The MSDIN family in amanitin-producing mushrooms and evolution of the prolyl oligopeptidase genes.

The biosynthetic pathway for amanitins and related cyclic peptides in deadly Amanita (Amanitaceae) mushrooms represents the first known ribosomal cyclic peptide pathway in the Fungi. Amanitins are found outside of the genus in distantly related agarics Galerina (Strophariaceae) and Lepiota (Agaricaceae). A long-standing question in the field persists: why is this pathway present in these phylogenetically disjunct agarics? Two deadly mushrooms, A. pallidorosea and A. subjunquillea, were deep sequenced, and sequences of biosynthetic genes encoding MSDINs (cyclic peptide precursor) and prolyl oligopeptidases (POPA and POPB) were obtained. The two Amanita species yielded 29 and 18 MSDINs, respectively. In addition, two MSDIN sequences were cloned from L. brunneoincarnata basidiomes. The toxin MSDIN genes encoding amatoxins or phallotoxins from the three genera were compared, and a phylogenetic tree constructed. Prolyl oligopeptidase B (POPB), a key enzyme in the biosynthetic pathway, was used in phylogenetic reconstruction to infer the evolutionary history of the genes. Phylogenies of POPB and POPA based on both coding and amino acid sequences showed very different results: while POPA genes clearly reflected the phylogeny of the host species, POPB did not; strikingly, it formed a well-supported monophyletic clade, despite that the species belong to different genera in disjunct families. POPA, a known house-keeping gene, was shown to be restricted in a branch containing only Amanita species and the phylogeny resembled that of those Amanita species. Phylogenetic analyses of MSDIN and POPB genes showed tight coordination and disjunct distribution. A POPB gene tree was compared with a corresponding species tree, and distances and substitution rates were compared. The result suggested POPB genes have significant smaller distances and rates than the house-keeping rpb2, discounting massive gene loss. Under this assumption, the incongruency between the gene tree and species tree was shown with strong support. Additionally, k-mer analyses consistently cluster Galerina and Amanita POPB genes, while Lepiota POPB is distinct. Our result suggests that horizontal gene transfer (HGT), at least between Amanita and Galerina, was involved in the acquisition of POPB genes, which may shed light on the evolution of the α-amanitin biosynthetic pathway.

INTRODUCTION

Amatoxins and related cyclic peptides produced by deadly Amanita and Galerina mushrooms are biosynthesized through a ribosomal cyclic peptide pathway (Hallen , Luo ). α-Amanitin, the major cyclic peptide toxin, is responsible for the vast majority (>90%) of deadly mushroom poisonings worldwide (Bresinsky & Besl 1990). The biosynthesis of this toxin and related cyclic peptides begins with activation of genes that encode a precursor peptide of 34–37 amino acids, named the MSDIN gene family after the highly conserved first five residues (Hallen ). The precursor peptides are cleaved and macrocyclized into 7–10 amino acid cyclic peptides by a specialized prolyl oligopeptidase enzyme, POPB (Luo , 2014, Riley ). POPB is the key enzyme of the cyclic peptide pathway, catalyzing both hydrolysis of the peptide bond and transpeptidation (Luo ).

Besides certain species of Amanita and Galerina, a few species of Lepiota also produce α-amanitin (Haines , Mottram , Sgambelluri ). Some Conocybe species are also reported to produce similar toxins, but we failed to detect any cyclic peptides in recently collected C. apala, reported to produce phallotoxins, and therefore we did not continue further with that study. Phylogenetically, the genera Amanita, Galerina and Lepiota are distantly related, belonging to three disjunct agaric families, Amanitaceae, Strophariaceae, and Agaricaceae, respectively. This non-continuous distribution of amatoxins has raised major questions in the field: why does this pathway occur in these isolated lineages and not others? Did amatoxin biosynthesis evolve independently on multiple occasions, or did it originate from a common ancestor followed by gene loss or horizontal gene transfer (HGT)? Researchers had tried to infer the evolution of the pathway, but major hurdles existed and have prevented significant progress. One serious problem is the lack of suitable molecular markers, specifically for those genes related to the pathway, to infer essential and well resolved phylogenetic frameworks for the target mushroom groups, which are critical for providing evolutionary evidence on how this pathway evolved among the groups. A second problem is that phylogenetically important species have been difficult to obtain, especially fresh samples suitable for genetic and genomic analyses. Amatoxin-producing Amanita species are obligately mycorrhizal and grow slowly in culture, necessitating the use of wild-collected basidiomes. Thirdly, sequenced relevant agaric genomes were insufficient to support comparative studies among the target mushroom groups. This lack of data motivated more and deeper genome sequencing of deadly Amanita species by our group. Recent evidence showed that POPB is a key biosynthetic gene for the amatoxins and related cyclic peptides of lethal mushrooms (Luo ), and these data have provided clues for rigorously elucidating the evolution of the pathway, even though genomic data on the mushrooms remained incomplete. Not only can the phylogeny of POP genes resolve the relationships among these genes, it can assist with HGT detection, as the most reliable method for HGT detection is based on phylogenetic inference (Ragan 2001, Fitzpatrick 2012). In the absence of experimental systems to track HGT, the standard method for identifying putative HGT events has relied on phylogenetic incongruence — a strongly supported disagreement between a well-supported gene phylogeny and the species phylogeny is often used to justify the acceptance of one or more putative HGT events as the cause of the phylogenetic conflict (Andersson 2005, Keeling & Palmer 2008).

Prolyl oligopeptidases (POPs; EC 3.4.21.26) are present in most phyla of life (Venalainen , Kaushik & Sowdhamini 2014). They play important but varied house-keeping functions. In mammals (including humans), POPs are apparently multifunctional enzymes involved in the maturation and degradation of peptide hormones and neuropeptides (Polgar 2002). As such, POPs play important roles in a number of physiological processes, including: learning and memory (Yoshimoto , Garcia-Horsman ), cell signaling (Williams , Duan ), sperm motility (Yoshida , Kimura ), and cell proliferation and differentiation (Ohtsuki , Moreno-Baylach , Sakaguchi ). Furthermore, abnormalities in POP activity are associated with diseases (Momeni ). The majority of reported POPs are intracellular enzymes, while mushrooms produce some extracellular POPs (Chen ) that can only be speculated to perform more specific peptide hydrolyzation roles compared to more general proteases.

While many POPs perform housekeeping functions, one POP has a specialized role as a biosynthetic enzyme for the MSDIN family of cyclic peptides. Deadly Amanita and Galerina species carry two copies of POP genes, POPA and POPB, in contrast to only a single copy in other basidiomycete genomes; most ascomycetes do not contain POPs. POPB is a specialized form involved in the toxin biosynthesis (Luo ). By comparison, POPA is considered to carry out housekeeping roles as the homologs are present in all the mushrooms examined to date, poisonous or not (Luo , 2012, 2014). In G. marginata, evidence shows that GmPOPA does not catalyse cyclization of the precursor peptide for α-amanitin, and therefore is not involved in the biosynthesis of the cyclic peptides (Luo ).

As for taxonomic distribution, to date, evidence indicates that POPB is strictly confined to mushrooms producing MSDIN-family cyclic peptides, a family which includes phallotoxins (not toxic to humans on ingestion) as well as the dangerous amatoxins. Our initial investigation indicated that, unlike many other single-copy genes, the phylogeny of POPB does not reflect that of the agaric species that harbor this gene. Rather, they tend to cluster according to functional chemical diversity, contradicting their species phylogeny. During our efforts to sequence more amanitin-producing mushrooms, we tried to gain insights into the evolution of the pathway using comparative genomics, but no clear conclusion has yet been reached. However, this has provided more putative POPB gene sequences over time. Recently Jonathan D. Walton's laboratory at Michigan State University started sequencing a deadly Lepiota, L. subincarnata, and kindly sent us the genomic sequences of two POPB sequences, one from each of the two strains sequenced (the species contains only one POPB, and no POPA). As a result, we now have POP gene sequences from all three taxonomic groups confirmed to produce MSDIN-family cyclic peptides, which makes it possible to reconstruct the evolutionary histories of these genes in a well-represented species composition, and to perhaps shed light on the history of the biosynthetic pathway.

In this research, two amanitin-producing mushrooms, A. subjunquillea and A. pallidorosea, were sequenced through the Beijing Genomics Institute (BGI) in Wuhan, China, and the genomes were surveyed for MSDIN genes. MSDIN sequences were also cloned from two L. brunneoincarnata strains, and toxin MSDINs (defined as MSDIN genes encoding amatoxins or phallotoxins) from all three genera were compared. Furthermore, DNA and amino acid sequences of POP genes were mined from the genome assemblies. Together with two POPB genes from L. subincarnata, predicted coding and amino acid sequences for POPs from genome mining and databases were used for phylogenetic analyses. A POP gene tree was compared with the species tree for incongruency analysis. Distances and substitution rates were compared among the three genera. A topology test was performed to determine the robustness of the POPB phylogeny. Gene structure was analyzed by examining intron placement in POPB and toxin MSDIN genes, and conducting k-mer analyses on di-, tri- and tetranucleotide frequencies on POP and POPB genes. Based on the results, we assessed the evolutionary history of POPB.

METHODS

Mushroom samples

Fresh wild basidiomes of Amanita subjunquillea, A. pallidorosea and Lepiota brunneoincarnata were harvested, stored at –80 °C, and lyophilized. Upon collection, all specimens were immediately put on dry ice after they were removed from soil. These species are a major cause of lethal mushroom poisonings in Eastern Asian countries (Chen , 2016).

Genome sequencing and assembly

High molecular weight DNA was extracted from lyophilized basidiomes using Genomic-tip 100/G (Qiagen 10243), following the manufacturer's protocols. The sequencing strategy for A. subjunquillea and A. pallidorosea used Illumina HiSeq 4000 and PacBio RSII at Beijing Genomics Institute (BGI) with 250 bp, 10 Kb and 20 Kb libraries constructed and sequenced using the company's standardized pipeline. In both cases, PacBio polymerase reads < 1000 bp, or with quality score less than 80 %, were removed. Subreads were extracted from polymerase reads, and adapter filtered. Subreads were corrected using Pbdagcon (https://github.com/PacificBiosciences/pbdagcon), Falcon (https://github.com/PacificBiosciences/FALCON-integrate) and Proovread (Hackl ). Corrected reads were assembled with Celera Assembler (Myers ) (v. 8.3, parameters: doTrim_initialQualityBased = 1, doTrim_finalEvidenceBased = 1, doRemoveSpurReads = 1, doRemoveChimericReads = 1, -d properties -U) or Falcon (v. 0.3.0, parameters: -v -dal8 -t32 -h60 -e.96 -l500 -s100 -H3000). Scaffolds were constructed through SSPACE Basic (v. 2.0) (Boetzer & Pirovano 2014) and gap closing with PBJelly2 (English ) (15.8.24 with default settings). GATK (https://www.broadinstitute.org/gatk/) and SOAP tool packages (SOAP2, SOAPsnp, SOAPindel) (Li , b) were applied for single-base corrections.

Cloning of MSDINs from Lepiota brunneoincarnata

Primers targeting conserved regions of MSDINs were designed based on genes from G. marginata and A. bisporigera. For PCR amplification, four primers out of 18 primers tried were used in four combinations. The forward primers were 5’-GGCTACCTCATGTCTGCTCTCG-3’ and 5’-CAATCCGTCTGACTACCCACTC-3’. The reverse primers were 5’-ACCGAGCGTTGTATAGGGAGAA-3’ and 5’-GCAAAGGCTAGCAGACAATACG-3’. PCR reactions were conducted under standard conditions, and products with predicted correct sizes directly sequenced.

Mining for MSDIN and prolyl oligopeptidase genes

Nucleotide sequences of MSDINs and/or POPs from the genomes of A. subjunquillea and A. pallidorosea (this study), and A. phalloides (Pulman ), were obtained through standalone BLAST searches (NCBI BLAST+ 2.4.0) with corresponding query MSDIN and POPB sequences from A. bisporigera and G. marginata, which are well characterized by our molecular and biochemical approaches (Luo , 2012). In order to obtain reasonably reliable coding and amino acid sequences of the POP genes from the sequenced genomes, the genomic DNA sequences of the genes were compared to those of well characterized cDNA sequences from A. bisporigera and G. marginata. It quickly became apparent that the intron and exon structures are highly conserved among both POPA and POPB genes. Coding sequences were predicted using POPB cDNA from A. bisporigera as the reference. Similarly, POPA coding sequences were retrieved using AbPOPA cDNA for comparison. In above cases, GT-AG intron borders were predicted by aligning the gDNA sequences with the cDNAs, and the resulting amino acid sequences were further assessed by examining conservation among the amino acid sequences along the full length. In all cases, we obtained undisrupted ORFs after deleting the introns, and the amino acid sequences were conserved throughout the full length. With this method, final exon-intron structure was resolved without ambiguity. The same approach was applied to L. subincarnata POPB genomic DNA by comparing its only POP, LsPOPB, to the cDNAs of those in G. marginata. After the introns were predicted and removed, amino acid sequences were retrieved through translation with no ambiguity found. The resulting sequences were then used for phylogenetic analysis. The Amanita POP sequences can be found in Suppl. File 1. To generate a well-represented POP pool for macrofungi, POP coding (or cDNA) and protein sequences were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/) and JGI MycoCosm (http://genome.jgi.doe.gov/programs/fungi/index.jsf) (Table 1). For MSDIN gene comparison, additional sequences were obtained from previously published sources (Li , Pulman ).

Table 1.

Accession numbers of prolyl oligopeptidase gene and amino acid sequences included in the phylogenetic study.

Taxon	Strain	Prolyl Oligopeptidase	Source	Amino Acids	Coding Sequence
Agaricus bisporus var. bisporus	H97	POP	JGI	219134
Agrocybe pediades	AH 40210	POP	JGI	736488
Anomoporia bombycina	ATCC 64506	POP	JGI	1333562
Antrodia sinuosa	LB1	POP	JGI	705706
Artolenzites elegans	CIRM-BRFM 1663	POP	JGI	887929
Auricularia subglabra	–	POP	JGI	1169635
Auriculariopsis ampla	NL-1724	POP	JGI	534627
Amanita bisporigera	–	POPA	NCBI	ADN19204	HQ225840
	–	POPB	NCBI	ADN19205	HQ225841
Amanita muscaria	Koide	POPA	JGI	74086
Amanita pallidorosea	–	POPA	Genome	–
	–	POPB	Genome	–
Amanita phalloides	–	POPA	Genome	–
	–	POPB	Genome	–
Amanita rimosa	–	POPA	Genome
	–	POPB	Genome
Amanita subjunquillea	–	POPA	Genome	–
	–	POPB	Genome	–
Amanita thiersii	Skay4041	POPA	JGI	193040
Beauveria bassiana	ARSEF 2860	POP	JGI	1657
Bolbitius vitellinus	SZMC-NL-1974	POP	JGI	1302751
Calocera cornea	–	POP	JGI	490308
Ceraceosorus bombacis	MCA 4658	POP	JGI	333060
Cerrena unicolor	–	POP	JGI	313626
Clitocybe gibba	IJFM A 808	POP	JGI	1443961
Colletotrichum nymphaeae	SA-01	POP	JGI	1020605
Conocybe apala	–	POP	NCBI	ACQ65797	FJ906819
Coprinopsis cinerea	okayama7#130	POP	NCBI	XP_001841244	XM_001841192
Cortinarius glaucopus	AT 2004 276	POP	JGI	7259917
Crepidotus variabilis	CBS 506.95	POP	JGI	870504
Crucibulum laeve	CBS 166.37	POP	JGI	718333
Cyathus striatus	AH40144	POP	JGI	1424557
Cytidiella melzeri	FP 102339	POP	JGI	1412396
Dichomitus squalens	LYAD-421 SS1	POP	JGI	160828
Fibulorhizoctonia sp.	CBS 109695	POP	JGI	740187
Fomitiporia mediterranea	–	POP	JGI	141658
Galerina marginata	–	POPA	JGI	70906
	–	POPB	JGI	146341
Gloeophyllum trabeum	ATCC 11539	POP	NCBI	XP_007862595	XM_007864404
Gymnopilus chrysopellus	PR-1187 v1.0	POP	JGI	1688417
Hebeloma cylindrosporum	h7	POP	JGI	444548
Heliocybe sulcata	OMC 1185	POP	JGI	1641899
Hydnomerulius pinastri	–	POP	JGI	174977
Hydnopolyporus fimbriatus	CBS384.51	POP	JGI	983183
Hypsizygus marmoreus	51987-8	POP	NCBI	KYQ30898	LUEZ01000114
Laccaria bicolor	–	POP	JGI	303722
Lentinus tigrinus	–	POP	JGI	581405
Leiotrametes sp.	BRFM 1775	POP	JGI	1337173
Lepiota subincarnata	–	POPB	JDW*	–
	–	POPB	JDW*	–
Lepista nuda	CBS 247.69	POP	JGI	1271729
Leucogyrophana mollusca	KUC20120723A-06	POP	JGI	1128529
Macrolepiota fuliginosa	MF-IS2	POP	JGI	799170
Malassezia pachydermatis	CBS 1879	POP	NCBI	KOS13970	LGAV01000004
Marssonina brunnea f. sp. ‘multigermtubi’	MB_m1	POP	JGI	1261
Metarhizium robertsii	ARSEF 23	POP	JGI	9962
Ophiocordyceps sinensis	CO18	POP	NCBI	EQL04271	KE652172
Panus rudis	PR-1116ss-1	POP	JGI	1587789
Paxillus adelphus	Ve08.2h10	POP	JGI	30278
Phanerochaete carnosa	HHB-10118-Sp	POP	JGI	191914
Pisolithus tinctorius	Marx 270	POP	JGI	969495
Pleurotus ostreatus	PC 15	POP	JGI	1092697
Plicaturopsis crispa	–	POP	JGI	170617
Pluteus cervinus	NL-1719	POP	JGI	759937
Polyporus brumalis	BRFM 1820	POP	JGI	1399521
Rhizoctonia solani	AG-1 IB	POP	JGI	7338
Rhizopogon vinicolor	AM-OR11-026	POP	JGI	794021
Schizophyllum commune	Tattone D	POP	JGI	421410
Serpula himantioides	SHA21-2	POP	JGI	40692
Suillus decipiens	EM49	POP	JGI	1139982
Trametes cingulata	BRFM 1805	POP	JGI	1559098
Trametes versicolor	–	POP	JGI	117177
Xerocomus badius	84.06	POP	JGI	1422158
Wolfiporia cocos	MD-104 SS10	POP	JGI	89370
Stereum hirsutum	FP-91666 SS1	POP	JGI	59870

*Provided by Jonathan D. Walton at Michigan State University.

Genome: obtained from Amanita genomes (Supplementary file 1).

Sequence alignment and phylogenetic analysis

Three datasets, the coding sequences (CDSs) and amino acid sequences of the selected POP genes, and the CDSs of selected toxin MSDINs, were compiled. Sequences were aligned using Muscle 3.6 (Edgar 2004) with default settings, and then manually adjusted with BioEdit (Hall 1999, Suppl. Files 2–4). For the amino acid alignment, LG+ G was selected as the best-fitting empirical model by ProTest 3 (Darriba ) under Akaike Information Criterion (AIC). For the nucleotide alignment, GTR + I + G and GTR + G were inferred as the best substitution models for the CDSs of POP and MSDIN genes by using MrModeltest v. 2.3 (Nylander 2004) under AIC, respectively. Maximum likelihood (ML) tree searching and bootstrapping (1 000 replicates) were done in RAxML v. 7 (Stamatakis 2006). Bayesian inference was carried out in MrBayes v. 3.2.6 (Ronquist & Huelsenbeck 2003) with two independent Markov chain Monte Carlo (MCMC) runs and four chains each. Runs were performed for 2 M generations, with trees sampled every 100 generations. Chain convergence was determined using Tracer v. 1.5 (http://tree.bio.ed.ac.uk/sofware/tracer/) to ensure convergence and sufficiently large effective sampling size values (>200). Subsequently, the sampled trees were summarized by discarding the first 25 % of trees as burn-in using the ‘sump’ and ‘sumt’ command implemented in MrBayes (Ronquist & Huelsenbeck 2003).

Gene tree vs. species tree

A POP gene tree and a species tree were prepared as follows. For the POP gene tree, 6 species containing POPB and 12 other related agarics were selected (Suppl. File 5). For the species tree, 30 taxa encompassing POPB-possessing species were chosen, and the taxa covered those in the gene tree (Suppl. File 6). CDS sequences of POP and rpb2 marker were applied for the two ML reconstructions. The rpb2 sequences were obtained from our custom genomes, deposited in GenBank and JGI genomes (blastp) using A. subpallidorosea rpb2 (KP691703) as the query. Alignments, model selection (GTR + I + G), and ML phylogeny were performed as described above. Comparison of distances and substitution rates of the resultant gene and species trees among the three amanitin-producing genera were performed using codeml in PMAL4.9 (Yang 2007) with REV model and runmode set to –2. Both trees were further analyzed by Notung 2.9 (Chen ) with Divergence-Loss (DL) and Divergence-Transfer-Loss (DTL) models under default settings. The statistics were recorded and compared. The model with lower “Event Score” was chosen to show predicted evolutionary events.

Topology test

In order to test how congruent the topology of the POPB clade is with that expected based on the species tree, alternative topologies/hypotheses were generated using PAUP (4.0b10, the following CONSEL analysis does not support RAxML outputs) for comparison with the best tree generated by this program. These hypotheses are: (1) POPBs were monophyletic with Amanita POPA; (2) POPBs were monophyletic with Amanita POPA and Galerina POPs; and (3) POPBs were monophyletic with Galerina POPs. The alternative trees with site-wise log-likelihoods were input into CONSEL (V0.1i) to perform approximately unbiased tests (Shimodaira & Hasegawa 2001).

K-mer analyses

The sequences used to generate the POP gene tree were compared using the k-mer profiles for 2-, 3-, and 4-mers. K-mer profiles were used to calculate the pairwise distances (cosine distance), and the resulting distance matrices were used to generate neighbor-joining trees. Trees were drawn using the itol webserver (https://itol.embl.de).

RESULTS

Amanita genomes

Fresh tissues of Amanita subjunquillea and A. pallidorosea were obtained in excellent condition (Fig. 1). The resultant two draft genomes have the following statistics: A. subjunquillea N50 is 680 kb with the genome size at 53 Mb (GC content = 46.57 %) with the assembly containing 148 scaffolds; A. pallidorosea N50 is 450 kb, and the genome reached 56 Mb (GC content = 46.24 %) and the assembly comprised 251 scaffolds. Our assemblies had improved on N50s compared with previous genome sequencing of deadly Amanita mushrooms (Pulman ). Because MSDIN genes are short and POP genes are intron-rich, reliable automated annotation cannot be achieved through standardized pipelines; these were therefore annotated manually. Further details of the genomes will be discussed elsewhere.

Fig. 1.

Deadly Amanita subjunquillea and A. pallidorosea collected in China for genome sequencing: A. A. subjunquillea (HKAS 54509). B. A. pallidorosea (HKAS 82350); Note that the preserved specimens were different basidiomes from the same collection sites. The inset in B shows the characteristic light rose-tinged colour on the cap of the species.

MSDINs and POP genes

Biosynthetic gene mining in the genomes revealed MSDIN and POPB genes in both of the newly sequenced Amanita species, in agreement with results from other amanitin-producing mushrooms in Amanita and Galerina. Tables 2 and 3 list the predicted MSDINs in both genomes, respectively. Amanita subjunquillea yielded 18 MSDINs while A. pallidorosea produced 29, each containing four toxin MSDIN genes that code for amatoxins or phallotoxins, namely the ones encoding α-amanitin, β-amanitin, phallacidin and phalloidin. Overall, the precursor genes have similar structures (leader peptide, core peptide, and recognition sequence; Arnison ) with strong conservation in leader peptide and recognition sequence regions. Exon and intron structures are conserved among all MSDIN genes, and the alignments indicate that the genes span four exons and three introns, with the first intron disrupting the coding region.

Table 2.

MSDIN peptide sequences from Amanita subjunquillea (As).

No.	Leader peptide	Core Peptide	Recognition Sequence	Scaffold	Coding and Notes
1	MSDINATCLP	IWGIGCNP	IWGIGCNP	1	α-amanitin
2	MSDINATRLP	IWGIGCNP	IWGIGCNP	1	β-amanitin
3	MSDINATRLP	IWGIGCDP	CIGDDVTALLTRGEALC	1	β-amanitin
4	MSDINATRLP	AWLVDCP	CVGDDVTALLTRGEALC	1	phallacidin
5	MSDINTARLP	HFASFIPP	CIGDDVTALLTRGEALC	1
6	MSDINTARLP	TFLPPLFVPP	CVSDDVTALLTRGEALC	1
7	MSDINATRLP	AWLATCP	CAGDDVTALLTRGEALC	2	phalloidin
8	MSDINATRLP	LNILPFMLPP	CVGDDVTALLTRGEALC	54
9	MSDMNATRLP	LIQRPFAP	CVSDDVTALLTRGEALC	41
10	MSDMNATRLP	LIQRPYAP	CVSDDVTALLTRGEALC	10
11	MFDINITRLP	IFWFIYFP	CVGDDVTALLTRGEALC	8
12	MSDINTARLP	IGRPESIP	CVGDDVTALLTRGEALC	13
13	MSDINTARLP	LRLPPFMIPP	CVGDDVTALLTRGEALC	13
14	MSDMNVARLP	ISDPTAYP	CVGGDVTALLTRGEALC	13	2 copies, in Apa
15	MSDINTVCLP	LQKPWSRP	CVGDDVTALLTRGEALC	13
16	MSDVNATRLP	FNFFRFPYP	CIGDDVTALLTRGEALC	38
17	MSDINATRLP	SSVLPRP	CVGDDVTALLTRGEALC	20
18	MSDMNVARLP	ISDPTAYP	CVGGDVTALLTRGEALC	22	2 copies, in Apa

Core peptides (correspond to cyclic peptide toxins) are in bold.

The conserved MSDIN (except for the ones with variations) is underlined.

No. 14 and 18 have two copies and are also in A. pallidorosea.

Table 4.

Comparison of distances and substitution rates of gene and species trees among the three amanitin-producing genera.

Species	Distance (Gt)	Distance (St)	dN (Gt)	dN (St)	dS (Gt)	dS (St)	dN/dS (Gt)	dN/dS (St)
A. rimosa vs. G. marginata	12.238	64.374	0.1572	0.0650	10.858	72.961	0.1448	0.0089
A. rimosa vs. L. subincarnata	17.565	43.247	0.1917	0.0916	16.058	52.356	0.1194	0.0175
G. marginata vs. L. subincarnata	14.226	34.662	0.1760	0.0836	12.357	49.097	0.1424	0.0170

Note: Gt = gene tree; St = species tree.

In A. subjunquillea and A. pallidorosea, both POPA and POPB were unambiguously identified. The sequences share high similarity with those in A. bisporigera, which allowed the introns to be determined by aligning the cDNAs of AbPOPA and AbPOPB to their counterpart genomic DNAs in the newly sequenced Amanita genomes. All the genomic DNA sequences have highly conserved exon-intron structures, each having 19 predicted exons and 18 introns. POP genes from A. phalloides were obtained similarly, and the conservation is higher among Amanita species compared with Galerina and Lepiota POPs. POPA and POPB gene and aa sequences from the three Amanita species are given in Suppl. File 1. Surprisingly, only one POP gene has been identified in L. subincarnata (Jonathan D. Walton, pers. comm.). The gene has strongest homology to known POPB sequences and is therefore named LsPOPB for reference purpose. The alignment of GmPOPB cDNA in Galerina and LsPOPB genomic DNA predicted the gene's structure to have 17 exons and 16 introns.

MSDINs from L. brunneoincarnata were obtained by PCR using conserved sequences found in poisonous species of Amanita or Galerina. So far, only two MSDINs have been retrieved and they both code for α-amanitin.

Comparison of toxin MSDIN genes in three amanitin-producing agaric genera

In this study, toxin MSDINs are defined as the precursor genes in the MSDIN family that encode amatoxins or phallotoxins, the major cyclic peptides in these agarics. The MSDIN sequences in Amanita are generally conserved (Fig. 2), with highlighted variations compared to the best represented consensus sequence from Amanita (not including the core peptides), although some variations are found in A. phalloides and two Asian Amanita species, A. subjunquillea and A. pallidorosea. This result also shows that Asian Amanita species share higher conservation with lowest amount of variation (red letters). In contrast, A. phalloides from Europe showed higher variation in its recognition sequences. Gene duplications are found in many species: in A. bisporigera, each copy for the listed two MSDIN genes is identical, indicating duplication happened recently without any accumulated variations. In A. phalloides and Asian amanitas, duplicates usually present some variations, indicating these duplications formed some time ago. In Lepiota, the leader peptides are 9 aa in length, while all others are 10 aa. In general, leader peptides are more conserved than the other sequences. In this case, the variations in this region reflect their generic position: MSDIN is specific to Amanita, MFDTN to Galerina, and MDAN to Lepiota (including L. subincarnata genomes). These sequences are highly conserved within the same genus, and true MSDIN sequences only exist in deadly Amanita species (underlined). Some sequences are highly conserved even across genera, including NATRLP in the leader peptide and LC (IC in G. marginata) at the very end of the recognition sequence. In addition, LTRG in the recognition sequence is conserved between the genera Lepiota and Amanita. Sequences of Galerina are closer to those of Lepiota than to those of Amanita: 10–11 variations vs. 15–16.

Fig. 2.

Overview of MSDIN family genes in the three amanitin-producing agaric genera. Upper left indicates the best represented consensus sequence, with a general schematic structure of MSDIN genes with leader peptide, core peptide, and recognition sequence (not including the core peptide, the hypervariable region). Coloured boxes harbor species from specific geological locations. Coloured (red and green) letters indicate variations (differences) compared with the consensus sequence. In G. marginata and L. brunneoincarnata, green letters also designate conserved amino acids only between the two species. MSDIN sequences are underlined, and true MSDIN sequences are only found in Amanita species.

Phylogeny of POPs in macrofungi

The aligned coding sequences of POP genes comprised 75 taxa (Table 1) with 3489 bp in length (Suppl. File 2), and the dataset of amino acid sequences included sequences of the same 75 species with 1035 aligned sites (Suppl. File 3). The nucleotide alignment of MSDIN genes consisted of 21 sequences with 376 bp (Suppl. File 4). ML and BI analyses yielded identical tree topologies, and thus only the trees inferred from the ML analysis are shown (Figs 3, 5). With regard to the POP genes, the lineage POPB from CDS (Fig. 4) had stronger statistical support than that from amino acid sequences (Fig. 3). This might be due to the higher conservation at the amino acid level due to degeneracy, quenching some of the phylogenetic signal that is present in coding sequences. The good support in the terminal clades including POPA and POPB allowed us to delineate boundaries of these genes. In the phylogenetic trees, POPB consistently formed a clade with strong support, while the remaining POP genes, including POPA, generated multiple strongly supported clades. Galerina POPA and Amanita POPA did not cluster together, but, rather, with POP genes from taxonomically related species, suggesting that “POPA” is simply the generic POP gene present in most basidiomycetes (Hibbett 2006, Matheny , 2007, Justo ). Notable examples include one clade representing all POPs from the order Boletales (Binder & Hibbett 2006) and another from the order Polyporales (Justo ). Strikingly, POPB displayed a very different pattern: all POPB genes, from three genera belonging to three disjunct families, clustered together forming a well-supported monophyletic clade (lineage POPB). Topology within POPB reflects species phylogeny; however, the apparent single origin of POPB within the POP tree (as opposed to derivation of POPB from each species’ POPA) requires further explanation.

Fig. 3.

Phylogeny of macrofungi inferred from maximum likelihood (ML) analysis based on amino acid sequences of POP gene. Maximum likelihood bootstraps over 50 % and Bayesian posterior probabilities over 0.90 are given at the internodes. Asterisks indicate orthologs.

Fig. 5.

Phylogeny of MSDIN genes in Amanita, Galerina and Lepiota inferred from maximum likelihood (ML). Maximum likelihood bootstraps over 50 % and Bayesian posterior probabilities over 0.90 are given at the internodes.

Fig. 4.

Phylogeny of macrofungi inferred from maximum likelihood (ML) analysis based on coding sequences of POP gene. Maximum likelihood bootstraps over 50 % and Bayesian posterior probabilities over 0.90 are given at the internodes. Asterisks indicate orthologs.

Phylogeny of MSDIN genes

The MSDIN phylogenetic tree was constructed without an outgroup. Fig. 5 shows that the MSDIN sequences from Lepiota and Amanita are separated and well supported. MSDINs are rather short and the hypervariable region (encoding cyclic peptides) interferes with phylogenetic analysis. As shown in the figure, Amanita MSDINs do not cluster according to species phylogeny, but do so based on chemical properties, in this case the toxins they encode. As a consequence, in Amanita, MSDINs encoding α-amanitin, β-amanitin, phalloidin, and phallacidin group together, respectively.

In order to investigate the relationship between POP gene tree and the corresponding species tree, a gene tree based on POPs (Fig. 6A) and a species tree based on rpb2 marker (Fig. 6B), were constructed. The pairwise distances and substitution rates among three species representing the three disjunct genera were calculated (Table 4). Consistent with the general hypothesis that genes acquired via HGT would show significantly less divergence compared with house-keeping genes, the result indicated significantly smaller distances (to 1:6) and substitution rates (to 1:7) from the gene tree vs. the species tree. This result also allowed the discounting of the massive gene loss hypothesis, in which case the distances and substitution rates are expected be similar. In light of this, the non-POPB-containing taxa in the species cannot be removed in the topology comparison, and then incongruency was shown in 6A and 6B, with the POPB subclade marked in red. The strong statistical support ruled out the possibility of aligning conflicting clades under the settings. For example, Galerina marginata (POPB clade) will not cluster with Gymnopilus chrysopellus POP in the gene tree as the species do in the species tree. With Notung, the DL model returned the following general statistics: Event Score = 36.0, Dups = 4, Losses = 30, and Numbers of optimal solutions = 1. The DTL model produced: Event Score = 23.0, Dups = 0, Transfers = 5, Losses = 8, and Numbers of optimal solutions = 4. The DTL score was significantly smaller and therefore further analysis was based on DTL results. The DTL reconciled tree with one of the four optimal solutions is shown in Fig. 6C; their differences are at the gene transfer points (green circles). The illustrated transfer events in the POPB clade (T1, T2, and T3) indicate the possibility that the HGT happened from L. subincarnata to G. marginata (T1), then to an unknown species between G. marginata and Amanita rimosa (T2), and followed by another transfer within Amanita (T3). Three other optimal solutions have slightly different routes, but all indicated gene transfer.

Fig. 6.

Comparison of POP gene tree and species tree. A. POP gene tree. POPB lineage was highlighted in red; POPA lineage in blue. B. Species tree based on rpb2. Corresponding species for the POPB lineage were highlighted in red. Amanita lineage was indicated by a black bar. C. Reconciled tree by Notung. Yellow arrows indicated gene transfer. The green circles showed where the alternative transfer events occured. The transfer events in the strongly supported POPB lineage were marked as T1, T2 and T3.

Topology tests

The phylogenetic trees generated above shows that the POPB clade is not congruent with the species tree. The robustness of the POPB clade was assessed in this study. The best tree generated by PAUP was consistent with those by RAxML (Figs 3, 4, 6) and is not shown here. With the three alternative trees, Table 5 shows that the best tree by PAUP is highly supported over the alternative topologies for competing hypotheses, with both approximately unbiased p-values (AU) and bootstrap probability (NP) at 1. This result strongly suggested the monophyletic POPB clade is highly supported, rejecting a de novo origin of POPB from POPA within species.

Table 5.

Approximately unbiased test on alternative POP trees.

Rank	Best tree vs. Hypothetical trees	obs	AU	NP
1	Best tree	–3346.3	1.000	1.000
2	POPA and POPB monophyletic	3346.3	100000000000,00E-7	3000000000000,00E-6
3	POPA, POPB and Galerina POP monophyletic	4866.1	2000000000000,00E-8	2000000000000,00E-7
4	POPB and Galerina POP monophyletic	5232.1	100000000000,00E-51	3000000000000,00E-18

obs: observed log-likelihood difference

AU: approximately unbiased p-values

NP: bootstrap probability

Gene structure

As phylogenetic data cannot fully rule out an ancestral origin of POPB followed by multiple independent losses, we examined the intron structure of POPB and toxin MSDIN genes, and evaluated di-, tri-, and tetranucleotide frequencies of representative POP genes. Toxin MSDIN genes each contain three introns with a conserved size and placement, one near the 3’ end of the coding region and two within the 3’ UTR (Hallen , Luo ), while POPB genes contain 17 introns of similar size and placement (one additional intron is present in Amanita bisporigera; Fig. 7). The k-mer profiles of POP genes consistently group Galerina POPB with Amanita POPB and Amanita POPA, and are distinct from Galerina POPA (Fig. 8). Lepiota POPB has distinct k-mer profiles from the other POPB genes.

Fig. 7.

POPB intron and exon structure for Amanita, Galerina, and Lepiota.

Fig. 8.

Tetranucleotide frequency analysis of the POP sequences used in Fig. 6A. Tree given as rooted (left) and unrooted (right). Galerina POPB is placed within Amanita POPs, while Lepiota POPB is distinct. Di- and trinucleotide analyses give similar results (data not shown).

DISCUSSION

All known amatoxin-producing Amanita species belong to section Phalloideae, which has at times been restricted to only lethal species. Recently the number of taxa in the section has undergone a minor expansion, and now includes at least four non-poisonous species in basal positions of the clade. Phylogenetic evidence also indicates a single origin of the cyclic peptide pathway within Amanita (Cai , Cui ).

Diversity of MSDIN genes in three agaric genera

The genes discovered to date clearly share similar structures (including exon and intron structure), with leader peptide, core peptide and recognition sequence (Arnison ), indicating they shared a common ancestor. The Asian Amanita species possess a similar pool of these genes compared to their European and North American relatives. In general, the toxin MSDINs (genes encoding amatoxins or phallotoxins, i.e. α-amanitin, β-amanitin, phallacidin, and phalloidin) are shared among amanitas, while the rest do not overlap significantly. Much less diversity of MSDIN genes was observed in Galerina and Lepiota species. Galerina marginata only possesses one GmAMA1 gene in two copies; a rigorous PCR search only found two AMA1 genes in two L. brunneoincarnata strains (our initial genome assembly now confirms this). Besides the structural similarities, other variations were evident. The actual “MSDIN” motif is restricted to Amanita, and in Lepiota and Galerina, the variations in this leader peptide region are distinctive, and likely genus-specific. In the recognition sequence region, there are conserved aa residues across genera but with a significant number of variations (Fig. 2).

Evolution of POPB

With limitations, phylogenetic reconstruction methods remain the only way to reliably infer historical events from gene sequences as they are the only methods that utilize large, comprehensive data sets (Eisen 2000). Non-tree-based (“surrogate”) methods are increasingly used in identifying instances of lateral genetic transfer, but in many cases they lack reliability compared to rigorous phylogenetic analysis; further, phylogenetic methods are less dependent on subtle nucleotide-level signatures that could be unevenly distributed and subject to amelioration (Ragan ). For these reasons, we took the phylogenetic route to assess the evolutionary history of the key biosynthetic gene POPB. In the distance and substitution rate analyses, significantly smaller numbers in both indicate the POPB lineage evolved at a much lower speed, incongruent with that of the three disjunct genera in the rpb2 species tree, but consistent with the hypothesis of HGT. We therefore continued the topology comparison without the consideration of massive gene loss. From the ML tree topologies, all POPBs reside exclusively in a well-supported monophyletic clade (lineage POPB), with the saprotrophic species L. subincarnata and G. marginata in basal positions. In contrast, all POPBs from Amanita species are terminal (Figs 3, 4, 6). This scenario indicates POPBs from the saprobes are ancestral whereas Amanita POPBs are newer entities. The highly supported topologic incongruency between the POP gene tree and the species tree strongly suggest the acquisition of POPB in those lineages were likely the result of HGT. We also tested the incongruency using different markers, such as LSU, and the results were consistent, although some were with weak statistical support. The illustrated incongruency between POP gene trees and the species tree strongly suggested an HGT cause of the POPB distribution among Lepiota, Galerina, and Amanita. In addition, the topology test showed strong support for the POPB clade, rejecting all other three competing hypotheses. In Notung analysis, all predicted four best DTL solutions involve HGT with only minor variations, lending more support to the hypothesis.

Since POPB is a single-copy gene and is at the centre of the cyclic peptide biosynthesis, its phylogeny may reflect the evolutionary history of the α-amanitin biosynthetic pathway. In contrast, MSDIN sequences are less usable as they are short and have a highly variable region (core peptide) that interfers with the phylogenetic methods by pulling genes for same cyclic peptides together, although coding sequence analysis showed less of this problem than aa phylogeny (Fig. 5). The existence of the core peptides also partly causes low statistical support. Lacking a proper outgroup is another reason not enough information was obtained through the analysis.

Hypothesis for transferring the cyclic peptide pathway

At least three possible hypotheses for how the cyclic peptide pathway evolved in the three disjunct agaric genera can be envisioned: (1) the pathway arose independently in each of the genera; (2) the pathway originated in a common ancestor but was lost in most of the descendants except for the three genera; or (3) the pathway formed in a common ancestor and was then transferred through HGT to other recipients. If the pathway was due to independent origins as a result of convergent evolution, little resemblance among the pathways in the three genera would be expected. However, all pathways use MSDIN genes for the precursor peptides, and all the MSDINs share a conserved structure that features leader peptide, core peptide, and recognition sequence (Arnison ). Furthermore, they all possess a specialized POPB gene that clusters into a single lineage, and they too, like the MSDINs, share exon and intron structures, indicating these genes are from a common ancestor (Fig. 7). Regarding the second hypothesis, our analyses on substitution rates and distances indicated that POPB genes evolve much slower than the house-keeping gene rpb2, consistent with the HGT hypothesis while conflicting with massive gene loss. The species trees only have a small subset of taxa, and the differences in distances and rates would only increase if more species are included. In addition, among the three families, we counted over 2000 species (or significantly more as the count was not complete). If there was a common ancestor in which the pathway originated, then thousands of agarics would have to lose at least two genes (MSDIN and POPB) to accommodate the toxin distribution. While this is not entirely impossible, we consider the chance slim. Further, one would expect some pathway remnants detectable in the closely-related genomes, but in our BLAST searches and comparative genomic study, none has been found. As discussed above, our study now lends some support for the third hypothesis. Multiple phylogenetic reconstructions, comparison of substitution rates and distance, analyses of gene tree and species tree, predicted evolutionary events, and topology test, all suggest HGT was the underlining cause for the disjunct toxin distribution.

Other information pertaining to the HGT hypothesis

We have cloned one Class II transposon close to GmPOPB in G. marginata (unpubl.), and both Class I and Class II transposons (~ 60 within the 50 kb range) were detected in our initial assembly of L. brunneosubincarnata (unpubl.). Therefore, HGT of POPB could be assisted by transposons. It is known that at least some degree of gene clustering in this pathway is present in A. bisporigera and G. marginata (Luo , 2012). Clustering of genes is considered to be able to assist in HGT. Supernumerary chromosome transfers can also be explained by interspecific mating rather than HGT, but our comparative genomic study using Symap was negative on this assumption. K-mer analysis (Fig. 8) suggests strongly that Galerina POPB is the result of HGT from Amanita, while Lepiota POPB clusters neither with the other POPB genes, nor with the POPs from related species (Macrolepiota fuliginosa and Agaricus bisporus).

Fungal HGT

Most early reports on HGT cases involve bacterial donors, however recent studies show-cased an increasing number of HGT events between fungi and other eukaryotes (Slot 2017). Some good examples include depudecin biosynthesis in Alternaria brassicicola (Reynolds ), ergotamine and loline biosyntheses in fungi belonging in Clavicipitaceae (Marcet-Houben & Gabaldon 2016), chaetoglobosin-like compound biosynthesis in Mycosphaerella populorum (Dhillon ), fumonisin cluster in Aspergillus niger (Khaldi & Wolfe 2011), avirulence gene ACE1 cluster in Magnaporthe grisea, Chaetomium globosum, Stagonospora nodorum, and A. clavatus (Khaldi ), four pathogenic chromosomes in Fusarium (Ma ), eight chromosomes in the wheat pathogen M. graminicola (Goodwin ), and a virulence gene called ToxA in Pyrenophora tritici-repentis (Friesen ). Recent phylogenomic studies indicate that in a “typical” fungal genome, between 0.1–2.8 % of the genes may be the result of HGT (Wisecaver , Wisecaver & Rokas 2015), a much lower proportion than that in bacteria and archaea (Koonin ). The structures of the MSDIN and POPB genes include multiple introns, strongly suggesting the fungal origin of these genes. Therefore, the evolution of POPB should be an example of fungal HGT. Based on the phylogenetic positions of POPBs from Lepiota, Galerina, and Amanita, one can conjecture the rough direction of the gene acquisition could also be in this direction. Substantial work is needed to provide more insights into this aspect.

Possible impact of the biosynthetic pathway

A long-observed phenomenon in ectomycorrhizal Amanita species is the rapid speciation within the lineage of lethal amanitas of the sect. Phalloideae, while the number of the non-amanitin-producing taxa in the section is significantly lower (Cai , Cui ). A logical question regarding these opposing trends is: did the pathway drive the rapid speciation in deadly Amanita? At present, we have little clue regarding the target organisms, biological roles or selective advantages of amanitins, even less on many other cyclic peptides made by these agarics. Hopefully, the fast development in genome research combined with our transformation system will point us to the right directions.

Table 3.

MSDIN peptide sequences from Amanita pallidorosea (Apa).

Name	Leader peptide	Core Peptide	Recognition Sequence	Scaffold	Coding
1	MSDINATRLP	IWGIGCNP	CVGDDVTALLTRGEALC	52	α-amanitin, 2 copies
2	MSDINATRLP	IWGIGCNP	CVGDDVTALLTRGEALC	52	α-amanitin, 2 copies
3	MSDINATRLP	IWGIGCDP	CVGDDVTALLTRGEALC	52	β-amanitin
4	MSDVNATRLP	MAFPEFLA	CVGDDVTALLTRGEALC	52
5	MSDINATRLP	AWLMTCP	CVGDDVTALLTRGEALC	7
6	MSDVNATRLP	AWLVDCP	CVGDDVTALLTRGEALC	62	phallacidin
7	MSDINASRLP	FFPEVGFFP	CVGDDVTALLTRGEALC	16
8	MSDINAIRAP	LPIFSLNP	CVGDDVTALLTRGEALC	16
9	MSDINATRLP	NWHAGPTRPP	CVADDVTALLTRGEALC	16
10	MSDINAARLP	NLFVWIPP	CISDDVTALLTRGEALC	16
11	MSDINASRLP	AWLATCP	CAGDDVTALLTRGEALC	16	phalloidin
12	MADINASRLP	LNILPFHLPP	CVSDDVTALLTRGEALC	16
13	MSDINATRLP	LGRPESLP	CVGDDVTALLTRGEALC	3
14	MSDINATRLP	HPFPLGLQP	CAGDDVTALLTRGEALC	73	2 copies
15	MSDINATRLP	HPFPLGLQP	CAGDDVTALLTRGEALC	101	2 copies
16	MSDMNVVRLP	ISDPTAYP	CVGDDVTALLTRGEALC	102	2 copies, in As
17	MSDINVIRLP	IF*FIYFP	CVGDNVTALLTRGEALC	60
18	MSDINTARLP	MHILAPPP	CVSDDVTALLTRGEALC	10
19	MSDMNVVRLP	ISDPTAYP	CVGDDVTALLTRGEALC	64	2 copies, in As
20	MSDTNATRLP	SIFIVYPP	CVSDDVTALLTRGEALC	86
21	MSDINAARLP	LVYMILFP	SVGDDVTALLTRGEALC	111
22	MSETNAARLP	TIHLFSAP	SVGDDVTALLTRGEALC	111
23	MSDINTARLP	EFIVFGIFP	CVGDDVTALLTRGEALC	78
24	MSDINTTRLP	YFFNDHPP	CASDDVTALLTRGEALC	78
25	MSDLNATRLP	FNLFRFPYP	CIGDDVTALLTRGEALC	9
26	MSDINTARLP	VFFMPPFIPP	CVSDDVTALLTRGEALC	1
27	MSNINTARLP	FLVPSFPP	CVSDDVTALLTRGEALC	1
28	MYDINTTRLP	HFFNLTPP	CVRDDVTALLTRGEALC	39
29	MSDINTARLP	FVIIPPFIFP	CVSDDVTALLTRGEALC	12

Core peptides (correspond to cyclic peptide toxins) are in bold.

The conserved MSDIN (except for the ones with variations) is underlined.

No. 14 and 15 have two copies.

No. 16 and 19 have two copies and are also in A. subjunquillea.

*indicates a stop codon exists in the sequence.