Hyperoxia Results in Increased Aerobic Metabolism Following Acute Brain

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by UCL Discovery

Original Article Journal of Cerebral Blood Flow & Metabolism 0(00) 1–11 Hyperoxia results in increased aerobic ! Author(s) 2016 Reprints and permissions: metabolism following acute brain injury sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0271678X16679171 jcbfm.sagepub.com Arnab Ghosh1,*, David Highton1,*, Christina Kolyva2, Ilias Tachtsidis2, Clare E Elwell2 and Martin Smith1,2,3

Abstract Acute brain injury is associated with depressed aerobic metabolism. Below a critical mitochondrial pO2 cytochrome c oxidase, the terminal electron acceptor in the mitochondrial respiratory chain, fails to sustain oxidative phosphorylation. After acute brain injury, this ischaemic threshold might be shifted into apparently normal levels of tissue oxygenation. We investigated the oxygen dependency of aerobic metabolism in 16 acutely brain-injured patients using a 120-min normobaric hyperoxia challenge in the acute phase (24–72 h) post-injury and multimodal neuromonitoring, including transcranial Doppler ultrasound-measured cerebral blood flow velocity, cerebral microdialysis-derived lactate-pyruvate ratio (LPR), brain tissue pO2 (pbrO2), and tissue oxygenation index and cytochrome c oxidase oxidation state (oxCCO) measured using broadband spectroscopy. Increased inspired oxygen resulted in increased pbrO2 [ÁpbrO2 30.9 mmHg p < 0.001], reduced LPR [ÁLPR 3.07 p ¼ 0.015], and increased cytochrome c oxidase (CCO) oxidation (Á[oxCCO] þ 0.32 mM p < 0.001) which persisted on return-to-baseline (Á[oxCCO] þ 0.22 mM, p < 0.01), accompa- nied by a 7.5% increase in estimated cerebral metabolic rate for oxygen (p ¼ 0.038). Our results are consistent with an improvement in cellular redox state, suggesting oxygen-limited metabolism above recognised ischaemic pbrO2 thresh- olds. Diffusion limitation or mitochondrial inhibition might explain these findings. Further investigation is warranted to establish optimal oxygenation to sustain aerobic metabolism after acute brain injury.

Keywords Brain ischaemia, energy metabolism, mitochondria, near infrared spectroscopy, neurocritical care

Received 2 June 2016; Revised 2 September 2016; Accepted 17 October 2016

Introduction reflects oxygen deprivation or a non-ischaemic meta- bolic crisis.8,9 The brain relies on aerobic metabolism to meet its sub- Mitochondria exist and function normally in a near stantial energy needs and, in health, various mechan- anoxic environment, facilitating a diffusion gradient for isms ensure that oxygen (and metabolic substrate) oxygen transport from the microvasculature, and supply is balanced to meet metabolic demand. This bal- ance is often disturbed after acute brain injury in which cerebral hypoxia–ischaemia is a key injury mechanism 1Neurocritical Care, University College London Hospitals, National Hospital for Neurology & Neurosurgery, London, UK associated with poor outcome, irrespective of brain 2 1–3 Department of Medical Physics and Biomedical Engineering, University injury type. Specific neuroprotective therapies have College London, London, UK 4,5 failed to translate into clinical benefit and treatment 3University College London Hospitals National Institute for Health of severe acute brain injury therefore focuses on Research Biomedical Research Centre, London, UK avoiding or minimising secondary cerebral hypoxia– *These authors contributed equally to this work. ischaemia and consequent mitochondrial energy failure by maintaining cerebral oxygen delivery at a level that Corresponding author: 6,7 David Highton, Neurocritical Care, University College London Hospitals, is sufficient to meet metabolic demand. Debate con- National Hospital for Neurology & Neurosurgery, Queen Square, tinues whether depressed aerobic metabolism, which is London WC1N 3BG, UK. common following acute brain injury, predominantly Email: [email protected] 2 Journal of Cerebral Blood Flow & Metabolism offering protection from oxidant damage. Cytochrome limitation or an altered mitochondrial ischaemic c oxidase (CCO), the terminal electron acceptor in the threshold could equally explain these findings. While mitochondrial respiratory chain, is responsible for CCO oxidation status reflects the activity of the respira- reducing oxygen to water. Its low Michaelis–Menton tory chain, it is also dependent on metabolic substrate constant (Km) for oxygen means that oxidative phos- supply, ATP, oxygen, and mediators which modify the 24 phorylation may continue unimpeded in isolated mito- Km for oxygen such as nitric oxide. Understanding 10 chondria with a pO2 less than 1 mmHg. Below a the changes in CCO oxidation status may therefore critical ischaemic threshold, CCO is reduced and, be a useful adjunct for the in-vivo investigation of dif- importantly, oxygen then becomes a rate limiting sub- fusion limitation and mitochondrial dysfunction after strate decreasing oxidative phosphorylation.11,12 acute brain injury. In health, changes of brain tissue pO2 within the We have developed an in-house optical technique, physiological range are not believed to influence cere- incorporating hybrid spatially resolved broadband bral oxygen consumption13 but, following acute brain and frequency domain near infrared spectroscopy injury, a range of disturbances to oxygen transport and (NIRS), optimised for the measurement of the oxida- its utilisation may complicate the relationship between tion state of CCO [oxCCO] in adult brain-injured 25 pO2 and metabolism. Classical ischaemia describes a patients. Spatially resolved cerebral tissue oxygen sat- situation of insufficient oxygen delivery, and therefore uration, also called the tissue oxygenation index (TOI), of maximal extraction of oxygen from haemoglobin, in association with concentrations of oxyhaemoglobin and is characterized by a combination of large oxygen ([HbO2]), deoxyhaemoglobin ([HHb]) and [oxCCO] extraction fraction (OEF) measured by positron emis- may be used to investigate oxygenation of both the sion tomography (PET), cerebral oligaemia, and falling microvasculature and mitochondria.12 A comprehen- 14 cerebral metabolic rate for oxygen (CMRO2). sive multimodal neuromonitoring array, including While prevalent early after acute brain injury, this pic- pbrO2, microdialysis, transcranial Doppler-measured ture is less common beyond the immediate period of cerebral blood flow velocity and NIRS therefore injury after stroke and traumatic brain injury (TBI).15 covers the entire oxygen cascade from the microvascu- Metabolic dysfunction has been identified in the pres- lature (TOI, HbO2, HHb) through the tissue intersti- ence of apparently acceptable tissue oxygenation, where tium (pbrO2) to the mitochrondria ([oxCCO]), and has both diffusion limited oxygen transport and mitochon- potential to predict cellular redox status (microdialysis drial dysfunction have been implicated as alternative LPR, [oxCCO]) and CMRO2 changes estimated using forms of restriction to oxidative metabolism in the pres- NIRS and transcranial Doppler,26 and might therefore 16 ence of a normal interstitial tissue pO2 or OEF. differentiate between diffusion limited oxygen transport Multimodal neuromonitoring with brain tissue pO2 and mitochondrial dysfunction. (pbrO2) and cerebral microdialysis-derived lactate:pyru- The aim of this study was to investigate the oxygen vate ratio (LPR) have enabled investigation of the dependence of mitochondrial metabolism in vivo relationship between oxygen delivery, cerebral tis- following acute brain injury. We hypothesised that nor- sue oxygenation, and cellular redox status in vivo fol- mobaric hyperoxia-induced increases in cerebral lowing TBI,16 aneurysmal subarachnoid haemorrhage oxygen availability would lead to an increase in CCO (SAH),17 and intracerebral haemorrhage (ICH).18,19 oxidation and reduction in microdialysate LPR, sug- Clinical therapy protocols guided by changes in pbrO2 gesting oxygen-limited mitochondrial oxidative metab- seek to maintain oxygen delivery and availability above olism at baseline. 20 a ‘critical’ pO2 threshold for anaerobic metabolism. Although overt ischaemia and anaerobic metabolism has typically been described when pbrO2 falls below Materials and methods 10 mmHg, normobaric hyperoxia and hyperbaric Study participants and protocol hyperoxia may improve LPR and CMRO2 after TBI in the presence of pbrO2 values that are within or After approval by the Research Ethics Committee of above the normal physiological range.21–23 Vespa the National Hospital for Neurology and Neurosurgery et al.16 demonstrated metabolic dysfunction without and Institute of Neurology (04/Q0512/67) and repre- classical ischaemia after TBI based on observation of sentative consent, recordings were carried out in elevated LPR and PET-derived OEF > 0.75. Others 16 sedated, mechanically ventilated acute brain injury have described a similar picture of metabolic dysfunc- patients requiring invasive neuromonitoring to guide tion ‘without hypoxia’ in SAH and ICH.17 However, it clinical management on the neurocritical care unit. is difficult to entirely rule out hypoxia as a cause of such This was performed in accordance with the observations because of the absence of a subcellular Declaration of Helsinki. Inclusion criteria also included marker of oxygenation in these studies. Diffusion baseline inspired fraction of oxygen (FiO2) Ghosh et al. 3

Baseline inspired oxygen Inspired oxygen fraction Inspired oxygen fraction Baseline fraction (typically 30%) 60% 100%

Continuous NIRS monitoring: broadband and frequency-domain spectrometers

2

Microdialysis lactate-pyruvate ratio measured

Arterial blood gases measured

0 60 120 180 210 Time (minutes)

Figure 1. Normobaric hyperoxia protocol and measured variables. less than 0.5. The patients were subject to a normobaric Chromophore concentration was derived from the hyperoxia protocol which consisted of a 60-min epoch broadband spectroscopy component that incorporates of baseline recording followed by two 60-min epochs in a 50 W halogen light source and lens-based spectrograph which FiO2 was increased first to 0.6 and then to 1.0, based on a charge-coupled device camera (PIXIS and a final 30-min epoch during which FiO2 was 512f, Princeton instruments) using the UCLn algo- returned to baseline values (Figure 1). rithm.29 Measurements were made simultaneously at four source-detector separations (20/25/30/35 mm). The Monitored parameters broadband spectrometer-derived concentrations of HbO2, HHb and oxCCO from the 35 mm separation Systemic physiological monitoring included invasive are reported here as we have previously shown that arterial blood pressure (ABP) and pulse oximetry this source-detector separation has the highest brain- 25 (SpO2) measured continuously, and measurement of specificity, particularly for the measurement of oxCCO. arterial blood gases (ABGs), including carbon dioxide The frequency domain component of the system utilises an and oxygen partial pressures (paCO2 and paO2, OxyPlexTS device (ISS Inc., Champaign, IL, USA) mod- respectively). Middle cerebral artery blood flow velocity ified with diodes emitting light at four wavelengths (690, (Vmca) was measured using transcranial Doppler ultra- 750, 790 and 850 nm), and was used to derive the absolute sonography (DWL Doppler Box, Compumedics, optical absorption and reduced scattering coefficients Singen, Germany). Invasive cerebral monitoring com- (maandms, respectively) as previously described,28 and prised pbrO2 (Licox, Integra Neurosciences, Plainsboro, derivation of the differential pathlength factor using the USA) and measurement of LPR by cerebral microdia- diffusion approximation.30 In this study, we report msrec- lysis (M Dialysis AB, Stockholm, Sweden), with cath- orded at 790 nm. An individual differential pathlength eters implanted via a cranial access device (Technicam, factor was calculated for each patient, based on the ma Newton Abbot UK or Licox IP2, Integra and ms measured by the frequency domain spectrometer Neurosciences) or surgically at time of craniotomy. In during the initial minute of recording of the baseline accordance with consensus guidelines, catheters were epoch. The TOI – defined as [HbO2]/([HbO2]þ[HHb]) – placed in peri-lesional tissue in patients with focal was calculated using spatially resolved spectroscopy.31 The TBI or ICH, in the right frontal lobe in patients with 740 nm–900 nm wavelength range was used to resolve for diffuse TBI, or tissue thought to be at risk of ischaemia HbO2, HHb, and water, and TOI calculated using indi- from vasospasm in patients with aneurysmal SAH.27 vidual scattering values measured with the frequency All non-invasive cerebral monitoring was conducted domain system.25 ipsilateral to the invasive monitoring. NIRS data analysis was performed in Matlab 2010b (Mathworks, Natick, MA). Differential concentrations NIRS instrumentation and processing of HbO2, HHb and oxCCO (Á[HbO2], Á[HHb] and Á[oxCCO], respectively) were calculated using the The NIRS apparatus used in this study has been UCLn algorithm.28–30 Changes in total haemoglobin described in detail elsewhere.28 In brief it comprises concentration (Á[HbT]) were calculated as two components – a multidistance broadband spectrom- Á[HbO2] þ Á[HHb] and in haemoglobin difference eter and a multidistance frequency domain spectrometer. concentration (Á[HbDiff]) as Á[HbO2]–Á[HHb]. 4 Journal of Cerebral Blood Flow & Metabolism Data processing Results After manual identification and linear interpolation to remove NIRS signal artefacts, mean values for each The full study protocol was completed in all 16 monitored variable were calculated for individual patients. Patient characteristics are shown in Table 1. epochs for each patient. The continuously monitored Technical failure resulted in the loss of ABP and pbrO2 systemic and cerebral variables (including NIRS) were recordings for one patient but this patient was included synchronized, and a mean value from a period in all analyses, excluding these parameters. Baseline comprising 50% of the epoch which was free from levels for the physiological variables are shown in noise was used for analysis. For intermittently sampled Table 2, the epoch effect for each variable in Table 3, variables (i.e. ABGs and microdialysate LPR), the mean and changes in measured variables in Table 4 and of all readings per epoch (minimum two per epoch) was Figure 2. used as the summary variable for that epoch. Normobaric hyperoxia was associated with statistic- Relative estimated changes in CMRO2 (rCMRO2) ally significant increases in paO2, SpO2 and pbrO2, were estimated for the return-to-baseline epoch com- but there was no change in paCO2 during the study. pared to the baseline epoch using the NIRS Fick equa- While a significant overall epoch effect for Vmca tion (equation (1)) described by Roche-Labarbe et al.32  Vmca SpO2TOI Table 1. Demographic data. rCMRo2¼ : ð1Þ Vmca0 SpO20TOI0 Age (years) 46.5 (39.3–51.5) Sex 6 male, 10 female Primary diagnosis TBI 7 Statistical analysis SAH 8 ICH 1 We used GLIMMPSE, a validated model for power 33 Time to study 36 (25.5–45) calculation in linear mixed models, to conduct a (hours post injury) sample size calculation. Assuming an Á[oxCCO] stand- Admission GCS 7 (4–9) ard deviation of 0.2 mM in each epoch, a total of 16 patients are required to provide a power of 90% in Note: Data expressed as median with IQR. detecting Á[oxCCO] changes of þ 0.1, þ 0.2 and GCS: Glasgow coma score; ICH: intracerebral haemor- þ0.05 mM during the FiO ¼0.6, FiO ¼1.0 and return- rhage; SAH: subarachnoid haemorrhage; TBI: traumatic 2 2 brain injury. to-baseline epochs. Statistical analyses were carried out in R.34 Parameters of interest were analysed using a mixed effects model,35 modelling individual subjects as Table 2. Physiological & optical variables at random and epochs as fixed effects. The significance baseline. of the fixed epoch effect for each variable (i.e. the prob- Variable Baseline value (IQR) ability that the variable was the same across all four epochs) was then estimated using the Likelihood Ratio FiO2 0.325 (0.28–0.35) Test, comparing the mixed effects model to a null model paO2 (kPa) 15.7 (12.5–18.0) comprising only random effects. In variables with an paCO2 (kPa) 4.85 (4.65–4.97) epoch effect probability of <0.05, subsequent pairwise SpO2 (%) 99 (98–99) comparison between baseline and subsequent FiO2 MAP (mmHg) 91.5 (83.2–96.8) epochs (0.6 and 1.0 and return-to-baseline), was per- Vmca (cm.s1) 52.1 (48.7–71.8) formed using Bonferroni-corrected Wilcoxon signed- pbrO2 (mmHg) 17.5 (12.0–24.4) rank tests. The Hodges–Lehman estimate was used to Microdialysate LPR 25.3 (23.5–33.5) calculate the (pseudo)median and per-epoch 95% con- TOI (%) 72.8 (65.5–77.0) fidence intervals. Relative changes in CMRO were 2 DPF 9.33 (6.75–10.8) similarly treated, but no Bonferonni correction was m 1 applied since only the baseline and return-to-baseline s (cm ) 11.1 (6.76–12.4) epochs were compared. All data are expressed as (pseu- DPF: differential pathlength factor; FiO2: inspired oxygen do)median (95% confidence interval) unless otherwise fraction; IQR: inter-quartile range; MAP: mean arterial stated. A Spearman correlation was used to assess the blood pressure; LPR: lactate:pyruvate ratio; paO2: arterial pO2; paCO2: arterial pCO2; pbrO2, brain tissue pO2; TOI: relationship between baseline pbrO2 and LPR and the tissue oxygenation index; SpO2: arterial oxygen saturation; ÁLPR response to normobaric hyperoxia. Statistical ms: optical reduced scattering coefficient; Vmca: middle significance was inferred at p < 0.05. cerebral artery blood flow velocity. Ghosh et al. 5

Table 3. Epoch effects from likelihood ratio test. [oxCCO] and reduced LPR, suggesting a change in mitochondrial redox status and the presence of Variable Chi-squared p oxygen dependent metabolism above traditionally

paO2 171 <0.001 described ischaemic thresholds. Our findings are con-

pCO2 5.63 0.131 sistent with oxygen-limited metabolism in this cohort of patients with acute brain injury, and suggest the pres- SpO2 64.8 <0.001 LPR 9.28 0.026 ence of either oxygen diffusion limitation or mitochon- drial dysfunction and hypoxia–ischaemia despite pbrO 44.6 <0.001 2 ‘normal’ values for p O . Importantly, the [oxCCO] Vmca 8.31 0.04 br 2 and LPR changes are sustained when FiO is returned < 2 HbDiff 27.0 0.001 to baseline after the period of hyperoxia, while the HbT 4.5 0.21 markers of microvascular and brain tissue oxygenation oxCCO 15.1 0.002 (TOI, [HbO2], [HHb], pbrO2) return to their pre- TOI 22.3 <0.001 hyperoxia values. This suggests that improvement in ms 1.06 0.787 cellular metabolism persists beyond the immediate HbDiff: haemoglobin concentration difference; HbT: total period of normobaric hyperoxia, a supposition sup- haemoglobin concentration; LPR: lactate:pyruvate ratio; ported by the elevation in estimated CMRO2 during oxCCO: cytochrome c oxidation state; paO2: arterial the return-to-baseline FiO2 epoch. pO2; paCO2: arterial pCO2; pbrO2: brain tissue pO2; Although the mean baseline pbrO2 of 17.5 mmHg TOI: tissue oxygenation index; SpO2: arterial oxygen sat- in our study lies within some definitions of hypoxia– uration; ms: optical reduced scattering coefficient; Vmca: 36 middle cerebral artery blood flow velocity. ischaemia (<20 mmHg), the majority of previous studies highlight <10 mmHg as a particular risk for elevated LPR and PET markers of ischaemia.37,38 In was observed, post hoc testing identified no single our study, both epochs of the hyperoxia protocol epoch difference from baseline. Normobaric hyperoxia resulted in elevation of pbrO2 well into its ‘normal’ was also associated with statistically significant physiological range, and there was a stepwise increase increases in Á[oxCCO] and reductions in microdialy- in [oxCCO] and reduction in LPR as FiO2 was sate LPR during the 0.6 FiO2 (Á[oxCCO] þ 0.18, increased from 0.6 to 1.0. These findings are not con- p < 0.01; ÁLPR 1.16, p < 0.01) and 1.0 FiO2 sistent with classical hypoxia–ischaemia. There was also (Á[oxCCO] þ 0.32, p < 0.001; ÁLPR 3.07, p < 0.01) no correlation between baseline pbrO2 or LPR and the epochs. These changes persisted in to the return-to- change in LPR, suggesting that hypoxia/ischaemia, baseline epoch (Á[oxCCO] þ 0.22 [p < 0.01] and defined by pbrO2 or LPR, does not affect the brain’s ÁLPR 0.254 [p < 0.01]). Estimated CMRO2 was response to normobaric hyperoxia in this patient higher in the return-to-baseline epoch compared to group. This finding is unsurprising since the LPR was the baseline epoch [ÁCMRO2 107.5% of baseline consistently reduced (3.07 95% CI 4.38–1.61) (95% CI 100.3% – 119.0%, p ¼ 0.039)]. during normobaric hyperoxia despite different baseline There were no changes in Á[HbT] during the study. values for pbrO2 and LPR. [HbDiff] increased during the 0.6 and 1.0 FiO2 epochs Both oxygen diffusion abnormalities and mitochon- (Á[HbDiff] þ 1.18 mM and þ2.17, respectively, both drial dysfunction have been proposed as mechanisms p < 0.001), but there was no significant change during for oxygen becoming a rate limiting substrate for the return-to-baseline epoch compared to baseline. metabolism.15,16,21 Delivery of oxygen to the mitochon- There was a significant increase in TOI during the 0.6 dria is dependent on the gradient of oxygen tension as and 1.0 FiO2 epochs (ÁTOI 2.8% and 6.0% respect- well as the conductance of the tissues. During the study ively, both p < 0.001), with no significant change during period (24–72 h after ictus), cerebral oedema and hence the return-to-baseline epoch. There were no significant perivascular/cellular swelling and microvascular col- changes in optical scattering measured at 790 nm lapse are important factors which increase the diffusion (epoch effect p ¼ 0.786). There was no correlation distance from the microvasculature to mitochondria between baseline pbrO2 or LPR and the ÁLPR response and might necessitate increased oxygen tension to sus- to normobaric hyperoxia (r ¼0.04, p ¼ 0.89; r ¼ 0.01 tain the rate of mitochondrial oxygen delivery. p ¼ 0.98, respectively). However, our findings of sustained metabolic improve- ment ([oxCCO], LPR, CMRO2) on return-to-baseline Discussion FiO2 and therefore baseline paO2, and predicted oxygenation gradients (see below), are not entirely con- We have demonstrated that normobaric hyperoxia- sistent with diffusion limitation as the only patho- induced increase in pbrO2 is associated with increased physiological process. They may also indicate reversal 6 Journal of Cerebral Blood Flow & Metabolism

Table 4. Changes from baseline for measured variables data presented as (pseudo)median (95% confidence interval.

Epoch

FiO2 0.6 FiO2 1.0 Return-to-baseline Á[HbDiff] (mM) 1.18 2.17 0.30 (0.59–2.12) (1.17–4.13) (0.22–1.28) Á[HbT] (mM) 0.13 0.46 0.44 (0.52–0.22) (1.16–0.13) (0.01–0.90) Á[oxCCO] (mM) 0.18 0.32 0.22 (0.08–0.47) (0.11–0.76) (0.06–0.62) ÁTOI (%) 2.8 6.0 0.31 (1.8–5.6) (3.4–10.9) (2.6–2.8)

Relative CMRO2 (%) – – 107.5 (100.3–119.0) Á LPR 1.16 3.07 2.54 (1.93–0.455) (4.38–1.61) (4.38–0.475)

Á pbrO2 (mmHg) 8.44 30.9 2.72 (5.19–12.2) (21.6–43.4) (1.76–9.46) Á MAP (mmHg) 1.19 1.48 0.56 (2.32–4.92) (3.73–8.8) (7.83–7.38)

Á pCO2 (kPa) 0.15 0.114 0.203 (0.0333–0.258) (-0.05–0.258) (0.1–0.425)

Á paO2 (kPa) 14.1 38.7 1.21 (11.3–17) (35–42.3) (1.99–0.1)

Á SpO2 (%) 1.5 1.5 0.831 (1.00–2.00) (1.01–2.00) (1.16–0.778) Vmca (cm.s1) 2.19 2.64 5.19 (1.23–5.62) (2.09–7.47) (0.45–11) ms (cm1) 0.0168 0.0201 0.0785 (0.201–0.173) (0.394–0.297) (0.423–0.354)

FiO2: inspired oxygen fraction; CMRO2: cerebral metabolic rate for oxygen; HbDiff: haemoglobin concentration difference; HbT: total haemoglobin concentration; LPR: lactate:pyruvate ratio; MAP: mean arterial blood pressure;

oxCCO: cytochrome c oxidation state; paO2: arterial pO2; paCO2: arterial pCO2; pbrO2: brain tissue pO2; TOI: tissue oxygenation index; SpO2: arterial oxygen saturation; ms: optical reduced scattering coefficient; Vmca: middle cerebral artery blood flow velocity of mitochondrial dysfunction by normobaric hyper- CCO oxidation increased by 0.32 mM during a mean oxia. Baseline TOI was 73% in our study and this lies pbrO2 change of 30.9 mmHg in the 1.0 FiO2 epoch, and within a physiologically ‘normal’ range for NIRS- returned to 0.22 mM during return-to-baseline FiO2. derived regional cerebral saturation.39 Assuming one- Although the total concentration of CCO in the adult quarter of blood volume is saturated arterial blood, this human brain is unknown, it is approximately 5 mMin predicts a venous saturation of 64% and approximate rats.40 The CCO changes that we observed are therefore venous pO2 of 33 mmHg (using the calculation from likely to reflect an approximate 6% change in its oxi- Menon et al.15), and thus an average difference of 15.5 dation, which is higher than that observed in healthy mmHg between venous blood (33 mmHg) and pbrO2 volunteers during increases in cerebral oxygen delivery (17.5 mmHg). Similar comparisons using PET and or during functional activation,25,28,41,42 but equivalent 22 pbrO2 have described gradients of 10 mmHg and to those described previously in TBI. The oxidation 27 mmHg in normal and impaired brain regions,15 so status of CCO is modified by both mitochondrial our observations are consistent with only a moderate pO2 and metabolic factors (ADP, NAD:NADH), and diffusion distance between the microvasculature and our findings are consistent with both an increase in aer- tissue interstitium. This further supports the notion obic metabolism and/or increased mitochondrial pO2. that isolated diffusion limitation is not the sole mech- Earlier studies have shown an association between cere- anism implicated in oxygen becoming a rate limiting bral oxygen delivery and CCO oxidation in healthy vol- substrate for metabolism after acute brain injury. unteers25,43 and, in animal models, with brain ATP40 Ghosh et al. 7

50 20 paO2 pbrO2 *** 40

30 ) (mmHg)

2 10 -1 pbrO 20 *** *** Vmca (cm.s (kPa) or 2 10 paO *** 0 0

10

6

*** HbT HbDiff 10

4 ***

*** 5 2 *** TOI(%) [HbT]or [HbDiff] (µM)

0 0

0.8 ***

0 0.6 ** * *

1 **

0.4 * 2 LPR oxCCO (µM)

0.2 3

4 0

Baseline FiO2 60% FiO2 100% RTB Baseline FiO2 60% FiO2 100% RTB

Figure 2. Changes in markers of cerebral oxygen delivery and aerobic metabolism during normobaric hyperoxia showing (pseu- do)median changes and 95% confidence interval error bars. *p < 0.05; **p < 0.01; ***p < 0.001. 8 Journal of Cerebral Blood Flow & Metabolism

44 and lactate concentrations. Likewise, the persistent reduced CMRO2 at baseline. However, a smaller CCO oxidation in the return-to-baseline epoch in our study of eight patients with TBI, showed no improve- study suggests either increased aerobic metabolism ment in LPR during 3 h of normobaric hyperoxia.46 (consistent with the measured LPR and estimated Our study demonstrated an improvement in markers CMRO2) and/or an altered Km for O2. It is interesting of aerobic metabolism during a short (120 min) graded to note that nitric oxide is known to increase the Km of hyperoxia challenge. Although [oxCCO], LPR and CCO, and the proposed mechanism of action for nor- CMRO2 remained partially elevated following return- mobaric and hyperbaric hyperoxia is the reversal of this to-baseline FiO2, further assessments were not made nitric oxide effect thereby reducing the threshold at beyond this period so the longevity of the potential which oxygen becomes a rate limiting step in oxidative metabolic benefit of hyperoxia is uncertain. It must be metabolism.24 Thus, our results could theoretically rep- noted that hyperoxia has a variety of deleterious effects resent the breakdown of NO rather than a direct effect through generation of reactive oxygen species, induc- of elevated mitochondrial pO2. Hypoxia-inducible tion of cytotoxic cytokines and immunosuppression, factor 1a (HIF-1a) is another major hypoxia signalling and concerns exist regarding its prolonged use. pathway which inhibits pyruvate dehydrogenase activ- Kilgannon et al.47 identified an association between ity, a branch point controlling oxidative/anaerobic supranormal paO2 and worsened outcome following metabolism, as well as a range of glycolytic enzymes.45 cardiac arrest, while Quintard et al.48 demonstrated Although these effects may also be modified by hyper- an association between normobaric hyperoxia and oxia, it has also been suggested that a reduction in LPR increased microdialysate glutamate, a key mediator of (as seen in our study) is more consistent with an cerebral excitotoxicity, following TBI. While reactive increase in oxidative metabolism since HIF-1a should oxygen species are a major injury mechanism impli- reduce lactate while maintaining LPR.21 While the cated following cerebral ischaemia, they may be gener- median reduction in LPR (3.07) during normobaric ated both by hypoxia (excess of reductive substrate) as hyperoxia in our study is not large, and of unlikely well as by the delivery of excessive oxygen.49,50 Thus, clinical significance in itself, it might reflect a small oxygen therapy must be carefully controlled after acute volume of ischaemic tissue within the larger tissue brain injury. Time-limited application of hyperoxia or volume monitored by the microdialysis catheter. the use of pbrO2 to guide oxygen administration may Furthermore, when considered within the context of limit the potential deleterious effects of unrestrained the increases in aerobic metabolism shown by the oxygen use while minimising the risk of cerebral hyp- other monitoring modalities during normobaric hyper- oxia/ischaemia. A higher pbrO2 might be warranted oxia, it is possible that this small improvement in LPR given the concerns with oxygen diffusion and mito- might indicate patients with an oxygen-dependent def- chondrial inhibition 24–72 h post injury. Normobaric icit in aerobic metabolism that is amenable to treat- hyperoxia frequently results in restoration of pbrO2 ment. Further clinical studies are required to assess into what is usually considered to be a ‘normoxic’ the clinical relevance of such changes in LPR when range, and the consistent increases observed in markers interpreted in association with other monitored vari- of aerobic metabolism even when pbrO2 is greater than ables of aerobic metabolism. the physiological range21–23 might suggest the need for Our findings are consistent with those of several pre- a higher target. Future research should focus on the vious studies. Diringer et al. found no significant relevance of higher pbrO2 targets and additional moni- change in CMRO2 in a PET study of normobaric tored variables that can inform oxygen therapy after hyperoxia in five TBI patients, but the small sample acute brain injury. size precludes definitive conclusions. A pilot study by Our study has several limitations. First, the individ- our group of eight patients with TBI found similar ual monitoring modalities are designed to measure changes in LPR (1.6) and Á[oxCCO] (þ0.21 mM) different aspects of the oxygen cascade and cellular bio- during normobaric hyperoxia to those we report energetics, but each is subject to its own limitations. here.22 Our current study builds on that pilot in three Although differential spectroscopy NIRS methodolo- important aspects – by including a larger number of gies, such as the one we used to measure haemoglobin patients, using an improved NIRS apparatus with and [oxCCO] in this study, are subject to significantly patient-specific measures of differential pathlength more extracranial ‘contamination’ than spatially 39 factor, and incorporating an estimate of CMRO2. resolved spectroscopy techniques, we have previously Nortje et al.21 also demonstrated that normobaric shown that [oxCCO] is a brain-specific signal with neg- hyperoxia in patients with TBI was associated with a ligible contribution from extracranial tissues.25 similar reduction in LPR (mean LPR reduced from 34.1 Furthermore, by measuring scattering and optical path- to 32.5) to our current study, but PET-measured length, we can place greater confidence on the accuracy 40 CMRO2 increased only in regions of interest with a of the measured change. The differential spectroscopy Ghosh et al. 9 methodology that we used to measure changes in CCO assist in predicting oxygen diffusion gradients and util- oxidation is based on the modified Beer–Lambert law isation, exploiting measurement of TOI, pbrO2 and and therefore only able to quantify relative changes CCO, and should be incorporated into future studies. in chromophore concentration from an unknown base- line rather than measure absolute concentrations of Conclusion oxidised and reduced CCO. Nevertheless, changes in CCO have been evaluated in animal models and Standard clinical therapy following acute brain injury shown to be a reliable measure of intracellular energy fundamentally aims to avoid mitochondrial hypoxia in status.40,44,51 Microdialysate LPR is an imperfect meas- order to minimise secondary tissue ischaemia and worse ure of cerebral aerobic metabolism. It reflects the activ- clinical outcomes. pbrO2 and microdialysis-measured ity of cytosolic lactate dehydrogenase, which is in large LPR have been used as surrogates of mitochondrial part reflective of intracellular NADH:NADþ ratio and oxygen availability and its effect on mitochondrial thus related to the ability of mitochondria to produce redox status at the bedside, but clinical application ATP. There are therefore circumstances, including elec- and interpretation of such techniques requires clearly tron leak from the electron transport chain, during defined thresholds for ‘ischaemia’. Our results demon- which LPR can be unchanged in the face of an inability strate an increase in oxCCO, reduction in LPR and 52 of cells to generate energy. Secondly, although we increase in estimated CMRO2 during and following placed the microdialysis catheters in ‘at risk’ tissue in normobaric hyperoxia. These findings are consistent 27 line with consensus guidelines, used the same cranial with increased aerobic metabolism at pbrO2 levels access device for pbrO2 monitoring, and applied the higher than those typically recognised as ‘ischaemic’ NIRS optodes over the frontal region as close as pos- thresholds. Such oxygen-limited metabolism suggests sible to the insertion site of the invasive monitors, it is that hypoxia–ischaemia secondary to oxygen diffusion likely that different tissue volumes and regions were limitation or mitochondrial dysfunction might be interrogated by each device. Similarly, it is difficult to prevalent after acute brain injury, and complicate know exactly what region of interest is represented by assessment of ischaemia using measurement of pbrO2 our estimate of CMRO2 which is derived from Vmca in isolation. Simultaneous measurement of microvascu- (a relatively global measure of hemispheric CBF) and lar, tissue and cellular oxygenation and metabolism has TOI (a regional measure of tissue oxygenation).39 potential to redefine our understanding of ischaemia Finally, we investigated only 16 patients with mixed after acute brain injury. Measurement of the oxidation pathology, but this is a larger patient cohort than status of CCO as a bedside, continuous assessment of many investigations in this field – for example, those mitochondrial energetics over multiple regions of inter- cited above – and the concept of an ischaemic pbrO2 est has considerable potential to guide treatment after threshold is relevant across all the pathologies included. acute brain injury. Spatial heterogeneity is a limiting factor in the study of most acute brain injury types, and we have specifically Funding targeted analysis of the injured hemisphere using a The author(s) disclosed receipt of the following financial sup- comprehensive array of monitoring modalities. port for the research, authorship, and/or publication of this Overall, our findings highlight the difficulties in article: MS is supported in part by the Department of defining thresholds for hypoxia–ischaemia in the Health’s Institute for Health Research Centre’s funding injured brain, and the potential risks to an individual scheme via the UCLH/UCL Biomedical Research Centre. of utilising generic targets to guide clinical manage- AG was supported by a UK Medical Research Council ment. Diffusion limitation and mitochondrial dysfunc- Clinical Research Training Fellowship (G1000292). IT is sup- tion may disrupt the normal relationship between OEF, ported by a Wellcome Trust senior fellowship (104580/Z/14/ Z). This work was supported by the Engineering and Physical p O , and mitochondrial redox status, and this might br 2 Sciences Research Council (EP/K020315/1) and Medical explain some observations of metabolic dysfunction Research Council (17803). ‘without hypoxia’ when either a diffusion barrier to oxygen is present (when pbrO2 and OEF may not reflect Declaration of conflicting interests mitochondrial pO2) or much higher mitochondrial pO2 is required to maintain ATP generation. Further studies The author(s) declared no potential conflicts of interest with employing multimodal monitoring, including [oxCCO], respect to the research, authorship, and/or publication of this might shed further light on the exact nature of this article. metabolic derangement, and in the understanding of oxygen diffusion within the microvascular and intracel- Authors’ contributions lular environments, and oxygen utilisation. Extension AG and DH are joint first authors and contributed equally. of existing computational physiological models53 may AG, IT, CEE, MS designed the study. AG performed the 10 Journal of Cerebral Blood Flow & Metabolism research. AG, CK and DH analysed the data. AG, DH, IT, 15. Menon DK, Coles JP, Gupta AK, et al. Diffusion limited CEE, MS wrote the paper. oxygen delivery following head injury. Crit Care Med 2004; 32: 1384–1390. 16. Vespa P, Bergsneider M, Hattori N, et al. Metabolic crisis without brain ischemia is common after traumatic brain References injury: a combined microdialysis and positron emission 1. Kunz A, Dirnagl U and Mergenthaler P. Acute patho- tomography study. J Cereb Blood Flow Metab 2005; 25: physiological processes after ischaemic and traumatic 763–774. brain injury. Best Pract Res Clin Anaesthesiol 2010; 24: 17. Chen HI, Stiefel MF, Oddo M, et al. Detection of cere- 495–509. bral compromise with multimodality monitoring in 2. Cahill WJ, Calvert JH and Zhang JH. Mechanisms patients with subarachnoid hemorrhage. Neurosurgery of early brain injury after subarachnoid hemorrhage. 2011; 69: 53–63; discussion 63. J Cereb Blood Flow Metab 2006; 26: 1341–1353. 18. Vespa PM. Metabolic penumbra in intracerebral hemor- 3. Qureshi AI, Mendelow AD and Hanley DF. rhage. Stroke 2009; 40: 1547–1548. Intracerebral haemorrhage. Lancet 2009; 373: 1632–1644. 19. Ko S-B, Choi HA, Parikh G, et al. Multimodality moni- 4. Loane DJ and Faden AI. Neuroprotection for traumatic toring for cerebral perfusion pressure optimization in brain injury: translational challenges and emerging comatose patients with intracerebral hemorrhage. therapeutic strategies. Trends Pharmacol Sci 2010; 31: Stroke 2011; 42: 3087–3092. 596–604. 20. Geukens P and Oddo M. Brain tissue oxygen monitoring 5. Tymianski M. Neuroprotective therapies: preclinical in neurocritical care. In: Vincent J-L (eds.) Annual update reproducibility is only part of the problem. Sci Trans in intensive care and emergency medicine 2012. Berlin, Med 2015; 7: 299fs32. Heidelberg: Springer, 2012, pp.735–745. 6. Connolly ES, Rabinstein AA, Carhuapoma JR, et al. 21. Nortje J, Coles JP, Timofeev I, et al. Effect of hyperoxia Guidelines for the management of aneurysmal subarach- on regional oxygenation and metabolism after severe Crit Care noid hemorrhage: a guideline for healthcare professionals traumatic brain injury: preliminary findings*. Med 2008; 36: 273–281. from the American Heart Association/American Stroke 22. Tisdall MM, Tachtsidis I, Leung TS, et al. Increase in Association. Stroke 2012; 43: 1711–1737. cerebral aerobic metabolism by normobaric hyperoxia 7. Brain Trauma Foundation, American Association of after traumatic brain injury. J Neurosurg 2008; 109: Neurological Surgeons, Congress of Neurological 424–432. Surgeons. Guidelines for the management of severe trau- 23. Kumaria A and Tolias CM. Normobaric hyperoxia matic brain injury. J Neurotrauma 2007; 24(Suppl 1): therapy for traumatic brain injury and stroke: a review. S1–S106. Br J Neurosurg 2009; 23: 576–584. 8. Vespa PM. Brain hypoxia and ischemia after traumatic 24. Cooper C. Competitive, reversible, physiological? brain injury. JAMA Neurol 2016; 73: 504. Inhibition of mitochondrial cytochrome oxidase by 9. Veenith TV, Carter EL, Geeraerts T, et al. nitric oxide. IUBMB Life 2003; 55: 591–597. Pathophysiologic mechanisms of cerebral ischemia and 25. Kolyva C, Ghosh A, Tachtsidis I, et al. Cytochrome c diffusion hypoxia in traumatic brain injury. JAMA oxidase response to changes in cerebral oxygen delivery in Neurol 2016; 73: 542–550. the adult brain shows higher brain-specificity than 10. Krab K, Kempe H and Wikstro¨m M. Explaining the haemoglobin. Neuroimage 2014; 85(Pt 1): 234–244. enigmatic KM for oxygen in cytochrome c oxidase: a 26. Roche-Labarbe N, Carp SA, Surova A, et al. kinetic model. BBA – Bioenergetics 2011; 1807: 348–358. Noninvasive optical measures of CBV, StO2, CBF 11. Rolett EL, Azzawi A, Liu KJ, et al. Critical oxygen ten- index, and rCMRO2 in human premature neonates’ sion in rat brain: a combined (31)P-NMR and EPR oxim- brains in the first six weeks of life. Hum Brain Mapp etry study. Am J Physiol Regul Integr Comp Physiol 2000; 2009; 31: 341–352. 279: R9–R16. 27. Hutchinson PJ, Jalloh I, Helmy A, et al. Consensus state- 12. Scheufler K-M, Lehnert A, Rohrborn H-J, et al. ment from the 2014 international microdialysis forum. Individual value of brain tissue oxygen pressure, micro- Intens Care Med 2015; 41: 1517–1528. vascular oxygen saturation, cytochrome redox level, and 28. Kolyva C, Tachtsidis I, Ghosh A, et al. Systematic inves- energy metabolites in detecting critically reduced cerebral tigation of changes in oxidized cerebral cytochrome c oxi- energy state during acute changes in global cerebral per- dase concentration during frontal lobe activation in fusion. J Neurosurg Anesth 2004; 16: 210–219. healthy adults. Biomed Opt Exp 2012; 3: 2550–2566. 13. Ainslie PN, Shaw AD, Smith KJ, et al. Stability of cere- 29. Matcher SJ, Elwell CE, Cooper CE, et al. Performance bral metabolism and substrate availability in humans comparison of several published tissue near-infrared during hypoxia and hyperoxia. Clin Sci 2014; 126: spectroscopy algorithms. Analyt Biochem 1995; 227: 661–670. 54–68. 14. Coles JP, Fryer TD, Smielewski P, et al. Defining ische- 30. Duncan A, Meek JH, Clemence M, et al. Optical path- mic burden after traumatic brain injury using 15O PET length measurements on adult head, calf and forearm imaging of cerebral physiology. J Cereb Blood Flow and the head of the newborn infant using phase resolved Metab 2004; 24: 191–201. optical spectroscopy. Phys Med Biol 1995; 40: 295–304. Ghosh et al. 11

31. Suzuki S, Takasaki S, Ozaki T, et al. Tissue oxygenation human brain during hypoxemia. J Biomed Opt 2007; 12: monitor using NIR spatially resolved spectroscopy. Proc 024002. SPIE 1999; 3597: 582–592. 44. Bainbridge A, Tachtsidis I, Faulkner SD, et al. Brain 32. Roche-Labarbe NEG, Fenoglio A, Aggarwal A, et al. mitochondrial oxidative metabolism during and after Near-infrared spectroscopy assessment of cerebral cerebral hypoxia-ischemia studied by simultaneous phos- oxygen metabolism in the developing premature brain? phorus magnetic-resonance and broadband near-infrared J Cerebr Blood F Met 2012; 32: 481–488. spectroscopy. Neuroimage 2014; 102(Pt 1): 173–183. 33. Guo Y, Logan HL, Glueck DH, et al. Selecting a sample 45. Kietzmann T, Knabe W and Schmidt-Kastner R. size for studies with repeated measures. BMC Med Res Hypoxia and hypoxia-inducible factor modulated gene Methodol 2013; 13: 100. expression in brain: involvement in neuroprotection and 34. R Core Team. R: A language and environment for statis- cell death. Eur Arch Psychiatr Clin Neurosci 2001; 251: tical computing. 170–178. 35. Laird NM and Ware JH. Random-effects models for lon- 46. Magnoni S, Ghisoni L, Locatelli M, et al. Lack of gitudinal data. Biometrics 1982; 38: 963–974. improvement in cerebral metabolism after hyperoxia in 36. Longhi L, Pagan F, Valeriani V, et al. Monitoring brain severe head injury: a microdialysis study. J Neurosurg tissue oxygen tension in brain-injured patients reveals 2003; 98: 952–958. hypoxic episodes in normal-appearing and in peri-focal 47. Kilgannon JH, Kilgannon JH, Jones AE, et al. tissue. Intens Care Med 2007; 33: 2136–2142. Relationship between supranormal oxygen tension and 37. Johnston AJ, Steiner LA, Coles JP, et al. Effect of cere- outcome after resuscitation from cardiac arrest. bral perfusion pressure augmentation on regional oxy- Circulation 2011; 123: 2717–2722. genation and metabolism after head injury*. Crit Care 48. Quintard H, Patet C, Suys T, et al. Normobaric hyper- Med 2005; 33: 189–195. oxia is associated with increased cerebral excitotoxicity 38. van den Brink WA, van Santbrink H, Steyerberg EW, after severe traumatic brain injury. Neurocit Care 2014; et al. Brain oxygen tension in severe head injury. 22: 243–250. Neurosurgery 2000; 46: 868–878. 49. Altemeier WA and Sinclair SE. Hyperoxia in the inten- 39. Ghosh A, Elwell C and Smith M. Review article: cerebral sive care unit: why more is not always better. Curr Opin Crit Care 2007; 13: 73–78. near-infrared spectroscopy in adults: a work in progress. 50. Guzy RD and Schumacker PT. Oxygen sensing by mito- Anesth Analg 2012; 115: 1373–1383. chondria at complex III: the paradox of increased react- 40. Cooper CE and Springett R. Measurement of cyto- ive oxygen species during hypoxia. Exp Physiol 2006; 91: chrome oxidase and mitochondrial energetics by near- 807–819. infrared spectroscopy. Philos Trans R Soc Lond B Biol 51. Cooper CE, Cope M, Springett R, et al. Use of mitochon- Sci 1997; 352: 669–676. drial inhibitors to demonstrate that cytochrome oxidase 41. Heekeren HR, Kohl M, Obrig H, et al. Noninvasive near-infrared spectroscopy can measure mitochondrial assessment of changes in cytochrome-c oxidase oxidation dysfunction noninvasively in the brain. J Cereb Blood in human subjects during visual stimulation. J Cereb Flow Metab 1999; 19: 27–38. Blood Flow Metab 1999; 19: 592–603. 52. Gjedde A, Gjedde A, Marrett S, et al. Glycolysis in neu- 42. Tachtsidis I, Tisdall MM, Leung TS, et al. Relationship rons, not astrocytes, delays oxidative metabolism of between brain tissue haemodynamics, oxygenation and human visual cortex during sustained checkerboard metabolism in the healthy human adult brain during stimulation in vivo. J Cereb Blood Flow Metab 2001; hyperoxia and hypercapnea. Adv Exp Med Biol 2009; 21: 1384–1392. 645: 315–320. 53. Caldwell M, Hapuarachchi T, Highton D, et al. 43. Tisdall MM, Tachtsidis I, Leung TS, et al. Near-infrared BrainSignals revisited: simplifying a computational spectroscopic quantification of changes in the concentra- model of cerebral physiology. PLoS One 2015; 10: tion of oxidized cytochrome c oxidase in the healthy e0126695.