In Arabidopsis hybrids and Hybrid Mimics, up‐regulation of cell wall biogenesis is associated with the increased plant size

Abstract Hybrid breeding is of economic importance in agriculture for increasing yield, yet the basis of heterosis is not well understood. In Arabidopsis, crosses between different accessions produce hybrids with different levels of heterosis relative to parental phenotypes in biomass. In all hybrids, the advantage of the F1 hybrid in both phenotypic uniformity and yield gain is lost in the heterogeneous F2. F5/F6 Hybrid Mimics generated from a cross between C24 and Landsberg erecta (Ler) ecotypes demonstrated that the large plant phenotype of the F1 hybrids can be stabilized. Hybrid Mimic selection was applied to Wassilewskija (Ws)/Ler and Col/Ler hybrids. The two hybrids show different levels of heterosis. The Col/Ler hybrid generated F7 Hybrid Mimics with rosette diameter and fresh weight equivalent to the F1 hybrid at 30 DAS; F7 Ws/Ler Hybrid Mimics outperformed the F1 hybrid in both the rosette size and biomass. Transcriptome analysis revealed up‐regulation of cell wall biosynthesis, and cell wall expansion genes could be a common pathway in increased size in the Arabidopsis hybrids and Hybrid Mimics. Intercross of two independent Hybrid Mimic lines can further increase the biomass gain. Our results encourage the use of Hybrid Mimics for breeding and for investigating the molecular basis of heterosis.


| INTRODUC TI ON
Hybrids have proved to have great value in agriculture as they produce large gains in biomass and seed yield in a number of crops (Cheng, Zhuang, Fan, Du, & Cao, 2007;Crow, 1998;Dan et al., 2014).
The gain in the hybrid plants compared with the performance of parents is referred to as hybrid vigor or heterosis. In Arabidopsis (Fujimoto, Taylor, Shirasawa, Peacock, & Dennis, 2012;Groszmann et al., 2014), maize (Birchler, Auger, & Riddle, 2003;Li, Yang, et al., 2017), and Chinese cabbage (Saeki et al., 2016), hybrid plants have an architecture with larger leaves and the plants are taller compared with the parents. The increased leaf size is due to increased number and size of leaf cells (Fujimoto et al., 2012;Groszmann et al., 2014).
In Arabidopsis, different hybrids differ in growth pattern and level of heterosis, suggesting multiple genetic routes for hybrid vigor. Heterosis is generated through interactions between the two parental genomes and epigenomes in the nucleus of the hybrid.
Transcriptome studies revealed that salicylic acid (SA)-mediated down-regulation of defense pathway genes contribute to the heterotic phenotype of the C24/Ler hybrid through increased expression of growth-promoting genes . In C24/Ler hybrids, up-regulation of the transcription factor PHYTOCHROME INTERACTING FACTOR 4 (PIF4) results in increased auxin biosynthesis which promotes plant growth by targeting downstream cell expansion and division genes (Wang et al., 2017). A decreased level of ethylene and its effect in delaying of senescence can also contribute to the extra growth in the hybrid plants (Gonzalez-Bayon et al., 2019;Song et al., 2018).
Hybrids are an end point to self-fertilization breeding because of the genomic heterogeneity in the F2 and subsequent generations . This provides a challenge to the hybrid industry of how to extend the hybrid advantage beyond the F1. In 1971, Busch, Lucken, and Frohberg, (1971) reported that in wheat pure breeding F5 lines derived from the hybrid plants are equivalent to the F1 hybrid in performance. Similar observations were reported in pea and tomato (Sarawat, Stoddard, & Marshall, 1994;Williams, 1959). No molecular studies were explored beyond these observations. In Arabidopsis, crosses between the C24 and Landsberg erecta (Ler) ecotypes produce hybrids with performance superior to the parental lines in biomass and seed yield (Groszmann et al., 2014;Wang, Liu, et al., 2018).
We selfed F2 individuals and coupled these crosses with selection based on the phenotype of the F1 hybrid. These procedures, repeated in successive generations gave, in the F6 and later generations, lines with rosette diameter, biomass, and seed yield comparable to the F1 hybrid (Wang et al., , 2017. Because of genome homozygosity (Wang et al., , 2017, the lines maintained the high yielding phenotype in successive generations. The interactions of the two parental genomes in the F1 hybrid set the level of hybrid vigor that could be achieved by the component alleles of the parents through levels of gene expression and interactions between sequences in the two parental genomes.
These Hybrid Mimics overcame the F1/F2 hurdle, providing a seed source for high yielding crops based on kept-seed planting.
None of the C24/Ler Hybrid Mimics outperformed the F1 hybrids (Wang et al., 2017), suggesting that heterosis in these F1 hybrids is the maximum level of vigor which can be achieved by gene interaction between the two parent genomes. To investigate whether Hybrid Mimics can be selected from other Arabidopsis hybrid combinations and to understand the molecular basis of increased plant size, we selected Hybrid Mimics from two hybrid systems involving other ecotypes of Arabidopsis. We found we

| Plant Material and growth conditions
Arabidopsis hybrid seeds [Wassilewskija (Ws)/Landsberg erecta (Ler), Col/Ler and C24/Ler] were produced by hand-pollination between parental accessions. Ws/Ler and Col/Ler Hybrid Mimic lines and F7 small plant lines were produced from the recurrent selection protocol (Wang et al., , 2017 (Figures S1-S4). Seeds of parental lines, Hybrid Mimics, and small plant lines were obtained through natural pollination without restricting the number of siliques unless specified. In Figure 7, the F1 hybrids (Ws/ Ler and Col/Ler) and intercross offspring of Hybrid Mimics were produced by hand-pollination; the silique-restricting procedure was applied for producing seeds of the control lines: Ws, Ler, Col, and parental Hybrids Mimics (Meyer, Torjek, Becher, & Altmann, 2004  In the selection of the F3 generation, 30 F3 plants from each line were grown for selection, and the recurrent selection process was performed with the largest/smallest plants again selected based on the criteria of flowering initiation time and rosette diameter at 30 DAS. The same selection processes were performed in the F4 and in the subsequent generations. Some F3 plant lines produced by large F2 plants were not selected due to their unsatisfactory phenotype of small plant sizes or flowering initiation time later than both parents in the F3 generation. In the F6, plant lines having a F1-like phenotypes in rosette diameter and uniformity were termed Hybrid Mimics and used to produce F7 lines. For the growth pattern of the parents, hybrids, and F7 lines, rosette diameters of each plant lines were measured at several time-points during the growth (10,15,20,25,and 30 DAS). n = 12-20.

| Germination rate
To minimize the impact of seed age on seed germination, we grew the parental lines (Ws, Ler, and Col), Ws/Ler, and Col/Ler F6 Hybrid Mimics and small plant lines under the same condition. Seeds of each plant were collected at similar time when all siliques were yellow and dry. In the same experiment, crosses between parents (Ws/Ler and Col/Ler) were made for hybrid seed production. The rates of seed germination of all plant line (Ws, Ler, Col, Ws/Ler, and Col/Ler F1 hybrids, and Hybrid Mimics and small lines) were examined at approximately 4 weeks after seed collection. At least, 10 seeds per line were scored as one replicate. The data represent the average value from two to four replicates.

| Transcriptome sample preparation and RNA extraction
For the sample set of Ws/Ler system, the rosette leaves of 25day-old parents Ws and Ler, two reciprocal hybrids Ws × Ler and Ler × Ws, seven Hybrid Mimic lines (WL_HM1-7), and two small lines (wl_sml1-2) were collected at time Zeitgeber Time (ZT) = 6-8 (ZT = 0 refers to dawn). For the same set of Col/Ler system, the rosette leaves of 25-day-old parents Col and Ler, two reciprocal hybrids Col × Ler and Ler × Col, six Hybrid Mimic lines (CL_HM1-6), and two small lines (cl_sml1-2) were collected.

| Transcriptome analysis
The quality control reports of each sequencing sample were provided by the sequencing provider. Alignment of sequenced reads was performed using STAR version 2.5.3a against the TAIR10 ref-  showing the same up-or down-regulation in the hybrids and both small lines were unlikely to be important for heterotic phenotypes, so were excluded.

| Real-time PCR
RNA samples were treated with DNase I during the RNA extraction process, and then, the products were reverse transcribed using SuperScript® III Reverse Transcriptase (Invitrogen, 18080044).
The resulting cDNA was diluted 50-200 times in nuclease-free water. For real-time PCR, 10 µl diluted cDNA was used as template in a 20 µl reaction. Real-time PCR with SYBR green detection was performed using the real-time PCR instrument ROTOR-GeneQ (QIAGEN). The expression data for each gene were normalized to the expression of the housekeeping gene At4g26410 (Czechowski, Stitt, Altmann, Udvardi, & Scheible, 2005). The following primers were used: As in the C24/Ler hybrids (Zhu et al., 2016), both Ws/Ler and Col/ Ler hybrids germinated earlier than the parent lines (Table 1) (Table 1).

| Hybrid Mimics were selected from Ws/Ler and Col/Ler hybrids
F I G U R E 1 Hybrid Mimics selected from Ws/Ler and Col/Ler systems had increased rosette sizes and fresh weight at 30 days after sowing (DAS). Rosette diameter (RD) (a) and fresh weight (FW) (b) of the parents Ws and Ler, WsxLer hybrids, seven F7 Ws/Ler Hybrid Mimic lines (WL_HM 1-7), and two F7 small lines (wl_sml1-2) selected from Ws/Ler system at 30 DAS. (c) Rosette phenotypes of the parents Ws and Ler, Ws/Ler hybrids, two representative Hybrid Mimic lines (WL_HM 4 and 7), and one small line (wl_sml1) at 30 DAS. Scale bar = 5 cm. Rosette diameter (d) and fresh weight (e) of parents Col and Ler, ColxLer hybrids, four Col/Ler Hybrid Mimic lines (CL_HM 1-4), and two small lines (cl_sml1-2) selected from Col/Ler system at 30 DAS. (f) Rosette phenotypes of the parents Col and Ler, Col/Ler hybrids, two representative Hybrid Mimic lines (CL_HM 1 and 4), and one small line (cl_sml1) at 30 DAS. Scale bar = 5 cm. The black dotted line represents MPV. For the fresh weight measurements, the fresh weights of rosette leaves and shoots of each plant were measured separately. p value is generated using Student's t test. * indicates p < .05; + indicates RD/FW > F1, p < . 05 Table S1).
In the Col/Ler system, the two earlier germinating Hybrid Mimics had plant sizes comparable to the F1 and larger than the parents  (Figure 1d,e). The two small F7 lines in the Col/Ler system (cl_sml1-2) had plant size and rosette biomass approximately half to two-thirds that of the parents (Figure 1a-f).  (Table S2). Approximately 18,000 genes were expressed with a read cut-off of 30. Differentially expressed genes (DEG) were defined by a significance value of p < .05 from the MPV.

| Identification of differentially expressed genes in the Hybrid and Mimics
A total of 8,681 genes, approximately 50% of expressed genes, were differentially expressed between Ws and Ler; fewer DEGs (2,053 genes, 11%) were identified between Col and Ler. 1,740 DEGs (9.5%) were identified in the Ws/Ler hybrids compared with the MPV (Figure 3a, Table S3). Of the 1,740 DEGs, approximately half (782 genes) were differentially expressed between the two parents ( Table S3). The numbers of differentially expressed genes in the Ws/Ler Hybrid Mimics and small lines ranged from 5,310 to 9,459 (29% −52%) (Figure 3a). Of the 876 DEGs (5% of the expressed genes) in the Col/Ler hybrids, 171 genes were differentially expressed between the two parents ( Figure 3a, Table S3).

| Up-regulation of cell wall biosynthesis in the hybrids and Hybrid Mimics
Of the 37 DEGs annotated in "cell wall organization or biogenesis", 32 were up-regulated in the Ws/Ler hybrids and Hybrid Mimics, indicating increased activity of cell wall biosynthesis and/or cell wall expansion genes (Table S8) (Figures 1 and 4).   Hybrid Mimics. In both small lines cl_sml1 and 2, the majority of cell wall biosynthesis genes were down-regulated (Figure 4b).

GAE
In crosses between the C24 and Ler ecotypes, the F1 hybrids had substantial levels of hybrid vigor in vegetative biomass and plant size (Groszmann et al., 2014). In the Hybrid Mimic line L2 (referred as HM-G here) selected from the C24/Ler hybrid system , at 28 days the cell wall biosynthesis genes CesA, UDG, and AXS had higher levels of transcripts in the rosette leaves of F1 hybrids and Mimics than the MPV (Wang et al., 2015, Figure S9).

| Up-regulation of XYLOGLUCAN ENDOTRANSGLUCOSYLASE (XTH) genes in the Hybrid and Mimics
XTHs cut and ligate xyloglucans as a means of integrating new xyloglucans into the cell wall and are important for loosening existing wall material and enabling cell expansion (Becnel, Natarajan, Kipp, & Braam, 2006;Rose, Braam, Fry, & Nishitani, 2002 In the two small control lines, half of the expressed XTH genes had decreased transcript levels or showed a trend of down-regulation relative to MPV (Figure 5a,b, Table S9). The up-regulation of four cell wall-related genes in Ws/Ler Hybrid and Mimics was validated by quantitative real-time PCR ( Figure S10).
Three XTH genes (XTH4, XTH8, and XTH9) were up-regulated or had a trend of up-regulation in two other Hybrids (Col/Ler and C24/ Ler) and in the Mimic lines; these loci were not up-regulated in the small plants (Figure 5c and Figure S9).
Four WRKY genes (WRKY26, 46, 50, and 51) were down-regulated in the Ws/Ler hybrids, and three or more WRKY genes were down-regulated in the Mimics (Figure 6a, Table S11). WRKY genes have roles in regulating pathogen-induced defense responses (Eulgem et al., 2000). Of the 72 WRKYs in the Arabidopsis genome, the overall level of gene activities of WRKYs was downregulated in the Ws/Ler F1 and in the Mimics, but they were up-regulated in the small plant lines (Table S11), consistent with the concept of a trade-off between plant growth and the level of defense response gene expression (Denance, Sanchez-Vallet, Goffner, & Molina, 2013).
In the Hybrid Mimic with the lowest biomass (Figure 1b), WL_HM3, two PR genes PR1, and PR5 were up-regulated (Figure 6a). The small plant line wl_sml1 had down-regulation of the PR2 gene, but PR genes were not down-regulated in wl_sml2 (Figure 6a).
In the Col/Ler hybrid system, the majority of the defense response genes were down-regulated in the F1 hybrids. In the Mimics, most  Figure 6 and Figure S11).

| Flowering genes play a role in producing the large rosette phenotype
Flowering plays an important role in plant growth and development in Arabidopsis (Jung, Pillen, Staiger, Coupland, & Korff, 2017). Plants  (Table S1).
Genes in "regulation of shoot system development" and "regu-  (Boss, Bastow, Mylne, & Dean, 2004;Simpson & Dean, 2002). In our datasets, LFY was expressed at a low level making a change in its gene expression difficult to score. In agreement with the Hybrid Mimics having a slight delay of flowering initiation (Table S1), FT and SOC1 were down-regulated in the two sets of Hybrid Mimics ( Figure S12). Down-regulation of the flowering associated genes also occurred in the Ws/Ler and Col/Ler F1 hybrids ( Figure S12).

| Intercrosses of Hybrid Mimics
In the Ws/Ler Hybrid Mimics, cell wall-related genes were up-regu-

| D ISCUSS I ON
The Hybrid Mimics and hybrids of both hybrid systems all germinated earlier than the parents or small lines (Table 1). In both the Ws/Ler and Col/Ler systems, there was no correlation between rosette diameter at 30 DAS and timing of germination (Figure 1a,d,  (Sarkissian, Harris, & Kessinger, 1964).
The early germination of hybrid seeds is likely to be due to heterosis occurring during embryogenesis (Alonso-Peral et al., 2017).
The two genomes in the hybrid may interact as early as the singlecell zygote stage to produce more vigorous growth and development during embryogenesis priming the seeds for more rapid germination.

FC vs.MPV Ws/Ler or Col/Ler
In our experiments, the rates of seed germination were examined on Murashige and Skoog (MS) medium supplemented with 3% (wt/ vol) sucrose using freshly collected seeds (approximately 4 weeks after seed collection). On moist soil, C24/Col hybrids had germination times similar to the faster germinating parent Col (48 hr after sowing), while parent C24 germinated approximately 20 hr later (Meyer et al., 2012). The observations of different germination times of Arabidopsis hybrids compared with their parents can be due to differences in growth condition, the age of the seeds, or different hybrid genotypes.

| Common pathways up-regulated in all three hybrid systems
Plant cell walls are composed primarily of cellulose associated with hemicelluloses and pectin (Thompson, 2005 expressing CesA2, 5, or 6 were taller than the wild type and produced 20% more biomass in 7-week-old mature plants (Hu et al., 2018). UDG genes were up-regulated in each of the hybrids and Mimics. These genes are important in specifying plant size. The double mutant ugd2 and 3 lacking two of the four UGD genes has a dwarf phenotype (Reboul et al., 2011). Transgenic Arabidopsis plants overexpressing a UGD ortholog from Larix gmelinii has an increased content of hemicelluloses and enhanced vegetative growth (Li, Chen, et al., 2017).
Another contributor to plant cell wall biology is the XTH gene family. A number of XTH genes are up-regulated in all three hybrid systems but not in the small plant lines. XTH gene products participate in cell wall growth and remodeling by endolytically cleaving xyloglucan polymers and joining the newly generated end to another xyloglucan chain in the plant cell wall (Rose et al., 2002). The Arabidopsis genome encodes 33 XTH genes expressed in every developmental stage from seed germination through flowering (Becnel et al., 2006).

| Hybrid Mimics differ in defense response pathway genes
In C24/Ler F1 hybrids, changes in defense and stress response gene expression are consistent with a reduction in transcription of basal defense genes Miller et al., 2015). The which is involved in the control of the balance between growth and defense. C24 has high levels of salicylic acid (SA) which do not affect its growth but hybrids with C24 as one parent have a decreased level of SA relative to C24 and down-regulated defense pathway genes (Bechtold et al., 2010;Groszmann et al., 2015). In Ws/Ler and pointing to an overlap between defense and senescence ( Figure 6 and Figure S11). Some high yielding hybrids in Arabidopsis and crop species (stay-green mutants) show delayed senescence which allows for more photosynthate production at the grain filling stage (Spano et al., 2003;Thomas & Howarth, 2000;You et al., 2007).
In Arabidopsis, we predict that we could generate high yielding Hybrid Mimics where a hybrid has a high level of hybrid vigor and where out-crossing is excluded. Selections of hybrid Mimiclike plants have been reported in a number of other species; bread wheat, field pea, and tomato have all been reported to have F5 -F6 lines with the same characteristics as the parental F1 hybrids and were stable in their properties in successive generations (Busch et al., 1971;Cregan & Busch, 1978;Sarawat et al., 1994;Williams, 1959  height (Birchler et al., 2003;Groszmann et al., 2014). Cell wall biosynthesis genes have increased expression in hybrids relative to parents.
In C24/Ler hybrids and Hybrid Mimics, up-regulation of the transcription factor PHYTOCHROME-INTERACTING FACTOR 4 (PIF4) results in increased auxin biosynthesis and signaling. Several  (a) and (e) auxin-responsive genes including cell expansion genes were up-regulated in the F1 hybrids and hybrid mimics, suggesting that increased auxin biosynthesis and signaling contribute to the hybrid phenotype by promoting leaf growth (Wang, Liu, et al., 2018;Wang et al., 2017).
Apart from the expression level of plant cell wall biosynthesis genes, the defense/growth balance is likely to be important. A reduction in the expression level of PR genes can lead to more energy being channeled to pathways which contribute to growth.

ACCE SS ION NUMBER S
RNA-seq data from this article are available in the GenBank database (accession no. GSE131682).

ACK N OWLED G M ENTS
The authors wish to thank Dr. Jean Finnegan for reviewing the manuscript, Dr. Ming-Bo Wang for suggestions on the project, Dr. Anyu Zhu for help in manuscript reviewing and data processing, and Dr.
Aihua Wang for technical support.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.