A fractionation protocol for the isolation of a highly purified polypeptide fraction from plant biomass is described. The procedure dereplicates ubiquitous substance classes known to interfere with bioassays often used in natural product-based drug discovery programs. The protocol involves pre-extraction with dichloromethane, extraction with ethanol (50%), removal of tannins with polyamide, removal of low-molecular-weight components with size-exclusion chromatography over Sephadex G-10, and final removal of salts and polysaccharides with solid-phase extraction using reversed-phase cartridges. The method has been applied to the aerial parts of Viola arvensis, resulting in the isolation of a peptide fraction that on further separation yielded a novel 29-residue macrocyclic polypeptide named varv peptide A, cyclo(-TCVGGTCNTPGCSCSWPVCTRNGLPVCGE-).

Seven novel macrocyclic polypeptides, designated as varv peptides B-H, have been isolated from the aerial parts of Viola arvensis. Their primary structures have been elucidated by automated Edman degradation and mass spectrometry. They all consist of 29 or 30 amino acid residues, covalently cyclized via the amide backbone and by three internal disulfide bridges. Their amino acid sequences are as follows: varv peptide B, cyclo-(TCFGGTCNTPGCSCDPWPMCSRNGLPVCGE); varv peptide C, cyclo-(TCVGGTCNTPGCSCSWPVCTRNGVPICGE); varv peptide D, cyclo-(TCVGGSCNTPGCSCSWPVCTRNGLPICGE); varv peptide E, cyclo-(TCVGGTCNTPGCSCSWPVCTRNGLPICGE); varv peptide F, cyclo-(TCTLGTCYTAGCSCSWPVCTRNGVPICGE); varv peptide G, cyclo-(TCFGGTCNTPGCSCDPWPVCSRNGVPVCGE); and varv peptide H, cyclo-(TCFGGTCNTPGCSCETWPVCSRNGLPVCGE). The varv peptides B-H exhibited high degrees of homology with the hitherto known macrocyclic peptides varv peptide A, kalata B1, violapeptide I, circulins A and B, and cyclopsychotride A.

Chosen to reflect biodiversity in a phylogenetic sense, 100 fractionated plant extracts were screened in vitro for cytotoxicity following extraction and fractionation (polypeptide isolation). Of these 100 extracts, 30 were selected and then characterized preliminarily for antitumor potency and mode of action by testing them on two cell lines and primary cultures of human tumor cells. On the basis of cytotoxicity potency, 10 of the extracts were further characterized for anticancer activity in 10 human tumor cell lines. This final testing resulted in seven potential lead plants with superior evidence of antitumor potential: Colchicum autumnale L. (Colchicaceae), Digitalis lanata Ehrh. and Digitalis purpurea L. (Plantaginaceae), Helleborus cyclophyllus Boiss. (Ranunculaceae), Menyanthes trifoliata L. (Menyanthaceae), and Viola arvensis Murr. and Viola patrinii Ging. (Violaceae). Within a database of antitumor compounds, the activity profiles of the extracts from these seven plants were compared, by correlation analysis, with those of more than 100 other compounds, including 39 standard drugs from different classes of cytotoxic mechanisms. The activity profiles of six of these candidates were uncorrelated with those of the standard drugs, possibly indicating new pathways of drug-mediated cell death.

The saponin digitonin, the aglycone digitoxigenin and five cardiac glycosides were evaluated for cytotoxicity using primary cultures of tumor cells from patients and a human cell line panel (representing different cytotoxic drug-resistance patterns). Of these seven compounds, proscillaridin A was the most potent (IC(50): 6.4--76 nM), followed by digitoxin, and then ouabain, digoxin, lanatoside C, digitoxigenin and digitonin. Correlation analysis of the log IC(50) values for the cell lines in the panel showed that compound cytotoxicity was only slightly influenced by resistance mechanisms that involved P-glycoprotein, topoisomerase II, multidrug resistance-associated protein and glutathione-mediated drug resistance. Digitoxin and digoxin expressed selective toxicity against solid tumor cells from patients, while proscillaridin A expressed no selective toxicity against either solid or hematological tumor cells. The results revealed marked differences in cytotoxicity between the cardiac glycosides, both in potency and selectivity, and modes of action for cytotoxicity that differ from that of commonly used anticancer drugs.

Chosen to reflect biodiversity in a phylogenetic sense, 100 fractionated plant extracts were screened in vitro for cytotoxicity following extraction and fractionation (polypeptide isolation). Of these 100 extracts, 30 were selected and then characterized preliminarily for antitumor potency and mode of action by testing them on two cell lines and primary cultures of human tumor cells. On the basis of cytotoxicity potency, 10 of the extracts were further characterized for anticancer activity in 10 human tumor cell lines. This final testing resulted in seven potential lead plants with superior evidence of antitumor potential: Colchicum autumnale L. (Colchicaceae), Digitalis lanata Ehrh. and Digitalis purpurea L. (Plantaginaceae), Helleborus cyclophyllus Boiss. (Ranunculaceae), Menyanthes trifoliata L. (Menyanthaceae), and Viola arvensis Murr. and Viola patrinii Ging. (Violaceae). Within a database of antitumor compounds, the activity profiles of the extracts from these seven plants were compared, by correlation analysis, with those of more than 100 other compounds, including 39 standard drugs from different classes of cytotoxic mechanisms. The activity profiles of six of these candidates were uncorrelated with those of the standard drugs, possibly indicating new pathways of drug-mediated cell death.