Unless otherwise stated, all Arabidopsis thaliana (L.) Heynh. plants studied in this work were homozygous for the mutations indicated. The Nottingham Arabidopsis Stock Centre (NASC) provided seeds for the wild-type accessions Columbia-0 (Col-0; N1092), S96 (N914), and Wassilewskija-2 (Ws-2; N1601), as well as the following mutants: icu11-1 (N242) in the S96 background; curly leaf-2 (clf-2; N8853) in the Landsberg erecta (Ler) background; arabidopsis trithorax1-2 (atx1-2; N649002), arabidopsis trithorax-related protein 5 (atxr5; N630607), atxr6 (N866134), atxr7-1 (N667600), cp2-1 (N861581), cp2-2 (N828642), cp2-3 (N826626), demeter-like 2-3 (dml2-3; N631712), dml3-1 (N556440), dna methyltransferase-2-2 (dnmt2-2; N836854), domains rearranged methylase 1-2 (drm1-2; N521316), drm2-2 (N650863), histone acetyltransferase of the cbp family 1-3 (hac1-3; N580380), histone acetyltransferase of the myst family 1-1 (ham1-1; N655396), methyltransferase 1-4 (met1-4; N836155), repressor of silencing 1-4 (ros1-4; N682295), ros3-2 (N522363), terminal flower2-2 (tfl2-2; N3797), and embryonic flower2-3 (emf2-3; N16240) in the Col-0 background; icu2-1 (N329) and fasciata1-1 (fas1-1; N265) in the Enkheim2 (En-2) background; methyl-cpg-binding domain10-1 (mbd10-1; N872244) and variant in methylation 3-2 (vim3-2; N804664) in the Col-3 background; and histone deacetylase 6-6 (hda6-6; N66153) and hda6-7 (N66154) in the Col background. Seeds for early bolting in short days-1 (ebs-1, in the Ler background; Piñeiro et al., 2003) were provided by Manuel Piñeiro (CBGP, UPM-INIA-CSIC, Madrid, Spain), those of gigantea supressor5 (gis5, in the Col-0 background; Iglesias et al., 2015) by Pablo D. Cerdán (Fundación Instituto Leloir, IIBBA-CONICET, Buenos Aires, Argentina), and those of icu11-2 (in the Ws-2 background) by the Versailles Arabidopsis Stock Center (Brunaud et al., 2002). The presence and positions of all T-DNA insertions were confirmed by PCR amplification using gene-specific primers, with the LbB1.3 and LB1 primers used for the SALK and SAIL T-DNA insertions, respectively ( Supplementary Table S1 ).

Unless otherwise stated, all seeds were surface sterilized, plated onto 140-mm (diameter) Petri dishes containing 100 ml half-strength Murashige and Skoog (MS) plant agar medium with 1% (w/v) sucrose at 20 ± 1°C, 60-70% relative humidity, and continuous illumination at ~75 µmol m^–2 s^–1, as previously described (Ponce et al., 1998). Crosses were performed as previously described (Quesada et al., 2000).

Photographs showing morphology were taken with a Nikon SMZ1500 stereomicroscope equipped with a Nikon DXM1200F digital camera and the ACT-1 software (Nikon). Pollen grains were stained with Alexander red solution for 5–10 min before observation and photographed using a Leica DMRB microscope equipped with a Nikon DXM1200 digital camera.

The pKI1.1R-ICU11_sgRNA1 plasmid was constructed as described by Tsutsui and Higashiyama (2017). In brief, the pKI1.1R plasmid (85808; Addgene) was linearized by restriction digest with AarI (Thermo Fisher Scientific), and treated with FastAP alkaline phosphatase (Thermo Fisher Scientific). The ICU11_sgRNA1_F/R oligonucleotides were phosphorylated using T4 polynucleotide kinase (New England Biolabs) and hybridized in a thermal cycler (Bio-Rad Laboratories T100). The ligation reaction was performed with T4 DNA ligase (Thermo Fisher Scientific), and the ligation product was transformed into chemically competent Escherichia coli DH5α cells using the heat-shock method. Plasmid and insert integrity were verified by Sanger sequencing using an Applied Biosystems 3500 Genetic Analyzer (Thermo Fisher Scientific). Putative off-targets were identified using the default parameters of the ChopChop tool (Labun et al., 2019). The pKI1.1R-ICU11_sgRNA1 plasmid was mobilized into Agrobacterium tumefaciens GV3101 (C58C1 Rif^R) cells, which were used to transform Arabidopsis S96 and Col-0 plants via the floral dip method (Clough and Bent, 1998). T₁ Arabidopsis transgenic plants were selected on plates with half-strength MS medium containing 15 mg l^–1 hygromycin B (Thermo Fisher Scientific). Absence of the transgene in T₂ plants was verified by selecting DsRED negative seeds and confirmed by negative PCR amplifications using Cas9 specific primers ( Supplementary Table S1 ).

Flowering time was determined based on the total number of rosette leaves (counted when internode elongation was visible) and the number of days to bolting (Bouveret et al., 2006). To determine flowering time, all plants were grown on half-strength MS medium for five days and transferred to soil (a 2:2:1 mixture of perlite, vermiculite and sphagnum peat moss) in individual pots in a TC30 growth chamber (Conviron).

Sequence data from this article can be found at The Arabidopsis Information Resource (http://www.arabidopsis.org) under the following accession numbers: ICU11 (At1g22950), CP2 (At3g18210), EBS (At4g22140), FAS1 (At1g65470), GIS5 (At5g63960), ICU2 (At5g67100), CLF (At2g23380), TFL2 (At5g17690), EMF2 (At5g51230), DML2 (At3g10010), DML3 (At4g34060), DNMT2 (At5g25480), DRM1 (At5g15380), DRM2 (At5g14620), MBD10 (At1g15340), MET1 (At5g49160), ROS1 (At2g36490), ROS3 (At5g58130), VIM3 (At5g39550), ATX1 (At2g31650), ATXR5 (At5g09790), ATXR6 (At5g24330), ATXR7 (At5g42400), HAC1 (At1g79000), HDA6 (At5g63110), and HAM1 (At5g64610).

We previously characterized two loss-of-function icu11 alleles: icu11-1 and icu11-2 (Mateo-Bonmatí et al., 2018). A third allele, icu11-3, was identified in an Ac/Ds transposon-tagging mutagenesis screen (Bancroft et al., 1993; Bloomer et al., 2020). Although these three mutants appear to carry null alleles ( Figure 1 ), the hyponasty of icu11-1 leaves is stronger than that caused by the icu11-2 and icu11-3 alleles, which is likely due to their different genetic backgrounds (S96, Ws-2, and Ler, respectively). Different or hybrid genetic backgrounds may not facilitate proper comparisons of the morphological and molecular phenotypes of single and multiple mutants (Page and Grossniklaus, 2002; Chandler et al., 2013; Taylor and Ehrenreich, 2015). Therefore, as all cp2 alleles were in the Col-0 background but there were no available icu11 alleles in this background, we subjected Col-0 plants to CRISPR/Cas9 mutagenesis targeting ICU11. We also mutagenized S96 plants as a control. Accordingly, we designed a 20-nt single guide RNA (sgRNA) that targets the first exon of the ICU11 gene, with a predicted targeting efficiency of 49.7% ( Figures 1A, B ; Supplementary Figure S1A ). We obtained seven chimeric T₁ plants, from which we isolated four T₂ independent homozygous lines: the icu11-4 and icu11-7 mutants in the S96 background, carrying deletions of 4 bp and 1 bp, respectively, and icu11-5 and icu11-6 in the Col-0 background, carrying a 1-bp insertion and a 1-bp deletion, respectively ( Figure 1C ; Supplementary Figure S1B ). All these mutations are predicted to cause frameshifts that introduce premature stop codons, producing truncated proteins of only 86 (icu11-4), 44 (icu11-5), or 87 (icu11-6 and icu11-7) amino acids, instead of the 397 amino acids of wild-type ICU11 ( Supplementary Figure S1C ). We sequenced Cas9-free lines by Sanger sequencing to examine the two most likely off-targets, both of which presented four mismatches with at least one mismatch located in the 5 bp adjacent to the protospacer adjacent motif (PAM) of our sgRNA ( Supplementary Figure S2A ); neither off-target was mutated in subsequent generations of our mutant lines ( Supplementary Figure S2B ).

As expected from their S96 background, the icu11-4 and icu11-7 mutants exhibited a morphological phenotype indistinguishable from that of icu11-1 ( Figures 1D–G ). The icu11-5 and icu11-6 mutants showed wavy leaves that did not reach hyponasty ( Figures 1H–J ), different to that of icu11-2 ( Figures 1K, L ) and icu11-3 (Bloomer et al., 2020). In addition, icu11-5 and icu11-6 shared other phenotypic traits with the other icu11 mutants, such as cotyledon epinasty ( Figures 1D–L ) and early flowering ( Figures 2A, B ). We observed significantly more unfertilized ovules per half silique in the icu11-4, icu11-5, icu11-6 and icu11-7 mutants compared to their respective wild types ( Figures 2C–L ). Taken together, these observations indicate that our new icu11 mutants are phenotypically similar to previously reported mutants, with only minor differences caused by their genetic background, which are particularly visible in their rosette leaf morphology.

Although the morphological phenotypes caused by the icu11 alleles are relatively mild, and the cp2 null mutants are indistinguishable from wild type, the phenotypes of icu11 cp2 double mutants are synergistic: they are seedling lethal, as might be expected for the genetic combination of mutations in two close paralogs with a high degree of functional redundancy (Ohno, 1970; Nowak et al., 1997; Cusack et al., 2021). Our previously obtained icu11 cp2 double mutants had hybrid genetic backgrounds, which prevented a clear conclusion as to the seedling lethality presented by these double mutants. Here, we thus crossed icu11-5 and icu11-6 with the cp2-1 and cp2-2 hypomorphic alleles of CP2, and with the cp2-3 null allele, all of which are in the Col-0 genetic background. The double homozygous mutant combinations between icu11-5 or icu11-6 and cp2-1 or cp2-2 exhibited an embryonic-flowering seedling-lethal phenotype, as did the icu11/icu11;CP2/cp2-3 sesquimutants ( Figure 3 ). We obtained no icu11-5 cp2-3 or icu11-6 cp2-3 double mutants, as was previously published for icu11-1 cp2-3 (Mateo-Bonmatí et al., 2018). The ICU11/icu11;cp2-3/cp2-3 sesquimutants were indistinguishable from the wild type ( Figures 3I, N ), as were those we previously published in hybrid genetic backgrounds.

Synergistic phenotypes visualized in double mutants shed light on functional relationships between genes (Pérez-Pérez et al., 2009), including those encoding components of the epigenetic machinery of Arabidopsis. For example, PWWP-DOMAIN INTERACTOR OF POLYCOMB1 (PWO1) is a histone reader that recruits PcG proteins (Hohenstatt et al., 2018), and BLISTER (BLI) is a PRC2 interactor and a regulator of stress-responsive genes (Kleinmanns et al., 2017). Both pwo1 and bli loss-of-function mutations display synergistic phenotypes when combined with strong mutant alleles of CURLY LEAF (CLF), which encodes a PRC2 core component responsible for the deposition of H3K27me3 repressive marks (Goodrich et al., 1997; Schatlowski et al., 2010).

With the aim to expand the spectrum of genes demonstrated to genetically interact with ICU11, we crossed icu11-1 to loss-of-function mutants of 17 genes encoding components of the epigenetic machinery that include proteins involved in DNA or histone methylation, acetylation, or deacetylation (Pikaard and Mittelsten Scheid, 2014). We identified 18 double mutants with additive phenotypes ( Supplementary Table S2 ). Also we obtained double mutant combinations of icu11-5 with loss-of-function alleles of CLF, LIKE HETEROCHROMATIN PROTEIN 1 (LHP1; also named TFL2), FASCIATA1 (FAS1), EARLY BOLTING IN SHORT DAYS (EBS), GIGANTEA SUPRESSOR5 (GIS5), and ICU2, previously found to genetically interact with icu11-1 (Mateo-Bonmatí et al., 2018). FAS1 is a component of the Chromatin Assembly Factor 1 (CAF-1) complex that promotes the deposition of histone H3 and H4 at newly synthesized DNA during replication (Schönrock et al., 2006). EBS is an H3K27me3 and H3K4me3 reader that regulates the floral phase transition (Piñeiro et al., 2003; López-González et al., 2014; Yang et al., 2018). GIS5 and ICU2 are the catalytic subunits of DNA polymerase δ and α, respectively (Barrero et al., 2007; Iglesias et al., 2015). The double mutant combinations of icu11-5 with clf-2, ebs-1, gis5, icu2-1, tfl2-2, or fas1-1 exhibited strong synergistic phenotypes consisting of dwarf rosettes; extreme leaf hyponasty in the cases of gis5, icu2-1, and fas1-1; and some degree of anthocyanin accumulation in the icu11-5 tfl2-2 rosette center ( Figure 4 ). The phenotypes of the double mutant combinations of icu11-5 with clf-2, gis5, and icu2-1 were similar to those previously reported using icu11-1; however, those involving tfl2-2, ebs-1, and fas1-1 were milder in combination with icu11-5 ( Figures 4L–N ) than with icu11-1 (Mateo-Bonmatí et al., 2018).

We also combined the mutations mentioned above with the cp2-3 mutation, which lacks a distinctive phenotype; the resulting F₂ progeny showed rosette phenotypes that were indistinguishable from those of their corresponding phenotypically mutant parent ( Figures 4P–U ). We conclude that, unlike ICU11, the loss of CP2 function is not sufficient to modify the phenotypes caused by mutations in other genes involved in epigenetic modifications, confirming the previously proposed unequal functional redundancy between ICU11 and CP2 (Mateo-Bonmatí et al., 2018).

There is an expanding list of Arabidopsis mutants exhibiting morphological phenotypes that are partially or fully rescued by the exogenous supplementation of sucrose, such as phosphatidylglycerol phosphate synthase 1 (pgp1) and cyclophilin 38 (cyp38), which are defective in the assembly and proper function of photosystem II protein complexes, respectively (Fu et al., 2007; Kobayashi et al., 2016; Duan et al., 2021). This phenomenon is true for other Arabidopsis genes that are functionally unrelated, but whose mutations directly or indirectly impair or abolish photosynthesis. We observed the same effect in icu11 cp2-1 and icu11 cp2-2 double mutant seedlings, in which photosynthesis is diminished because they do not form true leaves, instead developing embryonic flowers immediately after germination. These mutants had a very slow growth rate, did not develop further, and died 20–40 days after stratification (das; Figures 5A–C ). When we raised the concentration of sucrose in the growth medium from 1% (w/v) to 3%, icu11-5 cp2-1 double mutant plants developed main and axillary shoots with long internodes, small cauline leaves, disorganized flowers, and short siliques. Most of the disorganized flowers showed homeotic transformations of sepals and petals into carpels and had few stamens ( Figures 5D–H ).

The emf2-3 single mutant (Yoshida et al., 2001), whose morphological phenotype is similar to that of icu11 cp2 double mutant plants, also exhibited a partial rescue under increased sucrose supplementation. These plants growing on 3% sucrose medium produced main and axillary shoots lacking apical dominance, as well as flowers with an altered structure that developed into short siliques with extended white sectors reminiscent of petal tissue ( Figures 5I–K ). When grown on growth medium supplemented with 3% sucrose, 36.5% of icu11-5 cp2-1 (n = 136) and 34.1% of emf2-3 (n = 110) plants developed stems with flowers, although even with hand pollination only 15.3% of icu11-5 cp2-1 plants and no emf2-3 plants produced mature seeds ( Figure 5L ). Only one out of the 23 seeds obtained in this manner from icu11-5 cp2-1 plants germinated on medium supplemented with 3% sucrose, and still developed embryonic flowers.

An alteration in pollen viability may explain why only some icu11-5 cp2-1 double mutant plants produced seeds. To assess pollen grain viability in the icu11-5, cp2-1, and icu11-5 cp2-1 mutants, we stained their anthers with Alexander solution. We detected no aberrations in anther shape or size or in pollen grain viability for the cp2-1 and icu11-5 anthers when compared to Col-0 ( Figures 6A–F ). However, 66% of the observed icu11-5 cp2-1 anthers were smaller and carried fewer but viable pollen grains ( Figures 6G, H, K ). The remaining 34% of anthers were also small and contained only non-viable pollen that turned blue/gray upon staining ( Figures 6G–K ).

A common task in developmental genetics is the analysis of phenotypic differences between mutants carrying different alleles of a given gene, as well as the comparison of their corresponding double mutants. Many studies have shown that morphological phenotypes can vary depending on the genetic background; for example, the gibberellin biosynthesis (ga5) mutant in Ler displays a much more pronounced drop in shoot fresh weight than the ga20ox1-3 (another loss-of-function allele of the GA5 gene) mutant, in the Col-0 background (Barboza-Barquero et al., 2015). Such phenotypic difference can be due to allele specifity, the erecta mutation carried by Ler, or to any other of the many differences between the genetic backgrounds of Ler and Col-0. Another example is the transfer of transgenes carrying either the functional Ler allele or the loss-of-function allele from the Japanese accession Fuk of the MADS box transcription factor gene SHORT VEGETATIVE PHASE (SVP), which regulates flowering time, into five Arabidopsis accessions. In some of these accessions, the presence of SVP-Fuk accelerated flowering, whereas SVP-Ler delayed it; however, in other accessions, no noticeable differences were observed (Méndez-Vigo et al., 2013). The genetic modifiers partially or fully responsible for genetic background effects have been identified in a few cases. For example, the short stem and midrib (ssm) mutant in Col-0 is semi-dwarf and has wavy leaves; in the progeny of crosses of Ler to ssm, such phenotype was rescued by a gene apparently present in Ler but absent from Col-0. ssm was found to be a null allele of SYNTAXIN OF PLANTS 22 (SYP22), which encodes a syntaxin-related protein required for vacuolar assembly. The Ler allele of SYP23, a close paralog of SYP22, is functional and complements ssm, whereas the Col-0 allele of SYP23 is mutated (Ohtomo et al., 2005).

The icu11-1 mutant exhibits a developmental phenotype previously observed in other mutants that carry alleles of genes encoding components of the epigenetic machinery (Mateo-Bonmatí et al., 2018). These characteristic aberrations, such as leaf hyponasty and early flowering, are in some cases associated with a reduced deposition of the epigenetic mark H3K27me3 across a large number of genes (Förderer et al., 2016). The gene-edited icu11-5 and icu11-6 null alleles of ICU11 that we obtained in the Col-0 genetic background in this study exhibited a leaf incurvature milder than that of icu11-1 in the S96 background. These differences are most likely due to differences in the genetic backgrounds, as we also obtained two CRISPR/Cas9 mutants in the S96 background, icu11-4 and icu11-7, which displayed identical leaf curvature as icu11-1. Furthermore, it is worth mentioning here that icu11-6 and icu11-7 carry independently obtained but identical mutations: a deletion of one nucleotide immediately downstream of the PAM; therefore, their differences in leaf phenotype can only be due to their different genetic backgrounds.

CP2 and ICU11 are unequally redundant paralogs, as we inferred from the phenotype of the icu11-1/icu11-1;CP2/cp2-3 sesquimutant, which develops lethal embryonic flowers, while the reciprocal ICU11/icu11-1;cp2-3/cp2-3 sesquimutant is phenotypically wild type (Mateo-Bonmatí et al., 2018). The icu11-1 cp2-1 and icu11-1 cp2-2 double mutants also developed embryonic flowers, whereas the ICU11/icu11-1;cp2-1/cp2-1 and ICU11/icu11-1;cp2-2/cp2-2 sesquimutants were indistinguishable from wild type plants, and the icu11-1/icu11-1;CP2/cp2-1 and icu11-1/icu11-1;CP2/cp2-2 sesquimutants were indistinguishable from icu11-1 single mutant plants. While the genetic background of these double mutants and sesquimutants was hybrid (S96/Col-0), all combinations between the icu11-5, icu11-6, cp2-1, cp2-2, and cp2-3 mutations that we obtained here were in a single genetic background (Col-0). Furthermore, as previously shown for icu11-1/icu11-1;cp2-3/cp2-3, no icu11-5/icu11-5;cp2-3/cp2-3 or icu11-6/icu11-6;cp2-3/cp2-3 double mutants were obtained likely because of their gametic or early embryonic mortality. The genetic combination of our new icu11 alleles with cp2 alleles, all in the Col-0 background, confirmed that CP2 behaves as a haploinsufficient locus in a homozygous icu11 background (Pérez-Pérez et al., 2009; Meinke, 2013; Navarro-Quiles et al., 2023), and in particular, that the embryonic flower phenotype of the double mutant and some sesquimutant combinations of their alleles are not influenced by the S96, Ws-2, or Col-0 genetic backgrounds.

We also provided further evidence for the unequal functional redundancy between ICU11 and CP2 through the analysis of their genetic interactions with loss-of-function alleles of other epigenetic machinery components. We established that icu11-5 synergistically interacts with clf-2, ebs-1, gis5, icu2-1, tfl2-2, and fas1-1 in the presence of two wild-type CP2 copies, as previously reported for icu11-1. There were some minor differences between the synergistic phenotypes of the double mutant combinations of ebs-1 (in the Ler background), tfl2-2 (Col-0), and fas1-1 (En-2) with icu11-1 (S96) or icu11-5 (Col-0), which can be attributed to the genetic background. By contrast, the double mutant combinations of ebs-1, tfl2-2, or fas1-1 with the cp2-3 null allele, in the presence of two ICU11 wild-type copies, resulted in phenotypes indistinguishable from those of the ebs-1, tfl2-2, or fas1-1 single mutants. Apparently, the presence of a wild-type allele of ICU11 impedes the identification of CP2 genetic interactors, as they are redundant. Thus, the lethality of the icu11 cp2 double mutants is an obstacle for an independent characterization of CP2.

The phenotype of a large number of mutants in genes that are primarily associated with photosynthesis or processes closely linked to it can be partially or completely alleviated by supplementation with an exogenous carbon source. Indeed, photosynthesis, primarily occurring in plant leaves, serves as the ultimate source of sugars, which are the primary carriers of both sunlight energy and carbon required for metabolism. Sugar availability is crucial for Arabidopsis development, leading to early developmental arrest in mutants with impaired photosynthesis, such as pgp1 (Kobayashi et al., 2016), and fructokinase-like 2-4 (fln2-4; Huang et al., 2013) or white cotyledons (wco; Yamamoto et al., 2000). Since WCO is only required for chloroplast biogenesis in cotyledons but not in true leaves, wco mutants can be supplied with 3% sucrose for a few days to enable them to survive and produce true leaves, after which the plants develop normally. Other mutations of genes involved in photosynthesis are not seedling or plant lethal but delay growth; for example, CYP38 participates in the assembly and maintenance of photosystem II, and loss-of-function cyp38 alleles show retarded growth and pale green leaves. A 1% sucrose supplementation was sufficient to rescue rosette growth, although cyp38 still displayed hypersensitivity to light, while higher concentrations of sucrose increased the length of the primary root (Fu et al., 2007; Duan et al., 2021).

Although several genes repress flowering in Arabidopsis, the emf single mutants are unique because they completely skip the vegetative phase after seed germination without generating true leaves (Sung et al., 1992; Yang et al., 1995). The icu11 cp2 double mutants resembled the emf single mutants, as they also lacked true leaves. Impaired carbon fixation results in the early-development arrest of these mutants, so they cannot produce reproductive structures beyond a disorganized embryonic flower. These mutants represent a class completely different from those mentioned above, as they are not mutated in genes primarily related to photosynthesis. Instead, they skip vegetative development and die because they cannot develop leaves, which does not allow them to photosynthesize properly.

The increase from 1% to 3% sucrose in the growth medium was sufficient to allow the formation in our icu11 cp2 double mutants of axillary shoots, cauline leaves, and disorganized inflorescences exhibiting flowers with homeotic transformations. Analysis of these structures revealed that ICU11 and CP2 are not only required for vegetative development, as previously described (Mateo-Bonmatí et al., 2018), but also to ensure proper reproductive development. ICU11 and CP2 are expressed in both vegetative and reproductive organs, with CP2 showing higher expression levels than ICU11 in the flowers and siliques of wild-type plants (Mateo-Bonmatí et al., 2018).

Similar to the homeotic transformations that we observed in icu11-5 cp2-1 plants grown with 3% sucrose, the loss-of-function apetala2 (ap2) mutations and the overexpression of AGAMOUS (AG) and SEPALLATA3 (SEP3) result in the transformation of sepals and petals into carpelloid structures (Mizukami and Ma, 1992; Riechmann and Meyerowitz, 1997; Chen, 2004; Castillejo et al., 2005). Moreover, the weak emf2-10 allele causes the appearance of carpelloid sepals (Chanvivattana et al., 2004) reminiscent of those of the icu11-5 cp2-1 double mutant. Indeed, previous studies have described that ICU11 plays a role in repressing AG, SEP1, SEP2, and SEP3 expression during the vegetative phase (Mateo-Bonmatí et al., 2018; Bloomer et al., 2020). Therefore, based on our observations of icu11-5 cp2-1, we propose that ICU11 and CP2 are required to regulate the expression of floral identity genes during reproductive development.

The homeotic transformations of the flowers produced by icu11-5 cp2-1 and emf2-3 plants grown on medium supplemented with 3% sucrose impeded self-pollination. Hand pollination of the icu11-5 cp2-1 double mutant, but not the emf2-3 single mutant, resulted in a few seeds, only one of which germinated and developed an embryonic flower when grown on growth medium containing 3% sucrose. In this growth condition, icu11-5 cp2-1 anthers were smaller than those of Col-0, and only 64% contained small but viable pollen grains. When grown under short-day conditions, the mutant emf2-3 also develops a small shoot with two or three flowers and short siliques that did not produce mature seeds (Chen et al., 1997).

The moderate reduction in pollen viability that we observed by Alexander staining does not explain the extremely low production and viability of icu11-5 cp2-1 seeds. This clearly suggests the existence of additional problems with either ovule formation, fertilization or embryonic development. Further research would be needed to determine the causes of icu11-5 cp2-1 reduced fertility.