Annexin A2 mediates secretion of collagen VI, pulmonary elasticity and apoptosis of bronchial epithelial cells

The basement membrane is an electron-dense, 50- to 100-nm-thick, sheet-like structure composed of intricately assembled extracellular matrix proteins that separate epithelial cell monolayers from their underlying connective tissue (LeBleu et al., 2007). Not only do basement membranes provide structural support and serve as barriers to toxins and pathogens, they also provide a conduit for signaling and mechanotransduction (Bateman et al., 2009; Lu et al., 2011). Polarized epithelial cells, such as those in the lung, continuously remodel the basement membrane by secreting and proteolyzing its component proteins; in turn, basement membrane proteins support epithelial cell-adhesion, survival and differentiation (Pozzi and Zent, 2011). Alterations in basement membrane composition accompany an array of human lung disorders, including asthma, chronic obstructive pulmonary disease and interstitial fibrosis (Cox and Erler, 2011; Kranenburg et al., 2006; Lambrecht and Hammad, 2012; Proud and Leigh, 2011).

Most basement membranes consist of a collagen IV and laminin superstructure, whose stability and compliance are refined by expression of additional proteins (LeBleu et al., 2007). In the lung, the basement membrane scaffolding is buttressed by agrin, fibulin, netrin, nidogen, perlecan, tropoelastin (Manabe et al., 2008) and collagen VI (COL6), which together impart elasticity, mechanical support and signaling capability to the bronchial wall (Bober et al., 2010; Gelse et al., 2003; Specks et al., 1995). COL6, a microfibrillar collagen, forms filaments that anchor basement membranes to other collagen fibers (Bonaldo et al., 1990; Engvall et al., 1986; Keene et al., 1988). Abnormalities in COL6 lead to myopathy in the mouse (Bonaldo et al., 1998; Irwin et al., 2003) and zebrafish (Telfer et al., 2010), and to Ullrich and Bethlem congenital muscular dystrophy in humans (Baker et al., 2005; Bönnemann, 2011).

Many basement membrane polypeptides undergo intracellular synthesis and maturation, followed by extracellular assembly (LeBleu et al., 2007). Laminin, for example, consists of polypeptide chains that form intracellular αβγ trimers (∼800 kDa), which aggregate into a hexagonal polymer only after secretion (LeBleu et al., 2007). Nidogen and perlecan are secreted without prior complex formation (LeBleu et al., 2007). Collagen IV, which accounts for about 50% of most basement membrane protein mass, assembles into triple helical, kilodalton-scale protomers (∼500 kDa), which are also secreted before extracellular assembly into a stable network (Gelse et al., 2003). This is similar to many other collagens that form classical fibrils (Tooley et al., 2010). However, in sharp contrast, COL6 undergoes extensive intracellular assembly of highly crosslinked megadalton-scale 12-subunit complexes (∼2000 kDa), which then, upon secretion, associate into a microfibrillar network (Bateman et al., 2009; Colombatti et al., 1995; Engvall et al., 1986).

Annexin A2 (ANXA2), a component of a cell surface fibrinolytic receptor complex (Dassah et al., 2009; Flood and Hajjar, 2011), belongs to a family of conserved, membrane-binding proteins that has been implicated in intracellular membrane-fusion events (Gerke et al., 2005). Intracellularly, ANXA2 and S100A10 form a heterotetramer, which appears to have a calcium-regulated membrane-bridging function. In cultured cells, ANXA2 regulates secretion of surfactant-containing lamellar bodies in the type II pneumocyte (Wang et al., 2007), release of catecholamines from dense core vesicles in adrenal chromaffin cells (Umbrecht-Jenck et al., 2010), and exocytosis of ultra-large von Willebrand factor multimers from endothelial cell Weibel–Palade bodies (Knop et al., 2004). In these processes, ANXA2 appears in vitro to crosslink secretory granules to the plasma membrane by interacting with soluble N-ethylmaleimide-sensitive factor acceptor proteins 23 and 25 (SNAP-23 and SNAP-25), as well as vesicle associated membrane protein 2 (VAMP2), which are both components of the SNAP receptor (SNARE) complex (Nakata et al., 1990; Umbrecht-Jenck et al., 2010; Wang et al., 2007).

On the basis of previous findings that implicate ANXA2 in membrane-fusion events in vitro, we sought to define the role of ANXA2 in basement membrane secretion in vivo, and its physiological relevance. We noted that AnxA2^−/− mice exhibit reduced exercise tolerance and abnormal pulmonary stiffness. Examination of lung tissue from resting mice revealed dysmorphic bronchial epithelial cells, high-level apoptosis and cell dropout. Specific absence of COL6 within the AnxA2^−/− basement membrane correlated with retention of COL6 bronchial epithelial cell secretory vesicles. In cultured AnxA2^−/− fibroblasts, secretion of COL6 was impaired, and COL6 was retained within a late-Golgi, VAMP2-positive compartment. By immunoelectron microscopy and immunoprecipitation, ANXA2 and COL6 were identified to be associated with VAMP2 and SNAP-23 within a secretory vesicle membrane complex. Interestingly, skeletal muscle from AnxA2^−/− mice also demonstrated accumulation of COL6 in the interstitium, its site of synthesis. Together, these data show that ANXA2 enables the specific secretion of COL6, and that absence of ANXA2 leads to epithelial cell dropout, apoptosis and a physiologically significant restrictive pulmonary ventilatory defect.

When littermate mice were enrolled in an exercise protocol consisting of daily 90 minute swims over a 3 week period, AnxA2^−/− mice exhibited significantly longer post-swim recovery times (49.2±3.5 versus 70.5±7.8 minutes; mean ± s.e., n = 5 per group, P = 0.018; Fig. 1A). Compared with AnxA2^+/+ mice, AnxA2^−/− mice from five separate heterozygous matings showed reduced post-exercise grooming activity and distinctly longer times to complete drying (Fig. 1B). Evaluation of cardiac function by M-mode echocardiography revealed no significant difference in resting heart rate or cardiac output, and heart-to-body-mass ratios at baseline and post exercise did not differ between the two genotypes (not shown). Upon evaluation of pulmonary biomechanics, however, AnxA2^−/− mice showed significant decreases in both dynamic and static compliance (Fig. 1C,D; P<0.0001), and corresponding increases in both whole lung and tissue elastance (Fig. 1E,F; P<0.0001). Total dynamic resistance was also increased in the AnxA2^−/− lung (Fig. 1G; P<0.0001), whereas airway resistance showed no significant change (Fig. 1H; P = 0.732), even following increasing doses of methacholine (supplementary material Fig. S1A). Tissue dampening, finally, was increased (Fig. 1I; P<0.0001), whereas inspiratory capacity was significantly reduced (Fig. 1J; P<0.0001). Together, these data indicated significantly increased tissue stiffness and ventilatory restriction in the AnxA2^−/− mouse lung.

Hematoxylin and eosin staining of AnxA2^+/+ and AnxA2^−/− lung tissue from resting mice showed normal cellularity and arborization of bronchioles and alveoli. However, cross-sections through terminal bronchioles, as well as sections at other levels of airway ramification, revealed increased basophilia and dense, flattened bronchiolar epithelial cells (BECs) (Fig. 2A,B). At the same time, basement membrane autofluorescence, which is imparted by collagens and elastin (Evans et al., 2000), was indistinct in AnxA2^−/− lung tissue (Fig. 2C,D). Examination of Toluidine-Blue-stained 1 μm sections revealed the absence of a discrete, compact basement membrane, lack of apical Clara cell (or non-ciliated bronchiolar secretory cell) protrusions and a paucity of cilia in AnxA2^−/− BECs (Fig. 2E,F). Electron microscopy further demonstrated the indistinct nature of the AnxA2^−/− bronchiolar basement membrane (supplementary material Fig. S1B,C). Scoring of cells lining bronchioles 65–85 µm in diameter confirmed a reduction in the proportion of ciliated cells within AnxA2^−/− bronchioles (8.3±0.34%, mean ± s.e.; n = 6) compared with AnxA2^+/+ bronchioles (47.5±3.7, n = 8; P<0.001) (Fig. 2G). These data suggest abnormalities in both bronchiolar epithelial cells and their basement membranes.

Immunostaining of bronchioles from AnxA2^+/+ and AnxA2^−/− mice confirmed that ANXA2 was localized mainly within the basal portions of bronchial epithelial cells, in close proximity to the autofluorescent basement membrane (Vishwanatha et al., 1995) (Fig. 2H). To determine the expression level and distribution of bronchiolar basement membrane proteins in the AnxA2^−/− lung, a hierarchical approach was adopted (supplementary material Table S1). From a database of mouse lung matrix proteins, we selected those for which basement membrane-specific expression had been demonstrated (16 proteins), for which staining antibodies were available (11 proteins), and for which expression in the adult mouse lung had been documented (7 proteins) (Manabe et al., 2008). We then compared the expression pattern of these seven proteins (tropoelastin, keratin 5, collagen I, collagen IV, laminin A, fibrillin-1 and COL6) in lung tissue from resting AnxA2^+/+ versus AnxA2^−/− (supplementary material Fig. S2; Fig. 3A–C). Although six proteins showed no difference in staining pattern (supplementary material Fig. S2), basement membrane staining for COL6 was indistinct in the AnxA2^−/− lung and mirrored the loss of basement-membrane-associated autofluorescence. Upon quantification of staining intensity over a series of stained sections, we noted that peribronchiolar basement membrane staining of COL6 in AnxA2^+/+ was, on average, 3.3-times greater than that of AnxA2^−/− tissue (18.4±2.4 versus 5.6±1.6 arbitrary units; mean ± s.e., n = 6, P<0.01) (Fig. 3A–C). These data suggest a COL6-specific defect in synthesis, processing or secretion in AnxA2^−/− bronchial epithelial cells.

Evaluation of pulmonary biomechanics in Col6a1^−/− versus Col6a1^+/− mice revealed a profile very similar to that observed in the AnxA2^−/− mouse. Dynamic, but not static, compliance was significantly decreased (Fig. 3D,E; P<0.0001 and P = 0.578, respectively), whereas both whole lung and tissue elastance were correspondingly increased (Fig. 3F,G; P = 0.0076 and P<0.0001, respectively). As in the AnxA2^−/− mouse, total dynamic resistance was increased in the Col6a1^−/− lung (Fig. 3H; P<0.0001), but airway resistance showed no significant change (Fig. 3I; P = 0.276). Tissue dampening was also increased (Fig. 3J; P<0.0001), whereas inspiratory capacity was not significantly affected (Fig. 3K; P = 0.0639). Because Col6a1^−/− mice lack immunodetectable triple helical COL6 (Bonaldo et al., 1998), these data indicate that loss of mature COL6 leads to a significant reduction in dynamic compliance and an increase in hyperelastic recoil, thus providing a potential explanation for the pulmonary dysfunction in the AnxA2^−/− mouse.

In evaluating nonciliated BECs in lung tissue from resting mice, electron-dense secretory granules were almost twice as abundant in AnxA2^−/− BECs compared with AnxA2^+/+ BECs (Fig. 4A). To determine the potential role of ANXA2 in secretion of COL6, we stained AnxA2^+/+ and AnxA2^−/− primary mouse embryonic fibroblasts (mEFs) at 90% confluence with antibody directed against COL6 (Fig. 4B). Immunoreactive material was identified within 200–1000 nm punctate structures that were, on average, 4-times more abundant in AnxA2^−/− than AnxA2^+/+ cells (62.7±2.9 vs 16.5±3.2, mean ± s.e., n = 6–8, P<0.001). These data were consistent with the increased abundance of electron-dense secretory granules in AnxA2^−/− BECs observed by electron microscopy (Fig. 4A).

We assessed steady state synthesis of COL6 in AnxaA2^+/+ and AnxaA2^−/− lung tissue. Semi-quantitative and quantitative real-time RT-PCR analyses of Col6a1, Col6a2 and Col6a3 transcripts, which encode the COL6 alpha1, alpha2 and alpha3 isoforms, respectively, revealed no significant differences in transcript levels (Fig. 4C). However, immunoblots of reduced whole-lung homogenates revealed that total protein levels of COL6a1 and COL6a2 were diminished in AnxA2^−/− lungs (Fig. 4D). Interestingly, no difference in the abundance of protomeric, intracellular COL6, which is soluble in Triton X-100 (Engvall et al., 1986), was detected, whereas highly crosslinked extracellular COL6, which is insoluble in Triton X-100 (Engvall et al., 1986), was markedly reduced in AnxA2^−/− compared with AnxA2^+/+ tissues (Fig. 4E). Based on these data, we concluded that the reduction in total COL6 protein in AnxA2^−/− lung tissue reflected a defect in its post-transcriptional processing, deposition or stability within the basement membrane.

To identify the location of COL6 within AnxA2^−/− cells, we probed subcellular microsomal and cytosolic fractions by immunoblotting. In AnxA2^+/+ lung homogenate, ANXA2 was present, as expected, in both compartments (Fig. 4F). However, although some COL6 was present in AnxA2^−/− cytosol, it was highly enriched in AnxA2^−/−, but not AnxA2^+/+, microsomes. Blotting for calnexin (CNX), a microsomal marker, and lactate dehydrogenase (LDH), a cytosolic marker, demonstrated complete separation of the two fractions. To investigate whether microsome-associated COL6 and ANXA2 occupied intra- or extravesicular compartments, we next subjected microsomes from lung homogenates to limited proteolysis using a trypsin-protection assay (Fig. 4G). Whereas full-length, native COL6 is sensitive to tryptic digestion (not shown), microsomal COL6 was not affected by tryptic hydrolysis. However, microsome-associated ANXA2 disappeared almost completely within 60 minutes. From these data, we conclude that COL6 accumulates within a trypsin-protected, microsomal environment, whereas ANXA2 occupies an unprotected, but neighboring location.

To address the potential physical association between ANXA2 and COL6, we conducted immunoprecipitation assays. Anti-COL6 IgG, but not isotype-matched control IgG, precipitated both COL6 and ANXA2 from whole AnxA2^+/+ lung extracts, and COL6 from whole AnxA2^−/− lung extracts (Fig. 5A). Similarly, anti-ANXA2 IgG immunoprecipitated both ANXA2 and COL6 from whole AnxA2^+/+, but neither protein from AnxA2^−/− lung extracts (Fig. 5B). These results suggest that ANXA2 and COL6 undergo a physical association that could be either direct or indirect.

To further evaluate the physical proximity between ANXA2 and COL6 within the bronchiolar epithelial cell in vivo, we conducted immunoelectron microscopy studies. In BECs from lung tissue of resting mice, anti-ANXA2- and anti-COL6-conjugated immunogold beads (10 nm and 6 nm diameter, respectively) co-associated in clusters located at the membrane borders of large, intracellular vesicles enclosing electron dense material (Fig. 5C). Parallel analyses in AnxA2^−/− BECs revealed the absence of anti-ANXA2-labeled beads, and a reduction in the total number of clustered anti-COL6-conjugated beads. Within AnxA2^+/+ bronchial epithelial cells, we found an average of six COL6 beads per ANXA2–COL6 cluster, whereas clusters in AnxA2^−/− BECs contained an average of only two 6 nm beads (Fig. 5D). In addition, whereas 71% of COL6 beads in AnxA2^+/+ BECs were associated with membrane-bound, secretory vesicle-like structures, only 31% were vesicle associated in AnxA2^−/− BECs (Fig. 5E). These data also suggest a physical association between ANXA2 and COL6 within secretory vesicles.

To analyze COL6 secretion in the presence and absence of ANXA2, we transduced primary AnxA2^−/− mEFs with adenoviral constructs encoding either GFP alone (AdV-null) or a GFP–ANXA2 fusion construct (AdV-A2) (Fig. 5F). At 72 hours, immunoblot analysis revealed that COL6 was expressed within AnxA2^+/+ and AnxA2^−/− cells, regardless of viral transduction (Fig. 5F, left panel). COL6 was also detected in matrix secreted by AnxA2^+/+ cells or AnxA2^−/− cells transduced with AdV-A2, but not in matrix produced by AnxA2^−/− cells or AnxA2^−/− cells transduced with AdV-null. In the matrix, there was, on average, 7-times more COL6 in AnxA2^−/− AdV.rA2 samples compared with AnxA2^−/− samples, and 14-times more COL6 compared with AnxA2^−/− AdV-null samples (4.2±0.9 versus 0.6±0.3 and 4.2±0.9 versus 0.3±0.1 arbitrary units, respectively, mean ± s.e., n = 3, P<0.05). There was no significant difference between COL6 in AnxA2^−/− AdV.rA2 samples, compared with AnxA2^+/+ samples (P>0.1) (Fig. 5F, right panel). Similarly, primary AnxA2^+/+ mEFs transduced with a GFP–COL6 fusion construct secreted GFP into the extracellular space, where it was detected by Rhodamine-conjugated anti-GFP following fixation under nonpermeabilizing conditions (Fig. 5G, top panel). By contrast, extracellular GFP tag was not detected in association with AnxA2^−/− mEFs transfected with the same construct (Fig. 5G, bottom panel). Cells transduced with a construct encoding GFP alone did not release GFP into the extracellular milieu (not shown). Together, these data demonstrate that expression of ANXA2 is required for efficient secretion of COL6 by primary mEFs.

In cultured neuroendocrine and type II alveolar pneumocytes, ANXA2 has been reported to interact with members of the SNARE family of membrane fusion proteins (Umbrecht-Jenck et al., 2010; Wang et al., 2007). In pull-down assays in AnxA2^+/+ whole-lung extracts, similarly, we found that anti-COL6 IgG specifically precipitated ANXA2, as well as SNAP-23, VAMP2 and syntaxin 2 (Fig. 6A). In parallel samples from AnxA2^−/− lung extracts, anti-COL6 precipitated SNAP-23, VAMP2 and syntaxin 2, indicating that the association between COL6, SNAP-23, VAMP2 and syntaxin2 was independent of ANXA2. Immunoelectron microscopy studies, furthermore, revealed that ANXA2 specifically colocalized with both SNAP-23 and VAMP2 in BECs from AnxA2^+/+ lungs, whereas fewer SNAP-23 and VAMP2-related beads (52±1%, n = 10 versus 14±1%, n = 16; means ± s.e.) were associated with secretory-vesicle-like structures in AnxA2^−/− BECs (Fig. 6B). Together, these data demonstrate that ANXA2 associates with SNARE complexes, thereby enabling secretion of COL6.

We used standard markers for specific intracellular compartments to analyze the location of COL6 at multiple points along the secretory pathway in AnxA2^+/+ and AnxA2^−/− mouse embryonic fibroblasts (mEFs) (Fig. 6C and supplementary material Fig. S3C,D). We found no colocalization of anti-COL6 immunoreactive material and MitoTracker in either AnxA2^+/+and AnxA2^−/− mEFs (supplementary material Fig. S3C). In addition, we found slight colocalization of anti-COL6 immunoreactivity with the endoplasmic reticulum (ER) marker GRP94, but this did not differ between AnxA2^+/+ and AnxA2^−/− mEFs (Fig. 6C). Similar results were obtained upon staining for a second ER marker, anti-calnexin (supplementary material Fig. S3D). However, staining for the trans-Golgi marker (TGN38) revealed significant colocalization with anti-Col6 reactivity within AnxA2^−/− cells, but only minimal colocalization in AnxA2^+/+ cells (Fig. 6C). Finally, staining for VAMP2, generally considered to reside within post-Golgi secretion compartments, colocalized with anti-COL6 in a manner that was much more prominent in AnxA2^−/− than in AnxA2^+/+ mEFs. At the same time, immunoelectron microscopy of BECs in situ showed COL6 to be present in VAMP2-positive electron-dense structures in AnxA2^+/+, but not AnxA2^−/−, cells (supplementary material Fig. S4A,B). Together, these data indicate that COL6 is retained within a late-Golgi VAMP2-positive compartment.

By immunofluorescence microscopy, siRNA knockdown of SNAREs known to interact with ANXA2 led to accumulation of intracellular COL6 (Fig. 7). Specifically, and as seen in AnxA2^−/− cells, we found increased association of anti-COL6 and anti-TGN38 immunoreactivity in cells depleted of either SNAP-23 (Fig. 7A,B) or VAMP2 (Fig. 7C,D). These data indicate that both SNAP-23 and VAMP2 play a functional role in the secretion of COL6.

We observed abnormalities in both BECs and their basement membranes, in association with a reduced proportion of ciliated cells in the AnxA2^−/− mouse (Fig. 2A–G). To determine whether an alteration in the bronchiolar basement membrane compromises epithelial cell adhesion and/or survival in the AnxA2^−/− mouse, we carried out bronchoalveolar lavage (BAL) (Fig. 8A). BAL fluid from AnxA2^+/+ and AnxA2^−/− mice contained equivalent concentrations of cells (152±14 vs 158±14 cells/µl, respectively; mean ± s.e., n = 6) with similar total volumes of BAL fluid retrieved. Differential cell counts, however, revealed an excess of epithelial cells in fluid recovered from AnxA2^−/− mice (34±12 vs 10±4%, n = 6, P = 0.001), and a corresponding reduction in macrophages (62±10 vs 87±3%, n = 6, P = 0.0002). This result suggested a higher rate of bronchial epithelial cell disadhesion in response to the lavage procedure.

Loss of cell–matrix adhesion might induce epithelial cell apoptosis (anoikis) (Frisch and Screaton, 2001; Taddei et al., 2012). We, therefore, conducted TUNEL staining of lung sections, which revealed dramatically increased nucleic acid breakage within bronchial epithelial cells of AnxA2^−/− compared with AnxA2^+/+ lungs (Fig. 8B). Further staining revealed that AnxA2^−/− bronchiolar epithelial cells contained twice as many activated caspase-3-positive epithelial cells than AnxA2^+/+ bronchioles, confirming an increased rate of apoptosis (Fig. 8C,D).

To determine whether the secretion-enabling role of ANXA2 is unique to the bronchial epithelial cell or seen in other tissues, we examined skeletal muscle from AnxA2^+/+ and AnxA2^−/− mice (Fig. 8E). In this tissue, interstitial fibroblasts are the main source of COL6 (Bönnemann 2011; Zou et al., 2008). Interestingly, immunofluorescence staining revealed an accumulation of COL6 between individual myofibers cut in cross-section, the precise location where interstitial fibroblasts are located. This finding is consistent with our studies on mEFs, which also accumulated and failed to secrete COL6 (Fig. 4B and Fig. 5F,G). These data suggest that ANXA2-mediated secretion of COL6 is not limited to the bronchial epithelial cell, but also occurs in tissue fibroblasts.

To determine whether cell adhesion to extracellular matrix and basement membrane might be COL6-dependent, we conducted a series of cell-adhesion assays using (Fig. 8F). Whereas 60–70% of cells of both genotypes adhered well to matrices produced by AnxA2^+/+ mEFs, only 20–30% of each genotype adhered to matrices produced by AnxA2^−/− mEFs. In addition, AnxA2^+/+ mEFs were superior to AnxA2^−/− mEFs in adhering to uncoated or BSA-coated polystyrene. Interestingly, however, supplementation of COL6 completely corrected the defect in AnxA2^−/− mEF adhesion to BSA. These data indicate that COL6 supports cell adhesion to extracellular matrix substrates, and suggest that the reduced matrix adhesion displayed by AnxA2^−/− bronchial epithelial cells is associated with a deficiency of COL6.

These data reveal for the first time that deficiency of ANXA2 results in aberrant secretion of COL6, but not other collagens, from bronchial epithelial cells into basement membranes. Failure to secrete COL6 is associated with attenuated cell–matrix adhesion, apoptosis and physiologically significant pulmonary stiffness with reduced exercise tolerance. In the AnxA2^−/− lung, histological studies revealed bronchial epithelial cell dysmorphology and apoptosis, together with an altered basement membrane that lacked COL6. Despite equivalent mRNA levels for COL6 in AnxA2^−/− and AnxA2^+/+ lung tissue, we found diminished overall COL6 protein levels, reduced matrix-associated COL6 and retention of COL6 within a trypsin-protected, microsomal fraction in AnxA2^−/− lung cells. In cultured AnxA2^−/− fibroblasts, moreover, we uncovered a block to COL6 secretion that could be reversed by restoration of ANXA2 expression. Co-immunoprecipitation and immunoelectron microscopy experiments revealed that ANXA2 and COL6 form a complex with the SNARE proteins SNAP-23, VAMP2 and syntaxin 2, which are all components of the epithelial exocytotic machinery. In the absence of ANXA2, COL6 accumulated in a late-Golgi, VAMP2-positive compartment, strongly suggesting impaired fusion of secretory vesicles with the plasma membrane. We found a similar accumulation of COL6 in cells in which either SNAP-23 or VAMP2 had been depleted. Together, these data indicate that ANXA2 promotes bronchial epithelial cell survival and normal biomechanics in the mouse lung, by mediating a unique secretion pathway for macromolecular COL6.

An important question is whether the impairment of COL6 secretion in the AnxA2^−/− mouse is sufficient to explain the observed respiratory dysfunction and exercise intolerance. Humans with Ullrich congenital muscular dystrophy, which can result from mutations in the COL6A1, COL6A2 or COL6A3 loci, exhibit a restrictive ventilatory defect that is frequently out of proportion to the degree of myopathy. In a recent study, nearly 50% of still-ambulatory patients suffered from respiratory insufficiency, a distinct difference from Duchenne muscular dystrophy in which respiratory failure typically follows loss of ambulation (Nadeau et al., 2009). Similarly, 16% of patients with Bethlem myopathy, all of whom were still ambulatory, had a forced vital capacity that was less than 70% of the predicted value (van der Kooi et al., 2006). These studies suggest that the earlier development of respiratory insufficiency in disorders due to COL6 mutations might reflect loss of pulmonary compliance in addition to compromise of muscles of respiration.

In our pulmonary function studies, elastance and compliance were more severely perturbed in AnxA2^−/− mice compared with Col6a1^−/− mice. This result suggests that ANXA2 might play a larger role in regulating secretion of additional matrix proteins that must undergo high-level intracellular assembly. Our unpublished data are consistent with this possibility, and indicate that at least one additional basement membrane protein, fibulin-2, is abnormally secreted. Because COL6 secretion involves a high level of pre-assembly that is not required for secretion of other collagens or most other matrix proteins, it is perhaps not surprising that COL6 must be prepackaged within specialized secretory vesicles (Bateman et al., 2009). COL6 is composed most often of three genetically distinct polypeptide chains (COL6a1, COL6a2 and COL6a3), which assemble intracellularly into monomers that form higher-order polymers (Furthmayr et al., 1983). Antiparallel pairing of monomers gives rise to disulfide-stabilized dimers, which then form 12-chain complexes. This extensive polymerization occurs prior to secretion within secretory vesicles (Colombatti and Bonaldo, 1987; Colombatti et al., 1995; Engvall et al., 1986) and mutations that perturb COL6 assembly, as in Bethlem myopathy, also prevent its secretion (Lamandé et al., 1999).

ANXA2 has been found to participate in the regulated exocytosis of at least three classes of molecules through interactions with SNARE proteins. During release of chromaffin granule-associated catecholamines, ANXA2 translocates in an S100A10-dependent manner to a SNARE protein complex, consisting of SNAP-25, VAMP2 and syntaxin, at points of plasma-membrane-secretory granule contact (Creutz et al., 1978; Umbrecht-Jenck et al., 2010). In the endothelial cell, ultralarge (∼20,000 kDa) multimers of the procoagulant protein von Willebrand Factor (vWF) are pre-packaged within specialized structures, the Weibel–Palade bodies, and released by exocytosis in response to specific stimuli, such as thrombin, histamine or complement factors that trigger fusion with the plasma membrane (Pimanda and Hogg, 2002; Sadler, 2009). In primary cultures of endothelial cells, this process requires SNARE proteins, including SNAP-23, syntaxin 4 and VAMP3, as well as the annexin-A2–S100A10 heterotetramer (Knop et al., 2004; Rojo Pulido et al., 2011). In the lung, type II alveolar pneumocytes release surfactant, a complex of phospholipids and proteins SP-A and SP-D, upon fusion of lamellar bodies with the plasma membrane (Dietl et al., 2012). In this process, ANXA2 binds to SNAP-23 in a Ca²⁺-dependent manner (Wang et al., 2007). Thus, it appears that ANXA2 might be among the diverse proteins that regulate or support SNARE-related exocytosis (Kasai et al., 2012; Rizo and Südhof, 2012).

In the biosynthetic-secretory pathway, proteins shuttle from the ER to the Golgi in 60–90 nm coat protein complex (COP)II-coated vesicles, and then from the Golgi to the plasma membrane in similar-sized COPI-coated vesicles, prior to vesicle fusion with the plasma membrane. However, to accommodate COL7, and possibly other collagens, which form 300–400 nm triple-helical rod-like ‘monomers’, proteins such as TANGO (transport and Golgi organization) and cTAGE5 (cutaneous T-cell lymphoma-associated antigen 5) appear to assemble at ER-exit sites and delay COPII membrane coating, resulting in larger-diameter vesicles (Malhotra and Erlmann, 2011). Our data suggest that an additional post-Golgi adaptation facilitates export of COL6, which undergoes even higher-level, tetrameric assembly. We propose a model in which intravesicular COL6 interacts with VAMP2, a transmembrane vesicular protein, which in turn interacts with extravesicular SNAP-23 and ANXA2. ANXA2 appears to enable plasma membrane docking of vesicles and/or fusion by interacting with SNAREs in the final step of COL6 secretion (Kasai et al., 2012). In the absence of ANXA2, membrane fusion fails and COL6 accumulates within VAMP2-positive late-Golgi vesicles, thus defying export. This mechanism might be unique to COL6, because COL1 and COL4 appear to be normally localized in the bronchial epithelial basement membrane of the AnxA2^−/− mouse.

The high level of apoptosis in AnxA2^−/− bronchial epithelial cells suggested cell loss that is due to anoikis, whereby disadhesion induces apoptotic cell death in anchorage-dependent cells (Chiarugi and Giannoni, 2008; Frisch and Screaton, 2001; Taddei et al., 2012). Interestingly, soluble COL6 appears to prevent apoptosis of serum-starved fibroblasts by suppressing the pro-apoptotic protein, Bax (Rühl et al., 1999). In avian corneal fibroblasts, cell viability depends specifically on the interaction between matrix-associated COL6 and cell-membrane-associated β1 integrins (Howell and Doane, 1998), and COL6 also appears to protect chondrocytes, neurons and cardiac myocytes from apoptosis due to chemical inducers, ultraviolet irradiation or early, but not late, tissue infarction, respectively (Cheng et al., 2011; Luther et al., 2012; Peters et al., 2011). In Col6a1^−/− mice, myofibers undergo apoptotic cell death in a process thought to reflect opening of the cyclosporine-A-sensitive, mitochondrial permeability transition pore, which might be triggered by integrin-mediated induction of reactive oxygen species or defective regulation of autophagy (Grumati et al., 2010; Irwin et al., 2003).

In our studies in the AnxA2^−/− mouse, changes in bronchiolar epithelial cells were among the most dramatic pulmonary findings. These cells displayed a dense, flattened appearance, a high apoptotic index and a striking reduction in the proportion of ciliated cells, possibly indicating replacement by Clara-like progenitor cells. Further studies confirmed dropout of airway epithelial cells in response to bronchoalveolar lavage, and experiments with embryonic fibroblasts revealed impaired adhesion of cells of either genotype to AnxA2^−/− mEF-derived matrix. Impaired adhesion of AnxA2^−/− mEFs to albumin–matrix substrate could be restored to the level seen for AnxA2^+/+ mEFs by addition of COL6. This is a finding consistent with previous studies, indicating that cellular adhesion to COL6 is dependent upon β₁ integrins (Pfaff et al., 1993). Furthermore, the COL6-related loss of cells secreting basement membrane could compromise deposition of additional normal matrix proteins that contribute to normal pulmonary compliance and elastance. In summary, our data show that ANXA2 is fundamental to secretion of COL6, subsequent cell adhesion to basement membrane, prevention of apoptotic cell dropout and normal pulmonary function.

AnxA2^−/− and AnxA2^+/+ mice were generated by homologous recombination as described previously (Ling et al., 2004). Either littermate pairs or mice cross-bred extensively to the C57BL/6 background were used. Col6a1^−/− mice, which lack α chain expression and fail to form triple-helical COL6 monomers, were generated as described and bred to a mixed FVB-C57BL/6 background (Bonaldo et al., 1998). All animal procedures were performed in accordance with Institutional Animal Care and Use Committee approval.

Swim testing took place in a plexiglass tank containing stationary room-temperature (23±1°C) water at a depth of 20 cm. After swimming, mice were placed in cages containing normal bedding and maintained at 21°C. AnxA2^−/− and AnxA2^+/+ littermate mice underwent a 3 week training period consisting of supervised swim sessions 4–5 times per week and increasing in duration by 10 minute intervals to a plateau of 90 minutes per session (Godínez-Victoria et al., 2012; Kregel et al., 2006). Thereafter, mice were subjected to 90 minute daily swim sessions over a 3 week period. Recovery time, defined as duration from removal from the swim tank to the time of complete coat dryness, was recorded by a masked observer (Meeks and Larson, 2012). At the conclusion of each experiment, the mice were sacrificed and heart to body weight ratios determined.

After 3 days of acclimatization to handling, prewarmed ultrasound gel was applied to the depilated anterior chest of seven AnxA2^−/− and seven AnxA2^+/+ mice (Donovan et al., 2000). An Acuson Sequoia™ C512 15-MHz linear array probe (Siemens Medical Solutions, Malvern, PA) was placed on the anterior chest wall and the position of the transducer confirmed using two-dimensional long- or short-axis images. An M-mode cursor was then placed at the level of the papillary muscles and M-modes recorded at a sweep speed of 200 cm/second. Left ventricular dimensions in diastole (LVIDd) and systole (LVIDs), as well as septal and posterior wall thickness and heart rate, were measured over four to six cycles using the Digisonics Offline Analysis System. Left ventricular volumes in diastole (VolD) and systole (VolS) were calculated using a cubed formula. Fractional shortening was calculated using the formula, FS (%)  =  100 × (LVIDd-LVIDs)/LVIDd. Stroke volume (SV) was calculated as the difference between VolD and VolS. Ejection fraction (EF) was calculated using the formula: EF (%)  =  100 × (VolD−VolS)/VolD. Cardiac output was derived from the formula SV × HR. Left ventricular mass was calculated using the prolate ellipse formula (Lang et al., 2005).

Respiratory mechanics were evaluated on anaesthetized mice (125 mg/kg pentobarbital, 15 mg/kg xylazine; i.p.) that were tracheostomized and mechanically ventilated using a computer-controlled animal ventilator (Sireq, Montreal, Canada). Three perturbations were used to assess baseline respiratory mechanics and were analyzed using Flexivent software (Scireq). Static compliance was determined using the Salazar–Knowles equation to the plateau pressure measurements obtained between total lung capacity and functional residual capacity. Single-frequency forced oscillations were used to determine tissue resistance (R), elastance (E) and dynamic compliance (C) using a single compartment model. Broadband forced oscillations served to determine Newtonian (airway) resistance (R_n), tissue dampening (G) and tissue elastance (H) using a constant phase model. R_n was also assessed following increasing doses of nebulized methacholine (3.125, 12.5 and 50 mg/ml). Inspiratory capacity was determined between the plateau pressure measurements of the total lung capacity and functional residual volume (Ding et al., 2011).

Sections were initially stained with hematoxylin and eosin. Unstained sections were also imaged by autofluorescence using a green FITC-filter channel and UV excitation. For immunohistology, 4% paraformaldehyde (PFA)-fixed frozen lung sections were treated with blocking solution (DPBS/5% BSA, 1 hour), and then incubated overnight in blocking solution containing rabbit polyclonal anti-human ANXA2 IgG (clone H-50; Santa Cruz Biotech), or containing goat polyclonal anti-mouse COL6 IgG (clone S-16; Santa Cruz Biotech). Samples were washed and incubated with donkey anti-rabbit or anti-goat IgG coupled to Cy3, Alexa Fluor 569 or Alexa Fluor 647 (Jackson Immunochemicals and Invitrogen). Coverslips were mounted with Vectashield mounting medium (Vector Laboratories, Inc., Burlingame, CA). Mounted coverslips were imaged by fluorescence microscopy (Retiga 1300i), or by confocal microscopy (Zeiss LSM 5 Live). For immunochemistry, 4% PFA-fixed frozen lung tissue sections were treated as above, and incubated with secondary donkey anti-rabbit or anti-goat peroxidase-conjugated antibodies (blocking solution, 1 hour), rinsed, and counterstained with DAPI.

For detection of apoptosis, in situ terminal deoxynucleotidyl transferase-mediated DNA nick-end labeling (TdT-TUNEL) staining was performed using the TACS 2 in situ kit (Trevigen Inc., Gaithersburg, MD). Briefly, fixed lung sections were incubated with TdT Blue labeling buffer (0.01% Thimerasol, 0.5% BSA, <0.01% 2-mercaptoethanesulfonic acid, TES sodium salt, pH 7.5; 5 minutes), followed by either labeling reagents (10 µM biotinylated dNTP, terminal deoxynucleotidyl transferase (TdT) enzyme, and Co²⁺ cation) or TACS-nuclease for positive controls (1 hour, 37°C). The reaction was stopped with termination buffer (10 mM EDTA, pH 8.0, 15 minutes, 21°C) and blue labeling of cell nuclei imaged. For identification of cleaved caspase-3 activity, fixed lung cryosections were blocked (5% BSA) and incubated with rabbit monoclonal anti-human activated caspase 3 (Cell Signaling), followed by polyclonal goat anti-E-cadherin IgG (R&D Systems). Secondary antibodies were Cy5-labeled goat anti-rabbit and Cy3-labeled donkey anti-goat IgG (Jackson Immunochemicals).

RNA was extracted from lung tissue with Trizol (Invitrogen) and purified by standard phenol–chloroform and ethanol precipitation. To generate cDNAs from extracted RNA, reverse primers complementary to a portion of Col6a1, Col6a2 and Col6a3 were used in reverse transcriptase reactions. Semi-quantitative PCR was performed with the resulting cDNA templates, with primer pairs corresponding to exons 1 and 2 in Col6a1 and Col6a2, exons 1 and 3 in Col6a3, and Taq polymerase (Denville Scientific, NJ). Primers for GAPDH were used as internal controls (primer sequences available upon request). PCR cycles were as follows: 94°C, 5 minutes for denaturation, 32 cycles of 94°C, 1 minute; 59°C, 90 seconds for annealing; and 72°C, 2 minutes for extension. For real-time PCR, LightCycler 480 reaction mastermix was prepared according to the manufacturer's protocol (SYBR I Green Master; Roche Diagnostics) and added to 1 μl cDNA in individual wells of a 384-well plate. PCR cycles were as follows: 95°C for 5 minutes (denaturation), 95°C for 30 seconds, 60°C for 30 seconds, 76°C for 30 seconds (amplification and quantification) for 40 cycles, 60°C–95°C with a heating rate of 0.1°C per second and a continuous fluorescence measurement (melting curve) and 40°C (final cooling). Relative expression ratios were determined by ‘Fit Point Method,’ using LightCycler 480 Software (Roche Diagnostics) (Pfaffl, 2001).

Protein samples were analyzed on 12.5% SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose membranes (18 hours, 120 mA, 4°C), which were then blocked (5% milk, 1 hour, 21°C), and probed with rabbit polyclonal IgGs directed against ANXA2 (clone H-50; Santa Cruz Biotech), mouse tropoelastin (Abcam), human β-tubulin (Abcam), mouse SNAP23 (clone G-16) (Santa Cruz Biotech), rat VAMP2 (Abcam), rat syntaxin-2 (Abcam) or mouse monoclonal IgG directed against human β-actin (Santa Cruz Biotech). Membranes were developed with peroxidase-conjugated anti-rabbit and anti-goat IgGs (GE Healthcare), and ECL Plus detection reagents (GE Healthcare).

Primary cultures of mouse embryonic fibroblasts (mEFs) were established from E14.5 embryos that were rinsed with 1× PBS, minced with a razor blade, and incubated in 0.1% trypsin, 1 mM EDTA (37°C, 15 minutes, gentle agitation). The resulting supernatant was centrifuged (200 g, 5 minutes) and the pellet resuspended in prewarmed mEF medium (DMEM, 10% fetal bovine serum (FBS), 0.01% L-glutamine, 0.01% sodium pyruvate, and 0.01% penicillin-streptomycin). For transfection, mEFs were incubated with mouse cDNA expression plasmids encoding GFP-tagged versions of COL6 (accession no. NM_009933.4; OriGene, Rockville, MD) using TurboFectin 8.0 reagent (OriGene) per the manufacturer's guide. For adenoviral infection, mEFs were incubated with an adenoviral construct containing recombinant ANXA2 and a GFP marker (AdV-A2), or an empty GFP-encoding control vector (AdV-null), in serum-free mEF medium (2 hours, 37°C). Twenty-four hours post infection, fresh complete mEF medium was added, and cells incubated for 48 hours.

mEFs were transfected with either mock, SNAP-23- or VAMP2-directed siRNA (Santa Cruz Biotech) in siRNA Transfection Medium, as per the manufacturer's protocol (Santa Cruz Biotech). Five to seven hours later, complete medium containing twice the normal serum and antibiotic concentration was added. Twenty-four hours post transfection, the medium was replaced with normal growth medium, and the cells assayed 24 hours later.

For immunocytology, cells were fixed with 2% PFA and 2.5% sucrose in DPBS plus calcium and magnesium (C/M; 15 minutes, 37°C). Cells were blocked with either DPBS C/M plus MOM (Vector Laboratories) or DPBS C/M with 5% normal donkey serum, each containing 0.3 M glycine, and 0.1% Tween-20. Cells were blocked for 1 hour, and then incubated with primary antibody (2 hours, 37°C), or overnight in blocking solution plus primary antibody (supplementary material Table S2). Samples were washed and incubated with secondary antibody (supplementary material Table S3). Coverslips were mounted with Fluoromount G medium (Electron Microscopy Sciences, Hatfield, PA), and imaged by fluorescence (Retiga 1300i) or confocal microscopy (Zeiss LSM 5 Live). Fluorescence signals were quantified using Image J (1.48c, National Institutes of Health, USA).

Mouse lungs were flash frozen in liquid N₂ and ground manually into a fine powder with a mortar and pestle. Washed, cultured cells were scraped and pelleted (200 g, 21°C). Pelleted cells or tissue powder was resuspended in lysis buffer [25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 mM complete mini protease inhibitor cocktail (Roche), 1 mM DTT, 1 mM PMSF]. Lysates were passed sequentially through 26.5-, 28- and 30-gauge needles, nutated (4°C, 2 hours) and then centrifuged (16,060 g, 4°C, 15 minutes) for collection of supernatants. Pellets containing Triton-X-insoluble proteins were extracted with ECM buffer (8 M urea, 500 mM NaCl, 50 mM Tris-HCl, pH 7.5, 0.5% Triton X-100, 1 mM complete mini protease inhibitor cocktail, 0.02 mM PMSF). Protein concentrations were estimated using the bicinchoninic acid assay (BCA).

Flash-frozen mouse lungs were Dounce-homogenized (30 strokes, conical microcentrifuge pestle, Thomas Scientific) in cell extraction buffer (50 mM Tris-HCl, pH 7.3, 250 mM sucrose, 2 mM EDTA, 2 mM protease inhibitor cocktail), and centrifuged twice (2800 g, 4°C, 20 minutes) to remove unbroken cells and nuclei. Ice-cold aqueous 0.1 M Na₂CO₃ (pH 11) was added to the supernatant homogenate and agitated gently (4°C, 1 hour), prior to two additional centrifugations (100,000 g, 1 hour, 4°C). Pooled supernatants (cytosol) and pellets (microsomes) were resuspended in cell extraction buffer containing 8 M urea and analyzed by immunoblot.

Protein extracts (750 µg) from whole-lung homogenates were mixed with pre-washed protein A/G beads (Santa Cruz Biotech), resuspended at 100 µg/µl and coupled with 10 µg of rabbit polyclonal anti-human COL6 (Rockland) or rabbit polyclonal anti-human ANXA2 antibodies (Santa Cruz Biotech) and incubated (5 hours, 4°C). Immunocomplexes were washed four times with cell lysis buffer and once with DPBS, and then analyzed by immunoblot.

Microsomes (75 µg) underwent limited digestion with trypsin (Worthington, 0.1 µg/ml, 24°C, 60 min) (Sarkadi et al., 1986). Reactions were quenched with 1 mM protease inhibitor cocktail and 1 mM PMSF. Untreated and trypsin-treated samples were analyzed by SDS-PAGE and immunoblot.

To determine areas of interest for electron microscopy imaging, ultra-thin 1.0 µm tissue sections were fixed in 4% PFA and 0.1% glutaraldehyde in DPBS (Cellgro, Manassas, Virginia), stained with Toluidine Blue, rinsed with dH₂0 and imaged by brightfield microscopy (Retiga 1300i). For electron microscopy imaging, the sections were post-fixed in 1% OsO₄, contrasted in 1% uranyl acetate, dehydrated and embedded in acrylic resin. For immunogold labeling, ultrathin sections were transferred to grids, incubated with either primary polyclonal rabbit anti-human ANXA2 IgG or goat anti-human ANXA2 IgG (Santa Cruz Biotech), goat anti-mouse COL6 (Santa Cruz Biotech), goat anti-human SNAP-23 (Santa Cruz Biotech) or rabbit anti-rat VAMP2 (Abcam) in various combinations, and incubated with anti-rabbit 10-nm-gold-conjugated and anti-goat 6-nm-gold-conjugated antibodies (Aurion). Grids were examined in a JEM 1400 transmission electron microscope (JEOL Ltd, Japan).

PBS (0.5 ml) was delivered via endotracheal tube and then immediately withdrawn using gentle suction. This procedure was repeated once, and the recovered fluid, typically ∼600 µl, pooled and stained with Trypan Blue for total cell counts in a hemocytometer. Differential counts were performed on cytospins stained with Diff-Quik (Dade Behring, Düdingen, Switzerland); at least 500 cells were evaluated for each preparation (Walters et al., 2000).

Biosynthesized ECM was prepared by incubating AnxA2^+/+ and AnxA2^−/− mEFs on BSA-coated wells (8 hours, 37°C, 5% CO₂), with prewarmed mEF medium) until confluent. The cells were washed with 1 ml PBS C/M, treated with 1 ml lysis buffer (0.5% Triton X-100, 0.02 M NH₄OH), and rinsed once with PBS C/M before use. To evaluate cell adhesion, mEFs were plated (∼50,000 cells/well, 24-well plates, 8 hours) on uncoated polystyrene wells or wells coated with BSA, COL6 (Abcam) or biosynthetic ECM (Rickard et al., 1991). The supernatant medium was then removed, and unattached cells counted by hemocytometer. Attached cells were rinsed once with PBS C/M to remove loosely attached cells, collected after incubation with 1 ml 0.25% trypsin in EDTA (Cellgro; 2 minutes, 37°C), and counted.

Graphs and associated statistics were generated with GraphPad Prism software (v. 4.00, San Diego, CA) and Microsoft Excel 2007. Data are provided as mean ± s.e. Significance of differences was calculated using the Student's unpaired t-test.

We thank Ralph L. Nachman, Biin Sung, Andrew T. Jacovina, and Lee Cohen-Gould and Becky Smith for helpful discussions and assistance with pulmonary mechanics, exercise testing, and immunoelectron microscopy, respectively.