Morphological and morphometric specializations of the lung of the Andean goose, Chloephaga melanoptera: A lifelong high-altitude resident

High altitude flight in rarefied, extremely cold and hypoxic air is a very challenging activity. Only a few species of birds can achieve it. Hitherto, the structure of the lungs of such birds has not been studied. This is because of the rarity of such species and the challenges of preparing well-fixed lung tissue. Here, it was posited that in addition to the now proven physiological adaptations, high altitude flying birds will also have acquired pulmonary structural adaptations that enable them to obtain the large amounts of oxygen (O2) needed for flight at high elevation, an environment where O2 levels are very low. The Andean goose (Chloephaga melanoptera) normally resides at altitudes above 3000 meters and flies to elevations as high as 6000 meters where O2 becomes limiting. In this study, its lung was morphologically- and morphometrically investigated. It was found that structurally the lungs are exceptionally specialized for gas exchange. Atypically, the infundibulae are well-vascularized. The mass-specific volume of the lung (42.8 cm3.kg-1), the mass-specific respiratory surface area of the blood-gas (tissue) barrier (96.5 cm2.g-1) and the mass-specific volume of the pulmonary capillary blood (7.44 cm3.kg-1) were some of the highest values so far reported in birds. The pulmonary structural specializations have generated a mass-specific total (overall) pulmonary morphometric diffusing capacity of the lung for oxygen (DLo2) of 0.119 mlO2.sec-1.mbar-1.kg-1, a value that is among some of the highest ones in birds that have been studied. The adaptations of the lung of the Andean goose possibly produce the high O2 conductance needed to live and fly at high altitude.

The respiratory physiology of birds that fly at high altitude differs profoundly from that of low level flying ones [6]. For example, in the bar-headed goose (Anser indicus) [11,12,19], the Andean goose (Chloephaga melanoptera) [22,23], the Tibetan chicken (Gallus gallus) [24] and the Ruppell's griffon vulture (Gyps rueppellii) [25], mutations of hemoglobin have generated molecular configurations that increase O 2 affinity [23][24][25][26][27][28]. Bar-headed geese in particular have evolved increased capacity of transporting O 2 along the path from the atmosphere to the mitochondria [19]. Recently, it was shown that Andean geese, lifelong residents of the Andes, have evolved a different strategy of matching O 2 supply to demand in hypoxic environments: while bar-headed geese match O 2 supply and demand during rest in low O 2 environments with large increases in ventilation [6,28], the Andean geese exhibit very modest ventilatory response and instead greatly increase the amount of O 2 extracted from each breath [29]. The latter suggests that Andean geese possess pulmonary structural adaptations that increase the capacity for O 2 acquisition.
While the unique functional design of the avian respiratory system [30-39] undoubtedly increases O 2 uptake efficiency, details on the structure of the lungs of the elite high-altitude flying birds are totally lacking. This is because of rarity of such species and the challenges posed in obtaining adequately fixed lung tissues. Of the known 63 species and 15 orders of volant and non-volant bird species on which pulmonary morphometric data exist [33,36,37,39], none of them is an extreme high-altitude flyer. Here, we examined the lungs of the Andean goose, a large bird that lives at altitude between 3,000 and 6,000 meters of the Andes [40] to determine possible morphological specializations that may support its lifestyle.
The abbreviations used in the text are defined in Table 1 and the morphometric values are given as mean ±SD. Surface area of the blood-gas (tissue) barrier (S (t) ) per unit volume of the lung parenchyma (V (LP) ), i.e., the surface density of the respiratory surface area V L Volume of the lung V (LP) Volume of the lung parenchyma V (PCB) Volume of the pulmonary capillary blood V (PCB) .S (t) -1 Volume of the pulmonary capillary blood (V (PCB) ) per unit surface area of the blood-gas (tissue) barrier τ hb Harmonic mean thickness of the total barrier, i.e., the distance between the respiratory surface and the surface of the erythrocyte cell membrane (the air-hemoglobin pathway) τ ht Harmonic mean thickness of the blood-gas (tissue) barrier

Materials and methods
Ethical clearance, location and method of capture of the birds Lungs were obtained from three wild Andean geese (Chloephaga melanoptera) captured and raised in San Pedro de Casta, Perú at an altitude of 3180 meters. Animals were collected under permit # 36087-2012 from the Gestión Florestral y de Fauna Sylvestre, Ministerio de Agricultura, Republica del Perú. They were two years old when sampled. All procedures were conducted according to guidelines approved by the Animal Care Committee at the University of British Columbia (A16-0019) in accordance with the Canadian Council on Animal Care.

Fixation of the lungs
The birds were weighed and then killed by intravenous injection of propofol into the tibiotarsal vein (>20 mg/kg). With the body in a supine position, the lungs and the air sacs were fixed by intratracheal instillation with 2.3% glutaraldehyde buffered with sodium cacodylate (osmolarity 350 mOsm and pH 7.4) at a pressure head of 3000 Pa (1 cm H 2 O = 1 mbar = 10 2 Pa). The top of the funnel, which was constantly topped up, was held 30 cm above the sternum of the supine bird. When it stopped flowing by gravity, the body wall was repeatedly gently squeezed to expel the air from the lung and the air sacs to achieve better penetration of the fixative. When the fixative finally stopped flowing, the trachea was ligated and the fixative left in the respiratory system for~4 hours. Thereafter, the lungs were removed from their costovertebral attachments and immersed in fixative.

Determination of the lung volume (V L )
In the laboratory, the extrapulmonary primary bronchus was trimmed close to the hilum and the adhering fat and connective tissue elements removed before the V L was determined by the weight displacement method of Scherle [41]. Sampling of the lung and microscopic and morphometric analyses of the lung. Details on lung sampling, tissue processing for microscopy and morphometric analyses are given in Maina [42]. They are succinctly outlined below.
Sampling and analysis of the main structural components of the lung at the light microscopic level of magnification. The left lung of one of the birds was cut into five slices along the costal sulci and the slices in turn cut into halves just dorsal to the primary bronchus. The ten half-slices were processed and embedded in paraffin wax with the cranial face directed anteriorly. Sections were cut at 8 μm thickness and the first technically adequate slice stained with hematoxylin and eosin. The volume densities of the lung parenchyma (LP), the lumina of the parabronchi (tertiary bronchi) and the secondary bronchi, the blood vessels larger than blood capillaries (BCs) and the primary bronchus were determined field-by-field by pointcounting using a 100-point Zeiss integrating graticule at a final magnification of x100. The absolute volumes of the structural parameters were calculated from the V L , the reference space.
Sampling and analysis of the structural components of the LP at the transmission electron microscopic level of magnification. The right lungs of the three birds were cut into five slices along the costovertebral sulci and the slices then cut into halves just dorsal to the primary bronchus. From the cranial surfaces of each of the ten half slices, 1 mm thick slices were cut and laid out flat. A transparent acetate paper with a quadratic lattice grid on which squares were numbered was then dropped onto the surface of the slice to avoid bias. Six random numbers that fell within the range of those on the part of the grid that lay on the slice were generated from a free computer software (https://www.random.org/integers) and small pieces of lung tissue taken from the areas (squares) where the numbers were located. When a random number fell on a large blood vessel or an airway (components of the lung that were analyzed at light microscopic level), an extra random number was generated until six pieces of tissue were sampled from the LP. The pieces of tissue were cut to a size of~1 mm 3 and processed for transmission electron microscopy. From the group of blocks processed from the pieces of tissue from each half slice of the lung (normally more than six in number), one block was picked at random and semithin sections cut and stained with toluidine blue. The block face was trimmed to an appropriate shape and size by removing the non-parenchymatous structures. Ultrathin sections were cut, mounted on 200-square wire-mesh coated copper grids and stained with lead citrate and counter-stained with uranyl acetate. For determination of the volume densities and the surface areas of the components of the LP, eight electron micrographs were taken from a predetermined corner (top right) of the grid (to avoid bias) from each section, at a primary magnification of x4,400. The same number of electron micrographs (80) was taken from the same areas of the grid squares at a higher primary magnification of x13,000 for the determination of the harmonic mean thickness of the blood-gas (tissue) barrier (τ ht ) and the harmonic mean thickness of the total barrier, i.e., the distance between the respiratory surface and the erythrocyte cell membrane (τ hb ). The images were enlarged by a factor of x2.5 and a quadratic lattice grid superimposed on top. The volume densities of the components of the LP, namely the BCs, the air capillaries (ACs) and the structural tissue of the LP were determined by point-counting, the surface areas by intersection counting and the τ hb and the τ ht by intercept length measurement using a logarithmic scale [43,44]. The absolute volumes of components of the LP were calculated as the product of their volume densities and the volume of the LP, the reference space.
Morphometric modeling of the lung of the Andean goose for conductance of oxygen. After determining the relevant morphometric parameters, the membrane diffusing capacity (DMo 2 ), the morphometric diffusing capacity of the erythrocyte (Deo 2 ) and the total morphometric pulmonary diffusing capacity of the lung for oxygen (DLo 2 ) were determined by applying the appropriate O 2 permeation coefficient (Kto 2 ) and the O 2 uptake coefficient of the whole blood (

Morphological findings
The cone-shaped lungs ( Fig 1A) of the Andean goose were firmly attached to the ribs and the vertebrae across six costovertebral sulci (Fig 1A and 1B). In some parts of the lung, interparabronchial septa were present while in others they were lacking (Fig 1C-1F). Interparabronchial blood vessels gave rise to relatively smaller intraparabronchial ones (Figs 1C-1E, 2A and 2B), the atria were conspicuous (Figs 1C, 1D and 2A-2D) and the infundibulae were intensely vascularized (Fig 2C-2F). This is an uncharacteristic feature of the lungs of birds that have so far been studied where the surface of the infundibulae is nonvascular (Fig 3A) with BCs being located in the LP. Probably to optimize respiratory surface area, occasionally, single BCs surround ACs (Fig 3B). Fig 3C shows an extravasated erythrocyte that measures~13 μm in diameter lying next to an AC and one of the cells lodged into an AC (Insert). While most of the BCs connected by means of epithelial-epithelial cell retinaculae that also separated the ACs (Fig 3D  and 3E), some BCs attached directly (Fig 3D and 3E). The blood-gas (tissue) barrier (BGB) comprised an epithelial cell, an endothelial cell and a common basement membrane ( Fig 3F) and was rather uniform in thickness (Fig 3D-3F).   The mean body mass and V L of the Andean geese were respectively 2,636±144 g and 112±5.23 cm 3 (S1 Table). The volume densities of the main structural components of the lung were: lung parenchyma (LP) = 61.7±7.4%; parabronchi and secondary bronchi = 25.8±3.7%; blood vessels larger than BCs = 9.5±1.4% and; primary bronchus = 3.0±5.2% (S2 Table). Concerning the structural components of the LP, the volume densities of the ACs, the BCs and the structural tissue were respectively 63.4±3.0, 28.2±2.9 and 8.4±0.13% (S3 Table). The surface area of the BGB was 2.52±0.025 m 2 (S4 Table) and the τ ht and the τ hb (S5 Table), were respectively 0.222±0.01 μm and 0.474±0.08 μm and the Deo 2 , the DMo 2 and the DLo 2 (S6 Table) were respectively 0.221±0.02, 3.87±0.89 and 0.312±0.03 mlO 2 .sec -1 .mbar -1 .
Comparison of some pulmonary morphometric parameters of the lung of the Andean goose with those of some species of birds is shown on Table 2 (S7 Table) and with those of the population of birds on which data are available (S8 Table) on Figs 4-9.

Discussion
This study reports that certain morphological-and morphometric specializations occur in the lung of the Andean goose (Chloephaga melanoptera). Such features should increase the gas exchange efficiency of the lung and may explain the high pulmonary O 2 extraction rates that were recently reported in this species by Lague et al. [28]. The Andean goose is one of the largest birds in the Western Hemisphere that lives at an altitude of above 3000 m in valleys and plateaus of the Andes from Peru to southern Chile. The species is non-migratory and may breed at elevations as high as 5,500 meters but rarely ever descends below 2,000 m [40].

Morphological findings
Compared with the lungs of the other species of birds that have so far been studied where the infundibulae are avascular [33, 36] (Fig 3A), in the lung of the Andean goose, the structural units are intensely vascularized (Fig 3C-3F). To an undermined extent, this feature should substantially increase the respiratory surface area and the volume of the pulmonary capillary blood (V (pcb) ). Location of infundibular BCs next to the parabronchial lumen (PL) (Fig 2C-2F) results in the following features that increase respiratory efficiency: (a) the infundibular dead-space air is reduced or possibly eliminated, (b) the infundibular BCs are exposed to inspired air with the highest possible partial pressure of O 2 and (c) the BCs are not affected by the process of 'O 2 screening' [46,47].
The significance of the sparse distribution of the interparabronchial septa in the lung of the Andean goose (Figs 1C, 1D and 2A) is unclear. Scheid [48] contended that since no air pressure differences exist between the parabronchi in the avian lung, little if any air moves between them. If the assertion is correct, the sparse distribution of the interparabronchial septa may be of no functional consequence in the lung of the Andean goose. The septa may be a phylogenetic relic carried over during the evolution of the lung. Interparabronchial septa are welldeveloped in the lungs of the galliform birds [49-53] while they are totally lacking in those of the passerine ones [53,54]. transmission electron micrograph showing the structure of the blood-gas (tissue) barrier. It [blood-gas (tissue) barrier] consists of an epithelial cell (Ep), an endothelial cell (En) and a basement membrane (BM). ?, blood capillary; *, air capillary. The insert shows a logarithmic scale that was used to determine the harmonic mean thickness of the blood-gas (tissue) barrier and that of the total barrier. In the avian lung, the mean volume density of the LP is 49% [33, 36, 39, [52][53][54]. The values range from 18 to 78% respectively in the lungs of the emu (Dromaius novaehollandiae) [55] to that of the ostrich (Struthio camelus) [56]. For the Andean goose, the mean value of 62% lies on the upper end of the range in birds. The highest mass-specific respiratory surface areas in birds of 87.1 and 86.5 cm 2 .g -1 have respectively been reported in the lungs of a 7.3 g violeteared hummingbird (Colibri coruscans) [57] and a 13.7 g African rock martin (Hirundo fuligula) [33, 36, 37, 39, 58]. The value of 96.5 cm 2 .g -1 for the lung of the Andean goose is the highest so far reported in a bird ( Table 2). The very high mass specific respiratory surface area of Table 2. Comparison of some pulmonary morphometric parameters of the Andean goose (Chloephaga melanoptera) with those of other species of birds. BM, body mass; τ ht , harmonic mean thickness of the blood-gas (tissue) barrier; VL.BM -1 , volume of the lung per unit body mass; V (PCB) .BM -1 , volume of the pulmonary capillary blood per unit body mass; V (PCB) .S(t)-1, volume of the pulmonary capillary blood per unit surface area of the blood-gas tissue barrier; S (t) .BM -1 , surface area of the blood-gas (tissue) barrier per unit body mass; S (t) .V(LP)-1, surface area of the blood-gas (tissue) barrier per unit volume of the lung parenchyma; DLo 2 .kg -1 , total (overall) morphometric pulmonary diffusing capacity of the lung per unit body mass. 800 cm 2 g -1 that was reported for an unnamed species of hummingbird by Stanislaus [59] should be treated with caution since it is not explained how the value was determined.
In vertebrate lungs, respiratory surface area is increased by subdivision of the LP [33, 36, 39] and/or gross enlargement of the lung [60]. In the avian lung, the respiratory units (the ACs) range in diameter from 3 μm in the smaller species [54] to 20 μm in that of the ostrich [56]. In this study, while the diameters of the ACs were not directly determined, their sizes corresponded to those of the erythrocytes of~13 μm (Fig 3C).
The large S (t) .V (LP) -1 of 330±70.6 mm 2 .mm -3 that exceeds the values of most birds (Table 2) shows that the very small ACs are generated by intense subdivision of the LP. In birds that have so far been investigated, a S (t) .V (LP) -1 of 82 mm 2 .mm -3 was reported in the lung of the emu [55] and the highest ones of 389 mm 2 .mm -3 in those of the violet-eared hummingbird  [54][55][56] and are given in data supporting this paper (S8 Table). . The extreme subdivision of the LP of the avian lung explains the unexpected outcome that while the avian lung is 27% smaller compared to that of a mammal of equivalent body mass and the volume density of the LP is about one-half that of a mammalian lung [1,33,36,39], the respiratory surface in a bird lung is~15% greater. In birds, the thinnest BGBs (τ ht ) have been reported in the lungs of the violet-eared hummingbird (0.099 μm) [57], the house sparrow (0.096 μm) and the African rock martin (0.090 μm) [33, 36, 39]. The thickest BGBs occur in the ostrich (0.56 μm) [56] and the Humboldt penguin (Spheniscus humboldti) (0.53 μm) [61] lungs. For the Andean goose, the mean τ ht of 0.222 μm was not relatively remarkably thin (Table 2; Fig 8). A τ ht of 0.118 μm was reported in the lung of a 3.84 kg body mass low altitude flying greylag goose (Anser anser) [33,36,37,58]. For a bird that lives and flies in a cold and hypoxic environment where heart rate and cardiac output may need to be constantly high, the thicker BGB in the lung of the Andean goose may help avert structural failure of the BGB, a feature that has been reported to occur in the avian lungs [62,63]. The particularly thick BGB of the lung of the Humboldt penguin [61] [54][55][56] and are given in data supporting this paper (S8 Table). https://doi.org/10.1371/journal.pone.0174395.g005 Lung of the Andean goose and the presence of plentiful connective tissue elements, especially of collagen in the BGB [64], was attributed to the capacity of the lung tolerating high hydrodynamic pressures during dives.
Gas exchangers, including the lungs, are characteristically intensely vascularized [65][66][67][68][69]. The greater the degree of vascularization the larger the volume of blood in the BCs [33,36,37,39,54]. In the African rock martin, 29% of the volume of the lung consists of blood, with 79% of it located in the BCs [33,36,37,39,54]. For the Andean goose, the volume of blood comprised 30% of the V L , with 65% of it in the BCs. Among the data available on the bird lungs, the mass-specific V (PCB) of the lung of the Andean goose (7.44 cm 3 .kg -1 ) is only surpassed by that of the diving Humboldti penguin (8.02 cm 3 .kg -1 ) [60] ( Table 2). The ratio of the V (PCB) to the total respiratory surface (SA), i.e., V (PCB) .S (t) -1 , the so-called capillary loading [70], indicates the degree of exposure of pulmonary capillary blood to air [65,70], with low values indicating high gas exchange efficiency. In the avian lungs that have been investigated, V (PCB) .S (t) -1 ranges from 0.7 cm 3 .m -2 in the African rock martin [33, 36, 37, 39, 54] to 4.4 cm 3 .m -2 in the Humboldt penguin [55]. For the lung of the Andean goose, the particularly low value of 0.8 cm 3 .m -2 points out to a specialization for gas exchange. Among the data available on the avian lungs, the mass-specific DMo 2 of the lung of the Andean goose of 1.479 mlO 2 .sec -1 .mbar -1 .kg is only surpassed by that of the violet-eared hummingbird [57] while the mass-specific DLo 2 (0.119 mlO 2 .sec -1 .mbar -1 .kg) is only surpassed by that of the lung of the African rock martin [33,36,37,39,54] (Table 2). Because the DLo 2 integrates the diffusing capacities, i.e., the conductances, of the structural components that form  [54][55][56] and are given in data supporting this paper (S8 Table). https://doi.org/10.1371/journal.pone.0174395.g006 Lung of the Andean goose the air-hemoglobin pathway [42][43][44][45], structurally, the parameter it is the most comprehensive estimator of the gas exchange capacity of a lung. The large mass-specific DLo 2 of the lung of the Andean goose is consistent with and undoubtedly contributes to the high pulmonary O 2 extraction rates reported for this species by Lague et al. [28].
Compared with data on lungs of species of bird that have so far been investigated, the Andean goose has relatively larger lungs (Fig 4), larger surface area of the BGB (Fig 5), larger V (PCB) (Fig 6), greater surface density of the BGB (Fig 7) and the BGB is moderately thin (Fig  8). These specializations have generated relatively larger DLo 2 of the lung of the Andean goose (Fig 9).
Comment on the modeling of the oxygen diffusing capacity of the avian lung and future research directions. Here, the revised model of Weibel et al. [45] was used to determine the DMo 2 and the DLo 2 of the lung of the Andean goose. To compare the value of the DLo 2 of the lung of the Andean goose with those of species of birds that had been calculated using the older model of Weibel [44], the values were adjusted by a factor of 31.6%. This was the ratio of  [54][55][56] and given in data supporting this paper (S8 Table). https://doi.org/10.1371/journal.pone.0174395.g007 Lung of the Andean goose the value of the DLo 2 of the Andean goose calculated by the revised model of Weibel et al. [45] to that calculated by the older model of Weibel [44]. Because the current study is the only one where the two models have been applied and compared on the lung of a species of bird, the adjustment factor should be considered tentative. Taking into account the large structural and functional differences between the avian (parabronchial) lung and the mammalian (bronchioalveolar) one, it was considered judicious to adjust the avian DLo 2 values that had been determined with the older model of Weibel [44] with an avian lung derived adjustment factor rather than by applying the adjustment factors that were derived from the mammalian lung in Weibel et al. [45]. The application of the two models on the avian lung should be assessed on lungs of a larger number of species birds.
Following this study, it will be of great interest to find out whether the pulmonary specializations that have been found in the lung of the Andean goose characterize the lungs of all high  [54][55][56] and given in data supporting this paper (S8 Table).
https://doi.org/10.1371/journal.pone.0174395.g008 altitude flying birds or different adaptive strategies exist in the lungs of the other birds that fly to high elevations.  [54][55][56] and given in data supporting this paper (S8 Table).

Supporting information
https://doi.org/10.1371/journal.pone.0174395.g009 Lung of the Andean goose S4 Table. Surface areas of the Blood-Gas Barrier (BGB), the total surface area of the airand the blood capillaries (AC+BC), Red Blood Cells (RBC) and the Capillary Endothelium (CE). (DOCX) S5 Table. Harmonic mean thicknesses (μm) of the blood-gas (tissue) barrier (τht) and the harmonic mean thickness of the total barrier, i.e., the distance between the respiratory surface and the erythrocyte membrane (τhb), the air-hemoglobin pathway. (DOCX) S6 Table. Mean pulmonary diffusing capacities, namely the diffusing capacity of the blood-gas tissue barrier (Dto 2 ), the membrane (Dmo 2 ), the erythrocytes (Deo 2 ) and the total morphometric diffusing capacity (DLo 2 ). (DOCX) S7 Table. Surface area of the blood-gas barrier per unit body mass, surface area of the blood-gas barrier per unit volume of the exchange tissue, pulmonary capillary blood volume per unit surface area of the blood-gas barrier, volume of the lung per unit body mass, diffusing capacity of the blood-gas barrier per unit body mass and total pulmonary diffusing capacity per unit body mass. (DOCX) S8 Table. Pulmonary morphometric parameters of the bird lung on which the graphs (regression lines) were plotted and the sources of data. (DOCX)