Hypothalamic Mitochondrial Dysfunction Associated with Anorexia in the Anx/Anx Mouse

Hypothalamic mitochondrial dysfunction associated with anorexia in the anx/anx mouse

Charlotte Lindforsa,b,1, Ida A. K. Nilssona,b,c,1,2, Pablo M. Garcia-Rovesd,e, Aamir R. Zuberif, Mohsen Karimib,g, Leah Rae Donahuef, Derry C. Roopenianf, Jan Mulderh,i, Mathias Uhlénj, Tomas J. Ekströmb,g, Muriel T. Davissonf, Tomas G. M. Hökfeltc,i,2, Martin Schallinga,b, and Jeanette E. Johansena,b

Departments of aMolecular Medicine and Surgery and gClinical Neuroscience, Karolinska Institutet, 171 76 Stockholm, Sweden; bCenter for Molecular Medicine, Karolinska University Hospital, 171 76 Stockholm, Sweden; Departments of cNeuroscience and dPhysiology and Pharmacology, Karolinska Institutet, 171 77 Stockholm, Sweden; eDiabetes and Obesity Laboratory, Institute for Biomedical Research August Pi i Sunyer and Spanish Biomedical Research Centre in Diabetes and Associated Metabolic Disorders, 08036 Barcelona, Spain; fThe Jackson Laboratory, Bar Harbor, ME 04609; hEuropean Neuroscience Institute at Aberdeen, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom; iScience for Life Laboratory, 171 21 Stockholm, Sweden; and jDepartment of Biotechnology, AlbaNova University Center, 106 91 Stockholm, Sweden

Contributed by Tomas G. M. Hökfelt, September 16, 2011 (sent for review August 23, 2010)

The anorectic anx/anx mouse exhibits disturbed feeding behavior fibers immunoreactive (ir) for AGRP and NPY (7, 8, 10, 11, 16), and aberrances, including neurodegeneration, in peptidergic neu- as well as α-melanocyte–stimulating hormone (αMSH) and anx/anx rons in the appetite regulating hypothalamic arcuate nucleus. Poor CART (7, 9). In the mouse, the orexigenic AGRP system feeding in infants, as well as neurodegeneration, are common develops normally until P12, after which the normal, continuous increase in AGRP-ir fiber density in the main projection areas phenotypes in human disorders caused by dysfunction of the ceases, and in some areas even decreases. These changes overlap mitochondrial oxidative phosphorylation system (OXPHOS). We both temporally and spatially with activation of microglia (10) therefore hypothesized that the anorexia and degenerative phe- and expression of MHC class I gene (H2-Db) by both neurons notypes in the anx/anx mouse could be related to defects in the and glia cells (15). In addition, we have shown increased OXPHOS. In this study, we found reduced efficiency of hypotha- amounts of apoptotic cells and presence of activated caspase 6 in lamic OXPHOS complex I assembly and activity in the anx/anx NPY-ir fibers in the hypothalamus of these mice, indicating de- mouse. We also recorded signs of increased oxidative stress in generative processes (15). anx/anx anx/anx hypothalamus, possibly as an effect of the decreased Interestingly, the mouse shares many symptoms with mitochondrial oxidative phosphorylation system (OXPHOS) hypothalamic levels of fully assembled complex I, that were dem- fi Ndufaf1 complex I (CI) de ciencies, including poor feeding, neuro- onstrated by native Western blots. Furthermore, the degeneration, and muscle weakness (17), including the eyelid gene, encoding a complex I assembly factor, was genetically map- muscles (OMIM 252010). In the late stage of disease, the anx/anx ped to the anx interval and found to be down-regulated in anx/ mice keep their eyes partly shut, possibly because of weakening anx mice. These results suggest that the anorexia and hypotha- of the eyelid muscles. Taking these data together, we hypothe- lamic neurodegeneration of the anx/anx mouse are associated sized that the starvation and neurodegeneration observed in the with dysfunction of mitochondrial complex I. anx/anx mouse could be related to defects in the mitochondrial OXPHOS (17) and possibly CI deficiency. neuropeptides | reactive oxygen species | agouti-gene related protein | In the present study, using Affymetrix microarray, we found food intake | neuroinflammation that the most prominent cellular pathways likely to be affected in Arc of the anx/anx mouse include mitochondrial function and the OXPHOS. We therefore assessed the efficiency of OXPHOS he anorectic anx/anx mouse (1) is an attractive model for dis- and CI in anx/anx mice by high-resolution respirometry. The Tturbed feeding behavior, as it exhibits phenotypes associated levels of fully and partly assembled CI were studied using native with failure to thrive in infants and young children (2), anorexia Western blots, and markers for reactive oxygen species (ROS) nervosa (3), and cachexia (4). This mouse arose by a spontaneous and oxidative stress were analyzed by histochemical methods. mutation (anorexia, allele symbol anx) and is characterized by Mapping of the interval of the anx gene and mutation revealed poor appetite and reduced stomach content, and dies (likely be- that one of the CI assembly factors, the NADH dehydrogenase cause of the severe starvation) around 3 wk after birth (1). The anx/ (ubiquinone) 1α-subcomplex, assembly factor 1 (Ndufaf1)-gene, anx mice eat significantly less than their wild-type littermates, and is located in the interval. This gene was also found to be down- by postnatal day (P) 21 they weigh half as much, rendering them an regulated in anx/anx mice, specific for the anx-allele. emaciated appearance (1). These mice also exhibit a number of neurological abnormalities, such as body tremors, head weaving, Results hyperactivity, and uncoordinated gait (1). fi anx/anx – Mitochondrial CI De ciency in the Mouse Hypothalamus. Several neurotransmitter (5, 6) and neuropeptidergic (7 11) A microarray analysis of Arc from anx/anx (n = 3) and +/+ systems involved in the regulation of food intake and energy (wild-type) (n = 3) mice revealed 132 up-regulated genes and 73 metabolism are disturbed in the anx/anx mouse. The majority of anx/anx ≥ fi down-regulated genes in mice, with 1.4-fold average these ndings are centered around the hypothalamus, the origin increase or decrease (Table S1). Through an Ingenuity Pathway of a neuronal network important for the control of initiation and termination of food intake, as well as diet-induced thermo- genesis and energy expenditure. The arcuate nucleus of hypothalamus (Arc), situated at the interface between the periphery Author contributions: C.L., I.A.K.N., P.M.G.-R., M.K., T.J.E., T.G.M.H., M.S., and J.E.J. de- and brain, receives signals about energy status from the periph- signed research; C.L., I.A.K.N., P.M.G.-R., A.R.Z., M.K., L.R.D., D.C.R., J.M., M.T.D., M.S., and J.E.J. performed research; M.U. contributed new reagents/analytic tools; C.L., I.A.K.N., ery, such as circulating leptin and insulin levels, resulting in an- P.M.G.-R., and J.E.J. analyzed data; and C.L., I.A.K.N., P.M.G.-R., T.G.M.H., M.S., and J.E.J. orexigenic or orexigenic behavior. Two main populations of food wrote the paper. intake-regulating neurons reside in the Arc, both expressing The authors declare no conflict of interest. leptin and insulin receptors (12). One group coexpresses the two orexigenic neuropeptides neuropeptide Y (NPY) and agouti- Freely available online through the PNAS open access option. gene related protein (AGRP) (11, 13), whereas the other popula- 1C.L. and I.A.K.N. contributed equally to this work. tion expresses anorexigenic proopiomelanocortin (POMC) and 2To whom correspondence may be addressed. E-mail: [email protected] or ida.nilsson@ cocaine- and amphetamine-regulated transcript (CART) (14). ki.se. Both these groups undergo degeneration in the anorexic anx/anx This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. mouse (15). By P21, anx/anx mice exhibit fewer hypothalamic 1073/pnas.1114863108/-/DCSupplemental.

18108–18113 | PNAS | November 1, 2011 | vol. 108 | no. 44 www.pnas.org/cgi/doi/10.1073/pnas.1114863108 Downloaded by guest on September 28, 2021 Analysis (IPA) of these genes, we identified the top canonical assembly pattern of the different mitochondrial complexes in pathways most likely to be involved in the phenotype of the anx/anx anx/anx hypothalamus. This analysis showed lower levels of fully mice. The pathway analysis identified several genes related to assembled CI in the anx/anx hypothalamus compared with +/+ OXPHOS pathways and mitochondrial functions (Fig. S1 and mice, revealed by reduced amounts of protein complexes of ∼700 Table S2). Genes related to mitochondrial function and oxidative kDa, with CI-specific activity (0.66 vs. 1.64, P = 0.04; n = 10) stress (e.g., Sod1, Prdx1, Bcl211, and Cox5B) are shown in Table 1. (Fig. 1 B and C). The levels of CIII were also decreased in anx/ In addition, the top canonical pathways identified included func- anx compared with +/+ hypothalamus (0.56 vs. 1.7, P = 0.02; tions, such as cell death and morphology, cellular growth, and n = 10) (Fig. 1B). However, the amount of CII, CIV, and CV proliferation and inflammatory responses (Table S2). were unchanged (CII: 0.95 vs. 1.2, P = 0.4; n = 10; CIV: 1.0 vs. Because both the phenotype and the microarray results suggest 1.2, P = 0.7; n = 6; CV: 0.85 vs. 1.25, P = 0.5; n = 4) (Fig. 1B). hypothalamic mitochondrial dysfunction in the anx/anx mouse, we The in-gel staining confirmed CI activity of the 700-kDa band and assessed mitochondrial respiration in mechanically permeabilized revealed lower activity of CI in the hypothalamus of anx/anx (n =4) hypothalami from anx/anx and +/+ mice (n =6–9 per genotype) compared with +/+ (n = 3) mice (9.7 vs. 14.4, P = 0.01) (Fig. 1C). using high-resolution respirometry (Fig. 1A). Oxygen flux normalized to hypothalamus wet weight was significantly reduced in Increased Hypothalamic ROS Production and Signs of Oxidative Stress anx/anx mice compared with +/+ littermates. The anx/anx mice in the anx/anx Mouse. CI deficiency is often accompanied by oxi- showed a 40% reduction in oxygen consumption, when mito- dative stress (18). We therefore measured hypothalamic ROS chondrial respiration was assessed at the leak state by addition of production in anx/anx and +/+ mice using intravenous dihy- CI substrates, malate and pyruvate (ADP not added). We could droethidium (DHE) injections. Microscopic analysis revealed also detect a significant reduction (29%) in CI respiration in anx/ small and dot-like red fluorescent staining, indicating ROS, to anx mice, when the oxidative phosphorylation system was evalu- a higher number in anx/anx (n = 3) (Fig. 2 A, B, and D) than in ated after the subsequent addition of ADP and glutamate (cou- +/+ (n = 3) Arc at P21 (Fig. 2C) (4.1 vs. 2.2, P < 0.05). The red pled respiration from CI), and after the addition of the exogenous fluorescent staining was localized to the area harboring the uncoupler carbonylcyanide-4-(trifluoromethoxy)-phenylhydrazone Arc AGRP/NPY and POMC/CART neurons and occasionally (FCCP) to test the capacity of the electron transfer system I appeared to be localized within neurons, as seen in Y1- (Fig. 2B) (ETSI). To discern whether the defects observed in the OXPHOS and NeuN-positive (Fig. S3) cells, as well as in microglia (Fig. 2D). were specific for CI or a more general defect of the respiratory Furthermore, immunohistochemistry (IHC) staining for su- chain, we evaluated mitochondrial respiration through CII by peroxide dismutase 2 (SOD2), a major ROS-scavenging antiox- addition of succinate. We did not find any significant differences idant in mtochondria (19), revealed increased immunoreactivity between anx/anx and +/+ mice, when oxygen consumption was in Arc of anx/anx mice at P15 and P21 (n = 4 per genotype and measured after this addition, neither in the coupled state (CIIa) age) (Fig. 2 E–G). In anx/anx, 12.5% of the Arc area was covered nor after the addition of the exogenous uncoupler FCCP (ETSII), by SOD2-ir, versus 5.0% in +/+ (P = 0.05), indicating increased suggesting a specific CI defect. In other words, if the defect ob- oxidative stress in the anx/anx Arc. On some occasions we found served in mitochondrial respiration was because of a malfunction SOD2+ cells in Arc that also expressed AGRP (Fig. 2E), al- of complex II, III, IV, or V, we should have observed a decrease in though this peptide could not be detected in the majority of oxygen consumption also after succinate addition. However, the the SOD2+ cells. We never found SOD2+ cells that convincingly increase in oxygen consumption generated by succinate addition expressed β-endorphin, a marker for Arc POMC/CART neu- was the same in anx/anx and +/+ mice, both in the coupled state rons. Occasionally, low-intensity SOD2 labeling was also seen in (CIIa) and after addition of FCCP and inhibition of CI by rotenone anx/anx mice at P10 and in +/+ mice of all ages. Preadsorption (ETSII). To ascertain that the differences in the OXPHOS system of SOD2 antiserum with the immunogenic SOD2 fragment resulted in anx/anx mice were not related to an abnormal amount of mi- in a virtually complete disappearance of immunoreactivity. tochondria, we calculated the proportions of mitochondrial (cytochrome B and cox1) vs. nuclear (cyclophyllin) DNA with SYBR IHC Analysis of Substantia Nigra in the anx/anx Mouse. As the anx/ green real-time PCR. We found no difference between anx/anx anx mice display motor disturbances, and because CI inhibition (n = 4) and +/+ (n = 4) hypothalami (Fig. S2). —as well as oxidative stress—have been implicated in the To evaluate if the decreased CI activity was caused by ab- pathogenesis and degeneration of nigral dopaminergic neurons normal amounts or assembly of CI, we performed 1D blue na- in Parkinson disease (20), we investigated the prevalence of tive-polyacrylamide gel electrophoresis (BN-PAGE) to study the neurodegenerative and neuroinflammatory processes in the sub-

Table 1. Selected mitochondria and oxidative stress-related genes showing altered expression when comparing anx/anx with +/+ mice Symbol Full gene name Fold-change

Bcl2l11 BCL2-like 11 (apoptosis facilitator) 2.1 NEUROSCIENCE Atf4 Activating transcription factor 4 (tax-responsive enhancer element B67) 2.0 Gstm5 GST mu 5 1.8 Sod1 Superoxide dismutase 1, soluble 1.7 Ndufa5 NADH dehydrogenase (ubiquinone) 1 α subcomplex, 5, 13 kDa 1.7 Cox6a1 Cytochrome c oxidase subunit VIa polypeptide 1 1.6 Ubb Ubiquitin B 1.6 Cox4i1 Cytochrome c oxidase subunit IV isoform 1 1.6 Cox5b Cytochrome c oxidase subunit Vb 1.5 Prdx1 Peroxiredoxin 1 1.5 Cox6c Cytochrome c oxidase subunit VIc 1.5 Ndufc2 NADH dehydrogenase (ubiquinone) 1, subcomplex unknown, 2, 14.5 kDa 1.5 Cox17 COX17 cytochrome c oxidase assembly homolog (Saccharomyces cerevisiae)1.4 Gpd2 Glycerol-3-phosphate dehydrogenase 2 (mitochondrial) −1.4 Map2k4 Mitogen-activated protein kinase kinase 4 −1.6 Atp5f1 ATP synthase, H+ transporting, mitochondrial F0 complex, subunit B1 −2.2 Ndufaf1 NADH dehydrogenase (ubiquinone) 1 α subcomplex, assembly factor 1 −2.2

Lindfors et al. PNAS | November 1, 2011 | vol. 108 | no. 44 | 18109 Downloaded by guest on September 28, 2021 both sides of the anx locus. The anx locus was thus mapped to a 0.2-cM interval, between markers D2Mit133 and Jojo5 (Chr 2: bp 118,889,896–120,175,108; www.ensembl.org)(Fig. S4). One recombination event in an F2 progeny from cross 2 (B6C3Fe-anx A/+ a × CAST/Ei) identified D2Mit133 as the proximal flanking marker. Three recombination events in three F2 offspring from cross 1 (B6C3Fe-anx A/+ a × B6C3H F1) identified Jojo5 as the distal flanking marker. Both flanking markers were supported by a number of recombination events on each side of the anx interval in both crosses. The anx mutation was found to cosegregate with markers D2Mit104, D2Mit395, and Jojo8 (Fig. S4). The microarray analysis of Arc from anx/anx and +/+ mice identified one of the 73 down-regulated genes as the CI assembly factor Ndufaf1 (−2.220 fold-change). This gene was also mapped to the anx gene interval (Chr 2: bp 119,481,182–119,488,563 reverse strand; www.ensembl.org), and was thus considered a strong anx-gene candidate. However, sequencing of Ndufaf1, both genomic (exons) and cDNA, revealed no unique alterations in the anx/anx mice (n = 3); thus the mutation does not appear to be located in any of the coding regions of the Ndufaf1 gene. The Ndufaf1 down-regulation was confirmed by Taqman real- time PCR. In the anx/anx brain, liver, and lung the Ndufaf1 mRNA levels were approximately half of +/+, whereas in pancreas the difference was even larger, down to 20% (n = 6 for each genotype and tissue) (Fig. 3A). No significant difference in Ndufaf1 expression was found in the heart or kidney. To determine if the reduced Ndufaf1 expression was linked to the anx allele, we analyzed the expression levels of wild type (WT) and anx allele expression of Ndufaf1 in four heterozygous mice, using a silent T/C SNP to separate the two alleles. The ratio between anx and WT alleles [(peak heightanx allele [cDNA]/ peak heightwt allele [cDNA])/(peak heightanx allele [gDNA]/ peak Fig. 1. Hypothalamic complex I deficiency and mitochondrial dysfunction in heightwt allele [gDNA])] were in all four instances lower than 1, the anx/anx mouse. Complex I respiratory capacity (A) is impaired in anx/anx indicating a lower expression of the anx allele (Table 2). On mice. Mitochondrial respiration was analyzed in hypothalamus samples average, we detected an ∼22% reduction associated with the anx P n (n =6–9 per genotype). For basal O2 flux from CI of the respiratory chain, allele compared with WT in heterozygous mice ( = 0.015; = malate and pyruvate were added to the Oxygraph-2k chamber (Leak). To 4), suggesting a 44% reduction of the Ndufaf1 transcript in ho- maximize oxidative phosphorylation from CI, ADP and glutamate were added mozygous anx/anx mice, in accordance with the twofold re- in sufficient amounts and in subsequent steps to couple electron transfer to duction detected by microarray and real-time PCR. This result ATP production (i.e., coupled respiration from CI). To ascertain O flux from thus indicates that the down-regulation of Ndufaf1 expression in 2 anx/anx anx Ndufaf1 CI and CII, succinate was added (CI+II). We calculated additional O2 flux mice is indeed associated with the allele of , through complex II, subtracting CI from CI+II (CIIa). To reach maximal ca- rather than a secondary effect of the phenotype. In addition, pacity of the electron transfer system by CI and CII, the uncoupler FCCP was Ndufaf1 and RNA polymerase II associated protein 1 (Rpap1) titrated to the chamber (ETSI+II). CI was inhibited by the addition of rote- were the only genes in the anx interval that showed altered none (ETSII). Electron transfer trough CI was calculated by subtracting ETSII expression (both down-regulated) in the microarray analysis. from ETSI+II.The remaining O flux after inhibition of CIII with antimycin A Ndufaf1 and Rpap1 are separated by four genes with no observed 2 ∼ (O2 flux independent of the electron transfer system) was subtracted from alterations in expression and 10 Mb, indicating that it is not each of the previous steps. Error bars represent SE of mean. BN analysis (B) a region-specific down-regulation. Rpap1 was not included in the revealed lower levels of fully assembled CI in the hypothalamus of anx/anx top canonical pathways, identified by IPA, and was thus regarded mice (n = 3, exemplified by lanes 2 and 4) compared with +/+ mice (n =4, unlikely to be involved in the phenotype of the anx/anx mouse. exemplified by lanes 1 and 3) and CIII, but the levels of levels of CII and CIV Western blot analysis showed that the down-regulation of the were unchanged. In-gel activity assay (C) revealed lower activity of fully Ndufaf1 mRNA is also accompanied by an ∼50% decrease (0.8 assembled CI in anx/anx mice (n = 4, exemplified by lane 2) compared with vs. 1.3, P < 0.001) of the corresponding protein level in the anx/ +/+ mice (n = 3, exemplified by lane 1). Abbreviations are listed in Table S4. anx brain (Fig. 3 B and C). The analysis was repeated three times, *P < 0.05. including three mice for each genotype, and a similar decrease was always observed. Actin was used as a loading-control and revealed no difference between samples (Fig. 3 B and C). stantia nigra (SN) of anx/anx mice. Degenerative processes in the SN can be expected to cause an increased activation of microglia, Discussion revealed by increased ionized calcium-binding adapter 1 (Iba1)- There are a number of possible explanations for the anorectic labeling of bushy-shaped microglia, as seen in several hypotha- phenotype in the anx/anx mouse (1) as well as for the aberrant lamic areas of the anx/anx mice (10). However, no obvious in- anx/anx appearance and degeneration of the arcuate food intake-regulat- crease was seen in mice after Iba1 staining. Neither did ing circuitries in this mouse (5, 7–10, 16). Here we provide evi- IHC analysis of cholecystokinin (CCK) and tyrosine hydroxylase dence for mitochondrial dysfunction in the hypothalamus of the (TH), two markers expressed in mesencephalic dopamine neu- anx/anx mouse. More specifically, we have detected a reduction in rons (21), reveal any apparent deviation in the staining pattern in the CI assembly factor Ndufaf1 expression and lower levels of fully anx/anx compared with +/+ mice (n = 2 per genotype). assembled CI, accompanied by a 29% reduction of CI-specific respiration efficiency in the anx/anx mouse. This finding is in Identification of Ndufaf1 as an anx Candidate Gene. To determine agreement with previous findings, showing that reduction of the genetic cause behind the phenotype of the anx/anx mouse, we Ndufaf1 mRNA and protein levels results in decreased CI set up two intercrosses to genetically map the anx gene and mu- assembly and activity (22). We also detected a decrease in CIII tation. The anx gene was previously mapped to ∼20 cM proximal of levels, but not in CII, CIV, and CV. The reason why CIII also show the nonagouti locus on mouse Chromosome (Chr) 2 (1). In the reduced levels in anx/anx hypothalamus is not understood. How- present study, a total of 4,844 meioses from the two different ever, there are reports on mutations affecting CI that also signif- intercrosses identified a number of recombination events on icantly reduce the amounts of CIII and CIV (23–25), which is in

18110 | www.pnas.org/cgi/doi/10.1073/pnas.1114863108 Lindfors et al. Downloaded by guest on September 28, 2021 Fig. 2. ROS and SOD2 in anx/anx hypothalamus. Micrographs show ROS staining and immunoﬂuo- rescent labeling for AGRP (A), Y1 (B and C), Iba1 (D) in Arc of anx/anx (n =3,A, B, and D) and +/+ mice (n =3,C) at P21. Immunoﬂuorescence micrographs show SOD2 (E–G) and AGRP (E and F) in Arc of an anx/anx mouse (n =4,E and G) and absence of staining in +/+ mouse (n =4,F), both at P21. Arrows in A and B indicate ROS-staining; arrowhead in D indicates possible ROS staining within Iba1+ cell. Arrowhead in E indicates a SOD2+ neuron also expressing AGRP. Asterisk in G marks a small part of a section of an adjacent brain that was mounted and cut together with the anx/anx brain. [Scale bar: 50 μminA (applies to A–C); 10 μminD;20μminF (applies to E and F); 200 μminG.] Abbreviations are listed in Table S4.

accordance with the theory that CI, CIII, and CIV form super- The OXPHOS of the mitochondria is the major site for ATP complexes to reduce the diffusion distance of the substrates, and production in cells, and CI is the largest of the five protein thereby are able to affect the stability of each other (26, 27). An- complexes (CI–CV) of this system. CI oxidizes NADH and other possible explanation is that the increased ROS production thereby releases electrons that are transferred via the electron anx/anx fl carriers ubiquinone and cytochrome c, as well as CIII and CIV, observed in the hypothalamus is in uencing the CIII fi assembly. to the nal acceptor, molecular oxygen. The energy generated by this transfer is used to build up the electrochemical gradient necessary for generation of ATP by CV (ATP synthase). In addition, CI is the main intracellular source of ROS in the mammalian cell during normal conditions, by “leaking” electrons instead of transferring them to the electron carriers. Even under normal conditions it is estimated that ∼2% to 5% of the oxygen •− consumed by mitochondria are converted to O2 (28). ROS are therefore natural byproducts of the normal oxygen metabolism and have several roles in cell and metabolic signaling (24, 25). However, ROS production can increase dramatically during different conditions, such as CI dysfunction (23), resulting in oxidative stress and thereby significant damage of cell structures, eventually contributing to cell death (29). This finding is in agreement with our findings of increased levels of ROS in the hypothalamus of the anx/anx mouse. Furthermore, we show increased SOD2 immunoreactivity specifically in Arc of the anx/anx mouse, providing indications of oxidative stress, which is possibly caused by the increased levels of ROS, as SOD2 is a major ROS- scavenging antioxidant in mitochondria (19). The microarray performed in this study also shows that several other genes related to oxidative stress responses (e.g., the ROS scavengers Sod1, Prdx1, Bcl211, and Cox5B) are up-regulated in the anx/anx mouse Arc (Table 1). Increased ROS levels have been implicated in the pathology of several diseases, such as cancer, atherosclerosis, inflammation, NEUROSCIENCE and neurodegenerative disorders, and have been shown to in- duce apoptosis or necrosis (18). However, ROS may also act as a metabolic signaling molecule, in addition to being an oxidative

Table 2. Allele-speciﬁc expression of anx and WT alleles in heterozygous mice analyzed by pyrosequencing Fig. 3. Relative expression levels of Ndufaf1 in anx/anx and +/+ mice (A), normalized against two endogenous controls (n = 6 for each genotype and anx allele/WT anx allele/WT anx allele/WT tissue). Because of the large intertissue variation, all values were also nor- Mouse ID allele (cDNA) allele (gDNA) allele (cDNA/gDNA)* malized against the +/+ mean. Exact P values (normalized against): brain (B) P = 0.07, (G) P = 0.003; pancreas (B) P = 0.03, (T) P = 0.00004; muscle (P) P = 0.003, (T) 2602 0.88 1.10 0.80 P = 0.06; lung (T) P = 0.003, (B) P = 0.008; liver (B) P =0.1;(T)P = 0.002. Error bars 2618 0.88 1.04 0.85 represent ranges, adjusted for the normalization against the +/+ mean. *P ≤ 2623 0.89 1.38 0.65 ≤ 0.05; **P 0.01. Western blot (B) revealed down-regulation of Ndufaf1 in 2626 0.87 1.08 0.81 brain of anx/anx mice (n = 9, lane 1) compared with +/+ mice (n = 9, lane 2). (C) Quantiﬁcation of Ndufaf1 Western blot. ***P ≤ 0.001. B, β-actin; G, GAPDH; P, *P = 0.0015, as calculated by one-sample t-test [mean value anx allele/WT PPIA; T, TATAbp. Abbreviations are listed in Table S4. allele (cDNA/gDNA) = 0.7775].

Lindfors et al. PNAS | November 1, 2011 | vol. 108 | no. 44 | 18111 Downloaded by guest on September 28, 2021 stressor (30). For example, ROS-signaling pathways play a sig- sitive to even a moderate CI deficiency, and more prone to se- nificant role in hypothalamic metabolic sensing of both glucose lective degeneration, during the first 3 postnatal weeks, when the and lipids (31, 32). Thus, high glucose concentrations increase anx/anx mice develop their symptoms. Because the dopaminergic hypothalamic ROS production, and ROS influence neuronal subpopulation of neurons in SN projecting to striatum also have activation of Arc neurons. Hypothalamic ROS are also essential been shown to be sensitive to CI deficiency, and in view of the for the reduced food intake observed in intralipid-injected rats motor disturbances seen in the anx/anx mouse (1), we studied SN (31). In addition, uncontrolled generation of ROS in NPY/ of anx/anx and +/+ mice by IHC. No apparent difference was AGRP neurons causes impaired firing of these neurons (33, 34). detected with any of the markers used, Iba1, CCK, and TH, but it Consequently, the increased hypothalamic ROS levels that we is possible that changes could be detected in SN with other have observed in the anx/anx mouse could be causally involved in markers and by using more sensitive approaches (e.g., stereol- the initial starvation and anorectic phenotype of this model. In ogy), or if the anx/anx mice would survive longer. It should also the longer perspective, a chronic overproduction of ROS could be mentioned that altered dopaminergic transmission has indeed lead to the signs of oxidative stress and neuronal degeneration of been detected in the striatum of the anx/anx mice (6). both the orexigenic and anorexigenic arcuate neurons that we To what extent hypothalamic neurodegeneration occurs in ro- have previously observed in the anx/anx mouse (7, 15). In addi- tenone-treated rodents remains to be explored. It is possible that tion, the canonical pathways generated in our microarray anal- the Arc neurons are less sensitive to CI-inhibition by, for example ysis suggest that cell death and inflammatory processes occur in rotenone, after the first 3 postnatal weeks, when they have dif- the hypothalamus of the anx/anx mouse. This finding is in ferentiated into adult state. Interestingly, selective atrophy of, agreement with previous studies suggesting inflammation and among other parts, the hypothalamus and the striatum was cell death in the anx/anx hypothalamus (10, 15, 35, 36). reported in a patient with a null mutation in the NDUFA12L gene, The anx locus is still unknown. However, the present study encoding another assembly factor for CI (45), indicating that suggests that the Ndufaf1 gene is equivalent to the anx gene. We the hypothalamus could indeed be sensitive to CI deficiency have mapped the anx locus to a 0.2-cM interval on Chr. 2, in humans. containing more than 40 genes, including the CI assembly factor Like the anx/anx mouse, children with CI deficiencies are often Ndufaf1. Thus, Ndufaf1 is approximately twofold down-regulated born normal and develop disease during the first year of life (38), in anx/anx mice as shown by Affymetrix microarray analysis, real- indicating that the organism is protected during the prenatal time PCR, and Western blots. Furthermore, down-regulation of period. However, there is evidence for a markedly increased Ndufaf1 expression in anx/anx mice is associated with the anx demand of OXPHOS-derived energy postnatally (45, 46). In- allele of Ndufaf1, rather than being a secondary effect of the terestingly, Dunning et al. described an 11-mo-old baby, who phenotype. Sequencing of the coding parts of the Ndufaf1 gene showed markedly reduced food intake, with two heterozygous did not identify the anx mutation, suggesting that it is located in mutations in the NDUFAF1 gene (47). These mutations resulted a regulatory element, such as the promoter-region or enhancers. in reduced CI activity, similar to what we have observed in the Unfortunately, we have not been able to identify the promoter anx/anx mouse. region of Ndufaf1, using standard in silico methods to reveal con- The process of inactivated appetite-regulating neurons and served regions among different species. In addition, this region on subsequent degeneration, observed in the anx/anx mouse, may Chr. 2 is very gene-dense, and it is possible that Ndufaf1 en- occur also in some human conditions. We hypothesize that hancers are located within or in between other genes in the in- a subclinical CI deficiency could become overt in the presence of terval, thus making it difficult to identify them and evaluate the cellular stress during, for example, severe starvation. This could effect of possible sequence alterations. Further analysis of the lead to increased oxidative stress and possibly cause irreversible 1.3-Mb anx-interval by resequencing is likely needed to identify damage of the sensitive hypothalamic neurons, resulting in per- the anx-mutation. In addition, further studies of, for example, manent damage of the appetite-regulating neuronal networks. Ndufaf1 knockout mice or transgenic rescue of anx/anx mice are Further studies on patients with such conditions are needed. needed to finally prove that Ndufaf1 is indeed the anx gene. Why the phenotypic core features of the anx/anx mouse are Materials and Methods related to brain functions remains to be answered, as well as Animals. Animal experiments were approved by the Stockholms Norra why only certain neurons seem to be affected by the Ndufaf1 down- anx/anx Ndufaf1 Djurförsöksetiska Nämnd ethical committee. Heterozygous anx breeding regulation in the mouse even though is down- pairs (B6C3Fe–a/a–anx A/+ a) were originally obtained from The Jackson regulated in the whole brain as well as other tissues. One Laboratory. For more information, see SI Materials and Methods. explanation could be related to a mutation located in a tissue/cell- specific promoter or other regulatory elements. Furthermore, various tissues—and possibly also cell-types—may have different Microarray Analysis and IPA. Total RNA was prepared, cRNA synthesized from demands on OXPHOS-derived energy, and the tissue/cell-specific anx/anx and +/+ basal hypothalamus, mainly Arc, and hybridized to an need of CI activity may therefore vary (37, 38). In addition, the Affymetrix Mouse Genome 430 2.0 Array (Affymetrix Inc.). For information development of more subtle or slowly developing symptoms in less regarding data analysis and details, see SI Materials and Methods. energy-requiring tissues/cells may be hidden behind the rapid development of more severe phenotypes in tissues/cells with high Oxygraph Measurements. Mitochondrial oxygen consumption of anx/anx and energy demand, such as brain and muscle. Interestingly, CI in- +/+ hypothalami was assessed by high-resolution respirometry (Oxygraph- hibition has long been implicated in the pathogenesis and de- 2K; Oroboros Instruments), in particular the consumption via CI and CII with generation of dopaminergic neurons of Parkinson disease (20). slight modification of protocols previously described (48). For details, see SI Thus, chronic infusion of low doses of the CI inhibitor rotenone Materials and Methods. induces selective dopaminergic neurodegeneration and Parkin- son-like symptoms in rats, although the infusion causes a uniform Mitochondria Count. Real-time PCR with SYBR green was used to quantify the CI inhibition throughout the brain (39). Moreover, selective amount of mitochondrial DNA (cytochrome B and cox1) /genomic DNA neuronal damage and gliosis, as observed in the anx/anx mouse (cyclophyllin) in hypothalami from anx/anx and +/+ mice using the Sequence (10), seem to be common features in both human disorders with CI detector system ABI-prism 7000 (Applied Biosystems Inc.). For details see SI deficiencies (38) and animal models of CI-deficiency, such as the Materials and Methods. Ndufs4 knockout mouse (40). This finding suggests that different brain areas/neurons, with different energy demands, are more or BN-PAGE and in-Gel Activity Assay. Crude hypothalamic mitochondria were less sensitive to CI deficiency and oxidative stress, possibly resulting in isolated from anx/anx and +/+ mice as previously described (49), followed by the selective neuronal damage seen in both animal models and hu- either Western blotting, to detect CI, CII, CIII, CIV, CV, and TOM40, or CI in- man patients with CI deficiencies. gel activity staining. For details, see SI Materials and Methods. During the early postnatal period the Arc neurons differenti- ate into their adult state (41–43) and they are highly active and DHE Injections. Intravenous injection of DHE (Invitrogen) was used to measure thus probably very energy demanding (44). Our hypothesis is that hypothalamic ROS production in anx/anx and +/+ mice. See SI Materials and the Arc NPY/AGRP and POMC/CART neurons are more sen- Methods for details.

18112 | www.pnas.org/cgi/doi/10.1073/pnas.1114863108 Lindfors et al. Downloaded by guest on September 28, 2021 Immunohistochemistry. IHC was performed as described previously (8). See SI from the WT-allele. DNA and cDNA of the two alleles from heterozygous Materials and Methods for detailed description and antibody information. anx/+ mice were PCR amplified over the T/C SNP and quantified using PyroSequencing (PSQ 96; Qiagen). Imbalanced expression of the two alleles Mapping of the anx Gene. Two intercrosses were set up to map the anx in- was assessed by comparing the ratio of mean peak heights from the two × terval: cross 1 (B6C3Fe-anx A/+ a B6C3H F1) and cross 2 (B6C3Fe-anx A/+ alleles in cDNA to corresponding peak heights in genomic DNA, as described a and CAST/Ei). For information on genetic markers, PCR conditions, and in SI Materials and Methods. calculation of recombination frequencies, see SI Materials and Methods. Western Blot Analysis. Protein expression of Ndufaf1 was analyzed in total Sequencing of Ndufaf1. To identify Ndufaf1 sequence alterations between brain from anx/anx and +/+ mice according to standard procedures described anx/anx and +/+ mice, genomic DNA and cDNA were prepared by standard in SI Materials and Methods. procedures, followed by PCR amplification and sequencing using ABI 3730 DNA Analyzer and BigDye Terminator v3.1 Cycle Sequencing kit (Applied ACKNOWLEDGMENTS. We thank Drs. H. Tegel and S. Hober for donation of Biosystems). For more information on DNA/RNA preparation, PCR, sequenc- superoxide dismutase 2 antiserum and peptide, Prof. L. Andersson for ing conditions, and analysis see SI Materials and Methods. valuable comments on the manuscript, and Dr. P. Pavlov for helpful advice regarding blue native-polyacrylamide gel electrophoresis. These studies Real-Time PCR. Total RNA was prepared from seven different tissues, and were supported by funding from Karolinska Institutet (KI), the Karolinska cDNA was synthesized using standard methods. Samples were analyzed for University Hospital, the AFA Insurance Company, National Institutes of expression of Ndufaf1 with Taqman real-time PCR (7900 HT Fast Real-time Health Grant P40 RR001183, The Swedish Research Council, Hjärnfonden, PCR system, Applied Biosystems Inc.) For details see SI Materials and Meth- the Swedish Medical Association, Drottning Silvias Jubileumsfond, the fol- ods. Assays are listed in Table S3. lowing foundations: Fredrik and Ingrid Thuring, Åhlén, Marianne and Mar- cus Wallenberg, Knut and Alice Wallenberg, Åke Wiberg, Magnus Bergvall, Allele Specificity. A silent T/C SNP (Ensembl gene ID ENSMUSG00000027305 Torsten och Ragnar Söderberg, and Erik and Edith Fernström, and unre- Ndufaf1 +87 T/C) was used to separate the expression level of the anx-allele stricted grants from Scandinavian Clinical Nutrition AB and Bringwell AB.

1. Maltais LJ, Lane PW, Beamer WG (1984) Anorexia, a recessive mutation causing 25. Schägger H, et al. (2004) Significance of respirasomes for the assembly/stability of starvation in preweanling mice. J Hered 75:468–472. human respiratory chain complex I. J Biol Chem 279:36349–36353. 2. Kasese-Hara M, Wright C, Drewett R (2002) Energy compensation in young children 26. Schägger H, Pfeiffer K (2000) Supercomplexes in the respiratory chains of yeast and who fail to thrive. J Child Psychol Psychiatry 43:449–456. mammalian mitochondria. EMBO J 19:1777–1783. 3. Kaye W (2008) Neurobiology of anorexia and bulimia nervosa. Physiol Behav 94: 27. Genova ML, Bianchi C, Lenaz G (2003) Structural organization of the mitochondrial 121–135. respiratory chain. Ital J Biochem 52:58–61. 4. Ramos EJ, et al. (2004) Cancer anorexia-cachexia syndrome: Cytokines and neuro- 28. Boveris A, Chance B (1973) The mitochondrial generation of hydrogen peroxide. peptides. Curr Opin Clin Nutr Metab Care 7:427–434. General properties and effect of hyperbaric oxygen. Biochem J 134:707–716. 5. Jahng JW, Houpt TA, Kim SJ, Joh TH, Son JH (1998) Neuropeptide Y mRNA and 29. Barrientos A, Moraes CT (1999) Titrating the effects of mitochondrial complex I serotonin innervation in the arcuate nucleus of anorexia mutant mice. Brain Res 790: impairment in the cell physiology. J Biol Chem 274:16188–16197. 67–73. 30. Diano S, et al. (2011) Peroxisome proliferation-associated control of reactive oxygen spe- 6. Johansen JE, et al. (2001) Altered dopaminergic transmission in the anorexic anx/anx cies sets melanocortin tone and feeding in diet-induced obesity. Nat Med 17:1121–1127. mouse striatum. Neuroreport 12:2737–2741. 31. Leloup C, et al. (2006) Mitochondrial reactive oxygen species are required for hypo- 7. Broberger C, et al. (1999) Changes in neuropeptide Y receptors and pro-opiomela- thalamic glucose sensing. Diabetes 55:2084–2090. nocortin in the anorexia (anx/anx) mouse hypothalamus. J Neurosci 19:7130–7139. 32. Benani A, et al. (2007) Role for mitochondrial reactive oxygen species in brain lipid 8. Broberger C, Johansen J, Schalling M, Hökfelt T (1997) Hypothalamic neuro- sensing: redox regulation of food intake. Diabetes 56:152–160. histochemistry of the murine anorexia (anx/anx) mutation: Altered processing of 33. Andrews ZB, et al. (2008) UCP2 mediates ghrelin’s action on NPY/AgRP neurons by neuropeptide Y in the arcuate nucleus. J Comp Neurol 387:124–135. lowering free radicals. Nature 454:846–851. 9. Johansen JE, et al. (2000) Hypothalamic CART and serum leptin levels are reduced in 34. Horvath TL, Andrews ZB, Diano S (2009) Fuel utilization by hypothalamic neurons: the anorectic (anx/anx) mouse. Brain Res Mol Brain Res 84:97–105. Roles for ROS. Trends Endocrinol Metab 20:78–87. 10. Nilsson I, Lindfors C, Fetissov SO, Hökfelt T, Johansen JE (2008) Aberrant agouti- 35. Mercader JM, et al. (2008) Hypothalamus transcriptome profile suggests an anorexia- related protein system in the hypothalamus of the anx/anx mouse is associated with cachexia syndrome in the anx/anx mouse model. Physiol Genomics 35:341–350. activation of microglia. J Comp Neurol 507:1128–1140. 36. Lachuer J, Ouyang L, Legras C, Del Rio J, Barlow C (2005) Gene expression profiling 11. Broberger C, Johansen J, Johansson C, Schalling M, Hökfelt T (1998) The neuropeptide reveals an inflammatory process in the anx/anx mutant mice. Brain Res Mol Brain Res Y/agouti gene-related protein (AGRP) brain circuitry in normal, anorectic, and mon- 139:372–376. osodium glutamate-treated mice. Proc Natl Acad Sci USA 95:15043–15048. 37. Rossignol R, et al. (2003) Mitochondrial threshold effects. Biochem J 370:751–762. 12. Schwartz MW (2006) Central nervous system regulation of food intake. Obesity (Silver 38. Distelmaier F, et al. (2009) The antioxidant Trolox restores mitochondrial membrane Spring) 14 (Suppl 1):1S–8S. potential and Ca2+ -stimulated ATP production in human complex I deficiency. J Mol 13. Hahn TM, Breininger JF, Baskin DG, Schwartz MW (1998) Coexpression of Agrp and Med (Berl) 87:515–522. NPY in fasting-activated hypothalamic neurons. Nat Neurosci 1:271–272. 39. Betarbet R, et al. (2000) Chronic systemic pesticide exposure reproduces features of 14. Vrang N, Larsen PJ, Clausen JT, Kristensen P (1999) Neurochemical characterization of Parkinson’s disease. Nat Neurosci 3:1301–1306. hypothalamic cocaine- amphetamine-regulated transcript neurons. J Neurosci 19:RC5. 40. Quintana A, Kruse SE, Kapur RP, Sanz E, Palmiter RD (2010) Complex I deficiency due 15. Nilsson IA, et al. (2011) Evidence of hypothalamic degeneration in the anorectic anx/ to loss of Ndufs4 in the brain results in progressive encephalopathy resembling Leigh anx mouse. Glia 59:45–57. syndrome. Proc Natl Acad Sci USA 107:10996–11001. 16. Fetissov SO, et al. (2005) Alterations of arcuate nucleus neuropeptidergic develop- 41. Bouret SG, Draper SJ, Simerly RB (2004) Formation of projection pathways from the

ment in contactin-deficient mice: comparison with anorexia and food-deprived mice. arcuate nucleus of the hypothalamus to hypothalamic regions implicated in the NEUROSCIENCE Eur J Neurosci 22:3217–3228. neural control of feeding behavior in mice. J Neurosci 24:2797–2805. 17. Luft R, Landau BR (1995) Mitochondrial medicine. J Intern Med 238:405–421. 42. Grove KL, Smith MS (2003) Ontogeny of the hypothalamic neuropeptide Y system. 18. Orrenius S (2007) Reactive oxygen species in mitochondria-mediated cell death. Drug Physiol Behav 79:47–63. Metab Rev 39:443–455. 43. Nilsson I, Johansen JE, Schalling M, Hökfelt T, Fetissov SO (2005) Maturation of the 19. Melov S, Coskun PE, Wallace DC (1999) Mouse models of mitochondrial disease, hypothalamic arcuate agouti-related protein system during postnatal development in oxidative stress, and senescence. Mutat Res 434:233–242. the mouse. Brain Res Dev Brain Res 155:147–154. 20. Hirsch EC, Faucheux B, Damier P, Mouatt-Prigent A, Agid Y (1997) Neuronal vulner- 44. van den Top M, Spanswick D (2006) Integration of metabolic stimuli in the hypo- ability in Parkinson’s disease. J Neural Transm Suppl 50:79–88. thalamic arcuate nucleus. Prog Brain Res 153:141–154. 21. Hökfelt T, et al. (1980) A subpopulation of mesencephalic dopamine neurons pro- 45. Smeitink J, et al. (1992) Maturation of mitochondrial and other isoenzymes of crea- jecting to limbic areas contains a cholecystokinin-like peptide: Evidence from immu- tine kinase in skeletal muscle of preterm born infants. Ann Clin Biochem 29:302–306. nohistochemistry combined with retrograde tracing. Neuroscience 5:2093–2124. 46. Minai L, et al. (2008) Mitochondrial respiratory chain complex assembly and function 22. Vogel RO, et al. (2005) Human mitochondrial complex I assembly is mediated by during human fetal development. Mol Genet Metab 94:120–126. NDUFAF1. FEBS J 272:5317–5326. 47. Dunning CJ, et al. (2007) Human CIA30 is involved in the early assembly of mito- 23. Ugalde C, Janssen RJ, van den Heuvel LP, Smeitink JA, Nijtmans LG (2004) Differences chondrial complex I and mutations in its gene cause disease. EMBO J 26:3227–3237. in assembly or stability of complex I and other mitochondrial OXPHOS complexes in 48. Kuznetsov AV, et al. (2002) Evaluation of mitochondrial respiratory function in small inherited complex I deficiency. Hum Mol Genet 13:659–667. biopsies of liver. Anal Biochem 305:186–194. 24. Grad LI, Lemire BD (2006) Riboflavin enhances the assembly of mitochondrial cyto- 49. Lapidus RG, Sokolove PM (1993) Spermine inhibition of the permeability transition of chrome c oxidase in C. elegans NADH-ubiquinone oxidoreductase mutants. Biochim isolated rat liver mitochondria: An investigation of mechanism. Arch Biochem Biophys Biophys Acta 1757:115–122. 306:246–253.

Lindfors et al. PNAS | November 1, 2011 | vol. 108 | no. 44 | 18113 Downloaded by guest on September 28, 2021