Oxidation-induced Structural Changes of Ceruloplasmin Foster NGR Motif Deamidation That Promotes Integrin Binding and Signaling

Background: Asparagine deamidation at Asn-Gly-Arg (NGR) sites leads to the isoAsp-Gly-Arg (isoDGR) integrin-binding motif formation. Results: Ceruloplasmin (Cp), which contains two NGR sites and is oxidized in cerebrospinal fluid (CSF) in neurodegenerative diseases, can, undergo oxidation-induced structural changes fostering NGR deamidation with gain of integrin binding and signaling properties, in vitro and ex vivo in pathological CSF. Conclusion: Cp NGR motifs can deamidate acquiring integrin-binding functions. Significance: Cp structural changes favor NGR deamidation. Asparagine deamidation occurs spontaneously in proteins during aging; deamidation of Asn-Gly-Arg (NGR) sites can lead to the formation of isoAsp-Gly-Arg (isoDGR), a motif that can recognize the RGD-binding site of integrins. Ceruloplasmin (Cp), a ferroxidase present in the cerebrospinal fluid (CSF), contains two NGR sites in its sequence: one exposed on the protein surface (568NGR) and the other buried in the tertiary structure (962NGR). Considering that Cp can undergo oxidative modifications in the CSF of neurodegenerative diseases, we investigated the effect of oxidation on the deamidation of both NGR motifs and, consequently, on the acquisition of integrin binding properties. We observed that the exposed 568NGR site can deamidate under conditions mimicking accelerated Asn aging. In contrast, the hidden 962NGR site can deamidate exclusively when aging occurs under oxidative conditions, suggesting that oxidation-induced structural changes foster deamidation at this site. NGR deamidation in Cp was associated with gain of integrin-binding function, intracellular signaling, and cell pro-adhesive activity. Finally, Cp aging in the CSF from Alzheimer disease patients, but not in control CSF, causes Cp deamidation with gain of integrin-binding function, suggesting that this transition might also occur in pathological conditions. In conclusion, both Cp NGR sites can deamidate during aging under oxidative conditions, likely as a consequence of oxidative-induced structural changes, thereby promoting a gain of function in integrin binding, signaling, and cell adhesion.

Asparagine deamidation occurs spontaneously in proteins during aging; deamidation of Asn-Gly-Arg (NGR) sites can lead to the formation of isoAsp-Gly-Arg (isoDGR), a motif that can recognize the RGD-binding site of integrins. Ceruloplasmin (Cp), a ferroxidase present in the cerebrospinal fluid (CSF), contains two NGR sites in its sequence: one exposed on the protein surface ( 568 NGR) and the other buried in the tertiary structure ( 962 NGR). Considering that Cp can undergo oxidative modifications in the CSF of neurodegenerative diseases, we investigated the effect of oxidation on the deamidation of both NGR motifs and, consequently, on the acquisition of integrin binding properties. We observed that the exposed 568 NGR site can deamidate under conditions mimicking accelerated Asn aging. In contrast, the hidden 962 NGR site can deamidate exclusively when aging occurs under oxidative conditions, suggesting that oxidationinduced structural changes foster deamidation at this site. NGR deamidation in Cp was associated with gain of integrin-binding function, intracellular signaling, and cell pro-adhesive activity. Finally, Cp aging in the CSF from Alzheimer disease patients, but not in control CSF, causes Cp deamidation with gain of integrin-binding function, suggesting that this transition might also occur in pathological conditions. In conclusion, both Cp NGR sites can deamidate during aging under oxidative conditions, likely as a consequence of oxidative-induced structural changes, thereby promoting a gain of function in integrin binding, signaling, and cell adhesion.
Asparagine deamidation is a spontaneous chemical reaction that leads to the formation of aspartate and isoaspartate (isoAsp) 3 in proteins. This reaction can occur both in vitro, e.g., during protein storage, and in vivo during protein aging (1). The rate of Asn deamidation is determined by several factors, including protein sequence and secondary/tertiary structure that, juxtaposing Asn to specific functional groups, can inhibit or accelerate the reaction (1,2). For example, the presence of a glycine residue following Asn accelerates the deamidation reaction (3). Other elements that affect Asn deamidation are related to the environment like pH, ionic strength, and temperature (1). Asn deamidation has been mostly considered as a proteindamaging reaction because it introduces negative charges into a protein and, in the case of isoAsp formation, causes a change in the peptide backbone length (1). These modifications can alter protein function and stability and render the proteins more susceptible to degradation (1). Therefore, Asn deamidation has been viewed as a sort of molecular timer that marks the spontaneous protein aging (1,3). In vivo, Asn deamidation has been observed to occur in pathological conditions associated with oxidative stress (4 -7). For example, in Alzheimer disease (AD), deamidated forms of Tau protein and ␤-amyloid peptides have been reported to accumulate in characteristic pathological aggregates (6 -9).
However, Asn deamidation may also have other potential biological consequences. In fact, it has been shown that the tripeptide Asn-Gly-Arg (NGR) motif can undergo rapid deamidation reactions that result in the formation of isoDGR (isoAsp-Gly-Arg), a motif that can mimic RGD and recognize the RGD-binding site of integrins (10 -14). A blast search in a nonredundant Homo sapiens database for protein containing NGR sites indicates that ϳ5% of all proteins contain at least one NGR motif, and ϳ0.5% contain more than one NGR motif, suggesting a regulated functional role for this motif. Nevertheless, depending on the protein sequences and structures and on environmental conditions, only certain NGR sites with suitable features are likely to undergo deamidation.
The copper protein ceruloplasmin (Cp), a ferroxidase enzyme present in the cerebrospinal fluid (CSF) (15,16), contains two NGR sites ( 568 NGR and 962 NGR). We previously reported that Cp undergoes oxidative modifications in the CSF of Parkinson disease and AD patients (17), as a consequence of changes in the environmental redox status of pathological CSF (18,19). Cp oxidation causes a decrease in its ferroxidase activity that may have pathological implications (17). Because oxidative conditions might cause structural changes that accelerate Asn deamidation, we investigated whether the Cp NGR motifs can deamidate during aging under oxidative conditions and whether, as a consequence, Cp acquires integrin binding properties mediated by isoDGR.
Here, we show that although both Cp NGR sites can deamidate, the 962 NGR motif undergoes deamidation only when Asn aging occurs under oxidative conditions, in which structural changes are induced. The NGR to isoDGR transition in Cp induces gain of integrin binding function, integrin-mediated intracellular signals, and cell pro-adhesive activity. Finally, we show that Cp deamidation is faster in the CSF from AD patients compared with the one from healthy subjects, suggesting that this Cp modification, and the consequent gain of function, might occur in vivo in pathological conditions.

Patients
Samples were obtained from the Institute of Experimental Neurology INSPE-Biobank (San Raffaele Scientific Institute, Milan, Italy). After approval from the hospital's ethical committee and informed consent from patients, CSF samples (0.8 -1 ml) were collected by lumbar puncture. The analyzed groups were: Alzheimer disease patients (n ϭ 16) and healthy controls (n ϭ 16) ( Table 1). All patients were at first diagnosis and drugfree. Current criteria for the diagnosis of AD (20) were used for patients admission into the study. Exclusion criteria were: HIV or Hepatitis C virus seropositivity; the appearance of other neurodegenerative diseases or previous cerebral ischemic events; and severe metabolic disorders. Control CSF was from patients who underwent lumbar puncture on account of a suspected neurological disease and who proved to be normal and free from pathological alterations after complete CSF analysis and thorough clinico-neuroimaging assessment. Samples were centrifuged (800 ϫ g, 10 min at 4°C) to eliminate cells, then were either immediately processed, or stored at Ϫ80°C in an N 2 -supplemented atmosphere to avoid oxidation.

Oxidation and Asparagine Accelerated Aging Treatments
Oxidation and accelerated Asn aging treatments were performed by incubating purified Cp at 37°C in various buffers as described below. Oxidized Cp was prepared as follows: purified Cp was diluted at 1 mg/ml in PBS buffer containing in 10 mM H 2 O 2 solution and incubated for 16 h at 37°C. Deamidated Cp was prepared as follows: purified Cp was diluted 1 mg/ml in 100 mM ammonium bicarbonate (AmBic) buffer, pH 8.5, incubated for 16 h at 37°C, and stored at Ϫ20°C until analysis. This condition is known to favor Asn deamidation at the NGR site (10). Before use in the assays, Cp was dialyzed against PBS using Slide-A-lyzer dialysis cassette 10,000 molecular weight cutoff (Pierce). Oxidized and deamidated Cp was obtained by combining the above treatments. These products are hereinafter referred to as "oxidized Cp" (Cp-ox), "deamidated Cp" (Cp-AmBic), and "oxidized/deamidated Cp" (Cp-ox/AmBic), respectively. Aging of Cp in CSF was performed by adding purified Cp (final concentration, 20 g/ml) to CSF either from healthy subjects or from AD patients. CSF samples were then left to incubate for different time (0, 3, 6, 9, and 12 days) at 37°C under nitrogen conditioned atmosphere, to avoid the exposure to atmospheric oxidative environment.

Binding of Cp to Integrins and Competition with isoDGR
Peptide ␣5␤1, ␣v␤3, ␣v␤5, ␣v␤6, and ␣v␤8 integrins (1 g/ml in PBS with Ca 2ϩ/ Mg 2ϩ ; DPBS, Cambrex) were added to 96-well polyvinylchloride plates (BD Biosciences) and incubated 12 h at 4°C. Subsequent steps were carried out at 20°C. After blocking (3% BSA-PBS), the plates were filled with: Cp solutions (2-20 g/ml in binding buffer: 25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM MnCl 2 , 0,05% Tween, 1% BSA) or CSF samples from healthy subjects or AD patients supplemented with Cp (prepared as described above, diluted in binding buffer) and incubated for 2 h. Binding was detected using a polyclonal anti-human Cp Ab (Abcam ab8813) followed by a secondary HRP conjugate Ab (Abcam) and by o-phenylendiamine chromogenic substrate. Competitive binding assays were performed by mixing 10 g/ml of either the acetyl-CisoDGRCGVRSSSRTPSDKY peptide (isoDGR peptide) or the control peptide CARACGVRSSSRTPSDKY (ARA peptide) (12) with Cp-ox/AmBic solutions (20 g/ml) and mixing isoDGR peptide (30 g/ml) with CSF samples spiked with Cp (20 g/ml). The mixtures were added to microtiter plates coated with integrin, and the binding assay was carried out as described above. Cp bound to the ␣v␤6-coated plates was analyzed as follow; after the binding, the plates were washed and incubated with Laemmli buffer (5 min at 100°C) to promote protein detachment, the solution was then collected and subjected to SDS-PAGE and Western blot analyses as described (17) using a polyclonal anti-human Cp Ab (Abcam). Images were acquired using a laser densitometer (Molecular Dynamics).

Cell Adhesion Assay and Protein L-Isoaspartyl Methyltransferase (PIMT) Treatment
96-well polyvinyl chloride microtiter plates were coated with either untreated or treated Cp or with Cp that was immunoprecipitated (see below) after aging in healthy or AD CSF (1-20 g/ml in 50 mM Na 3 PO 4 , pH 7.3, 150 mM NaCl, 16 h at 4°C). After washing and blocking with 3% BSA-DPBS, the plates were seeded with HaCaT, EA.hy926, and T98G cells diluted in 0.1% BSA-DMEM (40,000 cells/well) and left to adhere for 3 h at 37°C. Adherent cells were fixed and stained with crystal violet as described (14), and adhesion was evaluated as adsorbance at 570 nm. For each treatment condition, cell spreading was evaluated comparing cell area in different microscopy fields. To measure cell area, pixels were automatically detected by image software analysis (Progenesis PG240; Nonlinear Dynamics). For PIMT treatment, plates coated as described above, were washed and filled with 45 l of 0.02 mM S-adenosyl-L-methionine in 50 mM Na 3 PO 4 , pH 6.8, and 5 l of PIMT solution (from IsoQuant isoaspartate detection kit; Promega) and incubated at 37°C for 16 h. Then plates were washed, and cell adhesion assay was performed as described above.

Cp Immunoprecipitation from CSF
After incubation in healthy or AD CSF, spiked purified Cp was immunoprecipitated using protein G-agarose beads (Invitrogen) coated with an anti-Cp antibody (Abcam ab8813). The antibodies were cross-linked to the beads with 20 mM dimethyl pimelimidate (Sigma) and added to the CSF supplemented with Cp; the solution was then incubated for 24 h at 4°C under gentle stirring. Beads were washed, and Cp was eluted with 0.1 M glycin solution, pH 2.5; samples were immediately diafiltered with PBS using an Amicon filter device.

Bathophenanthroline Assay
Ferroxidase activity of Cp after aging in CSF was evaluated on immunoprecipitated Cp by bathophenanthroline (Btp) assay as in Ref. 17. Briefly, immunoprecipitated Cp (1.25 g) was incubated with 80 mM FeSO 4 (ferrous form) and analyzed after 1 h with a solution of 1 mM Btp in acetate buffer, pH 6.2. A decrease in Btp-Fe 2ϩ complex absorbance at 535 nm derives from ferrous iron oxidation into ferric form (Fe 3ϩ ).

Mass Spectrometry Analysis
Samples were dialyzed against digestion buffer and incubated 2 h at 20°C with trypsin (10 ng/l; Roche Applied Science). Digested samples were desalted (Stage tips C18; Ther-moScientific) and injected in a capillary chromatographic system (EasyLC; Proxeon Biosystem). Peptide separations occurred on a 25-cm reverse phase silica capillary column, packed with 3-m ReproSil 100 Å C18 AQ. A gradient of acetonitrile eluents was used to achieve separation (flow rate, 0.15 l/min). MS analysis was performed by nanoLC-MS/MS using an LTQ-Orbitrap (ThermoScientific) equipped with a nanoelectrospray ion source (Proxeon Biosystems). Full scan spectra were acquired with the lock mass option, resolution set to 60,000, and mass range from m/z 350 to 1700 Da. The six most intense doubly and triply charged ions were selected and fragmented in the ion trap. All MS/MS samples were analyzed using Mascot (v.2.2.07; Matrix Science) and X!Tandem (within Scaffold software, v.2_06_00, 2007; Proteome Software Inc.) search engines to search the UniProt_Human Complete Proteome_ cp_hum_2012_07. Searches were performed with three missed cleavages allowed, N terminus acetylation, methionine oxidation, and deamidation of asparagine/glutamine as variable modifications. Mass tolerance was set to 5 ppm and 0.6 Da for precursor and fragment ions, respectively. Scaffold was used to validate MS/MS-based peptide and protein identifications. Protein thresholds were set to 99.0% and two peptides minimum, whereas peptide thresholds were set to 95% minimum.

Circular Dichroism and Melting Curves
CD spectra were acquired for Cp and Cp after oxidative and aging treatments (at 1 M in 50 mM PBS, pH 7) on a Jasco J-815 CD spectrometer at 20°C. Each spectrum was averaged using four accumulations collected in 0.1-nm intervals with an average time of 0.5 s. The protein spectra were corrected by sub-tracting the corresponding buffer spectra and then smoothed. The observed ellipticity (mdeg) was converted into molar residue ellipticity [] (deg cm 2 dmol Ϫ1 ). Melting curves were calculated recording CD spectra in continuous from 20 to 96°C (1°C intervals) in a wavelength range from 190 to 260 nm to achieve the better resolution. Melting temperature (T m ) was calculated by nonlinear fitting Boltzmann sigmoidal with Prism V4.03 software (GraphPad Inc.).

Molecular Dynamics Simulations and Docking Calculations
Homology Modeling-I-TASSER server was used to predict the secondary and tertiary structure of ␣v␤6 integrin (21).
Molecular Dynamics (MD) Simulations-MD simulations were performed on Cp wild type (P00450; Protein Data Bank code 2J5W), using the GROMACS 4.5.4 package (22) with the optimized parameters for liquid simulation force field. The 568 isoDGR-Cp and ␣v␤6 systems have been simulated in the same manner. Three independent 50-ns-long MD simulations (150-ns production run) were performed for Cp to allow for better conformational sampling, whereas three independent 10-ns-long MD simulations (30 ns) were performed for ␣v␤6 to get a set of structures to be used in docking calculations. All the analysis were performed using the GROMACS utilities on the last 40 ns of each simulation concatenated in a single trajectory (120 ns total). In particular, cluster analysis were performed using the Gromos algorithm.
Docking Calculations-Docking calculations of 568 isoDGR-Cp on the globular head of the extracellular part of ␣v␤6 have been performed using the docking program HADDOCK2.0 (23) with the optimized parameters for liquid simulation force field. We have docked a bundle of 27 568 isoDGR-Cp structures onto a bundle of 10 ␣v␤6 structures, which correspond to the centers of the clusters obtained from MD simulations. The protocol follows a three-stage docking procedure: (a) randomization of orientations and rigid body minimization, (b) simulated annealing in torsion angle space, and (c) refinement in Cartesian space with explicit water. Ambiguous interaction restraints (␣v␤6: residues 150, 218, 506, 547, 548, 601, and 815; 568 isoDGR-Cp: residues 568 -570) were derived from the known interactions of the isoDGR motif of the cyclic peptide with ␣v␤6 (11). The best 200 solutions in terms of intermolecular energies were selected for a semiflexible simulated annealing in which the side chains of ␣v␤6 and of the 568 isoDGR-Cp located at the binding interface were allowed to move in a semirigid body docking protocol to search for conformational rearrangements. The models were then subjected to a water refinement step. The single best docked solutions were analyzed according to hydrogen bonds, salt bridge contact, and buried surface accessibility.

Statistical Analysis
Categorical data were analyzed by using Fisher's exact test and two-tailed p value. Continuous data were evaluated by unpaired Student's t test, if the data passed the normality test for Gaussian distribution as assessed by the Kolmogorov-Smirnov test or were evaluated by Mann Whitney test; a two-tailed p value was used for the comparison of two means and standard error. In all analyses, p Ͻ 0.05 was considered to be statistically significant. The analysis was performed with Prism V4.03 software (GraphPad Inc.).

Reverse Phase Protein Array Analysis
HaCaT cells were incubated (30 or 120 min) with either Cp or aged/oxidized Cp. Cell lysates obtained with Zeptosens CLB1 TM lysis buffer (Bayer) were spotted on arrays and analyzed using Zeptosens custom service (Bayer). A set of 40 Abs, specific either for protein expression or for residue phosphorylation, that identify 25 pairs of phosphorylation rate for 16 different proteins belonging to the integrin signaling pathway (Kyoto Encyclopedia of Genes and Genomes; ko04510 and ko04810) were reacted in a standard Zeptosens profiling assay. Mean referenced fluorescence intensity (RFI) obtained from Ab reactivity were used to evaluate the proportion of specific residues phosphorylation (RFI Ϫ (phos-protein-x)/RFI Ϫ (proteinx)) for each protein. Then the phosphorylation proportion of the cell treated with aged/oxidized Cp was compared with that of the time-related control Cp (Cp-ox/Cp, 30 min; Cp-ox/Cp, 120 min) to define whether the phosphorylation rate was unaffected (value ϭ 1), induced (value Ͼ 1), or inhibited (value Ͻ 1) by Cp-ox. Phosphorylation rate was analyzed by hierarchical clustering performed with Mev4.6 v10.2 software (24); the 25 data series of protein phosphorylation rate in the two different conditions (Cp-ox/Cp, 30 min; Cp-ox/Cp, 120 min) were clustered into groups on the basis of the distances between the data series reported as covariance correlation.

Accelerated Aging of Cp under Oxidative Conditions Causes
Deamidation of NGR Sites-Sequence analysis of purified Cp by tandem MS showed that both 568 NGR and 962 NGR motifs were not deamidated ( Fig. 1A and B, untreated) and that protein oxidation by incubation in 10 mM H 2 O 2 at 37°C cannot convert NGR to DGR/isoDGR ( Fig. 1A and B, ox). In contrast, protein incubation in 100 mM AmBic or AmBic plus 10 mM H 2 O 2 , i.e., solutions that respectively mimic accelerated asparagine aging conditions (AmBic) (10) and accelerated asparagine aging under oxidative conditions (ox/AmBic), the 568 NGR motif was partially deamidated (Fig. 1A). Interestingly, the 962 NGR motif was partially deamidated exclusively following Cp aging under oxidative conditions (Fig. 1B ox/AmBic). The presence of Asp in place of Asn in the AmBic-treated Cp indicates that a deamidation reaction occurred at the two NGR sites of Cp, although MS did not ascertain whether DGR or isoDGR isoform was present.
A possible explanation for the different behavior of the 568 NGR and 962 NGR motifs may arise from analysis of the crystallographic Cp structure (Protein Data Bank code 2J5W) (25), which reveals that the 568 NGR motif is exposed on the surface of the protein, whereas the 962 NGR motif is less exposed and therefore less accessible to the solvent ( Fig. 2A and B). The two sequences display differing tertiary contexts ( Fig. 2C and D) that might in turn influence deamidation rates differentially. We then attempted to clarify the structural differences between the two sequences by a computational approach. Molecular dynamics simulations on wild type Cp (Protein Data Bank code 2J5W) showed a relatively stable structure with large fluctua-tions of loop regions (Fig. 2E). Notably, the root mean square fluctuations, as calculated throughout the simulation, suggest that the two NGR sequences are more rigid than the other residues (Fig. 2F). Importantly, the 568 NGR sequence is steadily accessible to the solvent throughout the simulation, whereas 962 NGR is buried inside the protein and is blocked in a stable conformation by polar interactions with neighboring amino acids ( Fig. 2C and D). Accordingly, we hypothesized that oxidation might affect the Cp structure and promote the exposure of the 962 NGR motif, which in turn adopts a favorable confor-mation for deamidation. Indeed, it has been previously inferred that Cp structure is affected by oxidation (17, 26 -28).
Cp Oxidation Induces Secondary Structure Changes-CD spectra obtained for purified Cp indicated that the secondary structure of Cp, which is principally composed of ␤-strands (45.5%, as evaluated by Uniprot tool), was partially affected by Cp-AmBic treatment (Fig. 2G, black line versus blue line); notably, large spectra changes occurred upon Cp-ox and -ox/AmBic treatment, suggesting structural alterations (Fig. 2G, black line  versus green and orange lines). Interestingly, the spectra profile obtained with heat-denatured Cp (Fig. 2G, red line), which showed a pronounced minimum toward random coil structure (ϳ200 nm), differed from that of Cp obtained in -ox/AmBic conditions. These data suggest that Cp oxidation or accelerated Asn deamidation under oxidative conditions induces secondary structure changes but does not unfold the protein. Comparison of the protein melting temperatures (T m ) before and after the different treatments shows that oxidation induces the strongest destabilizing effect, as assessed by the decreased T m after oxidation (Cp, 54.2°C; Cp-ox/AmBic, 46.5°C; Cp-ox, 40.4°C); in contrast, the AmBic treatment did not change the Cp thermostability (Cp-AmBic, 53.3°C) (Fig. 2G).
Cp Aged under Oxidative Conditions Binds Integrins-We investigated whether Cp Asn accelerated aging under oxidative conditions, i.e., the condition that may occur in some neurodegenerative diseases, can induce integrin binding via isoDGR formation. In vitro binding assays to purified integrins showed negligible binding of untreated Cp, whereas after accelerated Asn aging in AmBic alone or in the presence of oxidative conditions (-ox/AmBic), Cp acquired clear binding properties to ␣v␤5, ␣v␤6, ␣v␤3, and ␣v␤8, albeit to differing degrees (Fig.  3A). Of note, oxidation alone was ineffective (not shown). Based on the observed additional increase of binding of Cp aged under oxidative conditions to ␣v␤6 compared with Cp aged alone, and with the ␣v␤6 integrin expressed by epithelial cells, this integrin was used for further analysis. Interestingly, the -ox/ AmBic treatment increased the binding to ␣v␤6 (Fig. 3A) in a dose-dependent manner (Fig. 3B). Competitive binding experiments, performed either with the isoDGR peptide capable to bind the RGD binding pocket of integrins (12) or with the control ARA peptide, showed that the isoDGR peptide could abolish the binding of oxidized-aged Cp to the ␣v␤6 integrin (Student's t test, p Ͻ 0.0001) (Fig. 3C), supporting the role of Cp NGR deamidation in the acquisition of binding activity after aging. The effect of Cp accelerated aging under oxidative conditions was also investigated by Western blot analysis. Untreated Cp showed a principal band of 150 kDa, as expected for a full-length protein (Fig. 3D), whereas the treatment caused the appearance of an additional band at Ͼ250 kDa. These data suggest that the treatment generating reactive succinimide intermediates on Cp may cause intermolecular aggregation, likely caused by reaction with the amino group present in neighbor Cp molecules (29). Remarkably, both products, full-length Cp and aggregates, bind ␣v␤6 integrin (Fig. 3D). Taken together, these results indicate that accelerated Asn aging under oxidative conditions promotes the Cp ability to bind integrin RGD-binding pockets, presumably via NGR to isoDGR conversion.

Molecular Dynamics Simulations and Docking Calculations Indicate That 568 isoDGR-Cp Can Interact with ␣v␤6 Integrin-
Our results indicated that the accelerated Asn aging under oxidative conditions leads to the deamidation of both 568 NGR and 962 NGR motifs in the Cp sequence. Computational approaches have been exploited to evaluate the possibility for deamidated

. NGR motifs structural analysis, Cp molecular dynamics modeling, and circular dichroism analysis indicate that 962 NGR is buried inside
Cp structure and that secondary structure changes induced by oxidative conditions might allow its surface exposure. A and B, surface representation of Cp. In red are highlighted the 568 NGR and 962 NGR sequences, respectively, as calculated by the GROMACS g_sas tool on crystallographic Cp structure (Protein Data Bank code 2J5W). The higher solvent-accessible surface of 568 NGR is represented by the larger size of the red-colored region. C and D, cartoon representation of 568 NGR and 962 NGR sequences in Cp. The NGR amino acids are shown as orange sticks. The hydrogen bonds are shown as red dashed lines. 568 NGR is engaged in short range polar interactions between GLY and ASN, whereas 962 NGR sequence is mostly involved in strong interactions with the rest of the protein. E, C␣ root mean square deviations from the initial Cp structure. The three different trajectories calculated by molecular dynamics modeling, shown in red, black, and green, respectively, indicate a relatively stable structure. Each spectrum was averaged using four accumulations collected in 0.1-nm wavelength intervals with an average time of 0.5 s. The observed ellipticity (mdeg) was converted into molar residue ellipticity [] (deg cm 2 dmol Ϫ1 ). For each treatment, the melting temperature (T m ) is also indicated. FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6

JOURNAL OF BIOLOGICAL CHEMISTRY 3741
Cp to bind ␣v␤6 integrin. However, the experimental evidence reported in Fig. 2 suggests that the deamidation of 962 NGR in oxidative conditions causes a considerable structural rearrangement that cannot be predicted by computational methods. Thus, we decided to perform molecular dynamics simulation only on 568 isoDGR-Cp protein to generate a set of reliable structures to be docked onto ␣v␤6.
Cluster analysis on the 568 isoDGR-Cp molecular dynamics allowed extraction of a number of representative structures of the simulations. A bundle of 27 isoDGR structures, corresponding to the center of the clusters, were docked onto a bundle of 10 ␣v␤6 structures, which are the center of the clusters obtained from 10 ns of molecular dynamics of the free ␣v␤6.  (Fig. 4B).  Structures belong to the same cluster if they differ by less than 2 Å in the pairwise root mean square deviation matrix. The HADDOCK score corresponds to the weighted sum of different energy terms (van der Waals, electrostatic, and restraint energies). Cluster 3 is the best solution that is represented in B. B, HADDOCK model of isoDGR-Cp/␣v␤6 integrin binding site. Surface representation of ␣v␤6 binding pocket (green) in complex with isoDGR-Cp (cyan cartoon). The side chains of the residues forming stable interactions (as reported in text) are shown in licorice. Orange spheres correspond to Cu 2ϩ ions; magenta spheres correspond to Ca 2ϩ cation; green spheres correspond to Na ϩ ions. Red dotted lines denote the hydrogen bonds of the Cp with ␣v and ␤6 subunits. In magenta licorice is shown the 962 NGR motif. The arginine of 568 isoDGR predominantly interacts with Asp 150 of ␣v domain, whereas the same arginine can transiently interact with Asp 218 . The isoaspartyl residue of isoDGR interacts with Ca 2ϩ MIDAS ion located in the ␤6 domain (see zoomed area). Other interactions between 568 isoDGR-Cp and ␣v␤6 have been highlighted; in particular we noticed a strong and stable network of salt bridges between Lys 562 and Lys 369 of isoDGR-Cp and Asp 146 and Asp 147 of ␤6 domain of ␣v␤6 integrin (interaction reported in text).

Deamidated Cp Mediates Cellular Adhesion and Spreading-
The binding to integrins observed in vitro might not reflect the real binding properties of deamidated Cp, given the possibility that adsorption of purified integrins on microtiter plates might alter their conformation. Furthermore, we have previously shown that the isoDGR site of fibronectin but not its corresponding NGR site can promote cell adhesion (10 -12, 14). To assess whether Cp can generate an isoDGR motif able to recognize integrins in physiological conformations, we analyzed the effect of Cp aged under oxidative conditions on the adhesion of human epithelial (HaCaT), endothelial (E.A.hy926), and glioblastoma (T98G) cell lines. Preliminary flow cytometry analysis showed ␣v␤6 expression by HaCaT, ␣v␤3 and ␣v␤5 expression by EA.hy926, and ␣v␤8 expression by TG98 (data not shown).
Plates coated with ox/AmBic-Cp increased in a dose-dependent manner the adhesion and spreading of HaCaT cells, whereas untreated Cp induced little cell adhesion (Fig. 5A and  B). In HaCat cells adhering to ox/ambic-Cp coated plates, cell spreading showed a significant (p ϭ 0.0087) area increase (ϳ20%) as compared with cells adhering to Cp coated plates (Fig. 5C). Similar pro-adhesive effects were observed with glioblastoma ( Fig. 5D and E) and endothelial cell lines (data not shown).
To assess whether cell adhesion was mediated by isoDGR, we incubated ox/AmBic-treated Cp with PIMT, an enzyme that converts L-isoAsp residues to L-Asp (10,30). PIMT treatment almost completely inhibited the pro-adhesive activity of ox/AmBic-treated Cp and of isoDGR peptide, the latter serving as a control (Student's t test, p Ͻ 0.0001) (Fig. 5D and E). These results suggest that isoAsp formation, presumably at NGR site(s) of Cp, is associated with a "gain of function" in cell adhesion assays.
Cp Aged under Oxidative Conditions Transduces Intracellular Signal through Integrin Engagement-Having observed that Cp after accelerated Asn aging under oxidative conditions was able to bind integrins expressed on the cell surface, we investigated whether this binding could also induce intracellular signaling. Signal transduction pathway analysis, as performed by reverse phase protein arrays on epithelial cells, indicated that cell incubation with ox/AmBic-treated Cp can induce coordinated phosphorylation events that recapitulate many of the steps of classical integrinmediated signal transduction. Hierarchical clustering analysis of the comparison of treatment with either ox/AmBic-Cp or untreated Cp revealed that after 30 min: (a) there was an increase in the phosphorylation of the activation residues of several molecules, e.g., p-Tyr 397 FAK1, p-Thr 514 PKC␥, p-Ser 217/221 MEK1, p-Thr 185/202 Tyr 204 ERK1/2, p-Ser 241 PDK1, p-Ser 473 Akt, and (b) the phosphorylation of inhibitory residues, e.g., p-Ser 9/21 GSK3␤, which in turn maintain ␤-catenin activity by preventing its phosphorylation ( Fig. 6A and B). Several other proteins were unaffected (such as CrKl, SAPK/JNK, c-Jun, and Raf1) or slightly inhibited (PTEN and PKC␣) (Fig. 6A and B). These results suggest that early signals mediated by deamidated Cp mostly addressed gene activation regulation, cell cycle, and MAPK signaling pathway (Fig. 7). In contrast, late signals (i.e., immediately after 120 min of treatment) seemed to sustain actin cytoskeleton rearrangement rather than cell survival, proliferation, and MAPK pathway activation (Fig. 7). Src and CrkL were acti-vated, and the phosphorylation of FAK1 was inhibited, concomitantly with the slight phosphorylation of PTEN and Raf1 inhibitory molecules, which block MEK1 and ERK1/2 activation ( Fig. 6A and B; Fig. 7). Similarly, Akt inactivation results in ␤-catenin inhibition by Ser 33/37/45 Thr 41 phosphorylation, as mediated by GSK3␤ activation (Fig. 6A and B; Fig. 7).

Pathological CSF from Alzheimer Disease Fosters Integrin Binding of Spiked Purified Cp and Inhibits Its Ferroxidase
Activity-To investigate whether Cp deamidation might occur also in vivo, in particular in pro-oxidant pathological CSF environments, we added purified Cp to CSF from healthy subjects or CSF from AD patients and incubated the mixture for 0, 3, 6, Cp-ox/AmBic, untreated Cp. Cell spreading is visible at higher magnification showed in the boxes. C, cell spreading was evaluated based on the pixel area automatically detected by image software analysis. Statistical significance was evaluated on cells measurement in three different microscopy fields for each condition (Cp, n ϭ 112; Cp-ox/AmBic n ϭ 208). **, p ϭ 0.0087. D, adhesion of glioblastoma T98G cells to plastic coated either with BSA (None), aged/ oxidized Cp (ox/AmBic), or the isoDGR peptide. Inhibition of cells adhesion occurred after treatment of coating proteins with PIMT (ϩ) compared with the untreated wells (Ϫ). E, microscopy images of stained cells adherent to ox/AmBic-Cp treated with (ϩ) or without (Ϫ) PIMT enzyme. All the adhesion assays were performed in triplicate for three independent experiments (n ϭ 3). Statistical significance reported as p values was evaluated by Student's t test. ***, p Ͻ 0.0001. 9, and 12 days at 37°C. Cp was added in large amounts because of the presence of other integrin-binding molecules in CSF (e.g., fibronectin, tenascin) that can compete the binding of Cp. Aging in a pathological milieu was able to induce a time-dependent Cp significant binding to ␣v␤6 (p Ͻ 0.0001) (Fig. 8A, CSF/ AD), whereas binding was absent or very weak when aging occurred in normal CSF (Fig. 8A, CSF). Thus, Cp deamidation showed a faster kinetic in the CSF from AD patients than from healthy subjects. By comparison with the binding ability of the same amount of in vitro chemically aged Cp (Cp-ox/AmBic), we ascertained that ϳ40% of the added Cp was converted to pro-adhesive Cp after 12 days of aging in AD patients CSF (Fig.  8B). Also in the case of purified Cp aging in the CSF, the gain in integrin binding properties was mediated by Cp deamidated NGR motifs, as demonstrated by competition with an isoDGR peptide, which abolished binding to ␣v␤6 (p ϭ 0.0004) (Fig. 8C, gray bars). To investigate whether Cp aged in pathological conditions was also able to recognize integrins in their physiological conformation, we performed cell adhesion experiments. HaCat cells were seeded on plates coated with Cp immunoprecipitated from either AD CSF or control CSF at time 0 or after 9 days of aging. The Cp aged in pathological CSF was able to promote the HaCat cell adhesion (Fig. 8D, black bars), whereas little or no adhesion was observed on Cp aged in control CSF (Fig. 8D, white bars). In addition to the acquisition of integrin binding function, Cp aged in AD CSF showed a complete loss of ferroxidase activity (p Ͻ 0.0001), whereas the Cp aged in healthy CSF showed a reduction of ϳ40% (Fig. 8E, 9 days).

DISCUSSION
Asparagine deamidation during protein aging depends on neighboring residues as well as on secondary and tertiary structural elements. The Asn deamidation rate is also affected by environmental factors such as pH, temperature, and ionic strength (1,2). The combination of these features controls the kinetic of deamidation reactions that can range from hours to years (1,2). Thus, for each protein it is important to assess whether specific Asn residues can undergo or not this post-translational modification. For example, we previously reported that the deamidation of an NGR site of fibronectin does not occur in the intact protein, whereas it can rapidly occur in protein fragments (10). The results reported here show that deamidation of the 568 NGR site, which is exposed on the protein surface, does not require structural changes. In contrast, deamidation of the 962 NGR site, which is buried inside two ␤-strands of the protein structure and is blocked in a stable conformation by polar interactions with neighboring amino acids, can occur only after structural changes induced by an oxidative microenvironment. Protein oxidation and conformational changes similar to those reported here have been shown to cause copper ions to release from Cp (17,26,28). Noticeable, copper ions are necessary for the Cp ferroxidase activity (17,26,28). Because protein aging under oxidative stress is a condition that may occur in some neurodegenerative diseases (19,(31)(32)(33), we hypothesize that Cp aging under pathological conditions promotes modifications that favor structural changes that in turn foster both copper ion release and NGR deamidation. Copper ions released from Cp might contribute to the reported increase of the pool of free copper and to the decrease of active Cp found in the CSF of AD patients (34 -36). It is conceivable that this increase, affecting the redox environment (37), might contribute to the observed Cp modifications by a feedback mechanism. The observed loss of ferroxidase activity of Cp aged in AD CSF suggests that changes in the Cp structure can be fostered also ex vivo. Given the role of Cp, ferroxidase activity reduction may affect the iron homeostasis (38), causing an increase in iron retention as seen in (17,26,28).
A relationship between oxidation and protein deamidation in neurodegenerative disease is also suggested by the observation that several proteins found to be deamidated in the brain of PIMT knock-out mouse were oxidized in the brain of AD patients (30). We previously reported that in the CSF from Par-kinson disease and AD patients, Cp showed modifications derived from protein oxidation (17). Similar modifications have been reported for serum Cp during human aging (26). Thus, the Cp alterations observed in Parkinson disease and AD patients might reflect accelerated protein aging, as a consequence of changes in the environmental redox status of pathological CSF (18,19). Therefore, it is plausible that the reported Cp oxidation-induced changes (17) also include Asn deamidation of NGR motifs.
Another major finding of this work is that the NGR to isoDGR conversion in Cp induces a gain of function in terms of binding to several ␣v integrins and, above all, the fact that the interaction of deamidated Cp with integrins can trigger an intracellular signaling cascade. We previously demonstrated that the NGR to isoDGR transition in fibronectin can work as a molecular switch for integrin-ligand recognition; the same mechanism seems to explain the gain of integrin binding function in Cp aged under oxidative conditions (10 -12, 14). It is of relevance that, in contrast to fibronectin, the binding to integrin occurs for the intact full-length deamidated Cp. Consistently with binding data, our docking studies show that at least the 568 isoDGR motif of intact Cp can interact with the canonical RGD binding pocket of the ␣v␤6 having the stereochemical and electrostatic requirements for a correct recognition. Because it is not possible to predict a priori the putative structural rearrangement induced by the oxidation, we cannot describe in silico the possible interactions occurring between integrin and the 962 isoDGR sequence in the context of full-length Cp. The capability of the intact deamidated Cp to bind integrins might explain the ability to promote integrin activation.
These findings suggest that the Cp deamidation can produce a biological relevant gain of function. The functional importance of the NGR sites of Cp is also suggested by the high conservation of these tripeptide sequences across differing species from hamster to human. Remarkably, very few substitutions in various species were also shown in the 962 NGR flanking sequences, which are likely crucial for regulating the deamidation rate.
Notably, Cp NGR deamidation might occur in a relevant manner in certain pathological conditions, as suggested by the observation that Cp ex vivo aged in the CSF from AD patients acquires integrin binding and cell adhesion properties. The faster kinetic of Cp deamidation observed in CSF from AD patients compared with the CSF from healthy subjects supports the hypothesis of a generally accelerated protein aging in neurodegenerative diseases, as a consequence of pathological environment (18,19). Based on the observation that after 12 days of aging in the pathological CSF ϳ40% of the Cp were able to bind integrin and given that the Cp concentration in the CSF is ϳ2 g/ml and that its physiological half-life is ϳ5.5 days in vivo (15), it is conceivable that NGR deamidation might occur in patients to an extent that yields functionally relevant concentration of deamidated Cp (ϳ200 ng/ml). The observed induction of NGR modification in Cp incubated in the CSF from AD patients implies the presence of factors that foster/accelerate deamidation, either directly or indirectly as a consequence of the induction of structural changes. However, further studies are necessary to define the differential composition of healthy and AD CSFs as to identify factors underlying the increased deamidation rate observed in pathological CSF. A, binding to ␣v␤6 of purified Cp added to CSF from healthy subjects (CSF, white squares) or AD patients (CSF/AD, black squares) aged for 0, 3, 6, 9, and 12 days at 37°C. B, quantitation of the Cp bound to ␣v␤6 integrin after ex vivo aging in CSF evaluated by the comparison with binding of chemically in vitro aged Cp (Cp-ox/AmBic, 100%). Aging from 0 to 9 days at 37°C of purified Cp in the CSF from AD patients (CSF/AD) showed ϳ40% of the protein binding ability to ␣v␤6, whereas the same aging conditions in the CSF from healthy subjects showed negligible binding ability (3-4%). C, competition with isoDGR peptide (gray bars) of the binding to ␣v␤6 of Cp aged in CSF from healthy subjects or AD patients (CSF and CSF/AD, respectively, white bars). D, HaCat cell adhesion to plates coated with immunoprecipitated purified Cp (5 g/ml) after ex vivo aging (0 or 9 days) in CSF from healthy subjects (CSF, white bars) or AD patients (CSF/AD, black bars). Adhesion was evaluated as absorbance (570 nm) of the crystal violet-stained cells and was expressed as a percentage of control (cells in adhesion on BSA-coated wells). E, Cp ferroxidase activity after incubation in CSF from healthy subjects (CSF, white bars) or AD patients (CSF/AD, black bars). Ferroxidase activity of immunoprecipitated Cp (1.25 g) was evaluated measuring the decrease in Btp-Fe 2ϩ complex absorbance at 535 nm. The values are expressed as percentages of total ferroxidase activity measured for 1.25 g of untreated Cp. All the assays were performed in triplicate for two independent experiments, by using CSF from n ϭ 8 different subjects each group (Table 1). Statistical p value was evaluated by Student's t test. ****, p Ͻ 0.0001; ***, p Ͻ 0.001; **, p Ͻ 0.01.
It is known that integrin binding may activate intracellular signaling cascades, which in turn can affect cell differentiation, survival, growth, and division (39). Thereby, the hypothesis that in neurodegenerative diseases Cp might deamidate and transduce unusual intracellular signals opens new investigation perspectives in the AD field. It has been reported that an aberrant signaling mediated via integrin engagement (e.g., ␣V␤1, ␣2␤1, ␣V␤3, and ␣V␤5) by fibrillar amyloid-␤ in the CNS can result in unconventional FAK activation and downstream signaling (i.e., MAPK, GSK␤3, etc.) (40 -42). A similar signaling mechanism is triggered by deamidated Cp in epithelial cells: we observed inhibition of FAK and MAPK signaling, which could lead to cell cycle arrest, proliferation inhibition, and cytoskeleton reorganization. Various cell types in the CNS might be activated via integrins, including microglial cells, which contribute to the inflammation mechanisms in neurodegenerative diseases (43). Interestingly, it has been reported that Cp can activate microglial cells (44), although the receptor underlying this mechanism has not yet been identified: one possibility is that deamidated Cp interacts with integrins. Other potential cell targets of deamidated Cp include the specialized epithelial cells of the ependymal layer and choroid plexus, which are directly in contact with the CSF and have been reported to be altered in AD (45)(46)(47).
In conclusion, the results show that Asn deamidation reactions, which may occur in Cp upon aging, can convert this protein into an isoDGR-containing ligand of ␣v integrins. Furthermore, integrin-mediated intracellular signaling can be transduced by the modified protein. Oxidation-induced structural changes foster the NGR to isoDGR transition at the NGR site buried in the protein structure. Even though the mechanism we described results mainly from in vitro observations, the ex vivo evidence that the CSF from AD patients promotes Cp deamidation suggests that Cp modifications might also occur in patients. Further investigations are needed to assess the biological role of in vivo Cp deamidation and the consequences for neurodegenerative diseases.