Critical Re-Evaluation of Neuroglobin Expression Reveals Conserved Patterns Among Mammals

Neuroscience 337 (2016) 339–354

CRITICAL RE-EVALUATION OF NEUROGLOBIN EXPRESSION REVEALS CONSERVED PATTERNS AMONG MAMMALS

ANDREJ FABRIZIUS, a,b DANIEL ANDRE, a a a c TILMANN LAUFS, ANNE BICKER, STEFAN REUSS, Key words: mRNA expression, Hypothalamus, RNA-Seq. ELENA PORTO, a THORSTEN BURMESTER b AND THOMAS HANKELN a* INTRODUCTION a Institute of Molecular Genetics, Johannes

Gutenberg-University Mainz, D-55099 Mainz, Germany Neuroglobin (Ngb) is an O2-binding respiratory protein b Institute of Zoology, Biocenter Grindel, University of Hamburg, (Burmester et al., 2000) predominantly expressed in the D-20146 Hamburg, Germany central and peripheral nervous system (CNS and PNS), c Department of Nuclear Medicine, University Medical Center, and is also found in some endocrine tissues (Reuss Johannes Gutenberg-University, Langenbeckstr. 1, D-55101 et al., 2002, 2016; Wystub et al., 2003; Laufs et al., Mainz, Germany 2004). Ngb is a monomer with about 150 amino acids and binds O2 with a similar affinity as myoglobin (Mb) Abstract—Neuroglobin (Ngb) is a respiratory protein that is (P50 1–2 Torr). It is a structurally highly conserved pro- almost exclusively expressed in the vertebrate nervous tein that occurs in almost all vertebrates except Agnatha system. Despite many years of research, the exact function (hagfish and lampreys) and Chondrichthyes (sharks and and even the expression sites of Ngb are still a matter of rays) (Schwarze et al., 2014; Opazo et al., 2015). Despite debate. However, to investigate hypotheses surrounding 15 years of research, the function of Ngb is still a matter of the potential roles of Ngb, a detailed knowledge of its debate, and a large variety of alternative and competing major and minor expression sites is indispensable. We hypotheses have been put forward (for reviews, see: have therefore evaluated Ngb expression by extensive Hankeln et al., 2005; Burmester and Hankeln, 2009, bioinformatic analysis using publicly available transcrip- 2014). Like other intracellular globins, Ngb may enhance tome data (RNA-Seq). During mammalian brain development, we observed low embryonic expression of Ngb O2 supply to metabolically highly active neurons mRNA and an increase after birth, arguing against a role (Burmester et al., 2000) or may be instrumental in the of Ngb in fetal hypoxia tolerance. In adult mouse brain, detoxification of reactive oxygen species (ROS) (Fordel we found highest Ngb mRNA levels in the hypothalamus, et al., 2006). Ngb may also be involved in NO metabolism where expression was up to 100-fold stronger than in cere- by detoxification of harmful NO under normoxic conditions bral cortex, cerebellum or hippocampus, as confirmed by (Brunori et al., 2005), or act as a nitrite reductase and qRT-PCR and Western blotting. High Ngb expression in generate NO under hypoxic conditions, thus preventing the hypothalamus was found conserved in humans and mitochondrial respiration in low-oxygen environment other mammals. Thus, Ngb mRNA is expressed at a basal (Tiso et al., 2011). Ngb may further interact with cyto- level in many mammalian brain regions, but shows distinc- chrome c, which is released from mitochondria upon an tive regional peaks. RNA-Seq analysis further revealed only low levels of Ngb mRNA in retina and testes and no apoptotic stimulus, thereby inhibiting the intrinsic apop- signal in standard tumor cell lines, thus raising questions totic pathway (Fago et al., 2006; Raychaudhuri et al., concerning previous studies and functional hypotheses. In 2010). It has also been proposed that Ngb interacts with conclusion, this broad-scale expression study may point the phosphatase and tensin homolog/protein kinase B to distinct Ngb functions for high- and low-expressing signaling pathway (PTEN/PI3K/AKT), implying a signaling cells and tissues and argues against a single, generic role role of Ngb in neuronal development (Li et al., 2014). of Ngb as an oxygen supplier or as an endogenous protec- Regardless of its true function(s), it has repeatedly been tant in all nerve cells. Ó 2016 Published by Elsevier Ltd on suggested that Ngb serves as a neuroprotective agent behalf of IBRO. (Sun et al., 2001; Khan et al., 2006; Greenberg et al., 2008; Raida et al., 2013). However, this hypothesis has also received criticism (Schmidt-Kastner et al., 2006; Kelsen et al., 2008; Raida et al., 2012; Di Pietro et al., 2014). Evidence has been reported that Ngb expression *Corresponding author. Address: Institute of Molecular Genetics, Johannes Gutenberg-University Mainz, J.-J. Becher-Weg 30a, levels are positively correlated with oxidative metabolism D-55099 Mainz, Germany. (Bentmann et al., 2005; Mitz et al., 2009) and hypoxia tol- E-mail address: [email protected] (T. Hankeln). erance of a species (Roesner et al., 2008; Nayak et al., Abbreviations: NGB/Ngb, human/mouse neuroglobin gene and mRNA; 2009; Avivi et al., 2010; Schneuer et al., 2012). NGB/Ngb, human/mouse neuroglobin protein; qRT-PCR, quantitative real-time reverse-transcriptase PCR; RNA-Seq, RNA sequencing; Detailed knowledge of Ngb expression is an essential RPKM, reads per kilobase exon model per million mapped reads. prerequisite for the understanding of Ngb function(s) http://dx.doi.org/10.1016/j.neuroscience.2016.07.042 0306-4522/Ó 2016 Published by Elsevier Ltd on behalf of IBRO.

339 340 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

(Hankeln et al., 2004; Burmester and Hankeln, 2014). Western blotting However, there is no consensus concerning the levels Tissues were homogenized in heated RIPA lysis buffer and patterns of Ngb expression in the mammalian CNS (10 mM Tris, 1 mM CaCl , 150 mM NaCl, 10 mM NaF, (cf., e.g., Mammen et al., 2002; Hankeln et al., 2004; 2 25 mM b-Glycerophosphate, 0.5% NP-40, 0.5% Hundahl et al., 2010). The distribution and relative expres- Deoxycholic acid, 0.1% SDS, pH 7.4). Total protein sion levels vary between different studies, probably extracts were cleared by centrifugation and protein depending on the methods applied to evaluate protein concentrations were determined by a Bradford assay and/or mRNA expression. In fact, many techniques for (Bradford, 1976). Extracts were supplemented with an detecting protein or mRNA levels are potentially error- equal volume of 2 Laemmli buffer and heat denatured prone, since they rely on indirect staining and hybridiza- at 95 °C for 5 min before they were loaded onto a 12%- tion techniques, potentially causing misleading interpreta- SDS polyacrylamide gel. Proteins were transferred onto tions. We therefore decided to evaluate (and re-assess) nitrocellulose membranes using the Mini Trans Blot Sys- the published, diverging results of Ngb expression pat- tem (Bio-Rad, Hercules, California, USA). The mem- terns and levels via a hybridization-free approach. We branes were incubated overnight with anti-Ngb serum made use of the vast and steadily growing amount of tran- diluted 1:150 in 5% non-fat dry milk in PBS-T (80 mM Na - scriptome sequencing (RNA-Seq) datasets, which pro- 2 HPO , 20 mM NaH PO , 100 mM NaCl, 0.2% Tween 20, vide gene-specific mRNA expression information via 4 2 4 pH 7.5) at 4 °C. For Western blot analysis of the develop- digital read-counting and mapping of cDNA sequences, ing mouse brain, we used anti-Ngb1, a polyclonal rabbit obtained by Illumina next-generation sequencing. More antibody raised against a peptide that corresponds to than 380 RNA-Seq data sets were analyzed bioinformat- amino acids 55–70 of human NGB (tested as described ically to determine the expression of Ngb in the CNS of in Schmidt et al., 2003, 2005; Wystub et al., 2003). Wes- mouse and man, other land-living mammals and, for com- tern analysis of mouse adult brain regions was performed parison, the zebrafish. In addition, we studied Ngb using a polyclonal rabbit anti-Ngb serum raised against a expression during embryonic development of the mouse combination of peptides that correspond to amino acids brain, and in human cancer cell lines. Selected transcrip- 55–70 and 138–151 of human NGB. The latter was suc- tome sequencing results were additionally confirmed by cessfully tested for specific Western blot binding to Ngb qRT-PCR and Western blotting. We observed a con- using tissues from our Ngb knockout mouse model (to served pattern of regionalized Ngb mRNA expression in be described elsewhere). As loading control, an anti-b- adult mammalian brains, with low expression levels in actin antibody raised against the C11-peptide (Sigma– many regions and consistent expression peaks in the Aldrich, #A2066) was used. After incubation with a sec- hypothalamus. Ngb transcription in embryonic brain ondary goat anti-rabbit antibody coupled with development and in the retina was very low. Altogether, horseradish-peroxidase (HRP) (Biozym, Hamburg, Ger- the data suggest different functions of Ngb, depending many) for 1 h at RT, detection was carried out using on its strength of expression, and argue against a single enhanced chemiluminescence ECL reagent. Signals generic role in oxygen supply or neuroprotection. were collected by exposure to an X-ray film (Kodak, Bio- Max). Quantification of Western blot signals was performed with ImageJ (Schneider et al., 2012)by EXPERIMENTAL PROCEDURES measuring the area under curve (AUC) or calculating the number of pixels for each specific signal and the load- Animals ing control. Ngb signals were normalized to either the Animals were handled according to the Directive 2010/63/ used loading controls or the unspecific band produced EU EEC for animal experiments. To quantify Ngb by the anti-Ngb serum. expression in developing mouse brain by qRT-PCR and Western blotting, total RNA and proteins were extracted Quantitative real-time RT-PCR from BALB/c embryos and adult mice. Mouse embryos were collected at days 10, 15 and 19 of pregnancy Total RNA was isolated from mouse tissues that had been (E10, E15, E19) and shortly after birth (D1). For stages snap-frozen in liquid nitrogen using the RNeasy kit E10 and E15 cranial regions were dissected. In E19 (Qiagen, Hilden, Germany). A DNase I digestion step stage embryos and adults, the brains were removed and was included according to the manufacturer’s processed individually. For the analysis of Ngb instructions. RNA concentration and integrity were expression pattern in the developing mouse CNS, a determined using the BioAnalyzer total RNA Nano Chips total of 32 samples were used (see legend Fig. 3). To (Agilent, Santa Clara, USA) or assessed photometrically analyze the Ngb expression in different regions, three in combination with denaturing formaldehyde gel adult mice (3 months old, C57BL/6 background) were electrophoresis (brain development samples). cDNA used. The brains, the retinae, and testis were dissected, synthesis was carried out with the SuperScript III the brain regions separated and all tissues were snap- reverse transcriptase (Invitrogen, Carlsbad, USA), frozen in liquid nitrogen. For all brain regions except the generally using 1 lg total RNA per reaction. qRT-PCR hypothalamus, three biological replicates were carried was carried out on the ABI Prism 7500 Fast and 7000 out. For hypothalamus, retina and testis, only one Sequence Detection Systems (SDS, Applied biological sample and three technical replicates were Biosystems, Carlsbad, USA). We used the amount of used. cDNA equivalent to 25–50 ng of total RNA in a 10 ll A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 341 reaction containing SYBR Green (GoTaq Real-Time SRX038748, SRX099867, 099868, ERX022792, Master Mix, Promega or QuantiTect SYBR Green PCR ERX022778, ERX022783, ERX022777, ERX069445, MM, Qiagen) with a final primer concentration of SRX003937, SRX003938, SRX003939, ERX004472, 0.33 lM each. The following primer combinations were SRX113365, SRX113364, ERX008572, ERX008614, used for amplification: Ngb-qPCR-Dev-For: 50-GAAG ERX016244, ERX016245, ERX022788, ERX022789, CAYCGGGCAGTG-30, Ngb-qPCR-Dev-Rev: 50-AGRCA SRX026529, SRX026530, SRX042342. In addition, we CYTCTCCAGCATGTASAG-30; Ngb-qPCR-Reg-For: 50- investigated pre-processed UCSC hg19 browser tracks GCTCAGCTCCTTCTCGACAG-30, Ngb-qPCR-Reg-Rev: (n = 16) from the ENCODE consortium which derived 50-CAACAGGCAGATCAACAGAC-30; HistH2a-qPCR- from DNaseI HS, FAIRE-Seq, and ChIP-Seq experiments For: 50-GTCGTGGCAAGCAAGGAG-30, HistH2a-qPCR- on MCF-7 cells (https://genome.ucsc.edu/). Rev: 50-GATCTCGGCCGTTAGGTACTC-30.Amplification was performed in a three-step protocol: 94 °C for 15 s, 60 °C for 30 s, 72 °C for 30 s, measuring the RESULTS fluorescence during the last step of each cycle. The Analysis of Ngb mRNA expression in mammals via mRNA levels were calculated by the standard-curve RNA-Seq approach by measuring Ct-values. Assay performance was calculated by a serial dilution of a standard-plasmid A quantitative picture of Ngb expression was obtained via containing the amplicon sequence. Ngb was quantified read-mapping of RNA-Seq transcriptome datasets using the absolute quantification method with serial (n = 381). In human tissue data obtained from the dilutions of a known standard, and copy numbers per Illumina BodyMap 2.0 project (GSE30611), NGB was 1 lg RNA were calculated. The calculations of means found expressed in brain tissue (with RPKM values of and standard deviations of the technical and biological 0.7 and 3.1), but was also detected in colon (RPKM replicates were carried out with the Microsoft Excel 0.5/1.9) and substantially weaker in testes (RPKM 0.3), spreadsheet program. prostate and lymph node (RPKM 0.1; Appendix Table A1). A better spatial resolution of human NGB expression and a higher statistical reliability (most often RNA-Seq and analysis of epigenetic modifications with >70 biological replicates per tissue) was evident in RNA-Seq transcriptome datasets freely available at data from the Genotype Tissue Expression Project Sequence Read Archives (SRA) from NCBI, EBI or (www.gtexportal.org). Strongest NGB transcription was DDJB were mapped using the CLC Genomics observed in the human hypothalamus with a median Workbench against the reference genomes of mouse RPKM of 47 (n = 96) and a peak value of 150. Further (build 38), human (build 37), zebrafish (build 5), rat NGB-expressing brain regions included frontal cortex, (build 5.0), dog (build 3.1), vole (build 1.0), pig (build cortex, anterior cingulate cortex, amygdala (median 10.2), cattle (build 3.1) and sheep (build 3.1). Only NGS RPKM values between 22 and 4) and several others data generated with Illumina technology were used. with RPKMs below 2. Additional NGB expression with Mapping was performed using the CLC Workbench RPKM >1 was observed in the pituitary gland, colon RNA-Seq analysis tool. Short reads (below 56 nt) were and esophagus. allowed two mismatches. For long reads, 95% of the In mouse, Ngb was detected in all analyzed brain read and 95% of all nucleotides were required to match regions, including the neocortex, hippocampus, the reference for the read to be included in the cerebellum and the brainstem (Fig. 1). The analysis mapping. Because of lack of further information again showed strong quantitative differences in Ngb regarding the library preparation, a paired read distance expression among different brain regions, ranging from between 1–500 nucleotides was allowed. Only uniquely RPKM values of <1 in cerebellum to >60 in the mapping reads in the genome were used for RPKM preoptic area (as part of the hypothalamus) and >160 (reads per kilobase exon model per million mapped in hypothalamus (Fig. 2A, Appendix Table A2). In reads; Mortazavi et al., 2008) calculations, thereby elimi- mouse retina, Ngb mRNA levels were low with RPKM nating repetitive sequence bias in read quantification. only in the range of two (Fig. 2A). This substantially To study expression of human NGB in MCF-7 human lower transcription of Ngb in retina versus brain was breast cancer cells with respect to its activity after confirmed in a vertebrate reference species, the estrogen treatments, the following datasets were used: zebrafish (Appendix Table A3). Ngb was also detected SRR1012917-18, SRR1012919-20, SRR1012921-22, at low levels in mouse tissues such as testis SRR1012923-24, SRR1012925-27, SRR1012928-30, (RPKM 1.2), adrenal gland (RPKM 2.1), pancreas SRR1012931-33, SRR1012934-36, SRR1012937-39, (RPKM 0.5) and lung (RPKM 2.8) (Appendix SRR1012940-42, SRR1012943-45, SRR1012946-48, Table A2). SRR1012949-51, SRR1012952-54. Datasets of To evaluate the relative expression levels, we epigenetic marks around the human NGB promoter in compared the RPKM of Ngb in mouse with those of MCF-7 human breast cancer cells, including ChIP-Seq, typical reference genes used, e.g., in qRT-PCR studies: FAIRE-Seq and ChIA-PET, were processed as b-actin (Actb), large ribosomal protein (Rplp1), pyruvate described (Bicker et al., 2015). IDs for epigenomic data- dehydrogenase phosphatase (Pdp1), and TATA box sets were: ERX008588, ERX008584, ERX008578, binding protein (Tbp). These genes were chosen ERX008597, SRX0038733, SRX038734, SRX038739, because they represent highly expressed genes SRX038740, SRX153146, SRX115153, SRX038747, important for the cytoskeleton (Actb) and the ribosome 342 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Fig. 1. Schematic overview of mouse brain regions used for Ngb expression analysis by various techniques (modiﬁed from: Heintz, 2004).

(Rplp1), respectively, and genes expressed at lower Experimental confirmation of Ngb expression levels, involved in mRNA transcription (Tbp) and patterns in mouse CNS oxidative phosphorylation (Pbp1). As expected, RPKM To confirm the quantitative differences of Ngb expression values of Actb and Rplp1 were high and in the range of in mouse brain regions as calculated by RNA-Seq, we 500–5000 (Appendix Tables A1 and A2). In contrast, performed qRT-PCR (Fig. 2B) and Western blotting RPKM values of Tbp and Pdp1 ranged between 2 and (Fig. 2C). The results of the qRT-PCR analysis matched 20. Their rather uniform and ubiquitous detection proved the Ngb expression levels as obtained from digital RNA- the sensitivity and reliability of our read-mapping Seq read-counting. In the cerebral cortex, hippocampus approach. and cerebellum Ngb was detected at comparably low Different neuronal densities (either by cell numbers or levels in the range of 104–105 transcripts/lg total RNA cell sizes) in different brain regions (Herculano-Houzel, (Fig. 2B). Ngb levels in the retina were about the same, 2014) may have a confounding effect on calculations of with 1.4 104 transcripts/lg total RNA. Further RPKM for the neuronal Ngb, since different ratios of neu- confirming the RNA-Seq results, the highest Ngb level ronal vs. glial mRNA would be sequenced. To account for was found in the hypothalamus (>106 transcripts/lg this potential bias, we analyzed the RPKM expression val- total RNA), which is roughly ten to a hundredfold higher ues of genes commonly used as cell type-specific mark- than in other brain regions (Fig. 2B). In a subsequent ers for astroglial and neuronal populations and Western blot analysis, weak levels of Ngb expression calculated the neuron/glia ratios. The expression of glial were detected at the expected position of 17 kDa in fibrillary acidic protein (Gfap) was used as an indicator mouse cerebral cortex, hippocampus and cerebellum for astroglial cells, the RNA binding protein, fox-1 homo- (Fig. 2C). In agreement with the RNA-Seq data, a strong log (Rbfox3, also known as NeuN) indicated neuronal Ngb signal was found in brain tissue, which (after gene expression. The neuron/glia ratios in the hypothala- removal of cerebral cortex, cerebellum and mus and the hippocampus were 0.8 and 0.7, respectively, hippocampus) comprised interbrain, midbrain and resulting in nearly equal neuron densities in these brainstem and included the hypothalamic and thalamic regions. The ratio in neocortex was higher with 1.8, and regions (Fig. 1, gray shading). in the cerebellum the neuron/glia ratio was calculated at 3.6, representing the highest neuron density of the analyzed regions (Fig. 2A, gray boxes). With these indirectly inferred, RNA-based neuron/glia ratios in mind, it can be Expression of Ngb during mouse development excluded that the high Ngb expression values observed The availability of RNA-Seq datasets facilitated the in the hypothalamus might be due to a dominance of neu- analysis of Ngb mRNA expression throughout mouse rons over (astro)glia. development. Read mapping (Fig. 3A) revealed very low To assess the phylogenetic conservation of Ngb expression of Ngb in embryonic stem cells (isolated at expression patterns, we analyzed Ngb levels in the stage E14) and in early embryonic brains (up to E14) brains of other mammals by exploring the publically with RPKMs <1, followed by a slight increase in available transcriptome data. In the prairie vole, dog, pig RPKMs of 1 and 2 in stages E15 and E18, respectively. and cattle, Ngb levels were higher in the hypothalamus Ngb mRNA and protein levels were also analyzed by than in any other brain region (Appendix Table A3). In qRT-PCR and Western blotting in cranial regions of the pig brain, Ngb levels in the hypothalamus ranged mouse embryos (E15, E19), brains shortly after birth from RPKM 16 to 69, whereas in the frontal cortex the (D1) and in adult brains (Fig. 3B, C). Ngb transcripts RPKM ranged only from 1.2 to 1.7. The lowest and protein were detected in all samples and increased, difference was found in sheep, which showed Ngb although not strictly proportionately in amount, from expression levels of RPKM = 6.7 in the hypothalamus early embryonic stages over later stage embryos and and RPKM = 4.7 in the cerebrum (Appendix Table A3). newborns to yield the highest values in young adults. A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 343

Fig. 2. Quantiﬁcation of Ngb mRNA and protein expression in mouse tissues. (A) Ngb mRNA expression (RPKM values) and the estimated neuron to glia (N/G) ratio (gray boxes) in selected brain regions, retina and testis as calculated from RNA-Seq experiments. Bars indicate means obtained from biological replicates, the calculated standard deviation is indicated. The N/G ratios are based on the relative expression of Gfap and Rbfox3 as (astro)glial and neuronal markers, respectively. Datasets used for calculation were: SRX186042, SRX023113, SRX023112, SRX023108, SRX023109; SRX151659, ERX012379, SRX171524, ERX012351, ERX012358, ERX012345, SRX151658; SRX216830, SRX216832, SRX147582, SRX216826, SRX186046; SRX088978, SRX103286, SRX103290, SRX103287, SRX072649, SRX103288; SRR594401, SRR594418, SRR594409. (B) Ngb expression in mouse tissues as determined via qRT-PCR. Bars for the hippocampus, cerebral cortex and cerebellum show the means from three biological replicates, the standard deviation is indicated. For hypothalamus, retina and testis the bars indicate the means of three technical replicates from one biological sample. For copy number calculation, the equivalent of 50 ng RNA per reaction was used. (C) Western blotting of Ngb protein in mouse brain regions. Diﬀerent amounts of total protein extracts of each region (n = 2, biological replicates for each region) were loaded per lane. Note the high Ngb expression in extracts containing the hypothalamus (imb) despite the lower protein amount loaded. Arrow points out a protein band of unknown identity recognized by the antibody, which was used as relative loading control. The anti-Ngb serum used for detection was tested on mouse Ngb knockout tissues to prove successful binding to the authentic Ngb band at 17 kDa (imb: interbrain + midbrain + brainstem, loaded protein amount 30 lg, co: cerebral cortex, 70 lg, hi: hippocampus, 70 lg, cb: cerebellum, 70 lg, Ngb: recombinant Ngb, 15 ng). 344 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Fig. 3. Ngb mRNA and protein expression during mouse development. (A) Digital expression values (RPKM values) for Ngb mRNA in diﬀerent mouse embryonic stages (ESC: embryonic stem cells, E11.5–E18 dissected embryonic brain tissue). RNA-Seq datasets were taken from NCBI- SRA (SRR530639, SRR567500, SRR567492, SRR567494). (B) qRT-PCR quantiﬁcation of Ngb mRNA expression in mouse embryonic, newborn and adult brain. Data were normalized to histone H2a mRNA expression as housekeeping gene The calculated Ngb/H2a ratio in embryonic stage E10 was set to 1 and the other developmental stages are shown in relation. Each bar represents the mean value ± standard deviation of at least three biological replicates (for E10 n = 6; E15 n = 5; E19 n =7;D1n = 6; Adult n = 5). (C) Ngb and b-actin protein detection via Western blotting in mouse embryonic and adult brain samples. Molecular weight is indicated. Each lane was loaded with 100 lg of whole protein extract; recombinant Ngb: 100 ng. For detection, the anti-Ngb1 and a commercial b-actin antibody (#2066, Sigma–Aldrich) were used. (D) Ngb/Actb relative protein ratios calculated from Western blots depicted above.

Ngb expression in tumor cell lines epigenetic marks indicating chromatin accessibility and bound transcription factors. Neither E2-treated (100 nM The expression of NGB protein in non-nervous tumors E2, 1 h) nor control cells featured DNaseI HS-derived and cell lines and in particular an inducibility of NGB in sequence enrichment marks that would indicate open human cancer cells after estrogen treatment (e.g. 10 nM chromatin at the NGB promoter (as defined by Wystub E2, 24 h) and ROS stress (H O ) has recently been 2 2 et al., 2004; UCSC genome browser interval reported, as documented by Western blot and RT-PCR 77.746.000–77.731.000), arguing against an accessibil- data (Emara et al., 2010; Fiocchetti et al., 2014, 2015, ity of the NGB promotor for transcription factors. Confir- 2016a,b). Trying to confirm these data suggesting ectopic mation of limited accessibility was given by FAIRE-Seq and inducible NGB expression, we investigated NGB and histone modification-specific ChIP-Seq data, which levels by RNA-Seq in the hepatocellular carcinoma line also did not show peaks in any E2-treated or control HepG2, SK-N-BE neuroblastoma cells, cervix sample around the NGB promoter. Specifically, we did carcinoma-derived HeLa cells, DLD-1 colon carcinoma not detect RNA polymerase II peaks by ChIP-Seq. Serv- cells (Appendix Table A4) and in estrogen (E2)-treated ing as a control, the generated UCSC tracks, in con- (10 nM E2, 24 h) and control human breast cancer trast, showed multiple ChIP-Seq, DNaseI and FAIRE- MCF-7 cells (not shown). Surprisingly, within these cell Seq peaks upstream of the SIAH2 gene (UCSC genome lines no (or almost no) NGB reads could be detected browser interval 150.500.000–150.445.000), the expres- (RPKMs < 0.06) within these cell lines, implying an sion of which was previously reported to be inducible by absence of NGB transcription. E2 in MCF-7 cells (Stender et al., 2007). Altogether, Based on this surprising result, we decided to these results are in possible contradiction to reports of examine the transcriptional status of the NGB ectopic NGB expression in human tumor cell lines and promoter in 34 publically available sequence datasets its estrogen-inducibility. originating from E2-incubated MCF-7 cells for A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 345

DISCUSSION somewhat different ratios and absolute RPKM values (Appendix Table A3). Conserved Ngb expression in the mammalian brain Ngb expression has been studied in much detail on In our study, we present the most comprehensive the RNA and protein level in mouse and human brain analysis of Ngb gene expression on the mRNA level to using Northern hybridization, radioactive/non-radioactive date, making use of more than 200 transcriptome mRNA in situ hybridization (ISH) and immuno-staining datasets from the mouse plus 130 datasets from other (e.g. Burmester et al., 2000; Mammen et al., 2002; species, including man, cattle, sheep and pig (Appendix Reuss et al., 2002, 2016; Wystub et al., 2003; Hundahl Table A1). In gene expression analysis, the et al., 2010, 2013). Each of these techniques has its quantification of gene transcripts in data obtained by own specificity, dynamic range and limits of detection, deep cDNA sequencing (RNA-Seq) allows a simple which partly explains the discussion whether Ngb is approximate measurement of mRNA abundance in focally or globally expressed (reviewed in Hankeln et al., tissue samples by digital read-counting (Wang et al., 2004). RNA-Seq as a method for measuring mRNA abun- 2009), thereby avoiding potential biases of other experi- dance is known to have a comparatively broad range of mental methods (introduced e.g. by antibody binding or detection (Wang et al., 2009). Our RNA-Seq analysis of nucleic acid hybridization). The calculation of an RPKM mouse and human brain regions in fact reveals both, value in RNA-Seq experiments is accompanied by the focally strong and globally lower Ngb transcript levels with normalization of the number of reads to the length of a a roughly 100-fold difference. We note that this result is transcript and the size of the dataset (Mortazavi et al., most closely reflected by Northern hybridization of human 2008). Such normalized read-count data readily allow brain RNA (Burmester et al., 2000), radioactive mRNA- the comparison of mRNA expression profiles among dif- ISH in mouse brain (Mammen et al., 2002) and ferent tissues, organisms and experimental treatments immuno-staining of mouse and human brain (Hundahl (Danielsson et al., 2015; Sudmant et al., 2015). As recom- et al., 2010, 2013), which consistently reported strongest mended by Danielsson et al. (2015), we have taken care signals in the hypothalamus region. In general, the con- to process all downloaded raw sequence data with the served regional focus of strong Ngb expression, con- same pipeline and parameters to eliminate biases from trasted by the widely observed lower-level expression, bioinformatics procedures and thus to ensure compara- raises the question of distinct Ngb functions in highly tiveness between datasets. Although more sophisticated and lowly expressing cell types (see below). read-count normalization methods are available (e.g. Lin et al., 2016), we obtained a rather consistent picture of Limited expression of Ngb mRNA in the retina Ngb mRNA abundance across many different studies, The retina was reported to be a primary Ngb-expressing replicates within these studies and data from different tissue in man, mouse (Schmidt et al., 2003; Bentmann species. Concerning data interpretation, it is clear that et al., 2005), rat (Lechauve et al., 2009) and dog mRNA levels only explain a fraction of protein abun- (Ostojic et al., 2006), as evidenced by Western blotting dances (e.g. Vogel et al., 2010). At the same time, how- and immunostaining. Retinal Ngb protein concentration ever, the presence or absence of an mRNA can well be was initially reported to be approximately 100-fold higher considered as an approximate measure for the presence than in mouse total brain, with a wide distribution of the or absence of a protein (Vogel and Marcotte, 2012), thus globin in the ganglion cell layer, the plexiform layers and guiding the design of additional experiments. inner segments of photoreceptors (Schmidt et al., 2003). The main result of this study is the highly regionalized Based on these data, it was suggested that Ngb might expression of Ngb mRNA in the mammalian brain (see serve as an O supply protein to sustain energy metabo- Fig. 1 for an overview). In mouse, Ngb transcripts are 2 lism in the visual process, analogous to the role of highly found in all regions of the brain, but there are expressed myoglobin (Mb) in the heart or skeletal mus- substantial differences in their abundances (Fig. 2A; cles (Wittenberg and Wittenberg, 2003). If the RPKM tran- Appendix Table A2). Ngb is most highly expressed in script levels inferred for retina by RNA-Seq (values of the hypothalamic region of the mouse brain (RPKM ca. 0.7–3.5 in mouse and 2.4–3.5 in cattle) correctly reflect 160), but only at approximately 30- to 100-fold lower the protein level, this would clearly argue against the levels e.g. in hippocampus, cerebral cortex and retina as the previously reported primary Ngb expression cerebellum (RPKM 0.1–4.8). Thus, we can distinguish site and would question our earlier hypothesis of an Mb- areas of rather low, basal Ngb mRNA expression found like role of Ngb in the mammalian retina. In line with this, globally in the brain from the hypothalamic region, which Ilmjarv et al. (2014) reported a lack of major alterations in clearly represents a focus of Ngb expression. This retinal gene expression of Ngb-deficient mice. Using an finding was confirmed using qRT-PCR and Western Ngb antiserum, which was validated on Ngb knockout blotting techniques (Fig. 2B, C). Note that in the latter mouse tissue, Hundahl et al. (2012) observed a rather approach, we used an Ngb antiserum, which was restricted expression of Ngb protein in the mouse retina, validated for specific binding using tissues from an Ngb i.e. in neurons of the ganglion cell and inner nuclear lay- knockout mouse. The results in mouse are further ers, which might better match the RNA-Seq results quan- confirmed by interspecies comparisons of Ngb mRNA titatively. We consider that in our earlier studies (Schmidt expression. In cattle, pig, sheep, dog and prairie vole, et al., 2003; Bentmann et al., 2005), and probably those of Ngb expression levels were consistently higher in the others as well, the binding specificity of Ngb antibodies hypothalamus than in other tissue samples, although at used (although tested for specificity by blocking with 346 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 recombinant Ngb) was insufficient. Notably however, Ngb Hankeln, 2014). In fact, there is growing consensus that is indeed expressed at still unknown intracellular levels in metazoan proteins often exhibit more than just one func- a subset of retinal neurons and functional studies aimed tion (Jeffery, 1999, 2003; Huberts and van der Klei, 2010; at its physiological role e.g. in ganglion cells (Lechauve Copley, 2012), and Ngb (like Mb) may be no exception. A et al., 2014) may still be valid. single, general function of Ngb in all nerve cells is indeed rather unlikely, considering the observed strong regional Expression of Ngb in mammalian development differences in Ngb mRNA expression levels. Instead, different Ngb functions in high- and low-expressing cell Ngb expression data during the development of types appear conceivable. Low expression levels of Ngb mammalian brain are still scarce, although such data may in fact be sufficient to sustain non-metabolic regula- may indirectly hint at the physiological role of Ngb. The tory roles, such as the potential involvement of Ngb in RNA-Seq, qRT-PCR and Western blot data suggested intracellular signaling pathways (e.g. PTEN/PI3K/AKT; Li low-level Ngb mRNA/protein expression in early stages et al., 2014). Early studies suggested a function of Ngb of mouse brain development with a slow, but steady rise in brain O2 supply and homeostasis (Burmester et al., from prenatal day E19 to newborns (D1) and further on to 2000). O2 storage and transport probably require sub- young adults. This is in line with (Hu¨ mmler et al., 2012) stantial amounts of protein in the cell (e.g. in the mM who demonstrated an incremental increase in Ngb mRNA range for Mb in muscle). The very low mRNA expression via qRT-PCR during mouse brain maturation immediately of Ngb in most mammalian brain regions (including the after birth. In humans we studied NGB expression in brain retina) evidently argues against an O2 supply role, since development as presented by transcriptome data from the Ngb RPKM values were one to three orders of magnitude BrainSpan Atlas database (www.brainspan.org). Interest- lower than, for comparison, Mb expression levels in mur- ingly, several subregions of the cerebral cortex and the ine skeletal muscle (RPKM = 500) or heart amygdaloid complex revealed low NGB transcription (RPKM = 5000). This conclusion may, however, not hold throughout fetal development and a substantial rise of for other mammals and fish, where increased Ngb expression in early infancy (4–10 months), often with a ten- expression in brain correlates with their degree of dency to slightly decline thereafter toward adulthood. hypoxia-tolerance (Roesner et al., 2008; Avivi et al., Expression in the hippocampus, cerebellar cortex and tha- 2010; Schneuer et al., 2012). In fact, in our study we lamus did not follow this time course and showed a more noted Ngb mRNA levels in total brain of the hypoxia- even developmental distribution of read-counts. tolerant zebrafish (RPKM 50), which were 10-fold higher The comparatively low Ngb expression in early than in the brain of hypoxia-sensitive mammals (mouse: development implies that Ngb is possibly not involved in RPKM 5). the known hypoxia tolerance of mammalian fetuses and Our study clearly suggests that the prime research neonates (Bickler and Donohoe, 2002). Notably, Ngb efforts should be directed toward solving the role of Ngb expression does not peak at birth time, when cellular O2 in the area of its strongest expression, the hypothalamic levels rise approximately fourfold from a relatively hypoxic region of the mammalian brain. This evolutionarily old intra-uterine milieu to the extra-uterine environment, lead- brain region controls essential vegetative body functions ing to a peak in Hif1a expression (between E15 and D1; such as the hormone status, circadian rhythms, (Madan et al., 2002), increased generation of ROS and thermoregulation, food and water intake and stress a compensatory peak in activity of selected antioxidant response (c.f. Hall, 2010). Intracellular Ngb protein levels enzymes (Khan and Black, 2003). This may indicate that within hypothalamic neurons are currently unknown, but Ngb rather does not serve a primary general role in medi- (at 1/5 of the transcriptional level of Mb mRNA in muscle) ating hypoxia tolerance by O2 binding or in ROS defense. they might still be sufficient to buffer fluctuating O2 Since hypothalamic neurogenesis predominantly occurs demands and to ensure proper O2 supply to mitochondria. between stages E10 and E16 in the developing mouse The integration of hormone and nutritional signals by brain, a primary involvement of Ngb e.g. in initial neurito- hypothalamic neurons involves the production of poten- genesis (Li et al., 2014) seems difficult to be reconciled tially damaging ROS (review: Horvath et al., 2009), and with the expression data. Instead, the comprehensive, increased oxidative stress and apoptosis in the hypothala- microarray-based gene expression profiling of mouse mus are observed in a pathological condition like diabetes hypothalamic development (Shimogori et al., 2010) con- (Baquedano et al., 2016). It therefore seems conceivable firms the substantially elevated mRNA expression of that Ngb might specifically counteract such ROS stress Ngb in the preoptic area in the time interval between birth and apoptosis in hypothalamic neurons, and several and postnatal day D21. This points to a crucial involve- molecular mechanisms have been proposed for that ment of Ngb in functional maturation of the hypothalamus, (e.g. Fago et al., 2004; Raychaudhuri et al., 2010; Tiso which includes the regulation of most essential metabolic et al., 2011). Many in vitro and in vivo studies have indeed and behavioral processes (e.g. hunger, thirst, body tem- reported a protective effect of unknown molecular basis perature, sleep, circadian rhythms, parenting, sexual by Ngb against hypoxic/ischemic damage in nerve cells maturation). and even ectopically in heart muscle (review: Dietz, 2011). Importantly, this protective action was mostly Implications of expression data for Ngb function demonstrated using Ngb-overexpressing cell culture and animal models, probably producing abundant amounts A vast number of different functional hypotheses for Ngb of the globin. The recent knockout of endogenous Ngb have been proposed in the last 15 years (Burmester and A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 347 expression in adult mice, however, did not show an overt Bicker A, Brahmer AM, Meller S, Kristiansen G, Gorr TA, Hankeln T detrimental phenotype and in fact produced the initially (2015) The distinct gene regulatory network of myoglobin in paradoxical result that Ngb KO animals had reduced tis- prostate and breast cancer. PLoS One 10:e0142662. Bickler PE, Donohoe PH (2002) Adaptive responses of vertebrate sue damage in cortical brain regions after experimental neurons to hypoxia. J Exp Biol 205:3579–3586. stroke (Raida et al., 2012). It therefore seems unlikely that Blencowe BJ, Ahmad S, Lee LJ (2009) Current-generation high- Ngb evolved to serve as a generic endogenous neuropro- throughput sequencing: deepening insights into mammalian tective agent for all types of neurons. Unfortunately, data transcriptomes. Genes Dev 23:1379–1386. are currently lacking to study if Ngb would specifically Bradford MM (1976) A rapid and sensitive method for the quantitation contribute to a protection of its major site of expression, of microgram quantities of protein utilizing the principle of protein- dye binding. Anal Biochem 72:248–254. the hypothalamus, in vivo. Brunori M, Giuffre A, Nienhaus K, Nienhaus GU, Scandurra FM, Several publications have reported the ectopic Vallone B (2005) Neuroglobin, nitric oxide, and oxygen: functional expression of Ngb mRNA and protein in non-neuronal pathways and conformational changes. Proc Natl Acad Sci U S A tumor cell lines (e.g. MCF7; Emara et al., 2010; 102:8483–8488. Fiocchetti et al., 2014, 2015, 2016a,b) and a marked Burmester T, Hankeln T (2009) What is the function of neuroglobin? J inducibility of Ngb by estrogen treatment (Fiocchetti Exp Biol 212:1423–1428. et al., 2015, 2016a). In our RNA-Seq study, however, Burmester T, Hankeln T (2014) Function and evolution of vertebrate globins. Acta Physiol (Oxf) 211:501–514. we did not find convincing evidence for the presence of Burmester T, Weich B, Reinhardt S, Hankeln T (2000) A vertebrate Ngb transcription in MCF-7 breast cancer, HepG2 liver globin expressed in the brain. Nature 407:520–523. carcinoma, HeLa cervix carcinoma and neuroblastoma Copley SD (2012) Moonlighting is mainstream: paradigm adjustment cell lines. Moreover, no estrogen (E2) inducibility of Ngb required. BioEssays 34:578–588. transcription (in fact no transcripts at all) was observed Danielsson F, James T, Gomez-Cabrero D, Huss M (2015) in a large RNA-Seq study totaling >286 mio sequence Assessing the consistency of public human tissue RNA-seq data sets. Brief Bioinform 16:941–949. reads using MCF-7 cells. Since 80–100 mio reads are Di Pietro V, Lazzarino G, Amorini AM, Tavazzi B, D’Urso S, Longo S, considered sufficient to accurately quantify >99% of all Vagnozzi R, Signoretti S, Clementi E, Giardina B, Lazzarino G, expressed transcripts at the mRNA level (Blencowe Belli A (2014) Neuroglobin expression and oxidant/antioxidant et al., 2009; Sims et al., 2014) our findings may require balance after graded traumatic brain injury in the rat. Free Radical additional confirmation of published Ngb expression and Biol Med 69:258–264. regulation data in those tumor cell lines. Dietz GP (2011) Protection by neuroglobin and cell-penetrating peptide-mediated delivery in vivo: a decade of research. Comment on Cai et al.: TAT-mediated delivery of neuroglobin protects against focal cerebral ischemia in mice. Exp Neurol 227 CONFLICT OF INTEREST (1):224–231. Exp Neurol 231: 1–10. The authors declare no conflict of interest. Emara M, Turner AR, Allalunis-Turner J (2010) Hypoxic regulation of cytoglobin and neuroglobin expression in human normal and tumor tissues. Cancer Cell Int 10:33. Fago A, Hundahl C, Malte H, Weber RE (2004) Functional properties Acknowledgments—Bioinformatical work was supported by the of neuroglobin and cytoglobin. Insights into the ancestral Johannes Gutenberg University Center for Computer Sciences physiological roles of globins. IUBMB Life 56:689–696. (CSM/SRFN). Part of this study was supported by the Deutsche Fago A, Mathews AJ, Moens L, Dewilde S, Brittain T (2006) The Forschungsgemeinschaft (DFG Ha2103/3-4). TH wishes to thank reaction of neuroglobin with potential redox protein partners Dr. Vijay Tiwari and Dr. Hyobin Jeong (Institute of Molecular Biol- cytochrome b5 and cytochrome c. FEBS Lett 580:4884–4888. ogy, Mainz, Germany) for advice and critical reading of the Fiocchetti M, Nuzzo MT, Totta P, Acconcia F, Ascenzi P, Marino M manuscript. (2014) Neuroglobin, a pro-survival player in estrogen receptor alpha-positive cancer cells. Cell Death Dis 5:e1449. Author contributions-T.H. and A.F. conceived the study. A.F., D. Fiocchetti M, Camilli G, Acconcia F, Leone S, Ascenzi P, Marino M A., A.B., T.B., E.P. and T.H. performed bioinformatical data min- (2015) ERbeta-dependent neuroglobin up-regulation impairs ing and evaluated results. A.F., T.L., D.A. and S.R. contributed 17beta-estradiol-induced apoptosis in DLD-1 colon cancer cells wet-lab data. A.F., T.H., D.A., S.R., A.B. and T.B. wrote and upon oxidative stress injury. J Steroid Biochem Mol Biol revised the manuscript. All others read and edited the final 149:128–137. version. Fiocchetti M, Cipolletti M, Leone S, Ascenzi P, Marino M (2016a) Neuroglobin overexpression induced by the 17beta-Estradiol- Estrogen receptor-alpha Pathway reduces the sensitivity of MCF- REFERENCES 7 Breast cancer cell to paclitaxel. IUBMB Life. Fiocchetti M, Cipolletti M, Leone S, Naldini A, Carraro F, Giordano D, Verde C, Ascenzi P, Marino M (2016b) Neuroglobin in breast Avivi A, Gerlach F, Joel A, Reuss S, Burmester T, Nevo E, Hankeln T cancer cells: effect of hypoxia and oxidative stress on protein level, (2010) Neuroglobin, cytoglobin, and myoglobin contribute to localization, and anti-apoptotic function. PLoS One 11:e0154959. hypoxia adaptation of the subterranean mole rat Spalax. Proc Fordel E, Thijs L, Martinet W, Lenjou M, Laufs T, Van Bockstaele D, Natl Acad Sci U S A 107:21570–21575. Moens L, Dewilde S (2006) Neuroglobin and cytoglobin Baquedano E, Burgos-Ramos E, Canelles S, Gonzalez-Rodriguez A, overexpression protects human SH-SY5Y neuroblastoma cells Chowen JA, Argente J, Barrios V, Valverde AM, Frago LM (2016) against oxidative stress-induced cell death. Neurosci Lett Increased oxidative stress and apoptosis in the hypothalamus of 410:146–151. diabetic male mice in the insulin receptor substrate-2 knockout Greenberg DA, Jin K, Khan AA (2008) Neuroglobin: an endogenous model. Dis Model Mech 9:573–583. neuroprotectant. Curr Opin Pharmacol 8:20–24. Bentmann A, Schmidt M, Reuss S, Wolfrum U, Hankeln T, Burmester Hall JE (2010) Guyton and Hall textbook of medical T (2005) Divergent distribution in vascular and avascular physiology. Philadelphia: Elsevier Health Sciences. 12, pp 711– mammalian retinae links neuroglobin to cellular respiration. J 721. Biol Chem 280:20660–20665. 348 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Hankeln T, Wystub S, Laufs T, Schmidt M, Gerlach F, Saaler- differential expression analyses using RNA-Seq data from 726 Reinhardt S, Reuss S, Burmester T (2004) The cellular and individual Drosophila melanogaster. BMC Genomics 17:28. subcellular localization of neuroglobin and cytoglobin – a clue to Madan A, Varma S, Cohen HJ (2002) Developmental stage-specific their function? IUBMB Life 56:671–679. expression of the alpha and beta subunits of the HIF-1 protein in Hankeln T, Ebner B, Fuchs C, Gerlach F, Haberkamp M, Laufs TL, the mouse and human fetus. Mol Genet Metab 75:244–249. Roesner A, Schmidt M, Weich B, Wystub S, Saaler-Reinhardt S, Mammen PP, Shelton JM, Goetsch SC, Williams SC, Richardson JA, Reuss S, Bolognesi M, De Sanctis D, Marden MC, Kiger L, Moens Garry MG, Garry DJ (2002) Neuroglobin, a novel member of the L, Dewilde S, Nevo E, Avivi A, Weber RE, Fago A, Burmester T globin family, is expressed in focal regions of the brain. J (2005) Neuroglobin and cytoglobin in search of their role in the Histochem Cytochem 50:1591–1598. vertebrate globin family. J Inorg Biochem 99:110–119. Mitz SA, Reuss S, Folkow LP, Blix AS, Ramirez JM, Hankeln T, Heintz N (2004) Gene expression nervous system atlas (GENSAT). Burmester T (2009) When the brain goes diving: glial oxidative Nat Neurosci 7:483. metabolism may confer hypoxia tolerance to the seal brain. Herculano-Houzel S (2014) The glia/neuron ratio: how it varies Neuroscience 163:552–560. uniformly across brain structures and species and what that Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B (2008) means for brain physiology and evolution. Glia 62:1377–1391. Mapping and quantifying mammalian transcriptomes by RNA- Horvath TL, Andrews ZB, Diano S (2009) Fuel utilization by Seq. Nat Methods 5:621–628. hypothalamic neurons: roles for ROS. Trends Endocrinol Metab Nayak G, Prentice HM, Milton SL (2009) Role of neuroglobin in 20:78–87. regulating reactive oxygen species in the brain of the anoxia- Huberts DH, van der Klei IJ (2010) Moonlighting proteins: an tolerant turtle Trachemys scripta. J Neurochem 110:603–612. intriguing mode of multitasking. Biochim Biophys Acta Opazo JC, Lee AP, Hoffmann FG, Toloza-Villalobos J, Burmester T, 1803:520–525. Venkatesh B, Storz JF (2015) Ancient duplications and Hu¨ mmler N, Schneider C, Giessl A, Bauer R, Walkinshaw G, expression divergence in the globin gene superfamily of Gassmann M, Rascher W, Trollmann R (2012) Acute hypoxia vertebrates: insights from the elephant shark genome and modifies regulation of neuroglobin in the neonatal mouse brain. transcriptome. Mol Biol Evol 32:1684–1694. Exp Neurol 236:112–121. Ostojic J, Sakaguchi DS, de Lathouder Y, Hargrove MS, Trent 3rd JT, Hundahl CA, Allen GC, Hannibal J, Kjaer K, Rehfeld JF, Dewilde S, Kwon YH, Kardon RH, Kuehn MH, Betts DM, Grozdanic S (2006) Nyengaard JR, Kelsen J, Hay-Schmidt A (2010) Anatomical Neuroglobin and cytoglobin: oxygen-binding proteins in retinal characterization of cytoglobin and neuroglobin mRNA and protein neurons. Invest Ophthalmol Vis Sci 47:1016–1023. expression in the mouse brain. Brain Res 1331:58–73. Raida Z, Hundahl CA, Kelsen J, Nyengaard JR, Hay-Schmidt A Hundahl CA, Fahrenkrug J, Luuk H, Hay-Schmidt A, Hannibal J (2012) Reduced infarct size in neuroglobin-null mice after (2012) Restricted expression of Neuroglobin in the mouse retina experimental stroke in vivo. Exp Transl Stroke Med 4:15. and co-localization with Melanopsin and Tyrosine Hydroxylase. Raida Z, Hundahl CA, Nyengaard JR, Hay-Schmidt A (2013) Biochem Biophys Res Commun 425:100–106. Neuroglobin over expressing mice: expression pattern and Hundahl CA, Kelsen J, Hay-Schmidt A (2013) Neuroglobin and effect on brain ischemic infarct size. PLoS ONE 8:e76565. Cytoglobin expression in the human brain. Brain Struct Funct Raychaudhuri S, Skommer J, Henty K, Birch N, Brittain T (2010) 218:603–609. Neuroglobin protects nerve cells from apoptosis by inhibiting the Ilmjarv S, Reimets R, Hundahl CA, Luuk H (2014) Effect of light on intrinsic pathway of cell death. Apoptosis 15:401–411. global gene expression in the neuroglobin-deficient mouse retina. Reuss S, Saaler-Reinhardt S, Weich B, Wystub S, Reuss MH, Biomed Rep 2:780–786. Burmester T, Hankeln T (2002) Expression analysis of Jeffery CJ (1999) Moonlighting proteins. Trends Biochem Sci neuroglobin mRNA in rodent tissues. Neuroscience 115:645–656. 24:8–11. Reuss S, Banica O, Elgurt M, Mitz S, Disque-Kaiser U, Riemann R, Jeffery CJ (2003) Multifunctional proteins: examples of gene sharing. Hill M, Jaquish DV, Koehrn FJ, Burmester T, Hankeln T, Woolf NK Ann Med 35:28–35. (2016) Neuroglobin expression in the mammalian auditory Kelsen J, Hundahl CA, Hay-Schmidt A (2008) Neuroglobin: system. Mol Neurobiol 53:1461–1477. endogenous neuroprotectant or maintenance of homeostasis? Roesner A, Mitz SA, Hankeln T, Burmester T (2008) Globins and Stroke 39:e177–e178. author reply e179. hypoxia adaptation in the goldfish, Carassius auratus. FEBS J Khan JY, Black SM (2003) Developmental changes in murine brain 275:3633–3643. antioxidant enzymes. Pediatr Res 54:77–82. Schmidt M, Giessl A, Laufs T, Hankeln T, Wolfrum U, Burmester T Khan AA, Wang Y, Sun Y, Mao XO, Xie L, Miles E, Graboski J, Chen (2003) How does the eye breathe? Evidence for neuroglobin- S, Ellerby LM, Jin K, Greenberg DA (2006) Neuroglobin- mediated oxygen supply in the mammalian retina. J Biol Chem overexpressing transgenic mice are resistant to cerebral and 278:1932–1935. myocardial ischemia. Proc Natl Acad Sci U S A Schmidt M, Laufs T, Reuss S, Hankeln T, Burmester T (2005) 103:17944–17948. Divergent distribution of cytoglobin and neuroglobin in the murine Laufs TL, Wystub S, Reuss S, Burmester T, Saaler-Reinhardt S, eye. Neurosci Lett 374:207–211. Hankeln T (2004) Neuron-specific expression of neuroglobin in Schmidt-Kastner R, Haberkamp M, Schmitz C, Hankeln T, Burmester mammals. Neurosci Lett 362:83–86. T (2006) Neuroglobin mRNA expression after transient global Lechauve C, Rezaei H, Celier C, Kiger L, Corral-Debrinski M, brain ischemia and prolonged hypoxia in cell culture. Brain Res Noinville S, Chauvierre C, Hamdane D, Pato C, Marden MC 1103:173–180. (2009) Neuroglobin and prion cellular localization: investigation of Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to a potential interaction. J Mol Biol 388:968–977. ImageJ: 25 years of image analysis. Nat Methods 9:671–675. Lechauve C, Augustin S, Cwerman-Thibault H, Reboussin E, Schneuer M, Flachsbarth S, Czech-Damal NU, Folkow LP, Siebert U, Roussel D, Lai-Kuen R, Saubamea B, Sahel JA, Debeir T, Burmester T (2012) Neuroglobin of seals and whales: evidence Corral-Debrinski M (2014) Neuroglobin gene therapy prevents for a divergent role in the diving brain. Neuroscience 223:35–44. optic atrophy and preserves durably visual function in Harlequin Schwarze K, Campbell KL, Hankeln T, Storz JF, Hoffmann FG, mice. Mol Ther 22:1096–1109. Burmester T (2014) The globin gene repertoire of lampreys: Li L, Liu QR, Xiong XX, Liu JM, Lai XJ, Cheng C, Pan F, Chen Y, Yu convergent evolution of hemoglobin and myoglobin in jawed and SB, Yu AC, Chen XQ (2014) Neuroglobin promotes neurite jawless vertebrates. Mol Biol Evol 31:2708–2721. outgrowth via differential binding to PTEN and Akt. Mol Neurobiol Shimogori T, Lee DA, Miranda-Angulo A, Yang Y, Wang H, Jiang L, 49:149–162. Yoshida AC, Kataoka A, Mashiko H, Avetisyan M, Qi L, Qian J, Lin Y, Golovnina K, Chen ZX, Lee HN, Negron YL, Sultana H, Oliver Blackshaw S (2010) A genomic atlas of mouse hypothalamic B, Harbison ST (2016) Comparison of normalization and development. Nat Neurosci 13:767–775. A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 349

Sims D, Sudbery I, Ilott NE, Heger A, Ponting CP (2014) Sequencing signatures and mRNA concentration can explain two-thirds of depth and coverage: key considerations in genomic analyses. Nat protein abundance variation in a human cell line. Mol Syst Biol Rev Genet 15:121–132. 6:400. Stender JD, Frasor J, Komm B, Chang KC, Kraus WL, Wang Z, Gerstein M, Snyder M (2009) RNA-Seq: a revolutionary tool Katzenellenbogen BS (2007) Estrogen-regulated gene networks for transcriptomics. Nat Rev Genet 10:57–63. in human breast cancer cells: involvement of E2F1 in the Wittenberg JB, Wittenberg BA (2003) Myoglobin function regulation of cell proliferation. Mol Endocrinol 21:2112–2123. reassessed. J Exp Biol 206:2011–2020. Sudmant PH, Alexis MS, Burge CB (2015) Meta-analysis of RNA-seq Wystub S, Laufs T, Schmidt M, Burmester T, Maas U, Saaler- expression data across species, tissues and studies. Genome Reinhardt S, Hankeln T, Reuss S (2003) Localization of Biol 16:287. neuroglobin protein in the mouse brain. Neurosci Lett Sun Y, Jin K, Mao XO, Zhu Y, Greenberg DA (2001) Neuroglobin is 346:114–116. up-regulated by and protects neurons from hypoxic-ischemic Wystub S, Ebner B, Fuchs C, Weich B, Burmester T, Hankeln T injury. Proc Natl Acad Sci U S A 98:15306–15311. (2004) Interspecies comparison of neuroglobin, cytoglobin and Tiso M, Tejero J, Basu S, Azarov I, Wang X, Simplaceanu V, Frizzell myoglobin: sequence evolution and candidate regulatory S, Jayaraman T, Geary L, Shapiro C, Ho C, Shiva S, Kim-Shapiro elements. Cytogenet Genome Res 105:65–78. DB, Gladwin MT (2011) Human neuroglobin functions as a redox- regulated nitrite reductase. J Biol Chem 286:18277–18289. Vogel C, Marcotte EM (2012) Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat Rev Genet 13:227–232. APPENDIX A Vogel C, Abreu Rde S, Ko D, Le SY, Shapiro BA, Burns SC, Sandhu D, Boutz DR, Marcotte EM, Penalva LO (2010) Sequence

Table A1. RPKM values in human tissues

SRA-ID Tissue Uniquely mapped Reads total % mapped NGB ACTB RPLP1 PDP1 TBP

ERP000546 Colon 41,822,940 82,437,443 51 1.9 3118.3 986.0 3.5 2.9 41,066,840 80,257,757 51 0.5 5397.7 1571.3 6.1 4.5 Heart 38,584,211 82,918,784 47 0.0 287.2 406.3 12.6 2.8 35,942,679 76,766,862 47 0.0 540.2 621.7 26.1 5.2 Kidney 38,585,835 80,397,337 48 0.0 1051.2 512.8 9.1 5.9 40,340,226 79,772,393 51 0.0 1800.4 815.8 15.5 9.6 Liver 63,723,795 80,048,623 80 0.0 759.9 576.4 0.9 3.1 50,311,697 77,453,877 65 0.0 1012.6 642.3 1.6 4.0 Lung 50,056,678 79,296,905 63 0.0 3103.8 1545.1 4.3 1.8 51,817,959 81,255,438 64 0.0 4349.8 1952.4 5.9 2.4 Lymph node 43,321,329 82,078,157 53 0.1 1638.3 2073.6 6.3 5.2 43,910,590 81,916,460 54 0.0 2492.6 2879.4 10.0 7.9 Ovary 50,187,395 80,946,260 62 0.0 2363.7 1106.6 10.3 5.4 51,700,794 81,003,052 64 0.0 3292.9 1414.2 14.4 7.9 Prostate 60,043,051 82,334,076 73 0.1 3687.8 1714.0 7.5 4.4 53,660,229 83,319,902 64 0.0 5064.0 2130.5 10.7 7.3 Skeletal muscle 45,840,195 82,111,139 56 0.0 363.8 3165.2 6.1 2.4 46,954,871 82,864,636 57 0.0 580.7 4573.3 9.6 3.6 Testes 57,921,192 81,836,199 71 0.3 1614.7 734.3 14.9 19.2 62,520,093 82,044,319 76 0.3 1801.5 762.1 16.6 23.3 Thyroid 50,948,924 81,912,887 62 0.0 1011.0 523.5 16.2 6.0 52,616,953 80,246,657 66 0.0 1361.5 665.6 21.8 9.9 White blood cells 53,386,610 81,217,148 66 0.0 4757.8 1913.9 17.9 6.2 54,604,119 82,785,673 66 0.0 6560.4 2300.9 24.4 8.6 Adipose 41,473,621 77,300,072 54 0.0 2577.9 755.0 5.0 3.1 42,081,425 76,269,225 55 0.0 4195.4 1140.4 8.5 4.2 Adrenal 42,487,742 74,472,871 57 0.0 2655.1 1388.2 5.8 7.5 44,587,275 76,171,569 59 0.0 3953.3 1901.2 8.6 10.4 Brain 42,517,660 73,513,047 58 3.1 1239.4 169.0 43.5 4.2 39,415,994 64,313,204 61 0.7 1710.6 212.0 64.8 5.7 Breast 41,731,824 75,862,215 55 0.0 1669.5 562.5 8.9 4.4 44,261,533 77,195,260 57 0.0 2617.0 801.3 13.0 7.0 SRP007461 H1-hESC 28,713,176 41,419,823 69 0.3 3573.3 9.2 1792.6 2.9 Neuroblastoma 21,844,741 40,386,822 54 0.0 1513.9 10.9 757.4 14.4

The RPKM values for neuroglobin (NBG), b-actin (ACTB), large ribosomal protein (RPLP1), pyruvate dehydrogenase phosphatase (PDP1) and TATA box binding protein (TBP) in diﬀerent tissues and cell culture extracts were calculated. For quality purposes the number of total reads versus uniquely mapped reads is indicated for each analyzed dataset. 350 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Table A2. RPKM values of mouse tissues A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 351

The RPKM values for neuroglobin (Ngb), b-actin (Actb), large ribosomal protein (Rplp1), pyruvate dehydrogenase phosphatase (Pdp1) and TATA box binding protein (Tbp) in diﬀerent mouse tissues and cell culture extracts were calculated. Datasets with Ngb RPKM >5 are arbitrarily highlighted in gray. For quality purposes the number of total reads versus uniquely mapped reads is indicated for each analyzed dataset. 352 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Table A3. RPKM values for other species A. Fabrizius et al. / Neuroscience 337 (2016) 339–354 353

The RPKM values for neuroglobin (Ngb), b-actin (Actb), large ribosomal protein (Rplp1), pyruvate dehydrogenase phosphatase (Pdp1) and TATA box binding protein (Tbp) in diﬀerent tissues of rat, zebraﬁsh, dog, prairie vole, sheep, pig and cattle were calculated. Datasets with Ngb-RPKM >5 are arbitrarily highlighted in gray. For quality check purposes, the number of total reads versus uniquely mapped reads is indicated for each analyzed dataset. 354 A. Fabrizius et al. / Neuroscience 337 (2016) 339–354

Table A4. RPKM values for NGB in human tumor cell lines

Cell line SRA ID RPKM NGB # mapped reads % mapped

MCF-7 SRR521521 0 40,682,030 51.0 MCF-7 SRR534294 0 176,883,677 62.4 MCF-7 SRR534308 0 206,049,370 65.5 MCF-7 SRR629578 0 17,506,775 62.5 MCF-7 SRR629577 0 12,409,009 63.5 HepG2 SRR629576 0 18,612,234 56.0 HepG2 SRR629576 0 18,221,291 54.8 HepG2 SRR629573 0 15,313,216 55.1 HepG2 SRR629573 0 14,991,092 54.0 HepG2 SRR1107930 0 19,609,922 56.6 HeLa SRR065506 0 21,437,296 44.9 HeLa SRR629571 0 17,415,953 57.4 HeLa SRR629572 0 15,511,218 37.2 HeLa SRR867736 0 18,799,835 49.8 HeLa SRR922112 0 15,656,212 51.7 HeLa S3 SRR315334 0.02 50,463,543 64.0 SK-N-SH SRR534310 0 136,384,022 68.4 SK-N-SH SRR534309 0 196,074,153 69.9 SK-N-SH RA SRR315315 0 40,433,512 50.1 SK-N-SH RA SRR315316 0.05 44,859,361 58.1 SK-N-BE SRR1177127 0.02 35,244,020 59.4 SK-N-BE SRR1177128 0.01 36,241,670 59.0

Short Read Archive identiﬁcation numbers for datasets are given (SRA ID). For quality purposes, the number of total reads versus uniquely mapped reads is indicated for each analyzed data.

(Accepted 26 July 2016) (Available online 16 August 2016)