J Appl Physiol 118: 544–557, 2015. First published November 26, 2014; doi:10.1152/japplphysiol.00629.2014.

Simulating obstructive sleep apnea patients’ oxygenation characteristics into a mouse model of cyclical intermittent hypoxia X Diane C. Lim,1,2 Daniel C. Brady,2 Pengse Po,2 Li Pang Chuang,3 Laise Marcondes,4 Emily Y. Kim,2 Brendan T. Keenan,2 Xiaofeng Guo,2 Greg Maislin,1,2 Raymond J. Galante,2 and Allan I. Pack1,2 1

Division of Sleep Medicine, Department of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania; 2Center for Sleep and Circadian Neurobiology, University of Pennsylvania, Philadelphia, Pennsylvania; 3Department of Thoracic Medicine and Department of Sleep Center, Chang Gung Memorial Hospital, Taipei, Taiwan and Graduate Institute of Clinical Medical Sciences, Chang Gung University, Tauyan, Taiwan; and 4Superior School of Health Sciences, Brasilia, Brazil Submitted 14 July 2014; accepted in final form 24 November 2014

Lim DC, Brady DC, Po P, Chuang LP, Marcondes L, Kim EY, Keenan BT, Guo X, Maislin G, Galante RJ, Pack AI. Simulating obstructive sleep apnea patients’ oxygenation characteristics into a mouse model of cyclical intermittent hypoxia. J Appl Physiol 118: 544 –557, 2015. First published November 26, 2014; doi:10.1152/japplphysiol.00629.2014.—Mouse models of cyclical intermittent hypoxia (CIH) are used to study the consequences of both hypoxia and oxidative stress in obstructive sleep apnea (OSA). Whether or not a mouse model of CIH that simulates OSA patients’ oxygenation characteristics would translate into improved patient care remains unanswered. First we identified oxygenation characteristics using the desaturation and resaturation time in 47 OSA subjects from the Molecular Signatures of Obstructive Sleep Apnea Cohort (MSOSA). We observe that a cycle of intermittent hypoxia is not sinusoidal; specifically, desaturation time increases in an almost linear relationship to the degree of hypoxia (nadir), whereas resaturation time is somewhat constant (⬃15 s), irrespective of the nadir. Second, we modified the Hycon mouse model of CIH to accommodate a 15-s resaturation time. Using this modified CIH model, we explored whether a short resaturation schedule (15 s), which includes the characteristics of OSA patients, had a different effect on levels of oxidative stress (i.e., urinary 8,12-iso-iPF2␣-VI levels) compared with sham and a long resaturation schedule (90 s), a schedule that is not uncommon in rodent models of CIH. Results suggest that shorter resaturation time may result in a higher level of 8,12-iso-iPF2␣-VI compared with long resaturation or sham conditions. Therefore, simulating the rodent model of CIH to reflect this and other OSA patients’ oxygenation characteristics may be worthy of consideration to better understand the effects of hypoxia, oxidative stress, and their interactions. obstructive sleep apnea; cyclical intermittent hypoxia; oxidative stress; 8,12-iso-iPF2␣-VI

of cyclical intermittent hypoxia (CIH) are widely used to assess the consequences of obstructive sleep apnea (OSA). These studies build on the seminal work of Fletcher (18, 19). Such studies are important because OSA is not only a highly prevalent chronic disease (76) but is also associated with many chronic illnesses including cardiovascular (46, 54, 60), metabolic (33, 48), and neurodegenerative (2, 75) diseases, as well as cancer (44, 52). The association between OSA and seemingly different classes of disease sug-

RODENT MODELS

gests that OSA is a systemic disorder that contributes to the development of these chronic diseases. Chronic CIH is likely one of the major pathogenic components of OSA, since CIH results in downstream effects of both hypoxia (55) and oxidative stress (34, 74). Thus investigators use rodent models of CIH to generate gene, protein, and physiological data to understand mechanisms underlying the consequences of OSA (1, 41, 51, 55). While studies using rodent CIH models have led to major advances in knowledge, there are fundamental questions that have yet to be addressed. For example, it may be important to consider whether rodent models of CIH should better reflect oxygenation characteristics seen in OSA patients to more accurately simulate hypoxia and oxidative stress conditions. Therefore, our first aim was to use an OSA patient cohort to identify characteristics of oxygen desaturation/resaturation. We identified two features that are not currently utilized in CIH systems. First, the rate of resaturation is much faster than the rate of desaturation. Second, there is considerable variability in a single night of study, both within and between subjects. Simulating these characteristics represents a new approach to modeling CIH in mice that may more accurately reflect oxygenation characteristics within patients with OSA. Our second aim was to use the MSOSA data to design and implement a new CIH system that can simulate these oxygenation characteristics; this new system is the main focus of this report. We used this new model to establish arterial oxygen saturation (SaO2) (by MouseOx) at different levels of fraction of inspired oxygen (FIO2), demonstrating differences in zenith and nadir SaO2 values between four different inbred strains. Our third aim was to use this newly developed model and assess whether urinary 8,12-iso-iPF2␣-VI levels (a marker of oxidative stress) increased in mice that were exposed to a more rapid resaturation compared with a slower resaturation. We hypothesized that a shorter resaturation time would generate more oxidative stress as demonstrated in in vitro studies (40, 66) and in vivo studies (70). Our new CIH system allows us to simulate characteristics of desaturation/resaturation that were identified in patients with OSA. METHODS

Human Studies Address for reprint requests and other correspondence: D. C. Lim, Dept. of Medicine, Univ. of Pennsylvania, Translational Research Laboratory, 125 S. 31st St., Ste. 2100, Philadelphia, PA 19104 (e-mail: [email protected]. edu). 544

Molecular Signatures of Obstructive Sleep Apnea cohort. To assess characteristics of oxygen desaturation/resaturation time, we utilized data from the Molecular Signatures of Obstructive Sleep Apnea study

8750-7587/15 Copyright © 2015 the American Physiological Society

http://www.jappl.org

Simulating OSA with CIH

(MSOSA). MSOSA is an institutional review board-approved study at the University of Pennsylvania that prospectively recruited newly diagnosed, untreated patients with OSA from March 2008 to January 2013 at the Hospital of the University of Pennsylvania. Subjects underwent a written informed consent process and in-laboratory polysomnogram-recorded electroencephalograms; eye, chin, and pretibial muscle activity; electrocardiography; oximetry; chest and abdominal respiratory effort; and airflow by nasal cannula (nasal pressure) and oral thermistor (Compumedix, Safiro) that was scored using Sandman (Natus) software. Studies with less than 4 h of sleep (n ⫽ 3) were repeated. Only the second study was used in the analyses if studies were repeated. Automatic oxygen desaturation and respiratory modules were run, and results were manually reviewed to remove or reclassify events produced by autoscoring, add missing events, and mark bad data in 5-min epochs. Analyses presented in this manuscript are based on 47 subjects who were diagnosed with OSA [apneahypopnea index (AHI) ⬎ 5 events/h, an index used to indicate severity of sleep apnea and is the average number of apneas and hypopneas per hour of sleep], based on their in-laboratory study. We also calculated the oxygen desaturations index 3% (ODI3), which is the average number of desaturations events with ⱖ3% fall in oxygen saturation per hour of sleep. MSOSA polysomnograms underwent additional measurements for the rate of oxygen desaturation (downward slope) and resaturation

•

Lim DC et al.

545

(upward slope) for each apnea or hypopnea. Specifically, one technician manually scored all studies for the duration of 1) oxygen desaturation, with time starting at the SaO2 where an event starts to undergo a 3% or more oxygen desaturation and ending when the SaO2 was either at the nadir or sustained a 3-s plateau; and 2) oxygen resaturation, with time starting at the SaO2 undergoing a 3% or more oxygen resaturation and time ending when the SaO2 reaches a 3-s plateau. Apneas and hypopneas without a ⱖ3% oxygen desaturation were not used for this analysis. Mouse Studies HyCon system of CIH. Savransky initially developed the HyCon0520 to the technical requirements of Polotsky (62), which is commercially available (S. Savransky, West Des Moines, IA). The HyCon is fully automatic, using a closed-loop control mechanism to regulate air/N2 delivery by making direct comparisons between the oxygenation points set by the investigator to a Hycon O2 sensor (HO2S) placed 5 cm above the noses of sleeping mice. The HO2S samples the gas inside the home cage (Fig. 1, A and D) and is externally validated with a second external O2 sensor (EO2S), the O2 Analyzer (model 17620; VacuMed, Ventura, CA). To ensure rapid exchange of gas in the mouse cage, we delivered gasses simultaneously through six inlets (Fig. 1C), spraying the gasses downward with nozzles (Fig. 1B) and

Fig. 1. Modifications to the HyCon. A: the HyCon allows us to use commercially available cages that house 5 mice per cage, thus permitting normal socialization. Within the modified HyCon, mice consistently choose to sleep in the middle of the cage under the outlet holes near the HyCon O2 Sensor (HO2S) and the metal rod for the external O2 sensor (EO2S). B: a schematic of how the hardware inside the home cage directly opposite to inlets forces gasses to “spray” downward (denoted by arrows) to the level of the mouse. C: 6 inlets are equally spaced around the home cage at 10 cm above the bottom of the cage. This allowed mixing of gasses before reaching the mice at the bottom of the cage. D: a fixed metal rod allows us to feed the EO2S tube into the home cage and sample gasses close to the mouse’s nose and in close proximity to the HO2S. E: a schematic of a 2-valve system implemented to allow faster oxygen resaturation. Abbreviations: H1 represents the intermittent hypoxia for cage 1 and H2 represents the intermittent hypoxia for cage 2 within one HyCon system. F: the top of the cage has a uniform location and number of outlets that force gasses to move in only one flow pattern.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

546

Simulating OSA with CIH

then directing the gasses out of the cage through 10 outlet holes in the lid (Fig. 1F). Further modifications were made to the hardware (Fig. 1E) and software to allow us to simulate oxygen desaturation/resaturation characteristics identified and measured in the MSOSA cohort. Details are outlined in RESULTS, but in short, hardware and software upgrades enabled input of the following parameters: SH (the Set High FIO2); SL (the Set Low FIO2); TD (desaturation time); TR (resaturation time); TPH (time during plateau at set high FIO2); and TPL (time during plateau at set low FIO2). Validation of HyCon modifications: MouseOx study. All experiments were approved by the IACUC at the University of Pennsylvania. Mice were routinely housed on a 12:12-h light:dark cycle (7 am:7 pm) and given food and water ad libitum. To validate our modified HyCon system, SaO2 was first measured via a MouseOx in eight C57BL/6 (National Institute on Aging, Bethesda, MD) using a sinusoidal desaturation/resaturation program. Analysis of these data using a power analysis revealed that five mice per group were needed to give 80% power to observe a mean difference of at least 5% oxygen saturation (for details, see Data Analysis and Statistical Approaches: Mouse Studies). We then measured SaO2 in four of the founder strains from the Collaborative Cross (8). Therefore, 10 C57BL/6 (National Institute on Aging, Bethesda, MD), 10 129SH/SvImJ (Jackson Lab, ME), 10 A/J (Jackson Lab, ME), and 10 NZO/HiLtJ (Jackson Lab, ME) male mice, aged 4 mo, were used to explore the SaO2 levels in response to a short and a long resaturation program. Specifically, five mice per strain were scheduled to receive a nonsinusoidal short resaturation (TR of 15 s) program, and five mice per strain were scheduled to receive a nonsinusoidal long resaturation (TR of 90 s) program. To measure SaO2, a MouseOx collar sensor was placed over the carotid artery (Rev 6.3.10 MouseOx software) of an unanesthetized mouse, which was then gently placed inside a clear tube restrainer (Starr Life Sciences). The MouseOx sensor relays data to STARR-Link, a device outside the home cage that transmits an analog voltage representation of digital data to a computer that displays SaO2, heart rate, respiratory rate, and other measurements. Within the tube, the mouse was allowed to walk back-and-forth for 10 –15 min to acclimate to its new surroundings before recording started. An external oxygen sensor (EO2S) was placed directly at the level of the mouse’s nose right outside the restrainer. Oxygen saturation was recorded throughout continuous CIH cycles that desaturated from an FIO2 of Set High FIO2 (SH) to Set Low FIO2 (SL) over Time Desaturation (TD) seconds, then resaturated from SL to SH over Time Resaturation (TR) seconds. Table 1 outlines the sinusoidal and nonsinusoidal schedules that were used, with the nonsinusoidal TD and TR determined from the data on subjects with OSA. After obtaining at least 10 acceptable cycles (e.g., minimal movement artifact) the FIO2 at SL was stepped down by 0.01. For the nonsinusoidal CIH schedule, TD and TR were changed with each step down in SL, with time during plateau at SH (TPH) held constant at 3 s and time during plateau at SL (TPL) was 0. Effects of short vs. long resaturation times on oxidative stress. To explore whether CIH schedules differentially affected levels of oxidative stress, we measured urinary 8,12-iso-iPF2␣-VI, a validated measure of oxidative stress that is elevated in patients with OSA (36, 57). We exposed mice to one of three conditions: 1) a short resaturation (SR) time (TR ⫽ 15 s), a characteristic of the MSOSA cohort; 2) a long resaturation (LR) time (TR ⫽ 90 s), which has been used in other published mouse models of CIH (9, 12, 32, 39, 50, 56, 68, 69); and 3) sham, continuous air at flow rates similar to SR and LR. We examined the effect of each condition on urinary 8,12-iso-iPF2␣-VI over a 3-h period of CIH. Based on the observed variability and effect sizes seen in an independent pilot sample using pooled groups of mice undergoing each condition (for details, see Data Analysis and Statistical Approaches: Mouse Studies), 40 new C57BL/6 (National Institute on Aging) male mice, 6 mo old were used to measure urinary 8,12-iso-iPF2␣-VI levels. Twenty mice underwent all three conditions (sham, SR, and LR, with SR and LR separated by 1 wk). Five mice

•

Lim DC et al.

Table 1. HyCon CIH schedule for MouseOx study (sinusoidal and nonsinusoidal) and urinary 8,12-iso-iPF2␣VI study (nonsinusoidal only) TD-Desat (s)/TR-Resat (s) Nonsinusoidal CIH schedule* Hycon FiO2 SH (high FiO2)-SL (low FiO2)

Sinusoidal CIH schedule

Short resaturation (SR)

Long resaturation (LR)

0.21-0.18 0.21-0.17 0.21-0.16 0.21-0.15 0.21-0.14 0.21-0.13 0.21-0.12 0.21-0.11 0.21-0.10 0.21-0.09 0.21-0.08 0.21-0.07 0.21-0.06 0.21-0.05

NA NA NA 30/30 30/30 30/30 30/30 30/30 30/30 30/30 30/30 30/30 30/30 30/30

15/15 20/15 24/15 27/15 30/15 32/15 38/15 45/15 45/15 NA NA NA NA NA

15/90 20/90 24/90 27/90 30/90 32/90 38/90 45/90 45/90 NA NA NA NA NA

*Nonsinusoidal short resaturation time TD and TR are derived from MSOSA data shown in Fig. 3A; NA, not applicable because it was not performed. CIH, cyclical intermittent hypoxia.

underwent sham, then SR (separated by 1 day); five mice underwent sham and LR only (separated by 1 day); five mice underwent SR only; and five mice underwent LR only. In total, 90 urine samples were obtained: 30 sham, 30 SR, and 30 LR samples. Two-hundred microliters of each urine sample is required for measurement of 8,12-isoiPF2␣-VI, and 10 ␮l is required for creatinine; therefore, samples of ⬍210 ␮l were excluded from further analyses. HyCon settings for 8,12-iso-iPF2␣-VI study were selected to ensure results reflected resaturation time, compatible with human studies. We used data from the MouseOx study to select a schedule where the short and long resaturation programs resulted in similar SaO2 nadirs. Settings for SR schedule were FIO2 0.21 to 0.12 (SaO2 65%) over 38 s and from 0.12 to 0.21 over 15 s for 3 h, a total of 204 cycles; settings for LR schedule were FIO2 0.21 to 0.13 (SaO2 65%) over 32 s and from 0.13 to 0.21 over 90 s for 3 h, a total of 89 cycles. The sham condition involved placing a mouse in a HyCon home cage exposed to continuous room air at a flow similar to that in the CIH cages. Urine collection for 8,12-iso-iPF2␣-VI study was performed such that each sample represented one mouse exposed to 3 h (0800 to 1100) of SR, LR, or sham. Before placing a single mouse into one cage, each mouse was held by the scruff of the neck and a light bladder massage was performed (20, 42, 72) to obtain an initial void (hour 0). Three clear 96-well plates were placed on the bottom of each HyCon home cage to collect urine and collected within 5 min to prevent evaporation, then stored on ice in the dark. In addition, hourly bladder massage (at hours 1, 2, and 3) was performed and this urine was pooled with urine collected within the cage for that hour. After each hourly urine collection was completed, urine was immediately put on dry ice and stored at ⫺80°C. We combined urine (not including the initial void) so that one sample reflected a 3-h urine collection during CIH or sham. Pooling hourly samples was required to obtain a sufficient volume of urine for analyses. Results for 8,12-iso-iPF2␣-VI were normalized to creatinine and both were measured by LC-MS/MS as previously described (58, 64). In short, samples for 8,12-iso-iPF2␣-VI underwent solid-phase extraction (SPE) and were analyzed using Acquity UPLC and Waters Xevo TQ-S tandem quadrupole mass spectrometer (Waters, Milford, MA) used in the electrospray negative (ES⫺), multiple reaction monitoring (MRM) mode.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH

Data Analysis and Statistical Approaches: Human Studies Clinical characteristics. Continuous characteristics are summarized using means, standard deviations (SDs), and medians and compared among Apnea-Hypopnea Index (AHI) groups in the MSOSA data using a parametric analysis of variance (ANOVA) or nonparametric Kruskal-Wallis test, where appropriate. Categorical variables are summarized using frequencies and percentages and were compared among AHI groupings using chi-square or Fisher’s exact tests. AHI groups were defined based on clinically relevant definitions as controls (AHI ⬍ 5), mild OSA (AHI 6 –15), moderate OSA (AHI 16 –30), and severe OSA (AHI ⬎ 30). If a significant difference was observed in omnibus tests, we subsequently performed post hoc pairwise comparisons between groups to identify the specific pairwise differences contributing to overall significance. MSOSA: oxygen desaturation and resaturation characteristics. Within the MSOSA sample, each individual oxygen desaturation/ resaturation cycle was assessed for the amplitude, nadir, and duration. These measures were then used to 1) visually compare the characteristics of desaturations and resaturations and 2) assess the withinsubject variability of desaturation nadirs and the association with the total number of desaturations. To compare characteristics of desaturation/resaturation, event duration was plotted using the least squares mean and 95% confidence interval (CI) derived from a repeatedmeasures ANOVA model. The events were grouped in two ways: 1) to maintain the paired combinations of desaturations and resaturations, events were grouped based on the desaturation amplitude; and 2) to make the estimates comparable within a given amplitude, desaturations and resaturations were unpaired and grouped by their respective individual amplitudes. To assess desaturation variability, we ranked each subject based on their total number of desaturation events and plotted the subject-specific mean and SD of the desaturation nadirs against this ranking. We also calculated the coefficient of variation of the nadirs [CV ⫽ (SD of nadir)/(mean nadir)] and assessed its correlation with the total number of events using Spearman’s rank correlation coefficients. After filtering out desaturation/ resaturation events with amplitude ⬍3%, a total of 7,232 events were included in these analyses, with the median number of events per subject equal to 91 (range: 2– 639 events). Data Analysis and Statistical Approaches: Mouse Studies Assessment of changes in SaO2 to different FIO2 using the new CIH system: MouseOx study. Using the analysis from the sinusoidal desaturation/resaturation program, across FIO2 values relevant to our SR and LR schedules (e.g., 21-16 to 21-10), the SD of the zenith values ranged from 2.2 to 3.2 (mean ⫽ 2.7) and the SD of the nadirs ranged from 3.7 to 6.0 (mean ⫽ 4.9). For studies in different inbred strains, given these estimates, a total of five mice per group results in ⬎80% power to observe a mean difference of at least 5% in zenith O2 saturation values between strains and a mean difference of at least 9% in nadir O2 saturation levels at an ␣ ⫽ 0.05 using a standard t-test. Given that these analyses were meant to be exploratory and descriptive, we decided that powering the analysis based on this magnitude of O2 differences was sufficient. Data were carefully reviewed for cycles deemed as acceptable (i.e., without poor signal detection). Program-specific (SR and LR) least squares mean and standard error (SE) SaO2 values for both the zenith and nadir of all cycles were estimated and compared using a repeated-measures ANOVA model, accounting for multiple measures per mouse, separately for each FIO2 cycle and mouse strain. Given no strong a priori hypothesis regarding the differences in the two programs, two-sided P values were computed to assess whether the zenith or nadir were significantly different between the SR and LR programs. Assessment of effect of short and long desaturation times on oxidative stress. The number of mice used in analyses comparing urinary 8,12-iso-iPF2␣-VI levels among CIH conditions was based on estimates of urinary 8,12-iso-iPF2␣-VI collected in an independent

•

Lim DC et al.

547

pilot study (unpublished), collecting pooled urine from five mice undergoing each of the three conditions in 2-h bins across an 8-h period. Using these data, we estimated the mean urinary 8,12-isoiPF2␣-VI levels for each condition based on the first two 2-h bins, and treated all data as replicates to estimate the pooled SD used in the pairwise comparison. Based on these estimated mean differences and pooled SDs, and using a one-sided ␣ ⫽ 0.05, we estimated that 15 mice per group were required to have 90% power to observe the same difference between LR and SR as seen in the pilot data, 8 mice to have 90% power to observe the same difference between SR and sham, and 7 mice for 90% power to find a difference between LR and sham. Therefore, 15 samples per condition is an appropriate sample size to maintain power across all comparisons. Since we determined that on average a minimum of ⬃3 h was required to obtain a sufficient amount of urine from an individual mouse, rather than a pooled sample, we then doubled this number per condition to ensure we would have at least 15 samples with a large enough urine volume for 8,12-iso-iPF2␣-VI calculations. Samples where volumes were ⬍210 ␮l, the volume required for accurate assessment of 8,12-iso-iPF2␣-VI, were excluded (n ⫽ 24; 12 sham, 8 LR, and 4 SR) from final analyses. Program-specific (SR, LR, and sham) least squares mean and SE urinary 8,12-iso-iPF2␣-VI levels were estimated using repeated-measures ANOVA models. In addition, a linear mixed model analysis was performed to assess whether there was a linear dose response in the levels of 8,12-iso-iPF2␣-VI across the three programs, i.e., whether 8,12-iso-iPF2␣-VISR ⬎ 8,12-iso-iPF2␣-VILR ⬎ 8,12-iso-iPF2␣VISHAM. While studying higher frequency and faster rate of desaturation/resaturation has not been systematically performed in animal or in vitro models of CIH, assuming that relevant enzymes that are oxygen dependent (65) are still generating superoxide or hydrogen peroxide (45), the rate by which the electron acceptor (oxygen) is introduced should influence the rate of production of free radicals. Since the SR program introduces oxygen at a faster rate, we hypothesized that the rate of reoxygenation will affect the amount of free radicals produced, such that urinary 8,12-iso-iPF2␣-VI levels will be higher in the SR program than the levels in the LR program, and levels in the LR program will be higher than in sham. Given the directionality of this hypothesis, one-sided P values were computed to test the hypothesis of a linear dose response across programs, as well as to examine pairwise comparisons between SR, sham, and LR conditions. As an exploratory analysis based on the observed results of these a priori pairwise comparisons, we also compared the 8,12-iso-iPF2␣-VI levels in the short resaturation group vs. the average in the combined sham and long resaturation groups (discussed further in RESULTS, Mouse Studies). RESULTS

Human Studies MSOSA clinical characteristics. MSOSA subjects (n ⫽ 47) had a mean (⫾SD) age of 47.0 ⫾ 11.3 yr, a body mass index (BMI) of 31.8 ⫾ 4.7 kg/m2, and 68.1% were male. They had an AHI of 34.2 ⫾ 24.0 events/h. The oxygen-desaturation index (ODI) (the number of times per hour of sleep that the blood’s oxygen level drops by 3% or more from baseline), SaO2 nadir, and % time SaO2 ⬍ 90% were on average 17.2 ⫾ 19.5 events/hour, 82.9 ⫾ 7.7%, and 7.1 ⫾ 17.1% (Table 2). There were no statistically significant differences among the mild (AHI 6 –15), moderate (AHI 16 –30), and severe (AHI ⬎ 30) groups in sex (P ⫽ 0.226) or age (P ⫽ 0.790). There was a borderline nonsignificant difference in BMI (P ⫽ 0.060), with more severe OSA subjects being more obese on average.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

548

Simulating OSA with CIH

•

Lim DC et al.

Table 2. Demographics of 47 MSOSA subjects Measure

Overall (N ⫽ 47)

AHI 6–15 (N ⫽ 11)

AHI 16–30 (N ⫽ 16)

AHI ⬎30 (N ⫽ 20)

P†

Male Age, yr BMI, kg/m2 AHI ODI SaO2 nadir, % % Time SaO2 ⬍ 90

32 (68.1%) 47.0 ⫾ 11.3 [47.6] 31.8 ⫾ 4.7 [31.6] 34.2 ⫾ 24.0 [24.7] 17.2 ⫾ 19.5 [10.2] 82.9 ⫾ 7.7 [86.0] 7.1 ⫾ 17.1 [0.4]

5 (45.5%) 46.4 ⫾ 12.6 [50.0] 29.4 ⫾ 3.9 [28.8] 10.8 ⫾ 1.7 [11.4] 2.1 ⫾ 3.1 [0.9] 89.6 ⫾ 2.9 [90.0] 1.7 ⫾ 5.2 [0.0]

12 (75.0%) 45.8 ⫾ 11.3 [47.6] 31.4 ⫾ 4.9 [31.1] 21.3 ⫾ 4.3 [20.6] 7.9 ⫾ 4.4 [7.7] 85.1 ⫾ 3.8 [86.0] 1.3 ⫾ 3.2 [0.3]

15 (75.0%) 48.3 ⫾ 11.1 [49.1] 33.5 ⫾ 4.5 [33.5] 57.4 ⫾ 18.8 [57.8] 32.9 ⫾ 21.0 [28.9] 77.4 ⫾ 8.2 [78.5] 14.7 ⫾ 24.1 [4.3]

0.2255 0.7900 0.0596 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.0003

Results are presented as N (%) or means ⫾ SD [median]. MSOSA, Molecular Signatures of Obstructive Sleep Apnea Cohort; BMI, body mass index; AHI, apnea-hypopnea index; ODI, oxygen desaturation index. †P value from exact test (categorical outcomes) and ANOVA or Kruskal-Wallis test (continuous variables).

MSOSA: obstructive events have a resaturation time that stays relatively constant while oxygen desaturation time increases with increasing magnitude of desaturation. When looking at oximetry data, we can qualitatively see that an obstructive event results in an oxygen desaturation/resaturation cycle that is not sinusoidal (Fig. 2A). Using the MSOSA data, we provide quantitative measurements for these nonsinusoidal cycles. Figure 3A shows the result with desaturation and resaturation grouped by the desaturation amplitude for every desaturation/resaturation pair; Fig. 3B shows this relationship for desaturations and resaturations grouped by their respective amplitudes (i.e., unpaired). The data shown in these figures reveal two important characteristics: 1) as desaturation amplitude increases, time for desaturations also increases in an almost linear fashion; e.g., as the amplitude increases, average

desaturation time increased from 16 to 54 s; 2) in contrast, resaturation time remains relatively constant regardless of the degree of desaturation or resaturation amplitude; e.g., ranging from 11 to 19 s in the two figures. Thus the shape of the oxygen desaturation/resaturation cycle is not sinusoidal. Therefore, to more accurately simulate the human desaturation/resaturation curve in a mouse model of CIH, desaturation time should be longer than resaturation. For example, based on our data, if we wanted to program a schedule with an SaO2 desaturation of 25% and a corresponding SaO2 resaturation of 25% in a mouse CIH model, the desaturation time should be ⬃40 s, while the resaturation should be ⬃17 s. MSOSA: within- and between-subject variability in a night of sleep. Based on sleep study data, no subject desaturates to the same nadir all night; there is variability in the oxygen

Fig. 2. Patient with severe obstructive sleep apnea (OSA). A: oxygen desaturation/resaturation cycle is not sinusoidal, and each cycle does not reach the same nadir. A 10-epoch window of 30-s epochs from a patient with severe OSA demonstrating significant oxygen desaturations with a longer downward slope (% oxygen desaturation over time) and a more rapid upward slope (% oxygen resaturation over time). (F3)-(M2) ⫽ frontal lead grounded to mastoid 2; (C3)-(M2) ⫽ central region 3 grounded to mastoid 2; (O1)-(M2) ⫽ occipital lead 1 grounded to mastoid 2. Bar ⫽ 30 s. B: oxygen desaturation/resaturation has considerable variability in patients with OSA that includes different nadirs and periods of normoxia. Oxygen saturation levels over the course of a night of study in an OSA patient with severe OSA [apnea-hypopnea index (AHI) ⫽ 83.7 with arterial oxygen saturation (SaO2) nadir 53%], time lapse 6 h, 45 min. There is considerable variability in the oxygen saturation nadirs across the night, and there are periods of normoxia, even if it is during wake.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH

•

Lim DC et al.

549

Fig. 3. Molecular Signatures of Obstructive Sleep Apnea Cohort (MSOSA): average oxygen desaturation time increases with increased amplitude of desaturation but oxygen resaturation time stays constant. The mean and 95% confidence intervals for the duration are shown for desaturations and resaturations from 47 MSOSA subjects. A: desaturation/resaturation events are paired, grouping both the desaturations and resaturations based on the desaturation amplitude. B: a more direct comparison within a specific amplitude bin can be made by grouping events based on their individual amplitudes. Based on both of these graphs, we observe that desaturations increase in duration with larger amplitudes. Conversely, resaturations have a relatively constant duration across amplitudes. This lack of sinusoidal nature in the oxygen desaturation/resaturation curve should be accounted for in mouse models of cyclical intermittent hypoxia (CIH).

desaturation/resaturation characteristics across the night within a subject, including periods of normoxia (Fig. 2B) from, e.g., nocturia. We quantified this variability for each OSA subject in the MSOSA cohort. The mean and SD of desaturation nadirs for each of the 47 subjects is shown in Fig. 4A, ordered based on the total number of desaturation events throughout the night. In general, subjects with a fewer number of total desaturations had a higher mean nadir (largely above 90%) with a smaller SD; subjects with many total desaturations had lower mean nadirs with a larger SD. For example, in the most severe subject with the lowest SaO2 nadir, not all events were to the same nadir, but varied widely. This is illustrated in Fig. 4B, which shows subject-specific coefficients of variation for desaturation nadirs against the total number of desaturations. We observe a significant positive correlation between these two measures (Spearman’s rho ⫽ 0.668, P ⬍ 0.0001) (Fig. 4B). In

addition to quantifying variability within one subject, we see variability between subjects. There are subjects with few events that have little variability and other subjects with few events with a wide variability. Conversely, there are subjects with many events with little variability and subjects with many events with wide variability in SaO2 nadir. Mouse Studies Changes in HyCon hardware and software allow programming of oxygen desaturation/resaturation slopes and variability similar to those in OSA subjects. To accommodate both the mixing of gasses at a higher FIO2 and faster cycling, a twovalve system is required (Fig. 1E). Software upgrades enabled input of the following parameters: SH (the Set High FIO2, the desired “normoxia”); SL (the Set Low FIO2, the desired “hyp-

Fig. 4. MSOSA: variability of oxygen nadirs during one night of sleep. A: the subject-specific mean and SD of the desaturation SaO2 nadir is shown for participants ranked by the total number of desaturation events. In general, it appears that the mean nadir decreases and SD increases with an increasing number of events. B: the subject-specific coefficients of variation for the desaturation nadirs [CV ⫽ (SD of nadir)/(mean nadir)] are shown, plotted against the total number of desaturation events. As illustrated by the regression line, there is a significant positive correlation (Spearman’s rho ⫽ 0.652, P ⬍ 0.0001) between the variability in the nadirs and the number of events. J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

550

Simulating OSA with CIH

oxia” nadir); TD (desaturation time, time from SH to SL); TR (resaturation time, time from SL to SH); TPH (time during plateau at SH); and TPL (time during plateau at SL). Figure 5A is a schematic of how each cycle can have a unique SH, SL, TD, TR, TPH, and TPL. Figure 5B is a real-time strip demonstrating our ability to program oxygen desaturation and resaturation slopes similar to that seen in subjects (Figs. 2A and 3A).

•

Lim DC et al.

Specifically, this modified system allows slower desaturation with more rapid resaturation. Software upgrades in the HyCon also allow variation in O2 nadirs across sleep (Fig. 5C), similar to that seen in OSA subjects (Fig. 2B), including, if desired, periods of normoxia. Currently, it is possible to program up to 100 periods (P) per 24 h, with each period capable of delivering unique values of SH, SL, TD, TR, TPH, and TPL. This will

Fig. 5. HyCon oxygen desaturation, resaturation, and variability can be programmed to reflect patients with OSA. A: schematic defining fraction of inspired oxygen (FIO2) at the Set High Point (SH) as the FIO2 at the High set point, and Set Low Point (SL) as the FIO2 at the Low set point. TD is the time of desaturation (s), TR is the time of resaturation (s), TPH is the plateau at SH (s), and TPL is the plateau at SL (s). We are currently able to set up to 100 different periods (P), with each period having unique settings (SH, SL, TD, TR, TPH, TPL) and ranging from minutes to hours. P1 is period 1 and P2 is period 2. B: real-time strip of the HyCon demonstrating SH ⫽ 20.9%, SL ⫽ 9%, TD ⫽ 39 s, TR ⫽ 14 s, TPH ⫽ 5 s, and TPL ⫽ 2 s. C: programming variability of desaturation and resaturation throughout sleep within the HyCon. The HyCon is able to deliver variability throughout sleep as seen by the EO2S tracing; here the start time was 0900 and the end time was 1600, with a time lapse of ⬃7 h. J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH

•

551

Lim DC et al.

Fig. 6. HyCon settings overlap with External O2 sensor and MouseOx demonstrates dynamic 1:1:1 cycling rate. Top: simultaneous HyCon settings (solid line denotes HyCon settings and dashed line denotes what the HyCon O2 sensor measures to feedback to the software which then regulates the valves). Middle trace: external O2 sensor placed at the level of the mouse’s nose. Bottom trace: oxygen saturation measured by MouseOx. All tracings are overlapped in time and demonstrate a 1:1:1 cycling rate.

allow investigators to address whether variability in frequencies, amplitude, nadir, and durations of oxygen desaturation play differential roles in influencing changes in gene and protein expression. Measurement of changes in oxygen saturation with this new system: MouseOx study. We first investigated the temporal relationship between the various measures of oxygen. Figure 6 displays, in real time, the overlap of the HyCon settings (top panel, straight line), the HyCon O2 sensor (top panel, dotted line), the external O2 sensor (middle panel), and the saturation from the MouseOx sensor (bottom panel). The external O2 sensor signal lags behind that of the HyCon because gasses are vacuumed through a tube (⬃3 ft in length) before the oxygen concentration is measured and a chemical reaction transpires prior to evaluating the oxygen concentration. The MouseOx readings are also delayed behind the HyCon sensors, most likely due to the physiology of oxygen delivery. However, after accounting for these expected delays there is a cycling of 1:1:1, meaning that one cycle delivered by the HyCon corresponds to one cycle measured by the external O2 sensor, which corresponds to one cycle recorded by the MouseOx. We evaluated oxygen saturation with the MouseOx at different FIO2 values. We report the oxygen saturation averages (at the zenith and nadir) at each “dose” of CIH for both the sinusoidal (Table 3) and nonsinusoidal (Table 4) schedule. Our sinusoidal oxygen saturation measurements are similar to that described by Jun et al. (31). At CIH levels with very low nadirs, the oxygen saturation did not return back to the baseline of normoxia if the mouse was sleeping (observation). However, when the mouse woke up, i.e., had an arousal as reflected by movement artifact on the MouseOx, then there was a return

to normoxia. This likely occurs because hypoxia was so severe at the lower FIO2 levels that allowing only 30 s to come back to normoxia may not be adequate to completely normalize blood oxygen levels, unless there is a compensatory further increase in breathing and heart rate to increase oxygen delivery. Thus, with progressive decreases in FIO2, there is not only a progressive decrease in SaO2 nadir, but also a small effect on the zenith (from 94% to 90%). In our nonsinusoidal oxygen saturation programs within the C57Bl/6 strain, there were statistically significant differences in both the SaO2 zenith and nadir between the SR and the LR schedules at different levels of FIO2. In general, we see that the LR program resulted in Table 3. Assessment of changes in oxygen saturation in the new system: MouseOx study using a sinusoidal schedule (C57Bl/6 mice) Mean ⫾ SE FiO2

Zenith

Nadir

21-16 21-15 21-14 21-13 21-12 21-11 21-10 21-9 21-8 21-7 21-6 21-5

92.4 ⫾ 2.7 94.0 ⫾ 1.3 94.1 ⫾ 0.8 94.5 ⫾ 0.9 92.5 ⫾ 1.1 92.7 ⫾ 0.8 92.5 ⫾ 0.7 91.8 ⫾ 0.8 90.5 ⫾ 1.1 89.8 ⫾ 1.2 89.6 ⫾ 1.2 90.7 ⫾ 1.7

84.1 ⫾ 2.9 81.6 ⫾ 1.4 78.1 ⫾ 1.0 76.2 ⫾ 2.1 67.9 ⫾ 2.8 62.8 ⫾ 1.7 61.0 ⫾ 1.4 54.7 ⫾ 2.5 48.9 ⫾ 1.8 44.0 ⫾ 2.0 38.7 ⫾ 1.4 36.6 ⫾ 1.7

Values are means ⫾ SE SaO2 of zenith and nadir.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

552

Simulating OSA with CIH

•

Lim DC et al.

Table 4. Assessment of changes in oxygen saturation in the new system: MouseOx study using a nonsinusoidal schedule Zenith FiO2

SR

LR

21-18 21-17 21-16 21-15 21-14 21-13 21-12 21-11 21-10

97.6 ⫾ 0.6 96.4 ⫾ 0.7 95.0 ⫾ 0.5 94.3 ⫾ 0.4 93.8 ⫾ 0.5 94.0 ⫾ 0.6 93.0 ⫾ 0.8 92.1 ⴞ 1.0 91.8 ⴞ 0.7

97.0 ⫾ 0.6 96.2 ⫾ 0.7 95.4 ⫾ 0.4 95.2 ⫾ 0.4 95.1 ⫾ 0.5 94.5 ⫾ 0.6 94.1 ⫾ 0.8 94.8 ⴞ 1.0 93.9 ⴞ 0.7

21-18 21-17 21-16 21-15 21-14 21-13 21-12 21-11 21-10

92.2 ⫾ 1.3 90.5 ⫾ 1.3 91.1 ⫾ 1.2 90.9 ⫾ 1.4 89.7 ⫾ 1.4 89.3 ⫾ 1.4 88.7 ⫾ 1.0 88.2 ⫾ 1.7 87.3 ⫾ 1.6

91.6 ⫾ 1.4 91.8 ⫾ 1.5 91.2 ⫾ 1.4 90.3 ⫾ 1.6 90.5 ⫾ 1.6 89.8 ⫾ 1.6 88.8 ⫾ 1.1 88.2 ⫾ 1.9 88.6 ⫾ 1.8

Nadir P

SR

C57Bl/6 Mice 0.430 92.6 ⫾ 1.5 0.830 87.7 ⫾ 1.0 0.504 83.2 ⴞ 0.7 0.066 78.9 ⴞ 0.8 0.072 74.7 ⴞ 1.1 0.606 70.8 ⴞ 1.4 0.283 65.8 ⴞ 1.6 0.044 61.4 ⴞ 1.9 0.036 56.7 ⴞ 1.2 NZO/HiLtJ Mice 0.768 85.5 ⫾ 1.0 0.522 81.4 ⫾ 0.9 0.925 80.3 ⴞ 0.8 0.780 77.5 ⫾ 0.9 0.729 74.1 ⫾ 1.2 0.786 69.3 ⫾ 1.5 0.942 66.3 ⫾ 1.8 0.996 62.4 ⫾ 2.1 0.582 59.7 ⫾ 2.8

Zenith

LR

P

SR

LR

89.6 ⫾ 1.5 84.9 ⫾ 1.1 80.1 ⴞ 0.7 76.1 ⴞ 0.8 70.5 ⴞ 1.1 65.2 ⴞ 1.4 60.4 ⴞ 1.6 55.3 ⴞ 1.9 52.3 ⴞ 1.3

0.154 0.057 0.003 0.017 0.009 0.004 0.022 0.025 0.015

99.1 ⫾ 0.1 99.0 ⫾ 0.2 98.6 ⫾ 0.4 98.7 ⫾ 0.2 98.7 ⫾ 0.3 98.7 ⫾ 0.5 98.5 ⫾ 0.3 98.3 ⫾ 0.2 98.6 ⫾ 0.1

99.0 ⫾ 0.1 98.8 ⫾ 0.2 98.9 ⫾ 0.5 98.4 ⫾ 0.2 98.4 ⫾ 0.4 98.2 ⫾ 0.5 98.4 ⫾ 0.4 98.8 ⫾ 0.2 98.8 ⫾ 0.2

83.7 ⫾ 1.1 80.9 ⫾ 1.0 77.9 ⴞ 0.9 75.1 ⫾ 1 71.1 ⫾ 1.4 67.4 ⫾ 1.7 62.1 ⫾ 2 58.4 ⫾ 2.3 55.2 ⫾ 3.1

0.250 0.716 0.045 0.062 0.101 0.401 0.119 0.204 0.284

96.2 ⫾ 0.6 94.2 ⫾ 0.6 93.3 ⫾ 0.8 92.6 ⫾ 1.0 91.2 ⫾ 1.0 91.6 ⫾ 1.0 91.1 ⫾ 1.3 90.7 ⫾ 1.2 90.8 ⫾ 1.1

95.9 ⫾ 0.6 95.4 ⫾ 0.6 94.7 ⫾ 0.9 94.1 ⫾ 1.0 94.3 ⫾ 1.0 94.9 ⫾ 1.0 94.7 ⫾ 1.3 94.4 ⫾ 1.2 94.3 ⫾ 1.1

Nadir P

SR

A/J Mice 0.341 95.5 ⫾ 1.0 0.567 93.0 ⫾ 1.1 0.639 91.3 ⫾ 1.6 0.182 86.9 ⫾ 1.3 0.469 84.1 ⫾ 0.5 0.497 79.4 ⫾ 1.5 0.892 75.0 ⫾ 1.4 0.191 70.4 ⫾ 1.6 0.242 65.0 ⫾ 0.9 129SH/SvImJ Mice 0.768 89.9 ⫾ 1.3 0.522 86.3 ⫾ 1.1 0.925 83.7 ⫾ 1.3 0.780 80.4 ⫾ 1.8 0.729 77.3 ⫾ 2.0 0.786 74.2 ⫾ 2.1 0.942 70.6 ⫾ 2.1 0.996 66.7 ⫾ 1.8 0.582 63.8 ⫾ 1.5

LR

P

94.5 ⫾ 1.1 93.7 ⫾ 1.3 88.6 ⫾ 1.8 86.6 ⫾ 1.4 82.7 ⫾ 0.7 78.4 ⫾ 1.6 76.4 ⫾ 1.6 69.6 ⫾ 1.8 65.2 ⫾ 1.0

0.527 0.726 0.261 0.870 0.100 0.654 0.505 0.753 0.889

88.2 ⫾ 1.3 84.2 ⫾ 1.1 80.9 ⫾ 1.3 78.2 ⫾ 1.8 75.0 ⫾ 2.1 72.7 ⫾ 2.1 70.5 ⫾ 2.1 65.6 ⫾ 1.8 62.0 ⫾ 1.5

0.336 0.158 0.128 0.390 0.426 0.619 0.962 0.654 0.413

Values are means ⫾ SE SaO2 of zenith and nadir in both short (SR) and long (LR) resaturation times. P ⬍ 0.05 between SR and LR at zenith or nadir are in boldface.

significantly lower nadirs at FIO2 levels of 0.21– 0.16 and lower, and the SR program resulted in lower SaO2 zenith values at FIO2 of 0.21– 0.11 and 0.21– 0.10. These differences were not seen in the 129SH/SvImJ, A/J, and NZO/HiLtJ strains; there was only a borderline significant difference (P ⫽ 0.045) in the SaO2 nadirs at FIO2 of 0.21– 0.16 in the NZO. Not unexpectedly, delivering the same FIO2 to different strains does not result in the same SaO2; there were significant differences among the four strains at every level of FIO2 for both the SR (all P ⱕ 4.7 ⫻ 10⫺5) and LR (all P ⱕ 0.006). Differences in oxidative stress with short and long resaturation times. We assessed whether the three conditions had differential effects on oxidative stress, measured as urinary 8,12-iso-iPF2␣-VI. After removing urine samples where volume was not ⬎210 ␮l, the final analysis included 18 sham, 22 long resaturation, and 26 short resaturation samples. Untransformed least squares mean ⫾ SE urinary 8,12-iso-iPF2␣-VI levels normalized to milligram creatinine across the three conditions were 0.36 ⫾ 0.08 ng/mg, 0.29 ⫾ 0.04 ng/mg, and 0.64 ⫾ 0.19 ng/mg, for sham, LR, and SR, respectively. These levels were natural log transformed for our statistical analyses, to satisfy assumptions of normality. When comparing the three conditions, we find that the data are consistent with our a priori hypothesis that the SR program results in the greatest levels of urinary 8,12-iso-iPF2␣-VI (Table 5). When looking at pairwise comparisons between conditions, short resaturation increases mean 8,12-iso-iPF2␣-VI relative to sham by 1.30-fold (90% CI: 1.00 –1.69, P ⫽ 0.049), while long resaturation is not elevated relative to sham (P ⫽ 0.676). We observed a borderline, but nonsignificantly elevated 8,12-iso-iPF2␣-VI for the SR compared with LR (P ⫽ 0.067); we note that the estimated difference was larger between SR and LR and between SR and sham. Given that there was no difference between LR and sham, we combined these groups and compared against SR. In this comparison, the SR had 8,12-iso-iPF2␣-VI levels 1.36

times larger (90% CI: 1.02–1.82, P ⫽ 0.039) than the combined sham ⫹ LR group. DISCUSSION

We demonstrate with data from the MSOSA study of OSA subjects that a cycle of intermittent hypoxia is not sinusoidal with the desaturation time increasing in an almost linear relationship to the degree of hypoxia (based on amplitude), whereas resaturation time is somewhat constant and much faster (⬃15 s), independent of the amplitude. There is also within-subject variation in the degree of desaturations across the night. We implemented these characteristics into our new mouse model of CIH by modifying the Hycon to deliver a shorter resaturation time and allow for variability. The shorter resaturation time may have functional significance, as we demonstrated that even after only 3 h of CIH there is an increase in urinary 8,12-iso-iPF2␣-VI in the short resaturation Table 5. Effects of Sham, LR, and SR on oxidative stress using a mixed-model analysis comparing natural logtransformed 8,12-iso-iPF2␣-VI measurements among conditions in samples with ⬎210 ␮l ln(8,12-iso-iPF2␣-VI)

Comparison

Estimate ⫾ SE

90% CI

†Proportional difference (90% CI)

Dose Response (Sham ⬍ LR ⬍ SR) 0.12 ⫾ 0.10 ⫺0.05, 0.28 — LR-Sham ⫺0.09 ⫾ 0.19 ⫺0.42, 0.24 0.91 (0.66, 1.27) SR-Sham 0.27 ⫾ 0.16 0.00, 0.53 1.30 (1.00, 1.69) SR-LR 0.35 ⫾ 0.23 ⫺0.04, 0.75 1.43 (0.97, 2.11)

P1-sided

0.1211 0.6760 0.0478 0.0666

†Calculated as eEstimate and interpreted as the between-group multiplicative magnitude of difference in untransformed results; e.g., the value in the first group is eEstimate times larger than that in the second.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH

schedule compared with long resaturation plus sham. By designing a rodent model of CIH that is capable of simulating oxygenation characteristics seen in humans, we have a tool that can implement these and other oxygenation characteristics to study the effects of cyclical intermittent hypoxia on various outcomes of CIH. Lessons From the Clinical Cohort The MSOSA data illustrate several observations about apnea/hypopnea events. First, we demonstrate quantitatively that an event’s cycle of intermittent hypoxia is not sinusoidal, but has a much more rapid time for oxygen resaturation compared with that for oxygen desaturation. Furthermore, our measurements of desaturations/resaturation time support the idea that for obstructive apneic/hypopneic events, desaturation time (time to reach the nadir) is variable and is a function of the SaO2 nadir; but resaturation time is relatively constant (about 15–20 s) and rapid. Desaturation of obstructive events is terminated by an arousal at the end of the obstructive event. What stimulates the arousal is a combination of hypoxia and hypercapnia sensing both peripherally and centrally as well as increasing ventilatory effort (4, 21, 27). Whether there is genetic variation in determining one’s arousal threshold which shapes the duration of apnea is an exciting and active area of inquiry (14). Resaturation time of obstructive events is determined by ventilatory drive such that the longer the apneic episode the greater is the ventilatory drive at the end of the apneic episode. For example, a long apnea will result in a larger (faster) increase in ventilation when airway patency is reestablished. Conversely, a short apnea will result in a smaller (slower) increase in ventilation. Thus the relatively constant time for resaturation is likely to be explained by this relationship between apnea length and ventilatory response. The results we present, however, are specific to obstructive events. Central events in Cheyne-Stokes breathing are different in etiology in that they result from the unstable operation of a closed-loop feedback system with a sleep-dependent CO2 apnea threshold (13). The arousal, when it occurs, is not at the end of the apnea but rather at the peak of ventilation during the ventilatory cycle. Given the difference in etiology, it is unclear if resaturation time in Cheyne-Stokes respiration would also be constant and more rapid than desaturation time and requires further study. Ataxic respiration (damage to the medulla oblongata) and Biot’s respiration (damage to the pons that can be caused by opioid use) being of an altogether different etiology is another area that requires further study to assess whether resaturation time is constant and more rapid than desaturation. Another characteristic illustrated by MSOSA is that there is considerable within-subject variability in SaO2 nadir throughout a given night sleep, as well as clear variability between subjects, even with similar number of events. In general, the higher the number of events a subject experiences, the higher the variability of the SaO2 nadir. However, there are subjects with few events that have large variability in nadirs, and there are also subjects with a large number of events but low variability. Whether variability of oxygen nadir has an impact differentially on molecular consequences and disease outcome measures remains an open question for future study.

•

Lim DC et al.

553

Use of Mouse Models of CIH to Simulate Characteristics of Desaturation/Resaturation Found in Subjects with OSA Cyclical intermittent hypoxia systems are not standardized. To understand the variability and clinical relevance of current rodent models of CIH we performed a PubMed search (search terms “sleep,” “intermittent hypoxia,” and “rodent” with filters “Journal Articles,” “English”) identifying 326 articles resulting in 21 distinct rodent systems of CIH that correlated FIO2 to blood arterial oxygen levels (5–7, 9, 12, 15, 17, 22, 24, 32, 37–39, 43, 50, 53, 56, 62, 68, 69, 73). In short, there are many different mechanical delivery CIH systems being used to conduct OSA research in rodent models with no standardizations. CIH systems have a considerable variation in SaO2/PaO2 at a given FIO2. For example, at an FIO2 of 0.05, SaO2 can range from 32.5% (61) to 80 – 85% (50, 68). CIH systems also have considerable variation in schedule with desaturation time, ranging from 12 s (38) to 3 min (9), and considerable variation in resaturation time, ranging from 18 s (38) to 6 min (9). Most schedules use sinusoidal desaturation and resaturation times, e.g., 30 s-30 s (13, 15, 21, 37, 61, 62); 60 s-60 s (5, 43); 90 s-90 s (32, 39); or 120 s-120 s (56). In addition to mechanical differences of gas delivery between CIH systems, there are differences in assessing the performance of CIH systems, i.e., using either arterial blood gasses or continuous MouseOx (Starr Life Sciences, Oakmont, PA). Given mechanical differences between systems, consistently providing details regarding the SaO2 at a given FIO2 would allow data from different laboratories with different systems to be compared. Shorter resaturation results in increased 8,12-iso-iPF2␣-VI. It is well known that a consequence of cellular reoxygenation after sustained hypoxia (e.g., as seen in a stroke or heart attack) is a net increase in ROS (3, 16), with increases in superoxide observed instantly (40, 66) and reaching significance as early as 1 min after reoxygenation (26). Clinical studies indicate that subjects with OSA exhibit oxidative stress (29, 35), as well as the downstream effects of oxidative stress, like the activation of the transcription factor NF␬B (63, 71) and cytokine release (49). Animal studies also provide evidence that CIH leads to oxidative stress (30, 51, 59). Here, we present data that a CIH schedule that includes a faster resaturation time may affect the rate of free radicals produced, consistent with the notion that relevant enzymes are generating superoxide or hydrogen peroxide in response to the rate by which the electron acceptor (oxygen) is introduced into the cell. Thus, to more accurately simulate the effects of OSA, rapid resaturation is likely to be more relevant. Given that more rapid cycling occurs with faster resaturation, there is a difference in the number of cycles between the short resaturation schedule (n ⫽ 209) and the long resaturation schedule (n ⫽ 89). We had considered keeping the number of cycles constant by increasing the long resaturation schedule to 6 h and 40 min and comparing it to a short resaturation schedule for 3 h. This, however, would have confounding effects in that we would be comparing effects over much different time scales. The reason we chose to keep duration of 3 h constant was that, practically speaking, when investigators use a long resaturation schedule of 90 s they do not double the total duration of CIH exposure (per day) to accommodate their decision to do a long resaturation, e.g., they do not “match” the number of cycles to investigators that use a shorter resaturation

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

554

Simulating OSA with CIH

(e.g., 30 s resaturation or 45 s resaturation or 60 s resaturation). While our data are compatible with the hypothesis that faster resaturation will lead to increased oxidative stress, we cannot exclude the possibility that the large number of cycles played a role. Potential Relevance of Desaturation/Resaturation Time and Variability to Pathogenetic Consequences Oxygen sensors within a cell are generally classified to be either very fast responders such as those found within the mitochondria that produce superoxide radicals (28, 65) or those that take longer to respond to hypoxia such as the family of transcription factors, the hypoxia inducible factors (Hifs) (10, 11). The mitochondria are a major source of oxidative stress. An increase in oxidative stress in the shorter, faster resaturation schedule is likely due to an increase in production of oxidative species that over time overwhelms the clearance of oxidative stress, e.g., inadequate antioxidant production. In addition to the present study, there is another study that suggests a faster resaturation results in an increase in the production of oxidative stress. In an in vivo piglet study, hypoxia (PO2 ⬃ 25 mmHg) followed by cardiopulmonary bypass using a fast reoxygen˙ O, conjugated ation protocol demonstrated an increase in N dienes, and malondialdehyde (MDA) (all of which are markers of increased oxidative stress) compared with a slower reoxygenation protocol (47). Although biochemical mechanisms of increased oxidative stress production has been studied in response to sustained hypoxia, in intermittent hypoxia, this has yet to be fully explored. It is our hypothesis that a repetitive, shorter, and faster resaturation time produces more free radicals within the mitochondria compared with a longer resaturation time. Specifically, relevant enzymes that are responsible for generating superoxide or hydrogen peroxide do so in response to the rate by which the electron acceptor, oxygen, is introduced into the cell. Future intermittent hypoxia studies that can identify enzymes involved in both the production and the clearance of oxidative stress in the mitochondria and whether it differs during a faster or slower resaturation time would better clarify the mechanisms involved. Cellular oxygen sensors like the Hifs have been well studied in response to sustained hypoxia; however, fundamental questions surrounding the response of Hifs to intermittent hypoxia have yet to be answered. Specifically, how does each variable of intermittent hypoxia—frequency, amplitude, nadir, and duration, or a combination thereof—affect transcription, translation, and degradation of Hifs? Yuan et al. (77) used an in vitro model of CIH showing Hif1␣ accumulation persisting for at least 90 min into resaturation time compared with almost no persistent accumulation of Hif1␣ with sustained hypoxia. This work highlights two important concepts: 1) data using sustained hypoxia (e.g., cancer models) or sustained hypoxia with a single reoxygenation episode (e.g., ischemia-reperfusion models) are unlikely to have the same consequences of Hif1␣ as that seen in CIH models of OSA, and therefore, extrapolations should not be assumed; and 2) Hif1␣ may be persistently elevated even after returning back to normoxia which is not unlike periods of wake/normoxia during sleep and the period immediately following sleep; long-term effects of persistent Hif1␣ activation is a fascinating area of study. Schumacker et al. (67) demonstrated that another oxygen sensor, complex III

•

Lim DC et al.

of the mitochondria, releases superoxide during hypoxia, allowing ROS to cross into the cytosol to either trigger the release of intracellular calcium or inhibit prolyl hydroxylase from degrading Hif1␣. With so many potential CIH-associated mechanisms to explore, our new rodent model of CIH will allow us to study the effects of oxygen desaturation/resaturation frequency, hypoxic nadir, and duration on cellular oxygen sensors. Limitations Human studies. Data surrounding very low oxygen saturations should be acknowledged with caution due to the intrinsic and known limitations of the pulse oximeter at very low oxygen saturations. We believe that our results can be generalized to most OSA populations. However, additional studies with larger numbers and among populations with various ethnic, age, comorbidity, and BMI distributions would be useful to both better generalize and either confirm or refute the desaturation/resaturation characteristics observed here. Although the MSOSA was not designed to assess the impact of variable oxygen characteristics on differences in outcome measures, e.g., hypertension, stroke, etc., this will be an interesting future direction. Mouse studies. Mice have metabolic rates normalized to body weight that is 7– 8 times less efficient than a human. Mice have a 7- to 8-fold increase in heart rates and respiratory rate, although most mice are housed 10°C below their thermoneutrality and therefore the difference in heart rate may only be 5-fold (25). These scalar differences between mice and humans make it impossible to exactly replicate the patterns of desaturation/resaturation in individual subjects with OSA. However, one would argue that the major characteristics of oxygenation found in OSA subjects should be simulated in mouse models. By incorporating these characteristics, we may find new molecular consequences of CIH that help us better understand the pathophysiological consequences of OSA. When using continuous oxygen saturation devices like the MouseOx, the technology calculates oxygen saturation based on oxygen dissociation curves for mice or rats, not curves based on humans (23). Because Starr Life Sciences has improved their technology over the years, it is important to report which generation of hardware (sensors) and software are used due to notable differences (E. Ayers, Starr Life Sciences, personal communication). Evolving technology that allows continuous measurement of oxygenation would seem to be the optimal method of assessing the performance of a CIH system, allowing standardization of CIH systems. The MouseOx recordings in Tables 3 and 4 are specific to the strains of mice we tested, and specific to the CIH schedules we selected within the Hycon system. We recommend that SaO2 as well as FIO2 data be reported to allow comparison of the results with different schedules and systems when exposed to various conditions. Conclusions Rodent models of CIH are extremely valuable in helping us understand the mechanism for the consequences of OSA. Rather than assume that SaO2 nadir is the only important variable, we have questioned this premise and present quantitative characteristics of oxygen desaturation/resaturation and variability in subjects with OSA. We have implemented these

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH

characteristics into a mouse model of CIH that is practical and commercially available. One characteristic seen in OSA subjects, a shorter resaturation schedule, resulted in increased oxidative stress. Future studies that identify other relevant oxygenation patterns in OSA patients and incorporating them into mouse models of CIH may reveal novel pathways and mechanisms. ACKNOWLEDGMENTS We thank Dr. V. Y. Polotsky for introducing us to the HyCon and S. Savransky for expertise and willingness to implement the numerous requests to modify the HyCon. We thank the Penn team behind the MSOSA data [Trial name: Biomarkers for Obstructive Sleep Apnea (BOSA); URL: http://clinicaltrials.gov/cTR/show/ NCT00834509?term⫽BOSA&rank⫽1; Registration number: NCT00834509]: C. Walsh, F. Pack, B. Staley, W. Wieland, and R. Dietrich; as well as the personnel that assisted with collecting the MouseOx recordings: C. Cozzan, J. Gilliland, R. Dillinger, and C. Huh. Last, we are extremely grateful to J. Lawson for expertise and assistance in measuring 8,12-iso-iPF2␣-VI and to H. Zou and W. Li for running the LC-MS/MS. GRANTS We acknowledge funding sources for this work: National Heart, Lung, and Blood Institute Grants P01-HL-094307 (A. I. Pack), K12-HL-090021 (A. I. Pack); American Academy of Sleep Medicine Foundation grant no. 56-PA-10 (D. C. Lim); National Center for Advancing Translational Sciences Grant UL1TR000003; and the Science Without Borders Program-Coordination for the Improvement of Higher Education Personnel, Ministry of Education, Brazil (L. Marcondes). DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s). AUTHOR CONTRIBUTIONS Author contributions: D.C.L., L.P.C., R.J.G., and A.I.P. conception and design of research; D.C.L., D.C.B., P.P., L.P.C., L.M., and E.Y.K. performed experiments; D.C.L., D.C.B., P.P., L.M., E.Y.K., B.T.K., and A.I.P. analyzed data; D.C.L., B.T.K., X.G., G.M., and A.I.P. interpreted results of experiments; D.C.L., D.C.B., B.T.K., G.M., and A.I.P. prepared figures; D.C.L. and A.I.P. drafted manuscript; D.C.L., D.C.B., B.T.K., X.G., R.J.G., and A.I.P. edited and revised manuscript; D.C.L., D.C.B., P.P., L.P.C., L.M., E.Y.K., B.T.K., X.G., G.M., R.J.G., and A.I.P. approved final version of manuscript. REFERENCES 1. Almendros I, Montserrat JM, Ramirez J, Torres M, Duran-Cantolla J, Navajas D, Farre R. Intermittent hypoxia enhances cancer progression in a mouse model of sleep apnoea. Eur Respir J 39: 215–217, 2012. 2. Ancoli-Israel S, Palmer BW, Cooke JR, Corey-Bloom J, Fiorentino L, Natarajan L, Liu L, Ayalon L, He F, Loredo JS. Cognitive effects of treating obstructive sleep apnea in Alzheimer’s disease: a randomized controlled study. J Am Geriatr Soc 56: 2076 –2081, 2008. 3. Assaly R, de Tassigny A, Paradis S, Jacquin S, Berdeaux A, Morin D. Oxidative stress, mitochondrial permeability transition pore opening and cell death during hypoxia-reoxygenation in adult cardiomyocytes. Eur J Pharmacol 675: 6 –14, 2012. 4. Berry RB, Kouchi KG, Der DE, Dickel MJ, Light RW. Sleep apnea impairs the arousal response to airway occlusion. Chest 109: 1490 –1496, 1996. 5. Chen L, Zhang J, Gan TX, Chen-Izu Y, Hasday JD, Karmazyn M, Balke CW, Scharf SM. Left ventricular dysfunction and associated cellular injury in rats exposed to chronic intermittent hypoxia. J Appl Physiol 104: 218 –223, 2008. 6. Chodzynski KJ, Conotte S, Vanhamme L, Van Antwerpen P, Kerkhofs M, Legros JL, Vanhaeverbeek M, Van Meerhaeghe A, Coussement G, Boudjeltia KZ, Legrand A. A new device to mimic intermittent hypoxia in mice. PLos One 8: e59973, 2013. 7. Coleman CG, Wang G, Park L, Anrather J, Delagrammatikas GJ, Chan J, Zhou J, Iadecola C, Pickel VM. Chronic intermittent hypoxia induces NMDA receptor-dependent plasticity and suppresses nitric oxide signaling in the mouse hypothalamic paraventricular nucleus. J Neurosci 30: 12103–12112, 2010.

•

Lim DC et al.

555

8. Collaborative Cross Consortium. The genome architecture of the Collaborative Cross mouse genetic reference population. Genetics 190: 389 – 401, 2012. 9. Darnall RA, McWilliams S, Schneider RW, Tobia CM. Reversible blunting of arousal from sleep in response to intermittent hypoxia in the developing rat. J Appl Physiol 109: 1686 –1696, 2010. 10. Deschene K, Celeste C, Boerboom D, Theoret CL. Hypoxia regulates the expression of extracellular matrix associated proteins in equine dermal fibroblasts via HIF1. J Dermatol Sci 65: 12–18, 2012. 11. Ding G, Liu HD, Liang HX, Ni RF, Ding ZY, Ni GY, Hua HW, Xu WG. HIF1-regulated ATRIP expression is required for hypoxia induced ATR activation. FEBS Lett 587: 930 –935, 2013. 12. Douglas RM, Ryu J, Kanaan A, Del Carmen Rivero M, Dugan LL, Haddad GG, Ali SS. Neuronal death during combined intermittent hypoxia/hypercapnia is due to mitochondrial dysfunction. Am J Physiol Cell Physiol 298: C1594 –C1602, 2010. 13. Eckert DJ, Jordan AS, Merchia P, Malhotra A. Central sleep apnea: pathophysiology and treatment. Chest 131: 595–607, 2007. 14. Eckert DJ, White DP, Jordan AS, Malhotra A, Wellman A. Defining phenotypic causes of obstructive sleep apnea. Identification of novel therapeutic targets. Am J Respir Crit Care Med 188: 996 –1004, 2013. 15. Fang G, Song D, Ye X, Mao SZ, Liu G, Liu SF. Chronic intermittent hypoxia exposure induces atherosclerosis in ApoE knockout mice: role of NF-kappaB p50. Am J Pathol 181: 1530 –1539, 2012. 16. Fekete A, Vizi ES, Kovacs KJ, Lendvai B, Zelles T. Layer-specific differences in reactive oxygen species levels after oxygen-glucose deprivation in acute hippocampal slices. Free Radic Biol Med 44: 1010 –1022, 2008. 17. Feng SZ, Tian JL, Zhang Q, Wang H, Sun N, Zhang Y, Chen BY. An experimental research on chronic intermittent hypoxia leading to liver injury. Sleep Breath 15: 493–502, 2011. 18. Fletcher EC, Lesske J, Behm R, Miller CC 3rd, Stauss H, Unger T. Carotid chemoreceptors, systemic blood pressure, and chronic episodic hypoxia mimicking sleep apnea. J Appl Physiol 72: 1978 –1984, 1992. 19. Fletcher EC, Lesske J, Qian W, Miller CC 3rd, Unger T. Repetitive, episodic hypoxia causes diurnal elevation of blood pressure in rats. Hypertension 19: 555–561, 1992. 20. Garratt M, McArdle F, Stockley P, Vasilaki A, Beynon RJ, Jackson MJ, Hurst JL. Tissue-dependent changes in oxidative damage with male reproductive effort in house mice. Funct Ecol 26: 423–433, 2012. 21. Gleeson K, Zwillich CW, White DP. The influence of increasing ventilatory effort on arousal from sleep. Am Rev Respir Dis 142: 295–300, 1990. 22. Gonzalez-Martin MC, Vega-Agapito MV, Conde SV, Castaneda J, Bustamante R, Olea E, Perez-Vizcaino F, Gonzalez C, Obeso A. Carotid body function and ventilatory responses in intermittent hypoxia. Evidence for anomalous brainstem integration of arterial chemoreceptor input. J Cell Physiol 226: 1961–1969, 2011. 23. Gray LH, Steadman JM. Determination of the oxyhaemoglobin dissociation curves for mouse and rat blood. J Physiol 175: 161–171, 1964. 24. Greenberg HE, Sica A, Batson D, Scharf SM. Chronic intermittent hypoxia increases sympathetic responsiveness to hypoxia and hypercapnia. J Appl Physiol 86: 298 –305, 1999. 25. Hamlin RL, Altschuld RA. Extrapolation from mouse to man. Circ Cardiovasc Imaging 4: 2–4, 2011. 26. Hashimoto Y, Itoh K, Nishida K, Okano T, Miyazawa Y, Okinaga K. Rapid superoxide production by endothelial-cells and their injury upon reperfusion. J Surg Res 57: 693–697, 1994. 27. Horner RL, Rivera MP, Kozar LF, Phillipson EA. The ventilatory response to arousal from sleep is not fully explained by differences in CO2 levels between sleep and wakefulness. J Physiol 534: 881–890, 2001. 28. Jaeschke H, Mitchell JR. Mitochondria and xanthine-oxidase both generate reactive oxygen species in isolated perfused rat-liver after hypoxic injury. Biochem Biophys Res Commun 160: 140 –147, 1989. 29. Jelic S, Le Jemtel TH. Inflammation, oxidative stress, and the vascular endothelium in obstructive sleep apnea. Trends Cardiovasc Med 18: 253–260, 2008. 30. Jun J, Savransky V, Nanayakkara A, Bevans S, Li J, Smith PL, Polotsky VY. Intermittent hypoxia has organ-specific effects on oxidative stress. Am J Physiol Regul Integr Comp Physiol 295: R1274 –R1281, 2008. 31. Jun JC, Shin MK, Yao Q, Bevans-Fonti S, Poole J, Drager LF, Polotsky VY. Acute hypoxia induces hypertriglyceridemia by decreasing

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

556

32.

33.

34. 35. 36.

37.

38.

39. 40.

41.

42.

43.

44.

45. 46.

47.

48.

49.

50.

51.

Simulating OSA with CIH

plasma triglyceride clearance in mice. Am J Physiol Endocrinol Metab 303: E377–E388, 2012. Kaushal N, Ramesh V, Gozal D. Human apolipoprotein E4 targeted replacement in mice reveals increased susceptibility to sleep disruption and intermittent hypoxia. Am J Physiol Regul Integr Comp Physiol 303: R19 –R29, 2012. Lam JC, Lam B, Lam CL, Fong D, Wang JK, Tse HF, Lam KS, Ip MS. Obstructive sleep apnea and the metabolic syndrome in communitybased Chinese adults in Hong Kong. Respir Med 100: 980 –987, 2006. Lavie L. Obstructive sleep apnoea syndrome—an oxidative stress disorder. Sleep Med Rev 7: 35–51, 2003. Lavie L. Oxidative stress inflammation and endothelial dysfunction in obstructive sleep apnea. Front Biosci (Elite Ed) 4: 1391–1403, 2012. Lawson JA, Li H, Rokach J, Adiyaman M, Hwang SW, Khanapure SP, FitzGerald GA. Identification of two major F2 isoprostanes, 8,12-isoand 5-epi-8, 12-iso-isoprostane F2alpha-VI, in human urine. J Biol Chem 273: 29295–29301, 1998. Lee EJ, Woodske ME, Zou B, O’Donnell CP. Dynamic arterial blood gas analysis in conscious, unrestrained C57BL/6J mice during exposure to intermittent hypoxia. J Appl Physiol 107: 290 –294, 2009. Lesske J, Fletcher EC, Bao G, Unger T. Hypertension caused by chronic intermittent hypoxia—influence of chemoreceptors and sympathetic nervous system. J Hypertens 15: 1593–1603, 1997. Li C, Lu J, Zhang B. Development of a novel chronic intermittent hypoxia chamber. Sleep Breath 16: 177–179, 2012. Littauer A, de Groot H. Release of reactive oxygen by hepatocytes on reoxygenation: three phases and role of mitochondria. Am J Physiol Gastrointest Liver Physiol 262: G1015–G1020, 1992. Lob HE, Schultz D, Marvar PJ, Davisson RL, Harrison DG. Role of the NADPH oxidases in the subfornical organ in angiotensin II-induced hypertension. Hypertension 61: 382–387, 2013. Manoli I, Sysol JR, Li L, Houillier P, Garone C, Wang C, Zerfas PM, Cusmano-Ozog K, Young S, Trivedi NS, Cheng J, Sloan JL, Chandler RJ, Abu-Asab M, Tsokos M, Elkahloun AG, Rosen S, Enns GM, Berry GT, Hoffmann V, DiMauro S, Schnermann J, Venditti CP. Targeting proximal tubule mitochondrial dysfunction attenuates the renal disease of methylmalonic acidemia. Proc Natl Acad Sci USA 110: 13552– 13557, 2013. Martinez D, Vasconcellos LF, de Oliveira PG, Konrad SP. Weight loss and brown adipose tissue reduction in rat model of sleep apnea. Lipids Health Dis 7: 26, 2008. Martinez-Garcia MA, Campos-Rodriguez F, Duran-Cantolla J, de la Pena M, Masdeu MJ, Gonzalez M, Del Campo F, Serra PC, ValeroSanchez I, Ferrer MJ, Marin JM, Barbe F, Martinez M, Farre R, Montserrat JM, Spanish Sleep Network. Obstructive sleep apnea is associated with cancer mortality in younger patients. Sleep Med 15: 742–748, 2014. McCord JM. Oxygen-derived free radicals in postischemic tissue injury. N Engl J Med 312: 159 –163, 1985. McNicholas WT, Bonsigore MR. Sleep apnoea as an independent risk factor for cardiovascular disease: current evidence, basic mechanisms and research priorities. Eur Respir J 29: 156 –178, 2007. Morita K, Ihnken K, Buckberg GD. Studies of hypoxemic/reoxygenation injury: with aortic clamping. XII Delay of cardiac reoxygenation damage in the presence of cyanosis: a new concept of controlled cardiac reoxygenation. J Thorac Cardiovasc Surg 110: 1265–1273, 1995. Musso G, Olivetti C, Cassader M, Gambino R. Obstructive sleep apnea-hypopnea syndrome and nonalcoholic fatty liver disease: emerging evidence and mechanisms. Semin Liver Dis 32: 49 –64, 2012. Nadeem R, Molnar J, Madbouly EM, Nida M, Aggarwal S, Sajid H, Naseem J, Loomba R. Serum inflammatory markers in obstructive sleep apnea: a meta-analysis. J Clin Sleep Med 9: 1003–1012, 2013. Nanduri J, Makarenko V, Reddy VD, Yuan G, Pawar A, Wang N, Khan SA, Zhang X, Kinsman B, Peng YJ, Kumar GK, Fox AP, Godley LA, Semenza GL, Prabhakar NR. Epigenetic regulation of hypoxic sensing disrupts cardiorespiratory homeostasis. Proc Natl Acad Sci USA 109: 2515–2520, 2012. Nanduri J, Wang N, Yuan G, Khan SA, Souvannakitti D, Peng YJ, Kumar GK, Garcia JA, Prabhakar NR. Intermittent hypoxia degrades HIF-2alpha via calpains resulting in oxidative stress: implications for recurrent apnea-induced morbidities. Proc Natl Acad Sci USA 106: 1199 – 1204, 2009.

•

Lim DC et al.

52. Nieto FJ, Peppard PE, Young T, Finn L, Hla KM, Farre R. Sleepdisordered breathing and cancer mortality: results from the Wisconsin Sleep Cohort Study. Am J Respir Crit Care Med 186: 190 –194, 2012. 53. Nisbet RE, Graves AS, Kleinhenz DJ, Rupnow HL, Reed AL, Fan TH, Mitchell PO, Sutliff RL, Hart CM. The role of NADPH oxidase in chronic intermittent hypoxia-induced pulmonary hypertension in mice. Am J Respir Cell Mol Biol 40: 601–609, 2009. 54. Pack AI, Gislason T. Obstructive sleep apnea and cardiovascular disease: a perspective and future directions. Prog Cardiovasc Dis 51: 434 –451, 2009. 55. Peng YJ, Yuan G, Ramakrishnan D, Sharma SD, Bosch-Marce M, Kumar GK, Semenza GL, Prabhakar NR. Heterozygous HIF-1alpha deficiency impairs carotid body-mediated systemic responses and reactive oxygen species generation in mice exposed to intermittent hypoxia. J Physiol 577: 705–716, 2006. 56. Perry JC, D’Almeida V, Souza FG, Schoorlemmer GH, Colombari E, Tufik S. Consequences of subchronic and chronic exposure to intermittent hypoxia and sleep deprivation on cardiovascular risk factors in rats. Respir Physiol Neurobiol 156: 250 –258, 2007. 57. Pratico D, Clark CM, Lee VM, Trojanowski JQ, Rokach J, FitzGerald GA. Increased 8,12-iso-iPF2alpha-VI in Alzheimer’s disease: correlation of a noninvasive index of lipid peroxidation with disease severity. Ann Neurol 48: 809 –812, 2000. 58. Pratico D, Tillmann C, Zhang ZB, Li H, FitzGerald GA. Acceleration of atherogenesis by COX-1-dependent prostanoid formation in low density lipoprotein receptor knockout mice. Proc Natl Acad Sci USA 98: 3358 – 3363, 2001. 59. Ramond A, Godin-Ribuot D, Ribuot C, Totoson P, Koritchneva I, Cachot S, Levy P, Joyeux-Faure M. Oxidative stress mediates cardiac infarction aggravation induced by intermittent hypoxia. Fundam Clin Pharmacol 27: 252–261, 2013. 60. Redline S, Yenokyan G, Gottlieb DJ, Shahar E, O’Connor GT, Resnick HE, Diener-West M, Sanders MH, Wolf PA, Geraghty EM, Ali T, Lebowitz M, Punjabi NM. Obstructive sleep apnea-hypopnea and incident stroke: the sleep heart health study. Am J Respir Crit Care Med 182: 269 –277, 2010. 61. Reinke C, Bevans-Fonti S, Drager LF, Shin MK, Polotsky VY. Effects of different acute hypoxic regimens on tissue oxygen profiles and metabolic outcomes. J Appl Physiol 111: 881–890, 2011. 62. Reinke C, Bevans-Fonti S, Grigoryev DN, Drager LF, Myers AC, Wise RA, Schwartz AR, Mitzner W, Polotsky VY. Chronic intermittent hypoxia induces lung growth in adult mice. Am J Physiol Lung Cell Mol Physiol 300: L266 –L273, 2011. 63. Robinson MA, Baumgardner JE, Otto CM. Oxygen-dependent regulation of nitric oxide production by inducible nitric oxide synthase. Free Radic Biol Med 51: 1952–1965, 2011. 64. Rocca B, Loeb AL, Strauss JF 3rd, Vezza R, Habib A, Li H, FitzGerald GA. Directed vascular expression of the thromboxane A2 receptor results in intrauterine growth retardation. Nat Med 6: 219 –221, 2000. 65. Rouslin W. Mitochondrial complexes I, II, III, IV, and V in myocardial ischemia and autolysis. Am J Physiol Heart Circ Physiol 244: H743– H748, 1983. 66. Rymsa B, Wang JF, de Groot H. O2- release by activated Kupffer cells upon hypoxia-reoxygenation. Am J Physiol Gastrointest Liver Physiol 261: G602–G607, 1991. 67. Schumacker PT, Gillespie MN, Nakahira K, Choi AMK, Crouser ED, Piantadosi CA, Bhattacharya J. Mitochondria in lung biology and pathology: more than just a powerhouse. Am J Physiol Lung Cell Mol Physiol 306: L962–L974, 2014. 68. Skelly JR, Edge D, Shortt CM, Jones JF, Bradford A, O’Halloran KD. Respiratory control and sternohyoid muscle structure and function in aged male rats: decreased susceptibility to chronic intermittent hypoxia. Respir Physiol Neurobiol 180: 175–182, 2012. 69. Snow JB, Kitzis V, Norton CE, Torres SN, Johnson KD, Kanagy NL, Walker BR, Resta TC. Differential effects of chronic hypoxia and intermittent hypocapnic and eucapnic hypoxia on pulmonary vasoreactivity. J Appl Physiol 104: 110 –118, 2008. 70. Tan S, Zhou F, Nielsen VG, Wang ZW, Gladson CL, Parks DA. Increased injury following intermittent fetal hypoxia-reoxygenation is associated with increased free radical production in fetal rabbit brain. J Neuropath Exp Neur 58: 972–981, 1999. 71. Taylor CT, Kent BD, Crinion SJ, McNicholas WT, Ryan S. Human adipocytes are highly sensitive to intermittent hypoxia induced NF-

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org

Simulating OSA with CIH kappaB activity and subsequent inflammatory gene expression. Biochem Biophys Res Commun 447: 660 –665, 2014. 72. Thai KH, Thathireddy A, Hsieh MH. Transurethral induction of mouse urinary tract infection. J Vis Exp 2010. 73. Veasey SC, Davis CW, Fenik P, Zhan G, Hsu YJ, Pratico D, Gow A. Long-term intermittent hypoxia in mice: protracted hypersomnolence with oxidative injury to sleep-wake brain regions. Sleep 27: 194 –201, 2004. 74. Wang Y, Zhang SX, Gozal D. Reactive oxygen species and the brain in sleep apnea. Respir Physiol Neurobiol 174: 307–316, 2010.

•

Lim DC et al.

557

75. Yaffe K, Laffan AM, Harrison SL, Redline S, Spira AP, Ensrud KE, Ancoli-Israel S, Stone KL. Sleep-disordered breathing, hypoxia, and risk of mild cognitive impairment and dementia in older women. JAMA 306: 613–619, 2011. 76. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occurrence of sleep-disordered breathing among middle-aged adults. N Engl J Med 328: 1230 –1235, 1993. 77. Yuan G, Nanduri J, Khan S, Semenza GL, Prabhakar NR. Induction of HIF-1alpha expression by intermittent hypoxia: involvement of NADPH oxidase, Ca2⫹ signaling, prolyl hydroxylases, and mTOR. J Cell Physiol 217: 674 –685, 2008.

J Appl Physiol • doi:10.1152/japplphysiol.00629.2014 • www.jappl.org