Glycosylation of Erythrocyte Spectrin and Its Modification in Visceral Leishmaniasis

Using a lectin, Achatinin-H, having preferential specificity for glycoproteins with terminal 9-O-acetyl sialic acid derivatives linked in α2-6 linkages to subterminal N-acetylgalactosamine, eight distinct disease-associated 9-O-acetylated sialoglycoproteins was purified from erythrocytes of visceral leishmaniaisis (VL) patients (RBCVL). Analyses of tryptic fragments by mass spectrometry led to the identification of two high-molecular weight 9-O-acetylated sialoglycoproteins as human erythrocytic α- and β-spectrin. Total spectrin purified from erythrocytes of VL patients (spectrinVL) was reactive with Achatinin-H. Interestingly, along with two high molecular weight bands corresponding to α- and β-spectrin another low molecular weight 60 kDa band was observed. Total spectrin was also purified from normal human erythrocytes (spectrinN) and insignificant binding with Achatinin-H was demonstrated. Additionally, this 60 kDa fragment was totally absent in spectrinN. Although the presence of both N- and O-glycosylations was found both in spectrinN and spectrinVL, enhanced sialylation was predominantly induced in spectrinVL. Sialic acids accounted for approximately 1.25 kDa mass of the 60 kDa polypeptide. The demonstration of a few identified sialylated tryptic fragments of α- and β-spectrinVL confirmed the presence of terminal sialic acids. Molecular modelling studies of spectrin suggest that a sugar moiety can fit into the potential glycosylation sites. Interestingly, highly sialylated spectrinVL showed decreased binding with spectrin-depleted inside-out membrane vesicles of normal erythrocytes compared to spectrinN suggesting functional abnormality. Taken together this is the first report of glycosylated eythrocytic spectrin in normal erythrocytes and its enhanced sialylation in RBCVL. The enhanced sialylation of this cytoskeleton protein is possibly related to the fragmentation of spectrinVL as evidenced by the presence of an additional 60 kDa fragment, absent in spectrinN which possibly affects the biology of RBCVL linked to both severe distortion of erythrocyte development and impairment of erythrocyte membrane integrity and may provide an explanation for their sensitivity to hemolysis and anemia in VL patients.


Introduction
The erythrocyte membrane is supported by a well-structured cytoskeleton. This cytoskeleton comprises of a network of different proteins maintaining the structural integrity and rigidity of the red blood cell (RBC) and of the RBC membrane [1]. Spectrin is a major cytoskeletal protein present as tetramers of aand bsubunits associated with other cytoskeletal proteins forming a lattice that governs erythrocyte membrane properties. Alterations of spectrin have been associated with several congenital anomalies like hereditary hemolytic anemia and hereditary elliptocytosis leading to cellular distortion [2]. Biochemical modifications of spectrin, mainly glycation and oxidation, have been observed in diabetes mellitus indicating erythrocyte membrane changes [3][4].
VL caused by the intracellular kinetoplastid protozoa L. donovani accounts for an estimated 12 million infected humans with an incidence of 0.5 million cases per year [27][28]. Approximately 50% of the world's VL cases occur in the Indian subcontinent.
Along with other signature manifestations, VL is almost always associated with anemia [17][18]. However alteration of the RBC membrane architecture as one of the causes leading to anemia remains poorly understood.
We have detected the exclusive presence of eight distinct disease-associated 9-O-acetylated sialoglycoproteins (9-O-AcSGPs) on RBC VL [17], using the preferential specificity of a snail lectin, Achatinin-H for glycoproteins with terminal 9-O-acetyl sialic acid (9-O-AcSA) derivatives linked in a2-6 linkages to subterminal Nacetylgalactosamine (GalNAc) [29]. Interestingly, normal erythrocytes (RBC N ) are devoid of such 9-O-AcSGPs. Antibodies directed against O-acetylated sialic acids have also been demonstrated in VL [30,15]. Moreover enhanced pattern of altered sialylation demonstrated a direct correlation with the degree of complementmediated hemolysis of RBC VL providing a plausible basis for anemia associated with VL [18]. Taking into consideration the involvement of 9-O-AcSGPs in VL erythrocyte pathology, we report the presence, purification and identification of sialylation, Nand O-glycosylation of two high molecular weight O-acetylated sialoglycoproteins as human eythrocytic a and b-spectrin by analysis of the tryptic fragments using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF)/post source decay (PSD) mass spectrometry (MS). Purified of spectrin VL showed the presence of an additional 60 kDa band which is completely absent in spectrin N purified by the same procedure. Additionally we have demonstrated glycosylation of spectrin N purified from normal erythrocytes (RBC N ). Although the presence of Nand O-glycosylations was found both in spectrin N and spectrin VL , enhanced sialylation was largely induced only in spectrin VL . Controlled tryptic fragments of aand b-spectrin VL exhibited the presence of terminal linkage specific sialic acids. This enhanced sialylation is possibly related to the fragmentation of spectrin in VL as evidenced by an additional 60 kDa fragment in spectrin VL and totally absent in spectrin N . In summary, this is the first report of glycosylation in eythrocytic spectrin N and its modifications in diseased condition which possibly affects the biology of RBC VL and may provide an explanation for their sensitivity to hemolysis and anemia in VL patients.

Results
Eight distinct disease-associated 9-O-AcSGPs exclusively induced on RBC VL Disease-associated 9-O-AcSGPs were purified from RBC VL of clinically confirmed untreated VL patients (n = 30) using Achatinin-H-Sepharose 4B affinity matrix (Fig. 1A). Total membrane protein (1.2060.187 mg) obtained from 2610 10 RBC VL yielded 0.43260.025 mg of purified 9-O-AcSGPs separated as eight distinct bands on SDS-PAGE. Purification of the same from RBC N (2610 10 ) of normal healthy individuals (n = 30) by the same procedure yielded undetectable amount of protein.

Identification of spectrin by mass spectrometry
Molecular identification of VL-associated proteins is a key to their significance in the disease pathology. With this aim, peptide mass fingerprint (PMF) analysis and sequencing of tryptic fragments of the two high molecular weight bands of 9-O-AcSGPs by MALDI-TOF/PSD-MS led to the identification of erythrocytic aand b-spectrin with sequences coverage of 33.6% and 22.7% respectively ( Fig. 1B-C). These tryptic fragments were mapped on to the NCBI database sequences of human erythrocytic aand bspectrin sequences gi: 119573202 and gi: 67782321 respectively (shown in red color in Fig. S1).
Spectrin VL is specifically a 9-O-AcSGP present only in RBC VL The identification of erythrocytic spectrin by mass spectrometry as O-acetylated sialoglycoprotein, prompted us to explore the status of total spectrin in RBC VL (spectrin VL ). Accordingly, spectrins were separately purified by the method as described by Ungewickell et al. [31]. The yield of spectrin VL purified from ghost membrane (1.48960.064 mg) of RBC VL was 0.23160.017 mg. Purified spectrin from RBC N (spectrin N ) demonstrated only two bands corresponding to a-, b-spectrin on SDS-PAGE ( Fig. 2A, lane 1). In contrast, purified spectrin VL exhibited an additional 60 kDa fragment along with a-, b-spectrin bands ( Fig. 2A, lane 2). Purified spectrin VL was further allowed to bind with Achatinin-H-Sepharose 4B. Achatinin-H bound spectrin VL demonstrated three similar bands ( Fig. 2A, lane 3). These observations confirmed the presence of 9-O-AcSA on a-, b-spectrin and 60 kDa band of RBC VL , which was indicative of alteration of spectrin in VL.
Western blot analysis of purified spectrin VL separately by the method as described by Ungewickell et al [31] also showed reactivity with Achatinin-H reconfirming the identity of all three bands as spectrin containing 9-O-AcSA (Fig. 2B). In contrast, similar analysis of spectrin N demonstrated the absence of all these three bands suggesting lack or undetectable 9-O-AcSA on RBC N .
SDS-PAGE analysis (5 and 7.5%) of spectrin N and spectrin VL demonstrated slight variation in their electrophoretic mobility ( Fig. 2C) suggesting some changes in VL. Two dimensional gel electrophoresis of purified spectrin VL reveals that individual spots corresponding to a-spectrin and b-spectrin and 60 kDa band have multiple isoelectric points (pI) (Fig. 2D) suggesting microheterogeneity of each spot possibly due to the differential sialylation of the same protein resulting into different pI.

Identification of 60 kDa band of erythrocytic a-spectrin in RBC VL
The 60 kDa band was identified as a fragment of a-spectrin (gi: 119573202) by mass-spectrometric PMF analysis (Fig. 3A) as well as sequence determination of fragments from PSD spectra. Representative MS/MS spectra of tryptic fragments of m/ z = 1237.6 ( Fig. 3B) and m/z = 1709.8 ( Fig. 3C) are shown. The 24 detected and annotated tryptic fragments matched the Nterminal section of a-spectrin with sequence coverage of 18.4% for the entire spectrin sequence and of 36.1% for the N-terminal exons (Table 1).

Erythrocytic spectrin N and spectrin VL are glycosylated
The identification of sialic acids in spectrin VL prompted us to explore the status of glycosylation of spectrin in RBC N . The presence of comparable Nand O-glycosylation was demonstrated by a shift in the respective protein bands corresponding to aand b-spectrin following enzymatic deglycosylation of neuraminidase treated purified spectrin N and spectrin VL due to their reduced molecular mass (Fig. 4A).
In parallel, neuraminidase treated 60 kDa fragment was also exposed to Nand O-glycosidase F which evidenced a shift of the band corresponding to a reduction of molecular weight by ,13.12 kDa and ,3.12 kDa indicating presence of both Nand O-glycosydic bonds (Fig. 4B). The 60 kDa fragment also demonstrated a shift of ,1.25 kDa after desialylation. Hence, deglycosylation accounted for about ,16.24 kDa of the total mass of 60 kDa.
The existence of both Nand O-glycosylation was further confirmed by binding with Sepharose/agarose bound specific lectins using iodinated spectrin VL and spectrin N (Fig. 4C-D). The binding of 125 I-spectrin N with immobilized Concanavalin A (ConA), Ricinus communis agglutinin (RCA), Helix pomatia agglutinin (HPA) and Ulex europaeus agglutinin (UEA) clearly suggested the existence of Nglycosylation. Similarly the binding of Dolichos biflorus agglutinin (DBA) and Jacalin reflected the presence of O-glycosylation in 125 Ispectrin N . These lectins also showed affinity towards the Nand Oglycosylated sugars present in 125 I-spectrin VL .
Immobilized ConA and UEA showed comparable binding with 125 I-spectrin N and 125 I-spectrin VL suggesting equivalent glycosylation levels having a-Man (mannose), a-Glc (glucose) and a-L-Fuc (fucose) (Fig. 4C). However immobilized RCA, HPA, DBA and Jacalin showed higher binding towards 125 I-spectrin N as compared to 125 I-spectrin VL suggesting the presence of more terminal b-D-Gal (galactose) (GalNAc, b-Gal), a/b-D-GalNAc, a-GalNAc and b1-3GalNAc sugars in spectrin N than spectrin VL .

Erythrocytic spectrin VL is highly sialylated
Isoelectric focusing (IEF) of spectrin VL demonstrated four distinct bands within a pI range of 4.6-5.21 (Fig. 5A), which showed a considerable shift of their pI to a range of 6.25-7.95 after neuraminidase treatment indicating the presence of sialic acids. Furthermore the homogeneous shifts of the individual bands demonstrated the homogeneity of the proteins. In contrast shift in pI of spectrin N before and after neuraminidase treatment was less Enhanced presence of SA in spectrin VL as detected by biochemical and glycoanalytical methods The presence of total sialic acid was demonstrated in spectrin VL/N using DIG-glycan detection kit ( Fig. 5B-C). Spectrin VL showed enhanced (3-fold) sialylation as compared to spectrin N . The variable expression of linkage-specific sialic acids was demonstrated (Fig. 5D-E). Spectrin VL showed ,2.7-2.8 fold enhanced binding with Sambucus nigra agglutinin (SNA) and Maackia amurensis agglutinin (MAA) compared to spectrin N .
Sialylation were further demonstrated by binding with iodinated spectrin VL and spectrin N with Sepharose/agarose bound Wheat germ agglutinin (WGA), SNA and MAA. Spectrin VL showed significantly higher binding (,2.4 fold) with all three lectins than spectrin N (Fig. 5F). Achatinin-H also showed higher binding with spectrin VL whereas negligible with spectrin N (Fig. 5F).
With an attempt to search for the presence of sialylated tryptic fragments, the aand b-subunits of spectrin VL were partially digested with trypsin separately. This controlled digestion yielded many fragments though we might have missed many smaller fractions (Fig. 5G). Theses fragments were allowed to bind with SNA and MAA agarose separately. SNA and MAA bound fragments were analysed on SDS-PAGE. Approximately 25 fragments of a-spectrin VL showed a2,6 linked and 18 of them had a2,3 linked terminal sialic acid. b-spectrin VL showed comparatively less number of a2,6 and a2,3 linked terminal sialic acids containing fragments (Fig. 5H).
Glycosidically bound sialic acids (SA) liberated from spectrin VL when separated on a TLC (thin layer chromatography) plate  [31]. Purified spectrin VL was further passed through an Achatinin-H-Sepharose 4B affinity column and 9-O-acetylated sialic acid containing spectrin VL (2.0 mg, lane 3) was purified as described in Materials and Methods. Lane M shows molecular weight standards. B. Presence of 9-O-AcSA as detected by Western blot analysis. Equal amounts (2 mg) of purified spectrin VL and spectrin N were transferred onto nitrocellulose membrane after SDS-PAGE (8.5%). The blots were incubated overnight at 4uC with Achatinin-H and processed as described in Materials and Methods. C. Equal amount (2 mg) of purified spectrin VL and spectrin N were separated both on 5 and 7.5% SDS-PAGE under similar conditions. D. Two dimensional (2D) gel electrophoresis of spectrin VL . A representative 2D (pI range 4-7, 4-15% gradient) profile of purified spectrin (100 mg) from RBC VL after staining with Coomassie is shown. doi:10.1371/journal.pone.0028169.g002 demonstrated the presence of Neu5Ac (N-acetyl neuraminic acid) and Neu5,9Ac 2 (5,9-diacetyl neuraminic acid) as compared with standard Neu5Ac and free SA purified from bovine submandibular mucin (BSM) (Fig. 6A). The presence of these derivatives was also demonstrated in the chromatogram of the liberated SA from spectrin VL by fluorimetric high-performance liquid chromatography (HPLC) (Fig. 6B). The Neu5,9Ac 2 peak of spectrin VL completely disappeared on saponification. In contrast, spectrin N showed undetectable Neu5,9Ac 2 suggesting disease-associated modification. BSM-derived SA showing ,40% Neu5,9Ac 2 was used as standard. Each fraction corresponding to Neu5Ac and Neu5,9Ac 2 of spectrin VL was collected after fluorimetric-HPLC and was subsequently confirmed by MALDI-TOF-MS which yielded their expected molecular ion signals having m/z at 448.1 (Fig. 6C) and 490.1 (Fig. 6D) respectively.

Molecular modelling of glycosylated residues
In a-spectrin, four potential N-glycosylation sites were identified all of which contains the consensus sequence, Asn-Xaa-Ser/Thr (Table 2). However, only one potential O-glycosylation site was found at Thr-817 position with a score above the threshold value (0.35). However, in b-spectrin, two potential N-glycosylation sites were found and no potential O-glycosylation site ( Table 2).
Structural verification of the predicted models of the modules containing Nand Olinked glycosylation sites revealed that the backbone conformations were satisfactory as the allowed phi-psi combinations were above 90% in the allowed region of Ramachandran's plot. Verify 3-D results showed that the models have an average 85% of the residues with 3D-1D score greater than 0.2 which indicates good quality 3D structural parameters. From ERRAT analysis it was observed that most of the models have an overall quality factor greater than 90% indicating good structural quality.
Solvent accessible surface areas was calculated by ACCESS and the probable glycosylation sites shows that the residues are exposed enough needed for glycosylation at the sites (Table 2). Further, modelling of a representative sugar (b-GlcNAc) into one each of Nand Oglycosylation sites showed that the sugar moiety can go into the available space around the amino acid residues (Fig. 7).

Spectrin N and spectrin VL showed slight variations in their secondary structures
The CD (Circular Dichroism) spectra of spectrin N in far-UV (ultra violet) region showed that protein contains 51.71% of ahelix, 9.17% of b-sheet and 39.12% of random coil (Fig. 8A). The similar trend was observed in the secondary structure prediction using GOR 4 [32]. The sequences of aand b-spectrin N was taken as weighted averages; the values of a-helix, b-sheet and random coil are 71.61%, 4.98% and 23.28% respectively. In parallel, the values for a-helix, b-sheet and random coil are 77.9%, 3.03% and 18.71% respectively as predicted from the modelled structures using MODELYN [33].
However, spectrin VL demonstrated a slight increase of a-helicity (63.05%) and a minimal decrease of b-sheet (5.68%) structure suggesting higher degree of sialylation possibly playing a role for such minute changes in its secondary structure.
Binding of 125 I-spectrin N and 125 I-spectrin VL to spectrin-depleted inside-out membrane vesicles (IOV) In order to further demonstrate the modified structure of spectrin VL in comparison to spectrin N we have compared the binding status of iodinated spectrin VL and spectrin N with spectrindepleted IOV from normal RBC-ghost (IOV N ). The binding of 125 I-spectrin VL to spectrin-depleted IOV N increases with increasing amount of 125 I-spectrin VL (Fig. 8B). In contrast, under identical condition, 125 I-spectrin N showed much higher binding towards spectrin-depleted IOV N . Such differences in binding signifies that minute structural modifications due to enhanced sialylation in spectrin VL possibly make it less available for interacting with other associated proteins in the spectrin-depleted IOV N of RBC N .

Discussion
VL is often complicated by anemia. The exclusive presence of 9-O-AcSGPs on erythrocytes of active VL has been correlated to RBC hemolysis [18,20]. The functional attributes of erythrocytes The tryptic fragments matched the N-terminal portion of human erythrocytic a spectrin as compared to the protein sequences of the NCBI sequence database. The identification was confirmed by complete de novo sequencing of two fragments (shown in Figure 3 B-C). Mass [M+H] + denotes the mono-isotopic masses of the fragment ions; sequence range refers to the alignment of the sequence of the denoted fragments with the a-spectrin reference sequence (gi: 119573202); deviation from theoretical mass is the mass difference between the measured mass and the mass calculated from the corresponding database sequence; missed cleavage refers to the missed trypsin cleavage sites in the identified fragment; sequence is the fragment sequence in one-letter code, Mox is oxidized methionine. doi:10.1371/journal.pone.0028169.t001 are determined by the structural integrity of the membrane, which is often described in terms of alterations in the membrane characteristics like osmotic fragility, fluidity and hydrophobicity. Any kind of perturbation in the milieu of the erythrocytes like oxidative changes or ligand specific interaction culminates in changes in the membrane characters generally associated with pathological conditions [25]. Therefore, we considered it worthwhile to unravel the molecular determinants and implications of 9-O-AcSGPs on RBC VL . The major observation of this study is the demonstration of glycosylation in normal spectrin purified from RBC N , presence of higher degree of sialylation in spectrin purified from RBC VL and fragmentation of spectrin VL as a 60 kDa 9-O-AcSGP. Therefore, it may be envisaged that enhanced sialylation of spectrin VL is possibly responsible for the generation of this fragmented Oacetylated sialic acid-containing spectrin VL . Altered binding of highly sialylated spectrin VL with spectrin-depleted inside-out membrane vesicles of RBC N possibly suggested functional abnormality. Membrane characteristics of RBC VL were observed by enhanced hydrophobicity, fragility, fluidity as compared to RBC N hinting towards membrane damage [25].
We have purified eight distinct 9-O-AcSGPs from RBC VL using Achatinin-H as an affinity matrix indicating linkage specific terminal 9-O-AcSA in these sialoglycoproteins. Distinct multiple spots of individual 9-O-AcSGP suggested microheterogeneity possibly due to differential sialylation. As 9-O-AcSGPs are exclusively present on RBC VL , their identification through mass spectrometry was necessary to assess their possible implication in the disease pathology.
The analysis of two high molecular weight 9-O-AcSGPs by MALDI-TOF-MS evidenced a match with the NCBI entry of human erythrocytic a and b-spectrin with sequence coverage of 33.6% and 22.7% respectively. The amino acid sequences of tryptic fragments deduced from MS analysis confirmed the identification. Sequencing of two tryptic fragments and database-dependent Mascot as well as database-independent Sequit analyses made the identification unambiguous.
The cytoskeleton beneath the lipid bilayer of the membrane of RBC comprises of several proteins interconnected with each other providing stability and integrity to the membrane structure. Spectrin exists as a heterotetramer consisting of two subunits each of a-(280 kDa) and b-(246 kDa) spectrin oriented in an anti- Figure 5. Presence of Neu5Ac and Neu5,9Ac 2 in spectrin VL by biochemical methods. A. Enhanced sialylation demonstrated by IEF. Equal amounts (3.0 mg) of purified spectrin VL , 60 kDa band and spectrin N before and after removal of sialic acids were analyzed by IEF within a pH gradient of 3-10 and the respective bands visualized by silver staining. Lane M shows the pI markers. B-C. Enhanced sialylation in spectrin VL . Equal amount (1.0 mg) of purified spectrin VL and spectrin N was analyzed by using DIG-glycan detection kits and total sialylation was compared based on the densitometric scores of spots (B). Representative bar graph of densitometric scores of corresponding spots (C). D-E. Detection of linkage-specific terminal sialic acids in spectrin VL . Equal amount (2.0 mg) of spectrin VL and spectrin N was dot blotted on NC-paper and analyzed by DIG glycan and differentiation kit using SNA and MAA lectins following manufacturer's protocol (D). Densitometric scores of corresponding spots are shown as bar graph (E). F. Binding of 125 I-spectrin VL/N with several sialic acid binding lectins. To demonstrate the presence or absence of terminal sialic acids, a fixed concentrations of 125 I-spectrin VL/N were analyzed by binding with Sepharose/agarose bound WGA, SNA, MAA, Achatinin-H (25 ml bead volume) having specificity towards linkage specific sialic acids as described in materials and methods. Bound radioactivity of 125 I-spectrin VL/N was measured by Gamma-counter and represented as bar graphs. G-H. Detection of sialylated tryptic fragments in spectrin VL . The a and b subunits of purified spectrin VL were digested separately by restricted amount of trypsin. Such controlled digested and extracted tryptic fragments were dried and redissolved and an aliquot was separated in SDS-PAGE (7.5%-15% gradient) (G). Subsequently the presence of sialic acids on resulting tryptic fragments was analyzed by binding with SNA-agarose and MAA-agarose separately and followed by electrophoresis on SDS-PAGE (7.5%-15% gradient) (H) as described in Materials and Methods. Lane M shows the molecular weight standerds. doi:10.1371/journal.pone.0028169.g005 parallel arrangement. The presence of both Nand O-glycosylation was indicated by shifts in molecular mass after the respective glycosidase treatments of neuraminidase-treated spectrin VL and spectrin N . Binding with several lectins specific for Nand Oglycosylation also supported the presence of such glycosylation in spectrin VL and spectrin N .
Molecular modelling studies also supported both N-& Oglycosylations of a-specrin. However, only N-glycosylation was found in b-spectrin. Modelling the sugar moiety to the predicted glycosylation sites suggested that glycans could fit into these sites without any steric clashes, thus signifying the probability of glycosylation of spectrin.
Cell surface sialic acids have been widely associated with different pathological conditions. Enhanced presence of sialic acids in spectrin VL has been convincingly exhibited by lectin binding, which was further confirmed by TLC, fluorimetric-HPLC and MALDI-TOF-MS. Demonstration of pI of spectrin VL in acidic region and a huge shift of pI after neuraminidase treatment Figure 6. Presence of Neu5Ac and Neu5,9Ac 2 in spectrin VL by analytical methods. A. Thin layer chromatography (TLC). Glycosidically bound sialic acids of spectrin VL were subjected to acid hydrolysis, purified, separated on a TLC plate and detected by staining with orcinol/HCl spray reagent and baking at 180uC. Similarly processed free sialic acids released from BSM served as standard. Additionally, commercially available Neu5Ac was used as references. For comparison liberated sialic acids from purified spectrin N were similarly analyzed. B. Enhanced presence of Neu5Ac and Neu5,9Ac 2 in spectrin VL as determined by fluorimetric HPLC. Glycosidically bound sialic acids released from spectrin VL by acid hydrolysis were derivatized with DMB and analyzed by fluorimetric HPLC before and after saponification as described in Materials and Methods. A representative chromatogram of the spectrin VL and spectrin N derived sialic acids showed the presence of fluorescent derivatives of free sialic acids. In parallel sialic acids of BSM similarly analyzed under identical conditions served as standard. C-D. Identification of sialic acids by MALDI-TOF MS. Fractions corresponding to peaks of Neu5Ac (C) and Neu5,9Ac 2 (D) were collected after fluorimetric HPLC, spotted and analyzed by MALDI-TOF MS using DHBA matrix as described in Materials and Methods. Positive ion mode was used for mass-spectrometric analysis with 1000 laser shots per spot. doi:10.1371/journal.pone.0028169.g006 established enhanced sialylation compared to spectrin N . More importantly, exclusive presence of Neu5,9Ac 2 in spectrin VL suggested disease-associated enhanced sialylation in VL. Enhanced sialylation in spectrin VL compared to spectrin N possibly causes structural modification of spectrin VL . Such structural changes were perhaps the basis for the reduced capacity of spectrin VL to complex with other associated cytoskeletal proteins in normal environment as demonstrated by its less binding with spectrin-depleted IOV N .
Interestingly, purified 60 kDa fragment demonstrated the presence of two distinct bands in IEF and each of the bands depicted a distinct shift in their pI after removal of SA showing the presence of two sialylated proteins of similar molecular mass. In contrast spectrin N having comparable glycosylation, showed complete absence of such fragmentation, which suggested that alteration of spectrin mainly enhanced sialylation may be associated with VL pathology.
The cleaved 60 kDa fragment which belongs to a-spectrin contains two potential N-linked glycosylation sites, at Asn-633 & Asn-657 and one O-linked glycosylation site at position Thr-817 with sufficient surface accessibility. The remaining portion of aspectrin although contains two potential N-linked glycosylation sites, but no O-linked glycosylation site was found. On the other hand, we could not identify any potential O-linked glycosylation site in the b-spectrin. Therefore, it may be envisaged that the exclusive presence of O-linked glycosylation site in the N-terminal region of a-spectrin with high surface accessibility tends to have higher sialylation for interaction with each other. All these factors combined may play important role in the cleavage of a-spectrin to 60 kDa fragment in VL.
Production of erythrocytes requires synthesis of red cell proteins specially cytoskeleton proteins. During terminal differentiation of erythroid progenitor cells in culture it retains the key components of the endoplasmic reticulum protein translocation, glycosylation,  and protein folding machinery, chaperones, calreticulin and Hsp90 for red cell glycoprotein biosynthesis [34]. Non-enzymatic glycation and oxidation of spectrin were reported under several physiological conditions [3]. Such nonenzymatic changes are a result of glyco-oxidation, where the oxidative stress within, surpasses the antioxidant defense system of the cell [35]. The associated biochemical alterations affect the structure, aggregation and integrity of the membrane and membrane-associated proteins. Such changes have been witnessed in erythrocytic spectrin of subjects suffering from long-term diabetes mellitus. Here the elevated glucose concentration increases oxidation and advanced glycation end product formation of structural and membrane proteins of erythrocyte [36]. The demonstration of spectrin VL with enhanced sialylation in VL patients raised questions regarding the basis of these modifications. Therefore, it may be envisaged that oxidative modification of spectrin affects membrane morphology of the erythrocytes. Enhanced fragility, membrane fluidity and hydrophobicity of RBC VL as compared to RBC N were demonstrated earlier [25]. Hence, the evidence of altered spectrin reported here may provide an explanation for the known-impaired stability of erythrocytes in VL.
The presence of elevated levels of serum sialic acids in cardiovascular diseases and their relation to evaluated myocardial cell damage have been documented and it has been suggested that either the shedding or secretion of cell membrane sialic acids determines their accumulation in serum [37,38]. Furthermore, the importance of elevated serum sialic acids and soluble sialyltransferases in the diagnosis of Down-syndrome affected pregnancy and oral cavity cancer has been documented [39,40]. The presence of sialyltransferases in human serum may provide a possible way of changes in serum proteins with terminal a2-6 sialic acid [40]. Bulai et al. have characterized a transport system and demonstrated the uptake of free sialic acids into human erythrocytes [41]. Therefore, free sialic acids could be transported across the membrane into RBC through a sialic acids transport system. Interestingly, the presence of enhanced sialic acids in the serum of VL patients probably hinted towards a possible mechanism of transport of these free sialic acids under the influence of sialyltransferases. Hence presence of sialic acids in VL serum could essentially serve as a source for the erythrocyte sialic acids, which could use a transport system for their entry. The uptake of sialic acids was monitored by measuring free sialic acids and ManNAc produced by cytosolic sialate pyruvate-lyase in human erythrocytes that indicated the presence of a sialic acid transport system [42]. Furthermore, in VL, peripheral hematopoietic cells have increased sialic acids [26] that could be shaded in the serum and be transported across the erythrocyte membrane. The presence of serum sialyltransferases in VL patients and testing this hypothesis as well as the elucidation of the mechanisms of enhanced sialylation of spectrin in RBC VL demands extensive studies and will be a subject of future investigations.
Taken together the current study provides evidence for the first time not only for glycosylation of spectrin N but also enhanced sialylation in diseased condition i.e. spectrin in RBC VL . Additionally, we have demonstrated fragmented spectrin VL which could be triggered by such enhanced sialylation. Therefore, we hypothesize that the higher sialylation along with exclusive presence of 9-O-AcSA on RBC VL may in turn, cleaves spectrin, ultimately resulting in destabilization and functional inability of the RBC. From this entire study, we contend that these 9-O-AcSGPs trigger membrane damage and may serve as an important factor leading to anemiaassociated with VL. Hence the study successfully dissects one of the causal mechanisms leading to anemia, a common manifestation in VL.

Clinical samples
Blood sample of clinically confirmed active VL patients (n = 30; 21 males, 9 females, median age: 30 years) based on microscopic demonstration of Leishmania sp. amastigotes in splenic aspirates were collected from School of Tropical Medicine, Kolkata and immediately processed for the separation of RBC VL at Indian Institute of Chemical Biology. The diagnosis was validated by two in-house techniques, in which the increased presence of linkagespecific 9-O-AcSGPs was quantified by erythrocyte binding assay [17] and anti-9-O-AcSGPs antibodies were detected by enzymelinked immunosorbent assay (ELISA) using BSM known to contain a high percentage of 9-O-AcSAs, as coating antigen [30,23]. The hematological parameters evidenced anemia in these patients and ruled out any other blood cell disorder ( Table 3). Existence of high level of sialic acid in VL serum was observed.
Peripheral blood from normal human donors from endemic (n = 15) and non-endemic areas (n = 15) was processed similarly to obtain RBC N for the study. The Institutional Human Ethical Committee had approved the study and samples were taken with the consent of the donors, patients, or in case of minors from their parents/guardians. protease inhibitor cocktail, pH 7.2, sonicated (three pulses, 10 sec each) in ice-mixture and incubated at 4uC for 1 hr [43]. After centrifugation at 82006g, 4uC the supernatant was collected and dialyzed against Tris-HCl (0.05 M, pH 7.2) saline (TBS) containing 0.03 M Ca 2+ (TBS-Ca 2+ ), 0.01% (w/v) detergent (CHAPS: BOG 1:1), sodium azide (0.02%). The dialyzed protein was processed for affinity chromatography and the protein content was quantified by Lowry method [44].
RBC VL ghost membrane protein fraction (1.85 mg) was passed through Achatinin-H-Sepharose-4B affinity column (2.0 mg/ml) equilibrated with TBS-Ca 2+ containing sodium azide (0.02%) at 4uC as described elsewhere [18]. After extensive washing, Achatinin-H-bound 9-O-AcSGPs were eluted at 25uC with TBS containing sodium citrate (0.04 M, pH 7.2), dialyzed against TBS at 4uC and stored at 270uC for future use. As the binding of Achatinin-H towards 7-O-and/or 8-O-AcSA cannot be ruled out, therefore, presence of such linkages in O-acetylated sialoglycoproteins are also possible.

Protein/peptide mass spectrometry
The identification of the glycoprotein was done by mass spectrometry using Bruker-Daltonics MALDI-TOF mass spectrometer Reflex IV (Bruker Daltonics, Bremen, Germany). The samples were prepared by dried-droplet procedure using 2,5dihydroxybenzoic acid (DHBA) as matrix. Calibration was done externally with a mixture of Angiotensin I, Angiotensin II, Substance P, Bombesine, ACTH clip 1-17 and ACTH clip 18-39. Subsequently, peptide samples were used for analysis. Tryptic fragments were generated by overnight in-gel digestion of two high molecular bands in ammonium hydrocarbonate (5 mM) using 10 ng trypsin (Promega, Mannheim, Germany) per sample. Sequence analysis of selected tryptic fragments was done with an Ultraflex III MALDI-TOF-TOF mass spectrometer (Bruker Daltonics). PSD spectra were acquired using default LIFT method for MS/MS spectra acquisition with manually adjusted laser energy accumulating data from 1500-2000 laser shots. Spectra annotation was done using the FlexAnalysis 3.0 (Bruker Daltonics) software. PMF analyses and MS/MS ion searches were done with MASCOT (Matrix Science Ltd., London, UK). Database searches through Mascot with PMF and MS/MS data were done with the BioTools 3.1 software (Bruker Daltonics). For database searches the following parameters were used. Taxonomy: Homo sapiens; database: NCBI; enzyme: trypsin; variable modifications: oxidation on methionine and one missed cleavages. Database searches for PMF spectra were done at the fragment mass tolerance 60.3 Da. For the MS/MS searches mass tolerances for precursor was 60.2 Da and 0.4 Da for fragment masses were used. The identification of the fragments and thereby of the protein was confirmed by database-independent de novo sequencing using the Sequit! Software [45].

Purification of spectrin
The spectrin N from RBC N (1.25 mg total ghost membrane protein) was purified following the method of Ungewickell et al. with slight modifications [31]. Briefly, the ghosts were washed twice and resuspended in 3 volume of sodium phosphate (0.3 mM, pH 7.2) containing ethylene diamine tetraacetic acid (EDTA; 0.2 mM), (extraction buffer) and incubated for 20 min at 37uC. The fragmented ghosts were pelleted by centrifugation at 800006 g for 1 h at 2uC. Water-soluble proteins in the supernatant were immediately applied to a Sepharose 4B column (9062 cm), equilibrated with Tris (25 mM), EDTA (5 mM), NaCl (0.1 M), pH 7.6 (Tris/EDTA/saline buffer) at 4uC. The column was eluted at 10 ml/h and 4-ml fractions were collected. Protein in the effluent was monitored by absorbance at 280 nm. Fractions containing purified spectrin dimer were pooled, concentrated and dialyzed overnight at 4uC against TBS-Ca 2+ buffer containing sodium azide. In parallel, spectrin VL was similarly purified from Anti-9-O-AcSGP antibody was detected by using BSM as coating antigen as described elsewhere [30]. d Parasite specific antibody was detected by using parasite lysate as coating antigen as described elsewhere [16]. e Sialic acid content in serum was estimated by thiobarbituric acid method [54]. doi:10.1371/journal.pone.0028169.t003 RBC VL . Additionally, the purified total spectrin VL was further passed through an Achatinin-H-Sepharose 4B affinity column and 9-O-AcSA containing spectrin VL was purified as described above.
Purified spectrin N , spectrin VL or affinity-purified spectrin VL and gel eluted 60 kDa fragment and were analysed by SDS-PAGE (5 and 7.5%) in a minigel apparatus (Bio-Rad, USA) [25] and the gels were stained. Nor O-linked glycosylation of spectrin VL , spectrin N and 60 kDa fragment was demonstrated after deglycosylation with specific glycosidases using deglycosylation kit (Roche Applied Science, Mannheim, Germany) according to the manufacturer's protocol [26].
Western blot analysis [26] of spectrin VL and spectrin N was performed by semidry method at 15 V for 20 min. After blocking, the membranes were incubated with Achatinin-H (100 mg/ml) in the presence of Ca 2+ (0.03 M). Subsequently, the Achatinin-H probed membrane was incubated with polyclonal rabbit anti-Achatinin-H antibodies (1:400) at 4uC. Both the blots were developed using HRP-conjugated goat anti-rabbit IgG (1:5000, Cell signaling) and detected using diaminobenzidine (Sigma, St. Louis, MO) as substrate.
To obtain pure 60 kDa protein, Coomassie-stained bands corresponding to 60 kDa were excised from the electrophoresis gels and the proteins eluted using an Electro-Eluter Model 422 (Bio Rad, USA). IEF of purified spectrin VL , gel-eluted 60 kDa fragment and spectrin N was performed in capillary tubes within a pH range 3.0-10.0 using Mini-PROTEAN II tube cell apparatus (Bio-Rad, USA) and silver stained. The samples were desialylated overnight with Arthrobacter ureafaciens neuraminidase (0.2 mU/mg) at 37uC and processed similarly. The isoelectric points were determined from the pI of known proteins used as standards [26].

Analysis of carbohydrates
DIG-glycan detection. Equal amounts (1.0 mg) of spectrin VL and spectrin N were dot blotted on nitrocellulose paper (NC-paper) and total sialic acid content was analyzed by using DIG-glycan detection kit (Roche Applied Science, Mannheim, Germany) following manufacturer's protocol [8]. Densitometric measurement of spots was done by using ImageQuantTL software (GE Healthcare).
Immobilized lectin binding assays to 125 I-spectrin VL/N . To analyze the terminal sugar linkages, spectrin VL and spectrin N were separately iodinated with 125 I (Bhabha Atomic Research Centre, Mumbai, India) yielding specific activity of spectrin N 1.96610 6 cpm/ mg and of spectrin VL 1.8610 6 cpm/mg respectively. Fixed concentrations of 125 I-spectrin VL/N were incubated separately with Sepharose/agarose bound lectins (25 ml bead volume) of different sugar-linkage specificity. Con A (specific for a-Man and a-Glc), RCA (specific for b-D-Gal (GalNAc, b-Gal)), HPA (specific for a-or b-D-GalNAc), UEA (specific for a-L-Fuc) were used to demonstrate the presence of N-glycosylation. Similarly Jacalin (specific for b1,3GalNAc) and DBA (specific for a-GalNAc) were used to illustrate the presence of O-glycosylation. The lectins WGA (specific for GlcNAc and Neu5Ac), SNA, MAA, Achatinin-H (specific for 9-Oacetyl sialoglycosyl residue) were used to show the presence of sialic acids. Unbound radioactivity was removed using TBS-bovine serum albumin (1 mg/ml) and bound 125 I-spectrin VL/N was measured by Gamma-counter (Electronic Corporation, India) [47].
Identification of sialoglycopeptides from spectrin VL containing a2,6 and a2,3 linked sialic acids. Purified spectrin VL (150 mg) was run in 5% SDS-PAGE and a-spectrin & b-spectrin bands were digested partially by trypsin (200 ng) using in-gel trypsin digestion kit (Pierce, Rochford, USA) following the manufacturer's protocol. Digested and extracted tryptic fragments were dried under speed-vac and redissolved in 0.1% TFA (100 ml). An aliquot of the redissolved tryptic fragments were seperated on 7.5%-15% gradient SDS-gel. Remaining portion was neutralized by Tris-HCl (pH 8.0) and an aliquot was incubated separately with SNA-agarose (10 ml) and MAA-agarose (10 ml) for overnight at 4uC under mild shaking. The mixture was centrifuged at 5000 rpm for 10 min, supernatant collected as unbound fraction. Pellet was suspended in cold phosphate buffered saline (0.02 M, pH 7.0) and centrifuged at 5000 rpm for 10 min. Washed pellet was boiled with SDS-PAGE sample buffer, centrifuged at 5000 rpm for 5 min and the supernatant was ran in 7.5%-15% gradient gel.
MALDI-TOF MS. Each fraction corresponding to different forms of sialic acids was collected from fluorimetric HPLC and subsequently analyzed by MALDI-TOF-MS (Applied Biosystem, USA) using DHBA as matrix as described previously [8,26,48]. Positive ion mode was used for analysis. The acquired spectra were accumulations of 1000 laser shots.
3-D structural modelling of spectrin modules. 3D structure of the modules containing the potential glycosylation sites were modelled using Swiss Model software [50]. The quality of the models was validated using Structural Analysis and Verification Server (http://nihserver.mbi.ucla.edu/SAVES/). Solvent accessibility of probable glycosylation sites. Solvent accessibility surface area of all amino acids residues of the models were calculated using ACCESS software [51]. The probable Nand Olinked glycosylation sites were identified by their percentage of surface exposure. Sites falling within the identified segments of aand b-spectrin sequences by MASCOT program with significantly high intensity were eliminated from the list of probable sites as their nonglycosylated status was confirmed.
Attachment of carbohydrates to probable glycosylation sites. The structures of the modules attached to the carbohydrate at assigned Asn and Thr of the identified N-& Oglycosylation sites were optimized using molecular modelling software suite InsightII (2005) of Accelrys (San Diego, CA) by repeated energy minimization and molecular dynamics simulations with DISCOVER module. Energy minimization was performed alternatively with steepest descent and conjugate gradient methods (200 steps each using cff91 force field). Molecular dynamics simulation run was done with 10,000 steps of 1 fs after 1000 steps of equilibration with a conformation sampling of one in 100 steps at 300 K. At the end of the molecular dynamics simulation, the lowest potential energy conformation was picked using ANALYSIS module of Insight II for further energy minimization. The molecular dynamics simulation followed by energy minimization was performed on the glycosylation site residues attached with the sugar moiety while keeping the rest of the protein molecule fixed by applying positional constraints. This process was continued until satisfactory conformational parameters were achieved [26].
Physicochemical studies CD spectra of spectrin. Far-UV CD spectra (between 190 nm and 250 nm) measurement of equal amount (0.05 mg/ml) of spectrin VL and spectrin N were performed at 25uC on a JASCO J-715 spectropolarimeter using a quartz cuvette of path length 1 mm under continuous flush of nitrogen gas. The spectra shown are the average of ten data collected in continuous scan mode. The individual secondary structural contents of a-helix, b-sheet, and random coil were analyzed from the far-UV CD spectra using the K2D2 software [52].
To demonstrate the binding of spectrin N with spectrin-depleted-IOV N , 125 I-spectrin N (0-10 mg/ml) were incubated for 90 min at 0uC in a buffer A (100 ml) containing 20 mg/ml spectrin-depleted-IOV N protein [53] and centrifuged at 50,000 g for 25 min at 4uC. Membrane-bound 125 I-spectrin was washed with Buffer A and the radioactivity was counted by a Gamma-counter (Electronic Corporation, India). Nonspecific binding at each 125 I-spectrin N concentration was determined by the use of heat-denatured (70uC, 15 min) spectrin, and this value (10-28% of total counts) was routinely subtracted [53].
Estimation of sialic acid (SA) in serum. Estimation of total SA in serum was carried out colorimetrically by the thiobarbituric acid method after hydrolysis with 0.1 N sulfuric acid at 80uC for 1 hr [54]. The absolute value of sialic acid in serum was obtained from standard curve of authentic Neu5Ac.
Results are expressed as means 6 S.D for individual sets of data. Each experiment was performed at least 3 times. Figure S1 Sequence of a-spectrin and b-spectrin with the identified and annotated fragments in red and the sequenced fragments underlined. (TIF)