Protein Expression and Purification 82 (2012) 97–105

Contents lists available at SciVerse ScienceDirect

Protein Expression and Purification

journal homepage: www.elsevier .com/ locate /yprep

Purification and characterization of riproximin from Ximenia americana fruit kernels

Helene Bayer a, Noreen Ey b, Andreas Wattenberg c, Cristina Voss b, Martin R. Berger a,⇑a Toxicology and Chemotherapy Unit, German Cancer Research Center, Im Neuenheimer Feld 581, 69120 Heidelberg, Germanyb Department of Biochemistry, Heidelberg Pharma AG, Schriesheimerstraße 101, 68526 Ladenburg, Germanyc PROTAGEN AG, Otto-Hahn-Straße 15, 44227 Dortmund, Germany

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 October 2011and in revised form 28 November 2011Available online 8 December 2011

Keywords:RiproximinPlant lectinType II RIPXimenia americanaChromatographyIsoforms

1046-5928/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.pep.2011.11.018

⇑ Corresponding author. Fax: +49 6221 42 3313.E-mail address: m.berger@dkfz.de (M.R. Berger).

1 Abbreviations used: DEAE, diethylaminoethyl; FPLtography; LS, lactosyl–Sepharose; MALDI, matrix-assition; MW, molecular weight; MWCO, molecularspectrometry; RCA, Ricinus communis agglutinin; RIP, rRpx, riproximin; SA, specific activity; SDS-PAGE, sodiuamide gel electrophoresis; TOF, time of flight; X.a., Xim

Highly pure riproximin was isolated from the fruit kernels of Ximenia americana, a defined, seasonallyavailable and potentially unlimited herbal source. The newly established purification procedure includedan initial aqueous extraction, removal of lipids with chloroform and subsequent chromatographic puri-fication steps on a strong anion exchange resin and lactosyl–Sepharose. Consistent purity and stable bio-logical properties were shown over several purification batches. The purified, kernel-derived riproximinwas characterized in comparison to the African plant material riproximin and revealed highly similar bio-chemical and biological properties but differences in the electrophoresis pattern and mass spectrometrypeptide profile. Our results suggest that although the purified fruit kernel riproximin consists of a mix-ture of closely related isoforms, it provides a reliable basis for further research and development of thistype II ribosome inactivating protein (RIP).

� 2011 Elsevier Inc. All rights reserved.

Introduction

Riproximin is a plant lectin that was recently identified as theactive component of a powdered plant material used in African tra-ditional medicine. Molecular phylogenetic analysis identified itssource as the semiparasitic plant Ximenia americana.1 Riproximinhas been shown to exhibit potent anticancer activity in vitro andin vivo and was classified as a ribosome inactivating protein (RIP)of type II [1,2].

RIPs of type II are heterodimeric proteins consisting of two pep-tide chains held together by a disulfide bridge. The binding chain(B-chain) is a lectin with affinity to various sugar structures thatare specific for each RIP. Following binding to the cell surface, typeII RIPs are transferred into the cell [3,4]. The active chain (A-chain)is an RNA N-glycosidase, which is able to hydrolyse a specific ade-nine of the ribosomal large subunit [5]. The mechanism of type IIRIPs’ cellular toxicity has been ascribed to the depurination ofrRNA, which is catalysed by the A-chain and results in an irrevers-ible arrest of cellular protein synthesis [6,7]. Recently, this hypoth-esis has been challenged by the finding that riproximin and otherRIPs of type II induce the unfolded protein response [8].

ll rights reserved.

C, fast protein liquid chroma-sted laser desorption/ ioniza-

weight cut off; MS, massibosome inactivating protein;m dodecyl sulfate–polyacryl-enia americana.

Representatives from RIPs of type II, like ricin or viscumin, arehighly toxic lectins and have been investigated as anticancer drugs.The recombinant lectin aviscumin, which was produced in Esche-richia coli, has demonstrated immunomodulatory and cytotoxicactivity. In clinical phases I/II studies, it achieved disease stabiliza-tion in some cases [9]. In addition, the RIP A-chain has been used toconstruct toxic antibody conjugates targeting cancer specific anti-gens. For example, the immunotoxin Combotox consists of the ri-cin A-chain coupled to an antibody directed against cell surfaceantigens CD19 and CD22 and has been investigated in a phase Istudy as a candidate for treatment of children with refractory leu-kaemia [10].

Due to its highly potent antineoplastic activity, the type II RIPriproximin has received interest as a potential anticancer drug can-didate [1]. Riproximin was purified from an undefined Africanplant material of limited availability. The purification procedureestablished for riproximin from this material consisted of foursteps. First, the plant material was extracted with acetone to de-plete tannins. An aqueous extract was prepared from the dried,tannin-free powder. This extract was subsequently purified by an-ion exchange chromatography. At neutral pH, riproximin bound toan anion exchange resin and eluted with high-salt buffer. The finalaffinity purification step was performed on a matrix containingfree galactose residues, which had been prepared by partial hydro-lysis of Sepharose.

For further investigations the availability of riproximin had tobe ascertained. X. americana fruit kernel extracts were shown tobe highly cytotoxic [1]; these kernels were considered to providea reliable source of riproximin. In the present study, we describe

98 H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105

the isolation procedure of riproximin from this source. In addition,we detail our results regarding separation and characterization ofX. americana riproximin isoforms.

Materials and methods

Aqueous extraction

Fruit kernels from X. americana originating from Florida, USA,were cracked and 10 mg of the inner soft kernel was homogenizedin 50 ml 20 mM Tris–HCl-buffer, pH 7.0. The resulting viscous sus-pension was transferred into a 50 ml tube and centrifuged (4000g,4 �C, 10 min). After separation from the upper fat layer, the loweraqueous supernatant was filtrated through a glass frit, resultingin a turbid aqueous extract.

Removal of lipids

Two extraction procedures and different organic solvents, suchas chloroform, acetic ester and n-heptane, were compared to opti-mize the removal of lipids either from the turbid aqueous extractor the initial kernel material. For removing the fats prior to theaqueous extraction of the proteins, 10 g kernel material werehomogenized directly in 50 ml of each organic solvent and centri-fuged (4000g, 4 �C, 10 min). This pre-extraction step was repeatedtwice. The fat-free kernel material was resuspended in 100% ace-tone, filtrated through a glass frit and vacuum-dried. The driedmaterial was resuspended in 50 ml 20 mM Tris–HCl-buffer, pH7.0 to obtain the final, clear aqueous extract.

Alternatively, the turbid aqueous extract prepared from the ker-nels as described above was clarified from remaining lipids byextraction with an equal volume of solvent. The extraction stepwas repeated 2–3 times. To remove traces of the solvent, whichinterfered with the chromatographic separation, the extractedaqueous solution was vacuum–freeze–dried. Prior to anion ex-change chromatography, the lyophilisate was redissolved in dou-ble distilled water (ddH2O) to the initial volume.

Pre-purification on CaptoQ

The pre-purification step was performed on the strong anionexchange resin, CaptoQ (GE Healthcare, Uppsala, Sweden). Theaqueous extract was loaded onto pre-equilibrated resin and thecolumn was washed with 20 mM Tris–HCl-buffer, pH 7.0. Riproxi-min containing fractions were eluted by increasing the salt concen-tration up to 500 mM NaCl. Collected fractions were analyzed bySDS–PAGE. Protein concentrations were determined colorimetri-cally with the Bradford assay (Roti-Nanoquant, Carl Roth,Karlsruhe, Germany).

Affinity purification on galactose resins

The affinity purification step was performed on a BeckmanCoulter PP2000 FPLC (Beckmann Coulter GmbH, Krefeld, Germany).The affinity purification of riproximin from fruit kernels on par-tially hydrolyzed Sepharose 4B (GE Healthcare, Uppsala, Sweden)followed the procedure described for riproximin from plant mate-rial [2]. Alternatively, lactosyl–Sepharose (Lactosyl Sepharose 4Fast Flow, GE Healthcare, Uppsala, Sweden), a prototype affinity re-sin providing lactose for binding of the lectins, was used. Lactosyl–Sepharose was equilibrated with 20 mM Tris–HCl-buffer, pH 7.0containing up to 500 mM NaCl. The riproximin-containing anionexchanger eluate from the pre-purification step was loaded andthe column was washed with equilibration buffer. Riproximin

was eluted with 20 mM Tris–HCl-buffer, pH 7.0 containing up to500 mM NaCl and 100 mM galactose.

The collected riproximin samples were pooled and concen-trated by ultrafiltration on 10,000 MWCO membrane filters,exchanging the elution buffer to a storage buffer containing20 mM Tris–HCl-buffer, pH 7.0 and 50 mM galactose. The purifiedproteins were analyzed by SDS–PAGE and their cytotoxic activitywas assessed in HeLa cell proliferation assays (see below).

Purified and lyophilized riproximin was used to determine itsextinction coefficient at 280 nm. Riproximin concentrations in thefractions were subsequently calculated from absorption at 280 nm.

Separation of the riproximin isoforms

A weak anionic resin, Toyopearl-DEAE-650S (Tosoh Bioscience,Stuttgart, Germany), was used for the separation of riproximin iso-forms. Riproximin protein mixture was loaded on the columnequilibrated with 20 mM piperazine-buffer, pH 5.7 and 50 mM gal-actose. After washing with the same buffer, elution was performedby increasing the NaCl concentration to 80 mM. The separatedriproximin isoforms were further analyzed by SDS–PAGE, gel filtra-tion, deglycosylation and their cytotoxic activity was assessed inHeLa cells.

Biological activity

The cytotoxic activities of riproximin or riproximin isoformswere characterized using the WST-1 viability assay (Roche Diag-nostics, Mannheim, Germany). Cells were propagated in humidatmosphere containing 5% CO2 at 37 �C. The media was supple-mented with 10% FCS, 2 mM L-glutamine, 100 IU/ml penicillinand 100 lg/ml streptomycin. For viability assays, the cells wereseeded into 96-well-plates (2500 cells/well in 100 ll) and allowedto settle down overnight. Riproximin containing fractions or puri-fied riproximin samples (100 ll) were then added to the cells andthe plates were incubated for 72 h. The cell growth was deter-mined by adding WST-1 and subsequent colorimetric detection.The cytotoxic activity of each riproximin sample/batch was charac-terized by its IC50 value in HeLa cells. For quantifying the fractionor batch purity, the specific activity (SA), defined as 1/IC50, wasused. One unit (1 U) was defined as the amount of the active com-pound (lg) in 1 ml that inhibits 50% of control cell proliferation.

Riproximin deglycosylation

For enzymatic deglycosylation of riproximin, the glycoproteinwas first incubated for 5 min at 99 �C in a denaturing and reducingbuffer (50 mM sodium phosphate, pH 5.0, 0.1% SDS and 0.05 mM b-mercaptoethanol). Afterwards 1 ll N-glycosidase F (1 U; RocheDiagnostics, Mannheim, Germany) was added and the mixturewas incubated for 3 h at 37 �C. The deglycosylated proteins wereanalyzed by SDS–PAGE.

Chemical deglycosylation was performed with the Glyco Pro-fileTM IV Chemical Deglycosylation Kit (Sigma Aldrich, Steinheim,Germany) according to the instructions of the manufacturer.Briefly, desalted and lyophilized riproximin material was dissolvedin a trifluoromethanesulfonic acid-anisole mixture, incubated for4 h at �20 �C and neutralized by the addition of pre-cooled 60%pyridine-solution. The deglycosylated proteins were dialyzedagainst 20 mM Tris–HCl-buffer, pH 7.5 and analyzed by SDS–PAGEand HeLa cell proliferation assay.

Mass spectrometric analysis

Both native and enzymatic deglycosylated riproximin isoformswere analyzed by MALDI mass spectrometry. A reducing

H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105 99

20 � 30 cm SDS–PAGE was used for separation [11]. The gel wascast with 12.5% acrylamide. After Coomassie staining, the sepa-rated bands were cut out, washed and digested in gel with trypsin(Promega, Madison WI, USA). Subsequently, the peptides were ex-tracted from the gel using 0.1% trifluoroacetic acid. The peptidepools were prepared onto a MALDI target using alpha-cyano-4-hy-droxy-cinnamic acid (Bruker Daltonik, Bremen, Germany) and ana-lyzed by an Ultraflex III TOF/TOF (Bruker Daltonik, Bremen,Germany). The data sets were imported into a relational database(ProteinScape, Bruker Daltonik, Bremen, Germany) and searchedagainst the NCBInr protein database using the Mascot algorithm(Matrix Science, London, UK). A 50 ppm mass tolerance was usedfor MS data, 0.2 Da mass tolerance for the parent mass in theMS/MS data and 1 Da tolerance for the fragment data. MS/MS spec-tra were obtained from selected peaks using the same instrumentand analyzed by manual de novo sequencing.

The clustering of the normalized MALDI data was performedusing the Clustal algorithm with hierarchical clustering, completelinkage and uncentered correlation [12].

Gene amplification

For RNA isolation, small pieces from the inner part of X. ameri-cana fruit kernels were powdered after freezing in liquid nitrogen.The powdered fruit kernel material was homogenized in 0.5 ml ofpre-cooled Concert Plant RNA Reagent (Invitrogen, Darmstadt,Germany). Further isolation followed the instructions of this man-ufacturer. For gene amplification, RNA was transcribed into cDNAand the riproximin gene was amplified by PCR using the riproximinspecific primers, Rpx1-frw (CATGCCGACTACTACCAAACCG) andRpx1-rev (GCATGTCAGACAACCACCATCC), which were designedto match within the non-translated region of the published riprox-imin cDNA sequence (Accession Number AM114537). The PCRproduct was sequenced by GATC company (Konstanz, Germany)using internal riproximin specific primers.

Results

Starting from the published procedure for riproximin fromAfrican plant material, the extraction as well as the chromato-graphic separation steps had to be modified and optimized forriproximin from the fruit kernels of X. americana. Furthermore,the purified fruit kernel riproximin was characterized with respectto its chain and isoform composition, as well as glycosylationpatterns using biochemical methods and mass spectrometry.

Aqueous extraction and removal of lipids by solvent extraction

The extraction step had to be modified to deal with the high fatcontent of the fruit kernels. This characteristic differed from theAfrican plant material, from which contaminating tannins had tobe removed. The optimal extraction ratio was 200 mg kernel mate-rial per 1 ml aqueous buffer. After centrifugation of the raw aque-ous extract, a thick lipid layer covered the supernatant and couldnot be mechanically removed. The subsequent filtration of the dec-anted supernatant yielded a cytotoxic turbid extract that could notbe cleared by additional centrifugation (Fig. 1A). Direct binding/elution experiments of the extract on partially hydrolyzed Sephar-ose 4B showed that no proteins were retained by the affinity ma-trix, most probably because of the fats interfering with binding.

To remove the lipids from the aqueous extract, chloroform, ace-tic ester and n-heptane were used in two different extraction pro-cedures: pre-extraction of lipids from the kernels prior to aqueousextraction of the proteins vs. a solvent extraction of the turbid ex-tract. The extraction and clean up efficacies were measured,

respectively, by comparing the extracts’ cytotoxic activity as wellas binding onto hydrolyzed Sepharose.

While the cytotoxic activity did not differ between the variousextracts (Fig. 1B), the galactose binding of riproximin varied con-siderably with the solvent and/or extraction procedure. Best bind-ing to hydrolyzed Sepharose was achieved with the aqueousextract cleared by chloroform extraction, for which only a smallproportion of the riproximin proteins was lost in the flow throughand wash fractions (Fig. 1C). To completely remove chloroformtraces, the clarified extracts were vacuum–freeze–dried and redis-solved in ddH2O.

Pre-purification on a strong anion exchange resin

As for the pre-purification of the African plant material aqueousextract, a strong anion exchange resin was used for the first coarsechromatographic purification step. This resin was chosen, becausethe previously used DEAE cellulose was not adequate for FPLC.

SDS–PAGE analysis of the flow through revealed that severalirrelevant proteins were removed from the extract, while the entirecytotoxic activity remained bound to the column (Fig. 2). Riproxi-min started to elute with buffer containing around 200 mM NaCl.Elution was complete at 500 mM NaCl. The yield of this step de-pended on the NaCl concentration used for elution and increasedat higher salt concentrations. When 200 mM NaCl were used forelution, an average of 76% of the loaded biological activity wasrecovered (Table 1).

Affinity purification on galactose containing resins

Purification of riproximin proteins was first performed onhydrolyzed Sepharose 4B as described for the purification ofriproximin from African plant material. However, the bindingcapacity of hydrolyzed Sepharose proved to be too low for the highriproximin content of the fruit kernel aqueous extract. A prototypelactosyl–Sepharose was tested as alternative affinity resin.

First, purification runs on lactosyl–Sepharose, performed aswith the hydrolyzed Sepharose 4B (binding and washing buffer:20 mM Tris–HCl, pH 7.0, elution buffer: 20 mM Tris–HCl, pH 7.0and 100 mM galactose), resulted in riproximin binding but no elu-tion. To assess whether this effect was due to ionic interactions,binding and elution of riproximin to/from lactosyl–Sepharose weresubsequently analyzed in a batch approach. Riproximin obtainedfrom hydrolyzed Sepharose purification was supplemented withincreasing NaCl concentrations (100–1000 mM) and applied ontolactosyl–Sepharose. For elution, 100 mM galactose was used incombination with the respective NaCl concentration. SDS–PAGEanalysis of the binding and elution supernatants showed thatriproximin in the sample without NaCl bound completely to lacto-syl–Sepharose, but failed to dissociate in the presence of 100 mMgalactose alone. With increasing NaCl concentrations, the bindingof riproximin decreased only slightly (Fig. 3A, binding). Elution ofriproximin was observed for buffers containing P100 mM NaClas well as 100 mM galactose (Fig. 3A, elution). A NaCl concentra-tion of 100 mM was therefore chosen for starting the gradient col-umn runs.

For further optimization, complete affinity purification runswere performed at NaCl concentrations of 200, 300 and 400 mMin both loading and elution buffers. With 200 mM NaCl a very pureriproximin was eluted from the column. However, the regenerationsteps with 1 M and 5 M NaCl indicated an incomplete elution, sincea small amount of riproximin was also detectable in these frac-tions. Higher concentrations of NaCl (300 and 400 mM) resultedin a higher yield of riproximin, but the eluted riproximin wasdetectably contaminated with other proteins (Fig. 3B). As expectedfrom the SDS–PAGE pattern, when analyzed for concentration and

Fig. 1. Protein extraction, cytotoxicity and binding behavior: Proteins extracted from Ximenia americana by aqueous and/or organic solvents were characterized by theirbinding to hydrolyzed Sepharose 4B. (A) SDS–PAGE analysis followed by silver staining showing proteins from aqueous extracts from American X.a. fruit kernels incomparison to an extract from African plant material. Lane 1: aqueous extract from X.a. fruit kernels; lane 2: chloroform clarified, aqueous extract from X.a. fruit kernels; lane3: lyophilized, chloroform clarified, aqueous extract from X.a. fruit kernels; lane 4: molecular weight marker; lane 5: acetone pre-extracted, aqueous extract from X.a. Africanplant material. The bands indicating riproximin proteins are framed. (B) The cytotoxic activity of solvent treated, aqueous extracts (Extracts 1–3) from X.a. fruit kernels wascompared by a WST-1 assay on HeLa cell proliferation at 72 h with that of the solvent untreated, aqueous extract (Control). Extract 1: chloroform extraction followed byaqueous extraction; Extract 2: acetic ester extraction followed by aqueous extraction; Extract 3: aqueous extraction followed by chloroform; Control: aqueous extractiononly. For all extracts, a starting dilution of 1:100,000 was chosen. (C) Comparison of riproximin purification yield depending on the extraction procedure. Three extractionprocedures are compared by the yield of riproximin after binding to hydrolyzed Sepharose 4B: (a) aqueous extraction only, (b) aqueous extraction followed by chloroformextraction, (c) chloroform extraction followed by aqueous extraction. Extract (500 ll, respectively) was loaded onto hydrolyzed Sepharose 4B. Flow throughs (Ft), washfractions (W) and eluates with 100 mM galactose (E) were separated and visualized by SDS–PAGE and silver staining.

Fig. 2. Optimization of the NaCl concentration for pre-purification on CaptoQ: Chloroform clarified, aqueous extract was loaded onto a column filled with the strong anionexchange resin, CaptoQ. The elution was performed with a NaCl gradient (0–500 mM). (A) Chromatogram of the optimization run with flow through (Ft) of the chloroformclarified, aqueous extract and elution peaks of the NaCl gradient (EP1-EP4), as indicated by the arrows. (B) SDS–PAGE analysis followed by silver staining of flow through andelution fractions with lane 1: molecular weight marker, lane 2: flow through of the chloroform clarified, aqueous extract, lane 3–4: eluted proteins from peak 1, lane 5–6:eluted proteins from peak 2, lane 7–8: eluted proteins from peak 3, lane 9–11: eluted proteins from peak 4. The bands indicating riproximin proteins are framed.

Table 1Representative run of the pre-purification of an aqueous extract on CaptoQa.

Volume (ml) Concentration (lg/ml) IC50b,c (lg/ml) Specific activityd (U/lg) Total activity (U) % of total activity

Loaded extracte 20 9900 8.1 � 10�4 1200 238 � 106 –Flow through 12 255 4.1 � 10�4 24 0.07 � 106 0.03Elutionf 48 1019 2.7 � 10�4 3700 181 � 106 76.15 M NaCl 12 1041 5.0 � 10�4 200 2.5 � 106 1.1

a Three different pre-purification runs yielded on average 75.8% (mean; SD = 0.67; n = 3) of the respective biological activity.b IC50 is the concentration inhibiting 50% of control cell proliferation.c IC50 was determined by WST-1 assay on HeLa cell proliferation at 72 h.d Specific activity SA (U/lg) was defined as 1/IC50 (lg/ml). One unit (1 U) was defined as the amount of the active compound (lg) in 1 ml that inhibits 50% of control cell

proliferation.e Aqueous extract pre-cleared with chloroform.f Four elution fractions were pooled.

100 H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105

biologic activity, the eluate from the 200 mM NaCl run showed thehighest specific activity, but contained less total protein than theother fractions (Table 2).

Using the optimized procedure, several purification runs wereperformed in the presence of 200 mM NaCl. The purification wasvery good reproducible, yielding on average 81% of the loaded

Fig. 3. Optimization of the salt concentration used for affinity purification of riproximin: The NaCl concentration used for final purification of riproximin on lactosyl–Sepharose (LS) was selected from a series of NaCl concentrations that were tested for their influence on binding to and elution from the resin. All samples from batch andchromatographic approaches were analyzed by SDS–PAGE and subsequent silver staining. (A) Investigation of binding and elution of riproximin in the presence of differentNaCl concentrations in a batch approach. Chloroform clarified, aqueous extract (100 ll; Load) was applied on 200 ll of LS resin. Binding of the extract to LS was investigatedin the presence of 0–1000 mM NaCl. For elution of the riproximin proteins, 100 mM galactose was added to the various NaCl concentrations (0–1000 mM). (B) Purification ofriproximin on a LS column with various NaCl concentrations. The CaptoQ eluate containing riproximin was affinity purified in three independent runs in the presence of 200,300 and 400 mM NaCl. For regeneration the column was washed with 1 M NaCl. Only elution and regeneration fractions are shown.

Table 2Comparison of the elution fractions of independent runs with varying NaClconcentrations.

NaCl concentration (mM) Eluates with 100 mM galactose

Total protein (lg) IC50a

(lg/ml)Specific activityb

(U/lg)

200 204 3.8 � 10�5 26,300300 312 5.0 � 10�5 20,000400 528 7.0 � 10�5 14,300

a IC50 was determined by WST-1 assay on HeLa cell proliferation at 72 h.b Specific activity SA (U/lg) was defined as 1/IC50 (lg/ml).

Table 3Purification runs on lactosyl–Sepharosea.

CaptoQeluate

IC50b

(lg/ml)Specificactivity(U/lg)

Totalactivity(U)

Yieldc

(%)

12 � 10�4 833 34 � 106 –

Run 1 1.1 � 10�4 9100 26 � 106 76Run 2 0.7 � 10�4 14300 29 � 106 85Run 3 1.0 � 10�4 10000 25 � 106 74Run 4 0.6 � 10�4 16700 32 � 106 94Run 5 0.4 � 10�4 25000 12 � 106 80

a CaptoQ eluate (40 ml) was purified in five runs (run 1–4: 9 ml each; run 5: 4 ml)on 9 ml lactosyl–Sepharose.

b The IC50 of purified riproximin was determined by WST-1 assay on HeLa cellproliferation at 72 h.

c The purification of riproximin on lactosyl–Sepharose yielded on average 81%

H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105 101

biological activity (Table 3). SDS–PAGE analysis revealed thatriproximin was eluted in a highly pure form (Fig. 4A, B).

Optimized purification protocol

The optimized purification protocol for riproximin from X.americana kernels consisted of the following steps: (1) aqueousextraction of riproximin proteins from homogenized fruit kernelmaterial; (2) lipid extraction by chloroform with subsequent vac-uum–freeze–drying; (3) coarse chromatographic purification ofthe redissolved extract on the strong anion exchange resin with

Fig. 4. Final purification of riproximin on lactosyl–Sepharose: Typical riproximin purificain the presence of 200 mM NaCl and elution of riproximin by adding 100 mM galactose toflow through (Ft), elution and NaCl regeneration peaks, as indicated by the arrows. (B) Pelution fractions were visualized by SDS–PAGE and silver staining. Lane 1: pre-purificatimarker, lane 4: pooled wash fraction (W), lanes 5–7: elution fractions, lane 8: regenera

200 mM NaCl elution and (4) affinity binding of the anion exchan-ger eluate to lactosyl–Sepharose followed by elution of pureriproximin with buffer containing 200 mM NaCl and 100 mMgalactose.

Purified riproximin eluates obtained from 10 to 20 g kernelmaterial were pooled to a batch. Riproximin samples were storedin a buffer containing 50 mM galactose at �20 �C. Table 4 gives

tion run on lactosyl–Sepharose using the optimized protocol, loading of riproximin200 mM NaCl, is shown. (A) Chromatogram of a representative purification run withroteins from the pooled flow through and wash fractions as well as the three singleon eluate (Ld, starting material), lane 2: flow through (Ft), lane 3: molecular weighttion fraction.

Table 4Comparison of riproximin batches.

Kernelmaterial(g)

Yield ofriproximin(mg)

Concentration(lg/ml)

IC50a

(lg/ml)Specificactivityb

(U/lg)

Batch1

10 28.9 3886 1.5 � 10�4 6700

Batch2

10 28.4 7100 1.2 � 10�4 8300

Batch3

10 28.4 8700 1.4 � 10�4 7100

Fig. 5. SDS–PAGE analysis of purified riproximin and its deglycosylated counter-parts: Native, enzymatically and chemically deglycosylated riproximin proteinswere separated by SDS–PAGE and visualized by Coomassie (A) or silver staining (B).Enzymatic deglycosylation was performed with PNGase F, chemical deglycosylationby treatment with trifluoromethanesulfonic acid. (A) High resolution (20 � 30 cmgel) reducing SDS–PAGE showing the chain patterns of native (lane 2) andenzymatically deglycosylated riproximin (lane 3), lane 1: molecular weight marker.Bands that were analyzed by MALDI-TOF are marked with spots. (B) Reducing SDS–PAGE showing native riproximin (lane 2) and chemically deglycosylated riproximin(lane 3), lane 1: molecular weight marker.

102 H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105

an overview on three batches for their protein content, biologicalactivity (IC50, SA) and total yield of riproximin. On average, 2–3 mg of purified riproximin were obtained from 1 g fruit kernelmaterial. This amount was >500-fold higher than the respectiveamount that could be purified from 1 g African plant material. Nev-ertheless, the biological activity of the fruit kernel riproximin wassimilar to that from the African plant material, as shown by an IC50

of 0.14 ng/ml in HeLa cells.Fruit kernel riproximin showed six different bands (Fig. 5A,

bands A1–A3 and B1–B3) following separation by reducing SDS–PAGE. The three lower clearly separated but neighbouring bandswere assumed to be A-chains, with two of them predominating(Fig. 5A, bands A1–A3). The upper three bands, presumably B-chains, showed a more diffuse, contiguous band pattern, thus indi-cating heterogeneous glycosylation (Fig. 5A, bands B1–B3).

Separation of riproximin isoforms

Analogous to the separation of the African plant materialriproximin isoforms, two of the kernel riproximin isoforms couldbe partially separated by anion-exchange chromatography. SDS–PAGE analysis of these fractions revealed that two of the three low-er protein bands, which were assumed to be the A-chains, could beseparated and assigned to each of two riproximin isoforms: Rpx-I,represented by the A-chain with medium MW (Fig. 6A, lane 1) andRpx-II with higher MW A-chain (Fig. 6B, lane 1). The A-chain withthe lowest MW (Fig. 5A, band A1) was no longer found in the sep-arated fractions. However, no clear isoform assignment could begiven to the B-chains. Each of the separated riproximin isoforms

contained heterogeneous B-chains, with the larger B-chains beingenriched in the fraction of Rpx-I (Fig. 6A, lane 1) and the lowerB-chains in the fraction of Rpx-II (Fig. 6B, lane 1).

Both isoforms demonstrated similar MWs when analyzed bynon-reducing SDS–PAGE (�56–60 kDa) (Fig. 6C, lane 1–2) and sizeexclusion chromatography (Rpx-I: 50 kDa; Rpx-II: 53 kDa), as wellas highly similar biological activity (Fig. 7).

Glycosylation pattern

For analyzing the glycosylation impact, native fruit kernel riprox-imin as well as its separated isoforms were treated with PNGase F.After this treatment the diffuse upper native riproximin bands(Fig. 5A, bands B1–B3), corresponding to B-chains, shifted to lowerMW and appeared in the SDS–PAGE gel as two, better separatedbands (Fig. 5A, bands B1d and B2d). When each isoform was degly-cosylated and analyzed separately, the presumed isoform B-chainsshowed again a decrease in their apparent size, but the intensity ofthe resulting bands differed from the pattern observed for the nativeriproximin mixture (Fig. 6A, B, bands B1d–B2d). In contrast, the low-er bands, corresponding to the A-chains, showed no MW shift afterenzymatic deglycosylation, neither for the native riproximin (Fig-ure 5A, bands A1d–A3d) nor for the separated riproximin isoen-zymes (Fig. 6A, B, bands A2d-A3d). Treatment of the proteins withother deglycosylating enzymes, like O-deglycosydase or the endo-glycosydase, EndoH, caused no additional shift of the bands (datanot shown). Because enzymatic deglycosylation was effective onlyunder denaturing conditions, the biological activity of enzymaticallydeglycosylated riproximin could not be assessed.

Since an incomplete deglycosylation cannot be excluded whenusing enzymes, a chemical deglycosylation was additionally per-formed. After this treatment, the bands of native riproximin shiftedstronger towards lower MW than after enzymatic treatment. Thevarious upper bands (B-chains) as well as the lower bands (A-chains) of native riproximin proteins converged into two singlebands at apparent sizes of 28 and 26 kDa, respectively (Fig. 5B).During the neutralization and dialysis steps, a significant amountof riproximin precipitated. Despite extensive dialysis to renaturethe proteins, chemically deglycosylated riproximin showed nodetectable cytotoxic activity.

Mass spectrometry analysis

MALDI MS analysis was performed with tryptic digest of eachSDS–PAGE band from native and enzymatically deglycosylatedriproximin resulting in mass spectra representing the peptide pro-file of each polypeptide. One prominent peptide mass at 1377 Daappeared in all of the analyzed polypeptides and was thereforeused for normalization of the quantitation. The occurrence of pep-tides of a specific mass allowed the analyzed bands to be classifiedinto three groups. All of the presumed A-chains classified withinthe groups 1 and 2, while the B-chains classified within group 3(Table 5).

The MS spectra of the three lower native bands as well as theirdeglycosylated forms (Fig. 5A, bands A1–A3 and A1d–A3d) sharedseveral peptide masses but were distinguishable by the disappear-ance of prominent peptides of 1453 and 1577 Da and the simulta-neous appearance of a 1591 Da peptide. Chains A1, A2, A1d andA2d were thus classified into group 1, while chains A3 and A3dconstituted a similar but different group 2. No prominent masschange was observed between the spectra of corresponding nativeand deglycosylated bands, i.e. between polypeptide A1 and A1d, A2and A2d or A3 and A3d (Table 5).

The polypeptides from group 3, into which all B-chains classi-fied, shared with the A-chains only the one prominent mass thathad been used for normalization. Moreover, the MS spectra of

Fig. 6. SDS–PAGE analysis of riproximin isoforms: Riproximin isoforms, Rpx-I and Rpx-II, were partially separated by anion exchange chromatography. Native andenzymatically deglycosylated riproximin isoforms were analyzed by reducing (A + B) or non-reducing (C) SDS–PAGE and visualized by silver staining. Enzymaticdeglycosylation was performed with PNGase F, which is visible in silver stain as a light band of� 32 kDa (E). (A) Reducing SDS–PAGE showing the chain pattern of native (lane1) and enzymatically deglycosylated (lane 2) riproximin isoform 1 (Rpx-I). (B) Reducing SDS–PAGE showing the chain pattern of native (lane 1) and enzymaticallydeglycosylated (lane 2) riproximin isoform 2 (Rpx-II). (C) Non-reducing SDS–PAGE showing both riproximin isoforms, Rpx-I (lane 1) and Rpx-II (lane 2). Lane 3: molecularweight marker.

Fig. 7. Biological activity of riproximin isoforms: The cytotoxic activity of theenriched riproximin isoforms (Rpx-I and Rpx-II) was compared by a WST-1 assay onHeLa cell proliferation at 72 h with that of the native riproximin (Rpx).

H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105 103

the three B-chains of native riproximin (Fig. 5A, bands B1–B3)showed high similarity to the two deglycosylated B-chains(Fig. 5A, bands B1d–B2d), indicating that these polypeptides areclosely related to each other but explicitly different from the lowerbands (Table 5). Apart from the most prominent peak, the MS sig-nals of the B-chain peptides were very low so that no additionalpeptides could be identified in the profile of the deglycosylatedupper bands. Most likely, the glycosylated peptides were not cov-ered by the MALDI-MS analysis.

A comparison of MS spectra of the riproximin polypeptides withdatabase proteins found no correspondence. However, a directcomparison of peptide sequences obtained after de novo sequenc-ing of the MS/MS data with the protein sequence of African plant

Table 5Analysis of MALDI-TOF mass lists of riproximin bands.

a Peptide signals were marked according to their intensity: 0.5–1.0 (dark grey), 0.2–0.5b The peptide mass at 1377 Da, which was present in all of the analyzed polypeptides, wc The clustering of the normalized MALDI data was performed using the Clustal algorith

material riproximin (CAJ38823) showed some sequence homologythough no identity. As an example, two peptides with masses of1377 and 1577 Da were manually sequenced as YVEQQVLAGTLRand QSGSYGSVVNNGDHR, respectively. The sequence tags LAGTwithin the former as well as NNGD within the latter were identicalto the published riproximin sequence, whereas the rest of the se-quence of these peptides did not show any homology to the pub-lished riproximin sequence. The peptide with the mass of1591 Da (sequence: QSGSYGAEVNPGAPTR), which is a marker ofgroup 2 peptides, showed close homology to the peptide withthe mass of 1577 Da.

Expression of riproximin at transcriptional level

Following RNA extraction and transcription into cDNA a clearband was obtained after PCR amplification with riproximin specificprimers. Sequencing of this template with internal riproximin spe-cific primers revealed identity with the published cDNA sequence(data not shown).

Discussion

Riproximin was initially isolated as the active antineoplasticcomponent from African plant material of undefined compositionand limited availability. A reliable source was a prerequisite forthe further development of riproximin as a potential new com-pound for treating cancer. A milestone in this development wasthe identification of the semiparasitic plant X. americana as the ori-gin of the African plant material. Next, the fruit kernels wereshown to exhibit high cytotoxic activity indicating high riproximin

(light grey), <0.2 (unmarked).as used for normalization of the quantitation.

m with hierarchical clustering, complete linkage and uncentred correlation.

104 H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105

concentrations [1]. Because of the regenerative nature of the fruitkernels, a well defined and potentially unlimited source had beenfound. The next milestone was to establish a robust, reproducibleand up-scalable purification protocol for riproximin from X. amer-icana fruit kernels. Furthermore, the polypeptide composition andbiological activity of the fruit kernel riproximin had to be com-pared to that of riproximin from the African plant material to dem-onstrate the equivalence of these proteins.

The established purification protocol for riproximin from Afri-can plant material was not applicable for the fruit kernels. Thus,a new purification method had to be developed.

While the African plant material was a dry and low-fat but tan-nin-rich powder, the fresh fruit kernel material contained a highamount of lipids that interfered with protein binding to the chro-matography resins. The optimal procedure involved a combinationof aqueous extraction and subsequent clearance of the extract bychloroform, since it provided the best binding and elution of kernelriproximin in the subsequent chromatographic step.

Analogous to the pre-purification of the aqueous extract obtainedfrom the African plant material, for the coarse chromatographic stepthe strong anion exchanger, CaptoQ, was chosen as a robust andFPLC-suitable alternative to DEAE cellulose. For the affinity chroma-tography a prototype lactosyl–Sepharose was used. It is a classicalSepharose with immobilized lactose, which additionally containscationic charges that lead to anionic interactions, typical for an anionexchange resin. A NaCl concentration of 200 mM was chosen for elu-tion from the strong anion exchanger, since this concentration pro-vided optimal binding to lactosyl–Sepharose as well as the highestpurity of the final riproximin eluates.

Fruit kernel riproximin showed the same cytotoxic activity(IC50 = 2.3 pM, HeLa) as that from the African plant material(IC50 = 1.1 pM, HeLa). However, the amount of riproximin obtainedfrom fruit kernels was considerably higher than that from the iden-tical net weight of African plant material. X. americana fruit kernelscan thus serve as an abundant and potentially unlimited riproxi-min source.

Riproximin purified from the X. americana fruit kernels showedthe same physico-chemical and biological properties as the Africanplant material riproximin, including solubility, charge, lectin bind-ing activity, cytotoxicity as well as specific ribosome depurination[8]. However, its SDS–PAGE band pattern differed considerablyfrom that of African plant material riproximin. The latter hadshown four distinct protein bands under reducing conditions,which had been assigned to the respective A- and B-chains oftwo riproximin isoforms [2]. In contrast, riproximin from fruit ker-nels showed three overlapping bands in the range of 30–35 kDa,that were assumed to be B-chains and three bands in the rangeof 25–29 kDa, presumably corresponding to the A-chains.

This chain assignment is in agreement with the typical type IIRIP structure, consisting of two chains, A- and B-, which are con-nected by a disulfide bond [13]. Moreover, the typical A- and B-chain pattern of fruit kernel riproximin was corroborated by MSanalysis.

The typical RIP B-chain is a lectin with specific sugar affinity.Characterized by a MW of approximately 35 kDa it is slightly largerthan the corresponding A-chain [4,14]. The B-chain of type II RIPscontains highly conserved asparagines [15], which are heavily gly-cosylated. Both potential glycosylation sites were shown to bepresent in the published riproximin B-chain sequence, too [2].

A size of 30–35 kDa as well as a diffuse appearance that is typ-ical for glycosylation also characterized the upper three bands inthe SDS–PAGE pattern of fruit kernel riproximin, which thereforewere considered as B-chains. As expected, enzymatic deglycosyla-tion lowered their MW and led to a decreased overall complexity ofthe B-chain band pattern, indicating that heterogeneous oligosac-charides present on the native polypeptidic backbones had been

removed. Moreover, in the MS analysis the native riproximin’s B-chains as well as their deglycosylated counterparts were recog-nized as a distinct group of very similar polypeptides clearly distin-guishable from the A-chains. The MS observations therefore notonly support the classification of theses bands as B-chains, but alsoprovide clues about their isoenzyme nature, when considered to-gether with the results of the analysis of the partially purifiedriproximin isoforms.

The type II RIP active A-chain with N-glycosidase activity isapproximately 30 kDa in size and can also be glycosylated [14].The A-chain of ricin, for example, was described to contain one ortwo glycosylation sites [16,17]. The published riproximin A-chainamino acid sequence showed only one potential N-glycosylationsite, which could be abolished by a single nucleotide polymorphism[2]. Accordingly, fruit kernel riproximin’s lower bands, which wereconsidered to be the A-chains, were of lower MW and clearly de-fined. Upon PNGase treatment, no MW shift was observed for thesepolypeptides. However, after chemical deglycosylation, all threebands converged at the level of the lowest one, indicating that theupper two bands possess PNGase-resistant glycosylation.

MS analysis assigned the A-chains into two similar but not iden-tical homology groups. This can be exemplified by the identifiedpeptides with the masses of 1577 and 1591 Da, which are keymarkers for groups 1 and 2, respectively and showed a high degreeof homology. Together, these findings strongly support the hypoth-esis that fruit kernel riproximin consists of a mixture of at least twoisoenzymes.

The fact that after chemical deglycosylation of riproximin onlytwo bands remained does not exclude the presence of isoforms,since a similar electrophoretic mobility might hide subtle differ-ences in amino acid composition. Accordingly, the two riproximinisoforms, which could be partially separated by chromatographicmethods, differed explicitly within their A-chains, while they stillshowed very similar physico-chemical and biological properties.

Type II RIP isoforms are common in plants expressing this groupof proteins [4,14]. For many of them, gene families coding for differ-ent isoenzymes with various homology degrees have been described[15]. For ricin, the best investigated type II RIP, several isoformsincluding ricin D, ricin E and Ricinus communis agglutinin (RCA) havebeen characterized [18]. Ricin D and RCA show 84% identity withintheir B-chains and 93% within their A-chains [19]. Ricin E appearsto be a gene recombination product of ricin D and RCA [20]. The ri-cin/RCA gene family is assumed to be composed of 7–8 members,of which at least three are non-functional [21,22]. Three lectin iso-forms have been isolated from mistletoe (Viscum album), MLI, MLIIand MLIII, which differ in their MW and sugar specificity. Accord-ingly, three different genes have been described to encode these lec-tins [23,24]. For the Korean mistletoe (Viscum album coloratum)several cDNA isoforms have been amplified using a single primerset, indicating that heterogeneity of the mistletoe lectins is not solelydue to posttranslational modifications [25]. Himalayan mistletoe re-vealed four different protein isoforms, which have not yet been char-acterized on DNA level [26]. Several different isoforms have alsobeen described for the type II RIPs expressed by plants of the Sambu-cus genus. Sambucus nigra, for example, produces the three lectinsSNAI, SNAI’and SNAV (nigrin b), and the two lectin-related proteinsSNALRP1 and SNALRP2 [15,27].

Interestingly, expression of the isoforms can vary with the planttissue [7,28,29], maturation status [18,30] and season [31–33].Moreover, the genetic drift observed in plants from the same spe-cies from different continents has been shown to lead to theexpression of isoforms. For example, a new lectin ricin E has beendescribed in Ricinus communis adapting from the tropical to thetemperate zone [20].

The finding that the MS analysis did not prove identity betweenfruit kernel riproximin and the published riproximin sequence

H. Bayer et al. / Protein Expression and Purification 82 (2012) 97–105 105

should be considered within this context. Molecular biology anal-ysis demonstrated that the published riproximin sequence waspresent in the fruit kernel RNA pool. Extensive posttranslationalprocessing is, however, an intrinsic part of type II RIP protein syn-thesis and expected to account for some dissimilarity between theprotein- and cDNA-derived peptide maps. RIPs of type II are syn-thesized as precursors from a single gene and posttranslationallymodified by proteolysis. The amino acid sequence of the N- andC-terminal peptides of the mature A- and B-chains is thereforenot identical with the theoretic sequence obtained by translationfrom the precursor mRNA sequence. Moreover, glycosylation is ex-pected to interfere with the MS identification of the internal pep-tides that contain the respective sites.

However, apart from these considerations, the fact that the pub-lished sequence of riproximin could be confirmed at RNA but not atprotein level strongly suggests that not only various riproximingenes exist in X. americana, but also that the encoded, different iso-forms are translated at different efficacies, as common for othertype II RIP expressing plants.

For better understanding the genetic identity of the riproximinisoforms, deeper analyses including de novo protein sequencingsupported by advanced molecular biology methods need to beemployed.

In summary, X. americana fruit kernels were a defined, season-ally available and potentially unlimited herbal source for riproxi-min. The newly established purification procedure wasreproducible, supplied highly pure riproximin and is suitable forup-scaling. The resulting product, fruit kernel riproximin, consistsof a mixture of closely related isoforms, which are strongly relatedto the riproximin isoforms that had been characterized from theAfrican plant material. The riproximin isoforms possessed verysimilar physical and biological properties that render them difficultto separate.

Nevertheless, the appearance, biochemical and biological prop-erties of the purified riproximin isoform mixture were consistentover many purification batches. Therefore, riproximin purifiedbased on the established protocol can and will be used for furtherresearch and development.

Acknowledgments

Helene Bayer was funded by a grant from the Federal Ministryof Economics and Technology (Pro Inno II, KF0425101UL6). Wethank Dr. Bernd Heiss from GE Healthcare for technical adviceand kindly providing the prototype resin lactosyl–Sepharose.

References

[1] C. Voss, E. Eyol, M.R. Berger, Identification of potent anticancer activity inXimenia americana aqueous extracts used by African traditional medicine,Toxicol. Appl. Pharmacol. 211 (2006) 177–187.

[2] C. Voss, E. Eyol, M. Frank, C.W. von der Lieth, M.R. Berger, Identification andcharacterization of riproximin, a new type II ribosome-inactivating proteinwith antineoplastic activity from Ximenia americana, FASEB J. 20 (2006) 1194–1196.

[3] S. Olsnes, The history of ricin, abrin and related toxins, Toxicon 44 (2004) 361–370.

[4] F. Stirpe, Ribosome-inactivating proteins, Toxicon 44 (2004) 371–383.[5] Y. Endo, K. Tsurugi, The RNA N-glycosidase activity of ricin A-chain. The

characteristics of the enzymatic activity of ricin A-chain with ribosomes andwith rRNA, J. Biol. Chem. 263 (1988) 8735–8739.

[6] Y. Endo, A. Gluck, I.G. Wool, Ribosomal RNA identity elements for ricin A-chainrecognition and catalysis, J. Mol. Biol. 221 (1991) 193–207.

[7] F. Stirpe, M.G. Battelli, Ribosome-inactivating proteins: progress and problems,Cell Mol. Life Sci. 63 (2006) 1850–1866.

[8] C. Horrix, Z. Raviv, E. Flescher, C. Voss, M.R. Berger, Plant ribosome-inactivatingproteins type II induce the unfolded protein response in human cancer cells,Cell Mol. Life Sci. 68 (2011) 1269–1281.

[9] H. Zwierzina, L. Bergmann, H. Fiebig, S. Aamdal, P. Schoffski, K. Witthohn, H.Lentzen, The preclinical and clinical activity of aviscumine: A potentialanticancer drug, Eur. J. Cancer 47 (2011) 1450–1457.

[10] L. Herrera, B. Bostrom, L. Gore, E. Sandler, G. Lew, P.G. Schlegel, V. Aquino, V.Ghetie, E.S. Vitetta, J. Schindler, A phase 1 study of Combotox in pediatricpatients with refractory B-lineage acute lymphoblastic leukemia, J. Pediatr.Hematol. Oncol. 31 (2009) 936–941.

[11] P. Weingarten, P. Lutter, A. Wattenberg, M. Blueggel, S. Bailey, J. Klose, H.E.Meyer, C. Huels, Application of proteomics and protein analysis for biomarkerand target finding for immunotherapy, Methods Mol. Med. 109 (2005) 155–174.

[12] M.B. Eisen, P.T. Spellman, P.O. Brown, D. Botstein, Cluster analysis and displayof genome-wide expression patterns, Proc. Natl. Acad. Sci. USA 95 (1998)14863–14868.

[13] J.D. Robertus, A.F. Monzingo, The structure of ribosome inactivating proteins,Mini. Rev. Med. Chem. 4 (2004) 477–486.

[14] L. Barbieri, M.G. Battelli, F. Stirpe, Ribosome-inactivating proteins from plants,Biochim. Biophys. Acta 1154 (1993) 237–282.

[15] M.R. Hartley, J.M. Lord, Genetics of ribosome-inactivating proteins, Mini Rev.Med. Chem. 4 (2004) 487–492.

[16] M.R. Hartley, J.M. Lord, Cytotoxic ribosome-inactivating lectins from plants,Biochim. Biophys. Acta 1701 (2004) 1–14.

[17] Y. Kimura, S. Hase, Y. Kobayashi, Y. Kyogoku, T. Ikenaka, G. Funatsu, Structuresof sugar chains of ricin D, J. Biochem. 103 (1988) 944–949.

[18] J.M. Lord, L.M. Roberts, J.D. Robertus, Ricin: structure, mode of action, andsome current applications, FASEB J. 8 (1994) 201–208.

[19] L.M. Roberts, F.I. Lamb, D.J. Pappin, J.M. Lord, The primary sequence of Ricinuscommunis agglutinin, comparison with ricin, J. Biol. Chem. 260 (1985) 15682–15686.

[20] T. Araki, G. Funatsu, The complete amino acid sequence of the B-chain of ricinE isolated from small-grain castor bean seeds. Ricin E is a gene recombinationproduct of ricin D and Ricinus communis agglutinin, Biochim. Biophys. Acta911 (1987) 191–200.

[21] J. Leshin, M. Danielsen, J.J. Credle, A. Weeks, K.P. O’Connell, K. Dretchen,Characterization of ricin toxin family members from Ricinus communis,Toxicon 55 (2010) 658–661.

[22] J.W. Tregear, L.M. Roberts, The lectin gene family of Ricinus communis: cloningof a functional ricin gene and three lectin pseudogenes, Plant Mol. Biol. 18(1992) 515–525.

[23] H. Franz, P. Ziska, A. Kindt, Isolation and properties of three lectins frommistletoe (Viscum album L.), Biochem. J. 195 (1981) 481–484.

[24] A.G. Kourmanova, O.J. Soudarkina, S. Olsnes, J.V. Kozlov, Cloning andcharacterization of the genes encoding toxic lectins in mistletoe (Viscumalbum L), Eur. J. Biochem. 271 (2004) 2350–2360.

[25] C.H. Park, D.W. Lee, T.B. Kang, K.H. Lee, T.J. Yoon, J.B. Kim, M.S. Do, S.K. Song,CDNA cloning and sequence analysis of the lectin genes of the Koreanmistletoe (Viscum album coloratum), Mol. Cells 12 (2001) 215–220.

[26] V. Mishra, R.S. Sharma, S. Yadav, C.R. Babu, T.P. Singh, Purification andcharacterization of four isoforms of Himalayan mistletoe ribosome-inactivating protein from Viscum album having unique sugar affinity, Arch.Biochem. Biophys. 423 (2004) 288–301.

[27] I. Kozlov, O.I. Sudarkina, A.G. Kurmanova, Ribosome-inactivating lectins fromplants, Mol. Biol. (Mosk) 40 (2006) 711–723.

[28] A. Gasperi-Campani, L. Barbieri, E. Lorenzoni, L. Montanaro, S. Sperti, E. Bonetti,F. Stirpe, Modeccin, the toxin of Adenia digitata. Purification, toxicity andinhibition of protein synthesis in vitro, Biochem. J. 174 (1978) 491–496.

[29] F. Stirpe, L. Barbieri, A. Abbondanza, A.I. Falasca, A.N. Brown, K. Sandvig, S.Olsnes, A. Pihl, Properties of volkensin, a toxic lectin from Adenia volkensii, J.Biol. Chem. 260 (1985) 14589–14595.

[30] R.J. Youle, A.H. Huang, Protein bodies from the endosperm of castor bean:subfractionation, protein components, lectins, and changes during germination,Plant Physiol. 58 (1976) 703–709.

[31] T. Girbes, J.M. Ferreras, F.J. Arias, R. Munoz, R. Iglesias, P. Jimenez, M.A. Rojo, Y.Arias, Y. Perez, J. Benitez, D. Sanchez, M.J. Gayoso, Non-toxic type 2 ribosome-inactivating proteins (RIPs) from Sambucus: occurrence, cellular andmolecular activities and potential uses, Cell Mol. Biol. (Noisy. -le-grand) 49(2003) 537–545.

[32] M.A. Rojo, L. Citores, F.J. Arias, J.M. Ferreras, P. Jimenez, T. Girbes, CDNAmolecular cloning and seasonal accumulation of an ebulin l-related dimericlectin of dwarf elder (Sambucus ebulus L) leaves, Int. J. Biochem. Cell Biol. 35(2003) 1061–1065.

[33] M. Nsimba-Lubaki, W.J. Peumans, Seasonal fluctuations of lectins in barks ofelderberry (Sambucus nigra) and black locust (Robinia pseudoacacia), PlantPhysiol. 80 (1986) 747–751.