Arabidopsis ribosomal RNA processing meerling mutants exhibit suspensor-derived polyembryony due to direct reprogramming of the suspensor

Abstract Embryo development in Arabidopsis (Arabidopsis thaliana) starts off with an asymmetric division of the zygote to generate the precursors of the embryo proper and the supporting extraembryonic suspensor. The suspensor degenerates as the development of the embryo proper proceeds beyond the heart stage. Until the globular stage, the suspensor maintains embryonic potential and can form embryos in the absence of the developing embryo proper. We report a mutant called meerling-1 (mrl-1), which shows a high penetrance of suspensor-derived polyembryony due to delayed development of the embryo proper. Eventually, embryos from both apical and suspensor lineages successfully develop into normal plants and complete their life cycle. We identified the causal mutation as a genomic rearrangement altering the promoter of the Arabidopsis U3 SMALL NUCLEOLAR RNA-ASSOCIATED PROTEIN 18 (UTP18) homolog that encodes a nucleolar-localized WD40-repeat protein involved in processing 18S preribosomal RNA. Accordingly, root-specific knockout of UTP18 caused growth arrest and accumulation of unprocessed 18S pre-rRNA. We generated the mrl-2 loss-of-function mutant and observed asynchronous megagametophyte development causing embryo sac abortion. Together, our results indicate that promoter rearrangement decreased UTP18 protein abundance during early stage embryo proper development, triggering suspensor-derived embryogenesis. Our data support the existence of noncell autonomous signaling from the embryo proper to prevent direct reprogramming of the suspensor toward embryonic fate.


Introduction
In sexually reproducing land plants, a seed normally contains a single mature embryo originating from a single fertilized egg cell, the zygote.Nevertheless, certain plant species can produce two or more embryos from one seed, termed polyembryony (Lakshmanan and Ambegaokar 1984).Polyembryos can emerge from nonzygotic (maternal) tissue without fertilization, a process called apomixis, or by cleavage from sexual-derived zygotic tissue (Lakshmanan and Ambegaokar 1984).Apomixis has attracted much attention since it allows the maintenance of commercially valuable traits that segregate during sexual reproduction of hybrid plants (Marimuthu et al. 2011;Kelliher et al. 2017;Wang et al. 2019;Zhong et al. 2019).Natural polyembryony by apomixis in citrus is a well-known example and was reported to be first described by Van Leeuwenhoek in 1719 (Lakshmanan and Ambegaokar 1984).Natural zygotic polyembryony is common in gymnosperms and occurs during early embryo development.For example, in Pinus species the initial proembryo divides to give rise to four embryos.Only one of the embryos develops further into a mature embryo, while the others are degraded by programmed cell death (Filonova et al. 2002;Merino et al. 2016).Other species show zygotic polyembryony derived from the usually quiescent suspensor that connects the embryo to the maternal tissues (Lakshmanan and Ambegaokar 1984).However, little is known about the genetic basis of zygotic polyembryony in plants and the molecular mechanisms are largely unexplored.
Whereas a universal cell division pattern during plant embryo development does not exist, the model plant Arabidopsis (Arabidopsis thaliana) presents an almost invariable sequence of early embryonic cell cleavages and therefore lends itself well for studies on embryonic patterning (West and Harada 1993;Jürgens and Mayer 1994).During Arabidopsis zygotic embryogenesis, an asymmetric division of the zygote forms a smaller apical cell and large basal cell.The apical cell is the major founder of the embryo proper and undergoes divisions to form the majority of tissues in the mature embryo and germinated seedling.The basal cell undergoes limited cell division and in large part differentiates into the suspensor.The embryo is connected to maternal tissue by the suspensor for structural support and nutrient exchange (Kawashima and Goldberg 2010;Peng and Sun 2018).The topmost suspensor cell generates the hypophyseal cell that becomes part of the embryonic and seedling root (Jürgens and Mayer 1994;Scheres et al. 1994;ten Hove et al. 2015).The remaining suspensor cells undergo programmed death at the heart stage of embryonic development (Bozhkov et al. 2005).
In Arabidopsis, the rule "one seed-one embryo" is executed well.However, the discovery of the twin1 (twn1) mutant, which displayed polyembryony with a 9% penetrance, showed that this rule can be broken (Vernon and Meinke 1994).Interestingly, the supernumerary embryos developed from the suspensor.This led to the hypothesis, developed in the 1990s, that the suspensor has a potential to form an embryo, but this potential is suppressed by the developing embryo proper (Schwartz et al. 1994;Vernon and Meinke 1994).Since then, this hypothesis has been supported by evidence from several studies.For example, disturbing the development of the embryo proper by genetic ablation can induce cell proliferation in suspensor cells, resulting in at least partial switching to embryonic fate (Weijers et al. 2003).More direct evidence came from laser ablation studies whereby damage of the embryo proper triggered suspensor cells to develop into an embryo (Gooh et al. 2015;Liu et al. 2015).Also, twin embryos were induced by reducing auxin activity through ectopic expression of a stabilized version of INDOLE-3-ACETIC ACID INDUCIBLE 12/BODENLOS (IAA12/BDL) (Rademacher et al. 2012;Radoeva et al. 2020).Additional mutants showing suspensor-derived polyembryony have been reported that may provide further insight into the genetics behind the communication mechanism between embryo proper and suspensor.In suspensor (sus) and raspberry mutants, the suspensor-derived embryo-like structures undergo limited cell division but are inviable (Schwartz et al. 1994;Yadegari et al. 1994).Hypomorphic allele combinations of iyo, mutated in a positive regulator of transcriptional elongation that is essential for differentiation onset, can develop ectopic embryos from the suspensor that successfully germinate.However, the resulting plants cannot complete their life cycle to produce seeds (Sanmartín et al. 2011).Finally, the twn2 mutation, a T-DNA insertion in the 5′ flanking region of a valyl-tRNA synthetase gene, caused early proembryo cell division arrest accompanied by suspensor produced twin embryos (Zhang and Somerville 1997).Shared between all these mutants producing suspensor-derived embryos is the arrest of the embryo proper.The only exception is twn1, in which both proembryo and suspensor can develop into normal healthy plants (Vernon and Meinke 1994), but for which the causal gene remains unclear.
Here we identify and characterize a recessive hypomorphic Arabidopsis mutant, meerling-1 (mrl-1), with high penetrance polyembryony.Supernumerary embryos originate from the suspensor and together with the proembryo can grow out to form healthy adult plants.We identified the causal gene, which encodes a conserved WD40-domain U3 small nucleolar RNA-associated protein 18 (UTP18) homolog, and revealed its function in the cleavage of the 18S preribosomal RNA.A UTP18 knock-out allele (mrl-2) showed arrested female gametophyte development which resulted in ovule abortion.Complementation analysis indicated that UTP18 functions noncell autonomously to maintain suspensor identity.Together our results support the hypothesis that the embryo proper inhibits the developmental potential of suspensor cells, not as a response to physical perturbation but as a monitoring system for developmental progression.

Results
Following Agrobacterium tumefaciens (Agrobacterium)mediated transformation of the pCB1 construct, carrying norflurazon and phosphinothricin resistance cassettes (Supplementary Fig. S1A) into the transgenic Arabidopsis ecotype Nossen (No-0) carrying HSP18.2 pro :CRE (HCN) (Sieburth et al. 1998;Heidstra et al. 2004), we observed that multiple seedlings germinated from single seeds in the progeny of line T2-10 (Fig. 1A).We next performed a series of crosses and selfings with these polyembryo-derived plants and found that the phenotype was stably inherited as a recessive trait.Since we observed more than two seedlings germinating from a single seed (Fig. 1A), we named this mutant meerling (mrl-1) ("multiple birth").However, the phenotype could not be linked to the presence of the pCB1 T-DNA based on the observation of norflurazon sensitive seedlings displaying polyembryony (Supplementary Fig. S1B) and the absence of T-DNA hybridizing fragments in the DNA of their progeny tested by DNA gel blot analysis (Supplementary Fig. S1, C and D).

mrl-1, a polyembryonic mutant
To characterize the polyembryo phenotype, we germinated seeds of five homozygous T3 lines (Supplementary Data Set 1) and observed a penetrance of 33% to 53% for polyembryonic seeds and 13% to 26% for tricotyledon seedlings.The remainder germinated as single WT-looking seedlings.To determine the stability of this phenotype, we carefully reexamined one of these lines after several generations of inbreeding (Supplementary Data Set 1).We germinated 228 seeds of inbred mrl-1 line and observed almost half of the seeds (105/228) producing multiple seedlings (Fig. 1, B and C).Of these polyembryonic seeds, less than half (45/105) produced twins and the remainder (60/105) produced triplets.
The majority of triplets (53/60) comprised one complete seedling with the other two seedlings fused together in the junction of the cotyledons (Fig. 1C), whereas a few triplets were separated upon germination (7/60).Together, these results showed that the penetrance of the mrl-1 polyembryony phenotype is around 46%. Apart from the polyembryo phenotype, about 20% of the seedlings showed polycotyledony, with the majority displaying three cotyledons (45/228) (Fig. 1D).Throughout these experiments we observed two seedlings germinating with four cotyledon-like structures and one with double shoots supported by a single root (Supplementary Fig. S2, A and B).The remaining seeds germinated as wild-type-like seedlings (76/228) (Fig. 1E).Notably, all mrl-1 seedlings developed into normal wild-type-looking plants and successfully completed their life cycle with the production of seeds (Supplementary Fig. S2C).
To investigate where these supernumerary seedlings originate, we compared late embryo stages in mrl-1 and WT.In WT, the heart shaped embryo is connected by a single file of suspensor cells to the maternal tissue (Fig. 1F).In mrl-1, it appears that embryos can develop from suspensors in addition to the primary embryo (Fig. 1G).This phenotype reminded us of the polyembryonic twn1 mutant, where ectopic embryos also develop from the suspensor (Vernon and Meinke 1994).We therefore crossed both mutants to test for allelism and observed that all F1 seeds germinated single seedlings, indicating twn1 and mrl-1 represent mutations in different genes (Supplementary Fig. S2D).

Development of extra-embryonically derived mrl-1 embryos
To investigate the origin and development of secondary embryos, we followed their formation from fertilization onward.First, we compared early embryo development in time in WT and mrl-1 following manual pollination to ensure synchronous fertilization.At 48 hap (hours after pollination), WT embryos were in the dermatogen stage whereas early globular stage embryos occurred at 72 hap (Supplementary Fig. S2, E and G).However, mrl-1 embryos were in the 2-or 4-cell stage at 48 hap and developed toward the octant stage at 72 hap (Supplementary Fig. S2, F and H).These results indicate a delay in early embryo development in mrl-1 compared to WT.
Next, we followed embryo development and counted suspensor cell number in relation to silique development after normal self-fertilization.We considered the stage 15 flower as silique 1 (sl1), and subsequently the older flower/silique was referred to as sl2 and so on (Fig. 2H) (Smyth et al. 1990).We mainly observed WT dermatogen stage embryos in sl4.In contrast, mrl-1 embryo development was delayed and mostly 2-to 4-cell stage embryos occupied sl4 (Fig. 2, A, I, and P; Supplementary Data Set 2).Similarly, while WT sl5 already contained early globular stage embryos, most mrl-1 embryos were at the octant stage in sl5 (Fig. 2, B, J, and P; Supplementary Data Set 2).Our results were consistent with the manual pollination experiment.Therefore, we continued tracing embryo stages and determined suspensor number in subsequent self-fertilized siliques to illustrate differences in embryo development between WT and mrl-1.
Embryo development in WT continued from the dermatogen stage at sl4 toward the torpedo stage within seven consecutive siliques (sl4 to sl10) (Fig. 2, A to G, P; Supplementary Data Set 2).However, mrl-1 embryo development slowed down toward the globular stage, after which development paused, only to continue much later (Fig. 2, I to O, P; Supplementary Data Set S2).For instance, most WT embryos within sl7 had reached the early heart stage whereas mrl-1 embryos remained mainly at the globular stage until at least sl17 (Fig. 2P; Supplementary Data Set 2).Nevertheless, fertilized ovules successfully developed mature embryos indicating embryo development catches up during later silique developmental stages.
In WT, the suspensor originates from the large basal cell after the first asymmetric division of the zygote.This basal cell undergoes several transverse divisions to form a mature suspensor consisting of 7 to 12 cells (Fig. 2, A to D, Q) (Yeung and Meinke 1993).However, the mrl-1 suspensor underwent more transverse divisions (Fig. 2I-N, Q, Supplementary Data Set S3).In addition, we observed longitudinal suspensor cell divisions from sl9 onwards (Fig. 2N), which was never observed in WT.At later stages, additional ectopic suspensor divisions occurred that were the prelude of future embryo-like structures (Fig. 2O).
To investigate whether and when the ectopically dividing suspensor cells in mrl-1 assumed embryo identity, we checked DÖRNROSCHEN (DRN) reporter expression.We transformed mrl-1 with a binary construct harboring DRN pro >> erGFP and crossed a selected T3 homozygous transgenic line with WT to compare expression patterns.In the F1 WT embryo, the DRN reporter is specifically expressed in the embryo proper in agreement with its reported expression pattern (Fig. 2V) (Cole et al. 2009).Prior to the globular stage the DRN reporter is only expressed in the embryo proper in mrl-1, suggesting that the suspensor identity is initially specified in mrl-1 (Fig. 2W).However, in addition to embryo proper expression, we observed ectopic expression in the transverse dividing suspensor cells around the globular stage (Fig. 2, X and Y).In fact, we could already observe suspensor expression of DRN before the suspensor underwent longitudinal cell divisions (Fig. 2X).Next, we examined the integrity of suspensor fate using M0171, an enhancer trap line specifically expressing erGFP in the suspensor of octant-until heart-stage embryos and which later becomes expressed in the cotyledon junction (Fig. 2Z, AA, DD) (Radoeva et al. 2016).However, in mrl-1 embryos, M0171 expression was absent from the suspensor (n > 20), and only the later expression in the cotyledon junction remained (Fig. 2BB, CC, EE).In addition, we tested the ATPase pro suspensor marker that was reported to be expressed from at least the 4-cell stage onwards (Radoeva et al. 2020).We confirmed its expression in the wild-type and mrl-1 suspensor (Fig. 2, R to U).The correct setup of the DRN marker expression together with the ATPase pro expression pattern indicate that suspensor fate is initially specified and becomes (partially) lost during early embryogenesis after which ectopic expression of embryo proper fate is followed by suspensor division.
Reprogramming of the suspensor could also be achieved upon inhibition of auxin response in these cells, similarly resulting in aberrant divisions and embryo-specific gene expression (Rademacher et al. 2012).Therefore, we compared auxin response in WT and mrl-1 embryos, visualized by the DR5 pro :nlsVENUS reporter (Heisler et al. 2005) (Supplementary Fig. S3, A to F).In WT, DR5 expression shifts from the apical lineage to the hypophysis and uppermost suspensor cells around the young globular stage and, following hypophyseal cell division, to the progenitors of QC and columella stem cells (Friml et al. 2003) (Supplementary Fig. S3, A to C).We could not detect DR5 reporter expression in the mrl-1 embryo proper before the globular stage (Supplementary Fig. S3D).Globular and later stage mrl-1 embryo proper displayed much reduced DR5 levels, corresponding with delayed development of the embryo proper.Instead, high DR5 expression was observed in the ectopic transversely divided suspensor cells in mrl-1 embryos (Supplementary Fig. S3, E and F), cells that later develop embryo structures.This indicates that auxin response in the suspensor is associated with a cell fate transition in mrl-1, and suggests a different mechanism of reprogramming compared to that observed upon auxin response inhibition.
Together, these results reveal that mrl-1 is a polyembryonic mutant in which both embryo proper and suspensor-derived embryos develop.

Molecular identification of the MRL gene
As an initial step to identify the causal mutation we roughly mapped the region conferring the mrl-1 phenotype on the Arabidopsis genome.We generated a mapping population by crossing the homozygous mrl-1 mutant in the No-0 background with Arabidopsis accession Landsberg erecta (Ler).F2 plants originating from polyembryonic seeds were genotyped using primers based on insertion-deletion polymorphisms (InDels) (Jander et al. 2002).This allowed us to map the mrl-1 mutation to the short arm of chromosome 5, between markers MXM12-Del15 (2.5 Mb) and Ciw8 (7.5 Mb).We then went ahead to perform whole genome sequencing of the T-DNA free mrl-1 mutant and its parental HCN line.We identified an approximately 2 Mb duplicated genomic DNA region of chromosome 1 that was reversely inserted into the first exon (−326 bp) of AT5g14050 (Fig. 3A, Supplementary File 1 and S2), which encodes UTP18.The promoter and part of the 5′-UTR of the AT1G01830 gene, encoding an Armadillo repeat superfamily protein, followed by a 44 bp unidentified sequence, was inserted in the 5′-UTR of AT5G14050, 326 bp upstream of its start codon (Fig. 3A).However, this 2 Mb insertion was also located at −246 bp from the start codon of the AT5g14040 gene encoding a MITOCHONDRIAL PHOSPHATE TRANSPORTER 3 (MPT3) in the opposite direction (Fig. 3A).
To determine which of these two genes is causal for the polyembryo phenotype, constructs harboring these genes were transformed into the mrl-1 mutant and the transgenic progeny was tested for complementation of the polyembryo phenotype.One construct contained AT5G14050 expressed from the 3,218 bp upstream region, named the UTP18 pro long promoter, and expressed both genes.The other construct only contained AT5G14050 expressed from the short 562 bp promoter, hence separating both genes (UTP18 pro short, Fig. 3A).Both constructs completely restored the mrl-1 phenotypes to WT (Fig. 3, B to E), indicating that AT5G14050 represents the causal gene for mrl-1.The AT5G14050 encoded UTP18 protein consists of 537 amino acids and contains four WD40 repeat domains (Supplementary Fig. S4A).Phylogenetic analysis based on amino acid alignment revealed that no UTP18 paralogs exist in the Arabidopsis genome, indicating that UTP18 represents a single-copy gene (Supplementary Fig. S4B).AlphaFold prediction of UTP18 homologs in three model species, Arabidopsis, rice (Oryza sativa), Saccharomyces cerevisiae and in Homo sapiens indicate they have a similar scaffold-like WD40 repeat structure suggesting a conserved function (Supplementary Fig. S4, C to F). UTP18 protein homologs in different species such as S. cerevisiae, Homo sapiens, and Drosophila melanogaster indicate its function as a UTP18 protein involved in nucleolar processing of pre-18S ribosomal RNA (Dragon et al. 2002;Bernstein et al. 2004;Fichelson et al. 2009).UTP18 is a conserved component associated with the U3 small nucleolar ribonucleoprotein (snoRNP) complex which is involved in cleavage of the primary precursor 18S rRNA at the P-site to cut off the 5′ external transcript spacer (ETS) (Fig. 6H) (Venema and Tollervey 1999;Sáez-Vasquez et al. 2004;Sáez-Vásquez and Delseny 2019).Loss-of-function utp18 alleles are unviable in Saccharomyces and Drosophila (Vandenbol and Portetelle 1999;Fichelson et al. 2009) which indicates a crucial function for the encoded proteins in development.

UTP18 localizes to the nucleolus and is ubiquitously expressed
To determine in which tissues the UTP18 gene is expressed, we performed RT-PCR with UTP18-specific primers, which resulted in the amplification of a single fragment with expected size from cDNA of roots, stems, cauline leaves, inflorescences, and siliques (Fig. 4A).Public transcriptome data also indicate ubiquitous presence of UTP18 mRNA albeit with higher levels in dividing tissues (Supplementary Fig. S5A).We generated reporter lines by transcriptionally fusing the 3,218 bp long UTP18 promoter region to β-Glucuronidase (GUS), creating UTP18 pro long:GUS, and similarly observed expression in all tissues of the plant (Fig. 4, B to F). Strongest GUS staining was observed in actively proliferating tissues such as leaf tip, root tip, and lateral root primordia (Fig. 4, B to F), consistent with a protein function in regions of high demand for ribosome biogenesis.
UTP18 is reported to process the pre-18S rRNA in the nucleolus of yeast cells (Dragon et al. 2002).To determine the subcellular localization of UTP18 in Arabidopsis, we generated a construct expressing a GFP-UTP18 protein fusion under control of the strong and constitutive Cauliflower Mosaic Virus 35S promoter.This construct was transiently expressed in Nicotiana benthamiana leaves by Agrobacterium-mediated infiltration, and protein accumulation was examined by confocal scanning laser microscopy.The GFP-UTP18 fusion protein localized to the nucleus with the highest concentration in the nucleolus (Fig. 4, G to I).To confirm the nucleolar accumulation in Arabidopsis, we transformed mrl-1 plants with a construct harboring UTP18 pro long:GFP-UTP18, and that was able to complement the polyembryo phenotype (Fig. 3F).We observed the fluorescent GFP signal predominantly in the nucleolus (Fig. 4, J to L).

UTP18 is essential for female gametophyte development
To investigate the incomplete penetrance of the polyembryo phenotype in mrl-1 and the corresponding mutation in the 5′-UTR of the UTP18 gene, we generated additional alleles in the coding sequence using CRISPR/Cas9 in the Arabidopsis ecotype Col-0 background.Two alleles were pursued: one with a 14 bp deletion 379 bp downstream of the start codon thereby introducing a premature stop codon (mrl-2, Fig. 3A), and one with a 12 bp homozygous deletion 380 bp downstream of the start codon (mrl-3).The mrl-3 allele did not display an obvious phenotype indicating that a four amino acid deletion at this position in the protein can sustain its function.For the mrl-2 allele we could not recover homozygous mutant offspring (n > 100), which corresponded well to approximately one-third of ovules being aborted in siliques of heterozygous mrl-2 +/− plants (Fig. 4, M and N; Supplementary Data Set 4).
To determine the origin of the observed homozygous lethality we performed reciprocal crosses between mrl-2 +/− and Col-0 plants.Fertilization of mrl-2 +/− flowers with Col-0 pollen resulted in WT F1 progeny only, suggesting that the mutation influences female gametophyte development thereby preventing fertilization.Vice versa, fertilizing Col-0 flowers with pollen of mrl-2 +/− plants and genotyping of eight F1 plants gave only one heterozygous plant.This skewed segregation ratio suggested that the mrl-2 mutation also influences the male gametophyte.To assess pollen vitality, we compared pollen from stamens of two heterozygous mrl-2 +/− plants with those from WT using a simplified Alexander's staining method (Peterson et al. 2010).This experiment showed that mrl-2 +/− plants produced pollen that all appear normal (Fig. 4O).

Altered embryonic UTP18 expression causes the polyembryo phenotype
Given that the homozygous mrl-2 mutant cannot be obtained, the polyembryo phenotype in mrl-1 may be due to altered gene expression because of the 2 Mb DNA insertion whereby the promoter and part of the 5′-UTR of the AT1G01830 gene now controls UTP18 expression (Fig. 3A).To investigate the role of the mutant promoter in conferring the polyembryo phenotype, we first compared public expression data available for AT1G01830 and UTP18 (AT5G14050).It appears that AT1G01830 is generally expressed at lower levels during Arabidopsis development, particularly during embryogenesis (Supplementary Fig. S5,  A and B).
To confirm that altered regulation of UTP18 is responsible for the polyembryo phenotype, we first transformed heterozygous mrl-2 +/− plants with the WT UTP18 coding sequence under control of the 2365bp mrl-1 mutant promoter (mrl-1 pro :UTP18) and genotyped for homozygous mrl-2 mutant progeny in the next generation.We identified two double homozygous mrl-2; mrl-1 pro :MRL T3 lines with polyembryo and polycotyledon phenotypes similar to mrl-1 (Fig. 3G; Supplementary Fig. S2, I to K).Interestingly, the penetrance of polyembryonic seeds (63% and 54%) was higher in the double homozygous lines than in mrl-1 (45%) (Supplementary Data Set 1).
We then compared activities of the wild-type UTP18 pro long and mutant mrl-1 pro promoters driving a 3xVENUSnls in three independent transgenic lines generated for each construct in the Col-0 background.In early embryonic stages, we observed much lower VENUS fluorescence in both suspensor and embryo proper for the mrl-1 pro :3xVENUSnls construct compared to that of the wild-type promoter construct UTP18 pro long:3xVENUSnls, using the same confocal settings (Supplementary Fig. S7, A and C).By the heart stage of embryogenesis, the VENUS signal could be detected at around the same level for both promoter constructs, albeit with a different, but consistent, pattern of expression depending on the genetic background (Supplementary Fig. S7, B and D).
Next, we aimed to compare the protein abundance of UTP18 during WT and polyembryo development.Using the same strategy as above we generated transgenic lines in the homozygous mrl-2 background expressing GFP-UTP18 from the UTP18 pro long promoter and a slightly truncated mrl-1 pro (1,964 bp) promoter.The introduction of the UTP18 pro long: GFP-MRL construct fully complemented the seed abortion phenotype of mrl-2 (compare Fig. 4N and Supplementary Fig. S8A).Correspondingly, GFP signal was observed in all embryonic cells at early embryonic stages (Fig. 5, A and B).To examine UTP18 abundance during polyembryony we used the reconstituted mrl-1 phenotype displayed by the mrl-2; mrl-1 pro (1964):GFP-UTP18 genotype.We pursued a T1 line for which ovules displayed asynchronous development within a silique (Supplementary Fig. S8B).These delayed ovules resemble the delay in ovule development displayed by mrl-1 when compared to the No-0 WT background in subsequent siliques (Supplementary Fig. S8, D to K).Indeed, we found suspensor-derived embryo structures formed in these delayed ovules (Fig. 5E), and additional seedlings germinated from single seeds (Supplementary Fig. S8C).Based on this, we considered the GFP signal to represent the true UTP18 protein abundance in the mrl-1 polyembryonic mutant.Interestingly, GFP-UTP18 was absent from the embryo proper during early embryo stages (Fig. 5, C and D).This altered expression during the early embryonic stages corresponds with the observed ectopic divisions in the suspensor in later stages (Fig. 5D) that served as a prelude to suspensor-derived embryogenesis.In these later embryo stages, both embryo proper and suspensor (which includes suspensor-derived embryos) expressed GFP-UTP18 protein (Fig. 5E).
Together these results indicate that the abundance of UTP18 is altered predominantly during early embryogenesis leading to halted embryo proper development and ectopic suspensor-derived embryo formation.

UTP18 functions noncell autonomously from the embryo proper to maintain suspensor identity
The embryonic transformation of the suspensor, as observed in twn1 mutants, was proposed to be due to loss of communication between embryo proper and suspensor (Vernon and Meinke 1994;Schwartz et al. 1997).Here we observe very similar phenotypes for mrl-1, and our expression and complementation experiments above suggest an important role for sufficient and local expression of UTP18 during development.Therefore, we tested where UTP18 expression is limiting for maintaining embryo homeostasis by complementation experiments utilizing specific promoters to drive UTP18 in the mrl-1 background.We used the DRN promoter to drive UTP18 expression in the apical cell lineage after asymmetric division of the mrl-1 zygote.DRN reporter expression was limited to the embryo proper in mrl-1 early embryo development before its ectopic expression from the globular stage onward (Fig. 2, W to Y), allowing its use in complementation experiments.Next, we introduced DRN pro : UTP18 into the mrl-1 mutant and observed in 8 out of 10 T2 lines a substantial reduced penetrance of polyembryo and polycotyledon phenotypes (Fig. 5F; Supplementary Data Set 5).Of these, we generated two lines homozygous for the transgene (T3-1-1 and T3-2-1) that showed 100% complementation (Supplementary Fig. S9A).
To drive UTP18 expression in the suspensor we used the ATPase promoter (Radoeva et al. 2020), that is active in the mrl-1 suspensor (Fig. 2, T and U).To test whether UTP18 expression in the suspensor can rescue the polyembryo phenotype of mrl-1, we generated plants with the ATPase pro >> UTP18 or ATPase pro :UTP18 transgene.None of the T2 transgenic lines nor a next generation homozygous T3 line showed reduced penetrance of polyembryony or polycotyledony (Fig. 5F; Supplementary Data Set 5, Supplementary Fig. S9B).
Taken together, our findings suggest that sufficient levels of UTP18 expression are required in and during early embryo proper development to nonautonomously prevent the formation of suspensor derived embryos that are causal to the polyembryo phenotype of mrl-1.

UTP18 is involved in processing of pre-18S rRNA
In plants, the 45S rDNA encodes the 18S, 5.8S, and 25S rRNAs on a single transcription unit, and exists in hundreds of copies in the genome.The resulting primary transcripts, the pre-rRNAs, harbor the individual rRNAs.These are separated by internal transcript spacers (ITS) and the outer boundaries are determined by the ETS.The spacers are subsequently removed in a complex maturation process inside the nucleolus through transient interaction with a large so-called small subunit processome or U3 snoRNP complex that also mediates the early endonucleolytic cut in the 5′-ETS upstream from the 18S rRNA (Fig. 6H) (reviewed in Tomecki et al. 2017; Sáez-Vásquez and Delseny 2019).To investigate whether the molecular function of UTP18 is consistent with being the Arabidopsis UTP18 homolog and thereby part of this U3 snoRNP complex, we chose a conditional CRISPR approach because of the lethal mrl-2 knock-out phenotype.We generated two constructs, one for induced constitutive and one for tissue-specific CRISPR/Cas9, that were introduced into Arabidopsis Col-0.
To test the efficacy of the constitutive CRISPR/Cas9induced mutations and the associated post-embryonic phenotypes, we germinated transgenic seeds or transferred seedlings of two inducible T2 lines (T2-4 and T2-5) onto medium containing 10 μM β-estradiol and compared these to a DMSO mock control.Seedlings stop growing upon Cas9 induction as visualized by loss of meristem activity 4 days postinduction (dpi), in contrast to normal-growing mock-treated control roots (Fig. 6, A to D). UTP18 expression was decreased in seedlings germinated on Cas9 induction media compared to in control seedlings (Fig. 6E), possibly due to the loss of dividing tissues that require high UTP18 levels.For the lateral root specific CRISPR/zCas9i T2 seedlings, we observed an absence of emerging lateral roots compared to the multiple lateral roots emerging from WT primary roots of the same age (Fig. 6, F and G).These results indicate that the conditional CRISPR/zCas9i targeting of UTP18 is effective, and that UTP18 is essential for post-embryonic (root) growth.
Previously, SLOW WALKER 1 (SWA1), a UTP15 homolog and one of the components of the U3 snoRNP, was shown to be involved in the endonucleolytic cleavage at the P-site of the 5′-ETS from the pre-18S rRNA.This was demonstrated using callus derived from roots (Shi et al. 2005).To investigate a similar function for Arabidopsis UTP18 as a yeast UTP18 homolog, we used the conditional mutant strategy to induce loss of gene function and subsequently test pre-rRNA cleavage in the affected tissues.Therefore, we generated callus from 9-d old transgenic roots of the same inducible constitutive CRISPR/Cas9 T2 lines used above.After culturing the roots on callus-inducing medium for 2 d, Cas9 was induced for the final 9 d with 10 μM β-estradiol or with DMSO as mock treatment.Total RNA was isolated from these 11-d-old calli and RT-qPCR was performed with primers P1 and P2 that flank the P-site (Fig. 6H).The results show that the amount of unprocessed pre-18S rRNA increased almost 2-fold and 1.5-fold in callus derived from CRISPR/ Cas9-induced lines T2-5 and T2-4, respectively, compared to the control (Fig. 6I).We also tested pre-18S rRNA cleavage in 14-d-old T2-4 and T2-5 seedling roots that were induced for 9 d upon transfer to 10 μM β-estradiol or DMSO mock medium (Supplementary Fig. S10, A to D). RT-PCR was performed using root total RNA with primers U1 and U2 as described in Shi et al. (2005).The results similarly show that the amount of uncleaved pre-18S rRNA increased in CRISPR/Cas9-induced roots compared to control roots (Fig. 6J).Taken together, these data indicate that the UTP18 protein is required for the primary cleavage of P-site during the processing of the pre-18S rRNA, consistent with its annotation as a homolog of yeast UTP18.

Discussion
In this study, we identified mrl-1 as a polyembryonic mutant.The mrl-1 mutant displays a high frequency of seeds containing two or more embryos.All viable seedlings gave rise to fertile plants that successfully complete their life cycle with the production of seeds.The supernumerary seedlings in mrl-1 seeds originate from the suspensor, which normally degenerates at late embryogenesis.The mutant phenotype is caused by an Agrobacterium transformation-induced duplication of a 2 Mb piece of DNA from Chr1 in the 5′ untranslated region of a UTP18 homolog.The high degree of sequence identity between this Arabidopsis UTP18 and its homologs in other species, together with our experimental data indicate that it is an essential component of U3 snoRNP involved in 18S preribosomal RNA cleavage.The insertion mutation in mrl-1 generally lowered the level of expression of the UTP18 gene, especially in dividing tissues and during early embryo proper development.Our results provide detailed molecular evidence for the hypothesis that normal progression of embryo proper development is required to remain suspensor quiescence.
In addition to UTP18, several other members of the U3 snoRNP have been identified as mutants with similar developmental defects.For example, the defective female gametogenesis phenotype causing lethality in the knockout mrl-2 allele, is similar to the mutant phenotype of swa1 (Shi et al. 2005).SWA1 encodes a homolog of the yeast UTP15 within the U3 snoRNP complex.However, the seed abortion segregation ratio in the heterozygous mrl-2 +/− plant does not follow Mendelian genetics, which suggests UTP18 is also required during fertilization.The TORMOZ (TOZ) and POPCORN (PCN) genes encode homologs of the yeast UTP13 and UTP4 U3 snoRNP members, respectively (Griffith et al. 2007;Xiang et al. 2011).The toz mutant did not appear to be a null allele and it displayed only a slight effect on rRNA processing in ex-plant cultured embryos.The mutant showed aberrant embryonic cell division planes and arrested development before the globular stage.The pcn mutant showed cotyledon phenotypes, delayed embryo proper development and longitudinal suspensor divisions similar to mrl-1.Auxin distribution and response were altered in pcn, exemplified by ectopic DR5 in suspensor cells, reminiscent of mrl-1 mutant embryos, although the suspensor successfully regenerates embryo(s) in mrl-1 as opposed to its developmental arrest in pcn.Together, these mutants indicate that phenotypic severity depends on the allele and corresponding protein inside the U3 snoRNP complex, with arrest in female gametogenesis as the most severe phenotype.These effects may be attributed to defective 18S rRNA processing resulting in a general arrest of growth.However, the molecular effects of mutations in the pre-rRNA processing may be diverse, ranging from apoptosis to reduced translation to changes in the translatome (reviewed in Aubert et al. 2018).
Based on the observed phenotypes in twn mutants and the ablation studies, the prevalent hypothesis stated that the presence of the embryo proper suppresses secondary embryo development from the suspensor (Vernon and Meinke 1994;Gooh et al. 2015;Liu et al. 2015).In our mrl mutant reconstruction line, whereby we introduced a GFP-UTP18 fusion expressed from the mrl-1 pro promoter in the mrl-2 knockout that successfully reproduced the polyembryo phenotype, we observed GFP-UTP18 fluorescence only in the suspensor at early embryo stages.This contrasts with MRL expression in both embryo proper and suspensor in WT.In addition, expression of UTP18 from the DRN embryo proper promoter complemented the polyembryo phenotype whereas the suspensor (ATPase pro ) driven UTP18 did not.Together with the delayed embryo proper development, these data indicate that UTP18 protein abundance in the embryo proper falls below the level to sustain embryo proper vitality.As a consequence, the suspensor initiates secondary embryogenesis.However, upon reaching the transition stage, GFP-UTP18 was again detected in the embryo proper as well as in suspensor-derived embryo, explaining the survival of the original embryo in addition to the formation of ectopic embryos.Together, these results merge into a model on the requirement for UTP18 during early embryogenesis.
Previous laser ablation studies showed that damage to the apical cell that is formed after the asymmetric division of the zygote can induce basal cell reprogramming (Gooh et al. 2015;Liu et al. 2015).Gooh et al. observed cell fate conversion of the upper one or two suspensor cells by subsequent ectopic transverse suspensor cell division followed by loss of suspensor properties (WOX8 expression) and gain of embryonic properties (DRN expression), prior to the first longitudinal embryonic cell division (Gooh et al. 2015).In their experiments, Liu et al. observed DR5 reporter accumulation in the free end of the suspensor (topmost one or two cells) after ablation, and this cell further developed into an embryo structure.The authors hypothesized that without the embryo proper, auxin transport from the suspensor to the embryo proper was blocked in the topmost suspensor cell, leading to reprogramming of cell fate (Liu et al. 2015).Similar to the above observations, the mrl-1 single file suspensor cells first undergo transverse divisions.Suspensor cells in mrl-1 express the ATPase pro marker but failed initiation of M0171 expression at the octant stage supporting the notion of reprogramming suspensor fate toward embryonic competence.Subsequently these reprogramming suspensor cells display high DR5 accumulation at the globular stage.Importantly, the embryo proper marker DRN is expressed before longitudinal divisions leading to embryo development occur, suggesting that reprogramming is the driver for the divisions leading to embryo formation.In agreement, induction of suspensor cell division through ectopic bHLH49 expression was not sufficient to induce embryogenesis (Radoeva et al. 2019), likely due to the absence of reprogramming.
The auxin response maximum in the ectopic suspensor cells observed in the ablation studies and in the mrl-1 suspensor corresponds with suspensor-derived embryo development but may not be the trigger of the suspensor reprograming.This is consistent with the formation of twin embryos upon auxin response inhibition (Rademacher et al. 2012).Using a set of ubiquitous and suspensor drivers to express a nondegradable bodenlos (bdl) protein, excessive aberrant divisions were observed in the suspensor.However, only the constitutive RPS5A promoter-driven bdl delivered occasional true twin embryos, which corresponded with the ectopic expression of embryonic markers SHOOT MERISTEMLESS (STM) and WUSCHEL (WUS).Together with evidence on additional ARF and IAA factors in suspensor development, it was concluded that auxin response is required to maintain suspensor cell identity (Rademacher et al. 2012).Suspensor reprogramming was also observed upon ectopic expression of embryo-related transcription factors RWP-RK DOMAIN CONTAINING 1 (RKD1), RKD4, and WUS (Radoeva et al. 2020).Here, suspensor identity is lost, which is followed by ectopic suspensor cell division and subsequent DRN embryo proper marker expression.This process shared similarity to somatic embryogenesis, and may resemble a regeneration-like mechanism whereby cells first dedifferentiate and proliferate followed by re-specification and expression of embryonic genes leading to the formation of somatic embryos (Ikeda-Iwai et al. 2003;Horstman et al. 2017;Radoeva et al. 2020).Together, our results suggest a case where suspensor-derived polyembryony as observed in mrl-1 follows a reprogramming mechanism similar to that observed in the ablation studies and different to the auxin response inhibition and ectopic transcription factor studies.
Embryo ablation studies indicated that the suspensor has the potential to form an embryo up to the globular stage only (Gooh et al. 2015;Liu et al. 2015).Taking advantage of the high penetrance of the mrl-1 phenotype, we show that delayed development of the embryo proper prior to the transition stage was accompanied by suspensor-derived secondary embryo formation.These observations, combined with the presence of the ATPase pro expression and failure of M0171 marker initiation at the octant stage in the mrl-1 mutant places the timing of suspensor reprogramming before or around the octant stage of embryo development.Indeed, DRN marker expression is clearly observed in suspensor cells in globular stage embryos.
The mrl-1 and twn1 mutants may represent a distinct clade of polyembryonic mutants because suspensor-derived embryogenesis does not depend on injury to the embryo proper.Specifically, in mrl-1, embryo development is asynchronous where approximately half of the delayed transition stage embryos exhibit suspensor-derived embryogenesis while the other half continues the typical normal embryo development including suspensor.Nevertheless, embryos from wild-type plants do not develop suspensor-derived embryos at this stage in development, suggesting crucial signaling between embryo and maternal tissues and/or between embryo proper and suspensor.The synchronization between embryo proper and suspensor development appears disturbed with the delay imposed specifically on embryo proper development by the mrl-1 mutation.The molecular function of the causal mutated gene in mrl-1 suggests that any factor causing a specific delay only in embryo proper development prior to the globular stage will result in suspensor reprogramming and subsequent embryo development.Such a scenario supports active signaling between the embryo proper and suspensor.If auxin is not the cue for reprogramming the suspensor the question remains what is the role and nature of this signaling?An analogy may be drawn to the opposing (nonautonomous) signaling in the maternal and embryonic control over embryonic root development mediated by a complex interplay of WIP gene expression (Du et al. 2022).Crosstalk between these genes may represent a module to deal with local conditions and shift resources toward defense or reproduction (Du et al. 2022;Wittmer and Heidstra 2022).Interestingly, WIPs act through so-called hub proteins able to interact with various transcription factors and mutation of these hub genes alleviated the growth defects in wip mutants (Du et al. 2022).In a similar way, the mrl-1 mutant phenotype may now be applied to screen for signaling components between embryo proper and suspensor, e.g.utilizing a suppressor screening in mrl-1 background for absence of suspensor derived embryo development or by single-cell RNA sequencing.
In conclusion, our study shows that embryo propermediated inhibition of suspensor reprogramming ensures the general one seed-one embryo rule.Our results suggest an active signaling of developmental progression between embryo proper and suspensor to sustain the development of a single dominant embryo.Because of its high penetrance of polyembryony, mrl mutants provide an excellent tool for unraveling communication mechanisms between the embryo proper and suspensor.

Plant material and growth conditions
Arabidopsis (A. thaliana) seeds were gas sterilized for 2 h as described in Lindsey et al. (2017), resuspended in sterile 0.1% (w/v) agarose and stratified at 4 °C for at least 2 d.Seeds were subsequently plated on 0.5× Murashige and Skoog medium (MS including vitamins), supplemented with 0.8% (w/v) plant agar, 1% (w/v) sucrose, and 0.5 g/L MES monohydrate pH5.8 (all from Duchefa Biochemie).Plates were positioned near vertical and seedlings were grown at 22 °C with 16 h light and 8 h dark cycle with a light intensity of ∼85 µmol m −2 s −1 from fluorescent lighting.Seedlings were transferred to soil and grown in a growth chamber under the same 22 °C and long-day conditions but with a light intensity of 120 to 140 µmol m −2 s −1 from LED lights (Lumeco).

Map-based cloning
Seedlings displaying the mrl-1 phenotype in the F2 population resulting from a cross between mrl-1 and Landsberg erecta (Ler) were used as the mapping population.Using primers based on InDels between Col 0 and Ler (Jander et al. 2002), we were able to roughly map the mutation in between markers MXM12-Del15 (MXM12-Del15F: 5′-GCCAATTTC AACAACGAAGG, MXM12-Del15R: 5′-ATTCGCCGTCGGAA TTATCT) and ciw8 (CIW8-F: 5′-TAGTGAAACCTTTCTC AGAT; CIW8-R: 5′-TTATGTTTTCTTCAATCAGTT) in a population of 28 plants selected for the mrl-1 phenotype out of a total of 60 F2 progeny.Subsequent fine mapping using additional mrl-1 type seedlings segregating from a total of 800 F2 progeny did not result in getting closer to the mutation.DNA from mrl-1, Ler, and the F2 population seedlings was isolated from leaves using the CTAB method (Lukowitz et al. 1996).PCR was carried out in a total volume of 20 μL, containing 0.5 μL of 5U/μL home-made Taq polymerase, 0.5 μL of 10 mmol l −1 forward and reverse primers, 2.5 μL of 10× PCR buffer, 0.5 µL of 250 ng/µL DNA, and 16 µL of Milli Q water.All PCR products were detected on a 1% (w/ v) agarose gel using electrophoresis.
Whole genome sequencing with paired-end reads was performed on genomic DNA isolated from the T-DNA free mrl-1 mutant and its parental HCN line on an Illumina HiSeq2000 sequencing machine (BioProject ID PRJNA1009023).Reads were mapped on the TAIR10 reference genome, with the pCB1 T-DNA and HCN sequences added as an additional sequence, using bwa mem (bwa Version: 0.7.12-r1039), and a BAM alignment file was produced with samtools Version: 0.1.19-44428cd.The BAM file was realigned using GATK, and GATK was used to call SNPs and indels.For structural variant analysis, the tool Delly was used (Rausch et al. 2012).Since neither analysis delivered leads as to the identity of the mutation, we resorted to visual inspection of the isolated genomic region in between the markers used for mapping using the Integrative Genome Browser (IGV) (Robinson et al. 2011) to identify the mrl-1 mutation.

Construct design, cloning, and plant transformation
All plasmids were constructed by Golden Gate cloning using the MoClo Toolkit and Plant parts (Addgene Kits #1000000044 and #1000000047) (Engler et al. 2014) or gateway cloning (Invitrogen) unless otherwise specified (Supplementary Data Sets 6 and 7).Transgenic plants were generated by Agrobacterium-mediated transformation by means of the floral dip method (Clough and Bent 1998).
To generate DRN pro >> erGFP and ATPase pro >> erGFP reporter lines, we amplified promoter sequences up to 4,771 bp for DRN and 982 bp for ATPase, all located upstream of their respective ATG start codons.For the DRN reporter, the pNOS-BAR selection cassette, DRN pro :GAL4VP16, and UAS pro :erGFP were then assembled into the destination vector pICSL4723 using BpiI Golden Gate assembly.For the ATPase reporter construct, we used the FAST-Red selection cassette instead of Basta.The reporter plasmids were transformed into mrl-1, and we confirmed that the reporter signals were consistent across three independent homozygous lines.
We constructed UTP18 pro long:GUS by gateway cloning the 3,218 bp UTP18 promoter region (Fig. 3A) in front of GUS in the pGII0229R4R3 pGreen vector harboring a basta resistance (Hellens et al. 2000;Heidstra et al. 2004).We used three independent T2 lines for GUS analysis and confirmed consistent GUS signals across all three lines.
To compare the expression patterns of UTP18 pro and mrl-1 pro , we constructed two promoter reporter lines: UTP18 pro long:3xVENUSnls and mrl-1 pro (2365bp):3xVENUSnls.Both were gateway cloned into pGII0229R4R3.Three independent T3 homozygous lines were used for confocal imaging and by using the same settings we confirmed that the expression patterns were consistent among the three lines.
mrl-1 pro (1964bp):GFP-UTP18 and UTP18 pro long:GFP-UTP18 complementation vectors were constructed with a FAST-Red selection cassette into pICSL4723.For these FAST-Red containing constructs transformed into mrl-2 +/− plants, next generation T1 seeds were selected based on red seed phenotype, sown on soil and subsequently genotyped for mrl-2 +/− heterozygosity by PCR before harvesting the T2 generation.We grew six plants from T2 generation for each (T2-5, T2-16) on soil and genotyped these for mrl-2 +/− heterozygosity and checked the siliques for complementation of the seed abortion phenotype.We also checked the GFP-MRL signals in these ovules by confocal microscopy.Similarly, we checked the siliques in one of the mrl-1 pro (1964bp):GFP-UTP18; mrl-2 +/− T1 plants (T1-15), and examined the GFP-MRL signals in the (delayed) embryos using the same confocal settings.Three plants from three independent UTP18 pro long:GFP-UTP18; mrl-1 T2 lines were grown in soil of which embryos containing GFP-UTP18 signal were imaged.
UTP18 pro short(536bp):UTP18 and UTP18 pro long(3218bp): UTP18 were constructed by gateway cloning the respective promoter fragments upstream of the UTP18 coding sequence in the pGII0229R4R3 vector.We then transformed the above plasmids into the mrl-1 mutant and selected positive plants by sowing T1 generation on soil and spraying a 100 mg/L basta solution three times a week.We used T2 seeds for statistics and a homozygous T3 line for imaging.
DRN pro :UTP18 was constructed by Gateway cloning the DRN promoter (4,771 bp) upstream of the MRL coding sequence in the pGII0229R4R3 vector.ATPase pro :UTP18 was constructed by Golden Gate cloning, combining FAST-Red selection cassette with the ATPase pro (982bp):UTP18 construct and the endlinker pICH41744 into the destination binary vector pICSL4723.The ATPase promoter constitutes a 982 bp sequence upstream of the open reading frame of the ATPase gene.The ATPase pro >> UTP18 was constructed by Golden Gate cloning, combining Fast-Red, UAS pro :erGFP, ATPase pro :GV, and UAS pro :MRLcds into pICSL4723.We introduced the constructs into the mrl-1 mutant and determined the penetrance of polyembryony and polycotyledon phenotypes in the T2 lines.One homozygous line (T3-7-4) was generated for imaging.
35S pro :GFP-UTP18 was generated by gateway cloning the UTP18 CDS into pGWB6.Next, we transformed this plasmid into A. tumefaciens.

Agrobacterium-mediated infiltration of Nicotiana benthamiana leaves
Agrobacterium-mediated infiltration was performed largely according to Diaz-Trivino et al. (2017).In brief: Agrobacterium colonies carrying the 35S pro :GFP-UTP18 plasmids were inoculated in 5 mL LB culture (with the appropriated antibiotics) and grown at 28 °C for ∼48 h.The same was done to Agrobacterium carrying the P19 silencing suppressor plasmid.The Agrobacterium cultures were then subcultured (1:100 ratio, v/v) into new 5 mL LB medium with 10 mM 2-(Nmorpholine)-ethanesulfonic acid (MES; pH 5.6) and 40 μM acetosyringone.Bacteria were grown at 28 °C until an OD600 of ∼3.0, and then gently pelleted (3,200 × g, 10 min).The pellets were resuspended in 10 mM MgCl 2 to an OD600 = 0.5, and acetosyringone was added to a final concentration of 200 μM.
The bacteria were kept at room temperature for at least 3 h without shaking.The Agrobacterium cultures containing the 35S pro :GFP-UTP18 and the p19-helper and 10 mM MgCl 2 were mixed in a v/v proportion of 3:1.With a 5 mL syringe, the mixed solution was infiltrated in young leaves of ∼14-d-old N. benthamiana plants.Two days after infiltration, the leaves were observed under the confocal microscope.

Tissue-specific and inducible CRISPR/Cas9
Inducible CRISPR/Cas9-mediated mutagenesis was used to generate the inducible and tissue-specific knock out of UTP18 in Col-0 roots.Induction was by transfer to 10 μM β-estradiol (dissolved in DMSO) and DMSO as a control.Five independent T2 lines were checked and T2-1 were used for imaging.

Histology and microscopy
Imaging of 3 to 7 d seedlings and flowering plants was performed using a Nikon 5300 camera and stereoscopic microscope Nikon SMZ745T.We screened transgenic seeds with the Fast-Red fluorescence marker using the Leica fluorescence binocular (Leica MZ16F).Root systems from the inducible CRISPR experiments were imaged using reflective scanning on an Epson Expression 11,000 XL scanner including an A3 transparency unit.Pollen viability was examined using a simplified Alexander staining (Peterson et al. 2010) on dissected anthers from WT and heterozygous mrl-2 flowers and imaged using the Zeiss Axioscope.We visualized GUS activity by staining tissues with a GUS solution containing 10 mM EDTA, 0.1% (w/v) Triton-X100, 1 mM potassium ferrocyanide (K 4 Fe(CN) 6 ), 1 mM potassium ferricyanide (K 3 Fe(CN) 6 ), 2 mM X-Gluc, and 0.1 M sodium phosphate buffer at pH 7.0.The inflorescences were incubated overnight in GUS solution at 37 °C, seedlings were incubated at 37 °C for 1 h.Images were taken using the Zeiss Axioscope.
To observe embryos within their embryo sac, these were treated with chloral hydrate solution (Lukowitz et al. 1996) and observed using the Zeiss Axioscope with a DIC module.To investigate female gametophyte development we treated ovules with Herr's solution 2:2:2:2:1 (85% acetic acid:chloral hydrate:clove oil:phenol:xylene) for 4 h (Aulbach-Smith and Herr 1984) and observed treated ovules using the Zeiss Axioscope.
We excited the green fluorescent protein (GFP) with 488 nm and detected it at a wavelength range of 490 to 595 nm, whereas we visualized Venus by excitation at 514 nm and detection between 515 and 625 nm.SR2200 was excited at 405 nm, and the emission was measured at 410 to 530 nm.

Protein sequence and phylogenetic analysis
UTP18 protein domains were predicted using the Conserved Domain tool through the NCBI (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi).Fourteen homologs of the UTP18 protein were obtained using the NCBI Homologene database (https://www.ncbi.nlm.nih.gov/homologene).We aligned the amino acid sequences using MEGA X with the MUSCLE algorithm set (https://www.ebi.ac.uk/Tools/msa/ muscle/) to default settings (Kumar et al. 2018), and generated a phylogenetic tree using the neighbor-joining method based on the Jones-Taylor-Thornton model (Jones et al. 1992) (Supplementary Files 3 and 4).To calculate Bootstrap support values, we performed 1,000 replicates of the tree topology

Figure 1 .
Figure 1 .Polyembryo and polycotyledon phenotypes displayed by the mrl-1 mutant.A to C) Multiple seedlings germinating from a single seed, exemplified by triplets (A, C) and twins (B, arrows).Triplet in (C) shows two seedlings fused at the shoot (arrow).D) Tricotyledon seedling.E) Wild-type-like seedling.Percentages indicate phenotype penetrance.F) WT heart stage embryo supported by a single file of suspensor cells.G) mrl-1 heart shape embryo with ectopic embryo structure (arrow) developing from the suspensor.Scalebars in (A) is 2.5 mm, in (B to E) is 10 mm, in (F, G) is 50 µm.

BFigure 3 .
Figure 3. Molecular identification of the UTP18 gene.A) Schematic illustration of the wild type and mutant UTP18 locus.UTP18 was mapped on chr 5 between markers MXM12-Del15 and Ciw8.The lines at the site of the triangle 326 bp upstream of the start codon (ATG), indicate the 2 Mb insertion including a 44 bp unknown sequence (bar) in the UTP18 gene of the mrl-1 mutant.The UTP18 transcribed region is represented by the pentagon with an intron indicated in front of the start codon (ATG), a triangle representing the mrl-2 CRISPR mutation at 379 bp, and TGA representing the stop codon.Numbers are base pairs relative to the start codon of UTP18.Introns not indicated in AT5G14040.Arrows indicate amplified promoter regions from wild type (UTP18pro) and mutant (mrl-1pro).B) Complementation of mrl-1 phenotypes by UTP18 driven by a short (563 bp, UTP18 pro short) or long (3,218 bp, UTP18 pro long) wild-type promoter sequence in segregating T2 lines.Bars are divided in percentage of WT-like seedlings (top section), percentage of polyembryo seedlings (middle), and percentage of polycotyledon seedlings (bottom section).Numbers above columns indicate germinating seeds counted per line.C) Germinating mrl-1 seeds.Arrows indicate multiple seedlings germinating from one seed.Number indicates polyembryonic seeds out of total shown in the image.D to F) Complementation of mrl-1 phenotype by introduction of UTP18 pro short:UTP18 (D), UTP18 pro long:UTP18 (E), UTP18 pro long:GFP-UTP18 (F).All lines depicted are homozygous for the mutation and construct.G) Reconstruction of the polyembryo phenotype by introduction of mrl-1pro:UTP18 into the mrl-2 knockout.Number indicates polyembryonic seeds out of total shown in the image.Scale bar: 10 mm.

Figure 4 .
Figure 4. UTP18 localizes to the nucleolus and is required for FM development.A) UTP18 mRNA expression relative to the E2FA control determined by RT-PCR in Arabidopsis roots (1), stems (2), cauline leaves (3), inflorescences (4) and siliques (5).B to F) UTP18 promoter activity examined by GUS staining of UTP18 pro long::GUS transgenic plants showing inflorescence (B), 7-d-old shoot (C), developing leaf (D), seedling root tip (E), and older root part developing root primordia (F).Arrow and inset indicate lateral root primordium (F).G to I) Transient expression of 35S pro :GFP-UTP18 in N. benthamiana leaf cells showing nuclear accumulation with highest levels in nucleolus.Overview (G) and single nucleus focus with overlay of bright field and fluorescence channel (H) and fluorescence channel only (I).(H) and (I) are representative of >10 imaged cells.J to L) Transgenic embryo expressing UTP18 pro long:GFP-UTP18 showing accumulation in nucleolus.Overview (G) and single nucleus focus (dashed line) in embryo proper with overlay (H) and fluorescence channel only (I).(J) to (L) is representative of >5 imaged embryos.M to N) Dissected siliques of WT showing normal ovules (M) compared to mrl-2 +/− silique segregating aborted ovules (N, arrowheads).O) Pollen vitality assay by Alexander staining of mrl-2 +/− stamen.Staining indicates viable pollen.P, Q) Asynchronous ovule development in mrl-2 +/− showing a FG6-7 stage ovule (P) and a FG2-3 stage ovule (Q) both from a stage 4 silique.Left arrow indicates a bigger secondary nucleus (fused polar nuclei) and a smaller nucleus (right arrow) in the FG6 stage (P).Arrows indicate two equal size nuclei in the FG2-3 stage (Q).Scale bars in B, E, F, M, N is 5 mm, C, D, G is 100 μm, J, O, P, Q and F inset is 50 μm, H, I, K, L is 5 μm.

Figure 5 .
Figure 5. Altered UTP18 expression pattern is causal to the polyembryo phenotype.A, B) Complementation of the loss of function mrl-2 by expression of UTP18 pro long:GFP-UTP18 exemplified by formation of WT looking octant (A) and dermatogen (B) stage embryos.C to E) Expression of mrl-1 pro(1964):GFP-UTP18 in the loss of function mrl-2 reconstitutes the mrl-1 polyembryo phenotype.Misexpression of UTP18 in octant (C) stage embryo is associated with delayed embryo proper development and suspensor reprogramming (D).Recovery of UTP18 expression in the primary embryo at later developmental stages is associated with polyembryo formation (E).F) Phenotypes of mrl-1 are complemented by embryo proper expression of DRN pro :UTP18, but not by suspensor specific ATPase pro :UTP18 and ATPase pro >> UTP18 expression.Numbers above columns indicate germinating seeds counted per line.Phenotypes are determined in transgenic T2 lines.Green represents germinating twin and triple seedlings, purple represents polycotyledon seedlings, and grey represents WT-looking seedlings.In A-E, n (observed embryo numbers) > 5. Scalebars are 50 µm, except in C 5 µm.
. The bootstrap values, representing the percentage of replicate trees where associated taxa clustered together, are shown next to the branches.The branch length representing the evolutionary meerling polyembryonic mutant THE PLANT CELL 2024: 36; 2550-2569 | 2565

Table 1 .
Developmental stages of FG in WT and mrl-2 +/− flowers