Our isolated lung preparation has been well-described in detail previously (Lovering et al., 2007; Stickland et al., 2007; Bates et al., 2012). Briefly, rats (n = 6–9, each group) were secured supine, the trachea was cannulated, the femoral vein was catheterized, and 6,000 U/kg heparin was injected intravenously. To prevent clot formation during lung isolation, blood was flushed from the rat by cutting the femoral artery and simultaneously perfusing the femoral vein with 10 mL of isotonic saline. The lungs were then inflated with 5 mL of air and the chest wall was removed. A fluid filled catheter was placed in the pulmonary artery and connected to a continuous flow reservoir, allowing the fluid column height to remain constant and the lungs to be gravitationally perfused. In this way, the pulmonary artery perfusion pressure is determined by the height of the fluid column. An additional fluid filled catheter was placed in the left atrium for the collection of the lung effluent and maintained at the height of the left atrium so that the left outlet pressure was 0 cmH₂O. The lungs were mechanically ventilated 30–45 times/min to a peak inflation pressure of 15 cmH₂O with 5 cmH₂O end expiratory pressure.

Perfusion with 3% hetastarch at 40 cmH₂O pulmonary artery pressure increased the rate of perfusate flow across the lung compared to 20 cmH₂O (6.9 ± 1.0 vs. 3.3 ± 1.1 mL/min, p < 0.001). Flow across lungs perfused with isoproterenol was similar to control lungs (4.3 ± 1.1 mL/min, p > 0.05). The injection of 50 μm microspheres decreased perfusate flow 29.7 ± 25.1%. Congruent with our previous observations (Bates et al., 2012), the percent passage of microspheres across the lung with 20 cmH₂O perfusion pressure was trivial (0.001 ± 0.002%) and this was not increased by raising the perfusion pressure (0.005 ± 0.005%, p > 0.05; Figure 1). Perfusion with isoproterenol at 20 cmH₂O increased microsphere passage compared to the hetastarch control (0.221 ± 0.269%, p < 0.05) and this was prevented with pre-treatment with propranolol (0.001 ± 0.003%, p > 0.05).

Hemodynamic parameters and blood gas values for our intact rat studies are given in Table 1. As expected, hypoxia depressed PaO₂ and hyperoxia raised PaO₂, compared to 21% O₂. Infusion with isoproterenol increased heart rate. As we demonstrated previously, hypoxic ventilation increased transpulmonary passage of 15 μm microspheres (14.5 ± 4.5%), compared to room air (0.11 ± 0.03%, p < 0.001; Figure 2). Similarly, isoproterenol increased microsphere passage in normoxic rats (18.9 ± 4.8%, p < 0.001). Isoproterenol-induced IPAVA recruitment was eliminated by ventilation with 100% O₂ and pre-treatmentwith propranolol (0.03 ± 0.04 and 0.07 ± 0.04%, p > 0.05 compared to room air control). Pre-treatment with propranolol also significantly reduced hypoxia-induced IPAVA perfusion (4.6 ± 2.2%, p < 0.05).

In experiments with multiple size microspheres, a trivial number of microspheres traversed the lung with normoxic and hyperoxic ventilation, and only the smallest microspheres were found in the liver (Table 2). Hypoxic ventilation and isoproterenol infusion increased the passage of all microsphere sizes, including 70 μm microspheres. The passage of larger microspheres with isoproterenol infusion and hypoxic ventilation was eliminated by hyperoxia and propranolol, respectively.

We were inspired to consider that IPAVA might be adrenergically regulated by (1) the observed epinephrine-induced recruitment of IPAVA in healthy humans (Laurie et al., 2012; Elliott et al., 2014), (2) our clinical observation of IPAVA in a young asthmatic patient treated with high doses of β₂-adrengergic agonists (inhaled albuterol plus i.v terbutaline; Bates et al., 2014b), and (3) the fact that increased sympathetic tone is a common feature to both high intensity exercise and hypoxia. Isoproterenol causes vasodilation by binding the β₂-adrengergic receptor, thereby increasing intracellular cAMP and relaxing vascular smooth muscle. Little is known about the three dimensional distribution of β₂-adrengergic receptors in the pulmonary vasculature, but mapping their location may offer some clue as to the location of IPAVA pathways.

Interestingly Laurie et al. found using saline contrast echocardiography that propranolol did not qualitatively reduce IPAVA recruitment in hypoxic humans. These studies used a qualitative scoring mechanism that may not be sufficiently sensitive to quantify changes in IPAVA recruitment. Using our quantitative microsphere method, we found that propranolol did not completely eliminate hypoxia-induced IPAVA recruitment, but substantially reduced it. Therefore, while adrenergic blockade can limit blood flow through IPAVA, IPAVA recruitment by hypoxia was not completely prevented by propranolol. It is possible that propranolol does not completely eliminate IPAVA recruitment and that the hypoxic vasoconstriction downstream of these pathways redirects flow through the remaining patent IPAVA pathways. Alternatively, these data point to the existence of additional co-factors that may be required for complete IPAVA recruitment by hypoxia. There may be additional vasoactive neural or humoral factors that are induced by hypoxia that regulate the recruitment of IPAVA that are not blocked by propranolol.

Importantly, this work is the first to demonstrate quantitatively, using microspheres of a discreet diameter, that hyperoxia prevents blood flow through IPAVA. A potential interpretation of our in-vivo data is that the 15 μm microspheres pass through distended capillaries with isoproterenol, and not IPAVA. However, the fact that hyperoxia decreases microsphere passage with isoproterenol infusion suggests that this is not the case. Furthermore, hyperoxia completely antagonizes IPAVA recruitment by isoproterenol. This is in contrast to propranolol, which only partially blocked the impact of hypoxia. This suggests that hyperoxia is working through an independent mechanism.

The structure and anatomic location of these pathways remains a topic of debate and defining them morphometrically is an important next step. Weibel carefully and systemically characterized the anatomy of the pulmonary vasculature by evaluating serial histological sections of the human lung and noted no evidence of direct arteriovenous connections (Weibel, 1963). In contrast, Tobin studied the pulmonary vascular tree in lungs infused with casting material and found glomerular-like structures early in the vascular tree with direct arteriovenous connections (Tobin and Zariquiey, 1950; Tobin, 1966). He suggested that these structures are not visible in traditional histological sections because they have little elastic tissue and collapse during sectioning. It certainly is possible that, IPAVA are simply passive structures and that flow through them is determined by the tone of the vasculature surrounding them. That is to say, in our study, isoproterenol and propranolol may be changing the tone of the pulmonary arteries distal to these pathways, thereby redirecting blood flow through the lower resistance IPAVA. In support of this, we recently found that the transpulmonary passage of ^99mTc-albumin macroaggregate particles is higher in conditions when the pulmonary vascular resistance is higher (Bates et al., 2014a). In contrast, simply increasing pulmonary artery pressure and flow in our isolated lung preparation was not sufficient to recruit IPAVA. Furthermore, hyperoxia closes IPAVA even when the cardiac output is elevated by catecholamine infusion and exercise. This supports the theory that, even if the distal pulmonary vascular resistance does influence the degree to which IPAVA are perfused, they are not simply distended or recruited. We doubled the perfusate flow in our isolated lung prep, and this did not translate to an increase in IPAVA recruitment. We propose that intrapulmonary arteriovenous anastomoses may be biochemically recruited with their perfusion determined by the downstream vascular resistance.

Recall that IPAVA are (1) recruited by hypoxia and catecholamines, (2) closed by hyperoxia, and (3) a large component of the hypoxia response is blocked by propranolol. We cannot help but be struck by how similar these responses are to those of systemic blood vessels, leading us to speculate that IPAVA may, in fact, be systemic-like vessels. Since we first compared the recruitment of IPAVA by hypoxia in healthy, intact animals and isolated lungs, Galambos et al. have published several important morphometric studies of arteriovenous connections in human lungs from pediatric patients with lung disease, including alveolar capillary dysplasia, pulmonary hypertension, and bronchopulmonary dysplasia (Galambos et al., 2013, 2014, 2015; Acker et al., 2015; Ali et al., 2015). This group found that the right-to-left passage of mixed venous blood may occur in pediatric patients with fatal hypoxemia via anastomoses between the pulmonary arterial and bronchial circulations. The substantial communications between the pulmonary and bronchial circulations are well-documented and pulmonary artery-to-bronchial artery and bronchial artery-to-bronchial vein anastomoses having been observed in healthy lungs (Aramendia et al., 1962; Hyman et al., 1975; Charan et al., 1986; Thapar et al., 1991; Hasegawa et al., 1999). Furthermore, bronchial arteries are systemic in nature, being dilated by hypoxia, and catecholamines and constricted by hyperoxia. It is possible that the IPAVA we have been interested in may be bronchopulmonary in nature. More work is needed to understand the hemodynamic interactions between these two circulations.

This is the first time that we have been able to significantly increase the transpulmonary passage of large microspheres across the isolated, perfused lung, although it only increased to a peak of 0.8% in our isolated lung prep. Microsphere passage is typically trivial in the isolated lung and these data represent a 10–400 fold increase in microsphere passage. We acknowledge that this is not as large as the increase in the intact animal and it is worthwhile to speculate about the differences between our in situ and in vivo studies that may have led to this difference. The decreased IPAVA perfusion in the isolated lung may have been a result of the larger particles used in this experiment. We used 50 μm microspheres in our isolated lung studies in order to be consistent with our previous isolated lung studies (Lovering et al., 2007; Stickland et al., 2007; Bates et al., 2012) and increase our confidence that our findings are not simply the result of capillary distension. It requires >100 mmHg perfusion pressure to distend a pulmonary capillary from 6 to 50 μm (Linehan et al., 1992; Bates et al., 2012). We did not use 50 μm particles in our intact animal studies because their injection frequently results in hemodynamic instability. We have shown previously that the injection of 15 μm particles does not substantially impact pulmonary hemodynamics. In our intact animal model, 15 μm microsphere passage was >50 μm microsphere passage and the lower IPAVA flow in our isolated lungs may be because there are fewer IPAVA pathways that are >15 μm. We used 15 μm microspheres to quantify IPAVA flow in our intact studies because we they do not traverse the normoxic lung (here and Eldridge et al., 2004). Future studies may consider combining isolated lung physiology studies with histology to better understand the size distribution of IPAVA pathways.

Secondly, it is possible that there is lower IPAVA recruitment in the isolated lung because of differences in the experimental conditions between the isolated lung and intact animal. We previously described additional limitations of our isolated lung technique and believe they are equally applicable in this study (5). Total pulmonary flow is less in the isolated lung than the intact rat and the lung is positive pressure ventilated. Our isolated lung preparation is also perfused at room temperature with a non-blood perfusate that is in equilibrium with room air, giving it a non-physiological PvO₂ and PvCO₂. It is possible that this contributes to sub-maximal IPAVA recruitment in the isolated lung prep.

A final limitation of this study is that, given our technical limitations, we were not able to simultaneously measure cardiac output and inject microspheres into the left ventricle in our intact studies. This motivated us to perform experiments in our isolated lung model where we can control pressure and flow.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.