Decreased arterial $P{}_{{\mathrm{O}}_2}$, not O₂ content, increases blood flow through intrapulmonary arteriovenous anastomoses at rest

Abstract

Key points

• The mechanism(s) that regulate hypoxia‐induced blood flow through intrapulmonary arteriovenous anastomoses (Q _IPAVA) are currently unknown.

• Our previous work has demonstrated that the mechanism of hypoxia‐induced Q _IPAVA is not simply increased cardiac output, pulmonary artery systolic pressure or sympathetic nervous system activity and, instead, it may be a result of hypoxaemia directly.

• To determine whether it is reduced arterial $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{a}{\mathrm{O}}_2}$) or O₂ content ($C_{\mathrm{a}{\mathrm{O}}_2}$) that causes hypoxia‐induced Q _IPAVA, individuals were instructed to breathe room air and three levels of hypoxic gas at rest before (control) and after $C_{\mathrm{a}{\mathrm{O}}_2}$ was reduced by 10% by lowering the haemoglobin concentration (isovolaemic haemodilution; Low [Hb]).

• Q _IPAVA, assessed by transthoracic saline contrast echocardiography, significantly increased as $P_{\mathrm{a}{\mathrm{O}}_2}$ decreased and, despite reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ (via isovolaemic haemodilution), was similar at iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$.

• These data suggest that, with alveolar hypoxia, low $P_{\mathrm{a}{\mathrm{O}}_2}$ causes the hypoxia‐induced increase in Q _IPAVA, although where and how this is detected remains unknown.

Abstract

Alveolar hypoxia causes increased blood flow through intrapulmonary arteriovenous anastomoses (Q _IPAVA) in healthy humans at rest. However, it is unknown whether the stimulus regulating hypoxia‐induced Q _IPAVA is decreased arterial $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{a}{\mathrm{O}}_2}$) or O₂ content ($C_{\mathrm{a}{\mathrm{O}}_2}$). $C_{\mathrm{a}{\mathrm{O}}_2}$ is known to regulate blood flow in the systemic circulation and it is suggested that IPAVA may be regulated similar to the systemic vasculature. Thus, we hypothesized that reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ would be the stimulus for hypoxia‐induced Q _IPAVA. Blood volume (BV) was measured using the optimized carbon monoxide rebreathing method in 10 individuals. Less than 5 days later, subjects breathed room air, as well as 18%, 14% and 12.5% O₂, for 30 min each, in a randomized order, before (CON) and after isovolaemic haemodilution (10% of BV withdrawn and replaced with an equal volume of 5% human serum albumin–saline mixture) to reduce [Hb] (Low [Hb]). $P_{\mathrm{a}{\mathrm{O}}_2}$ was measured at the end of each condition and Q _IPAVA was assessed using transthoracic saline contrast echocardiography. [Hb] was reduced from 14.2 ± 0.8 to 12.8 ± 0.7 g dl⁻¹ (10 ± 2% reduction) from CON to Low [Hb] conditions. $P_{\mathrm{a}{\mathrm{O}}_2}$ was no different between CON and Low [Hb], although $C_{\mathrm{a}{\mathrm{O}}_2}$ was 10.4%, 9.2% and 9.8% lower at 18%, 14% and 12.5% O₂, respectively. Q _IPAVA significantly increased as $P_{\mathrm{a}{\mathrm{O}}_2}$ decreased and, despite reduced $C_{\mathrm{a}{\mathrm{O}}_2}$, was similar at iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$. These data suggest that, with alveolar hypoxia, low $P_{\mathrm{a}{\mathrm{O}}_2}$ causes the hypoxia‐induced increase in Q _IPAVA. Whether the low $P_{{\mathrm{O}}_2}$ is detected at the carotid body, airway and/or the vasculature remains unknown.

Key points

• The mechanism(s) that regulate hypoxia‐induced blood flow through intrapulmonary arteriovenous anastomoses (Q _IPAVA) are currently unknown.

• Our previous work has demonstrated that the mechanism of hypoxia‐induced Q _IPAVA is not simply increased cardiac output, pulmonary artery systolic pressure or sympathetic nervous system activity and, instead, it may be a result of hypoxaemia directly.

• To determine whether it is reduced arterial $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{a}{\mathrm{O}}_2}$) or O₂ content ($C_{\mathrm{a}{\mathrm{O}}_2}$) that causes hypoxia‐induced Q _IPAVA, individuals were instructed to breathe room air and three levels of hypoxic gas at rest before (control) and after $C_{\mathrm{a}{\mathrm{O}}_2}$ was reduced by 10% by lowering the haemoglobin concentration (isovolaemic haemodilution; Low [Hb]).

• Q _IPAVA, assessed by transthoracic saline contrast echocardiography, significantly increased as $P_{\mathrm{a}{\mathrm{O}}_2}$ decreased and, despite reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ (via isovolaemic haemodilution), was similar at iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$.

• These data suggest that, with alveolar hypoxia, low $P_{\mathrm{a}{\mathrm{O}}_2}$ causes the hypoxia‐induced increase in Q _IPAVA, although where and how this is detected remains unknown.

Abbreviations

A–aDO₂

alveolar‐to‐arterial difference in $P_{{\mathrm{O}}_2}$

$C_{\mathrm{a}{\mathrm{O}}_2}$

arterial O₂ content

HbCO%

carboxyhaemoglobin

Hct

haematocrit

HR

heart rate

HPV

hypoxic pulmonary vasoconstriction

HSA

human serum albumin

IPAVA

intrapulmonary arteriovenous anastomoses

LVOT

left ventricular outflow tract

LVOT_VTI

LVOT velocity time integral

PASP

pulmonary artery systolic pressure

$P_{\mathrm{A}{\mathrm{O}}_2}$

alveolar $P_{{\mathrm{O}}_2}$

$P_{\mathrm{a}{\mathrm{O}}_2}$

arterial $P_{{\mathrm{O}}_2}$

$P_{\mathrm{aC}{\mathrm{O}}_2}$

arterial $P_{\mathrm{C}{\mathrm{O}}_2}$

PFO

patent foramen ovale

Q_IPAVA

blood flow through IPAVA

Q_T

cardiac output

$S_{\mathrm{a}{\mathrm{O}}_2}$

arterial O₂ saturation

$S_{\mathrm{p}{\mathrm{O}}_2}$

peripheral estimate of arterial O₂ saturation

TTSCE

transthoracic saline contrast echocardiography

Introduction

Blood flow through intrapulmonary arteriovenous anastomoses (Q _IPAVA) is known to occur in humans and animals under a number of physiological and experimental conditions, including exercise (Eldridge et al. 2004; Stickland et al. 2004; Dujic et al. 2005; Stickland et al. 2007; Lovering et al. 2008 a,b, 2009; Elliott et al. 2011; Bates et al. 2014), i.v. inotropic drug infusion (Bryan et al. 2012; Laurie et al. 2012; Elliott et al. 2014 a) and hypoxia (Lovering et al. 2008 a; Laurie et al. 2010; Bates et al. 2012; Bates et al. 2014; Norris et al. 2014). Despite such findings, as well as many others, the precise mechanism(s) regulating Q _IPAVA under each respective condition remain incompletely known (Lovering et al. 2015).

Transthoracic saline contrast echocardiography (TTSCE) is a non‐invasive method for studying Q _IPAVA because it allows for serial injections in the same individual (Duke et al. 2015) and has demonstrated findings identical to those those obtained using quantitative methods in humans (Whyte et al. 1992; Lovering et al. 2009; Bates et al. 2014) and animals (Stickland et al. 2007; Bates et al. 2012). Using TTSCE in humans, we have recently demonstrated that Q _IPAVA is increased in healthy humans in normoxia during i.v. inotropic drug infusion via an increase in cardiac output (Q _T), with little or no dependence on pulmonary artery systolic pressure (PASP) (Elliott et al. 2014 a). With respect to breathing hypoxic gas at rest, the mechanism(s) regulating Q _IPAVA are unknown, although we have suggested that IPAVA could be passively recruited via an active redistribution of pulmonary blood flow (i.e. hypoxic pulmonary vasoconstriction or another unknown mechanism) (Lovering et al. 2015).

Acutely breathing hypoxic gas at rest elicits a number of physiological changes depending on the severity and duration of the hypoxic stimulus [e.g. increased Q _T, increased PASP, increased sympathetic nervous system activity, hypoxic pulmonary vasoconstriction (HPV), and both reduced arterial $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{a}{\mathrm{O}}_2}$) and reduced arterial O₂ content ($C_{\mathrm{a}{\mathrm{O}}_2}$)]. Hypoxia‐induced increases in Q _IPAVA, as detected with TTSCE in humans, strongly correlate with decreased arterial O₂ saturation measured via pulse oximetry ($S_{\mathrm{p}{\mathrm{O}}_2}$), such that the greatest degree of Q _IPAVA occurs with the lowest $S_{\mathrm{p}{\mathrm{O}}_2}$ and is present in the absence of a significant increase in PASP (Laurie et al. 2010). Additionally, Tremblay et al. (2015), using acetazolamide to block HPV, found that isocapnic hypoxia‐induced Q _IPAVA occurred independently of significant increases in PASP. i.v. infusion of a non‐specific β‐receptor antagonist (propranolol) when subjects were breathing 10% O₂ at rest did not abolish or reduce the hypoxia‐induced increase in Q _IPAVA as detected with TTSCE (Laurie et al. 2012). In summary, this previous work suggests that hypoxia‐induced Q _IPAVA is not a result of increased PASP, increased β‐receptor activation and/or exclusively HPV and, instead, may be a result of hypoxaemia directly because of its association with $S_{\mathrm{p}{\mathrm{O}}_2}$.

We have recently suggested that IPAVA are regulated similar to the systemic vasculature because hypoxia increases and hyperoxia decreases Q _IPAVA, which is the same direction as that by which the systemic vasculature responds to hypoxia/hyperoxia (Lovering et al. 2015). With respect to the systemic circulation, vasodilatation and blood flow distribution in response to hypoxia in skeletal muscle has been demonstrated to be regulated by decreases in $C_{\mathrm{a}{\mathrm{O}}_2}$ rather than decreases in $P_{\mathrm{a}{\mathrm{O}}_2}$ (Roach et al. 1999; Calbet, 2000; Gonzalez‐Alonso et al. 2002; Gonzalez‐Alonso et al. 2006). The signal by which the effect of $C_{\mathrm{a}{\mathrm{O}}_2}$ affects vasodilatation and thus increased skeletal muscle blood flow is probably via ATP being released from erythrocytes (Ellsworth & Sprague, 2012; Gonzalez‐Alonso, 2012). Again, because of the similarities in responses to hypoxia/hyperoxia between the systemic vasculature (i.e. increased skeletal muscle blood flow) and Q _IPAVA, the possibility that hypoxia‐induced Q _IPAVA is also regulated by reduced $C_{\mathrm{a}{\mathrm{O}}_2}$, rather than $P_{\mathrm{a}{\mathrm{O}}_2}$, is an intriguing and provocative hypothesis.

Accordingly, the present study aimed to determine whether it is a reduction in $C_{\mathrm{a}{\mathrm{O}}_2}$ or $P_{\mathrm{a}{\mathrm{O}}_2}$ that regulates hypoxia‐induced Q _IPAVA. Testing the effect of reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ on Q _IPAVA independent of the simultaneous reduction of $P_{\mathrm{a}{\mathrm{O}}_2}$ requires an experimental paradigm to uncouple these related variables. To do so, in the present study, we detected Q _IPAVA with TTSCE when individuals were breathing room air and under various levels of alveolar hypoxia (18%, 14% and 12.5% O₂) where $P_{\mathrm{a}{\mathrm{O}}_2}$ and $C_{\mathrm{a}{\mathrm{O}}_2}$ were simultaneously reduced (CON; Part A) (Fig. 1). We then uncoupled these variables by reducing $C_{\mathrm{a}{\mathrm{O}}_2}$ and keeping $P_{\mathrm{a}{\mathrm{O}}_2}$ constant when individuals were breathing room air and at each level of hypoxia. This was carried out via an isovolaemic haemodilution of 10% (Low [Hb]; Part B) (Fig. 1). This experimental paradigm allowed us to compare hypoxia‐induced Q _IPAVA via TTSCE in conditions of reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ but iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$, as well as iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ but reduced $P_{\mathrm{a}{\mathrm{O}}_2}$ (e.g. CON 14% O₂ vs. Low [Hb] 18% O₂ and CON 12.5% O₂ vs. Low [Hb] 14% O₂). We hypothesized that a reduction in $C_{\mathrm{a}{\mathrm{O}}_2}$, independent of $P_{\mathrm{a}{\mathrm{O}}_2}$, would regulate hypoxia‐induced Q _IPAVA.

Figure 1. Schematic of protocol .

A, protocol used for the control condition with normal [Hb] (CON) (A). B, protocol used for the Low [Hb] condition following isovolaemic haemodilution (B). C, each individual bout of hypoxia. Saline contrast signifies an injection of agitated saline contrast (3 ml of saline + 1 ml of air) to obtain bubble score. PASP signifies an injection of agitated saline contrast (3 ml of saline + ∼0.1 ml of air) to obtain an enhanced tricuspid regurgitation signal to quantify PASP. Arterial sample signifies when an arterial blood sample was obtained.

Methods

Ethical approval

Eighteen college‐aged individuals (12 males and six females) volunteered to participate in the present study after being advised both verbally and in writing about the nature of the experiments. The informed consent form was approved by the University of Oregon Research Compliance Services. All studies were performed in accordance with the Declaration of Helsinki. Of the initial 18 individuals that volunteered to participate, five dropped out before completing all study visits for reasons not related to the study protocol (e.g. time commitment) and three did not qualify because the presence of a patent foramen ovale (PFO). The prevalence of PFO in the present study is much lower than that reported by ourselves and others (Woods & Patel, 2006; Woods et al. 2010; Elliott et al. 2013; Marriott et al. 2013) in that 12 individuals who volunteered for the study were invited to participate because they had previously been identified as not having a PFO.

Visits 1–3

Echocardiographic screening, fasting serum iron and ferritin, and pulmonary function

An initial echocardiographic screening and bubble study was performed on all subjects as reported previously (Lovering & Goodman, 2012; Elliott et al. 2013). Briefly, saline contrast was injected through an i.v. catheter in a peripheral vein with and without a Valsalva manoeuvre to confirm that a PFO was not present (i.e. bubbles appear in the left ventricle ≤3 cardiac cycles) as described previously (Lovering & Goodman, 2012; Elliott et al. 2013).

Because of the known effect of iron‐deficiency on the pulmonary vascular pressure response to hypoxia (Smith et al. 2008; Smith et al. 2009), we needed to ensure that participants were not iron‐deficient. Therefore, approximately 7 days before the main study visit (visit 4; see below), and on a separate day, serum iron and ferritin were assessed in each subject following a 12 h overnight fast. Blood was drawn from an antecubital vein into a sterile tiger‐top vaccutainer and immediately sent to PeaceHealth Laboratories in the Sacred Heart Medical Centre at RiverBend (Springfield, OR, USA) for iron and ferritin determination. All 10 subjects had values within the normal clinical range (Wu, 2006) (Table 1).

Table 1.

Anthropometric, serum iron and ferretin, and blood volume data

Age (years)	28 ± 8
Height (cm)	180 ± 7
Mass (kg)	81 ± 10
Serum iron (μg dl−1)	107.9 ± 44.6	(50–170)
Serum ferritin (ng ml−1)	50.2 ± 14.0	(10–300)
Blood volume (mL)	6455 ± 1166

Anthropometric, iron, ferritin and blood volume data are shown as the mean ± SD. Values in parentheses are normal values according to a clinical reference textbook (Wu, 2006).

During a separate visit, each subject performed complete spirometry including forced and slow vital capacity manoeuvres in accordance with American Thoracic Society/European Respiratory Society standards (Miller et al. 2005; Wanger et al. 2005). Whole body plethysmography (MedGraphics Elite Plethysmograph, St Paul, MN, USA) was performed to determine lung volumes and capacities (Miller et al. 2005; Wanger et al. 2005). The single‐breath, breath‐hold technique was used for the determination of lung diffusion capacity for carbon monoxide (CO) (MacIntyre et al. 2005). All subjects had pulmonary function measurements ≥90% of predicted.

Visit 4

Blood volume determination

Haemoglobin mass and blood volume were measured using the optimized CO rebreathing method (Schmidt & Prommer, 2005; Prommer & Schmidt, 2007). Upon arrival at the laboratory, an i.v. catheter was placed in an antecubital vein and subjects rested when seated upright for 15–20 min. Then, a 5 ml venous blood sample was obtained to measure haemoglobin concentration ([Hb]), baseline carboxyhaemoglobin (HbCO%) and haematocrit (Hct) in sextuplicate. Hct was measured via microcentrifugation (M24 Centrifuge; LW Scientific, Lawrenceville, GA, USA) and [Hb] and HbCO% were measured via CO‐oximetry (OSM3; Radiometer, Copenhagen, Denmark). HbCO% values were corrected for oxygen saturation using the equation of Hutler et al. (2001). End‐tidal CO was measured by having subjects exhale slowly and completely into a portable CO detector (Draeger Pac 7000; Draeger, Lübeck, Germany).

Prior to the test, the custom‐built spirometer (Spico‐CO Respirations‐Applikator; Blood Tec, Bayreuth, Germany) was flushed and prefilled with 3–5 litres of 100% O₂. A bolus of 99.9% pure CO (males: 1.0 ml kg⁻¹ body mass; females: 0.8 ml kg⁻¹ body mass) was administered into the spirometer from a calibrated syringe as the subjects breathed from residual volume to total lung capacity on the rebreathing circuit. After CO administration, subjects held their breath for 10 s and began breathing normally for an additional 1 min 50 s. At the end of the 2 min period, subjects exhaled to residual volume and continued resting while seated. Post‐rebreathing end‐tidal CO was measured 4 min after initial CO inhalation in the same manner as above. Seven minutes after the initial CO inhalation another venous blood sample was obtained from the i.v. catheter to determine the post‐rebreathing HbCO%, again in sextuplicate. The volume of CO remaining in the spirometer was measured using a calibration syringe and the portable CO detector. During the rebreathing procedure, the portable CO detector was used to monitor for potential CO leaks (Ryan et al. 2011). A leak was detected at the mouth in one subject; this test was discarded and we administered the rebreathing test a second time (1 week later) prior to proceeding with the study.

All data were compiled and used to calculate haemoglobin mass in accordance with previously published formulas (Schmidt & Prommer, 2005; Prommer & Schmidt, 2007). Blood volume was calculated from haemoglobin mass, [Hb] and Hct using equations in accordance with previous studies (Burge & Skinner, 1995; Schmidt et al. 2002; Ryan et al. 2014):

$$\begin{array}{ccc}\hfill \mathrm{Red}\mathrm{cell}\mathrm{volume}& =& \mathrm{haemoglobin}\mathrm{mass}\times \mathrm{Hct}\times {\left[\mathrm{Hb}\right]}^{-1}\hfill \\ & & \times {100}^{-1}\hfill \\ \hfill \mathrm{Blood}\mathrm{volume}& =& \mathrm{red}\mathrm{cell}\mathrm{volume}\times 100\times {\left[\mathrm{Hct}\right]}^{-1}\hfill \\ & & \times 0.9\mathrm{litre}{\mathrm{s}}^{-1}\hfill \end{array}$$

Haematocrit was multiplied by 0.96 to account for trapped plasma. The constant of 0.91 is included in the blood volume calculation to account for the ratio of body haematocrit to peripheral haematocrit (Chaplin et al. 1953).

Visit 5

Comparison of the effect of arterial O₂ content and arterial $P_{{\mathrm{O}}_2}$ on Q _IPAVA

Protocol

Subjects reported to the laboratory in the morning (06.00–07.00 h) at least 24 h, but less than 5 days, following the blood volume determination visit (see above) to be re‐familiarized with the study procedures and to begin instrumentation (see below). After instrumentation was complete, subjects were positioned in a reclining chair in the left lateral decubitus position, where they would remain for the duration of the protocol, with the exception of a break between Parts A and B (Fig. 1).

After resting quietly in a chair for ∼15 min when breathing room air through a mouthpiece connected to a non‐rebreathing valve (Series 2700; Hans Rudolph, Shawnee, KS, USA), the first measurements were made (see below) to obtain resting, baseline data. Next, subjects were connected to large bore tubing connected to a 70 litre non‐diffusing gas bag (Hans Rudolph) that was pre‐filled with the prescribed hypoxic gas mixture (18%, 14% and 12.5% O₂) assigned in a random order for Parts A and B. The non‐diffusing gas bag containing hypoxic gas was connected to an airtight bucket filled with warm water to humidify the gas. Subjects breathed each gas for 30 min and then had a break, during which they breathed room air for 20–30 min before breathing the next gas. A description of each data collection method is described in detail below. Metabolic data were collected continuously throughout each hypoxic bout, TTSCE was used to detect Q _IPAVA at time 29 min 15 s, PASP was measured at 29 min 30 s, and a 3 ml arterial blood sample was drawn anaerobically from the radial artery catheter immediately after PASP was obtained (∼29 min 45 s). Q _T was assessed simultaneously with the arterial blood sample (see below).

The first portion of the protocol (Part A) was performed with normal [Hb] (CON) and the second portion of the protocol (Part B) was performed with Low [Hb] following 10% isovolaemic haemodilution. For the Low [Hb] condition, we removed 10% of each subject's blood volume (calculated from the blood volume measured above) and immediately infused an equal volume of a 5% human serum albumin (HSA)‐saline mixture to maintain isovolaemia. The 5% HSA and saline solution was mixed immediately prior to each study from a 25% HSA solution (Buminate; Baxter, Deerfield, IL, USA) and sterile saline (0.9% NaCl). Removal occurred over a period of ∼60 min and infusion of the 5% HSA‐saline solution was conducted over a period of ∼15–20 min. For obvious technical reasons, Part B was always performed after Part A. In Part A, subjects breathed three levels of hypoxic gas (18%, 14% and 12.5% O₂) for 30 min each in a random order. In Part B, subjects breathed the same three levels of hypoxic gas, also in a random order, with the exception of two subjects who did not breathe 12.5% O₂ in the Low [Hb] condition.

To determine whether decreased $P_{\mathrm{a}{\mathrm{O}}_2}$ is the mechanism regulating hypoxia‐induced Q _IPAVA, our experimental protocol created four comparisons at iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$ under CON and Low [Hb] conditions: room air (21%), 18%, 14% and 12.5% O₂. To determine whether decreased $C_{\mathrm{a}{\mathrm{O}}_2}$ is the mechanism regulating hypoxia‐induced Q _IPAVA, we chose these specific hypoxic gas mixtures to create two comparisons at which $C_{\mathrm{a}{\mathrm{O}}_2}$ between CON and Low [Hb] was approximately equivalent depending on the subject's individual hypoxic ventilatory response, but with a different $P_{\mathrm{a}{\mathrm{O}}_2}$. For example, using typical values for a male breathing 14% O₂ in the CON condition, it could be approximated that $P_{\mathrm{a}{\mathrm{O}}_2}$ = 50 Torr, [Hb] = 14 g dl⁻¹ and $S_{\mathrm{a}{\mathrm{O}}_2}$ = 85%, and thus $C_{\mathrm{a}{\mathrm{O}}_2}$ = 16.1 ml O₂ (dl blood)^–1. Following isovolaemic haemodilution in this same individual breathing 18% O₂, they would have $P_{\mathrm{a}{\mathrm{O}}_2}$ = 77 Torr, [Hb] = 12.6 g dl⁻¹ and $S_{\mathrm{a}{\mathrm{O}}_2}$ = 94%, and thus $C_{\mathrm{a}{\mathrm{O}}_2}$ = 16.1 ml O₂ (dl blood)^–1. Thus, we have a comparison at iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$, but a different $P_{\mathrm{a}{\mathrm{O}}_2}$. Our second comparison with comparable $C_{\mathrm{a}{\mathrm{O}}_2}$ but differing $P_{\mathrm{a}{\mathrm{O}}_2}$ was 12.5% O₂ in the CON condition compared to 14% O₂ in the Low [Hb] condition (example math not shown).

Instrumentation

Following local anaesthesia [1% lidocaine, 2% by volume nitro‐glycerine (5 mg ml⁻¹) to minimize vasospasm], a 20 Ga × 1.75 inch radial artery catheter (Arrow International, Reading, PA, USA) was placed under aseptic conditions by a board certified cardiologist (JAH). Patency of the arterial catheter was maintained using a pressurized flush system of normal saline. The volume of saline delivered was minimal as reflected by the minimal change in Hct and [Hb] across CON and across Low [Hb] conditions. An i.v. catheter (18–22 Ga) was placed into an antecubital vein for the injection of agitated saline contrast, removal of whole blood and infusion of the 5% HSA‐saline mixture (see above). Measurement of core body temperature was performed using an oesophageal temperature probe (Mon‐A‐Therm; Medtronic, Dublin, Ireland). The temperature probe was inserted into the oesophagus via nasal intubation following the application of an anaesthetic gel (1 ml of 2% lidocaine) to numb the naris and throat. The depth of the probe was determined as 0.479 × (sitting height in cm) as performed previously (Mekjavic & Rempel, 1990; Davis et al. 2015). In two subjects, the placement of an oesophageal probe was not possible because of a lidocaine allergy (n = 1) and intolerance of the placement of the probe (n = 1). In these subjects, we measured the core temperature using an ingestible temperature pill (CorTemp; HQInc., Palmetto, FL, USA). Unpublished observations from our laboratory demonstrate that the temperatures obtained using an oesophageal probe and the ingestible core temperature pill are in good agreement, particularly at rest (r ² = 0.85).

Measurements

Metabolic, respiratory variables, $S_{\mathrm{p}{\mathrm{O}}_2}$ and heart rate

Breath‐by‐breath metabolic data were collected (CardiO₂; MedGraphics, St Paul, MN, USA) and presented as the mid 5 of 7. Oxygen consumption (${\dot{V}}_{{\mathrm{O}}_2}$), carbon dioxide production (${\dot{V}}_{\mathrm{C}{\mathrm{O}}_2}$), minute ventilation (${\dot{V}}_{\mathrm{E}}$) and tidal volume (${\dot{V}}_{\mathrm{T}}$) were collected continuously throughout each hypoxic bout. Alveolar ventilation (${\dot{V}}_{\mathrm{A}}$) was calculated using the directly measured ${\dot{V}}_{\mathrm{C}{\mathrm{O}}_2}$ and the temperature‐ and tonometry‐corrected $P_{\mathrm{aC}{\mathrm{O}}_2}$ using the equation:

$${\dot{V}}_{\mathrm{A}}=\frac{({\dot{V}}_{\mathrm{C}{\mathrm{O}}_2}\times 863)}{P_{\mathrm{aC}{\mathrm{O}}_2}}$$

Gases with known O₂ and CO₂ concentrations within the physiological range were used to calibrate the gas analyser before each bout. Peripheral estimate of arterial oxygen saturation ($S_{\mathrm{p}{\mathrm{O}}_2}$) and heart rate (HR) were continuously measured using a forehead sensor (Nellcor Oximax N‐600 pulse oximeter; Tyco, Mansfield, MA, USA).

Q _T and PASP

Q _T was estimated using echocardiography and determined at the level of the left ventricular outflow tract (LVOT) as reported previously (Elliott et al. 2014 b). Briefly, at rest in the left lateral decubitus position, LVOT was determined and recorded in each subject. This value was used to estimate the cross‐sectional area of the outflow tract. Pulse‐waved Doppler ultrasound (ie33; Philips, Eindhoven, The Netherlands) was used in all studies to determine LVOT velocity time integral (LVOT_VTI) of blood flow through the outflow tract. LVOT_VTI measures were recorded in triplicate, averaged and multiplied by the cross‐sectional area of the LVOT. The product is equal to stroke volume, which is then multiplied by HR to obtain Q _T. Strong agreement has been reported between transesophageal LVOT and pulmonary artery thermodilution (r > 0.95 and SE of estimate = 0.87–0.97 litres min⁻¹) (Stoddard et al. 1993). Unpublished data collected as part of recent work from our laboratory (Elliott et al. 2014 a) showed a strong agreement between transthoracic LVOT and the open‐circuit acetylene wash‐in method (r = 0.78), with LVOT consistently estimating Q _T to be 1–2 litres min⁻¹ greater than the acetylene wash‐in method.

PASP was determined via ultrasound (ie33; Philips) by measuring the peak velocity (v) of the tricuspid regurgitation jet and estimating right atrial pressure (P _RA) based on the collapsibility of the inferior vena cava, and applying these to the modified Bernoulli equation 4v ² + P _RA (Yock & Popp, 1984; Currie et al. 1985; Himelman et al. 1989; Rudski et al. 2010). Doppler ultrasound estimates of PASP have been shown to be strongly correlated with direct measurements of PASP (r = 0.88–0.97) (Yock & Popp, 1984; Berger et al. 1985; Currie et al. 1985; Allemann et al. 2000) with Allemann et al. (2000) reporting a difference of less than 1 mmHg between estimates and direct measurements.

Q _IPAVA detection

TTSCE was used to detect blood flow through IPAVA as reported previously (Lovering et al. 2008 a; Laurie et al. 2010; Elliott et al. 2011, 2014 a,b; Norris et al. 2014). Briefly, agitated saline contrast was produced by combining 3 ml of saline with 1 ml of room air and agitating for ∼15 s prior to the injection. Each agitated saline contrast injection was visualized in the apical, four‐chamber view and recorded at >30 frames s^–1 for 20 cardiac cycles post‐appearance of saline contrast in the right ventricle. The single frame within the 20 cardiac cycle recording with the greatest density and spatial distribution of contrast was qualitatively assessed using a previously published scoring system (Lovering et al. 2008 b), similar to other studies (Barzilai et al. 1991) (0 = no bubbles; 1 = 1–3 bubbles; 2 = 4–12 bubbles; 3 = >12 bubbles appearing as a bolus; 4 = >12 bubbles heterogeneously filling the left ventricle; and 5 = >12 bubbles homogeneously filling the left ventricle). At present, TTSCE provides only an estimate of Q _IPAVA and direct comparisons between Q _IPAVA detected via TTSCE and quantified using radiolabelled macroaggregates in humans or microspheres in animals are yet to be performed. Nevertheless, hypoxia‐induced and exercise‐induced Q _IPAVA studied using all of these methods in a number of studies all demonstrate almost identical findings as discussed in recent reviews (Duke et al. 2015; Lovering et al. 2015). Additionally, there is strong agreement between bubble score and PFO size measured invasively by intracardiac echocardiography (Fenster et al. 2014).

Arterial blood gas analysis

Arterial blood samples (3 ml) were drawn anaerobically via the radial artery over 10–15 s, immediately after each PASP measurement, into a heparinized syringe and immediately analysed in duplicate (and triplicate if necessary) for $P_{\mathrm{a}{\mathrm{O}}_2}$, arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ ($P_{\mathrm{aC}{\mathrm{O}}_2}$) and pH. The blood‐gas analyser (RAPIDLab 248; Siemens, Munich, Germany) was calibrated daily with tonometered whole human blood by equilibrating three 6 ml samples of human blood, positioned within a heated block (37°C), with gases of known concentrations of O₂ and CO₂ for ∼120 min. The gases were chosen to span the expected range of $P_{\mathrm{a}{\mathrm{O}}_2}$ and $P_{\mathrm{aC}{\mathrm{O}}_2}$ anticipated in subjects breathing room air and hypoxic gas at rest ($P_{{\mathrm{O}}_2}$ = 100–42 Torr; $P_{\mathrm{C}{\mathrm{O}}_2}$ = 49–15 Torr). Each sample was run in triplicate, and the values were used to create a predicted vs. measured slope. A correction factor based off of the inverse slope of this relationship was applied to measured values in the present study. Arterial blood gases were also corrected for body temperature (Kelman & Nunn, 1966; Severinghaus, 1966; Dempsey & Wagner, 1999). Direct measures of arterial O₂ saturation ($S_{\mathrm{a}{\mathrm{O}}_2}$) and [Hb] were measured with co‐oximetry. Haematocrit was analysed using the microcapillary tube centrifugation method.

Calculations

Alveolar $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{A}{\mathrm{O}}_2}$) was calculated using the ideal gas equation as reported previously (Lovering et al. 2013; Duke et al. 2014; Elliott et al. 2014 a). Briefly, we used temperature‐ and tonometry‐corrected $P_{\mathrm{aC}{\mathrm{O}}_2}$, barometric pressure, core body temperature (for correction of water vapour pressure) and respiratory exchange ratio from a 15 s average of breath‐by‐breath metabolic data corresponding to the time and duration of the arterial blood sampling. The alveolar‐to‐arterial difference in $P_{{\mathrm{O}}_2}$ (A–aDO₂) was calculated as $P_{\mathrm{A}{\mathrm{O}}_2}$ – $P_{\mathrm{a}{\mathrm{O}}_2}$. Additionally, we calculated $C_{\mathrm{a}{\mathrm{O}}_2}$ as reported previously (Lovering et al. 2013; Duke et al. 2014; Elliott et al. 2014 a) via the equation:

$$C_{\mathrm{a}{\mathrm{O}}_2}=\left[1.34\times \left[\mathrm{Hb}\right]\times \left(\frac{S_{\mathrm{a}}{{}_{\mathrm{O}}}_{{}_2}}{100}\right)\right]+(0.003\times P_{\mathrm{a}{\mathrm{O}}_2}).$$

using an O₂‐carrying capacity of 1.34 ml O₂ (g haemoglobin)^–1, $S_{\mathrm{a}{\mathrm{O}}_2}$ measured via co‐oximetry that was corrected for the portion of haemoglobin that cannot effectively bind O₂ (i.e. [Hb] = total [Hb] – HbCO – methaemoglobin) and temperature‐ and tonometry‐corrected $P_{\mathrm{a}{\mathrm{O}}_2}$.

Statistical analysis

All statistical analyses were performed using Prism, version 6.0f (GraphPad Software Inc., San Diego, CA, USA). P < 0.05 was considered statistically significant for all tests. We compared [Hb] in the CON condition (before the start of the first hypoxia bout in Part A) and Low [Hb] condition (before the start of the first hypoxia bout in Part B) using a paired samples t test. To compare bubble scores across each level of hypoxia within CON and Low [Hb], we used a Friedman's test (non‐parametric repeated measures ANOVA). To determine whether bubble scores differed within a specific level of hypoxia between CON and Low [Hb], we computed a total of four paired samples non‐parametric t tests (Wilcoxon signed rank test) and adjusted α for multiple comparisons within a family of comparisons using the Bonferroni correction, such that α is divided by the number of comparisons (i.e. 0.05/4 = 0.0125). To determine whether there was a significant difference in bubble scores between CON and Low [Hb] conditions in the two iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons (14% O₂ in CON vs. 18% O₂ in Low [Hb] and 12.5% O₂ in CON vs. 14% O₂ in Low [Hb]), we computed two non‐parametric t tests (Wilcoxon signed rank test). Because these pairwise comparisons were made within the same family, we adjusted α via the Holm–Bonferroni correction to control the Type I error rate. To test for differences in our continuous variables (i.e. all non‐bubble score data) across levels of hypoxia and between CON and Low [Hb], we computed a two‐way repeated measures ANOVA with a Holm–Sidak multiple comparison test when appropriate.

Results

Anthropometrics, fasting iron and ferritin, blood volume, and pulmonary function

Anthropometrics, fasted serum iron and ferritin, and blood volume for all 10 subjects are presented in Table 1. Pulmonary function, lung volumes, and lung diffusion capacity for CO data for all 10 subjects are presented in Table 2. All pulmonary function parameters were within the normal range.

Table 2.

Pulmonary function, lung volumes, and diffusion capacity data

FVC (litres)	5.7 ± 0.9	(101 ± 7)
SVC (litres)	5.9 ± 0.8	(105 ± 8)
FEV1 (litres)	4.6 ± 0.7	(99 ± 8)
FEV1/FVC (%)	81 ± 6	(97 ± 8)
FEF25—75 (litres s–1)	4.4 ± 1.1	(93 ± 20)
FRC pleth (litres)	3.5 ± 0.6	(101 ± 12)
IC (litres)	3.8 ± 0.6	(106 ± 10)
ERV (litres)	2.1 ± 0.4	(102 ± 12)
TLC (litres)	7.3 ± 0.9	(104 ± 5)
RV (litres)	1.4 ± 0.4	(87 ± 19)
DLCO (ml min–1 Torr–1)	41.5 ± 7.0	(124 ± 19)
DLCO/VA (ml min–1 Torr–1 litre–1)	5.7 ± 0.7	(119 ± 15)

Pulmonary function data are shown as the mean ± SD and values in parentheses are the mean ± SD % of predicted. FVC, forced vital capacity; SVC, slow vital capacity; FEV₁, forced expiratory volume in 1 s; FEV₁/FVC, ratio of forced expiratory volume in 1 s to forced vital capacity; FEF_25–75, forced expired flow rate from 25% to 75% of FVC; IC, inspiratory capacity; ERV, expiratory reserve volume; FRC pleth, functional residual capacity determined by whole body plethysmography; TLC, total lung capacity; RV, residual volume; DL_CO, diffusion capacity for carbon monoxide; DL_CO/V_A, diffusion capacity for carbon monoxide per litre alveolar volume.

Arterial $P_{{\mathrm{O}}_2}$ vs. arterial O₂ content

Haemoglobin concentration

In CON, [Hb] was 14.2 ± 0.8 g dl⁻¹ and decreased significantly to 12.8 ± 0.7 g dl⁻¹ in the Low [Hb] condition (P < 0.0001). The magnitude of this decrease was 10 ± 2%, demonstrating that we successfully induced a 10% isovolaemic haemodilution in our subjects.

$P_{\mathrm{a}{\mathrm{O}}_2}$ and $C_{\mathrm{a}{\mathrm{O}}_2}$

$P_{\mathrm{a}{\mathrm{O}}_2}$ and $C_{\mathrm{a}{\mathrm{O}}_2}$ data in CON and Low [Hb] conditions are shown in Fig. 2. For $P_{\mathrm{a}{\mathrm{O}}_2}$, there was no significant interaction effect between the level of hypoxia and condition (P = 0.9), nor was there a significant main effect for condition (CON vs. Low [Hb]; P = 0.24). However, there was a significant main effect for the level of hypoxia (P < 0.001). $P_{\mathrm{a}{\mathrm{O}}_2}$ in hypoxia (all levels) was significantly lower (P < 0.05) than for room air in CON and Low [Hb] conditions. For $C_{\mathrm{a}{\mathrm{O}}_2}$, there was no significant interaction effect between the level of hypoxia and condition (P = 0.29). However, there was a significant main effect for condition (P < 0.0001) and the level of hypoxia (P < 0.0001). $C_{\mathrm{a}{\mathrm{O}}_2}$ was significantly lower (P < 0.05) in the Low [Hb] condition compared to CON breathing room air and every level of hypoxia. In CON and Low [Hb] conditions, $C_{\mathrm{a}{\mathrm{O}}_2}$ was significantly lower when breathing 14% and 12.5% O₂ compared to room air (P < 0.05), although there was no difference between 18% O₂ and room air (P > 0.05). For our two iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons (i.e. 14% CON vs. 18% Low [Hb] and 12.5% CON vs. 14% Low [Hb]), $P_{\mathrm{a}{\mathrm{O}}_2}$ was significantly lower in the CON condition for both (P < 0.001), although there was no difference in $C_{\mathrm{a}{\mathrm{O}}_2}$ between CON and Low [Hb] (P = 0.15 for both).

Figure 2. Arterial O₂ tension and O₂ content .

Arterial $P_{{\mathrm{O}}_2}$ ($P_{\mathrm{a}{\mathrm{O}}_2}$) in (A) and arterial O₂ content ($C_{\mathrm{a}{\mathrm{O}}_2}$) in (B) in the CON (solid bars) and Low [Hb] conditions (hashed bars). There was no difference in $P_{\mathrm{a}{\mathrm{O}}_2}$ between CON and Low [Hb] at any % O₂. ^†Significantly lower $P_{\mathrm{a}{\mathrm{O}}_2}$ in CON compared to Low [Hb] in the iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons (14% O₂ CON vs. 18% O₂ Low [Hb] and 12.5% O₂ CON vs. 14% O₂ Low [Hb]). ^*Significantly lower $C_{\mathrm{a}{\mathrm{O}}_2}$ at all % O₂ between CON and Low [Hb]. There was no difference in $C_{\mathrm{a}{\mathrm{O}}_2}$ in either iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparison. ^‡Significantly lower $C_{\mathrm{a}{\mathrm{O}}_2}$ compared to room air (21% O₂) in CON and Low [Hb].

Bubble scores

Bubble scores in CON and Low [Hb] are shown in Fig. 3. In both CON and Low [Hb], there was a statistically significant effect of the level of hypoxia on bubble scores (P < 0.0001 for both conditions), which parallels our previous findings (Laurie et al. 2010). Under both CON and Low [Hb] conditions, bubble scores when breathing 14% and 12.5% O₂ were significantly greater compared to room air, although bubbles scores when breathing 18% O₂ did not differ from those when breathing room air. There was no difference in bubble scores at any level of hypoxia between CON and Low [Hb] conditions (P = 0.38–0.99). Additionally, at iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ (14% O₂ in CON vs. 18% O₂ in Low [Hb] and 12.5% O₂ in CON vs. 14% O₂ in Low [Hb]) (Fig. 3), bubble scores were significantly greater in the CON condition (P = 0.03 and P = 0.04, respectively). These data demonstrate that there is an effect of $P_{\mathrm{a}{\mathrm{O}}_2}$, but not $C_{\mathrm{a}{\mathrm{O}}_2}$, on Q _IPAVA as detected with TTSCE.

Figure 3. Bubble scores in room air and in each level of hypoxia in CON (solid circles) and Low [Hb] (open squares) .

^*Significantly greater bubble scores in 18%, 14% and 12.5% O₂ compared to room air (21% O₂) in CON and Low [Hb]. There was no difference in bubbles scores between CON and Low [Hb] at any level of hypoxia or room air. Iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons are denoted by the brackets (14% O₂ CON vs. 18% O₂ Low [Hb] and 12.5% O₂ CON vs. 14% O₂ Low [Hb]). ^†Significantly greater bubble score in CON compared to Low [Hb].

Arterial blood gas, cardiopulmonary and metabolic data

There was a significant main effect for the level of hypoxia on A–aDO₂, $P_{\mathrm{A}{\mathrm{O}}_2}$, $P_{\mathrm{aC}{\mathrm{O}}_2}$, pH and $S_{\mathrm{a}{\mathrm{O}}_2}$, although no differences between CON and Low [Hb] on any of these variables, with the exception of $P_{\mathrm{aC}{\mathrm{O}}_2}$, which was significantly lower in the Low [Hb] condition when breathing room air (Table 3). For the iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons, $P_{\mathrm{A}{\mathrm{O}}_2}$ was significantly lower in CON compared to Low [Hb]. Hct was significantly lower in Low [Hb] compared to CON at every level of hypoxia and there were no differences in lactate between any level of hypoxia or between conditions (Table 3). There was a significant main effect for level of hypoxia on Q _T, PASP, HR, respiratory exchange ratio, ${\dot{V}}_{\mathrm{A}}$, ${\dot{V}}_{\mathrm{T}}$ and respiratory rate. There was a significant difference between CON and Low [Hb] on Q _T (all levels of hypoxia), PASP (all but room air), HR (all levels of hypoxia and room air) and ${\dot{V}}_{\mathrm{T}}$ (12.5% O₂ only) (Table 4).

Table 3.

Arterial blood gas data

	CON				Low [Hb]
	% O2
	21%	18%	14%	12.5%	21%	18%	14%	12.5%2
A–aDO2 (Torr)	1 ± 1	3 ± 1	5 ± 2	6 ± 2	1 ± 1	3 ± 2	4 ± 2	6 ± 2†
$P_{\mathrm{A}{\mathrm{O}}_2}$ (Torr)	98 ± 2	81 ± 3	55 ± 4**	46 ± 3**	99 ± 3	81 ± 3	55 ± 4	46 ± 2†
$P_{\mathrm{aC}{\mathrm{O}}_2}$ (Torr)	40 ± 2	39 ± 2	37 ± 2	35 ± 3	38 ± 2*	38 ± 2	36 ± 2	34 ± 2†
pH	7.40 ± 0.01	7.41 ± 0.02	7.42 ± 0.02	7.43 ± 0.03	7.39 ± 0.02	7.40 ± 0.03	7.41 ± 0.02	7.43 ± 0.03†
$S_{\mathrm{a}{\mathrm{O}}_2}$ (%)	98 ± 1	97 ± 1	89 ± 2	81 ± 4	98 ± 1	97 ± 1	89 ± 3	82 ± 3†
Hct (%)	41 ± 2	41 ± 2	41 ± 2	42 ± 2	36 ± 2*	36 ± 2*	36 ± 2*	37 ± 2*
Lactate (mmol l−2)	0.7 ± 0.2	0.6 ± 0.1	0.6 ± 0.1	0.6 ± 0.1	0.8 ± 0.3	0.7 ± 0.2	0.7 ± 0.2	1.3 ± 1.8

Arterial blood gas data are shown as the mean ± SD. pH = the arterial pH. Lactate = arterial lactate concentration. ¹ n – 1. ² n – 2. ^*Significant difference between CON and Low [Hb]. ^†Significant main effect for level of hypoxia. ^**Significant difference between CON and Low [Hb] at iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ (i.e. 14% O₂ in CON vs. 18% O₂ in Low [Hb] and 12.5% O₂ in CON vs. 14% O₂ in Low [Hb]).

Table 4.

Cardiopulmonary and metabolic data

	CON				Low [Hb]
	% O2
	21%	18%	14%	12.5%	21%	18%	14%	12.5%2
Q T (litres min−1)	4.8 ± 0.7	4.6 ± 0.4	5.2 ± 0.9	5.8 ± 0.9	5.6 ± 0.9*	5.5 ± 0.9*	6.0 ± 1.1*	6.5 ± 1.1*, †
PASP (mmHg)	27.0 ± 3.3	28.6 ± 4.1	34.0 ± 7.6	37.2 ± 8.8	26.3 ± 2.3	33.2 ± 6.1*	39.3 ± 7.8*	41.3 ± 9.5*, †
HR (beats min–1)	58 ± 9	56 ± 5	60 ± 7	66 ± 12	62 ± 8*	61 ± 8*	67 ± 9*	71 ± 71*, †
${\dot{V}}_{{\mathrm{O}}_2}$ (litres min−1)	0.37 ± 0.08	0.36 ± 0.10	0.36 ± 0.12	0.37 ± 0.07	0.37 ± 0.10	0.36 ± 0.07	0.39 ± 0.10	0.39 ± 0.06
RER	0.78 ± 0.06	0.83 ± 0.06	0.87 ± 0.18	0.85 ± 0.06	0.77 ± 0.05	0.82 ± 0.08	0.84 ± 0.08	0.83 ± 0.06†
${\dot{V}}_{\mathrm{A}}$ (litres min−1)	6.4 ± 1.4	6.7 ± 1.8	7.3 ± 2.2	8.0 ± 1.1	6.5 ± 2.1	7.0 ± 1.3	7.5 ± 2.0	8.0 ± 0.8†
${\dot{V}}_{\mathrm{T}}$ (litres breath–1)	0.60 ± 0.11	0.62 ± 0.16	0.65 ± 0.21	0.72 ± 0.18	0.65 ± 0.12	0.64 ± 0.14	0.70 ± 0.16	0.86 ± 0.25*, †
RR (breaths min–1)	20 ± 5	20 ± 4	21 ± 6	18 ± 5	20 ± 4	21 ± 5	20 ± 4	17 ± 6†

Cardiopulmonary and metabolic data are show as the mean ± SD. RER, respiratory exchange ratio; RR, respiratory rate. ¹ n – 1. ² n – 2. ^*Significant difference between CON and Low [Hb]. ^†Significant main effect for level of hypoxia.

Discussion

The present study aimed to determine whether hypoxia‐induced Q _IPAVA is associated with a reduction in $C_{\mathrm{a}{\mathrm{O}}_2}$ or a reduction in $P_{\mathrm{a}{\mathrm{O}}_2}$. Previously, we have demonstrated that Q _IPAVA increases at rest with alveolar hypoxia and during exercise with alveolar normoxia or hypoxia, but decreases during exercise with alveolar hyperoxia (Lovering et al. 2008 b; Laurie et al. 2010; Elliott et al. 2011). Therefore, we have suggested that IPAVA may be regulated similarly to the peripheral systemic vasculature and opposite to the conventional pulmonary circulation (Lovering et al. 2015). Skeletal muscle vasodilatation and increased blood flow was demonstrated to be a response mediated by $C_{\mathrm{a}{\mathrm{O}}_2}$ rather than $P_{\mathrm{a}{\mathrm{O}}_2}$ (Roach et al. 1999; Calbet, 2000; Gonzalez‐Alonso et al. 2002; Gonzalez‐Alonso et al. 2006). Thus, we hypothesized that hypoxia‐induced Q _IPAVA would be the result of a reduction in $C_{\mathrm{a}{\mathrm{O}}_2}$ rather than $P_{\mathrm{a}{\mathrm{O}}_2}$. However, our data do not support this hypothesis and, instead, they demonstrate that reduced $P_{\mathrm{a}{\mathrm{O}}_2}$ or $P_{\mathrm{A}{\mathrm{O}}_2}$, by some currently unknown mechanism(s), stimulates the hypoxia‐induced increase in Q _IPAVA.

As discussed above, there are a number of changes that occur when an individual acutely breathes hypoxic gas at rest and all of them, independently or in total, could play a role in hypoxia‐induced Q _IPAVA. For reasons discussed previously (Lovering et al. 2015) and as outlined above, we have ruled out elevated pulmonary blood flow, elevated PASP and increased β‐receptor activation in acute hypoxia as the primary mechanism regulating hypoxia‐induced Q _IPAVA. However, there are a number of other possible mechanisms that relate directly to the data obtained in the present study.

Role of hypoxaemia independently

We hypothesized that a reduction in $C_{\mathrm{a}{\mathrm{O}}_2}$ would be the mechanism regulating hypoxia‐induced Q _IPAVA because of the similarities between IPAVA and the peripheral systemic vasculature. Specifically, skeletal muscle blood flow is regulated via $C_{\mathrm{a}{\mathrm{O}}_2}$ (Roach et al. 1999; Calbet, 2000; Gonzalez‐Alonso et al. 2002; Gonzalez‐Alonso et al. 2006) with the effect probably mediated by ATP being released from erythrocytes (Ellsworth et al. 2009; Ellsworth & Sprague, 2012). Although we did not observe an effect of $C_{\mathrm{a}{\mathrm{O}}_2}$ on hypoxia‐induced Q _IPAVA, an effect of ATP on pulmonary blood flow distribution cannot be entirely ruled out. ATP release from erythrocytes is known to occur with a reduction in O₂ tension in various animals (Ellsworth et al. 2009), as well as mechanical deformation of the erythrocyte, which is known to occur in the pulmonary circulation (Sprague et al. 1996). ATP released from the erythrocyte would not circulate as a mediator, but would act on the pulmonary vascular endothelium and regulate local vessel calibre (Sprague et al. 1996). This is supported by the identification of several ATP receptors located in the pulmonary vasculature (Communi et al. 1995).

Therefore, alveolar hypoxia sensed either in the airways/alveoli (see below) and/or the $P_{{\mathrm{O}}_2}$ of the mixed venous blood entering the pulmonary circulation, in combination with mechanical deformation of the erythrocyte, would cause ATP release and subsequent vasodilatation of IPAVA and/or the region(s) of the lung where IPAVA are located. ATP causes smooth muscle dilatation/relaxation, pulmonary vascular resistance is decreased, and blood flow is directed to areas of the lung that are theoretically well ventilated (Walley, 1996). Our hypothesis was that IPAVA are present in these areas of the pulmonary circulation where blood is allowed to flow as a result of ATP‐mediated reduced pulmonary vascular resistance. This is highly speculative and will require studies to be performed where NO synthesis is blunted or blocked completely (e.g. by infusion of l‐NAME) in the animal as has been reported previously (Sprague et al. 1996).

Effect of pulmonary blood flow

Under normoxic conditions, we have demonstrated that Q _IPAVA is a result of increased Q _T (Laurie et al. 2012; Elliott et al. 2014 a) and so it is conceivable that this is also the case under hypoxic conditions. Our previous work has suggested that, under normoxic conditions, a Q _T ‘threshold’ of ∼10 litres min⁻¹ may need to be exceeded before significant Q _IPAVA can occur (Laurie et al. 2012). Similarly, Lovering et al. (2008 a) did not observe an onset of Q _IPAVA during exercise in normoxia until the workload exceeded ∼40% of VO_2peak, which corresponds to an estimated Q _T of 12–13 litres min⁻¹. In the present study, Q _T never exceeded our proposed necessary threshold (5.8 ± 0.9 and 6.2 ± 1.0 litres min⁻¹ in CON and Low [Hb] when breathing 12.5% O₂) to induce significant Q _IPAVA as detected via TTSCE. This is consistent with our previous work, where Q _T increased from 4.2 litres min⁻¹ in normoxia to 5.0 litres min⁻¹ in hypoxia (12% O₂) in young individuals (Norris et al. 2014). Although hypoxaemia may alter the proposed necessary Q _T threshold, the existing data do not support Q _T as a primary mechanism causing hypoxia‐induced Q _IPAVA primarily because we observed a significantly greater Q _T in Low [Hb] at all levels of hypoxia and room air, but no difference in Q _IPAVA between conditions.

Effect of hypoxic pulmonary vasoconstriction and blood flow redistribution

HPV is a phenomenon occurring during hypobaric or normobaric alveolar hypoxia where pulmonary arteries constrict and redistribute pulmonary blood flow, as first demonstrated by von Euler and Liljestrand (1946). We have concluded previously that hypoxia‐induced Q _IPAVA was not caused by HPV per se because hypoxia‐induced Q _IPAVA occurs in the absence of a significant increase in PASP (Laurie et al. 2010). Increased PASP is the hallmark manifestation of HPV occurring 2–4 h after the onset of hypoxaemia (Talbot et al. 2008). Support for this is provided in the present study with respect to the significant effect of condition (i.e. decreasing [Hb]) on PASP. Although PASP was significantly greater in Low [Hb], there was no effect on Q _IPAVA. Additionally, we have observed significant hypoxia‐induced Q _IPAVA even when HPV was blocked with acetazolamide (Tremblay et al. 2015). However, it is important to note that HPV is known to begin within 5 min of the onset of hypoxaemia (Talbot et al. 2005) and could therefore still contribute to a redistribution of pulmonary blood flow without any vascular changes occurring in IPAVA. Accordingly, hypoxia‐induced Q _IPAVA may occur as a result of the passive recruitment of IPAVA secondary to any active redistribution of pulmonary blood flow that would be expected when breathing hypoxic gas at rest (Lovering et al. 2015).

HPV occurs in the absence of neural connections because it occurs in the isolated lung preparation and following lung transplantation (Robin et al. 1987), and is the result of a ‘stimulus’ $P_{{\mathrm{O}}_2}$ that is a combination of $P_{\mathrm{A}{\mathrm{O}}_2}$ and mixed venous $P_{{\mathrm{O}}_2}$, with $P_{\mathrm{A}{\mathrm{O}}_2}$ having the greatest effect (Marshall et al. 1992). Therefore, it is possible that $P_{\mathrm{A}{\mathrm{O}}_2}$, instead of or in combination with $P_{\mathrm{a}{\mathrm{O}}_2}$, is the mechanism regulating hypoxia‐induced Q _IPAVA. Our data support this potential conclusion because the $P_{\mathrm{A}{\mathrm{O}}_2}$ data, unsurprisingly, mirror that of $P_{\mathrm{a}{\mathrm{O}}_2}$ (Fig. 2A and Table 3). An important aspect of the existing literature to consider in this context is the comparison of isolated lung and intact animal preparations as it pertains to hypoxia‐induced Q _IPAVA. Specifically, hypoxia‐induced Q _IPAVA in isolated rat lungs was <0.01% of Q _T, but was 1.1% of Q _T in the intact rat under the same hypoxic conditions (Bates et al. 2012). If the mechanism were only $P_{\mathrm{A}{\mathrm{O}}_2}$, then one would expect Q _IPAVA to be equivalent between the isolated lung and intact animal preparations. Additionally, Q _IPAVA under normoxic conditions measured in isolated lungs from healthy humans and baboons, ventilated and perfused under physiological conditions equivalent to rest, was <0.1% of Q _T (Lovering et al. 2007). This is less than the 0.4% of Q _T measured using macroaggregates of albumin in healthy humans at rest in normoxia (Lovering et al. 2009). These data demonstrate that there is at least some reliance on arterial and/or mixed venous $P_{{\mathrm{O}}_2}$. Nevertheless, the present study was not designed to separate $P_{\mathrm{A}{\mathrm{O}}_2}$ and $P_{\mathrm{a}{\mathrm{O}}_2}$ to identify which is the regulation mechanism. To address such a question specifically, a study would be needed in an intact animal to independently quantify the effects of alveolar hypoxia and arterial hypoxaemia on Q _IPAVA (i.e. by ventilating an animal with hypoxic gas, but perfusing it with normoxic blood and vice versa).

Based on the discussion above, the role of HPV in hypoxia‐induced Q _IPAVA is appealing but is probably an incomplete explanation for hypoxia‐induced Q _IPAVA. Although HPV is a complex system, it is clear that the change in pulmonary vascular tone is a result of the contraction of pulmonary artery smooth muscle. The contractile response in pulmonary artery smooth muscle cells can occur via non‐L‐type Ca²⁺ channels (Shimoda et al. 2007), L‐type Ca²⁺ channels (Franco‐Obregon & Lopez‐Barneo, 1996) and reduced nitric oxide and/or endothelin receptors (Sylvester et al. 2012). Recently Tremblay et al. (2015) detected Q _IPAVA in isocapnic hypoxia using TTSCE and used oral acetazolamide to blunt the hypoxia‐induced increase in PASP. It was found that the degree of hypoxia‐induced Q _IPAVA did not differ between isocapnic hypoxia with and without acetazolamide, suggesting little or no role for non‐L‐type Ca²⁺ channels in hypoxia‐induced Q _IPAVA. We have recently investigated the independent effect of sildenafil (nitric oxide), nifedipine (L‐type Ca²⁺ channels) and acetazolamide (non‐L‐type Ca²⁺ channels) with respect to the hyperoxia‐induced reduction in Q _IPAVA during exercise (Elliott et al. 2014 b). In combination, data from Elliott et al. (2014 b) and Tremblay et al. (2015) suggest that there is no independent effect of these smooth muscle cell vasoconstrictor pathways with respect to the increase or decrease in Q _IPAVA observed during alveolar hypoxia and hyperoxia.

An important aspect of the existing literature to consider is the comparison of isolated lung and intact animal studies with respect to Q _IPAVA. Q _IPAVA has been measured using hypoxic, isolated rat lungs, as being less than 0.01% of Q _T but as 1.1% of Q _T in the intact rat under the same hypoxic condition (Bates et al. 2012). Furthermore, Q _IPAVA was reported to increase by 4.6% of Q _T from normoxic rest to hypoxic rest (10% O₂) in healthy humans (Bates et al. 2014). Similarly, under normoxic conditions, Q _IPAVA measured in isolated lungs from healthy humans and baboons, ventilated and perfused under physiological conditions equivalent to rest, was < 0.1% of Q _T (Lovering et al. 2007), which is less than the 0.4% of Q _T measured using macroaggregates of albumin in healthy humans at rest in normoxia (Lovering et al. 2009). This is important because HPV is not dependent upon lung neural innervation (Robin et al. 1987), which suggests that HPV does not completely explain hypoxia‐induced Q _IPAVA.

Role of the peripheral chemoreceptors (carotid bodies)

Reduced $P_{\mathrm{a}{\mathrm{O}}_2}$ corresponded with an increase in Q _IPAVA in CON and Low [Hb], with no difference in $P_{\mathrm{a}{\mathrm{O}}_2}$ between the two conditions (Fig. 2); therefore, it is possible that hypoxia‐induced Q _IPAVA is mediated via the peripheral chemoreceptor. Glomus cells (type I) of the carotid body sense a reduction in $P_{\mathrm{a}{\mathrm{O}}_2}$ and cause a subsequent increase in ventilation. In addition to its association with $P_{\mathrm{a}{\mathrm{O}}_2}$, there is also a known interaction between peripheral chemoreceptors and the pulmonary vasculature. Specifically, pulmonary vascular pressure and resistance are decreased with peripheral chemoreceptor stimulation via hypoxia (Daly & Daly, 1959; Levitzky et al. 1977; Naeije et al. 1989; Wilson & Levitzky, 1989; Fitzgerald et al. 1992) and peripheral chemoreceptor denervation exacerbates the increase in pulmonary vascular pressure and resistance observed in acute hypoxia (Levitzky et al. 1977; Chapleau et al. 1988; Naeije et al. 1989; Wilson & Levitzky, 1989). However it is important to note that peripheral chemoreceptor denervation, without maintaining a constant alveolar ventilation to the control condition, results in a lower $P_{\mathrm{a}{\mathrm{O}}_2}$ (Bisgard et al. 1976), which would probably result in a greater stimulus $P_{{\mathrm{O}}_2}$.

Nevertheless, there is some indirect support for hypoxia‐induced Q _IPAVA being a carotid body mediated response in intact healthy humans. It is well known that there is a hyperbolic relationship between peripheral chemoreceptor output (i.e. ventilation) and $P_{\mathrm{A}{\mathrm{O}}_2}$/$P_{\mathrm{a}{\mathrm{O}}_2}$, such that there is an ‘effective hypoxia’ necessary to result in a meaningful chemoreceptor response (Rahn & Otis, 1949). There is also considerable intra‐individual variability in the ventilatory responsiveness to hypoxia where some have a ‘brisk’ response and others have a ‘slow’ response (Weil et al. 1970). We have previously reported similar findings with respect to the degree of hypoxia‐induced Q _IPAVA where some subjects are ‘high shunters’ and others are ‘low shunters’ and some individuals have Q _IPAVA with a mildly hypoxic gas and others do not have Q _IPAVA until more severe levels of hypoxic gas are breathed (Laurie et al. 2010; Norris et al. 2014). Coincidentally, almost all individuals (>99%) have some degree of hypoxia‐induced Q _IPAVA when their measured or estimated $P_{\mathrm{a}{\mathrm{O}}_2}$ is less than 50 Torr, which corresponds to the break point in the relationship between peripheral chemoreceptor output (i.e. ventilation) and $P_{\mathrm{A}{\mathrm{O}}_2}$/$P_{\mathrm{a}{\mathrm{O}}_2}$. The present study was not designed specifically to address the potential role of the peripheral chemoreceptors, although we can calculate a Spearman correlation between the change in ventilation from baseline to 30 min when breathing 12.5% and the bubble score at the same time point. Doing so resulted in a significant correlation (r = 0.76; P = 0.03), which suggests that an increase in chemoreceptor output (i.e. ventilation) is related to hypoxia‐induced Q _IPAVA. This is far from definitive proof, although does provide a rationale to investigate the potential role of the peripheral chemoreceptor in mediating hypoxia‐induced Q _IPAVA.

Limitations

There are several limitations associated with the present study. One limitation is that we were only able to obtain two iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons between bubble scores. Ideally, we would have progressively removed whole blood throughout the study and obtained several iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparisons to enhance the strength of our findings. However, logistically, this was not possible because it was imperative to maintain isovolaemia between both arms of the study. Because we have demonstrated that Q _IPAVA increases via catecholamine‐induced increases in cardiac output (Laurie et al. 2012; Elliott et al. 2014 a), it is conceivable that hypovolemia would have the opposite effect. We could have used CO loading to alter $C_{\mathrm{a}{\mathrm{O}}_2}$ as has been reported previously (Ekblom et al. 1975) but, because CO is a known signalling molecule, its use could have impacted our data interpretation (i.e. Q _IPAVA might be mediated via increased HbCO or decreased $P_{\mathrm{a}{\mathrm{O}}_2}$). Specific to the present study, low dose CO inhalation has been demonstrated to alter the pulmonary vascular response to hypoxia in sheep (Nachar et al. 2001). An additional limitation was that we found significant variability in our bubble scores. Ideally, all subjects would have had significant Q _IPAVA (i.e. bubble score of 4–5) when breathing 14% and 12.5% O₂, which we may have observed had we used a lower level of hypoxia (i.e. 10% O₂) as employed previously (Laurie et al. 2010). Doing so would have provided us with an additional comparison between iso‐$P_{\mathrm{a}{\mathrm{O}}_2}$ but would have not provided us with an additional iso‐$C_{\mathrm{a}{\mathrm{O}}_2}$ comparison because the content when breathing 10% O₂ in the Low [Hb] would not have matched up with a $C_{\mathrm{a}{\mathrm{O}}_2}$ in CON (see example math above).

Summary

The present study aimed to determine whether hypoxia‐induced Q _IPAVA was a result of reduced $C_{\mathrm{a}{\mathrm{O}}_2}$ or $P_{\mathrm{a}{\mathrm{O}}_2}$. Our data demonstrate that hypoxia‐induced Q _IPAVA is mediated by a reduction in $P_{\mathrm{a}{\mathrm{O}}_2}$ and/or $P_{\mathrm{A}{\mathrm{O}}_2}$ that is not dependent on $C_{\mathrm{a}{\mathrm{O}}_2}$ acting through a number of potential mechanisms: (1) HPV in the presence or absence of increased PASP; (2) carotid body efferent influence over pulmonary vascular tone because, without lung innervation (i.e. isolated lung preparation), there is little or no Q _IPAVA; and (3) hypoxaemia stimulating ATP release from erythrocytes causing a reduction in pulmonary vascular resistance. One important point is that the changes in $P_{\mathrm{A}{\mathrm{O}}_2}$ mirrored those of $P_{\mathrm{a}{\mathrm{O}}_2}$ and so it could be either or a combination of both. Studies in isolated lungs compared to those in intact animals suggest that it may not be solely a $P_{\mathrm{A}{\mathrm{O}}_2}$ mechanism (Bates et al. 2012), although the present study is unable to completely rule out a potential role of $P_{\mathrm{A}{\mathrm{O}}_2}$. However, these considerations are speculative at this point and will require additional follow‐up studies for clarification. In conclusion, the data obtained in the present study support our working hypothesis that Q _IPAVA is the consequence of active or passive changes in pulmonary blood flow distribution (Lovering et al. 2015).

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

JWD, BJR, JEE and ATL conceived and designed the experiments. ATL, JTD and JEE were responsible for financial support. JWD, JTD, BJR, JEE, KMB, JAH and ATL collected and assembled data. JWD, JTD, JEE, KMB and ATL were responsible for data analysis and interpretation. JWD, JTD, BJR, JEE, KMB, JAH, WCB and ATL were responsible for writing the manuscript. JWD, JTD, BJR, JEE, KMB, JAH, WCB and ATL approved the final manuscript submitted for publication. All authors agree to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

This research was supported by the Defense Medical Research & Development Program, Department of Defense Grant W81XWH‐10–2‐0114; Eugene & Clarissa Evonuk Memorial Graduate Fellowship in Environmental, Exercise, or Stress Physiology (JTD and JEE).