Citation for published version (APA):
Breurken, M. (2011). Molecular probes for imaging tissue remodeling. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR693440
DOI:
10.6100/IR693440
Document status and date: Published: 01/01/2011
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
Molecular probes for imaging tissue
remodeling
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de
Technische Universiteit Eindhoven, op gezag van de
rector magnificus, prof.dr.ir. C.J. van Duijn, voor een
commissie aangewezen door het College voor
Promoties in het openbaar te verdedigen
op dinsdag 25 januari 2011 om 16.00 uur
door
Monica Breurken
Dit proefschrift is goedgekeurd door de promotor:
prof.dr. E.W. Meijer
Copromotor:
dr. M. Merkx
A catalogue record is available from the Eindhoven University of
Technology Library
ISBN: 978-90-386-2410-5
Cover design: Koen Pieterse
Prof. dr. E.W. Meijer Dr. M. Merkx
Prof. dr. R.J. Pieters Prof. dr. T.M. Hackeng Prof. dr. C.V.C. Bouten
Chapter 1 Visualization of the extracellular matrix 1
Chapter 2 From phage display to dendrimer display: 29
testing a strategy to obtain high affinity probes
Chapter 3 Collagen targeting using multivalent protein-functionalized 49 dendrimers
Chapter 4 Protease activatable targeting based on protein cyclization 65
Chapter 5 Hybrid of CNA35 and a synthetic collagen mimic allows 79 protease activatable collagen targeting
Summary 99
Samenvatting 102
Curriculum Vitae 107
Chapter
1
Visualization of the extracellular matrix
Most cells in multicellular organisms are surrounded by the extracellular matrix (ECM), which consists of numerous macromolecules and has various functions like providing structural support and enabling cell communication and migration. The ECM is always slowly adapting, but excessive turnover is usually associated with pathological conditions. Therefore, imaging of the ECM can play a key role in the understanding and diagnosis of diseases and the design and evaluation of therapeutics. This chapter provides an overview of imaging approaches to visualize the dynamics of ECM turnover. Attention is paid to the structure and function of the ECM and the process of remodeling. The principle of multivalency and examples of multivalent ligands are discussed as multivalent interactions can be valuable for the design of molecular probes. The most abundant protein in the ECM is collagen. An interesting protein that can be used for collagen targeting is CNA35. Structure and function of both collagen and CNA35 are outlined and progress in research that focusses on collagen targeting is reviewed. In the final part of this chapter molecular imaging of enzymatic activity is discussed.
1.1 Extracellular matrix in health and disease
1.1.1 Structure and functionThe extracellular matrix (ECM) is a complex, highly organized macromolecular network of large and multifunctional proteins surrounding all cells in multicellular organisms.[1] ECM
proteins are typically composed of multiple domains that are frequently repeated within in the same molecule and have evolved with the emergence of multicellular organisms millions of years ago. A number of genes encoding these proteins show a surprisingly good conservation in time. In vertebrates, proliferation of ECM proteins occurred by expansion with novel proteins as well as evolution of pre-existing families.[2-4]
The ECM serves as a structural support for cells, thereby providing strength and elasticity to tissues. Moreover, it also determines cellular functions through cell-matrix interactions to contribute to essential processes, like survival, migration and proliferation of the adjoining cells.[5-8] The macromolecules that constitute the extracellular matrix are mainly
produced locally by cells in the matrix. In most connective tissues, ECM molecules are secreted by cells called fibroblasts.[1] Two main classes of molecules that constitute the matrix are
proteoglycans and glycoproteins.[1, 9] Most glycoproteins form fibrillar structures that in turn
tend to form fibers and other three dimensional heterogeneous aggregates or networks. Collagens are the most abundant glycoproteins in mammals. Being the major component in for instance skin and bone, these triple-helical proteins are responsible for providing tensile strength to tissues.[1, 10, 11] Elastin is the main component of elastic fibers, which give tissues the
required resilience.[1, 11] Besides elastin, also fibrillin forms a structural component of these
elastic microfibrils.[9] Among the group of important ECM proteins are also fibronectin and
laminin, which can associate with multiple ECM factors and components.[9-11] The
proteoglycans in connective tissue form a highly hydrated, gel-like structure in which the fibrous proteins are embedded. The arrangement and concentration of different macromolecules gives rise to a wide diversity of ECM forms in various types of tissue.[1, 9]
1.1.2 Remodeling by matrix metalloproteinases
In the seemingly static extracellular matrix there is always a slow turnover, with a continuous degradation and resynthesis of ECM molecules. This continuous turnover is necessary to enable cell division and cell migration, as well as adaptation of the matrix to changing stresses that the tissues are subjected to.[1] However, extensive remodeling of the
characteristic feature of many diseases, including plaque formation in atherosclerosis, angiogenesis, myocardial infarction and tumor metastasis.[12-15]
A family of Zn2+ dependent enzymes called matrix metalloproteinases (MMPs) play an
important role in the catabolism of ECM components, during both normal tissue turnover like development, homeostasis and wound healing, and a variety of diseases associated with abnormal turnover. They are thought to be the major enzymes responsible for ECM degradation. There are at least 25 members of the MMP family. MMPs generally consist of a prodomain, a catalytic domain, a hinge region, and a hemopexin domain (Figure 1.1). Prior to their activity in the extracellular space, MMPs are activated by the removal of their propeptide domain. Although the exact functions and interactions of the hemopexin domain are not completely unraveled yet, the hemopexin domain is known to mediate binding to native collagens, possibly by trapping the triple helix into the active site cleft of the catalytic domain. Furthermore, it is involved in the interactions with specific MMP inhibitors.[16, 17]
Figure 1.1 Domain structures of the matrix metalloproteinases (MMPs). Common domains within the MMP family are a prodomain, a catalytic domain, a hinge region and the hemopexin domain. All MMPs are Zn2+
dependent.
MMPs are categorized into six groups based on their substrate specificity and domain structures. MMP-1, MMP-8, MMP-13 and MMP-18 belong to the group of collagenases, which can cleave interstitial collagens (type I, II and III) into characteristic ¾ and ¼ length fragments of the fibers, providing the first step in the degradation process.[17-19] Cleavage of collagen at the
proteases. The collagenases can also digest other ECM molecules and soluble proteins, however.[19] MMP-2 and MMP-9 are gelatinases. These enzymes can digest denatured
collagens, gelatins.[19] Three repeats of a type II fibronectin are present in the catalytic domain
of the gelatins, which bind to gelatin, collagens and laminin.[20] However, MMP-2 (but not
MMP-9) can also cleave soluble or fibrillar collagen, although this collagenolytic activity is not a general property of gelatinases.[21-23] Among the other groups of MMPs are the stromelysins,
which can digest a number of ECM components and participate in pro-MMP activation.[17, 18]
Studies using synthetic peptide substrates and X-ray crystallographic studies of MMPs have shown that the MMP substrate specificity is not only depending on the collagen primary structure.[24, 25] Therefore, a model of the cleavage sites in interstitial collagens (type I, II and III)
was developed based on the combination of primary and supersecondary structures. This model suggests that the cleavage site regions are distinguished by a low content of charged residues in addition to being tightly helical prior to the cleavage site and loosely triple-helical following the cleavage site (Figure 1.2).[26]
Figure 1.2 Model of the mammalian collagenase cleavage site in interstitial collagens. The four triplet region that precedes the scissile (Gly – Ile/Leu) bond is rich in imino acids, whereas the four triplet region that follows this bond is imino acid deficient. The overall 24 amino acid residue region is hydrophobic, containing <5% charged residues. Figure reproduced with permission from [26].
Most MMPs are synthesized as pre-proenzymes. The signal peptide is removed during translation and inactive proMMPs are generated, also called latent MMPs or zymogens. Activation of these zymogens is thus an important regulatory step of MMP activity and can be induced in the extracellular space by serine proteases (e.g. plasmin) or already activated MMPs.[19, 27, 28] Proteolytic activation of MMPs is mostly stepwise. The initial proteolytic attack
the propeptide is removed, this probably destabilizes the rest of the propeptide, including the interaction between the cysteine in the prodomain and the zinc in the catalytic domain (cysteine switch), allowing further processing by other enzymes.[29-31] The activation of MMP-2
is regulated differently and takes place on the cell surface.
Another level of MMP regulation involves tissue inhibitors of metalloproteinases (TIMPs). These specific inhibitors can bind MMPs in a 1:1 stochiometry. Their expression is regulated during development and tissue remodeling. Under pathological conditions associated with unbalanced MMP activities, changes in TIMP levels are considered to be important because they directly affect the level of MMP activity. TIMPs have an N- and C-terminal domain with each containing three disulfide bonds. The N-C-terminal domain folds as a separate unit with MMP inhibitory activity.[17, 19]
The level of active MMPs is thus depending on a delicate balance including the level of gene transcription, proteolytic cleavage of the propeptide for activation, the secretion of specific inhibitors and also the ability of MMPs to degrade non-specific protease inhibitors.[32]
The tightly regulated activity of MMPs is essential for matrix remodeling and breakdown and therefore, quantification and visualization of the activity of MMPs is important for understanding of many biological and pathological processes. Currently available methods include the use of labeled antibodies or zymography.[33, 34] However, labeled antibodies do not
probe enzymatic activity and zymography can only be used for ex vivo tissue analysis. To better understand the involvement of MMP activity in healthy and diseased tissue, MMP activatable imaging probes are required that allow direct monitoring of MMP activity in vivo.
1.1.3 MMP activity during disease processes
Remodeling of the ECM is involved in many disease events. MMPs play an important role in the response to infection by degrading components of the ECM, but also by for instance modulating cytokine and chemokine activity.[32] The migration of immune cells to sites of
inflammation requires proteolysis of the basement membrane and it is likely that the migration of all inflammatory cells also require MMP activity, although this remains to be proven in vivo.[32] On the other hand, excess MMP activity can result in the adverse effect, leading to host
tissue damage instead of facilitation of an effective immune response. Increased MMP secretion or decreased TIMP secretion may result in immunopathological phenomena, such as pulmonary cavitation, increased blood brain barrier permeability and neurotoxin production. Tuberculosis, HIV infection, Hepatitis B and Lyme disease are examples of infections where MMP activity is implicated to be responsible for the disease progress.[32] MMPs are also
a wound matrix to provide a scaffold to direct cells to the injury as well as to stimulate them to proliferate and synthesize new ECM. As healing proceeds, the initial ECM of the scar tissue undergoes remodeling and eventually the injured tissue is repaired, thereby partially restoring structure and function of the tissue.[35]
Besides pathological processes derived from infections and the corresponding immune responses, excess ECM turnover is known to occur after myocardial infarction, especially in the left ventricle.[15] A myocardial infarction occurs when a previously atherosclerotic coronary
artery becomes totally occluded. This occlusion can lead to necrosis of the cells in the myocardium due to metabolic distress. In response to this process, remodeling of the ECM by MMPs is induced to enable migration of cells towards the diseased tissue and scar tissue formation to replace affected tissue. ECM remodeling after myocardial infarction can in turn lead to dysfunction of the heart and the scarring process can eventually result in the occurrence of severe atrial fibrosis by excessive accumulation of fibrillar collagen deposits.[14, 15, 27]
Eventually, this can disturb the coordinated contraction of the heart muscles, a phenomenon called atrial fibrillation. Other well-known pathological processes implicated with abnormal ECM remodeling are tumor metastasis, vascular remodeling of hypertension and atherosclerosis.[11, 36, 37] In case of atherosclerosis, plaques are formed in the artery walls that are
mostly formed by macrophages that contain compounds like lipids and fibrous connective tissue, usually as a result of a decrease in coronary flow. These plaques are surrounded by a fibrous cap consisting of connective tissue. In absence of inflammation of injury, these fibrous caps show little evidence of active ECM synthesis and degradation. However, when tissues are injured, a repair process starts involving ECM turnover. Diminished collagen synthesis by overexpression of MMPs will weaken the strength of the fibrous cap resulting in an increased risk of plaque rupture.[38, 39]
1.1.4 Importance of extracellular matrix imaging
Since extensive ECM remodeling is an indication of a variety of disease events, molecular imaging of structural ECM components has gained a lot of interest. Determination of the ECM changes that are trigger points for remodeling is important for the development of therapeutics. ECM fibers can be visualized by physical methods allowing high resolution, such as electron microscopy, X-ray diffraction and atomic force microscopy. The overall structure of the heterogeneous fibres or fibre-fragments is therefore rather well-known, but characterization of individual components is much more difficult.[9] Ideally, one should be able to visualize
different ECM components over time and location to assign key roles to specific ECM components instead of detecting overall morphological changes in the matrix. The ECM
proteins are attractive targets for molecular imaging, since they are abundant in the human body and extracellular and thus well accessible for targeting. Moreover, the abundance as well as the characteristic repetitive structure of most ECM proteins provide a lot of potential binding sites, making the concept of multivalency a promising approach for the development of high affinity molecular imaging probes.
Instead of using structural changes as a marker for disease processes, enzymatic activity can also serve as a biological marker. Imaging MMP activity would allow even more distinctive and sensitive detection of tissue remodeling. The ability to repeatedly and noninvasively map the protease activity in real time in vivo can be an important tool for early disease diagnosis, even before gross structural changes in the ECM occur. Moreover, understanding the activities of MMPs can be the key to successful drug development and therapy, in light of evidence that some MMPs may have protective effects in disease, while others can cause deleterious effects.[40, 41]
1.2 Multivalent ligands for targeting and imaging
1.2.1 Principle of multivalencyInteractions in many biological systems often rely on multiple low affinity interactions, a phenomenon called multivalency. Multivalent ligands can undergo multiple, identical interactions simultaneously in order to increase affinity and specificity, resulting in unique properties that are fundamentally different from properties displayed by their monovalent analogues.[42-44] These ligands play an important role in cell-surface interactions (e.g. adhesion
of a virus to the cell surface), the immune system (e.g. target binding of antibodies), cell-matrix interactions and cell-cell interactions.[43, 45]
Different mechanisms of action for multivalent ligands can be distinguished (Figure 1.3). The structural valency of a ligand refers to the number of binding sites present. This structural valency is not necessarily the same as the functional valency, which indicates the number of binding sites that are actually involved in the interaction and thus functionally active.[42] When a divalent ligand interacts with a monovalent receptor, the affinity is already
increased by a factor 2 due to a statistical effect. Besides this statistical effect, an additional increase in affinity can occur upon binding of a multivalent ligand to multiple sites of a receptor, if the geometries of both ligand and receptor match, which is known as the chelate effect. In addition to an increase in affinity, clustering receptors on cell membranes can take place upon multivalent binding of a ligand, which often activates distinct intracellular signal transduction pathways.
Figure 1.3 A) Binding of a multivalent ligand to a single receptor resulting in an increase in the apparent affinity due to a statistical effect. B) Binding of a multivalent ligand with multiple identical binding sites, which is called the chelate effect. C) Binding of a multivalent ligand causing receptors to cluster. Figure adapted with permission from [45].
The enhancement in strength of a multivalent association in comparison to the monovalent association is sometimes expressed by the ratio β, which is defined as Kamulti/Kamono.
Typically, in a multivalent interaction β is much higher than the valency of the ligand, demonstrating the practical value of multivalency for designing a high affinity system.[42, 43]
Another approach to compare multivalent interactions is by using the effective molarity, which is defined as the ratio between the intermolecular dissociation constant (Kdinter) and the
intramolecular dissociation constant (Kdintra).[46] This term gives more insight into the effect of
the length and flexibility of the linker present in the multivalent ligand and represents the enhancement in affinity of the multivalent interaction. In a bivalent interaction, the dissociation is initiated by an intramolecular dissociation of binding part of the ligand followed by the intermolecular dissociation of the complete ligand (Figure 1.4). The overall dissociation constant is determined by the monovalent Kd (Kdinter), the effective molarity (Meff = Kdinter / Kdintra)
Figure 1.4 A divalent interaction can be described using the effective molarity, which is defined as the ratio between the intermolecular dissociation constant Kdinter and the intramolecular dissociation constant Kdintra. The
dissociation is initiated by an intramolecular dissociation of a binding domain of the ligand followed by intermolecular dissociation of the complete divalent molecule. Figure adapted with permission from [45, 46].
1.2.2 Synthetic multivalent peptide and protein constructs
Using multivalent display is an attractive strategy for the development of (semi)synthetic ligands to enhance the affinity and specificity for biological targets. Carbohydrates, peptides, proteins or other recognition elements can be coupled multiple times to a variety of scaffolds that differ in size, shape and physical characteristics. Although binding of a multivalent ligand is critically depending on the recognition elements displayed on the scaffold, also the architecture of the scaffold itself can strongly influence its biological activity
[44, 45, 47], depending on the local concentration, distribution and orientation of the binding sites
of the target. The spatial arrangement of these binding sites should be considered for choosing the appropriate scaffold to enable multivalent binding.
A variety of scaffolds have been described in the literature for generating multivalent ligands. Polymers and dendrimers constitute an important class of these scaffolds.[48] Whereas
the generation of well-defined linear polymers to serve as a scaffold is still difficult [49], the
number of end groups of dendrimers can easily be controled during synthesis.[50] The
homogeneity makes dendrimers particularly suitable for biological applications, although the quite elaborate synthesis can be a serious drawback.
Design and synthesis of multivalent carbohydrate ligands for carbohydrate-protein interactions have been widely explored.[51-62] However, multivalent peptides and proteins have
been investigated less extensively. The first report of multivalent peptides was display of multiple antigen peptides (MAP).[63] Eight peptides were attached to a dendritic lysine core and
were found to be more immunogenic than the corresponding monovalent peptide. Dirksen and coworkers extended this application by ligating two poly(lysine) wedges together at the focal point via native chemical ligation (NCL) with one wedge displaying target binding peptides and the other wedge multiple DTPA moieties.[64] The oligopeptide containing the
arginine-glycine-asparagine-serine (RGDS) sequence, which is known to bind an integrin receptor that is overexpressed in tumors, was used as a model peptide for the synthesis of these multivalent peptide-based non-symmetric dendrimers. Successful synthesis of a dendrimer consisting of four RGDS peptide units and four as well as eight DTPA moieties was reported.
Another example of peptide multimerization on dendrimers is the multivalent display of cyclic RGD (cRGD) peptides on a dendritic scaffold to specifically target tumor tissue.[65-67] The cRGD peptides were divalently and tetravalently displayed and subsequently
conjugated with DOTA. A competitive binding assay showed that the affinity was enhanced from 120 nM for the monomeric cRGD, 69.9 nM for the dimer, to 19.6 nM for the tetramer. Tetravalent display of this peptide thus resulted in an approximately five-fold decrease in KD in
comparison to the monovalent analogue, and tumor uptake of the tetramer was significantly higher as compared to that of the monomer, as was shown in biodistribution studies.[66] This
five-fold increase in affinity corresponds primarily to the statistical effect of multivalent display, however.
Besides the covalent scaffolds, dynamic self-assembling structures, like micelles[68-73],
liposomes[74-77], and also amphiphilic nanofibers[78-80], have been successfully used to serve as
multivalent platforms for targeted drug delivery and molecular imaging. With the use of poly(ethyleneglycol) (PEG) modified phospholipids in these assemblies, micelles and liposomes gained a reasonable stability making them attractive for biological applications.[81]
The advantage of using liposomes or micelles as multivalent scaffolds is the relatively easy functionalization. After coupling of the recognition elements to the phospholipids, these can self-assemble to form multivalent ligands, thus no complicated synthesis steps are involved. However, unlike dendritic multivalent structures, liposomes and micelles are not well-defined, since there is a constant exchange of phospholipids between the dynamic assemblies.
A beautiful example of multivalent dynamic self-assembling structures is the functionalization of liposomes with peptides to develop anthrax and cholera toxin inhibitors. [74-76, 82] Anthrax toxin forms a heptameric pore in the cell membrane to enter cells. Phage display
was used to select peptides that inhibit binding of the toxins to cells.[82, 83] Liposomes were
functionalized with these peptides at various densities and tested for inhibition of toxin binding. Upon increase in peptide density, the IC50 value decreased from ~ 150 µM for the
monovalent peptide to an IC50 value in the low micromolar range. When a certain spacing
between the peptides displayed on the liposomes was reached, the IC50 values did not change
anymore, suggesting an optimal peptide geometry for interaction with the heptameric pore formed by the anthrax toxin.[75]
Less work has been performed on multivalent protein domains. An important example was the multimerization of single-chain antibodies to form structures with a valency up to four, which was reported in 2001.[84] Single-chain antibodies were found to spontaneously
self-assemble into multivalent structures with a valency depending on the length of the linker. However, this method for multimerization is limited to single chain antibody fragments. Fusing the protein of interest to self-associating domains or domains that can form hetero-dimers (e.g. use of the biotin-streptavidin interaction) is another useful method to obtain multivalent proteins. Although these methods are rather simple, the low structural valency that can be reached (typically divalent or tetravalent structures) limits the number of possible applications. Moreover, modification of the multivalent proteins to incorporate imaging moieties is difficult and usually involves non-specific coupling to the protein domains.
Few reports are available on protein-functionalized dendrimers and these typically involve nonspecific conjugation of monoclonal antibodies.[85-87] However, to prevent reaction of
the protein with the dendrimer near the active site and to keep control over the orientation of the protein after ligation, site-specific conjugation of the protein to the multivalent scaffold is preferred. The first general method for the synthesis of well-defined peptide and protein dendrimers was reported in 2005 by Van Baal and colleagues, who used native chemical ligation (NCL) to site-specifically couple peptides or proteins to poly(propylene imine) dendrimers with cysteine functionalized end groups.[88] Whereas up to eight peptides could be
ligated to the scaffold, this was more difficult for protein functionalization. Site-specific ligation of Green Fluorescent Protein (GFP) resulted in a mixture of di-, tri- and tetravalent protein dendrimers.[88] More recently, generation of semi-synthetic protein dendrimers based on a
strong, but non-covalent interaction between a peptide and complementary protein was reported.[89] The interaction between S-peptide and S-protein that together form the active
enzyme ribonuclease S was used as a model system. Four S-peptides bearing a C-terminal thioester were coupled to a tetravalent cysteine-functionalized dendritic wedge using NCL. Subsequently, the ribonuclease S tetramer was obtained by spontaneous association of an S-protein to each of the four S-peptides presented on the dendritic scaffold. The correct formation
of the multivalent enzyme complex was verified using isothermal titration calorimetry (ITC), surface plasmon resonance (SPR), mass spectrometry and an enzymatic activity assay.[89] This
semi-synthetic approach thus enables synthesis of well-defined protein assemblies in high yields.
1.3
Collagen imaging
1.3.1 Collagen structureCollagens are the most abundant proteins present in the ECM and are responsible for providing tensile strength to tissues. This heterogeneous family consists of at least 28 different types, each of these distributed differently in specific tissues.[90] All members of the collagen
family have one characteristic feature, a right-handed triple helical structure formed by three polypeptide chains. The helix can be formed by either three identical chains (homotrimers) as in collagen type II, III and others, or by two or three different chains (heterotrimers) as for instance in collagen type I, IV and V.[90-92] Each peptide chain within the molecule forms a
left-handed helix (α-chain) and these in turn supercoil in a right-left-handed manner to form the triple helix (Figure 1.5). Collagens are extremely rich in prolines and glycines, which are both important for the formation of the triple helix. Long stretches of the polypeptide sequence are repeats of the Gly-X-Y motif, where X is frequently proline (Pro) and Y is frequently hydroxyproline (Hyp).[90, 91, 93] Pro and Hyp stabilize the helical conformation in each α-chain,
because of their ring structure, while glycine is regularly spaced at every third residue throughout the central region of the triple helix, allowing close packing.[91] In each α-chain,
triple helical regions, called Col domains, are flanked by non-collageneous (non-Gly-X-Y) regions, which are called NC domains. The NC domains often show similarity to regions found in other matrix molecules.[90, 94]
Collagens are grouped based on their domain organization in fibril-forming collagens, fibril-associated collagens (FACIT), network-forming collagens, anchoring fibrils, transmembrane collagens and others, although 90% of the total collagen is represented by fibril-forming collagens.[91] Fibril-forming collagens are synthesized as soluble procollagens.
When secreted into the extracellular space, procollagen metalloproteinases convert these into collagens by cleavage of terminal propeptides.[92, 95] The collagen triple helices (~1.5 nm
diameter and 300 nm long) assemble into five-stranded microfibrils of approximately 4.5 nm in diameter. The microfibrils aggregate end-to-end and laterally, forming macrofibrils (10-100 nm diameter).[92, 96, 97] The mechanical and physical properties of a tissue depend on the diameter as
architectures (1 µm-2 mm, figure 1.5), which can be visualized with a light microscope on the scale of micrometers.[92, 95, 97] Large networks of fibrils and fibers can be formed by the formation
of crosslinks, determining the maturity of the collagen structures.
Figure 1.5 Structure of collagen. Each peptide chain within the molecule forms helical α-chain and these in turn supercoil to form a right-handed triple helix. The collagen triple helices can assemble into collagen fibrils, which can aggregate into larger cable-like bundles, called collagen fibers. Figure adapted from [98].
1.3.2 Conventional collagen imaging techniques
Monitoring in vivo changes in local collagen structures in time is important for understanding ECM related diseases. Its abundance and easy accessibility makes collagen an attractive marker for these diseases. Histological techniques or biochemical assays are available for measuring amounts of collagen. However, these methods are not suitable for in vivo research, because they require cell fixation or sacrificing tissue. Collagen can be visualized in tissues without the use of specific probes using techniques such as birefringence under polarized light[99], autofluorescence[100], and second harmonic generation.[101-103] In general, these
approaches are suitable to detect gross overall changes in collagen structures. However, these techniques do not provide enough sensitivity and specificity to visualize small, newly formed fibrils and fibers.
A small fluorescent collagen probe has been reported, namely dichlorotriazinyl aminofluorescein (DTAF). DTAF has been used for visualizing collagen fibers in live tissue. [104-107] However, DTAF is a non-specific fluorescent dye and is thus expected to have a low
neutral pH.[108-112] Another drawback is that DTAF binds collagen via a covalent bond, thereby
possibly influencing the collagen properties.
Fluorescently labeled antibodies are available to image specific types of collagen. However, the high affinity binding might affect tissue function and development. This is especially a problem for real time measurements to monitor collagen structures. For these experiments, binding of the imaging probes should not interfere with or influence the normal development and maturation of the ECM. Moreover, the relatively large size of the antibodies could limit tissue penetration and commercially available collagen antibodies are relatively expensive.
1.3.3 Collagen targeting probe based on the protein CNA35
CNA is a collagen binding protein that originates from Staphylococcus aureus. S. aureus is a pathogenic bacterium that is involved in various types of infections. CNA belongs to a family of bacterial surface proteins called MSCRAMMs (Microbial Surface Components Recognizing Adhesive Matrix Molecules) that mediate adherence of bacteria to the host tissue.[113] The
collagen binding protein CNA is approximately 135 kDa and it contains a signal peptide, an A domain followed by a varying number of B repeats, the cell wall anchoring region, a transmembrane segment and a short cytoplasmic tail (Figure 1.6A).[114-116] The A domain is
responsible for binding to several types of collagen, whereas the B repeats have no known effect on binding.[117-119]
Xu and colleagues showed that the N-terminal part of the collagen binding A domain of CNA (two-third of the complete A domain) has a 100-fold higher affinity for collagen than the full length A domain.[120, 121] This part of the protein is called CNA35 (residue 31-343) as it has a
molecular weight of approximately 35 kDa. The minimal collagen binding part is CNA19 (residues 151-318). However, CNA19 shows a lower binding affinity than CNA35.[119, 120]
In 1997 Symersky and colleagues elucidated the crystal structure of CNA19.[115] This
subdomain consists of two anti-parallel β-sheets and two short α-helices. The molecular surface of the collagen binding domain shows an apparent groove on one of the β-sheets, indicating a putative collagen binding region. Site-specific mutations showed that disruption of the walls of this binding trench results in a protein with a significant lower affinity for collagen than the wild type protein. Mutation of the tyrosine at position 175 to a lysine even results in total loss of the collagen binding capability.[114, 115, 122]
More insight into the mechanism of collagen binding has been provided by the crystal structures of CNA35 in the absence and presence of a collagen fragment (Figure 1.6B).[121] These
a long linker region. The N2 domain corresponds to the previously identified CNA19 structure. The C-terminal extension of this domain stretches towards the N1 domain and forms a β strand that complements one of the β sheets of the N1 domain. In this way a distinct hole between the two domains is created through which the collagen peptide penetrates upon binding. No electron density was observed for residues 332-344 and they are therefore assumed to be disordered.[121]
Figure 1.6 A) Schematic representation of a CNA protein containing a signal peptide (S), an A domain, three B repeats and the cell wall-anchoring region (CW), a transmembrane segment (TM) and a cytoplasmic tail (CT). The N-terminal two-thirds of the A domain is called CNA35 and the minimal binding part is CNA19. B) Crystal structure of CNA35 in complex with a collagen peptide. CNA35 consists of an N1 (green) and an N2 (yellow) domain that are connected via a linker (blue). The collagen peptide (purple) penetrates through the hole formed by the two domains upon binding. Figure reproduced with permission from [121]. C) A collagen binding model based on
the X-ray structure. The collagen triple helix first docks on the N2 domain (1), the collagen is then wrapped by the linker and the N1 domain (2), the N1 domain interacts with the N2 domain via multiple hydrophobic interactions and finally the C-terminal latch is introduced in the N1 domain to secure the ligand in place (3). Figure reproduced with permission from [121].
When CNA35 is in complex with a collagen peptide, the collagen triple helix penetrates through the hole between the N1 and N2 domains. The interactions between CNA35 and the collagen fragments are primarily hydrophobic. A collagen hug model was proposed for collagen binding of CNA35 (Figure 1.6C).[121] The apo-form of CNA35 is proposed to exist in
equilibrium between an open and closed conformation in solution. Only the open conformation is thought to be able to bind collagen. The binding of CNA35 to collagen is initiated by a low affinity interaction between the collagen triple helix and the N2 domain, followed by reorientation and restabilization of the N1 domain to ‘wrap’ the protein around the triple helix. Finally, the N1 domain interacts with the N2 domain by introducing the C-terminal latch into the N1 domain.
CNA35 binds to various collagen types and binding studies showed the presence of multiple non-identical CNA35 binding sites in collagen, exhibiting different affinities.[119, 122]
Different interactions with dissociation constants ranging from approximately 20 nM to 35 µM were observed in surface plasmon resonance (SPR) binding experiments for binding to type I collagen.[122] The presence of high affinity and low affinity binding sites in collagen results in a
complex overall binding behavior that can be attributed to the heterogeneous nature of collagen.
The generic collagen binding properties make CNA35 an attractive targeting ligand for imaging collagen. Within our group, a collagen specific fluorescent probe was developed by labeling CNA35 with Oregon Green 488 (OG488) via the amines present at the protein exterior.[123] Approximately three dye molecules were ligated per protein molecule. It was
shown that labeling of wild type (wt) CNA35 with OG488 by reacting with the N-hydroxysuccinimide ester does not significantly affect collagen binding and that no lysines near the collagen binding site are modified by coupling of a dye molecule.[123, 124] The lysines
that are favored for fluorescent labeling were identified and are located at the C-terminus of the protein and in a loop of the N-terminal subdomain.[124] The fluorescent probe was tested ex vivo
in various samples (cell culture, blood vessels and engineered tissues) and was proven to be successful in high resolution imaging of collagen organization.[123, 125] Structures ranging from
very young fibrils to mature collagen fibers could be visualized. Moreover, collagen imaging was possible without the need to remove non-bound CNA35 because of the high density of CNA35 binding sites in collagen. The formation of collagen could even be monitored in live tissues over prolonged periods of time.[125] Besides ex vivo experiments, the fluorescent collagen
probe was also used in vivo as a molecular imaging agent for detection of atherosclerosis in mice.[126] Atherosclerotic plaques showed a strong fluorescent labeling, illustrating the
The high density of CNA35 binding sites makes collagen attractive as a multivalent target. Due to its repetitive primary sequence, collagen is involved in many multivalent interactions in nature, such as platelet aggregation.[127, 128] Multivalent CNA35 structures have
been synthesized by coupling CNA35 to micelles en liposomes using NCL. SPR binding studies using the multivalent CNA35 micelles showed a drastic decrease in dissociation rate and interestingly, micelles en liposomes functionalized with multiple copies of a so-called ‘non-binding’ variant of CNA35[115] containing a Y175K mutation in the collagen binding site showed
a remarkable ‘restoration’ of the collagen binding capacity.[129, 130] Collagen targeting using MRI
as an imaging modality was explored by the development of a liposomal MR contrast agent that was functionalized with CNA35.[131] A set of in vitro experiments demonstrated the
potential of these CNA35 liposomes for in vivo molecular imaging of collagen. Using CNA35 as a molecular probe thus allows generation of high resolution images for visualization of collagen architectures.
1.3.4 Peptide-based collagen targeting
The relatively large size of CNA35 limits tissue penetration and the probe is not very specific, since it binds to various types of collagen. Collagen targeting using peptides is another promising strategy for molecular imaging. An advantage of peptides is their small size in comparison to proteins, which can increase tissue penetration depth, and their relatively easy and inexpensive synthesis. Folding is typically not required for their activity, which makes them attractive targeting ligands. However, a serious drawback of peptides is that they usually do not exhibit high affinities for their target, probably caused by their flexibility and small binding interface.
Phage display can be used to identify specific targeting peptides for a broad variety of substrates.[132] The phages displaying the peptides are screened for binding to a specific target
by multiple rounds of selection and amplification. Previously, two collagen binding cyclic hexapeptides (CVWLWENC en CVWLWEQC) were selected indirectly by phage display screening against a monoclonal antibody against von Willebrand factor.[133] Competition
experiments showed the ability of both cyclic peptides to inhibit binding of von Willebrand factor to rat tail collagen type I. IC50 values of 100 µM for CVWLWENC and 35 µM for
CVWLWEQC were reported.
Caravan and coworkers performed direct phage display screening on collagen type I, revealing a collagen type I specific peptide sequence, which was subsequently modified to improve collagen affinity.[134] A conjugate consisting of a peptide of 16 amino acids
10 amino acids that is formed through a disulfide bond. Biphenylalanine (Bip) was added at the C-terminus, since this was found to improve the collagen affinity. Moreover, three Gd-DTPA moieties were conjugated to obtain a collagen targeted MRI contrast agent. The collagen targeted contrast agent showed an affinity for collagen type I in the low µM range. Evaluation of the probe in a mouse model of aged myocardial infarction showed high contrast for fibrotic scar tissue consisting of collagen type I, type III and fibronectin versus viable myocardium.[134, 135]
Within our group we have recently selected peptides for binding to rat tail collagen type I and collagen type V using phage display.[136] For collagen type I, a 7-mer peptide
sequence (HVWMQAP) was found that showed an apparent Kd of approximately 60 µM. A
12-mer peptide (SPVSFTRYLSNW) was found to bind specifically to collagen type V, although with a low affinity.[137] The peptides derived from phage display usually show significantly
weaker binding than their respective high affinity phage. Therefore, multivalent display was used to mimic the multivalent architecture of the original phages. [136] The selected peptide
sequences were ligated five times to a pentavalent dendritic wedge to mimic the pentavalent phages (Figure 1.7). Multivalent display of the selected weakly binding peptides was shown to enhance binding, while keeping the size of the construct relatively small. Compared to their monovalent analogue, the pentavalent collagen type I and collagen type V peptides showed an increase in affinity of a 100-fold and 500-fold, respectively.
Figure 1.7 From phage display to dendrimer display: phage display to collagen type I revealed a collagen binding peptide of seven amino acids, which was translated into a high affinity, versatile synthetic collagen specific probe by mimicking the original pentavalent phage architecture on a dendritic wedge. Figure reproduced with permission from [136].
1.4
Molecular imaging of proteolytic activity
1.4.1 Optical probesIn many diseased tissues MMP levels are elevated. The high regional expression of these enzymes can be measured in immunohistological assays using labeled antibodies against the proteases. However, this method does not probe the activity of the enzymes, only the presence. Zymography is an extensively used method to test the activity of the proteases, but only useful for ex vivo tissue analysis.
Noninvasive in vivo detection of MMP activity could be an important tool for early disease diagnosis.[36] Common in the design of many protease sensitive probes is that they act
as a substrate for their target protease, in this way amplifying the signal. Most molecular imaging probes for MMPs and other proteases are based on optical techniques. Fluorescently labeled substrates have been developed to report protease activity in living biological systems. The principle mechanism of activation is based on either quenching due to the proximity of the fluorophores (self-quenching) or Förster Resonance Energy Transfer (FRET) to quench the fluorescent signal that is then restored upon proteolytic cleavage.[138-145]
The first demonstration of in vivo detection and imaging of protease activity was reported by Weissleder and coworkers.[146] Near-infrared (NIR) fluorophores were attached to a
linear poly-lysine-polyethyleneglycol copolymer, resulting in quenching of the fluorescent signal due to the location of the fluorophores on the polymer substrate. Upon proteolytic cleavage of the poly-lysine peptide linker, the fluorescent signal was enhanced producing an optically detectable near-infrared fluorescence (NIRF) signal associated with the tumor. Optimizing substrates specific for MMPs using the same concept has been done to develop self-quenched and FRET pair proteolytic beacons to specifically image tumor associated MMPs. The first probe for imaging MMP activity was developed by Bremer and colleagues.[147] A peptide
substrate (GPLGVRGK(FITC)C-NH2) for MMP-2 was labeled with Cy5 NIR fluorophores at the
amino terminus. Quenching of the closely positioned fluorophores in the native construct results in the ability to sense and image MMP activity in vivo.[147, 148] The MMP sensitive probe
was shown to be able to act as an activatable reporter probe to sense MMP activity in tumor bearing mice. A significant difference in fluorescent signal was observed between the mice injected with the specific MMP sensitive probe and mice injected with the control probe. Moreover, the effect of MMP inhibition by a potent MMP inhibitor could be directly imaged within hours after treatment using the near-infrared MMP sensitive probe. [147, 148] However, a
disadvantage of FRET-based probes is that they can only be used for in vivo imaging in small animals, since the penetration depth of the fluorescent signal is small.
A more advanced optical protease sensitive probe was developed by cyclization of the bioluminescent enzyme firefly luciferase.[149] This enzyme consists of an N-terminal and a
C-terminal domain that were connected to each other via a substrate sequence for the protease caspase-3 to force the protein into a conformation resulting in loss of bioluminescence. Activation of this cyclic probe was mediated by opening of the circular structure upon protease cleavage resulting in restoration of the enzymatic activity. In vivo real-time imaging of caspase-3 activity in living mice was shown. Stimulation of caspase-caspase-3 activity could be successfully detected in these mice. The main advantage of bioluminescence imaging lies in the fact that there is no need for excitation to obtain a measurable signal. In addition, bioluminescence imaging enables detection of very low levels of its target due to its virtually background-free light emission, it requires short acquisition times and exhibits ease of operation, which makes it ideal for high-throughput screening.
1.4.2 Protease sensitive targeting probes for MMP imaging
Molecular imaging of MMP activity has thus far been limited to optical imaging. However, optical protease sensitive probes are not ideal for in vivo imaging due to the limited imaging penetration depth. Ideally, one would like to develop a protease activatable probe that can be used in combination with other imaging modalities, such as magnetic resonance imaging (MRI) or single photon emission computed tomography (SPECT) that allow in vivo molecular imaging. In 2004 a new strategy for MMP sensitive selective delivery of contrast agents (or drugs) to tumor cells was reported, independent of the imaging technique (Figure 1.8).[150] Tsien and coworkers used polyarginine-based, cell-penetrating peptides (CPPs) that
were blocked by fusion to negatively charged inhibitory domain to prevent cellular association. Upon cleavage of the linker between the two domains by MMP-2 or MMP-9, the CPP labeled with the contrast agent can dissociate and thus bind cells. The activatable CPPs showed a 3-fold increase in cellular uptake for tumors secreting MMP-2 and MMP-9 relative to normal tissue.[150] However, once the CPPs are activated they will enter any cell, so the probes still lack
dual-specificity.
An alternative strategy for protease sensitive molecular imaging is the development of specific targeting probes that only become active after proteolytic cleavage. Recently, Suzuki and coworkers developed an MMP activatable targeting probe for which uptake into cells is synergistically activated by both scavenger receptors overexpressed in activated macrophages and active MMP-9 (Figure 1.9).[151]
Figure 1.8 Schematic diagram of activatable CPPs. Cellular uptake induced by a cationic peptide (blue) is blocked by a short stretch of acidic residues (red) attached by a cleavable linker (green). Once the linker is cleaved by an MMP, the acidic inhibitory domain drifts away, and the cationic CPP is free to carry its cargo (yellow) into cells. Figure reproduced with permission from [150].
Figure 1.9 Dual-specific MMP activatable targeting probe that consists of an inhibitory domain (magenta), a cleavable linker for MMP-9 (light blue), a receptor binding domain (ApoB547-735, yellow), and a reporter domain
(green). Once the targeting probe encounters activated macrophages that express both the scavenger receptors (SR) and MMP-9, the inhibitory domain of the probe dissociates upon cleavage by MMP-9 and then the ApoB547-735
domain is exposed to bind to SR. After binding the receptors, the probe is incorporated into the cells expressing both SR and MMP-9. Figure reproduced with permission from [151].
These overexpressed scavenger receptors are the marker receptors of atherosclerosis and active MMP-9 is the marker protease for atherosclerosis. The synergistic activation thus allows accurate targeting of disease cells. The targeting probe consists of a receptor binding domain (ApoB547-735) fused to both an inhibitory domain via a cleavage linker, and a reporter
domain. The intact probe will not associate with the cells. However, in presence of MMP-9 the inhibitory domain can dissociate resulting in exposure of ApoB. ApoB fused to the reporter domain can then bind the receptors displayed on activated macrophages for incorporation into the cells. Thus, the MMP-9 sensitive probe will bind selectively to disease cells by targeting of the overexpressed scavenger receptors after activation by MMP-9, which was successfully demonstrated in ex vivo cell experiments.
1.5 Aim and outline
Remodeling of the ECM is involved in many disease processes, including angiogenesis, myocardial infarction and atherosclerosis. The aim of the research described in this thesis is the development of specific imaging probes to visualize ECM remodeling during disease. Molecular imaging of structural changes can help to understand the complex dynamics of those diseases, which can lead to rational design of therapeutics. Moreover, visualization of the different components of the ECM can be applied for monitoring and characterizing the response of patients to therapy. Molecular targets for imaging agents are structural components, such as collagen and fibronectin, but also MMPs, which are crucial for remodeling of the highly organized ECM structure and composition, and whose levels are often upregulated during disease.
In Chapter 2, a new method for specific peptide-based targeting that was previously developed for collagen type I and V was explored to test whether it can be used as a generic approach to target other structural ECM proteins. Selection using phage display revealed peptide sequences specifically binding to fibronectin and fibrinogen. Multivalent display of these target binding peptides on a pentavalent dendritic wedge resulted in a significant increase in affinity compared to the monovalent weakly binding peptide in surface plasmon resonance (SPR) binding studies. Ex vivo molecular imaging using the pentavalent fibronectin binding probes was explored using sections of human kidney tissue. Instead of specific fibronectin structures, background binding to epithelial cells was observed, emphasizing the need to test the performance of every multivalent dendritic structure in native tissue for its specificity and affinity for the target protein.
Chapter 3 describes the preparation of well-defined protein dendrimers as multivalent collagen targeting ligands by coupling of CNA35 to divalent and tetravalent dendritic wedges.[152] The binding of these multivalent protein constructs was studied on
collagen-immobilized chip surfaces as well as to native collagen in rat intestinal tissues. To understand the importance of target density we also created collagen-mimicking surfaces by immobilizing synthetic collagen triple helical peptides at various densities on a chip surface. Multivalent display of a weakly binding variant of CNA35 resulted in an impressive increase in collagen affinity and the dissociation of the multivalent CNA35 dendrimers was found to be strongly attenuated.
In Chapter 4, the development of a dual-specific MMP sensitive collagen probe is reported.[153] This probe is based on the concept of cyclization of CNA35 via a protease
recognition site. In line with the proposed mechanism for collagen binding, binding to mature collagen should be topologically inhibited for circular CNA35. Upon MMP cleavage the cyclic structure is opened restoring its ability to bind collagen. Circular CNA35 was synthesized and binding experiments using SPR as well as ex vivo tissue experiments showed that the collagen binding affinity was significantly hampered by cyclization of the protein. Addition of MMP-1 resulted in restoration of the collagen binding capacity. This is the first report showing that protein cyclization is a new strategy to integrate two different markers for molecular targeting.
The strategy outlined in chapter 4 depends on the specific binding mechanism of CNA35. In this strategy the binding to collagen was not blocked, but topologically inhibited. A more generally applicable principle for generation of dual-specific enzyme sensitive imaging probes is based on intramolecular blocking, which is outlined in chapter 5. CNA35 was ligated to a synthetic collagen mimic via a cleavable linker to block the collagen binding site. Successful synthesis of the completely blocked CNA35 construct was performed allowing binding studies to assess the applicability of the sensor. SPR experiments showed a 5-fold increase in collagen binding affinity upon 1 digestion. Activation of the sensor by MMP-1 could also be visualized in ex vivo experiments.
1.6 References
[1] B. Alberts, A. Johnson, J. Lewis, M. Raff, K. Robers, P. Walter, Molecular Biology of the Cell, 4th ed., London: Garland, 2002.
[2] A. Aouacheria, C. Geourjon, N. Aghajari, V. Navratil, G. Deléage, C. Lethias, J. Y. Exposito, Molecular Biology and Evolution 2006, 23, 2288.
[3] R. O. Hynes, Science 2009, 326, 1216.
[4] E. Hohenester, J. Engel, Matrix Biology 2002, 21, 115. [5] P. Bruckner, Cell and Tissue Research 2009, 1.
[6] L. Bruckner-Tuderman, K. Von Der Mark, T. Pihlajaniemi, K. Unsicker, Cell and Tissue Research 2010, 339, 1.
[7] K. Y. Tsang, M. C. H. Cheung, D. Chan, K. S. E. Cheah, Cell and Tissue Research 2010, 339, 93. [8] T. L. Adair-Kirk, R. M. Senior, International Journal of Biochemistry and Cell Biology 2008, 40, 1101. [9] I. Vakonakis, I. D. Campbell, Current Opinion in Cell Biology 2007, 19, 578.
[10] S. L. K. Bowers, I. Banerjee, T. A. Baudino, Journal of Molecular and Cellular Cardiology 2010, 48, 474.
[11] A. M. Briones, S. M. Arribas, M. Salaices, Current Opinion in Nephrology and Hypertension 2010, 19, 187.
[12] E. Adiguzel, P. J. Ahmad, C. Franco, M. P. Bendeck, Vascular Medicine 2009, 14, 73. [13] J. Myllyharju, K. I. Kivirikko, Annals of Medicine 2001, 33, 7.
[14] J. Pellman, R. C. Lyon, F. Sheikh, Journal of Molecular and Cellular Cardiology 2010, 48, 461. [15] R. Zamilpa, M. L. Lindsey, Journal of Molecular and Cellular Cardiology 2010, 48, 558. [16] G. Murphy, V. Knäuper, Matrix Biology 1997, 15, 511.
[17] H. Nagase, R. Visse, G. Murphy, Cardiovascular Research 2006, 69, 562.
[18] J. L. Lauer-Fields, D. Juska, G. B. Fields, Biopolymers - Peptide Science Section 2002, 66, 19. [19] R. Visse, H. Nagase, Circulation Research 2003, 92, 827.
[20] J. A. Allan, A. J. P. Docherty, P. J. Barker, N. S. Huskisson, J. J. Reynolds, G. Murphy, Biochemical Journal 1995, 309, 299.
[21] R. T. Aimes, J. P. Quigley, Journal of Biological Chemistry 1995, 270, 5872.
[22] E. H. M. Kerkvliet, A. J. P. Docherty, W. Beertsen, V. Everts, Matrix Biology 1999, 18, 373. [23] M. L. Patterson, S. J. Atkinson, V. Knäuper, G. Murphy, FEBS Letters 2001, 503, 158. [24] H. Nagase, G. B. Fields, Biopolymers 1996, 40, 399.
[25] W. Bode, C. Fernandez-Catalan, H. Tschesche, F. Grams, H. Nagase, K. Maskos, Cellular and Molecular Life Sciences 1999, 55, 639.
[26] G. B. Fields, Journal of Theoretical Biology 1991, 153, 585.
[27] Y. Y. Li, C. F. McTiernan, A. M. Feldman, Cardiovascular Research 2000, 46, 214.
[28] M. D. Sternlicht, Z. Werb, in Annual Review of Cell and Developmental Biology, Vol. 17, 2001, pp. 463.
[29] H. Nagase, Biological Chemistry 1997, 378, 151.
[30] H. Nagase, J. J. Enghild, K. Suzuki, G. Salvesen, Biochemistry 1990, 29, 5783.
[31] K. Suzuki, J. J. Enghild, T. Morodomi, G. Salvesen, H. Nagase, Biochemistry 1990, 29, 10261. [32] P. T. G. Elkington, C. M. O'Kane, J. S. Friedland, Clinical and Experimental Immunology 2005, 142,
12.
[33] J. B. Catterall, T. E. Cawston, Methods in molecular biology (Clifton, N.J.) 2003, 225, 353. [34] P. A. M. Snoek-van Beurden, J. W. Von Den Hoff, BioTechniques 2005, 38, 73.
[35] G. S. Schultz, G. Ladwig, A. Wysocki, World Wide Wounds 2005, 2005.
[36] R. L. Scherer, J. O. McIntyre, L. M. Matrisian, Cancer and Metastasis Reviews 2008, 27, 679. [37] C. E. Brinckerhoff, L. M. Matrisian, Nature Reviews Molecular Cell Biology 2002, 3, 207. [38] L. H. Arroyo, R. T. Lee, Cardiovascular Research 1999, 41, 369.
[39] S. Katsuda, T. Kaji, Journal of Atherosclerosis and Thrombosis 2003, 10, 267. [40] C. M. Overall, O. Kleifeld, Nature Reviews Cancer 2006, 6, 227.
[41] C. J. Morrison, G. S. Butler, D. Rodríguez, C. M. Overall, Current Opinion in Cell Biology 2009, 21, 645.
[42] S. K. Choi, Synthetic Multivalent Molecules, John Wiley & Sons Inc., 2004.
[43] M. Mammen, S. K. Choi, G. M. Whitesides, Angewandte Chemie - International Edition 1998, 37, 2755.
[44] L. L. Kiessling, J. E. Gestwicki, L. E. Strong, Angewandte Chemie - International Edition 2006, 45, 2348.
[46] V. M. Krishnamurthy, V. Semetey, P. J. Bracher, N. Shen, G. M. Whitesides, Journal of the American Chemical Society 2007, 129, 1312.
[47] E. Fan, Z. Zhang, W. E. Minke, Z. Hou, C. L. M. J. Verlinde, W. G. J. Hol, Journal of the American Chemical Society 2000, 122, 2663.
[48] Y. Lee, N. S. Sampson, Current Opinion in Structural Biology 2006, 16, 544. [49] M. Okada, Progress in Polymer Science (Oxford) 2001, 26, 67.
[50] A. W. Bosman, H. M. Janssen, E. W. Meijer, Chemical Reviews 1999, 99, 1665. [51] E. A. Merritt, W. G. J. Hol, Current Opinion in Structural Biology 1995, 5, 165.
[52] E. Fan, E. A. Merritt, C. L. M. J. Verlinde, W. G. J. Hol, Current Opinion in Structural Biology 2000, 10, 680.
[53] P. I. Kitov, J. M. Sadowska, G. Mulvey, G. D. Armstrong, H. Ling, N. S. Pannu, R. J. Read, D. R. Bundle, Nature 2000, 403, 669.
[54] E. A. Merritt, Z. Zhang, J. C. Pickens, M. Ahn, W. G. J. Hol, E. Fan, Journal of the American Chemical Society 2002, 124, 8818.
[55] Z. Zhang, E. A. Merritt, M. Ahn, C. Roach, Z. Hou, C. L. M. J. Verlinde, W. G. J. Hol, E. Fan, Journal of the American Chemical Society 2002, 124, 12991.
[56] G. L. Mulvey, P. Marcato, P. I. Kitov, J. Sadowska, D. R. Bundle, G. D. Armstrong, Journal of Infectious Diseases 2003, 187, 640.
[57] P. I. Kitov, T. Lipinski, E. Paszkiewicz, D. Solomon, J. M. Sadowska, G. A. Grant, G. L. Mulvey, E. N. Kitova, J. S. Klassen, G. D. Armstrong, D. R. Bundle, Angewandte Chemie - International Edition 2008, 47, 672.
[58] C. B. Calson, P. Mowery, R. M. Owen, E. C. Dykhuizen, L. L. Kiessling, ACS Chemical Biology 2007, 2, 119.
[59] R. M. Owen, C. B. Carlson, J. Xu, P. Mowery, E. Fasella, L. L. Kiessling, ChemBioChem 2007, 8, 68. [60] H. M. Branderhorst, R. M. J. Liskamp, G. M. Visser, R. J. Pieters, Chemical Communications 2007,
5043.
[61] R. J. Pieters, Drug Discovery Today: Technologies 2009, 6. [62] R. J. Pieters, Organic and Biomolecular Chemistry 2009, 7, 2013.
[63] J. P. Tam, Proceedings of the National Academy of Sciences of the United States of America 1988, 85, 5409.
[64] A. Dirksen, E. W. Meijer, W. Adriaens, T. M. Hackeng, Chemical Communications 2006, 1667. [65] I. Dijkgraaf, A. Y. Rijnders, A. Soede, A. C. Dechesne, G. W. Van Esse, A. J. Brouwer, F. H. M.
Corstens, O. C. Boerman, D. T. S. Rijkers, R. M. J. Liskamp, Organic and Biomolecular Chemistry 2007, 5, 935.
[66] I. Dijkgraaf, J. A. W. Kruijtzer, S. Liu, A. C. Soede, W. J. G. Oyen, F. H. M. Corstens, R. M. J. Liskamp, O. C. Boerman, European Journal of Nuclear Medicine and Molecular Imaging 2007, 34, 267. [67] I. Dijkgraaf, A. J. Beer, H. J. Wester, Frontiers in Bioscience : a Journal and Virtual library 2009, 14,
887.
[68] P. P. Karmali, V. R. Kotamraju, M. Kastantin, M. Black, D. Missirlis, M. Tirrell, E. Ruoslahti, Nanomedicine: Nanotechnology, Biology, and Medicine 2009, 5, 73.
[69] W. J. M. Mulder, G. J. Strijkers, K. C. Briley-Saboe, J. C. Frias, J. G. S. Aguinaldo, E. Vucic, V. Amirbekian, C. Tang, P. T. K. Chin, K. Nicolay, Z. A. Fayad, Magnetic Resonance in Medicine 2007, 58, 1164.
[70] D. Peters, M. Kastantin, V. R. Kotamraju, P. P. Karmali, K. Gujraty, M. Tirrell, E. Ruoslahti, Proceedings of the National Academy of Sciences of the United States of America 2009, 106, 9815.
[71] D. Sutton, N. Nasongkla, E. Blanco, J. Gao, Pharmaceutical Research 2007, 24, 1029. [72] V. P. Torchilin, Pharmaceutical Research 2007, 24, 1.
[73] V. P. Torchilin, A. N. Lukyanov, Z. Gao, B. Papahadjopoulos-Sternberg, Proceedings of the National Academy of Sciences of the United States of America 2003, 100, 6039.
