Annexins – insights from knockout mice

Introduction

Annexins represent a multigene family of evolutionary conserved and structurally related Ca²⁺ and phospholipid-binding proteins. They are found in a large variety of species and are classified into five groups (A–E), with 12 members (AnxA1–A13, AnxA12 is unassigned) making up the human and vertebrate annexin family (Moss and Morgan, 2004; Gerke et al., 2005). Purification and crystal structure analysis of AnxA5, and later other annexins, identified that all annexins are composed of two domains: a C-terminal conserved protein core and a variable N-terminus. The C-terminal core of all annexins consists of four (or eight in AnxA6) highly conserved annexin repeats, each with Ca²⁺ binding sites that enable binding to negatively charged phospholipids in a Ca²⁺-dependent manner (Figure 1). Numerous in vitro studies with purified annexins and model membranes provided the structural concepts that determine their biochemical properties, three-dimensional folding, Ca²⁺- and membrane binding characteristics. Their affinities to various phospholipids found in biological membranes, as well as interactions with other proteins, in particular members of the S100 family have also been determined (Bandorowicz-Pikula, 2005). In addition, the availability of antibodies, the cloning and overexpression of annexins, the development of recombinant peptides, and more recently, RNAi-mediated knockdown approaches, enabled a plethora of cell and animal studies addressing the localization and function of individual annexins.

Figure 1:

Domain structure of annexins.

The domain structures of annexins AnxA1, A2, A4, A5, A6 and AnxA7 are illustrated. The N-terminal leader domain (blue), C-terminal core with annexin repeats I–IV (and I–VIII for AnxA6) and length (in amino acids) are indicated.

Annexins rapidly translocate to the plasma membrane or intracellular membranes upon Ca²⁺ elevation (Gerke and Moss, 2002; Gerke et al., 2005; Grewal and Enrich, 2009). Hence, most of the postulated annexin functions have been linked to their dynamic, transient and reversible membrane binding behavior. However, their conserved structure, similar membrane binding properties, coinciding localizations and shared interaction partners have made it difficult to elucidate their precise functions. This implicated that redundancy within the annexin family may exist. It is now thought that specificity is conferred through the unique N-terminal domain of each annexin, sequence variations in their conserved Ca²⁺-binding modules, posttranslational modifications, different spatio-temporal and Ca²⁺-sensitive membrane binding kinetics. In addition, pH and other lipids, such as phosphatidylinositol-4,5-bisphosphate, cholesterol and ceramide, facilitate the opportunity for differential function and behavior of annexins (Gerke and Moss, 2002; Hayes et al., 2004; Rescher and Gerke, 2004; Gerke et al., 2005; Grewal and Enrich, 2009; Monastyrskaya et al., 2009).

Numerous gain- and loss-of-function experiments in cell culture have associated annexins with a wide variety of cellular processes relevant for proliferation, differentiation, apoptosis, migration, membrane repair and inflammatory response. As predicted, the proposed mechanisms are often related to Ca²⁺ signaling and membrane function. Although as outlined below, some prominent extracellular activities of several annexins may not require these aspects (Gerke et al., 2005). Overexpression studies were the initial method of choice. To date, the preferred methodology in a lot of laboratories is the RNAi-mediated downregulation of annexin gene expression. While these approaches have given valuable information in cell culture, targeted gene disruption in mice is considered to ultimately provide insight into protein function in vivo. Since 1999, mice deficient of AnxA1, A2, A4, A5, A6 and A7 have been produced (Gerke and Moss, 2002; Gerke et al., 2005) (Figure 1, Table 1). With the exception of one study, all mice strains lacking one or even two annexins were viable, enabling studies under various stress conditions that often revealed striking phenotypes. Several of those animal models support Ca²⁺ and membrane binding as a fundamental theme in annexin biology. In fact, phenotypes often center around the scaffolding function of annexins that enables organization of membrane domains, signaling platforms and the formation of complex protein networks (Gerke et al., 2005; Grewal and Enrich, 2009; Hoque et al., 2014). However, several KO strains also identified extracellular functions that appear independent of intracellular membrane trafficking or Ca²⁺ signaling events. The prominent phenotype of AnxA1 and AnxA2-deficient mice based on extracellular activities has been discussed in detail elsewhere (see below), while a comparison and discussion of other annexin KO models is still missing. In the following, we will summarize the research that utilized KO animals to examine annexin function (Figure 1, Table 1). Interestingly, and in contrast to the proposed redundancy within the annexin family, a substantial amount of studies utilizing various disease models identify underlying mechanisms that require functions that appear specific for individual annexins.

AnxA1 KO mice

AnxA1 was originally discovered in the late 1970s when aiming to identify factors that mediate the anti-inflammatory action of glucocorticoids (D’Acquisto et al., 2008; Gavins and Hickey, 2012). AnxA1 is found in most tissues with significant amounts commonly expressed in differentiated cell types, in particular macrophages and neutrophils, as well as cells from the nervous and endocrine system (Gerke et al., 2005; D’Acunto et al., 2014) (Table 1). Like several other annexins, AnxA1 is found at intra- and extracellular locations. Inside cells, AnxA1 is localized at the plasma membrane and on endosomal and secretory vesicles, but association with the cytoskeleton and nucleus has also been observed. In these locations, AnxA1 participates in cellular events relevant for proliferation, differentiation, migration and survival. This includes roles in endo- and exocytosis, signal transduction, cytoskeletal rearrangements and regulation of metabolic enzymes (Gerke and Moss, 2002; Hayes et al., 2004; Rescher and Gerke, 2004; Gerke et al., 2005; Grewal and Enrich, 2009; D’Acunto et al., 2014). However, based predominantly on cell culture but also in vivo studies utilizing recombinant full-length AnxA1, N-terminal AnxA1 fragments or neutralizing AnxA1 antibodies, extracellular activities of AnxA1 were identified to mediate multiple anti-inflammatory aspects of glucocorticoid action. This involves AnxA1 translocation to the cell surface upon cell activation, followed by secretion into the extracellular milieu. It is now thought that glucocorticoids positively regulate AnxA1 expression as well as AnxA1 translocation and secretion (Perretti and D’Acquisto, 2009). The transport mechanisms facilitating AnxA1 export from cells are not fully understood as AnxA1 lacks a signal sequence for secretion. However, depending on the cell type analysed, several non-conventional secretion pathways seem to contribute to the cellular export of AnxA1 (Gerke et al., 2005; Perretti and D’Acquisto, 2009; Gavins and Hickey, 2012).

After the successful generation of viable AnxA6 and AnxA7 KO strains (Hawkins et al., 1999; Herr et al., 2001; see below), Flower, Hannon and colleagues generated the AnxA1 KO mice (AnxA1^-/-) to provide proof-of-principle in vivo that AnxA1 confers glucocorticoid-dependent activation of innate immune cells to limit pro-inflammatory response (Hannon et al., 2003) (Table 2). AnxA1 KO mice were healthy and bred normally without obvious physical or behavioral difference compared to control animals. Interestingly, several other annexins, including AnxA2, A4, A5 and A6 were up- or downregulated in various tissues, indicating compensatory action through these annexins for some AnxA1-mediated physiological tasks. Furthermore, cyclooxygenase 2 (COX2) and cytoplasmic phospholipase A₂ (cPLA₂) levels were also elevated. The latter observations are in line with earlier models that implicated AnxA1 (and other annexins) to mediate anti-inflammatory actions through enzymes that control prostanoid synthesis (Hannon et al., 2003).

In these initial studies, two models of acute inflammation, carrageenan-induced paw edema and zymosan-induced peritonitis, served to evaluate the effects of AnxA1 deficiency. Strikingly, and supporting earlier in vivo studies with recombinant AnxA1 tools (Perretti and Dalli, 2009; Sugimoto et al., 2016), AnxA1^-/- mice displayed a boosted immune response. This was characterized by an increased leucocyte migratory behavior and a substantial resistance to the anti-inflammatory action of glucocorticoids (Hannon et al., 2003). Altogether, this prominent extracellular AnxA1 activity suggested the existence of an AnxA1 receptor. Indeed, Gerke and colleagues identified antagonists to seven-membrane-spanning G-protein coupled receptors (GPCRs) known as formyl peptide receptors (FPRs) that blocked the ability of AnxA1 and AnxA1 peptides to reduce neutrophil transmigration across monolayers of endothelial cells (Walther et al., 2000). This fundamentally new insight instigated further research that established the functional links between AnxA1 and FPRs. FPRs are expressed on a large variety of cell types, including neutrophils, monocytes, endothelial and epithelial cells, activating multiple signaling pathways. In cell models, all three members of the human FPR family showed activation by N-terminal AnxA1-derived peptides (Ernst et al., 2004). Then, after initial in vivo studies in mouse strains lacking the structurally related FPR1 (Perretti and D’Acquisto, 2009), analysis of the FPR2^-/- mice (Dufton et al., 2010; Dalli et al., 2013; Buss et al., 2015; Drechsler et al., 2015; Fredman et al., 2015; McArthur et al., 2015; Smith et al., 2015) provided more conclusive information that AnxA1 mediated innate immune reactions via FPR2 (also known as lipoxin A4 receptor or ALX) (Table 2). How the interaction of AnxA1 with FPR2 translates into the multiple and diverse molecular events that drive immune response is not fully understood, but activation of mitogen-activated protein kinases (MAPK), including extracellular signal-regulated kinases 1/2 (Erk1/2) and p38MAPK, Akt, c-Jun N-terminal kinase, increase of intracellular Ca²⁺ levels appear to be involved. Other contributing candidates include several chemokine receptors (Drechsler et al., 2015; Machado et al., 2016) and transcription factors (D’Acquisto et al., 2008; Perretti and D’Acquisto, 2009; Yang et al., 2013a).

Since the generation of the AnxA1 KO mice (Hannon et al., 2003), numerous studies employed this strain as a disease model to examine the anti-inflammatory roles of AnxA1 (Table 2). It would go beyond the scope of this review to summarize all this research. Thus we point the reader to excellent reviews with comprehensive lists of in vivo investigations using AnxA1^-/- mice or AnxA1-derived peptides from Perretti and colleagues (D’Acquisto et al., 2008; Perretti and D’Acquisto, 2009; Perretti and Dalli 2009), and more recently from others (Gavins and Hickey, 2012; Yang et al., 2013a; Sugimoto et al., 2016). Overall these studies have provided an enormous depth of information on the variety of AnxA1/FPR2-mediated immunological consequences in various cell types driving the innate immune response. In particular the decreased neutrophil adhesion to endothelium, increased detachment of adherent cells and inhibition of neutrophil transmigration are critical outcomes. In addition, AnxA1 induces apoptosis and clearance of neutrophils at inflammatory sites and promotes monocyte recruitment. Furthermore, AnxA1 downregulates the production of pro-inflammatory mediators including eicosanoids, nitric oxide, and several interleukins (ILs). More recently, AnxA1/FPR2 has also been linked to the adaptive immune response (D’Acquisto et al., 2008; Perretti and D’Acquisto, 2009; Gavins and Hickey, 2012; Yang et al., 2013a,b). However, a clear picture is still to emerge, as AnxA1 deficiency may have stimulatory or inhibitory effects on the adaptive immune system depending on the experimental settings and disease models analysed (D’Acquisto et al., 2008; Yang et al., 2013a,b).

Hence, the AnxA1/FRP2 complex has become a therapeutic target for the development of anti-inflammatory strategies. To date, administration of the AnxA1 derived N-terminal Ac2-26 peptide is the most common approach (D’Acquisto et al., 2008; Perretti and D’Acquisto, 2009; Perretti and Dalli, 2009; Gavins and Hickey, 2012; Yang et al., 2013a; Locatelli et al., 2014; Sugimoto et al., 2016). Overall, exogenous administration of AnxA1 or AnxA1-derived peptides are effective to limit or resolve inflammation in a large range of disease models in rodents, including stroke, myocardial ischemia, non-alcoholic steatohepatitis, rheumatoid arthritis, multiple sclerosis, colitis and asthma. In addition, AnxA1 deficiency is detrimental in epithelial wound repair (Leoni et al., 2013). Loss of AnxA1 also has a role in lung fibrosis (Damazo et al., 2011), obesity and insulin resistance (Akasheh et al., 2013), as well as adrenal steroidogenesis in sepsis (Buss et al., 2015) or allergic conjunctivitis (Gimenes et al., 2015). Given the limitations of peptide delivery for therapeutic use in vivo, efforts have been made to develop small molecule AnxA1 mimetics. These approaches remain difficult, as AnxA1 undergoes conformational change in the presence of Ca²⁺. Recent mapping of the AnxA1/FPR2 interaction domains indicate that AnxA1 may simultaneously interact with several amino acids within the N-terminal domain but also the extracellular loop II of FRP2 (Bena et al., 2012). It has yet to be determined how these multiple interaction sites contribute to differential signaling events depending on the cell type, physiological and micro-environmental context.

The understanding that AnxA1 promotes a resolving phase to counteract the pro-inflammatory response has revealed novel therapeutic applications for AnxA1 to delay or even regress the progression of chronic diseases (Perretti and D’Acquisto, 2009; Neymeyer et al., 2015; Soehnlein, 2015; Rackham et al., 2016). This has become increasingly evident in cardiovascular settings. For example, in low density lipoprotein receptor (LDLR) deficient mice on a western type diet, the intraperitoneal injection of human recombinant AnxA1 significantly reduced FPR2-dependent neutrophil rolling and adhesion to endothelial cells and attenuated progression of atherosclerotic plaques (Kusters et al., 2015). Conversely, crossing AnxA1- or FPR2-deficient strains with apolipoprotein E (apoE)^-/- mice enhanced atherosclerotic lesion formation and arterial myeloid cell adhesion. In vivo administration of Ac2-26 strongly reduced FPR2-dependent recruitment of myeloid cells, reduced atherosclerotic lesion size and macrophage accumulation in lesions. In this setting, the AnxA1/FRP2 axis inhibited chemokine-dependent integrin activation required for leukocyte recruitment (Drechsler et al., 2015). Along these lines, Fredman and coworkers packaged Ac2-26 onto nanoparticles that target collagen type IV, which is highly enriched in atherosclerotic plaque, thereby reducing mistargeting and systemic side effects. Advanced atheroslerotic lesions from mice receiving these nanoparticles showed reduced lesion instability in an FPR2-dependent manner (Fredman et al., 2015). Nevertheless, the mechanisms enabling AnxA1 to reduce the risk of lesion rupture are still not fully understood. Yet, the combination of nanotechnology-based novel delivery techniques with biological effects mediated by AnxA1-derived peptides may offer pharmacological means to treat very complex chronic diseases such as atherosclerosis.

As outlined above, the AnxA1/FPR2 axis has been explored extensively. Thus, to date only a limited number of studies have examined the AnxA1 KO strain in the context of intracellular AnxA1 locations and activities (Table 2). Interestingly, several consequences of AnxA1 deficiency inside cells might even potentiate the extracellular anti-inflammatory mode of AnxA1 action. On the one hand, altered activity of multiple signaling pathways in AnxA1^-/- macrophages contribute to increased production of inflammatory cytokines, such as interferon-β and tumor necrosis factor α (Bist et al., 2013; Nair et al., 2015). On the other hand, the secretome of mesenchymal stromal cells from the AnxA1 KO mice lacks the ability to stimulate glucose-induced insulin secretion of pancreatic β cells (Rackham et al., 2016). This indicates that compromised secretion upon AnxA1 deletion may also occur and contribute to other disease models investigated in the AnxA1 KO strain. In addition, an impact of AnxA1 on transcriptional activation, mRNA transport and stability has also been described (Bist et al., 2013; Nair et al., 2015). It is tempting to speculate that yet unknown AnxA1-dependent protein-protein or protein-lipid interactions that are lacking in the AnxA1 KO strain add to altered membrane trafficking in the secretory pathway.

Furthermore, AnxA1^-/- lung fibroblasts exhibited an accumulation of intracellular organelles and overexpressed COX2 and cPLA₂. In line with whole animal data (Hannon et al., 2003), this coincided with enhanced and glucocorticoid-insensitive eicosanoid production (Croxtall et al., 2003). Futter and colleagues utilized these AnxA1^-/- fibroblasts to examine the biogenesis of internal vesicles in multivesicular bodies (MVBs). Indeed, epidermal growth factor was unable to increase internal vesicle numbers in MVBs of AnxA1^-/- fibroblasts. Thus, AnxA1 appears critical for inward vesiculation in the MVB compartment (White et al., 2006). This observation still needs further mechanistic insight, but might also be relevant for other biological events that require uptake and trafficking through endosomal compartments. For instance, AnxA1^-/- mice are partially protected against influenza A virus infection, involving reduced uptake and exit of internalized virus from late endosomes/MVBs (Arora et al., 2016). In contrast, AnxA1 deficiency interferes with cell fusion but not intracellular vesicle fusion, both needed for repair of myofiber sarcolemma, during repair and regeneration of injured skeletal myofiber (Leikina et al., 2015). Altogether, the latter studies add to the complexity of intracellular AnxA1 functions with different molecular interactions contributing to membrane organization and trafficking also at the plasma membrane and in the endocytic compartment (Table 2).

AnxA2 KO mice

AnxA2 is expressed in a wide variety of cells, and in mice, high AnxA2 levels are found in the lung, pancreas, colon, ileum and adrenal tissues, while spleen, testis, kidney and liver express low AnxA2 levels (Seidah et al., 2012) (Table 1). In most cell types, including endo- and epithelial cells, monocytes and macrophages, AnxA2 is found predominantly at the plasma membrane, on endosomal and secretory vesicles, but also in the nucleus (Hedhli et al., 2012; Bharadwaj et al., 2013). Inside cells, AnxA2 regulates a spectrum of functions related to membrane organization and trafficking. This includes the endo- and exocytic pathway, microdomain formation and membrane repair at the plasma membrane as well as RNA export from the nucleus. The complex interactions of AnxA2 with the actin cytoskeleton and membranes enriched with phosphatidylinositol-4,5-bisphosphate probably contribute to the multiple membrane-related AnxA2 activities relevant for cell growth, differentiation, apoptosis and migration (Hayes et al., 2004; Rescher and Gerke, 2004; Gerke et al., 2005; Grewal and Enrich, 2009; Bharadwaj et al., 2013).

Importantly, the majority of AnxA2 exists as a heterotetramer associated with p11 (S100A10), a member of the S100 protein family, at the intra- but also extracellular leaflet of the plasma membrane. After formation and translocation of the intracellular AnxA2/p11 complex to the plasma membrane, Src kinase-mediated phosphorylation of AnxA2 at tyrosine 23 is required for AnxA2/p11 export to the outer membrane leaflet. At the cell surface, in particular on endothelial cells, AnxA2/p11 act as receptors or docking sites for plasminogen and tissue plasminogen activator (tPA), both key factors to generate plasmin and thereby promote vascular fibrinolysis. Different models of AnxA2/p11-mediated plasminogen activation have been proposed. We recommend several reviews that have extensively discussed the formation, structural organization, translocation and mode of action of the extracellular AnxA2/p11 complex (Flood and Hajjar, 2011; Grieve et al., 2012; Hedhli et al., 2012; Bharadwaj et al., 2013; Luo and Hajjar, 2013; Bydoun and Waisman, 2014).

To address the in vivo role of AnxA2 in vascular homeostasis, the Hajjar laboratory generated AnxA2-deficient mice (Ling et al., 2004) (Table 3). Similar to the other Anx-deficient strains, AnxA2^-/- mice were viable, fertile and showed normal development and life span. Reduced p11 levels in these mice supported studies that loss of AnxA2 promotes p11 ubiquitination and degradation (He et al., 2008). Strikingly, AnxA2-deficient animals showed substantial deposition of fibrin in all tissues analyzed. Moreover, after induction of acute carotic artery thrombosis, AnxA2 KO animals displayed increased thrombosis. This provided initial proof-of-concept in vivo that AnxA2 mediates fibrin clearance. Follow-up studies examining AnxA2^-/- mice in the carotid artery injury model revealed that besides Src kinase, protein kinase C (PKC) was also required for AnxA2/p11 translocation to the cell surface (He et al., 2011). Further supporting a role for AnxA2/p11 in fibrinolysis, de-regulated fibrinolysis was also evident in p11 KO mice (Surette et al., 2011).

In addition, most likely also due to the lack of AnxA2-dependent plasmin generation, major defects in the ability of AnxA2^-/- mice to form new blood vessels in several in vivo assays were observed (Table 3). This comprises matrigel implant, corneal pocket and oxygen-induced retinopathy, the latter mimicking aspects of diabetic retinopathy (Ling et al., 2004; Huang et al., 2011; Hedhli et al., 2012).

The consequences of high or low plasma levels of AnxA2/p11 in several human diseases linked to vascular homeostasis and angiogenesis have been discussed (Flood and Hajjar, 2011; Hedhli et al., 2012; Bharadwaj et al., 2013; Luo and Hajjar, 2013; Bydoun and Waisman, 2014). For instance, diet-induced hyperhomocysteinemia, which is implicated in thrombotic and atherosclerotic disease, shows fibrin accumulation and defects in neoangiogenesis resembling those observed in the AnxA2^-/- mice (Jacovina et al., 2009). AnxA2 isolated from mice fed a high methionine diet failed to bind tPA and activate plasminogen, while intravenous injection of recombinant AnxA2 partially restored angiogenetic activity in this mouse model (Jacovina et al., 2009). In ischemic cerebral disease, recombinant AnxA2 may serve as an adjunct drug to amplify tPA-mediated thrombolysis to prevent stroke (Fan et al., 2010). However, although inhibition of plasmin generation has therapeutic potential in atherosclerosis, crossing the AnxA2^-/- mice into an apoE-deficient background did not reduce lesion development compared to the control apoE KO animals (Hedhli et al., 2012).

High antibody titers directed against AnxA2 seem to be associated with thrombotic complications in the antiphospholipid syndrome (Romay-Penabad et al., 2009) and cerebral venous thrombosis (Flood and Hajjar, 2011). High levels of AnxA2 correlate with hyperfibrinolysis in acute promyelocytic leukemia (Flood and Hajjar, 2011; Luo and Hajjar, 2013). In addition, data from p11 KO mice support findings that AnxA2/p11-dependent plasmin generation is involved in tumor progression (Bharadwaj et al., 2013; Luo and Hajjar, 2013; Bydoun and Waisman, 2014).

In recent years, several studies in AnxA2 KO mice aimed to investigate extracellular AnxA2 activities not related to plasmin generation (Table 3). This includes the interaction of AnxA2 with proprotein convertase subtilisin/kexin-9 (PCSK9), a potent inducer of LDLR degradation (Mayer et al., 2008; Seidah et al., 2012). In cell models, AnxA2 or the AnxA2/p11 complex inhibit PCSK9-mediated LDLR downregulation, leading to elevated LDLR levels at the cell surface (Mayer et al., 2008). In line with these findings, AnxA2^-/- mice displayed elevated LDL-cholesterol and circulating PCSK9 levels (Seidah et al., 2012). AnxA2^-/- mice showed reduced LDLR expression in extrahepatic tissues while adenoviral AnxA2 overexpression was capable to elevate hepatic LDLR levels in vivo. Monoclonal antibodies against PCSK9 have recently been clinically approved for additional use together with statins in hypercholesterolemic patients. Hence these findings could identify modulation of AnxA2 levels as a potential endogenous inhibitor of PCSK9.

In addition, in sensory neurons, AnxA2 is a ligand for the transient receptor potential ion channel A1 (TRPA1) (Avenali et al., 2014), which is essential for detection of harmful chemicals, tissue damage and chronic pain. Increased levels of TRPA1 at the plasma membrane of AnxA2-deficient sensory neurons and enhanced TRPA1-dependent pain in AnxA2^-/- mice were observed. This may indeed be the first in vivo evidence to support several examples that linked AnxA2/p11 with ion channels (Gerke and Moss, 2002; Gerke et al., 2005; Avenali et al., 2014). Finally, inoculation of prostate cancer cells mixed with bone marrow stromal cells from the AnxA2 KO mice revealed that AnxA2 binding to stromal-derived factor 1 is crucial for recruitment, growth and survival of prostate cancer cells in the bone marrow microenvironment (Jung et al., 2015).

Similar to the limited numbers of studies addressing AnxA1 functions inside cells in vivo (see section ‘AnxA1 KO mice’ above), the multiple cellular locations and interaction partners have probably prevented researchers using the AnxA2 KO mice to examine intracellular events. However, several possibly clinically relevant observations with this strain have been made in recent years (Table 3). Crossing a sophisticated mouse model for the development of pancreatic ductal adenocarcinoma (PDA) with the AnxA2^-/- strain revealed a prominent role for AnxA2 in metastatic events driven by pancreatic cancers (Foley et al., 2015). While primary pancreatic tumors from AnxA2-deficient animals and controls were comparable, the metastatic potential of the PDA tumors from AnxA2 KO animals was substantially reduced. It is unknown if this striking phenotype requires p11, but it appears unrelated to plasmin generation and neoangiogenesis. Yet it involves secretion and autocrine signaling of semaphorin 3D, suggesting a critical role for AnxA2 in exocytic pathways contributing to the aggressive behavior of pancreatic cancer cells (Foley et al., 2015).

Further support for a critical in vivo function of AnxA2 in the secretory pathway come from studies examining pulmonary function. AnxA2^-/- display reduced exercise tolerance and impaired lung tissue elasticity that resembles the phenotype of collagen VI (COL6) deficiency, which anchors basement membranes to other collagen fibers (Dassah et al., 2014). Closer inspection identified AnxA2^-/- lung basement membrane to lack COL6 due to its retention in the Golgi apparatus (Dassah et al., 2014). Previous cell culture studies had identified AnxA2 to regulate exocytosis through interaction with at least three soluble NSF attachment protein receptor (SNARE) proteins, synaptosomal-associated protein 23 (SNAP23), SNAP25 and vesicle-associated membrane protein 2 (VAMP2) (Knop et al., 2004; Umbrecht-Jenck et al., 2010). In line with these findings, pulmonary dysfunction in the AnxA2^-/- mice correlates with association of AnxA2 with COL6 as well as SNAP23 and VAMP2 in secretory vesicles from bronchial epithelial cells. It still remains difficult to envisage how AnxA2 may interact simultaneously with these three proteins to exert these regulatory effects. COL6 is located in the lumen of secretory vesicles, VAMP2 is a trans-membrane protein and SNAP23 is membrane-bound facing the cytoplasm. Nevertheless, these studies are the first in vivo evidence that AnxA2 is critical for membrane transport along the secretory pathway. This may be relevant for pulmonary function, but may extend to other disorders associated with COL6 deficiency (Dassah et al., 2014). In addition, this could also be related to migratory and invasive properties of pancreatic cancer cells (Foley et al., 2015) (Table 3).

A few studies in the AnxA2 KO strain have now also provided initial evidence that AnxA2 contributes to membrane organization and transport along endocytic routes. For instance, Toll-like receptor 4 (TLR4) activation at the plasma membrane and on endosomes is critical for pathogen recognition and activation of innate immunity in macrophages. Interestingly, AnxA2^-/- mice were highly susceptible to bacteria-induced pulmonary inflammation. This correlated with enhanced TLR4 signaling and reduced TLR4 endocytosis (Zhang et al., 2015). Further in vivo evidence that AnxA2 is linked to membrane organization during cellular uptake come from studies addressing autophagy. In autophagy, budding and fusion of vesicles, often at the plasma membrane, generate a phagosome for encapsulation of organelles and biomolecules. These phagosomes are then targeted to endosomes/lysosomes for degradation. Dendritic cells, which have the highest plasma membrane turnover, were used to isolate vesicles from the AnxA2 KO mice and showed reduced amounts of phosphatidylserine and phosphatidylinositides. Phagosome biogenesis and maturation in dendritic cells from the AnxA2^-/- mice was also strongly reduced (Morozova et al., 2015). Along these lines, autophagy can also serve as a defense mechanism against bacterial infection, and was de-regulated in a mouse infection model using AnxA2^-/- animals (Li et al., 2015). Altogether this supports the large amount of cell culture based literature that have proposed AnxA2 to organize membrane domains at the plasma membrane and endocytic compartment and enable vesiculation.

AnxA4 KO mice

AnxA4 is expressed in many tissues, in particular in the cytoplasm of secretory epithelia in the lung, intestine, stomach and kidney (Kaetzel et al., 1994) (Table 1). After Ca²⁺ elevation, AnxA4 translocates to the plasma membrane or the nuclear membrane. Like some other annexins, proposed functions include self-association on membrane surfaces to participate in Ca²⁺-dependent aggregation of vesicles. Oligomerization of AnxA4 is also believed to modulate the mobility of membrane-associated proteins (Gilmanshin et al., 1994; Piljic and Schultz, 2006; Crosby et al., 2013). Based on a single transcript originally reported in mice, Li and coworkers generated a mouse with a gene trap inserted into the first intron upstream from the first coding exon of AnxA4 (Li et al., 2003). Somewhat unexpectedly, AnxA4 expression was not disrupted in all tissues. This led to the identification of three AnxA4 transcripts, AnxA4a, AnxA4b and AnxA4c, with unique tissue expression patterns. The AnxA4 KO mice lacked the major transcript AnxA4a, which is broadly distributed. AnxA4b is only found in the digestive track, while AnxA4c expression is restricted to solitary chemosensory cells. In follow-up studies, the function of AnxA4 in the epithelium of the bladder was investigated. However, loss of AnxA4a in the urothelium did not reveal alterations in bladder function. This suggests that AnxA4a is not involved in the integrity or the regulation of the bladder as a permeability barrier (Hill et al., 2008) (Table 4).

Based on AnxA4 upregulation in failing hearts, and GPCRs being influenced by lipid-protein interactions and membrane domain properties, Heinick and colleagues (Heinick et al., 2015) postulated a role of AnxA4 in β-adrenoreceptor (β-AR)/cAMP signaling. Cardiomyocytes from adult AnxA4 KO mice displayed increased cAMP levels, likely due to the loss of the inhibitory action of AnxA4 on adenylyl cyclase 5, which controls conversion of ATP into cAMP. Moreover, AnxA4A-KO mice treated with β-AR agonists showed increased cAMP levels, associated with enhanced cardiac contraction force in the heart. Annexins A5, A6 and A7, which are abundant in cardiomyocytes, were not upregulated in AnxA4^-/- heart tissues. Taken together, these studies have for the first time linked an intracellular annexin function with GPCR activity, which could go well beyond β-AR signaling in cardiomyocytes. As several annexins are believed to contribute to membrane microdomain formation, this novel mechanism could open exciting new opportunities in the annexin field, examining their impact on GPCR signal transduction (Table 4).

AnxA5 KO mice

AnxA5 is the most abundant member of the annexin family and expressed in most cells and tissues, except neurons (Boersma et al., 2005; Gerke et al., 2005) (Table 5). Inside most cells, upon Ca²⁺ elevation, AnxA5 translocates to the plasma membrane and nucleus. Other locations, including the Golgi, endoplasmic reticulum, late endosomes/lysosomes, phagosomes, and mitochondria, have also been reported (Dubois et al., 1998; Gerke et al., 2005; Ghislat et al., 2012). At these cellular sites, AnxA5 has been associated with membrane trafficking and organization, Ca²⁺ signaling, regulation of ion channels and Ca²⁺-influx as well as cell cycle regulation and apoptosis (Hawkins et al., 2002; Wang et al., 2003; Gerke et al., 2005; Monastyrskaya et al., 2007; Faria et al., 2011).

Extracellular AnxA5 has become a diagnostic tool to detect apoptotic cells, as AnxA5 exhibits strong Ca²⁺-dependent binding affinity towards phosphatidylserine in the outer leaflet of the plasma membrane (Boersma et al., 2005). In addition, extracellular AnxA5 has been linked to blood coagulation, cancer, phagocytosis, viral infection, membrane invagination and membrane repair (Boersma et al., 2005; Rand et al., 2010; Bouter et al., 2011; Peng et al., 2014; Bouter et al., 2015).

AnxA5 is highly expressed in cartilaginous tissues and bone. Given its proposed role in cellular Ca²⁺ influx, the Pöschl laboratory generated the AnxA5 KO mouse to examine if loss of AnxA5 would affect the calcification process during skeletal development (Brachvogel et al., 2003) (Table 5). In line with all other Anx-deficient mouse models, AnxA5-deficient mouse were viable and fertile, lacking any obvious defects in glucose and lipid metabolism, liver function or the development of bone and cartilage. Thus, despite the high expression levels in cartilage and bone, AnxA5 did not appear essential for the calcification process during skeletal development. It was concluded that other annexins may functionally compensate for the AnxA5 deficiency. In particular AnxA6, which colocalizes with AnxA5 in several tissues (Thorin et al., 1995; Matteo and Moravec, 2000) appeared the most promising candidate. Although some changes in pituitary hormone expression in the AnxA5 KO mice were reported, their physiological relevance are unclear (Mittag et al., 2007). Yet, most important within the context of compensatory effects, upregulation of AnxA2, AnxA6 or AnxA7, all of which with overlapping expression, localization and activity pattern compared to AnxA5, was not observed in the AnxA5 KO mice. Nevertheless, the hypothesis of functional redundancy within the annexin family led to the generation of double KO mice lacking AnxA5 and AnxA6 (Belluoccio et al., 2010; Grskovic et al., 2012), which is described in more detail below (see the section ‘AnxA6 KO mice’ below; Tables 5 and 6).

AnxA5 is believed to be involved in pregnancy loss and thrombosis associated with the antiphospholipid syndrome (Rand et al., 2010). Initial studies with the AnxA5 KO strain did not reveal any impact of AnxA5 deficiency on the developing fetus or litter size. However, on closer inspection, reduced litter size and increased foetal loss in AnxA5 KO mice was reported recently (Ueki et al., 2012). Strikingly, smaller foetus size was already apparent upon maternal loss of AnxA5. Moreover, administration of anti-coagulant heparin to prevent thrombi formation in the placental circulation reduced pregnancy loss in the AnxA5 KO mice (Table 5). Mechanistically, the ability of AnxA5 to form two-dimensional crystal lattices on phosphatidylserine-rich membranes in a Ca²⁺-dependent manner in vitro seems most relevant (van Genderen et al., 2008; Rand et al., 2010). Hence, the AnxA5 KO animals could become a suitable model to further investigate therapeutic approaches to prevent inappropriate thrombogenesis causing infertility by the antiphospholipid syndrome in humans (Rand et al., 2012).

The targeted disruption of the AnxA5 gene involved homologous recombination and generation of an AnxA5-lacZ fusion gene (Brachvogel et al., 2003). This enabled detailed AnxA5 expression pattern analysis during embryonal mouse development, identifying a subset of AnxA5-LacZ-positive perivascular cells with mesenchymal stem cell-like properties. These cells provided many opportunies to investigate formation of blood vessels (Brachvogel et al., 2005; Zhou et al., 2016). Other aspects of AnxA5 biology, for instance its ability to recognize phosphatidylserine exposed by apoptotic cells, have also been investigated using the AnxA5 13 KO strain. In contrast to control animals, AnxA5-KO mice showed a strongly reduced allogeneic reaction against primary and secondary necrotic cells as well as carcinoma cells (Munoz et al., 2007; Frey et al., 2009). In this setting, a role for AnxA5 in anti-inflammatory pathways that enable the phagocytic process and clearance of necrotic or apoptotic cells can be envisaged. This could also be relevant for preventing viral infections, including human immunodeficiency virus, which together with infected monocytes/macrophages, exposes increased amounts of phosphatidylserine on the surface (Munoz et al., 2007) (Table 5).

AnxA6 KO mice

Out of the annexin family, AnxA6 is the only member that consists of eight annexin repeats, most likely due to the duplication and fusion of the genes encoding AnxA5 and A10. High AnxA6 levels are well documented in most mammalian tissues, only epithelial cells of the small intestine, colon and parathyroid gland express low to undetectable AnxA6 levels (Grewal and Enrich, 2009; Enrich et al., 2011; Grewal et al., 2010) (Table 1). AnxA6 is predominantly found at the plasma membrane and endosomes (Grewal et al., 2005; Cubells et al., 2007; Vilà de Muga et al., 2009). Other AnxA6 locations include mitochondria, lipid droplets, and the secretory pathway (Gerke and Moss, 2002; Freye-Minks et al., 2003; Gerke et al., 2005; Turró et al., 2006; Chlystun et al., 2013). These various cellular locations, together with multiple interaction partners identified in tissues, cell culture and in vitro studies likely reflect the many different regulatory AnxA6 functions. Indeed, AnxA6 is engaged in endo- and exocytosis, plasma membrane microdomain rearrangements, signal transduction, cholesterol homeostasis, actin dynamics, phospholipase activity and stress response. Hence over the years, AnxA6 has been implicated in cell proliferation, survival, migration, differentiation, and inflammation (Grewal and Enrich, 2009; Enrich et al., 2011; Koese et al., 2013; García-Melero et al., 2016), as well as membrane repair (Skrahina et al., 2008; Roostalu and Strähle, 2012) and viral infection (Ma et al., 2012; Musiol et al., 2013).

The generation of transgenic mice overexpressing AnxA6 in the heart represented the first effort to address AnxA6 function in vivo (Gunteski-Hamblin et al., 1996) (Table 6). This gain-of-function approach revealed enlarged dilated hearts, acute diffuse myocarditis and lymphocytic infiltration. In addition, moderate-to-severe fibrosis throughout the heart, and mild fibrosis around the pulmonary veins of the lungs, ultimately causing heart failure at a relatively young age, was observed. Reduced amplitude of Ca²⁺ transients, impaired contractility and cardiomyopathy suggested that elevated AnxA6 levels in the heart interfered with Ca²⁺ homeostasis, possibly through regulation of membrane-associated ion pumps and/or exchangers of cardiomyocytes.

However, despite these encouraging first insights, the generation of the AnxA6 KO mice by Moss and coworkers resulted in animals lacking a clear phenotype (Hawkins et al., 1999) (Table 6). Loss of the AnxA6 gene was not associated with noticeable morphological changes and did not interfere with overall viability and fertility. In addition, despite the prominent cardiovascular function of the transgenic AnxA6 mice (Gunteski-Hamblin et al., 1996), the AnxA6 KO mice displayed a heart rate, blood pressure and cardiovascular response to septic shock that was comparable to controls. Furthermore, although AnxA6 expression is strongly upregulated during the differentiation of B and T lymphocytes (Clark et al., 1991), AnxA6^-/- mice displayed normal B and T cell development without any obvious immunological phenotype. Overall, this clearly challenged the various functions proposed for AnxA6, in particular in respect to lymphocyte development and heart function. Hence, anticipated AnxA6 functions were considered to be restricted to cultured cells. Alternatively, compensatory mechanisms were proposed to alleviate the loss of AnxA6 in vivo. However, while several annexins were found to be upregulated in the AnxA1 KO mice (Hannon et al., 2003), AnxA1, A2 and A5 protein levels in the heart, liver and spleen of the AnxA6^-/- mice remained unchanged (Hawkins et al., 1999). This indicated that at least these annexins were not upregulated to compensate for the loss of AnxA6 in these organs.

These initial findings from the AnxA6 KO mice supported a view that within the annexin family structural and functional similarities may lead to redundancy. Consequently, mice lacking multiple annexins was contemplated to identify critical functions of individual annexins in vivo. Based on this hypothesis, the Brachvogel group generated mice deficient of both AnxA5 and AnxA6 to investigate their involvement in the calcification within the growth plate during the formation of the mineralized skeleton (Belluoccio et al., 2010). Similar to mice lacking either AnxA5 or AnxA6, immunohistochemistry and tomographic analysis showed normal skeletal development in the double KO mice. Gene expression analysis from the (pre-)hypertrophic zones from cartilage did not identify upregulation of other annexins in the AnxA5^-/-AnxA6^-/- mice. Only minor changes in mRNA levels from growth- and metabolism-related genes were observed. It was therefore concluded that AnxA5 and AnxA6 were not essential for mineralization in vivo. As both annexins were speculated to initiate the calcification of cartilage through collagen X-dependent pathways, Grskovic et al. generated triple KO mice deficient of AnxA5, AnxA6 and collagen X (Grskovic et al., 2012). However, these mice did not show obvious abnormalities. Thus, the interaction of AnxA5, AnxA6 and collagen X was not considered essential for the calcification of the growth plate cartilage (Table 6).

Despite the lack of a prominent phenotype, primary cells isolated from the AnxA6 KO mice identified interesting features. First, AnxA6^-/- cardiomyocytes exhibited higher contractility and accelerated removal of diastolic Ca²⁺ from the cytoplasm. This was consistent with AnxA6 modulating Ca²⁺ channels and exchangers (Song et al., 2002). In line with these findings, using transgenic mice overexpressing a truncated dominant-negative AnxA6 mutant in the heart (Kaetzel and Dedman, 2004), changes in Ca²⁺ signaling and stimulus-response coupling were observed. Similar N-terminal AnxA6 deletion mutants cloned by others interfered with LDL endocytosis (Kamal et al. 1998; Pons et al., 2001) as well as epidermal growth factor receptor and Ras/MAPK signaling (Grewal et al., 2005; Vilà de Muga et al., 2009). One could speculate that the truncation of 5–6 C-terminal annexin repeats leads to cytosolic mutant AnxA6 proteins that may mistarget membrane-binding interaction partners to the cytosol (Table 6).

Second, chondrocytes isolated from newborn AnxA6 KO mice exhibited delayed terminal differentiation. This coincided with markedly decreased intracellular Ca²⁺ levels upon ascorbic acid stimulation (Minashima et al., 2012). PKCα membrane translocation and activity was reduced in AnxA6-deficient chondrocytes, which may decrease MAPK signaling required for chondrocyte differentiation. Subsequent analysis of the skeletal phenotype of AnxA6 KO newborns revealed reduced growth plate length and number of chondrocytes. Mineralization of growth plate cartilage was reduced in newborn AnxA6 KO mice and associated with decreased MAPK activities. Hence, during chondrocyte differentiation, AnxA6 may promote PKCα membrane translocation and activity through direct interaction, which then modulates PKCα-dependent MAPK activities. We observed similar scaffold functions for AnxA6 in PKCα-dependent EGFR inactivation in A431 cancer cells (Koese et al., 2013). This is likely responsible for the reduced xenograft growth of A431 cells in vivo upon AnxA6 overexpression (Theobald et al., 1995; Hoque and Grewal, unpublished data).

Third, skin fibroblasts, as well as liver and retinal cells from the AnxA6 KO mice showed abnormal mitochondrial morphology, decreased mitochondrial Ca²⁺ uptake and increased cytosolic Ca²⁺ transients compared to controls (Chlystun et al., 2013). The underlying mechanisms are not fully understood. It was proposed that a pool of AnxA6 may associate with mitochondria to interact with Drp1, a protein with a role in mitochondrial fusion and fission (Table 6).

Based on the overall mild phenotype observed with the AnxA6 KO mice, it was speculated that a physiological challenge mimicking disease could trigger defective biological responses due to the lack of AnxA6. Therefore, Kirsch and colleagues analyzed cartilage destruction that occurs in osteoarthritis, comparing knee joints after interleukin 1β (IL-1β) injection or partial meniscetomy (Campbell et al., 2013). In both settings, knee cartilage destruction was markedly reduced in AnxA6 KO mice. These findings correlated with reduced nuclear factor kappa B (NF-κB) activity and expression of NF-κB target genes, major drivers in the pathogenesis of osteoarthritis. As AnxA6 was found to co-immunoprecipitate with p65 of the NF-κB complex, one could envisage that lack of AnxA6 may interfere with the translocation of activated NF-κB from the cytosol to the nucleus (Table 6). It remains to be determined if this interaction and a possible role of AnxA6 in the nuclear translocation of proteins is also relevant in other scenarios.

We recently initiated studies to examine AnxA6 KO mice after immune challenge. In agreement with earlier studies (Hawkins et al., 1999), unchallenged AnxA6^-/- had a normal repertoire of CD3⁺, CD4⁺ and CD8⁺, naïve (CD62L⁺CD44^-), effector (CD62L^-CD44⁺) and memory (CD62L⁺CD44⁺) T lymphocytes (Cornely et al., 2016). To elicit an immune response, mice were then subjected to a delayed-type contact hypersensitivity reaction. In this setting, the local tissue swelling and clonal expansion of T cells in the lymph nodes was determined after dermal exposure to an irritant. In lymph nodes from AnxA6 KO mice, significantly lower levels of proliferating CD4⁺, but not CD8⁺, T lymphocytes were observed. After immune challenge, AnxA6-deficiency neither affected the ability of lymphocytes to migrate towards the site of inflammation, nor T cell receptor signaling. However, IL-2 signaling, which is essential to drive T cell proliferation, was impaired in AnxA6^-/- T cells. We speculated that the underlying mechanisms for the inadequate IL-2 response could involve assembly and signaling of the IL-2 receptor in lipid-ordered domains (lipid rafts). Although it has long been proposed that AnxA6 and other annexins regulate membrane organization (Gerke et al., 2005; Enrich et al., 2011), evidence to support this hypothesis in live cells and organisms has been lacking. Yet, AnxA6^-/- T cells displayed reduced membrane order at the plasma membrane. Likewise, membrane order and distribution of proteins in rafts and non-rafts was also affected in mesenchymal mouse embryo AnxA6^-/- fibroblasts (Alvarez-Guaita et al., 2015). Thus, the lack of AnxA6 may perturb the organization of cholesterol-rich domains required for appropriate IL-2 receptor signaling during immune response. The compromised immune response might also be relevant for other settings. Data available from the Wellcome Trust Mouse Genetics Project identified AnxA6-deficient mice to display reduced bacterial clearance after oral challenge with Citrobacter rodentium, indicating a higher susceptibility to infection (The Wellcome Trust Sanger Institute Mouse Genetics Project, Database Release 2011; http://www.informatics.jax.org/reference/J:175295). Taken together, these studies revealed for the first time that members of the annexin family can modulate the organization of the plasma membrane in vivo, with possible consequences for receptor signaling in T cell proliferation and immune response (Table 6).

AnxA6 is abundantly expressed in the liver (Tagoe et al., 1994), and involved in cholesterol and lipoprotein metabolism (Grewal et al., 2000; de Diego et al., 2002; Cubells et al., 2007). We therefore recently examined disease settings with a focus on liver function. Most strikingly, after partial hepatectomy a strongly reduced survival rate was observed for the AnxA6 KO mice (Rentero et al., 2015). Compromised liver regeneration in AnxA6^-/- mice was associated with delayed steatosis and prolonged hypoglycaemia, and correlated with impaired liver gluconeogenesis. Mice as well as primary hepatocytes lacking AnxA6 were unable to produce glucose from alanine, a major hepatic gluconeogenic substrate under metabolic stress. The underlying mechanisms are under investigation. In line with a regulatory role of AnxA6 in endo- and exocytic pathways (Enrich et al., 2011; Grewal et al., 2010; García-Melero et al., 2016), AnxA6^-/- primary hepatocytes display de-regulated membrane targeting of amino acid transporters, which is essential during liver regeneration and fasting.

Further support that AnxA6 is linked to hepatic lipid and glucose homeostasis come from studies examining AnxA6 KO mice after high-fat diet feeding. Under these conditions, mice eventually develop a fatty liver and insulin resistance. When examining hepatic glucose metabolism, we identified an inability of high-fat diet-fed AnxA6 KO mice to properly downregulate hepatic gluconeogenesis and lower blood glucose levels. Preliminary results suggest that loss of AnxA6 regulate membrane transport and cellular signaling events that contribute to metabolic complications associated with fatty liver disease and insulin resistance (Cairns, Rentero, Enrich and Grewal, unpublished data) (Table 6).

Finally, besides all the abovementioned observations in rodents, AnxA6 deficiency in zebrafish leads to myopathy, making it an attractive model for muscle dystrophy (Roostalu and Strähle, 2012). Cell models have implicated several annexins, including AnxA6, in membrane repair (Skrahina et al., 2008; Draeger et al., 2014). In zebrafish, AnxA6 is rapidly recruited to damaged sarcolemma and may cooperate with dysferlin to repair membrane lesions. This also appears relevant in mice, as genetic screening in a mouse model of muscular dystrophy identified the AnxA6 locus as a modifier of membrane damage in abdominal muscle and right ventricle mass, both relevant to cardiopulmonary function (Swaggart et al., 2014). Similar to zebrafish, mouse AnxA6 is rapidly recruited to damaged sarcolemma, contributing to the formation of a tight cap over a vesicle-rich repair zone. However, low expression of an AnxA6 splice variant that lacks the four C-terminal annexin repeats interfered with membrane recruitment of full-length AnxA6 and substantially decreased efficiency of membrane repair. It is tempting to speculate that this mutant acts in a dominant-negative fashion like similar AnxA6 deletion mutants described above (Kamal et al. 1998; Pons et al., 2001; Grewal et al., 2005; Vilà de Muga et al., 2009).

AnxA7 KO mice

Within the annexin family, AnxA7 is the only member that contains a 100 amino acid long hydrophobic N-terminus. Alternative splicing gives rise to a 47 kDa isoform being expressed in all tissues except skeletal muscle, while a larger 51 kDa isoform is found in the heart, brain and myotubes (Selbert et al., 1995) (Table 1). AnxA7 is predominantly located on secretory vesicles, the plasma membrane and the nuclear envelope. Like other annexins, AnxA7 translocates from the cytosol to these locations in a Ca²⁺-dependent manner (Kuijpers et al., 1994). Many reports support a role of AnxA7 in multiple aspects of Ca²⁺/GTP-dependent exocytic pathways (Srivastava et al., 1999; Caohuy and Pollard, 2002; Gerelsaikhan et al., 2012; Taniuchi et al., 2012; Chander et al., 2013). Related to Ca²⁺ homeostasis, GTPase activity and possibly other cellular functions, AnxA7 participates in prostaglandin production, cardiac remodeling and inflammatory myopathies, but also cell survival and tumor growth (Voelkl et al., 2014; Luo et al., 2015). Although mechanistically not fully understood, the latter findings correlate with the tumor suppressor activity of AnxA7 in prostate, breast and several other cancers (Srivastava et al., 2003, 2007).

The Pollard laboratory generated the first mouse model lacking AnxA7 (Srivastava et al., 1999) and in striking contrast to all other annexin KO strains (see the sections on AnxA1- AnxA6 KO mice, above), the targeted disruption of the AnxA7 gene led to embryonic lethality due to cerebral hemorrhage. However, heterozygous AnxA7^+/- mice were viable and fertile. A closer inspection of the pancreas revealed islet hyperplasia, β-cell hypertrophy and aberrant metabolic gene expression patterns. Moreover, reduced inositol 1,4,5-trisphosphate (InsP₃) receptor expression was determined. Consequently, anomalous Ca²⁺ release from intracellular stores, causing defects in Ca²⁺ signal transduction that drive insulin secretion in pancreatic β-cells, was observed. Follow-up studies identified that abnormal insulin secretion in β-cells from AnxA7^+/- mice gene was further compromised by a change in the sensitivity to activate ryanodine receptor (RyR) – mediated Ca²⁺ release (Mears et al., 2012) (Table 7).

In line with the tumor suppressor activity of AnxA7, the heterozygous Anx7^+/- strain also revealed a cancer-prone phenotype, with more than 20% of the mutant mice developing spontaneous neoplasms (Srivastava et al., 2003). This was accompanied by reduced expression of several tumor suppressors, DNA repair- and apoptosis-related genes, indicating the development of genomic instability driving disease progression upon partial loss of AnxA7. Altogether, these studies support multiple roles for AnxA7 in vivo, some of which probably include Ca²⁺-dependent signal transduction. The underlying mechanisms are not fully understood, and may involve loss of direct protein-protein interactions between AnxA7 and other key players (Colotti et al., 2006). In addition, how reduced AnxA7 levels in the heterozygous AnxA7^+/- mice interfere with InsP₃ receptor expression and RyR activity in β-cells or tumor progression in prostate and breast cancer remains to be determined.

Soon after the lethal phenotype of the first AnxA7 KO model was described (Srivastava et al., 1999), Noegel and coworkers independently generated another mouse model with a targeted disruption of the AnxA7 gene (Herr et al., 2001). However, these AnxA7 KO mice were healthy, viable and comparable to control animals in respect to Ca²⁺-induced and cAMP-mediated insulin secretion. The AnxA7^-/- phenotypes reported by the two groups are strikingly different and still represent a challenge when interpreting studies assessing AnxA7 functions in vivo. Different design and integration sites of the targeting construct, orientation of the inserted neo gene and possible consequences for neighboring genes are possibilities. The different genetic background used to generate the two KO strains has also been discussed (Herr et al., 2001). Nevertheless, although the latter studies did not reveal a role for AnxA7 in Ca²⁺-dependent insulin secretion, analysis of adult cardiomyocytes isolated from these mice identified decreased frequency of cell shortening. This pointed at impaired Ca²⁺ homeostasis or the misfunctioning of the apparatus responsible for contraction. Interestingly, AnxA7 interacts with Sorcin and RyR (Verzili et al., 2000), both proteins involved in coupling Ca²⁺ channels with the contractile machinery in cardiac muscle. It was therefore speculated that impaired interaction of these proteins in the adult AnxA7^-/- mice might occur (Herr et al., 2001; Franceschini et al., 2008) (Table 7).

As impaired Ca²⁺ homeostasis in cardiomyocytes is critical for heart function, the impact on cardiac electrophysiological properties in the viable AnxA7^-/- strain was examined in more detail (Schrickel et al., 2007). These studies identified an elevated susceptibility to atrial fibrillation and ventricular tachycardia, both underlying causes for heart arrhythmia. Interestingly, examination of atrial and ventral tissues by electron microscopy revealed impaired integrity of the basement membrane of cardiomyocytes and lack of collagen fibers in intracellular spaces. Together with the defects in Ca²⁺ homeostasis, and the interactions discussed above, these alterations in membrane structure might be due to de-regulated conductive properties of the AnxA7^-/- mice.

Recently, studies in the AnxA7^-/- mice suggest that the molecular functions of AnxA7 in the heart related to Ca²⁺ homeostasis also affect cardiac remodeling (Voelkl et al., 2014). AnxA7^-/- and control mice were studied following pressure overload by transverse aortic constriction. This is a well-established surgical procedure, which increases the resistance to blood flow out of the heart. As the heart then needs to generate more force to maintain normal cardiac output, this eventually leads to cardiac hypertrophy. In these studies, the increase in heart-to-body weight was significantly more pronounced in AnxA7 KO mice, and associated with increased expression of several genes, possibly induced by the Ca²⁺-regulated cardiac nuclear factor of activated T cells (NFAT) in an AnxA7-dependent manner.

In addition to the abovementioned studies addressing AnxA7 functions in the pancreas, heart, and tumor development, lack of AnxA7 affects Ca²⁺ homeostasis and proliferation of primary astrocytes isolated from the viable AnxA7 KO strain (Clemen et al., 2003). These observations are not well understood, but may be related to several studies that point at major alterations in the haematopoetic system of these animals. Initial characterization of red blood cells from AnxA7^-/- mice identified an altered cell shape and increased resistance to osmotic shock (Herr et al., 2003). Follow-up studies then addressed how the lack of AnxA7 impacts on regulatory circuits that promote the suicidal death of erythrocytes, also termed eryptosis (Lang et al., 2010). In erythrocytes from AnxA7 KO mice, hyperosmotic shock, Cl^- removal or energy depletion triggered enhanced prostaglandin E₂ (PGE₂) formation. This led to increased cytosolic Ca²⁺ activity, followed by elevated exposure of phosphatidylserine at the cell surface, and consequently, higher apoptotic death. Interestingly, pharmacological inhibition of COX1/2 enzymes significantly attenuated the enhanced production of PGE₂ in AnxA7^-/- erythrocytes, indicating that AnxA7 negatively regulates COX-dependent PGE₂ formation. This observation may have clinical significance, as accelerated eryptosis of plasmodium-infected erythrocytes from the AnxA7 KO mice confers partial protection to malaria in vivo (Lang et al., 2009). Intriguingly, although earlier research linked several annexins with inhibition of PLA₂ (Gerke and Moss, 2002), only COX but not PLA₂ inhibitors abolished the differences in parasitemia and survival between AnxA7^-/- and wildtype animals (Table 7).

The inhibitory action of AnxA7 on PGE₂ formation was then analyzed in other in vivo disease settings. First, isolated glands from AnxA7 KO mice revealed that glucocorticoid-induced gastric acid secretion, which involves inhibition of COX-mediated PGE₂ production, was compromised in AnxA7^-/- mice (Pasham et al., 2013). Second, as PGE₂ induces insulin resistance in hepatocytes, hepatic COX activity and glucose tolerance in AnxA7 KO mice was examined (Luo et al., 2015). In these studies, prostaglandin levels in plasma and hepatic COX activity were significantly elevated in AnxA7 KO mice. This probably contributes to the decreased glucose tolerance in these animals. Glucose-induced insulin secretion was elevated in the AnxA7 KO strain. Pharmacological COX inhibition, using aspirin, abrogated differences in insulin levels in wildtype and AnxA7 KO animals and improved glucose tolerance in the AnxA7-deficient strain (Table 7). As pointed out by the authors, this does not rule out other mechanisms. Thus, altered RyR receptor signaling in pancreatic β-cells observed in the heterozygous AnxA7^+/- model as underlying cause for enhanced glucose-stimulated insulin secretion should also be considered (Mears et al., 2012). Taking into account the complex involvement of COX enzymes in inflammatory and other diseases, and the wide expression of IP₃ and RyR receptors affecting cellular signaling and Ca²⁺ homeostasis in numerous cells, future studies will have to clarify the contribution of the various candidates affected by AnxA7 deficiency, including COX, IP₃ or RyR receptors.

Conclusions

Since the generation of the first annexin KO mouse more than 15 years ago, efforts in the field have led to a substantial number of mouse models deficient of one or even two members of the annexin family (Figure 1, Table 1). Their ability to bind negatively charged phospholipids in a Ca²⁺-dependent manner has been considered fundamental in membrane transport and organization, protein complex formation and signal transduction. However, mice lacking AnxA1, A2, A4, A5, A6 and A7 are all viable, fertile and appear normal (Table 1). These observations initially supported a view that annexins may only be modifiers of biological functions rather than essential mediators or effectors. Nevertheless, researchers have been taking advantage of the various Anx-deficient KO strains over the years, and assessing physiological stress in a plethora of disease models. This facilitated the identification of important biological functions of individual annexins that often argue against their minor biological contribution, but also against redundancy within the annexin family (Tables 2–7).

Strikingly, the prominent phenotypes of mice lacking AnxA1 or AnxA2 revealed extracellular activities quite different to the central theme in annexin biology (Tables 2 and 3). The resistance of AnxA1 KO mice to the anti-inflammatory action of glucocorticoids was proof-of-principle that ultimately identified extracellular AnxA1 as a ligand for the FPR2 receptor. This interaction is critical for the innate and adaptive immune response, with therapeutic applications in acute and chronic diseases. The AnxA2 KO mice exposed the critical function of extracellular AnxA2/p11 in plasmin generation, therapeutically relevant for fibrinolysis and neoangiogenesis in several diseases. More recently, studies in AnxA1- or AnxA2-deficient mice have also identified significant intracellular functions, both in endo- and exocytic pathways. The underlying mechanisms are still unclear, but at least for AnxA2, a critical in vivo role in SNARE-dependent exocytic vesicle transport seems to evolve. It remains to be clarified if and to what extend intracellular functions of AnxA1 and AnxA2, for example in the secretory pathway, contribute to the prominent phenotypes that are considered to be driven by extracellular AnxA1 and AnxA2 activities.

In the case of AnxA5, findings from the KO model support cell culture and in vitro data that showed strong affinity for phosphatidylserine, with diagnostic and possibly therapeutic applications for recombinant AnxA5 tools. Mice lacking AnxA4, AnxA6 and AnxA7 identified phenotypes that are often linked to Ca²⁺ homeostasis and the regulation of receptors, transporters, or ion channels at the cell surface (Table 5). This points at their membrane binding and organizing properties and their ability to stabilize or create functional protein networks. Although further mechanistic insights are needed, at least in the case of AnxA6, its involvement in membrane organization in vivo is emerging (Table 6). A remaining major challenge for the field is the multifunctionality of annexins and the difficult task to dissect the contribution of individual protein-protein or protein-lipid interactions to certain biological outcomes. Future experiments, exploring state-of-the-art technology such as three dimensional, intravital imaging, should aim to visualize annexin interactions and signaling in live tissue. In combination with (phospho-) proteomics and lipidomics, this could identify the de-regulated protein-protein or protein-lipid interactions that are compromised upon depletion of these annexins.

The different phenotypes observed with the various Anx KO models implicate vital biological functions (Tables 2–7). Yet the identification of genetic evidence in humans to support their relevance in disease is still a critical issue for annexin researchers. The changes observed in annexin expression levels in cancer and other diseases have been reviewed extensively (Gerke and Moss, 2002; Boersma et al., 2005; Gerke et al., 2005; Grewal and Enrich, 2009; Rand et al., 2010; Grieve et al., 2012; Bharadwaj et al., 2013; Hoque et al., 2014). However, in the majority of studies these findings have remained descriptive. This makes it still difficult to assess if changes in annexin expression are a cause or secondary to events that trigger disease. Favorably, single nucleotide polymorphisms (SNPs) within annexin genes have been identified in genome-wide analyses of patient cohorts. Although their functional relevance is yet to be determined, this includes SNPs in the AnxA1 locus linked to Alzheimer’s disease (Lee and Song, 2015) as well as SNPs associated with psoriasis in the vicinity of the AnxA6 locus (Sun et al., 2010). More closely related to data obtained from the knockout (KO) models, SNPs for AnxA2 in sickle cell disease (Luo and Hajjar, 2013) and possibly cardiovascular disease (Seidah et al., 2012; Grewal et al., unpublished data) seem to exist. SNPs in the AnxA5 gene are associated with the risk of pregnancy-related venous thrombosis (Dahm et al., 2015) and malignant melanoma (Arroyo-Berdugo et al., 2014). Finally, in a mouse model for muscle dystrophy, a SNP in the coding region of the AnxA6 gene creates a dominant-negative deletion mutant that interferes with membrane repair (Swaggart et al., 2014). These observations will require further functional analysis in cell culture and in vivo. Advanced intravital imaging techniques and -omics approaches in combination with knock-in mouse models could open exciting perspectives to identify mutations in human disease that are associated with loss of function of annexins.