The genome of the dicotyledonous model plant Arabidopsis thaliana can be mutagenized using transfer DNA (T-DNA) of the soil bacterium Agrobacteria tumefaciens. Plant genome engineering techniques such as these predate CRISPR-Cas9 technology by a quarter of a century (Cong et al., 2013; Feldmann and Marks, 1987; Feldmann et al., 1989; Jinek et al., 2012; Kwon, 2016), and have enabled researchers to link mutant phenotype to genotype, contributing a powerful and comprehensive catalog of mutations, a tool to plant biologists previously not available in other organisms (O’Malley and Ecker, 2010). Currently more than 700,000 mutant lines with gene affecting insertions have been generated, representing potential disruption mutants for most of the 27,000 Arabidopsis genes (Jupe et al., 2019).

The process of T-DNA excision from the tumor inducing (Ti) plasmid and passage through the Agrobacterial membrane is well established; however, the exact mechanism of T-DNA integration into the host genome is not well understood. It is known that multiple T-DNA insertions as direct or inverted repeats occurs frequently during integration, and that this multiple repeat structure often results in intra- and inter-chromosomal rearrangements (Errampalli et al., 1991; Nacry et al., 1998; Thomashow et al., 1980; Ulker et al., 2008; Zambryski et al., 1982; Zhu et al., 2006). In directed T-DNA mutant screens, approximately 17% of the insertion mutants were found to contain chromosomal rearrangements (Castle et al., 1993), although they are often aphenotypic and thus not generally detected. However, there are a number of scenarios whereby T-DNA mediated chromosomal rearrangements can obscure or bias mutant inheritance data, in which case they must be taken into account. For example, intra-chromosomal rearrangements that result in the inversion of chromosomal fragments normally lead to recombination suppression during homologous chromosomal pairing in meiosis. Inter-chromosomal rearrangement results in translocation of the chromosomal fragment and distortion of the linkage group (Curtis et al., 2009). If mutant loci are linked to the rearranged portions of chromosomes, non-Mendelian distribution of segregating populations can be observed during genetic crosses (Clark and Krysan, 2010).

ASF1 is an evolutionarily conserved histone chaperone of the H3/H4 family involved in several fundamental cellular processes, including nucleosome rearrangement and DNA damage repair (English et al., 2005; Le et al., 1997; Mousson et al., 2005; Schulz and Tyler, 2006; Singer et al., 1998). Two Arabidopsis homologues, ASF1A and ASF1B, are associated with T-DNA mutant lines asf1a (CS393977) and asf1b (Salk_105822C) that are loss-of function alleles (Zhu et al., 2011). Single homozygous mutants do not have a phenotype, but ASF1 homologues act redundantly, whereby double homozygotes show a drastic inhibition of vegetative and gametophytic growth (Min et al., 2019; Zhu et al., 2011). Here, we describe our observations of mutants that are heterozygous for the asf1b T-DNA insertion. We found completely random, non-Mendelian distributions of F2 progeny from selfed F1 double heterozygous plants (ASF1A/asf1a ASF1B/asf1b and asf1a ASF1B/asf1b, hereafter AaBb and aaBb, respectively). Our cytological and genetic data showed that inheritance of the mutant allele in the gamete did not determine whether seeds were viable or not. Further, the AaBb and aaBb abortion phenotypes were not rescued by expression of functional ASF1. Taken together, our results suggest that the gametophytic abortion observed in heterozygotes, but not in homozygotes, is most likely from a T-DNA mediated chromosome rearrangement that includes the ASF1B locus.

Plant material and growth conditions

asf1a and asf1b alleles have been described previously (Min et al., 2019). We obtained T-DNA insertion mutants for ASF1A and ASF1B from the Arabidopsis Biological Resource Center (ABRC): asf1a-2 (CS393977), asf1b-1 (Salk_105822C). All asf1 double mutants (AaBb, aaBb, Aabb, aabb) were obtained by crossing the two single mutants and allowing self-pollination. All genotypes were determined by polymerase chain reaction (PCR) analysis and plants were grown in long-day conditions (16-h light/8-h dark) at 22°C with 60% relative humidity.

Seed-set analysis and whole-mount clearing

Seeds, and any undeveloped ovules, were dissected and the number of each counted on a Leica stereoscopic microscope (Leica Microsystems, Germany). For whole mount clearing, pistils of mature FG7 stage and developing siliques were dissected and mounted in a clearing solution containing chloral hydrate in distilled water. Cleared ovules and seeds were observed using an Axio Imager A1 microscope (Carl Zeiss, Germany) under DIC optics and photographed using an AxioCam HRc camera (Carl Zeiss) as previously described (Min et al., 2019).

Analysis of GUS and GFP expression

GUS histochemical analysis was performed as previously described (Yadegari et al., 2000). Female gametophytes from plants expressing GFP fluorescence were analyzed using an LSM700 (Carl Zeiss) confocal laser microscope.

Confocal laser scanning microscopic analysis

For the analysis of embryo sac development in wild type and asf1 mutants, confocal laser scanning microscopic (CLSM) analysis of ovules was performed as previously described (Min et al., 2019).

Pollen viability analysis using Alexander’s staining and DAPI staining

For analysis of pollen viability, pollen grains were mounted with Alexander’s staining solution as previously described (Park et al., 2014). To visualize nuclei in pollen grains, DAPI (4′,6-diamidino-2-phenylindole) staining was performed. The Alexander-stained slides and DAPI fluorescence were examined using an Axio Imager A1 microscope under DIC optics and were photographed using an AxioCam HRc camera.

ASF1B/asf1b heterozygosity causes defects during reproduction

The two Arabidopsis thaliana ASF1 homologues, ASF1 (At1g66740) and ASF1B (At5g38110), play redundant roles in replication-dependent chromatin assembly during vegetative and gametophytic growth (Min et al., 2019; Zhu et al., 2011). By crossing the single mutants, asf1a and asf1b (Supplementary Fig. S1), which do not display developmental defects, AaBb double heterozygous plants were generated and then self-pollinated. The F2 segregating population showed distorted Mendelian segregation (Supplementary Table S1), and the AaBb and aaBb double mutants displayed distinctly reduced fertility compared to any of the other genotypes (Fig. 1A, Table 1). aaBb mutants exhibited 47% seed viability (n = 1,351) (Table 1). Intriguingly, however, self-fertilized AaBb double heterozygous plants also showed 52% seed viability (n = 2,009) (Table 1). Approximately half of the ovules from AaBb and aaBb mutants aborted (47% and 50%, respectively) compared to 8% ovules aborted from Aabb plants. This result indicates that, intriguingly, Bb heterozygosity is more detrimental to seed viability than the bb homozygous state.

Mitotic defects and a failure of proper cell identity during female gametogenesis were observed in aaBb and AaBb mutants

In aaBb and AaBb mutants, approximately half of the ovules developed only a short embryo sac and subsequently aborted (Fig. 1B, Table 1). We introduced cell-specific markers (DD45::GFP for the egg cell and DD7::GFP for the central cell; Steffen et al., 2007) into aaBb and AaBb mutant plants to ascertain the identity of the cells present within normal and aborting embryo sacs. Normal embryo sacs demonstrated appropriate fluorescence to indicate the correct egg and central cell identities (Figs. 2A-2D); however, arrested ovules did not (Figs. 2E-2H). To determine at which developmental stage the aberrant ovules had aborted, we introduced into aaBb and AaBb mutants a marker for functional megaspores, the FM2::GUS transgene (Olmedo-Monfil et al., 2010). FM2::GUS was not expressed in normal embryo sacs (Figs. 2I and 2J), but was expressed in arrested ovules (Figs. 2K and 2L), indicating that arrested embryo sacs have functional megaspore identity. Hence, meiosis occurred to produce functional haploid megaspores, but the subsequent haploid mitotic divisions, so called megagametogenesis, to produce functional gametes, were not completely processed, thereby a developing embryo sac was impaired in aaBb and AaBb mutants.

To delineate the cause of ovule arrest in aaBb and AaBb mutants, we analyzed the development of the female gametophyte using confocal microscopy. Consistent with FM2:GUS marker gene expression, no discernible differences during meiotic division were observed within the aaBb and AaBb mutant ovules. Megagametogenesis of wild-type plants proceeds from FG1 to the FG7 mature female gametophytic stage (Christensen et al., 1997) (Figs. 3A-3C). In contrast, aaBb and AaBb mutants universally failed to undergo three rounds of haploid mitotic division of the megaspore, resulting in either a single nucleus (no division, Fig. 3E) or two nuclei (one mitotic division; Fig. 3F) inside the short embryo sac. Even though we observed ovules that had reached the FG3 stage (two nuclei and a small vacuole), these did not develop further (Fig. 3F). This phenotype is distinct from that we observed in aabb mutants previously, which was shown to be due to the loss of both ASF1A and ASF1B (Min et al., 2019). In these plants, the length of aabb embryo sacs were comparable to normal wild-type embryo sacs; therefore, haploid mitotic divisions were occurring, but the subsequent maturation and acquisition of cell identity were defective, resulting in a collapsed embryo sac without visible nuclei in about 30% of the aabb ovules (Min et al., 2019). In contrast, ovule arrest in aaBb and AaBb mutants occurs much earlier than in aabb mutants. Taken together, our data indicate that megasporogenesis occurs normally in aaBb and AaBb mutants, but subsequent mitotic divisions do not proceed beyond the FG3 stage in aaBb and AaBb mutant ovules.

aaBb and AaBb mutants exhibit mitotic defects during male gametogenesis

ASF1 was shown previously to not affect pollen viability, as demonstrated in double aabb mutants (Min et al., 2019). Most aabb mature pollen grains reached the tri-cellular stage, although subsequent pollen tube elongation was shown to be defective (Min et al., 2019). However, in aaBb mutant anthers both viable (purple) and aborted (green) pollen were present when analysed using Alexander’s staining (Fig. 1C). In constrast, the viability of Aabb pollen was comparable to that of wild-type (Fig. 1C). Similar to our observations in the female gametophyte, the AaBb double heterozygote phenotype was identical to aaBb, with both viable and aborted pollen present (Fig. 1C), leading us to speculate the mutant b allele in the homozygous state, as in Aabb mutants, somehow alleviates the phenotype observed in AaBb or aaBb mutants (Table 1).

To analyze at what stage aaBb and AaBb pollen arrested, DAPI (4’,6-diamidino-2-phenylindole) staining was performed. Virtually all mature wild-type pollen was at the tri-cellular stage with fluorescently stained DNA contained within two sperm nuclei (Fig. 4). However, only half of the pollen from aaBb and AaBb plants had reached the tri-cellular stage, the rest had shrunk (Fig. 4, asterisks). We determined that although no obvious defects were observed from the diploid pollen mother cell (PMC) to microspore stages prior to pollen mitosis I (PMI), differences between wild type and mutant pollen became apparent during PMI, from the early bi-cellular stage (Fig. 4). aaBb and AaBb microspores failed to complete PMI (Fig. 4, arrows) exhibiting abnormal pollen morphology with no DAPI staining that quickly became shrunken (Fig. 4, asterisks). This indicated that the AaBb and aaBb genotypes primarily affect the early stages of male gametophyte development, resulting in a failure of PMI in their half of the pollen.

To visualize mitotic defects in the AaBb and aaBb pollen, the ProHTR12:HTR12:GFP transgene, a centromere marker (Fang and Spector, 2005; Ingouff et al., 2007; Talbert et al., 2002), was introduced into the AaBb and aaBb mutants by genetic crossing. Then, chromosome segregation during pollen development was observed under the confocal microscope (Fig. 5). In wild-type haploid microspores, HTR12:GFP was detected at all five chromosomal centromeres (Fig. 5A) and in each dividing cell during PMI (Fig. 5B). Thereafter, GFP was no longer detected in the vegetative cell nucleus (Fig. 5C). Only the generative cell and subsequent sperm cells showed HTR12:GFP expression (Figs. 5C and 5D), as previously reported (Chen et al., 2009) (Fig. 5). aaBb and AaBb mutant microspores showed an expression pattern similar to that of the wild-type microspores (Figs. 5E and 5I). However, during PMI, HTR12:GFP expression disappeared in half of the pollen grains from the aaBb and AaBb mutants, and these arrested at the microspore stage without nuclear division (Figs. 5F and 5J, arrows). Occasionally, division did occur in the abnormal pollen grains, with only the vegetative cell nucleus visible (Fig. 5G, arrow), or the pollen appeared empty (Fig. 5G, arrowhead) or had strong staining indicative of being aborted (Fig. 5K, arrow), plus larger more diffuse areas of variegated HTR12 staining (Supplementary Fig. S2). These data demonstrate that progression of the AaBb and aaBb mutant pollen was prevented through PMI, thus, causing defects in haploid mitosis during pollen development.

Inheritance of the asf1b allele does not cosegregate with the defects in female and male gametophytes

In both male and female gametophytes, we observed that Aabb mutants exhibited surprisingly higher viability than AaBb mutants, which suggested that a possible chromosomal defect had occurred, unlinked to ASF1 gene function. To investigate this, we carried out segregation analysis of the asf1b allele. aaBb pistils contain approximately 1:1 (50.4% vs 49.6%, n = 1,351) fully matured ovules and small arrested ovules before fertilization (Fig. 1B, Table 1); therefore, we could hypothesize that only either the aB or the ab female gametophytes generated from aaBb pistils would be viable, and could thus undergo fertilization, producing progeny. Thus, either AaBB or AaBb progeny should survive when aaBb pistils were pollinated with wild-type AABB pollen. Surprisingly, however, among 152 progeny, we obtained 81 that were AaBb and 71 that were Aabb, suggesting that female gametophytes inheriting ab mutant alleles were fertilized and developed at similar frequencies to aB female gametophytes (Table 2). This meant that in aaBb pistils, ab and aB female gametophytes functioned well enough to transmit their alleles. Similarly, both ab and aB female gametophytes are likely to be present in the aborted ovules. In summary, these data indicate that the ASF1B genotype of the female gamete does not determine whether it will be viable or aborted.

During male gametogenesis, we observed many aborted pollen grains in aaBb anthers (Figs. 1C and 5H). When we crossed wild-type ovules (AABB) with aaBb mutant pollen, among 127 F1 progeny, 47 plants were AaBb while 80 plants were AaBB, meaning that ab pollen transmission is reduced compared to aB pollen (Table 2). Thus, the mutant pollen genotype appears to segregate with mutant phenotype. To further explore the pollen defect and mutant allele segregation, we used QUARTET (qrt) mutants that fail to undergo microspore separation, releasing viable pollen tetrads (> 95% shown by Alexander Staining; Fig. 6A), allowing us to examine the fate of each of the four progeny from individual PMCs (Preuss et al., 1994). We generated aaBb, Aabb, and AaBb mutant plants in the qrt1-4 homozygous mutant background by genetic crosses and then analyzed pollen generated from individual PMCs. Compared to wild type, whose pollen grains were virtually all viable, Aabb (Fig. 6B, purple bar, n = 1,570) exhibited a slight reduction in viability (89%), reminiscent of the slight reduction in ovule viability in this genotype. Conversely, aaBb (Fig. 6B, yellow bar, n = 1,539) and AaBb (Fig. 6B, green bar, n = 1,226) mutants showed much lower viability (43% and 46%, respectively) and a distribution of tetrads with all possible viable and nonviable ratios (Fig. 6B). Strikingly, 4:0 viable:nonviable ratios were the most common tetrads in aaBb and AaBb pollen (43% and 46%, respectively), followed by the tetrads with 0:4 (32% and 33%, respectively). In theory, if ASF1A and ASF1B are completely redundant, we would expect to get 100% 2:2 viable:nonviable tetrads in aaBb and AaBb pollen, with their mutations each displaying gametophytic defects during male reproduction. However, we observed only 18% and 15% tetrads with 2:2 ratios in aaBb and AaBb pollen, respectively. The presence of 4:0 or 0:4 tetrads representing 4 haploid cells that are either all alive, or all dead, in aaBb and AaBb pollen demonstrate that, in male gametogenesis, neither mutant allele cosegregates with pollen abortion. In addition, the frequent occurrence of 4:0 or 0:4 tetrads is a characteristic feature of two possible results of meiosis-generated reciprocal chromosomal translocations. A cruciform tetravalent can be formed to maximize chromosome paring in a translocation heterozygote. If an alternate disjunction occurs, this can produce euploid gametes, resulting in all 4 viable haploids. If an adjacent disjunction occurs, this produces aneuploid gametes, resulting in all 4 dead pollen grains (Fig. 7). Therefore, our pollen tetrad analysis strongly supports the hypothesis that the observed phenotypes are due to segregation in a translocation heterozygote.

Complementation analysis further suggests that phenotypes observed in the heterozygous Bb state are not linked to ASF1 gene function

To examine whether the reproductive phenotypes observed in AaBb and aaBb were from the T-DNA insertions in ASF1 genes, we investigated whether the phenotypes were rescued by introducing functional ASF1A or ASF1B. We introgressed ProASF1A::ASF1A-GFP and Pro ASF1B::ASF1B-GFP transgenic plants (Min et al., 2019) respectively into AaBb and aaBb mutant backgrounds and analyzed whether the mutant defects were complemented. Surprisingly, neither ASF1A-GFP nor ASF1B-GFP, even in the homozygous transgene state, rescued the abortion phenotype (Table 3). This result led us to suspect that lethality might exist in the Bb heterozygous state, but not in the bb homozygous state, since there was no mutant phenotype observed in bb single homozygous plants (Fig. 1A). To confirm this, we crossed Bb single heterozygous plants with wild type and their seed set was examined in the progeny. While BB wild-type progeny produced virtually all viable seeds, Bb single heterozygous plants exhibited 51% (n = 441) ovule abortion (Table 3). Furthermore, the ASF1B-GFP transgene, even in the homozygous state, could not rescue the abortion phenotype of the Bb plants (Table 3). These results strongly suggest that the reproductive phenotypes in AaBb and aaBb plants were not from the T-DNA insertion in the ASF1B gene. These data, combined with a lack of Mendelian inheritance of the Bb alleles, and the characteristic meiotic defects that we observe on both the male and female side, suggest that these phenotypes were caused by a T-DNA insertion-mediated chromosome rearrangement. These observations highlight the importance of including complementation analyses when using T-DNA mutants to assign gene function.

ASF1 is an evolutionarily conserved histone chaperone required for H3/H4 nucleosome assembly and disassembly and essential cellular processes such as DNA replication, transcription, gene silencing, and DNA damage checkpoint and repair. Previous reports of ASF1 function in Arabidopsis focused on the involvement of ASF1 in vegetative growth and reproductive organ formation, UV-induced DNA damage repair, and heat stress-induced expression of several key heat-shock related genes (Lario et al., 2013; Weng et al., 2014; Zhu et al., 2011). Since the biological function of ASF1 activity during mega- and microgametogenesis were largely unknown, we aimed to uncover ASF1 function during reproduction using detailed cytological and genetic analyses during sporogenesis, gametogenesis, and seed development.

From single aa and bb mutants, we generated AaBb double heterozygous plants. The AaBb plants were then self-fertilized to obtain different combinations of the double mutant genotype. ASF1A and ASF1B are located on Arabidopsis chromosomes I and V, respectively; therefore, they should show independent assortment in F2 progeny. However, the F2 populations did not show Mendelian segregation (Supplementary Table S1). Moreover, one interesting phenomenon was observed; aabb double homozygous plants were underrepresented (dashed-line box in Supplementary Table S2), and AABB, AAbb and aaBB were overrepresented (solid-line box in Supplementary Table S2). The underrepresented aabb double homozygous progeny can be explained by an essential but redundant ASF1A and ASF1B function as a histone chaperone during gametogenesis and embryogenesis, so that double homozygous mutant cannot be recovered easily in the F2 progeny (Min et al., 2019).

From aaBb plants, aB and ab alleles were equally transmitted and represented among surviving seeds, even after 50% of the gametes from the female side were dead. We also observed Mendelian segregation among the reduced number of progeny in the selfed Bb heterozygotes or Bb heterozygous plants crossed reciprocally with wild type. Bb heterozygous plants show 50% seed abortion but homozygous bb plants produce all viable seeds. This is in contrast to the typical gametophytic or embryonic lethal genes which show reduced transmission and thus underrepresentation of those alleles among the surviving seeds.

The phenotype we observe in AaBb and aaBb mutants is dominated by a failure of post-meiotic cell divisions during both male and female gametogenesis. On the female side, ovules were aborted with short embryo sacs and functional megaspore identity. During male gametophyte development, pollen grains were arrested and shrunken, having failed to undergo nuclear division. A similar 50% post-meiotic arrest and gamete abortion was previously reported in Atrev3-2 heterozygous plants (Curtis et al., 2009). AtREV3 (Arabidopsis thaliana Recovery protein3) encodes a catalytic subunit of DNA polymerase zeta involved in UVB-induced DNA damage repair by translesion synthesis. Heterozygous Atrev3-2 plants exhibited 50% ovule and pollen grain abortion, resulting in a 50% reduction of the seed set. The surviving 50% plants displayed a Mendelian distribution of mutant alleles, with a 1:2:1 ratio of wild type:heterozygous Atrev3-2:homozygous Atrev3-2. Since the reduced seed set (but Mendelian distribution of the progeny) in selfed Atrev3-2 heterozygotes, but not in homozygotes, was observed, the authors hypothesized that a T-DNA-mediated translocation might have occurred that is associated with the Atrev3-2 allele, located in chromosome V. They examined chromosome spreads using fluorescence in situ hybridization (FISH) in meiocytes and showed that a tetravalent was formed between chromosome I and V. This indicates that a reciprocal translocation of chromosomal fragments had occurred between chromosomes I and V, leading to the pairing of four chromosomes to form a cruciform structure during meiosis. Two modes of segregation are then possible, with the gametes resulting from adjacent segregation being all sub-haploid, and thus inviable; and those resulting from alternate segregation all being viable (Fig. 7). In the current study, the overrepresented genotypes from our mutant allele segregation analyses; AABB, AAbb, and aaBB, were homozygous for each allele of ASF1A and ASF1B gene, with either both alleles occurring in the wild type (AA, BB) or mutant (aa, bb) state. This means that when undergoing meiosis, homologous chromosomes can pair successfully, even if they contain rearranged chromosomal fragments. The reduced viability and seed set in Bb heterozyotes but not in bb homozygotes (but with Mendelian segregation) suggests a strong possibility of a translocation of the chromosomal fragment associated with the asf1b allele. In our quartet analysis, the two most common phenotype distributions were all 4 viable pollen grains (43% in aaBb and 46% in AaBb) but also all 4 dead pollen grains (32% in aaBb and 33% in AaBb). This strongly supports the possibility that a T-DNA associated reciprocal translocation occurred associated with the asf1b allele in chromosome V, thereby forming a tetravalent cruciform structure in asf1b meiocytes. When alternate segregation occurs, then we see all 4 meiotic products viable (4 purple pollen grains in Fig. 6A), and if adjacent segregation occurs, then we see all 4 pollen grains dead (4 green pollen grains in Fig. 6A). To directly uncover the precise structural abnormalities in Bb meiocytes, further experimentation including microscopic observations by FISH and immunolocalization experiments, or genome-wide genetic mapping, could be carried out.

ASF1A and ASF1B proteins are both expressed from the beginning of male and female gamete formation, whereby they are present in the PMC and megaspore mother cell (MMC), and then expressed throughout gametogenesis (Min et al., 2019). Since their expression patterns and the defects of both genuine ASF1 mutant plants (aabb mutants), and plants we suspect to have a translocation (AaBb and aaBb mutants), are overlapping, without performing complementation tests, we could have mis-assigned the role of ASF1 during reproduction. T-DNA-associated chromosomal rearrangement is not uncommon; therefore, our observations further highlight the importance of complementation tests and the use of multiple mutant alleles where possible.

The work was supported by NRF of Korea (2017R1A2B2 007067) and the Next-Generation BioGreen 21 Program Grant (PJ013127). Y.M. was supported by the Stadelmann-Lee Scholarship Fund, Seoul National University.

AUTHOR CONTRIBUTIONS

Y.M. and Y.C. conceived and designed the experiments. Y.M. performed the experiments. Y.M., J.M.F., and Y.C. analyzed the data and wrote the paper.

CONFLICT OF INTEREST

The authors have no potential conflicts of interest to disclose.

Analysis of ASF1 mutants seed viability

Transmission efficiency of the asf1a and asf1b alleles through reciprocal crosses

Bb heterozygous phenotype and complementation analysis