| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* German Cancer Research Center, Toxicology and Chemotherapy Unit, E100, Heidelberg, Germany; and
German Cancer Research Center, Central Spectroscopic Department, B090, Heidelberg, Germany
1Correspondence: Toxicology and Chemotherapy Unit, E100, Deutsches Krebsforschungszentrum Heidelberg, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany. E-mail: m.berger{at}dkfz.de
ABSTRACT
The aim of this study was to identify and characterize the active component(s) of Ximenia americana plant material used to treat cancer in African traditional medicine. By a combination of preextraction, extraction, ion exchange and affinity chromatography, a mixture of two cytotoxic proteins was isolated. Using degenerated primers designed on the de novo sequence of two tryptic peptides from one of these proteins, a DNA fragment was amplified and the sequence obtained was used to determine the complete cDNA sequence by the RACE method. Sequence analysis and molecular modeling showed that the new protein, riproximin, belongs to the family of type II ribosome inactivating proteins. These results are in good agreement with the ability of riproximin to inhibit protein synthesis in a cell-free system, as well as with the cytotoxicity of riproximin, as demonstrated by its IC50 value of 0.5 pM in MCF7, 1.1 pM in HELA and 0.6 pM in CC531-lacZ cells. To assess the antineoplastic efficacy of the purified riproximin in vivo, the CC531-lacZ colorectal cancer rat metastasis model was used. Significant anticancer activity was found after administration of total dosages of 100 (perorally) and 10 (intraperitoneally) pmol riproximin/kg. These results suggest that riproximin has distinct potential for cancer treatment. —Voss, C., Eyol, E., Frank, M., von der Lieth, C.-W., Berger, M. R. Identification and characterization of riproximin, a new type II ribosome-inactivating protein with antineoplastic activity from Ximenia americana.
Key Words: anticancer agent • plant protein • lectin • peroral activity • liver metastasis model
POWDERED MATERIAL used in African traditional medicine has been shown recently to exert highly potent anticancer activity in vitro and in vivo (1)
. The source plant of this material was determined to be the semiparasitic subtropical plant Ximenia americana. Physical and chemical characterization of the active component(s) strongly hinted to protein(s) with galactose affinity. An initial mass spectrometry analysis of the active protein components identified a peptide identical to a tryptic peptide from the ribosome-inactivating protein ricin.
Ricin belongs to the family of type II ribosome-inactivating plant proteins (RIPs) consisting of two polypeptide chains termed A- and B-chain, which are hold together by a disulfide bridge (2
, 3)
. The A-chain of type II RIPs is an rRNA N-glycosidase able to hydrolyze a specific adenine from the ricin/sarcin loop of the ribosome large subunit. The B-chain is a lectin which shows affinity for certain sugar moieties, mainly for galactose. Based on their toxicity to mammals, this family is divided into two groups: the toxic and nontoxic type II RIPs. The former group includes, e.g., ricin, abrin, viscum lectin I, and volkesin, which are among the most potent plant toxins. Conversely, Ricinus agglutinin (RCA), Sambucus RIPs, and cinnamomins belonging to the latter group show little or no toxicity in higher animals (4)
. The toxic effects of type II RIPs are based on a mechanism involving the cellular uptake of the protein mediated by the binding of the B-chain to sugar moieties on the cell surface, followed by an internalization of the A-chain, which inactivates the ribosomes and thus terminates protein synthesis (5)
.
Interest in type II RIPs as anticancer agents rose as early as 1970, when Lin et al. showed that ricin and abrin were more toxic to tumor than to normal cells (6)
. However, ricin’s high unspecific toxicity prevented a clinical use. To improve the selectivity, ricin or ricin A-chain were conjugated to antibodies raised against specific tumor antigens (7
, 8)
. Another type II RIP, which recently has been developed for application in cancer treatment is rViscumin (aviscumine), a recombinantly produced viscum lectin I, is currently tested in phase I/II clinical studies (9
, 10)
.
Here we report the identification and characterization of a new type II RIP from Ximenia americana. This new RIP, termed riproximin, is the active component of the plant material used in African traditional medicine to treat some forms of cancer.
MATERIALS AND METHODS
Materials
Plant material was obtained from Tanzania (1)
. MTT was obtained from Serva (Heidelberg, Germany), culture media from Invitrogen (Karlsruhe, Germany). Sepharose was partially hydrolyzed to obtain free galactose ends as described in ref 11
. In brief, Sepharose 4B (Amersham GE Healthcare, Freiburg, Germany) was incubated with 1M HCl for 3 h at 50°C, washed with distilled water and equilibrated with transfer buffer (see below).
Purification
Dry plant material was extracted three times with 70% acetone followed by one extraction with 100% acetone to deplete tannins. The residue was dried by air stream and extracted with extraction buffer (20 mM Tris-HCl, pH=7.0).
The extract was applied on a diethylaminoethyl (DEAE) -cellulose (Whatman, Maidstone, UK) column pre-equilibrated with extraction buffer and the column washed with extraction buffer until no proteins could be detected in the eluate. For eluting all biologically active components, 500 mM NaCl were added to the extraction buffer. Alternatively, a step gradient elution was performed with extraction buffer containing increasing NaCl concentrations (100, 200, 300, and 400 mM NaCl, respectively).
Eluted fractions containing the biological activity were applied on a column containing partially hydrolyzed Sepharose. The column was washed with transfer buffer until the eluate was free of proteins. The bound proteins were eluted with elution buffer (20 mM Tris-HCl, 500 mM NaCl, 100 mM galactose, pH=7.0). The protein fraction was collected and concentrated by ultrafiltration on a Microcon YM10 (MW cutoff 10 kDa) membrane (Millipore, Schwalbach, Germany). Where possible, the dry protein weight was determined by replacing the elution buffer with water and subsequent lyophilization. In addition, protein molar concentrations were calculated from the respective 280 nM absorbance (A280 nM) of the purified samples by using a molecular absorption coefficient which was estimated based on the riproximin sequence (12)
. Purification degree and yield were monitored by SDS-PAGE and in vitro tests for cytotoxic activity. To analyze the purified fractions, SDS-PAGE was performed under reducing (addition of dithiothreitrol to the sample in order to reduce the disulfide bonds) or nonreducing (without dithiothreitrol, keeping the disulfide bonds intact) conditions. The purified proteins were analyzed by mass spectrometry as described below.
Deglycosylation assay
The affinity-purified proteins were heated for 5 min to 99°C in a buffer containing 50 mM sodium phosphate buffer, pH = 7.0, 0.1% SDS, and 0.05% ß-mercaptoethanol. To obtain a better visualization of the effects of deglycosylation on each of the four proteins, the bands separated under reducing conditions by SDS-PAGE were cut out after visualization by Coomassie blue staining and heated as described above. After cooling to room temperature, Triton X-100 and N-glycosidase F (Calbiochem, Merck Biosciences GmbH, Bad Soden, Germany) were added to a final concentration of 0.75% and 100 U/ml, respectively. The reaction mix was incubated over night at 37°C. The deglycosylated proteins were analyzed by SDS-PAGE under reducing conditions.
Cytotoxic activity
In brief, MCF7 human breast carcinoma, HELA cervix carcinoma cells or CC531-lacZ rat colon carcinoma cells were seeded into 96-well microtiter plates at an initial cell density of 8000 (MCF7) or 4000 cells/well (HELA; CC531-lacZ). On the next day, medium with (treated) or without (control) extract or purified fractions was added to the wells. The cells were incubated for three more days in an incubator under standard culture conditions (humidified atmosphere, 37°C and 5% CO2 in air). The viable cell count was measured by the MTT (3-(4,5-dimethyldiazol-2-yl)-2,5 diphenyl tetrazolium bromide) assay as described in (13)
with some modifications. In brief, 0.1 x vol MTT solution (10 mg/ml in PBS) was added to each well at the end of the incubation period except for the blank. After 3 h incubation the medium containing the excess MTT was removed. The formazan crystals were dissolved in 200 µl isopropanol containing 0.4 M HCl. Extinctions were measured by a microplate reader at 540 nM (reference 690 nM).
Inhibition of protein synthesis
The efficacy in inhibiting the protein synthesis was determined for both native and reduced riproximin (riproximin treated with 1% ß-mercaptoethanol for 1h at 37°C, to dissociate the A- and B-chains), obtained by affinity purification, as described above.
A nonradioactive method based on the TNT Quick Coupled Transcription/Translation System (Promega, Mannheim, Germany) was used to determine the protein synthesis-inhibiting activity (14)
. Equal amounts (8x28.5 µl) of a transcription/translation mix (240 µl TNT Quick Master Mix, 6 µl 1 mM methionine, 12 µl T7 luciferase control DNA, and 27 µl H2O) were distributed into separate tubes (two control and six inhibition reactions) and incubated at 30°C. After 5, 7, 9, 11, 13, and 15 min incubation, 2 µl samples of the first control tube were diluted in 80 µl bioluminescence buffer (20 mM Tricin, 0.05% (w/v) BSA, pH=7.8) and immediately frozen in liquid nitrogen to stop the reaction. Thereafter, 1.5 µl Tris-HCl buffer (20 mM Tris-HCl, pH=7.0, control) or Tris-HCl buffer containing appropriate riproximin dilutions (inhibiton reactions) were added to the reaction mix in the remaining seven tubes and further incubated at 30°C. Samples (2 µl) were collected from each reaction at 18, 20, 22, 24, 26, 28, and 30 min, respectively, and treated as described above. The relative luciferase content of the samples was determined with the Luciferase Assay System (Promega, Mannheim, Germany) on a MicroLumat Plus luminometer (Berthold Technologies, Bad Wildbach, Germany). Five microliters of each diluted sample were mixed with 50 µl luciferase reagent and the chemoluminescence (cpm) was measured for 10 s after an initial delay of 2 s. For each reaction, the absolute luminescence count of the 18 min sample was subtracted from that of the 24 min sample (values in sqrt), yielding the relative amount of luciferase generated within this interval. The results were calculated as percent of the luciferase amount generated in the control reaction (100%) and plotted against the riproximin concentration.
In vivo efficacy
For determining the effect of the purified protein (riproximin) in a rat liver metastasis model, the rat colon cancer cell line CC531-lacZ was used (15
,16)
. In short, 4 x 106 CC531-lacZ cells were implanted via the portal vein on day 0 into male Wag/Rij rats (Charles-River, Sulzfeld, Germany). Tumor-bearing rats were treated perorally (by gavage) or intraperitoneally (i.p.) with the aqueous extract starting on day 1, as shown in Table 1
. Three weeks later (day 21) the experiment was terminated, the liver of the animals was removed, weighed, and kept at –80°C until analysis. The number of tumor cells per liver was determined by the ß-galactosidase assay (Applied Biosystems, Darmstadt, Germany) as described earlier (15)
, by comparing to a standard curve established with a mixture of healthy liver tissue and rising numbers of tumor cells.
|
Mass spectroscopy
After nonreducing SDS-PAGE separation of the affinity-purified protein mixture and staining with Coomassie blue, the upper protein band was cut out from a nonreducing gel and an in-gel digestion was performed with trypsin as described in (17)
. The eluted tryptic peptides were analyzed by electrosprayionisation mass spectrometry (MS-MS) on a hybrid Q-TOF mass spectrometer type Q-Tof2 (Waters Micromass, Manchester). For the obtained mass spectra, a database search was performed using the program Mascot search from Matrix science (18)
. Additional de novo peptide sequences were obtained from the MS-MS spectra analysis as described in (19)
(courtesy of Dr. W. Lehmann).
Cloning and sequencing
RNA was isolated from fresh Ximenia americana leaves and cDNA synthesized. A schematic overview of the sequencing procedure is given in Fig. 1
. Degenerated primers (deg-sense and deg-antisense, Table 2
) were designed from peptide sequences obtained by mass spectrometry and a polymerase chain reaction (PCR) amplification was performed (Fig. 1a
). The amplified fragment (
400 bp) was sequenced and gene-specific primers were designed (Table 2)
. The specific antisense primers p31rew3Xa and p31rew4Xa (nested) were used together with the degenerated primer RMLA2-frw (20)
to 5' extend the sequence (Fig. 1b
). Amplification of the 5' and 3' cDNA ends was performed by the RACE method using the SMART-RACE kit from BD Biosciences (Heidelberg, Germany) as shown in Fig. 1c
. The sequence was confirmed by amplification with a proofread polymerase and subsequent sequencing. Sequencing of the PCR-amplified DNA fragments was performed on an ABI PRISM 310 Genetic Analyzer (Applied Biosystems, Darmstadt, Germany) as described by the manufacturer.
|
|
For N-terminal sequencing by Edman degradation, the affinity-purified proteins were separated by SDS-PAGE under reducing conditions, blotted onto a PDVF membrane and visualized with Coomassie blue. Each of the four protein bands was cut out from the membrane and sequenced on a Procise 494 protein sequencer (Applied Biosystems, Darmstadt, Germany, courtesy of Dr. H. Heid).
An independent mass spectrometric investigation was performed for each of the four protein bands separated by reducing SDS-PAGE. The Coomassie blue-stained protein bands were submitted for ESI MS-MS analysis to the Proteome factory, (Berlin, Germany).
Sequence analysis
For sequence analysis, programs from the HUSAR package (DKFZ, Heidelberg, Germany) were used. Database search with newly obtained peptide/DNA sequences was performed with the program FASTA. The respective DNA sequences were assembled with the program CAP. The resulting protein sequence was aligned with other RIP sequences with the program CLUSTAL. Homology and identity to known RIP-sequences were calculated with the program GAP.
Homology modeling of the riproximin 3D structure
Based on the alignment of the riproximin precursor sequence with known type II RIP-sequences, the putative sequences for the A and B riproximin chains were chosen and a FASTA-search of the PDB-database (www.pdb.com) was performed. For modeling the riproximin A-chain, the structure of the recombinant ricin R213D A-chain was chosen (PDB accession no. 1UQ4; 21
). Modeling of the riproximin B-chain was based on the structure of the mistletoe lectin I B-chain, refined to 2.05 (PDB accession no. 1SZ6). The A- and B-chain models were calculated following the approach of comparative protein modeling on the SWISS-MODEL Automated Protein Modeling Server (22)
. The quality of the models was checked with the programs WHAT_CHECK (23)
as well as PROCHECK (24)
.
Galactose docking simulation
A "blind docking" approach using AUTODOCK 3.05 (25)
was applied to screen the surface of the B-chain for potential ß-Gal binding sites. The various files required as input for AUTODOCK were created with the help of "AutoDockTools" (http://www.scripps.edu/sanner/python/adt/) and the Conformational Analysis Tools (CAT) program (http://www.md-simulations.de/cat/). The genetic algorithm with local search option (GA-LS) as implemented in AUTODOCK was used to dock the flexible ligand. Two docking strategies where employed: one using the rigid homology model as target for the docking and a second introducing protein flexibility into the docking protocol by generating an ensemble of protein conformations using molecular dynamics (MD) simulation with the homology model as starting structure. For the MD simulation the GROMACS software (www.gromacs.org) was used (GROMOS 96 force field, 300 K, 1 ns simulation time, integration step 2 fs, particle-mesh Ewald electrostatics, explicit water molecules as solvent, ions added to counterbalance charges of protein). For the rigid protein 100 AUTODOCK jobs were started each performing 256 GA-LS runs giving rise to 20510 docked ß-Gal structures. The docking protocol with flexible protein implied 1000 starting structures derived from the MD simulation and 20 GA-LS runs were performed for each of them resulting in 27631 docked solutions. CAT was used to merge the output data of the AUTODOCK runs to perform the analysis of the entire dataset and organize the results in such a way that areas on the protein surface exhibiting a strong binding affinity can be easily visualized using standard display programs.
RESULTS
Purification
Purification of riproximin was performed according to a four-step procedure. Preextraction of the raw material with 70% acetone reduced the concentration of tannins in the subsequent water extract below detection limit (1)
. Reducing SDS-PAGE of this extract as well as of the subsequently obtained one-step DEAE eluate (500 mM NaCl) revealed a complex pattern of protein bands which was overlaid by a Schiff’s-reagent-positive smear (Fig. 2
a, lanes 1 and 2). Following affinity chromatography, the biologically active fraction showed only four bands with apparent MW of 34, 31, 29 and 27 kDa, respectively, in a reducing SDS-PAGE gel (Fig. 2a
, lane 3). Nonreducing SDS-PAGE revealed the presence of only two proteins with apparent MW of 56 and 51 kDa, respectively (see ref. 1
), implying that the affinity-purified sample consists of two heterodimeric proteins. In a typical preparation with one-step elution from DEAE-cellulose, 90 µg affinity-purified proteins with cytotoxic activity (see below) were obtained from 20g raw material. This affinity-purified protein fraction (shown in Fig. 2a
, lane 3) was used for all subsequent experiments. One of the two proteins, for which the cDNA sequence was determined, was termed "riproximin."
|
When a stepwise elution from the DEAE-cellulose with increasing NaCl concentrations was performed, only the fractions following 200 and 400 mM NaCl showed detectable cytotoxic activity. These fractions showed a very complex band pattern with only slight differences when analyzed by reducing or nonreducing SDS-PAGE (data not shown). When the 200 mM NaCl and the 400 mM NaCl DEAE fractions were subsequently processed by affinity chromatography, the purified samples contained different ratios of the four (reducing SDS-PAGE, Fig. 2b
) or two bands (nonreducing SDS-PAGE, Fig. 2c
), respectively. The 34 and 29 kDa bands (reducing SDS-PAGE, Fig. 2b
, lane 1) and the 56 kDa band (nonreducing SDS-PAGE, Fig. 2c
, lane 1) prevailed in the affinity-purified sample originating from the 200 mM NaCl DEAE fraction, while the 31 and 27 (reducing SDS-PAGE, Fig. 2b
, lane 2) and the 51 kDa band (nonreducing SDS-PAGE, Fig. 2c
lane 2) prevailed in the affinity-purified sample originating from the 400 mM NaCl DEAE fraction. Thus it became obvious that the protein with the apparent MW of 56 kDa consists of two subunits of 34 and 29 kDa, whereas the 31 and 27 kDa bands are subunits of the smaller protein (apparent MW 51 kDa). Since both fractions showed comparable cytotoxic activity (see below), it can be assumed that both proteins are cytotoxic.
Deglycosylation experiments
On treatment with N-glycosidase F, there was a shift of the four bands toward lower MW. To better define the effect on each subunit, they were treated individually after excision from gel. This experiment showed that the 34 and 31 kDa protein bands were shifted to an apparent MW of 30 and 29 kDa, respectively (Fig. 2d
, lanes 4 and 5), indicating an extensive N-glycosylation of the original protein subunit. Conversely, the subunits with the apparent MW of 29 and 27 kDa are probably not N-glycosylated, since no MW shift was observed upon treatment with N-glycosidase F (Fig. 2d
, lanes 2 and 3).
Cytotoxicity
The antiproliferative activity of the affinity-purified protein fractions was determined in MCF7 human breast cancer, HELA cervix carcinoma and CC531-lacZ rat colon cancer cells (Fig. 3
).
|
The affinity-purified sample obtained after a one-step DEAE-cellulose elution with 500 mM NaCl showed a distinct cytotoxic effect in all three cell lines (Fig. 3a
), as evidenced by IC50 values of 0.5 pM (MCF7), 1.1 pM (HELA) and 0.6 pM (CC531-lacZ). Despite the very similar IC50, the dose-response curves of the three cell lines differed in their slope, as reflected by their IC90/IC10 ratios of 83 (MCF7), 6.6 (HELA), and 20 (CC531-lacZ).
When the total biological activity was separated by step gradient DEAE elution and the activity of the 200 and 400 mM NaCl fractions tested after affinity purification, comparable cytotoxic effects were observed in MCF7 cells. The concentration of the two fractions was below the detection limit (A280nm<0.02). However, since the SDS gel lanes showed a comparable protein content (Fig. 2c, d
), the respective volumes were used for a dilution series. IC50 values in MCF7 cells were achieved by a 30,000-fold dilution of the affinity-purified sample originating from the 400 mM NaCl fraction, and by a 15,000-fold dilution of that from the 200 mM NaCl fraction (Fig. 3b
). Given that the detection limit (A280=0.02) corresponds to a concentration of 0.1 µM, the IC50 values derived from the respective dilutions are below 3.3 and 6.6 pM for the affinity-purified samples originating from the 400 and 200 mM NaCl fractions, respectively. These values relate well to the IC50 value measured for the complete protein fraction obtained by affinity purification of the 500 mM NaCl fraction (0.5 pM in MCF7 cells, see above).
Inhibition of protein synthesis
The affinity-purified protein sample clearly inhibited the synthesis of luciferase in an in vitro reticulocyte lysate transcription/translation assay (Fig. 4
). A clear concentration-dependent translation inhibition was seen in response to 0.17 to 50 nM nonreduced or reduced protein (Fig. 4a, b
), reflecting the activities of the heterodimeric protein and the separated A-chain, respectively. For calculating the relative luciferase amounts synthesized in each reaction, the linear phase of the kinetics was chosen (18–24 min). The IC50 for translation-inhibition was achieved at a riproximin concentration of 5.5 nM (nonreduced) or 2.6 nM (reduced), (Fig. 4c
).
|
In vivo experiments
As shown in Table 1
, the affinity-purified protein sample obtained after a one-step DEAE-cellulose elution was administered to male tumor-bearing Wag Rij rats via the peroral and the i.p. route. No toxicity was observed after total dosages of i.p. up to 10 pmol/kg and p.o. up to 100 pmol/kg. Intraperitoneal administration of 0.25, 0.5, and 1 pmol/kg every second day dose-dependently reduced the increase in tumor cell number seen in untreated tumor-bearing rats as indicated by T/C % (treated/controlx100) ratios of 47.4, 34.9, and 22.6 (P<0.05), respectively (Table 1)
. The effect on the liver wt was less prominent, with T/C % ratios of 57.5, 50.4, and 51.6, respectively. Peroral administration of 10 pmol/kg every second day was surprisingly effective: compared wtih the untreated controls, the mean liver wt of the treated animals was significantly lower (T/C %=30.5, P<0.05, Table 1
) and their mean tumor cell number was reduced by two orders of magnitude (T/C %=1.0, P<0.05, Table 1
).
Identification of the affinity-purified protein and cDNA sequencing
Two peptides showing homology to B-chain sequences of type II ribosome-inactivating proteins (RIPs) were identified by database search and de novo sequencing, respectively. Aligning of these peptides allowed the design of degenerated primers and the amplification of a ca. 400 bp DNA fragment containing a continuous open reading frame (ORF). The translation of this frame resulted in a protein sequence showing high homology to the B-chain of known type II RIPs. The complete sequence of the respective cDNA (1990 bp), containing an ORF of 1814 bp, was obtained as described under Materials and Methods and as shown in Fig. 1
. The sequence preceding the frame-start codon ATG (pos. 177–179) contains several stop-codons in frame with the ORF, as common for 5' untranslated regions. At 28 positions of the cDNA sequence single nucleotide polymorphisms (SNPs) were identified, 10 of which lead to the translation of different amino acids.
The homology of the translated protein sequence to known type II RIP precursor protein sequences demonstrates that the new protein termed "riproximin" is a so far unknown member of this class. The riproximin cDNA sequence was submitted to the EMBL database (accession no. AM114537).
A second MS-MS analysis for each of the four protein subunits separated by SDS-PAGE under nonreducing conditions resulted in information on tryptic peptides that are identical or highly homologous with the cDNA-derived riproximin sequence. Two tryptic peptides, CLSTSFGR and SNTDANQLWILK, were obtained from the 31 kDa band; both are identical to stretches of the cDNA-derived riproximin sequence (amino acids 396 – 403 and 374 – 385, respectively). The protein band with an apparent MW of 34 kDa revealed a peptide with the sequence SNTDANQLWTLK, which differs by a single amino acid (T at position 10 of the peptide, shown in bold) from the cDNA-derived sequence (amino acids 396 – 403 with I at position 383). It can be therefore assumed that the protein subunits with apparent MW of 34 and 31 kDa are B-chains of the type II RIP molecules riproximin and of a so far unknown homologue. For the respective A-chains, the N-terminal sequences were obtained by Edman degradation (see below).
Figure 5
shows an alignment of the riproximin sequence to the database sequences of ricin (SwissProt: riciricco, P02879) and viscum lectin I (SwissProt: ml1_visal, P81446). At the positions affected by the cDNA SNPs, both amino acids were specified. Two of the 10 amino acid variations are contained in the N-terminal signal peptide, six in the sequence of the A-chain and two in that of the B-chain.
|
Sequence analysis
The sequence of the riproximin precursor protein shares 55% similarity and 47% identity with that of the ricin precursor as well as 53% similarity and 45% identity with the sequence of the viscum lectin I precursor, both of which belong to the subgroup of toxic type II RIPs. On the other hand, the sequence homology to the nontoxic Sambucus type II RIPs ebulin 1 (EMBL: SEB400822, AJ400822) and nigrin b (Swiss-Prot: NIGRB_SAMNI, P33183) was lower (48% and 45% similarity as well as 40% and 37% identity, respectively). These homology data suggest that riproximin is a toxic member of the type II RIP family.
N-terminal blockade prevented determining the N-terminal sequence of the B-chains (apparent MW 34 and 31 kDa) by Edman degradation. The A-chains also showed a high extent of N-terminal blocking. However, by increasing the amount of protein, both of the A-chains (apparent MW=27 and 29 kDa, respectively) revealed the N-terminal sequence DYP, a sequence found at position 57–59 of the cloned riproximin precursor. The probable C-terminal end of the A- and the beginning of the B-chain were estimated based on homology to other type two RIPs including ricin, viscum lectin I, abrin, cinnamomin, ebulin, and nigrin, as described below. The sequence of the presumed A- and B-chains is indicated in Fig. 5
. However, since the exact positions of the A-chain C-terminus and B-chain NH2 terminus have not yet been experimentally verified, the amino acid numbering was based on the sequence of the precursor protein. The MW calculated for the riproximin A- and B-chains varied between 29 and 31 kDa, pending on the exact A-chain C-terminus and B-chain NH2 terminus positions and the allelic variation.
The presumed A-chain shows 49% similarity and 42–43% identity to that of viscum lectin I, 49–50% similarity and 41–42% identity to the ricin A-chain, and 42% similarity and 33–35% identity to that of nigrin b. The invariant residues within the active site are all conserved within the sequence of the riproximin A-chain: tyr73, arg81, tyr133, tyr172, glu230, arg233 and trp267 correspond to ricin tyr21, arg29, tyr80, tyr123, glu177, arg180, and trp211. In addition, the conserved C-terminal cysteine residue (cys259 in the A-chain of ricin) which is involved in the intermolecular disulfide bonding corresponds to cys317 in riproximin and signals the end of the riproximin A-chain. A second cysteine-residue within the A-chain, cys224, corresponds to cys171 in the A-chain of ricin but is not present in the viscum lectin I A-chain. While the ricin A-chain contains two glycosylation sites, the only potential glycosylation site of the riproximin A-chain was identified within the sequence of one of the alleles (asn159), and is replaced by thr159 in the other allele.
The riproximin B-chain shares 65% and 57–58% similarity as well as 57–58% and 49–50% identity with the B-chains of the toxic RIPs ricin and viscum lectin I, respectively. The homology to the nontoxic RIPs ebulin 1 and nigrin b is lower, as shown by 53% and 51% similarity as well as 45–46% and 43–44% identity, respectively. The eight cysteine residues known to be involved in four conserved intramolecular disulfide-bridges in the type II RIP B-chains (cys20–cys39, cys63–cys80, cys151–cys164, cys190–cys207 in ricin), as well as the B-chain N-terminal cysteine involved in the formation of the intermolecular disulfide-bridge between the A- and B-chains (cys4 in ricin), are all found in the riproximin B-chain as cys353–cys372, cys396–cys415, cys487–cys500, cys526 – cys546 (intramolecular bridges) and cys337 (N-terminal, binding A-chain cys317). In addition, the B-chain of riproximin contains another cysteine residue (cys529) aligning to cys194 of the viscum lectin I B-chain. All highly conserved residues previously reported to be involved in building the hydrophobic core of the B-chain are conserved or replaced by similar hydrophobic amino acids. The riproximin B-chain contains two potential glycosylation sites, asn430 and asn470, which correspond to highly conserved glycosylation sites within the B-chain sequence of other type II RIPs (asn95 and asn135 in the B-chain of ricin).
A comparison of the two sugar affinity domains found in the B-chains of type II RIPs shows that most of the amino acid residues involved in sugar binding are conserved in the B-chain of riproximin. The 1
subdomain amino acid residues asp22, trp37, asn46 and gln47 of the ricin B-chain correspond to asp355, trp370, asn379, and gln380 in the 1 subdomain of the riproximin B-chain, whereas gln35 in ricin is replaced by ile368 in riproximin. The 2
subdomain amino acid residues asp234, asn255 and gln256 of ricin correspond to asp573, asn594, and gln595 in the 2 subdomain of riproximin, whereas ile246 and tyr248 in ricin are replaced by leu585 and trp587 in riproximin, respectively.
Homology modeling
According to the PROCHECK* analysis (Ramachandran plot), 88.6% of the A-chain-amino acids were in favored, 11.0% in allowed and only 0.4% (arg155) in disallowed conformation. For the B-chain, 80.7% and 18.5% of the amino acids were in favored or allowed conformation, respectively, and only 0.8% (tyr413, ile417) was found within the disallowed area. The overall average G factor (ideally > –0.5) was calculated to be –0.03 for the A chain model and as –0.15 for the B chain model, respectively.
The 3-dimensional model of riproximin (Fig. 6
) clearly visualizes the A- and B-chains as well as the disulfide bridge linking the chains. The amino acid residues predicted to be involved in the ribonuclease activity of the A-chain cluster within a cleft on the surface and probably constitute the active catalytic site. The B-chain is characterized by a two-domain structure, each of them containing a potential sugar binding site. The location of eight cysteine residues allows the formation of the four predicted intramolecular disulfide bridges within the B-chain. The three potential glycosylation sites of riproximin are located on the surface of the molecule and thus are probably glycosylated in the mature protein. Except for asn159, which is the only glycosylation site on the A-chain, none of the allelic variations of riproximin shows a significant effect on the 3-dimensional structure, glycosylation, disulfide bridge formation, catalytic site, or sugar binding domains.
|
Galactose docking simulation
The blind docking docking calculation using the rigid homology model as target for docking resulted in three binding sites with maximum energy (Fig. 7
). Two of the binding sites found are analogous to those found for the mistletoe lectin I structure (next to trp370 and trp587, corresponding to the 1 and 2 subdomains, respectively). A third docking site of similar affinity was located close to gln583. A binding site found recently in the 1ß subdomain of Himalayan mistletoe RIP (26)
was predicted for riproximin from the docking as well (next to phe401 and tyr413) but with slightly lower affinity. Surprisingly, the analysis of the docking calculation with a flexible protein showed that the binding affinity of the trp370 binding site (1 subdomain) is significantly reduced compared with trp587 (2 subdomain). This seems not to be an artifact of the method used, since an application of the same protocol using the mistletoe lectin I structure (PDB code 1OQL) with an ile368 mutation resulted in a comparable affinity for both binding sites (data not shown).
|
DISCUSSION
A mixture of two new type II ribosome-inactivating proteins was identified to be the active principle of the antineoplastic activity contained in Ximenia americana plant material.
The initial characterization of an aqueous Ximenia americana extract had hinted at a ribosome-inactivating protein (1)
. Subsequently, a fraction containing two proteins could be purified, with both proteins showing cytotoxic activity.
Alignment of de novo sequenced peptides obtained by tryptic digestion of one of these proteins allowed the design of degenerated primers, which amplified 400 bp of a corresponding gene. This fragment in turn was starting point for sequencing the complete cDNA. The translated protein sequence was found to be homologous to type II RIP precursor sequences and thus confirmed the protein, which we termed riproximin, to be a member of this family.
All known type II RIPs are encoded by a single gene and post-translationally processed into the typical two-chain structure by cleaving off an N-terminal signal and a linker peptide. Moreover, many RIP expressing plants contain more than one RIP precursor gene and thus contain mixtures of RIP isoenzymes (27)
. Similarly, the Ximenia americana extract analyzed here contains a mixture of two homologous proteins, most probably type II RIP isoenzymes.
This conclusion is derived from the biological activity of each of the two proteins as well as from MS-MS and N-terminal sequence analysis, showing the presence of two B-chains and two A-chains in the mixture. Notably, the protein sequence information obtained from one of the B-chains as well as from the A-chain N-terminal sequences is identical to the cDNA derived sequence stretches of riproximin. MS-MS sequences derived from the other B-chain show high homology but no identity to the riproximin sequence, suggesting that the differences between the two homologous proteins are related to a different primary structure and not merely the result of differences in glycosylation. This is also supported by the fact that the bands show different MW even after deglycosylation. It is also unlikely that the differences observed in MW result from other types of posttranslational processing, since no other modifications have been described for type II RIPs so far. The observed cDNA allelic variation can not account for the presence of two discernable proteins in the mixture, since MW-differences resulting from the exchange of six (A-chain) or two (B-chain) amino acids are too low to be detected by SDS-PAGE separation.
On comparison with the sequences of ricin or viscum lectin I as well as after modeling the 3D structure of the riproximin A- and B-chains it became obvious that none but one of these amino acids affected by single nucleotide polymorphism is involved in the function or structure of the protein. The only exception is the asn/thr variation at position 159, which, when expressing asn159, could serve as a glycosylation site on the riproximin A-chain. Because of its singularity, the presence of this glycosylation site could contribute to a major difference between the A-chains alleles.
All of the invariant amino acids involved in the N-glycosidase activity of a RIP A-chain are conserved and cluster in the 3D model within a cleft likely to be the active site of the riproximin A-chain. It can be therefore assumed that the A-chain of riproximin is a fully active RNA N-glycosidase. Similar to other type II RIP A-chains, the A-chain of riproximin has very low lysine content (two of 263 amino acid residues). The low lysine content has been shown to be involved in the RIP-cytotoxicity, since it helps the RIP A-chain to escape ubiquitination and subsequent degradation on relocation to the cytosol (28
, 29)
.
The B-chain of type II RIPs consists of two homologous domains termed 1 and 2, each containing three subdomains termed 1, 1ß, 1 and 2, 2ß, 2 (30)
. The subdomains 1 and 2 are reported to be responsible for the sugar affinity of the B-chains of ricin (31)
, viscum lectin I (32)
and other type II RIPs (3)
. The B-chain of riproximin also shows this typical structure, including the conserved disulfide bridges and glycosylation sites. In addition, most of the amino acids which are responsible for sugar binding in ricin and viscum lectin I, are also present in the respective riproximin B-chain subdomains.
The cytotoxic effects of riproximin in vitro are characterized by a low IC50, which is comparable to that of the toxic RIP ricin (33)
, and a relatively steep, linear dose-response curve. The low IC50 is in line with classifying riproximin into the subgroup of toxic type II RIPs, as assumed from sequence homology. The IC50 for translation-inhibition in a cell-free system was higher for the heterodimer than for the A-chain separated by reducing the intermolecular disulfide bridge. This is typical for a type II RIP since the RNA-glycosydase activity of the A-chain is inhibited to some degree as long as the B-chain is attached. Moreover, the ratio is similar to values reported e.g., for Sambucus nigra or Iris hybrid type II RIPs (4)
.
Experiments with crude extract had shown that the in vitro IC50 in CC531-lacZ cells (3.3 µg/ml) corresponded to an in vivo effective dose of 5 mg/kg after i.p. therapy. Equivalent anticancer activity was seen after peroral administration of 100 mg/kg (1)
. By extrapolating, a similar relationship of pharmacodynamic activities was assumed for the affinity-purified protein sample. Therefore, the in vivo doses for intraperitoneal therapy (0.25, 0.5, and 1 pmol/kg) were selected from the respective in vitro IC50 (0.5 pM). The dose for the peroral treatment was selected 20-fold higher (10 pmol/kg), as had been done for the peroral therapy with the crude extract.
The mean number of CC531-lacZ tumor cells growing in the liver of Wag-Rij rats increased 600-fold in untreated controls. This increase corresponds formally to more than 9 cell divisions and was paralleled by a more than threefold increase in mean liver weight (15)
.
After p.o. treatment with the affinity-purified protein sample, both the liver weight and the tumor cell number were significantly reduced when compared wtih the untreated tumor-bearing control. The actual tumor cell number of p.o. treated rats corresponded to an 8-fold increase relative to the initial tumor cell implant, which is formally equivalent to 3 cell divisions. As observed for the extract, the i.p. route was associated with a lower effect than the p.o. route: A significantly reduced tumor cell number was seen in response to the highest dose only, causing >77% tumor growth inhibition (defined as 100% = T/C%), whereas the difference in mean liver weight was not significant when compared with the control. When considering the net increase in mean liver weight, which is a more specific indicator of tumor proliferation than the raw mean liver weight, the higher inhibition expressed in percent of untreated control (>69%) reflects the results obtained by determining the tumor cell number within the variation of a bioassay.
The high efficacy of the p.o. application is surprising, since proteins are commonly characterized by a low to very low bioavailability, which is due to degradation in the acidic stomach environment, to proteolytic enzymes as well as to a poor uptake through the intestinal wall. Accordingly, the viscum lectin I is only poorly absorbed via the gastrointestinal tract. Despite a low systemic LD50 in mice (5–10 µg/kg), dietary dosages as high as 500 mg/kg (mice) and 200 mg/kg (rats) are well tolerated (34
, 35)
. In line with this, cancer treatment with the recombinant rViscumin is based on systemic administration (9
, 10)
. Nevertheless, high doses of p.o. administered viscum lectin I are associated with some systemic activity: a diet containing up to 10 mg viscum lectin I per day and mouse was effective in reducing the tumor mass of a non-Hodgkin lymphoma (35)
. Notably, this dose is by more than six orders of magnitude higher than the effective p.o. dose of riproximin (0.6 µg/kg administered every 2nd day).
The differences in toxicity and specificity of type II RIPs are probably related to the sugar-specificity of the respective B-chains and their intracellular fate (36
37
38
39)
. It is interesting that only one of the two presumed sugar binding domains of the riproximin B-chain is supposed to bind galactose according to a flexible docking simulation assay. Whether this could be a reason for the high therapeutic efficacy of riproximin at a relatively low p.o. dose remains to be elucidated.
In conclusion, these results suggest that riproximin differs considerably in its pharmacological properties from other type II RIPs, thus indicating distinct potential for cancer treatment.
ACKNOWLEDGMENTS
We are indebted to Prof. Dr. M. Wink, Institute of Pharmacy and Biotechnology, University of Heidelberg, Germany, for his generous help in answering questions regarding plant biology. We thank furthermore Dr. W. Lehmann (Central Spectroscopy Unit, German Cancer Research Center, Heidelberg) for the mass spectrometrical analysis of the protein and his help with de novo sequencing of tryptic peptides as well as Dr. H. Heid (Division of Cell Biology, German Cancer Research Center, Heidelberg) for the N-terminal sequencing of the protein chains. We are grateful to Dr. C. Kliem and Prof. M. Wiessler (Department of Molecular Toxicology, German Cancer Research Center, Heidelberg) for kindly providing access to equipment for chromatographic separations. Finally, we thank Dr. D. Kübler (Department of Pathochemistry, German Cancer Research Center, Heidelberg) for helpful discussions.
Received for publication October 10, 2005. Accepted for publication January 17, 2006.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
