Cytotoxic protein from the mushroom Coprinuscomatus possesses a unique mode for glycanbinding and specificityPeilan Zhanga,1, Kunhua Lib,1,2, Guang Yanga, Changqing Xiac, Jane E. Polstond, Gengnan Lib, Shiwu Lic, Zhao Line,Li-jun Yangc, Steven D. Brunerb,3, and Yousong Dinga,3

aDepartment of Medicinal Chemistry, Center for Natural Products, Drug Discovery and Development, University of Florida, Gainesville, FL 32610;bDepartment of Chemistry, University of Florida, Gainesville, FL 32611; cDepartment of Pathology, Immunology and Laboratory Medicine, University ofFlorida, Gainesville, FL 32610; dDepartment of Plant Pathology, University of Florida, Gainesville, FL 32611; and eDepartment of Periodontics, VirginiaCommonwealth University, Richmond, VA 23298

Edited by Jerrold Meinwald, Cornell University, Ithaca, NY, and approved July 17, 2017 (received for review April 28, 2017)

Glycans possess significant chemical diversity; glycan binding pro-teins (GBPs) recognize specific glycans to translate their structures tofunctions in various physiological and pathological processes.Therefore, the discovery and characterization of novel GBPs andcharacterization of glycan–GBP interactions are significant to pro-vide potential targets for therapeutic intervention of many diseases.Here, we report the biochemical, functional, and structural charac-terization of a 130-amino-acid protein, Y3, from the mushroomCoprinus comatus. Biochemical studies of recombinant Y3 from ayeast expression system demonstrated the protein is a uniqueGBP. Additionally, we show that Y3 exhibits selective and potentcytotoxicity toward human T-cell leukemia Jurkat cells comparedwith a panel of cancer cell lines via inducing caspase-dependentapoptosis. Screening of a glycan array demonstrated GalNAcβ1–4(Fucα1–3)GlcNAc (LDNF) as a specific Y3-binding ligand. To providea structural basis for function, the crystal structure was solved to aresolution of 1.2 Å, revealing a single-domain αβα-sandwich motif.Two monomers were dimerized to form a large 10-stranded, anti-parallel β-sheet flanked by α-helices on each side, representing aunique oligomerization mode among GBPs. A large glycan bindingpocket extends into the dimeric interface, and docking of LDNFidentified key residues for glycan interactions. Disruption of resi-dues predicted to be involved in LDNF/Y3 interactions resulted inthe significant loss of binding to Jurkat T-cells and severely impairedtheir cytotoxicity. Collectively, these results demonstrate Y3 to be aGBP with selective cytotoxicity toward human T-cell leukemia cellsand indicate its potential use in cancer diagnosis and treatment.

glycan binding protein | Coprinus comatus | cytotoxicity | LDNF | crystalstructure

Polysaccharides (glycans) are one fundamental building blockof life, ubiquitously expressed in all organisms and essential

to numerous biological processes including adhesion and growth,signaling, infection, and tumor pathogenesis (1–3). In addition,aberrant glycosylation is directly linked with many human diseases(4). Of note, glycans serve as useful biomarkers of various cancersand targets of therapeutic intervention (4–7). Glycan bindingproteins (GBPs) read the diversity and complexity of glycans in arelatively specific manner and execute the physiological or path-ological information encoded by the polymers (1, 8). The speci-ficity of GBP–glycan interactions is determined by multiple factors,including composition, site-specific modifications, and tertiarystructure of the glycans. Lectins form one major group of GBPsand are widely distributed among organisms (e.g., viruses, bacteria,fungi, insects, plants, and animals) (9, 10). Lectins adopt at least14 different folds and common examples include the ricin-likeβ-trefoil, galectin-like fold, actinoporin-like fold, and β-propeller(11, 12). However, fold can show low sequence identities amongfamily members (13). Indeed, significant diversity in sequence andfolds highlights the divergent and convergent evolution of protein

functions and poses challenges to accurately and reliably annotatenew members (14, 15).A number of GBPs are small proteins, less than 150 amino

acids, and over the past decades the functions of many smallproteins have been discovered to be critical to various cellularprocesses, primarily through serendipitous studies (16). For ex-ample, galectin-1 with an affinity for β-galactose (Gal) was shownto be essential to neuronal cell differentiation and be associatedwith malignant tumor progression in human (17, 18). Small pro-teins also regulate essential cellular processes of bacteria (e.g., the43-aa SgrS) (19), yeast (20), and animals (16). In recent years,systems biology studies have demonstrated that organisms com-monly express hundreds of small proteins that often have nocharacterized homologs (16, 21–23). Accurately assigning a bio-chemical, cellular, or physiological function to these proteins is arapidly evolving research area whose advances require the in-tegration of multiple disciplines (16, 24).Mushrooms have been used for food, medicine, or other

purposes for thousands of years (25) but 90% of mushroomspecies in nature remain unexplored (26, 27). The potential ofthis untapped source in the discovery of useful substances is

Significance

Glycan binding proteins (GBPs) play an important and ever-emerging role in decoding the structural diversity of cell sur-face glycans into function. New GBPs provide useful tools toprobe and manipulate biological processes. Here we describethe characterization of the Y3 protein from the mushroomCoprinus comatus as a unique GBP that shows selective cyto-toxicity toward human T-cell leukemia Jurkat cells throughcaspase-associated apoptosis. Structural analysis along withglycan array screening of Y3 reveals a unique tertiary structureand a specific interaction with GalNAcβ1-4(Fucα1-3)GlcNAc, aglycan abundant in invertebrates but uncommon in humans.This work expands on promising novel GBPs available in less-explored sources for biomedical and research applications.

Author contributions: Y.D. designed research; P.Z., K.L., G.Y., C.X., J.E.P., G.L., S.L., Z.L.,and S.D.B. performed research; P.Z., K.L., C.X., J.E.P., Z.L., S.D.B., and Y.D. analyzed data;and P.Z., K.L., G.Y., J.E.P., L.-j.Y., S.D.B., and Y.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 5V6I and 5V6J).1P.Z. and K.L. contributed equally to this work.2Present address: Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA02215.

3To whom correspondence may be addressed. Email: bruner@chem.ufl.edu or YDing@cop.ufl.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1706894114/-/DCSupplemental.

8980–8985 | PNAS | August 22, 2017 | vol. 114 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1706894114

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0

exemplified by the semisynthetic analog of pleuromutilin (reta-pamulin) as clinically used antibiotic (28), illudins as anticancerdrugs (29), and peptidic omphalotins as nematicidal agents (30).Addition to low molecular weight (MW) secondary metabolites,mushrooms produce a variety of proteins, such as lectins, withantitumor, antiviral, antimicrobial, antioxidative, and/or immu-nomodulatory activities (31). Recent genomic and computa-tional studies have further identified an enormous number ofgenes encoding small proteins from fungal genomes (32). Thesesmall proteins awaiting assigned functions can become usefulbiotechnological and biomedical agents (31).The mushroom Coprinus comatus has an annual production of

about 0.4 million tons globally and shows a range of bioactivities(e.g., immunomodulation and anticancer) (33, 34). Y3 is a 130-aaprotein isolated from C. comatus and initial characterizationshowed it inhibits the infection and multiplication of the tobamo-virus, Tobacco mosaic virus (TMV) (35). Our search of publiclyavailable genome database yielded a limited number of Y3 ho-mologs with only low sequence similarity, and none has beencharacterized (Fig. 1A). Herein, we disclose the detailed bio-chemical, functional, and structural characterization of this fungalsmall protein. Our biochemical studies of recombinant matureY3 produced in a yeast expression system validated its anti-tobamovirus activity and revealed its glycan binding capability.Given the significance of glycans to cancer development and ap-optosis (4–6), we further demonstrated that Y3 selectively andpotently induced caspase-dependent apoptosis of human T-cellleukemia Jurkat cells. Glycan array screening identified a glycanantigen that Y3 showed strong specific interactions with. Finally,we determined a high-resolution crystal structure of Y3, andcharacterized its glycan binding pocket. These results lay thefoundation to identify and characterize Y3 homologs in otherfungal species and to develop this GBP scaffold for biomedical andresearch applications. This work also suggests the exploration ofuntapped potential of increasingly available small proteins in drugresearch using a multiple disciplinary approach.

Results and DiscussionBiochemical Characterization of Recombinant Y3. In an initial report,Y3 was isolated from the fungal fruiting body and showed anti-TMV activity which was retained after incubating the protein at90 °C for 10 min (35). A bioinformatic analysis of Y3 predictedan 18-aa signal peptide (SP) at its N-terminus and a potentialN-glycosylation site (Asn110) (Fig. 1A). Indeed, native Y3 wasreported to be a glycoprotein (35). Considering these putativepostmodifications, we expressed both full-length and SP-freegenes in Pichia pastoris (SI Appendix, Fig. S1). To preserve the

original structure and activity, no purification tag was included.Recombinant Y3 was produced only from the SP-free gene (Fig.1 B and C). The yield of recombinant Y3 was about 20 mg/Lafter filtration, concentration, and dialysis of the P. pastorisculture supernatant. The size-exclusion chromatography (SEC)analysis of purified Y3 led to a single peak with an estimatedMW of ∼39 kDa (Fig. 1B). Denaturing SDS/PAGE analysisrevealed a single band of ∼12 kDa, similar to that of nativeY3 directly isolated from C. comatus. Treatment with heat (90 °Cfor 20 min) or a range of acid/base (pH 3.0–11.5) had no effecton the SDS/PAGE profile of recombinant Y3. The impressivestability of Y3 may be ascribed to the presence of multipledisulfide bridges from its eight cysteine residues (Fig. 1A). In-deed, Ellman’s analysis excluded the presence of any free thiol inthe recombinant protein (36) (SI Appendix, Fig. S2).We next determined the MW of recombinant Y3 using

MALDI-TOF mass spectrometry (MS). The observed MW of12.20 kDa agreed well with the calculated MW of SP-free Y3(12.22 kDa) (Fig. 1D). Treating the recombinant Y3 with PNGaseF, which removes potential N-glycan of glycoproteins, led to noMW change in the MS analysis, further suggesting that Y3 is not aglycoprotein. Strikingly, we identified a 38% of total carbohy-drates in the Y3 sample (wt/wt) in phenol sulfuric acid analysis(37, 38) (SI Appendix, Fig. S3A), and dialysis of Y3 against PBSbuffer (vol/vol, 1:500) gradually reduced the carbohydrate contentto about 27% over the period of 6 d, suggesting a noncovalentinteraction. In light of the above result, we reexamined a broadpeak in the electrospray ionization mass spectrometry (ESI-MS)profile that did not possess proteinous UV-vis absorbance at260 nm (SI Appendix, Fig. S4 A–C). The peak consisted of a widerange of MWs and likely represented a mixture of potential car-bohydrate fragments bound to Y3, which possibly contribute tothe observed MW of Y3 in the SEC analysis (Fig. 1B). Copur-ification of endogenous carbohydrates with recombinant proteinshas been uncommonly encountered in previous studies (39) be-cause of generally relatively low binding affinity (∼millimolar)between noncognate glycans and GBPs (40). Nonetheless, ourstudies strongly suggest Y3 to be a GBP.We further evaluated the anti-TMV activity of recombinant

Y3 and showed it reduced the infectivity of purified Tobacco mildgreen mosaic virus (TMGMV) (26.5 μg/mL) by 50% at 0.12 μM(Fig. 1E), more effective than that reported for native Y3 andTMV (0.17 μM) (35). When mixed with TMGMV at room tem-perature and on ice, recombinant Y3 (0.078 μM) rapidly reducedthe infectivity of TMGMV by 40 and 10.7%, respectively. A20-min incubation period substantially increased the corres-ponding inhibition to 52.1 and 39.2%, suggesting the effects ofboth temperature and incubation time. Collectively, these data

Fig. 1. Biochemical characterization of Y3. (A) Se-quence alignment of Y3 and its homologs from fungalspecies. Conserved cysteine residues that potentiallyform disulfide bridges are shaded in dark- and lightgreen and all other conserved residues are indicatedwith dots (·). Potential N-glycosylation site (Asn110-Ser112) in Y3 is underlined with a dark line, andAsn110 is labeled with an asterisk (*). Residues poten-tially involved in ligand binding are shaded in yellow.Y3: C. comatus, GenBank: ADK35888.1; Agaricus:Agaricus bisporus var. burnettii JB137-S8, GenBank:XP_007333380.1; Galerina: Galerina marginata CBS339.88, GenBank: KDR84548.1; Gymnopus: Gymno-pus luxurians FD-317 M1, GenBank: KIK60251.1;Hebeloma: Hebeloma cylindrosporum h7, GenBank:KIM42673.1; Leucoagaricus: Leucoagaricus sp. SymC.cos,GenBank: KXN92904.1. (B) Recombinant Y3 was aputative dimer as shown in SEC analysis. Retentionvolumes of molecular standards of 1.35, 17, 44, and158 kDa are indicated. (C) SDS/PAGE analysis showedthe high purity of recombinant Y3. (D) The m/z value of purified Y3 at 12,204.85 in MALDI-MS analysis matched with calculated molecular weight of matureY3 monomer (12,229.54 Da). (E) The inhibitory curve of recombinant Y3 on TMGMV infection. Data are presented as mean ± SD (n = 6).

Zhang et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 8981

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0

demonstrated that recombinant Y3 from the yeast system retainedthe proven antiviral activity.

Y3-Induced Caspase-Dependent Apoptosis of Human T-Cell LeukemiaJurkat Cells. Given the diverse and important roles glycans play incancer biology, we assessed potential cytotoxicity of Y3 against apanel of human cancer cell lines (SI Appendix, Table S1). At 10 μMconcentration, Y3 showed only modest to weak growth inhibition ofcervical cancer HeLa cells, liver carcinoma HepG2 cells, pancreascarcinoma Dan-G cells, and prostatic cancer DU-145 cells (SI Ap-pendix, Fig. S5A). By contrast, however, Y3 exhibited potent activitytoward human T-cell leukemia Jurkat cells at the nanomolar level(SI Appendix, Fig. S5B). The high selectivity was further indicatedwith no observed cytotoxic effect on pancreas carcinoma MIA-PaCa-2, kidney HEK293, and head and neck squamous carci-noma UM-SCC-1 cells. To further probe the effects of Y3 on Jurkatcells, we examined the modes of cell death by using 7-amino-actinomycin D (7-AAD) and Annexin V double staining (41). Thisanalysis revealed that Y3 induced both early and late apoptosis ofJurkat cells in a dose-dependent manner (Fig. 2A). Treatment at0.1 μM for 4h induced 45 ± 2.5% of Jurkat cells to enter early stageapoptosis as indicated by Annexin V staining, while 23 ± 1.3% werestained by both 7-AAD and Annexin V, suggesting late-phase ap-optosis (Fig. 2A and SI Appendix, Fig. S6). These percentage valuesof early- and late-phase apoptosis were shifted to 17 ± 3.0% and73 ± 4.8%, respectively, at 20 h, giving apoptotic cells as 90% oftotal cells. These findings indicate a potent and rapid cytotoxicity ofY3 against leukemia cells. Of note, Y3 samples dialyzed to removebound carbohydrate exhibited the same extent of anti-Jurkat activity(SI Appendix, Fig. S3B), indicating a minimal effect of copurifiedcarbohydrates on the observed bioactivity.To further investigate this activity, we labeled recombinant

Y3 with a fluorescein-5-isothiocyanate (FITC) fluorescent probe.FITC-Y3 retained a similar level of anti-Jurkat activity as observedwith Y3 (SI Appendix, Fig. S7A), and demonstrated to bind to thecell surface of Jurkat cells in a dose-dependent manner (Fig. 2Band SI Appendix, Fig. S7B). We did not observe any transport ofFITC-Y3 into the cells. By contrast, FITC-Y3 showed minimalbinding to the control HEK293 cells (Fig. 2B). These results

suggest that Y3 triggered the apoptosis pathways through thebinding to the cell surface of Jurkat cells, and highlight potentialapplications in the management of acute T-cell leukemia.We next examined the potential mechanism of the cytotoxicity

of Y3. As the activation of caspases is essential in both intrinsicand extrinsic apoptotic pathways (42), we measured the levels ofcleaved caspases in Jurkat cells after the treatment of Y3 at 0.01–1 μM for 20 h (Fig. 2C). Western blotting analysis clearly dem-onstrated that Y3 induced the activation of caspases 3, 8, and 9 ina dose-dependent manner, while cotreatment with the pan-caspase inhibitor z-VAD-FMK (20 μM) significantly blocked theactivation of all three caspases, confirming a caspase-dependentmechanism of Y3′s action.

Glycan Binding Profile of Y3. As discussed, glycans are critical tovarious essential molecular and cellular processes related todiseases and health, for example as important biomarkers andpotential therapeutic and diagnostic targets of cancers (4–7). Toassess the glycan binding profile of Y3, we prepared biotinylatedY3 to screen against a mammalian glycan array consisting of600 glycans in replicates of 6 at the Consortium for FunctionalGlycomics (CFG). Detailed analysis of the concentration-dependent binding of Y3 showed the strongest binding affinity toGalNAcβ1–4(Fucα1–3)GlcNAc (LDNF) (Fig. 3A). LDNF is amember of the LacdiNAc (GalNAcβ1–4GlcNAc, LDN) familyof glycans, which are abundant in invertebrates such as parasitichelminths and insects (43). In humans, the LDN-type glycansare less common and may be associated with the developmentof some cancers (44, 45). The binding affinity of the LDNFglycan epitope and Y3 was 10–102 × higher than closely relatedLDN or sulfated analogs that were included in the array (Fig.3A). Additionally, 6′-sulfation of LDNF resulted in a weakerbinding (∼8× lower) while a 3-sulfation group completely

Fig. 2. Y3 induced the apoptosis of Jurkat cells. (A) Annexin V and 7-AADstaining quantitated cell populations in early and late apoptosis aftertreatment with serial concentrations of Y3 for 4 and 20 h. Data are shown asthe means ± SD (n = 3). Significant differences between control (0 μM) andtreatments are shown (*P < 0.05; **P < 0.01). (B) Y3-FITC showed strongbinding to the cell surface of Jurkat cells while minimal-to-no binding to HEK293 in fluorescence microscopy images. (C) Caspases-3, 8, and 9 in Jurkat cellswere activated by Y3 as indicated by Western blotting analysis using specificantibodies. Jurkat cells were treated with different concentrations of Y3 andpan-caspase inhibitor Z-VAD-FMK (20 μM) for 20 h.

Fig. 3. Glycan array screening identified LDNF as a specific ligand of Y3.(A) Y3 showed varying binding to glycans of the LDL family with LDNF as thebest ligand. Blue square: GlcNAc; yellow square: GalNAc; red triangle: Fuc; S:sulfo. (B) Fluorescence signals of other top glycans and common humanantigens in binding to Y3. Figures were generated from data with Y3 at50 μg/mL in the screening. Data are presented as mean ± SD (n = 6).

8982 | www.pnas.org/cgi/doi/10.1073/pnas.1706894114 Zhang et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0

blocked the glycan–Y3 interaction, illustrating a specific Y3–glycan interaction.In general, there was a distinct trend of sulfation in the top ligands

of Y3, such as disaccharides 3,6,6′-trisulfo-N-acetyllactosamine,6,6′-disulfo-lacto-N-biose, 6,6′-disulfo-lactose, 6,6′-disulfo-N-acetyllactosamine, and trisaccharide Fucα1–2(6S)Galβ1–3(6S)GlcNAc (6′,6′′-disulfo-trisaccharides H-type 1) (Fig. 3B). Infact, sulfo modification occurred on the majority of top 20 glycans,suggesting that this group could mediate specific interactions with Y3(SI Appendix, Table S2). GlcNAc, Glc, or their 6-sulfated analogsappeared at the reducing end of all top 20 glycans of Y3. In fact, β-6-sulfo-GlcNAc was the best monosaccharide glycan in the Y3 glycanscreening, whose signal was 11× lower than LDNF (Fig. 3B). Also,Y3 strongly favored terminal, nonreducing GalNAc or Gal thatis modified by mono- or disulfo groups at 6, 3, or 4 position (SIAppendix, Table S2). However, Y3 showed only weak binding toGalNAc itself, also known as Tn antigen that is commonly found oncancer cells (46). Other cancer-relevant Tn antigen derivatives in-cluding the Thomsen–Friedenreich antigen (Galβ1–3GalNAc) andsialyl Tn antigen (Neu5Acα2–6GalNAc) also showed minimal bind-ing to Y3 (Fig. 3B). Glycan profiling of Y3 further suggested thata β1,3-glycosidic bond was favored slightly more than a β1,4-linkage(e.g., 6,6′-disulfo-lacto-N-biose vs. 6,6′-disulfo-N-acetyllactosamine,Fig. 3B). In this regard, Y3 demonstrated minimal binding toneither β1–3 bond-containing histo-blood group antigensLewis a [Galβ1–3(Fucα1–4)GlcNAc] and b [Fucα1–2Galβ1–3(Fucα1–4)GlcNAc] nor β1–4 bond-related cancer antigensLewis x [Galβ1–4(Fucα1–3)GlcNAc] and y [Fucα1–2Galβ1–4(Fucα1–3)GlcNAc] (Fig. 3B). Strikingly, the only structural dif-ference between LDNF and Lewis x is the terminal GalNAc andGal, which varies their binding affinities toward Y3 at 50 μg/mL byover 17,000× (Fig. 3B). Overall, the glycan screening suggestedthat Y3 exclusively recognized the LDNF moiety and modestlyinteracted with sulfated di- or trisaccharides primarily consisting ofGalNAc, Gal, GlcNAc, Glc, Fuc, and their sulfated analogs. TwoCa2+-dependent lectins macrophage galactose-type lectin (MGL)and dendritic cell-specific C-type lectin DC-SIGN are othercharacterized GBPs that show a relatively tight interaction withLDNF (47, 48). However, MGL binds to the LDN antigen 1.3×more tightly than LDNF, while DC-SIGN shows significantlystronger interaction with Lewis x (49), marking Y3 as the onlyknown LDNF-specific GBP.

High-Resolution Crystal Structure of Y3. To gain structural insightsinto Y3 as a specific LDNF-binding GBP, we successfully de-termined its crystal structure by single-wavelength anomalousdiffraction analysis (SAD). Native Y3 crystal diffracted to 1.2-Åresolution bearing translational noncrystallographic symmetry (SIAppendix, Table S3). Heavy atom soaking of crystals resulted in anorphan dataset (Y3-Pt) in an alternative lattice used for experi-mental phasing, in the absence of any reference structure suitable

for molecular replacement (SI Appendix, Table S3). Depending onthe lattices, Y3 consists of one or two dimers per asymmetric unit.Y3 monomer forms a compact single-domain αβα-sandwich

consisting of three α-helices and a five-stranded β-sheet (Fig. 4).The β-sheet was linked as β1-β4-β5-β3-β2 from edge to center in anantiparallel orientation. Two long α-helices (α1 and α2) packagainst one side of the β-sheet, while a short C-terminal α3 is lo-cated on the opposite side. The monomer contains four-intramolecular disulfide bridges, Cys24-Cys101, Cys36-Cys69,Cys61-Cys126, and Cys44-Cys90 (Fig. 4 and SI Appendix, Fig.S8A), agreeing with the results of Ellman’s (SI Appendix, Fig. S2)and MS analysis (Fig. 1D). The majority of these disulfide bridgesare among conserved cysteines through Y3 homologs from otherfungal species (Fig. 1A). Another modification on Y3 was theformation of pyroglutamic acid from Gln19 presumably by a glu-tamine cyclase after translation (50) (SI Appendix, Fig. S8B); this isnot uncommonly observed on +1 residue that follows the signalpeptide. Also, one molecule ofN-cyclohexyl-2-aminoethanesulfonicacid from the crystallization buffer was clearly resolved between theβ3α2 and β4β5 loops (SI Appendix, Fig. S8C). Its anionic sulfategroup interacts with NPhe115, NSer86, and the Asn89 sidechain, whilethe hydrophobic cyclohexane extends into the bulk solvent. In thecrystallographic asymmetric unit, the four Y3 monomers have anrmsd range of 0.14–0.36 Å (among monomers) and differ primarilyin the β1α1-loop regions (residues 26–31), with a Cα (backbonecarbon) movement of up to 4.5 Å in this region (SI Appendix, Fig.S8D). The five-stranded, antiparallel β-sheets from two Y3 mono-mers assemble to a large, intermolecular 10-stranded, antiparallelβ-sheet with all α1 and α2 helices on one side and α3 on the op-posite side (Fig. 4 A and B). Additional dimeric interactions includeintermolecular hydrogen bonds between Arg′53 of the α1β2 loopand the neighboring Val63, and water-mediated H bonds betweenGln59/Gln′59 (SI Appendix, Fig. S8E).Structural homology searches (DALI server) (51) revealed that

Y3 shares limited structural similarity with known structures. Themost relevant in weak overall structural homology is an α-Galbinding lectin LDL from the mushroom Lyophyllum decastes[Protein Data Bank (PDB) ID code 4NDV, Z score 8.7, rmsd2.2 Å)] (52), which shares 12% sequence identity with Y3 (SIAppendix, Fig. S9A). The LDL structure contains a four-stranded,antiparallel β-sheet and two α-helices being packed against oneside of the β-sheet. However, the structure of Y3 shows significantdifferences from LDL (SI Appendix, Fig. S9B), including thenumber and locations of disulfide bridges, the number of β-sheets,and the α3 helix present in Y3. Importantly, LDL forms an al-ternate dimer by the stacking of the exposed sides of the β-strandsof the two subunits as an αβ2α-sandwich (53) (SI Appendix, Fig.S9C). By contrast, the Y3 monomer is featured with the presenceof a third α-helix (α3) and is incompatible of forming such β-stack-βconformation (SI Appendix, Fig. S9D). Finally, up to 1 mM of re-combinant Y3 did not agglutinate human and rabbit erythrocytes

Fig. 4. The X-ray crystal structure of Y3. (A–B) Y3 dimer forms a 10-stranded β-sheet with helices α1 and α2 on one side of the dimeric interface while twoα3 on the other side. (C) Structure topology illustration of Y3 dimer with the indication of disulfide bridges.

Zhang et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 8983

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0

(SI Appendix, Fig. S10), strikingly different from the character-istic feature of other reported lectins. Indeed, Y3 exhibited in-significant interactions with glycans (e.g., blood group glycans)other than LDNF in the microarray screening (Fig. 3). In thisregard, the reported hemagglutination activity of native Y3 pu-rified from the mushroom, in an early report (35), might be re-lated to impurities from the partially purified sample or possiblysuggest a different activity spectrum of Y3 from the nativesource. All of these structural and functional findings indicatethat Y3 adopts a unique glycan-binding mode.

Glycan Binding Site of Y3. The Y3 dimer forms a Janus conforma-tion with the majority of hydrophilic residues assembled on oneside. The other surface forms a hydrophilic pocket that sits on the10-stranded, antiparallel β-sheet and is surrounded by the twoα3 helices, one from each monomer (Fig. 5A). The large pocket is aunique feature compared with other reported GBPs and suggeststhat Y3 most likely interacts with complex glycan chains ratherthan mono- or disaccharides. Indeed, GalNAc, 6-sulfo-GlcNAc,D-Man, D-Glu, D-Gal, D-Fuc, or D-Lac did not show an interactionwith Y3 using isothermal titration calorimetry or biolayer in-terferometry analysis. Additionally, soaking of Y3 crystals with anyof these sugars did not allow clear interpretable ligand electrondifference maps. These results are consistent with the glycanbinding array data suggesting LDNF as a favored ligand.To provide additional insights into the Y3/glycan recognition,

the LDNF structure was modeled into the Y3 dimer (Fig. 5B).LDNF was aligned along the β1α1 loop in the top docking result.Key hydrogen bond interactions between the protein and sugarinclude Asp26, Asn30, Asp120, Asn122, and Asp123. The GalNAcmoiety sits in the cavity between the β3 and the β1α1 loop, and theterminal GlcNAc is near Asp120/Asn122, forming an interactionwith Tyr25 (Fig. 5B). The similar interaction mode with GlcNAc ispresent in the structures ofWisteria floribunda lectin (PDB ID code5KXC) and Clitocybe nebularis ricin B-like lectin (PDB IC code3NBE) (54) (SI Appendix, Fig. S11). In addition, the LDNF motifcould further extend to the dimeric interface of Y3 leading toimproved glycan binding affinity. Of note, we did observe un-assigned electron density near the GlcNAc interacting region inour high-resolution structure (Fig. 5A and SI Appendix, Fig. S12).The density most likely corresponds to a polysaccharide fragmentof bound endogenous glycan (likely mannose), indirectly sup-porting our docking model and agreeing with the results of phenolsulfuric acid and ESI-MS analysis (SI Appendix, Figs. S3 and S4).To provide a structure/function relationship of key residues in li-gand binding, we created Y3 D26A and N122A mutants (SI Ap-pendix, Fig. S13), whose carbohydrate contents were approximately33 and 28%, respectively (SI Appendix, Fig. S3C). Both mutantsnearly lost their ability to bind Jurkat cells (Fig. 5C). Furthermore,

the cytotoxicity of both mutants was significantly decreased com-pared with the level of a negative control BSA (Fig. 5D). Theseresults provided strong supportive evidence for our docking model,and offered insights into the specific interactions of Y3-LDNF.After binding to the glycan, Y3 may initiate apoptosis pathwaysin Jurkat cells.

ConclusionOur studies describe and characterize Y3 from the edible mush-room C. comatus as a GBP with a unique topology and tertiarystructure. Its cytotoxicity toward Jurkat cells is mediated by theactivation of caspase cascade that is presumably induced byspecific LDNF/Y3 interactions. T-cell acute lymphoblastic leu-kemia (T-ALL) is an aggressive hematological malignancy, andthe long-term survival of T-ALL patients remains extremely poordue to the remission and recurrence (55, 56). The critical need ofan improved treatment regimen of aggressive acute T-cell leu-kemia lends obvious significance to Y3 for developing potentialnovel treatment and diagnosis options. Future work includes thefunctional characterization of Y3 homologs from other fungalspecies, which will show considerable variations of key residuesin the glycan binding pocket (Fig. 1A), the detailed investigationof Y3 signaling in cancer cells (e.g., its receptor), and the de-velopment of Y3 and its homologs for biomedical applications.We envision that such work will discover GBPs as potentialtherapeutic agents toward other diseases and provide new insightinto the structure–function-relationship of this family of GBPs.

MethodsScreens for Anticancer Activity. The cells shown in SI Appendix, Table S1 werecultured in DMEM containing 10% FBS, 100 U·mL−1 penicillin and streptomy-cin, and maintained at 37 °C in a humidified incubator under 5% CO2. The cells(104 cells in 100 μL), seeded in 96-well plate, were treated with Y3 (10 μM).After incubation at 37 °C for 48 h, a thiazolyl blue tetrazolium bromide pro-tocol was followed to determine cell viability using a UV/vis microplate spec-trophotometer (BioTek). Six replications were performed per treatment andthe percent inhibition was calculated as (1 − test OD570/nontreated OD570) ×100%. Serial concentrations of Y3 (0, 0.016, 0.031, 0.063, 0.13, 0.25, 0.5,and 1 μM) were used to treat Jurkat cells.

Glycan Microarray Analysis of Y3′s Glycan Binding Profile. Glycan microarrayanalysis was performed by the Consortium for Functional Glycomics (Core H).TheMammalian Printed Array, version 5.3, consists of 600 glycans in replicatesof 6 and was used in this work. For screening the array, Y3 was biotinylatedusing EZ-Link NHS-PEG4-Biotinylation Kit (Thermo Scientific) according tomanufacturer’s instructions. Biotin-labeled Y3 at 5 and 50 μg/mL was ana-lyzed as previously described (57). Streptavidin-488 was used to detect bio-tinylated Y3 that bound to the glycans on the array. The average binding foreach glycan target as well as SD was calculated after the highest and lowestpoint from each set of six replicates were removed, and glycans were then

Fig. 5. Y3 contained a large pocket for interacting with LDNF primarily through H-bonding network. (A) The putative glycan binding pocket of Y3. Left, anelectrostatic surface representation of the Y3 dimer. The LDNF motif was modeled into the Y3 dimer and aligned along with the β1α1 loop. Hydrophilic andhydrophobic regions are shown with red and blue, respectively. Right, the dimer rotated by 180° to show the hydrophobic face. (B) A close-up view of the bindingsite with the docked LDNF. Key residues involved in H-bond interactions are labeled. (C) Fluorescence microscopy images of Jurkat cells after treatment withFITC-labeled Y3, Y3D26A, and Y3N122A. (D) Annexin V/7-AAD apoptotic assays revealed significantly decreased anti-Jurkat activity of Y3D26A and N122Amutants (1 μM). Data are presented as means ± SD (n = 3). Significant differences between WT and mutants are shown (*P < 0.05; **P < 0.01).

8984 | www.pnas.org/cgi/doi/10.1073/pnas.1706894114 Zhang et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0

ranked and sorted. The scanner response was linear to a maximal relativefluorescence unit value of about 50,000.

Structure Determination of Y3. Briefly, Y3 crystallization was performed in ahanging drop format. Heavy atoms derivative crystals for SAD were obtainedthrough cocrystallization or soaking. Y3 native and derivative crystal X-raydiffraction data sets were collected on beamlines 21-ID-G and 21-ID-F of theLife Sciences Collaborative Access Team (LS-CAT) facility at the AdvancedPhoton Source, Argonne National Laboratory (ANL) with a wavelength of0.9786 Å at 100 K. The structure of Y3 was determined with Pt-SAD withdetails shown in SI Appendix. Statistics on data collection and atomic structurerefinement are given in SI Appendix, Table S3. The refined coordinates havebeen deposited in the PDB (ID codes 5V6I and 5V6J).

Supporting Information. Materials and methods, including heterologousexpression, Ellman’s test for free thiol determination, quantitation of car-bohydrate content in protein samples, anti-tobamovirus assay, cell apoptosis

quantitation, cell imaging, Western blotting, hemagglutination assays,determination of the carbohydrate content, the carbohydrate binding to Y3,Y3 crystallization, diffraction data collection and processing, atomic struc-ture determination and refinement, docking, MS analysis, and statisticalanalysis are described in detail in SI Appendix, Materials and Methods.

ACKNOWLEDGMENTS. We thank Prof. Hui Feng (Boston University) forimportant experiment materials. We also thank Dr. Chen Liu and Dr. QunfengWu (Rutgers University) for informative discussions and Mr. Yi Zhang andMr. Julian Rashid for technical support. We thank the staff of LS-CAT, ANL forhelp with data collection and discussion. Octet RED384 is available at theCenter for Translational Research in Neurodegenerative Disease at Universityof Florida (UF). We acknowledge the participation of the Protein-Glycan In-teraction Resource of the CFG (NIH Grant R24 GM098791) and the NationalCenter for Functional Glycomics at Beth Israel Deaconess Medical Center, Har-vard Medical School (NIH Grant P41 GM103694). Part of this work was sup-ported by College of Pharmacy and UF. Y.D. is an Air Force Office of ScientificResearch Young Investigator.

1. Varki A (2017) Biological roles of glycans. Glycobiology 27:3–49.2. Moremen KW, Tiemeyer M, Nairn AV (2012) Vertebrate protein glycosylation: Di-

versity, synthesis and function. Nat Rev Mol Cell Biol 13:448–462.3. Hart GW (2013) Thematic minireview series on glycobiology and extracellular matri-

ces: Glycan functions pervade biology at all levels. J Biol Chem 288:6903.4. Pinho SS, Reis CA (2015) Glycosylation in cancer: Mechanisms and clinical implications.

Nat Rev Cancer 15:540–555.5. Stowell SR, Ju T, Cummings RD (2015) Protein glycosylation in cancer. Annu Rev Pathol

10:473–510.6. Dube DH, Bertozzi CR (2005) Glycans in cancer and inflammation–potential for

therapeutics and diagnostics. Nat Rev Drug Discov 4:477–488.7. Hung TC, Lin CW, Hsu TL, Wu CY, Wong CH (2013) Investigation of SSEA-4 binding

protein in breast cancer cells. J Am Chem Soc 135:5934–5937.8. Potapenko IO, et al. (2010) Glycan gene expression signatures in normal and malig-

nant breast tissue; possible role in diagnosis and progression. Mol Oncol 4:98–118.9. Sharon N (2007) Lectins: Carbohydrate-specific reagents and biological recognition

molecules. J Biol Chem 282:2753–2764.10. Sharon N, Lis H (2004) History of lectins: From hemagglutinins to biological recogni-

tion molecules. Glycobiology 14:53R–62R.11. Gabius HJ, André S, Jiménez-Barbero J, Romero A, Solís D (2011) From lectin structure

to functional glycomics: Principles of the sugar code. Trends Biochem Sci 36:298–313.12. Hassan MA, Rouf R, Tiralongo E, May TW, Tiralongo J (2015) Mushroom lectins: Speci-

ficity, structure and bioactivity relevant to human disease. Int J Mol Sci 16:7802–7838.13. Le Coq J, Ghosh P (2011) Conservation of the C-type lectin fold for massive sequence

variation in a Treponema diversity-generating retroelement. Proc Natl Acad Sci USA108:14649–14653.

14. Grant OC, Woods RJ (2014) Recent advances in employing molecular modelling todetermine the specificity of glycan-binding proteins. Curr Opin Struct Biol 28:47–55.

15. Taylor ME, Drickamer K (2014) Convergent and divergent mechanisms of sugar rec-ognition across kingdoms. Curr Opin Struct Biol 28:14–22.

16. Saghatelian A, Couso JP (2015) Discovery and characterization of smORF-encodedbioactive polypeptides. Nat Chem Biol 11:909–916.

17. Camby I, Le Mercier M, Lefranc F, Kiss R (2006) Galectin-1: A small protein with majorfunctions. Glycobiology 16:137R–157R.

18. Barondes SH, et al. (1994) Galectins: A family of animal beta-galactoside-bindinglectins. Cell 76:597–598.

19. Wadler CS, Vanderpool CK (2007) A dual function for a bacterial small RNA: SgrSperforms base pairing-dependent regulation and encodes a functional polypeptide.Proc Natl Acad Sci USA 104:20454–20459.

20. Kastenmayer JP, et al. (2006) Functional genomics of genes with small open readingframes (sORFs) in S. cerevisiae. Genome Res 16:365–373.

21. Ramamurthi KS, Storz G (2014) The small protein floodgates are opening; now thefunctional analysis begins. BMC Biol 12:96.

22. Ruiz-Orera J, Messeguer X, Subirana JA, Alba MM (2014) Long non-coding RNAs as asource of new peptides. eLife 3:e03523.

23. Ma J, et al. (2016) Improved identification and analysis of small open reading frameencoded polypeptides. Anal Chem 88:3967–3975.

24. Huang H, et al. (2015) A general strategy for the discovery of metabolic pathways:D-threitol, L-threitol, and erythritol utilization in Mycobacterium smegmatis. J AmChem Soc 137:14570–14573.

25. Erjavec J, Kos J, Ravnikar M, Dreo T, Sabotic J (2012) Proteins of higher fungi–fromforest to application. Trends Biotechnol 30:259–273.

26. Hibbett DS, et al. (2007) A higher-level phylogenetic classification of the Fungi. MycolRes 111:509–547.

27. Lindequist U, Niedermeyer TH, Jülich WD (2005) The pharmacological potential ofmushrooms. Evid Based Complement Alternat Med 2:285–299.

28. Kirst HA (2013) Developing new antibacterials through natural product research.Expert Opin Drug Discov 8:479–493.

29. Tanasova M, Sturla SJ (2012) Chemistry and biology of acylfulvenes: Sesquiterpene-derived antitumor agents. Chem Rev 112:3578–3610.

30. Buchel E, Mayer A, Martini U, Anke H, Sterner O (1998) Structure elucidation ofomphalotin, a cyclic dodecapeptide with potent nematicidal activity isolated fromOmphalotus olearius. Pestic Sci 54:309–311.

31. Xu X, Yan H, Chen J, Zhang X (2011) Bioactive proteins from mushrooms. BiotechnolAdv 29:667–674.

32. Kumar A, et al. (2002) An integrated approach for finding overlooked genes in yeast.Nat Biotechnol 20:58–63.

33. Li B, Lu F, Suo X, Nan H, Li B (2010) Antioxidant properties of cap and stipe fromCoprinus comatus. Molecules 15:1473–1486.

34. Fan J, et al. (2006) Structural elucidation of a neutral fucogalactan from the myceliumof Coprinus comatus. Carbohydr Res 341:1130–1134.

35. Wu L, Wu Z, Lin Q, Xie L (2003) [Purification and activities of an alkaline protein frommushroom Coprinus comatus]. Wei Sheng Wu Xue Bao 43:793–798. Chinese.

36. Ellman GL (1959) Tissue sulfhydryl groups. Arch Biochem Biophys 82:70–77.37. Masuko T, et al. (2005) Carbohydrate analysis by a phenol-sulfuric acid method in

microplate format. Anal Biochem 339:69–72.38. Dubois M, Gilles KA, Hamilton JK, Rebers PA, Smith F (1956) Colorimetric method for

determination of sugars and related substances. Anal Chem 28:350–356.39. Tateno H, Winter HC, Petryniak J, Goldstein IJ (2003) Purification, characterization,

molecular cloning, and expression of novel members of jacalin-related lectins fromrhizomes of the true fern Phlebodium aureum (L) J. Smith (Polypodiaceae). J BiolChem 278:10891–10899.

40. Liao JH, et al. (2016) A multivalent marine lectin from Crenomytilus grayanus possessesanti-cancer activity through recognizing globotriose Gb3. J Am Chem Soc 138:4787–4795.

41. George TC, et al. (2004) Distinguishing modes of cell death using the ImageStreammultispectral imaging flow cytometer. Cytometry A 59:237–245.

42. Elmore S (2007) Apoptosis: A review of programmed cell death. Toxicol Pathol 35:495–516.

43. Nyame AK, Leppänen AM, Bogitsh BJ, Cummings RD (2000) Antibody responses to thefucosylated LacdiNAc glycan antigen in Schistosoma mansoni-infected mice and ex-pression of the glycan among schistosomes. Exp Parasitol 96:202–212.

44. Machado E, et al. (2011) N-Glycosylation of total cellular glycoproteins from the hu-man ovarian carcinoma SKOV3 cell line and of recombinantly expressed humanerythropoietin. Glycobiology 21:376–386.

45. Hirano K, Matsuda A, Shirai T, Furukawa K (2014) Expression of LacdiNAc groups onN-glycans among human tumors is complex. BioMed Res Int 2014:981627.

46. Ju T, Aryal RP, Kudelka MR, Wang Y, Cummings RD (2014) The Cosmc connection tothe Tn antigen in cancer. Cancer Biomark 14:63–81.

47. van Vliet SJ, et al. (2005) Carbohydrate profiling reveals a distinctive role for theC-type lectin MGL in the recognition of helminth parasites and tumor antigens bydendritic cells. Int Immunol 17:661–669.

48. van Die I, et al. (2003) The dendritic cell-specific C-type lectin DC-SIGN is a receptor forSchistosoma mansoni egg antigens and recognizes the glycan antigen Lewis x.Glycobiology 13:471–478.

49. van Liempt E, et al. (2006) Specificity of DC-SIGN for mannose- and fucose-containingglycans. FEBS Lett 580:6123–6131.

50. Huang KF, Liu YL, Cheng WJ, Ko TP, Wang AH (2005) Crystal structures of humanglutaminyl cyclase, an enzyme responsible for protein N-terminal pyroglutamateformation. Proc Natl Acad Sci USA 102:13117–13122.

51. Holm L, Rosenström P (2010) Dali server: Conservation mapping in 3D. Nucleic AcidsRes 38:W545-9.

52. van Eerde A, Grahn EM, Winter HC, Goldstein IJ, Krengel U (2015) Atomic-resolutionstructure of the α-galactosyl binding Lyophyllum decastes lectin reveals a new proteinfamily found in both fungi and plants. Glycobiology 25:492–501.

53. Goldstein IJ, et al. (2007) A new alpha-galactosyl-binding protein from the mushroomLyophyllum decastes. Arch Biochem Biophys 467:268–274.

54. Pohleven J, et al. (2012) Bivalent carbohydrate binding is required for biological activityof Clitocybe nebularis lectin (CNL), the N,N′-diacetyllactosediamine (GalNAcβ1-4GlcNAc,LacdiNAc)-specific lectin from basidiomycete C. nebularis. J Biol Chem 287:10602–10612.

55. Maude SL, Teachey DT, Porter DL, Grupp SA (2015) CD19-targeted chimeric antigenreceptor T-cell therapy for acute lymphoblastic leukemia. Blood 125:4017–4023.

56. Belver L, Ferrando A (2016) The genetics and mechanisms of T cell acute lymphoblasticleukaemia. Nat Rev Cancer 16:494–507.

57. Smith DF, Song X, Cummings RD (2010) Use of glycan microarrays to explore speci-ficity of glycan-binding proteins. Methods Enzymol 480:417–444.

Zhang et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 8985

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

20,

202

0