Introduction

Arbuscular mycorrhizae (AM) are formed during symbiosis between host plants and soil fungi. In AM, the plant provides carbon in the form of lipids and sugar to the fungus, and receives water and nutrients in return (Oldroyd 2013; Rich et al. 2017). AM can also confer resistance to pathogens and abiotic stress (Ceballos et al. 2013; Aliyu et al. 2019; Begum et al. 2019; Ramírez-Flores et al. 2020). AM are conserved in the majority of plant species from a single origin through the present day, and are thought to have aided in the first colonization of land (Wang et al. 2010; Delaux et al. 2013; Genre et al. 2020). However, multiple independent losses of the AM trait and its genetic machinery have occurred in diverse plant clades (Cosme et al. 2018). Why a trait considered to be beneficial has been so often lost is a puzzle important for not only basic understanding of AM symbiosis' role in plant resilience, but also for potential crop improvement through re-engineering the trait (Hornstein and Sederoff 2023).

Arabidopsis thaliana is one of the ~ 7% of plant species that lack the ability to form AM and which have not replaced them with an alternative endosymbiosis (Veiga et al. 2013; Brundrett 2017; Brundrett and Tedersoo 2018; Cosme et al. 2018; Radhakrishnan et al. 2020). Other such non-mycorrhizal (nonAM) plants include economically important species in the Brassicaceae and Amaranthaceae (Brundrett 2017). Proposed reasons for AM loss include changes in root morphology and lifestyle, altered insect and pathogen resistance, and the carbon cost of supporting the symbiont (Sikes 2010; Field et al. 2012; Brundrett 2017; Ma et al. 2018; Poveda et al. 2019). Little experimental validation exists for drivers of AM loss in individual species, let alone an overarching explanation for which species lose AM, why, and when. The proximate genetic causes for mycorrhizal loss, however, are quite clear: a shared subset of specific genes is deleted in all independent nonAM evolutions (Delaux et al. 2014; Radhakrishnan et al. 2020).

Gene losses in nonAM plants include the Common Symbiosis Pathway (CSP) that mediates signal perception and transduction (Cope et al. 2019; Radhakrishnan et al. 2020; Genre et al. 2020). Member genes lost in Arabidopsis include the cell-surface receptor SymRK, the ion channel DMI1, the kinase DMI3, and the transcription factor IPD3 (Delaux et al. 2014; Bravo et al. 2016; Radhakrishnan et al. 2020). These genes mediate a signal transduction pathway leading from AMF perception to activation of DMI3 by calcium (Demchenko et al. 2004; Pan et al. 2018; Feng et al. 2019). DMI3 then phosphorylates IPD3, enabling its DNA-binding activity for regulation of downstream targets (Oldroyd 2013; Pimprikar and Gutjahr 2018). Outside the CSP, groups of genes involved in lipid flux from plant to AMF and in vesicle trafficking to the arbuscule are also lost (Radhakrishnan et al. 2020).

IPD3 (Interacting Protein of DMI3) is notable among lost genes in nonAM plants. While most such genes belong to families with non-AM-specific homologs, IPD3 does not, suggesting its function is distinct to symbiosis. Genetic work has also demonstrated the powerful role of this gene in turning CSP signaling into regulation of AM response genes. The ipd3 knockout phenotype is near-complete elimination of mycorrhization (Watts-Williams and Cavagnaro 2015). Ectopic expression of IPD3 constitutively activated via phosphomimicking (IPD3S50D) or truncation to the DNA-binding domain (IPD3Min), however, results in symbiosis-like gene regulation even in the absence of a microbial signal or upstream CSP genes (Singh et al. 2014; Pimprikar et al. 2016).

Here, we express IPD3 in Arabidopsis based on knowledge of its uniquely powerful function in AM. By restoring expression of this gene that was present in the mycorrhizal ancestors of Arabidopsis before loss of AM in the Brassicaceae, we apply a novel means of identifying retained or unrecognized connections between AM and other genetic networks conserved in nonAM plants. Characterization of IPD3-expressing Arabidopsis targets two specific questions. First, does IPD3 have roles outside of its canonically narrow function in AM establishment? Second, does IPD3 retain functional molecular relationships when restored to Arabidopsis, which lost its ancestral copy of this gene along with the trait approximately 65 million years ago (Hohmann et al. 2015)? If so, could reconnecting such relationships alter the response to AMF or even restore symbiosis?

We use transcriptomics to understand the gene-regulatory impact of restoring IPD3 expression to Arabidopsis, and we also compare the inverse case of IPD3 knockout in a mycorrhizal host plant, via the cyclops-4 mutant of Lotus japonicus. While Lotus has the upstream CSP that normally activates IPD3 in the presence of AMF, the cyclops-4 mutant lacks functioning IPD3 and results in Lotus that fails to form mycorrhizae despite having the upstream CSP elements intact (Yano et al. 2008; Singh et al. 2014; Pimprikar et al. 2016). This Lotus mutant phenotypically mirrors wild type Arabidopsis in lacking AM, though Arabidopsis further lacks many other genes as described above (Delaux et al. 2013). We subjected both genotypes of both species to AMF treatment, thus, our experiment generates a factorial set of transcriptome data for every combination of IPD3 genotype, [non]symbiotic species background, and AMF exposure (Fig. 1). Because AMF cannot colonize or survive in long-term monoculture with nonAM plants, we use a pre-germinated spore treatment as described by Fernández et al. to study the early phase of plant-AMF interaction after 48 h of co-culture (Fernández et al. 2019). This corresponds to the pre-symbiotic phase of AM host plant-AMF interaction where significant CSP activity takes place but little colonization is established (Gutjahr and Parniske 2013).

Fig. 1
figure 1

Experimental design comparing different species and genotypes. A The AM-host Lotus japonicus wildtype (ecotype Gifu) and IPD3 knockout (cyclops-4; cyc4) genotypes differ in their ability to establish AM. Arabidopsis thaliana as a non-host species was engineered to express the DNA binding domain of IPD3 (OX-IPD3Min). B Phenotypic and transcriptomic analysis was carried out on the genotypes with or without AM inoculation and C under high and low macronutrient-containing media. Two independent transgenic lines of Arabidopsis were used for analysis in B and 1 transgenic line of Arabidopsis was used for analysis in C

Full size image

We additionally examined a set of AMF-exposed Arabidopsis plants under low-macronutrient conditions (N, P, K) as a known regulator of symbiotic interactions in host plants (Bonneau et al. 2013; Nouri et al. 2014; Carbonnel and Gutjahr 2014). We perform correlation network and differential expression analyses to identify patterns in transcriptomic data from this experiment, and explore their functional implications using gene ontology (GO).

Materials and methods

Please see supplemental experimental procedures in Online Resource 1 for additional detail on all sections.

Generation of transgenic Arabidopsis

The coding sequences of MtIPD3, S50D-IPD3, and IPD3-Min (Genbank EF569224.1; Yano et al. 2008; Singh et al. 2014) were synthesized (Integrated DNA Technologies, Research Triangle Park, NC) and assembled by Hi-Fi assembly (New England Biolabs, Ipswich, MA) into a modified pCAMBIA0380 expression construct (Genbank AF234290.1) under control of the Arabidopsis Ubiquitin 10 promoter (Ivanov and Harrison 2014). 35S:mCherry amplified from pC-GW-mCherry (Genbank KP826771.1) (Dalal et al. 2015) was the selection marker. Arabidopsis were transformed as described by Davis et al. (2009). Seeds were screened by fluorescence and PCR during segregation and lines were brought to homozygosity.

Protein analysis

Protein was extracted from roots and shoots of 6-week-old seedlings grown on 1/2MS plates. For untargeted shotgun proteomics, 5 samples per line and tissue type were ground and lysed in SDT buffer, then prepared as described in Wiśniewski et al. (2009) before 1D-LC-MS/MS using an UltiMateTM 3000 RSLCnano Liquid Chromatograph and Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher) as described in Mordant and Kleiner (2021). Mass spectra were searched against a database of Col-0 A. thaliana proteins (Uniprot:UP000006548) using the SEQUEST HT algorithm in Proteome Discoverer. Protein abundance was quantified as normalized spectral abundance factor (NSAF) in Microsoft Excel (Zybailov et al. 2006).

For analysis of specific protein size ranges corresponding to bands observed on blots or predicted protein length, we used in-gel digestion of excised SDS-PAGE prior to LC-MS/MS analysis (GeLC-MS/MS). Only 1 sample per construct was used for GeLC-MS/MS. 30–40 µg of total protein in lysate generated for untargeted proteomics was denatured by heating with Laemmli buffer, then run on an SDS-PAGE 12% separating gel with 5% stacking gel. Excised bands were processed according to Shevchenko et al. (2006), and 10 μL of peptide mixture was injected to LC as described above, connected to an Orbitrap Exploris 480 mass spectrometer (Thermo-Fisher) for MS/MS with the same settings used for untargeted proteomics.

For Western blotting, frozen tissue was hand-ground with extraction buffer, centrifuged, and the supernatant denatured by heating in LDS sample buffer. Protein was run on a 12% Tris-Glycine acrylamide gel, then transferred to PVDF membranes and blocked in TBS+2% BSA with 4 μL/mL Tween-20, then incubated with 1:1000 dilution of custom rabbit anti-IPD3 polyclonal antibody (Genscript, China) followed by 1:2500 donkey anti-rabbit AlexaFluor 488-conjugated secondary antibodies (Thermo Fisher, Waltham, MA) and imaged on a GelDoc SR (Bio-Rad, Hercules, CA).

Growth phenotyping

Plants were grown in the NCSU phytotron under long day conditions in 8 oz pots filled with SunGro propagation mix (Sungro, Agawam, MA). Pots were hand-watered with deionized water. Seeds were harvested from dried mature plants and hand-cleaned before weighing. The seed yield experiment was repeated 3 times with n = 9–15. The flowering time experiment was performed once with n = 14 for line 312 and n = 15 for all other lines, with plants censused daily for onset of bolting.

Transcriptome experiment

Seeds were sterilized and grown for 6 weeks under long-day conditions (16 h light/8 h dark, 21/18 ℃) on sterile petri dishes containing either 1/2MS or low-nutrient MS media (Phytotech Labs, Lenexa, KS). Following the protocols established by Fernández et al. (2019) and Mukherjee and Ané (2011), sterile spores of R. irregularis (Premier Tech, Canada) were pre-germinated for 1 week at 26 ℃, and a suspension of ~ 200 germinated spores was added to the roots of each AMF-treated plant 48 hours prior to sample collection. Control plants were mock-inoculated with sterile water. Each replicate consisted of the pooled roots of 5 seedlings from the same plate.

RNA was extracted with the Purelink RNA Mini kit and treated with Turbo DNAse (Invitrogen, Waltham, MA), then sequenced by DNBSEQ at BGI Group (China) using strand-specific, poly-A enrichment to obtain 100 bp paired-end reads. Lotus and Arabidopsis alignments were performed using BBSplit, a multi-reference aligner, to separate plant and fungal reads (Bushnell 2014). Reference genomes used were Araport 11 for Arabidopsis (TAIR), the Joint Genomics Institute assembly for R. irregularis (Genbank: GCA_000439145.3), the Gifu V1.2 assembly for Lotus (GCA_012489685.2), the IPD3-Min transformation construct, and the LjCYCLOPS sequence. Reads were counted at the transcript level using featureCounts (Liao et al. 2014).

Gene expression networks were constructed using WGCNA v1.69 (Langfelder and Horvath 2008, 2012) with soft-threshold power of 16 (Arabidopsis) or 24 (Lotus). Correlation coefficients were calculated between the eigengenes of each module and treatment variables to identify significant module-trait relationships, with p < 0.05. Gene Ontology enrichment and hierarchical term clustering was performed with PANTHER (Mi et al. 2013). Terms were filtered by specificity and significance, then subjected to semantic similarity clustering in reviGO (Supek et al. 2011).

Differential expression analysis was performed in R using the edgeR package (Liao et al. 2014; Robinson et al. 2010). The estimateGLMCommonDisp function with FDR-adjusted p-value < 0.05 was used to test for differential expression. Cross-comparison of gene lists was done in Excel.

Results

Generation of transgenic Arabidopsis

The transcription factor IPD3 consists of an activation domain that can be phosphorylated on Serine-50 upon AM inoculation and a DNA-binding domain essential for transcriptional regulation of downstream effectors (Singh et al. 2014; Jin et al. 2018). To ensure that we can identify transcriptional regulation via IPD3 in the non-host that is likely lacking signaling and activation of IPD3, we generated 7 homozygous transgenic Arabidopsis lines expressing Medicago truncatula IPD3 (IPD3Mt), 4 lines of phosphomimic IPD3 (IPD3S50D), and 5 lines expressing IPD3Min (aa 254-513 of IPD3Mt) (Singh et al. 2014). All transgenes were expressed under the Arabidopsis UBIQUITIN10 promoter (Grefen et al. 2010; Ivanov and Harrison 2014). RT-PCR confirmed RNA expression of all versions of the transgene.

Western blot of root and shoot tissue of T3 individuals confirmed presence of IPD3Min protein, but not IPD3Mt and IPD3S50D (Online Resource 2). We noted that Western blot visualization was difficult to obtain, requiring the precise method described in Online Resource 2, and that root and shoot IPD3Min protein bands appear to run at slightly different size, which might represent influence of unknown post-translational modification or cellular localization that interferes with expression and running conditions. We therefore performed targeted protein sequencing using mass spectrometry to confirm that the protein bands observed on the blot were in fact IPD3Min. Excised SDS-PAGE gel bands at the expected size of IPD3Min in shoot and root samples were used for shotgun proteomics (Online Resources 1 and 3). IPD3Mt and IPD3S50D lines were analyzed in the same manner, however, and as with Western blot of these lines, protein expression was again not detected. To further confirm that our ability to detect and identify all forms of IPD3 was not affected by gel running conditions or extraction, we then performed untargeted shotgun proteomics of shoots and roots (n = 5) using independent processing methods (Online Resource 1). In untargeted proteomic analysis, IPD3Mt and IPD3S50D were again not detected, while IPD3Min was highly abundant (top 10% NSAF) in respective lines (Online Resource 3). Consequently, we focused only on characterizing the 5 IPD3Min lines (numbered 303; 308; 310; 312; 357).

In IPD3Min lines used we confirmed that protein and RNA of the gene of interest were highly expressed relative to native sequences (Online Resources 3 and 15), however, these quantitative data do not provide a direct measurement of transcription factor activity of the protein product.

Phenotypic effect of IPD3 Min expression in Arabidopsis

Plants were surveyed for differences in growth under long-day conditions (Fig. 2, Online Resource 4). All 5 transgenic IPD3Min lines were significantly slower to initiate the transition to flowering than wild type Col-0, measured as days to onset of bolting (Fig. 2A, B, Online Resource 5). Four of 5 lines (excluding 303) were also significantly slower to flower than a null segregant control (Fig. 2B). Despite flowering differences, no significant difference in seed yield was detected over 3 repetitions of the experiment (Fig. 2C). No difference in germination timing was observed (Online Resource 5). An unexpected strong phenotype was the pink coloration of four of the five IPD3Min transgenic plant roots, while this color in IPD3Mt and IPD3S50D lines was very weak in 4 out of the 11 plants, and visually absent in the remaining 7 (Online Resource 6). In addition, 7 of the 8 independent T1 empty vector lines showed a range of pink coloration in roots. Because anthocyanins or carotenoids can cause this coloration in plant tissues, we performed extraction protocols for those secondary metabolites, but were not able to confirm any measurable quantities. The most likely conclusion is that the color could be caused by the marker mCherry that was used as a reporter gene. This reporter gene was present in the IPD3Mt and IPD3S50D constructs as well, where little to no pink coloration was observed, which may relate to the lack of expression and potential silencing of those transgenes as assessed through proteomics (Online Resources 2, 3).

Fig. 2
figure 2

Phenotypes of independent homozygous transgenic lines expressing the DNA-binding domain of IPD3 (IPD3Min). A Shoots of most IPD3Min transgenic lines appear shorter, less branched and delayed in development; scale bar 10 cm. B IPD3Min Transgenic lines (blue) had a delayed transition to reproductive development (onset of bolting) relative to nontransgenic controls (orange) (ANOVA, p < 0.05, n = 15), but C showed no significant differences in total seed yield (MCM p < 0.05) (in B and C mean is marked with X and median marked with –, outliers shown as dots) (see Online Resources 4–6)

Full size image

We also monitored Lotus japonicus plants used in cross-species comparison for overt differences between wild type (Gifu ecotype) and the cyclops-4 knockout mutant of IPD3. Consistent with prior literature we noted no obvious growth phenotype in the ipd3 mutant (Yano et al. 2008; Horváth et al. 2011) (Online Resource 7). The genotype of cyclops-4 Lotus was confirmed by Sanger sequencing of the CYCLOPS genomic region.

Transcriptional consequences of manipulating IPD3

Before a functional AM host trait could be engineered into Arabidopsis as a reference plant for Brassicaceae (Hornstein and Sederoff 2023), the potential for evolutionarily conserved interactions and pathways can be explored via comparative transcriptome analysis of non-host and host plants with and without IPD3/IPD3Min. We sequenced total mRNA of 6-week-old roots of transgenic IPD3Min, mutant cyclops-4, and wild type Arabidopsis Col-0 & L. japonicus Gifu plants with or without 48 hours of AMF germinated spore treatment prior to collection (Fig. 1). Two independent transgenic lines of Arabidopsis (308 and 310) with similar phenotype (Fig. 2, Online resources 4, 6) and IPD3Min transcript abundance were included in the experiment to account for possible variation due to genomic insertion site effects and epigenetic consequences in transgenic lines (Schnell et al. 2015).

We used statistical methods like Principle Component Analysis (PCA) and Differential Gene Expression as well as machine learning for network analysis to identify genes, pathways and networks in Arabidopsis and Lotus plants that are responsive to AMF inoculation and dependent or independent of IPD3 regulation. The comparison of the species and genotype specific transcriptome analyses provides an insight into those pathways that are still conserved in Arabidopsis after the evolutionary loss of the AM host trait through IPD3 loss. Correlation network construction used quantitative expression of IPD3Min in each sample of both lines as the primary transgenic trait. In differential expression analysis an IPD3Min category including only genes independently attested as DEGs in both transgenic lines was used. We also compared Col-0 Arabidopsis to a single IPD3Min transgenic line (308) with AMF inoculation on low-nutrient (LN) MS medium containing 0.5% P, 1% N, and 1% K relative to 1/2 MS (Fig. 1).

In principal component analysis (PCA) of Lotus, Gifu transcriptomes are clustered according to AMF treatment along PC1 (Online Resource 8). cyclops-4 transcriptomes were separated from those of Gifu plants along PC2. Interestingly, both AMF-treated and untreated cyclops-4 transcriptomes clustered along PC1 with those of AMF-treated Gifu plants. In Arabidopsis, IPD3Min transgenic lines clustered relative to Col-0 along PC1 (Online Resource 8). Line 308 clustered farther from Col-0 than line 310, most likely due to the higher IPD3Min expression in individuals of this line (Online Resources 8, 9). Col-0 plants clustered by AMF treatment along PC2, but there was no clear separation by treatment among IPD3Min lines (Online Resource 8). PCA of the low-nutrient experiment also shows strong clustering by genotype (Online Resource 10). In conclusion, the variance in gene expression changes in wild type Lotus and Arabidopsis plants corresponds more to the inoculation with AMF, while gene expression variation in response to presence or absence of IPD3/IPD3Min is less affected by presence or absence of AMF and this variance is explained by a different component.

Correlation network and gene ontology analysis

We used weighted gene co-expression network analysis (WGCNA) as a transcriptome analysis tool which captures global expression patterns reflective of the underlying reality of coordinated expression of many genes affected by multiple factors (Langfelder and Horvath 2008, 2012). WGCNA clusters the transcriptome into modules with similar expression profiles which suggest they may regulate each other or are co-regulated. WGCNA also correlates expression of a representative eigengene for each module to traits and treatments, in this case quantitative IPD3Min expression, AMF treatment, and nutrient level. The two-step process of module construction and module-trait correlation makes WGCNA sensitive to links between potentially subtle expression changes at the gene level and experimental variables, and can be used to guide more targeted analysis. WGCNA analysis incorporates transgene expression agnostic of categorical information about the 2 transgenic lines, instead reflecting how well expression of native Arabidopsis genes correlate to IPD3Min across a range of variation in transgene expression levels in each individual plant. WGCNA of Arabidopsis identified 22 co-regulatory modules (Online Resources 11, 12). Figure 3A includes all modules significantly correlated to IPD3Min expression, and selected modules correlated to AMF treatment and nutrient level. We used gene ontology (GO) enrichment to link module membership to biological functions (Fig. 3B). We also performed targeted pathway analysis of specific gene members within modules to better connect these broad patterns to physiological functions (Fig. 4). Statistical significance in Fig. 4 is a hypothesis test of whether mean expression of the given gene differs between specific treatment groups, an independently calculated and complementary measure to the network analysis correlations which establish module membership.

Fig. 3
figure 3

Correlation network modules and gene ontology analysis for transcription of Arabidopsis thaliana IPD3Min and Col-0 genotypes across AMF and low-nutrient treatments. A Selected correlation modules, including all those significantly correlated to IPD3Min expression. Modules are labeled E–T in order of appearance in the full network (Online Resource 7). Color of cells within the heatmap reflects correlation of that module to trait values; p-value of the module-trait correlation is listed in parentheses. B Gene Ontology enrichment of all modules shown in A; terms have been reduced by applying a cutoff for the top 3 most-enriched and most-significant terms in each module followed by overlap analysis to establish representative terms. Color intensity corresponds to FDR-corrected p-value of term enrichment and circle area corresponds to scale of enrichment. (See Online Resource 11, Online Resources 12, 14, S10)

Full size image
Fig. 4
figure 4

Targeted analysis for pathways of interest within correlation network modules. All genes included in this figure are statistically significant members of the respective network module (having module-trait correlations at p_GS < 0.05 as indicated in Fig. 3) at p_kME < 0.05. Significance reflects 1-tailed t-test of group means for transcript counts of specific genes, a complementary measure which is calculated independently from both the correlation network and untargeted DEG analysis; n = 3–5, *p < 0.05; **p < 0.01; ***p < 0.005. Standard Error of all treatment groups is provided in Online Resource 15 in addition to t-test results. Some interactions in pathway maps have been omitted for clarity. A Expression of a subset of individual genes belonging to module O related to extracellular wax synthesis. B Expression of well-known circadian clock related genes in module T. C Pathway map of the WRKY70-SA-NPR1-mediated defense response with expression of individual genes. Genes belong to module S unless marked otherwise in parentheses. (See Online Resources 12, 15). Gene acronyms are expanded as follows: A LACS2, Long-chain Acyl-CoA Synthetase 2; KCS1, 3-Ketoacyl-CoA Synthase 1; CYP86B1, Cytochrome P450, Family 86, Subfamily B, Polypetide 1; GPAT5, Glycerol-3-phosphate sn-2-Acyltransferase 5; ABCG20, ATP-binding Cassette G20. B FKF1, Flavin-binding, Kelch Repeat, F Box 1; ELF3, Early Flowering 3; APRR1, Pseudo-response Regulator 1; Spatula; Gigantea. C BDA1, Bian Da 1; WRKY70, WRKY DNA-binding Protein 70; WRKY54, WRKY DNA-binding Protein 54; SARD1, SAR Deficient 1; NPR1, Nonexpressor of PR Genes 1; lnc_NPR1, antisense long non-coding RNA of NPR1; NIMIN1, NIM1-interacting 1; PR5, Pathogenesis-related Gene 5; PR2, Pathogenesis-related Gene 2

Full size image

The WGCNA analysis identified clusters of co-regulated transcripts that were either responsive to the nutrient level, the exposure to AMF, the expression of IPD3Min, or combinations of those treatments. Further GOterm analysis of those specific clusters showed the potential functional connections of genes that were differentially co-regulated in response to either treatment of combinations of those. Only one cluster of transcripts (cluster O) significantly correlated only with AMF inoculation in Arabidopsis, and three clusters (E, F, H) significantly correlated only with IPD3Min expression. One cluster (M) showed positive correlation with AMF inoculation but negative correlation with IPD3Min expression suggesting repression of those genes by IPD3Min.

The Arabidopsis response to AMF includes altered defense and lipid metabolism

Despite its nonAM status, Arabidopsis transcriptome networks showed 10 modules of co-regulated genes with trait correlations to AMF treatment, indicating a clear transcriptional response to the fungus (Online Resource 11). We considered modules O and R that correlate to AMF treatment but not IPD3Min expression to reflect portions of the native Arabidopsis AMF response that are unaffected by IPD3Min expression. Expression of genes in module R positively correlated with AMF treatment but negatively correlated with nutrient level. Top GO terms in this module R related to both nutrient stress (GO:0016036; phosphate starvation) and biotic interactions (GO:0010200; chitin response) (Fig. 4B, Online Resources 13, 14).

Module O expression correlated solely to AMF treatment, independent of IPD3Min genotype and was highly enriched for terms related to biosynthesis of extracellular lipids (GO:0010143; cutin biosynthesis, GO:0010345; suberin biosynthesis, GO:0042761; very long-chain fatty acid biosynthesis) (Fig. 3B). These genes fall along the biosynthetic pathway of 2-monoacylglycerols (2-MAGs), which affect pathogen and stress resistance as components of cutin and suberin (Fig. 4A, Online Resources 12, 14). 2-MAGs are also the primary form of carbon exported to AMF fungus from AM host plants (Luginbuehl et al. 2017; Rich et al. 2017). Mean expression of all tested genes in the pathway was significantly higher in AMF-treated Col-0 plants than in untreated plants (Fig. 4A). Significantly higher expression was also detected in some but not all AMF-treated transgenic lines for the same genes. This indicates that upregulation of lipid biosynthetic genes in AMF-treated Arabidopsis is IPD3Min-independent, consistent with the module’s correlation to AMF treatment but not IPD3Min expression in Fig. 3A. The gene level comparisons of transcript abundance for this module shows that both independent transgenic lines (308 and 310) have similar transcript abundance differences to Col-0 and are comparable in their response prior to AMF treatment (Fig. 4B, C). Other genes in module O also function in lipid synthesis, including GPAT6, a functional homolog of RAM2 which is strictly required for AM in host species (Online Resource 12) (Gobbato et al. 2013; Dai et al. 2022). While it is a member of module O, AMF-mediated increase of GPAT6 gene expression was only significant in the IPD3Min line 308, but not in Col-0 nor in the IPD3Min line 310 (Online Resource 15).

IPD3 Min expression imitates AMF treatment in Arabidopsis Col-0 plants

Transcription in module T correlated negatively to AMF treatment and IPD3Min expression (Fig. 3A). This is consistent with IPD3Min expression enhancing the effect of AMF treatment or replicating the effect of AMF in untreated samples. A similar pattern was present in modules N (both positive) and S (both negative) (Fig. 3A). Module T contains only 56 genes and was enriched solely for terms related to circadian rhythm (GO:0007623; circadian rhythm, GO:0048511; rhythmic process) (Fig. 3B, Online Resource 14). Members of this module include components of the CONSTANS-FT daylength sensing system (Online Resources 12, 15) (Takagi et al. 2023). As shown in Fig. 4B, AMF treatment of Col-0 plants results in significantly reduced expression for genes across the pathway, and IPD3Min expression replicates this effect even in the absence of AMF.

Module S transcription was negatively correlated with both IPD3Min and AMF treatment, and is enriched for GO terms related to pathogen defense and systemic acquired resistance (SAR) (GO:0002229, oomycete defense; GO:0002237, bacterial molecule response; GO:0009627, systemic acquired resistance) (Fig. 3). This module contains the key defense regulator WRKY70, which mediates selection of defense responses by positively regulating salicylic acid (SA) immunity and negatively regulating jasmonic acid (JA) related pathways (Li et al. 2006). As shown in Fig. 4C, genes acting up- and downstream of WRKY70 in SA-mediated defense are also present in this module, including cell-surface ankyrin protein BDA1, and pathogen response gene PR2 (Thomma et al. 2001; Yang et al. 2012a). NPR1, another key positive regulator of SA-mediated defense was detected in the transcriptome but not placed in a network module (Fig. 4C) (Thomma et al. 2001).

Several SA defense genes (BDA1; WRKY70; WRKY54; NIMIN1; PR5) were significantly downregulated in one or both transgenic lines even in the absence of biotic interaction with AMF (Fig. 4C). This suggests that IPD3Min expression in isolation can replicate the suppressive effect of AMF interaction on SA-related defense. Consistent with downregulation of SA-related genes, we also found evidence of antagonistic crosstalk between jasmonic and salicylic acid-mediated defenses (Li et al. 2006; Hou and Tsuda 2022). Module R, which in contrast to Module S was upregulated in response to AMF, is functionally enriched for jasmonic acid (JA) signaling (GO:009753) (Online Resource 14). JA genes upregulated by AMF in module R included upstream (jasmonate methlytransferase JMT) and downstream (chitinases PR3 and PR4) members of the JA defense pathway (Samac et al. 1990; Thomma et al. 1998; Seo et al. 2001). Module R also contained a set of genes enriched for salicylic acid signaling, some of which are negative regulators of SA defense such as At1g08667.1 (lnc_NPR1), a putative antisense transcript of NPR1 whose upregulation would correspond to downregulation of NPR1 and the SA response (Fig. 4C, Online Resource 15). Other SA-related genes were placed into module R due to expression correlation at the individual sample level, but upon comparison of treatment group means matched the pattern of downregulation in module S; this included WRKY54, a co-regulator with WRKY70 of SA synthesis (Chen et al. 2021b) (Fig. 4C, Online Resource 15).

Positive IPD3Min correlation with modules F, G, H, and N corresponded to GO enrichment for background processes, e.g. GO:0000466, rRNA maturation (Fig. 3). These functions are essential to the organism, but limited in GO analysis from being linked to specific effects. These modules contained 4132 genes, indicating that about 10% of all gene models in the Arabidopsis genome were upregulated in connection to IPD3Min (Online Resource 12). Module E had the strongest correlation to IPD3Min expression but resulted in no significant GO term enrichment (Fig. 3A, Online Resources 12, 14). Notably, 8 of 25 genes subsequently identified as IPD3-responsive via differential transcription analysis in Fig. 6B belonged to module E and are discussed in later sections (Online Resource 16).

Network analysis of Lotus transcriptomes

We also constructed a correlation network for Lotus, which is available in Online Resources 17 and 18. A distinctive feature of the Lotus network is that while expression of 8 modules correlated with IPD3 genotypes and 15 modules correlated with the AMF treatment, no modules correlated with both treatments. We confirmed transcription of CSP and AMF marker genes including CCaMK, NSP2, and PT4, however, all detectable CSP genes were placed into network modules not correlated to either treatment.

Cross-species comparison of differential transcription

The major question of our experiment was whether elements of the AMF response network are still present in the non-host Arabidopsis after the evolutionary loss of the transcription factor IPD3. To directly characterize transcript abundance changes related to the presence or absence of the respective IPD3/IPD3Min versions in a manner that enables cross-comparison of species, we conducted differential transcript abundance analysis for genotype and AMF treatment in Lotus and Arabidopsis.

The question to answer in this cross-species comparison was which genes are IPD3-dependent in their response to AMF inoculation in the host plant (Lotus) and then compare it to the non-host Arabidopsis (Fig. 5A, Online Resource #19) with the goal to identify genes that still can be regulated by IPD3Min. Transgenic expression used to create the +IPD3Min genotype in Arabidopsis introduces two additional factors which make this genotype only partially parallel to the Lotus wild type. First, the transgene is constitutively expressed under the Ubiquitin promoter and constitutively active at the protein level due to truncation of the activation domain, as opposed to inducible as in Lotus (Pimprikar et al. 2016; Singh et al. 2014). Second, Agrobacterium mediated transformation randomly inserts the transgene into different genomic locations in different independent lines, resulting in differing expression levels and potential effects on native genes (Edwards et al. 2022). To limit influence of such across-line variation, in the primary analysis we considered only those Arabidopsis DEGs independently attested for a given contrast in both transgenic lines (Fig. 5). Contrasts of individual transgenic lines 308 and 310 vs. Col-0 are provided in Online Resource 16. In the −ipd3 genotype of Lotus (cyclops-4), AMF exposure induced only 15 DEGs (Fig. 5A contrast A), while AMF treatment of +IPD3 Lotus induced 461 DEGs (Fig. 5A, contrast B). This is consistent with prior knowledge that AM symbiosis is strongly IPD3-dependent (Yano et al. 2008). In −ipd3 Col-0 Arabidopsis, however, AMF regulated 497 DEGs (Fig. 5A contrast C), while AMF treatment of +IPD3Min Arabidopsis regulated abundance of only 3 transcripts (contrast D). In -AMF Lotus the presence of IPD3 in +IPD3 Gifu produced 338 DEGs relative to −ipd3 cyclops-4 (Fig. 5A contrast E), while the equivalent contrast in Arabidopsis produced 25 DEGs (contrast F). Contrast of the two genotypes under AMF treatment resulted in 38 DEGs for Lotus and 77 DEGs for Arabidopsis (Fig. 5A, contrasts G and H). This comparison showed that regulation of only 15 transcripts in Lotus were independent of IPD3 when Lotus roots were challenged with AMF in contrast to 497 genes that did not require IPD3Min in Arabidopsis to be regulated in response to AMF inoculation. When we interpret these DEGs for the two Arabidopsis genotypes, it is important to keep in mind that the IPD3Min protein is constitutively active and the transgene was controlled by the constitutive UBI10 promoter and therefore both highly expressed and active prior to the AMF-inoculation, unlike in Lotus. If IPD3-responsive genes were already regulated by constitutive IPD3Min expression in Arabidopsis, further activation upon AMF exposure would not be expected.

Fig. 5
figure 5

Differential transcription contrasts for IPD3[Min] expression with and without AMF treatment in two species. A Overview of all contrasts executed and number of DEGs detected. Contrast labels A–H in B and C and Fig. 6 correspond to those noted in A. DEGs reported for the +IPD3Min genotype in Arabidopsis are those attested in both of the individually conducted contrasts for the 2 transgenic lines 308 and 310. B Comparison of DEGs resulting from AMF treatment in −ipd3 (cyc4) and +IPD3 (Gifu) genotypes of Lotus. L2FC is Log2(fold change) of mean expression. C Comparison of DEGs resulting from AMF treatment in −ipd3 (Col-0) and +IPD3 (OX-IPD3Min) genotypes of Arabidopsis. Differential transcription is expressed as log2(fold change) (L2FC); more-positive numbers indicate higher expression and more-negative numbers indicate lower. (See Online Resource 16)

Full size image
Fig. 6
figure 6

Cross-species comparison of AMF and IPD3[Min] responsive DEGs. A When the AMF-responsive DEGs within the respective +IPD3 and −ipd3 genotypes of Lotus and Arabidopsis are directly compared using annotated Arabidopsis homologs of Lotus genes, OBP4 is shared by −ipd3 Col-0 Arabidopsis with both Lotus genotypes, but oppositely regulated across species. Thirteen additional genes are shared by −ipd3 Col-0 Arabidopsis and +IPD3 Gifu Lotus, also with opposite regulation across species. B When the IPD3-responsive DEGs in non-AMF-treated Lotus and Arabidopsis are compared, LEA4-5 and His1-3 are shared across species, with LEA4-5 being oppositely regulated. C When IPD3-responsive DEGs in non-AMF-treated Arabidopsis are compared to AMF-responsive DEGs in −ipd3 Col-0 Arabidopsis, there is significant overlap. Eleven of 25 DEGs induced by IPD3Min in both transgenic lines 308 and 310 are also induced by AMF in −ipd3 Col-0. 103 out of 497 AMF-responsive DEGs (21%) in −ipd3 Col-0 are also responsive to IPD3Min in transgenic line 308, which has higher IPD3Min expression and a larger overall DEG set. (See Online Resource 18)

Full size image

Next, we asked if the genes that remained responsive to AMF in −ipd3 Lotus would be conserved in Arabidopsis. To detect candidate genes that are part of the Lotus AMF host response but independent of IPD3, we looked for overlap between AMF-responsive DEGs for both Lotus genotypes (Fig. 5B). Only 3 DEGs were shared: Lj1g0209300.1 (homolog of OBF Binding Protein 4 (LjOBP4)); Lj2g0329400.1 (homolog of Cinnamyl Alcohol Dehydrogenase 8 (LjCAD8)); and Lj5g0253700.1 (homolog of Binding to TOMV RNA 1 (LjBTR1)). All of these genes were regulated in the same direction and at similar scale in both Lotus genotypes, suggesting they are elements of the native Lotus AMF response that are governed by an IPD3-independent mechanism. Despite IPD3's effect on the scale of AMF response in Lotus, IPD3Min expression in Arabidopsis did not enable any new AMF-responsive DEGs, and the scale of response to AMF was much larger in the −ipd3 Col-0 genotype than in IPD3Min (Fig. 5C).

We then looked for overlaps in DEGs for AMF response between both genotypes of both species (Fig. 6). For direct cross-species comparisons Lotus genes were labeled with their closest Arabidopsis homologs as recorded in genome annotations; total DEG count for AMF response of Gifu was slightly reduced due to exclusion of 33 genes without an annotated Arabidopsis homolog (Online Resource 19). One of the 3 IPD3-independent genes characterized in Fig. 5B as a member of the IPD3-independent Lotus AMF response, OBP4 (Lj1g0209300; At1g68050), was also shared with the AMF response of Col-0 Arabidopsis (Fig. 6A). OBP4 is a transcription factor that negatively regulates lateral root and root hair development in response to nitrate and abscisic acid (Ramirez‐Parra et al. 2017; Rymen et al. 2017; Xu and Cai 2019). Interestingly, OBP4 was upregulated in the AMF response of Col-0 Arabidopsis, but downregulated in Lotus. OBP4 was not differentially regulated in IPD3Min plants.

In addition to OBP4, 13 genes were differentially regulated in response to AMF in Gifu Lotus and Col-0 Arabidopsis, but not cyclops-4 Lotus or IPD3Min Arabidopsis (Fig. 6A, Online Resource 19). In most of these genes, as for OBP4, the direction of regulation was reversed between species, including the next two most strongly regulated, multi-stress-responsive Blue Copper Binding Protein and pathogen response-related WRKY22 (Fig. 6A). Consistent with defense effects of AMF observed in the network analysis, WRKY22, which was downregulated in Arabidopsis, is implicated in systemic acquired resistance and is characterized as a positive regulator of the SA pathway in SA-JA defense crosstalk (Kloth et al. 2016).

We also compared the effect of IPD3 presence and absence across the two species in the absence of AMF treatment. Two genes differentially transcribed in +IPD3 Lotus relative to −ipd3 Lotus were also differentially transcribed in +IPD3Min Arabidopsis relative to −ipd3 (Fig. 6B). Of these, osmotic stress responsive Late Embryogenesis Abundant 4–5 was strongly downregulated by the presence of IPD3 in Lotus but upregulated by the presence of IPD3Min in Arabidopsis, while drought stress responsive Histone H1-3 was downregulated in both species by the presence of respective IPD3 versions.

Beyond specific genes, a shared and unexpected feature of the +IPD3/−ipd3 contrast in both species was a pattern of many differentially expressed genes resulting from manipulation of IPD3 even in the absence of AMF (Figs. 5A, 6B). Many of the affected genes are related to biotic and abiotic stress. These included strong downregulation in Lotus of Lj4g0393400.1 and Lj1g0155200_LC.1, both of which are homologs of AtFRS5, a gene involved in linking light perception to stress and pathogen defense via JA, and upregulation of Lj4g0167200.1, a homolog of the AtJASSY JA biosynthesis enzyme (Ma and Li 2018; Guan et al. 2019). The top most strongly regulated gene that was dependent on the presence of IPD3 in Lotus was Lj1g0431600.1; its homolog Lj1g0070100.1 was also upregulated, with an average L2FC of 7.14. These genes are homologous to an uncharacterized MYB/SANT transcription factor in Arabidopsis, At2g24960.2. In Arabidopsis, IPD3Min expression also resulted in strong upregulation of genes related to JA pathogen response and auxin signaling including Cytochrome P450 76C2 and Tryptophan Aminotransferase Related 1 (Fig. 6B) (Lorenzo et al. 2003; Stepanova et al. 2008). Notably, both of these genes were included in module E of the network analysis, which had the strongest correlation to quantitative IPD3Min expression (Online Resource 16).

While the differential response to AMF was nearly eliminated in −ipd3 Lotus, 337 Differentially Expressed Genes (DEGs) respond to IPD3 in the contrast of +IPD3 and −ipd3 Lotus without AMF inoculation (Fig. 5A). 135 of these DEGs resulting from knockout of IPD3 in the absence of AMF treatment, were also found among DEGs resulting from AMF treatment of Gifu plants, accounting for 29% of the wildtype Lotus AMF response (Online Resource 19). Thus, although AMF-responsive upregulation of genes by IPD3 is known to be essential for normal AM function, its deletion also appears to partially replicate the AM response under the conditions of this experiment. This suggests that IPD3 in Lotus may act as a context-dependent transcriptional repressor of a portion of its own targets.

A similar pattern is present in Arabidopsis, where 11 of 25 genes (44%) differentially expressed in response to IPD3Min in the absence of AMF treatment were also induced by AMF treatment of Col-0 plants (Fig. 6C). These genes included, for example, the FKF1 circadian clock gene identified in Fig. 5C as well as in the IPD3-Min responsive circadian clock module of the correlation network (Figs. 4, 5) (Online Resource 19). In transgenic line 308 where higher IPD3Min expression corresponds to a larger absolute number of DEGs (Fig. 5A, Online Resource 9), this effect is much larger. The no-AMF contrast of IPD3Min-308 vs Col-0 recapitulates differential regulation of 103 of the 497 (21%) AMF-responsive DEGs in Col-0 (Fig. 6C). These results confirm evidence in the correlation network that effects of IPD3Min expression in Arabidopsis overlap partially with the AMF response. They indicate that while the nature of the response may differ, expression of constitutively activated IPD3 in Arabidopsis can activate the transcriptional response to AMF even in the symbiont's absence, much as it does in AM host plants (Singh et al. 2014).

While the detailed comparisons of DEGs provide interesting insight into those genes and pathways that are apparently dependent on IPD3 activation in Lotus for a response to AMF challenge, the conservation of genes regulated by the constitutive expression of the DNA binding domain of IPD3 (IPD3Min) in Arabidopsis shows that there is still a set of genes regulated by this domain. While we used a conservative approach by only considering those DEGs that were regulated in both transgenic lines, the differential gene expression in the respective individual transgenic lines by IPD3Min was much larger and can provide targets for further studies.

Discussion

We successfully expressed IPD3Min, the DNA binding domain of the AM symbiosis-essential transcription factor IPD3, in the nonAM host plant Arabidopsis which lost both the AM trait and IPD3 about 60–70 million years ago (Hohmann et al. 2015; Radhakrishnan et al. 2020). To understand the impact of IPD3Min on remaining genetic pathways and biotic interactions, we compared the response of Arabidopsis genotypes to AMF inoculation, as well as the AM host plant Lotus and its ipd3 mutant cyclops-4. Our results indicate that despite the long intervening period as a nonAM plant and the further loss of related genes, expressing IPD3Min in Arabidopsis resulted in phenotypic and transcriptional effects. The preservation of molecular connections for IPD3 raises the prospect of re-wiring nonAM plants to restore AM symbiosis, with applications in agriculture (French 2017; Lynch 2019). Although they are a phylogenetic minority, nonAM crops are responsible for almost half a billion metric tons of agricultural harvest every year (Hornstein 2022). With reported yield increases from mycorrhizae of 20–100% in various host crops, significant economic advantages could result if these benefits can be conferred on nonAM crops (Eo and Eom 2009; Pellegrino et al. 2015; Berruti et al. 2016). In 2021, such an increase in US canola alone would have resulted in an additional 250,000–1.25 million metric tons of harvest worth 175–870 million dollars (USDA NASS 2022).

We also found evidence that IPD3 may be subject to targeted regulation in Arabidopsis. Two full-length versions of IPD3 transgenically expressed in Arabidopsis could not be detected at the protein level, while IPD3Min was successfully expressed in roots and shoots (Figs. 2, S1). This suggests IPD3 may be subject to silencing or degradation specific to the N-terminal portion excluded from IPD3Min. In prior work, we attempted to express IPD3Mt and IPD3S50D in the oilseed crop Camelina sativa, and observed low expression levels in multiple independent lines, consistent with evidence in the present study for silencing of these constructs in Arabidopsis (Hornstein 2022). We have also observed potentially deleterious effects for IPD3S50D in Camelina. T1 transformants were repeatedly lost to severe fungal infections, and later generations produced unusual dwarfed growth phenotypes at nonmendelian ratios (Hornstein and Sederoff unpublished). It is possible that both the present study and past work in Camelina were biased by survivorship of T1 individuals which avoid a deleterious effect due to silencing.

Mechanisms for functional AM go beyond the CSP to essential plant functions including lipid biosynthesis, membrane and vesicle functions, nutrient transport, and defense (Wang et al. 2012; Behie and Bidochka 2014; Luginbuehl et al. 2017). Although symbiosis-specific genes that affect these functions in AM hosts have been lost in nonAM species, most belong to families with non-symbiosis-specific members that have similar molecular functions. In our experiment we observed IPD3-independent upregulation of lipid and 2-MAG biosynthesis genes in response to AMF in Arabidopsis (Figs. 3, 4). Activation of 2-MAG biosynthesis via RAM1 and RAM2 is well-characterized as an essential feature of AM symbiosis that strictly depends on IPD3 in host species (Wang et al. 2012; Gobbato et al. 2013; Pimprikar and Gutjahr 2018). In our results, 2-MAG-related transcription appeared to occur in response to AMF, but by action of related genes which are not symbiosis-specific (Yang et al. 2010, 2012b). Given that AMF was the only microbe tested, the biotic interaction terms and linkage to phosphate starvation as shown in the WGCNA and GOterm analysis may represent a generalized microbial response consistent with known effects in Arabidopsis (Finkel et al. 2019). In the context of engineering AM as an agriculturally useful trait, this raises the interesting possibility that non-symbiosis-specific genes retained in nonAM plants can be recruited to perform symbiotic functions (Hornstein and Sederoff 2023).

Figure 7 lays out the main findings of this study in relation to the canonical knowledge of IPD3’s role in AM. In Arabidopsis, IPD3Min plants showed a > 99% reduced transcriptional response to AMF compared to wild type (Fig. 5). This is in part because IPD3Min expression places plants in an AMF-exposure-like transcriptional state prior to actual exposure to the fungus (Fig. 6C). This is reminiscent of results in symbiosis models where expression of phosphomimic IPD3 leads to a constitutive symbiotic response in the absence of biotic stimulus (Singh et al. 2014; Gobbato 2015). Our results suggest that although the nature of the transcriptional response may differ, IPD3Min acts analogously in Arabidopsis by activating the native AMF response even in the absence of the fungus. In Lotus, surprisingly, we found that −ipd3 knockout was associated with a similar priming effect. The AMF response in this Lotus genotype is ~ 90% reduced in our experiment, consistent with canonical understanding of IPD3 (Yano et al. 2008; Pimprikar et al. 2016). However, in our experiment this was partly because a set of genes in uninoculated plants is regulated similarly upon ipd3 knockout and AMF exposure (Fig. 5). This suggests IPD3 may have an unrecognized function as a direct or indirect repressor of some of its own targets.

Fig. 7
figure 7

Novel roles for IPD3 and responses to AMF in Arabidopsis and Lotus. In Lotus, we add to the current understanding of IPD3 by showing that its knockout has a large effect even in the absence of AMF, and that the reduction in overall scale of AMF response in the cyclops-4 genotype is due to a surprising partial replication of AMF-exposure-like transcription. In Arabidopsis we find evidence that AM-related lipid metabolism is altered by AMF independently of IPD3 and despite the plant's nonAM status. Expression of IPD3Min vastly reduced the amount of AMF-responsive differential expression in Arabidopsis, but not because IPD3Min has little effect. IPD3Min expression results in many differentially expressed genes in the absence of AMF, and the response to AMF in transgenic plants is small at least in part because many of the DEGs acting in the wild type Arabidopsis AMF response are similarly regulated in IPD3Min transgenics even before AMF treatment is applied. Genes affected by IPD3Min in Arabidopsis relate to functions including abiotic stress, pathogen defense, and circadian rhythm

Full size image

The striking inverse effects in Lotus and Arabidopsis also highlight two limitations of our results for drawing comparisons between the two species and with prior work. First, the IPD3 genotypes across the two species are not identical. The native LjIPD3/CYCLOPS gene in Lotus encodes a protein product that undergoes extensive protein-protein interactions and requires phosphorylation for activation, while the IPD3Min protein we used in Arabidopsis is constitutively active (Yu et al. 2014; Singh et al. 2014; Pimprikar et al. 2016; Jin et al. 2016, 2018). Second, the effects we observed for AMF interaction may be specific to the early time point sampled in this study (48 h). We selected an early time point for the whole experiment because Arabidopsis cannot sustain AMF in monoxenic culture, but most studies allow interaction with AMF on the scale of weeks to establish mature colonization; earlier stages can involve rapidly shifting responses not seen later (Siciliano et al. 2007; Genre et al. 2008; Gutjahr et al. 2009; Handa et al. 2015; Nanjareddy et al. 2017; Prihatna et al. 2018). An idealized experiment might use cyclops-4 Lotus rescued by expression of IPD3Min, and a nurse-pot colonization system to support Arabidopsis AM colonization (Veiga et al. 2013; Fernández et al. 2019).

Prior studies in AM model plants at this early phase of colonization have found 20–650 DEGs in response to AMF treatment, germinated spore exudates, and various AM signal molecules at timepoints from 6 h to 48 days from contact (Czaja et al. 2012; Hohnjec et al. 2015; Giovannetti et al. 2015; Siciliano et al. 2007). Importantly, CSP gene knockout has previously been found to significantly alter the AMF (and nodulating rhizobacteria) response even at this early time point (Hohnjec et al. 2015; Høgslund et al. 2009). However, early-stage results in this experiment should also be viewed in recognition of the fact that late-stage colonization of AM host plants, as well as nurse-pot supported late-stage colonization of Arabidopsis, implicate much larger plant gene sets in the AM relationship (Fernández et al. 2019).

One area where IPD3Min expression duplicates the effect of AMF treatment is downregulation of circadian clock genes that result in delayed flowering (Fig. 4) (Sawa et al. 2007; Para et al. 2007; Niinuma et al. 2008; Shim et al. 2017). This gene regulation may explain the delayed flowering phenotype in IPD3Min transgenic Arabidopsis seen in our experiment if the root transcriptome effect extends to aboveground tissue (Figs. 1, S1). AM symbiosis affects flowering time in host plants, and the circadian clock is implicated in maintenance of the symbiosis itself as well as AM-mediated abiotic stress resistance (Hernandez and Allen 2013; Lee et al. 2019; Bennett and Meek 2020; Liu et al. 2022). In Arabidopsis, circadian rhythm genes are also known to act in microbiome construction, pathogen defense, development, and abiotic stress (Lee et al. 2005; Nakamichi et al. 2016; Newman et al. 2022; Xu et al. 2022; Singh 2022a, b). Remarkably, in mosses, the sole group of nonAM plants that retain CSP genes, IPD3 specifically has been shown to mediate a stress-responsive reproductive transition (Kleist et al. 2022).

In uninoculated plants, two DEGs result from the respective +IPD3/−ipd3 contrasts in both Arabidopsis and Lotus: LEA4-5 and HIS1-3 (Fig. 6B). Both of these genes are associated with response to abiotic stress including drought, heat, and cold (Dalal et al. 2009; Olvera-Carrillo et al. 2010; Rutowicz et al. 2015). HIS1-3 is downregulated in the +IPD3 genotype of both species, and LEA4-5 is upregulated in Arabidopsis but downregulated in Lotus (Fig. 6).

A molecular function in stress response would subject IPD3 to selection on factors unrelated to its role in AM symbiosis, and might help explain both the reasons for this gene’s loss in nonAM species, and its presence in charophyte ancestors of land plants long before the existence of AM symbiosis (Delaux et al. 2015; Delaux and Schornack 2021).

We also identified AMF and IPD3-Min-responsive downregulation of Arabidopsis genes involved in salicylic acid-mediated defense (Fig. 4C). Reduced transcription of WRKY70, WRKY54, and PR5 by IPD3Min in the absence of AMF suggests that plants expressing IPD3Min have reduced baseline levels of SA-mediated defense to biotrophic pathogens (Glazebrook 2005; Blanco et al. 2005; Li et al. 2006; Yang et al. 2012a). Early AM colonization of host plants can involve a transient increased defense response followed by reduction, while Arabidopsis shows an early symbiosis-like response to AMF, followed by a strong defensive response under forced long-term interaction (Giovannetti et al. 2015) (Fernández et al. 2019; Cosme et al. 2021). While AM can confer resistance to pathogens and insects, genes acting in AM also enable infection by some pathogens (Wang et al. 2012; Siebers et al. 2016; Ried et al. 2019; Chen et al. 2021a; Dey and Ghosh 2022). The full mechanisms for such effects are not known, but comport with our evidence for perturbation of Arabidopsis defenses by a nominally AM-specific gene.

Recently, significant portions of the AMF genetic network were revealed to be under control of the phosphate-stress-responsive PHR-SPX system, which is conserved in Arabidopsis (Shi et al. 2021, 2022). In Arabidopsis, PHR also regulates the microbiome by reducing plant defenses when phosphate is low, resulting in recruitment of beneficial microbes (Finkel et al 2019; Martin-Rivilla et al. 2019; Castrillo et al 2017; Cho et al. 2013; Kotchoni and Gachomo 2006; Dangl 1998). Given knowledge that both AM and non-AM microbial relations are part of the PHR-regulated phosphate response network, it is possible that the IPD3Min effect in our experiment relates to conserved, non-AM-specific points of crosstalk with defense and symbiosis (Fig. 7).

Regardless of whether plants are formally labeled non-hosts in isolation, in natural settings they continue to interact with ubiquitous AM fungi and can display partial and transitional phenotypes (Ma et al. 2018; Cosme et al. 2018). It is not surprising that an ancient, conserved trait is interconnected to other aspects of plant life. Our results show that despite the loss of genes essential for AM symbiosis, connections to this trait remain in place in Arabidopsis, and can be highlighted, even re-activated by expression of IPD3Min. The transcriptional effects we identified suggest specific targets for follow-up studies to directly assess pleiotropic effects, including pathogen sensitivity and abiotic stress resilience.