Intrapulmonary arteriovenous anastomoses in humans – response to exercise and the environment

Abstract

Intrapulmonary arteriovenous anastomoses (IPAVA) have been known to exist in human lungs for over 60 years. The majority of the work in this area has largely focused on characterizing the conditions in which IPAVA blood flow (_IPAVA) is either increased, e.g. during exercise, acute normobaric hypoxia, and the intravenous infusion of catecholamines, or absent/decreased, e.g. at rest and in all conditions with alveolar hyperoxia ( = 1.0). Additionally, _IPAVA is present in utero and shortly after birth, but is reduced in older (>50 years) adults during exercise and with alveolar hypoxia, suggesting potential developmental origins and an effect of age. The physiological and pathophysiological roles of _IPAVA are only beginning to be understood and therefore these data remain controversial. Although evidence is accumulating in support of important roles in both health and disease, including associations with pulmonary arterial pressure, and adverse neurological sequelae, there is much work that remains to be done to fully understand the physiological and pathophysiological roles of IPAVA. The development of novel approaches to studying these pathways that can overcome the limitations of the currently employed techniques will greatly help to better quantify _IPAVA and identify the consequences of _IPAVA on physiological and pathophysiological processes. Nevertheless, based on currently published data, our proposed working model is that _IPAVA occurs due to passive recruitment under conditions of exercise and supine body posture, but can be further modified by active redistribution of pulmonary blood flow under hypoxic and hyperoxic conditions.

Introduction

Arteriovenous anastomoses provide a vascular conduit for blood flow to bypass a capillary bed. First described over 100 years ago (Sappey, 1879), these anastomotic connections have been demonstrated to exist not only in various systemic vascular beds (Brown, 1937; Prinzmetal et al. 1947, 1948; Simkin et al. 1948; Prichard & Daniel, 1953, 1956), but also in the lungs of animals (Prinzmetal et al. 1948; Rahn et al. 1952; Sirsi & Bucher, 1953; Niden & Aviado, 1956; Niden et al. 1960; McMullan et al. 2004; Lovering et al. 2007; Stickland et al. 2007; Bates et al. 2012) including humans (Tobin & Zariquiey, 1950; Tobin, 1966; Wilkinson & Fagan, 1990; Lovering et al. 2007). Despite the anatomical evidence for the existence of intrapulmonary arteriovenous anastomoses (IPAVA), potential physiological and pathophysiological roles for these vascular conduits are only beginning to be described, and therefore, remain controversial. Current interests focus on the ideas that IPAVA blood flow (_IPAVA) may (1) be involved in regulating pulmonary pressure, (2) reduce pulmonary gas exchange efficiency, and (3) explain some forms of cryptogenic stroke. Supporting data are largely correlational, due in part to the inherent limitations associated with the available methodology used to study _IPAVA, and the present inability to isolate the potential effect of _IPAVA from other possible effects. Here we will present the case for the existence, and importance, of _IPAVA and attempt to reconcile the controversy surrounding the proposed physiological and pathophysiological roles of _IPAVA by presenting anatomic-based data that are consistent with physiological consequences. We also present a working model that helps to uniformly explain the accumulating data in this area of investigation, although we expect this working model to evolve as new data are published and better techniques to study _IPAVA become available.

Techniques, and their limitations, for detecting and quantifying blood flow through IPAVA

The ‘anatomic-based’ techniques that can be used to study _IPAVA, all exploit the role of the lungs as a biological filter. The premise of these techniques is that microspheres, microbubbles, or radiolabelled macroaggregates of albumin (MAA) injected into either the venous circulation or directly into the pulmonary artery are trapped by the pulmonary microcirculation unless they flow through IPAVA or a patent foramen ovale (PFO) that are larger than the objects injected. These intravenously injected objects can subsequently be either collected in the pulmonary venous effluent (microspheres), visualized in the left heart using echocardiography (microbubbles), or imaged using radioactive labels attached to the particles injected (radiolabelled MAA and microspheres). In all cases, the transpulmonary passage of microspheres, microbubbles and MAA is dependent upon the diameter of these objects being larger than pulmonary capillaries (mean diameter of 6.5 μm, maximum diameter of 13 μm; Glazier et al. 1969; Rosenzweig et al. 1970). Accordingly, if bubbles have the ability to squeeze through capillaries, or if MAA breakup into smaller pieces, or if large diameter microspheres can pass though distended capillaries or corner vessels, then these techniques may either overestimate the percentage of _IPAVA or suggest it is present when it is not. Important to this discussion, theoretical estimates suggest that capillaries can distend 2% per 1 Torr increase in pressure (Krenz & Dawson, 2003; Reeves et al. 2005). Applying this to high intensity exercise, a 10 μm capillary may be able to increase in diameter to between 16 μm and 25 μm at mean pulmonary artery pressures between 40–60 Torr. Thus, using microspheres greater than 25 μm and/or when mean pulmonary pressure is less than 60 Torr, will allow for the use of microspheres to detect and quantify _IPAVA assuming no pulmonary capillaries are greater than a 10 μm initial diameter. We refer the reader to previous reviews on this topic for more detailed descriptions of these techniques and their limitations (Lovering et al. 2006, 2009a, 2010; Stickland & Lovering, 2006).

The technique used most often for investigations of _IPAVA is transthoracic saline contrast echocardiography (TTSCE). Although TTSCE does not quantify the volume of _IPAVA, several scoring systems have been developed (Barzilai et al. 1991; Zukotynski et al. 2007; Lovering et al. 2008b; Gazzaniga et al. 2009; van Gent et al. 2009), with the intent to grade or score the degree of left heart contrast observed under different conditions. Although quantitative measures of blood flow cannot be determined from bubble scores, greater degrees of left heart contrast (i.e. increased bubble scores) appear to correspond well with an increase in the volume of blood flow through pulmonary arteriovenous malformations (PAVM) and impairments in pulmonary gas exchange efficiency (Fischer et al. 2010). Bubble grades/scores also correlate with the size of PFO determined using intracardiac echocardiography (Fenster et al. 2014). However, a direct correlation between saline contrast grades/scores and _IPAVA in healthy humans remains to be established.

Because of the limitations with the currently used techniques, novel techniques and/or approaches that overcome this current set of limitations will be required to advance this field significantly. One place to start would be the development of biodegradable microspheres too large (>25 μm) to squeeze through pulmonary capillaries in humans, used in conjunction with novel imaging or detection techniques. Additionally, the development of novel approaches to visualize _IPAVA under various conditions in animal models will allow for mechanistically based approaches to study the direct regulation of blood flow through these pathways.

Casting of the pulmonary vasculature has been done in an attempt to visualize and locate IPAVA within isolated lungs (Tobin & Zariquiey, 1950; Rahn et al. 1952; Tobin, 1966), and these studies have found that IPAVA are ‘located at the apex of and within the lobular divisions of the lung’ (Tobin & Zariquiey, 1950). Assuming a 100 μm diameter and 300 μm length of an IPAVA, there would need to be ∼200 IPAVA in order to accommodate the estimated 2% of the cardiac output _t during high-intensity exercise (cardiac output ∼25 l min^-1), see exercise section below. These calculation estimates assume uniformity of size, but there may be many sizes and the sizes may be distributed differently within the lung, e.g. large diameter IPAVA at the bottom and small diameter IPAVA at the top of the lung. Recent work in human fetal lungs demonstrates the potential to create three-dimensional reconstructions of arteriovenous connections in the lung (Galambos et al. 2013, 2014, 2015), providing a novel approach for studying IPAVA. A significant amount of work remains to establish the precise size range and distribution, of IPAVA as these data are currently lacking in this area.

Regulation of blood flow through IPAVA in healthy humans and in animals

Rest

For detecting _IPAVA in intact healthy humans, TTSCE offers a minimally invasive, sensitive technique with the capacity to perform multiple serial injections. Using TTSCE to detect _IPAVA, we have recently shown that 28% of 174 healthy, young subjects demonstrated _IPAVA at rest, at sea level (Elliott et al. 2013). These data are supported by other work using TTSCE in otherwise healthy humans at rest with a history of migraine headache (n = 104) that shows an identical prevalence of _IPAVA of 28% (Woods et al. 2010). Importantly, both of these data sets included a rigorous evaluation for PFO in their respective subject populations and both report a PFO prevalence of 38%. A very similar prevalence of PFO (35%) has also been demonstrated using TTSCE in a larger sample (n = 1162) of adults (Marriott et al. 2013).

Our data were collected in human subjects while reclined at 45 deg in the left lateral decubitus position (Elliott et al. 2013). This information on body positioning is important because there is an effect of posture on the detection of _IPAVA. First reported by Stickland et al. (2004), two out of eight subjects demonstrated _IPAVA at rest in the supine position, but were clear of left-sided contrast after standing upright, suggesting that the distribution of pulmonary blood flow may be important for the detection of _IPAVA (Stickland et al. 2004). Similarly, we have shown that in 18 subjects with _IPAVA at rest when supine, 17 were clear of left-sided contrast after standing upright (Elliott et al. 2013). Note though that, although standing upright, subjects leaned forward slightly to facilitate obtaining clear apical, four-chamber echocardiograms. Why body position alters the perfusion of IPAVA in some, but not all, individuals is unknown but it may be due to changes in regional lung perfusion that accompany changes in body posture. Note also that the use of the vasodilatory drugs nitroglycerine and aminophylline in humans at rest does not induce _IPAVA (Lozo et al. 2014). Additionally, the use of noradrenaline (norepinephrine) to acutely increase pulmonary artery systolic pressure (PASP) in humans at rest does not induce _IPAVA (Lozo et al. 2014) and the use of lower body positive pressure to acutely elevate PASP does not consistently result in _IPAVA (Stickland et al. 2006). Accordingly, the consistent findings with these studies is that these vessels appear to not be perfused in the majority of subjects tested under resting conditions despite acute increases in PASP and/or in the presence of well known pulmonary vascular dilators, suggesting that something else is responsible for mediating _IPAVA in humans at rest breathing air (Fig.1A). Theoretically it remains possible that saline contrast bubbles used to detect _IPAVA may be squeezing through pulmonary capillaries, and this has yet to be conclusively proven or disproven. A comprehensive review of the bubble physics literature is beyond the scope of this review, and as such we refer the reader to prior work in this area (Yang, 1971; Yang et al. 1971a,b; Butler & Hills, 1979; Meltzer et al. 1980, 1981; Roelandt, 1982; Meerbaum et al. 1993). Despite these apparent consistencies across numerous studies, the saline contrast bubbles used to detect _IPAVA may be getting through pulmonary capillaries in some individuals but not others, for reasons that we do not yet understand.

Figure 1. IPAVA working model.

Diagram summarizing the working model for how active and passive regulation of the pulmonary circulation could determine the recruitment and perfusion of intrapulmonary arteriovenous anastomoses (IPAVA) under various conditions. IPAVA are considered to be few in number and >25 μm in diameter (top vessel branch) while capillaries are considered to be many in number and <10 μm in diameter (all others); each panel is within an isogravitational plane so we are not suggesting top or bottom blood flow preference is dependent upon gravity. Small/thin vessels with pink background void of erythrocytes and microspheres represent no pulmonary perfusion but the potential for recruitment is possible, whereas areas with erythrocytes and/or microspheres represent areas that are being perfused. Yellow spheres represent intravenously injected microspheres. A, at rest breathing air. _T is ∼5 l min^-1 and there is only passive recruitment of a small portion of the available pulmonary vasculature with little, or no _IPAVA, and consequently there is no appearance of microspheres in the pulmonary venous effluent and/or left ventricle. Large black arrow represents direction of blood flow. B, increased _T during exercise or intravenous inotropic drug infusion breathing air. _T is at least double resting values (≥10 l min⁻¹), which causes passive recruitment of a larger proportion of the pulmonary vasculature that contains IPAVA, such that IPAVA are passively recruited/perfused. Subsequently microspheres appear in the pulmonary venous effluent and/or left ventricle and the volume of _IPAVA increases as the intensity of exercise increases. Large black arrow represents direction of blood flow. C, at rest breathing hypoxic gas. _T is slightly increased to ∼7 l min⁻¹, depending upon the severity of hypoxia. Hypoxia causes an active redistribution of pulmonary blood flow via varying degrees of hypoxic pulmonary vasoconstriction (or other unknown active redistribution mechanism) towards those areas of the lung with IPAVA, and consequently, IPAVA are passively recruited/perfused. Subsequently microspheres appear in the pulmonary venous effluent and/or left ventricle and the volume of _IPAVA increases as the severity of hypoxia increases (i.e. decreases). Large black arrow represents direction of blood flow. D, increased _T during exercise or intravenous catecholamine infusion breathing 100% O₂. _T is at least double resting values (≥10 l min^-1), but with alveolar hyperoxia ( = 1.0). Hyperoxia causes an active redistribution of pulmonary blood flow, either hyperoxic vasoconstriction of areas containing IPAVA or hyperoxic vasodilatation of areas without IPAVA (or other unknown active redistribution mechanism). Consequently, IPAVA are not recruited/perfused and no microspheres appear in the pulmonary venous effluent and/or left ventricle. Large black arrow represents direction of blood flow.

Exercise

In contrast to resting conditions, >95% of healthy humans demonstrate _IPAVA during exercise (Fig.1B). In 2004, two seminal studies using TTSCE to detect _IPAVA were published demonstrating in healthy humans without a PFO and/or _IPAVA at rest that blood begins to flow through IPAVA during submaximal exercise and continues to flow through IPAVA up to maximal exercise at sea level (Eldridge et al. 2004; Stickland et al. 2004). Since then, several studies using subjects without a PFO have replicated these findings (Dujic et al. 2005; Lovering et al. 2008a,b; Elliott et al. 2011; Madden et al. 2013, 2014; Thom et al. 2013; Norris et al. 2014). Two of these studies using TTSCE and bubbles scores have also suggested that _IPAVA is increased and occurs at lower relative workloads with exercise while breathing hypoxic gas, when compared to exercise breathing air (Lovering et al. 2008a; Elliott et al. 2011). Additionally, there do not appear to be any differences between men and women in the occurrence of _IPAVA during exercise (Kennedy et al. 2012). A consistent finding with these studies is that during exercise, most humans (>95%) appear to have increased _IPAVA as detected by TTSCE (Fig.1B), but, as mentioned above with the resting studies, this may be due to bubbles getting through distended capillaries.

The main shortcoming of these studies, and TTSCE in general, is that the volume of _IPAVA during exercise is not known nor does it provide us with definitive information about the size and location of IPAVA within the lung. However, intravenously injection of 25 and 50 μm microspheres in exercising dogs confirmed that _IPAVA does occur during exercise and was calculated to be 1.4% (range 0.2–3.1%) of _T; (Stickland et al. 2007). Technetium-99m-labelled MAA (^99mTc-MAA) is a technique that can be used in humans to quantify _IPAVA because the proportion of radioactivity outside of the lungs represents the proportion of MAA that were not trapped in the pulmonary capillaries, i.e. they travelled through IPAVA and became trapped in systemic capillaries. Studies in healthy male humans using ^99mTc-MAA and gamma camera imaging measured _IPAVA to be 1.3% (range −0.3 to 2.7%) of _t during maximal treadmill exercise (Lovering et al. 2009b). Mean _IPAVA in seven subjects (3 female) has been measured to be 2.5% (±2.6%; range ∼ −0.5 to 7%) with _IPAVA increasing in 5/7 subjects during cycle ergometer exercise at 85% of maximal capacity (Bates et al. 2014). These studies used MAA with a mean diameter of 20–40 μm and ∼90% of MAA between 10 and 90 μm in size. One study prior to these used ^99mTc albumin microspheres with a size range of 7–25 μm in diameter and these authors found an increase in shunt of 2.4% of _T in five normal subjects during exercise, with a range of −2.4 to 5.9% (Whyte et al. 1992). A key point of summary for these studies is that the onset of, and the amount of, _IPAVA is variable between subjects but it appears to consistently increase in healthy subjects and happens in almost everyone during exercise (Fig.1B). Furthermore, and perhaps most importantly, _IPAVA during exercise has been consistently detected using three different anatomically based techniques (microspheres, TTSCE, and ^99mTc-MAA) that all report identical findings. Nevertheless, because of the limitations of these techniques, MAA may be breaking up and travelling through pulmonary capillaries as smaller pieces in the human studies and therefore may be overestimating _IPAVA.

An area of inconsistency in this field is in the exercising thoroughbred horse. In one study using microspheres, the authors found that microspheres did not travel through the lungs of thoroughbred horses during exercise (Manohar & Goetz, 2005). However, we have previously detailed the technical limitations of this study where injecting too few microspheres and/or sampling low volumes of blood may result in conditions where microsphere detection may not be possible even if IPAVA are present and being perfused (Stickland & Lovering, 2006). Nonetheless, recent work has found that bubble contrast is able to traverse the pulmonary circulation in exercising horses (La Gerche et al. 2013). Whether or not this resulted from bubble contrast travelling through IPAVA or from distended pulmonary capillaries is unknown, but future studies using the appropriate numbers of large diameter microspheres and blood sampling volume will help to answer this interesting question.

The underlying factors and mechanisms that regulate _IPAVA during exercise at sea level are currently unknown. Key suggested candidates include increased _T, pulmonary artery pressures, or left atrial pressure (Stickland et al. 2004; Bryan et al. 2012; Laurie et al. 2012; Elliott et al. 2014a). Work using intravenous catecholamine infusion to increase _T and PASP (Bryan et al. 2012; Laurie et al. 2012) demonstrates that, regardless of the inotropic drug used, a doubling of _T results in a significant increase in _IPAVA as detected with TTSCE (Fig.1B). Furthermore, recent work by our group suggests that increased _T, independent of PASP, can increase _IPAVA in human subjects breathing room air (Elliott et al. 2014a). Accordingly, increases in _T that accompany exercise or are induced by infusion of inotropic drugs, may allow for blood to perfuse IPAVA through simple vascular recruitment of under-perfused areas of the lung (Fig.1B). Thus, the consistent finding with these investigations using TTSCE is that increased _T is likely important, but not increased PASP, which is consistent with the resting data obtained using TTSCE (Fig.1A and B). Despite these consistencies, until mechanistically based studies using large diameter microspheres that are not able to pass through distended pulmonary capillaries are utilized, this area of research will remain inconclusive. Clearly, performing these mechanistic studies will significantly advance this area of research.

Hypoxia

Using TTSCE and bubble scoring, _IPAVA is suggested to increase when healthy humans breathe hypoxic gas at sea level (normobaric hypoxia) at rest (Lovering et al. 2008a; Laurie et al. 2010; Norris et al. 2014; Tremblay et al. 2014; Fig.1C). Additionally, left-sided bubble contrast increases with greater decreases in arterial O₂ saturation (Laurie et al. 2010). To date, the majority of work in humans has used TTSCE to detect _IPAVA; however, recent work has demonstrated that _IPAVA, measured using ^99mTc-MAA, is ∼5% of _T in healthy humans breathing 10% O₂ (Bates et al. 2014). Additionally, studies in dogs (Niden & Aviado, 1956) using microspheres 60–420 μm in diameter, and in rats (Bates et al. 2012) using microspheres 25, 50 and 70 μm in diameter, demonstrate that normobaric hypoxia also increases the proportion of large diameter microspheres that pass through IPAVA, identical to the work using TTSCE and ^99mTc-MAA in humans. The consistent finding in this area is that hypoxia increases _IPAVA, but because of the techniques utilized for these investigations additional work using large diameter microspheres is needed to verify if this is the case in humans and not just in dogs and rats.

Interestingly, the work by Bates and colleagues demonstrated that hypoxia-induced _IPAVA was less or absent in isolated rat lungs, but was significantly greater in the intact animal, suggesting there is some inconsistency in this area. Thus, there may be either some blood-borne factor or innervation to the lung that is responsible for increasing _IPAVA in hypoxic conditions. Considering the clear difference in _IPAVA between these two experimental conditions, it would appear most beneficial to focus time and resources on studying _IPAVA in the intact animal. However, directing future work to the intact animal would not preclude studies aimed at isolating IPAVA for histological purposes in isolated lungs. Also, if neurally mediated active control plays a role in regulating _IPAVA then a neurologically intact preparation may be required to continue to advance this area.

Another area of inconsistency in this area comes from data obtained in field studies. These studies at high altitude have investigated the occurrence of _IPAVA at rest in Sherpas and healthy sea level inhabitants after acclimatization to 5050 m (Foster et al. 2014). One noteworthy finding of this study was that _IPAVA in hypobaric hypoxia was less than expected for the level of arterial hypoxemia compared to studies in normobaric hypoxia. Specifically, there was a reduction in bubble scores at altitude, as detected by TTSCE, compared to sea level. The authors speculated that several possibilities might explain the lower than expected _IPAVA including pulmonary vascular remodelling and/or reduced _T with acclimatization compared to acute normobaric hypoxia. Additionally, it remains possible that the in vivo microbubble dynamics of saline contrast are altered in hypobaric environments. Specifically, compared to normobaria, microbubble stability may be impaired, accelerating microbubble time to dissolution. For example, during studies at both sea level and at 5050 m all saline contrast injections consisted of 4 ml of saline combined with 1 ml of air. In accordance with the ideal gas law (PV = nRT), for the same volume of air the ∼50% reduction in barometric pressure would correspond to a ∼50% reduction in the absolute moles of gas within the 1 ml volume. Yet in all cases at altitude there was complete right heart opacification following bubble injection, suggesting that a sufficient number of bubbles were initially present in the right heart. Nevertheless, it remains unknown to what extent, or even if, this change in barometric pressure would affect microbubble stability. However, as mentioned above, due to the limitations of using TTSCE with saline contrast bubbles, future investigations should be performed using large diameter microspheres to determine if in fact _IPAVA is reduced at altitude, or if it is simply an artefact of the techniques currently employed.

Work by Laurie et al. was the first to suggest that _IPAVA increased with higher bubble scores as detected by TTSCE as peripheral arterial oxygen saturation () decreased (Laurie et al. 2010). Breathing hypoxic gas at rest elicits a number of physiological responses that may cause the increases in _IPAVA, including decreasing arterial O₂ content (), arterial partial pressure of O₂ (), and mixed venous partial pressure of O₂ (), and increasing sympathetic activity and hypoxic pulmonary vasoconstriction (HPV), but it is currently unknown how the reduction in O₂ functions to regulate _IPAVA. Work using intravenous propranolol in healthy humans breathing hypoxic gas ( = 0.12) at rest suggests that the hypoxaemia-induced increase in _IPAVA is not a β-receptor-mediated response (Laurie et al. 2012). As with sea level exercise, increased _T and/or increased pulmonary artery pressure may be responsible for increasing _IPAVA. However, hypoxia-induced _IPAVA is probably not due to increased left atrial pressure because it does not increase in subjects breathing hypoxic gases at rest (Groves et al. 1987). Hypoxia-induced _IPAVA could be related to HPV and a subsequent increase in pulmonary vascular pressure, specifically an uneven pulmonary vasoconstriction where some vessels dilate (Hultgren et al. 1964, 1971; Fig.1C). Hypoxia-induced _IPAVA has been shown to occur within 30 min of breathing hypoxic gas (Laurie et al. 2010), although acute hypoxia induces only minor increases in PASP whereas more significant increases in PASP occur at approximately 45 min and later (Talbot et al. 2005). Work using acetazolamide to block HPV found that isocapnic hypoxia-induced _IPAVA occurred independently of significant increases in PASP (Tremblay et al. 2014). It has been shown that hypoxia alters pulmonary blood flow distribution after 10 min so it is possible that hypoxia is able to regulate _IPAVA independently of HPV or as a result of minimal increases in PASP caused by the acute phase of HPV. Bates and colleagues have suggested that a reduced may be responsible for increasing _IPAVA, which may be the case with hypoxia (Bates et al. 2012). However, with inotropic drug infusion at rest, would increase so a reduction in would not explain the increase in _IPAVA that occurs with inotropic drugs. Future work using drugs such as acetazolamide to block HPV, and separating the effect of from , may help to elucidate the mechanisms responsible for hypoxia-induced _IPAVA. Developing an experimental paradigm where _IPAVA can be prevented in hypoxic conditions would help to elucidate the specific mechanisms that induce blood flow through IPAVA with alveolar hypoxia. Although the mechanisms responsible for the regulation of _IPAVA remain elusive, understanding the regulation of hypoxia-induced _IPAVA could have broad implications for millions of individuals residing at high altitude and individuals who are hypoxaemic secondary to lung disease, such as COPD patients who have been reported to have intrapulmonary shunting as detected with TTSCE (Shaikh et al. 2014). Furthermore, until these investigations are performed using large diameter microspheres in humans and animals or better techniques become available, this area will remain inconclusive.

Hyperoxia

Of particular interest in this field are the studies demonstrating that _IPAVA is either prevented or reduced in healthy humans breathing 100% O₂ with lower bubbles scores or no bubbles in the left heart as detected by TTSCE (Lovering et al. 2008b; Fig.1D). The effect of breathing 100% O₂ on regulating _IPAVA has been demonstrated during exercise (Elliott et al. 2011, 2014b; Ljubkovic et al. 2012), post exercise (Madden et al. 2013) and at rest with inotropic drug infusion when breathing 100% O₂ (Bryan et al. 2012; Laurie et al. 2012), but not when breathing 40% O₂ (Elliott et al. 2014a). One criticism of these findings is that the reduced appearance of left-sided contrast (i.e. lower bubble scores) is due to changing the internal and/or external partial pressure environment of the intravenously injected bubbles. Elliott et al. (2011) addressed this criticism by injecting bubbles of different gaseous compositions (air, O₂, CO₂, N₂, and He) during exercise breathing air, 14% O₂, and 100% O₂ and found identical bubble scores within a given regardless of the injected bubble composition (Elliott et al. 2011). Similar findings with respect to the effect of hyperoxia reducing _IPAVA have been demonstrated using an embolization model of increased _IPAVA and intravenously injected microspheres 60–420 μm in diameter to detect _IPAVA (Niden & Aviado, 1956). A consistent finding of these studies using 100% O₂ is that using either TTSCE or microspheres provided identical results. One area of inconsistency with these data was that using of 0.4 did not prevent _IPAVA as detected by TTSCE, suggesting that the effect of hyperoxia may be a dose-dependent phenomenon. However, the bubbles used in these studies with 40% O₂ may be travelling through distended capillaries so more work in this area using large microspheres in humans and animals is needed.

The mechanisms responsible for the hyperoxia-induced reduction in _IPAVA during exercise at sea level remain unknown. According to the Fick principle of mass balance, breathing 100% O₂ during maximal exercise could result in a decrease in _T proportional to the increase in , assuming a constant . The 10% increase in when breathing 100% O₂, could therefore correspond to a reduction in _T of no more than 10% at a given submaximal exercise intensity. If _IPAVA was in part regulated by increases in flow, then this reduction in _T during exercise breathing 100% O₂, could, theoretically, help explain the reduction in _IPAVA during exercise breathing 100% O₂. However, with a 10% reduction in _T from 20 l min⁻¹ to 18 l min⁻¹, _IPAVA during submaximal exercise breathing air when _T < 18 l min⁻¹ is still greater than _IPAVA during exercise while breathing 100% O₂ when _T = 18 l min⁻¹. Thus, it is unlikely that an expected ∼10% reduction in _T is responsible for the observed effect of hyperoxia on _IPAVA. Other possible explanations are that hyperoxia either vasoconstricts IPAVA as previously hypothesized (Lovering et al. 2009a) or pulmonary blood flow is redistributed by hyperoxia-induced vasodilatation such that blood flows through areas of the lung without IPAVA (Fig.1D). Recent work by our group using nifedipine, acetazolamide and sildenafil, independently, to target various pulmonary vascular smooth muscle pathways to prevent or reduce vasoconstriction during exercise breathing 100% O₂, were ineffective in altering bubble score as detected by TTSCE. These data suggest that the mechanisms targeted by these drugs are not independently involved in the effect of 100% O₂ on _IPAVA (Elliott et al. 2014b). Whether or not there are multiple redundant pathways mediating _IPAVA is unknown, but studies using combinations of drugs that target the pulmonary vascular pathways may be helpful in elucidating this question. There is an increase in the perfusion heterogeneity of the lung, measured using radioactive microspheres when animals are ventilated with hyperoxia ( = 0.4–0.5; Melsom et al. 1999; Hlastala et al. 2004), so an active redistribution of pulmonary blood flow caused by hyperoxia is not without precedent. Nevertheless, the effect of hyperoxia on the redistribution of microspheres may be due to vasodilatation of areas that are closed or vasoconstriction of areas that are open. The majority of work in this area has used TTSCE, and considering the limitations to this technique that were mentioned previously, future work in this area should seek to quantify _IPAVA with large diameter microspheres (e.g. >25 μm). Furthermore, future studies should seek to determine the effect of alveolar hyperoxia on pulmonary blood flow distribution and _IPAVA at rest and/or during exercise using the existing methodologies as well as novel approaches. Additionally, as outlined above for hypoxia, developing an experimental paradigm where _IPAVA is possible in hyperoxic conditions would help to elucidate the specific mechanisms that induce or prevent _IPAVA in these conditions.

Effect of age on blood flow through IPAVA

It is well known that the pulmonary vasculature changes with age (Taylor & Johnson, 2010) and this may be true of IPAVA as well. Work in a fetal lamb preparation has demonstrated that _IPAVA occurs in the fetus, and then gradually declines as the newborn lamb ages to approximately 6 months old (McMullan et al. 2004). Of particular interest, children with congenital heart diseases who underwent a ‘classical’ Glen Shunt surgical procedure (unilateral superior vena cava to right pulmonary artery shunt) developed pulmonary arteriovenous anastomoses (Lovering et al. 2013). However, if both the superior and inferior vena cavas are anastomosed to the pulmonary artery, the inclusion of the hepatic blood flow from the inferior vena cava prevents the formation of pulmonary arteriovenous anastomoses in these children, suggesting that there is an unknown hepatic factor capable of regulating pulmonary arteriovenous anastomoses. To our knowledge there have not been any studies in healthy human new-borns examining _IPAVA, but the existence of IPAVA in human fetal lungs has been characterized previously using isolated lungs and large diameter microspheres (Wilkinson & Fagan, 1990). It is intriguing that _IPAVA occurs in the fetus but that it stops with normal postnatal development. It could be hypothesized that IPAVA are part of the normal fetal circulation, which also includes the ductus arteriosus, because they are regulated similarly, with hyperoxia preventing blood flow and hypoxia allowing for blood flow (Coceani & Olley, 1988). However, the ductus closes shortly after birth whereas IPAVA appear to remain patent in humans as adults.

Compared to humans <50 years old, lower bubble scores are detected with TTSCE in humans >50 years old during exercise breathing air and when breathing hypoxic gas at rest (Norris et al. 2014). These data suggest that _IPAVA is reduced in older subjects. In these studies, caparisons between old and young subjects were made during iso-workload exercise to avoid potential ageing confounders on exercise capacity. The consistent theme with these data is that _IPAVA is a normal occurrence early in life and that the body retains the ability to recruit these pathways during early life and middle age but that _IPAVA declines with age, for reasons that have yet to be determined. The majority of this work in older subjects has been obtained using TTSCE, so these results may be due to bubbles getting through pulmonary capillaries and/or bubbles not getting through pulmonary capillaries in older subjects. As such, work in developmental physiology models using techniques that can quantify _IPAVA will help to move this area forward.

Proposed physiological and pathophysiological roles of IPAVA

Defining the physiological and/or pathophysiological roles of IPAVA remains an exciting and wide-open area of study because these roles are only beginning to be described. Blood flow through IPAVA is proposed to bypass the pulmonary capillaries during conditions of increased pulmonary blood flow such as exercise (Recavarren, 1966). In doing so, it has been hypothesized that there may be multiple effects of _IPAVA. First, _IPAVA may act as a pressure relief mechanism to assist in pulmonary pressure regulation. Second, if _IPAVA is bypassing the gas-exchanging units within the lung, then this blood flow may reduce pulmonary gas exchange efficiency either as a shunt or a diffusion-limited vessel. Third, if blood flow is travelling through large diameter IPAVA, then this may cause a breach in the pulmonary capillary filter allowing emboli to enter into the systemic arterial circulation whereby the embolus can reach the brain or heart subsequently causing adverse neurological and/or cardiovascular sequelae.

Physiology

Pulmonary pressure

As mentioned above, pulmonary pressure and _IPAVA could be related (Stickland et al. 2004), such that IPAVA act as low resistance vascular conduits that help to regulate pulmonary pressure. In this manner, IPAVA would act as ‘pop-off’ valves by allowing _IPAVA, thereby preventing excessive increases in pulmonary vascular resistance and pressure (Stickland et al. 2004; Norris et al. 2014). Stickland and colleagues were first to demonstrate, in one subject without a PFO, that no _IPAVA during exercise was associated with high pulmonary artery pressure, compared to subjects with _IPAVA who had low pulmonary artery pressure (Stickland et al. 2004). However, it is unlikely that the volume of _IPAVA suggested to occur during exercise (<5% of the _T) being diverted through IPAVA would have a significant, much less measurable, effect on pulmonary artery pressure. Rather, we would offer the hypothesis that areas of the lung that are recruited during conditions of increased pulmonary blood flow may also contain IPAVA such that when these areas are recruited, they result in lower pressures and also allow for _IPAVA. In this way, it is the pulmonary vascular recruitment and distension that results in keeping pulmonary vascular resistance low and not the direct perfusion of IPAVA. Although _IPAVA occurs simultaneously under these conditions, it is simply due to the association of these vessels with areas of the vasculature that are recruited under these conditions and direct _IPAVA has no measureable effect on pulmonary pressure as we have previously suggested (Norris et al. 2014).

Pulmonary capillary distension

Others have investigated the relationship of the degree of pulmonary transit of agitated succinylated gelatin contrast and pulmonary vascular resistance during exercise (La Gerche et al. 2010; Lalande et al. 2012). Both of these studies concluded that transpulmonary passage of agitated gelatin contrast was through distended pulmonary capillaries rather than IPAVA. Although these data (La Gerche et al. 2010; Lalande et al. 2012) lend support to the notion that a greater _IPAVA is related to a lower pulmonary vascular resistance during exercise, and therefore may allow for the detection of capillary distension, these authors did not report that subjects were screened for the presence of a PFO. Given the discussion above with respect to the prevalence and importance of PFO in these anatomically based approaches to collecting data, using succinylated gelatin contrast should be interpreted with the understanding that contrast could have either traversed a PFO or IPAVA, or squeezed through pulmonary capillaries as those authors suggested. The differences in in vivo saline contrast bubble dynamics and succinylated gelatin bubble contrast dynamics remains unknown, but the possibility exists that gelatin bubbles are more stable and can therefore squeeze through capillaries but saline bubbles are larger and less stable so they cannot. Studies examining these different contrast agents in conjunction with microspheres of known sizes will help to determine whether or not bubbles are squeezing through pulmonary capillaries.

Pulmonary gas exchange efficiency

The idea that _IPAVA may act as a shunt is not new, yet this idea remains highly contentious. Seminal work by Nomoto et al. demonstrated that intravenous infusion of adrenaline (epinephrine) increased the number of large diameter microspheres (15–30 μm) able to traverse the lung (Nomoto et al. 1974). Concomitant with an increase in microspheres travelling through IPAVA, was a reduction in , suggesting the possibility that blood flowing through IPAVA contributes to pulmonary gas exchange efficiency and may act as a shunt (Nomoto et al. 1974). Stickland and colleagues were the first to show a correlation between _IPAVA detected with TTSCE, and a widening of the alveolar-to-arterial difference () during exercise (Stickland et al. 2004). Work using ^99mTc-MAA suggests that at maximal exercise is negatively correlated with _IPAVA (Lovering et al. 2009b) and that _IPAVA is positively correlated with the (Bates et al. 2014); however, correlation does not mean causation.

Investigations in healthy humans have used intravenous infusions of inotropic drugs to increase _IPAVA while at rest breathing air at sea level. Bryan and colleagues report that with increased IPAVA blood flow there was an increase in calculated shunt fraction (_S/_T; Bryan et al. 2012). These data are in agreement with those obtained with microsphere studies in animals (Nomoto et al. 1974), although the work by Bryan et al. did not explicitly state that their subjects were free of PFO and the increase in _S/_T may have been the result of blood flow through a PFO or increased ventilation to perfusion (/) heterogeneity. Elliott et al. have recently used intravenous adrenaline to increase _IPAVA in subjects breathing air and 40% O₂ (Elliott et al. 2014a). In this experimental paradigm, _IPAVA was the same in both conditions, yet when subjects are breathing 40% O₂ contributions to the from diffusion limitation and / heterogeneity are minimized or prevented (Wagner & West, 1972). These authors calculated that the venous admixture required to explain the was ∼2% of _T in both conditions, breathing either air or 40% O₂, when _IPAVA was present (Elliott et al. 2014a). Taken together, these data support the idea that with diffusion limitation and / heterogeneity minimized or prevented, only postpulmonary shunt and _IPAVA could have played a role in determining pulmonary gas exchange efficiency because these subjects did not have a PFO. Nevertheless, these data were obtained with TTSCE and as such the presence of bubbles in the left heart may be due to passage through pulmonary capillaries and the increase in calculated shunt fraction may have been the result of increases in postpulmonary shunt only, rather than from intrapulmonary shunt as we previously suggested (Elliott et al. 2014a). Accordingly, directly demonstrating deoxygenated blood flow travelling through these pathways and mixing with oxygenated blood will be beneficial in moving this area beyond correlational data to clearly demonstrate cause and effect. Until that time, these data will remain controversial because they are inconsistent with data obtained using the multiple inert gas elimination technique (MIGET). The contemporary explanation for determinants of pulmonary gas exchange efficiency in healthy humans suggest that intrapulmonary and/or intracardiac shunt does not significantly contribute to the (Dempsey & Wagner, 1999). This supposition is supported by data obtained using gas exchange-dependent techniques such as the MIGET and 100% O₂ technique (Torre-Bueno et al. 1985; Hammond et al. 1986; Wagner et al. 1986; Vogiatzis et al. 2008). There are presently no other technical approaches with the capacity to isolate the effect of each determinant of pulmonary gas exchange efficiency. Others, have previously outlined, in detail, the limitations associated with the MIGET technique (Hlastala, 1984) and the reasons why there may be discrepancies between data obtained with anatomically based techniques and those obtained with gas exchange-dependent techniques (Genovesi, 1976; Wagner, 1995; Stickland & Lovering, 2006). Thus we refer the reader to these previous works on the discrepancies between these two technical approaches, which is beyond the scope of this review.

Work by some authors using succinylated gelatin bubble contrast, rather than saline bubble contrast, has suggested that if _IPAVA participated in pulmonary gas exchange then it would ‘…immediately cause true shunt and hypoxemia’ (Lalande et al. 2012). However, this statement, and similar statements by others (La Gerche et al. 2010), do not take into account fundamental concepts of pulmonary gas exchange efficiency (Dempsey & Wagner, 1999). For example, during exercise, when subjects are hyperventilating and alveolar is increased, _IPAVA may act as a true shunt resulting in a widening of the , yet would remain unchanged from resting conditions and therefore arterial hypoxaemia would not occur (Dempsey & Wagner, 1999; Lovering et al. 2005). This would be true of any shunt, not just _IPAVA. It should also be mentioned that arterial oxygen saturation () could be reduced under conditions of constant , by reductions in pH and increases in body temperature that accompany exercise (Dempsey & Wagner, 1999; Lovering & Stickland, 2010), independent of any pulmonary gas exchange impairment.

It is possible that _IPAVA may contribute to pulmonary gas exchange efficiency in some cases, and not represent a source of true shunt in other cases. In this way, IPAVA would represent a vessel where some degree of gas exchange can occur, most likely around the perimeter of the vessel lumen (Conhaim & Staub, 1980) but would be critically dependent upon the size of the vessel and the . Pre-capillary/non-capillary gas exchange is not a new concept (Sobol et al. 1963; Conhaim & Staub, 1980) and has recently been demonstrated to occur in vivo in the murine model (Tabuchi et al. 2013). Tabuchi et al. demonstrated that during normoxic ventilation, complete O₂ equilibration occurs in ∼0.1 s and begins in arterioles ≤30 μm in diameter. In light of these data, blood flow through IPAVA >50 μm in diameter would probably not participate in pulmonary gas exchange, thereby impairing pulmonary gas exchange efficiency. Whereas blood flow through IPAVA ≤30 μm in diameter may participate in pulmonary gas exchange, thereby not impairing pulmonary gas exchange efficiency. Whether or not this is the case in humans is unknown and will remain unknown until this is directly visualized as it has been in the mouse. As such, the size of IPAVA and the are important and would need to be taken into consideration when determining if these vessels, or any pulmonary vessel for that matter, played a role in pulmonary gas exchange efficiency.

Pathophysiology

Pulmonary capillaries filter out particles (i.e. emboli) >7–15 μm in diameter. IPAVA are known to be large diameter pathways and _IPAVA may represent a breach in the pulmonary capillary filter (Abushora et al. 2013). If an embolus were to flow through an IPAVA it could end up in the heart or brain and result in myocardial infarction, transient ischaemic attack (TIA), or stroke. Prizel et al. demonstrated that ∼13% of the myocardial infarcts in their autopsy study were the result of coronary artery embolic infarct (Prizel et al. 1978). In support of this, Clark et al. (2013) has recently suggested that cardiac ischaemia in Hereditary Haemorrhagic Telangiectasia (HHT) patients with known PAVMs is the result of emboli traversing the lung via PAVMs and becoming lodged in a coronary artery (Clark et al. 2013). Stoddard and colleagues demonstrated that _IPAVA in individuals at rest breathing air occurred five times more frequently in a subgroup of individuals who had suffered a cerebrovascular accident that was cryptogenic in nature compared to healthy individuals (Abushora et al. 2013). In this study, _IPAVA at rest breathing air was a stronger predictor of stroke than both hypertension and hyperlipidaemia (Abushora et al. 2013). The TTSCE data in the aforementioned studies are supported by data in rats using large diameter microspheres, where the microspheres that traversed the pulmonary microcirculation, presumably through IPAVA, became lodged in the brain (Bates et al. 2012). The presence of a breach in the pulmonary capillary filter, however, only provides a potential pathway for emboli, and by no means guarantees adverse neurological or cardiovascular sequelae.

Seemingly inconsistent with these above data, most people are reported to have _IPAVA during exercise as detected with TTSCE, yet most people do not have stroke or TIA when they exercise. Chronic aerobic activity reduces coagulation potential and therefore lowers the probability for clot formation (Womack et al. 2003) which probably explains why most healthy people do not have a stroke or TIA when they go for a run. Nevertheless, until large diameter microspheres are injected into an animal under conditions that increase _IPAVA and the animal subsequently has a stroke and/or TIA, there will be no direct demonstration of causation.

Additional support for IPAVA acting as a breach in the pulmonary capillary filter comes from work by Dujic and colleagues suggesting that _IPAVA could allow arterialization of gas emboli that formed during the diving ascent (Ljubkovic et al. 2012). In this study, all subjects with post-dive arterialization of gas emboli had larger degrees of _IPAVA during exercise, while subjects without post-dive arterialization of gas emboli were individuals with no _IPAVA during exercise, as assessed using TTSCE (Ljubkovic et al. 2012). Whether or not this contributes to arterial gas embolism and subsequently decompression sickness remains controversial but exercise after a dive increases arterial gas embolism (Madden et al. 2013).

As mentioned above, _IPAVA has been suggested to play a role in pulmonary gas exchange efficiency in healthy humans, but this has not been directly established. Nonetheless, a recent case report of a patient with Beriberi heart implicates _IPAVA in the pathophysiology of this disease as well (Nakano et al. 2013). At presentation the patient had a _T of 14 l min⁻¹ with only a mild increase in pulmonary pressure along with of 93% on air and _IPAVA was present as detected by TTSCE. After treatment with thiamine repletion, _T decreased to 6.6 l min⁻¹, was 99% on air and _IPAVA was absent as detected by TTSCE. Although this is only a case report, the increase in _T concomitant with the presence of _IPAVA closely resembles the conditions required to increase _IPAVA in healthy humans (Elliott et al. 2014a). Using an animal model of this disease and large diameter microspheres would directly address the question of whether or not IPAVA are involved in this disease. Clearly, these and other potential pathophysiological roles of _IPAVA should be investigated but much more work is required in various patient populations in this area to confirm and/or refute these roles, which at present are only hypothesized roles.

Conclusions and future directions

In summary, we have presented a case that IPAVA exist and that _IPAVA may be involved in both physiological and pathophysiological processes. However, this line of research requires more work in well-designed studies using appropriately screened subjects without PFO to continue to make advances that will establish these vessels as relevant in human health and disease. Pertinent areas of future research include determining the precise location of IPAVA in healthy human lungs using modern anatomically based techniques, which will facilitate mechanistically based approaches to determining the regulation of these pathways. For example, real-time visualization of IPAVA and _IPAVA in vitro and in vivo will be required for studies of these pathways during the developmental and ageing processes. Additionally, there are humans with _IPAVA at rest who have been excluded from most studies outlined in this review. Work in these individuals may provide insight into the mechanisms regulating these blood vessels and the impact of this blood flow on physiological processes, if any. Important areas of future focus must be on quantifying _IPAVA during various conditions in intact humans and animals and directly measuring the impact of _IPAVA on pulmonary gas exchange efficiency. Notably, developing the ability to prevent _IPAVA under conditions when it is normally present, e.g. during exercise, and measuring the accompanying effect on pulmonary gas exchange efficiency would allow for demonstration of a cause and effect, if one existed. Lastly, the development of large animal models to investigate the mechanistic regulation of IPAVA and to understand the data discrepancies between anatomically based techniques and gas exchange-dependent techniques will be instrumental in resolving existing controversies.

Glossary

Abbreviations

A−aDO2

alveolar-to-arterial partial pressure of O₂ difference

CaO2

arterial O₂ content

FIO2

fraction of inspired oxygen

HHT

hereditary haemorrhagic telangiectasia

HPV

hypoxic pulmonary vasoconstriction

IPAVA

intrapulmonary arteriovenous anastomoses

MAA

macroaggregates of albumin

MIGET

multiple inert gas elimination technique

Pv¯O2

mixed venous partial pressure of O₂

PaO2

arterial partial pressure of O₂

PASP

pulmonary artery systolic pressure

PAVM

pulmonary arteriovenous malformation

PFO

patent foramen ovale

SaO2

arterial O₂ saturation

SpO2

peripheral estimate of arterial O₂ saturation

^99mTc-MAA

technetium-99m-labelled MAA

TIA

transient ischaemic attack

TTSCE

transthoracic saline contrast echocardiography

Q̇T

cardiac output

Q̇IPAVA

blood flow through IPAVA

Q̇SQ̇T

shunt fraction

V̇O2

O₂ consumption, ventilation to perfusion ratio (V̇/Q̇)

Biographies

AndrewLovering,PhD, is an integrative physiologistwhohas investigated awide range of respiratory systemquestions from understanding how medullary respiratory neurons reconfigure their activity during hypoxia-induced periodic breathing in sleeping cats to understanding the roles of intracardiac and intrapulmonary shunt on the regulation of pulmonary gas exchange efficiency in humans exercising above 5000 m; he thinks that the patent foramen ovale is highly underrated.

Joseph (JJ) Duke, PhD, was a postdoctoral fellow in Dr Lovering’s laboratory and investigated the regulation and quantification of blood flow through intrapulmonary arteriovenous anastomoses in healthy humans during exercise and at rest breathing hypoxic gas.He is currently an Assistant Professor.

Jonathan Elliott, PhD,was a graduate student in Dr Lovering’s Laboratory and investigatedmechanisms regulating blood flow through intrapulmonary arteriovenous anastomoses and quantifying the associated physiological impact of this blood flow on pulmonary gas exchange efficiency. He is currently a postdoctoral fellow.

Additional information

Competing interests

None declared.

Funding

Our research is supported by: American Heart Association Scientist Development Grant No. 228023; American Physiological Society's Giles F. Filley Memorial Award for Excellence in Respiratory Physiology & Medicine; American Lung Association in Oregon and administered through the American Thoracic Society (ATS Grant No. C-10-014); Defense Medical Research and Development Program Grant No. W81XWH-10-2-0114/No. DM1027581JTCG5 TATRC (A.T.L.); American Heart Association Pre-doctoral Fellowship (J.E.E.); Eugene and Clarissa Evonuk Graduate Fellowship in Environmental and Exercise Physiology (J.E.E.).