The Arabidopsis INNER NO OUTER (INO) gene acts exclusively and quantitatively in regulation of ovule outer integument development

Abstract

The INNER NO OUTER (INO) gene is essential for formation of the outer integument of ovules in Arabidopsis thaliana. Initially described lesions in INO were missense mutations resulting in aberrant mRNA splicing. To determine the null mutant phenotype, we generated frameshift mutations and found, in confirmation of results on another recently identified frameshift mutation, that such mutants have a phenotype identical to the most severe splicing mutant (ino-1), with effects specific to outer integument development. We show that the altered protein of an ino mRNA splicing mutant with a less severe phenotype (ino-4) does not have INO activity, and the mutant is partial because it produces a small amount of correctly spliced INO mRNA. Screening for suppressors of ino-4 in a fast neutron-mutagenized population identified a translocated duplication of the ino-4 gene, leading to an increase in the amount of this mRNA. The increased expression led to a decrease in the severity of the mutant effects, indicating that the amount of INO activity quantitatively regulates outer integument growth. The results further confirm that the role of INO in Arabidopsis development is specific to the outer integument of ovules where it quantitatively affects the growth of this structure.

INTRODUCTION

As the structures that develop into seeds, ovules are an essential component of sexual reproduction in seed plants. Angiosperm ovules comprise a combination of diploid maternal tissues and the haploid megagametophyte that develops to form the embryo and endosperm following double fertilization. Most angiosperm ovules include two sheathing maternal structures, the integuments, which appear to be essential for aspects of fertilization and/or embryo development and that further develop themselves, forming the seed coat. Genes involved in regulation of integument formation have been described in Arabidopsis thaliana through mutant and molecular studies, with many found to encode transcription factors. While having important roles in other aspects of plant development, the AINTEGUMENTA (ANT) and WUSCHEL (WUS) genes are necessary for integument formation, with mutation in either gene, leading to a complete absence of integuments (Baker et al., 1997; Gross-Hardt et al., 2007; Modrusan et al., 1994). The pattern of WUS expression is regulated by NOZZLE/SPOROCYTELES (NZZ/SPL) (Sieber et al., 2004). Integument identity is under the control of BELL1 (BEL1) (Modrusan et al., 1994; Ray et al., 1994; Reiser et al., 1995; Robinson-Beers et al., 1992) and a combination of the MADS-box genes SEEDSTICK (STK) and SHATTERPROOF 1 and 2 (SHP1 and SHP2) with mutations leading to formation of amorphous or carpelloid structures in place of integuments (Pinyopich et al., 2003). Determinants of polarity also play roles in development of the bifacial integuments. Adaxial polarity determinants such as members of the Class III HDZIP family play important roles in the final shapes of the two integuments (Kelley et al., 2009). Two KANADI (KAN) genes, KAN1 and KAN2, are abaxial determinants necessary for planar extension of the outer integument (OI) (Eshed et al., 2001; McAbee et al., 2006), and ABERRANT TESTA SHAPE (ATS/KAN4) partners with ETTIN/ARF3 to regulate inner integument (II) extension and integument separation (Kelley et al., 2012; McAbee et al., 2006). Only one member of the YABBY gene family of abaxial determinants, INNER NO OUTER (INO), is involved in integument development, and it acts as a master gene in development of the OI with strong ino mutants leading to complete absence of this structure (Baker et al., 1997; Vijayan et al., 2021; Villanueva et al., 1999). Genetic and physical interactions between these transcription factors have been studied in many cases (Gasser & Skinner, 2019; Reviewed in Barro-Trastoy et al., 2020).

Arabidopsis ovule morphogenesis has been characterized by scanning electron microscopy, sectioning (Robinson-Beers et al., 1992), clearing (Schneitz et al., 1995), and more recently by elegant confocal and computational analysis (Vijayan et al., 2021). During morphogenesis, the OI, and hence its primary regulator, INO, is a major determinant of ovule asymmetry and form. Asymmetric development of the ovule begins with curvature of the funiculus in the direction of the base of the gynoecium (the gynobasal direction) (Figure 1a). The II initiates in a symmetrical ring of cells around the ovule primordium (Figure 1a). INO expression and OI development initiate on the gynobasal side of the ovule on the inside of the funicular curve (Robinson-Beers et al., 1992; Vijayan et al., 2021; Villanueva et al., 1999) (Figure 1a). OI development continues with maximal growth on the gynobasal side of the ovule, resulting in curvature of the distal part of the ovule toward the apex of the gynoecium (the gynoapical direction), producing the final “S” shape of the ovule with the micropyle oriented gynoapically (Figure 1b and Figure 2a). Curvature of the embryo sac is driven by cell divisions in a chalazal region internal to the surrounding OI (Figure 1b). In the strong ino-1 mutant, funicular curvature occurs in the gynobasal direction just as in the wild-type (Figure 1c), but the complete absence of OI initiation or subsequent OI growth leaves the micropyle oriented in the gynobasal direction (Figure 1d and Figure 2b). The proliferation of the chalazal region occurs in ino-1, forming a thick collar of tissue below the exposed II (Figure 1d and Figure 2b).

FIGURE 1. Diagrams of ovule sections. A, B wild-type; C, D ino-1 mutant. All ovules are oriented with the gynoapical side to the right. (a) Wild-type ovule at Stage 2-IV where inner integuments (IIs) and outer integuments (OIs) have been initiated from the chalaza. (b) Wild-type ovule at Stage 3-VI where the OI encloses the II, nucellus, and chalazal region. Asymmetric expansion of the OI places the micropylar opening on the gynoapical side of the ovule. (c) ino-1 mutant ovule at Stage 2-IV where only the II has initiated from the chalaza. (d) ino-1 mutant ovule at Stage 3-VI, the II has covered the nucellus, but absence of the OI leaves the micropyle directed toward the gynobasal side of the ovule. Bar in (a) is 50 μm in all panels. Diagrams based on Baker et al. (1997)and Vijayan et al. (2021). Stages from (Schneitz et al., 1995). c, chalaza; f, funiculus; i, inner integument; n, nucellus; o, outer integument; *, micropyle.

FIGURE 2. Scanning electron micrographs of Arabidopsis ovules. False color has been added to emphasize ovule structures. All ovules are from flowers at or just following anthesis and are oriented with the gynoapical side to the right. (a) Wild-type ovule where the curved outer integument has covered the inner integument. (b) ino-1 mutant ovule where the outer integument is absent, exposing the inner integument, which remains erect due to this absence, and the chalazal region. (c) ino-4 mutant ovule has a rudimentary outer integument structure, which fails to cover the inner integument, and from which the expanding chalazal region protrudes. (d) Ovules of ino-1 mutant plant containing transgene transcribing the mutant ino-4 cDNA (pJC02) that fails to complement so that the ovules retain the ino-1 phenotype. (e) ino-9 mutant ovules that are phenotypically indistinguishable from those of ino-1. (f) i20 partially restores outer integument growth of ino-4 so that in the ino-4 i20 double mutant the outer integument can largely cover the inner integument and chalaza, leaving only a small part of the inner integument protruding. Bar in (a) is 50 μm in all panels. c, chalaza; f, funiculus; i, inner integument; o, outer integument; *, micropyle.

Only two different lesions in the Arabidopsis INO locus were identified in extensive forward mutant screens (Baker et al., 1997; Villanueva et al., 1999). Both are single-base substitutions in introns that result in new upstream splice acceptor sites. While being independently isolated, the ino-1 and ino-2 alleles have identical substitutions in intron five, creating an AG doublet that when utilized as a splicing acceptor site adds 11 bases to the mRNA, leading to a frameshift near the start of the essential YABBY domain (Villanueva et al., 1999). Absence of the OI in homozygous mutants results in near complete female infertility. That this is a secondary effect of the integument absence is shown by normal segregation of the recessive mutation in heterozygous progeny, where the haploid ino mutant embryo sacs effectively produce seed when they form within the wild-type phenotype heterozygous maternal tissue (Baker et al., 1997; Villanueva et al., 1999). Another ino allele, ino-4, has a substitution in intron four, also introducing an AG doublet that acts as a splice acceptor site. In this case, 15 bases are inserted into the mRNA, resulting to addition of five amino acids in the YABBY domain of the translation product, while still maintaining the reading frame for the remainder of the protein (Villanueva et al., 1999). ino-4 has a less severe phenotype than ino-1 (Villanueva et al., 1999). In ino-4, the OI initiates, but OI development arrests after only a small amount of growth, leaving the II largely exposed (Figure 2c). Chalazal cell proliferation occurs within the abortive OI and sometimes bulges out from this incomplete structure on the gynoapical side (Figure 2c). ino-4 also produces more seeds that ino-1 but fewer seeds than the wild-type (Villanueva et al., 1999). The weaker phenotype was hypothesized to be due to partial activity of the INO protein with inserted amino acids (Villanueva et al., 1999).

The altered splicing nature of the only two ino mutations initially described in Arabidopsis raised several questions. Both described mutations resulted in the introduction of splice acceptor sites that produce aberrant mRNAs. However, both also retain the original splice acceptors, raising the possibility that both could still produce wild-type INO mRNA. The assay used to indicate that normal INO mRNA was not produced, and direct sequencing of synthesized cDNA (Villanueva et al., 1999) is not especially sensitive, and low levels of wild-type mRNA could have been present. Thus, the actual ino null mutant phenotype in Arabidopsis was not clear. It is possible that such a mutation could produce deleterious pleiotropic effects precluding isolation. To assess this, we set out to identify null mutant alleles and used CRISPR (clustered regularly interspaced short palindromic repeats)-mediated mutagenesis to create null mutants. Prior to completion of this manuscript, another group used the same method to identify a putative null mutant, ino-5, and showed that it did not have such pleiotropic effects (Vijayan et al., 2021). Another question arising from the nature of the initially identified mutants is whether the ino-4 weak mutant effect results from a protein with reduced activity or from partial leakage of wild-type mRNA formation in this mutant. Finally, we do not know if levels of INO activity differentially affect OI formation.

To address these questions, we have reexamined the existing ino mutants by additional means, have obtained and characterized additional mutants that include likely null mutants, and have isolated a second-site mutant that addresses the question of quantitative expression of INO. As also shown in a recent study (Vijayan et al., 2021), we find that newly isolated null mutants duplicate the phenotypic effects of the ino-1 mutant, indicating that ino-1 is also effectively a null mutant. We show that the altered protein of ino-4 does not have INO activity, and the phenotype is weak because it produces a small amount of normal INO mRNA. A gene duplication leading to a quantitative increase in the amount of this mRNA leads to a decrease in the severity of the effects, indicating that the amount of INO activity quantitatively regulates the developmental effects of INO. The results further confirm that the role of INO in Arabidopsis development is specific to the OI of ovules.

MATERIALS AND METHODS Plant material

Arabidopsis plants were grown under long-day conditions as previously described (McAbee et al., 2006). Preexisting alleles used in this study are ino-1 (in the Ler background), ino-4 (in the Col-0 background) (Villanueva et al., 1999), and ino-9 (SALK_0063818) (Alonso et al., 2003). Plant transformation utilized Agrobacterium GV3101 (Koncz & Schell, 1986) and the floral dip method (Clough & Bent, 1998).

For ovule and seed counts, six fruits were collected after abscission of the other floral organs, starting from the fourth fruit produced. Fruits were collected from at least two different plants for each examined genotype. Aborted ovules and ovules forming seeds were counted in each of the two fruit locules (carpels) using a Zeiss (Oberkochen, Germany) SV8 stereomicroscope. Statistical analysis was performed in GraphPad Prism (San Diego, CA).

Plasmid manipulation

pJC02—To clone the cDNA for the aberrantly spliced ino-4 mRNA, cDNA was synthesized from ino-4 mRNA using Superscript II reverse transcriptase (RT) and oligo deoxythymidine (dT) primers (Thermo Fisher Scientific, Waltham, MA). The specific cDNA was amplified by polymerase chain reaction (PCR) using Phusion DNA polymerase (New England Biolabs, Ipswich, MA) and primers AtINOcDNA-f and AtINOcDNA-r (Supporting information Table S1), which introduce flanking BamHI and XbaI sites, respectively. The resulting PCR product was inserted into pJet2.1 using the CloneJet cloning kit (Thermo Fisher Scientific, Waltham, MA), and one sequence-verified ino-4-specific cDNA clone was designated pSR06. The cDNA in pSR06 was transferred into pJRM42 (a derivative of pRJM33—the expression vector for the wild-type INO cDNA from the wild-type INO promoter; Meister et al., 2005) as a BamH1/Xba1 fragment at these same sites, creating pJC01. pJC01 is identical to pRJM33 except having the ino-4 cDNA replacing the wild-type INO cDNA. The NotI insert fragment of pJC01 containing the complete promoter cDNA expression cassette was inserted into the NotI site of pMLBART (Gleave, 1992) to create the plant transformation vector pJC02, which can be selected in plants for phosphinothricin resistance.

CRISPR mutagenesis—For CRISPR-mediated mutagenesis, we utilized the dual guide RNA approach of Pauwels et al. (2018). Two guide RNA targeting sequences are inserted into two similar vectors at BpiI sites where they are adjacent to Arabidopsis U6 RNA promoters and other constant sequences of the guide RNA, and these two transcription units are subsequently inserted into a plant transformation vector by site-directed recombination using LR Clonase II (Thermo Fisher Scientific, Waltham, MA) (Pauwels et al., 2018). In our case, the synthetic guide RNA targeting sequences INOCRIS4 and INOCRIS5 (Supporting information Table S1) were inserted into the BpiI sites of pMR218 (Pauwels et al., 2018) and a derivative of pMR217 (Pauwels et al., 2018) with a synthetic U6 RNA promoter. LR Clonase II was used to insert the transcription units from these two plasmids into a derivative of plant transformation plasmid pDE-CAS9Km (Pauwels et al., 2018) in which the kanamycin resistance gene is replaced by a gene expressing phosphinothricin acetyltransferase (conferring resistance to BASTA herbicide) and also incorporating a gene for production of the green fluorescent protein (eGFP) in the embryo under control of the Arabidopsis oleosin promoter (Plant et al., 1994), resulting in pGD60. GFP fluorescence in the embryo from the P-Oleosin::eGFP gene allows for the differentiation of seeds harboring the CRISPR plasmid from those which have lost the plasmid through genetic segregation (Aliaga-Franco et al., 2019).

PCRs

PCRs utilized DreamTaq polymerase (Thermo Fisher Scientific, Waltham, MA) and the supplied reaction buffer components. Superscript II RT (Thermo Fisher Scientific, Waltham, MA) with oligo dT primers was used for cDNA synthesis for reverse transcriptase-polymerase chain reaction (RT-PCR).

Microscopy

Scanning electron microscopy was performed as described (Kelley et al., 2009). Dark field images were acquired on a Zeiss Axioplan microscope (Carl Zeiss Inc, Oberkochen, Germany) with a Moticam 5 digital camera (Motic Inc., San Antonio, TX). Images were cropped and matched for contrast in Adobe Photoshop CS5 (Adobe, San Jose, CA).

Fast neutron mutagenesis and mutant identification

Homozygous ino-4 seeds were mutagenized with fast neutrons (60 Gy) at the University of California, Davis McClellan Nuclear Research Center, Sacramento, CA. The treatment resulted in 20%–30% lethality, and 10%–20% of the surviving seed produced fully infertile plants. M1 seeds were planted in pools of five plants. M2 seeds from each pool were collected, and 50 seeds from each pool were planted and grown to anthesis. Ovules were examined under a stereomicroscope to detect deviations from the ino-4 mutant phenotype.

Genetic mapping and whole-genome analysis

To map the location of the mutation responsible for the suppressed ino-4 phenotype (i20), we outcrossed the i20/i20 ino-4/ino-4 plants to Ler. The F2 progeny from this cross were used to map the location of i20 to Chromosome 3 using Cleaved Amplified Polymorphic Sequences (CAPS), and Simple Sequence Length Polymorphism (SSLP) markers (Jander et al., 2002) and InDel markers (Hou et al., 2010). The suppressed phenotype observed in the i20 mutation acted semi-dominantly; thus, individuals used for fine mapping were progeny-tested, and only homozygous individuals were used for further analysis. The mapped location of i20 was between marker nga126 (3.098 Mb) and 3-4608 (4.608 Mb). A total of 12 individuals from F2 and F3 mapping populations, homozygous for ino-4 and i20, were used to extract DNA for sequencing.

Leaf samples from ino-4 and ino-4 i20 plants were separately pooled and DNA was extracted with Nucleon Phytopure DNA extraction kit (GE Healthcare, Pittsburgh, PA), according to the provider's instructions. NEBNext DNA library components (New England Biolabs, Ipswich, MA) were used for fragmentation to 200–500 bp and library preparation with Nextflex-96 indexes (Bioo Scientific, Austin, TX).

Libraries were pooled and sequenced using the PE100 bp protocol on a single lane of Hiseq 2000 (Illumina, Inc., San Diego, CA) at the Vincent J. Coates Genomics Sequencing Laboratory at UC Berkeley. A total of 67.5 and 99.9 million reads were obtained for ino-4 and ino-4 i20, respectively. Reads were preprocessed to remove tag and poor-quality regions using Sickle (Version 1.33, https://github.com/najoshi/sickle) and Scythe (https://github.com/najoshi/scythe), respectively. Reads were aligned with the Col-0 TAIR 10 reference sequence (Lamesch et al., 2012) using BWA-MEM (Li, 2013). Aligned sequences were viewed in Tablet (Milne et al., 2013) to visualize changes relative to the wild-type Col-0 sequence in the region identified as including the i20 suppressor based on the mapping.

RESULTS ino-4 mis-spliced mRNA cannot support OI growth

Whereas the ino-1 mutant exhibits no growth of the OI, the ino-4 mutant shows partial OI growth ((Villanueva et al., 1999), Figure 2b,c and Figure 3b,c). This was hypothesized to result from partial INO activity of the protein product of the mis-spliced ino-4 mRNA. This product has a fifteen-nucleotide insertion that would result in insertion of five amino acids at Glu-147 in the translated protein and no other change in the reading frame (Villanueva et al., 1999). To determine whether the mutant protein has activity, we tested transgenic complementation of ino-1 with a construct that produces the ino-4 mutant protein. The wild-type cDNA was shown to be effective at complementing ino-1. A construct, pRJM33 (Villanueva et al., 1999), including 2.3 kb of INO 5′-flanking DNA, INO cDNA, and 2.0 kb of INO 3′-flanking DNA, led to the formation of fully wild-type ovules in 28 of 42 independently produced transgenic homozygous ino-1 plants (Meister et al., 2005). When the wild-type INO cDNA was replaced in pRJM33 with the aberrantly spliced cDNA of ino-4 and the resulting construct (pJC02) was introduced into plants, 29 of 29 independently transformed homozygous ino-1 transgenics retained the ino-1 ovule phenotype (Figure 2d and 3d). These results differ significantly (p < .0001, two-tailed Chi²), demonstrating the failure of the ino-4 cDNA sequence to complement the mutant and hence the failure of the extended ino-4 protein to function.

FIGURE 3. Dark field micrographs of Arabidopsis ovules. All ovules are from flowers at or just following anthesis and are oriented with the gynoapical side to the right. (a) Wild-type. (b) ino-1. (c) ino-4. (d) ino-1 transformed with pJC02 transcribing ino-4 mutant cDNA. (e) ino-9. (f) ino-4 i20 double mutant. (g) ino-6. (h) ino-7. (i) ino-8. Bar in (a) is 50 μm in all panels. c, chalaza; i, inner integument; o, outer integument.

ino-4 produces a small amount of wild-type INO mRNA

The finding that the ino-4 aberrant mRNA/protein could not complement ino-1 indicates there must be another source of active INO provided by the ino-4 allele. RT-PCR analysis of mRNA from ino-4 detected a prominent product of the expected size for the aberrant ino-4 mRNA but also a smaller amount of a product that co-migrated with the product derived from mRNA from wild-type plants (Figure 4). To test if this product is properly spliced INO mRNA, we devised primers that cross the splice junction and would amplify only the properly spliced INO cDNA (Supporting information Table S1). We found that ino-4 does produce wild-type mRNA, along with the mis-spliced ino-4-specific product (Figure 4).

FIGURE 4. Reverse transcription polymerase chain reaction (PCR) detection of INO mRNA. (a) Lanes 1–5 show PCR products from primers flanking Intron 4 (ino-4 FOR, ino-4 REV) that will amplify cDNA from both mRNAs that utilized the wild-type splice site (WT) or the ectopic ino-4 splice site (ino-4). Lanes 6–8 show products from primers that selectively amplify only cDNA from mRNA using the wild-type splice site due to one primer crossing the junction between Exons 4 and 5 (inoex45F2 with inoRTR1). Input DNA in lanes: (1) plasmid pRJM33, which includes the wild-type INO cDNA. (2) pSR06, which includes the cDNA to the aberrantly spliced ino-4 cDNA. (3) and (6) cDNA to mRNA from ino-4 flowers. (4) and (7) wild-type Col-0 flower cDNA. (5) and (8) cDNA to mRNA from ino-4 i20 double mutant flowers. (b) PCR of cDNA from mRNA from flowers of the indicated plants or plasmid control (pRJM42) using primers selective for only the wild-type form of the INO mRNA. Samples were taken and evaluated following 27 cycles (Lanes 1–3), 30 cycles (Lanes 4–6), or 33 cycles (Lanes 7–10). Sizes of DNA markers are indicated in base pairs at the center (a) and left (b).

INO quantitatively affects ovule development

The effect of the apparent reduced level of functional INO in ino-4 suggests that the quantity of INO differentially affects OI growth. A test of this hypothesis was serendipitously produced by our identification of a second-site suppressor of ino-4 mutants. Ino-4 plants produce small numbers of viable seeds, which served as a starting point for suppressor screening. We mutagenized ino-4 seeds with fast neutrons and screened M2 plants for increased OI growth and increased fertility. One homozygous ino-4 line, i20, had significantly more OI growth than ino-4 single mutants, usually producing a nearly wild-type OI (Figure 2f and 3f). We evaluated the presence of wild-type INO mRNA in ino-4 and ino-4 i20. ino-4 i20 also produced both aberrant ino-4 mRNA and wild-type mRNA (Figure 4), but both were at an apparently higher level than in ino-4 (Figure 4).

Using segregating plants from a cross of ino-4 i20 (in the Col background) to Ler, we mapped the mutation in i20 to the top of Chromosome 3. We used whole-genome sequencing to compare ino-4 i20 sequence to the TAIR10 Col reference sequence (Lamesch et al., 2012). We found an apparent deletion within the mapped region on the basis of the absence of sequence reads (Figure 5a). This 57 bp deletion is between At3g10490 and At3g10500. Further work showed that there was a DNA insertion associated with the deletion. Individual reads that spanned this region were not present, and examination of the sequence reads at the borders of the deletion on Chromosome 3 showed that the reads paired with these reads were not aligned with Chromosome 3. Instead reads from one side of the deletion paired with reads mapped to a region on Chromosome 1, and those at the other end of the deletion paired with reads mapped to a region on Chromosome 2. Examination of mapped reads in these regions of Chromosomes 1 and 2 showed an approximately twofold increase in depth of sequence reads, suggesting that these regions were duplicated in the genome. Two areas of Chromosome 1 showed this duplication: a ~40 kb region close to the INO locus and a second ~1.1 Mb region that included the ino-4 sequence (Figure 5b). Similarly, this section of Chromosome 2 showed twofold more reads mapped over a 2.2 kb region (Figure 5c). The simplest explanation for this would be for regions of Chromosomes 1 and 2 to be duplicated and inserted into a small deletion on Chromosome 3. We used BLAST (Altschul et al., 1990) searches of our entire set of i20 reads to identify individual reads that crossed the borders between the deletion site and the putative insertions and were able to find the expected numbers of such reads. Applying this method recursively, we were able to determine that a small region of Chromosome 2 and a larger portion of Chromosome 1, which included the ino-4 locus, were duplicated and inserted into Chromosome 3 (Figure 5d and Supporting information Figure S1). Thus, the genome of i20 included an extra copy of the ino-4 locus on Chromosome 3. Using variable cycle counts in PCR, we found that the level of wild-type mRNA was apparently higher in ino-4 i20 than in ino-4 but was still lower than in wild-type plants (Figure 4b). Thus, an increase in the level of wild-type mRNA in ino-4 i20 over the minimal level present in ino-4 induced increased OI growth and produced ovules closer in form to those of wild-type plants. This demonstrates that the role of INO activity in ovule development is not a binary switch, but rather, the activity quantitatively affects development of the OI.

FIGURE 5. The structure of the insertion in the i20 region. (a)–(c) Illumina sequence reads of homozygous ino-4 i20 DNA aligned with wild-type Columbia Arabidopsis chromosomal regions as visualized in Tablet (Milne et al., 2013) software. Each ~100 base sequence read is shown as a colored line (appearing more as colored spots at the zoomed-out view in (b) and (c)) aligned below the corresponding sequence of the indicated Arabidopsis chromosome. (a) An absence of reads aligning with sequence between bases 3,270,821 and 3,270,878 of Chromosome 3 indicates deletion of this region, which is in the area within which i20 was mapped. Reads immediately to the left and right side of the deletion were paired with locations on Chromosomes 1 and 2, respectively, indicating the presence of inserted DNA. (b) Two regions of Chromosome 1 show approximately twofold more reads than expected. Reads at 7,312,475 paired with reads on Chromosome 3 at Position 3,270,821. Reads at Position 8,455,220 paired with reads on Chromosome 2 at Position 8,332,118. (c) A region of Chromosome 2 also showed approximately twofold as many reads as an average region of the genome. Reads at 8,334,334 paired with reads on Chromosome 3 at Position 3,270,878. (d) Diagram of the structure of the insertion into Chromosome 3 of i20 that is composed of regions of Chromosomes 1 and 2 (numbers indicate source chromosomes of the regions). Lower case letters at the top of the figure correspond to the chromosomal regions indicated by the same letters in panels (a)–(c). The location of the ino-4 gene in an inserted Chromosome 1 region is indicated by the green arrow.

The variable degree of OI formation in the different mutants showed corresponding differential effects on fertility and seed formation, and we quantitated these effects. The total numbers of ovules initiated in the different lines were similar, averaging between 26 and 33 per carpel (Supporting information Figure S2 and Supporting information Data S1). The fraction of ovules eventually forming seeds was high in the wild-type averaging 96% and 98% in Col and Ler, respectively (Figure 6 and Supporting information Data S1). ino-1 was largely infertile averaging only .3% of ovules setting seed, whereas ino-4 was substantially more fertile with 12% of ovules forming seeds (Figure 6 and Supporting information Data S1). The addition of the i20 translocation to ino-4 restored a wild-type level of seed set with an average of 92% of ovules forming seeds (Figure 6 and Supporting information Data S1).

FIGURE 6. Percent of initiated ovules that go on to form seeds. Seeds and aborted ovules were counted in 12 carpels for each genotype. Percentage seed set was determined for each carpel by dividing the number of seeds formed by the total of seeds formed plus aborted ovules observed (X100). Black dots indicate actual determinations from each carpel; bars represent the mean of all of the determinations for that line; and whisker lines indicate the standard deviations of the means. Different letters (a, b, and c) indicate different sets of means that statistically differ from each other (p [less than] .01) by Turkey's multiple comparison test.

Insertion mutant of INO

Because we had failed to isolate a clear null mutant of INO from chemically mutagenized material, we examined the set of available Arabidopsis insertion mutants for insertions in the INO locus. We found only one line, SALK_0063818, which we designated ino-9. Sequencing of genomic DNA flanking the inserted T-DNA shows that the insertion is 162 bp 5′ of the translation start site of the INO coding region (Supporting information Figure S3). ino-9 has a phenotype that precisely duplicates that of ino-1 (Figure 2e and 3e). A seed set fraction of .4% was not statistically different from that of ino-1 (Figure 6 and Supporting information Data S1). The insertion is between the regulatory region of the promoter and the coding sequence (Simon et al., 2012), apparently interrupting expression. However, like the ino-1 mutation, this mutation does not preclude the possibility of formation of active INO protein as initiation of transcription induced within the insertion could still produce wild-type INO protein. As a result, we still did not have certainty that the ino null mutant phenotype had yet been observed.

CRISPR-induced null ino mutants

In the absence of identifying a verified null mutant from random mutagenesis, we used CRISPR technology to induce deletion and insertion mutations in the coding region of INO that would cause frameshifts and preclude formation of the full INO protein. A plant transformation plasmid, pGD60, was constructed to produce the CAS9 protein and two guide RNAs directed to two locations in the coding region of INO and used to transform wild-type plants. The vector used included a cotyledon-expressed cassette producing GFP. This enabled simple isolation of M2 seeds, which had segregated out the CRISPR vector on the basis of the absence of GFP fluorescence (Aliaga-Franco et al., 2019). Multiple new ino mutant lines were isolated in the Ler background, and three were chosen for complete characterization and designated ino-6, −7, and −8 (Supporting information Figure S3). ino-6 included a single-base deletion of a “T” residue at Position 80 (in the first exon) and a second mutation inserting a “T” residue after Position 883 (in Exon 5) of the coding region of INO. This would introduce a frameshift and cause mistranslation of the entire zinc-finger region and the first half of the YABBY domain (Figure S3). ino-7 had an insertion of an “A” residue after Position 883 (in Exon 5) of the coding region, causing a frameshift of the second half of the YABBY domain (Villanueva et al., 1999) and all subsequent amino acids (Supporting information Figure S3). ino-8 had a deletion of Bases 77 and 78 of the coding region (in Exon 1) of INO. This two-base deletion would cause a frameshift leading to mistranslation of all of the INO protein except the first 25 amino acids (Supporting information Figure S3). All three of these mutants eliminate regions of INO shown to be critical to its function (Meister et al., 2005) and hence are null mutations. Homozygous plants of all three mutants were indistinguishable from ino-1 plants and wild-type plants in vegetative habit, and all three produced ovules that duplicated the ino-1 and ino-9 phenotypes (Figure 3). The seed set was examined in ino-6 with no seeds observed, and this result was not statistically different from the extremely low seed set of ino-1 and ino-9 (Figure 6 and Supporting information Data S1). These results show that ino-1 and ino-9 are also null mutants. Thus, the effects of ino complete loss-of-function mutants are confined to the OI of ovules, confirming the results of Vijayan et al. (2021).

DISCUSSION ino null mutants

Until 2021, the two identified ino lesions had base substitutions that caused ectopic utilization of upstream splicing sites in INO pre-mRNA (Villanueva et al., 1999). Both retained the normal splice junctions and so had the potential for production of wild-type mRNA. This raised the possibility that null mutants had not been isolated. This failure to identify null mutants could result from such mutants being selected against due to pleiotropic effects. We first searched for null mutants in libraries of existing insertion mutants, but there, we found only a mutant with an insertion in the 5′-untranslated region. We could not say this was certainly null. We followed this by using CRISPR-based mutagenesis to produce frameshift mutations in INO exons. Following completion of our mutagenesis, Vijayan et al. (2021) published work that included isolation of a putative null CRISPR-induced allele (ino-5). They reported the phenotype of ino-5 as not significantly different from ino-1. Confirming this result, all three of our CRISPR-induced null mutants and also the insertion mutant duplicated the ino-1 phenotypic effects. This indicates that ino-1 also was null.

The null mutants all exhibited complete failure in formation of an OI ((Vijayan et al., 2021; Villanueva et al., 1999); Figure 1). All of the existing null mutants also exhibit only rare, sporadic formation of functional embryo sacs (Skinner & Gasser, 2009; Vijayan et al., 2021; Villanueva et al., 1999) and so rarely set seed (Figure 6 and Supporting information Data S1). In addition, the detailed imaging used by Vijayan et al. (2021) revealed subtle alterations in the formation of the nucellus and II. However, expression of INO has only been observed in the abaxial layer of the OI and never in these other affected cells and locations (Balasubramanian & Schneitz, 2000; Villanueva et al., 1999). Vijayan et al. (2021) hypothesize that absence of the OI would alter the hormonal environment during nucellar development. We note that allocation of nutritional resources would also be altered by absence of the OI, which would be expected to be a significant nutrient sink. Additionally, cell wall tension and other hydrostatic forces within the developing ovule would be altered by the absence of the confining OI, and such forces also have roles in morphogenesis (Green, 1996). On this basis, we concur with Vijayan et al. (2021) that the effects of ino mutations on the development of the II, nucellus, and embryo sac are most likely to be secondary effects of the absence of the OI and not direct effects of absence of active INO protein.

It is interesting to note that the only other plant in which an ino mutant has been identified to date was Annona squamosa, the sugar apple tree. In this Thai seedless (Ts) mutant, the embryo sac still formed in the majority of ovules despite the complete absence of an OI (Lora et al., 2011). In Arabidopsis, the tenuinucellate ovules have only a single nucellar cell layer surrounding the embryo sac, and this is sheathed in a thin two cell layer thick II (Robinson-Beers et al., 1992; Vijayan et al., 2021). In contrast, the crassinucellate ovules of A. squamosa are surrounded by a robust nucellus of at least four cell layers and an II comprising multiple cell layers at their micropylar end (Lora et al., 2011). Thus, even when the OI is absent, there is substantial maternal tissue surrounding the developing embryo sac that could insulate it from the effects of the absence of an OI. This could explain the further development of the embryo sac in the ino mutant in this species relative to Arabidopsis. The ino mutation in A. squamosa appears to be a complete deletion of the INO locus and hence is a null mutation (Lora et al., 2011). The absence of pleiotropic effects on other aspects of development in A. squamosa is thus consistent with the absence of pleiotropic effects now demonstrated for Arabidopsis ino null mutants.

INO expression level correlates with the degree of OI growth

It had been previously hypothesized that the partial OI growth in ino-4 resulted from a partially active protein produced from the mis-spliced but in-frame ino-4 mRNA (Villanueva et al., 1999). The failure of a transgene transcribing the cDNA of the mis-spliced mRNA to induce any detectable OI growth demonstrated that this was not the case. The mis-spliced mRNA produces a protein with a five amino acid insertion in the YABBY domain (Villanueva et al., 1999), and our result shows that this protein is not active, supporting the importance of the integrity of this domain for function of the INO YABBY transcription factor. Instead, we were able to show that ino-4 produces a small amount of properly spliced mRNA (Figure 4). That this low amount of INO mRNA is able to promote a minimal growth of the OI is an indication that INO quantitatively affects OI growth. Further evidence of this is provided by our isolation of a line with a translocated duplicate ino-4 gene. As a result of the presence of this additional gene copy, this line has an increased level of wild-type mRNA over ino-4 (Figure 4) and an apparent proportional increase in OI growth (Figure 2f and 3f) with a corresponding increase in seed set (Figure 6). Thus, INO is a direct quantitative regulator of OI growth rather than being a binary switch. This is consistent with the observed function of INO in wild-type plants. In situ hybridization and reporter transgene studies (Meister et al., 2002; Villanueva et al., 1999) show that INO is expressed at the highest level on the gynobasal side of the ovule where maximal OI growth takes place. The level of INO expression shows a progressive decrease toward the gynoapical side of the ovule, correlating with the decreased OI growth in this area that leads to the hood-like OI of wild-type ovules (Meister et al., 2002). In superman (sup) mutant ovules, there is increased OI growth on the gynoapical side of the ovule relative to wild-type (Gaiser et al., 1995). This correlates with an increase in INO expression in this region in such mutants (Meister et al., 2002). The correlation between developmental evidence of quantitative effects of INO expression and our mutant studies linking quantity of INO expression and amount of OI growth provides robust evidence that the level of INO expression is a primary determinant of OI growth.

Conclusion

The phenotypes of the null ino mutants demonstrate that this gene's role is specific to OI development in Arabidopsis. The specificity of the effects is mirrored in the confinement of INO expression to the outer (abaxial) layer of cells of the OI (Balasubramanian & Schneitz, 2002; Meister et al., 2002; Villanueva et al., 1999). This pattern of expression is conserved for INO orthologs among all examined angiosperms including other eudicots such as Impatiens and tomato (McAbee et al., 2005; Skinner et al., 2016), members of earlier diverging groups such as magnoliids in the Annona genus (Lora et al., 2011), and the very earliest diverging angiosperms including Cabomba and Amborella (Arnault et al., 2018; Yamada et al., 2003). Functional analysis of INO activity has been evaluated through virus-induced gene silencing in tobacco (Skinner et al., 2016) and through analysis of an INO deletion mutant in Annona (Lora et al., 2011) where effects similar to effects on Arabidopsis ovules were observed. Together, these results indicate that INO has had a primary role as a regulator of OI development from the inception of the angiosperms. The OI is a synapomorphy of angiosperms and is one of the defining features separating angiosperms from gymnospermous progenitors (Gasser & Skinner, 2019; Shi et al., 2021). Phylogenetic analysis of the YABBY gene family shows that the origin of the INO clade correlates with the origin of the angiosperms (Gasser & Skinner, 2019; Yamada et al., 2011). Thus, the origin of this critical or master gene for OI development (as shown by developmental and genetic analysis) may have been an essential step in the evolution of the OI and the origin of angiosperms (Gasser & Skinner, 2019; Shi et al., 2021).

AUTHOR CONTRIBUTIONS

Work was planned and conceived by Debra J. Skinner and Charles S. Gasser. Debra J. Skinner performed fast neutron mutagenesis and analysis. Charles S. Gasser and Trang Dang performed CRISPR mutagenesis and analysis. Charles S. Gasser wrote the manuscript with editing by Debra J. Skinner. All authors approved the manuscript.

ACKNOWLEDGMENTS

We thank Mily Ron and Anne Britt (Dept. of Plant Biology, University of California, Davis) for the gift of vectors and protocols for CRISPR mutagenesis; Judy Callis for aid in statistical analysis and advice on microscopy; and Selina Rodrigues, Jennifer Zonka, and George Day for excellent technical assistance.

CONFLICT OF INTEREST STATEMENT

The authors did not report any conflict of interest.

DATA AVAILABILITY STATEMENT

Newly derived plant lines have been deposited at the Arabidopsis Biological Resource Center at Ohio State, and other plant materials are available from the cited sources from which they were obtained. Raw genomic sequence data have been deposited in the NCBI SRA database under accession PRJNA866995.