A century of exercise physiology: lung fluid balance during and following exercise

The story of interstitial fluid balance dates back to the pioneering work of Starling and Tubby (1894) and Starling (1896) concerning the clearance of fluid from serous cavities. At the time, transcapillary fluid exchanges were considered to depend upon the difference between capillary hydrostatic (P_C) and colloidosmotic (Π_C) pressures. Accordingly, the “Starling gradient” was given by:

Based on this concept, the idea was put forward that the longitudinal distribution of pressure in a capillary would cause fluid filtration at the arterial side and reabsorption at its venous end (Landis 1934). Shortly afterward, Danielli (1940) introduced the concept of permeability in relation to the development of edema.

Agostoni et al. (1957) found that the visceral pleura is capable of absorbing fluid against negative pressures down to ~ − 8 mm Hg in the open chest and assumed that this value reflected the imbalance between pulmonary capillary hydraulic and colloid osmotic pressure, a conclusion still based on pressure values existing in the capillary, thus Δ\(P = P_{{\text{c}}} - \Pi_{{\text{c}}}\).

The Starling model went to a profound revision by the work of Kedem and Katchalsky (1958) who considered the all set of variables existing in two compartments separated by a membrane; as an example the capillary and the interstitial compartment separated by the endothelium, or the interstitial and the alveolar compartment separated by the epithelium. The reformulation of the “Starling gradient” considered the hydraulic and the colloidosmotic pressures in the two compartments (identified as 1 and 2) and a factor σ, named reflection coefficient, expressing the permselectivity of the membrane to plasma proteins. The new formulation defines trans-membrane water flow across two adjacent compartments (J_v), by introducing also the filtration coefficient K_f = L_p∙A, being L_p the hydraulic conductance and A the capillary surface area available for fluid exchange:

The value of σ varies between 0 and 1 and is a function of the molecular sieve of membrane pores relative to the size of the proteins. From experiments on the whole lung, endothelial σ values were found of 0.65 and 0.5 for total proteins and albumin, respectively (Taylor and Parker 1985).

Remarkable contribution were provided to describe the morpho-functional features of the air-blood barrier, representing the interface for gas exchanges at capillary level (Weibel and Knight 1964; Weibel 1973). Further, indications concerning permeability came from the description of a “gel like, fuzzy substance, the glycocalyx, that may play an important role in the regulation of endothelial trans-vascular flows” (Ito 1969). For extended historical references, the reader is referred to Weinbaum et al. (2007) and to Jin et al. (2021).

Contrary to Landis hypothesis, fluid reabsorption at the venous end of lung capillaries could not be uphold; in fact, reabsorption was found to be a self-limited mechanism due to the increase of the colloidosmotic pressure at the level of the glycocalyx generated by protein accumulation on the interstitial side, thus preventing further absorption (Michel and Phillips 1987). Around the same years, the role of lymphatics was being considered as the main mechanism to balance the microvascular filtration (Unruh et al. 1984; Mitzner and Sylvester 1986). Further, based on the experimental data, a model of pleural fluid turnover was developed based on Starling dependent fluid filtration and a tight control operated by pleural lymphatics, so as to maintain the pressure of the liquid in the sub-atmospheric range, thus close to a minimum volume of pleural fluid (Miserocchi et al. 1993b). The key point was to demonstrate that lymphatics are able to generate an absorption pressure in the sub-atmospheric range (Miserocchi et al. 1989).

Guyton’s group estimated the pulmonary interstitial pressure via perforated capsules implanted in a lung lobe and kept in place up to 8 weeks (Meyer et al. 1968); as the authors recognized, data were affected by “operative trauma of the lungs and infection and inflammation in and around the implanted capsules”. Despite these limitations, values were recorded ranging from sub-atmospheric up to positive immediately after the surgery, the latter values being attributed by the authors to developing lung edema. Accordingly, the clear message from these experiments was that in the intact lung interstitial pressure was sub-atmospheric, while persisting inflammation would result in positive pressures.

Interest was being focused on the macromolecular assembly of the interstitial compartment, in particular concerning the non-fibrillar component represented by molecules from the proteoglycans family (Fessler 1957). For a complete historical review refer to Comper and Laurent (1978).

The role of proteoglycans has been largely extended as these molecules play a pivotal role in control of fluid exchanges, tissue development and repair by interacting with inflammatory cells, proteases and growth factors (Hardingham and Fosang 1992; Roberts et al. 1997; Chambers and Laurent 1996; Dunsmore and Rannels 1996). Proteoglycans fill the space left free by the fibrillar component (collagen and elastic fibers) and act as link proteins through low energy non-covalent bonds. Microvascular permeability is controlled by the low molecular weight heparan-sulphate proteoglycan (HS-PG, 300–500 kDa) coating the basement membrane and by small peptidoglycans of the glycocalyx normally assuring high resistivity to water flow down the paracellular route (Reitsma et al. 2007). Large molecular weight chondroitin-sulphate proteoglycans (CS-PG > 1000 kDa) bound to hyaluronan provide rigidity to the interstitial compartment.

The Multiple Inert Gas Exchange Technique (MIGET) (Yokoyama and Farhi 1967) allows measuring the pulmonary exchange of a set of six different inert gases infused intravenously. Data are used to estimate the distribution of the ventilation/perfusion ratio (\({\dot{V}}_{\mathrm{A}}/\dot{Q}\)), the O₂ alveolar-capillary gradient (ΔA − a) and the diffusion limitation. The mechanics leading to the enlargement of the alveolar-arterial oxygen pressure gradient have been extensively described by Ferretti et al (2017). Up to 1989, accumulating evidence for exercise induced \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) mismatch was found in the horse and in humans, the authors stated that mechanism as well as its clinical significance remained unknown. Further, no reference to lung edema, as a possible cause, was postulated (Wagner et al. 1989). Reference to interstitial lung edema as a potential cause of \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) mismatch during maximal exercise was made by Schaffartzik et al. (1992). Further, Burnham et al. (2009) state that spatial perfusion and \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) heterogeneity in heavy exercise in humans are “consistent with but not proof of interstitial pulmonary edema”.

The pioneering work of Meyer et al. (1968) stimulated further research, in particular because the key variable to estimate the transendothelial and transepithelial Starling gradients was the value of the interstitial hydraulic pressure (P_i). The obvious experimental difficulty was represented by the subtlety of the air-blood barrier, as shown in Fig. 1A. It was considered that a good compromise would be to measure the pressure in the peri-microvascular interstitial space, using micropipettes with a tip bevelled down to 2–3 µm (Fig. 1B) (Miserocchi et al. 1990). The important point of the experimental approach was that to maintain the physiological condition of the lung expanded in the chest thus exposed to a transpulmonary pressure with zero alveolar pressure (the physiological condition indeed). Micropipettes, advanced through the pleura provided a sub-atmospheric pressure of ~ – 10 cmH₂O at the end-expiratory volume (Miserocchi et al. 1990) (Supplementary Conceptual diagram 1).

Figure 2 is a schematic simplified picture of the macromolecular assembly filling the interstitial space: one can identify the fibrillary collagen I, the laminar collagen IV coating the capillary surface and the molecules of perlecan (heparan sulfate proteoglycan) filling the voids of the collagen IV, then hyaluronan, a long molecule snaking through the interstitium bound at several points with versican (chondroitin sulfate proteoglycan). Versican has a core protein to which many glycans are attached. The hyaluronan-versican complex is highly hydrophilic and can easily bind free water to form gel.

Knowledge of the colloidosmotic pressure in the blood and in the interstitial space (Negrini et al. 2001), on the assumption of a reasonable value for σ, allowed an estimate of the trans-endothelial and the trans-epithelial Starling gradients. These data are reported in Table 1 for end-expiration and end-inspiration (30 and 70% of total lung capacity): note that trans-capillary Starling gradient is always positive (favoring capillary filtration), while trans-epithelial is always negative (favoring alveolar reabsorption).

At interstitial level, the fluid dynamic equilibrium is set at a subatmospheric interstitial pressure resulting from the balance between capillary filtration down a high resistance pathway (low endothelial permeability) and a low resistance draining pathway represented by lymphatic drainage (Fig. 2). So, the model being proposed for the control of the extravascular water in the lung is the same as the one for the pleural space, both compartments being kept at subatmospheric pressure with a minimal water volume (Miserocchi et al. 1990; Miserocchi et al. 2001b; Miserocchi 2009).

Concerning the microvascular district, a functional compartmentation exists between an “extra-alveolar” compartment and a peripheral “true alveolar” compartment, that are placed in series (Parker 2007). The former has a lower surface area but higher L_p, it allows fluid accumulation in the prerivascular and peribronchial cuffs in edemagenic conditions and is served by an extended lymphatic network. The “true alveolar” compartment is designed to favour gas diffusion: besides the thinness of the air-blood barrier (see Fig. 1A), the L_p is very low and σ is as high as 0.97; these features minimize fluid exchange. The low permeability coefficients are considered as “safety factors” against development of edema. Despite an incredibly very extended surface, the “true alveolar” compartment is not served by an extended lymphatic network.

One shall now consider a further “safety factor” residing in the mechanical response of the lung interstitial matrix in developing edema. The micropuncture technique allowed to monitor the change in interstitial pressure in edemagenic conditions; three models were developed: saline loading, hypoxia exposure and a lesional model induced by injection of proteases (Miserocchi et al. 1993a, 2001a; Negrini et al. 1996).

Figure 3 shows that pulmonary interstitial pressure (P_i) increases remarkably on increasing the wet weight to dry weight ratio (W/D) of the lung by about 10% in edemagenic conditions (point B), relative to control (point A). W/D was considered a precise and reliable index of the lung fluid balance, as shared also by Parker and Townsley (2004). What happens is that increased free water filtering in the interstitium is captured by the hyaluronan-versican complex to form gel, whose increase in steric hindrance causes a remarkable increase in P_i (from − 10 cmH₂O up to ~  + 5 cmH₂O (Miserocchi et al. 1993a; Miserocchi et al. 2001a; Negrini et al. 1996). The increase in P_i compared to the small increase in W/D ratio reflects the low tissue compliance (~ 0.5 ml·mmHg⁻¹·100 g of wet weight⁻¹) (Miserocchi et al. 1993a). Therefore, the increase in P_i nullifies the Starling filtration gradient: so a further “safety factor” is operating opposing the progression of edema (Supplementary Conceptual diagram 2). Shifting towards point C is discussed in the next paragraph.

A sustained edemagenic condition is a cause of progressive loss of integrity of the matrix (Fig. 3C). This was demonstrated by the increase in proteoglycans fragments recovered from the lung tissue using a weak extraction agent; this technique allowed to prove the weakening of the non-covalent intermolecular bonds of degraded proteoglycans with the other components of the matrix.

Figure 4A shows, as an example, the time course of the recovery of fragments of proteoglycan families on 12% O₂ exposure (Miserocchi et al. 2001a). Figure 4B shows the negative correlation between the loss of integrity of hyaluronan and W/D ratio.

Differences were found in the sequence of matrix fragmentation among various models of experimental pulmonary edema: saline loading (Negrini et al. 1996), exposure to hypoxia (Miserocchi et al. 2001a) and a lesional model injecting proteases (Negrini et al. 1998; Passi et al. 1999). It was estimated that the decline in the force of the non-covalent bonds occurs when the inter-fibrillar distance exceeds 30 nm (Conforti et al. 2002).

The critical phase of aggravation of edema pivots around a W/D of ~ 6 to 6.5. It was hypothesized that for W/D exceeding 6, capillary filtration occurs by the concomitance of three factors (Supplementary Conceptual diagram 3):

It can be promptly understood that an increase in Kf (filtration coefficient) occurs through a multiplicative effect due to the increase of both L_p and surface area for filtration (see Eq. 2): these events occurring at the level of the “true alveolar” compartment lead to the accelerated phase of lung edema.

A recent study reported an increase in plasma HS-PG in subjects affected by high altitude lung edema after climbing up to 4558 m (Swenson et al. 2020). Although it was impossible to specify where HS-PG came from, it could not be excluded that part would come from partial degradation of the pulmonary matrix. This finding would confirm the one reported twenty years ago (Miserocchi et al. 2001a).

The micropuncture technique also allowed to describe the lung microvascular pressure profile in physiological conditions (normoxia) (Negrini et al. 1992) and during developing lung edema (Negrini et al. 2001).

The dashed line in Fig. 5 shows the drop in pressure occurring at the level of the capillaries reflecting the arteriolar precapillary resistance in control conditions (normoxia). The continuous line (closed symbols) shows how the microvascular pressure profile was modified in the edemagenic model (saline infusion) (Negrini 1995). It is clear that a further increase in precapillary resistance allows maintaining capillary pressure within the physiological value. The small increase in left atrial and post capillary pressure is compatible with a greater recruitment of the capillary network due to the increase in the circulating fluid mass.

One can now consider the role of precapillary arteriovenous shunts that are located in the “extra-alveolar” compartment. Their activation in edemagenic condition during exercise (Hopkins et al. 2009; Lovering et al. 2008, 2009; Stickland et al. 2019) would avoid an increase in alveolar capillary pressure, a further factor protecting against the development of alveolar edema. There is also evidence from direct observation of in situ lung surface with intact pleural space, that, during development of edema, vasoconstriction occurred for arterioles ~ 80 µm in diameter (Negrini et al. 2001).This vascular adjustment represents further “safety factors” protecting against development of edema.

Interestingly, the presence of peribronchial and perivascular edemagenic cuffs occurring in the “extra-alveolar” compartment have little effect on gas exchanges, while only alveolar flooding (thus the involvement of the “true alveolar” compartment) leads to a rapid drop in arterial PO₂ (Taylor and Parker 1985).

Edema development in the lung is patchy although favoured by gravity; noteworthy, redirection of blood flow from edematous to non edematous regions has been demonstrated (Rivolta et al. 2011). Blood flow switching among capillaries has been described as a passive mechanism reflecting the branching geometry, the transcapillary pressure gradient as well as the longitudinal profile of capillary pressure (Mazzuca et al. 2019). As an example, Fig. 6 shows the capillary perfusion pattern with a color-coded log-scale in alveolar districts after hypoxia exposure (Mazzuca et al. 2019). Progressive closure of microvessels (blue colour) was observed over time in alveolar regions where interstitial edema was more marked, as judged from the thickness of the perivascular interstitial space. Accordingly, besides precapillary vasoconstriction, capillary closure could result due to the compressive action of the increase in interstitial pressure in developing edema (Mazzuca et al. 2019). De-recruitment of alveolar capillaries may also occur on increasing lung volume simply due to increase in tensile parenchymal stretching, a phenomenon may that may be favoured by peri-microvascular edema (see  paragraph Alveolar unfolding-folding).

Microvascular filtration flow could also be modelled on changing blood flow by entering the adequate parameters (Mazzuca et al. 2016): giving the heterogeneous distribution of blood flow, it was found that the filtration flow was also heterogeneously distributed. Given the heterogeneities in capillary flow distribution, it appeared tempting to consider that blood flow diversion from regions being less resistant to edema to more resistant ones represents an intrinsic mechanism to control lung fluid balance and preserve the efficiency of the diffusion/perfusion function (Rivolta et al. 2011; Mazzuca et al. 2016, 2019).

Derecruitment due to perivascular edema may well occur also in humans, as administration of a vasodilator agent cannot restore blood flow in edematous lung regions (Scherrer et al. 1996).

An interesting point from the work of Mazzuca et al. (2019) is that lower fluid extravasation was observed around smaller alveoli. The question remains on how alveolar morphology might influence the stability of extravascular fluid control. Fractal geometry has been invoked to describe blood flow and ventilation in terminal units (Glenny 2011): the model considered that small asymmetries along the branching system might account for heterogeneity in both airways and blood flow distribution.

A comprehensive and integrated model has been developed to account for the different contribution of the various factors appearing in Eq. 2 to the developing phase of edema (Mazzuca et al. 2016). Based on the results from the different experimental models of edema, it appears that several factors might contribute to the kinetics of edema formation, although to a different extent. Yet, the kern of edema formation is the loss of the “safety factors” that assure in physiological conditions the mechanical resistance to fluid extravasation (low microvascular permeability) and tissue resistance to mechanical stress (low tissue compliance). The model suggests that increasing pulmonary capillary pressure (from 9 to 25 cmH₂O) is, by and large, more edemagenic compared to an eightfold increase in water permeability (L_p): the difference can be explained considering that the increase in capillary pressure contributes to capillary recruitment, so that a greater Starling gradient is acting on a large surface area. The same model showed a remarkable acceleration in edema formation if the increase in capillary pressure was coupled with the increase in L_p, the specific case of exercise in hypoxia.

Exercise implies a further edemagenic factor relating with hyperventilation, namely the cyclic alveolar “unfolding-folding” process. To explain this mechanism one shall recall the specific topology of Epithelial cells Type 1 (Epi1) (Weibel 2015). Any Epi1 cell covers about 4 endothelial cells and present numerous pleats on its surface that are subjected to the cyclic “unfolding–folding” process on changing lung volume (Bachofen et al. 1987). Based on data from electron microscopy, unfolding was found to be completed in healthy lung on reaching ~ 65% of vital capacity, corresponding to lung distending pressure ~15cm H₂O.

Figure 8 describes the unfolding–folding process in relation with lung volume, lung distending pressure and alveolar morphology. Septa are exposed to increase in parenchymal tension on increasing lung volume and clearly, as the unfolding process is being completed, the septa become exposed to increased mechanical tissue stretching (causing a decrease in lung compliance) that may result in lesion of the alveolar structure (Knudsen and Ochs 2018). Note also, recalling data from Table 1, that at end-inspiration P_i, due to parenchymal stretching, becomes rather sub-atmospheric (Miserocchi et al. 1991), thus potentially increasing the endothelial filtration gradient.

Other data confirm the edemagenic nature of increasing lung volume: data by Egan (1982) suggest that acute lung distension may cause epithelial albumin leak; further, increasing transpulmonary pressure to about 20 cmH₂O over 6 h was found to cause a fivefold increase in capillary filtration (Tarbell et al. 1999; Tarbell 2010) and moreover W/D was found to increase well above 6 for Palv > 20 cmH₂O kept for 2 h (Yoshikawa et al. 2004). The unfolding–folding process has been confirmed also for the human lung for a similar range of lung distending pressure (Miserocchi et al. 2008; Beretta et al. 2021).

This is a pure case of stress failure of pulmonary capillaries leading to edema rapidly developing during immersion time (on the average 2–3 min). Convincing modelling strongly based on respiratory mechanics showed that transcapillary pressure gradient is remarkably increased at depth due to the large drop in alveolar air pressure, resulting from the outward elastic recoil of the chest wall undergoing compression by the increase in hydrostatic pressure at depth (Fitz-Clarke 2007a, b). An estimate of 50 mmHg of transcapillary pressure was obtained for record of immersion at 183 m. Obviously, capillary stress failure entails haemorrhage and protein leak. Diving also implies a further edemagenic factor given by the remarkable blood pooling in the lung (Arborelius et al. 1972).

This is a case of acute lung edema developing for all-out exercise lasting about 2 min, the average racing time for a horse. Lung haemorrhage has been widely documented in competing thoroughbred horses (Whitwell and Greet 1984). Increasing power output requires an increase in oxygen delivery that is accomplished by an increase in cardiac output coupled with the increase in pulmonary oxygen uptake. The latter is favoured by lowering the precapillary arteriolar tone allowing capillary recruitment. In a horse race, 100% \(\dot{V}{\text{O}}_{2\max }\) (about 250 ml/Kg/min) is attained within 15–20 s from the start through an increase in cardiac output up to about 600 ml/kg/min, and in ventilation in the range of 1500L/min; the maximal speed attained in the last 400 m is in the range of 17 m/s. At a speed of 13 m/s, pulmonary artery and capillary pressure were found to range about 90 mmHg and 75 mmHg, respectively (Manohar 1993). The high value of capillary pressure stems for a relatively lower upstream resistance allowing indeed extensive capillary recruitment. The capillary recruitment favours gas diffusion and avoids excessive shortening of the blood capillary transit time; however, the price to be paid is an increase in capillary filtration down an increased pressure gradient and increased capillary surface area. Accordingly, the likely pathophysiological interpretation on the origin of acute capillary stress failure in racing horses may well relate with the remarkable increase in capillary pressure (mechanical factor), increase in surface area and shear (permeability factors); the latter represent the adaptive response to reach \(\dot{V}{\text{O}}_{2\max }\) in a very short time (Manohar 1994). Further the unfolding-folding might also play a substantial role given the incredibly high ventilation. The possible co-existence of an increase in L_p relating with developing hypoxemia could be considered (Wagner et al. 1989), although it is short lasting.

Maximum power output in humans at \(\dot{V}{\text{O}}_{2\max }\) (such as in middle distance running, 800–1500 m) have not been reported to cause edema (Zavorsky 2007). It remains difficult to scale power output at \(\dot{V}{\text{O}}_{2\max }\) from horse to humans, noting however that, unlike humans, horses are vigorously whipped!

One shall now consider exercise at high altitude that is known to potentially cause lung edema (Swenson and Bärtsch 2012). Unlike the case of thoroughbred horses, high altitude pulmonary edema (HAPE) it is not an acute phenomenon as it develops over 2–5 days; accordingly, over this time the process of matrix degradation triggered by hypoxia should have advanced leading to a progressive increase in microvascular permeability (due to fragmentation of HS-PG), as well as weakening of the mechanical resistance of the capillary wall and surrounding macromolecular structure (due to fragmentation of CS-PG) (see paragraph Lung cellular signalling of developing edema and Fig. 6).

Pulmonary artery pressure (PAP) values have been reported in the range 40–50 mmHg with corresponding capillary pressure (from wedge pressure) of 20–25 mmHg (Maggiorini et al. 2001). Hultgren et al. (1964) reported an increase in PAP reaching about 60 mmHg in subjects developing lung edema at an altitude of 3750 m but wedge pressure never exceeded 10 mmHg, thus revealing a remarkable increase in precapillary pulmonary vascular resistance. In 1987 an exhaustive paper was published describing the functional adaptation of the pulmonary circulation in subjects maintained in a hypobaric chamber for 40 days where barometric pressure was progressively decreased down to that corresponding to the summit of Mt. Everest (240 mmHg) (Groves et al. 1987). From sea level to an ambient pressure of 240 mmHg, PAP increased at rest from \(\sim\) 15 to \(\sim\) 34 mmHg, while at maximal exercise PAP attained \(\sim\) 34 and \(\sim\) 54 mmHg, respectively. The increase precapillary arteriolar resistance allowed to maintain pulmonary wedge pressure below 10 mmHg, both at rest and during exercise. Interestingly, 100% O₂ breathing lowered cardiac output and PAP, but did not change pulmonary vascular resistance, a finding suggesting a stable increase in pre-capillary flow resistance compatible with a mechanical constraint on capillary patency due to interstitial edema. Arteriolar vasoconstriction has been documented on hypoxia exposure in lowlanders (Dunham-Snary et al. 2017; Moudgil et al. 2005). It is of interest that chronic exposure to high altitude induces pulmonary vascular remodelling leading to fibrosis and development of sustained pulmonary hypertension, representing an increased load on the right ventricle. There is no reported evidence for development of lung edema in these subjects (Sydykov et al. 2021). One may comment that the excessive deposition of fibrotic tissue in the lung parenchyma as well as pulmonary artery fibrosis provides a form of maladaptive protection against development of lung edema, as it occurs at the expense of increasing right-ventricular afterload.

Stress failure of pulmonary capillaries has been invoked as the main cause of HAPE (West et al. 1991), by analogy with capillary stress failure in thoroughbred horses. However, concerning capillary resistance to wall mechanical stress in experiments similar to those of West, different results were obtained (Bachofen et al. 1993; Wu et al. 1995): endothelial lesions were rare in high pressure experiments (about 30 mmHg), while there were distinct and numerous lesions of the epithelial Epi1 cells. As already noted, Epi1 cells are exposed to the risk of lesion by the unfolding-folding mechanism characteristically accompanying the hyperventilation. One may add that when Epi1 cells are lesioned, they no longer cover the underlying network of intercellular junctions that represents a potential paracellular filtration route. This point has been so far overlooked.

In High Altitude Pulmonary Edema Sensitive (HAPE-S) subjects, a more marked vasoconstrictor response to hypoxia in the pulmonary circulation was also described in the systemic circulation, a finding attributed to impaired vascular endothelial function due decreased bioavailability of NO (Berger et al. 2005). Yet, a decrease in exhaled NO was also found in HAPE-S subjects on exposure to normobaric hypoxia (Busch et al. 2001) as well as in patients with HAPE (Duplain et al. 2000). On a causative basis, it remains to be established whether the low bioavailability of NO reflects an impairment of the biochemical pathway or, conversely, represents the functional response to counteract edema formation. Based on physio-pathological evidence, pulmonary hypertension developing during exercise may well be regarded not as the cause of a perturbation in lung fluid balance, but rather as the consequence of a protective functional response to limit edema formation (see PROPOSAL 1).

Concerning the hypothesis of pulmonary hypertension caused by “excessive rise in pulmonary vascular resistance…… leading to increased microvascular pressures” (Swenson and Bärtsch 2012), it remains counter-intuitive how a rise in pulmonary vascular resistance could lead to downstream increased microvascular pressure. So, the above hypothesis still requires the validation by a fluid dynamic model. The likely most important factors causing edema during work in exercise are the increase in microvascular permeability (increase in L_p and decrease in σ) and the unfolding-folding mechanism. As high altitude edema occurs preferentially in some individuals, one could invoke an inborn capillary fragility possibly reflecting higher microvascular permeability (see paragraph The kinetics of alveolar-capillary oxygen equilibration). It is therefore doubtful that stress failure does occur as initial acute phenomenon in high altitude edema. This point of view is shared by Hopkins who considers that “capillary stress failure is the extreme point on the continuum of the lung’s response to stress and is not required for the development of pulmonary edema” (Hopkins 2010b). Indeed, there may be a variable time dependent contribution to edema formation due the concomitance of changes in pressure/surface of capillaries, increase in microvascular permeability and/or increase in tissue compliance (Miserocchi et al. 2001b). It is important to note here that considering a sharp distinction between permeability and hydraulic edema represents a widely used but simplistic and possibly erroneous concept. In fact, as already noted in paragraph Lung cellular signalling of developing edema, the progressive increase in microvascular permeability may well occur concurrently with changes in capillary pressure/surface, although relative intensity of effects and timing may differ. This clarification might contribute to provide an answer to a puzzling point made by Swenson et al. 2002 who maintained that “high permeability edema is not readily explained by classical hydrostatic (Starling forces), which should generate a protein poor alveolar edema”: the comment is that the case hypothesized is a non-existing one; further, since the definition of Starling forces 64 years ago (Kedem and Katchalsky 1958), there is no conundrum in accepting that any combination of changes of the terms appearing in Eq. 2 may justify the development of lung edema.

An experimental study in rats exposed to simulated altitude of 4700 m, forced by electrical stimulation to walk for 48 h (with 15–20 min breaks every 4 h), at a speed of 12 m/min gave the following results: 37% died before; in survivors, the lungs were highly compromised (Bai et al. 2010) showing complete rupture of the blood-gas barrier including both the endothelium and the epithelium. The authors did not find signs of inflammation and propose that stress failure plays a major role in the findings. One can share the concept of stress failure, however it should be interpreted as the consequence of a “lung’s response to stress” in terms of overuse: this might well include a persistent increase in capillary pressure and recruitment as well as the infinite sequence of unfolding-folding events during hyperventilation. Such a case could well occur in other long/ultra-long performances. Interstitial lung edema may be induced by marathon running (Zavorsky et al. 2014). There is a report of pulmonary edema after the Bicycle Race Across America (4800 km), a relatively low-intensity effort at about 50% of \(\dot{V}{\text{O}}_{2\max }\) (Luks et al. 2007), while good tolerance was reported for another participant (Schumacher et al. 2011). Further there is a report for two cases of pulmonary edema in a ultra-marathon of 90 km (McKechnie et al. 1979).

Most of the evidence for developing some degree of lung edema during exercise was based on the data coming from the MIGET technique based on the dispersion of the \(\dot{V}_{{\text{A}}} /\dot{Q}\) distribution. MIGET actually cannot determine the morpho-functional or mechanical causes for the impairment in gas exchange. The hypothesis is that fluid accumulation around the microvessels, in the small airways as well as in the alveoli, is the likely cause substantially contributing to \(\dot{V}_{{\text{A}}} /\dot{Q}\) heterogeneity and diffusion limitation during exercise. Regional variability in the development of edema have been provided in experimental models; it appears therefore conceivable that the progressive dispersion of \(\dot{V}_{{\text{A}}} /\dot{Q}\) over time might include alveolar flooding (causing a decrease in \(\dot{V}_{{\text{A}}} /\dot{Q}\)) and capillary blood flow limitation (causing an increase in \(\dot{V}_{A} /\dot{Q})\) (Supplementary Conceptual diagram 2 and 3). Non-ventilated and non-perfused alveoli contribute to diffusion limitation. Hammond et al. (1986b) hypothesized that development of mild interstitial edema could contribute to the \(\dot{V}_{{\text{A}}} /\dot{Q}\) dispersion during work in normoxia and hypoxia. Reference to interstitial lung edema as a potential cause of \(\dot{V}_{{\text{A}}} /\dot{Q}\) mismatch was made by Schaffartzik et al. (1992) during maximal exercise. Log standard deviation of perfusion and ventilation distributions (Log SDQ, and log SDv, respectively) were correlated with the increase in ΔA-aO₂ and diffusion limitation (Hopkins et al. 1998a; Hopkins 2005, 2020). More recent data confirm that 45 min at about 80% \(\dot{V}{\text{O}}_{2\max }\) increased \(\dot{V}_{{\text{A}}} /\dot{Q}\) mismatch in the basal lung after exercise, likely reflecting gravitationally dependent interstitial pulmonary edema (Tedjasaputra et al. 2019).

In humans it was found that exercise induces remarkable redistribution of pulmonary blood flow, in particular in the apical regions (Harf et al. 1978) possibly reflecting re-direction of blood flow away from basal regions, that are more exposed to development of edema. Heterogeneity in pulmonary perfusion in sustained heavy exercise in humans was confirmed (Burnham et al. 2009). Techniques available for measuring perfusion are dealt with in the comprehensive review by Hopkins et al. (2012).

In the experimental model, the hypothesis was put forward to assess whether the mechanical properties of lung would be significantly affected when the lung tissue matrix is put under tension in interstitial edema (point B in Fig. 3). The problem was afforded by adopting the non-invasive “frequency oscillation technique” to estimate the changes in the mechanical properties of the respiratory system. This technique allows to derive the impedance of the respiratory system by oscillating it at different frequencies. Based on oscillation frequency imposed at the mouth, one can derive indications of the mechanical behaviour of the lung without the cooperation of the subject (Goldman et al. 2005). The system provides in particular: (a) the resistance at 5 Hz (R5), (b) the resistance at 20 Hz (R20), (c) the difference R5−R20, expression of the frequency dependence of resistance, (d) the reactance at 5 Hz (X5) that reflects the visco-elastic properties of the distal lung portion. A whole oscillation cycle lasts about 30 s and the manoeuver could be repeatedly performed during the experiment. Results indicated that the progressive changes in the impedenzometric indexes occurring during developing interstitial edema likely reflected the changes in the visco-elastic properties of the lung tissue in relation with the specific mechanical condition relating with the increase in extravascular water not exceeding 10% (Dellacà et al. 2008). Extending the use of the impedenzometric technique from the experimental animal to humans, one could actually confirm a greater lung water perturbation in some subjects exercising at sea level and at altitude. The result was interpreted as corresponding to a greater individual proneness to develop lung edema (Bartesaghi et al. 2014). See PROPOSAL 2.

Lung ultrasound (LUS) is a potential semiquantitative operator dependent tool for detecting lung edema, possibly discriminating between interstitial edema and alveolar edema (Picano and Pellikka 2016). The no-echo signal (black lung in physiological conditions) shifts towards a black and white pattern on increasing extravascular water by the appearance of peculiar echo fingerprints, named B-lines. LUS was used to identify signs of pulmonary edema in scuba divers (Castagna et al. 2018), as well as in swimmers over a distance between 1000 and 3000 m (Hårdstedt et al. 2020). Concerning HAPE, the work of Nowadly et al (2021) reported no LUS signs in 15 climbers during the ascent to the Mount Everest base camp (5.380 mt).

One can quote an interesting research development aiming to investigate temperature, relative humidity and water evaporation in exhaled air (Mansour et al. 2019). The results of the study define the ranges of temperature and relative humidity of exhaled breath and the correlation with clinical and environmental factors such as gender, BMI and age. These factors ought to be taken into consideration in order to increase the reproducibility and reliability of a wide variety of measurements. It was proposed that humidity in the exhaled air was significantly lower in rats developing acute pulmonary edema (induced by a-naphthylthiourea) compared to healthy rats. The result supports the hypothesis that humidity in expiratory air can be considered an important parameter to follow-up pulmonary edema in patients (Adar et al. 2022). See PROPOSAL 3.

A convincing demonstration that a perturbation induced by exercise takes some time to disappear is the persistence of \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) inequality with increasing \(\dot{V}{\mathrm{O}}_{2}\) and diffusion limitation at least transiently following exercise relative to rest with no change after giving 100% O₂ (Hammond et al. 1986a, Podolsky et al. 1996).

Sheel and McKenzie (2010) stated that “attempts to determine the presence of pulmonary edema in humans with modern, objective imaging techniques and other methods have failed to provide supportive evidence for its existence”.

Most of the controversies concerning the development of exercise induced edema reflect the sensitivity of the imaging methods available to detect an increase in extravascular water (Hopkins 2020).

We report a comparison of a new proton MRI technique to measure regional \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) ratio against the multiple inert gas elimination technique (MIGET). The study reports good relationships between measures of heterogeneity derived from MIGET and those derived from MRI. Although currently limited to a single slice acquisition, these data suggest that single sagittal slice measures of \({\dot{V}}_{\mathrm{A}}/\dot{Q}\) ratio provide an adequate means to assess heterogeneity in the normal lung (Sá et al. 2017). One can mention a recent study allowing to define the spatial/temporal dynamics of blood flow in the human lung by MRI in subjected developing HAPE, exhibiting known differences in vascular reactivity (Buxton et al. 2022).

Other MRI data reported increased ventilation-perfusion mismatch only in the basal lung of athletes following 45 min of cycling exercise, consistent with the development of lung edema in the gravitationally dependent lung during heavy exercise (Tedjasaputra et al. 2019).

Data from CT scan also reveal development of lung edema in highly trained athletes after performing a triathlon (Caillaud et al. 1995). A shareable point is made that a likely cause of edema would be an increase in capillary permeability that is independent of high vascular pressures (Lovering 2010).

The clearance rate of aerosolized 99mTc-labeled diethylenetriaminepentaacetic acid as an index for the permeability of the lung blood-gas barrier (Hanel et al. 2003) was found to increase after maximal workout in rowing, suggesting an increase in microvascular permeability.

So far, imaging data have not been correlated with the corresponding regional values of W/D ratios. It would be useful to define such correlation (see PROPOSAL 4) to detect the earliest stages of perturbation in lung fluid balance before the condition becomes life-threatening. In particular, it would be useful to identify intrapulmonary lung regions with a W/D ~ 6.5 (corresponding to a 35% increase in extravascular lung water) likely representing the critical threshold for developing severe edema (Beretta et al. 2021; Cressoni et al. 2013; Negrini et al. 2001; Taylor and Parker 1985).

Figure 9 shows that at W/D of ~ 6.5, the capacity of the interstitial compartment for fluid accumulation (Wint) is saturated while alveolar flooding (Walv) will continue to increase on increasing W/D (Beretta et al. 2021).

Most of the interest in estimating BAL contents pivots around finding markers of inflammation; their presence would explain the increase in permeability leading to various degree of leakyness of the capillary wall, justifying the finding of plasma proteins (in particular the largest one, albumin) and erythrocytes. One should mention that innate immunity is activated by sterile type of conditions, including tissue/vascular stretch and hypoxia that indeed represent inflammatory agents activating signalling molecules named HMGB1 (High mobility group box 1) (Andersson and Tracey 2011; Yang et al. 2020). So, one cannot exclude that this early detection system is activated during exercise and hypoxia exposure.

Potent mediators of inflammation, namely leukotriene B4 and C5a, were found in BAL on reaching 4400 m after some days of trekking together with high-molecular-weight proteins and erythrocytes. The correct biophysical interpretation was that high altitude edema involves a leak through "large pores" (a specific term modelling barrier permselectivity), thus a permeability type of edema reflecting an inflammatory process (Schoene et al. 1986). These results were confirmed by a following paper (Schoene et al. 1988). Increase of leukotriene B4, following maximal exercise lasting 6–8 min was also found in BAL 1 h after exercise. Further, in subjects with a history suggestive of lung edema, higher concentrations of red cells and proteins were found in BAL (Hopkins et al. 1998b).

A further report referring to cycle ergometer at 90% maximal effort in hypobaric hypoxia indicated a significantly greater leakage of red blood cells and protein into the alveolar space compared to normoxia, data extending up to 26 h after exercise. Results were interpreted as being due to progressive increase in microvascular permeability possibly evolving towards capillary stress failure (Eldridge et al. 2006).

All these data are obviously in line with the mRNA activation triggering the pre-inflammatory response to the initial phase of developing edema (Sabbadini et al. 2003).

One shall now discuss the data presented by Swenson et al. (2002) referring to post exercise ascent to Margherita Hut (4550 m). It was found that at altitude, HAPE resistant (HAPE-R) subjects had a small but significant increase in total protein and albumin in BAL (but not red blood cells), in agreement with data from Schoene et al. (1986, 1988). One can explain these findings by a progressive increase in microvascular permeability. Unfortunately, no data were provided as to the time course of the increase in albumin as well as erythrocytes in HAPE-S subjects. It is logic to consider that, as much as in the HAPE-R, the data would have shown a progressive increase, although likely reflecting a more tumultuous kinetics. The other factors reported in BAL cannot be considered as pre-inflammatory or inflammatory. Note that neutrophil accumulation in the BAL was already shown not to be characteristic in HAPE (Schoene et al. 1986). Further, uphold the stress-failure hypothesis by showing a positive correlation between the increase in pulmonary artery pressure and albumin concentration in BAL, cannot be taken as a cause-effect result, as it is simply a phenomenological process.

The conclusion for non-inflammatory process by Swenson et al. (2002) is questionable, as it neglects the progressive damage to the interstitial-capillary complex occurring on exposure to hypoxia during the time elapsing before overt edema is diagnosed. Noteworthy, the conclusion is at variance with data previously presented, providing the evidence of inflammatory response with increased permeability in HAPE (Bärtsch 1999), where Swenson was the last author.