Cytokinin‐induced promotion of root meristem size in the fern Azolla supports a shoot‐like origin of euphyllophyte roots

Summary The phytohormones cytokinin and auxin orchestrate the root meristem development in angiosperms by determining embryonic bipolarity. Ferns, having the most basal euphyllophyte root, form neither bipolar embryos nor permanent embryonic primary roots but rather an adventitious root system. This raises the questions of how auxin and cytokinin govern fern root system architecture and whether this can tell us something about the origin of that root. Using Azolla filiculoides, we characterized the influence of IAA and zeatin on adventitious fern root meristems and vasculature by Nomarski microscopy. Simultaneously, RNAseq analyses, yielding 36 091 contigs, were used to uncover how the phytohormones affect root tip gene expression. We show that auxin restricts Azolla root meristem development, while cytokinin promotes it; it is the opposite effect of what is observed in Arabidopsis. Global gene expression profiling uncovered 145 genes significantly regulated by cytokinin or auxin, including cell wall modulators, cell division regulators and lateral root formation coordinators. Our data illuminate both evolution and development of fern roots. Promotion of meristem size through cytokinin supports the idea that root meristems of euphyllophytes evolved from shoot meristems. The foundation of these roots was laid in a postembryonically branching shoot system.


Introduction
Roots are a key innovation of vascular plants. They probably arose twice, once in the fern lineage and once in lycopods (Pires & Dolan, 2012). Hence, the emergence of the complex sporophyte body was accompanied by the emergence of roots (Graham et al., 2000;Doyle, 2013). Phylogenetically early-branching bryophytes also form root-like rhizoids. Regardles of whether they are uni-or multicellular (Jones & Dolan, 2012), rhizoids are elementary structures that grow through tip growth in the same manner as root hairs (Menand et al., 2007). Gametophytes of ferns form rhizoids, too (Banks, 1999), but in contrast to the bryophytes, their life cycle is dominated by the sporophytic stage (as in angiosperms), in which they build proper roots (Gunning et al., 1978;Banks, 1999). During the evolutionary transition to a sporophyte-dominated plant life cycle, regulators of gametophyte development were recruited for sporophyte development (Menand et al., 2007;Frank & Scanlon, 2015). Gametophyte and sporophyte development in basal land plants share control mechanisms (Landberg et al., 2013;Bennett et al., 2014;Viaene et al., 2014), but we know little about complex roots of basal vascular plants beyond that.
Two major apical meristems determine complex plant growth: the shoot apical meristem (SAM) and the root apical meristem (RAM). Exploring the root's origin is therefore exploring the origin of the RAM. Apical development through a SAM-like structure precedes the emergence of the RAM (Graham et al., 2000;Prigge & Bezanilla, 2010) and regulation of the SAM and RAM show striking similarity (Stahl & Simon, 2010). It has therefore been speculated that the initial RAM evolved from the SAM (Stahl & Simon, 2010), a transition that must have happened in the early vascular plants. To understand this transition, we explored the molecular regulation of the fern RAM in comparison to that of angiosperms.
The Arabidopsis RAM has become the model system to study angiosperm meristem function (Petricka et al., 2012). While Arabidopsis forms a taproot system (allorhizy) that is dominated by the embryonic primary root, other plants such as grasses develop a fibrous root system (secondary homorhizy). Regardless of the mature root system architecture, angiosperms establish a primary root (and RAM) during early embryogenesis, serving angiosperm seedlings of both allorhizic and homorhizic species (Bellini et al., 2014). Angiosperm embryos establish the apicalbasal polarity already at the two-cell stage, a process orchestrated by polar auxin flow (Friml et al., 2003;Petrasek & Friml, 2009). Next to auxin, root identity in embryogenesis is determined by master regulators such as WUSCHEL RELATED HOMEOBOX (WOX) genes (Haecker et al., 2004;Schlereth et al., 2010;Radoeva & Weijers, 2014). Angiosperm roots develop longitudinally in three different zones: the meristematic zone (MZ), from which all root tissues arise, the elongation zone (EZ), where the cells reach their final size, and the differentiation zone (DZ), where elongated cells reach full maturation according to their respective fate (Dolan et al., 1993).
Fern embryos do not develop in a bipolar manner (Hou & Hill, 2002). The embryonic RAM is only briefly present and the permanent fern roots arise adventitiously from a shoot, a process termed primary homorhizy (Goebel, 1930;Schneider, 2013) that is clearly distinguished from (secondary) homorhizic angiosperms with long-living embryonic roots (Hou & Hill, 2002;Bellini et al., 2014). Instead of bearing a QC-centred SCN, fern roots have a single root apical cell (RAC) that gives rise to root tissues (Gunning et al., 1978;Hou & Blancaflor, 2009). Our knowledge about the molecular mechanisms underlying the function of the fern RAC is limited, but Nardmann & Werr (2012) could show that WUSCHEL-related transcription factors act in the RAC of Ceratopteris, demonstrating them to be ancient master regulators of plant development that radiated even before the angiospermgymnosperm split (Hedman et al., 2013). Studies on Ceratopteris showed that auxin did not alter lateral root formation, but inhibited parent root growth (Hou et al., 2004). Yet the molecular framework that underpins fern meristem regulation remained hidden. We used Azolla filiculoides, famous for its unique symbiosis with a nitrogen-fixing cyanobacterium (Rai et al., 2000;Carrapic ßo, 2010), to characterize how the key developmental regulators auxin and cytokinin influence fern root development. While the longitudinal organization of the Azolla root resembles that of angiosperms, we found that the phytohormonal meristem regulation through auxin and cytokinin is reversed. Our results shed light on the evolutionary origin of the RAM in euphyllophytes.

Phenotypic analysis
For differential interference contrast (DIC) analyses, roots were mounted in a chloral hydrate solution (5 : 2 : 1, chloral hydrate : glycerol : H 2 O) and bleached (i.e. tissues were cleared) for several days. Images were taken on a Nikon Ti Eclipse microscope (Nikon DS-Qi1Mc camera). Root lengths (entire 3 dpc root) were determined via dissection microscopy (Nikon SMZ 745T; Nikon DS-Ri1 camera). Images were processed using the NIS-Elements BR 4.20.00 software (Nikon, Tokyo, Japan) and Adobe Photoshop and Illustrator CS6. Statistical analysis was performed using R 3.0.2 (R Core Development Team 2013), normality was tested using a Shapiro-Wilk test (Shapiro & Wilk, 1965) and, accordingly, a Mann-Whitney U-test (Mann & Whitney, 1947) was performed. Amyloplasts were stained using 5% Lugol's iodine and pictures were taken on a Zeiss Axiophat microscope with an AxioCam ICc 5 camera, using the ZEN 2012 software (Zeiss). Dissection micrographs were generated using a SteREO Discovery V8 (Zeiss) microscopy with a AxioCAM ICc 5 and processed using the ZEN 2012 software (Zeiss).

Global gene expression analysis
For RNAseq analysis, 66 hpc roots were treated as described above but for 6 h; 3-4 mm root tips were collected and extracted using the Spectrum TM Plant Total RNA Kit (Sigma) according to the manufacturer's instructions. RNA was extracted for each treatment in triplicate (n ≥ 100 root tips each), always at the same time of the day. RNA quality was assessed using a formamide gel and shipped on dry ice to BGI Tech Solutions (Hong Kong), where RNA quality assessment by BioAnalyzer (Agilent Technologies, Waldbronn, Germany), library preparation using the TruSeq kit (Illumina, San Diego, CA, USA), and 100 bp paired-end sequencing via the Illumina HiSeq2000 system were performed. After removal of low-quality reads, we obtained a total of 270 866 324 reads (Supporting Information Table S1). The Transcriptome Shotgun Assembly project has been deposited at DDBJ/EMBL/GenBank under the accession GBTV00000000 and the National Center for Biotechnology Information (NCBI) Single Read Archive (PRJNA264391). The version described in this paper is the first version, GBTV01000000.
Differential regulation and statistical analysis were performed based on negative binomial probability distribution and Benjamini-Hochberg correction using edgeR (Robinson et al., 2010). Wordles were generated using wordle.net. Phytohormone signalling pathways were manually curated based on Arabidopsis homologues (Liscum & Reed, 2002;Lamesch et al., 2011;Brenner & Schm€ ullig, 2012). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed using the Power SYBR TM Green kit and a StepOnePlus TM (Applied Biosystems) system and analysed according to Pfaffl (2001). AfTufA (AzfiRT00021) and AfCAM5 (AzfiRT00154) were selected as reference genes based on their steady expression (confirmed during qPCR analysis on equal amounts of RNA) in the RNAseq data (see Table S3 for primer sequences).

Hierarchical clustering of expansin expression and phylogeny
An expansin dataset was generated based on best-BLASTx hits against the Azolla root transcriptome and sequences from Sampedro & Cosgrove (2005), Li et al. (2002), the NCBI and the Arabidopsis Information Resource (TAIR ; Table S4). Protein sequences were generated using the sequence manipulation suite (Stothard, 2000), aligned using Multiple Alignment using Fast Fourier Transform (MAFFT) v.7.127b L-INS-I (Katoh & Standley, 2013) and MEGA5.2.2 (Tamura et al., 2011) and modified to include only the conserved middle region of the protein sequences. Sequences containing only parts of the conserved region were removed. A maximum-likelihood phylogeny was computed using WAG + G + I (best-fit substitution model determined by MEGA5.2.2; five discrete gamma categories, 1000 bootstrap replicates, partial deletion site coverage cutoff: 99%). Sequences were annotated based on best BLAST hits to Arabidopsis expansins, unless annotation was given. Hierarchical clustering on the normalized expression data was performed by using log 2 (FC) data. Data for 3 h CK-and IAA-treated Arabidopsis thaliana seedlings were extracted from The Bio-Analytic Resource for Plant Biology (BAR) (Toufighi et al., 2005;Winter et al., 2007). Clusters were generated using the CLC main workbench 7.0 (CLCbio, Qiagen) and Euclidean distance measures with single linkage.

Results
Azolla has adventitiously emerging shoot-borne roots Azolla roots emerge adventitiously as shoot-borne roots (cf. Groff & Kaplan, 1988) from a node that contains a ventral and a dorsal leaf lobe (Fig. 1a). Roots of angiosperms can be longitudinally separated into the DZ, EZ and MZ (Dolan et al., 1993), a partitioning that we also used for the roots of Azolla. The largest and uppermost segment of the Azolla root is made up by the DZ, where the cells reached their final size, the xylem is differentiated and root hairs are prominent. The EZ begins with cells that are approximately twice as long as those of the MZ (from 8.7 AE 2.5 lm of the three outer cortex cells before the doubling to 16.8 AE 4.5 lm of the three outer cortex cells after the doubling, marking the TZ; Fig. 2a). At the tip lies the MZ, housing the RAC. Trichoblasts, from which the root hairs later emerge, are already distinguishable in the MZ as a result of their triangular appearance; proper differentiation of the root hairs occurs at the end of the MZ and subtending the inner root cap (Fig. S1). Azolla roots bear an outer (

Size of the Azolla root MZ is increased by cytokinin and decreased by auxin
Auxin and cytokinin are key regulators of the MZ's activity in angiosperms and thus we analysed their effect on the Azolla root's MZ. To make sure that only meristems of the same age and developmental stage were analysed, roots were first removed. At 1 dpc, Azolla roots had replaced the removed roots but showed nonuniformal phenotypes. We used the distinct TZ to evaluate the Azolla MZ (starting from first emergence, cells of the outer cortex file were counted until they had doubled in size) and determined that 3 dpc roots showed the steadiest MZ size (Fig. S3), which were thereafter used for all analyses. Ferns were kept in floating culture for 2 dpc and transferred to a new culture containing either solvent (mock treatment) or 0.1 lM IAA or 0.5 lM trans-zeatin (CK) for 24 h. The concentrations were chosen based on previous studies on meristem size in Arabidopsis (R u zi cka et al., 2009;Dello Ioio et al., 2012;Moubayidin et al., 2013), as well as preliminary experiments determining the minimal concentrations resulting in clear phenotypes. Mock-treated roots (3 dpc) had an MZ cell number of 77 AE 12 ( Fig. 2a,b). Exogenous 0.1 lM IAA application led to roots with a significant reduction (P < 0.001) in MZ size (62 AE 12; Fig. 2a,b). By contrast, application of 0.5 lM CK increased the MZ size (111 AE 19, P < 0.001; Fig. 2a,b). Application of the phytohormones also influenced the lengths of the MZ cells (Fig. 2a). In mock-treated roots, the first 10 outer cortex cells had an average length of 5.4 AE 1.4 lm; IAA treatment increased this to 6.9 AE 1.4 lm (P < 0.001), while CK treatment decreased cell size to 4.7 AE 1.7 lm (P < 0.001). Control 3 dpc roots were 6.81 AE 3.1 mm in length. While 0.5 lM CKalthough changing the meristem cell numberled to no significant change in root length (6.81 AE 2.7 mm; n = 40), 0.1 lM IAA resulted in a significant decrease (P < 0.001), to 3.73 AE 1.6 mm (Fig. 2c,d).

Auxin and cytokinin affect xylem formation
Auxin and cytokinin are also key regulators of root vasculature in Arabidopsis (Bishopp et al., 2011) and we therefore analysed their influence on xylem morphology under the same conditions as applied earlier (n = 40). Azolla roots bear two narrow protoxylem and two wide metaxylem strands (Fig. 1b). The protoxylem is c. 4 lm in diameter, progressing from scalariform ornamentation to annular ornamentation (Fig. 3). Protoxylem differentiates, on average, seven stele cells above the apex. One protoxylem strand differentiates before the other (approx. two cells), and differentiating either earlier on IAA or later on CK treatment (Fig. 3d). Metaxylem shows a scalariform ornamentation. First metaxylem strands differentiate as wide cell files (c. 8 lm in diameter) more tipwards, and second metaxylem strands differentiate as even wider cell files (c. 13 lm in diameter) more shootwards, than the rim of the inner root cap (Fig. 1b). Upon either phytohormone treatment, both metaxylem strands differentiated more towards the tip than in the untreated roots. The relative appearance of the second compared with the first metaxylem strand was also altered such that both strands appeared to start closer to each other (CK and IAA), or even simultaneously (IAA; Fig. 3f). This was accompanied by a change in the perforation plate angle, resulting in orthogonal instead of rhomboid-shaped metaxylem cells (Fig. 3g). In addition, IAA caused the secondary cell wall deposition leading to the scalariform ornamentation being closer to each other (Fig. 3h). Both phytohormes resulted in a widening of the second metaxylem strand (Fig. 3i). Although varying marginally in size along the root, the second metaxylem strand was always wider than the first.

Embryophyte developmental regulators of the Azolla root
Plants colonized land at least 470 million yr ago (Gensel, 2008) and many regulators that shape plant bodies are ancient (Pires & Dolan, 2012). To understand which growth regulators support fern root development, > 270 million paired-end reads of mRNA isolated from 3-mm-long A. filiculoides root tipsand under three different conditions (see later)were sequenced. The de novo transcriptome data were assembled into 190 000 contigs, then filtered for those that were supported by ≥ 150 reads and an assembled length of ≥ 300 bp, resulting in 36 091 contigs ( Fig. 4a; Table S1). To test for the depth and breadth of key developmental regulators among Azolla transcripts, we clustered all proteins derived from available genome, RNAseq and EST data (≥ 100 amino acids long) of a variety of streptophytes ( Fig. 4) and Chlamydomonas reinhardtii as an outgroup using a bidirectional BLASTp approach. Screening this dataset for clusters that contained orthologues to 353 known A. thaliana developmental denominators resulted in 161 clusters that contained at least one orthologue from another species, 40 of which contained homologues expressed in the A. filiculoides root. In comparison, proteins from the genomes of Selaginella moellendorffii and Picea abies were detected in 36 and 58 clusters, respectively, and proteins from the transcriptomes of C. richardii and Pteridium aquilinum were detected in only 10 and 14 clusters, respectively. Thus, sequence depth of our de novo transcriptome was sufficient to detect developmental master regulators, such as homologues of WOX13 (AzfiRT13282, AzfiRT23190, AzfiRT23355; e-value < 10 À47 ), bearing the characteristic ancestral NVYNWFQN peptide sequence at the second turn of the recognition helix (Nardmann & Werr, 2012). Key denominators of true root meristems known from angiosperms, such as three homologues of PLT (AzfiRT20391, AzfiRT31772, AzfiRT33493; e-value < 10 À95 ), were identified (Table S5), accounting for the readout of the meristem regulon. But does differential expression of these key regulators translate into the different phenotypes observed upon phytohormone treatment?
To connect the phenotypic observations to molecular mechanism, RNA was sequenced in triplicate from 3 mm root tips treated for 6 h with 0.1 lM IAA, 0.5 lM CK or solvent (mock). The 6 h treatment was chosen to detect the established readout over initial signalling components of the phytohormone response (Goda et al., 2004). Observed trends in expression were confirmed by two-step qRT-PCR (Fig. S4). To our knowledge, this is the first global gene expression profiling on a fern root, so we first ran a GO term enrichment analysis (P < 10 À3 ) on the top 2000 highest expressed genes from all root datasets and compared this with sequence data available from the Azolla sporophyte (Brouwer et al., 2014) based on homology to Arabidopsis proteins (e-value < 10 À7 ). The most evident difference is the expression of photosynthesis and pigment metabolism GO terms, detected only in the photosynthesising sporophyte system ( Fig. 5a; Table S6). The root expression profiles were enriched for various kinds of transport GO terms, including multifaceted ion transport. Intriguingly, photorespiratory genes were highly expressed in the Azolla root, a phenomenon that was recently discussed for angiosperm roots as well (Nunes-Nesi et al., 2014). Using a ranked GO term enrichment analysis highlighted the differential responses towards IAA and CK (Fig. S5). GO terms for RNA biosynthesis were prominent, indicating that both treatments triggered strong and specific transcriptional responses by the fern. Performing a ranked comparison of genes up-and down-regulated by CK in comparison to IAA highlighted the antagonistic action of the phytohormones. Not only were GO terms for regulation of meristem growth up-regulated upon CK, but so too were cell cycle regulation and inositol metabolism, which are well-known growth regulators (Stevenson et al., 2000). Importantly, no GO terms associated with stress responses were highlighted. Hence, the reduction of meristem size observed upon auxin treatment is governed by developmental control and not stress.
Auxin and cytokinin differentially shape Azolla roots through cell wall modification To detect key players and candidate genes for future studies involved in governing Azolla root development upon IAA and CK, we pursued a top-down approach. Screening among contigs with homology to nuclear land plant genes, and which were significantly regulated (Benjamini-Hochberg corrected P-value < 0.05), identified several homologues (e-value < 10 À7 ) to Arabidopsis growth regulators ( Fig. 5b; Table S6). As indicated by the GO terms (Fig. S5) in particular, 'information processing' was up-regulated upon treatment. This included RNA processing such as the RNA splicing SR45 homologue (e-value 1.74 9 10 À09 ) and ribosomal proteins in general. Notable was the 54-fold up-regulation of a CYTOCHROME 450 71A20 (CYP71A20) homologue, whose orthologue in A. thaliana is down-regulated upon auxin treatment (Goda et al., 2004).
Both phytohormone treatments resulted in pronounced expression changes of the cell wall-modifying expansins (e.g. EXPA4 CK vs mock, 0.3-fold up in contrast to 2.1-fold down in IAA vs mock; P < 0.05). The a-expansins, in particular, were enriched among Azolla root transcripts, whereas only one b-expansin and one a-expansin-like were detected. Conserved regions of Azolla expansins (Fig. S6b) match well with those of higher embryophytes (Li et al., 2002;Sampedro & Cosgrove, 2005), including the aand b-specific insertions (Li et al., 2002). Phylogenetic analysis using 93 protein sequences from 10 embryophytes show a clear distinction between the aand b-expansins, the expansin-like sequences clustered with the b-expansins (Fig. S6). By contrast, other expansin phylogenies (Sampedro & Cosgrove, 2005), expansin sequences of

New Phytologist
Azolla filiculoides, Physcomitrella patens, Marsilea quadrifolia and Regnellidium diphyllum formed lower plant-specific clades instead of being equally dispersed throughout the a-sequences. We hierarchically clustered normalized expression data of a-expansins upon auxin and cytokinin treatment in A. thaliana and A. filiculoides. Four of the 10 clusters obtained were speciesspecific, and only one for Azolla (Fig. 6a). The remaining six clusters were dominated by one species, four out of six by A. filiculoides. To analyse putative subfunctionalization after species-specific duplications, we identified potential orthologues, co-orthologues and paralogues in A. filiculoides and A. thaliana and estimated pairwise distances within those groups using ORTHOMCL (Li et al., 2003). In agreement with our phylogenetic analysis (Fig. S6) we obtained only a few orthologues and a larger group of co-orthologues. When plotting the pairwise distances of each pair against its absolute expression differences, there was little correlation (Fig. 6b). All in all, the data support the conclusion that, based on a common expansin set in all land plants, fernsperhaps also mosseshave advanced their expansin repertoire, governing a differential use.
Other cell wall modifiers were specifically regulated by the phytohormones as well. A homologue of the cellulose biosynthesis-associated gene RADIAL SWELLING3 (RSW3; e-value 0) was down-regulated significantly (P < 0.05), when comparing CK with IAA treatment (3.6-fold down), and Arabidopsis rsw3 mutants display a severe root phenotype (Baskin et al., 1992;Burn et al., 2002). An IRX12 (IRREGULAR XYLEM 12, an enzyme involved in lignin biosynthesis) homologue (e-value 1.74 9 10 À166 ) was significantly (P < 0.05) down-regulated upon IAA (2.6-fold) and only marginally down-regulated upon CK. Arabidopsis irx12 mutants have strong alterations in xylem T F E x p a n s i n (a) (b) Fig. 4 RNAseq of Azolla root tips and identification of orthologues to known developmental regulators across 14 species of Chloroplastida. (a) Workflow of the RNAseq analysis of Azolla filiculoides root tips starting with the filtered reads obtained by illumina paired-end (PE) sequencing followed by assembly using the Trinity pipeline, annotation and filtering via BLASTx and read mapping using the CLC workbench. For downstream analyses (protein clustering, expression wordles of significantly regulated genes or gene ontology (GO) term enrichment via GORILLA), only contigs that had their best BLASTx hit to streptophytes were used. (b) Protein data were extracted from nine genomes (black species names) and five RNAseq and expressed sequence tag (EST) libraries (red species names), and clustered; Chlamydomonas reinhardtii serves as an outgroup. Clusters were generated from all-against-all bidirectional best BLASTp hits. Those clusters were screened for 353 Arabidopsis thaliana proteins from 10 gene families (top row) and 161 clusters that contained at least one other species were extracted. The global identity of the identified homologues is displayed as a gradient colour. Note the high number of identified homologues in the A. filiculoides root tip transcriptome. The phylogeny on the left is based on the National Center for Biotechnology Information taxonomy database. formation, displaying collapsing vessel elements (Brown et al., 2005), and regulation of Azolla IRX12 homologues might therefore be correlated with the observed xylem phenotypes. Key players for the correct xylem cell differentiation are cellulose deposition-guiding microtubules (Oda & Fukuda, 2012).
The expression of several MICROTUBULE ORGANISATION 1 homologues (MOR1; e-values < 1.18 9 10 À107 ) increased upon IAA and CK treatment (4.4-, 4.1-and 3.9-fold up and 4.1-, 3.8-and 3.7-fold up, respectively; Fig. 5b  The fold enrichment of the GO terms is depicted as a colour gradient and GO terms were manually sorted into umbrella categories (top). (b) Significantly differentially regulated (P < 0.05) homologues (e-value < 10 À7 ) to Arabidopsis expressed in 0.3 mm of the root tip of Azolla filiculoides, represented as a word cloud. The bigger the word, the stronger the differential regulation comparing 0.

Research
New Phytologist mediates microtubule-directed cell wall development (Buschmann et al., 2004) was significantly down-regulated by auxin (2.8-fold down; P < 0.05). Gene expression changes associated with vascular function were also detected. This included regulation of the ABC transporter ABCB15 (AzfiRT08887; evalue 0; 1.9-fold down upon IAA treatment) that is expressed in developing vasculature and associated with auxin transport and secondary cell wall development (Kaneda et al., 2011) and the phloem amino acid transporter BAT1 homologue (Az-fiRT35714; e-value 0), which was 3.7-fold up-regulated upon CK treatment (P < 0.05). Intriguingly, nitrate transport-associated homologues were up to eightfold down-regulated upon CK treatment (SLAH2; AzfiRT18705, AzfiRT19565), which could be explained by the link between CK biosynthesis and nitrogen availability (Takei et al., 2004). IAA and CK, therefore, shape the root of Azolla radially and longitudinally by the differential regulation of cell wall-loosening and -reinforcing enzymes.
Expression of three known auxin signalling factors was significantly altered upon auxin treatment (Figs 5, 7). Two Aux/ IAA homologues were detected, grouping with the B-group Aux/IAA proteins of Arabidopsis ( Fig. S7; Remington et al., 2004) that are known for their importance in primary and lateral root development. One Azolla Aux/IAA homologue showed its highest identity to IAA13 (AzfiRT01636; e value 6.64 9 10 À35 ; up-regulated almost two-fold upon auxin treatment), known to be important for root initiation during embryogenesis and later expressed in the stele (Weijers et al., 2005). Besides its phylogenetic placement in the B-group, the other homologue showed highest identity to IAA14 (Solitary root; AzfiRT00243; e value 1.84 9 10 À35 ; up-regulated almost two-fold upon auxin treatment). IAA14 is a known regulator of lateral root formation and Arabidopsis iaa14 mutants do not generate any lateral roots (Fukaki et al., 2002). Finally, a transcript of an AUXIN RESPONSE FACTOR was found to be down-regulated upon IAA treatment, showing highest identity to ARF7 that is known to be a crucial regulator of lateral root formation (Okushima et al., 2007). Also a homologue of the transcription factor SHOOT REDIFFERENTIATION DEFECTIVE 2 (SRD2; e value 1.40 9 10 À10 ) was found to be up-regulated upon CK treatment (5.4-fold; P < 0.05) and Arabidopsis srd2 mutants are known to have impaired lateral and primary root growth (Ohtani & Sugiyama, 2005). Thus, while several components of the CK and IAA signalling pathway are expressed by the Azolla root, only specific factors showed a strong reaction after 6 h of the respective phytohormone treatment, many of which are associated with lateral root formation in Arabidopsis.

Discussion
Understanding meristem regulation is key to understanding plant development. The RAM of Arabidopsis has become a priced model to uncover the basic principles of angiosperm development (Petricka et al., 2012), but our knowledge about the evolutionary earlier fern root development is rather scarce (Hou & Blancaflor, 2009). Our results highlight the fundamental differences in the control of the Azolla RAM by the key regulators cytokinin and auxin.
Auxin, cytokinin and the reciprocal regulation of the root system in Azolla Similar to Arabidopsis (Dolan et al., 1993), we found that Azolla has a DZ, EZ and MZ. While auxin is known to increase Arabidopsis MZ cell number (Blilou et al., 2005), cytokinin is known to reduce it (Dello Ioio et al., 2007). Similar applies to Arabidopsis' adventitious roots, which one can assume constitute the more obvious analogue to the Azolla root. Auxin is a known inductor and promoter of adventitious roots, while cytokinin acts antagonistically (Bellini et al., 2014). Surprisingly, application of IAA and CK to the Azolla adventitious roots resulted in the opposite effect. Root growth inhibition through auxin is well known in the fern Ceratopteris (Hou et al., 2004) and also in Arabidopsis, where it acts on cell elongation (Rahman et al., 2007). Yet the exogenous auxin naphthaleneacetic acid (NAA; acting in a similar manner to IAA on root meristems; Rahman et al., 2007) increases MZ cell number in the latter and shows inhibitory effects only at very high concentrations (R u zi cka et al., 2009). While decreasing MZ size through earlier differentiation (Dello Ioio et al., 2007), cytokinin affects root length in Arabidopsis only at high concentrations (R u zi cka et al., 2009). In the case of Azolla, CK did not alter root growth but it did alter MZ cell number. This might be explained by a CK-induced delay in differentiation and a simultaneous restraint on cell length.
In summary, IAA influenced both cell elongation (in MZ and TZ) and MZ cell number. This resulted in its significant impact on root length, whereas phenotypes triggered by CK were mainly governed by changes in cells of the MZ and not the elongation in Azolla roots. This is further supported by the xylem phenotypes that displayed either a delayed or a premature differentiation of proto-and metaxylem. More densely ornamented xylem and orthogonal instead of rhomboid metaxylem cells observed upon IAA treatment might be the simple consequence of a delayed or impaired elongation. All of these are linked to the relative position to the apex and are thus probably a direct cause of longitudinal alterations. Our information is limited with regard to understanding the upstream molecular regulation that elicits these phenotypic alterations. For example, the RNAseq analysis revealed CYP71A20 to be differently regulated upon auxin treatment in comparison to its A. thaliana homologue (Goda et al., 2004). This might suggest a possible difference in the upstream

Research
New Phytologist perception of auxin in the Azolla root, but it is just as likely that auxin and cytokinin signalling acts downstream on very specific factors. Azolla is currently not accessible for genetic studies. Nevertheless, based on sequence homology, a set of genes were detected that can be associated with the observed phenotypes induced by the phytohormones; that provide information about the Azolla root developmental regulon and identity; and that are valuable candidate genes for future studies on fern root development.

Regulon of Azolla roots
The rigid cell walls of plants pose a challenge when a change in phenotype is required. We found that the most strongly regulated genes in Azolla roots were those involved in cell wall modification, a major group of which were the expansins that are involved in virtually all processes during plant cell remodelling and growth (Sampedro & Cosgrove, 2005). Of all expansins, the a-expansins are thought to be the most gene-rich group in the common ancestor of embryophytes (Sampedro & Cosgrove, 2005). The same is true for the transcriptome of the Azolla root. We did not find many orthologous a-expansin genes between A. thaliana and A. filiculoides, but we did find co-orthologous ones. This suggests that after an initial round of expansin family inflation in the common ancestor of embryophytes, several subsequent rounds of lineage-specific radiations occurred. Expansins influence cell wall loosening, which is linked to the auxin-mediated acid-growth hypothesis (Rayle & Cleland, 1992;Link & Cosgrove, 1998;Catal a et al., 2000;Cosgrove, 2000). Yet exogenous auxin is known to reduce root length in Arabidopsis through altering the cell elongation (Rahman et al., 2007;R u zi cka et al., 2009). IAAinduced reduction in root length and down-regulation of expansins suggest that the same occurs in Azolla roots. Cell wall modification and cellulose deposition are further mediated through microtubule organization (Wasteneys, 2004). Intriguingly, both CK and IAA lead to similar changes with regard to xylem phenotypes and up-regulation of MOR1 homologues. Owing to the direct link between xylem phenotype and microtubule organization (Oda & Fukuda, 2012), both phenotypes could be explained through the same mechanism: genetically controlled cell wall modification that shapes the Azolla root.
Fern root systems are governed by adventitiously arising roots (primary homorhizy); their primary RAM and resulting primary root are only short-lived (Hou & Blancaflor, 2009). This stands in clear contrast to (secondary) homorhizy in angiosperms, where the primary root is long-living. Yet, as in angiosperms, each of these adventitious roots has, similar to lateral roots (De Smet et al., 2006) root's meristematic signature, not least because the control of adventitious and lateral root formation seems to largely overlap (Gutierrez et al., 2009;Bellini et al., 2014). Indeed, although Azolla does not have any lateral roots, a salient amount of differentially expressed Azolla root genes were homologous to lateral root developmental regulators. In angiosperms the determination of the lateral root founder cell identity is largely regulated by auxin (Fukaki et al., 2005). Yet, in Ceratopteris, auxin affects only the adventitious parent root, but not lateral root development (Hou & Hill, 2002;Hou et al., 2004). Consistent with the canonical Arabidopsis auxin signalling pathway (Liscum & Reed, 2002), expression of b-group Aux/IAAs was induced upon IAA, but apparently did not promote the activity of the adventitiously formed RAC in a manner that would result in a larger MZ. Yet, cytokinin can also regulate lateral root development. Homozygous Arabidopsis wol mutants display no lateral root emergence from the primary root, but bear only hypocotyl-borne roots (Scheres et al., 1995). We found WOL to be one of the most ubiquitous cytokinin signalling components in the Azolla root. In addition, a possible homologue of BREVIS RADIS (AzfiRT34406; e-value 1.01 9 10 À06 ; Table S5) was also detected, which is thought to be crucial for cytokinin's regulation of lateral root formation (Li et al., 2008). What could this imply for the evolutionary origin of the euphyllophyte root?
From thallus to cormus and from adventious to primary roots Apical growth took place in the common ancestor of all land plants and precedes the evolution of roots (Graham et al., 2000;Prigge & Bezanilla, 2010;Pires & Dolan, 2012). Even the most basal plants, including some streptophyte algae (Kenrick & Crane, 1997;Wodniok et al., 2011), develop through the action of an apical cell (Graham et al., 2000). Liverworts, or hornworts, are often discussed to be the most ancient land plants (Gensel, 2008;Wickett et al., 2014) and their thalli grow through marginal meristems (Hagemann, 1999), but also by the activity of apical cells (Goffinet & Buck, 2013). Both sporophytes and gametophytes of ferns (and mosses) bear apical cells mediating their growth (Hagemann, 1999;Prigge & Bezanilla, 2010;Schneider, 2013). Although the roots of lycophytes and euphyllophytes most probably stem from two independent origins, both develop through an RAC (Prigge & Bezanilla, 2010;Pires & Dolan, 2012). Sanders & Langdale (2013) found that auxin and cytokinin act antagonistically on branching in the lycophyte Selaginella kraussiana; in that case, auxin led to a higher degree of root branching. It was concluded that this mechanism could be homologous to lateral root development of angiosperms, in which auxin promotes lateral root formation. In the fern Ceratopteris, however, lateral root development was not influenced by auxin (Hou et al., 2004). This means that while the antagonism of auxin and cytokinin is conserved in lycophytes and euphyllophytes, its outcome for the respective organs is not.
A shoot-like origin for euphyllophyte roots?
Fossil records show that early vascular plant roots are strikingly similar to shoots. Roots were therefore predicted to have evolved through dichotomous branching from shoots (Graham et al., 2000;Jiang & Feldman, 2005). If we now recall that moss gametophytes and sporophytes share developmental regulators (Menand et al., 2007;Bennett et al., 2014;Viaene et al., 2014), it seems plausible that for the development of the basal fernsporophytic body, regulators also found in moss gametophore development were recruited. This process would have then occurred in shoot-like structures governing a shoot-first morphology. It has been hypothesized that the RAM arose from the SAM (Jiang & Feldman, 2005). Yet, plant development is often governed by the same regulators, which can act in very different tissues, leading to different outcomes (Prigge et al., 2005). We detected contigs with homologies to Arabidopsis genes that could be characteristic for either root or shoot. Whether these indicate a more root-like or shoot-like regulation of the Azolla RAC regulon cannot be stated, not least because of the similarity in RAM and SAM regulation (Stahl & Simon, 2010). Only genetic studies will be able to elucidate the detailed framework that regulates fern RAMs. The observed phytohormone effect, however, provides intriguing implications for the evolution of root identity. Cytokinin is known as a growth inductor of shoot branches, while auxin inhibits this process (M€ uller & Leyser, 2011). Both shoot branches and adventitious roots arise laterally from the shoot during postembryonic development. If we entertain the idea of a shoot-first origin of roots, there are two lessons to be learned from Azolla root development with regard to the basal sporophyte root system in euphyllophytes: their origin is adventitious and postembryonic, and they branch off as an altered shoot supported by cytokinin. Whether these adventitious roots are homologous to the embryonic primary roots of seed plants remains to be proven. Yet, it is clear that auxin's role in the dominant organs of the Arabidopsis root system does not apply in the same way to the fern Azolla.
Our data indicate that it might be cytokinin instead. The key will be to determine the identity of the transiently present RAM of the fern embryo. It further raises the question at what point during euphyllophyte evolution the first permanent embryonic roots emerged, and whether this process was the advent of auxin's dominance during (lateral) root developmental regulation.

Supporting Information
Additional supporting information may be found in the online version of this article.         Please note: Wiley Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.
New Phytologist is an electronic (online-only) journal owned by the New Phytologist Trust, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews. Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication 'as ready' via Early View -our average time to decision is <27 days. There are no page or colour charges and a PDF version will be provided for each article.
The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for table of contents email alerts.