Draft genomic sequence of Armillaria gallica 012m: insights into its symbiotic relationship with Gastrodia elata

Abstract

Armillaria species (Basidiomycota, Physalacriaceae) are well known as plant pathogens related to serious root rot disease on various trees in forests and plantations. Interestingly, some Armillaria species are essential symbionts of the rare Chinese medicinal herb Gastrodia elata, a rootless and leafless orchid used for over 2000 years. In this work, an 87.3-M draft genome of Armillaria gallica 012m strain, which was symbiotic with G. elata, was assembled. The genome includes approximately 23.6% repetitive sequences and encodes 26,261 predicted genes. In comparison with other four genomes of Armillaria, the following gene families related to pathogenicity/saprophytic phase, including cytochrome P450 monooxygenases, carbohydrate-active enzyme AA3, and hydrophobins, were significantly contracted in A. gallica 012m. These characteristics may be beneficial for G. elata to get less injuries. The genome-guided analysis of differential expression between rhizomorph (RH) and vegetative mycelium (VM) showed that a total of 2549 genes were differentially expressed, including 632 downregulated genes and 1917 upregulated genes. In the RH, most differentially expressed genes (DEGs) related to pathogenicity were significantly upregulated. To further elucidate gene function, Gene Ontology enrichment analysis showed that the upregulated DEGs significantly grouped into monooxygenase activity, hydrolase activity, glucosidase activity, extracellular region, fungal cell wall, response to xenobiotic stimulus, response to toxic substance, etc. These phenomena indicate that RH had better infection ability than VM. The infection ability of RH may be beneficial for G. elata to obtain nutrition, because the rhizomorph constantly infected the nutritional stems of G. elata and formed the hyphae that can be digested by G. elata. These results clarified the characteristics of A. gallica 012m and the reason why the strain 012m can establish a symbiotic relationship with G. elata in some extent from the perspective of genomics.

Electronic supplementary material

The online version of this article (10.1007/s42770-020-00317-x) contains supplementary material, which is available to authorized users.

Introduction

Gastrodia elata, commonly called Tian Ma in Chinese, is a notable Chinese medicinal herb used for over 2000 years to treat dizziness, headaches, migraine, rheumatism, epilepsy, neuralgia, and paralysis [1–4]. As a rootless, leafless, achlorophyllous, and fully mycoheterotrophic orchid, G. elata needs to establish a symbiotic relationship with Mycena and Armillaria strains to obtain nutrient supply for its life cycle. At early stages, Mycena strains are responsible for nutrient supply during seed germination, protocorm growth, and differentiation. At late stages, Armillaria strains replace Mycena strains as new symbionts of G. elata for its tuber enlargement, flowering, and fruit setting [5–7].

Armillaria species (Basidiomycota, Physalacriaceae) are well known as plant pathogens related to serious root rot disease on various trees in forests and plantations [8]. Most Armillaria species are facultative pathogens, which possess a parasitic phase when they colonize the cambium of living root and a saprophytic phase when they feed on dead tissues [9]. Interestingly, only few Armillaria species of the approximately 70 known species [10] can establish a symbiotic relationship with G. elata. In addition, the production and quality of G. elata tubers are affected by different Armillaria strains [11, 12].

In view of symbionts, most studies have focused on the perspective of G. elata to reveal the mechanisms of its symbiotic relationship with Armillaria species. Gastrodin biosynthesis–related genes have been identified through the comparative transcriptome analysis of G. elata in response to plant–fungus symbiosis [13]. Yuan et al. provided a high-quality reference genome assembly of G. elata and conducted transcriptome analysis. They found that strigolactone, as an important signal, could stimulate the hyphal branching and development of Armillaria; G. elata may preferentially guide Armillaria to colonize in its cortex layer and prevent it from excessive growth within the tubers; and ABC transporter transcripts, which mediate strigolactone secretion to the extracellular space, are highly abundant in the cortex layer associated with the secretion of some antimicrobial components (e.g., phytoalexin gastrodin and Gastrodia antifungal protein) [14]. Virulence experiments/tests showed that Armillaria mellea has a greater or equal virulence than Armillaria ostoyae, A. ostoyae has a greater virulence than Armillaria gallica and Armillaria cepistipes, and Armillaria tabescens has the weakest virulence among those five species [15–18]. Guo et al. hypothesized that weakly pathogenic and preferentially saprotrophic Armillaria species (e.g., A. gallica, and A. cepistipes) may readily establish a symbiotic relationship with Tian Ma rather than cause a disease to tubers [19].

Six draft genomes of Armillaria had been reported as wood pathogens [20–22]. Previous studies had revealed some genes associated with plant cell wall degradation (e.g., lignin-, cellulose-, hemicellulose-, and pectin-degrading enzymes), pathogenicity, (e.g., expansins, salicylate hydroxylases, and cerato-platanin genes), and secondary metabolites (e.g., terpene cyclases and non-ribosomal metabolites) in Armillaria genomes [23]. However, no specific study had been conducted at a gene level to clarify the characteristics of Armillaria strains and how they established a symbiotic relationship with G. elata.

Here, we present a draft genome of Armillaria strain 012m isolated from the tuber of G. elata collected from a plantation field in Yiliang, Zhaotong, Yunnan Province, China. The evolutionary relationships between Armillaria strain 012m and other 15 Agaricales strains, including four Armillaria strains identified as forest pathogens previously, had been studied. The difference of the gene families implied in pathogenicity/saprophytic phase between Armillaria strain 012m and four Armillaria strains identified as forest pathogens was investigated. In addition, the genome-guided analysis of differential expression between rhizomorph (RH) and vegetative mycelium (VM) was performed. This research provides genomic perspective to explore the symbiotic relationship between Armillaria strain 012m and G. elata.

Materials and methods

Material and sequencing

The experimental material of Armillaria gallica 012m was isolated from the G. elata tuber collected from a plantation field in Yiliang, Zhaotong, Yunnan Province, China, in 2017 (Fig. 1a). For genomic DNA, A. gallica 012m was grown in a modified Czapek–Dox medium (NaNO₃, 3 g; K₂HPO₄, 1 g; VB₁, 0.02 g; MgSO₄·7H₂O, 0.5 g; KCl, 0.5 g; FeSO₄, 0.01 g; sucrose, 30 g; yeast extract, 10 g; peptone, 13 g; ethanol, 20 g; malt extract, 10 g; and water, 1 L) in the dark at 25 °C for 15 days to produce fresh rhizomorph (Fig. 1b). High-quality genomic DNA was extracted using a Qiagen DNeasy plant kit (Qiagen, Hilden, Germany) in accordance with the instruction manual. Whole-genome sequencing involved a combination of Illumina HiSeq and Pacific Biosciences (PacBio) Sequel sequencing platforms. For RNA sequencing, strain was cultured in a modified Czapek–Dox medium to produce fresh RH and VM. Total RNA was isolated from tissues by using an RNeasy mini kit (Qiagen, Hilden, Germany). Sequencing was performed on Illumina HiSeq.

Fig. 1.

Armillaria gallica 012m colonized on Gastrodia elata (a). Armillaria gallica 012m grew on a modified Czapek–Dox medium (b)

Genome assembly, gene prediction, and annotation

The genome of A. gallica 012m was assembled with a custom pipeline. The strategy consisted of preparing corrected Illumina paired-end (PE) read data via Trimmomatic V0.36 [24] to trim reads and ErrorCorrectReads module from ALLPATHS-LG V52488 [25] to correct reads. The corrected PE reads were used as input data to LoRDEC V0.6 [26] to pre-correct PacBio subreads. MECAT2, an improved version of MECAT [27], was used to generate a primary assembly on the basis of pre-corrected PacBio subreads. An enhanced haploid sub-assembly was reconstructed from the primary assembly with Purge Haplotigs [28]. The haploid sub-assembly was upgraded by corrected PacBio subreads with FinisherSC V2.1 [29]. Final assembly was obtained via NextPolish V1.0.21 to polish upgraded assembly with PE data.

The tRNA genes were identified with tRNAscan-SE V1.3.1 [30], while the rRNA genes were predicted with RNAmmer V1.2 [31]. The repetitive sequences were detected and annotated by calling RepeatMasker V4.0.72 and RepeatModeler3 through RepBase-20181026 [32] and filtered by 0.25 coverage rate.

The draft gene models of A. gallica 012m were generated with GETA V2.4.24 annotation tool, which could integrate various evidences from ab initio gene finding and experimental data, including homologous proteins and RNA-Seq data. Homologous proteins were downloaded from Cylindrobasidium torrendii FP15055 [33], Omphalotus olearius VT 653.13 [34], and Gymnopus luxurians [35] at JGI MycoCosm [36]. Benchmarking Universal Single-Copy Orthologs (BUSCO) V3.0.2b [37] was used to assess the genome completeness with the lineage-specific profile library basidiomycota_odb9 (species: selected 41:25 with 1335 BUSCO groups).

All predicted protein-coding genes were functionally annotated based on their sequence similarity to proteins in the non-redundant database of Basidiomycota obtained from GenBank by DIAMOND V0.9.24 [38]. The automated pipeline InterProScan V5.44 [39] was utilized to determine motifs and domains. The results of DIAMOND (xml file) and InterProScan were imported into Blast2GO V5.2.5 [40] to retrieve the Gene Ontology (GO) terms. A summary image was drawn by WEGO 2.0 [41].

Genome comparison analyses

For the analysis of orthology and phylogenetics, the protein repertoires of fungi, namely, Armillaria ostoyae C18/9, Armillaria gallica Ar21-2, Armillaria cepistipes B5, Armillaria solidipes 28-4 [42], Cylindrobasidium torrendii HHB-15055, Fistulina hepatica ATCC 64428 [33], Amanita muscaria BX008, Gymnopus luxurians FD-317 M1, Laccaria amethystina LaAM-08-1 [35], Schizophyllum commune H4-8 [43], Agaricus bisporus H97 [44], Coprinopsis marcescibilis CBS121175 [45], Laccaria bicolor S238N-H82 [46], Coprinopsis cinerea okayama7#130 [47], and Pleurotus ostreatus PC15 [48], were downloaded from JGI MycoCosm and NCBI. Then, those were annotated with InterProScan V5.44 [39].

OrthoFinder V2.3.1 [49] was used to identify the groups of orthologs between the 16 strains. After that, KinFin [50], a software for taxon-aware analysis of clustered protein sequences, was employed to integrate the results from OrthoFinder and InterProScan to infer the functional annotation of clusters.

The translation elongation factor 1 alpha (tef1-α) was used to infer the phylogenetic relationships among 131 samples of Armillaria. The single-copy ortholog sequences extracted from the OrthoFinder V2.3.1 were clustered to construct a phylogenetic tree for 16 strains step by step. Firstly, the multiple sequence alignments were generated with MAFFT Alignment v7.221 [51]. Secondly, Gblocks V0.91b [52] and ProtTest V3.4 [53] were then utilized to extract conservative sites from MAFFT results and determine the best-fitting model, respectively. Finally, IQ-TREE V2.0-rc1 [54] was used to build a maximum likelihood (ML) phylogenetic tree with 1000 bootstraps (Supplementary Table S1). The best-fitting model was determined with jModelTest 2.1.7 [55], and the phylogenetic tree was plotted with FigTree.5

r8s V1.80 [56] was used to date the divergence times with the penalized likelihood method [57], the truncated Newton (TN) algorithm [58], and the optimal smoothing factor deduced via cross-validation among 50 values from 1 to 7.9e+04 in the ML tree. Two fossil calibration points [59, 60] were fixed in the molecule clock analysis, and the minimum age of 16 strains was set at 130 million years ago (Ma). The most recent common ancestor (tMRCA) of C. cinerea, L. bicolor, and S. commune was diverged at 122.74 Ma. Finally, the ultrametric tree was visualized with iTOL V4 [61].

The evolutionary relationships of orthologous gene families were calculated by CAFÉ V4.2.1 [62]. The downstream statistical analysis of CAFÉ was performed by following a tutorial.6 The expanded and contracted families were confirmed with Fisher’s exact test based on the results of the Pfam search. For each gene family, we compared the gene count of the tested family in A. gallica 012m (copy number of the tested family as numerator, total number of genes of the whole genome as denominator) with the frequency in A. gallica Ar21-2 and A. cepistipes B5.

Armillaria sp. was mostly described as pathogens harmful to plants. The genes related to pathogenicity such as cerato-platanin, expansin, deuterolysin, fungal hydrophobin, PR-1, PR-5/Thaumatin family, salicylate hydroxylase, and carboxylesterase [63, 64] were identified with InterProScan V5.44 [39] (default cutoff threshold) and BLASTP search in the Nr database with the cutoff threshold evalue ≤ 1e−5 and covered fraction ratio ≥ 0.2 (Supplementary Table S4).

CAZYme

The carbohydrate-active enzymes (CAZYmes) of 16 strains were identified and classified separately with a localized dbCAN service [65] and imported to CAFÉ V4.2.1 [62] for evolution analysis.

Differential expression analysis

Differential expression analysis of RH vs. VM was performed with the Trimmomatic-HISAT2-Stringtie-edgeR RNA-Seq pipeline. Firstly, we applied Trimmomatic V0.36 [24] to trim adaptor sequences and low-quality ends of all RNA-Seq reads. Subsequently, HISAT2 V2.1.0 and StringTieV1.3.3b were used to align clean reads to the assembled genome and assess gene expression, respectively. Finally, differentially expressed genes (DEGs) were defined by edgeR v3.28.1 [66] with |log2FoldChange| ≥ 1, p < 0.05 and corrected p value (padj) < 0.05. GO functional enrichment analyses of the DEGs were analyzed using GO EnrichMent module in TBtools v0.66836 [67]. GO terms with p < 0.05 and corrected p value (padj) < 0.05 were considered significantly enriched by DEGs.

Results and discussion

Genome assembly, gene prediction, and annotation

In the last few years, six draft genomes of Armillaria as pathogens of woods were reported, namely Armillaria fuscipes CMW2740, A. mellea DSM 3731, A. cepistipes B5, A. gallica Ar21-2, A. ostoyae C18/9, and A. solidipes 28-4. Those genomes were assembled to 106–29,300 scaffolds comprising 53–86 M [16–18]. However, the genomes of A. mellea DSM 3731and A. fuscipes CMW2740 were still highly fragmented, and the genomes of A. gallica Ar21-2 and A. solidipes 28-4 contained numerous ambiguous bases (Table 1). In our work, the genome of Armillaria gallica 012m was sequenced with a combination strategy of ~ 70× Illumina HiSeq and ~ 80× Pacific Biosciences Sequel. A total of 11.72 G raw subread data (1,109,518 subreads, with an average length of 10,562 bp) and 10.16 G PE raw read data were acquired. Through a custom assembly pipeline, we obtained an 87.3-M genome sequence of Armillaria gallica 012m, which consisted of 63 contigs, with a contig N50 of 2,159,699 bp and a GC content of about 47.38% (Table 1). There were approximately 23.6% (20,614,473 bp) repetitive sequences in the genome, while most of the repetitive sequences were unknown (Table 2).

Table 1.

Genome assembly statistics of Armillaria sp. 012m, A. gallica Ar21-2, A. cepistipes B5, A. ostoyae C18/9, A. solidipes 28-4, A. mellea DSM 3731, and A. fuscipes CMW2740

Information	A. gallica 012m	A. gallica Ar21-2	A. cepistipes B5	A. ostoyae C18/9	A. solidipes 28-4	A. mellea DSM 3731	A. fuscipes CMW2740
Scaffolds	63	319	287	106	229	29,300	24,403
Contigs	63	1866	740	106	848	65,823	27,509
Longest length	6,431,929	4,779,317	6,135,745	6,405,655	3,399,694	639,705	157,180
Shortest length	2287	1005	2753	4960	1069	105	200
Scaffold size	87,305,441	85,336,812	75,828,441	60,106,801	58,009,494	79,545,241	52,984,320
Contigs size	87,305,441	78,368,969	75,822,108	60,106,801	55,743,814	67,014,010	52,475,986
Rate of N	0.0000	0.0817	0.0001	0.0000	0.0391	0.1575	0.0096
Rate of GC	0.4738	0.4735	0.4771	0.4833	0.4835	0.4726	0.4768
Scaffold N50	2,159,699	1,035,263	3,291,351	2,283,935	715,667	24,647	5422
Contig N50	2,159,699	146,437	655,924	2,283,935	242,961	3268	4836
Scaffold N90	990,575	264,681	214,185	294,811	189,369	1708	866
Contig N90	990,575	23,431	149,086	294,811	42,686	376	775
Complete BUSCOs	95.8%	98.6%	95.1%	95.6%	98.4%	-	-
Duplicated BUSCOs	4.9%	3.8%	2.9%	2.2%	3.1%	-	-
Fragmented BUSCOs	3.3%	1.0%	4.0%	3.7%	1.3%	-	-
Missing BUSCOs	0.9%	0.4%	0.9%	0.7%	0.3%	-	-

Table 2.

The repetitive element length detailed statistics of A. gallica 012m on class or subclass

Class	Subclass	Number	Length (bp)
Total		-	20,614,473
DNA	IS3EU	202	358,705
	MULE-MuDR	41	36,955
	MULE-NOF	14	59,562
	PIF	2	46,769
	PIF-Harbinger	3	2298
	PiggyBac	8	27,537
	TcMar-Ant1	2	878
	TcMar-Fot1	1	848
	TcMar-Pogo	77	6196
	TcMar-Sagan	31	17,227
	TcMar-Tc1	175	229,228
	hAT-Ac	103	52,256
	hAT-Charlie	380	246,123
	hAT-hATw	13	25,608
LINE	CR1	29	82,173
	I-Jockey	34	19,810
	L1	154	65,969
	Penelope	2	2904
	RTE-X	87	25,282
LTR	Copia	321	394,715
	Gypsy	2062	4,868,997
	Ngaro	8	24,278
	Pao	56	43,154
RC	Helitron	149	342,344
	Helitron-2	7	97,122
LINE		38	27,013
LTR		18	60,641
Low_complexity		796	37,085
Satellite		3	5040
Simple_repeat		6995	294,800
Unknown		27,603	13,359,144
rRNA		12	27,314

A total of 26,261 genes were predicted in the genome by GETA. The genome completeness was 95.8% based on the BUSCO assessment, while 38 BUSCOs were fragmented and 12 BUSCOs were missing (Table 1). In addition, 12 rRNAs were defined with RNAmmer V1.2. Of these rRNAs, four were 8s rRNA, four were 28s rRNA, and four were 18s rRNA. tRNA V1.3.1 found 353 tRNAs and 284 tRNAs with introns.

At the first level, GO annotation divided genes into three groups, namely molecular function (MF), cellular component (CC), and biological process (BP) (Fig. 2). A total of 12,055 genes were mapped to 22,947 GO terms, including 8892 in the MF domain, 6961 in the BP domain, and 7094 in the CC domain. Based on molecular function, 5545, 472, and 5800 genes were annotated for catalytic activities, transporter activities, and binding, respectively. For the BP domain, metabolic process (5491), cellular process (4485), and biological regulation (1352) were the most abundant terms. Membrane (18.9%) and membrane part (16.4%) were the most abundant terms in the CC domain.

Fig. 2.

The summary of GO (Gene Ontology) terms under level 2

Genome comparison analyses

A total of 25,627 ortholog cluster groups (OCGs) of 16 strains were constructed with OrthoFinder V2.3.1 [49]. In total, 4145 OCGs were core gene families shared in all the species. Approximately 79.4% (20852) putative proteins of A. gallica 012m were grouped under 12,674 (approximately 49.5%) OCGs. In particular, there were 6446 strain-specific proteins that existed in A. gallica 012m.

A total of 651,650 conservative positions (40%) from 2032 single-copy orthologs were used to produce a maximum likelihood tree with best “LG+I+G+F” model and 1000 bootstraps in IQ-TREE V2.0-rc1. The phylogenetic tree showed a relatively close genetic relationship between Armillaria gallica 012m and A. gallica Ar21-2 with a bootstrap support of 100% (Fig. 3). Interestingly, the phylogenetic relationship among Armillaria samples inferred from tef1-α sequence revealed that A. gallica Ar21-2 and Armillaria sp. 012m were clustered into two distinct A. gallica clusters, although Armillaria sp. 012m was still clustered into A. gallica (Fig. 4).

Fig. 3.

Phylogenetic tree of Armillaria gallica 012m with other 15 fungal species. The topology of the phylogenetic tree was constructed by the maximum likelihood method (bootstrap = 1000, LG+I+G+F model), and all bootstrap values were 100%. Time scale was shown by Ma (million years ago)

Fig. 4.

Phylogenetic relationships between samples of Armillaria inferred from tef1-α sequences using the ML analysis. The topology of the phylogenetic tree was constructed by using the maximum likelihood method (bootstrap = 1000, TrNef+I+G model). The accession numbers for the sequences retrieved from the GenBank database are listed in Supplementary Table S1

r8s V1.80 was used to infer the divergence time of the A. gallica 012m strain. The results showed that the divergence time of A. gallica 012m and A. gallica Ar21-2 was at ~ 2.20 Ma (Fig. 3). Computational analysis on the gene family evolution was achieved with CAFÉ V4.2.1 [62]. A total of 378 OCGs, including 125 expansions and 253 contractions, significantly deviated from these genomic background rates on the A. gallica 012m branch (p < 0.01). Emphasis was placed on 62 OCGs, which were inferred to possess the same functional annotation in KinFin. The 44 contracted OCGs consisted of cytochrome P450 (n = 7), caspase domain (n = 7), fungal hydrophobin (n = 1), glucose–methanol–choline (GMC) oxidoreductase (n = 1), etc. (Table 3). The 13 expanded clusters were annotated as cytochrome P450 (n = 2), caspase domain (n = 1), GMC oxidoreductase (n = 1), etc. (Table 3). According to the results of the Pfam search, some extended and reduced families were reconfirmed with Fisher’s exact test (Supplementary Table S2).

Table 3.

Significantly expanded and contracted annotated orthogroups in A. gallica 012m

Family ID	KinFin annotation		p
OG0000003[-13*]	PF01185	Fungal hydrophobin	3.76E−15
OG0000008[-3*]	PF18759	Plavaka transposase	4.66E−04
OG0000009[-13*]	PF00732;PF05199	GMC oxidoreductase; GMC oxidoreductase	2.10E−16
OG0000010[-7*]	PF00067	Cytochrome P450	4.22E−08
OG0000012[-16*]	PF00067	Cytochrome P450	9.86E−26
OG0000027[-14*]	PF00724	NADH:flavin oxidoreductase/NADH oxidase family	4.77E−22
OG0000042[-3*]	PF13561	Enoyl-(acyl carrier protein) reductase	7.16E−04
OG0000047[-2*]	PF18759	Plavaka transposase	4.06E−03
OG0000059[-17*]	PF00067	Cytochrome P450	1.55E−29
OG0000071[-3*]	PF00067	Cytochrome P450	1.03E−03
OG0000085[-9*]	PF00248	Aldo/keto reductase family	2.26E−14
OG0000099[-8*]	PF00106	Short-chain dehydrogenase	7.78E−13
OG0000100[-2*]	PF05970	PIF1-like helicase	1.76E−03
OG0000140[-6*]	PF00656	Caspase domain	1.21E−07
OG0000145[-3*]	PF00005	ABC transporter	3.66E−04
OG0000168[-4*]	PF01734	Patatin-like phospholipase	6.02E−06
OG0000174[-3*]	PF00067	Cytochrome P450	1.01E−04
OG0000197[-11*]	PF00067	Cytochrome P450	6.78E−18
OG0000214[-2*]	PF18758	Kyakuja-Dileera-Zisupton transposase	5.01E−03
OG0000254[-5*]	PF01485	IBR domain, a half RING-finger domain	2.49E−02
OG0000273[-5*]	PF01713;PF08590	Smr domain; domain of unknown function (DUF1771)	7.41E−08
OG0000310[-2*]	PF00867	XPG I-region	1.76E−03
OG0000311[-10*]	PF00067	Cytochrome P450	9.19E−17
OG0000327[-2*]	PF00651	BTB/POZ domain	7.25E−04
OG0000399[-5*]	PF00106	Short-chain dehydrogenase	3.75E−08
OG0000486[-3*]	PF17921	Integrase zinc-binding domain	8.47E−06
OG0000506[-3*]	PF01753	MYND finger	6.49E−05
OG0000508[-3*]	PF10419	TFIIIC subunit triple-barrel domain	6.49E−05
OG0000556[-4*]	PF01026	TatD-related DNase	2.17E−07
OG0000696[-3*]	PF07714	Protein tyrosine kinase	3.82E−05
OG0000789[-3*]	PF08534	Redoxin	1.99E−05
OG0000824[-2*]	PF13359	DDE superfamily endonuclease	1.28E−04
OG0000897[-2*]	PF01765	Ribosome recycling factor	7.25E−04
OG0001014[-5*]	PF02902	Ulp1 protease family, C-terminal catalytic domain	4.52E−10
OG0001468[-3*]	PF01693	Caulimovirus viroplasmin	2.48E−06
OG0001494[-2*]	PF00656	Caspase domain	1.19E−03
OG0001687[-4*]	PF00656	Caspase domain	2.17E−07
OG0001708[-2*]	PF00656	Caspase domain	1.28E−04
OG0001709[-4*]	PF01753	MYND finger	6.69E−08
OG0004147[-2*]	PF11917	Protein of unknown function (DUF3435)	3.70E−04
OG0006593[-4*]	PF00656	Caspase domain	6.69E−08
OG0007387[-3*]	PF00656	Caspase domain	2.48E−06
OG0008912[-2*]	PF00656	Caspase domain	1.28E−04
OG0010195[-2*]	PF01693	Caulimovirus viroplasmin	1.28E−04
OG0000069[+3*]	PF13417	Glutathione S-transferase, N-terminal domain	8.63E−04
OG0000089[+3*]	PF12770	CHAT domain	4.89E-04
OG0000092[+13*]	PF00656	Caspase domain	9.66E−19
OG0000123[+2*]	PF17667	Fungal protein kinase	9.13E−03
OG0000233[+6*]	PF03171;PF14226	2OG-Fe(II) oxygenase superfamily; non-haem dioxygenase in morphine synthesis N-terminal	8.97E−09
OG0000320[+2*]	PF00067	Cytochrome P450	3.72E−03
OG0000448[+4*]	PF05699	hAT family C-terminal dimerization region	9.10E−07
OG0000461[+2*]	PF05686	Glycosyl transferase family 90	2.39E−03
OG0000509[+7*]	PF01040	UbiA prenyltransferase family	7.56E−11
OG0001226[+7*]	PF02182	SAD/SRA domain	2.72E−11
OG0001490[+15*]	PF00067	Cytochrome P450	3.96E−27
OG0002857[+3*]	PF00385	Chromo (CHRromatin Organization MOdifier) domain	4.60E−05
OG0003496[+4*]	PF18803	CxC2 like cysteine cluster associated with KDZ transposases	2.63E−08
OG0006849[+3*]	PF08284	Retroviral aspartyl protease	1.74E−06
OG0007274[+5*]	PF00732	GMC oxidoreductase	3.99E−10
OG0007470[+7*]	PF00075	RNase H	7.14E−13
OG0009953[+2*]	PF07727	Reverse transcriptase (RNA-dependent DNA polymerase)	5.67E−04
OG0011360[+3*]	PF00268	Ribonucleotide reductase, small chain	1.74E−06

CAZymes involved in plant cell wall degradation

A total of 824 CAZymes were predicted in the genome of A. gallica 012m, which included 331 glycoside hydrolases (GHs), 25 polysaccharide lyases (PLs), 142 carbohydrate esters (CEs), 82 carbohydrate-binding modules (CBMs), 170 auxiliary activity enzymes (AAs), and glycosyl transferases (GTs). Compared with the genomes of other Armillaria, A. gallica 012m had more CAZymes than A. solidipes 28-4 and A. ostoyae C18/9, while it had a similar number of CAZymes compared with A. cepistipes B5 and A. gallica Ar21-2. Therefore, it had the largest set of GHs and CBMs among the five strains (Supplementary Table S3).

Armillaria sp. is well known as saprophytic fungi under white-rot (WR) fungi; many CAZymes involved in the degradation of lignin, cellulose, hemicellulose, and pectin were found in the genomes of five Armillaria strains (Supplementary Table S3). The computational analysis on the gene family evolution showed that the AA3 family was significantly contracted (5 losses compared with tMRCA of A. gallica 012m and A. gallica Ar21-2) in A. gallica 012m. AA3 enzymes, which belong to the GMC oxidoreductase family, contain a flavin–adenine dinucleotide–binding domain and were divided into four subfamilies [68]. AA3 enzymes of Armillaria were mostly classified under the AA3-2 subfamily (including both aryl alcohol oxidase and glucose 1-oxidase). The AA3-2 subfamily had been extensively studied in several well-known white-rot fungi and considered as important enzymes in the biodegradation of lignocellulose [69, 70]. The AA3-2 subfamily of A. gallica 012m had only 45 representatives, while A. gallica Ar21-2 had 58 representatives. This copy number was also lower than that of A. cepistipes B5, A. ostoyae C18/9, and A. solidipes 28-4 (Supplementary Table S3). Previous research revealed that certain CAZyme families have contracted during the evolution of mycorrhizal fungi compared with their saprotrophic ancestors [35]. Our result coincided with previous research.

Cytochrome P450 monooxygenase, as a multifunctional oxidoreductase, is distributed in living organisms [59]. In fungi, P450 enzymes not only participate in the production of a variety of metabolites but also play a key role in adaptation to specific ecological and/or nutritional niche such as wood degradation [71–73]. A total of 271 cytochrome P450s were identified in A. gallica 012m. The number was close to that of A. ostoyae C18/9 and A. solidipes 28-4 but much less than that of A. gallica Ar21-2 (n = 330) and A. cepistipes B5 (n = 324) (Supplementary Table S2).

Genes implied in pathogenicity

Some genes implied in pathogenicity of the five Armillaria strains were investigated. Compared with A. gallica Ar21-2 and A. cepistipes B5, the number of hydrophobin (HP) gene of A. gallica 012m was significantly reduced (Supplementary Table S4). HPs, as small secreted cysteine-rich amphiphilic proteins even isolated from Ascomycetes, Basidiomycetes, and Zygomycetes, might play several roles during fungal growth and plant infection [74–76]. HPs could self-assemble at water–air interfaces to provide water-repellent properties to surfaces of conidia, spores, hyphae, and fruiting structures [76–79]. Therefore, HPs were involved in conidium germination, fruit body development, infectious structure formation, and fungal pathogenicity [80–83]. Cerato-ulmin (CU) was a kind of HPs from Ophiostoma ulmi and Ophiostoma novoulmi that cause Dutch elm disease. The sensitivity of different plants, correlated with the presence of CU, could induce embolisms by stabilizing air bubbles in xylem vessels [84].

In a previous research, an idea was put forward that G. elata established an association with Armillaria spp. to obtain nutrition, and the HPs of these strains provide protection against chemical and enzymatic attacks by their hosts [76]. We hypothesized that HPs may have played an important role in the symbiotic relationship between strains and G. elata. Less hydrophobins of Armillaria sp. were beneficial for G. elata to obtain nutrition and get less injuries.

Differential expression analysis

In the life cycle of Armillaria, there are three morphological differences: fruit bodies, VM, and RH, of which VM and RH are related to the symbiotic relationship with G. elata [10, 85]. Six RNA samples of RH and VM were sequenced by Illumina HiSeq platform to obtain 20,625,024; 22,920,597; 19,855,909; 19,544,248; 20,817,573; and 19,964,304 pairs of PE reads, respectively. After removing the reads with adapter sequences and of low quality, an average of 20,474,936 and 19,510,486 pairs of clean reads were retained from RH and VM, respectively. These clean reads were then aligned to the reference genome, and they all had an overall alignment rate of over 93.5 (Table 4).

Table 4.

Summary statistics for sequencing for six libraries from RH and VM

Sample_name	Raw_reads	Clean_reads	Read mapped (%)
RH1	20,625,024	19,955,816	94.53
RH2	22,920,597	22,263,584	94.25
RH3	19,855,909	19,205,408	93.88
VM1	19,544,248	18,959,106	93.79
VM2	20,817,573	20,206,083	93.96
VM3	19,964,304	19,366,270	94.24

In the RH vs. VM comparison, a total of 2549 genes were differentially expressed, including 1917 upregulated genes and 632 downregulated genes (Fig. 5). Thirty-one DEGs associated with pathogenicity were significantly upregulated, including 2 cerato-platanins, 10 carboxylesterases, 9 fungal hydrophobins, 5 salicylate hydroxylases, 2 PR-1, 1 expansins, 1 PR-5, and 1 deuterolysin, while three DEGs associated with pathogenicity were significantly downregulated, including 1 carboxylesterase, 1 fungal hydrophobin, and salicylate hydroxylase (Supplementary Table S5).

Fig. 5.

Volcano map of differentially expressed genes. Differentially expressed genes (DEGs) were defined by edgeR with |log2FoldChange| ≥ 1, p < 0.05 and corrected p value (padj) < 0.05

To further elucidate gene function, the GO enrichment analysis was carried out on DEGs. The upregulated DEGs were significantly grouped into the 24 molecular functions, such as monooxygenase activity (GO:0004497), oxidoreductase activity (GO:0016491), hydrolase activity (GO:0016787), glucosidase activity (GO:0015926), and coenzyme binding (GO:0050662); 8 cellular components, such as membrane part (GO:0044425), membrane (GO:0016020), extracellular region (GO: 0005576), and fungal cell wall (GO:0009277); and 10 biological processes, including oxidation–reduction process (GO:0055114), carbohydrate metabolic process (GO:0005975), xenobiotic metabolic process (GO:0006805), response to xenobiotic stimulus (GO:0009410), and response to toxic substance (GO:0009636) (Fig. 6). For downregulated DEGs, only five membrane-related GO terms, namely, membrane component, membrane, transmembrane transport (GO:0055085), membrane component (integral component of membrane (GO:0016021)), and intrinsic component of membrane (GO:00312245), were significantly enriched. These phenomena indicate that RH had better infection ability than VM.

Fig. 6.

Gene Ontology (GO) enrichment analysis of upregulated differentially expressed genes (DEGs)

A previous study had shown that the RH of Armillaria infected the nutritional stems of G. elata and formed hyphal stream. The hyphal stream would infect the cells of the cortex layer in an outward direction and infect the large cell layers in an inward direction. However, the hyphae that infected the cortical cells and the large cell layers could be digested by G. elata as nutrition. When the RH of Armillaria connected with G. elata was cut off, the new corms would stop growing [86]. Therefore, the infection ability of the rhizomorph of Armillaria is very important for G. elata to obtain nutrition.

Conclusion

Here, we report an 87.3-M genome of the Armillaria 012m strain, which was symbiotic with G. elata. Repetitive sequences represent approximately 23.6% (20,614,473 bp) of the genome. A total of 26,261 genes were predicted in the genome. The genome comparison analyses with other 15 Agaricales strains showed that the genes related to pathogenicity/saprophytic, encoding cytochrome P450 monooxygenases, the carbohydrate-active enzyme AA3 family, and hydrophobins, were significantly contracted in A. gallica 012m. These characteristics may be beneficial for its symbiotic relationship with G. elata.

The analysis of differential expression between RH and VM showed that a total of 2549 genes were differentially expressed, including 632 downregulated genes and 1917 upregulated genes. In the RH, most DEGs related to pathogenicity were significantly upregulated. To further elucidate gene function, the GO enrichment analysis showed that the function of the upregulated DEGs significantly grouped into the regulation of twenty-four molecular functions, such as monooxygenase activity, oxidoreductase activity, hydrolase activity, and glucosidase activity; 8 cellular components, such as membrane part, membrane, extracellular region, and fungal cell wall; and 10 biological processes, including oxidation–reduction process, carbohydrate metabolic process, xenobiotic metabolic process, response to xenobiotic stimulus, and response to toxic substance. For downregulated genes, only five membrane-related terms were significantly enriched. These phenomena indicate that RH had better infection ability than VM. The infection ability of rhizomorph is very important for G. elata to obtain nutrition, because the rhizomorph of Armillaria constantly infected the nutritional stems of G. elata and formed the hyphae that can be digested by G. elata, which could continuously provide nutrition for the whole growth period of G. elata. These results clarified the characteristics of A. gallica 012m and the reason why the strain 012m can establish a symbiotic relationship with G. elata in some extent from the perspective of genomics.

Nucleotide sequence accession numbers

The sequences of genome of A. gallica 012m was deposited in GenBank under the accession number VTST01000000. The RNA-Seq data of rhizomorph and vegetative mycelium were available in the NCBI SRA repository, with the accession number PRJNA633976 (https://www.ncbi.nlm.nih.gov/sra/PRJNA633976).

Funding information

This research was financially supported by the National Natural Science Foundation of China (Nos. 81860624 and 31760096) and Yunnan Innovative Research Team for Discovery and Comprehensive Utilization of Functional Small Molecules in Medicinal Plants.