Integrins are cell-adhesion receptors that mediate cell–extracellular-matrix (ECM) and cell–cell interactions by recognizing specific ligands. Recent studies have shown that the formation of isoaspartyl residues (isoAsp) in integrin ligands by asparagine deamidation or aspartate isomerization could represent a mechanism for the regulation of integrin–ligand recognition. This spontaneous post-translational modification, which might occur in aged proteins of the ECM, changes the length of the peptide bond and, in the case of asparagine, also of the charge. Although these changes typically have negative effects on protein function, recent studies suggested that isoAsp formation at certain Asn-Gly-Arg (NGR) sites in ECM proteins have a gain-of-function effect, because the resulting isoAsp-Gly-Arg (isoDGR) sequence can mimic Arg-Gly-Asp (RGD), a well-known integrin-binding motif. Substantial experimental evidence suggests that the NGR-to-isoDGR transition can occur in vitro in natural proteins and in drugs containing this motif, thereby promoting integrin recognition and cell adhesion. In this Commentary, we review these studies and discuss the potential effects that isoAsp formation at NGR, DGR and RGD sites might have in the recognition of integrins by natural ligands and by drugs that contain these motifs, as well as their potential biological and pharmacological implications.
Integrins are αβ heterodimeric transmembrane receptors involved in the regulation of many aspects of cell behavior, including adhesion, growth, survival, proliferation, migration and invasion (Barczyk et al., 2010). The integrin family comprises 24 members made up of different combinations of α and β subunits. The specific combination of subunits in each heterodimer determines the function of the integrin and its ligand specificity. Individual integrins may bind more than one ligand and individual ligands can bind to more than one integrin (Humphries et al., 2006; Plow et al., 2000). For example, the αvβ3 heterodimer can recognize the sequence Arg-Gly-Asp (RGD) that is present in fibronectin, vitronectin, fibrinogen, osteopontin, vonWillebrand factor, tenascin and thrombospondin, whereas a subset of integrins, such as α5β1-, αvβ1-, αvβ3-, αvβ5-, αvβ6-, α8β1- and αIIbβ3-integrins, can recognize the sequence RGD in fibronectin (Pankov and Yamada, 2002). Other integrins recognize different sequences in the extracellular matrix (ECM) or other proteins. For example, α4β1-, α4β7- and α9β1-integrins, can recognize the sequence LDV in fibronectin and SVVYGLR in osteopontin, whereas α1β1-, α2β1-, α10β1- and α11β1-integrins recognize the GFOGER sequence in collagen (Humphries et al., 2006). Besides these well-characterized binding motifs, it has been proposed that the NGR and DGR sequences also have a role in integrin recognition, although controversial data have been reported with regard to the integrin-binding properties of these motifs (Corti et al., 2008; Hautanen et al., 1989; Jullienne et al., 2009; Koivunen et al., 1999; Liu et al., 1997; Rusnati et al., 1997; Soncin, 1992; Yamada and Kennedy, 1987). Interestingly, these motifs and many other integrin recognition sequences (reviewed by Ruoslahti, 1996) contain aspartate (Asp) or asparagine (Asn), i.e. two amino acid residues that spontaneously can undergo post-translational modifications to form isoaspartyl residues (isoAsp or isoD). There is substantial evidence that formation of this non-standard β-amino acid can occur during tissue aging in the ECM and in other secreted proteins with a slow turnover, e.g. in collagen, fibronectin and α-crystallin (Lanthier and Desrosiers, 2004; Lindner and Helliger, 2001; Weber and McFadden, 1997b). Although this protein modification typically causes a loss-of-function, recent data suggest that isoAsp formation at NGR or DGR sites, by Asn deamidation or Asp isomerization, might have a gain-of-function effect, because the resulting isoDGR motif can mimic RGD and recognize the RGD-binding site of integrins (Curnis et al., 2006).
In this Commentary, we discuss the potential role that isoAsp formation in proteins of the ECM have as a molecular switch for negative or positive regulation of integrin recognition by ligands, particularly those containing RGD, NGR and DGR sequences. Furthermore, we discuss the structural basis of the molecular mimicry of RGD by isoDGR in integrin recognition, the regulatory mechanisms, and the potential functional implications of isoAsp formation in natural ligands and in synthetic drugs that contain these motifs.
Formation of isoAsp in ECM proteins and its functional effects
Formation of isoAsp and ECM protein damage
The formation of isoAsp residues at Asn or Asp residues in proteins is generally considered a degradation reaction as a consequence of protein aging. IsoAsp formation has been shown to occur in ECM proteins, including collagen types I and III, in fibronectin, in α-crystallin and in the ECM of the mammalian brain (Lanthier and Desrosiers, 2004; Lindner and Helliger, 2001; Reissner and Aswad, 2003; Weber and McFadden, 1997a). IsoAsp formation can occur not only in vitro, e.g. during protein storage, but also in vivo, as has been documented for the extracellular matrix of mammalian brain or the vascular wall (Reissner and Aswad, 2003; Weber and McFadden, 1997b). This post-translational modification might cause changes in the charge or the length of the peptide bond (see Fig. 1A), typically causing loss of function. For example, in collagen-I or in peptides containing RGDSR and KDGEA (a recognition sequence for α2β1-integrin), the formation of isoAsp reduces their capability to promote cell motility (Lanthier and Desrosiers, 2004). Treatment of aged collagen I with protein-L-isoAsp-O-methyltransferase (PIMT), an enzyme that can convert isoAsp into Asp, partially recovers cell migration. In addition, fibronectin, an adhesive protein that has important roles in haemostasis, thrombosis, inflammation, wound repair, angiogenesis and embryogenesis (Humphries et al., 1989; Pankov and Yamada, 2002), also accumulates isoAsp during in vitro aging (Lanthier and Desrosiers, 2004). Similarly, treatment of aged fibronectin or retronectin – a fibronectin fragment containing RGD – with PIMT increases its pro-adhesive activity on endothelial cells (Curnis et al., 2006) (A.C. and F.C., unpublished data). Taken together, it therefore appears that the biological properties of collagen, fibronectin and their fragments are impaired by Asp-to-isoAsp transitions that occur at certain sites in these proteins and are subsequently restored by transforming isoAsp back into Asp.
Formation of isoAsp at NGR sites with gain-of-function
Although isoAsp formation in aged proteins is generally viewed as a deleterious event with regard to their function, a growing body of evidence suggests that this reaction, in some instances, constitutes a mechanism for the intentional modification of a protein structure with a useful function. For example, it has been proposed that isoAsp formation, which might occur faster in proteins that have lost their structural integrity, can provide a type of molecular indicator of protein damage, which is capable to activate selective degradation mechanisms (Lanthier and Desrosiers, 2004; Pepperkok et al., 2000; Reissner and Aswad, 2003; Robinson and Robinson, 2001; Weintraub and Deverman, 2007). Another striking example of this concept is the formation of isoAsp in fibronectin fragments. Fibronectin (FN) contains three types of repeating homologous modules, in addition to alternatively spliced modules (Mohri, 1997; Pankov and Yamada, 2002). Notably, four FN modules, i.e. the fifth type I (FN-I5), the first type II (FN-II1), the seventh type I (FN-I7) and the ninth type III (FN-III9) repeats, contain NGR sequences. It has been suggested that isoAsp formation at the NGR site of FN-I5 activates a latent αvβ3-integrin-binding site (Curnis et al., 2006; Takahashi et al., 2007). Here, deamidation of Asn263 leads to a transformation of NGR into isoDGR and DGR. Structure–activity relationships studies have shown that isoDGR and not DGR is the integrin-binding motif.
Mechanisms of isoAsp formation
The formation of isoAsp in proteins can occur by non-enzymatic Asn deamidation or Asp isomerization reactions (Geiger and Clarke, 1987; Weintraub and Deverman, 2007). Asn deamidation occurs by a nucleophilic attack of the backbone NH center at the carbonyl group of Asn side chain, which leads to formation of a succinimide ring (Fig. 1B). The formation of a succinimide ring is also the first step for Asp isomerization (Fig. 1B). Hydrolysis can then occur at both carbonyl groups of the cyclic imide, leading to the formation of a mixture of Asp and isoAsp residues, typically with a ratio of approximately 1:3. Racemization and hydrolysis of the succinimide intermediate can also lead to the formation of Asp and isoAsp in D-configuration, but this typically occurs with much lower efficiency (Geiger and Clarke, 1987; Stephenson and Clarke, 1989; Tyler-Cross and Schirch, 1991). Thus, the resulting Asp and isoAsp in L-configuration are more relevant. The above deamidation and isomerization reactions can take hours, days or even years – depending on several factors. For example, secondary, tertiary and quaternary structural elements in a protein can place Asn or Asp residues in proximity to functional groups that either inhibit or accelerate these reactions (Robinson, 2002; Robinson and Robinson, 2001; Robinson et al., 2004). In particular, the presence of a Gly residue following Asn or Asp generally accelerates these reactions owing to its small size and flexibility (Robinson and Robinson, 2001). Other crucial factors are the pH, the ionic strength and the temperature (Robinson and Robinson, 2001; Tyler-Cross and Schirch, 1991).
Kinetics of the NGR-to-isoDGR transition
Approximately 17% of proteins that are classified by using the keyword ‘adhesion’ in vertebrate protein databanks contain NGR sites (Corti et al., 2008). Considering the known effects of 3D-structural protein elements on the deamidation rate of Asn residues discussed above, only certain NGR sites with suitable accessibility, flexibility and side chain orientation are likely to undergo deamidation under physiological conditions and, subsequently, bind integrins. Remarkably, NGR deamidation in the FN-I5 fragment is very rapid, with a half-life of the NGR site in cell culture medium of only ~4 hours (Curnis et al., 2006). The surprising velocity of this reaction is probably owing to the fact that here Asn is followed by a Gly residue. Furthermore, the Asn and Gly residues are part of the GNGRG loop that is well exposed on the surface of this fibronectin fragment, thereby possibly favoring the correct orientation of the Asn side chain and peptide backbone for a nucleophilic attack (Di Matteo et al., 2006). Similarly, fast deamidation rates have been observed in peptides that contain the disulfide-bonded CNGRC loop (Curnis et al., 2010; Curnis et al., 2006).
By contrast, full-length plasma fibronectin was shown to be considerably more resistant to Asn deamidation than short FN fragments or peptides (Lanthier and Desrosiers, 2004) and, accordingly, freshly isolated plasma fibronectin contains only 0.03–0.05 pmol of isoAsp per pmol of protein (Curnis et al., 2006). Interestingly, cystine reduction and alkylation enhanced isoDGR formation in a 70 kDa N-terminal fragment of fibronectin (Xu et al., 2010). It therefore appears that rapid deamidation of NGR in fibronectin requires proteolytic processing and/or conformational changes. Nevertheless, the in vivo formation of isoDGR sites in fibronectin, e.g. after deposition in tissues, still remains to be demonstrated.
Rapid NGR-to-isoDGR transformation also occurs in recombinant proteins that contain CNGRC fused to the N-terminus of tumor necrosis factor α (TNF-α) (Curnis et al., 2006; Curnis et al., 2008) or to the C-terminus of interferon γ (IFNγ) (Curnis et al., 2005), suggesting that this phenomenon is not specific for the GNGRG site in FN-I5. Thus, it is possible that other proteins also use this mechanism to activate latent integrin recognition sites, and this should be taken into account when analyzing potential integrin-binding sites of proteins that contain the NGR sequence.
Recognition of RGD-dependent integrins by isoDGR and its functional role
Peptides that contain the isoDGR motif can recognize members of the RGD-dependent integrin family, such as αvβ3-, αvβ5-, αvβ6-, αvβ8- and α5β1-integrins, but not other, such as α1β1-, α3β1-, α4β7-, α6β4- or α9β1-integrins (Curnis et al., 2010). The affinity and specificity of the interaction between isoDGR and integrin depend on the molecular scaffold of isoDGR (Curnis et al., 2010). For instance, cyclic peptides that contain isoDGR flanked by Cys residues can bind αvβ3-integrin with an affinity that is 10–100-fold higher than that for other members of the RGD-dependent integrin family (for example, Kd is ~9 nM for αvβ3-integrin). Replacement of the Cys by two Gly residues leads to a marked loss of affinity for all integrins and a change of specificity (Curnis et al., 2010). It appears, therefore, that isoDGR, like RGD, can differentially recognize integrins depending on its molecular scaffold.
These findings and the results of a series of in vitro experiments performed with fibronectin and fibronectin fragments suggest that NGR deamidation has a functional role in the regulation of cell adhesion. For example, a deamidated FN-I4-5 fragment enhanced the adhesion of endothelial cells to microtiter plates, but not when its NGR sequence was replaced with SGS (Curnis et al., 2006). Similarly, replacement of NGR in the FN-I5 and FN-I7 modules of a large N-terminal fragment of fibronectin (FN-70 kDa) by QGR, a sequence that can deamidate but with a much slower rate (Robinson and Robinson, 2001), reduced its cell adhesion activity, further supporting a functional role for the NGR sequence in fibronectin modules (Xu et al., 2010).
Although isoDGR-containing compounds can promote cell adhesion when they are adsorbed onto microtiterplates, they can exert inhibitory effects in other assays. For example, deamidated FN-I5 and the peptide CisoDGRC inhibited endothelial cell proliferation and adhesion to vitronectin when these peptides were added as soluble ligands in the cell culture medium (Curnis et al., 2006). These peptides also recognized αvβ3-integrin-positive endothelial cells in tumor vessels, and inhibited tumor growth when they were systemically administered to tumor-bearing mice (Curnis et al., 2010; Curnis et al., 2006). Given the growing body of evidence that implicates αvβ3-integrin in angiogenesis (Brooks et al., 1995; Eliceiri and Cheresh, 2000), it is possible that the interaction of the isoDGR site in fibronectin or fibronectin fragments with αvβ3-integrin has an important role in cancer and in other diseases that involve angiogenesis.
Tissue fibronectin is organized in fibrils, which mediate a variety of cellular interactions with the extracellular matrix (Humphries et al., 1989; Pankov and Yamada, 2002). The formation of fibronectin fibrils in the ECM is a cell-mediated process that involves the rearrangement of the cytoskeleton and integrins in order to transmit the forces necessary to unfold fibronectin and assemble it into fibrils (Geiger, 2001; Ohashi et al., 2002; Pankov et al., 2000; Zamir et al., 2000). Mice that were genetically engineered to express a fibronectin variant with a non-functional RGE in place of RGD and that are, thus, unable to bind integrins through this site, can still assemble fibrils in vivo (Takahashi et al., 2007). The same authors also showed that a cyclic peptide containing isoDGR inhibits fibril formation of a RGE-fibronectin mutant in vitro and they proposed that the GNGRG of FN-I5 can function in fibril assembly. However, other investigators showed that fibronectin mutants, in which NGR is replaced with QGR in the FN-I5 and FN-I7 modules, can still form fibrils, thus arguing against a crucial role of these sites in fibril formation (Xu et al., 2010). As the role in fibril formation of those NGR sites located in FN-II2 and FN-III9 has not yet been investigated, further studies are necessary to clarify the function of different NGR sites in the formation of fibronectin fibrils.
The DGR-to-isoDGR transition and integrin recognition
The frequency of the DGR sequence element in proteins classified as having a role in cell adhesion is higher than that of other xGR sequences (where x represents any amino acid residue other than Asn or Asp) and is similar to that of NGR and RGD (Corti et al., 2008). Interestingly, it has been shown that DGR has also a role in integrin recognition and cell adhesion (Gao and Brigstock, 2006; Rusnati et al., 1997; Yamada and Kennedy, 1987). For example, SDGR-containing peptides were found to inhibit the attachment of fibroblasts to laminin and collagen type I when added to the cell culture medium (Yamada and Kennedy, 1987). Furthermore, fibroblast growth factor 2 (FGF2), a pro-angiogenic heparin-binding cytokine that contains two DGR sites, can bind to αvβ3-integrin and promote endothelial cell adhesion (Rusnati et al., 1997). Mapping of the cell-adhesive domains has shown that the regions from amino acids 38 to 61 and 82 to 101 in FGF2, which contain the DGR sequences, are responsible for this activity (Rusnati et al., 1997). In addition, it was also shown that the GVCTDGR sequence of connective tissue growth factor can bind α5β1-integrin (Gao and Brigstock, 2006). However, other studies found SDGR and other DGR peptides to be inactive in cell attachment assays, which argues against a functional role of this motif in integrin recognition (Hautanen et al., 1989; Liu et al., 1997; Soncin, 1992).
Given that DGR can be converted into isoDGR by the isomerization of Asp, one explanation for these apparently controversial data may be that DGR, like NGR, can function as a latent integrin-binding site that is differentially activated dependent on its molecular context and the conditions. In agreement with this hypothesis, we have observed that the interaction of FGF2 with αvβ3-integrin is inhibited by treatment with PIMT (A.C. and F.C., unpublished data), supporting the idea that isoAsp residues in FGF2 are crucial for this interaction.
Considering that a number of proteins contain DGR it is possible that this mechanisms is more general. However, as discussed above for NGR, it should be born in mind that the residues flanking DGR and its microenvironment can crucially affect the kinetics of Asp isomerization and its transformation into isoDGR. Furthermore, the rate of Asp isomerization is generally slower than that of Asn deamidation in similar peptides (Geiger and Clarke, 1987). This reaction is, therefore, likely to have a significant role in only a small subset of proteins.
Structural basis of the molecular mimicry of RGD by isoDGR in integrin recognition
The structural basis of ligand recognition by RGD-dependent integrins has been studied intensively (Luo et al., 2007; Takagi, 2004). Biochemical studies have shown that both α- and β-integrin subunits span the plasma membrane and have short cytoplasmic domains. Outside the plasma membrane, the α- and β-subunits lie close together and form the RGD-binding pocket (Xiong et al., 2001). Substantial experimental evidence suggests that isoDGR binds to the RGD-binding site of integrins. First, the cyclic CisoDGRCGVRY peptide (called isoDGR-2C) might inhibit the binding of a cyclic RGD peptide to αvβ3-integrin in a competitive manner (Curnis et al., 2006). Notably, the affinities of cyclic isoDGR-2C and RGD-2C peptides for αvβ3-integrin were similar (Curnis et al., 2006). Replacement of isoAsp with Asp (as in DGR-2C) or with its enantiomer in D-configuration (D-isoAsp) causes a marked loss of binding affinity, pointing to highly selective and stereospecific interactions (Curnis et al., 2010; Curnis et al., 2006). Second, NMR structure analysis of cyclic isoDGR-2C, RGD-2C, DGR-2C and NGR-2C peptides, and αvβ3-integrin docking experiments showed that isoDGR, but not DGR and NGR, fit into the RGD-binding pocket and that it favorably interacts with this integrin (Curnis et al., 2006; Spitaleri et al., 2008) (Fig. 2). Notably, isoDGR docks onto αvβ3-integrin in an inverted orientation compared with RGD docking to αvβ3-integrin. This orientation allows isoDGR to bind the integrin α- and β-subunits through the isoAsp and Arg residues. The acidic and basic side chains of these residues are at the correct distance and orientation to engage stabilizing interactions with the polar regions of the integrin, thus reproducing the canonical interactions between RGD and αvβ3-integrin (Spitaleri et al., 2008). These data are in line with the results of binding studies showing that isoDGR, but not DGR and NGR, can efficiently interact with αvβ3-integrin. Therefore, isoDGR, unlike DGR and NGR, is a natural fit for the RGD-binding pocket of αvβ3-integrin and supports the hypothesis that transforming DGR or NGR into isoDGR can result in a molecular switch able to activate integrin recognition.
Such a model might also explain the controversial data reported in the literature for the interaction of integrin with compounds that contain DGR – as discussed above – or NGR. The suggestion that NGR and integrin interact originated from several observations. For example, in vitro panning of phage-display libraries against αvβ3- and α5β1-integrins selected peptides that contain NGR (Healy et al., 1995; Koivunen et al., 1993; Koivunen et al., 1994). Furthermore, adeno-associated virus type 2 contains an NGR site that is crucial for α5β1-integrin binding and viral cell entry (Asokan et al., 2006), and adenoviruses that have been genetically engineered to express the linear NGR sequence on their surface bind to αvβ3-integrin-expressing cells (Majhen et al., 2006). However, in vitro binding studies with purified αvβ3-integrin and peptides that contain the CNGRC or GNGRG sequences show only a modest interaction, if any at all (Curnis et al., 2010). Although we cannot totally exclude that NGR sites that are embedded in viral particles behave in a different manner, we hypothesize that integrin binding observed with certain NGR-containing ligands can be explained by binding of isoDGR generated from NGR. For example, phages that have been selected as carrying the NGR sequence might instead have displayed isoDGR on their surface owing to deamidation reaction during library preparation and in vitro selection. Thus, we think it is important that any isoAsp formation must be ruled out before integrin recognition of NGR-containing compounds is proposed.
Although isoDGR- and RGD-containing ligands can share the same binding site on integrins, their effects on integrin function are not necessarily identical for the following reasons. It is well established that the interaction between integrins and ligands is regulated by conformational changes (Arnaout et al., 2005; Luo et al., 2007). Conformational changes and integrin activation can be transmitted bidirectionally across the membrane through interactions with intracellular molecules (inside-out signaling) and with extracellular ligands (outside-in signaling) (reviewed in Askari et al., 2009). It has been suggested that different RGD-containing ligands induce different signaling pathways (Askari et al., 2009). Thus, it is possible that RGD and isoDGR ligands induce differential conformational changes and signaling in vivo. Furthermore, although peptides containing CRGDC or CisoDGRC sequences recognize purified αvβ3-integrin with similar affinities in vitro (Curnis et al., 2006), their kon and koff, which have not yet been measured, might not necessarily be similar. Considering the importance of these parameters for their biological properties, kinetic analysis of binding data and characterization of the signaling mechanisms induced by isoDGR-containing ligands are needed to define the extent to which isoDGR mimics – or differs – from RGD.
Regulation of isoAsp formation in tissues
As discussed above, Asp isomerization and Asn deamidation are thermodynamically spontaneous reactions that do not require enzymatic catalysis. It is currently unknown whether specific Asn deamidases or Asp isomerases exist that could further accelerate this process. However, the observation that NGR deamidates faster in short fibronectin fragments compared with soluble plasma fibronectin (discussed above) and the finding that isoDGR can be converted into DGR by PIMT (Curnis et al., 2006) point to possible enzymatic regulatory mechanisms.
In humans, PIMT is encoded by PCMT1, which is localized on chromosome 6 (6q24–q25). This enzyme (EC 184.108.40.206) catalyzes the transfer of a methyl group from S-adenosyl-L-methionine to the free carboxyl groups of D-Asp and L-isoAsp residues (Reissner and Aswad, 2003). The resulting methyl ester can then form a cyclic succinimide that, after hydrolysis, can generate Asp and isoAsp (Fig. 1B). In peptide substrates, there appears to be a preference for isoAsp residues that are flanked by bulky hydrophobic group at the position preceeding isoAsp and by neutral or positive groups at the positions that follow isoAsp (Lowenson and Clarke, 1991). PIMT is present in all vertebrate tissues and represents the major protein methyltransferase in the brain, testis, erythrocytes and eye (Yamamoto et al., 1998). PIMT is normally an intracellular enzyme that exists in two isoforms due to differential mRNA splicing (Takeda et al., 1995). The C-terminus of one isoenzyme ends with the sequence RWK, whereas the other isoform ends with RDEL, a known endoplasmic reticulum retention sequence (MacLaren et al., 1992). It has been shown that calmodulin, tubulin, histone H2B and synapsin I are natural substrates of PIMT action in vivo (Furuchi et al., 2010). However, experimental evidence suggests that this enzyme also targets extracellular proteins, such as collagen type I and type II (Weber and McFadden, 1997b). Interestingly, it has been demonstrated that PIMT is released into the extracellular environment by damaged vessels and by injured tissues, and that it becomes entrapped in the extracellular space (Weber and McFadden, 1997a; Weber and McFadden, 1997b). In uninjured blood vessels, up to 90% of total isoAsp residues are inaccessible to endogenous intracellular PIMT (Weber and McFadden, 1997b). However, after vessel injury PIMT is released into the extracellular environment and methylation of extracellular proteins, such as aged collagen type I and type II, can then ensue (Weber and McFadden, 1997b). The biological importance of PIMT is also supported by the observation that PIMT-deficient mice have a smaller body and spleen, a larger brain, an abnormal neuronal organization and atypical patterns of motor activity, and that they die from progressive epilepsy at 4–10 weeks after birth (Kim et al., 1997; Yamamoto et al., 1998).
An important point to keep in mind is that PIMT can restore the original sequence in the case of Asp isomerization, but not in the case of Asn deamidation, because isoAsp is converted into Asp in both cases. Thus, whereas PIMT can potentially repair RGD or other damaged functional sites that undergo Asp isomerization, it can lead to inactivation of isoDGR sites derived from NGR or DGR.
Pharmacological implications and therapeutic opportunities for peptides that target RGD-dependent integrins
Integrins are involved in many physiological and pathological processes, such as inflammation, thrombosis, osteoporosis, angiogenesis and cancer. The role of integrins in cancer biology and angiogenesis is one of the most studied functions of these cell adhesion receptors. For example, αvβ3-integrin, an integrin overexpressed in the tumor vascular endothelium, has an important role in angiogenesis and tumor growth (Brooks et al., 1995; Eliceiri and Cheresh, 2000; Friedlander et al., 1996; Kumar, 2003). RGD-containing peptides with a variable degree of affinity and selectivity for this integrin have been developed and used for delivering a variety of drugs and nanoparticles to tumor vessels (Desgrosellier and Cheresh, 2009; Liu et al., 2008; Ruoslahti et al., 2010). Given the structural and functional similarities between RGD and isoDGR it is conceivable that peptides containing the isoDGR motif could also be exploited as specific ligands for the targeted delivery of drugs, imaging agents or other compounds to tumors. Consistent with this, a cyclic CisoDGRC peptide that was coupled to fluorescent nanoparticles (quantum dots) was shown to bind αvβ3-integrin and colocalize with antibodies against αvβ3-integrin in vessels of human renal cell carcinoma (Curnis et al., 2008). Furthermore, extremely low doses (1–10 pg) of a recombinant protein made up of CisoDGRC fused to TNF-α, a cytokine capable of inflicting damages to tumor vessels, induced anti-tumor effects in tumor-bearing mice through specifically targeting TNF-α to tumor sites (Curnis et al., 2008).
The isoDGR motif might also be exploited, similar to RGD, for the generation of integrin antagonists, which are currently being investigated for the treatment of cancer, osteoporosis and coagulation disorders (Desgrosellier and Cheresh, 2009; Liu et al., 2008). For example, c(RGDf[NMe]V) (also known as Cilengitide) is a cyclic RGD-peptide that binds αvβ3- and αvβ5-integrins and inhibits angiogenesis. This drug is currently tested in a phase III trial in patients with glioblastoma (Desgrosellier and Cheresh, 2009). In principle, new integrin antagonists could be developed by using isoDGR instead of RGD. However, it should be born in mind that simply replacing RGD with isoDGR within the context of an existing compound is unlikely to work, because isoDGR has to interact with the RGD-binding pocket in an inverted orientation (see Fig. 3 for a schematic representation) and, moreover, the flanking residues also contribute to integrin-binding affinity and specificity as discussed above. The potential use of small molecules that contain isoDGR is supported by the results of our recent study, which shows that a cyclic CisoDGRC peptide alone inhibits the growth of tumors in mouse models (Curnis et al., 2010). Further work is necessary to assess the potential use of isoDGR compared with RGD, regarding stability (also to proteases), integrin-binding properties (Kd, kon and koff), signaling and pharmacological properties.
The capability of isoDGR peptides to interact with integrins might also have important implications for another class of drugs, i.e. drugs containing the NGR motif. NGR peptides have been used by different investigators to deliver cytokines, chemotherapeutic drugs, liposomes, anti-angiogenic compounds, viruses, imaging agents and DNA complexes to tumor neovasculature, in order to improve their therapeutic or imaging properties (reviewed by Corti and Curnis, 2011). For example, the CNGRCG peptide, fused to the N-terminus of TNF-α (NGR-TNF), and the GNGRAHA-peptide, fused to the C-terminus of tissue factor, are currently tested in several phase I, II and III clinical studies (Bieker et al., 2009; Gregorc et al., 2010a; Gregorc et al., 2010b; van Laarhoven et al., 2010; Corti et al., 2011). The rationale for using these peptides is the observation that NGR can recognize a membrane-bound form of aminopeptidase N (CD13) that is expressed by angiogenic vessels (Curnis et al., 2002; Pasqualini et al., 2000). Remarkably, in these compounds the NGR-to-isoDGR transition, which may occur during purification, storage or even in vivo for molecules that are circulating a long while, might simultaneously switch off CD13 and switch on integrin binding (Curnis et al., 2010) – with potentially important pharmacological and toxicological implications.
The formation of isoAsp in proteins of the ECM might represent a new mechanism for the regulation of integrin recognition by their ligands. Although, this protein modification typically impairs the functional properties of ECM proteins, it might however – in certain cases – generate new integrin-binding sites. The spontaneous transformation from NGR to isoDGR in fibronectin fragments, through Asn deamidation, and the consequent activation of integrin-binding sites is an example of how isoAsp formation can confer new functional properties to proteins. Spontaneous formation of isoAsp in proteins might occur also at Asp residues through isomerization reactions, suggesting that also the DGR motif can work, in principle, as an inactive precursor of isoDGR. Whereas the formation of isoDGR may represent a molecular switch for ‘turning on’ integrin binding, the release of PIMT from damaged cells could represent an enzyme-dependent mechanism for ‘turning off’ the switch by promoting conversion of isoDGR to DGR. However, to be physiologically relevant, the kinetics of these on–off changes need to be faster than protein turnover and the resulting isoDGR site accessible to integrins. Thus, although substantial experimental evidence suggests that proteins of the extracellular matrix can accumulate isoAsp in vivo, further studies are necessary to demonstrate that this reaction can, indeed, occur at NGR sites in vivo, and that it generates isoDGR in an amount that is able to substantially affect cell adhesion. As isoAsp formation can take hours, days or even years, depending on the molecular microenvironment, this mechanism is likely to have a significant role only in certain proteins. Further work is necessary to identify those proteins with sequence motifs that are transformed into isoDGR, to quantify its formation in normal and pathological tissues, and to assess their physiological roles.
The observation that isoDGR is generated in purified proteins and in drugs that contain the NGR motif suggests that this reaction also has important implications for their pharmacological and toxicological properties. Furthermore, the ability of synthetic isoDGR-containing peptides to mimic RGD and to bind to tumor vessels suggests that these compounds can be exploited, similar to RGD, as tools for targeted delivery of drugs and diagnostic nanoparticles to tumor vessels or for the development of new integrin antagonists.
This work was supported by Associazione Italiana per la Ricerca sul Cancro (AIRC) and Alleanza Contro il Cancro (ACC) of Italy, and FIRB.