimize this by using a sample size (KO = 6, Atx = 5, WT = 5) that is adequate to enable quantitative comparisons across groups (Section 2.8). Furthermore, one limitation of our approach is the use of 10 KDa molecular weight cutoff filter prior to LC-MS/MS, which led to the loss of low molecular weight proteins including orexin A (~3.5 KDa) and orexin B ( ~3 KDa). However, we have confirmed the presence (WT) and absence (KO and Atx) of orexin by genotyping and gene expression (Section 2.1). Another limitation is the absence of multiple statistical comparisons; however, we designed the study with sufficient power to detect 1.5-fold changes (Section 2.8).In addition to these technique-associated challenges, the effects of orexin loss may be different when the genetic background is more heterogeneous. Moreover, the hypothalamus is comprised of cells and systems coordinating metabolic and hormonal responses beyond that of just the orexinergic system. These systems may respond to orexin loss and contribute to the protein differences.

In summary, the present discovery-based label-free expression data identifies pleiotropic effects of orexin absence (Fig. 5). For instance, downstream effects and the action of regulatory and counter regulatory systems include a broad prospective beyond orexin on sleep and metabolism. Overall, the data show alterations in proteins related to: (i) insulin resistance, obesity, and metabolic abnormalities, (ii) dopaminergic and monoaminergic imbalance, (iii) mitochondrial dysfunction and oxidative stress, and (iv) inflammatory demyelination as well as axonal, neuronal, and synaptic degeneration. In conclusion, our findings confirm and broaden what is known about the pathophysiology of narcolepsy and firmly support mechanistic insights fostered by animal models.

Figure 5.

Schematic diagram illustrating the dysregulated proteins and pathways associated with orexin loss. Differentially expressed proteins are related to: (1) insulin resistance, obesity, and metabolic abnormalities, (2) dopaminergic and monoaminergic imbalance, (3) mitochondrial dysfunction and oxidative stress, and (4) inflammatory demyelination and axonal, neuronal, and synaptic degeneration. The star indicates clinically known alterations and responses. Green, blue, and red colors indicate altered proteins in both, orexin/ataxin-3, and orexin-KO models, respectively.

Significance of the study.

Narcolepsy is a disabling neurological disorder of sleepiness linked to the loss of neurons producing orexin neuropeptides in the hypothalamus. Two well-characterized phenotypic mouse models of narcolepsy, loss-of-function (orexin-knockout), and progressive loss of orexin (orexin/ataxin-3) exist. In this study, we utilized a label-free quantitative approach followed by pathway and network analysis to profile each disease model and to pinpoint commonalities and differences of the hypothalamic proteome associated with orexin loss by different genetic mechanisms. Our work identified altered proteins and dysregulated pathways that provide a mechanistic insight into the functional consequences of orexin loss at a cellular level on energy homeostasis and neurotransmitters systems, and identified new commonalities in inflammation and degeneration. Overall, this study represents the first quantitative work on a global hypothalamus proteome in a narcolepsy model, and to our knowledge, it provides proteomics data for narcolepsy model in any species.

Abbreviations:

Atx

orexin/ataxin-3

IDE

insulin degrading enzyme

KO

orexin-knockout

MAOB

monoamine oxidase enzyme

PPP1R1B

protein phosphatase-1 regulatory subunit-1B

PKM

pyruvate kinase

RT

retention time

SIRT2

Sirtin-2

SNAP25

synaptosomal-associated protein-25