The phylogenetic original function of surfactant in vertebrates can be deduced from studies of non-mammalian vertebrates such as fish, lungfish, amphibia, and reptiles (for reviews, see [18,21]). It has been proposed to be that of an 'anti-glue' to prevent adhesion of the surfaces of gas-containing organs, such as swim bladder and lungs, which might occur during collapse. There are some indications that surfactant acts as an anti-oedema factor in non-mammalian lungs, too [21]. In mammals its primary function is to regulate alveolar surface tension in relation to alveolar size, which is an important clue to efficient ventilation and alveolar stability (for reviews, see [19,24]). According to the equation of Young and Laplace, the actual surface tension is much lower in small alveoles than would be expected from pure geometry. Because neighbouring alveoles communicate with each other via alveolar ducts and pores of Kohn (Fig. 1a), their surface tensions must be different (if they are different in size) in order to prevent the collapse of small alveoles in favour of large ones. Mechanical coupling of alveoles via the interstitial tissue of the septum acts as an additional mechanism to prevent alveolar collapse [25]. However, absence or inactivation of surfactant alone results in alveolar collapse at end-expiration and in atelectasis [26].

Although regulation of surface tension can be considered as the primary function of pulmonary surfactant in mammals, this is only one of a number of different functions [24]. Some critical aspects of current points of view have recently been discussed in detail [19].

Surfactant has long been postulated to prevent the formation of alveolar oedema through the effect of surface tension acting as an additional force to direct net fluid flow across the air–blood barrier [1,27]. The maintenance of fluid homeostasis in the alveolus is considered to represent one of its phylogenetically ancient functions [18]. A comprehensive discussion of the mechanism of surface-tension-dependent alveolar fluid balance predicted by different surfactant models is given by Hills [19].

In order to be effective in keeping the alveolar space free of excess fluid, ions and serum proteins, the AE2 cell is equipped with a number of membrane-bound water channels and ion pumps as well as an albumin-binding immunoglobulin receptor (for review, see [28]; Supplementary Table 2). However, instead of removing fluid completely, a very thin aqueous film is preserved, termed the hypophase, covering the alveolar surface. The hypophase is delimited at the alveolar face by the surfactant lining layer and at the septal face by the alveolar epithelium. It was estimated to comprise ≅0.37 ± 0.15 ml/kg body weight in sheep [29]. The hypophase can be considered as a reaction milieu for extracellular biochemical processes as well as a 'medium' for intra-alveolar cells such as alveolar macrophages. AE2 cells are thought to control various properties of this extracellular aqueous milieu, for example pH [30] and [Ca²⁺] [31]. Since many biochemical processes, such as the extracellular transformation of surfactant (see below), depend on the actual pH and [Ca²⁺], regulation of these parameters is important for controlling what happens in the alveolus. Furthermore, within a certain distance, any factor secreted into this continuous film is likely to reach other cells within the alveolus.

Another function of alveolar surfactant postulated by Macklin [1], host defence, has attracted major scientific interest in recent years (for reviews, see [32,33]). This function of surfactant relies on the nature of SP-A and SP-D as collectins. Both proteins are able to bind to the surface of various pathogens, thus acting as opsonins to facilitate their elimination by alveolar macrophages [32,33,34]. Moreover, AE2 cells are able to secrete several other products that are involved in host defence, such as the bacteriolytic lysozyme [35,36]. In rat lungs, lysozyme was detected in lamellar bodies of AE2 cells [36], whereas in humans it was identified in serous submucosal glands but not in alveolar AE2 cells [35].

Although the bronchiolar Clara cells synthesise and release the mature proteins SP-A, SP-B, and SP-D (Fig. 2a) [37,38], the AE2 cell is the only type of pulmonary cell that produces all the surfactant components (phospholipids [Fig. 3] as well as all four surfactant proteins). The mature 3.5–3.7 kDa small SP-C (Fig. 2b) is thought to be released by AE2 cells only [39,40].

The lamellar bodies of AE2 cells have long been recognised as storage granules from which surfactant is released into the alveolus [41,42]. The biochemical composition of this intracellular storage form is largely identical to the composition of the extracellular material obtained by BAL [43]. The stored phospholipids are bound by a limiting membrane (Fig. 3), which is characterised both by typical lysosomal/endosomal [44] as well as by specific integral membrane proteins [45] probably involved in intra-cellular trafficking. The lamellar body membrane is further equipped with transport proteins for regulation of internal acidic pH and high [Ca²⁺] [46]. High levels of Ca²⁺ interspersed between the stacks of phospholipids were demonstrated by microanalytical techniques [31].

The pathway of lipid and protein synthesis has been traced by means of electron microscopic autoradiography [47] to involve the organelles of the classical pathway, ie rough endoplasmic reticulum, Golgi apparatus, multivesicular bodies, and lamellar bodies. Immunoelectron microscopy confirmed this pathway for SP-B and SP-C, and by means of double- and triple-labelling, the different steps of processing and maturation of SP-B and SP-C were localised to specific intracellular structures [40,48,49,50]. Although the synthetic pathway of both SP-A and SP-D also involves endoplasmic reticulum and Golgi apparatus, mature SP-D is barely detectable in lamellar bodies [38]. It is thought that SP-D is released via a constitutive pathway [34], and a subpopulation of lamellar bodies has been proposed to be involved in recycling of SP-D [51].

The problem of differentiating between newly synthesised and recycled proteins is reflected in the controversy of whether or not SP-A is present in lamellar bodies (for review, see [52]). Although SP-A was detected in lamellar bodies by immunoelectron microscopy [48] and lamellar bodies have been reported to be enriched in SP-A [53,54], other studies reported only a relatively low amount of SP-A [55,56,57]. These contradictory data may result from the fact that most of the SP-A released into the alveolar hypophase is taken up again by the AE2 cell (see also below). The captured SP-A is directed to the lamellar bodies [57,58], while newly synthesised SP-A is likely to follow a constitutive pathway of secretion [59]. Re-secretion of internalised SP-A may be very rapid, at least in vitro, and may be achieved via a different pathway than the one used by internalised lipids [60]. Little attention has been given to potential species-specific differences, which may be another source for controversial data.

Surfactant material is released from its intracellular stores by exocytosis upon various stimuli. A number of physiologic and pharmacologic agents act via β-adrenergic receptors (epinephrine, terbutaline, isoproterenol), P1-purinoreceptors (receptors of adenosine and its analogues) or P2-purinoreceptors (ATP, UTP, ATP analogues; Supplementary Table 1), while several membrane-permeable substances act intracellularly, such as cholera toxin, forskolin, phorbol esters, and calcium ionophores (for review, see [52]). A number of agents have been reported to stimulate surfactant secretion, such as arachidonic acid, prostaglandins, histamine, and endothelin-1 [52]. Ventilation of the alveolus is a major physiologic stimulus of surfactant secretion and a single deep breath is considered to be sufficient [61,62]. An elegant in vitro study indicated that direct mechanical stretching of AE2 cells can trigger the release of surfactant [63]. However, a recent real-time study examining exocytosis in situ by means of vital stains in isolated perfused rat lungs demonstrated that lung expansion induced synchronous intracellular [Ca²⁺]-oscillations in all alveolar cells and lamellar body exocytosis in AE2 cells, with the exocytosis rate correlating with the frequency of the oscillations [64]. The authors' exciting conclusion is that AE1 cells may act as mechanotransducers that translate the mechanical stimulus into an intracellular Ca²⁺ signal, which is transmitted via gap junctions to the AE2 cell to regulate surfactant secretion.

Three pathways of signal transduction are now known (for a comprehensive review, see [52]). The first acts through activation of adenylate cyclase, formation of cyclic AMP and activation of cAMP-dependent protein kinase A. This pathway is followed, for example, by agents binding to β-adrenergic receptors or adenosine receptors A_2b. The second pathway acts through activation of protein kinase C (PKC), either by direct interaction with permeable substances or indirectly as a consequence of the activation of membrane receptors. Direct activation of PKC can be achieved by 12-O-tetradecanoylphorbol-13-acetate (TPA) and membrane permeable diacylglycerols (DAGs), while ATP and UTP, for example, activate the PKC pathway after binding to purine receptor P2Y₂. The third known pathway acts via an increase in intracellular Ca²⁺ levels, through either the uptake of extracellular calcium (using ionophores, for example), the transmission of calcium through gap junctions from neighbouring AE1 cells, or the release of calcium from intracellular stores. All of these may activate the Ca²⁺-calmodulin dependent protein kinase. The release of calcium from intracellular stores, for example, can be induced by binding of ATP to purine receptor P2Y₂ and subsequent formation of inositol-3-phosphate.

Activation of one of these signal cascades results in an increase in surfactant secretion by about two- to threefold (adenylate cyclase, Ca²⁺-ionophores) or about fivefold (TPA, PKC-activating agonists). Simultaneous activation of several pathways using several agonists, by mastoparan or ATP, which may activate all three pathways, results in a 5- to 12-fold increase above basal secretion (see references in [52]). Thus, an enormous redundancy is achieved through the existence of these different pathways of signal transduction and the great number of agonists, which guarantees a high degree of safety in the regulation of surfactant release and underlines the great importance of surfactant delivery to the alveolus.

The final step of the secretory pathway is accomplished via the classic mechanism of secretion by exocytosis, which results in the release of surfactant material from lamellar bodies into the alveolus. While it is well established that cytoskeletal components, such as microtubules [65] and actin filaments [66], are necessary for transport of the granules to the cell membrane and release of their contents, nothing is known about the mechanisms of release of constitutively formed SP-A and SP-D. Fusion of the lamellar body limiting membrane with the AE2 cell plasma membrane is mediated by annexins [67]. Single cell monitoring may provide new insights into the details of how exocytosis is regulated [68]. Secreted surfactant lipids as well as SP-A may inhibit subsequent surfactant release by negative feedback mechanisms [69,70], although this has not yet been proven in vivo.

Once released into the alveolar aqueous hypophase, the lamellar body material transforms into tubular myelin. This is an amazingly regular phospholipid/SP-A assembly (Fig. 4), which gives rise to the surface-active lining layer from which, in turn, small vesicular forms derive that are thought to represent spent surfactant (for review, see [71]). These categories of surfactant subtypes were defined by early ultrastructural studies and were consistently seen in both chemically and cryofixed surfactant [72,73]. By differential centrifugation of BAL material, surfactant is separated into large and small aggregates, while equilibrium buoyant density gradient centrifugation separates light, heavy and ultraheavy fractions. Correlative studies showed that large aggregates and the ultraheavy fraction correspond to tubular myelin and freshly secreted lamellar bodies, while small aggregates and the light fraction largely represent vesicular surfactant forms [43,55,74]. However, neither do the individual subfrac-tions represent a single ultrastructural subtype [74,75] nor is there congruence of fractions obtained by differential centrifugation and equilibrium buoyant density gradient centrifugation [76].

Being an extracellular process, transformation or conversion of surfactant can be studied in vitro. Surfactant subtypes can be reconstituted from individual components [55,77], and surfactant conversion can be mimicked by surface area cycling [74,78,79]. Thus, surfactant transformation was demonstrated to depend on various characteristics of the hypophase milieu, such as concentration of electrolytes [80], in particular of Ca²⁺[81], pH [82], and the presence of surfactant proteins, especially of SP-A [83].

The first step of transformation of freshly secreted lamellar body material into tubular myelin requires an increased [Ca²⁺] (probably derived from lamellar bodies [31]) and SP-A [84] which is finally observed at the corners of tubular myelin lattices [36,85]. The presence of tubular myelin is thought to be associated with the ability of surfactant lipids to rapidly adsorb to the lining layer at the gas/liquid interface. This second step of conversion appears to be promoted by SP-B (for review, see [71]). Refinement of the lining layer is the next step that results in an increase in its dipalmitoylphosphatidylcholine fraction, thereby achieving minimal surface tension [86]. This process is thought to involve both SP-B [87] and SP-C (for review, see [71,86]). The final step of conversion, from surface-active surfactant into inactive vesicular forms, appears to depend on an AE2-cell-derived enzyme termed convertase [88,89].

The balance between large aggregates and small aggregates has turned out to be an important parameter in assessing the functional integrity of alveolar surfactant obtained by BAL (for review, see [90]). This is corroborated by quantitative ultrastructural studies. While normal lungs showed little quantitative variation in the relative amount of tubular myelin under different ventilation strategies [91], tubular myelin was considerably decreased in different lung injury models [17,92]. In the context of lung injury, the ultrastructural approach offers the unique opportunity to examine surfactant retained in situ [93], which allows for the analysis of local surfactant inhomogeneities in relation to other structural changes [17].

Absence of tubular myelin was associated with reduced intracellular labelling for SP-A and with severe respiratory dysfunction in neonatal respiratory distress syndrome [94]. Paradoxically, targeted SP-A deletion in mice had minor effects on pulmonary function despite a severe depletion of tubular myelin [95]. This discrepancy is still a matter of debate.

Today it is established that most of the secreted surfactant — estimated at about 85% [24] — is taken up again, metabolised and re-secreted by the AE2 cells. Re-uptake and recycling have been demonstrated for surfactant lipids [58] and all four surfactant proteins [51,58,96,97]. SP-A, SP-B, and SP-C have been reported to enhance the uptake of phospholipids by AE2 cells in vitro; in the case of SP-A at least, this may be a receptor-mediated process [98,99]. SP-D, however, appeared to be ineffective in enhancing lipid uptake [51]. The significance of lipid uptake enhanced by surfactant protein in vivo is still unclear. The intracellular processes of metabolism and recycling are essentially associated with multivesicular bodies, which may exist as functionally heterogeneous populations [58]. Electron microscopic autoradiography [58] and confocal fluorescence microscopy [60] indicated that internalised lipids and SP-A are rapidly re-secreted by AE2 cells, probably along different pathways.

The degradation of surfactant is accomplished by the alveolar macrophages with only minimal contribution, if any, from AE2 cells. Phospholipids and SP-A appear to be degraded along different pathways [100]. Failure of surfactant removal and degradation may be one reason for alveolar proteinosis observed in transgenic mice lacking granulocyte-macrophage colony-stimulating factor (GM-CSF) [101].