Hemoglobin and cerebral hypoxic vasodilation in humans: Evidence for nitric oxide-dependent and S-nitrosothiol mediated signal transduction

Cerebral hypoxic vasodilation is poorly understood in humans, which undermines the development of therapeutics to optimize cerebral oxygen delivery. Across four investigations (total n = 195) we investigated the role of nitric oxide (NO) and hemoglobin-based S-nitrosothiol (RSNO) and nitrite ( NO2− ) signaling in the regulation of cerebral hypoxic vasodilation. We conducted hemodilution (n = 10) and NO synthase inhibition experiments (n = 11) as well as hemoglobin oxygen desaturation protocols, wherein we measured cerebral blood flow (CBF), intra-arterial blood pressure, and in subsets of participants trans-cerebral release/uptake of RSNO and NO2− . Higher CBF during hypoxia was associated with greater trans-cerebral RSNO release but not NO2− , while NO synthase inhibition reduced cerebral hypoxic vasodilation. Hemodilution increased the magnitude of cerebral hypoxic vasodilation following acute hemodilution, while in 134 participants tested under normal conditions, hypoxic cerebral vasodilation was inversely correlated to arterial hemoglobin concentration. These studies were replicated in a sample of polycythemic high-altitude native Andeans suffering from excessive erythrocytosis (n = 40), where cerebral hypoxic vasodilation was inversely correlated to hemoglobin concentration, and improved with hemodilution (n = 6). Collectively, our data indicate that cerebral hypoxic vasodilation is partially NO-dependent, associated with trans-cerebral RSNO release, and place hemoglobin-based NO signaling as a central mechanism of cerebral hypoxic vasodilation in humans.

For the Investigation 1 (isovolumic hemodilution) and Investigation 2 (nitric oxide synthase inhibition) the same general experimental measures were utilized for cardiovascular and respiratory monitoring. The protocols also shared the same cerebral blood flow monitoring techniques as well as arterial and jugular venous catheterization and blood sampling protocols. For the hemodilution study, participants attended the laboratory at 0600h (n=5) and 1200h (n=5) having fasted for greater than six hours as well as having abstained from exercise (24 hours), alcohol (24 hours), and caffeine (12 hours). Participants were further instructed to avoid foods high in nitrates for 48 hours prior to experimentation with specific instructions to avoid fruits, salads and cured meats. 2 For the nitric oxide synthase inhibition study, participants attended the laboratory at 0600h having fasted for greater than six hours, abstained from exercise (24 hours), alcohol (24 hours), and caffeine (12 hours). Participants were also instructed to avoid foods high in nitrates for 48 hours prior to experimentation.
Using sterile technique, a 20G arterial catheter (Arrow, Markham, ON) was advanced into the left radial artery under local anesthesia (Lidocaine, 1.0%). This technique was assisted via the use of ultrasound guidance. Subsequently, a 13G central venous catheter (Cook Medical, Bloomington, IN) was advanced into the right internal jugular vein, again under sterile conditions and with the use of local anesthesia and ultrasound guidance. The catheter was then advanced up to 15 cm cephalad. 3 The same technique, performed by the same clinician (DBM) has been previously demonstrated to lead to catheter tip placement in the jugular bulb, which is importantly proximal to the facial vein. 4 Further, correct placement was additionally determined by participants noting a sensation in their ear upon full insertion of the catheter. 3 Fulfilment of these techniques leads to ≤3% contamination by extra-cerebral blood. 3 Finally, an 18G venous catheter (Insyte TM Autoguard TM , Becton Dickinson, USA) was inserted into the median ante-cubital vein. See Supplemental Figure 1.
Cardiovascular and respiratory variables were sampled continuously throughout the protocol at 1 KHz via an analog-to-digital converter (Powerlab, 16/30, ADInstruments, Colorado Springs, CO). Heart rate (HR) was measured by a lead-II configuration electrocardiogram (ADI bioamp ML132). Both the partial pressure of end-tidal CO 2 (P ET CO 2 ) and O 2 (P ET O 2 ) were sampled at the mouth and recorded by a calibrated gas analyzer (ML206, ADInstruments, Colorado, CO), while respiratory flow and minute ventilation (VE) were measured by a pneumotachograph (HR800L, HansRudolph, Shawnee, KS) connected in series with a bacteriological filter. The radial arterial and jugular venous catheters were both attached to an in-line and waste-less sampling system (VAMP system, Edwards Life Sciences). This allows for serial blood sampling and the continuous measurement of radial arterial and jugular venous blood pressure (Truwave Transducer, Edwards Life Sciences). We calculated cerebral perfusion pressure (CPP) as the difference between mean arterial pressure (MAP) and jugular venous blood pressure (JBP), which is reflective of ICP in the pressure range of the current study 5 .

CPP = MAP − JBP
End-tidal forcing was utilized to control the partial pressure of end-tidal oxygen (P ET O 2 ) for the purpose of targeting a specific partial pressure of arterial oxygen (PaO 2 ). 6,7 We predicted the necessary P ET O 2 to target PaO 2 and subsequently achieve a desired arterial oxygen saturation (SaO 2 ) value by accounting for the known end-tidal-to-arterial gradient with our breathing apparatus 7 and the known relationship between PaO 2 and SaO 2 8 (formalism presented below). Pilot testing confirmed the suitability of our P ET O 2 targets for the study. Blood velocity through the right MCA (MCAv) and left PCA (PCAv) was measured with a 2-MHz transcranial Doppler ultrasound (TCD) (Spencer Technologies, Seattle, WA). The TCD probes were attached to a specialized headband (Model M600 Bilateral Head Frame, Spencer Technologies) and secured in place. Insonation was achieved through the trans-temporal window superior to the zygomatic arch and rostral of the pinna by using previously described location and standardization techniques (these data are not reported). 9 Blood velocity and vessel diameter of the ICA and VA were measured with a 10-MHz multi-frequency linear array duplex ultrasound (Terason uSmart 3300, Teratech, Burlington, MA). Arterial diameter was measured with B-mode imaging. Dynamic range (i.e., contrast) and gain (i.e., brightness) were consistent throughout the protocol within each vessel. Pulse-wave mode was used to simultaneously measure peak blood velocity. Insonation angle was constant (60 degrees) while sample volume was adjusted to span the entire intra-luminal diameter. Measures of ICA (ICA FLOW ) and VA (VA FLOW ) blood flow were made ipsilateral to the MCA and PCA, respectively. The ICA diameter and velocity were measured at least 1.5 cm distal to the common carotid bifurcation to eliminate recordings of turbulent and retrograde flow, while VA diameter and velocity were measured between C4-C5, C5-C6, or proximal to entry into the vertebral column. The location was determined on an individual basis in an attempt to select the most reproducible measures, with the same location repeated within participants and between trials.
Ultrasound recordings were screen captured and stored as video files for offline analysis. Concurrent outputs for arterial diameter and peak blood velocity were acquired at 30 Hz with customized edge detection and wall tracking software designed to mitigate observer bias 10 . One minute averages were used to determine ICA FLOW and VA FLOW . Volumetric cerebral blood flow (CBF) was calculated using the following formula: It is important to note that ~3% of the blood accounted for in this measurement of CBF is diverted from the brain via the ophthalmic artery. 11 To account for MAP in our analysis of the CBF responses, cerebrovascular conductance (CVC) was subsequently calculated (e.g., CBF/CPP). We conducted ultrasound scanning in accordance with published technical recommendations 12 and have previously reported a withinand between-day coefficient of variation for extracranial artery scanning of 1.5 and 4.4%, respectively. 13 At each stage of hypoxia ~1.0mL of radial arterial and jugular venous blood were simultaneously drawn into pre-heparinized syringes (SafePICO, Radiometer, Copenhagen, Denmark) and analyzed immediately using a commercial blood gas analyzer (ABL90 FLEX, Radiometer). This analysis included measurement of PaO 2 , PaCO 2 , SaO 2 , CaO 2 , pH, [H + ], [HCO 3 -], [Hb], Hct, and other parameters. Jugular venous blood was also analyzed for whole blood viscosity. Jugular venous blood was drawn into a Lithium Heparin Vacutainer® (Becton Dickinson, USA).
At specific experimental time points, arterial and jugular bulb blood samples were drawn into K 2 EDTA Vacutainers® (Becton Dickinson, USA). Whole blood was then immediately centrifuged at 600 g for 10 minutes at a temperature of 4.0°C. Plasma was aliquoted into cryovials, flash frozen in liquid N 2 , and stored at -80°C. Samples were later batch analyzed to measure plasma NO 2 − using tri-iodide reductive ozone based chemiluminescence, as previously described by our research group. 14-16 Samples were rapidly thawed at 37°C for 3-minutes, at which time 540 uL of plasma was mixed with 60 uL of 5% acidified sulphanilamide (9:1 ratio) and left to incubate in the dark at 21°C for 15-minutes to remove NO 2 − before injection into the tri-iodide reagent for the measurement of s-nitrosothiols (RSNO). Immediately following the mixing of plasma with acidified sulphanilamide (i.e., during the incubation period), two 100 uL plasma samples from the original cryovial were subsequently injected into the tri-iodide reagent for the combined measurement of NO 2 and RSNO. These injections were separated by ≥2min to ensure recordings had reached a new stable baseline before subsequent injections. The tri-iodide reagent was then replaced prior to injection of the 400uL of the sample treated with acidified sulfanilamide. The average of the two 100uL injections was calculated. NO 2 − was calculated by subtracting the concentration of RSNO from the total NO 2 − +RSNO value. Our coefficient of variation for these measures is <10%. All calculations were performed using Origin 8 (OriginLab Corps, Massachusetts, USA) and smoothed using a 150-point averaging algorithm. The Peak Analysis package was used to calculate the area under the curve (mV), which was subsequently converted to a concentration, using standard curves of known concentrations measured using identical injection volumes. All chemicals were of the highest purity available from MilliporeSigma.
We combined our measures of the differing NO species with CBF to determine their net exchange as follows:

Supplemental Methods: Hemodilution Protocol
Following the first isocapnic hypoxemia test, the isovolumic hemodilution commenced. Whole blood was transferred into an Anticoagulant Citrate Phosphate Dextrose Solution BLOOD-PACK TM (4R0012MC, Fenwal, USA) through a non-pyrogenic plasma transfer kit (4C2240, Fenwal, USA) to collect up to 450mL of whole blood per BLOOD-PACK for short term storage. This blood was stored at room temperature on an orbital agitator.
The hemodilution protocol was conducted in two stages aimed at removing/replacing blood in 10% increments (i.e., 10% of whole blood volume). Thus approximately 10% of whole blood was removed and then replaced with an equal volume of 5% human serum albumin (Alburex 5%). Adequate hemodilution (±1% of target) was confirmed with an arterial [Hb] and Hct measurement. Subsequently, end-tidal forcing was utilized to return end-tidal gases to normal resting values in the event that any respiratory changes and end-tidal gas deviations had occurred during hemodilution. Here, a duplex ultrasound measures of CBF were conducted simultaneous to a radial arterial and jugular venous blood sample. The first stage of blood removal took 27±7 minutes, while replacement with albumin took 15±3 minutes. The second stage of blood removal took 26±6 minutes, while blood replacement with albumin took 14±5 minutes. These two stages of hemodilution were designed to match the hypoxemic stimulus at 90% and 75-80% SaO 2 .

Supplemental Methods: Nitric Oxide Synthase Inhibition Protocol
Following instrumentation with a radial artery, jugular venous, and ante-cubital venous catheter (Supplemental Figure 1) participants had ≥20 minutes to rest while the remaining experimental set-up was completed. Thereafter, either a control (volume matched saline, 0.9% NaCl), nitric oxide synthase inhibitor (N G -monomethyl-L-arginine, L-NMMA) or α 1 -adrenoceptor agonist (phenylephrine) infusion was initiated. We utilized a singled-blinded counter-balanced experimental design where trials were separated by 5 h, the equivalent of five half-lives for l-NMMA. 17 The L-NMMA infusion was initiated with a 5 mg/kg bolus delivered over 5 min followed by a 50μg/kg/min maintenance dose throughout the protocol until completion. Further, in a subset of five of the same participants, we also completed a phenylephrine infusion (0.1-0.6μg/kg/min constant infusion; 18 ). Three participants received phenylephrine as their first infusion, and two as their second. The phenylephrine infusion was conducted as an equipotent vasoconstrictor control for the L-NMMA trial 18 During each drug infusion we conducted an isocapnic hypoxemia protocol with a baseline period followed by two hypoxic stages lasting 5-minutes each where SaO 2 was targeted to 85% and 75%. At baseline and each stage of hypoxia, V E , P ET CO 2 , P ET O 2 , HR, MAP, JVBP, MCA, PCA, VA, ICA, and blood gas measures were made.

Supplemental Methods: Baseline hemoglobin concentration and hypoxic cerebral vasodilation protocol
Data were analyzed from a repository made available through a collaboration with Duke University (Durham, NC, USA). The data repository analyzed for this study consisted of nine separate experiments conducted between 2004 and 2016 with a total of 199 study participants. A breakdown of the studies included is presented in Supplemental Table 1. Each study consisted of a graded isocapnic hypoxemia test to a final SaO 2 of 70%. However, as the data are derived from separate protocols, the number of stages was variable and ranged from four to seven stages. End-tidal gas control during the hypoxemia tests was conducted using the RespirAct TM (Thornhill Medical, Toronto) 19 or end-tidal forcing. 6 Subjects underwent the previously described radial arterial and jugular venous cannulation (Supplemental Figure 1). This blood was analyzed as previously described with an ABL90 FLEX for PaO 2 , PaCO 2 , SaO 2 , CaO 2 , pH, [H + ], [HCO 3 -], [Hb], Hct. Trans-cerebral blood draws were acquired serially at each stage of hypoxemia. In these studies there were no ultrasound measurements conducted. Thus, percentage changes in CBF at each stage of hypoxemia were determined based upon the Fick Principle. That is, the cerebral metabolic rate of oxygen (CMRO 2 ) is the product of CBF and the arterio-venous difference of oxygen (AvDO 2 ). Therefore, CBF is the quotient of CMRO 2 and the AvDO 2 . As CMRO 2 is constant in hypoxia (see Figure 3 in the manuscript) 4 , CBF is inversely proportional to AvDO 2 and percent reductions in AvDO 2 with hypoxemia indicate percent increases in CBF (see Figure 3 in the manuscript). The formalism guiding these calculations is provided below: Under the assumption of a constant CMRO 2 As noted, the number of hypoxemic stages was variable; however, given the CBF response in this range of hypoxemia is linear 20 , hypoxic CBF reactivity slopes are comparable across all experiments included. Therefore, hypoxemic cerebrovascular reactivity (i.e. magnitude of hypoxic cerebral vasodilation) was calculated as the slope of the linear regression between percent changes in CBF and absolute reductions in CaO 2 [i.e., %ΔCBF · ΔCaO 2 (mL/dL)]. Prior to further data and statistical analyses, this data repository was scrutinized for data quality control purposes. The following data exclusion criteria were used: 1) Data were excluded if one or more of the primary parameters for analysis (CaO 2 , CjvO 2 , & [Hb]) were missing (n=37); 2) If the participant terminated the isocapnic protocol prior to reaching 70% SaO 2 (n=5); 3) If the relationship between CBF and CaO 2 was not linear (R 2 <0.7 arbitrary cutoff; n=6); 4) If there was a PaCO 2 deviation of ≥3mmHg during any stage, which would largely confound the %change in CBF (n=17).
This resulted in the exclusion of 65 participants and a final sample size of 134 (79 male, 55 female).

Supplemental Methods: Peru Study
High-altitude native Andeans with (n=15; all male; Age: 39±15 years; body mass index: 25±3 kg/m 2 ) and without (n=25; 2 female; Age: 29±12 years; body mass index: 24±3 kg/m 2 ) excessive erythrocytosis (EE+ / EE-) were recruited to participate in the Cerro de Pasco region of Peru (4,300 m above sea-level). Upon initial enrollment a venous blood sample was collected for the measurement of Hct and [Hb]. Duplex ultrasound measures of VA and ICA flow, peripheral oxyhemoglobin saturation (SpO 2 , pulse oximetry), and resting blood pressure were made (automated brachial blood pressure cuff). Male participants were defined as EE+ with a [Hb] ≥21 g/dL while female participants were defined as EE+ with a [Hb] ≥19 g/dL. 21 To estimate CaO 2 in these participants, we used the following equation: Thereafter, 15 participants (EE-, n=9; EE+, n=6) participated in a substudy assessing the CBF response to isocapnic hypoxemia. An isocapnic hypoxemia test was conducted, wherein P ET O 2 was targeted to 100 mmHg to restore SaO 2 to normal (i.e. 98%), and then reduced to a P ET O 2 of 50 and 40 mmHg, with each stage lasting 5-minutes following the attainment of steadystate. At each stage, V E , P ET CO 2 , P ET O 2 , HR, MAP, SpO 2 , VA, and ICA measures were made.
In the six EE+ participants, this isocapnic hypoxemia test was conducted as part of a larger hemodilution protocol. Thus, following the initial isocapnic hypoxemia test, six EE+ participants completed an isovolumic hemodilution protocol (Age: 44±19 years; body mass index: 25±3 kg/m 2 ). Therein, 25% of blood volume was removed and replaced with human serum albumin (5%), The removal and replacement was conducted simultaneously and in 1-stage until completion. Following hemodilution these participants underwent a follow up isocapnic hypoxemia test, wherein P ET O 2 was targeted to 100 mmHg to restore SaO 2 to normal (i.e. 98%), and then reduced to a P ET O 2 of 50 and 40 mmHg, with each stage lasting 5-minutes following the attainment of steady-state. At each stage, V E , P ET CO 2 , P ET O 2 , HR, MAP, VA, and ICA measures were made. Further, these six participants were instrumented with a radial artery catheter so that blood gases could be collected at each baseline and stage of hypoxemia (thus CaO 2 was not estimated, but rather quantified by arterial blood gas analyses in these six participant).

Supplemental Methods: Statistical Analyses
For all statistical tests, significance was set at an alpha level of α=0.05. Data were assessed for normality using Shapiro-Wilks W tests. Normal data are reported as mean±standard deviation. When data significantly differed from a normal distribution, non-parametric analyses were utilized and data are were presented as median [interquartile range].
Investigation 1. Differences in variables at baseline prior to and following hemodilution were compared with two-tailed paired t-tests. Changes in CBF and CDO 2 during the initial hypoxemia trial, hemodilution, and the post-hemodilution hypoxemia trial were analyzed using linear mixed effects modelling. Fixed effects were the individual stages and the experimental trial (hypoxemia pre, hemodilution, hypoxemia post), with subject included as a random effect. When significant effects were detected, post-hoc comparisons were conducted using a Bonferroni correction. To compare the slope of the relationship between gCBF / CDO 2 and CaO 2 across each trial (i.e. cerebral hypoxic vasodilation), we used linear mixed effects modelling. 22 Fixed effects were CaO 2 and the experimental trial (hypoxemia pre, hemodilution, hypoxemia post), with subject included as a random effect. This model was tested against the same model, but with CPP included as a co-variate. Inclusion of CPP significantly improved the model. Thus, CPP was accounted for as a co-variate in our analyses of CBF reactivity. Residuals were examined for parity to normality. Changes in the a-v differences and cerebral net exchange of NO 2 − and RSNOs were analyzed with a 1-way repeated measures analysis of variance (ANOVA). Finally, the relationship between the a-v difference of NO 2 − and RSNOs with gCBF was assessed with repeated measures correlations. 23 Investigation 2. Plasma NO levels during saline and L-NMMA infusion were compared with a twotailed paired samples t-test. The CPP, CBF, CDO 2 , and CVC responses to hypoxemia during saline and L-NMMA infusions were assessed with two-way repeated measures ANOVAs. When significant effects were detected, post-hoc comparisons were conducted using a Bonferroni correction. The slope of the relationship between CBF / CDO 2 / CVC with CaO 2 was determined using linear regression and compared between saline and L-NMMA infusion with two-tailed paired t-tests. Comparisons between the saline and phenylephrine trials followed the same statistical models. Hypoxic reactivity slopes were compared between the saline and L-NMMA infusion trials using two-tailed paired t-tests or Wilcoxin Signed ranks tests where appropriate. Investigation 3. The data repository consisted of 199 participants. It was assessed for missing data, prematurely terminated protocols or technical difficulties, which led to the exclusion of 65 participants (see Supplemental Methods) and a final sample size of 134 (79 male, 55 female). The relationship between resting [Hb] and hypoxic reactivity was assessed using Pearson r correlations. Males and females were also analyzed separately. Further, baseline variables and reactivity were compared between males and females using unpaired t-tests. Investigation 4. Baseline characteristics were compared between EE-and EE+ participants with two-tailed unpaired t-tests. The relationships between CaO 2 and CBF as well as arterial [Hb] and CBF reactivity were assessed using Pearson r correlations. In the isovolumic hemodilution protocol, pre-post hemodilution comparisons were made using two-tailed paired t-tests. Changes in CBF and CDO 2 with hypoxemia prior to and following hemodilution were assessed with twoway repeated measures ANOVAs. When significant effects were detected, post-hoc comparisons were conducted using a bonferroni correction. The slope of the relationship between CBF and CDO 2 with CaO 2 was determined using linear regression and compared between pre-and posthemodilution with two-tailed paired t-tests. The association between changes in CBF and CDO 2 reactivity and baseline arterial [Hb] were assessed with Pearson r correlations.

Supplemental Figure 1. Experimental model for measuring cerebral exchange kinetics.
Panel A and B depict duplex ultrasound images of the internal carotid artery (ICA) and vertebral artery (VA) respectively, the measurements of which were utilized to calculate cerebral blood flow (CBF = 2 x (ICA FLOW + VA FLOW ). Panel C is an X-ray of a participant demonstrating the internal jugular venous (IJV) bulb catheter tip placement, while Panel D depicts insertion of a radial arterial line for the acquisition of arterial blood samples. Collectively, measures of cerebral blood flow and paired sampling of arterial and jugular venous blood are utilized to calculate trans-cerebral exchange kinetics (see calculations in supplemental methods).