Do antiangiogenic protein fragments have amyloid properties?

Angiogenesis, the formation of new blood vessels from preexisting vasculature, occurs in physiologic and pathologic processes, including embryonic development, the menstrual cycle, wound healing, inflammation, and tumor growth (for a review, see Carmeliet and Jain¹ ). In 1971, Folkman²  postulated that tumor growth is dependent on angiogenesis. Accumulating evidence indicates that drugs that target the growing vasculature are promising new therapeutics for the treatment of cancer. Based on the concept that a primary tumor produces inhibitors of angiogenesis that can inhibit the outgrowth of metastases, the first endogenous angiogenesis inhibitor, termed angiostatin, was purified in the laboratory of Folkman in 1994.³  The identification of angiostatin prompted the search for other angiogenesis inhibitors, resulting in the isolation of endostatin,⁴  prothrombin fragment 1 and 2,⁵  cleaved antithrombin III,⁶  fibrin(ogen) fragments (Brown et al⁷  and A.R., unpublished observations, September 1999) and many other inhibitory polypeptide fragments. Commonly, such antiangiogenic polypeptides are proteolytically cleaved or denatured derivatives of endogenous proteins. Endostatin, angiostatin, and a thrombospondin peptide are currently being tested in clinical trials.^8-10 

In 1997, Folkman and colleagues reported that treatment of experimental tumors with endostatin induced complete regression.¹¹  These promising results from animal studies fueled the great expectations that accompanied the proposed use of antiangiogenic therapy with endogenous inhibitors of angiogenesis. However, the initial results could not be reproduced by others and lessened the initial enthusiasm.¹²  Controversies regarding the antitumor effects of endostatin arose and initiated a heated discussion about its efficacy.¹²  The variability in endostatin bioactivity is still unexplained. Originally, endostatin was purified from tissue culture supernatant. In the in vivo experiments that showed dramatic tumor regression a recombinant denatured and insoluble form of endostatin produced in Escherichia coli was used.¹¹  In a number of subsequent studies endostatin has been used from a number of different sources. These different endostatin preparations caused less dramatic, minor, or even no effects.^13-16  For studies in patients, a soluble globular form is used, produced in Pichia pastoris.⁸  However, this globular form has no or very little effect on endothelial cells.¹⁷  We found that only denatured endostatin is toxic to cells.¹⁸  We wondered whether our findings might explain the observed variability in endostatin bioactivity and to what extent our observations could be extrapolated to other angiogenesis inhibitors.

Similar to certain antiangiogenic polypeptides, toxic polypeptides involved in conformational diseases such as Alzheimer disease (AD), light-chain amyloidosis, pancreatic islet amyloidosis, and spongiform encephalopathies are proteolytic fragments or conformationally changed forms of generally harmless proteins. Properly folded proteins form a stable 3-dimensional structure. However, protein fragments (such as endostatin) are often prone to (partial) denaturation and undergo sequential aggregation steps leading to the formation of highly ordered aggregates. Initially, (partly) unfolded proteins or protein fragments associate with each other and form small soluble aggregates, preceding the formation of protofibrils. Ultimately, mature fibrils accumulate and are deposited as plaques. Fibrillar aggregates are classified as amyloid fibrils based on the presence of cross-β structure. Cross-β structures in amyloid fibrils are stacked β sheets composed of flat and nontwisted β strands. This will result in a unique and flat 2-dimensional β-sheet surface, not seen in globular proteins (Figure 1). The list of amyloidogenic peptides arising from proteolytic processing of harmless proteins is long and still increasing (for a review, see Selkoe¹⁹ ). In addition to disease-associated amyloid formation, intriguing physiologic roles for amyloid fibril formation have been elucidated.^20,21  These recent reports highlight such a behavior of specific proteins or protein fragments in microorganisms as well as in eukaryotes. It is expected that numerous other proteins will be found to form amyloid as a natural product.

The formation of amyloid aggregates and their relation to disease are studied intensively (for reviews, see Selkoe¹⁹  and Dobson²² ). Just recently a common, although unknown, mechanism of cell toxicity has been implied for protein misfolding diseases. Toxicity is an inherent property of aggregates of denatured proteins and is not related to the amino acid sequence composition of the protein.²³  Although the nature of the toxicity remains unclear, recent studies suggest that the toxicity of amyloid proteins is enclosed in soluble oligomers, rather than in fibrils.^24-26  This notion is supported by the finding that antibodies that detect early amyloidogenic aggregates can inhibit their toxicity.²⁷  Many studies have indicated a close relationship between the presence of amyloid depositions and vascular damage. Transmissible spongiform disease is associated with the presence of the abnormal form of prion protein that aggregates into amyloid fibrils. Prion deposits are toxic to a wide variety of cells, including cerebral endothelial cells.²⁸  Damage to endothelial cells is also seen in AD. Patients with AD exhibit significant cerebrovascular pathology in addition to neuronal degeneration.^29,30  Furthermore, it is known that patients with AD suffer from stroke and that hemorrhages are present.³¹  In vitro experiments revealed that amyloid β increases permeability of endothelial cell monolayers and induces apoptosis.^32,33  Other studies have shown that smooth muscle cells and endothelial cells are damaged in cerebral blood vessels of those with AD.^33-35  Finally, the diabetes-related amyloidogenic peptide amylin and the circulating amyloidogenic peptide β₂-microglobulin damage blood vessels.³⁶  We recently showed that insoluble endostatin forms amyloid fibrils, suggesting that endostatin may induce apoptosis as a consequence of its amyloid structure.^18,37  In AD endostatin colocalized with amyloid β-positive plaques that were surrounded by focal gliosis.³⁸  These studies indicate that the cross-β structure has direct effects on endothelial cells. Is the cross-β structure then a common denominator in polypeptide fragments with antiangiogenic activity?

In general, denatured proteins can stimulate plasminogen activation by tissue-type plasminogen activator (tPA).³⁹  Amyloid fibrils such as amyloid β and prion protein markedly stimulate plasminogen activation by tPA.^{40, 41}  We found that the presence of cross-β structure is a prerequisite for tPA binding and for activation of plasminogen.³⁷  We noticed that many antiangiogenic compounds, generated through proteolytic cleavage or denaturation, stimulate tPA-mediated plasminogen activation (Table 1).

It is intriguing that unrelated proteins lacking sequence homology acquire antiangiogenic activity on proteolytic cleavage or denaturation and at the same time acquire the ability to stimulate tPA-mediated plasminogen activation. Toxicity of aggregates of denatured proteins is also independent of the amino acid sequence.^23,27  Taken together, this suggests that a common antiangiogenic pathway may exist that is induced by tPA-binding proteins. We hypothesize that antiangiogenic polypeptide fragments contain the cross-β structure and that this structure is the common denominator responsible for their antiangiogenic and antitumor effects. If correct, the presence of cross-β structure in antiangiogenic proteins could represent yet another example of a physiologic function of protein aggregation as has been published recently for microbial and eukaryotic proteins (for a review, see Huff et al²⁰ ).

Can tPA contribute to the antiangiogenic effects of these compounds?⁴¹  tPA is historically known for its role in blood clot lysis (fibrinolysis). tPA cleaves the zymogen plasminogen into the active serine proteinase plasmin. Pericellular plasmin subsequently degrades a fibrin clot. Apart from their well-established role in clot lysis, tPA and plasmin also function in other processes, including long-term potentiation⁴²  and neuronal cell migration.⁴³  tPA has also been identified as a mediator of neuronal cell death following ischemia or excitotoxic injury in the brain.⁴⁴  Intracerebral injection of excitotoxins such as glutamate and focal cerebral ischemia provoke degradation of the extracellular matrix protein laminin and cause neuronal death.^45-47  We found that tPA activation by amyloid endostatin induced plasmin formation, degradation of the extracellular matrix protein vitronectin, and cell detachment.¹⁷  Matrix proteins including vitronectin, fibronectin, and laminin play an important role in cell survival⁴⁸  and their breakdown can result in apoptosis. Induction of plasminogen activation leads to endothelial cell detachment,⁴⁹  inhibition of cell adhesion,⁵⁰  endothelial cell destruction,⁵¹  or regression of capillary tubes.⁵²  Most recently, Angles-Cano and coworkers showed that plasminogen activation induces detachment and apoptosis of Chinese hamster ovary (CHO) fibroblasts⁵³  and smooth muscle cells.⁵⁴  tPA is constitutively expressed by CHO fibroblasts and smooth muscle cells and appears to be responsible for pericellular plasminogen activation.^53,54  Finally, a possible role for tPA and plasmin in amyloid-mediated toxicity is further supported by the finding that serpins (a family of protease inhibitors) can inhibit amyloid toxicity. Taken together, activation of tPA by amyloid polypeptides may induce excessive degradation of extracellular matrix with subsequent loss of cell attachment sites resulting in cell detachment and apoptosis.

If our hypothesis is correct, other cross-β structure receptors in addition to tPA may mediate the antiangiogenic response. Good evidence indicates that at least 2 other receptors that bind amyloid polypeptides, CD36 and the receptor for advanced glycation end products (RAGE), could be involved.^55,56  Many potential mechanisms for amyloid-induced cell toxicity have already been proposed. It is well established that amyloid-induced neuronal cell toxicity involves the production of reactive oxygen species and disruption of calcium homeostasis. Endothelial cell toxicity caused by Aβ is also associated with the production of superoxide free radicals, causing oxidative lipid and protein damage, and increased intracellular calcium concentration.^32,57,58 

If correct, our hypothesis implies that (1) it is relevant to look at structural properties of angiogenesis inhibitors, rather than at their primary amino acid sequence and to consider the possibility that an active drug requires a cross-β structure; (2) the inhibitory effect of the drugs listed in Table 1 may be improved by adding factors that are known to increase amyloid-mediated toxicity, such as lipopolysaccharide (LPS), tumor necrosis factor α (TNF-α), and interferon γ (IFN-γ)^59,60 ; and (3) it is necessary to use these polypeptide fragments with caution because they may form toxic aggregates.

Moreover, tumor-derived endogenous circulating inhibitors may also possess cross-β structure. Besides their antiangiogenic activity, these circulating proteolytic fragments may, however, then also contribute to some of the clinical complications of cancer, such as amyloidosis or bleeding events.

Note added in proof. After this article was accepted for publication, we became aware of new results obtained by Paris et al.⁷⁷  The authors showed that amyloid β peptides inhibit angiogenesis. Their results add amyloid β peptides to the list of angiogenesis inhibitors that interact with tPA, and support our hypothesis that antiangiogenic fragments in general may have amyloid properties.