Long-Term Intermittent Hypoxia Elevates Cobalt Levels in the Brain and Injures White Matter in Adult Mice

Abstract

Study Objectives:

Exposure to the variable oxygenation patterns in obstructive sleep apnea (OSA) causes oxidative stress within the brain. We hypothesized that this stress is associated with increased levels of redox-active metals and white matter injury.

Design:

Participants were randomly allocated to a control or experimental group (single independent variable).

Setting:

University animal house.

Participants:

Adult male C57BL/6J mice.

Interventions:

To model OSA, mice were exposed to long-term intermittent hypoxia (LTIH) for 10 hours/day for 8 weeks or sham intermittent hypoxia (SIH).

Measurements and Results:

Laser ablation-inductively coupled plasma-mass spectrometry was used to quantitatively map the distribution of the trace elements cobalt, copper, iron, and zinc in forebrain sections. Control mice contained 62 ± 7 ng cobalt/g wet weight, whereas LTIH mice contained 5600 ± 600 ng cobalt/g wet weight (P < 0.0001). Other elements were unchanged between conditions. Cobalt was concentrated within white matter regions of the brain, including the corpus callosum. Compared to that of control mice, the corpus callosum of LTIH mice had significantly more endoplasmic reticulum stress, fewer myelin-associated proteins, disorganized myelin sheaths, and more degenerated axon profiles. Because cobalt is an essential component of vitamin B₁₂, serum methylmalonic acid (MMA) levels were measured. LTIH mice had low MMA levels (P < 0.0001), indicative of increased B₁₂ activity.

Conclusions:

Long-term intermittent hypoxia increases brain cobalt, predominantly in the white matter. The increased cobalt is associated with endoplasmic reticulum stress, myelin loss, and axonal injury. Low plasma methylmalonic acid levels are associated with white matter injury in long-term intermittent hypoxia and possibly in obstructive sleep apnea.

Citation:

Veasey SC; Lear J; Zhu Y; Grinspan JB; Hare DJ; Wang S; Bunch D; Doble PA; Robinson SR. Long-term intermittent hypoxia elevates cobalt levels in the brain and injures white matter in adult mice. SLEEP 2013;36(10):1471-1481.

INTRODUCTION

Occurring in 3-5% of adults in developed countries, obstructive sleep apnea (OSA) involves intermittent, sleep state-dependent, brief reductions or cessations in ventilation.¹ Each disruption is associated with a partial desaturation of hemoglobin followed by reoxygenation. The brain is particularly vulnerable to hypoxia/reoxygenation, and neuroimaging studies of patients with moderate to severe levels of OSA have revealed brain injury, especially within white matter regions of the forebrain.^2–4

Exposure of healthy animals to the patterns of deoxygenation and re-oxygenation in OSA results in the carbonylation of proteins and the peroxidation of membrane lipids, indicating that intermittent hypoxia can cause oxidative stress.^5,6 It is thought that the reoxygenation phase involves an increased production of superoxide by mitochondria.⁶ Although super-oxide is normally detoxified by superoxide dismutase, redox-active metal ions, such as iron or copper, can interact with superoxide to produce potent hydroxyl radicals that readily oxidize proteins and lipids.⁷ Thus, redox-active metals may contribute to the brain injury observed in intermittent hypoxia. It is not known whether intermittent hypoxia is associated with regional changes in the distribution of metals in the brain. This possibility is a concern because OSA involves increased levels of systemic inflammation,⁸ and some redox-active metals, such as iron, can accumulate in tissues in response to inflammation.⁹

In the current study, adult mice were subjected to long-term intermittent hypoxia (LTIH), using an established model of OSA that causes systemic inflammation¹⁰ and oxidative stress in the brain.^6,11,12 A control group of mice was subjected to sham intermittent hypoxia (SIH). We examined whether LTIH is associated with changes in the regional distribution of iron, copper, zinc, or cobalt. Iron and copper have been commonly linked to the presence of oxidative stress in neurodegenerative diseases,^13,14 whereas zinc and cobalt have been found to induce oxidative stress in brain cells.^15,16 Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) was used to quantitatively map the elemental distribution of these metals in coronal sections of mouse brains. Iron, copper, and zinc did not change, whereas a massive increase was observed in the concentration of cobalt in white matter regions of LTIH mice, coincident with a significant disruption of myelin integrity and endoplasmic reticulum stress.

MATERIALS AND METHODS

Animals

Adult male C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) were 8 weeks old at the start of experiments. Mice were confirmed pathogen free. The methods and study protocols were approved in full by the Institutional Animal Care and Use Committee of the University of Pennsylvania and conformed with the revised National Institutes of Health Office of Laboratory Animal Welfare Policy. Food and water were provided ad libitum.

LTIH Experiments

Mice were randomized to LTIH across 10 h of the lights-on period or to the control condition (SIH), which involved similar cages and noise but was accompanied by normoxia.¹² Throughout both LTIH (n = 20) and SIH (n = 21), home cages of mice were placed within Plexiglas chambers (64 cm × 50 cm × 50 cm; Biospherix, Redfield, NY). The flow rates of > 99% nitrogen and > 99% oxygen into the chambers were varied with an automated oxygen profile system (Oxycycler model A84XOV; Biospherix) to produce episodic reductions from an ambient oxygen level. The oxygen changes occurred once every 90 sec from 08:00 until 18:00, every day for 8 weeks. The light cycle lasted from 06:00 to 18:00. In the LTIH condition, ambient oxygen concentration fluctuated between 21% and 9-11%, with the < 12% nadir maintained for < 3 sec. This protocol causes oxyhemoglobin saturation to fluctuate between 94-98% and 76-84%,¹² which is comparable to the magnitude of desaturation observed during apneic episodes in humans. The SIH mice were housed under identical conditions except that the ambient oxygen level fluctuated between 21% and 19% over the same cycle length as in the LTIH protocol. This protocol causes oxyhemoglobin saturation to fluctuate very slightly (96-98% to 96-97%¹²). Diet, humidity (50-60%), ambient carbon dioxide, and environmental temperature (21-22°C) were held constant within and across exposures.

Trace Metal Measurement

At the conclusion of the experiment, mice in the LTIH (n = 5) and control (n = 6) groups were killed by an intraperitoneal injection of sodium pentobarbital (80 mg/kg). The mice were then perfused transcardially with 0.1 M phosphate buffered saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were postfixed, cryoprotected in sucrose solutions, and then sectioned in the coronal plane at 40 μM thickness on a cryotome. Sections were stored at 4°C in 0.1 M phosphate buffer containing sodium azide. Two sections/mouse passing through the hippocampus were mounted onto glass microscope slides and air-dried.

Measurement of trace elements was performed on an Agilent Technologies 7500ce Series ICP-MS (Forrest Hill, VIC, Australia) coupled to a New Wave Research UP213 laser ablation unit (Kennelec Scientific, Mitcham, VIC, Australia), equipped with a frequency quintupled neodymium:yttrium aluminum garnet (Nd:YAG) laser emitting a 213-nm nanosecond laser pulse. A New Wave Research Large Format Cell (LFC) with an x-y-z stage was fitted to the UP213 unit. To increase sensitivity, the Agilent ‘ce’ lens assembly was replaced with the ‘cs’ design in the ICP-MS interface. The 7500ce instrument was equipped with an octopole collision/ reaction cell. All experiments used high-purity liquid Ar (Ace Cryogenics, Castle Hill, NSW, Australia) as the carrier gas and plasma source. Ultrahigh-purity hydrogen (99.999%) was used as the reaction gas (BOC, North Ryde, NSW, Australia).¹⁷ The LA-ICP-MS was tuned daily and for both standard mode and reaction mode using NIST 612 trace elements in glass for maximum sensitivity and to ensure low oxide formation. Low oxide production was assured by an m/z 248/232 ratio (representing ²³²Th¹⁶O⁺/²³²Th⁺) that was consistently less than 0.3%. The instrument was fine-tuned for tissue analysis using matrix-matched tissue standards, prepared as described previously.^18,19 Typical operational parameters for LA-ICP-MS are presented in Table 1.

Table 1.

Typical LA-ICP-MS operational parameters

Brain sections were ablated using a laser spot diameter of 30 μm and a laser scan speed of 127 μm sec^-1. A scan cycle = 0.2372 sec was chosen in order to maintain the relative dimensions of the sample according to the speed equation proposed previously.¹⁹ Constructed images had a spatial resolution of 30 μm. Quantification for each analysis was performed using standard calibration curves constructed from a representative ablation of matrix-matched tissue standards. Data were reduced using a Microsoft Visual Basic macro in Microsoft Excel 2003/2007, which arranged the data in a format suitable for the hyperspectral imaging software ITT Visual ENVI 4.2 (Visual Information Systems, Boulder, Colorado, USA). Background signals for each m/z obtained from the gas blank were subtracted. Linear regression analysis of the average background-corrected signal intensity of each ablated standard was performed to produce calibration data, which were then used to convert all pixels from signal intensity to μg/g wet weight of tissue.

Assessment of Vitamin B₁₂ Activity

The serum levels of methylmalonic acid (MMA) provide an index of vitamin B₁₂ activity, with increased levels of MMA being present in vitamin B₁₂ deficiency. Mice were exposed for 8 weeks to either LTIH (n = 10) or SIH (n = 10), and then killed as described previously. A small incision was made in the left atrium, and blood was collected from the pericardial region and immediately centrifuged. The plasma was stored at -80°C until assayed. The MMA content of plasma samples was measured using a previously published liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay, using appropriate quality controls.²⁰ Technical limitations prevented us from obtaining reliable quantitative estimates of MMA in homogenates of brain tissue.

Electron Microscopy

Electron microscopy was used to discern ultrastructural injury to myelin sheaths and axons. Following LTIH (n = 5) or SIH (n = 5), mice were transcardially perfused with heparin and acrolein (3.8%) in 2% paraformaldehyde in phosphate buffered saline, pH 7.4, and postfixed as previously described.²¹ Coronal sections (50 μm) were cut through the anterior corpus callosum with a vibrotome. Tissue sections were immersed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, rinsed in 0.1 M sodium cacodylate buffer, and postfixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer. After washing, sections were dehydrated in acetone and embedded in resin; polymerized and semithin sections (0.35 μm) were cut from the blocks to enable the examination of axon bundles with transmission electron microscopy. Sequential ultrathin sections (70-80 nm) were made using a Leica Ultracut E ultra-microtome (Deerfield, IL). Ultrathin sections were collected onto mesh grids and counterstained with Reynold lead citrate and uranyl acetate, and examined using a transmission electron microscope (Phillips CM10; Eindhoven, Netherlands). For each brain, four electron micrographs (6300× magnification) of axon fascicles in the anterior corpus callosum were examined for evidence of damage to myelin sheaths. The g-ratio is widely used as a structural index of optimal axonal myelination and is defined as axon diameter / (axon + sheath) diameter.²² Higher ratios within a group of common axons signify demyelination.²² Ballooning (greater than paranodal splitting) of the myelin lamellae, as previously defined, was assessed as a percentage of all axons/images.^23,24 Degenerative fibers were identified by their darkened cytoplasm and irregularly shaped cross-sections.²⁵

Immunohistochemistry

Following exposure to LTIH or SIH, 12 mice from each group were deeply anesthetized as described previously and transcardially perfused with 4% paraformaldehyde. Brains were removed and postfixed overnight, prior to cryopreservation in 30% sucrose. Forebrains were sectioned coronally at 40 μM thickness on a cryotome, then immunolabeled with primary antibodies for either proteolipid protein (PLP) or aspartocylase (ASPA) as detailed in Table 2. Secondary antibodies were labeled with Alexa Fluor 488 (green) or 594 (red) dye and sections were processed and analyzed as recently described.²⁶ Sections labeled with ASPA (to reveal mature oligodendrocytes) were counterstained with diamidino-2-phenylindole (DAPI) to show all nuclei within the corpus callosum. All counts of labeled cells were made from the medial genu portion of the anterior corpus callosum. PLP immunolabeling was analyzed using ImageJ to integrate density, with constant settings used across sections, as previously detailed.²⁶

Table 2.

Primary antibody and primer/probes sets used in the current study

Real-Time Polymerase Chain Reaction

Real-time polymerase chain reaction was used to measure an oligodendrocyte-specific messenger RNA (mRNA OLIG1), and two specific myelin-associated proteins: PLP and myelin-associated glycoprotein (MAG), and to assess endoplasmic reticulum stress binding immunoglobulin protein (BIP) and C/ EBP homologous binding protein (CHOP). For these measures, mice were deeply anesthetized as described earlier and transcardially perfused with sterile phosphate buffered saline containing RNAse inhibitor. Anterior corpus callosum tissue dissected from LTIH and SIH mice was used to purify RNA; 2 μg total RNA was loaded¹² and samples were normalized to 18S ribosomal RNA.

Western Blot Analysis

To ascertain the effects of LTIH on myelin proteins, Western blots were performed on corpus callosum tissue as procured as previously described for real-time polymerase chain reaction. Using methods recently described,²⁶ we detected PLP, the 21 kDa isoform of myelin basic protein (MBP), MAG, and 2',3'-cyclic-nucleotide 3'-phosphodiesterase (CNPase) using antibodies listed in Table 2. Tissue punches from individual subjects were homogenized on ice in lysis buffer containing a proteinase inhibitor cocktail and centrifuged. In addition, 20 μg of total protein (measured with a Pierce micro BCA protein assay kit) was run on sodium dodecyl sulfate polyacryl-amide gel electrophoresis (SDS-PAGE) gels (10% Tris-HCl, Bio-Rad). Gels were transferred to polyvinylidene difluoride membranes. Details of primary antibodies are presented in Table 2. Odyssey Application software, version 3.0.16 (Li-Cor) was used to identify appropriately sized bands and measure mean integrated densities. Integrated densities of MBP, PLP, MAG, and CNPase were normalized to the β-tubulin integrated density in each sample.

Statistical Analysis

The mean concentrations of cobalt, iron, copper, and zinc were calculated for brain sections from each of the five LTIH mice, to obtain a group mean and standard deviation for each of these trace metals. A similar analysis was performed for the six control mice, and then data from the two groups were compared using two-tailed Student t-tests (α < 0.05). Data obtained from the analysis of plasma MMA levels, protein expression, and mRNA expression were compared using two-tailed Student t-tests (α < 0.05). The percentage of axons with ballooning and degenerative changes were averaged for each mouse and analyzed with one-way analysis of variance (ANOVA), Bonferroni-corrected for the two parameters (α < 0.05). Western blot data were analyzed with one-way ANOVA and Bonferroni-corrected for four parameters (α < 0.05).

RESULTS

Cobalt

There was a highly significant effect of LTIH on brain cobalt levels. In the LTIH mice, cobalt was concentrated in regions that contain a preponderance of myelin: internal capsule, cerebral peduncles, medullary white matter of cortex, medial lemniscus, and the corpus callosum. By contrast, brain regions that contain high densities of neurons (cortical gray matter, hippocampus and hypothalamus) contained relatively low concentrations of cobalt (Figure 1). Sections from the control mice contained very low levels of cobalt that were uniformly distributed throughout the brain sections. The mean concentration of cobalt was 5600 ± 600 ng/g in LTIH mice and 62 ± 7 ng/g in control mice. The difference between these two groups is highly significant (t = 23.76; df = 9; P < 0.0001), and none of the LTIH mice had mean cobalt concentrations below 4,900 ng/g wet weight, whereas none of the control mice had mean levels of cobalt above 71 ng/g wet weight.

Figure 1.

Individual maps of elemental cobalt distribution in coronal sections through rostral hippocampus from control mice (n = 6) that received sham long-term intermittent hypoxia (SIH, left column) and mice that received long-term intermittent hypoxia (n = 5, LTIH, right column).

Iron

There was no effect of LTIH on the regional distribution or concentration of iron. The mean concentration of iron was 8.9 μg/g (± 0.7) in LTIH mice and 10.4 μg/g (± 1.5) in control mice. The difference between these two groups was not significant (t = 1.92; df = 9; P = 0.087). The levels of iron tended to be highest in gray matter regions that contain a high density of neuronal cell bodies and were lowest in white matter regions. Iron was most strongly concentrated in the granule cell layer of the dentate gyrus, pyramidal cell layer of the hippocampus, the ventromedial hypothalamus, and layers one through four of parietal cortex (Figure 2).

Figure 2.

Individual maps of elemental iron distribution in coronal sections through rostral hippocampus from control mice (n = 6) that received sham long-term intermittent hypoxia (SIH, left column) and five mice that received long-term intermittent hypoxia (LTIH, n = 5, right column).

Copper

No differences in the overall quantity or distribution of copper were observed between sections from LTIH mice or control mice (data not shown). The mean concentration of copper was 7.1 μg/g (± 1.3) in LTIH and 7.0 μg/g (± 0.6) in SIH. The difference between these two groups was not significant (t = 0.169; df = 9; P = 0.870). The levels of copper were highest in the periventricular regions of the hypothalamus, thalamus, and inferotemporal cortex. The lowest levels of copper were observed in the corpus callosum, deep white matter of parietal cortex, and lateral thalamus.

Zinc

LTIH did not affect the distribution or concentration of zinc. The mean concentration of zinc was 17.2 μg/g (± 1.4) in the LTIH group and 17.3 μg/g (± 1.3) in the control group. The difference between these two groups was not significant (t = 0.118; df = 9; P = 0.909). Zinc was concentrated in sectors CA2 and CA3 of the hippocampus, as well as in the amygdala and inferotemporal cortex. Zinc was uniformly distributed throughout the remaining tissue. No differences in the overall distribution of zinc were observed between sections from treated mice with LTIH or SIH (data not shown).

Methylmalonic Acid

The standard curve of the MMA assay was linear (r = 0.989), with a lower limit of quantification of 26 nM. The signal-to-noise ratio was > 10 for all samples. The intra-assay coefficient of variation was 3.28-3.96%. Plasma samples from control mice had a mean MMA concentration of 695 ± 117 nM (mean ± standard deviation), whereas samples from LTIH mice contained 422 ± 78 nM. The concentration of MMA in LTIH mice was significantly lower than that in SIH mice (t = 6.139; df = 18; P < 0.0001). Indeed, plasma MMA concentrations for all 10 LTIH mice were lower than any of the values obtained for the control mice (Figure 3).

Figure 3.

Scatterplots showing individual values for plasma methylmalonic acid (MMA) concentration in mice exposed to long-term intermittent hypoxia (LTIH, n = 10) and sham LTIH control mice (SIH; n = 10). The lower values in LTIH mice indicate a higher activity of vitamin B₁₂. A horizontal line indicates mean values for each group. The LTIH versus SIH difference = P < 0.0001.

Microscopy and Electron Microscopy

Total cells (DAPI-labeled nuclei) within sections of corpus callosum yielded similar counts of nucleated cells in the 4 mm² counting frame in both SIH mice and LTIH mice (155 cells ± 28 versus 141 cells ± 43, t = 1.2, not significant). However, the number of ASPA-immunolabeled cells within the same grid as the DAPI cells was higher in SIH mice than in LTIH mice (88 cells/counting frame ± 6 versus 31 cells ± 2, t = 5.0, P < 0.001). An example of the DAPI and ASPA labeling is provided in Figure 4. Because ASPA is normally enriched in mature oligodendrocytes and plays an essential role in providing acetyl groups for myelin synthesis,²⁷ these results suggest that myelinogenesis is impaired by LTIH.

Figure 4.

LTIH effects on ASPA and PLP within the corpus callosum. (A) Representative immunofluorescent images of ASPA (red) within the anterior mid corpus callosum in SIH and LTIH mice. ASPA (upper panels, red) extends from densely labeled mature oligodendrocyte processes outward to many myelinated tracks, resulting in a blurred edge extending from aligned oligodendrocyte cell bodies. Lower panels show DAPI labeling (blue) of all nuclei in the same region. Aligned configuration of nuclei suggests that most nuclei are oligodendrocyte nuclei. Scale bar, 50 μm. (B) Mean ± standard error (SE) cell counts within the mid-anterior corpus callosum show DAPI and ASPA cell counts. Asterisks indicate a P < 0.001 difference in ASPA SIH versus LTIH cells. (C) Confocal images of proteolipid protein (PLP, red) in corpus callosum; upper panels show lateral corpus callosum (CC) ventrolateral to caudate putamen (CPu). Left panels are SIH and right panels are LTIH. Lower panels show mid-corpus callosum (CC). Scale bar, 100 μm. (D) PLP confocal integrated densities (I.D.) normalized to background immunofluorescence signal. Shown are mean ± SE data for n = 5 mice/group. Asterisk denotes P < 0.001. ASPA, aspartocylase; DAPI, diamidino-2-phenylindole; LTIH, long-term intermittent hypoxia; SIH, sham intermittent hypoxia.

Electron microscopy revealed a significant effect of LTIH on both the myelin ultrastructure and the number of degenerated axons per image (n = 5 mice/group). In SIH mice, 19% ± 5% of myelin sheaths/image (20-40/image) showed ballooning, whereas in LTIH mice 61% ± 4% of myelin sheaths displayed evidence of ballooning (t = 6.3, P < 0.001). In SIH mice, 4% ± 0.7% axons showed degenerative changes, whereas in LTIH mice, 25% ± 6% of axons showed degenerative changes (t = 3.1, P < 0.05). Examples are provided in Figure 5. In addition, the g-ratio for corpus callosum axons was increased in LTIH, t = 5.2, P < 0.001 (Figure 5G). Examining individual fibers in all animals, this difference was readily apparent across all diameters measured (Figure 5H). In contrast to the changes in morphology and g-ratios, LTIH had no effect on axon diameter (t = 0.08, not significant, Figure 5I). Thus, the LTIH increase in g-ratio may be attributed to a reduced quantity of myelin.

Figure 5.

Ultrastructural effects of long-term intermittent hypoxia (LTIH). (A) Electron microscopic sagittal section of rostral corpus callosum in a sham-LTIH control mouse (SIH), showing myelin wrapped tightly around axons of varying sizes. Axons in SIH mice contain very little space between axons and myelin. (B) Corpus callosum in a mouse exposed to LTIH shows loosely wrapped myelin, and many hypodense regions (blebs and balloons) between axons and myelin, that distort the cross-sectional shape of the axon. Red arrows highlight irregularly shaped axon cross-sections and hyperdense cytoplasm, indicative of neurodegeneration. (C-F) Electron micrographs of axons in the corpus callosum. Those in control mice (C, E) have healthy mitochondria and tightly wrapped myelin, whereas axons in LTIH mice show enlarged mitochondria, numerous blebs, and balloonings of myelin (D) and loosely wrapped myelin (F). White arrows delineate balloonings and debris. (G) G-ratios for anterior corpus callosum axons (sagittal), for SIH (red) and LTIH (blue). Column boxes show mean values for each intermittent hypoxia condition. Solid circles and solid boxes delineate average g-ratios for individual mice (n = 5/group) with 95^th percentile confidence intervals for each group. (H) Scattergram of data points collected for SIH (red circles) and LTIH (blue squares) to reveal g-ratios across all axon diameters in the fields analyzed. (I) Mean ± standard error for all average axon diameters/animal analyzed for n = 5/group for SIH (red) and LTIH (blue).

Protein and mRNA Expression

Homogenates from the corpus callosum were processed by Western blotting, then the bands were scanned and analyzed with β-tubulin loading controls in order to quantify the amounts of major proteins present in each sample, using recently described approaches.²⁶ LTIH mice were found to have significantly lower expressions of the myelin-associated proteins CNPase, MAG, and PLP (Figure 6). The mean expression of MBP was unchanged across groups.

Figure 6.

LTIH reduces the titer of myelin-associated proteins in the corpus callosum. (A) Mean ± standard error data (n = 6-10/group) for integrated densities normalized to 20 μg protein, β-tubulin loading control, and to SIH mean data/gel for SIH (light gray) and LTIH (black). Single asterisk denotes P < 0.05 and double asterisks denotes P < 0.01 differences between conditions. (B) Representative blots for myelin proteins are analyzed. The β-tubulin (β-tub) shown was the loading control for MAG, MBP, and PLP. CNP, 2',3'-cyclic-nucleotide 3'-phosphodiester; LTIH, long-term intermittent hypoxia; MAG, myelin-associated glycoprotein; MBP, myelin basic protein; PLP, proteolipid protein; SIH, sham LTIH.

Expression of mRNAs for myelin-associated proteins was similarly altered in LTIH. Thus, compared with SIH mice, LTIH mice contained fewer copies of mRNA for PLP and OLIG1; however, mice from both groups had a similar number of copies of mRNA for MAG (Table 3). The expression of marker proteins of endoplasmic reticulum stress was increased in LTIH mice. Specifically, LTIH mice had significantly more copies of mRNA for the transcription factor CHOP; chaperone BIP trended toward an increase but was not significantly upregulated (Table 3). Collectively these data show that LTIH is associated with a reduced expression of major myelin proteins and there is evidence of endoplasmic reticulum stress within white matter.

Table 3.

Long-term intermittent hypoxia effects on markers of endoplasmic reticulum stress and myelin in the corpus callosum

DISCUSSION

The mean concentrations of iron, copper, and zinc in sections from control mice (SIH) were approximately 10, 4, and 14 μg/g wet weight, respectively. These values are very similar to those reported by a previous ICP-MS study of adult mouse brain,²⁸ and the regional distributions of these elements closely resemble those reported previously,²⁹ thereby confirming the reliability of the current assay. The brain levels of iron, copper, and zinc have been reported not to change in response to 3 days of mild hypoxia.²⁸ The current study extends that finding by showing that the distributions and concentrations of these trace elements are not significantly affected by 8 weeks of LTIH. This result makes it likely that the oxidative stress observed in the brain after LTIH is not due to the sequestration of redox-active iron, unlike what is reported to occur in inflammation³⁰ and chronic neurodegenerative disorders.³¹

Biochemical assays of homogenized tissue have estimated that the healthy human brain contains 50-70 ng/g of cobalt (as cobalamin).^32,33 An elemental analysis of cobalt in the normal mouse brain reported 9.3 ng/g wet weight,³⁴ whereas in adult rats the concentration of cobalt has been estimated to be 29 ng/g in the hippocampus³⁵ and 100 ng/g in homogenates of whole brain.³⁶ We found that control mice contained approximately 60 ng/g wet weight of cobalt, which lies within the range of published values for cobalt concentration in the mammalian brain. As far as we are aware, the current study is the first to map the spatial distribution of cobalt in brain sections.

In healthy mammalian tissues, vitamin B₁₂ (cobalamin) accounts for virtually all of the cobalt present.³⁷ Thus, the distribution of cobalt in brain sections is expected to correspond to the distribution and concentration of vitamin B₁₂. In LTIH mice, cobalt was concentrated in white matter tracts of the brain, which is consistent with earlier reports that cobalamin levels are higher in homogenates of human white matter compared to gray matter.^32,33

Intermittent hypoxia disrupts myelinogenesis in neonatal mice³⁸ and the current study extends this finding by demonstrating that LTIH causes a widespread disruption of myelin in adult mice. A loss of mature oligodendrocytes was evidenced by the marked reduction in ASPA+ cells in the corpus callosum. ASPA deacetylates N-acetylaspartate, a necessary step in myelination. Specific mutations in the ASPA gene result in a fatal leukodystrophy called Canavan disease.³⁹ In the developed brain, ASPA is a marker of mature oligodendrocytes.⁴⁰ Collectively, these findings indicate an impaired ability to maintain myelination and/or to remyelinate. Expression of the myelin-associated proteins CNPase, MAG, and PLP was also reduced in the corpus callosum of LTIH mice. PLP mRNA was also reduced in LTIH. Interestingly, although MAG protein levels were substantially lower in LTIH, MAG mRNA levels were unchanged, suggesting either impaired translation or increased protein degradation. Reductions in OLIG1 mRNA, as observed in the current study, also indicate the presence of fewer mature oligodendrocytes and an impaired capacity to repair myelin.^41,42 OLIG1 is expressed in the oligodendrocyte progenitor as well as the mature cell, although OLIG1 becomes progressively more expressed in the cytoplasm versus the nucleus with maturation.^41,42 As the decrease in OLIG1 mRNA following LTIH was measured as mRNA alterations, we do not know from which compartment OLIG1 was lost. This protein is clearly associated with myelin repair so the specific loss of OLIG1 mRNA is consistent with an impaired regenerative capacity.

Increased expression of mRNAs for CHOP, as observed, indicates the presence of increased levels of uncompensated or severe endoplasmic reticulum stress⁴³ and highlights a potential reason for significant injury to oligodendrocytes and/or axons. Whether the endoplasmic reticulum stress in myelinated tracts is primary or secondary cannot be determined in the current study. We have recently observed that increased CHOP occurs throughout the brain, including the cortex and hippocampus, in response to LTIH, and is a critical factor in the oxidative stress observed in LTIH.⁴⁴ Collectively, the preceding observations demonstrate that the injury caused to myelin by LTIH is multifaceted and affects the maturation of oligodendrocytes as well as a variety of pathways involved in the maintenance and repair of myelin. These results are consistent with evidence from other models showing that oxidative stress is associated with impaired oligodendrocyte maturation.⁴⁵

Most of the large, myelinated axons in LTIH mice exhibited clear evidence of injury, with numerous examples of blebbed myelin, destabilized myelin sheaths, and irregular axonal profiles. Many of the smaller unmyelinated axons were swollen. Compared to axon fascicles in control mice, the axons in LTIH mice had more extracellular space, which is indicative of axonal loss. Our findings are consistent with reports from magnetic resonance imaging studies that cortical white matter is prone to injury in people with severe OSA.^2–4

A major finding of the current study is that the average brain concentration of cobalt is 93-fold higher in LTIH mice than in control mice. In mammals, most of the cobalt in the body is found within vitamin B₁₂. Because vitamin B₁₂ is required for DNA synthesis, cell repair, and the stabilization of myelin,⁴⁶ the elevated levels of cobalt in LTIH may indicate that the brain sequesters vitamin B₁₂ in order to assist the repair of tissue that has been damaged by the increased production of reactive oxygen species in LTIH.⁶ Specialized carrier systems responsible for transporting the B-group vitamins into the brain from the blood are found on the blood-brain barrier and choroid plexus.⁴⁷ The transcobalamin receptor (TCblR) is responsible for actively transporting vitamin B₁₂ into the brain from the blood via clathrin-mediated endocytosis.⁴⁸ Although this transporter can concentrate vitamin B₁₂ within cells, it seems unlikely that myelin repair would require a 93-fold elevation in the concentration of vitamin B₁₂. Since vitamin B₁₂ is readily inactivated by oxidative stress,^49,50 we speculate that in LTIH the brain may sequester vitamin B₁₂ in order to compensate for the rapid inactivation of vitamin B₁₂ by oxidative stress. If so, the titer of active vitamin B₁₂ in the brains of LTIH mice may be far lower than is implied by the concentration of cobalt.

To explore the relationship between myelin injury and vitamin B₁₂ activity, the current study measured MMA levels in the plasma of LTIH mice. Vitamin B₁₂ is an essential cofactor in the conversion of methylmalonyl coenzyme A to succinyl coenzyme A, which is an intermediate of the tricarboxylic acid cycle. When vitamin B₁₂ activity is deficient, the surplus methylmalonyl coenzyme A is hydrolyzed to MMA, allowing elevated MMA levels to serve as a specific marker of vitamin B₁₂ deficiency.^51–53 Our study found that plasma MMA levels in LTIH mice were lower than those in control mice, indicating that circulating vitamin B₁₂ activity is not deficient in LTIH, and that adequate supplies of vitamin B₁₂ should be available for uptake into the brain. These findings raise the intriguing possibility that low plasma MMA levels may be indicative of myelin injury. Important next steps will be to examine this relationship more closely in mouse models using graded severities of LTIH, whereas in humans with OSA, serum MMA levels will be assessed in tandem with in vivo imaging of white matter.

Cell culture studies have demonstrated that elemental cobalt is neurotoxic at concentrations of 300-500 μM^16,54 and the current study found that cobalt attains concentrations of 500 μM in white matter regions of mice exposed to LTIH. White matter appears to be vulnerable to the toxicity of cobalt,⁵⁵ possibly because cobalt antagonizes the calcium-dependent transfer of proteins to the myelin membrane,⁵⁶ which are necessary for remyelination. It is noteworthy that cobalt is used experimentally as a hypoxia-mimetic agent, due to the fact that cobalt intoxication stabilizes hypoxia inducible factor (HIF-1α) in cells and leads to the increased expression of genes that are also upregulated by hypoxia.⁵⁴ In addition, cobalt damages mitochondria, depleting adenosine triphosphate supplies, and increasing the production of reactive oxygen species.⁵⁴ Therefore, if oxidized cobalamin is degraded to release elemental cobalt, it will add to the myelin damage and injury of oligodendrocytes caused by LTIH.

CONCLUSION

The current results demonstrate that the brain levels of cobalt are dramatically elevated in response to intermittent hypoxia, particularly in the white matter, which is grossly injured. We speculate that the brain sequesters vitamin B₁₂ in order to facilitate the repair of injured axons and disrupted myelin. However, vitamin B₁₂ is readily inactivated by oxidative stress, and the accumulation of oxidized cobalamin may account for the elevated levels of cobalt found in the brains of LTIH mice. If some of this oxidized cobalamin is subsequently degraded to release elemental cobalt, it will compound the oxidative stress and myelin disruption caused by LTIH.

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.