INTRODUCTION

Top Abstract Introduction Summary References

PULMONARY ENDOTHELIUM regulates the exchange of fluid, solutes, macromolecules, and cells between vascular and tissue spaces. With inflammation, the endothelial barrier becomes more permissive for exchange as the pathway for transport is increased. Studies using whole animal, isolated lung, and cultured endothelial cell models of inflammation suggest that neurohumoral mediators, oxidants, and leukocytes increase endothelial cell cytosolic Ca²⁺ concentration ([Ca²⁺]_i). This change in [Ca²⁺]_i is believed essential for the generation of endothelial cell paracellular gaps. Whereas elevated [Ca²⁺]_i increases endothelial barrier permeability, increased cAMP has the opposite effect. Increased cAMP prevents or reverses permeability-induced pulmonary edema in nearly every animal species and model studied. Thus permissiveness of the endothelial barrier for exchange may depend on the relative levels of [Ca²⁺]_i and cAMP at any given time.

Due to the significance of [Ca²⁺]_i and cAMP levels on both pulmonary and systemic endothelial cell barrier integrity, this review focuses on putative sites of cross talk between these intracellular signals and their relationship to pulmonary endothelial permeability. Individual areas of focus are addressed in turn, and a summary of recent observations regarding [Ca²⁺]_i and cAMP permeability regulation for pulmonary arterial endothelial cells (PAECs) vs. pulmonary microvascular endothelial cells (PMVECs) is given. To focus the review, however, recent emerging data linking these signal transduction pathways to the mechanisms of paracellular gap formation, such as activation of myosin light chain kinase, focal adhesion kinase phosphorylation, and decreased cell-cell and cell-matrix tethering, are not addressed. For recent reviews on general Ca²⁺ signaling, the cAMP pathway, or other signal transduction pathways having an influence on endothelial permeability, references are provided throughout the text.

	CYTOSOLIC FREE CALCIUM

Many endothelial cell first messengers activate specific receptors and increase [Ca²⁺]_i. The [Ca²⁺]_i response is characterized by two distinct phases, including a transient rise corresponding to the release of Ca²⁺ from intracellular stores and a more sustained increase due to entry of Ca²⁺ across the plasmalemma. Each phase, Ca²⁺ release and entry, can regulate discrete cellular functions. As an example, activation of endothelial cell phospholipase A₂ depends on Ca²⁺ release, whereas activation of nitric oxide synthase and inhibition of adenylyl cyclase both require Ca²⁺ entry. In addition, [Ca²⁺]_i is also determined by the activity of Ca²⁺ reuptake and extrusion enzymes. Altering the activities of these enzymes may result in prolonged or abridged Ca²⁺-mediated changes in cell function.

Intracellular Ca²⁺ Release

Inositol 1,4,5-trisphosphate, [Ca²⁺]_i, and cAMP. Endothelial cell ligands, coupled through their receptors to the activation of phospholipase C, generate inositol 1,4,5-trisphosphate [Ins(1,4,5)P₃] and diacylglycerol as breakdown products of the membrane phospholipid phosphatidylinositol 4,5-bisphosphate. Ins(1,4,5)P₃ is a highly hydrophilic molecule that faces little molecular or steric hinderance to diffusion in the cytosol (5). Limited hinderance to diffusion makes Ins(1,4,5)P₃ an excellent second messenger because it has access to cytosolic compartments. Both nuclear and smooth endoplasmic reticulum (sER) membranes possess Ins(1,4,5)P₃ receptors (60, 106-109, 113, 137, 166, 190, 194) that, when bound to Ins(1,4,5)P₃, form an ion channel possessing high Ca²⁺ conductance (in order of divalent cations, Ba²⁺ > Sr²⁺ > Ca²⁺ > Mg²⁺). The Ins(1,4,5)P₃ receptor likely exists as a tetramer that forms a Ca²⁺ channel on activation (28, 106, 107, 117, 118). The receptor protein possesses six-membrane-spanning domains so that both amino and carboxy termini reside in the cytosol. These physical properties, four subunits and six-membrane-spanning domains, support placement of the Ins(1,4,5)P₃ receptor in the superfamily of voltage-gated and second messenger-gated channels. Recent evidence suggests that at least three distinct isoforms of the Ins(1,4,5)P₃ receptor exist plus additional products that represent splice variants (18, 111, 151, 173, 201). It is presently unclear how the different isoforms and splice variants translate into different functions. Although highly conserved among species, expression of Ins(1,4,5)P₃-receptor isoforms within a species is heterogeneous among organs. Expression of Ins(1,4,5)P₃-receptor isoforms in endothelial cells is poorly characterized.

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Effects of Ca2+ concentration (pCa) on open probability of inositol 1,4,5-trisphosphate [Ins(1,4,5)P3] receptor. Values were normalized to maximum channel activity observed in each individual experiment. At cytosolic Ca2+ concentration levels of ~750 nM, a negative feedback effect on Ins(1,4,5)P3 channel is observed. [Modified from Bezprozvanny and Ehrlich (14).]

1

1

1

1

Ca²⁺ Entry

Unstimulated endothelial cells normally maintain a low [Ca²⁺]_i, around 60-110 nM. In the environment of the endothelial cell, higher extracellular Ca²⁺ concentration ([Ca²⁺]_o) levels are found, with values in the low millimolar range (1.8 mM), thereby establishing an endothelial [Ca²⁺]_o-to-[Ca²⁺]_i gradient of 20,000:1. Clearly, the resting endothelial membrane is restrictive to Ca²⁺ influx. However, many neurohumoral first messengers alter endothelial membrane Ca²⁺ permeability to allow for increased [Ca²⁺]_i that is critical for intraendothelial processes. The change in Ca²⁺ permeability is accomplished by the opening of Ca²⁺-permeable cationic channels. Several different Ca²⁺-permeable channels have been suggested to exist in endothelial cells, although no gene product for an endothelial Ca²⁺ channel has been identified.

Leak channels. Conditions dictating the [Ca²⁺]_o-to-[Ca²⁺]_i gradient are not static. Rather, a dynamic equilibrium exists as Ca²⁺ cycles from the extracellular millieu through the cell, through the intracellular storage organelles, and back into the vascular and interstitial spaces (13). Therefore, Ca²⁺ leaks across the endothelial cell membrane, and a leak rate has been estimated at 16 pmol · 10⁶ cells¹ · s¹ for bovine (B) PAECs (82). The rate of Ca²⁺ leak for endothelial cells depends on membrane potential and pH because depolarization decreases (71, 82, 169) and alkalosis increases (41), respectively, Ca²⁺ leak.

The precise role for passive Ca²⁺ leak in regulating endothelial function in situ is unclear, although it is known that constitutive Ca²⁺ leak in cultured BPAECs regulates both basal and stimulated endothelial cAMP production (171) (Fig. 2, A and B). In addition, BPAEC monolayer permeability to [³H]sorbitol is decreased when the Ca²⁺ leak pathway is blocked with La³⁺ (171) (Fig. 2C). The effects of Ca²⁺ leak on cAMP levels and macromolecular permeability are believed to be mediated, in part, by Ca²⁺ inhibition of type VI adenylyl cyclase (171). Thus normal Ca²⁺ leak may help establish in situ pulmonary endothelial barrier permeability by controlling some proportion of endothelial cAMP production. This tonic effect of Ca²⁺ leak on cAMP levels would provide the endothelial barrier with the ability to undergo bidirectional changes in vascular permeability because inhibiting or increasing Ca²⁺ leak would decrease or increase, respectively, permeability. Changes in endothelial barrier fluid and solute permeability could then be accomplished in response to changes in the electrochemical driving force for Ca²⁺ entry.

View larger version (21K): [in this window] [in a new window]   Fig. 2.   A: regulation of basal pulmonary arterial endothelial cell (PAEC) membrane adenylyl cyclase (AC) activity. Basal cAMP accumulation (measured in absence of Ca2+) was 5.0 ± 0.3 pmol · mg1 · min1 (n = 6 membranes) and was increased by 25 µM isoproteronol and 15 µM forskolin (P < 0.05; n = 9 membranes). Combination of isoproteronol and forskolin additively increased cAMP accumulation (P < 0.05; n = 9 membranes). * Significantly different from control.  Significantly different from individual treatments. B: regulation of isoproteronol- and forskolin-stimulated PAEC membrane AC activity. These studies were conducted without calmodulin to test whether AC activity was directly sensitive to submicromolar Ca2+. cAMP accumulation was stimulated to 29.5 ± 1.4 pmol · mg1 · min1 and was dose dependently inhibited by Ca2+ (P < 0.05; n = 9 membranes). Data are means ± SE. * Significantly different from isoproteronol and forskolin stimulation. C: role of type VI AC on endothelial permeability. Potential role of Ca2+-inhibitable AC in endothelial permeability was tested by assessing transfer ratio of [3H]sorbitol across 12-day postconfluent monolayers of PAECs under conditions of altered cytosolic Ca2+ concentration and cAMP. Data are means ± SE. Basal 2-h transfer ratio of [3H]sorbitol was 0.089 ± 0.003. La3+ (500 µM) and 3-isobutyl-1-methylxanthine (IBMX; 500 µM) reduced transfer ratio (P < 0.05; n = 6 membranes/group). Ionomycin (1 µM), cAMP-dependent protein kinase inhibitor Rp diastereomer of 8-bromoadenosine 3',5'-cyclic monophosphothioate (RpcAMPS; 3 mM), and their combination increased monolayer leak to [3H]sorbitol (P < 0.05; n = 6 membranes/group). * Significantly different from basal [3H]sorbitol transfer. [Modified from Stevens et al. (171).]

To date, the molecular structure of the endothelial leak channel(s) is not known. One possibility is that mammalian counterparts of the transient receptor potential-like (trp-L) gene product, a constitutively active cation (including Ca²⁺) conductance pathway in the Drosophila melanogaster retina (142), mediate Ca²⁺ leak in the endothelium. Overexpression of Drosophila trp-L in nonexcitable cell lines increases unstimulated cation conductance (76, 95, 97). The mammalian Trp3 channel shares a high sequence homology (147) and ion selectivity profile (16, 76, 206) with trp-L, i.e., nonselective with respect to Na⁺ and Ca²⁺. Overexpression of human Trp3 (hTrp3 or TRPC3) in HEK 293 (16) or Chinese hamster ovary cells (206) increases unstimulated cation transmembrane currents. Thus Trp3 would seem to be an attractive candidate for mediating endothelial Ca²⁺ leak. Recent data indicate that endothelial cells of systemic vascular origin express Trp3 and that the rat lung also possesses the Trp3 message (58). However, Trp3 may not be expressed in rat (R) or human (H) PAECs (123). Future studies are certainly necessary to determine the gene product(s) responsible for forming pulmonary endothelial leak channels.

Mechanosensitive Ca²⁺-permeable channels. Increased vascular shear stress is known to mediate changes in vascular tone via increased nitric oxide production (37, 50, 93, 143, 195). One mechanism that partly explains shear stress regulation of nitric oxide production involves activation of mechanosensitive Ca²⁺-permeable cation channels in endothelial cells. Indeed, shear stress induces cationic currents in human umbilical venous endothelial cells with a Ca²⁺-to-Na⁺ permeability ratio of 12 (156). Interestingly, the shear stress-activated Ca²⁺ influx is lost when sialic acid is removed from the endothelial glycocalyx, suggesting that channel activation is dependent on a physical coupling of the channel to some stress-sensing glycocalyx structure (74).

Store-operated Ca²⁺ entry. Whereas endothelial first messengers promote increased [Ca²⁺]_i via receptor-coupled Ins(1,4,5)P₃ generation and/or receptor-gated channel activation, the depletion of intracellular Ca²⁺ stores appears to stimulate Ca²⁺ entry by activation of capacitative or SOCs. Specifically, Ins(1,4,5)P₃-mediated liberation of stored Ca²⁺ promotes intracellular store depletion. In turn, Ca²⁺-permeable membrane channels are activated to allow for refilling of the Ca²⁺ storage pools. This Ca²⁺ entry pathway may be distinct from the previously discussed receptor-operated pathway because store depletion, independent of receptor activation, can be accomplished with Ca²⁺-ATPase inhibitors and membrane cationic currents can be measured (134).

Ca²⁺ Reuptake

Sarco(endo)plasmic reticulum Ca²⁺-ATPase. Termination of a Ca²⁺ signal, whether due to Ca²⁺ release or entry, involves [Ca²⁺]_i resequestration into intracellular stores. Biochemical data demonstrate that intracellular Ca²⁺ stores in excitable and nonexcitable cells are maintained by a class of ion-motive ATPases known as sarco(endo)plasmic reticulum Ca²⁺-ATPase (SERCA). To date, five SERCA isoforms have been characterized, and these are derived through alternative splicing of three gene products: SERCA1, -2 and -3. SERCA1a is the major adult fast-twitch skeletal isoform expressed in high levels in striated muscle sarcoplasmic reticulum (33). SERCA1b is developmentally regulated and only expressed in neonatal fast-twitch skeletal muscle (7). SERCA2a is the major isoform found in slow-twitch and cardiac muscles, whereas SERCA2b is ubiquitously expressed and is considered to be the "housekeeping"-type Ca²⁺ pump of nonexcitable cells (104, 182). SERCA3 is also expressed in nonexcitable cells (104, 182); however, SERCA2b appears to be the major candidate to act as the Ca²⁺ pump for Ins(1,4,5)P₃-sensitive Ca²⁺ stores (48). SERCA3 expression is most abundant in the intestine, thymus, and cerebellum, with lower levels of expression in the spleen, lymph node, and lung (196). Quantification of SERCA isoform transcripts from rat lung reveals that SERCA2b (59%) is the predominant isoform followed by SERCA3 (28%) and SERCA2a (13%) (196). Although there is no direct evidence, SERCA3 expression may be limited to a subpopulation of airway secretory epithelial cells (6, 196). However, SERCA3 expression has been demonstrated in endothelial cells from the rat aorta and cardiac microvascular circulation (6, 196). Therefore, the SERCA isoform(s) relevant to lung endothelial cells has yet to be determined.

	ADENOSINE 3',5'-CYCLIC MONOPHOSPHATE

The adenine nucleotide cAMP is a conserved, ubiquitous intracellular second messenger. Intracellular cAMP content at any given time is a reflection of the balance between production and degradation (35, 179, 180). Production of cAMP is regulated by ligand-receptor-G protein activation of adenylyl cyclases, both stimulatory via G_s and inhibitory via G_i. The production of cAMP is therefore susceptible to regulation by ligand concentrations, ligand-receptor occupancies, the degree of G_s versus G_i activation, processes that regulate the intrinsic cycling of G proteins (e.g., GTP vs. GDP bound state), and the quantity and activity of adenylyl cyclase isoform(s). cAMP degradation is due to phosphodiesterase (PDE) activity, although membrane-associated cAMP transporters that extrude cAMP have been identified (68). Emerging evidence indicates that both cAMP production and degradation may be influenced by [Ca²⁺]_i.

The effect of cAMP on endothelial cell barrier properties, at this time, seems unambiguous. Whereas [Ca²⁺]_i and the degree of endothelial barrier "leakiness" often are not related, cAMP levels and pulmonary endothelial barrier integrity are usually highly coupled, i.e., lower cAMP concentration = relatively leaky and higher cAMP concentration = tight barrier. Elevation of cAMP due to receptor-coupled or direct activation of adenylyl cyclase, activation of G_s proteins, inhibition of PDEs, or application of cAMP mimetics all appear to increase cell-cell and cell-matrix tethering, decrease isometric tension development, decrease intercellular gap formation, and decrease permeability in multiple experimental preparations (2, 3, 10, 26, 27, 29, 47, 56, 72, 86, 87, 90, 98, 119, 120, 131, 139, 141, 160-162, 167, 171). Often, the cAMP target that confers the barrier-enhancing effect is poorly understood, but the message that elevated cAMP decreases permeability is clear.

cAMP Production

Recent studies (29, 170, 171) using lung endothelial cells have been conducted to determine whether elevated [Ca²⁺]_i, as occurs during inflammation, decreases cellular cAMP content by inhibiting adenylyl cyclase activity. Because increased [Ca²⁺]_i promotes endothelial disruption and increased cAMP opposes endothelial disruption, Ca²⁺ inhibition of adenylyl cyclase may provide a mechanism by which [Ca²⁺]_i could decrease cAMP content and thus permissively increase permeability. RT-PCR cloning has revealed expression of multiple isoforms, including the type VI Ca²⁺-inhibited adenylyl cyclase in cultured PAECs (170, 171) and PMVECs (29, 170), whereas immunostains similarly revealed expression of the type VI enzyme in endothelial cells throughout the intact pulmonary circulation (29). Studies using PAEC membrane fractions demonstrated that adenylyl cyclase activity is in fact inhibited by Ca²⁺, and studies using intact PAECs and PMVECs demonstrated that elevations in [Ca²⁺]_i also decreased cAMP content (29, 170, 171). These data substantiate the idea that Ca²⁺ inhibition of adenylyl cyclase regulates endothelial cell cAMP content.

One important regulatory feature of type VI adenylyl cyclase is its inhibition by compartmentalized Ca²⁺ responses. Nonspecific elevations in [Ca²⁺]_i with ionophores, for example, are capable of inhibiting enzyme activity (56, 171). However, Ca²⁺ release from intracellular stores is insufficient but activation of Ca²⁺ entry across the cell membrane is sufficient to inhibit adenylyl cyclase activity (35). The reason for the disparate effect was not initially apparent, although recent elegant work by Cooper et al. (35) demonstrated that activation of Ca²⁺ entry results in sustained increases in membrane-associated Ca²⁺ levels ranging from 1 to 40 µM, whereas activation of Ca²⁺ release resulted in transient increases in membrane-associated Ca²⁺ levels in the range of 0.5-1 µM. Therefore, Ca²⁺ entry rather than Ca²⁺ release most likely regulates type VI adenylyl cyclase activity because 1) high concentrations of membrane-associated Ca²⁺ levels are achieved with stimulation of Ca²⁺ entry and 2) diffusion limitations of Ca²⁺ released from storage organelles are not in close proximity to the type VI adenylyl cyclase.

Certain receptor-coupled agonists may also activate G_i proteins to decrease cAMP. Work by Manolopoulos et al. (110) suggested that thrombin is negatively coupled to adenylyl cyclase through a G_i protein that decreases cAMP independent of Ca²⁺ entry and inhibits isoproterenol stimulation of adenylyl cyclase. Experiments conducted by other investigators (187) using thrombin-challenged macro- and microvascular pulmonary endothelial cells have shown thrombin-induced inhibition of cAMP accumulation requires the presence of extracellular Ca²⁺. Thus agonists receptor coupled to phospholipase C may decrease adenylyl cyclase activity by activation of a G_i protein and/or stimulation of Ca²⁺ entry. Future studies will be required to determine whether extracellular Ca²⁺ influences G_i protein coupling to adenylyl cyclase.

As indicated, cAMP responses in endothelial cells are complicated because of the expression of multiple adenylyl cyclase isoforms, and thus the specific contribution of type VI adenylyl cyclase to cAMP content is difficult to assess. Recently, pertussis toxin was shown to possess differential sensitivities for adenylyl cyclase isoforms (179, 180). G_i greatly suppresses the activity of Ca²⁺-inhibited adenylyl cyclases compared with other isoforms, indicating that types V and VI adenylyl cyclase are most likely the targets for pertussis toxin-sensitive G_i proteins. Interestingly, pertussis toxin 1) is generally considered a potent cAMP-elevating agent in endothelial cells and 2) prevents thrombin-induced inhibition of cAMP accumulation in cells expressing type VI adenylyl cyclase (110). These data suggest that a pertussis toxin-sensitive adenylyl cyclase, likely type VI, regulates endothelial cell cAMP content. These data may also explain how both Ca²⁺ entry and G_i-coupled receptors synergistically or redundantly regulate cAMP content, i.e., by converging on the same isoform of adenylyl cyclase.

Ogawa et al. (135) have provided a link between pertussis toxin-sensitive adenylyl cyclase activity and endothelial cell permeability. These investigators have shown that exposure of endothelial cells to hypoxia results in decreased cAMP content and increased macromolecular permeability. Pertussis toxin apparently uncouples hypoxia from decreasing cAMP and thus preserves endothelial barrier function. A theoretical enigma exists because other investigators (169, 171) have shown that hypoxia reduces Ca²⁺ leak into endothelial cells and thus might be expected to increase cAMP levels. However, the conflicting data may be resolved by examining the time dependence of both [Ca²⁺]_i and cAMP changes. Still, these data provide additional support for the idea that type VI adenylyl cyclase plays a privotal role in regulating cAMP content, important for control of endothelial cell barrier function.

Some diversity in responsiveness to pertussis toxin has been seen between endothelial cells from different vascular beds, different sites within a vascular bed, between species, and between experimental conditions. However, if pertussis toxin regulates type VI adenylyl cyclase and represents an important regulatory mechanism of endothelial barrier properties, then differential sensitivities to pertussis toxin should predict the endothelial permeability response. In support of this idea, pertussis toxin produces barrier disruption in endothelial cells, which respond with little to no increase in cAMP, and barrier enhancement in endothelial cells, which respond with a large increase in cAMP (139, 140). Future studies will be required to carefully dissect the nature of interaction among pertussis toxin, type VI adenylyl cyclase activity, and endothelial barrier function.

In addition to a Ca²⁺-inhibited isoform of adenylyl cyclase, endothelial cells express two adenylyl cyclase isoforms (types II and VII) that are stimulated by PKC. Indeed, RT-PCR studies using RNA from cultured cells indicated expression of types II and VII adenylyl cyclase, whereas immunostains of lung sections detected type II adenylyl cyclase in macro- and microvascular pulmonary endothelial cells. These observations are significant because both arms of the phosphoinositide pathway, Ca²⁺ and PKC, appear to influence cAMP responses. The contribution of types II and VII adenylyl cyclases to endothelial cell cAMP homeostasis is generally poorly understood, although recent work indicates that inhibition of PKC reduces cAMP content and direct stimulation of PKC increases cAMP content, consistent with the idea that PKC regulates adenylyl cyclase activity (170). Furthermore, preliminary data indicate that neurohumoral inflammatory mediators coupled to G_q proteins and phosphoinositide turnover first cause Ca²⁺ inhibition of cAMP content that is followed by PKC-dependent stimulation of cAMP content (Stevens, unpublished observations). Future studies will be required to more carefully address the independent and combined effects of Ca²⁺-inhibited and PKC-stimulated isoforms of adenylyl cyclase on endothelial cell function because the relative expression of these isoforms may have profound effects on the endothelial cell permeability responses to neurohumoral inflammatory mediators.

cAMP Degradation

Seven gene families express >50 cyclic nucleotide PDEs in various mammalian tissues and species (34), thereby making biochemical studies on endothelial PDEs difficult. Endothelial cells have an added complexity in that different complements of PDE isoforms are expressed in different vascular beds or even areas of the same tissue. Although the nonselective PDE inhibitor 3-isobutyl-1-methylxanthine has been widely used to inhibit cAMP degradation in many studies, the development of selective inhibitors to characterize different isoforms has progressed rapidly in recent years. Agents such as rolipram and indolidan may be extremely useful for identifying specific roles for various isoforms in intact cells and tissues but, as yet, selective inhibitors have not been widely employed for endothelial functional analysis. The present data on the regulation of cAMP (and cGMP) metabolism in endothelial cells, however, do support a pivotal role for PDEs.

Data are limited on the expression of PDE enzymes in human endothelial cells. To the authors' knowledge, there are no detailed PDE characterizations in human endothelial cells. However, PDE3, characterized by its high affinity for cAMP as a substrate and the unique regulatory property of inhibition by cGMP, has been implicated as mediating cGMP-dependent reduction in thrombin-induced increased permeability of human umbilical venous endothelial cell monolayers (44, 175). A similar role for PDE3 in human aortic endothelial cells has not been observed (44). These observations emphasize the differences in PDE expression throughout the vasculature, although differences in umbilical venous and aortic endothelial cell isolation and culture may be considered.

Cyclic nucleotide PDE2 and -4 appear to be the major enzymes found in bovine and porcine endothelial cells, whereas PDE1, -3, -5, and -7 show relatively minor activity (8, 89, 91, 101, 165). Significant PDE3 activity has been detected in the particulate fraction of porcine endothelial cells, implicating a potential role for this isozyme in endothelial cAMP level regulation (177). PDE2 represents a major cGMP-receptor site and a potential site for cross talk between the cGMP and cAMP pathways. PDE2 hydrolyzes both cAMP and cGMP equally at saturating substrate concentrations. However, the enzyme contains two noncatalytic, high-affinity cGMP-binding sites (12), the occupation of which causes a conformational change in the enzyme and increases cAMP hydrolysis at low substrate concentrations. The effect of cGMP binding is to alleviate positive cooperativity of cAMP hydrolysis in the absence of cGMP. Because of this mechanism, activation occurs only at low cAMP substrate concentrations. PDE2 has been identified as the major endothelial cGMP-degrading enzyme in porcine PAECs (176) based on the observation of a synergistic reduction in monolayer permeability with atrial natriuretic peptide and the PDE2 inhibitor erythro-9-(2-hydroxy-3-nonyl)adenine. PDE2 cloning from brain tissues has shown two splice variants with hydrophobic amino-terminal substitutions that are thought to allow membrane association (202). As yet, an endothelial PDE2 has not been cloned.

The PDE4 gene family is far more complex than the PDE2 family. Four subfamilies designated PDE4A-D are known, with 15-20 splice variants occurring in these subfamilies. Subfamilies A-D are not necessarily expressed in all human or rat tissues (45). All isoforms show high-affinity cAMP-specific catalysis, and endothelial PDE4 exhibits Michaelis-Menten kinetic behavior and lacks the negative cooperativity or multiple affinity kinetics characteristic of some PDE4 enzymes. It has been proposed that one variant of PDE4 shows a membrane locus (99), therefore supporting the possibility of subcellular compartmentalization, presumably related to regional cAMP-controlled phosphorylation cascades.

Studies on the regulation of PDE2 and -4 in endothelial cells and their role in regulating endothelial barrier function are even more limited than biochemical characterizations. However, some studies (11, 91, 177) have shown that both of these isoforms may participate in agonist-induced changes in endothelial function. TNF- treatment of bovine aortic endothelial cell monolayers increases permeability and decreases thrombomodulin activity in a time- and dose-dependent manner. Both of these effects are associated with a decrease in cAMP and correlate with increased PDE2 and -4 activities, although variation in the PDE2 and -4 contributions to the TNF- effect has been reported (91). PDE3 and -4 inhibitors have been shown to attenuate H₂O₂-induced permeability alterations in porcine PAECs (177). Furthermore, a specific PDE4 inhibitor, rolipram, both prevents and reverses ischemia-reperfusion (I/R)-induced increased vascular permeability in isolated rat lungs (11). Thus at least PDE4 activation may be important for allowing increased lung vascular permeability.

The importance of differential PDE expression and activity within endothelial cells in conjunction with fluctuations in [Ca²⁺]_i raises an interesting question. Because Ca²⁺ influx regulates cAMP production to affect endothelial barrier properties (171), can increased [Ca²⁺]_i influence cAMP degradation to regulate permeability? The effects of [Ca²⁺]_i on PDE activity potentially involve two opposing cellular events: 1) increased [Ca²⁺]_i may synergistically promote permeability via PDE activation or 2) it may inhibit PDE activity and thus limit a Ca²⁺-induced permeability response. Significant Ca²⁺/calmodulin-stimulated cAMP PDE activity (PDE1) exists in smooth muscle cells (89, 144), and PDE1 has been detected in lysates of cultured porcine aortic endothelial cells (157). However, ionomycin has been reported to inhibit PDE4 in permeabilized endothelial cells (85). Therefore, future studies are needed to determine the precise effects of [Ca²⁺]_i on PDE activities in endothelial cells and how Ca²⁺ regulation of PDE function relates to regulation of pulmonary vascular permeability.

	CONDUIT-VESSEL AND MICROVASCULAR PULMONARY ENDOTHELIAL CELLS

Even though substantial data indicate that Ca²⁺ promotes endothelial permeability and cAMP reduces permeability, it remains unclear how the intact pulmonary vascular endothelial barrier specifically responds to different inflammatory events. Considerable effort has been made to better understand the normal physiology of pulmonary endothelium and to elucidate mechanisms responsible for producing endothelial dysfunction with the following experimental tools: 1) whole animal and isolated lung models that can be studied both in the physiological and pathological settings with respect to describing the biophysical characteristics of the entire vascular bed and 2) cultured endothelial cells that can be used to describe specific mechanisms of various inflammatory mediator effects.

Until recently, only conduit vessel-derived (i.e., pulmonary artery) endothelial cells have been readily available for mechanistic studies related to endothelial cell permeability regulation. A recurrent concern, however, has been that the data obtained from these cells do not adequately represent cells in situ existing at the putative edemagenic sites, i.e., the microvasculature. Indeed, microscopic evaluation of intact pulmonary circulations has shown microvascular endothelial cells are phenotypically distinct from conduit-vessel endothelial cells (40). Several laboratories have now isolated and cultured endothelial cells from the pulmonary microcirculation (168) and are providing evidence to show that microvascular cells are, in fact, distinct, even in culture, from pulmonary artery-derived endothelial cells.

In regard to normal barrier function, cultured PMVECs are more restrictive to basal macromolecular diffusion than similarly cultured PAECs (84). Although macromolecular convective flux and fluid flux measurements have yet to be performed, it is not unexpected that cultured PMVECs form a highly restrictive monolayer on reaching confluence. In situ estimates of the microvascular filtration coefficient (K_f,c) from intact pulmonary circulations from numerous species (181) indicate that pulmonary hydraulic conductances average 0.21 ± 0.05 ml · min¹ · mmHg¹ · 100 g lung wet weight¹. Because this average K_f,c value is much greater than the values reported for the intact skeletal muscle and cerebral vascular beds but is comparable to K_f,c estimates from the liver, heart, and intestinal circulations, the lung at first appears to be a relatively "leaky" organ. However, this is not at all the case because the lungs contain a much higher surface area for fluid exchange relative to other vascular beds. When quantitative estimates of the total capillary exchange area for the lung are compared with those of the skeletal muscle and the hydraulic conductivity (L_p,c; in cm³ · dyn¹ · s¹) values are calculated, the microcirculation of the lung (L_p,c = 0.51 ± 0.09) is actually much "tighter" than that of skeletal muscle (L_p,c = 2.2). It is not surprising that PMVECs form tight barriers because normal gas exchange would be impaired without a microcirculation highly restrictive to fluid and solute leakage.

Two important questions now are: 1) how do inflammatory mediators affect microvascular endothelial function? and 2) are PMVEC responses to classic inflammatory mediators different from conduit vessel-derived endothelial cell responses, especially when phenotype differences exist? These questions are just beginning to be answered, and some initial data specifically related to [Ca²⁺]_i regulation and permeability responses are emerging.

Responses to both the receptor-coupled agonist ATP and the receptor-independent, SOC-activating agonist thapsigargin have been compared with the use of cultured PMVECs and PAECs (84, 170). [Ca²⁺]_i responses to both agonists are decreased in microvascular endothelial cells compared with the conduit vessel-derived endothelial cells. In particular, store-operated Ca²⁺ entry is greatly suppressed in the microvascular cells under identical physiological conditions (84). Investigation into the mechanisms responsible for the suppressed Ca²⁺ entry reveals that microvascular cells exist at a relatively depolarized resting membrane potential and exhibit an attenuated hyperpolarization in response to agonist stimulation (170). Thus this decreased electrochemical driving gradient for Ca²⁺ entry may underlie the suppressed Ca²⁺ responses in the pulmonary microvascular endothelium, consistent with the previously reported effects of membrane potential on agonist-induced [Ca²⁺]_i responses in endothelium of isolated microvessels (38, 71, 73).

Elevated [Ca²⁺]_i has been clearly linked to increased permeability in conduit vessel-derived cells (62, 84, 102, 103, 122, 152, 153, 155, 158) as well as in isolated perfused lungs (29, 80, 88). However, the findings that pulmonary microvascular cells exhibit attenuated agonist-induced [Ca²⁺]_i responses suggest that these cells may display attenuated permeability responses. Store-operated Ca²⁺ entry mediates cell shape change and increases macromolecular flux across cultured PAEC monolayers (84, 122, 123). However, activation of store-operated Ca²⁺ entry does not promote these effects in identically treated PMVEC monolayers (84). Interestingly, when experimental measures are taken to produce equivalent [Ca²⁺]_i levels in both cell types by SOC activation, the pulmonary microvascular cell monolayers are still resistant to cell shape change and increased macromolecular flux. Therefore, one could speculate that inflammatory agonists activating SOCs may have preferential actions for different pulmonary vascular segments in situ, and the functional consequences of SOC activation may show heterogeneity between microvascular and conduit-vessel endothelial cells. Important future studies will be necessary to determine whether ex vivo culture conditions influence the differential pulmonary endothelial [Ca²⁺]_i and permeability responses.

The observation that equivalent [Ca²⁺]_i levels are achieved by SOC activation in pulmonary conduit-vessel endothelial cells and PMVECs without the production of identical effects on cell shape or macromolecular permeability suggests that microvascular endothelial cell shape and barrier regulation may be relatively Ca²⁺ insensitive. This concept deviates from studies that show that PAEC shape change and barrier disruption occurs in response to Ca²⁺-driven, myosin light chain-dependent active "contraction" (54, 61, 130, 131, 160, 198, 199). However, nonspecific elevations in [Ca²⁺]_i with ionophore treatment at doses sufficient to produce cultured PAEC shape change fail to produce cell shape change and increase macromolecular flux in cultured PMVECs (57, 84). These data suggest that microvascular endothelial cell shape may not be influenced directly by changes in [Ca²⁺]_i and are consistent with a previous report (43) that showed that inhibition of phosphatase activity in microvascular endothelial cells increases macromolecular permeability via a myosin light chain-independent mechanism.

In addition to differing [Ca²⁺]_i responses, cAMP responses vary between conduit-vessel and microvascular endothelial cells, although both PAECs and PMVECs express a Ca²⁺-inhibitable adenylyl cyclase (29, 170). In PAECs, endogenous cAMP levels appear to decrease in response to activation of store-operated Ca²⁺ entry (29, 170). However, neurohumoral inflammatory mediators that activate store-operated Ca²⁺ entry do not decrease cAMP content in microvascular endothelial cells (170), although stimulation of cAMP production or inhibition of cAMP degradation decreases permeability in both cell types (128). We predict that preserved cAMP responses represent an important determinant of the more restrictive barrier formed by PMVECs. This speculation is supported by previous key studies indicating the importance of cAMP in regulating vascular permeability because strategies aimed at elevating cAMP prevent or reverse agonist-evoked and I/R-induced leak (2, 3, 10, 26, 27, 29, 47, 56, 72, 86, 87, 90, 98, 119, 120, 131, 139, 141, 160-162, 167, 171).

	IMPLICATIONS

How can we begin to interpret the data from cultured pulmonary endothelial cells and integrate these findings with those from whole organ studies to enhance our knowledge of the mechanisms responsible for permeability-induced pulmonary edema? This integration process is critical because the diagnosis of clinical acute lung injury and adult respiratory distress syndrome is based, in part, on a patient's level of refractory hypoxemia, which reflects alveolar edema and microvascular dysfunction. Postmortem assessment of lung structure indicates that patients whose death is associated with various forms of noncardiogenic pulmonary edema certainly have severely disrupted microvascular architecture. The precipitating factors causing this condition are many, and overwhelming inflammation is usually associated with severe permeability-induced pulmonary edema (42, 92). However, the precise mechanisms causing the pulmonary microvascular dysfunction are still not clear, and thus the therapeutic regimen for treating the adult respiratory distress syndrome patient is largely supportive care with ventilatory management only.

On the basis of preliminary data from cultured pulmonary microvascular cells, receptor-linked Ca²⁺-promoting inflammatory mediators may not induce significant PMVEC tension development, although Ca²⁺-driven tension development is associated with increased macromolecular and fluid flux across PAEC monolayers (54, 61, 130, 131, 160, 198, 199) and in isolated lungs (88), respectively. Experiments dedicated to resolving these puzzling observations may help to accomplish two goals: 1) identify specific in situ leak sites occurring with noncardiogenic pulmonary edema of different etiologies and 2) describe novel functions for pulmonary conduit-vessel endothelial cells with respect to cell shape change in response to physiological agonists.

Recent data support the notion that different inflammatory events can produce distinct patterns of pulmonary vascular leak. In one scenario, pulmonary conduit vessels appear to become leaky without a discernible contribution of the alveolar microcirculation to the leak. In another scenario, leak occurs in all segments of the pulmonary vasculature. Individual examples of each scenario can be represented by the thapsigargin-induced (store-operated Ca²⁺ entry-induced) (29) and the I/R-induced lung edema models (2, 3, 11, 87, 88, 124-129, 161), respectively. In isolated lungs, both thapsigargin and I/R produce significant increases in the K_f,c, a biophysical index of fluid and small-solute permeability of the vascular wall. In fact, when an EC₅₀ dose of thapsigargin is compared with a 30-min period of postischemic reperfusion, changes in total vascular permeability are nearly identical. However, morphological changes in the two lung edema models are not identical (Fig. 3). Histological inspection of thapsigargin-challenged lungs reveals perivascular cuffing of larger vessels, with no apparent disruption of microvessels. Electron micrographs appear to support this view, demonstrating that endothelial cells of larger arteries and veins possess gaps, whereas smaller (150 µm and less) arteriolar, venular, and capillary cells remain intact (P. M. Chetham, P. Babal, J. P. Bridges, I. F. McMurtry, and T. Stevens, unpublished observations). In contrast, a different histological profile is revealed on inspection of I/R-challenged lungs. Perivascular cuffing of larger vessels is still present, but this is accompanied by capillary disruption and alveolar infiltrate (127). Detailed transmission electron microscopy reveals that the capillary disruption is characterized by dissociation of endothelial cells from the underlying basement membrane but that intercellular gap formation is not discernable (Chetham et al., unpublished observations).

View larger version (97K): [in this window] [in a new window]   Fig. 3.   Light micrographs reveal different patterns of edema in thapsigargin-treated isolated rat lungs compared with rat lungs subjected to ischemia and reperfusion. A: section of untreated (control) rat lung. B: rat lung challenged with 100 nM thapsigargin for 20 min. Perivascular cuffing (arrow) is evident in thapsigargin-treated lung, but alveolar edema is absent (*). C: rat lung subjected to 45 min of normothermic, global ischemia followed by 30 min of reperfusion. Both perivascular cuffing and alveolar infiltrate are present.

In addition to the histological differences between the two models, mechanistic differences with respect to the induction of vascular leak are also apparent. As expected, the thapsigargin-induced increased permeability requires extracellular Ca²⁺, and interventions designed to decrease Ca²⁺ entry diminish the increase in vascular permeability (29). Furthermore, extracellular antioxidants do not prevent the thapsigargin-induced permeability increase (29). In contrast, I/R-induced increased permeability is oxygen radical dependent (4, 94, 127), and initial data indicate that I/R injury occurs even when extracellular Ca²⁺ levels are diminished (Chetham et al., unpublished observations). However, the Ca²⁺ regulatory component to pulmonary microvascular endothelial barrier integrity during I/R may be highly complex. Superoxide is known to inhibit some microsomal SERCA isozyme activity to promote Ca²⁺ release (116, 138). In addition, calmodulin antagonists block I/R-induced permeability changes in whole lungs (88), and an H₂O₂-induced pulmonary microvascular permeability increase in cultured monolayers is correlated to Ca²⁺ release. These observations suggest that a critical component to I/R-induced pulmonary microvascular damage may be endothelial microsomal Ca²⁺ release. In a preliminary report, it has been shown that depletion of intracellular Ca²⁺ stores before exposure of isolated lungs to I/R significantly reduces the I/R-induced permeability edema (Chetham et al., unpublished observations). Thus oxidant production during I/R and endothelial [Ca²⁺]_i may, in fact, be linked in promoting microvascular barrier dysfunction, but the mechanism may not involve Ca²⁺ entry.

These two models of noncardiogenic pulmonary edema, store-operated Ca²⁺ entry activation and I/R, have brought into focus critical issues related to lung endothelial cell function: 1) activation of Ca²⁺ entry may produce intercellular gap formation in selective vascular segments located away from gas-exchange areas, and 2) the pulmonary oxidant burden during a pulmonary inflammatory event may determine the degree of microvascular involvement in the overall permeability response. These observations are significant because multiple inflammatory stimuli, including thrombin, complement, platelet-activating factor, and substance P, may activate store-operated Ca²⁺ entry, promote oxidant generation, or both.

	SUMMARY

Top Abstract Introduction Summary References

This review identifies the sites of potential cross talk between [Ca²⁺]_i and cAMP in lung endothelial cells. It is clear that although such sites of cross talk exist, their relevance to endothelial cell function are, for the most part, poorly understood. In particular, although both [Ca²⁺]_i and cAMP clearly influence endothelial cell barrier function, little is known about how [Ca²⁺]_i and cAMP responses are balanced during development of intercellular gap formation and in the reformation of an intact barrier. Future studies will be required to address these important issues.

Perhaps most strikingly, emerging data support the