Mitochondrial adaptation in human mesenchymal stem cells following ionizing radiation.

Mitochondria are highly dynamic organelles that respond rapidly to a number of stressors to regulate energy transduction, cell death signaling, and reactive oxygen species generation. We hypothesized that mitochondrial remodeling, comprising both structural and functional alterations, following ionizing radiation (IR) may underlie some of the tenets of radiobiology. Mesenchymal stem cells (MSCs) are precursors of bone marrow stroma and are altered in acute myeloid leukemia and by radiation and chemotherapy. Here, we report on changes in mitochondrial remodeling in human MSCs following X-ray IR. Mitochondrial function was significantly increased in MSCs  4 h after IR as measured by mitochondrial oxygen consumption. Consistent with this elevated functional effect, electron transport chain supercomplexes were also increased in irradiated samples. In addition, mitochondria were significantly, albeit modestly, elongated, as measured by high-throughput automated confocal imaging coupled with automated mitochondrial morphometric analyses. We also demonstrate in fibroblasts that mitochondrial remodeling is required for the adaptation of cells to IR. To determine novel mechanisms involved in mitochondrial remodeling, we performed quantitative proteomics on isolated mitochondria from cells following IR. Label-free quantitative mitochondrial proteomics revealed notable changes in proteins in irradiated samples and identified prosaposin, and potentially its daughter protein saposin-B, as a potential candidate for regulating mitochondrial function following IR. Whereas research into the biologic effects of cellular irradiation has long focused on nuclear DNA effects, our experimental work, along with that of others, is finding that mitochondrial effects may have broader implications in the field of stress adaptation and cell death in cancer (including leukemia) and other disease states.-Patten, D. A., Ouellet, M., Allan, D. S., Germain, M., Baird, S. D., Harper, M.-E., Richardson, R. B. Mitochondrial adaptation in human mesenchymal stem cells following ionizing radiation.

Ionizing radiation (IR) was discovered at the end of the 19th century (1), and yet novel mechanisms by which cells and tissues respond to this quantifiable stress continue to be elucidated. IR is both a carcinogen and a therapeutic option for cancer patients, and thus, understanding biologic responses to IR is essential. Evidently, direct nuclear DNA damage by IR is a primary cause of DNA mutations as well as the initiation of cancer. Still, mounting evidence demonstrates that extranuclear elements, including mitochondria, are important contributors to the cellular response to IR (2–4).

Mitochondria are the primary sites of ATP generation in the majority of cell types and are also vital to cell death signaling, calcium handling, and many other functions (5). For ATP generation, mitochondria chiefly oxidize fuel substrates in the citric acid cycle, generating reducing equivalents. These high-energy molecules mainly donate electrons to the electron transport chain (ETC) at complex (C) I or II. Through a series of reactions, coupled to proton pumping at CI, CIII, and CIV, a proton motive force is generated, which is used to drive ATP production by the ATP synthase (CV). It is appreciated that generating ATP through this process, oxidative phosphorylation (oxphos), results in a greater amount of ATP generated per fuel substrate compared with the alternate ATP-generating pathway, glycolysis. Using glucose as an example, glycolysis yields a net value of 2 ATP molecules, whereas oxphos produces a maximum of 33.45 ATP molecules per molecule of glucose (6), with the amount depending on the efficiency of energy transduction by oxphos (7). A number of mechanisms alter this efficiency, including inner membrane proton leak and electron slippage, with the former decreasing and the latter increasing reactive oxygen species (ROS) generation, respectively, by the ETC. Both oxphos efficiency and ROS production may be altered by changes in higher-order assemblies of the ETC complexes, termed ETC supercomplexes (8, 9).

Another factor tightly correlated with energy efficiency is the overall structure of the mitochondrial reticulum, which is dynamically regulated though successive rounds of mitochondrial fission and fusion. Mitochondrial fusion promotes elongation of the reticulum and is executed by mitofusin 1 and 2 and optic atrophy 1 (OPA1); whereas mitochondrial fission is executed by dynamin-related protein 1 and associated factors (10). Imbalanced rates of fusion and fission can result in mitochondrial elongation or fragmentation. Generally, mitochondrial elongation is associated with increased ETC function, resistance to apoptosis, resistance to cell starvation, and protection from various cellular stresses, whereas mitochondrial fragmentation is associated with increased mitophagy, increased ROS generation, decreased ETC function, and cell death signaling (11). Mitochondrial structure responds to a growing list of extra- and intracellular cues, including apoptotic signals, inhibition of transcription, increased extracellular acidosis, and altered cellular bioenergetics (12–15). The role of IR in mitochondrial structural alterations has been incompletely investigated, especially when considering nonlethal doses. Moreover, the role of mitochondrial structural and functional remodeling in regulating the adaptive response to IR (i.e., the ability of cells to up-regulate intrinsic defense mechanisms by a low adaptive dose prior to the stress a higher challenging dose) has, to our knowledge, not been reported.

The use of radiation for diagnostic and therapeutic interventions continues to raise concerns regarding mechanisms of cellular function in tissue microenvironments such as bone marrow. Mesenchymal stem cells (MSCs) are supporting cells found in bone marrow and throughout the body that can differentiate primarily into chondrocytes, adipocytes, and osteoblasts, but under certain conditions can also differentiate into cardiomyocytes, myocytes, and neurons (16). MSCs have the unique attributes of being uncharacteristically radioresistant stem cells (17) and having intricately connected and elongated mitochondria. Understanding the effects of IR on mitochondrial bioenergetics in MSCs may have implications for protecting the marrow microenvironment during radiation treatment as well as further our basic understanding of mitochondrial function following IR.

Our work aims to test the hypothesis that mitochondrial structure and function are altered in response to IR, and this may dictate the adaptive response and, potentially, other tenets of radiobiology. We report here on altered mitochondrial respiration, ETC supercomplex associations, and slight mitochondrial elongation following IR in MSCs. Additionally, our use of subcellular quantitative proteomics identifies novel and potentially important protein regulators of mitochondrial function following IR.

MATERIALS AND METHODS

Cell culture

Human MSCs were isolated as previously described by Le et al. (18) from the filters of normal bone marrow harvests in accordance with the Ottawa Health Sciences Network Research Ethics Board (Ottawa, ON, Canada). Cells were grown in low-glucose DMEM (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% characterized fetal bovine serum (GE Healthcare, Waukesha, WI, USA) and 1% penicillin and streptomycin with fresh medium changes weekly. Mouse embryonic fibroblasts (MEFs), human fibroblasts (Ago1522), and human cervical cancer cells (HeLa) were cultured in DMEM supplemented with 10% fetal bovine serum (Wisent, Saint-Jean-Baptiste, QC, Canada), 1% penicillin and streptomycin, and l-glutamine (2 mM; Thermo Fisher Scientific). Irradiations were performed in an X-RAD 320 (Precision X-Ray, North Branford, CT, USA) biologic irradiator at 320 keV and 5 mA, administering a dose rate of 93 cGy/min, similar to previously published dose rates in MSCs (19). For experiments investigating the adaptive response, the amperage was set to 0.25 mA to attain a dose rate of 4.6 cGy/min; considerably lower than the 10 cGy/min proposed by Broome et al. (20) as the upper dose rate limit for the adaptive response.

Analysis of oxygen consumption rates

Cellular bioenergetics was assessed in MSCs irradiated with 0, 0.1, 1, or 10 Gy X-ray irradiation with the Seahorse XF24 (Agilent Technologies, Santa Clara, CA, USA), giving real-time data on both oxygen consumption rates (OCRs) and extracellular acidification rates (proxy measures of mitochondrial oxphos and cellular glycolysis). Because control wells must be run in parallel with irradiated cells within the Seahorse analyzer, sham-irradiated wells were shielded with 3 × 9-mm lead plates over control wells and 1-mm lead plates underneath to shield back-scattered X-rays. With this shielding, sham-irradiated cells received an estimated <<1 pGy.

Standard mitochondrial stress tests were performed on 50,000 cells per well using subsequent injections of oligomycin (1 µg/ml), Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; 1 μM), and antimycin A (1 μM) with rotenone (1 µM). HCO3-free DMEM was supplemented with 4 mM l-glutamine, 1 mM Na-pyruvate, and 5 mM d-glucose at pH 7.4. Remaining OCR following antimycin A and rotenone (nonmitochondrial OCR) was subtracted from the initial readings and those after oligomycin or FCCP treatments to yield resting, leak-dependent, and maximal OCR, respectively. The difference between resting and leak respiration is ATP-linked OCR. Finally, the difference between maximal and resting OCR is reserve capacity OCR, representing how much the cells under those conditions can alter their respiration if required. All data were normalized to protein levels via Bradford assay (Bio-Rad, Hercules, CA, USA) and compiled with Seahorse Wave (Agilent Technologies) and Excel (Microsoft, Redmond, WA, USA) software.

Western blot analysis

Cells or isolated mitochondria were lysed in Triton lysis buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% Triton X-100) supplemented with protease inhibitor cocktail (P8340; MilliporeSigma, Burlington, MA, USA) and phosphatase inhibitor cocktail (sc-45065; Santa Cruz Biotechnology, Dallas, TX, USA) when investigating phosphoproteins. Samples were lysed on ice with a 27 gauge needle; when investigating nuclear proteins [γH2A histone family member X (γH2AX)], samples were lysed via sonication. Protein was quantified via Bradford assay. The following antibodies were used: anti–cytochrome c (556432; BD Biosciences, San Jose, CA, USA); rabbit anti-TOM20 (11415; Santa Cruz Biotechnology); mouse anti-mitochondrial import receptor subunit TOM20l 70-kDa heat shock protein (MA3-028; Thermo Fisher Scientific); rabbit anti-prosaposin (PSAP) (ab180751; Abcam, Cambridge, United Kingdom); and rabbit anti–γH2AX–phosphorlyated S139 (ab11174; Abcam). For blue-native PAGE (BN-PAGE), the following antibodies were used: CI [NADH:ubiquinone oxidoreductase (NDUF) A9, 459100; Thermo Fisher Scientific], CII (Fp, 459200; Thermo Fisher Scientific), CIII (UQCRC2, Ab14745; Abcam), CIV (subunit I, 459600; Thermo Fisher Scientific), and CV (ATP5A, Ab14748; Abcam). Horseradish peroxidase–conjugated secondary antibodies (Promega, Madison, WI, USA) were used in conjunction with either autoradiography film (Bio-Rad) or a ChemiDoc Imaging System (Bio-Rad). Densitometry of Western blots were quantified using ImageJ (National Institutes of Health, Bethesda, MD, USA; ).

Mitochondrial structural determinations

For the quantification of mitochondrial structure, 2000 cells per well were plated on 384-well imaging quality dishes (Brooks Automation, Chelmsford, MA, USA) for 48 h prior to the indicated cell treatments. Following treatments, cells were fixed in 4% paraformaldehyde in PBS for 20 min and washed with PBS. For immunofluorescence, cells were incubated for 1 h in a primary antibody buffer (PBS + 1% Triton + 1% bovine serum albumin) with a 1:250 dilution of the antibody. Cells were washed 3 times with PBS and incubated for 1 h in secondary Oregon Green antibody (06381; Thermo Fisher Scientific) 1:500 in PBS with 1:4000 Hoechst 33342 (H1399; Thermo Fisher Scientific). Cells were washed again and imaged in PBS. Images were acquired with an automated confocal PerkinElmer microscope equipped with a ×40 water objective (PerkinElmer, Waltham, MA, USA).

For image analyses, 9 wells per condition were analyzed with 42 images/well in 4 independent experiments (4 × 378 fields/condition). For segment length analyses, images were processed with an embedded Acapella (v.4.1) script in the Columbus Image Analysis System v.2.8.1 (PerkinElmer). First, the find nuclei function was used; then, the find cytoplasm function around each nucleus was used based on the Tom20 staining (). The mitochondria image channel was put through a gaussian filter and inversed, and then a skeleton was made of the mitochondria in which the segment lengths of the mitochondria were averaged per cell, per field, and per well. Mitochondrion segments are any continuous length of mitochondria between an end, break point, crossover, or branch point. The mean segment length is then presented as the mean across multiple experiments. For analyses of mitochondrial interconnectedness, we segmented the images using the tubeness filter in ImageJ and then measured mitochondrial connectivity with both the ratio of junctions to ends and the number of elements per cluster (E) according to the method previously published by Ouellet et al. (21).

Figure 3

Mitochondrial length slightly increases in MSCs following IR. A) Representative images demonstrating how images were processed for mitochondrial segement length (see Materials and Methods). White: Tom20 (mitochondria). Blue: Hoechst (nucleus). B) Representative images taken by high-throughput automated confocal imaging of the indicated cell type. White: Tom20 (mitochondria). Blue: Hoechst (nucleus). C) Automated quantification of mitochondrial segment length performed by the Columbus Image Analysis System (means ± sem of 7 independent experiments). D) Representative images taken by high-throughput automated confocal imaging of MSCs treated as indicated. E) Automated quantification of mitochondrial segment length performed by the Columbus Image Analysis System (means ± sem of 4 independent experiments). Ctrl, control. Student’s unpaired t tests were performed as indicated. *P < 0.05, **P < 0.01, ***P < 0.005 compared with MSC ctrl. Scale bar, 20 µm.

Mitochondrial isolations

Mitochondrial isolations were performed on cells as previously described by Patten et al. (22), similar to Frezza et al. (23), entirely on ice or at 4°C. Cells were harvested, rinsed in PBS, and lysed in mitochondrial isolation buffer (200 mM mannitol, 70 mM sucrose, 10 mM HEPES, 1 mM EGTA, pH 7.4) with 20 strokes of a 17.5 gauge needle. Nuclei and cell debris were pelleted by centrifugation at 800 g for 9 min. Then, the supernatant was further centrifuged at 8600 g for 9 min to pellet mitochondria. Mitochondria were then resuspended in mitochondrial isolation buffer, and the 2 spins were repeated to further enrich the mitochondrial fraction.

Cytochrome C retention assay

As a proxy measure of either mitochondrial, cristae, or both, cristae ultrastructure, the ability of mitochondria to retain cytochrome c, was examined. Isolated mitochondria were incubated with digitonin (1 mg/mg of protein at 0.1%) at 4°C for 30 min with rotation to selectively permeabilize the mitochondrial outer membrane. Cytochrome c was then mobilized to the supernatant or not depending on the ultrastructure of mitochondrial cristae. Mitochondria were pelleted at 10,000 g for 10 min (4°C) to separate the released proteins. The resulting supernatants and pellets (volume equivalents from 12.5 µg of starting mitochondria) were analyzed by Western blot to quantify cytochrome c protein levels.

BN-PAGE

Analyses of ETC supercomplexes were performed by BN-PAGE according to Wittig et al. (24). MSCs were harvested, washed in PBS, and extracted in digitonin extraction buffer (50 mM imidazole/HCl pH 7.0, 50 mM NaCl, 5 mM 6-aminohexanoic acid, and 1 mM EDTA with 2% digitonin). The final cell wet weight:digitonin ratio was ∼7:1 w/w. Proteins were extracted on ice for 1 h with intermittent gentle pipetting. The resulting samples were cleared at 10,000 g for 30 min. Glycerol (5%) and a 1:4 dye:digitonin ratio of Coomassie blue G-250 (Bio-Rad) in 500 mM 6-aminohexanoic acid were added to the samples, which were then loaded onto in-house 3–13% large gradient gels. Gels were run for 2 h in a high Coomassie blue G-250 cathode buffer at 150 V and then overnight in a low Coomassie blue G-250 cathode buffer at 200 V. Pictures were taken of all gels to confirm proper loading (unpublished results), and gels were incubated for 30 min in transfer buffer containing 0.1% SDS and subsequently transferred to a nitrocellulose membrane at 400 mA for 2.5 h. Membranes were then analyzed by Western blotting as described above.

Quantitative mitochondrial proteomics

To investigate molecular mechanisms regulating mitochondrial structure and function, we performed label-free quantitative proteomics on isolated mitochondria. Mitochondrial proteins were extracted in 8 M urea in 50 mM ammonium bicarbonate (NH4HCO3), reduced with 10 mM DTT for 1 h, and alkylated with 20 mM iodoacetamide for 40 min in the dark. Samples were diluted 5 times with 50 mM ammonium bicarbonate and trypsin digested overnight at room temperature with agitation (protein:trypsin ratio of 50:1). Samples were then desalted with Sep-Pak columns (Waters, Milford, MA, USA) according to the manufacturer’s protocol and dried in a SpeedVac (Thermo Fisher Scientific). Peptides were analyzed by HPLC/electrospray ionization tandem mass spectrometry consisting of an 1100 micro-HPLC (Agilent Technologies) and an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific). Tandem mass spectrometry data were searched against the human International Protein Index (IPI, v.3.85; ) using MaxQuant software (25) for label-free quantitation.

Citrate synthase activity

Citrate synthase (CS) activity assays were performed as previously described by Pileggi et al. (26). Following the indicated treatments, cells were lysed on ice with CS homogenization buffer (50 mM KCl, 1 mM EDTA, 2 mM MgCl2, and 0.5% Triton X-100 in 25 mM Tris-HCl, pH 7.8). Reactions were performed in CS activity buffer [0.1 mM acetyl coenzyme A–trilithium salt (MilliporeSigma), 0.2 mM 5,5′-disthiobis 2-bitrobenzoic acid in 50 mM Tris-HCl, pH 8.0] and started with the addition of 0.5 mM oxaloacetate (MilliporeSigma). Samples were mixed and read on a 96-well plate reader spectrophotometer at 412 nm on a kinetic program at room temperature for 10 min. Rates of CS activity were normalized to protein content (Bradford), and expressed as percent of control lysates.

Cell death assays

Cells were treated as indicated and stained using propidium iodide (MilliporeSigma) and Hoechst 33342 (Thermo Fischer Scientific) for 20 min at 37°C. Images were acquired with an Evos FL Auto 2 Cell Imaging System using a ×20 objective (Thermo Fisher Scientific). Automated images were taken of 10 fields per condition containing a mean of 50 MSCs or >100 HeLa cells each (total cells per condition = ∼500 MSCs or ∼1000 HeLa cells). Cell death was quantified as a percentage of condensed nuclei to total nuclei and as PI+ cells to total nuclei.

Statistics

Unless otherwise stated, data are represented as means ± sem. Statistical analyses, including Student’s unpaired t tests and ANOVAs (as described in figure legends) were performed using Excel (Microsoft), Prism (Graphpad Software, La Jolla, CA, USA), and Perseus (27) (Max Planck Institute of Biochemistry, Planegg, Germany) software. Values of P < 0.05 were considered statistically significant. For gene ontology, Database for Annotation, Visualization and Integrated Discovery (DAVID, v.6.8) was utilized with Benjamini correction for multiple comparisons (28, 29).

RESULTS

Mitochondrial bioenergetics and ETC supercomplexes are altered in irradiated MSCs

The role of bioenergetics in influencing stem cell function is currently a hot topic in the stem cell field (30–33), but the impact of cellular bioenergetics on MSC function in particular is currently underappreciated (34). Moreover, the metabolic response of MSCs to irradiation, especially when considering the initial metabolic remodeling, is as of yet unknown. We began our studies by irradiating MSCs with low (0.1 Gy) to high (10 Gy) doses of X-rays with proper shielding for sham control wells (see Materials and Methods). At 4 and 8 h postirradiation, we investigated cellular bioenergetics. Interestingly, we observed a statistically significant increase in resting respiration as well as an increase in leak respiration in MSCs that were irradiated with 1 Gy of X-rays (). Leak respiration, which represents mitochondrial oxygen consumption not linked to ATP production, is often associated with the activation of proteins that uncouple the inner membrane and may help to limit ROS generation (35). This increase could be a transient adaptation because there was no significant increase in respiration at 8 h (Fig. 1). Furthermore, at 4 h post-IR, we observed a trend toward an increase in maximal respiration across all doses of IR. Maximal respiration represents the maximal activity of the ETC if the cells are further stressed. In many models, maximal respiration is correlated with levels of ETC supercomplex assembly, which can influence both superoxide generation (9) and respiration efficiency (8). Since conventional fractionated radiotherapy often uses doses close to 2 Gy (36), we further investigated the effects of 2 Gy exposures on MSC bioenergetics. Similar to initial results, 2 Gy of X-ray radiation significantly increased resting, ATP-coupled, leak (trend; P = 0.054), and maximal OCR at 4 h but not at 8 h post-IR (Supplemental Fig. S1). All subsequent experiments were performed at 4 h post-IR to investigate mechanisms of this functional alteration and to investigate early mitochondrial remodeling. Mitochondrial content, as determined by total cellular CS activity, was unaltered 4 h post-IR (Supplemental Fig. S1). Of note, as previously mentioned, MSCs are relatively radioresistant (compared with HeLa cells) to cell death by X-ray irradiation (reviewed in ref. 17 and Supplemental Fig. S2).

Figure 1

Cellular respiration is increased shortly following X-ray irradiation. A) MSCs were irradiated or not with the indicated dose of X-rays for 4 h (A, B) or 8 h (C, D). Cells were analyzed using a Seahorse XF24 analyzer with injections at the indicated times of oligomycin (Oligo), FCCP, and rotenone with antimycin A (Rot&AA). B) Quantification of the indicated parameter of respiration from A (means ± sem of 4 independent experiments). C) Matching experiments were also performed at 8 h postirradiation. D) Quantification of the indicated parameter of respiration from C (means ± sem of 3 independent experiments). Ctrl, control; non-mito, nonmitochondrial; RAD, radiation. One-way ANOVAs with repeated measures, Geisser-Greenhouse correction, and Tukey’s correction for multiple comparisons were performed. *P < 0.05.

To investigate ETC supercomplexes, cells treated as above with IR and 4 h post-IR proteins were extracted for BN-PAGE analysis. Following gel electrophoresis, we blotted membranes with antibodies against CI–V subunits. Interestingly, upon qualitative assessment, compared with our previously published BN-PAGE results (13, 37, 38), we found that CI- and CIII-containing subunits were often entirely associated with ETC supercomplexes in these cells (). For quantification purposes, all supercomplexes were controlled to the monomer of CII because its levels are stable (Supplemental Fig. S3) and it does not associate with ETC supercomplexes, as noted in previous analyses (13, 22, 37). Following quantification of CI-, CIII-, and CIV-containing supercomplexes, we found a significant increase in CIV-containing supercomplexes in MSCs irradiated with 1, 2, and 10 Gy X-rays (Fig. 2). Although there were dose-dependent trends toward increased CI and CIII-containing supercomplexes, we failed to observe statistically significant increased associations. This could be perhaps anticipated because, as noted above, CI and CIII were mostly already present in their supercomplexed form. Whereas we have previously demonstrated that other stressors affect cristae structure and CV assembly (13, 22), we were unable to detect significant differences in CV monomers following irradiations. Similar results were obtained when normalizing CI-, CIII-, and CIV-containing supercomplex to the CV monomer instead of CII as earlier (Supplemental Fig. S3). Similar to some other mammalian cells and tissues (13, 22, 37, 38), MSCs contained very little to no CV dimer under these experimental conditions.

Figure 2

X-ray irradiation increases CIV-containing ETC supercomplexes in MSCs. A) Representative BN-PAGE blots of the indicated ETC complexes of MSCs that were treated with X-rays (0–10 Gy) for 4 h. Monomers (mon) and supercomplexes (SCs) are indicated for each of the complexes blotted for: NDUFA9 for CI; SDHA for CII; UQCRC2 for CIII; COX4 for CIV; and ATP5A for CV. B) Data from A were quantified from 6 experiments as indicated and normalized to CII monomer (means ± sem of 6 independent experiments). Ctrl, control. Student’s unpaired t tests were performed relative to control. *P < 0.05.

Limited changes in mitochondrial reticular structure following MSC irradiation

Although it is well known that different cellular stressors can induce mitochondrial elongation and that this elongation is associated with metabolic remodeling (13–15), little is known about the effect of IR. To investigate the role of IR on mitochondrial structure in MSCs as a mechanism regulating the increased respiration observed in Fig. 1, we performed large-scale automated image acquisition coupled with automated image analysis. Across a number of cell lines investigated [MEFs, human primary fibroblasts (Ago1522), human cervical cancer (HeLa) cells, and MSCs], we found that MSCs had the longest and most interconnected mitochondrial reticulum (Fig. 3). Following computerized image analysis, MSCs also had the longest mitochondrial segment length (Fig. 3). While investigating the mitochondrial response to IR, we observed a slight increase in mitochondrial segment length trending for all doses but a significant increase at 2 and 10 Gy (Fig. 3). However, the segment length increase was very modest (up to 5%) and was not significantly increased in 1 Gy–irradiated MSCs, and therefore does not likely explain the changes in respiration observed in Fig. 1.

Although its functional significance remains to be elucidated, the branching or interconnectedness of the mitochondrial reticulum is another parameter of mitochondrial structure that is altered under numerous conditions including cell starvation (21). The functional significance of mitochondrial branching is incompletely understood but may allow for mitochondrial signaling events, including mitoflashes (39). Based on computational calculations using the previously developed robust probabilistic algorithm (21), we did not observe alterations in mitochondrial connectedness parameters of the ratios of junctions to ends and E (). Furthermore, no changes in mitochondrial ultrastructure were observed at 4 h following 0.1, 1, 2, or 10 Gy X-rays, as analyzed by the distribution of cytochrome c within cristae stores (Fig. 4 and Supplemental Fig. S4).

Figure 4

Mitochondrial interconnectedness and cytochrome c (CytC) distribution were unaltered in irradiated MSCs. A) Images accrued from experiments in Fig. 3 were analyzed for mitochondrial connectedness. J/E, junctions per end. B) Images accrued from experiments in Fig. 3 were analyzed for another parameter of mitochondrial connectedness, E. C) MSCs were irradiated or not for 4 h, and the distribution of cytochrome c was determined as an indicator of cristae structure. D) Quantification of data in C (means ± sem of 3 independent experiments). Ctrl, control; dig, digitonin; RAD, radiation; supers, supernatants. One-way ANOVAs with repeated measures, Geisser-Greenhouse correction, and Tukey’s correction for multiple comparisons were performed. **P < 0.01.

Mitochondrial remodeling is required for the adaptive response to IR

We next questioned the physiologic significance of mitochondrial remodeling in cells following IR. To this end, we investigated the adaptive response to IR in cells that cannot undergo functional and structural mitochondrial remodeling. In MEFs lacking the inner membrane fusion protein OPA1 [OPA1 knockout (KO)], mitochondria are completely fragmented and cannot undergo mitochondrial elongation or changes in respiration following cell stress (14, 22). γH2AX was used as a known measure of radiation damage to nuclear DNA (40), induced by IR and sensitive to the adaptive response (41). The adaptive dose of 10 cGy was preferred to 5 cGy to illicit a protective adaptive response 4 h prior to 4 and 10 Gy challenging doses (Supplemental Fig. S5). Under these conditions, there was a significant difference in the level of adaptation to IR in OPA1-KO cells compared with wild-type (WT) cells with a challenging dose of 4 Gy () or a challenging dose of 10 Gy (trend; P = 0.065; Supplemental Fig. S5). In fact, in contrast to WT cells, OPA1-KO MEFs seemed to display a negative adaptive response with a challenging dose of 4 Gy and a decreased adaptive response compared with WT MEFs at 10 Gy. Additionaly, although failing to reach statistical significance, OPA1-KO cells also generally had less γH2AX following IR. This last finding may be related to the oxygen effect, which posits that cells that are less reliant on mitochondria for ATP production are radioresistant because there is less secondary IR damage due in part to lowered induction of mitochondrial ROS (3). Taken together, our data suggest that functional mitochondrial remodeling is required for certain aspects of the biologic responses to IR.

Figure 5

Mitochondrial adaptation is required for the adaptive response to IR. A) WT and OPA1-KO cells were treated with the indicated doses and analyzed by Western blot for γH2AX and tubulin (loading control). B) Quantification of experiments in A displayed as percent of WT control (Ctrl). The right panel is expressed as the percent adaptation: the difference between the level of γH2AX following the challenging dose alone and the level of γH2AX that also received the adaptive dose (means ± sem of 3 independent experiments). A Student’s unpaired t test was performed on the relative adaptive response [4 Gy − (0.1 Gy + 4 Gy)]. *P < 0.05.

Subcellular proteomic changes in mitochondrially enriched samples following IR

To investigate possible molecular mechanisms regulating mitochondrial remodeling following cellular irradiation, we performed label-free subcellular quantitative proteomics. In 7 independent experiments, MSCs were treated for 4 h with the indicated dose of X-rays, and mitochondria were isolated and processed for quantitative proteomics (Supplemental Fig. S6). As exemplified in , many of the proteins identified were indeed mitochondrial, with some expected protein contaminants from lysosomes and the endoplasmic reticulum. Importantly, as visualized by volcano plots, the majority of proteins were aligned along a log2 ratio of the IR dose results to control of 0, signifying that the majority of proteins are not altered within the mitochondrially enriched proteome (Supplemental Fig. S6). Significantly altered proteins are identified in . We also analyzed results using a liberal statistical cutoff (paired Student’s t test with a nominal value of P < 0.05) followed by trend analyses (linear, polynomial, and logarithmic) across the different radiation doses ( and Supplemental Table S1). This allowed us to investigate proteins that change localization and expression across the different doses of radiation while reducing type II errors.

TABLE 1

Cellular components that were significantly enriched in the subcellular proteomic data following DAVID gene ontology

Term	Count	Percent	P value	Benjamin
Extracellular exosome	713	56.3	2.1E−273	1.5E−270
Focal adhesion	198	15.6	1E−124	3.7E−122
Membrane	454	35.8	2.3E−117	5.5E−115
Mitochondrion	298	23.5	4.7E−81	8.3E−79
Myelin sheath	90	7.1	1.7E−63	2.4E−61
Extracellular matrix	119	9.4	7.1E−60	8.5E−58
Cytosol	457	36.1	2.6E−56	2.7E−54
Cell-cell adherens junction	112	8.8	1.1E−48	9.9E−47
Melanosome	62	4.9	6.2E−45	5E−43
Ribosome	74	5.8	8.7E−41	6.2E−39
Mitochondrial inner membrane	118	9.3	1.5E−38	9.6E−37
Lysosomal membrane	90	7.1	6.1E−37	3.7E−35
Mitochondrial matrix	97	7.7	1.1E−35	6.3E−34
Cytosolic large ribosomal subunit	45	3.6	1.2E−34	6.4E−33
Endoplasmic reticulum membrane	163	12.9	4.2E−33	2E−31

TABLE 2

Candidate mitochondrial proteins in MSCs altered following different IR exposures

Protein name	Protein symbol	Traditional function	Fold change (relative to Ctrl)
			0.1 Gy	1 Gy	2 Gy	10 Gy
Membrane magnesium transporter 1	MMGT1	Golgi and endosome magnesium transporter	0.82	0.86	0.73	0.64a
Vesicle transport through interaction with t-SNAREs homolog 1B	VTI1B	Fusion of organelles, autophagy	0.81	0.76	1.13	0.64a
Protein family with sequence similarity 3 member C	FAM3C	Regulator of glucose and lipid metabolism, hepatokine	1.04	1.05	0.86	0.69a
Polypeptide N-acetylgalactosaminyl transferase 2	GALNT2	O-linked oligosaccharide biosynthesis	0.86	0.88	0.92	0.76a
Golgin subfamily A member 7	GOLGA7	Golgi protein trasport	1.04	0.92	0.71	0.81a
COX6C	COX6C	ETC	0.99	1.09	1.39a	0.98
Heme oxygenase 2	HMOX2	Heme catabolism, oxygen sensing	1.14	1.36a	1.08	1.03
NADPH–cytochrome P450 reductase	POR	P450 oxidoreductase	1.18	1.46a	1.18	1.06
Collagen α-1(V) chain	COL5A1	Connective fibrillar collagens	1.12	1.16	1.20a	1.07
Mitochondrial import inner membrane translocase subunit Tim8 A	TIMM8A	Mitochondrial protein import	1.16	1.12	1.28a	1.08
Peptidyl-prolyl cis-trans isomerase A	PPIA	Multiple functions: signaling, transport, transcription	1.13	1.41a	1.09	1.09
Ragulator complex protein late endosomal/lysosomal adaptor, MAPK and mammalian target of rapamycin activator 1	LAMTOR1	Nutrient sensing, autophagy, lysosomal function	1.22a	1.24a	1.10	1.10
Profilin-1	PFN1	Cytoskeleton, autophagy	1.08	1.27a	1.11	1.10
Aflatoxin B1 aldehyde reductase member 2	AKR7A2	Detoxification of aldehydes and ketones	1.13	1.41a	1.04	1.13
Atlastin GTPase 3	ATL3	Endoplasmic reticulum dynamics	1.13	1.18a	1.09	1.16
Neutral α-glucosidase AB	GANAB	Protein folding and quality control	1.10	1.17a	1.15	1.16a
Perilipin-3	PLIN3	Endosome and Golgi transport	1.29	1.46a	1.19	1.24
Peptidyl-prolyl cis-trans isomerase C	PPIC	Protein folding	1.26	1.50a	1.06	1.36
Dynactin subunit 2	DCTN2	Organellar transport	1.25	1.46a	0.98	1.42a
Nucleophosmin	NPM1	Multiple functions: cell cycle, protein chaperone	1.67	1.29	1.38	2.50a

Values represent fold change relative to control. Ctrl, control.

Values are significantly different than control following one-way ANOVA with multiple comparisons (false discovery rate = 0.2).

Figure 6

Dose-dependent trend analysis of candidate proteins altered by cellular IR. Paired Student’s t tests were performed between the control (Ctrl) group and the 4 irradiated groups to identify a large group of candidates (197) for trend analysis. A) Linear trend analysis of candidate proteins that had an association of r > 0.8. Additional linear trends that had an association of r2 > 0.6 are provided in Supplemental Table S1. B) Positive and negative polynomial second-degree associations of candidate proteins with an association of r2 > 0.8. C) Logarithmic trend analysis of candidate proteins having an association r2 > 0.8. Ctrl unirradiated samples were set to 0.1e−6 Gy to allow for logarithmic trend analysis.

Interestingly, following trend analyses, one of the most significantly altered proteins was PSAP, which had enriched expression at the lower 0.1–2-Gy doses, but that later decreased at 10 Gy. Previous research has demonstrated that one of the proteins made from PSAP, saposin (SAP)-B, is associated with mitochondria and furthermore can regulate the mitochondrial quantity of coenzyme Q (CoQ) (42–44). Additionally, the majority of PSAP peptides from our proteomic data set indeed mapped to SAP-B (). We thus used a commercially available antibody raised against the region of PSAP that includes SAP-A and SAP-B (Fig. 7). Results confirm that an immunoreactive band likely representing SAP-B is increased in mitochondrial isolates from 1 Gy–irradiated MSCs (Fig. 7). Moreover, in additional experiments, similar to the dose-dependent proteomic trend analyses, there tends to be a parabolic relationship between dose and mitochondrial SAP-B (Supplemental Fig. S7). Further studies will investigate the functional significance and regulation of SAP-B in mitochondria following IR and potentially other stressors.

Figure 7

Sap-B, the cleavage product of PSAP, is increased in mitochondrial isolates. A) Schematic diagram of the protein PSAP with its 4 daughter proteins, peptides identified from proteomic analyses, and targeting of the polyclonal antibody used in the present study. B) Western blot analysis of γH2AX in nuclear fractions and PSAP, SAP-B, and mitochondrial 70-kDa heat shock protein (mtHSP70) in mitochondrial fractions of MSCs irradiated or not with 2 Gy X-rays for 4 h. C) Quantification of samples analyzed in B (means ± sem of 5 independent experiments). Ctrl, control. A Student’s unpaired t test was performed. *P < 0.05.

DISCUSSION

Although the long-term response to high-dose IR on mitochondrial structure and function is appreciated (reviewed in ref. 4), much less was previously known about mitochondrion remodeling in the short term. We demonstrate here that mitochondrial function is increased following 4 h of IR in MSCs, and that this is correlated with changes in ETC supercomplexes and in mitochondrial length. Previous work has demonstrated that ROS and mitochondrial function are increased following IR in lung carcinoma cells (45). However, the researchers concluded that subsequent to changes in the cell cycle, changes in ROS and mitochondrial function were primarily affected in a cell cycle–dependent manner. In contrast, under our conditions, MSCs were very slow to proliferate and postirradiation times were short, suggesting that these effects are not cell cycle dependent.

There are multiple ways by which mitochondrial function may be altered following stress. One mechanism is by changes in ATP synthase assembly and the dimer formation regulating cristae structure and ATP production, which has been observed following multiple stimuli such as cell starvation and extracellular acidosis (13, 14, 22). Here, we did not observe similar changes in ATP synthase assembly or dimer formation following IR treatments in MSCs. Instead, ETC supercomplexes were increased, albeit only CIV-containing ETC supercomplexes were statistically significantly regulated, which is likely attributed to CI and CIII already present in their supercomplexed forms. Precisely how ETC supercomplexes are altered following IR remains unknown. Although cristae structure is an important factor regulating ETC supercomplex levels (46), we were unable to detect significant changes in cytochrome c distribution, which could signify that cristae structure is not altered. In addition, our proteomic analyses did not highlight changes in known shaping proteins or of known ETC supercomplex assembly factors.

In an effort to uncover proteins responsible for changes in mitochondrial function post-IR, we turned to quantitative subcellular proteomics. Our mitochondrial fraction contained standard contaminants including lysosomes and endoplasmic reticulum proteins. Indeed, a number of significantly altered proteins were of nonmitochondrial origin, including some proteins of interest: LAMTOR1, involved in sensing nutrients as part of the mammalian target of rapamycin complex 1–ragulator complex, and proteins involved in protein chaperoning and folding: nucleophosmin, PPIA, and PPIC. FAM3C was among those proteins that decreased following high-dose IR. Interestingly, FAM3C may play a metabolic role by decreasing both gluconeogenesis and lipogenesis, at least in the liver (47). Because of cell type, subcellular fractionation, and acute time points, our data differ considerably from others (48–50). Offspring of irradiated pregnant mice have been reported to up-regulate a number of ETC components, including proteins in CI and CIII (50). By contrast, in whole-heart tissue of irradiated mice, a large decrease in ETC components has been reported (49). We did not find any sweeping changes in ETC components, although there were small decreases at higher dose points of CI (NDUFB10) and CIV [cytochrome C oxidase (COX) 5B] proteins and slightly elevated CIV subunit (COX6C) at medium doses. Because CIV-containing supercomplexes were increased following IR, it is interesting to speculate that COX6C could play some role.

Trend analysis of quantitative subcellular proteomics revealed an increased expression of PSAP following IR. We confirmed that PSAP, and likely SAP-B, were significantly enriched in our mitochondrially enriched samples from MSCs. A body of literature by Jin et al. (43) has described a role for SAP-B in regulating mitochondrial function. Firstly, they found that SAP-B can both bind CoQ and then regulate its mitochondrial levels in liver HepG2 cells (42). More recently, they showed that the opposite, PSAP knockdown, decreases both CoQ and ATP levels (44). It is provocative to speculate that the increased levels of PSAP, and likely SAP-B, that we observed could affect mitochondrial function through modulating CoQ levels. Indeed, CoQ is an electron carrier that shuttles electrons to CIII, is a potent antioxidant, and also associates with ETC supercomplexes (51). It is perhaps not surprising that CoQ in itself protects cells again IR damage (52, 53).

In addition to this specific metabolic role of SAP-B, there is considerably more data on the protective role of the precursor glycoprotein PSAP. PSAP is mostly known for its role in the regulation of lysosomal function by acting as an activator of sphingolipid metabolizing enzymes (54). In the brain, PSAP is secreted following cellular stress and endocytosed by neighboring neural and glial cells, functioning as a prosurvival factor (55). PSAP may also provide neuroprotection by regulating progranulin levels (56). In cancer, PSAP may have an undesirable role by stimulating prosurvival pathways, activated by Homeobox protein Hox-C11 and modulating androgen receptor pathways (57). High-PSAP mRNA levels are correlated with decreased survival following treated breast cancer, an effect that is worsened when androgen receptor mRNA is also high. Establishing the function of PSAP up-regulation following IR in MSCs, and potentially other cell types, requires further experimentation. Since our mitochondrial isolations contain many lysosomal proteins, we cannot rule out that PSAP is up-regulated in the lysosomal fraction, which, again, requires further investigation. We are unable to speculate on the parabolic relationship between X-ray dose and mitochondrial PSAP, although it is not likely because of lethality at 10 Gy based on our data presented here, and others (17), demonstrating the relative radioresistance of MSCs.

The physiologic implications to changes in mitochondrial structure and function following IR are vast. Here, we investigated one implication and demonstrated that cells that cannot undergo changes in mitochondrial structure and function (OPA1-KO MEFs) have a decreased adaptive response to irradiation. The concept that preconditioning of cells with an adaptive dose of IR protects them from a subsequent challenging dose is an accepted phenomenon in IR (58) akin to many other fields such as ischemic preconditioning in the heart (59) or in the brain (60). In the IR-adaptive response, the priming dose is often low-dose radiation (1–10 cGy) at a low-dose rate (<10 cGy/min), and its beneficial, hormetic effects have been traditionally ascribed to the induction of antioxidant and DNA repair mechanisms (20, 58). Although mitochondrial remodeling has been observed and proven protective across a number of different stressors (13–15), its potential role in modulating the adaptive response had previously been uncharacterized. Interestingly, Lall et al. (61) described an HIF1-dependent metabolic adaptation at 5% O2 to the adaptive response that involves the induction of a glycolytic program at later time points (12 h). We believe that this adaptation may be attributed to changes in the radiation oxygen effect, in which greater cell death is observed at higher O2 tension, as previously described by Richardson and Harper (3). Our system differs from that of Lall et al. (61) in that experiments were performed at 21% O2 and the IR adaptation was observed at short time points postirradiation. Taken together, it could be speculated that 2 metabolic processes may regulate the adaptive response at different time points: first, in the short term, mitochondrial remodeling increases metabolic efficiency and ATP production, and secondly, in the longer term, processes induce a metabolic shift toward glycolysis (62, 63), which may limit the damaging ROS production from mitochondria.

Precisely how mitochondrial remodeling affects the adaptive response is beyond the scope of the present project. However, we speculate that these changes in mitochondrial shape, ETC supercomplexes, and ETC function likely affect both ROS production and ATP production based on our findings in MSCs. On decreasing ROS production, we have demonstrated an increase in proton leak OCR, which may limit electron slippage and subsequent superoxide production (35), and we have also demonstrated that ETC supercomplexes are increased following IR, which may limit ROS production at CI (9). Of note here, Miura (58) has previously demonstrated that the priming during the adaptive response can (in addition to IR) be accomplished using H2O2. On increasing ATP production, we have demonstrated an increase in ETC supercomplexes, which may increase the efficiency of energy conversion (64).

In addition to the adaptive response, the physiologic implications of the present work extend to the effect of radiation on stem cell function. Recent work demonstrates how acute myeloid leukemia (AML)–derived MSCs grown in culture are highly dysfunctional, leading to a proadipogenic phenotype (18). Moreover, AML MSCs altered hematopoetic stem cells in coculture by increasing their proliferation and decreasing their quiescence. These findings suggest that intrinsic MSC characteristics are altered in AML, which likely supports a procancerous phenotype, and that maintaining healthy MSC function could better support hematopoiesis and lower AML relapse. Furthermore, irradiated MSCs have a similar phenotype of increased adipogenic potential. It is of interest to speculate that prolonged irradiation would continue to promote our demonstrated mitochondrial remodeling, favoring OXPHOS for ATP production, and inducing an adipogenic lineage. Future experiments are thus aimed at manipulating cellular metabolism during the cellular response to IR to ascertain the relative impact on stem cell fate, especially in MSCs. Although the mechanisms underlying the AML MSC phenotype have yet to be elucidated, many lines of evidence point to mitochondrial bioenergetics as regulators of stem cell fate (65). Since mitochondrial respiration likely affects MSC differentiation through ROS signaling (31, 33), future experiments will be required to examine the role of mitochondrial remodeling on ROS signaling in our system.

Because current transplant strategies for the treatment of AML begin with high doses of chemotherapy and radiation, severely affecting the bone marrow microenvironment, killing off resident stem cells, and likely creating a hostile microenvironment prior to hematopoietic stem cell transplantation, understanding the underlying mechanisms regulating MSC health is pivotal to improving therapeutic interventions. Beyond AML, because of their repair and regenerative potential, MSCs are currently being investigated for therapeutic benefit in hundreds of clinical trials (66), including those involving radiation treatments (67). Interestingly, MSCs with high colony-forming units and long lives in vitro are characterized by lower mitochondrial activity (68). These characteristics are similar to those in in vivo bone marrow, in which active differentiating MSCs occupy perivascular niches and more stem cell–like quiescent MSCs pair up with hematopoietic stem cells in the hypoxic endosteal niche of trabecular bone surfaces (69). Moreover, increased ETC supercomplexes have been observed in MSCs undergoing adipogenic differentiation (70). Taken together, our results suggest that IR, especially at a high dose, may negatively affect therapeutic potential by increasing mitochondrial function and ETC supercomplexes, promoting an adipogenic lineage (18).

In summary, we have experimentally demonstrated a number of mitochondrial effects following IR in MSCs. Specifically, we have demonstrated increased mitochondrial respiration, mitochondrial structure, and ETC supercomplex formation in MSCs following IR. Furthermore, quantitative proteomics have revealed a number of interesting candidate proteins that may alter mitochondrial function following IR, including SAP-B. We believe that our observations reveal key mitochondrial remodeling events following IR and novel mechanisms potentially involved in this adaptation. Ultimately, understanding the biology of MSCs and their response to IR may elucidate exploitable pathways for treatments requiring stem cell transplants, including leukemia.

Supplementary Material

This article includes supplemental data. Please visit to obtain this information.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.