Stress-associated developmental reprogramming in moss protonemata by synthetic activation of the common symbiosis pathway

Summary Symbioses between angiosperms and rhizobia or arbuscular mycorrhizal fungi are controlled through a conserved signaling pathway. Microbe-derived, chitin-based elicitors activate plant cell surface receptors and trigger nuclear calcium oscillations, which are decoded by a calcium/calmodulin-dependent protein kinase (CCaMK) and its target transcription factor interacting protein of DMI3 (IPD3). Genes encoding CCaMK and IPD3 have been lost in multiple non-mycorrhizal plant lineages yet retained among non-mycorrhizal mosses. Here, we demonstrated that the moss Physcomitrium is equipped with a bona fide CCaMK that can functionally complement a Medicago loss-of-function mutant. Conservation of regulatory phosphosites allowed us to generate predicted hyperactive forms of Physcomitrium CCaMK and IPD3. Overexpression of synthetically activated CCaMK or IPD3 in Physcomitrium led to abscisic acid (ABA) accumulation and ectopic development of brood cells, which are asexual propagules that facilitate escape from local abiotic stresses. We therefore propose a functional role for Physcomitrium CCaMK-IPD3 in stress-associated developmental reprogramming


INTRODUCTION
During their early evolution, plants faced numerous challenges in the shift from freshwater to terrestrial environments. These problems included decreased water availability, the sparsity of nutrients, and increased UV radiation levels. The shared ancestor of extant land plants evolved several strategies to surmount these stressors. For example, arbuscular mycorrhizal fungi (AMF) and AMF-like interactions with fungal mutualists likely aided early land plants in acquiring water and nutrients Pirozynski and Malloch, 1975;Read et al., 2000). Arbuscular mycorrhizae are controlled infections of plant roots by fungi of the Glomeromycotina (Parniske, 2008;Spatafora et al., 2016). The establishment of intracellular arbuscules within cortical root cells enables the fungus to provide the plant host with greater access to resources such as phosphate, nitrogen, potassium, and water in exchange for host photosynthates (Garcia et al., 2017;Parniske, 2008;Smith and Read, 2010).
Endomycorrhizal, AMF-like interactions occur in early-diverging plant lineages, including some liverworts. Moreover, fossil samples provide evidence for ancient AMF-like associations. Endophytic structures with a striking similarity to arbuscules are present in the Early Devonian fossil record of the Rhynie chert (Remy et al., 1994;Strullu-Derrien et al., 2014, 2015. Fossilized fungal spores with similar morphology to extant AMF have been found in the Ordovician (Redecker et al., 2000).The broad phylogenetic distribution of AMF and AMF-like host lineages among land plants and the available fossil evidence point toward establishing plant-fungal symbioses early in land plant evolution (Wang and Qiu, 2006). Mosses are one of the earliest diverging and most diverse lineages of extant land plants. Whereas numerous pathogenic, saprotrophic, and commensal fungal interactions have been described in mosses (Davey and Currah, 2006), no convincing evidence has been published to date for bona fide mutualistic interactions among mosses and AMF with the possible exception of Takakia, which is distantly related to other extant mosses (Newton et al., 2000;Liu et al., 2019). A few reports describing observations of AMF within moss samples (e.g., Rabatin, 1980;Carleton and Read, 1991) were likely due to misinterpretation of fungal growth present in senescent or dead plant tissues.  Table S1 for details. (B) Domain architecture diagram of CCaMK and multiple sequence alignment of the region (green) surrounding the regulatory autophosphorylation site (green). CB: CaM-binding domain, PKD: protein-kinase domain, EF: Calcium-binding EF-hand. (C) Domain architecture diagram of IPD3 and multiple sequence alignments of regions (green) surrounding two regulatory phospho-sites (green) that are necessary and sufficient for activation of LjIPD3/CYCLOPS. AI: autoinhibitory domain, AD: activation domain, DBD: DNA-binding domain, CC: coiled-coil domain.
(D) PpCCaMK interacted with PpIPD3 in yeast two-hybrid assay, whereas PpCCaMKb or empty vector (EV) controls did not. The left panel shows growth on control (-LT) media; the right panel shows growth on the test (-LTHA) media to screen for physical interactions. AD: activating-domain, BD: DNA-bindingdomain. (E) Kinase assays using purified recombinant proteins showed that PpCCaMK but not PpCCaMKb exhibited kinase activity and that kinase activity is responsive to calcium (Ca 2+ ) and CaM. AR: autoradiogram, CB: Coomassie Brilliant Blue stain. (F) Calmodulin-binding assays show that PpCCaMK or positive control (MtCCaMK) binds calmodulin, whereas PpCCaMKb or negative control from Chlamydomonas reinhardtii (CrCDPK) does not. PS: Ponceau S staining.

OPEN ACCESS
Nearly all AMF-host plants that have been studied possess the full complement of this core signaling pathway, from angiosperms to liverworts (Delaux et al., 2015;Wang et al., 2010). In several instances, plant lineages that have lost the ability to host AMF have also lost several symbiosis pathway genes. This correlation is exemplified by the Brassicaceae in which many species, including the model plant Arabidopsis thaliana ( Figure 1A, Table S1), are unable to host AMF and have concomitantly lost many of the core common signaling components (Delaux et al., 2014;Garcia et al., 2015). The retention of symbiosis signaling genes in non-mycorrhizal mosses, including the model organism Physcomitrium patens (Physcomitrium, formerly Physcomitrella patens), provides a striking counter-example (Delaux et al., 2015;Rensing et al., 2020;Wang et al., 2010). Given that mosses have retained the vertically inherited symbiosis signaling pathway yet cannot establish AMF or AMF-like interactions, we pursued an investigation of the biochemical properties and physiological function(s) of these proteins in mosses using Physcomitrium as a model.
CCaMK and IPD3 are two of the genes whose presence or absence most strongly correlates with AMF host compatibility or incompatibility, respectively, in studied plant lineages (Delaux et al., 2014;Garcia et al., 2015;Wang et al., 2010). Moreover, genetic studies in legumes have elucidated mutational strategies to produce gain-of-function variants of either of these two proteins that can auto-activate root nodule development in the absence of symbionts or symbiont-derived signals, a phenomenon termed spontaneous nodulation. Expression of a constitutively active CCaMK, lacking the C-terminal autoinhibitory domain, in Medicago truncatula (Medicago) or Lotus japonicus (Lotus) is sufficient to cause the development of root nodules in the absence of rhizobia or rhizobial exudates (Gleason et al., 2006;Tirichine et al., 2006). Spontaneous nodule development can also be achieved by substituting an aspartate for a threonine residue in the kinase auto-activation loop of Medicago CCaMK. In Lotus, nuclear-localized and constitutively active CCaMK induced the partial development of the pre-penetration apparatus, a structure that facilitates hyphal entry of AMF into host roots Takeda et al., 2012). A pair of phosphomimetic substitutions in the IPD3 ortholog, CYCLOPS, in Lotus is likewise sufficient to induce spontaneous development of root nodules (Singh et al., 2014). These legume gain-of-function mutants revealed the pivotal role of the CCaMK-IPD3 module in this signaling pathway. We hypothesized that similar molecular genetic manipulations in Physcomitrium might lead to phenotypes that could provide clues to the possible biological relevance of these genes in mosses.
In this study, we investigated the evolutionary conservation, biochemical activities, and physiological function(s) of the two CCaMK and sole IPD3 homologs present in the Physcomitrium genome. We cloned the iScience Article coding sequence of each homolog from cDNA. We used yeast two-hybrid and biochemical assays to demonstrate that one of two CCaMKs and the sole IPD3 homolog from Physcomitrium have retained many of the biochemical properties required for CCaMK and IPD3 functionality in angiosperms. We further demonstrated that the Physcomitrium CCaMK, which shared biochemical properties with angiosperm CCaMKs, could restore both nodulation and mycorrhization when expressed in a Medicago ccamk-1 mutant background defective for both symbioses. Additionally, Physcomitrium IPD3 is capable of partially restoring nodulation defects in Medicago ipd3-1 mutants. Transgenic expression of modified forms of CCaMK and IPD3 predicted to show constitutive activation in Physcomitrium (but not the unmodified forms driven by the same promoter) promoted ectopic development of brood cells, a well-characterized developmental program of mosses in response to drought or osmotic stress. Brood cell development was accompanied by changes in abiotic stress-responsive LEA gene transcript levels and elevated amounts of abscisic acid (ABA). Whereas activation of PpCCaMK or PpIPD3 promoted brood cell development, genetic deletion of either the CCaMK or IPD3 loci from Physcomitrium was insufficient to block brood cell development in response to osmotic stress treatment, suggesting other pathways exist for activation of brood cell development. Unexpectedly, we observed prominent nuclear calcium oscillations in Physcomitrium protonemata in the absence of any experimental treatment (i.e., spontaneous). This is in stark contrast to published data on root cells of legume species (Ehrhardt et al., 1996;Chabaud et al., 2011). We therefore propose that CCaMK-IPD3 activation by nuclear calcium levels may be more complex than in studied model legumes, as changes in oscillation frequency or amplitude may trigger activation in moss protonemata. Our results collectively indicate that the Physcomitrium CCaMK-IPD3 signaling module has retained many of the biochemical properties that typify these components in symbiont host plants and that the CCaMK-IPD3 module regulates ABA levels and associated developmental reprogramming to promote escape from adverse environmental conditions.

Conservation of the CCaMK-IPD3 signaling module in Physcomitrium
Homologs of CCaMK encoded in the Physcomitrium patens genome sequence version 3.3 were identified by BLAST (Altschul et al., 1990) of the predicted proteome using Medicago CCaMK/DMI3 (MtCCaMK, Phytozome: Medtr8g043970) and Lotus CCaMK (LjCCaMK, Lotus Base: Lj3g3v1739280) as queries. The top five hits were used for reciprocal BLAST against the predicted Medicago or Lotus proteomes (Figures S1A, S1B). The two most significant BLAST hits in Physcomitrium, Phytozome: Pp3c21_15330V3 (E-value = 0, hereafter PpCCaMK) and Phytozome: Pp3c19_20580V3 (E = 6 3 10 À171 , hereafter PpCCaMKb), each returned MtCCaMK or LjCCaMK as the top reciprocal BLAST hit with highly significant E-values (E % 6 3 10 À176 ). The top reciprocal BLAST hits for other loci were identified as calcium-dependent protein kinases (CDPKs), which lack the distinctive CaM-binding site found in CCaMKs. Thus, it appears that up to two loci in the Physcomitrium genome may encode functional CCaMKs. Full-length coding sequences (CDS) were cloned from each locus to validate inferred gene models. Amino acid sequences were aligned using MUSCLE (Edgar, 2004), and the resulting sequence alignment corroborated that the protein kinase domain, auto-activation loop, predicted CaM-binding site, and three calcium-binding EF-hand domains were each conserved in candidate PpCCaMKs ( Figure S1C). Closer inspection of the auto-activation loop, which is required for MtCCaMK function, revealed that PpCCaMK and PpCCaMKb each have a serine residue at the position orthologous to the auto-phosphorylated threonine residue (T271) in MtCCaMK (Figure 1B), which suggests that PpCCaMK and/or PpCCaMKb may likewise be subject to regulatory autophosphorylation.
To identify potential IPD3 homologs encoded in the Physcomitrella genome, we employed a similar strategy. iScience Article and a C-terminal coiled-coil domain, are also present in PpIPD3 ( Figure S2C). In particular, two sequence motifs surrounding CCaMK-targeted phosphosites necessary and sufficient for activation of LjIPD3 are strongly conserved in PpIPD3 ( Figure 1C), suggesting that CCaMK-mediated phosphoregulation of IPD3 may be conserved in Physcomitrium.
If identified CCaMK and IPD3 homologs constitute a functional signaling module in Physcomitrium, the respective genes should be co-expressed in the same cell types. To determine and compare the relative expression patterns of PpCCaMK, PpCCaMKb, and PpIPD3, we mined their expression profiles from two Physcomitrium transcriptome atlas studies (Frank and Scanlon, 2015;Ortiz-Ramírez et al., 2016).Data from both studies confirmed that PpCCaMK, PpCCaMKb, and PpIPD3show overlapping expression patterns. Each is expressed in protonema, which we had expected based on our ability to clone each CDS from protonemal cDNA. Moreover, PpCCaMK showed greater transcript abundance than PpCCaMKb in all tested tissues ( Figure S3). Physical interaction between CCaMK and IPD3 has been demonstrated in multiple legume models (Messinese et al., 2007;Yano et al., 2008). We tested whether PpCCaMK or PpCCaMKb could interact with PpIPD3 in yeast two-hybrid (Y2H) assays. PpIPD3 was fused to the GAL4 split-transcription factor activation domain (AD) and tested in pairwise combination with PpCCaMK or PpCCaMKb fused to the GAL4 DNA-binding domain (BD). Co-transformation of PpIPD3-AD with PpCCaMK-BD facilitated robust growth on selective media, indicative of strong physical interaction. However, no growth or evidence for interaction was detected between PpIPD3-AD and PpCCaMKb-BD (Figure 1D). The lower expression levels of PpCCaMKb compared to PpCCaMK, along with the apparent inability of its gene product to bind PpIPD3, suggest that PpCCaMKb may encode a non-functional protein or function in a different context.
Legume CCaMKs have been characterized biochemically, and their autophosphorylation activity is known to be stimulated by elevated levels of calcium and inhibited by calmodulin (CaM) in the presence of high calcium levels (e.g., . Based on the conservation of the autoactivation loop shown in Figure 1B, we predicted that PpCCaMK and/or PpCCaMKb would lead to similar activities in vitro. To test if either Physcomitrium CCaMK homolog showed calcium/CaM-dependent protein kinase activity, we purified recombinant PpCCaMK and PpCCaMKb, along with positive and negative controls, and assayed autophosphorylation activity using radiolabeled ATP. Autophosphorylation of purified PpCCaMK was detectable and enhanced in buffer containing free calcium ions compared to the EGTA control ( Figure 1E). The autophosphorylation of PpCCaMK was attenuated in the presence of calcium and calmodulin, as described for Medicago CCaMK . No detectable kinase activity was observed for PpCCaMKb under the same conditions. These results demonstrated that PpCCaMK has retained similar calcium-and calmodulin-regulated kinase activity and suggested that PpCCaMKb may not be enzymatically active. To further assess whether PpCCaMK and/or PpCCaMKb are bona fide CCaMKs, we tested whether either could bind calmodulin (CaM) in vitro. Biotin-labeled CaM was applied to immobilized recombinant PpCCaMK and PpCCaMKb and detected by chemiluminescence to check for binding (Figure 1F). PpCCaMK showed similar CaM-binding activity levels to MtCCaMK, the positive control; however, PpCCaMKb showed nearly undetectable CaM-binding activity under the same conditions. Thus, consistent with gene expression and Y2H data, biochemical data supported a model wherein PpCCaMK but not PpCCaMKb comprises a functional signaling module with PpIPD3. Given the presence of this symbiosis signaling module in Physcomitrium, co-culture of wild-type moss with the model mycorrhizal fungus Rhizophagus irregularis was attempted. Still, no evidence of intracellular infection was obtained after six months of co-culture ( Figure S4), consistent with the prevailing interpretation that Physcomitrium is not an AMF host plant.

Heterologous expression of synthetically activated PpCCaMK stimulates symbiotic signaling in Medicago
Deleting the C-terminal autoinhibitory domains in Medicago or Lotus CCaMK leads to autoactivation and spontaneous nodule formation (Gleason et al., 2006). We investigated whether an equivalent deletion of the PpCCaMK C-terminus could promote spontaneous activation of the common symbiosis pathway by heterologous expression of native or modified PpCCaMK in Medicago roots. M. truncatula plants carrying the pENOD11::GUS reporter were transformed with constructs expressing MtCCaMK, PpCCaMK, or just the kinase domain of these proteins (MtCCaMK K and PpCCaMK K , respectively). Plants transformed with a vector control either treated with S. meliloti LCOs or not were used as positive and negative controls, respectively. Roots expressing MtCCaMK and PpCCaMK did not exhibit any detectable ENOD11 expression ( Figure S5A). In contrast, roots expressing MtCCaMK K and PpCCaMK K not only expressed MtENOD11 ll OPEN ACCESS iScience 25, 103754, February 18, 2022 5 iScience Article strongly but also elicited spontaneous nodules ( Figure S5B), indicating that PpCCaMK is functionally capable of activating the symbiosis signaling pathway in Medicago.
Complementation of symbiosis-defective phenotypes of Medicago ccamk loss-of-function mutants by heterologous expression of PpCCaMK To further interrogate the functionality of PpCCaMK or PpCCaMKb in vivo, within the functional context of the symbiotic signaling pathway, we tested their ability to rescue the phenotype of Medicago ccamk-1 mutants, which are defective for both nodulation and mycorrhization. 'Hairy root' genetic transformations mediated by Agrobacterium rhizogenes were used to introduce expression vectors containing the CDS from MtCCaMK (positive control), PpCCaMK, PpCCaMKb, or the empty vector (EV) negative control into roots of Medicago ccamk-1 plants. A red fluorescent protein (RFP) visual marker was used to confirm that transformations were successful. To test for AMF colonization, transformed roots were inoculated with Rhizophagus irregularis and grown in co-culture for six weeks. Trypan blue staining was used to visualize arbuscules and revealed that roots transformed with vectors containing PpCCaMK or MtCCaMK formed arbuscules indicative of colonization. In contrast, roots transformed with PpCCaMKb or the EV did not show any instances of arbuscule formation ( Figure 1G). To test for the ability to nodulate, transformed roots were inoculated with Sinorhizobium meliloti and co-cultured for two weeks. Whereas roots transformed with the EV or PpCCaMKb did not form any nodules, roots transformed with PpCCaMK formed nodules similarly to roots transformed with MtCCaMK ( Figure 1H). These data corroborate the conservation of key functional features of CCaMK between legumes and mosses and demonstrate that PpCCaMK can decode symbiotic signals when heterologously expressed in legumes.
PpIPD3 partially rescues the symbiotic defects of Medicago ipd3 mutants To determine the extent to whichPpIPD3 can functionally substitute for MtIPD3, we assessed the ability of heterologously expressed PpIPD3 to rescue the symbiotic defects of the M. truncatula ipd3-1mutant. Roots of the ipd3-1mutant were transformed with vectors driving transgenic expression of MtIPD3 or PpIPD3 or with an empty vector (EV) for a negative control. Roots of wild-type plants transformed with the empty vector were used as a positive control. All constructs also contained a tdTomato fluorescent reporter for the confirmation of transformation. In each case, roots were inoculated with the Sinorhizobium meliloti multireporter strain CL304 expressing a construct carrying both hemA::lacZ and PnifH::GUS (Lang et al., 2018a). As expected, based on the findings of Horvá th et al. (2011), roots of the Mtipd3-1 mutants transformed with the EV control developed nodules; however, few nodules were infected by rhizobia, and none of these nodules showed detectable expression nifH, in contrast to wild-type plants transformed with the same EV. Nodules produced on the Mtipd3-1 mutants transformed with MtIPD3 were similar to those on wildtype plants transformed with the EV, indicating a rescue of the symbiotic phenotype ( Figure S6). Interestingly, the transformation of the Mtipd3-1mutant roots with PpIPD3 only partially rescued the symbiotic defects with many colonized nodules observed, but none containing rhizobia showing expression of nifH ( Figure S6). These findings indicate that PpIPD3 contains some of the molecular features necessary for coordinating nodule infection but is not fully capable of restoring mutually beneficial symbiosis when heterologously expressed.

Developmental reprogramming and brood cell formation associated with synthetic activation of CCaMK-IPD3 in Physcomitrium
Previous studies in legumes have shown that mutated forms of CCaMK or IPD3 are sufficient to cause striking gain-of-function phenotypes: the development of nodules or the pre-penetration apparatus in the absence of rhizobial or mycorrhizal symbionts (Gleason et al., 2006;Singh et al., 2014;Takeda et al., 2012;Tirichine et al., 2006). We introduced equivalent amino acid substitutions or deletions into PpCCaMK or PpIPD3 to engineer predicted gain-of-function variants. Native or modified forms (hereafter referred to to as PpCCaMK K , PpCCaMK D , and PpIPD3 DD ) were transgenically expressed in Physcomitrium under the control of a maize ubiquitin (ZmUBI1) promoter. Transgenes were delivered by particle bombardment, as described in a previous study . A minimum of eight independently transformed lines were examined for phenotypic consistency (Table 1). Expression of unmodified PpIPD3 in this manner did not cause any noticeable effects on the development or morphology of protonemata or gametophores under standard axenic growth conditions, as these lines closely resembled wild-type Physcomitrium or empty vector controls ( iScience Article S7B and S7C). These features are diagnostic of brood cells, which are stress-resistant asexual propagules found in mosses (Correns, 1899;Duckett and Ligrone, 1992;Schnepf and Reinhard, 1997). Lines expressing PpIPD3 DD failed to form normal chloronema or caulonema and did not develop gametophores ( Figure S7D). Transformants expressing unmodified PpCCaMK displayed typical protonemal morphology and were able to develop gametophores, albeit with reduced frequency and size ( Figure 2D). Brood cell formation was not observed in lines expressing unmodified PpCCaMK under standard growth conditions. Transformants expressing a phosphomimetic variant, PpCCaMK D , developed mixed populations of phenotypically normal protonema and brood cells under standard growth conditions ( Figures 2E, S7E). Gametophores were rarely observed and, when present, were stunted and malformed ( Figure S7F). Lines expressing PpCCaMK K showed similar but more severe phenotypes, with frequent brood cell development and scarce instances of gametophore formation ( Figure 2F). Quantitative analysis of protonemal cell dimensions revealed highly statistically significant differences in cell length and width for lines expressing gain-of-function forms of PpCCaMK or PpIPD3 compared to untransformed lines or lines expressing the native form of PpCCaMK or PpIPD3 ( Figure 3, Table S2). Quantitative real-time PCR analysis showed elevated transcript abundances for each of the transgenically expressed forms of PpCCaMK and PpIPD3, demonstrating that transgenes were transcribed ( Figure S8). Expression of native IPD3 or gain-of-function PpIPD3 DD tagged with green fluorescent protein (GFP) using the same vector demonstrated that protein product is present in either case, causes similar phenotypes to expression of untagged forms, and fusion proteins shows preferential localization to nuclei ( Figure 4). The developmental phenotypes that we observed in CCaMK and IPD3 gain-of-function lines, particularly the constitutive development of brood cells, which generally only occurs in response to stress, led us to hypothesize that the Physcomitrium CCaMK-IPD3 module functions in developmental reprogramming to mediate resistance to or escape from stress conditions.

Elevated levels of ABA and LATE EMBRYOGENESIS ABUNDANT transcripts in physcomitrium expressing synthetically activated forms of CCaMK or IPD3
The stress-associated phytohormone abscisic acid (ABA) has long been linked to the induction of brood cells (Bopp, 2000;Schnepf and Reinhard, 1997). As expected, treatment of wild-type protonema with ABA phenocopied the gain-of-function effects of PpCCaMK K or PpIPD3 DD and stimulated the development of brood cells ( Figure 5A). Quantitative RT-PCR was used to test whether stress-associated, ABAinducible marker genes were likewise upregulated in CCaMK-IPD3 gain-of-function lines. We selected two previously described marker genes, LEA3-1 and LEA3-2, which encode late embryogenesis abundant (LEA) proteins (Shinde et al., 2012(Shinde et al., , 2013 and confirmed that transcript levels were elevated in wild-type protonemata treated with exogenously supplied ABA ( Figure 5B). Initially characterized in seeds, LEA proteins serve as osmoprotective molecules and are thought to confer abiotic stress resistance in brood cells, thereby enhancing their dispersal ability ( Figure 5C). Transgenic lines that constitutively form brood cells accumulated elevated levels of LEA3-1 and LEA3-2 under standard growth conditions (i.e., in the absence of any stress agent) relative to wildtype ( Figure 5D). Transcript levels were more abundant in gain-of-function lines compared to lines expressing unmodified PpCCaMK or PpIPD3. For example, expression of PpCCaMK K was associated with significantly higher levels of LEA3-1 transcript compared to lines expressing native PpCCaMK from the same promoter (p < .01).
As observed for developmental phenotypes, expression of PpIPD3 DD had the most substantial effect on LEA transcript abundance, and the accumulation of LEA3-1 and LEA3-2 transcripts was significantly The name, introduced mutations (if applicable), and predicted effects of introduced mutations are listed. For each construct, the number (#) of independent transformants that were generated and analyzed is provided. The reference for each study that guided our directed mutagenesis are provided and cited in the main text. Delta (D) indicates deletion. Dashes (À) indicate not applicable. iScience Article higher in lines expressing PpIPD3 DD compared to lines expressing PpIPD3 (p < .05), providing further evidence for a functional link between activation of the Physcomitrium CCaMK-IPD3 module and ABA signaling.
The observed phenotypic similarities between ABA-treated wild-type Physcomitrium and PpCCaMK-IPD3 gain-of-function lines may, in theory, be caused by an increase in ABA accumulation, an increase in ABA sensitivity, activation of a different pathway with similar effects, or a combination of these scenarios.
To further investigate the mechanism whereby the CCaMK-IPD3 module elicited these responses, we quantified ABA levels in tissues overexpressing IPD3 DD compared to wildtype by ELISA (enzyme-linked immunosorbent assays). The results showed that IPD3 DD gain-of-function lines contained significantly higher ABA levels than wildtype ( Figure 5E), suggesting that the ABA-associated responses we observed are likely due, at least in part, to increased ABA accumulation.

Brood cell formation in Physcomitrium ccamk and ipd3 loss-of-function mutants
To investigate if the CCaMK-IPD3 module is required to develop brood cells, we assayed responses to stress treatments in deletion lines lacking either the CCaMK or IPD3 genomic locus. Each locus was deleted by homologous recombination with antibiotic selective markers. Disruption of the respective locus was confirmed by PCR genotyping using genomic DNA and by RT-PCR ( Figure S9). Deletion lines displayed stereotypical protonematal and gametophore morphology when grown under standard conditions ( Figures  6A and 6B). When treated with ABA or hyperosmotic media supplemented with mannitol, multiple independently generated ccamk and ipd3 knockout lines responded similarly to wild-type controls by developing brood cells, which we did not observe in wild-type moss under standard laboratory growth conditions ( Figure 6B). These results suggest that while sufficient to stimulate ABA accumulation and brood cell formation, the CCaMK-IPD3 module is not required for brood cell development in response to ABA or osmotic stress treatments. We did not observe any noticeable differences in levels of brood cell formation between mutant and wild-type on both treatments, indicating that genetic perturbation of the CCaMK-IPD3 module does not substantially alter the sensitivity of Physcomitrium to ABA. The lack of phenotypic defects in ccamk or ipd3 deletion lines may imply functional redundancy in stress-induced developmental programming. Our results are collectively consistent with the hypothesis that the CCaMK-IPD3 module operates in the context of a broader signaling network that mediates stress-responsive developmental reprogramming in Physcomitrium ( Figure 6C).

Nuclear calcium oscillations in Physcomitrium protonemata
In legumes and other plants, the CCaMK-IPD3 signaling module is activated by nuclear calcium oscillations elicited by nod or myc factors. We therefore deployed nuclear targeted genetically encoded calcium indicators (GECIs) in Physcomitrium and sought to identify conditions or treatments that led to elicitation of nuclear calcium oscillations. In lines expressing nuclear-targeted intensiometric GCaMP6s (Chen et al., 2013), we unexpectedly observed prominent spontaneous calcium spiking in protonematal nuclei without performing any experimental treatment ( Figure 7A, Figure S10, Video S1). To corroborate these observations, we repeated these experiments using MatryoshCaMP6s, which contains a stable internal reference fluorophore (Ast et al., 2017). We observed similar spontaneous spikes using MatryoshCaMP6s ( Figure 7B, Video S2). Quantification showed that fluorescence intensity changes were pronounced in the reporter circularly permutated GFP channel but not in the reference LSSmOrange channel, which is consistent with spontaneous oscillatory calcium concentration changes in protonematal nuclei ( Figure 7C). Average oscillatory periods we observed were approximately seven and a half minutes (mean: 7.6 min, SEM: 0.45 min, median: 6.4 min, n = 37 nuclei). Average nuclear calcium spike duration was approximately 2 minutes (mean: 2.2 min, SEM: 0.078 min, median: 2.2 min, n = 40 spikes from eight different nuclei). We did not observe any statistically distinguishable differences in oscillation period between apical and subapical cells ( Figure 7D) nor in calcium spike duration, which we took as time from half-maximal rise to half-maximal iScience Article decay ( Figure 7E). Similarly, we did not observe statistically significant difference between apical and subapical cells in calcium spike amplitude ( Figure S10B). Thus, we uncovered robust evidence that nuclear calcium spiking occurs in protonemata without application of any chemical elicitors, suggesting unexpected complexity in the nuclear calcium signaling code in Physcomitrium.

Evolution of the CCaMK-IPD3 signaling module across land plants
Many of the critical components of the symbiosis pathway were present in the algal ancestors of land plants, indicating that they have been vertically inherited across land plants (Delaux et al., 2015). Across evolutionary time, spanning from the divergence of bryophytes to the emergence of angiosperms, AMF-like interactions have remained morphologically similar (Remy et al., 1994;Strullu-Derrien et al., 2014). In light of plant comparative genomics of AMF-host versus non-host lineages, the symbiosis pathway in the earliest land plants likely contributed to the recognition and intracellular infection of AMF, as the presence/absence of the symbiosis pathway is strongly correlated with host/non-host status, respectively (Delaux, 2017;Delaux et al., 2014;Garcia et al., 2015;Kamel et al., 2016). Among embryophytes, the moss clade is a striking exception to this genomic signature. To date, there has been no demonstration of the mutualistic transfer of nutrients between mosses and AMF. On the contrary, endophytic fungal interactions described in mosses have appeared restricted to dead or senescing tissues . This peculiarity piqued our interest in the conserved components of the common symbiosis pathway in mosses.
In the present study, we investigated the functional reason for retaining the CCaMK-IPD3 signaling module in non-mycorrhizal mosses, which stands in stark contrast to multiple independent losses of these genes in Note that each comparison was pre-planned (i.e., non-exploratory). One-way ANOVA analyses of the same datasets were performed for comparison. Asterisks indicate significance levels (*: p % 0.05; **: p % 0.01; ***: p % 0.001; ****: p % 0.0001). Full ANOVA results provided in Table S2. iScience Article non-mycorrhizal angiosperm and liverwort lineages. Biochemical and mutant-rescue assays demonstrate that the biochemical activities of CCaMK and IPD3 are conserved broadly throughout land plants, which suggests that CCaMK may decode similar oscillatory calcium signals in bryophytes. The biochemical similarity of Physcomitrium CCaMK to homologs in angiosperms is consistent with a previously published in vitro comparison (Okada et al., 2003). We used a gain-of-function strategy in Physcomitrium to shed light on the physiological consequences of CCaMK-IPD3 activation. The developmental phenotypes observed in Physcomitrium cells expressing CCaMK and IPD3 carrying predicted gain-of-function mutations imply a functional link between the CCaMK-IPD3 module and ABA signaling in mosses. The heterologous expression of native and engineered forms of PpCCaMK or IPD3 in Medicago roots corroborated the predicted effects of gain-of-function mutations and demonstrated partial functional conservation with Medicago homologs. The most striking finding in this study is the developmental phenotype of IPD3 DD -expressing Physcomitrium lines: IPD3 DD transgenics displayed prolific, nearly constitutive formation of brood cells, mainly to the exclusion of other cell types. The effect does not appear to be attributable merely to overexpression, as controls transformed with the same vectors differing only by two codon changes to introduce phosphomimetic substitutions putatively. The observation that gain-of-function lines expressing modified forms of CCaMK showed a less severe phenotype than lines expressing modified ones might simply reflect a signaling bottleneck of natively expressed IPD3 upon the activities of expressed CCaMK. Comparative

Calcium signaling in Physcomitrium protonemata
We propose a model wherein calcium-dependent activation of CCaMK and trans-phosphorylation of IPD3 in Physcomitrium leads to ABA accumulation and brood cell formation. The rescue of Medicago ccamk-1 mutants by PpCCaMK seemingly suggests that PpCCaMK may be activated by similar oscillatory calcium signals, as seen in legumes in response to symbionts. Additionally, the observed partial rescue of the Medicago ipd3-1mutant indicates partial conservation of function between PpIPD3 and MtIPD3, with PpIPD3 likely retaining the ability to be activated by CCaMK and stimulate expression of some (but perhaps not all) known downstream transcription factors. Whereas calcium oscillations in growing protonemal tips have been well documented (e.g., Bascom et al., 2018), nuclear calcium oscillations and potential elicitors have not been previously described in Physcomitrium. Hyperosmotic stress has been shown to elicit a pronounced transient elevation of cytosolic calcium levels. However, this response was neither oscillatory nor predominantly restricted to the nuclear region . Recently, Galotto et al. (2020) reported that chitin can elicit oscillatory calcium signals in Physcomitrium, although these oscillations appear to occur primarily in the cytosol.
In legume root cells, nuclear calcium spiking is elicited by specific chemical signals from potential symbionts (Ehrhardt et al., 1996;Chabaud et al., 2011). We therefore posited that nuclear-targeted GECIs may allow us to screen for treatments expected to activate Physcomitrium CCaMK-IPD3 but unexpectedly found that protonematal nuclei exhibited repetitive calcium spiking absent any experimental treatment. Observed oscillations were neither restricted to nor noticeably distinct in nuclei of apical cells, therefore, they did not appear to be associated with mechanical stimuli caused by tip growth, as has been described in pollen tubes (Moser et al., 2020). It will therefore be highly informative for future studies to examine the activation mechanism of Physcomitrium CCaMK and decipher the calcium signatures it decodes in the context of moss protonemata. It is noteworthy that excitation of GECIs with blue light may have inadvertently stressed cells during our calcium imaging experiments or participated in elicitation of nuclear calcium oscillations. We noticed increased autofluorescence from chloroplasts during acquisitions extending over roughly 30 min (Videos S1 and S2), suggesting cell stress that may also influence calcium oscillations. Blue light has been linked to calcium signaling in plants, including Physcomitrium (Russell et al., 1998;Stoelzle et al., 2003); and, interestingly, blue light can have inhibitory effects on nodulation in Lotus (Shimomura et al., 2016).
While our observations open new avenues for inquiry, further work is needed to identify factors involved in the coding and decoding stress-induced calcium signals in moss. Specifically, a central question for further investigation identified by this study is how nuclear calcium oscillations are coupled to CCaMK activation status and phospho-regulation of IPD3 in Physcomitrium. This topic has been extensively investigated and modeled in the context of legume symbioses with rhizobia and mycorrhiza (e.g., Kosuta et al., 2008;. Our findings collectively implicate a different physiological function and possibly a more complex decoding mechanism for CCaMK-IPD3 in Physcomitrium, given that nuclear calcium oscillations appear to occur constitutively without any apparent evidence for CCaMK-IPD3 activation (e.g., ABA-associated responses or brood cell formation). Inspired by engineering of optical reporters from mammalian CaMKII (e.g., Bossuyt and Bers, 2013), development of a fluorescent biosensor from Physcomitrium CCaMK that reports the active conformation of the kinase would provide a powerful Figure 5. Continued (D) RT-qPCR analyses indicated that activation of the CCaMK-IPD3 signaling module is associated with elevated transcript levels for the ABA response marker genes LEA3-1 and LEA3-2. Error bars indicate SEM among biological replicates. Results were statistically evaluated using the Tukey honestly significant difference (HSD) test. The p-values that indicated statistical significance below a threshold of 0.05 are annotated.
(E) Enzyme-linked immunosorbent assays (ELISAs) revealed substantially elevated levels of (+)-ABA in protonemal tissues of lines expressing PpIPD3 DD compared to wild-type controls under standard in vitro growth conditions. Error bars indicate SEM among three biological replicates each from two independently transformed lines. The p-value was obtained using a two-sample Student's t-test.

OPEN ACCESS
iScience 25, 103754, February 18, 2022 13 iScience Article tool to more precisely examine the physiological function of the moss CCaMK-IPD3 signaling module. A similar strategy could be pursued using IPD3, although could be complicated as almost nothing is known about IPD3 structure except for a coiled-coil domain. Fö rster resonance energy transfer (FRET) between fluorophores separately tagged to CCaMK and IPD3, as implemented in a previous study in Lotus (Singh et al., 2014), may also be an effective strategy to pinpoint stimuli that endogenously activate CCaMK-IPD3 in Physcomitrium. Figure 6. Neither CCaMK nor IPD3 is required for brood cell formation under ABA treatment or osmotic stress conditions (A) Physcomitrium ccamk andipd3 deletion mutants did not show any obvious phenotypic aberrations relative to WT under standard in vitro growth conditions in the BCDAT medium. Scale bar = 500 mm (B) Physcomitrium ccamk and ipd3 deletion mutants were able to develop brood cells when 100 mM ABA was added to BCDAT medium under standard growth conditions or when hyperosmotic stress was applied by addition of 500 mM mannitol to the growth medium. Treatments were performed for two weeks before images were taken. Two independently generated deletion mutant lines each were tested for IPD3 and CCaMK with similar results. Data from a single experiment are shown. Scale bar = 50 mm. The established physiological function of brood cells is to serve as stress-resistant asexual propagules that break away from parent plants and enable mosses to escape osmotic stress and dehydration. For this reason, they have also been referred to to as 'vegetative spores.' The phenotypes of Physcomitrium CCaMK and IPD3 gain-of-function lines suggest that these components may also be linked to osmotic stress and dehydration responses. Because osmotic stress and dehydration are closely related to oxidative stress, it is worth noting that CCaMK has been associated with oxidative stress responses in other plants. CCaMK expression is induced by ABA or oxidative stress in rice; CCaMK was also required for ABA-mediated antioxidant responses (Shi et al., , 2014. Similarly, CCaMK has been reported to be activated by nitric oxide and required for ABA-mediated antioxidant activity in maize (Ma et al., 2012;Yan et al., 2015). In wheat (Triticum aestivum), CCaMK expression is modulated by ABA and osmotic stress, likely through the activity of numerous predicted ABA-response elements in its promoter region (Yang et al., 2011). Moreover, we noticed using the Physcomitrella Expression Atlas Tool (PEATmoss) that CCaMK transcript abundance is elevated under heat stress conditions (Fernandez-Pozo et al., 2020). In light of these observations, the CCaMK-IPD3 may play a role in abiotic stress acclimation in Physcomitrium and other plants. This possibility may explain retention of the CCaMK-IPD3 signaling module in non-mycorrhizal mosses.
The cell wall thickenings, lipid reserves, and enhanced dispersal ability of brood cells could conceivably be useful in evading pathogenesis, although this idea is presently unsubstantiated. Fungal pathogens, as well as oomycetes, have been shown to induce reactive oxygen species (ROS) production, cell-wall depositions (including callose depositions mediated in part through ABA signaling), and altered fatty acid metabolism (Oliver et al., 2009;Ponce de Leó n, 2011;de Leó n et al., 2015). There are also mechanistic links between pathogen perception and the symbiosis pathway in angiosperms. In rice, chitin-receptor cerk1 mutants were impaired in both mycorrhizal and blast fungus infections (Miyata et al., 2014), and CERK1 is conserved as a chitin-induced immunity signaling receptor in Physcomitrium, possibly hinting at a further link between the common symbiosis pathway and immunity signaling (Bressendorff et al., 2016). These observations may warrant further investigation into the possibility that brood cells and the CCaMK-IPD3 pathway could also serve a heretofore unnoticed function in moss acclimation to biotic stresses.
Functional dissection of possible contributions of CCaMK-IPD3 to stress signaling in Physcomitrium will require further investigation and will likely rely on combining loss-of-function mutations and/or extensive screening of stress treatments (or combinations of stress treatments). We hypothesize that functional redundancy may occur through an ABA-dependent pathway in addition to the CCaMK-IPD3 pathway investigated in this study and that there may be crosstalk among stress signaling pathways upstream of stress-associated developmental reprogramming and brood cell formation. A similarly complex scenario has been described for stress signaling in angiosperm guard cells, wherein ABA and calcium function in a partially independent yet also synergistic manner (Webb et al., 2001;Huang et al., 2019;Schulze et al., 2021). In addition to identifying putatively functionally redundant signaling components, other interesting topics for future work include whether parallel pathways are calciumdependent or-independent and whether ABA hyperaccumulation, which we observed, is required for the developmental phenotypes of CCaMK-IPD3 gain-of-function lines. Calcium is a ubiquitous secondary messenger with a vast array of functions in plant cells. How specificity can be achieved and maintained when a common signal is employed for diverse functions has been a long-standing mystery. Parallel signaling pathways may be one mechanistic explanation; oscillatory calcium signals may be another mechanism for specificity and fidelity. Advances in calcium imaging and other biosensor technologies may empower future studies to demystify how calcium signals are coded in the model moss Physcomitrium.

Summary
In this study, we demonstrated that moss homologs of CCaMK and IPD3 have retained biochemical properties critical for functionality in legumes and are able to at least partially genetically complement cognate mutants in heterologous expression assays. Nonetheless, Physcomitrium does not appear to host canonical microbial symbionts such as mycorrhizal fungi. Synthetic activation of CCaMK or its downstream target transcriptional activator IPD3 in Physcomitrium induces ABA signaling and the constitutive formation of brood cells, which serve as asexual propagules that enable escape from abiotic stresses. The unexpected finding that protonematal nuclei exhibit spontaneous calcium spiking prompts questions about the ll OPEN ACCESS iScience 25, 103754, February 18, 2022 iScience Article regulation of CCaMK in Physcomitrium by calcium and provides fertile ground for future studies. Overall, our observations are consistent with a model wherein PpCCaMK-IPD3 functions to decode stress-associated calcium signatures and developmental reprogramming. Functional inquiries into CCaMK and IPD3 homologs in other early-diverging embryophytes such as the mycorrhizal host plant Marchantia paleacea and charophyte green algae are expected to complement these efforts and provide a fuller perspective of the evolutionary establishment of the molecular mechanisms underpinning the plant-microbe common symbiosis pathway.

Limitations of the study
Here, we have shown that synthetic activation of CCaMK or its target transcription factor triggers ABAassociated developmental reprogramming and formation of asexual propagules termed brood cells.
Notably, ccamk or ipd3 knockout mutants are still able to form brood cells in response to stress or ABA treatments, which would be consistent with parallel or alternative signaling processes; future work should target these putative components (e.g., by combining loss-of-function mutations). Mechanistic insight may also be gleaned by testing whether expression of synthetically activated forms of CCaMK or IPD3 is sufficient to trigger brood cell formation in mutants defective for ABA biosynthesis (e.g., Takezawa et al., 2015). Interpretation of the constitutive brood cell formation phenotype of gain-of-function lines is complicated by overexpression driven by a strong heterologous promoter. Expression of modified CCaMK or IPD3 forms from their native locus via homologous recombination may provide clearer insight into the endogenous function of CCaMK-IPD3 in Physcomitrium. Nonetheless, it is worth noting that seminal studies of CCaMK in legumes relied on constitutive strong (rather than native) promoters (Gleason et al., 2006;Tirichine et al., 2006). Deeper understanding of the role of CCaMK-IPD3 in elicitation of brood cell development may be facilitated by inducible expression of synthetically activated forms of CCaMK or IPD3 or engineering of light-controllable CCaMK or IPD3 derivatives (Zhou et al., 2012;Kubo et al., 2013). Further investigation of processes that govern CCaMK-IPD3 activation in Physcomitrium is needed, and observations of spontaneous nuclear calcium spiking hint that CCaMK-IPD3 regulation may be more complex in Physcomitrium protonemata than in legume root cells, wherein calcium spiking has been observed specifically in response to chemical elicitors (Ehrhardt et al., 1996;Chabaud et al., 2011). Except for extrapolation from legume homologs and heterologous complementation assays in this study, nothing is presently known about endogenous calcium signatures that lead to Physcomitrium CCaMK activation. The calcium imaging tools generated here provide a valuable route for further investigation. Next steps could examine effects of stress treatment on nuclear calcium oscillations; such efforts could be bolstered by inclusion of additional GECI with different spectral properties such as XCaMP-Yellow (Inoue et al., 2019), as blue and red light have been reported to trigger calcium signals in Physcomitrium protonemata (Ermolayeva et al., 1997;Russell et al., 1998).

Co-culture of Physcomitrium and Rhizophagus irregularis
Three-week-old gametophores were collected from cellophane-overlaid Knop agar plates. Gametophores were transferred to half-strength Knop semi-liquid medium containing 0.15% Phytagel (Sigma) in 24-well plates. In each well, 2 mL of semi-liquid Knop medium was poured, and approximately 50 spores of R. irregularis IRBV 0 95 were added. One gametophore was placed gently over the medium such that the rhizoids were immersed within the semi-liquid medium. This experimental setup was incubated at 25 C with a light intensity of 55mmolm À2 s À1 , with a 16 h photoperiod, for six months. Starting from one month after coculturing, AMF colonization was analyzed every week for up to six months using bright-field or confocal microscopy. Trypan blue staining of R. irregularis was performed as described (Koske and Gemma, 1989). For confocal microscopy, the fungal hyphae were stained using Wheat Germ Agglutinin Alexa Fluorâ 488 (Excitation: 488 nm; Emission: 520 nm), and the rhizoids, and gametophyte tissues were observed by chlorophyll autofluorescence (Excitation: 488 nm, Emission: 670 nm).

Bioinformatic analyses
To identify CCaMK and IPD3 homologs encoded in the Physcomitrium genome, the full-length protein sequences of CCaMK and IPD3 from Lotus japonicus (LjCCaMK, UniProt: A0AAR7, Lotus Base: Lj3g3v1739280; LjIPD3, UniProt: A9XMT3, Lotus Base: Lj2g3v1549600) and Medicago truncatula (MtCCaMK, UniProt: Q6RET7, Phytozome: Medtr8g043970; MtIPD3, UniProt: A7TUE1, Phytozome: Medtr5g026850) were retrieved from UniProt and used as BLASTp queries against the predicted Physcomitrium patens version 3.3 predicted proteome (Lang et al., 2018b) using Phytozome 12 (https:// phytozome.jgi.doe.gov/pz/portal.html). The BLOSUM62 scoring matrix was used. E-value thresholds were set to À1 for CCaMK searches and 1 3 10 4 for IPD3 searches. Other parameters followed default settings. Data were downloaded and analyzed from December 11 to 17, 2019. The top five hits were used as queries for reciprocal BLASTp searches against either the Lotus japonicus MG20 v3.0 protein database (https://lotus.au.dk/blast/#database-protein) using default settings or the Medicago truncatula Mt4.0 predicted proteome, accessed through Phytozome 12 and performed using default settings (Tang et al., 2014;Mun et al., 2016). For each reciprocal BLASTp search, the top hit was displayed. Multiple sequence alignments were made using MUSCLE (Edgar, 2004) version 3.8.425 plugin for Geneious Prime under default settings and were annotated manually or using the InterProScan feature in Geneious Prime (Biomatters).Accession numbers for putative orthologs referred to in Figure 1A can be found in Table S1.

Molecular cloning and plasmid construction
DNA and RNA were extracted from protonemal tissue by chloroform phase separation and cetrimonium bromide (CTAB) buffer as previously described (Chang et al., 1993;Kleist et al., 2014). The Quantitect (Qiagen) reverse transcription kit was used to synthesize cDNA for cloning and qPCR. PCR reactions were performed using Phusion (Thermo Fisher) or Primestar GXL DNA Polymerase (Clontech). The sequences of oligonucleotide primers used in this study are given in Table S3. Single-and multi-site Gateway (Thermo Fisher) cloning reactions were performed per the manufacturer's recommendations. The coding sequences (CDSs) of PpCCaMK, PpCCaMKb, and PpIPD3 were cloned into pDONR/Zeo and modified, as described, by site-directed mutagenesis using whole-plasmid amplification with anticomplementary primers followed by digestion with FastDigest DpnI (Thermo Fisher). The coding sequences of PpCCaMK, PpCCaMKb, and PpIPD3 were subcloned into pGAD-GH-GW or pGBT9-GW for yeast two-hybrid analysis. Assembly PCR was used to attach an eGFP tag and a polyglycine linker to the N-terminus of IPD3 or IPD3 DD . The vectors containing NLS-GCaMP6s or NLS-MatryoshCaMP6s (Ast et al., 2017) driven by Zea mays UBIQUITIN1 promoter were cloned by Gateway LR reaction into a modified pANIC5a vector with the region encoding Porites porites RFP deleted by PCR followed by In-Fusion reaction (Takara).

OPEN ACCESS
iScience 25, 103754, February 18, 2022 25 iScience Article Extraction and quantification of ABA. Abscisic acid (ABA) extraction and quantitation was done was previously described with slight modifications (Ondzighi-Assoume et al., 2016). Extraction from moss tissue was performed using methanol. For each sample, 100 mg of fresh tissue was collected, flash frozen in liquid nitrogen, and lyophilized. After lyophilization, three sterile 3 mm glass beads were added to the dried samples and the samples were macerated at max speed in a bead-beater for 3 min. 1 mL of chilled methanol supplemented with 2.5 mM citric acid monohydrate and 0.5 mM 2,6-di-ter-butyl-4-methlyphenol was added to the sample tubes and the samples were incubated for 16 h on a shaking platform at 4 C at 30 rpm. After extraction, sample tubes were centrifuged at 5,000 rcf for 10 min at 4 C, and the supernatant methanol extractions were transferred to clean tubes. Sterile deionized water was added to each extraction to adjust to 70% methanol. To remove chlorophyll and other assay-inhibiting compounds, the adjusted extractions were each passed through a C18 Sep-Pak cartridge (Waters) that was pre-equilibrated with fresh 70% methanol. The resulting eluates were centrifuged and dried in a speed vac at ambient temperature for 16 h. The resulting dried eluates were resuspended in TBS prior to 40-fold dilution with TBS and subsequent assay by colorimetric Phytodetek ELISA kit (Agdia) using the manufacturer's standard protocol. ELISA plate results were quantified using a Tecan Infinite M1000 Pro microplate reader.

Measurement of cell dimensions
Physcomitrium samples were taken from edges of approximately month-old cultures grown under standard conditions, as described above. Three independently transformed lines were analyzed per construct. Measurements were performed manually in Zen Blue software version 2.6 (Zeiss). Graphs were made using Origin Pro 2020.

Statistical analyses
For quantitative measurements of ABA content, seven biological replicates were analyzed for wild-type, and three biological replicates of three independently transformed lines were analyzed for IPD3 DD . Every biological replicate was tested in triplicate for the ABA ELISA. Data were analyzed and tested using a Student's T-Test using the R statistical programming language (R Core Team, 2014). For statistical analyses of RT-qPCR data, samples were compared via one-way ANOVA analysis using R (R Core Team, 2014). Levene's Test confirmed equality of variance for both sets of data (Levene, 1960). The Tukey honest significant difference (HSD) was used for posthoc analysis of ANOVA results (Tukey, 1949), and the p values that were reported were calculated using this method. For ipd3-1 rescue experiments, each treatment was compared for differences in total nodule number per root, the number of colonized nodules per root, and the number of nodules containing rhizobia expressing nifH per root using R (R Core Team, 2014). The sample size per treatment varied from 16 roots to 44 roots. Levene's test determined equal variance for total nodule number between samples, but not for either number of colonized nodules or number of nifH expressing nodules (Levene, 1960