Caveolae associated proteins and how they effect caveolae dynamics. Björn Morén

Transcription

1 Caveolae associated proteins and how they effect caveolae dynamics Björn Morén Department of Medical Biochemistry and Biophysics Umeå 2014

2 Responsible publisher under swedish law: the Dean of the Medical Faculty This work is protected by the Swedish Copyright Legislation (Act 1960:729) ISBN: ISSN: Cover image: Basic representation of caveolae structure Electronic version available at Printed by: Print & Media, Umeå Universitet Umeå, Sverige, 2014

3 To my family and friends

4

5 Table of contents Abstract... ii List of papers... iv Abbreviations... v Basic concepts of cellular organization... 1 Endocytosis... 1 Plasma membrane composition... 4 Endosomal sorting and protein degradation... 5 Cytoskeleton... 6 Caveolae... 8 Caveolae biogenesis... 9 Functions of caveolae Signaling Lipid homeostasis Mechanosensation and mechanorelief Role of caveolae in disease Caveolae coat proteins Caveolin Cavins EHD Pacsin Implications for the function of caveolae Aims Brief methodological account Cell lines and basic function of the model cells Fluorescence microscopy TIRF Tracking FRAP AFM microscopy Purification of caveolae EPR and spin- labeling Results Paper I Paper II Paper III Paper IV Summary of results and concluding remarks Future prospects Acknowledgements References i

6 Abstract Caveolae are a type of invaginated membrane domain that has been shown to be involved in several disease states, including lipodystrophy, muscular dystrophies and cancer. Several of these diseases are caused by the lack of caveolae or caveolae- related signaling deficiencies in the tissues in which the caveolar domain are abundant such as lung, adipose, muscle and their related endothelial cells. Caveolae are formed through the assembly of the membrane inserted protein caveolin, cholesterol and the recently described family of cavin proteins, which together form the caveolae coat. The work in this thesis focuses on understanding the protein components and mechanisms that control the biogenesis and dynamics of caveolae. We have found that the protein EHD2 is an important regulator and stabilizer of the caveolar domain at the cell membrane. EHD2 is a dimeric ATPase known to oligomerize into ring- like structures around lipid membranes to control their shape. We have characterized the domain interactions involved in the specific targeting and assembly of this protein at caveolae. We propose a stringent regulatory mechanism for the assembly of EHD2 involving ATP binding and switching of the EH domain position to release the N- terminus and facilitate oligomerization in the presence of membrane species. We show that loss of EHD2 in cells results in hyper- dynamic caveolae and that caveolae stability at the membrane can be restored by reintroducing EHD2 into these cells. In a study of the protein cavin- 3, which is known to be an integral component of the caveolar coat, we showed that this protein is targeted to caveolae via direct binding to the caveolar core protein caveolin1. Furthermore, we show that cavin- 3 is enriched at deeply invaginated caveolae and regulate the duration time of caveolae at the cell surface. ii

7 In combination with a biochemical and cellbiological approach, the advanced fluorescence microscopy techniques, like Fluorescence Recovery After Photobleaching (FRAP), Total Internal Reflection microscopy (TIRF), combined with correlative Atomic Force Microscopy (AFM) have allowed us to characterize distinct caveolae- associated proteins and their respective functions at caveolae. Keywords: caveolae, caveolin, EHD2, cavin, microdomain, microscopy, TIRF, AFM iii

8 List of papers I. Morén, B., Shah, C., Howes, M.T., Schieber, N.L., McMahon, H.T., Parton, R.G., Daumke, O., and Lundmark, R. (2012). EHD2 regulates caveola dynamics via ATP- driven targeting and oligomerization. Mol Biol Cell. Reprinted with permission. II. Shah, C., Hegde, B.G., Morén, B., Behrmann, E., Mielke, T., Moenke, G., Spahn, C.M.T., Lundmark, R., Daumke, O., and Langen, R. (2014). Structural Insights into Membrane Interaction and Caveolar Targeting of Dynamin- like EHD2. Structure. Reprinted with permission. III. Larsson, E., Morén, B., Shah, C., Daumke, O., Lundmark, R. An EH- domain switching mechanism regulates stable membrane association of EHD2. Manuscript. IV. Mohan, J., Morén, B., Larsson, E., Holst, M.R., Lundmark, R. Cavin3 interacts with cavin1 and caveolin1 to increase surface dynamics of caveolae. Manuscript. iv

9 Abbreviations AFM Atomic Force Microscopy AP- 2 Activating Protein 2 Arf6 ADP- ribosylation factor 6 ATP/GTP Adenosine- Tri- Phosphate/Guanosine- Tri- Phosphate BAR Bis- Amphysin- Rvs Cdc42 Cell division cycle 42 CliC Clathrin Independent Carrier CME Clathrin Mediated Endocytosis DAG Diacylglycerol EEA1 Early Endosome Antigen 1 EHD2 Eps15 Homology Domain containing protein 2 EPR Electron Paramagnetic Resonance Eps15 Epidermal growth factor receptor pathway substrate 15 FRAP Fluorescence Recovery After Photobleaching GPI Glycosylphosphatidylinositol Graf1 GTPase regulator associated with focal adhesion kinase- 1 PIP2 Phosphatidylinositol 4,5- bisphosphate QNM Quantitative Nanomechanical Mapping TIRF Total Internal ReFlection microscopy v

10 Basic concepts of cellular organization Endocytosis Multicellular organisms consist of a multitude of different cell types that perform specialized functions. All of these cells require that molecules such as proteins, lipids and vitamins are absorbed by and transported within the cells to maintain the homeostasis and function of the cells. Therefore, examination of these transport pathways and their general organization is crucial for our understanding of cell biology and its role in human disease [1]. Signaling deficiencies in cells, such as the inhibition of repressive signaling or an increase in hormonal uptake have been determined to be very important for the formation of cancer [2, 3]. It has also been shown that viruses and other infectious agents hijack endogenous pathways to Fig 1. Basic representation of a cell depicting an internal network of cytoskeletal components. enter the cell and cause damage, which also highlights the importance of investigating these pathways [4, 5]. The cell consists of many components that contribute to the formation of the cellular structure and organization of its internal components, all of which constitutes an intricate network (Fig 1). The first barrier to cell entry is the plasma membrane, which consists of different lipids and proteins sorted into functional units that are responsible for the signaling and transport of molecules into the cell. Uptake and transport at the cellular membrane occurs via endocytosis, which is a 1

11 regulated invagination of the membrane that forms a defined membrane compartment, or vesicle, containing the molecules which are destined for transport. The bending of the membrane during the generation of the invaginated vesicle is typically regulated by proteins that display a slightly bent conformation, such as Bin- Amphiphysin- Rvs (BAR) domain proteins or clathrin (which accumulates to form vesicles), that cluster at the site at which endocytosis should occur. There are several endocytic pathways; each pathway is characterized by its carrier molecules, which enable the cell to bind to and direct various molecules into the appropriate sorting pathway (Fig 2). These carrier molecules envelop the vesicle and are typically designated as the vesicle "coat". The most characterized endocytic pathway is Clathrin Mediated Endocytosis (CME) [1, 6, 7], which displays a clearly visible coat in electron micrographs and requires different adaptor and sorting proteins to bind to and transport cargo. Clathrin clusters at the site at which endocytosis should occur and interlocks, forming a cage that comprises the vesicle. Clathrin is recruited to the site of endocytosis by adaptor proteins, most notably Activating Protein 2 (AP- 2), which is important for CME and is also responsible for cargo binding and initiation of transport. Accessory proteins are responsible for the recruitment of dynamin, which is a protein that is responsible for the scission process during CME [8, 9]. This process is coupled to other proteins that participate in the tubulation and bending of the membrane, including BAR- domain proteins [10, 11]. Dynamin- mediated membrane constriction is achieved via GTP hydrolysis, although whether constriction is sufficient to induce scission or whether other factors are required remains unknown [8]. Less well characterized endocytic pathways include the Clathrin Independent Carrier (CliC) pathway and the macropinosomal pathway, which is responsible for the very rapid uptake of large amounts of fluid and large particles, such as viruses. Although little is known about the protein components that constitute these carriers, some proteins have been 2

12 identified to function in these machineries, including GTPase regulator associated with focal adhesion kinase- 1 (Graf1) and ADP- ribosylation factor 6 (Arf6) [12-15]. A specialized variant of plasma membrane invaginations are caveolae, which will be discussed at length in this thesis. There have been multiple reports on caveolar stability and the stable association between caveolae and the membrane [16, 17], as well as reports of caveolar endocytosis and transport [5]. It was reported that caveolae are an important uptake vector for viral particles, including the SV40 virus [4]. We hypothesize that this structure is generally stably associated with the plasma membrane and that this structure is mobile in the membrane but that endocytosis of the caveolar structure is a slow process that requires a multitude of signaling cues. Fig 2. Different types of endocytic carriers. Adapted with permission from [18]. 3

13 Plasma membrane composition Lipids, which are the primary constituent the cellular membrane, are very important for all cellular processes, including protein clustering, endocytosis and signaling [19]. Because it is very difficult to study lipid dynamics and because the lipid composition of biological cells is very dynamic, the importance of lipids in cellular processes has only recently been understood. The possibility of "lipid rafts" and higher order lipid domains has been suggested for a long time [20-23], and there have been many studies examining these structures using different techniques and stains, such as Laurdan [24, 25]. Caveolae is also considered as a high- order microdomain that contains high levels of sphingolipids and cholesterol [26]. These structures are dependent on certain types of lipids to exist, and caveolae, for example, cannot exist without cholesterol [26]. This finding demonstrates that lipids, such as cholesterol, are just as important for the organization of the cell as proteins and that these molecules most likely function cooperatively in many processes [27-29]. Moreover, cholesterol is important in many cellular structures aside from caveolae and has been associated with several diseases, such as atherosclerosis and neurodegenerative disorders [30]. The cholesterol molecule is also a precursor of several important signaling molecules in the human body [31]. Phosphoinositides is another important class of lipids that regulates many functions in cells. Phosphoinositides are known to be involved in protein anchoring and signaling and to mediate various downstream cellular responses via their cleavage and specific phosphorylation [32]. One example of a downstream effector molecule is diacylglycerol (DAG), which is known to be important for Phosphatidylinositol 4,5- bisphosphate (PIP2)- mediated signaling [33]. In addition, membrane proteins are an integral contributor to cell function, performing activities such as transport and signaling. Membrane proteins have also been suggested to be partitioned into high order lipid domains or lipid rafts or even to aid in the formation of 4

14 lipid rafts by binding to rigid types of lipids, such as sphingolipids. Membrane proteins can be transmembrane proteins, which span the membrane. This type of protein is typically a receptor that mediates signaling into the cytosol using downstream effectors or transporters. Alternatively, some membrane proteins are partitioned into one side of the membrane, such as glycoproteins, which function as membrane protectors or structural proteins specifically intended to stabilize membrane structures [23, 34, 35]. Endosomal sorting and protein degradation Inside the cell, there are different specialized compartments in which proteins are sorted not only when taken up, but also when they are being synthesized. The directionality of these sorting mechanisms is primarily mediated by Rab proteins. The proteins that have been discovered to be responsible for the bulk of protein targeting and sorting in the endocytic machinery are Early Endosome Antigen 1 (EEA1), Rab5, Rab7 and Rab11 [36, 37]. EEA1 and Rab5 are markers of the early endosome, which is the vesicle that is formed just after scission. Rab7 is involved in the maturation of the endosome into a late endosome, a vesicle typically destined for the lysosomal pathway. Rab11 is predominantly involved in endosomal recycling to the plasma membrane. Most of these proteins are not associated with caveolar structure, but previous work has suggested that the proteins EEA1 and Rab5 colocalize with caveolar vesicles at the beginning of the caveolar endocytic pathway, implicating a role of these proteins in caveolar transport and endocytosis [38]. Products taken up from outside the cell, as well as large objects such as organelles, are typically broken down in the lysosome. The lysosome is a low ph compartment filled with proteases and other hydrolases that completely break down all lysosomal contents into their basic components, such as lipids and amino 5

15 acids. Products can be transported to the lysosome, but typically, a preformed lysosome fuses with an endosome or a similar compartment to initiate product breakdown. This breakdown process is similar to the autophagy pathway, in which large compartments containing organelles destined for destruction fuse with one or several lysosomes to initiate product breakdown [39]. Many proteins are targeted for breakdown by ubiquitination signals. Ubiquitin is a ubiquitously expressed small regulatory protein that attaches as single or multiple copies to proteins assigned for breakdown [40]. It has also been shown that caveolin1 is targeted for breakdown via ubiquitination and that caveolae/caveolin- 1 is degraded in lysosomes [41-43]. Cytoskeleton The transport of endocytic vesicles typically occurs along the cytoskeletal network, such as microtubules, to reach different cellular compartments. The cytoskeleton primarily consists of microtubules, actin filaments and intermediate filaments. Microtubules are slightly larger than the other cytoskeletal components and are primarily responsible for the cellular transport of vesicles and the migration of chromosomes during cell division. Actin is primarily responsible for the stabilization of cell morphology and is also important during cell migration by advancing the membrane. Actin filaments are slightly smaller than microtubules. Intermediate filaments stabilize the cell and function together with actin as the structural component of the membrane scaffold. Inside the cell, intermediate filaments and actin work closely together to stabilize the cell and protect it against mechanical stress. One notable characteristic of the cytoskeleton is that it consists of smaller subunits which can be combined into larger and smaller filaments as needed. The combined activities of the small assembled subunits make this system very dynamic and are a defining characteristic of 6

16 cytoskeletal function [44]. Free diffusion of larger structures would take too much time, and typically, only proteins that are free in solution diffuse freely. Most large assemblies or vesicles are directed to the appropriate site. The actual movement of vesicles on the cytoskeleton is mediated by myosin proteins [45-47], to which vesicles are connected via adaptor proteins. Myosin is situated directly on the actin filament and drags the vesicle along in an ATPase activity- dependent manner, providing directionality to to vesicle trafficking and accelerating vesicle delivery. 7

17 Caveolae Caveolae are membrane structures that have been observed under the microscope for more than 50 years [48], but their exact function remains unclear. The caveolae are rigid lipid raft- type domains that display a characteristic invaginated cup/omega- shape which is constantly open to the outside of the cell. It is similar in size to a clathrin vesicle, approximately nm, and its coat is not clearly visible in an EM micrograph, unlike a clathrin pit, which has a clearly visible coat. High in cholesterol and sphingolipids, the rigid caveolar structure primarily consists of the caveolin1 protein, which inserts into the membrane, and the helper proteins cavin- 1, cavin- 2 and EHD2, which comprise the outer coat of the caveolae (Fig 3) [26]. Not all cell types exhibit caveolae. Caveolae are primarily found Fig 3. Basic representation of caveolar structure. EHD2 oligomers are depicted as ring structures in green and red around the caveolae neck, and the blue lines represent the cavin coat complexes. The EM micrograph of the caveolar structure is adapted with permission from [49]. in endothelial cells, adipocytes and smooth muscle cells. There have been many suggestions for the possible functions of caveolae, including signaling, lipid homeostasis and mechanosensation [26, 50]. Caveolae are anchored to the plasma membrane via the cytoskeleton [44]. Actin, the primary 8

18 stabilizer of caveolae, forms intricate networks around the caveolae structure [44, 51]. Caveolae transport and dynamics have been suggested to occur along the microtubule network [52], assisted by myosin motor proteins. Caveolae biogenesis The primary structural protein of caveolae is caveolin1. Together with membrane lipids, caveolin1 comprises the bulk of the caveolae and is most likely responsible for inducing the curvature of caveolae and their binding to cholesterol, which is an essential component of caveolae. There are also Fig 4. Biogenesis of caveolae. Caveolin1 is synthesized in the ER and is subsequently transported to the Golgi where it becomes modified and aggregates. Then, the caveolin1 cluster is transported to the plasma membrane, where the complete caveolar domain formation occurs. Caveolin1 represented in red. helper proteins, such as cavin1-4 [53], EHD2 [54] and pacsin2 [55], which are closely associated with the caveolar structure and contribute to the formation of fully invaginated caveolae. The caveolin1 assemblies are believed to be formed in the Golgi before they are transported to the plasma membrane (Fig 4). There is a pool of caveolin1 in the Golgi that is not easily 9

19 labeled or visualized except in certain cell types, such as CV- 1 or COS- 7 cells, which contain a large Golgi pool [56], predominantly due to the high abundance of caveolin1 in the plasma membrane. There is also a pool of caveolin1 being localized to lipid droplets, but unless the regulation of caveolin1 is disturbed, this pool is not clearly detectable, except in adipocyte cell lines (Fig 5). Currently, the general consensus is that caveolin1 is synthesized in the ER and is transported for further Fig 5. The trafficking and turnover of caveolin. Caveolin1 clusters are labeled red. Caveolin1 is also localized to the lipid droplet coat. modification in the Golgi where it aggregates prior to its transport to the plasma membrane, where various helper proteins, together with caveolin1, form the fully invaginated caveolar structure. It is also believed that the association with cholesterol in the caveolar structure and the palmitoylation of caveolin1 occurs in different compartments of the Golgi [49, 57]. Cholesterol plays a major structural role in caveolae [49]; removal of cholesterol yields flat caveolae, and in fluorescence microscopy, the signal completely disappears [42]. Because membrane proteins are notoriously difficult to crystallize, obtaining the structural information of domains such 10

20 as caveolae can be challenging. Some studies have attempted to computationally model the caveolae and the caveolin1 protein [27, 49, 58], but because these are modeling results, the actual caveolin1 structure is unknown. Functions of caveolae Signaling It was the finding that GlycosylPhosphatidylInositol (GPI)- linked receptors appeared to cluster in caveolae that led Lisanti et. al. to propose the signaling hypothesis (Fig 6) [59]. They also suggested that their GPI- linkage and high glycosphingolipid content render caveolae detergent- insoluble. Since then, many receptors such as the insulin receptor have been shown to be present in caveolae [60, 61], which further suggests that caveolae are important for signaling. It appears likely that this structure is involved in signaling, as it is a lipid raft- type structure that could easily cluster receptors within its rigid membrane. Because removing this structure in mouse models, such as via caveolin1 knockdown results in a higher incidence of cancer, caveolar dysfunction is likely associated with signaling defects. Caveolae have also been suggested to be important for PIP2 signaling and precursor metabolites of PIP2 have been detected in caveolae. Additionally, the addition of Epidermal Growth Factor (EGF) has been shown to exert an effect on these metabolite pools, supporting the role of caveolae in PIP2 signaling [62]. 11

21 Fig 6. Basic representation of the caveolar structure. EHD2 oligomers are depicted as ring structures in green and red around the neck of the caveolae and the blue lines represent the cavin coat complexes. Signaling receptors are clustered in the membrane in green and orange. Lipid homeostasis Adipocytes contain an abundance of caveolae, covering a third of the cell surface, suggesting an important role of caveolae in this cell type. Knocking down caveolin1 or otherwise removing the caveolar structure in mice models yields lipid regulation defects [63, 64]. These mice become insensitive to diet- induced obesity and other insulin- related effects. In addition, these mice become leaner and contain less diet- induced LDL in the bloodstream [65]. These findings indicate that there are either transporters in the caveolae which transport lipids into adipocytes or that caveolae directly contact the lipid droplets in the adipocytes. This process could 12

22 mediate, either in the presence or absence of transporters, the translocation of lipid species from the cellular environment to the lipid droplet inside the adipocyte (Fig 7). A portion of the cellular caveolin1 pool is sequestered in lipid droplets, and it has been shown that an N- terminal truncation mutant Fig 7. Basic representation of the caveolar structure, depicting a speculative lipid transport pathway. Lipid transport could be mediated by a transport protein present in the caveolae or by a direct association between the lipid droplet and the caveolae. EHD2 oligomers are depicted as ring structures in green and red around the neck of the caveolae and the blue lines represent the cavin coat complexes. Speculated transport protein depicted as orange bar and lipid droplet in purple. of caveolin1 is localized to lipid droplets to a much greater extent, suggesting a direct role of caveolin1 in lipid binding and in the vesicular transport of lipids [66-68]. However, this pool is only a subset of the total caveolin1 in the cell. Therefore, this process is most likely coupled to a secondary transport system, as well. 13

23 Mechanosensation and mechanorelief Thanks to an array of experiments by different research groups a model for caveolae mechanosensation and mechanorelief in muscle cells has been proposed [26, 42, 69]. The caveolae are comprised of large amounts of membrane lipids, which have been proposed to be able to flatten upon mechanical stress, providing relief to the cell structure (Fig 8) [70]. It was conceived that this function could occur via proteins that are released when the caveolae flatten, such as a cavin protein, which are coupled to a "sensing" cascade in the cell [26]. This mechanism could be important for Fig 8. Image depicting the basic mechanics of caveolar mechanosensation. Upon mechanical stress (depicted by the prodding "finger") caveolae become flattened, thus relieving the pressure. This system could be coupled to a sensing cascade by releasing a caveolae- associated protein. smooth muscle cells, for which the pressure on the vessels might change rapidly and might also be coupled to Nitric Oxide Synthase (NOS) signaling for vasodilation [71, 72]. It has been suggested that a portion of the protein coat remains when the caveolae flatten, because the striations on the caveolae remain detectable after caveolar flattening. As mentioned 14

24 previously, a portion of the coat is most likely released upon stress, especially the cytosolically bound proteins, but a portion of the coat remains detectable after stress and continues to bind to antibodies [69, 73]. Furthermore, some reports have suggested that caveolin signaling is related Fig 9. Schematic representation of the involvement of CliC/caveolae in cell migration. Caveolae sequester inhibitory molecules, which are transported to the posterior region of the cell, whereas inducers of movement are transported together with the new membrane to the anterior region via CliC transport. to the activation and inhibition of cell migration [26, 74, 75]. It has been shown that in migrating cells, the Cell division cycle 42 (Cdc42) protein is localized to the front of the moving cell, whereas the caveolin is clustered at the back of the cell [75]. This organization has been suggested to occur due to the sequestration of regulatory proteins such as Rho GTPases in the caveolae, inducing an inhibitory effect on these proteins. Relocalizing the caveolae to the posterior region of the cell would release the inhibition at the anterior region, enabling migration (Fig 9). Cdc42 is known to regulate Graf1 and Graf1- mediated endocytosis [12-14], which has also been shown to be important for cellular migration. Knockdown of Graf1 yields similar 15

25 defects in migration and polarization to caveolin1 knockdown, further strengthening the roles of these proteins in cell migration [14, 75]. Role of caveolae in disease Certain diseases, such as cancer, lipodystrophy and caveolin3 mediated muscular deficiencies, have been associated with deficiencies in caveolae and the loss of caveolae from tissues exhibiting caveolae [63, 64, 76-79]. The term for these (caveolin3- related) disease deficiencies in muscle tissue is caveolinopathy [80]. In the case of cancer, it might be difficult to identify an exact reason for the observed increase in cancer incidence; however, certain molecules most likely use the caveolar structure as a basis for signaling. The increased cancer incidence could also be associated with the regulation of movement via the suggested inhibitory effect of caveolae on movement [75]. Cell migration has been shown to be important for cancer metastasis and tissue invasion. Lipodystrophy may also occur due to the dysfunction of this molecular transport mechanism. The high abundance of caveolae in adipose tissue suggests an important role of caveolae in these cell types, as demonstrated by lipid transport deficiencies, but whether a specific transport protein or the overall caveolar structure is responsible for these defects remains unclear [65]. It has been suggested that the caveolin protein is responsible for the transport of lipids by binding to lipid droplets. The lipid droplet coat has also been shown to contain small amounts of caveolin, and because caveolin has been suggested to bind to cholesterol, this conclusion appears to be logical. A mutant of caveolin1 or caveolin3 truncated at amino acid 53, referred to as Cav1 DGI and Cav3 DGV respectively, mislocalizes to lipid droplet- like structures when overexpressed, suggesting an important regulatory domain at the N- terminus of caveolin [66-68]. In the case of caveolin3 deficiency, particular mutations have been discovered 16

26 that result in deficiencies in muscle tone and control, and some of these mutations are hereditary [79]. This class of diseases includes limb- girdle muscular dystrophy and muscle rippling disease [78, 80, 81]. Caveolae coat proteins Caveolin The caveolae coat is thought to be comprised of caveolin and cavins, forming large assemblies in the membrane. Curvature- inducing proteins, such as EHD2, Pacsin2, are also associated with the caveolae that, together with caveolar core components, may contribute to the formation of the complete caveolar structure. The primary structural protein in caveolae is caveolin (Fig 10). Caveolin1 is a short protein that is partially inserted into the membrane and is believed to bind to cholesterol, facilitating the formation of the specialized caveolar lipid domain [49]. Caveolin1 include an N- terminal domain (1-82) that is believed to be responsible for caveolar regulation [82]. Removal of this domain causes the formation of lipid droplet- like structures in cells. Caveolin1 contains a scaffolding domain (82-102) which is thought to be responsible for protein- protein interactions and binding to cholesterol. Next, the membrane interacting domain ( ) is completely inserted into the membrane. It does not cross the membrane as a transmembrane protein; instead, both ends of the protein protrude into the cytosol. The C- terminal domain ( ) contains palmitoylation sites, which are believed to be modified in the Golgi, participating in protein stabilization and attachment. These palmitoylations may also help to localize the caveolin molecules to the rigid domain structures near cholesterol [49, 83]. The exposed portion of caveolin is in the cytosol. Therefore, all interactions with caveolin are believed to occur inside the cell, either cytosolically or directly in the membrane [50, 84]. It is also believed 17

27 that caveolin1 induces the curvature of caveolae. The exact mechanism underlying the curvature of caveolae remains speculative, but a recent finding indicated that the inserted portion of caveolin1 reaches the other side of the membrane and pulls the phospholipids, thus generating membrane curvature (Fig 11) [85]. The prediction data originally presented by the group of Rob Parton suggested that the primary component of the membrane- inserted caveolin1 consists of three helices [49]. More recent prediction data suggests that this region only consists of two helices [85], Fig 10. Representation of the structure of caveolin1. The membrane- inserted region and the scaffolding domain are presented in greater detail in the lower window. Adapted with permission from [49]. which would further complicate the signaling model based on the direct interaction with the scaffolding domain of caveolin. The original investigators of the structure of caveolin suggested that the scaffolding 18

28 domain is too close to the membrane to allow for direct interaction. Based on an analysis of proteins containing the scaffolding domain motif, there was no association between those proteins [84]. This result suggests that caveolin- based signaling occurs via some form of communication between proteins other than direct binding, or if signaling does occur via direct binding, this binding occurs near the N- terminal domain of caveolin1. Similar in sequence and domain structure to caveolin1, caveolin2 is also a component of the caveolae and forms heterooligomers with caveolin1. Fig 11. Model of caveolin1 membrane domain and scaffolding domain inserted into the membrane pulling on the phospholipids on the opposite side of the membrane. Adapted with permission from [85]. Normally co- expressed with caveolin1, caveolin2 cannot alone form the caveolar structure. Unlike overexpressed caveolin1, which forms caveolae in a cellular setting, caveolin2 requires caveolin1 for caveolae formation [86]. Caveolin3 is muscle- specific and most likely performs similar functions to caveolin1 in this tissue because removal or mutation of the caveolin3 gene causes severe effects similar to caveolin1 removal in other tissues [77, 87]. 19

29 Cavins Recently, cavins have been identified as integral structural components of caveolae. The various proteins in this family are believed to play different roles in the formation and dynamics of caveolar structure [88, 89]. The cavin proteins were initially discovered to participate in other aspects and functions of the cellular machinery and, thus, have been referred to by alternative terms that are related to their functions. Cavin- 1 is also called Polymerase I and Transcript Release Factor (PTRF), cavin- 2 can be referred to as Serum Deprivation response PRotein (SDPR), cavin- 3 is named PRotein Kinase C Delta- Binding Protein (PRKCDBP) or Serum deprivation Response factor- related gene product that Binds to C- kinase (SRBC) and finally cavin- 4 is also referred to as MUscle Restricted Coil- coil protein (MURC) (Fig 12) [53]. In the rest of the thesis, these proteins are hereafter referred to as cavin1-4. Cavin- 1 is essential for the formation of the caveolar structure. If this protein is removed, for example, via sirna knockdown, the typical invaginated caveolae do not form, and caveolin is downregulated [90]. Cavin- 2 and cavin- 3 appear to be nonessential because removing them does not prevent caveolar structure formation, but induces other effects. It has been shown that removal of cavin- 2 yields severe effects in lung tissue, suggesting an important role of cavin- 2 in the lungs [91]. Cavin- 3 has been suggested to play some role in caveolar localization and trafficking [92]. Cavin- 4 is muscle- specific and is expressed at high levels in this tissue, suggesting a specialized role for this protein. The other cavins are also expressed in muscle tissue, suggesting that cavin- 4 does not substitute the function of the other cavins but rather performs a distinct regulatory function in this tissue [53]. The coat visibility of the caveolae via EM is low, especially compared to the coat of clathrin- coated vesicles, but in electron microscopy, a low- visibility coat has been observed, which has been suggested to reflect protrusions of caveolin1 [49]. Using an electron microscopy technique coined "deep- etch", developed in the lab of John 20

30 Heuser [93], it is possible to detect the striated coat on the bulbs of the caveolae [42]. It has been suggested that this clearly visible coat primarily consists of the cavin proteins at that the caveolin protein is difficult to detect. EM analysis combined with single- molecule analysis Fig 12. Schematic representation of the cavin domain structure as prediction data show. NLS = Nuclear Localization Signal, LR = Leucine Rich domain and PEST = Pro- Glu- Ser- Thr rich sequences. PEST has been suggested to the important for protein degradation. Adapted from [53]. revealed that cavin- 1 forms distinct cavin1- cavin1, cavin1- cavin2 and cavin1- cavin3 subcomplexes, and that these subcomplexes can be present on the same caveolae but are always separate. This analysis also indicated that the different subcomplexes formed separate striations on the caveolae that are arranged in a meshwork that helps form the complete caveolar structure [94]. These coat striations appear to display different patterns depending on the expression level of each cavin protein [95]. This coat also appears to be present on flattened caveolae [69], suggesting an intricate interaction between these proteins. These data also suggest that the current theory of caveolae disassembly and caveolin breakdown upon perturbation [42] might be incorrect or at least uncertain and that most proteins remain present even after cholesterol removal/mechanical perturbation, which might explain the rapid recovery that we have detected after reintroducing cholesterol to this system. 21

31 EHD2 EHD2 is a dimeric ATPase known to tubulate liposomes in vitro [54]. It was shown to oligomerize around the tubules in evenly shaped rings of varying size providing flexibility to the system. EHD2 was also shown to heavily tubulate membranes when overexpressed in HeLa cells. Structurally, EHD2 (Fig 14) consists of a membrane- binding interface that is slightly curved, which most likely is of great importance for its membrane- bending capacity. EHD2 consists of a central G- domain which only displays binding specificity for ATP, and ATP hydrolysis has shown to be important for the oligomerization of EHD2 dimers. ATP hydrolysis and subsequent EHD2 oligomerization has also been shown to be important for the normal function of EHD2 because removal of these activities abolishes protein binding in a cellular setting. EHD2 contains an NPF- motif (a three amino acid Asn- Pro- Phe chain that is known to bind to the EH domain pocket) in a flexible linker at the side of the protein dimer that may be involved in protein- protein interactions. NPF- motifs are found in many proteins involved in endocytic transport and targeting. EHD2 also contains an EH- domain, which is known to interact with the NPF- motif. In EHD1, this interaction is important for endocytic targeting, but for EHD2, this interaction might not be important, as removal of the EH- domain from EHD2 does not prevent its binding to intracellular structures [54]. In addition to our own studies, the structural elements of EHD2 have been thoroughly examined [54, 96]. It has been suggested that EHD2 is localized to the neck of the caveolae (See Paper I) [95, 96]. This localization appears to be logical, as it is known that EHD2 oligomerizes in a ring structure and belongs to the dynamin- like protein superfamily [54]. The theory was that, at least initially, EHD2 would act as a facilitator of scission because it is a known tubulator. EHD2 may play this role, but most data would suggest it acts as a stabilizer of caveolae structure. Recently, PIP2 has been shown to be important for the localization of EHD2 to the membrane and that 22

32 inhibition of phospholipase D, which influences PIP2 production via PI4P conversion, reduces the membrane localization of EHD2 [97]. The importance of PIP2 in the membrane localization of EHD2 has been shown in the original study in which EHD2 was characterized [54]. PIP2 has been shown to be present in caveolae [98], and it has also been shown that PIP2 Fig 14. Ribbon representation of an EHD2 dimer. The N- terminal loop is modeled in the G- domain pocket. Adapted with permission from [54]. is important for the localization of EHD2 to the plasma membrane/caveolae [97]. In contrast to Epidermal growth factor receptor pathway substrate 15 (Eps15) that bind PIP2 via its EH2 domain, EHD2 is suggested to use another motif for PIP2 interaction [97, 99]. This motif has yet to be determined, although PIP2- containing membranes appear to be important for EHD2 binding. PIP2 may serve as a determinant of caveolar localization, and EHD2 specification and binding may be derived from other residues in its membrane- interaction domains. 23

33 Pacsin2 Pacsin2 (also referred to as syndapin2) is a F- BAR- domain- containing protein (Fig 13) that has been shown to tubulate membranes [100]. In addition, pacsin2 contains an NPF- motif, which has been shown to interact with the EH domain pocket. Overexpressed pacsin2 was shown to associate with caveolae and affect the morphology of the caveolar structure. It was also suggested that pacsin2, as a BAR- domain protein, is essential in the formation of the caveolar structure [55, 101]. Pacsin2 was also shown to bind to the EH domain of EHD2, but the exact nature of this interaction remains unclear. BAR- domain- containing proteins have been shown to be important for the dynamin- dependent scission of clathrin- coated pits [10, 11], and it could be that pacsin2 mediates the same role in caveolae scission. Although there have been reports of the dynamin- mediated scission of caveolae [102, 103], the occurrence of this event remains controversial, mostly because we cannot confirm the presence of dynamin in caveolar structures or any effect of dynamin inhibitors on caveolae. Fig 13. The generic structure of the pacsin BAR- domain, presented as a space fill model; the colors reflect the charge of the residue. Adapted from [104]. 24

34 Implications for the function of caveolae It is likely that the different proposed roles of caveolae [105] are valid and that these activities are regulated by distinct combinations of the caveolins and caveolae- related proteins in different tissues, reflecting the role of caveolae in each tissue. We and others have observed that different tissues contain distinct amounts of cavin1-4 and caveolin1-3 [91], which suggests that a different combination of caveolin and caveolae helper proteins define the function of caveolae in that tissue. However, this concept remains speculative and must be further evaluated and verified before any conclusion can be drawn. 25

35 Aims 1. To understand and characterize the cellular role of EHD2. The cellular role of the dynamin- like, large ATPase EHD2 was not known and we aimed to identify its cellular localization and thereby elucidate its role in the cell. 2. To characterize the N- terminus of EHD2 and its function. In the crystal structure from Daumke et al. (2007) [54] the N- terminus of EHD2 was missing. Given that the loop was closely situated to the G- domain our aim for the second paper was to characterize this domain and its role in EHD2 function. 3. To study the role of the EH- domain in EHD2 dynamics. The EH domain of EHD2 was previously proposed to interact with the loop of the G domain following oligomerisation. The aim of the third paper was to characterise the importance of this interaction for the stable membrane association of EHD2. 4. To discover the function of cavin- 3 at the caveolae, and how it associates with the other cavins. Cavin- 3 has been proposed not to be essential for caveolae but to rather play a role in caveolae trafficking. Our aim was to discover the function of cavin- 3 as a coat component and how it associates with the other cavin proteins. 26

36 Brief methodological account Cell lines and basic function of the model cells HeLa is a fibroblast- like cancer cell line derived from cervical cancer. Named after the patient it was harvested from, HEnrietta LAcks, the HeLa cell line has been used as a human cell model for decades, which has made it very useful for protein expression and microscopic analyses. 3T3- L1 is a pre- adipocyte fibroblast- like cell line derived from mouse embryo. 3T3- L1 cells can be differentiated into mature adipose cells, but at that stage, proliferation stops. Because of its adipocyte- like characteristics, the 3T3- L1 cell line contains a vast amount of caveolae and related proteins, making it ideal for examining protein expression and colocalization with the caveolar structure. If this cell line is differentiated it expresses even more caveolae structures and and proteins, but it becomes more tissue- like, making microscopy more difficult. Fluorescence microscopy There are many approaches to examine the clustering of proteins in internal structures and membrane structures via fluorescence microscopy. Traditionally, UV- lamps (mercury) have been used as a light source together with specific filters at the required wavelengths. However, for modern fluorescence microscopes specific lasers and mirrors are used. Fluorophores can be excited at specific wavelengths and then they release light at a different wavelength that can be observed. One of the most used fluorophores is the Green Fluorescent Protein (GFP), which is a naturally occurring fluorescent protein from jellyfish. There are many fluorescent alternatives which can be either overexpressed or attached as tags, but the benefit of using fluorophores of biological origin is that they are less damaging to cells. Typically, for fluorophores, the light is observed from the 27

37 side of the excitation source so that the light observed is only derived from the sample and not from the light source. For modern confocal microscopes, the light is observed through a variable pinhole, which enables the removal of unspecific background, meaning that, effectively, only a small slice of the cell is observed. For older microscopes, this selection of light is not possible, such that signal is also detected from outside the focal plane. Examining dynamic processes can be performed using fluorescence microscopy and a fast CCD- camera. However, the resolution of standard fluorescence techniques is limited to 200 nm, so other techniques can be used to complement standard fluorescence microscopy, such as high- resolution STochastic Optical Reconstruction Microscopy (STORM). STORM is a technique in which reactivatable fluorophores are sequentially imaged, bleached and reactivated for thousands of images, enabling the construction of a composite image from individual fluorescent points. This process enables the refinement of the signal from each fluorophore, resulting in reduced background [ ]. There are also hardware- based superresolution techniques, such as Stimulated Emission Depletion (STED) microscopy. This technique is based on laser depletion of the signal around the laser, which provides a more resolved fluorescent spot from the sample [112]. TIRF Total Internal Reflection Fluorescence Microscopy (TIRF or TIR- FM), is a technique in which light is reflected at the glass- water interface, generating an evanescent wave [113]. This wave only penetrates a short distance into the sample, approximately 100 nm, which enables clear visualization of the events that occur at or near the membrane without any background signal. This technique is particularly beneficial when performing live- cell experiments, especially because only part of the cell is illuminated, which 28

38 reduces the stress to the cell compared to a standard experiment. Because only a small part of the cell is illuminated, the general fluorophore pool remains longer- lived, enabling the exchange of fresh fluorophores for the bleached fluorophores, which increases the possible imaging time. Tracking Certain software and algorithms enable the detection and tracking of spots between frames by drawing lines along the trajectory of each fluorescent spot. Many of these algorithms developed in- house appear to be MATLAB (MathWorks, USA)- based, but we have primarily used the Imaris software package (BitPlane, Switzerland). Imaris allows us to track a multitude of fluorescent protein spots and gives a number of statistical parameters that can be plotted. In paper 3, track duration and track displacement were primarily examined, facilitating the measurement of the movement of the fluorescent spots and their lifetime. When used in combination with TIRF microscopy, Imaris is a powerful tool to measure the lifetime and dynamics of membrane- associated structures. 29

39 FRAP Fluorescence Recovery After Photobleaching (FRAP) is a technique in which an area of the cell, defined as a Region Of Interest (ROI) is bleached using high laser power, and then, the recovery of the fluorescent signal is observed (Fig 15). This recovery is used as a model of the exchange of the bleached fluorophore- protein with unbleached fluorophore- protein. Highly dynamic protein assemblies recover rapidly, whereas highly stable structures recover slowly, if at all, depending on the stability of the assembly. As an example, caveolin1 hardly recovers at all in the short- term Fig 15. Schematic image of FRAP operation. An ROI in the cell is bleached using high laser power, and then, the recovery of the signal is measured. because of its high stability, whereas a purely cytosolic protein recovers almost instantly. Proteins which are cytosolic in nature but peripherally bind to and release from membrane structures, such as cavin- 1, display slower recovery than purely cytosolic proteins but typically recover after a while. In my experience, stable protein assemblies never completely recover, mostly because not all of the protein is exchanged in the short- term. Complete exchange most likely takes much more time. 30

40 AFM microscopy Atomic Force Microscopy (AFM) is a technique in which a very small (nanoscale) tip attached to a cantilever is moved by applying picocurrents to change the piezoelectric element, causing the cantilever to oscillate. Then the tip is applied to the sample, which is probed using the tip [114]. The data obtained from AFM is typically topographic, but additional information, including attachment and interaction strength, can be obtained using recently developed analysis tools. It is also possible to use functionalized probes, facilitating the performance of experiments such as antibody binding affinity assays. It is also possible to measure on cells, but this approach requires much more tweaking because these samples are much more sensitive than comparable samples imaged in air [115]. Traditionally, AFM has been used to examine amyloid fibrils and sheets of chemically stable compunds, which are relatively simple to fix and measure in air. However, in the last few years, advancement in techniques and microscopes have enabled measurements of samples in liquid, such as cells in culture [116], force measurements of samples [117] and high- speed AFM [118, 119]. Measuring in the fluid on live cells using the force- volume technique enables the user to elucidate physical properties such as membrane stiffness [120]. Using high- speed AFM, the structural changes in membrane pores and other cellular machineries have been elucidated [121]. These advanced techniques enable the imaging of cellular structures at a resolution previously limited to electron microscopy, which only enables snapshot images of cellular processes. Using an unroofing technique originally developed for EM (removal of the upper section of the cell membrane, leaving only a portion of the cell membrane attached to the substrate intact) has enabled us to use AFM to measure the inside of the cellular membrane to elucidate the height of the caveolar structure (Fig 16) [93, 122, 123]. It may also be possible in the future to use functionalized probes to examine interactions with the caveolar structure, such as the 31

41 determination of which proteins are actually at these structures and the protein binding affinites to identify the proteins at the caveolae which are most likely to bind to or interact with other proteins [115, 116, ]. Fig 16. Figure depicting an overlay of a fluorescence image and an AFM micrograph performed on the same sample. Fluorescence microscopic images are in the top row, and AFM images are in the bottom row. In the lower right panel, a deflection and fluorescence overlay is presented. Adapted with permission from [122]. The primary technique that we have applied is Quantitative Nanomechanical Mapping (QNM), which is a recently developed technique that determines the mechanical properties of the sample by mathematically analyzing the input data from the microscope. Based on the Bruker peakforce tapping mode, in which the tip approaches the sample in a "tapping" motion, this technique records and determines the properties of the sample in every pixel of the imaging area. In addition to sample topography, adhesion, deflection (tip deflection error) and modulus (sample stiffness) can be measured. It is also possible to obtain tip force curve readouts from each pixel of the imaged area. 32