Ruminococcin C, a new anti-Clostridium perfringens bacteriocin produced in the gut by the commensal bacterium Ruminococcus gnavus E1

When colonizing the digestive tract of mono-associated rats, Ruminococcus gnavus E1 - a bacterium isolated from human faeces - produced a trypsin-dependent anti-Clostridium perfringens substance collectively named Ruminococcin C (RumC). RumC was isolated from the caecal contents of E1-monocontaminated rats and found to consist of two antimicrobial fractions: a single peptide (RumCsp) of 4235 Da, and a mixture of two other peptides (RumCdp) with distinct molecular masses of 4324 Da and 4456 Da. Both RumCsp and RumCdp were as effective as metronidazole in combating C. perfringens and their activity spectra against different pathogens were established. Even if devoid of synergistic activity, the combination of RumCsp and RumCdp was observed to be much more resistant to acidic pH and high temperature than each fraction tested individually. N-terminal sequence analysis showed that the primary structures of these three peptides shared a high degree of homology, but were clearly distinct from previously reported amino acid sequences. Amino acid composition of the three RumC peptides did not highlight the presence of any Lanthionine residue. However, Edman degradation could not run beyond the 11th amino acid residue. Five genes encoding putative pre-RumC-like peptides were identified in the genome of strain E1, confirming that RumC was a bacteriocin. This is the first time that a bacteriocin produced in vivo by a human commensal bacterium was purified and characterized.     

Leishmanicidal activity of synthetic antimicrobial peptides in an infection model with human dendritic cells

Different species of Leishmania are responsible for cutaneous, mucocutaneous or visceral leishmaniasis infections in millions of people around the world [14]. The adverse reactions caused by antileishmanial drugs, emergence of resistance and lack of a vaccine have motivated the search for new therapeutic options to control this disease. Different sources of antimicrobial molecules are under study as antileishmanial agents, including peptides with antimicrobial and/or immunomodulatory activity, which have been considered to be potentially active against diverse species of Leishmania [7] and [39]. This study evaluated the cytotoxicity on dendritic cells, hemolytic activity, leishmanicidal properties on Leishmania panamensis and Leishmania major promastigotes and effectiveness on parasite intracellular forms (dendritic cells infected with L. panamensis and L. major promastigotes), when each parasite in culture was exposed to different concentrations of a group of synthetic peptides with previously reported antimicrobial properties, which were synthesized based on their naturally occurring reported sequences. Dermaseptin, Pr-2 and Pr-3 showed inhibitory activity on the growth of L. panamensis promastigotes, while Andropin and Cecropin A (with a selectivity index of 4 and 40, respectively) showed specific activity against intracellular forms of this species. The activities of Andropin and Cecropin A were exclusively against the intracellular forms of the parasite, therefore indicating the relevance of these two peptides as potential antileishmanial agents. In the case of L. major promastigotes, Melittin and Dermaseptin showed inhibitory activity, the latter also showed a selectivity index of 8 against intracellular forms. These findings suggest Andropin, Cecropin A and Dermaseptin as potential therapeutic tools to treat New and Old World cutaneous leishmaniasis.     

A quantitative resazurin assay to determinate the viability of Trichomonas vaginalis and the cytotoxicity of organic solvents and surfactant agents

Trichomonas vaginalis causes trichomonosis, the most common, non-viral sexually transmitted disease. To test anti-Trichomonas agents, usually many with low water solubility, organic solvents and surfactant agents should be used. Therefore, the minimal inhibitory concentration (MIC) of acetone, methanol, ethanol, isopropanol, DMSO, Tween 20, Tween 80, and Triton X-100 was determined against T. vaginalis isolates using the quantitative resazurin method. Our results showed that solvents and surfactant agents can be employed as vehicles to test bioactive compounds at lower concentrations than MIC values and we suggest acetone and DMSO as preferential. Moreover, a new methodology is established to substitute or to complement the counting of viable trophozoites. The amount of resazurin reduced by T. vaginalis can be quantified by fluorescence spectroscopy, making the test a quantitative determination of cell viability. These results contribute for pharmacological investigations of bioactive compounds that need the use of solvents as solubilization vehicles to test anti-Trichomonas activity.     

Identification and characterization of a novel antimicrobial peptide from the venom of the ant Tetramorium bicarinatum

A novel antimicrobial peptide, named Bicarinalin, has been isolated from the venom of the ant Tetramorium bicarinatum. Its amino acid sequence has been determined by de novo sequencing using mass spectrometry and by Edman degradation. Bicarinalin contained 20 amino acid residues and was C-terminally amidated as the majority of antimicrobial peptides isolated to date from insect venoms. Interestingly, this peptide had a linear structure and exhibited no meaningful similarity with any known peptides. Antibacterial activities against Staphylococcus aureus and S. xylosus strains were evaluated using a synthetic replicate. Bicarinalin had a potent and broad antibacterial activity of the same magnitude as Melittin and other hymenopteran antimicrobial peptides such as Pilosulin or Defensin. Moreover, this antimicrobial peptide has a weak hemolytic activity compared to Melittin on erythrocytes, suggesting potential for development into an anti-infective agent for use against emerging antibiotic-resistant pathogens.     

Characterization of diverse antimicrobial peptides in skin secretions of Chungan torrent frog Amolops chunganensis

We have cloned, synthesized, and characterized 11 novel antimicrobial peptides from a skin derived cDNA library of the Chungan torrent frog, Amolops chunganensis. Seven of the 11 antimicrobial peptides were present in authentic A. chunganensis skin secretions. Sequence analysis indicated that the 11 peptides belonged to the temporin, esculentin-2, palustrin-2, brevinin-1, and brevinin-2 families. The peptides displayed potent antimicrobial activities against several strains of microorganisms. One peptide, brevinin-1CG5, demonstrated antimicrobial activity against all tested Gram-positive and Gram-negative bacteria and fungi, and showed high antimicrobial potency (MIC = 0.6 μM) against Gram-positive bacterium Rhodococcus rhodochrous. Some peptides also demonstrated weak hemolytic activity against human erythrocytes in vitro. Phylogenetic analysis based on the amino acid sequences of brevinin-1, brevinin-2, and esculentin-2 peptides from family Ranidae confirmed that the current taxonomic status of A. chunganensis is correct.     

Identification of LI-F type antibiotics and di-n-butyl phthalate produced by Paenibacillus polymyxa

Paenibacillus polymyxa strain JSa-9, a soil isolate that displayed antibacterial and antifungal activities in vitro, had been found to produce two types of antimicrobial substances. The two compounds were extracted from the fermentation broth of JSa-9 using ethyl acetate and subsequently purified by high performance liquid chromatography. By means of liquid chromatography-mass spectrometry and tandem mass spectrometry analysis, one of two antagonistic compounds was determined as di-n-butyl phthalate. And another was characterized as a mixture of related peptides of molecular masses of 883, 897, 911, 947, and 961 Da, with the most likely structure of them determined to be a cyclic depsipeptide with an unusual 15-guanidino-3-hydroxypentadecanoic acid moiety bound to a free amino group. These peptides were therefore members of the LI-F group of cyclic depsipeptides.     

Isolation and identification of Perkinsus olseni from feces and marine sediment using immunological and molecular techniques

Molecular and immunological probes were used to identify various life stages of Perkinsus olseni, a protozoan parasite of the Manila clam Ruditapes philippinarum, from a marine environment and decomposing clam tissue. Western blotting revealed that the antigenic determinants of the rabbit anti-P. olseni antibody developed in this study were peptides with molecular masses of 55.9, 24.0, and 19.2 kDa. Immunofluorescent assay indicated that the rabbit anti-P. olseni IgG was specific to all life stages, including the prezoosporangium, trophozoite, and zoospore. Perkinsus olseni prezoosporangium-like cells were successfully isolated from marine sediment collected from Hwangdo on the west coast of Korea, where P. olseni-associated clam mortality has recurred for the past decade. Purified cells were positively stained with the rabbit anti-P. olseni antibody in an immunofluorescence assay, confirming for the first time the presence of P. olseni in marine sediment. Actively replicating zoospores inside the prezoosporangia were observed in the decomposing clam tissue collected from Hwangdo. P. olseni was also isolated from the feces and pseudofeces of infected clams and confirmed by PCR. The clams released 1-2 prezoosporangia per day through feces. The data suggested that the fecal discharge and decomposition of the infected clam tissue could be the two major P. olseni transmission routes.     

Isolation of anti-Candida albicans compounds from Markhamia obtusifolia (Baker) Sprague (Bignoniaceae)

An increase in clinical cases of Candidiosis globally as well as fungal resistance to drugs prompted the search for novel anti-Candida albicans agents from plant sources. Leaf extracts of Markhamia obtusifolia were screened for activity against C. albicans in vitro. An acetone extract obtained following serial exhaustive extraction contained mainly the active components with at least four active zones on the bioautogram. Bioassay guided fractionation of this extract led to the isolation of three compounds which inhibited the growth of three C. albicans strains. Based on spectroscopy studies (NMR and MS), the compounds were identified as 3β-hydroxyurs-12-en-28-oic acid, ursolic acid (1) 3β, 19α-dihydroxyurs-12-en-28-oic acid, pomolic acid (2) and 2β, 3β, 19α -trihydroxy-urs-12-en-28-oic acid, 2-epi-tormentic acid (3). The most active compound was 3β, 19α-dihydroxy-12-ursen-28-oic acid (2) with a minimum inhibitory concentration (MIC) value of 12.5 µg/mL for C. albicans isolated from dog and 25.0 µg/mL for C. albicans from cat and ATCC 90028 at 24 h following incubation. However, at 48 h of incubation MICs were > 400 µg/mL for all the three compounds isolated. This study indicated that M. obtusifolia could be a potential source of active principles against C. albicans.     

A rapid microtiter plate assay for measuring the effect of compounds on Staphylococcus aureus membrane potential

We developed a homogenous microtiter based assay using the cationic dye 3, 3′-Diethyloxacarbocyanine iodide, DiOC₂(3), to measure the effect of compounds on membrane potential in Staphylococcus aureus. In a screen of 372 compounds from a synthetic compound collection with anti-Escherichia coli activity due to unknown modes of action at least 17% demonstrated potent membrane activity, enabling rapid discrimination of nuisance compounds.     

Epigallocatechin gallate inhibits intracellular survival of Listeria monocytogenes in macrophages

Epigallocatechin gallate (EGCg), the major tea catechin, is known as a potent anti-microbial and anti-tumor compound. The effects of EGCg on host defense mechanisms against Listeria monocytogenes infection were examined in vitro using mouse peritoneal exudate cells. The study showed that EGCg inhibited the intracellular growth of L. monocytogenes in macrophages. The enhancement of in vitro anti-L. monocytogenes activity by EGCg is not due to the modulation of reactive oxygen intermediates or the production of reactive nitrogen intermediates but due to the inhibition of its escaping from the phagosome into cytosolic space. Anti-L. monocytogenes of EGCg is through the inhibition of hemolytic and cholesterol-binding activity of listeriolysin O, which usually disrupts the phagosomal membrane in the escaping phase of L. monocytogenes.     

Purification and antibacterial activity of recombinant warnericin RK expressed in Escherichia coli

Warnericin RK is a small cationic peptide produced by Staphylococcus warneri RK. This peptide has an antimicrobial spectrum of activity almost restricted to the Legionella genus. It is a membrane-active peptide with a proposed detergent-like mechanism of action at high concentration. Moreover, the fatty acids content of Legionella was shown to modulate the peptide activity. In order to decipher the mode of action in details using solid-state NMR spectroscopy, large amount of an isotopic labeled peptide is required. Since it is less expensive to obtain such a peptide biologically, we report here methods to express warnericin RK in Escherichia coli with or without a fusion partner and to purify resulting recombinant peptides. The cDNA fragment encoding warnericin RK was synthesized and ligated into three expression vectors. Two fusion peptides, carrying polyhistidine tag in N- or C-terminal and a native peptide, without tag, were expressed in E. coli cells. Fusion peptides were purified, with a yield of 3 mg/l, by affinity chromatography and reverse-phase HPLC. The recombinant native peptide was purified using a two-step purification method consisting of a hydrophobic chromatography followed by a reverse-phase HPLC step with a yield of 1.4 mg/l. However, the anti-Legionella activity was lower for both tagged peptide probably because of structural modifications. So, the native recombinant peptide was preferentially chosen for ¹⁵N-labeling experiments. Our results suggest that the developed production and purification procedures will be useful in obtaining a large quantity of recombinant isotope-labeled warnericin RK for further studies.     

Honeybees and beehives are rich sources for fructophilic lactic acid bacteria

Fructophilic lactic acid bacteria (FLAB) are a specific group of lactic acid bacteria (LAB) characterized and described only recently. They prefer fructose as growth substrate and inhabit only fructose-rich niches. Honeybees are high-fructose-consuming insects and important pollinators in nature, but reported to be decreasing in the wild. In the present study, we analyzed FLAB microbiota in honeybees, larvae, fresh honey and bee pollen. A total of 66 strains of LAB were isolated from samples using a selective isolation technique for FLAB. Surprisingly, all strains showed fructophilic characteristics. The 66 strains and ten FLAB strains isolated from flowers in a separate study were genotypically separated into six groups, four of which being identified as Lactobacillus kunkeei and two as Fructobacillus fructosus. One of the L. kunkeei isolates showed antibacterial activity against Melissococcus plutonius, a causative pathogen of European foulbrood, this protection being attributable to production of an antibacterial peptide or protein. Culture-independent analysis suggested that bee products and larvae contained simple Lactobacillus-group microbiota, dominated by L. kunkeei, although adult bees carried a more complex microbiota. The findings clearly demonstrate that honeybees and their products are rich sources of FLAB, and FLAB are potential candidates for future bee probiotics.     

Phenolics from the culms of five bamboo species in the Tangjiahe and Wolong Giant Panda Reserves,Sichuan, China

Phenolic compounds were studied in the culms of five bamboo species collected in China: Yushania chungii, Fargesia robusta, Fargesia denudata, Fargesia rufa and Fargesia scabrida. All the species are eaten by giant panda (Ailuropoda melanoleuca). The culms contained phenolic acids and flavonoids in small concentrations, except for F. robusta, which did not contain flavonoids in detectable amounts. The species differed from each other in their phenolic composition. For example, F. rufa with the highest number of compounds clearly differed from other species. There were also differences among sampling sites, which reflect the differences among genotypes. Furthermore, there were clear ontogenetic differences in the culms: some compounds were present in mature culms but not in young (1–2 year old) culms, while the concentrations of other compounds decreased with increasing age. Over all, the composition and concentrations of soluble phenolic compounds in the bamboo culms were affected by species, age and site.     

Synthesis and evaluation of anti-tubercular activity of new dithiocarbamate sugar derivatives

The present study was undertaken to optimize the anti-tubercular activity of 2-acetamido-2-deoxy-β-d-glucopyranosyl N,N-dimethyldithiocarbamate (OCT313, Glc-NAc-DMDC), a lead compound previously reported by us. Structural modifications of OCT313 included the replacements of the DMDC group at C-1 by pyrrolidine dithiocarbamate (PDTC) and the acetyl group at C-2 by either propyl, butyl, benzyl or oleic acid groups. The antimycobacterial activities of these derivatives were evaluated against Mycobacterium tuberculosis (MTB). Glc-NAc-pyrrolidine dithiocarbamate (OCT313HK, Glc-NAc-PDTC) exhibited the most potent anti-tubercular activity with the minimal inhibitory concentration (MIC) of 6.25-12.5 μg/ml. The antibacterial activity of OCT313HK was highly specific to MTB and Mycobacterium bovis BCG, but not against Mycobacterium avium, Mycobacterium smegmatis, Staphylococcus aureus or Escherichia coli. Importantly, OCT313HK was also effective against MTB clinical isolates, including multidrug-resistant (MDR) and extensively drug-resistant (XDR) strains. Interestingly, OCT313HK was exerted the primary bactericidal activity, and it was also exhibited the bacteriolytic activity at high concentrations. We next investigated whether the mycobacterial monooxygenase EthA, a common activator of thiocarbamide-containing anti-tubercular drugs, also activated OCT313HK. Contrary to our expectations, the anti-tubercular activity of dithiocarbamate sugar derivatives and dithiocarbamates were not dependent on ethA expression, in contrast to thiocarbamide-containing drugs. Overall, this study presents OCT313HK as a novel and potent compound against MTB, particularly promising to overcome drug resistance.     

Genotypic and phenotypic diversity of rhizobia isolated from Lathyrus japonicus indigenous to Japan

Sixty-one rhizobial strains from Lathyrus japonicus nodules growing on the seashore in Japan were characterized and compared to two strains from Canada. The PCR-based method was used to identify test strains with novel taxonomic markers that were designed to discriminate between all known Lathyrus rhizobia. Three genomic groups (I, II, and III) were finally identified using RAPD, RFLP, and phylogenetic analyses. Strains in genomic group I (related to Rhizobium leguminosarum) were divided into two subgroups (Ia and Ib) and subgroup Ia was related to biovar viciae. Strains in subgroup Ib, which were all isolated from Japanese sea pea, belonged to a distinct group from other rhizobial groups in the recA phylogeny and PCR-based grouping, and were more tolerant to salt than the isolate from an inland legume. Test strains in genomic groups II and III belonged to a single clade with the reference strains of R. pisi, R. etli, and R. phaseoli in the 16S rRNA phylogeny. The PCR-based method and phylogenetic analysis of recA revealed that genomic group II was related to R. pisi. The analyses also showed that genomic group III harbored a mixed chromosomal sequence of different genomic groups, suggesting a recent horizontal gene transfer between diverse rhizobia. Although two Canadian strains belonged to subgroup Ia, molecular and physiological analyses showed the divergence between Canadian and Japanese strains. Phylogenetic analysis of nod genes divided the rhizobial strains into several groups that reflected the host range of rhizobia. Symbiosis between dispersing legumes and rhizobia at seashore is discussed.     

Effects of bleaching on the pharmacological and toxicological activities elicited by the aqueous extracts prepared from two “fire corals” collected in the Mexican Caribbean

Hydrocorals of the genus Millepora are abundant skeleton-forming inhabitants of coral reefs around the world. These species are popularly known as “fire corals” since contact with them causes severe pain, skin eruptions and blisters as a result of the release of unidentified toxins. Millepora species associate with photosynthetic dinoflagellates of the genus Symbiodinium (“zooxanthellae”), and up to now the role of these symbionts in the toxic effects induced by the “fire corals” is unknown. In this study, we compared the hemolytic, vasoconstrictor, and phospholipase A₂ (PLA₂) activities of the crude aqueous extracts prepared from normal and bleached specimens of two hydrocorals collected in the Mexican Caribbean, Milleporaalcicornis and Millepora complanata. Electrophoretic analysis revealed some differences between the protein profiles of the extracts prepared from normal and bleached specimens. Bleaching decreased, but not abolished, the hemolytic effect induced by the hydrocorals extracts and the phospholipase A₂ activity of M. complanata extract. Furthermore, it did not modify the enzymatic activity of M. alcicornis extract and vasoconstriction elicited by both extracts. Our results suggest that the presence of the symbionts does not importantly influence the pharmacological and toxic effects induced by Millepora ssp. extracts, and indicate that cnidarians are the main source of the bioactive compounds.     

Caissarolysin I (Bcs I), a new hemolytic toxin from the Brazilian sea anemone Bunodosoma caissarum: Purification and biological characterization

Two cationic proteins, C1 and C3, were purified to homogeneity from the hemolytic fraction of the venom of Bunodosoma caissarum sea anemone. The purification processes employed gel filtration followed by ion exchange chromatography, being the purity and molecular mass confirmed by SDS-PAGE and mass spectrometry. Protein C1 represented the second major peak of the hemolytic fraction and was previously believed to be a cytolysin belonging to a new class of hemolysins. The C1 protein has a molecular mass of 15495 Da and was assayed for hemolysis, PLA₂ activity and acute toxicity in crabs and mice, showing no activity in these assays. It has an amino terminal with no similarity to all known hemolysins and, therefore, should not be considered a toxin, being its function completely unknown. The protein C3 (19757 Da), that also lacks PLA₂ activity, was recognized by antiserum against Eqt II and presented high hemolytic activity to human erythrocytes (ED₅₀ of 0.270 μg/ml), being named Caissarolysin I (Bcs I). Its activity was inhibited by pre-incubation with sphingomyelin (SM) and also when in presence of erythrocytes pre-treated with the SMase P2, a phospholipase D from the brown spider Loxosceles intermedia, indicating that SM is the main target of Bcs I. Caissarolysin I is the first hemolysin purified from a sea anemone belonging to the genus Bunodosoma and belongs to the Actinoporin family of sea anemone hemolysins.     

Brevibacillus laterosporus strain BPM3, a potential biocontrol agent isolated from a natural hot water spring of Assam, India

A bacterial strain designated as BPM3 isolated from mud of a natural hot water spring of Nambar Wild Life Sanctuary, Assam, India, strongly inhibited growth of phytopathogenic fungi (Fusarium oxysporum f. sp. ciceri, F. semitectum, Magnaporthe grisea and Rhizoctonia oryzae) and gram-positive bacterium (Staphylococcus aureus). The maximum growth and antagonistic activity was recorded at 30 °C, pH 8.5 when starch and peptone were amended as carbon and nitrogen sources, respectively. In greenhouse experiment, this bacterium (BPM3) suppressed blast disease of rice by 30-67% and protected the weight loss by 35-56.5%. The maximum disease protection (67%) and weight loss protection (56.5%) were recorded when the bacterium was applied before 2 days of the pathogen inoculation. Antifungal and antibacterial compounds were isolated from the bacterium which also inhibited the growth of these targeted pathogens. The compounds were purified and on spectroscopic analysis of a purified fraction having R_f 0.22 which showed strong antifungal and antibacterial activity indicated the presence of C-H, carbonyl group, dimethyl group, -CH₂ and methyl group. The bacterium was characterized by morphological, biochemical and molecular approaches and confirmed that the strain BPM3 is Brevibacillus laterosporus.     

The endemic Genista versicolor from Sierra Nevada National Park in Spain is nodulated by putative new Bradyrhizobium species and a novel symbiovar (sierranevadense)

Genista versicolor is an endemic legume from Sierra Nevada National Park which constitutes one of the UNESCO-recognized Biosphere Reserves. In the present study, a collection of strains nodulating this legume was analysed in characteristic soils of this ecosystem. Most strains nodulating G. versicolor belonged to rrs group I within the genus Bradyrhizobium and only one strain, named GV137, belonged to rrs group II from which only a single species, B. retamae, has been described in Europe to date. Strain GV137, and some strains from rrs group I, belonged to putative new species of Bradyrhizobium, although most strains from group I belonged to B. canariense, according to the ITS fragment and atpD gene analysis. This result contrasted with those obtained in Genista tinctoria in Northeast Europe whose endosymbionts were identified as B. japonicum. The analysis of the symbiotic nodC and nifH genes carried by G. versicolor-nodulating strains showed that most of them belonged to symbiovar genistearum, as did those isolated from G. tinctoria. Nevertheless, strain GV137, belonging to rrs group II, formed a divergent lineage that constituted a novel symbiovar within the genus Bradyrhizobium for which the name sierranevadense is proposed. This finding showed that the Genisteae are not restrictive legumes only nodulated by symbiovar genistearum, since Genista is a promiscuous legume nodulated by at least two symbiovars of Bradyrhizobium, as occurs in Retama species.     

Comparative analyses of Legionella species identifies genetic features of strains causing Legionnaires’ disease

Background

The genus Legionella comprises over 60 species. However, L. pneumophila and L. longbeachae alone cause over 95% of Legionnaires’ disease. To identify the genetic bases underlying the different capacities to cause disease we sequenced and compared the genomes of L. micdadei, L. hackeliae and L. fallonii (LLAP10), which are all rarely isolated from humans.

Results

We show that these Legionella species possess different virulence capacities in amoeba and macrophages, correlating with their occurrence in humans. Our comparative analysis of 11 Legionella genomes belonging to five species reveals highly heterogeneous genome content with over 60% representing species-specific genes; these comprise a complete prophage in L. micdadei, the first ever identified in a Legionella genome. Mobile elements are abundant in Legionella genomes; many encode type IV secretion systems for conjugative transfer, pointing to their importance for adaptation of the genus. The Dot/Icm secretion system is conserved, although the core set of substrates is small, as only 24 out of over 300 described Dot/Icm effector genes are present in all Legionella species. We also identified new eukaryotic motifs including thaumatin, synaptobrevin or clathrin/coatomer adaptine like domains.

Conclusions

Legionella genomes are highly dynamic due to a large mobilome mainly comprising type IV secretion systems, while a minority of core substrates is shared among the diverse species. Eukaryotic like proteins and motifs remain a hallmark of the genus Legionella. Key factors such as proteins involved in oxygen binding, iron storage, host membrane transport and certain Dot/Icm substrates are specific features of disease-related strains.

Electronic supplementary material

The online version of this article (doi:10.1186/s13059-014-0505-0) contains supplementary material, which is available to authorized users.