Mechanotransduction and anoikis: Death and the homeless cell

Abstract

Developed organs display strict spatial organization of differentiated cells which is required for proper organ function. One important device that prevents tissue disorganization is the death of cells that lose anchorage to their native matrix, a signal that indicates potential loss of proper tissue context. Termed anoikis (Greek for Homelessness), this form of cell death is a specialized form of apoptosis. Interestingly, at certain stages of development and tissue repair, cells are required to migrate in an unanchored state, suggesting that anoikis must be strictly regulated at some level. Likewise, cellular transformation is often accompanied by an inappropriate loss of anoikis and subsequent acquisition of a metastatic phenotype. Despite its importance, the molecular pathways involved in the regulation of anoikis and the proximal signals reporting loss of anchorage are poorly understood. Recent studies suggest that attachment may be reported by a mechanosensory testing of the cell’s physical environment.

Introduction

Cells continually monitor their environment by sensing soluble mediators, cell-cell contact and mechanical signals, leading to fundamental changes in cell fate. One important example of such environmental sensing is attachment context, which allows survival of properly localized tissue cells, particularly of epithelial and endothelial lineages. In theory, allowing detached cells to survive would permit a kind of cellular vagrancy which could lead to ectopic placement of tissue cells in foreign locations within the body. The concept that differentiated tissue cells have a “home” within the body led to the recognition that detachment of such cells leads to a specific form of cell death, appropriately termed anoikis (homelessness).¹ Accordingly, inappropriate loss of anoikis is thought to facilitate wandering of cells to remote sites, as occurs during metastasis of cancer cells. In this review, we briefly cover current concepts regarding mechanisms of anoikis.

Attachment to Matrix is an Important Environmental Cue

Detachment-induced cell death was originally described in endothelial cells,² and was subsequently studied in detail in MDCK epithelial cells.¹ In this latter report, a number of far-reaching observations were made which continue to shed light on the biological significance of attachment context, yet have defied a unifying mechanistic model. First, tissue cells are not all equally sensitive to anoikis: normal fibroblasts, for instance, were found to resist death following detachment from matrix. Indeed, most mesenchymal cells require serum starvation in vitro to display anoikis. This finding indicated that predisposition to this specialized form of cell death must be specifically and actively regulated. Secondly, epithelial cell anoikis was found to be bypassed by oncogenic Ras or Src, and restored by adenovirus E1a, linking its regulation with transformation and differentiation pathways. Interestingly, anoikis was reduced by HGF/scatter factor, providing a link to cell motility. Consistent with this observation, it was also noted that anoikis was augmented in post-confluent, contact inhibited states.¹

For the most part, these seminal observations suggest a model whereby anoikis would be most pronounced in a differentiated epithelial phase and suppressed during normal or pathologic mesenchymal transitions. It should be noted, however, that exceptions may still exist. For instance, overexpression of the cyclin kinase inhibitor p16, commonly increased in differentiated and senescent states, restores anoikis in malignant Capan-1 cells;³ however, ectopic p16 expression instead confers resistance to anoikis in nontransformed MCF-10A cells.⁴ The reason for this discrepancy is unclear but may reflect a difference in the actions of p16 in malignant versus nonmalignant cells, or may be specific to the cell lines studied.

This spectrum of sensitivity to anoikis aligns with an epithelial-mesenchymal axis in normal states, in development, and in malignant transformation. Anoikis regularly ensures cell death during scheduled cell detachment, as occurs in both gut endoderm and skin ectoderm, epithelia with recognized spatial differentiation gradients. During cardiac development, the formation of the cardiac cushion is a well known endothelial to mesenchymal transition event requiring invasion of the cardiac jelly by endocardial cells. This transformation involves Jagged1/Notch signaling, which explains cardiac cushion abnormalities in both humans and mice defective in Jagged/Notch pathways.^5,6 More recently, Notch activation has been shown to suppress anoikis, presumably through Slug-dependent cadherin repression,⁷ again linking anoikis suppression with the acquisition of a mesenchymal phenotype.

During embryonic development, a variety of tube or branching tube structures are formed, either from the rolling up of established epithelial sheets or from cavitation of solid cords of cells. The latter form of tube morphogenesis is exemplified early in vertebrate development by cavitation of the inner cell mass. This process involves apoptotic regression of cells not anchored to the circumferential basement membrane, and is thus a form of anoikis.⁸ Anoikis also participates later in development during gland formation. This has been best studied in mammary gland formation, in which the lumen forms during ductal expansion through apoptosis of the inner cell population. Because the inner acinar cells lack attachment to the outer extracellular matrix shell, this has been considered a form of anoikis.⁹ Interestingly, the BH3-only proteins Bim and Bmf are induced upon loss of matrix attachment, and Bim knockdown inhibits anoikis in MCF-10A cells in vitro while Bim^−/− mice have delayed mammary duct clearance in vivo.^10,11

Finally, pathologic suppression of anoikis is thought to comprise one specific facet of malignant transformation. Clearly, not all neoplastic cells can metastasize despite being locally invasive (e.g., basal cell carcinomas of the skin). While a direct correlation between loss of anoikis and metastatic capacity has not been proven, the most aggressively metastatic carcinomas such as small cell lung cancer can survive in suspension in vivo (as in pleural fluid collections) and in vitro, thus bypassing anoikis. A more direct proof of concept study involved a genome-wide functional screen for genes conferring suppression of anoikis in vitro, which yielded the neurotrophin receptor TrkB. Expression of TrkB by nontransformed intestinal epithelial cells both abrogated anoikis and allowed aggressive metastatic behavior in vivo, suggesting that loss of anoikis may be sufficient to endow cells with metastatic capacity.¹²

Mechanisms of Cell Death in Anoikis

There is little debate that anoikis represents a specialized form of apoptosis, as detached cells display release of cytochrome c, DNA laddering and nuclear pyknosis. Interestingly, anoikis proceeds more consistently with induction of Bim and Bmf.^10,11,13 However, the events leading to this pathway are less clear. Most studies document activation of executioner caspases such as caspase 3 and 7, which may in part function through cleavage of MEKK1 into a proapoptotic protein.¹⁴ The role of more proximal caspases remains unproven. Caspase 8 is cleaved and activated in MDCK epithelial and in endothelial cells, suggesting participation of a death receptor.^15,16 Indeed, silencer of death domains (SODD/BAG-4) and a dominant negative FADD truncation both suppress anoikis, confirming the participation of death domain proteins. Again, however, the necessity for proximal caspase activation has not been definitively proven. In FSK-7 epithelial cells, for instance, anoikis is accompanied by a caspase-independent release of mitochondrial Bax.¹⁷ More recently, CHO cells were shown to undergo anoikis at least in part from release of a novel mitochondrial protein, Bit1, which facilitates apoptosis in a caspase-independent fashion as well.¹⁸

Besides activation of death pathways, withdrawal of survival signals clearly contributes to the initiation of anoikis, although here again, conflicting reports defy easy identification of a unifying pathway. Detachment often triggers dephosphorylation of FAK, and activated FAK renders MDCK cells anchorage independent,¹⁹ suggesting a key role for FAK suppression in anoikis. FAK dephosphorylation, however, also occurs following detachment of anoikis-insensitive osteosarcoma cells, which instead appear to rely on PI3K-dependent signals for survival in suspension.²⁰ Overexpression of integrin-linked kinase also blocks anoikis in mammary epithelial cells in a fashion not requiring FAK activity.²¹ Further, lung carcinoma cells appear to resist anoikis in both a FAK and PI3K-independent fashion, activating instead a Src/Pyk2 pathway,²² and oncogenic Ras confers anoikis resistance to rat intestinal epithelial cells through a PI3K and Akt-independent mechanism.²³ Overall, such studies seem to indicate that hyperactivation of any one of several survival pathways may disable detachment-induced death, although it is far from clear that deactivation of these pathways is the primary trigger for detachment-induced death in normal cells. The question of what initiates the death cascade also raises an equally puzzling question: what is the actual attachment sensor?

Attachment Sensation versus Shape Change

Although integrin function is known to be central to attachment sensation, a number of observations suggest that a unidirectional outside-in integrin pathway may be insufficient to relay attachment context. Hepatocytes adherent to poly-L-lysine or other synthetic non-integrin-binding matrices, for instance, have suppressed integrin signals yet spread, attach and survive.²⁴ While such experiments do not exclude integrin binding to endogenously secreted matrix, they suggest that binding and flattening may itself allow survival. Likewise, integrin ligation without structural matrix rigidity is insufficient to prevent anoikis. Endothelial cells forced into suspension are not rescued from death by either soluble matrix proteins or solid-phase RGD peptides bound to microbeads,²⁵ and epithelial cells likewise die in suspension despite integrin ligation with RGD peptides;¹ thus integrin ligation and clustering alone, which does not prevent cell rounding, also does not block anoikis. In a similar fashion, fibro-blasts grown in a stressed collagen matrix undergo apoptosis when the matrix is mechanically unloaded, causing abrupt shape changes despite presumably continued attachment to the matrix.²⁶

Thus, an alternative view of attachment sensation is that cell rounding—that is, shape change—may be a physical cue that cells have lost anchorage. The importance of cell shape in determining cell fate has been well described. By plating cells on micropatterned matrix islands of differing geometry, death of endothelial cells was clearly shown to be related to the inability of cells to spread as opposed to matrix surface area available for binding.²⁷ In this study, cells allowed to spread by stretching between small islands of matrix survived, whereas cells forced to round up by attachment to a single matrix patch of equivalent total surface area underwent apoptosis.

The teleologic reasoning behind the significance of shape change is not known but may be related to the necessity of tissue cells to gauge the stiffness of their surrounding environment. Naive mesenchymal stem cells plated on collagen I-coated matrices of varying elasticity follow strikingly different differentiation programs, with osteogenic, myogenic or neurogenic profiles being induced by decreasing levels of substrate stiffness.²⁸ In a separate study, mesenchymal stem cell differentiation was determined by altering cell shape, with rounded cells becoming adipocytes and cells allowed to flatten undergoing osteogenesis.²⁹ Importantly, in both studies the sensation of either matrix stiffness or cell shape was dependent upon the generation of cytoskeletal tension. In the latter study, constitutive activation of the RhoA effector ROCK induced osteogenesis independent of cell shape, suggesting that cell shape may be a secondary readout of the ability of cells to generate tension. Thus matrix attachment, shape change and cytoskeletal tension may ultimately all reflect the cell’s attempts to assess its physical environment.

Mechanosensation and Anoikis

Returning to anoikis, one may ask whether detachment from a solid surface presents one extreme in matrix stiffness (essentially infinite elastance), causing cell rounding and unopposed cytoskeletal tension. The concept that anoikis is triggered by a mechanosensory event fits well with the importance of integrins, which are well known mechanotransducers. Integrin clusters such as focal adhesions respond dynamically to applied external or cytoskeletal internal forces,³⁰ and matrix-coated microspheres restrained with an optical trap induces proportionate internal cytoskeletal tension through integrin sites.³¹ Thus integrins are well positioned to serve both efferent and afferent arms of an “inside-outside-in” loop thought to be initiated by cells to test local matrix stiffness.³²

In this scenario, anoikis may be viewed as the outcome of a continual tension test in which candidate attachment sites, preferentially integrin complexes, fail to support an internal contractile force due to loss of external anchorage, much like a tug-of-war game in which one side suddenly releases the rope. Returning to the original observations surrounding anoikis, several questions arise: first, since anoikis is regulated, what protein(s) initiate the tension test, and second, how does load failure initiate apoptosis?

Clues may be offered by hematopoietic cells which, by virtue of their peripatetic lifestyle, are both anchorage independent and insensitive to substrate stiffness, despite expressing a variety of integrins and having adhesive capacity. One integrin-associated scaffold, Shc, displays differential expression in parenchymal versus hematopoietic cells. The SHC1 gene expresses three proteins which differ in the amino terminal length, resulting in p66^Shc, p52^Shc and p46^Shc. Whereas the two shorter isoforms appear to be constitutively expressed, p66^Shc is repressed in hematopoietic cells.³³ Shc is also known to be involved in mechanotransduction,³⁴ rendering it an interesting candidate to transduce attachment signals. The role of p66^Shc in mediating anoikis was recently studied in immortalized 293 cells, which repress p66^Shc and escape anoikis, and primary endothelial cells, which express p66^Shc and display robust anoikis.¹³ Indeed, forced expression of p66^Shc restores anoikis in 293 cells whereas siRNA-mediated knockdown of p66^Shc suppresses endothelial cell anoikis. Importantly, p66^Shc expression also correlates with signs of cytoskeletal tension such as focal adhesions, stress fibers and focal RhoA activation, and truncations or point mutations which delocalize p66^Shc off of focal adhesions also fail to initiate anoikis.¹³ These morphologic features are consistent with the rounding and loss of actin stress fibers in myoblasts subjected to p66^Shc antisense treatment.³⁵ Suppressing cytoskeletal tension with inhibitors of actinomyosin contraction decrease p66^Shc-dependent anoikis, consistent with the concept of a tension test.¹³ These data square well with other studies in breast epithelium in which expression of the actin-binding protein tropomyosin-1 correlates with development of stress fibers and sensitivity to anoikis, whereas loss of tropomyosin-1 expression is seen in breast cancer specimens.^36,37

Answers to the second question, how mechanical load failure can lead to apoptosis, are not known. Certainly, internal cytoskeletal tension in an unanchored state may be expected to be borne entirely by internal struts such as intermediate filaments and microtubules. Keratin 15 and 17 themselves are subject to caspase-dependent cleavage during anoikis,³⁸ perhaps further weakening the inner scaffolding. More interestingly, the two BH3 domain-only proteins implicated in physiologic anoikis during mammary gland development, Bim and Bmf, are sequestered on microtubules and microfilaments through association with respective dynein motor complexes, and become active upon release from the cytoskeleton.^39,40 In the case of Bmf, cell detachment causes release of Bmf from actin filaments with neutralization of Bcl-2 and consequent cell death.⁴⁰ While it is not clear that these proteins are necessary for the execution of anoikis, they suggest that the cell has at least two mechanisms to respond to cytoskeletal collapse with apoptosis, and these data strongly implicate cytoskeletal perturbation in distal anoikis signaling.

Future Directions

One of the more striking observations in reviewing the literature of anoikis is the number of apparently discrepant results, suggesting that no single linear pathway is responsible for cell death caused by matrix detachment. Thus, a principal challenge is to recognize major themes which explain anchorage sensation and initiation of apoptosis. While the “tension test” hypothesis put forth above does much to integrate several large areas of study, a number of broad questions remain. For instance, is p66^Shc a universal explanation for the initiation of cytoskeletal tension, or can other proteins initiate this response? The p66^Shc−/− mouse apparently develops normally, and has a principal phenotype of longevity and reduced oxidative stress,⁴¹ suggesting the latter. There is also the riddle of how p66^Shc is linked to control of RhoA. Notably, the scaffold 14-3-3 has been shown to bind p66^Shc,⁴² and is known to regulate cytoskeletal architecture.⁴³ Of potential significance, knockdown of 14-3-3 restores sensitivity of lung cancer cells to anoikis.⁴⁴ With respect to malignancy, RhoA activation is generally associated with increased proliferation and invasion and not cell death,⁴⁵ raising the question of whether different cellular subpopulations within a tumor can be responsible for metastasis as opposed to local invasion, or whether some tumor cells have suppressed a distal part of the pathway specific for anoikis. Finally, it is as yet unclear whether anoikis is truly turned on during epithelial and endothelial cell differentiation, and whether this represents an irreversible commitment step. Neurospheres, for instance, retain stem-like capacity until they become adherent, at which point differentiation proceeds irreversibly. It is plausible that attachment context and thus anoikis pathways only become induced when a progenitor cell reaches its ultimate destination, its new “home.”