Address correspondence and reprint requests to David A. Greenberg, Buck Institute for Research on Aging, Novato, CA 94945, USA. E-mail:email@example.com
Neuroglobin is a hypoxia-inducible O2-binding protein with neuroprotective effects in cell and animal models of stroke and Alzheimer’s disease. The mechanism underlying neuroglobin’s cytoprotective action is unknown, although several possibilities have been proposed, including anti-oxidative and anti-apoptotic effects. We used affinity purification–mass spectrometry methods to identify neuroglobin-interacting proteins in normoxic and hypoxic murine neuronal (HN33) cell lysates, and to compare these interactions with those of a structurally and functionally related protein, myoglobin. We report that the protein interactomes of neuroglobin and myoglobin overlap substantially and are modified by hypoxia. In addition, neuroglobin-interacting proteins include partners consistent with both anti-oxidative and anti-apoptotic functions, as well as with a relationship to several neurodegenerative diseases.
Neuroglobin (Ngb) and myoglobin (Mgb) are ancestrally related hemoproteins with a variety of shared features, including ∼20% amino acid sequence homology (Burmester et al. 2000). Both occur as monomers, bind O2 with high affinity, scavenge NO, contain hypoxia-inducible mRNA stabilization signals in their 3′-untranslated regions, are expressed preferentially in particular (neuronal and skeletal muscle) tissues and are cytoprotective (Ordway 2004; Greenberg et al. 2008; Yu et al. 2009). These parallels have led to the suggestion that Ngb and Mgb may have similar functions, including O2 storage or transport and hypoxic adaptation (Brunori and Vallone 2006).
One approach for investigating Ngb function and exploring its relationship to that of Mgb is to examine protein–protein interactions (Droit et al. 2005). A small number of studies to date have reported interactions between Ngb and other proteins. Yeast two-hybrid screens were used to demonstrate as interactors the Na/K ATPase β1, β2, and β3 subunits; flotillin-1; cytochrome c1; ubiquitin C; disheveled homolog 1; synaptotagmin I; electron-transferring flavoprotein α subunit; GABAA receptor-associated protein-like 1; microtubule-associated protein 1A; and voltage-dependent anion channel 1 (Xu et al. 2003; Wakasugi et al. 2004a; Yu et al. 2012). Surface plasmon resonance identified the guanine nucleotide-binding protein Gαi subunit (Wakasugi et al. 2003), cystatin C (Wakasugi et al. 2004b), and prion protein (Palladino et al. 2011). Co-immunoprecipitation showed interaction between Ngb and 14-3-3 (Jayaraman et al. 2011). Finally, redox reactions with Ngb have been documented for cytochromes b5 and c (Fago et al. 2006) and thioredoxin reductase (Trandafir et al. 2007). Of these Ngb-interacting proteins, only cytochromes b5 (Liang et al. 2002) and c (Wu et al. 1972) have also been shown to interact with Mgb.
To further investigate Ngb protein–protein interactions and compare them to interactions with Mgb, we incubated normoxic or hypoxic murine neuronal (HN33) (mouse hippocampal neuron × N18TG2 neuroblastoma) cell lysates with recombinant mouse Ngb or horse Mgb, isolated Ngb- or Mgb-immunoreactive complexes, and analyzed the Ngb and Mgb interactomes using affinity purification–mass spectrometry.
Materials and methods
Dulbecco’s modified Eagle’s medium with L-glutamine, phosphate-buffered saline (PBS), culture dishes, protease inhibitor, cobalt purification columns, biotin buffer, biotin labeling kit, monomeric streptavidin columns, Zeba spin columns, biotin quantification assay, dialysis cassette, and protein concentrator were purchased from Thermo Scientific (Rockford, IL, USA). Fetal bovine serum, SDS (sodium dodecyl sulfate)-PAGE (polyacrylamide gel electrophoresis) gels, dithiothreitol, pET expression vector, BL21 competent cells, SDS western blot loading buffer, 0.05% Trypsin-EDTA, penicillin, and streptomycin were purchased from Invitrogen (Carlsbad, CA, USA). Horse heart myoglobin, Luria broth, imidazole, acetonitrile, ammonium bicarbonate, formic acid, and isopropyl-β-D-thiogalactopyranoside were purchased from Sigma (St. Louis, MO, USA). Trypsin was purchased from Promega (Madison, WI, USA).
HN33 cells, provided originally by Dr. Bruce H. Wainer, were plated at 80% confluency on 100-mm culture dishes and maintained at 37°C in humidified 5% CO2/95% air, in Dulbecco’s modified Eagle’s medium supplemented with heat-inactivated fetal bovine serum (10%) and 50 U/mL penicillin/streptomycin, as described previously (Haines et al. 2012). For hypoxia treatment, cells were placed in a chamber flushed with 95% N2/5% CO2 for 16 h at 37°C.
Recombinant protein expression
Full-length mouse Ngb cDNA was cloned into a pET protein expression vector and transformed into BL21 cells. Fresh Luria broth medium was inoculated with a starter culture grown the day before. Around OD 0.4–0.6, protein expression was induced with 1 mM isopropyl-β-D-thiogalactopyranoside. After 6 h of expression, bacteria were pelleted, frozen at −80°C, reconstituted with buffer containing 20 mM Tris/10 mM NaCl, sonicated for 30 s, and cleared by centrifugation at 1000 g for 5 min. Lysates were incubated with cobalt spin columns overnight on a rocker at 4°C. Protein was washed and later eluted using increasing concentrations of imidazole. Recombinant Ngb was then dialyzed in PBS overnight.
Protein preparation, binding, and pulldown
Recombinant mouse Ngb and horse heart Mgb were biotin-labeled using an EZ-Link Sulfo-NHS-Biotinylation Kit (Thermo Scientific) and the remaining free biotin was removed using a Zeba spin column. Biotin labeling was confirmed using a fluorescent biotin quantification kit (Thermo Scientific). Protein concentration was determined by the bicinchoninic acid protein assay (Biorad, Hercules, CA, USA).
Sixteen 100-mm dishes of HN33 cells were scraped and pelleted for each (hypoxia or normoxia) condition. Cells were lysed with radioimmunoprecipitation assay buffer and protease inhibitors and rocked at 4°C for 2 h; lysates were cleared by centrifugation at 1000 g for 5 min and later dialyzed in PBS with protease inhibitors.
Biotin-labeled Ngb or Mgb (1 mg) was incubated with 8 mg of total protein from hypoxic or normoxic cell lysates for 2 h at 37°C. The combined protein mixtures were passed over a reversibly bindable monomeric streptavidin column (Thermo Scientific), washed with PBS to remove unbound protein, and eluted with 2 mM biotin in 2-mL fractions. Samples were measured at 280 nm using a Nano-drop spectrophotometer (Thermo Scientific) to determine the fractions that contained the eluted protein, and fractions containing protein were pulled and concentrated to < 250 μL. For each condition, 200 μg of protein was denatured with SDS and dithiothreitol, incubated at 70°C for 10 min, loaded into four separate wells of a 12% SDS-PAGE gel, and run at 200 V for 25 min. A separate gel and running buffer was used for each condition. The gel was incubated with SYPRO Ruby stain (Life Technologies, Grand Island, NY, USA) and illuminated with UV light, and 12 sections were cut out corresponding to size.
In-gel protein digestion
Protein digestion, liquid chromatography (LC)/mass spectroscopy (MS), and database searching were performed by the Keck Biotechnology Resource at Yale University (New Haven, CT, USA). Additional data analysis was provided by the Buck Institute’s Bioinformatics Core. Proteins isolated in SDS-PAGE gels were subjected to in situ enzymatic digestion. Gel slices were washed with 250 μL of 50% acetonitrile/50% water for 5 min, followed by 250 μL of 50 mM ammonium bicarbonate in 50% acetonitrile/50% water for 30 min. A final wash was done using 10 mM ammonium bicarbonate in 50% acetonitrile/50% water for 30 min. After washing, gels were dried in a Speedvac and rehydrated with 0.1 μg of modified trypsin per ∼15 mm3 of gel in 15 μL of 10 mM ammonium bicarbonate. Samples were digested at 37°C for 16 h.
LC–MS/MS was performed using an LTQ Orbitrap (Thermo Scientific) equipped with a Waters nanoAcquity UPLC system (Waters, Milford, MA, USA), Waters Symmetry® C18 180 μm × 20 mm trap column, and a 1.7-μm, 75 μm × 250 mm nanoAcquity™ UPLC™ column (35°C) for peptide separation. Trapping was done at 15 μL/min, in 99% Buffer A (100% water/0.1% formic acid) for 1 min. Peptide separation was performed at 300 nL/min with Buffer A and Buffer B (100% acetonitrile/0.075% formic acid). A linear gradient (51 min) was run with 5% buffer B under initial conditions, 50% buffer B at 50 min, and 85% buffer B at 51 min. MS was acquired in the Orbitrap using one microscan, and a maximum inject time of 900, followed by four data-dependent MS/MS acquisitions in the ion trap. For phosphopeptide analysis, Neutral loss scans (MS3) were also obtained for 98.0, 49.0, and 32.7 atomic mass units.
MS/MS spectra were searched using the Mascot algorithm (version 2.2.0) for uninterpreted MS/MS spectra, after using the Mascot Distiller program to generate Mascot-compatible files. The Mascot Distiller program combines sequential MS/MS scans from profile data that have the same precursor ion; charge states of +2 and +3 are located preferentially with a signal : noise ratio of ≥ 1.2, and a peak list is generated for database searching. A protein was considered to have been identified when Mascot listed it as significant and when more than two peptides matched the same protein. The NCBI non-redundant (NR) database was used. The Mascot significance score match is based on a MOWSE score and relies on multiple matches to more than one peptide from the same protein. Typical parameters used for searching are partial methionine oxidation and propionamide (cysteine modification), a peptide tolerance of +20 ppm, MS/MS fragment tolerance of +0.6 Da, and peptide charges of +2 or +3. Normal and decoy database searches are run.
Proteins identified with an expectation score of > 0.01 were excluded. Proteins listed in the common Repository of Adventitious Proteins (cRAP, http://www.thegpm.org/crap/index.html) were excluded with two exceptions—Cyc and Thio—which were previously reported Ngb interactors and did not appear in all four data sets (Ngb/normoxia, Ngb/hypoxia, Mgb/normoxia, Mgb/hypoxia).
A total of 331 proteins (Table S1) were identified as interactors with Ngb or Mgb based on the presence of more than one peptide fragment (Fig. 1). Of these, 216 (65%) interacted with Ngb, 299 (90%) with Mgb, and 184 (56%) with both; 277 interactors (84%) were detected under normoxic and 248 (75%) under hypoxic conditions. Viewed alternatively, of 277 proteins identified in normoxic cells, 3% interacted only with Ngb, 32% only with Mgb, and 65% with both; of 248 proteins detected in hypoxic cells, 10% were exclusively Ngb interactors, 20% were exclusively Mgb interactors, and the remaining 70% interacted with both (Fig. 2).
Among Ngb-interacting proteins, we were interested to observe whether our sample included previously reported interactors or related proteins. Ngb has guanine nucleotide dissociation inhibitor (GDI) activity (Wakasugi et al. 2003), and proteins associated with Ngb under normoxic conditions included two GDIs: Rho GDIα (Gdir1) and translationally controlled tumor protein (Tctp). Ngb is a redox partner of cytochrome c (Fago et al. 2006), and both cytochrome c (Cyc) and cytochrome c2 (Cyc2), which shares 86% amino acid homology with Cyc, associated with Ngb under hypoxic conditions. Ngb is reduced by thioredoxin reductase (Trandafir et al. 2007), and interacted with thioredoxin reductase (Thio) under hypoxic and with thioredoxin-like protein 1 (Txnl1) under both normoxic and hypoxic conditions. Ngb co-immunoprecipitates with 14-3-3 (Jayaraman et al. 2011), and several 14-3-3 isoforms were detected as Ngb interactors in normoxic and hypoxic cells.
Because Ngb levels are increased by hypoxia (Sun et al. 2001) via hypoxia-inducible factor-1 (Hif1) (Haines et al. 2012), we looked for other hypoxia-inducible proteins among the interactors we identified. Examples found included the Hif1 transcriptional targets, enolase 1 (Enoa) and lactate dehydrogenase A (Ldha) (Semenza et al. 1996), under hypoxic and both normoxic and hypoxic conditions, respectively. Another interactor was prefoldin subunit 3 (Pfd3), a protein that binds to von Hippel-Lindau (Vhl), which promotes the proteasomal degradation of the Ngb inducer, Hif1 (Maxwell et al. 1999).
The neuroprotective effects of Ngb have been attributed in some reports to anti-apoptotic effects mediated through binding to cytochrome c (Cyc) or apoptosis-inducing factor (Aif) (Fago et al. 2006). Cyc, Cyc2, and Aif (Aifm) were all identified as Ngb interactors under hypoxic conditions.
Another hypothesis regarding the mechanism of Ngb’s cytoprotective action involves attenuation of oxidative injury, based on the ability of Ngb to protect against injury induced by, e.g., H2O2 (Fordel et al. 2006; Li et al. 2008) or NO (Brunori et al. 2005; Jin et al. 2008). Ngb interacted with the anti-oxidant enzyme Thio in hypoxic cells and with the mitochondrial anti-oxidant thioredoxin-dependent peroxide reductases (peroxiredoxins), Prdx3, Prdx4, and Prdx6, in normoxic or both normoxic and hypoxic cells.
Although the neuroprotective action of Ngb was identified first in relation to hypoxic and ischemic injury (Sun et al. 2001, 2003), subsequent investigations have pointed to a possible connection with Alzheimer’s disease (Khan et al. 2007; Li et al. 2008; Chen et al. 2012). In this regard, thimet oligopeptidase (Thop1), a protein with Alzheimer precursor protein β-secretase activity (Papastoitsis et al. 1994), was an Ngb interactor in hypoxic cells. Additional neurological disease-related proteins found to interact with Ngb included the zinc-finger protein Zpr1, which has been proposed as a modifier of disease severity in spinal muscular atrophy (Ahmad et al. 2012); DJ-1 (Park7), which is associated with autosomal recessive familial Parkinson’s disease; ubiquitin carboxyl-terminal esterase L1 (Uchl1), which may modify Parkinson’s disease risk (Lee and Liu 2008); and fused in sarcoma (Fus), mutations in which have been implicated in behavioral variant frontotemporal dementia and motor neuron disease (Mackenzie and Neumann 2012).
As noted, there was substantial overlap between Ngb- and Mgb-interacting proteins: 65% in normoxic and 70% in hypoxic cells. Cytochrome c, one protein with which both Ngb and Mgb have been reported to interact (Wu et al. 1972; Fago et al. 2006) and with which Ngb interacted in this study, was not found among Mgb interactors. Proteins mentioned above for Ngb, with which Mgb also interacted, included 14-3-3 isoforms, Aifm, Enoa, Fus, Ldha, Park7, Pfd3, Prdx6, Thio, Txnl1, and Uchl1.
One small, but potentially interesting group of proteins consists of those that interact with both Ngb and Mgb, but only under hypoxic conditions. These included brain acid-soluble protein 1 (also known as brain-abundant membrane-attached signal protein 1; Basp1), insulin-degrading enzyme (Ide), prolyl endopeptidase (Prep), selenocysteine insertion sequence (SECIS)-binding protein 2 (Sbp2), and vacuolar proton pump subunit F (Vatf). Basp1 is involved in the formation of membrane lipid microdomains (rafts) (Epand et al. 2005), to which the Ngb interactor flotillin-1 is localized (Wakasugi et al. 2004a), and which is inhibited in hypoxic neurons by Ngb over-expression (Khan et al. 2008). Ide contributes to clearance of β-amyloid and reduced Ide activity has been linked to some pedigrees with familial Alzheimer’s disease (Kim et al. 2007). Prep expression is altered in hippocampus of Alzheimer’s disease transgenic mice (Rossner et al. 2005). Sbp2 is involved in protection from oxidative injury, and Sbp2 knockdown triggers caspase- and cytochrome c-dependent apoptosis (Papp et al. 2010). Vatf is part of a protein complex that helps to regulate Hif1 levels by competing with Vhl (Lim et al. 2007).
The main finding of this study is that Ngb interacts with numerous proteins, including proteins related to known aspects of Ngb function, as well as proteins with which Mgb also interacts. Two key properties of Ngb—hypoxic induction and neuroprotection—were particularly reflected in the observed interactome. Thus, Ngb interactors included other hypoxia- and Hif1-inducible proteins, namely Enoa and Ldha, as well as Pfd3, a protein involved in regulating Hif1 levels. The neuroprotective effects of Ngb have been attributed variously to anti-oxidative effects (Fordel et al. 2006), anti-apoptotic effects (Fago et al. 2006), and inhibition of guanine nucleotide dissociation from G proteins (Wakasugi et al. 2003). Accordingly, Ngb interactors found in this study included anti-oxidative proteins (Prdx3, Prdx4, Prdx6), pro-apoptotic proteins (Cycs, Cyc2, Aifm), and GDI inhibitors (Gdir1, Tctp). Ngb attenuates brain injury in animal models of acute (stroke) and chronic (Alzheimer’s disease) neurodegeneration, and several interactors implicated in neurodegenerative disorders (Fus, Park7, Thop1, Uchl1, Zpr1) were also identified. Finally, two interactors demonstrated in previous studies, Cycs and 14-3-3 isoforms, were recognized in this study as well.
Mgb shared 65% of Ngb interactors under normoxic and 70% under hypoxic conditions, indicating substantial overlap. Although Mgb is normally found primarily in skeletal muscle and Ngb in neurons, they are related in structure (Burmester et al. 2000) and perhaps also in function (Brunori and Vallone 2006). Thus, Mgb interacted with some of the same hypoxia-inducible (Enoa, Ldha), anti-oxidative (Prdx6), pro-apoptotic (Aifm), and disease-related (Fus, Park7, Uchl1) proteins observed for Ngb. Mgb is not normally expressed in neurons, but the sharing of interactors with Ngb may argue that it has functions in muscle analogous to those of Ngb in neurons.
A small group of novel interactors—Basp1, Ide, Prep, Sbp2, Vatf—were distinguished by associating with both Ngb and Mgb exclusively under hypoxic conditions. These interactions could contribute to cytoprotection against hypoxia. For example, sequestration by Ngb of proteins that participate in the formation of (Basp1) or localize to (flotillin-1) lipid rafts, which have been implicated in hypoxic cell death (Khan et al. 2008), could confer protection by inhibiting raft formation. Conversely, Ngb could activate or otherwise promote the effects of interactors with cytoprotective effects, such as the anti-oxidative and anti-apoptotic protein Sbp2 or the Hif1-stabilizing protein Vatf.
The results of protein interaction studies should be interpreted with circumspection, because of limitations in their completeness, sensitivity, and precision (Vidal et al. 2011). For example, conditions under which these studies are conducted do not necessarily reflect the protein concentrations or compartmentalization that occurs in vivo (Sardiu and Washburn 2011). Indirect interactions, such as between non-interacting members of interacting protein complexes, may be detected, whereas genuine but low-affinity interactions or interactions involving low-abundance proteins may be missed. In comparing Ngb and Mgb interactomes in this study in particular, additional potential confounds are that Mgb is not normally expressed in neuronal cells, and that homology between horse Mgb (which we used) and mouse Mgb (with which mouse proteins would normally interact) is incomplete (82%). Nevertheless, studies of protein–protein interactions may have value for illuminating mechanisms involved in the cytoprotective effects of proteins like Ngb.
Supported by NIH grants R01 NS62040 and U54 HG04028 and CIRM grant TG2-01155. The authors have no conflicts of interest to declare.