Detection of Tumor-Associated Glycopeptides by Lectins: The Peptide

Dec 2, 2014 - Vassiliki Peppa , Hemalatha Venkat , Anastassia Kantsadi , Shashikala Inamdar , Ganapati Bhat , Sachin Eligar , Anupama Shivanand , Vish...
1 downloads 10 Views 5MB Size
Articles pubs.acs.org/acschemicalbiology

Detection of Tumor-Associated Glycopeptides by Lectins: The Peptide Context Modulates Carbohydrate Recognition David Madariaga, Nuria Martínez-Sáez, Víctor J. Somovilla, Helena Coelho,† Jessika Valero-González,‡ Jorge Castro-López,‡ Juan L. Asensio,§ Jesús Jiménez-Barbero,∥ Jesús H. Busto, Alberto Avenoza, Filipa Marcelo,† Ramón Hurtado-Guerrero,‡ Francisco Corzana,* and Jesús M. Peregrina* Centro de Investigación en Síntesis Química, Departamento de Química, Universidad de La Rioja, E-26006 Logroño, Spain S Supporting Information *

ABSTRACT: Tn antigen (α-O-GalNAc-Ser/Thr) is a convenient cancer biomarker that is recognized by antibodies and lectins. This work yields remarkable results for two plant lectins in terms of epitope recognition and reveals that these receptors show higher affinity for Tn antigen when it is incorporated in the Pro-Asp-Thr-Arg (PDTR) peptide region of mucin MUC1. In contrast, a significant affinity loss is observed when Tn antigen is located in the Ala-His-Gly-ValThr-Ser-Ala (AHGVTSA) or Ala-Pro-Gly-Ser-Thr-Ala-Pro (APGSTAP) fragments. Our data indicate that the charged residues, Arg and Asp, present in the PDTR sequence establish noteworthy fundamental interactions with the lectin surface as well as fix the conformation of the peptide backbone, favoring the presentation of the sugar moiety toward the lectin. These results may help to better understand glycopeptide−lectin interactions and may contribute to engineer new binding sites, allowing novel glycosensors for Tn antigen detection to be designed.

L

example of this concept in the context of natural MUC1 glycopeptides carrying the Tn antigen. The extracellular domain of MUC1 consists of tandem repeats comprising 20 amino acids of the sequence HGVTSAPDTRPAPGSTAPPA with five possible O-glycosylation sites and three relevant regions of MUC1 glycoproteins: the GVTSA sequence, present in compound 1 (Figure 1), is an effective substrate for GalNAc transferases;11,23 the PDTR fragment, found in compound 2, is the most immunogenic domain of MUC1 and, consequently, is a well-known epitope recognized by several anti-MUC1 antibodies;24 and the GSTA region, included in compound 3, which is recognized by different antibodies and represents a potential tool for use in diagnosis and therapeutic applications25 (Figure 1). To investigate how the peptide sequence influences MUC1 antigen−lectin recognition, we have elaborated a multidisciplinary approach involving (glyco)peptide synthesis, enzyme-linked lectin assay (ELLA) tests, isothermal titration microcalorimetry (ITC) studies, X-ray analysis, saturationtransfer difference (STD) NMR experiments, and molecular modeling simulations. Fortunately, our efforts yielded noteworthy results in terms of epitope recognition, showing evidence that the residues that flank the Tn moiety play an active role in the molecular recognition process by two plant lectins: soybean agglutinin (SBA) and Vicia villosa-B-4 agglutinin (VVA).26−30 According to our data, Tn antigen is

ectins are a class of carbohydrate-binding proteins that trigger several important cellular processes.1 They have been successfully employed to recognize malignant tumors, and a few are indicated for the reduction of treatment-associated side effects as adjuvant agents in chemotherapy and radiotherapy. It has been observed that certain lectins present antitumor activity and can directly kill human cancerous cells.2,3 These features make lectins a current niche of research that may be adopted as an alternative cancer therapy.4−6 Within the context of cancer, Tn antigen (α-O-GalNAc-Ser/ Thr) is one of the most specific human tumor-associated structures.7 Expression of this motif occurs early in tumor cells. In fact, there is a direct correlation between carcinoma aggressiveness and density of this antigen. In general, Tn antigen is presented in cancer cells as a part of modified glycoproteins, such as mucins, and in particular mucin MUC1.8−12 Therefore, MUC1 derivatives have become attractive molecules for the treatment of cancer.8,13−17 Logically, the premature detection of Tn antigen is essential to properly treat and eradicate tumors. In this regard, a variety of lectins that bind to glycopeptides bearing Tn motifs have been successfully employed as biomarkers18,19 of this antigen. From a molecular recognition viewpoint, it is well-documented that lectins recognize the sugar moiety as a free form or linked to peptides.1 Although there is evidence that the natural residue (Ser or Thr) bearing the sugar unit does indeed have a clear effect on binding,20 little is known about the role that the amino acids flanking the glycosylation point play in the recognition process.21,22 Herein, we study at the structural level a relevant © XXXX American Chemical Society

Received: October 21, 2014 Accepted: November 26, 2014

A

dx.doi.org/10.1021/cb500855x | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

almost undetectable by the lectins when it is presented in the specific peptide sequence AHGVT(αGalNAc)SA. In contrast, the glycosylated APDT(αGalNAc)R sequence is the best epitope for the two selected lectins. Our work shows chemical evidence indicating that this is not merely a matter of chance. As we describe in the present work, the APDTR sequence fixes the bioactive conformation, establishing direct interactions with the protein to properly present the sugar moiety to the corresponding lectin. Indeed, it is shown that the charged residues of this sequence (Asp and Arg amino acids) play a fundamental role in the binding process. The results presented here can help to better understand glycopeptide−lectin interactions and may contribute to engineer new building sites or to design specific mutants of the lectins, allowing the specificity to be tuned and the structure-based design of novel glycosensors that, within the cancer context, are mandatory to early detection of tumor cells.

Figure 2. ELLA strategy used in this work together with binding affinity of glycopeptides 1−3 to SBA lectin. Absorbance signals are the average of three replicate wells, and the error bars show the standard deviation of these measurements.



compounds 1 and 2 differ by 2 orders of magnitude. Thermodynamic binding parameters (Table 1 and Supporting Information Figure S1) for glycopeptide 2 and SBA highlight a binding process highly favored by enthalpy, with a reduce entropy penalty when compared to that of the corresponding glycopeptide 3 counterpart. These results indicate that more intermolecular interactions should take place between compound 2 and the protein surface of SBA. In addition, conformational behavior, in solution and complexed to SBA, could be distinct when glycopeptides 2 and 3 are compared. Encouraged by these results, we decided to investigate whether the affinity of these short molecules could be modified by the presence of the rest of the amino acids that comprise the MUC1 tandem repeat domain. Consequently, we synthesized and studied the glycopeptides shown in Scheme 1. The results obtained from the ELLA performed on these glycopeptides were equivalent to those obtained for short glycopeptides 1−3, with derivative m2 being the compound that clearly showed the best binding to SBA lectin (Supporting Information Figure S2). Taking into account that compound m1, with the lower affinity, has its epitope close to the plastic surface and to prevent artifacts due to accessibility issues, we synthesized glycopeptide m1′, with the epitope situated opposite the amino acid linked to the plastic surface (Supporting Information Figure S3). The comparable results obtained for m1 and m1′ ruled out the existence of accessibility difficulties in the recognition of glycopeptide m1 by SBA lectin. To generalize our results, we also estimated the binding of glycopeptides m1−m3 to Vicia villosa agglutinin (VVA) lectin. The results were in agreement with those observed for SBA lectin (Supporting Information Figure S4). All of these tests disclose that the underlying peptide sequence does play an active role in the molecular recognition process. In fact, our preliminary studies with glycine-rich glycopeptides bearing Tn antigen indicate a clear decrease of affinity with SBA lectin.20 To rationalize these experimental findings, and considering the similar results found for the short (1−3) and large glycopeptides (m1−m3), we performed a thorough conformational analysis in solution of glycopeptides 1−3 in both the free and bound states with SBA lectin. First, full assignment of the protons of the three compounds was performed using COSY and HSQC experiments (Supporting Information). Of note, in glycopeptides 2 and 3, a second set of signals (a small percentage) was observed, especially in the NH region. They correspond to the cis disposition of the amide bond of proline

RESULTS AND DISCUSSION We have used two model plant lectins as receptors: soybean (SBA) and Vicia villosa-B-4 (VVA) agglutinins. These lectins recognize GalNAc26−30 and are fairly stable and commercially available. The selected glycopeptides shown in Figure 1 comprise the three important regions of MUC1 glycoproteins.

Figure 1. Small glycopeptides synthesized and studied in this work that contain the three relevant epitopes of mucin MUC1.

The synthesis of compounds 1−3 was carried out following standard solid-phase peptide synthesis (SPPS) procedures (Methods). First, we investigated the binding properties of these derivatives to SBA lectin by using an enzyme-linked lectin assay (ELLA) (Figure 2). For this purpose, different amounts of glycopeptides 1−3 (from 0 to 1.5 mM) were covalently attached to a maleic anhydride activated surface. Biotinylated SBA lectin was then selected for estimating the binding. Horseradish peroxidase− 3,3′,5,5′-tetramethylbenzidine (TMB) was chosen as the detection system, which was quantified by measuring the absorbance of the wells at 450 nm (Methods). According to ELLA measurements, SBA lectin showed the maximum affinity for compound 2. In contrast, binding of glycopeptides 3 and 1 was very low and negligible, respectively. ELLA affinity evaluation was subsequently corroborated by ITC measurements. The results are in-line with those estimated using ELLA (Table 1 and Figure 2). Affinity of SBA lectin for compound 2 was calculated to be around 15-fold higher than that for glycopeptide 3. In addition, binding affinities between B

dx.doi.org/10.1021/cb500855x | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Table 1. Thermodynamic Binding Parameters Obtained by ITC for SBA Lectin and Glycopeptides 1−3 at 25 °C and pH 7.5 KD (μM) 1 2 3

≥800 6.6 ± 1.9 100.7 ± 20.1

ΔG (kcal mol−1)

ΔH (kcal mol−1)

−7.1 ± 0.3 −5.5 ± 0.2

−20.6 ± 2.9 −8.7 ± 1.6

TΔS (kcal mol−1)

n

−13.5 ± 3.2 −3.3 ± 1.8

1 0.9 ± 0.3 1.02 ± 0.2

Scheme 1. Synthesis of Glycopeptides m1−m3 Using SPPSa

a

HBTU = O-(1H-benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate. DIPEA = N,N-diisopropylethylamine. TFA = trifluoroacetic acid. TIS = triisopropylsilane.

residues.31 Next, the proton−proton distances with conformational information deduced from 2D-NOESY experiments (Figure 3) were used as restraints in MD simulations. In all MD simulations, only the set of signals corresponding to the trans disposition of the amide bond of Pro was considered. Due to the presumable flexibility of the glycopeptides, we used MD simulations with time-averaged restraints (MD-tar). These simulations have been successfully applied to flexible systems32−37 and provide a distribution of low-energy conformers able to quantitatively reproduce the NMR data. A good agreement between experimental and theoretically derived distances was obtained for the three glycopeptides (Supporting Information Tables S1−S3). Remarkably, while glycopeptide 2 is rather rigid in solution, compounds 1 and 3 are flexible (Figure 3). Compound 2 exhibits a main conformation (populated around 71%) characterized by an inverse γ-turn that comprises Pro2 and Asp3 residues. The conformer with an additional γ-turn between Asp3 and Thr4 residues, reported for other MUC1like glycopeptides in aqueous solution,24,31,38−40 is present in

about 29% of the total trajectory time (Supporting Information Figure S5). The main conformer of glycopeptide 2 appears to assist the presentation of the GalNAc residue to the target protein. No significant interactions between the side chain of the Asp3 and the peptide backbone or the GalNAc unit were found. Additionally, the electrostatic contact between Arg5 and Asp3 (distance Nε-Arg5/Cγ-Asp3 < 6.0 Å) appears ca. 30% of the total trajectory time. The geometry of the glycosidic linkage is fairly similar in the three derivatives. Actually, in all cases, ψ torsional angle takes values around 120°, characteristic of an eclipsed conformation.41 This geometry is supported by a key NOE cross-peak between the NH group of the threonine residue and the NH of GalNAc (Supporting Information Figure S6). As a next step, conformation of glycopeptides 1−3 complexed to SBA lectin was studied. We obtained the X-ray structure of the SBA:2 complex (Figure 4 and Table S4 in the Supporting Information), which is the first time that a structure between SBA lectin and a Tn antigen derivative has been solved and described. Of note, 12 complexes form the cell unit with C

dx.doi.org/10.1021/cb500855x | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 3. 2D-NOESY (H2O/D2O, 9:1, pH = 6.5, 293 K) spectra of glycopeptides 1 (left), 2 (middle), and 3 (right), together with the ensembles obtained from 20 ns MDH2O-tar simulations for these derivatives. Diagonal peaks and exchange cross-peaks connecting NH protons and water are negative (red color). The NOE contacts are represented as positive cross-peaks (blue color). The root-mean-squared deviation (rmsd) values for the peptide backbone, together with the population (%) and the peptide geometry of the main conformation in solution, is also shown.

two different binding modes for glycopeptide 2 complexed to SBA lectin (Figure 4, binding modes A and B). The electronic density for Ala1 was not observed in any of the complexes, indicating that this residue is exposed to the solvent and shows high flexibility due to the lack of interactions with the lectin. Although the 3D dispositions of glycopeptide 2 in binding modes A and B slightly differ in the geometry of their glycosidic linkage, the major changes are associated with the peptide backbone conformation, particularly with Thr4. As in solution, the ψ torsional angle of the glycosidic linkage of glycopeptide 2 adopts the expected eclipsed conformation in the bound state41 in both binding modes, with values around 120°. Accordingly, the sugar moiety has an almost perpendicular arrangement with respect to the peptide. In contrast, while Thr4 adopts a folded conformation in binding mode A, it displays an extended geometry in mode B. Nevertheless, the overall shape of both conformations of glycopeptide 2 bound to the lectin is rather similar to a root-mean-squared deviation (rmsd) value for the heavy atoms of 0.94 Å (Supporting Information Figure S7). Interestingly, these two conformations for Thr4 of glycopeptide 2 are also found in solution (Figure 3 and Supporting Information Figure S8). Analysis of the X-ray structures reveals

that the interaction pattern between glycopeptide 2 in binding mode A differs from that observed in binding mode B. In mode A, the hydroxyl groups of the GalNAc unit interact with the lectin through different hydrogen bonds. In particular, O3 and O4 form hydrogen bonds with the side chain of Asp88. Additionally, O3 participates in a hydrogen bond with Asn130 and O4 with Leu214. The GalNAc unit establishes an extra hydrogen bond between O6 and Asp215. The carbonyl group of the N-acetyl moiety of GalNAc interacts with the NH of Gly106 (Figure 4). This latter interaction may explain the higher affinity of SBA lectin for GalNAc in comparison to that for Gal.42,43 A CH···π contact between the α-face of the sugar and Phe128 is also established (Figure 4). It is important to point out that the calculated interactions between the sugar and the lectin by Qasba and co-workers are in agreement with those observed in our crystal structure.44 Regarding the interactions between the peptide and the lectin, the methyl group of Thr4 is involved in a hydrophobic pocket with Phe128. This contact may be responsible for the higher affinity observed for SBA lectin for Tn antigen when the underlying amino acid is a threonine residue.20 Furthermore, Arg5 participates in a salt bridge with Glu113 and a hydrogen bond with the hydroxyl of D

dx.doi.org/10.1021/cb500855x | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 4. X-ray structure of SBA lectin in complex with compound 2. Two binding modes for glycopeptide 2 were observed. The experimental electron density map obtained for glycopeptide 2 (in both binding modes) is shown, together with the geometry of the peptide backbone and the glycosidic linkage. A schematic representation of the hydrogen-bond network between the SBA lectin and glycopeptide 2 is shown, highlighting the differences between the two binding modes. The hydrophobic contact between the methyl group of Thr and Phe128 is shown in yellow. A CH···π interaction between GalNAc and Phe128 is also present in the X-ray structure. Two different molecular representations for each binding model are shown on the right.

The analysis indicates that the main epitope of the glycopeptide is located at the sugar moiety, with strong STD signals for H2, H3, H4, and the N-acetyl group of GalNAc. Additionally, these experiments show a clear interaction between the lectin and the peptide backbone. Hence, the methyl group of Thr4, ring of Pro2, and side chain of Arg5 are in closer contact with the protein. The STD experiments are compatible with the two binding modes observed in the X-ray structure for glycopeptide 2. The weak STD signal detected for the methyl group of Ala1 (