Utilizing MIKC-type MADS-box protein SOC1 for yield potential enhancement in maize

Key message Overexpression of Zea mays SOC gene promotes flowering, reduces plant height, and leads to no reduction in grain production per plant, suggesting enhanced yield potential, at least, through increasing planting density. Abstract MIKC-type MADS-box gene SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) is an integrator conserved in the plant flowering pathway. In this study, the maize SOC1 (ZmSOC1) gene was cloned and overexpressed in transgenic maize Hi-II genotype. The T0 plants were backcrossed with nontransgenic inbred B73 to produce first generation backcross (BC1) seeds. Phenotyping of both transgenic and null segregant (NT) BC1 plants was conducted in three independent experiments. The BC1 transgenic plants showed new attributes such as increased vegetative growth, accelerated flowering time, reduced overall plant height, and increased grain weight. Second generation backcross (BC2) plants were evaluated in the field using two planting densities. Compared to BC2 NT plants, BC2 transgenic plants, were 12–18% shorter, flowered 5 days earlier, and showed no reduction in grain production per plant and an increase in fat, starch, and simple sugars in the grain. Transcriptome comparison in young leaves of 56-day-old BC1 plants revealed that the overexpressed ZmSOC1 resulted in 107 differentially expressed genes. The upregulated transcription factor DNA BINDING WITH ONE FINGER 5.4 (DOF5.4) was among the genes responsible for the reduced plant height. Modulating expression of SOC1 opens a new and effective approach to promote flowering and reduce plant height, which may have potential to enhance crop yield and improve grain quality. Supplementary Information The online version contains supplementary material available at 10.1007/s00299-021-02722-4.


Introduction
Increasing crop production is key to feeding the future (Ash et al. 2010). Breeding efforts in both public and private sectors in agriculture have been made to increase yield through genetic manipulation of various traits such as biotic or abiotic stress resistance (Nelson et al. 2007;Tester and Langridge 2010). Of the proposed flowering pathway gene networks (Fornara et al. 2010;Hill and Li 2016), MADS-box genes play significant roles in the formation of floral meristem and floral organs (e.g., male and female gametophyte), the control of floral transition and flowering time, and the development of seed and fruit (Becker and Theissen 2003;Causier et al. 2002;Garcia-Marotoet al. 2003; Theissen 2010, 2013;Heijmans et al. 2012;Masiero et al. 2011;Ng and Yanofsky 2001;Parenicova et al. 2003;Theissen et al. 2000). In addition, MADS-box genes function in root growth, lateral root formation, and morphogenesis of other organs (Smaczniak et al. 2012;Tapia-Lopezet al. 2008;Teo et al. 2019;Yu et al. 2014;Zhang and Forde 1998). Manipulation of these MADS-box genes provides an alternative approach to modulating plant reproductive growth, with the potential to influence crop yield (Castelan-Munoz et al. 2019;Hill and Li 2016;Trevaskis 2018).
Due to their significance in plant development (Gramzow and Theissen 2010;Schilling et al. 2018), several MADSbox genes have been patented for their regulatory roles in enhancing agronomic traits (Bae et al. 2011;Cacharron et al. 2000;Danilevskaya and Bruce 2008;Giovannoni et al. 2013;Podila et al. 2005;Takatsuji and Kapoor 2002), of which a maize ZMM28 MADS-box gene (patent application # WO2008148872A1) has been patented for yield increase through its overexpression (Danilevskaya and Bruce 2008). The ZMM28 gene is a homolog of the Arabidopsis FRUIT-FUL (FUL)/ AGAMOUS-like 8 (AGL8) gene, and it regulates maize flowering (Munsteret al. 2002). Increasing the expression of the ZMM28 gene enhanced grain yield in the field (Wu et al. 2019). In 2019, USDA APHIS released a petition from DU PONT Pioneer Hi-Bred for determination of nonregulated status for enhanced grain yield potential and glufosinate-ammonium resistant DP202216 maize after safety evaluation of the maize ZMM28 protein (Anderson et al. 2019a, b;Catron 2019).
The SOC1 protein is a MIKC protein. In Arabidopsis, the SOC1 gene is a positive regulator of the downstream MADSbox genes such as APETALA1 (AP1) and FUL/AGL8 (Lee and Lee 2010). Maize (Zea mays) SOC1 gene (ZmSOC1 or ZmMADS1) is a flowering activator (Alter et al. 2016). The effect of overexpression of the ZmSOC1 on maize grain production has not been reported. In this report, we describe the potential of using the ZmSOC1 to enhance maize grain yield. We present phenotypic data of the first (BC 1 ) and second (BC 2 ) generations of backcross (BC) plants. These data provide evidence that the constitutive expression of ZmSOC1 can result in enhanced yield potential through (1) hastening plant growth, (2) promoting flowering, (3) reducing overall plant height, (4) shortening overall time (about 2 weeks) needed from sowing to harvest, and (5) increasing or having no reduction in grain weight per plant. We also provide RNA-seq data derived from young leaves of six transgenic and three nontransgenic BC 1 lines to reveal the overall impact of ZmSOC1 overexpression on expression of other associated genes.

Constructs and plant transformation
Seeds of maize inbred line B104 were germinated to grow to 2-week old plants to harvest leaves. Isolation of total RNA was conducted using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA). The RNA sample was treated by DNase. Reverse transcription of RNA to cDNA was performed using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA).
Sequence of the maize SOC1 gene (ZmSOC1 or ZmMADS1) has been published in GenBank (accession number: NM_001111682.1) (Alter et al. 2016). The ZmSOC1 sequence is also available in GenBank (accession number: HQ858775.1). Two pairs of polymerase chain reaction (PCR) primers included in Table S1 were designed to amplify the ZmSOC1 from the cDNA of the leaf tissues of maize inbred line B104. The first pair of primers, MK_F and MK_R, were used to amplify the exact coding region of ZmSOC1. Using the second pair of primers, MK_F1_Kpn I and MK_R1_Xba I, unique restriction enzymes Kpn I and Xba I were added to the 5′-and 3′-end of the ZmSOC1, respectively. The PCR products were double digested using Kpn I and Xba I. Meanwhile, the pKANNIBAL plasmid was digested using Kpn I and Xba I to remove the PDK intron. The digested PCR products and the pKANNIBAL without the intron were ligated to insert the ZmSOC1 gene between the CaMV 35S promoter and the Ocs terminator (Fig. S1). The ligated pKANNIBAL-ZmSOC1 plasmid was digested using Not I to release the CaMV 35S-ZmSOC1-Ocs expression cassette. The released cassette was blunted using Klenow enzyme. Binary vector pTF101.1 (Paz et al. 2004) was double digested by Hind III and EcoR I, and the sticky ends of the digested pTF101.1 plasmid were blunted using Klenow enzyme. The blunt end pTF101.1 fragment and the blunt end CaMV 35S-ZmSOC1-Ocs expression cassette were ligated to generate pTF101.1-ZmSOC1 for overexpression of the ZmSOC1 (herein ZmSOC1-OX) (Fig. S1). The pTF101.1-ZmSOC1 contains the bialaphos resistance (bar) gene under the CaMV 35S promoter for selection of transformed plant cells using glufosinate (GS) herbicide. The ZmSOC1 in the resulting pTF101.1-ZmSOC1 were sequenced. Sanger sequencing data confirmed that a 696bp ZmSOC1 derived from the cDNA of the maize inbred line B104 was successfully inserted into the binary vector pTF101.1-ZmSOC1. In the 696-bp ZmSOC1 sequence, 694 bp are identical to the published 696-bp reference deposited in the GenBank (accession numbers HQ858775.1 and NM_001111682.1). The protein sequence of the cloned ZmSOC1 is identical to that derived either from the HQ858775.1 or from a part of the NM_001111682.1. The pTF101.1-ZmSOC1 verified through sequencing was used for maize transformation (Fig. S1).
The pTF101.1-ZmSOC1 was transformed into Agrobacterium tumefaciens strain EHA101. The construct was introduced into maize using Agrobacterium-mediated immature embryo infection method of Hi-II (A188 × B73) genotype (Frame et al. 2015). The T 0 transgenic Hi-II plants were backcrossed with nontransgenic inbred line B73 to produce first generation of backcross (BC 1 ) seeds, which have about 75% of the B73 genetic background.

Phenotyping of the BC 1 plants grown in pots
In each experiment, ten BC 1 lines, each having more than one hundred seeds, were used for phenotyping in this study. BC 1 seeds were germinated in water-soaked Suremix Perlite planting medium (Michigan Grower Products Inc., Galesburg, MI) in 4-inch plastic pots (8.9 cm width × 12.7 cm height) in a greenhouse in the springs of 2018 and 2019. Three experiments were conducted, including two in 2018 and one in 2019. Ten seeds per selected line for each experiment were sowed on May 17 and June 11 in 2018 and 16 seeds per line were sowed on May 11 in 2019. In each experiment, poor germinating lines with less than two plants were excluded for further phenotypic data collection in this study. Individual BC 1 plants were transplanted to a 4-gallon pot (top diameter 30 cm, bottom diameter 24 cm, depth 27 cm) and the plants were grown in a secured courtyard under natural environmental conditions at Michigan State University, East Lansing, Michigan (latitude 42.701847, longitude −84.482170). All of the plants were irrigated every other day and fertilized once a week using fertilizer (N:P:K = 20:20:20). Young leaves of 30 to 40-day old plants, 0.5 g per plant, were collected separately for DNA isolation, frozen in liquid nitrogen, and stored in a freezer at −80 °C.
Phenotypic data collection included plant height, seed germination date, date of tassel appearance, date of silk appearance, the total number of stem nodes and leaves, the number of cobs, dry weight of aerial parts without ears, dry weight of ear(s) excluding husk(s), and dry weight of grain. Plant heights measured during plant growth refer to stalk heights from the soil surface to the node of the highest leaf. The final heights of the maize plants refer to stalk heights from the soil surface to the top of tassels at harvest time. All of the plants for each experiment were harvested at the same time after they reached full physiological maturity in late October. The aerial parts of each plant were collected in a paper bag and dried at 25 °C for over 2 months in the lab prior to weighing the dry weights of total aerial parts, cob(s), and grain. For the third experiment in 2019, the date that an ear reached mature color ( Fig. S2) was recorded for each plant. To reduce bias during data collection, identification of transgenic and nontransgenic plants using PCR was mostly performed about 1 month before plant harvesting. In total, plants from nine transgenic lines in three experiments were used for phenotypic data collection.

Field test of BC 2 plants
BC 1 seeds of transgenic lines c7 and c9 were sown in a greenhouse with temperatures above 21 °C and a 16-h photoperiod. For further seed production, B73 was used as the pollinator. Ten seeds for each of the B73, c7, and c9 were sown at one time for four times at 5-day intervals to make sure the flowering time of B73 was able to meet that of the c7 and c9. The presence/absence of the transgene in the seedlings was determined using PCR. Flowering time for all plants was recorded. All tassels of BC 1 plants were removed. Pollen from the B73 plants was used to pollinate the transgenic plants. The ears of transgenic plants were covered using pollinating bags. BC 2 seeds from each individual plant were harvested separately. BC 2 seeds from three c7 lines and three c9 lines were grown in the field. For each line, 30 plants were randomly grown in each of the six plots, including three plots at a high planting density of 40,000 plants/acre and another three at a low planting density of 32,000 plants/acre. Two extra rows of B73 plants for each plot were used as protective borders. A drip irrigation system was installed in the field for plant irrigation as needed. Phenotypic data collection, including flowering time, plant height, leaf number, ear dry weight, and grain dry weight, were conducted using the same procedures described above for the BC 1 plants. In addition, grain quality from 50 plants were measured using a Grain Analyser (Infratec ™ 1241, FOSS Analytical AB, Denmark).

Detection of the transgenic plants
DNA was isolated from about 200 mg of leaf tissue for each sample using the cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle 1987). Maize Actin gene primers ZmAct_F and ZmAct_R were used as a control to verify the template DNA. Two pairs of primers, bar-F and bar-R for the bar gene, 35S-F (3′ portion of the CaMV 35S promoter) and MK_R for the ZmSOC1 gene (Table S1), were used to detect the presence of transgenes in each sample. PCR reaction conditions for all primer pairs started with an initial denaturation for two min at 94 °C, 30 cycles of 45 s at 94 °C, 60 s at 58 °C and 90 s at 72 °C, and a final extension for 10 min at 72 °C. All amplified PCR products were separated on 1% agarose gel containing ethidium bromide and visualized and photographed under UV light (Fig. S3).

RNA sequencing and transcriptome analysis
For RNA isolation, leaves from 56-day-old plants at a vegetative growth stage were harvested ( Fig. S2), frozen immediately in liquid nitrogen, and stored at −80 °C in a freezer. Two transgenic lines, c7 and c9, were used. A total of nine samples from nine plants were used, including three transgenic and three nontransgenic null segregants from the c7 transgenic event and 3 transgenic plants from the c9 transgenic event. Total RNA of each sample was isolated from about 500 mg of young leaf tissue using a separate CTAB method (Zamboni et al. 2008) and purified using RNeasy Mini Kit (Qiagen, Valencia, CA, USA). On-Column DNase digestion with the RNase-free DNase Set was used to remove DNA in the RNA samples (Qiagen). RNA quality was determined using the High Sensitivity RNA ScreenTape system (Agilent technologies, Santa Clara, CA). All of the RNA samples used for RNA sequencing had a quality score greater than 5.0, and cDNA of all these samples were synthesized for reverse-transcription of 3-5 μg RNA per sample using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA). Regular RT-PCR was used to verify the expression of the transgenes before RNA sequencing. The reaction conditions for RT-PCR were 94 °C for 2 min, 35 cycles of 45 s at 94 °C, 60 s at 62 °C and 60 s at 72 °C, with a final 10 min extension at 72 °C. RT-PCR products were separated and visualized on 1.0% agarose gel containing ethidium bromide.
The RNA samples were sequenced (150 bp-paired end reads) using the Illumina HiSeq4000. All sequencing was performed at the Research Technology Support Facility at Michigan State University (East Lansing, Michigan, USA). The FastQC program (www. bioin forma tics. babra ham. ac. uk/ proje cts/ fastqc/) was used to assess the quality of sequencing reads for the per base quality scores. About 32-42 million reads (MR) for each of the nine biological samples with average scores ranging from 37 to 40 were obtained for transcriptome analysis. The paired-end reads (~ 100 MR in total) combined from one sample of the NT line and one sample for each of the two TR lines were used for transcriptome assembly to develop a maize transcriptome reference ZmTrinity using Trinity/2.8.5 (Haas et al. 2013). This transcriptome reference was anticipated to cover all potential isoforms in the sequence reads. The RNA-seq reads of three biological replicates for each of the c7 and c9 transgenic lines and one c7 line were analyzed. The paired reads were aligned to the transcriptome reference ZmTrinity, and the abundance for each of a single read was estimated using the Trinity command "align_and_estimate_abundance. pl". The Trinity command "run_DE_analysis.pl-method edgeR" was used to conduct differential expression analysis (Haas et al. 2013). Differentially expressed transcripts (DETs) with a false discovery rate (FDR) value below 0.05 (p value < 0.001) were used for further analysis of different pathway genes of maize.
Pathway genes of nine phytohormones in Arabidopsis, including auxin, cytokinin, ABA, ethylene, gibberellin, brassinosteroid, jasmonic acid, salicylic acid, and strigolactones, were retrieved from RIKEN Plant Hormone Research Network (http:// hormo nes. psc. riken. jp/). Similarly, pathway genes of sugar in Arabidopsis were identified. These Arabidopsis hormones, MADS-box, and sugar genes (Table S2) were used as queries to blast against the transcriptome reference ZmTrinity and the isoforms showing e values less than −20 were identified and used for transcriptome comparisons. Flowering pathway genes in Arabidopsis and cereals (Walworth et al. 2016) were used to analyze flowering-related DETs identified in this study. Cytoscape 3.8.2 was used to construct gene networks of overrepresented gene ontology (GO) terms for the selected DETs under BiNGO's default parameters with selected ontology file 'GOSlim_Plants' or 'GO_Full' and selected organism 'A. thaliana' (Maere et al. 2005;Shannon et al. 2003). Most of the analyses were performed using the resources at the High Performance Computing Center at Michigan State University.
Quantitative RT-PCR (RT-qPCR) using SYBR Green system (LifeTechnologies, Carlsbad, CA) were conducted to check the selected transcripts. The primers were designed according to the RNA-seq data, and ZmActin1 was used to normalize the RT-qPCR results (Table S1). RT-qPCR was performed on a Roche LightCycler ® 480 Instrument II. The reaction conditions for RT-qPCR were 95 °C for 5 min, 45 cycles of 30 s at 95 °C, 45 s at 62 °C and 30 s at 72 °C. Transcript levels within samples were normalized to Actin. Foldchanges were calculated using 2 −∆∆Ct , where ∆∆Ct = (Ct TARGET -Ct NOM ) transgenic -(Ct TARGET -Ct NOM ) nontransgenic . Three biological samples and three technical replicates were used for the analysis of each transgenic and nontransgenic line.

Statistical analysis
Statistical analysis of the phenotypic data was conducted using ANOVA and TukeyHSD in RStudio (Version 1.0.136).

The overexpressed ZmSOC1 causes phenotypic changes
Inbred maize genotypes remain recalcitrant for genetic transformation (Yadava et al. 2016). In this study, we performed transgene analysis in a transformable Hi-II genotype, which has a hybrid genetic background of A188 and B73 (Frame et al. 2015). To produce seeds, T 0 transgenic plants were pollinated with nontransgenic inbred B73 to obtain BC 1 seeds. A total of 16 independent ZmSOC1 overexpression events were brought to maturity. BC 1 and BC 2 seeds should have ~ 75 and 87.5% of B73 genetic background, respectively. Because of the complex hybrid background, each of the BC 1 and BC 2 plants of Hi-II was not genetically identical. To evaluate the effect of the ZmSOC1 transgene in diverse genetic backgrounds, nontransgenic, null segregant plants from nine independent BC 1 lines were identified and used as negative controls in the BC 1 experiments. These null segregants (NT) shared the similar genetic background as well as the tissue culture effects with the transgenic counterparts (TR) because they were derived from the same ears. For the field test of the BC 2 plants, six lines were derived from two BC 1 lines (c7 and c9), for which RNA-seq was conducted.
Three BC 1 experiments under three environmental conditions were conducted in two summers to evaluate the impact of ZmSOC1-OX on the following physiological or agronomic traits: (1) seed germination time, (2) time of tassel appearance, (3) time of silk appearance, (4) days between tassel and silk appearance, (5) leaf number, (6) plant height, (7) dry weight (plant aerial material excluding ears), (8) dry weight (ear), and (9) dry weight (grain). These selected physiological or agronomic traits are important criteria to evaluate yield potential. Of the nine traits analyzed (Table 1, Table S3), germination time of the TR seeds (5.3 ± 1.7 days) was similar to that of the NR seeds (5.1 ± 9.5 days); this was the only unchanged trait observed consistently in all three experiments. Two lines, c7 and c9, which had higher numbers of both TR and NT plants (Table S3), were selected for a detailed phenotypic and transcriptomic analysis. For the field test of the 1080 BC 2 plants, the results confirmed the accelerated flowering and reduced plant stature in the TR plants. In addition, the BC 2 TR plants showed unreduced grain production per plant and increased contents of starch, fat, and simple sugars in the grain. These phenotypic changes, as well as the ZmSOC1-OX-induced differentially expressed genes (DEGs), provide encouraging data that implicate the potential of utilizing the ZmSOC1-OX gene to increase maize yield.

Overexpression of ZmSOC1 promotes flowering
ZmSOC1-OX resulted in early flowering in both BC 1 and BC 2 plants. Upon close examination of BC 1 TR and NT plants, we observed that the TR plants were taller than the NT plants before flowering in the first two experiments, but ended up shorter than the NT plants when they reached maturity (Table 1). For the BC 1 lines tested in experiment #1 and #2 in 2018, the seeds were sown in mid-May and mid-June, respectively. To verify the observation, a third experiment was conducted in the summer of 2019. Seeds for experiment #3 were sown in mid-May, similar to experiment #1. We measured the heights of all plants from nine independent BC 1 lines every 10 days from the 20 th to the 90 th day after sowing. The comparisons again showed that TR plants were taller than NT plants during the vegetative growth period (between 20 and 70 days after sowing) (Fig. 1A). After around day 70, NT plants grew taller, and mature TR plants were shorter than NT plants in the end (Table 1).
For the BC 1 plants, slightly accelerated flowering, indicated by the time of tassel and silk appearance, was observed in TR plants in two of the three experiments. Tassels appeared 1 day earlier for TR (61.3 ± 7.2 days) than NT (62.8 ± 11.4 days) with no significant difference ( Fig. 2A). Silks appeared three days earlier for TR (67.3 ± 9.2 days) than NT (70.9 ± 11.4 days) with a significance value of p = 0.05 (Fig. 2B). In addition, the time from the appearance of tassels to the appearance of silks was about two days shorter for the TR plants (6.0 ± 3.3 days) than that of the NT plants (8.2 ± 4.0 days) in all three experiments, which is highly significant (p = 0.001). The reduced nicking time between male and female flowers can facilitate an effective pollination that affects grain production. We also made backcrosses between B73 and TR plants to produce BC 2 seeds for c7 and c9. All of the BC 1 TR plants flowered 5-10 days earlier than the BC 1 NR and B73 plants in the greenhouse. For the field test of the BC 2 plants, dynamic changes of plant height at early plant growth stages were not measured. Regardless of the planting densities, early flowering (~ 5 days for the appearance of both tassels and silks) was observed for the TR plants (Figs. 1B, 2C). Overall, overexpression of ZmSOC1 was able to promote flowering. The early flowering of ZmSOC1-OX plants grown under natural conditions either in pots for the BC 1 or in the field for the BC 2 TR plants is consistent with the results observed in the greenhouse-grown ZmSOC1-OX plants (Alter et al. 2016).

Overexpression of ZmSOC1 reduces plant height
Plant height of TR plants was significantly reduced (Fig. 3A, 3B, Table 1, Table S3). For mature BC 1 plants, the average height from soil to tassels was 134.9 ± 23.8 cm for TR plants but 159 ± 30.2 for NT (p = 0.001) (Fig. 3A). For BC 2 plants, the height of TR plants was shorter than NT plants at the time of both tassel appearance (TR 117.4 ± 26.9 cm vs. NT 140.0 ± 25.7 cm) and maturity (TR 152.3 ± 27.5 cm vs. NT 171.2 ± 28.1 cm) (Fig. 3B).
For leaf number in the BC 1 TR plants, a significant decrease and a significant increase were observed in experiment #1 and experiment #2, respectively (Table 1). For the c7 and c9 lines, TR plants had 1 and 1.4 fewer leaves than NT plants, respectively, in experiment #1 (seed sown in mid-May 2018; Fig. 2B). In contrast, TR plants showed 1.8 and 2 more leaves than the NT plants, respectively, in experiment #2 (seed sown in mid-June 2018). The inconsistency is likely due to the planting times that caused variations in the plant growth conditions under natural environmental conditions. Consequently, the averages of the leaf numbers for BC 1 plants in two experiments were similar for both TR (11.2 ± 1.5 leaves) and NT (11.1 ± 1.4 leaves) plants (Fig. 3C). For the BC 2 plants, the leaf number of TR plants (11.4 ± 1.7 leaves) was less than the NT plants (12.9 ± 1.9 leaves) (Fig. 3D). Overall, it was reasonable to conclude that overexpression of the ZmSOC1 reduces leaf number because of the larger population of the BC 2 plants. The results of both reduced plant height and leaf number in the BC 2 TR plants grown under the natural conditions are consistent with those reported in the greenhouse-grown BC 1 plants (Alter et al. 2016).

Overexpression of ZmSOC1 increases grain dry weight per plant
Yield potential of the ZmSOC1-OX plants was not reported (Alter et al. 2016). For the BC 1 plants, the increased dry weight for both ear and grain was recorded in TR plants in all three experiments, with significant difference in one experiment (Table 1, Table S3). Average dry weight of ear was over 27 g heavier for TR plants (174.3 ± 93.2 g) when compared to NT plants (146.9 ± 87.4 g). A nearly 23 g increase in average grain weight in TR plants (147.5 ± 82.7 g) versus NT plants (124.6 ± 77.9 g) was observed. For the c7 line, the grain dry weight of the TR plants showed a slight increase in experiment #1 and #3 and a significant increase in experiment #2 (Figs. 2F, 3,  Fig. S2C). For the c9 line, grain dry weight of the TR plants showed no significant difference in all three experiments (Fig. 3E).
On the other hand, plants with shorter stature are desirable for high-density planting in the field to increase yield per acre. This laid the foundation for our field test of the BC 2 plants using two planting densities. For each of the TR and NT BC 2 plants, the two planting densities did not result in any significant effect on flowering time, plant height, leaf number, average of grain dry weight per plant, or grain quality (Figs. 2, 3). For all of the BC 2 plants cross two planting densities, the average of grain dry weight per plant for TR plants (67.9 ± 45.5 g) was almost the same as that for NT plants (67.8 ± 53.4 g) (Fig. 3F). Notably, while the randomly mixed TR and NT plants grown in each plot were helpful to minimize the errors in this study, the NT plants had advantages over the TR plants in competing for light during the growing season because they were taller after flowering. The unreduced grain weight in the TR plants in the mixed NT/TR field was encouraging. This suggests that a higher yield potential could be achieved for the TR plants when a uniform population were tested in the field. The BC 1 plants were from two independent BC 1 lines (c7 and c9). Rep1, rep2, and rep3 represent experiment #1, #2 and #3. A Tassel appearance time of BC 1 lines. B Silk formation time of BC 1 lines. Star(s) on each bar represent the comparison result between the TR and the NT plants of the same BC 1 line in the same experiment, n for c7_18rep1 NT = 3, n for c7_18rep1 TR = 3, n for c7_18rep2 NT = 4, n for c7_18rep2 TR = 4, n for c7_17rep3 NT = 8, n for c7_19rep3 TR = 6, n for c9_18rep1 NT = 2, n for c9_18rep1 TR = 5, n for c9_18rep2 NT = 5, n for c9_18rep2 TR = 3, n for c9_19rep3 NT = 3, n for c9_19rep3 TR = 8. C Tassel and silk appearance time of BC 2 lines. BC 2 seeds from three c7 lines and three c9 lines were grown in a field. For each line, 30 plants were randomly grown in each of the six plots, including three plots for a high planting density of 40,000 plants/acre (H) and another three for a low planting density of 32,000 plants/acre (L). Star(s) on each bar represent the comparison result between the TR and the NT plants of the same planting density. Bars indicate standard deviation. Significance codes: *** < 0.001, ** < 0.01, *< 0.05

Overexpression of ZmSOC1 affects grain quality
Whether or not an overexpressed MADS-box gene is able to change grain quality is not known. We tested the grain quality from 35 BC 2 (12 NT and 23

Overexpression of ZmSOC1 functions at the transcript level
To reveal the potential impact of ZmSOC1-OX on transgenic plants at transcript levels, we analyzed the transcriptomes in leaves from three TR plants for each of the c7 and c9 E Average grain dry weight per BC 1 plant; average grain dry weight per BC 2 plant. Star(s) on each bar represent the comparison result between the TR and the NT plants of the same BC 1 line in the same experiment, n for c7_18rep1 NT = 3, n for c7_18rep1 TR = 3, n for c7_18rep2 NT = 4, n for c7_18rep2 TR = 4, n for c7_17rep3 NT = 8, n for c7_19rep3 TR = 6, n for c9_18rep1 NT = 2, n for c9_18rep1 TR = 5, n for c9_18rep2 NT = 5, n for c9_18rep2 TR = 3, n for c9_19rep3 NT = 3, n for c9_19rep3 TR = 8. For the BC 2 plants, star(s) on each bar represent the comparison result between the TR and the NT plants of the same planting density. Bars indicate standard deviation. Significance codes: *** < 0.001, ** < 0.01, * < 0.05 line and three NT plants for the c7 line, in which the c7 NT plants were genetically similar to c9 NT plants. Transcriptome analysis of BC 1 TR and NT plants was conducted on two major purposes, including to verify the expression of the transgenes and to reveal the genes that responded to the expression of ZmSOC1. Three comparisons were made to reveal DETs induced by ZmSOC1-OX. The comparison of c7 TR and NT resulted in 473 DETs, of which 322 were annotated to 249 unique genes; c9 TR and NT comparison resulted in 2576 DETs, of which 1692 were annotated to 1136 unique genes; and c9 TR and TR comparison resulted in 1,127 DETs, of which 676 were annotated to 485 unique genes (Table S4). The difference of the total number of DETs in the comparisons of the c9 TR and c7 TR likely reflects the genetic background variations and potential transgene insertion elicited changes in the two lines. The shared DETs in the two comparisons, including c9_TR versus NT and c7_TR versus NT, could be the transcripts responding to the overexpression of ZmSOC1.
The Venn diagram in Fig. 4A illustrates the overlap of the DETs from the three transcriptome comparisons. There are 130, 277 and 1,221 DETs that are unique for c7_TR versus NT, c7_TR versus c9_TR and c9_TR versus NT, respectively. Fifty-eight DETs were found from both c7_ TR versus NT and c7_TR versus c9_TR, 337 DETs were shared for both c7_TR versus c9_TR and c9_TR versus NT, and 130 DETs were shared for both c9_TR versus NT and c7_TR versus NT. Four DETs appeared to be present in all three transcriptome comparisons (Fig. 4A). These four DETs are NSE4A_ARATH (non-structural maintenance of chromosomes element 4), CP26B_ARATH (peptidyl-prolyl cis-trans isomerase), CHS2_MAIZE (chalcone synthase C2), and RA213_ARATH (ethylene-responsive transcription factor).
Of the annotated DETs, 134 DETs from 107 genes were shared in the two comparisons of c7_TR versus NT and c9_TR versus NT. Among them, 133 out of the 134 DETs were consistent in either up-regulation or down-regulation in the two comparisons (Table S5), including 96,861-fold and 127,973-fold increases of the overexpressed ZmSOC1 (MAD50_ORYSJ) in c7 and c9 TR, respectively ( Table 2). The high abundance of ZmSOC1 transcript accumulation verified strong expression of ZmSOC1 under the 35S promoter in both the c7 and c9 lines. The 134 shared DETs in the Venn diagram revealed the major genes affected by ZmSOC1-OX at the level of transcription (Table S5). From the 134 DETs, we picked primers to test seven selected DETs using RT-qPCR; the consistency of the RT-qPCR and RNA-seq results suggested that the RNA-seq data were reliable in this study (Fig. 4B, C).
Based on the phenotypic changes observed in BC 1 TR (such as earlier flowering and reduced plant stature, Table 1), we further analyzed the expression of flowering pathway genes and hormone-related genes. Twenty-one DETs of 14 genes were shared in the two comparisons of c7_TR versus NT and c9_TR versus NT (Table 2). In the flowering pathways, expression of two endogenous ZmSOC1 genes  (Table 2, Table S5). −∆∆Ct is an average of three biological and three technical replicates for each DET. ZmActin1 (SAC1_ARATH) was used to normalize the RT-qPCR results. Bars indicate standard deviation Table 2 Differentially expressed transcripts (DETs) of flowering pathway and hormone genes in maize leaves during the fast-growing stage of the plants before flowering [Log 2 FC: Log 2 (Fold change) = Log 2 (TR/NT) The bold DETs were verified by RT-qPCR. None of the DETs included in the table were the DETs in the comparison of the c7 TR and c9 TR N/A not annotated to known gene (s) (Fornara et al. 2010). GIGANTEA (GI) is an upstream regulator of CONSTANS (CO) and FT in the circadian clock-controlled flowering pathway of Arabidopsis, and promotes the expression of the downstream flowering-time genes (Mizoguchi et al. 2005). GIGAN_ORYSJ, an ortholog of GI, showed increased expression of 530-and 465-fold for the c7 TR and c9 TR, respectively (Table 2). Overall, the highly upregulated GIGAN_ORYSJ, MAD15_ ORYSJ (23-126 fold), and HD3A_ORYSJ (14-16 fold) and markedly reduced flowering time in both the c7 TR and c9 TR are likely the results of the overexpressed ZmSOC1 in these plants (Fig. 5A). The detected DETs in the pathways of auxin, brassinosteroid, ethylene, and gibberellin contributed to many observed phenotypes (e.g., reduced plant stature and the increased growth of the c7 and c9 TR plants during their vegetative growth) and nonvisible biochemical and physiological changes. DNA binding with one finger 5.4 (DOF5.4) is a transcription factor that negatively regulates cell cycle and cell expansion; enhanced expression of DOF5.4/OBP4 leads to plant dwarfing in Arabidopsis (Xu et al. 2016). In the c7 and c9 TR plants, DOF54_ARATH expression was upregulated to 143-and 104-fold, respectively (Table 2), supporting that the increase in expression of DOF54_ARATH may play a significant role in the reduced plant height of the c7 and c9 TR plants (Fig. 5A). Brassinosteroids (BRs) can promote plant growth and BR-deficient mutant plants often exhibit dwarfing (Fujioka and Yokota 2003;Grove et al. 1979;Kim et al. 2000;Mussiget al. 2003;Tanabe et al. 2005). In the BR pathway of Arabidopsis, UDP-GLYCOSYLTRANS-FERASE 73C5 (UGT73C5) catalyzes inactivation of BRs and can lead to dwarfed BR-deficient plants (Poppenberger et al. 2005), which is often undesirable for crop yield. The DETs of both 708A6_MAIZE and URT1_FRAAN had high similarities (e < −20) to UGT73C5, and both transcripts showed decreased expression of 23-42% and had the potential for an increase in BRs that often benefits plant development (Table 2). It is possible that the decreased expression of 708A6_MAIZE and URT1_FRAAN was responsible for the increased growth of c7 and c9 TR plants during their vegetative growth. Of the auxin-related DETs, expression of TRPA_MAIZE with its function unknown was downregulated to 3-12%; expression of C78A6_ARATH was upregulated to 2.6-3.8 fold in the c7 and c9 TR plants (Table 2). Notably, the increased expression of C78A6_ARATH may result in increased size of leaves, flowers and seeds, but defects in reproductive development (Fang et al. 2012). Both the c7 and c9 TR plants did not exhibit defects in reproductive development, but leaf and seed sizes were not investigated in this study. PROBABLE PROTEIN POSPHATASE 2C 27 (P2C27_ORYSJ) triggers the expression of stressresponsive genes in Arabidopsis (Liu et al. 2012); the increased expression of maize P2C27 could enhance salt tolerance in transgenic plants. Two ethylene-related, one gibberellin-related, and one sucrose-related DETs were also detected, although none of them seemed to have a close correlation with the phenotypic changes observed (Table 2). Overall, these major DETs driven by ZmSOC1-OX were the most likely candidates responsible for the phenotypic changes.
In addition to the DETs of flowering pathway and hormone-related genes, the other DETs also likely played a role in the overall phenotypic changes of the TR plants ( Table 2, Fig. 5A, Table S5). To reveal the overall impact of the DETs detected in the c7 TR plants, GOSlim_Plants in Cytoscape 3.8.2 was used to visualize the overrepresented Gene Ontology (GO) terms. As shown in Fig. 5, a total of nine GO terms were overrepresented or enriched (p < 0.05). Four overrepresented terms in the biological process are "metabolic process", "cellular process", "protein metabolic process", and "anatomical structure morphogenesis", supporting that multiple biological processes were affected by the ZmSOC1-OX. Three overrepresented terms in the cellular component include "cell", membrane", and "plasma membrane", suggesting that the phenotypic changes may be related to these cellular components. In the molecular function, two overrepresented GO terms are "transferase activity" and "catalytic activity" (Fig. 5B). More details of the overrepresented GO terms and their networks can be visualized using GO_full (Fig. S4). At a gene network level, the overrepresented GO terms reveal the overall impact of ZmSOC1 overexpression on plant growth, flowering, yield potential, or grain quality.

Discussion
Genetically modified (GM) crops for yield boosting are desirable to increase productivity without increasing land use, although they are not yet available on the market (ISAAA's GM Approval Database. http:// www. isaaa. org/ gmapp roval datab ase/). In our efforts to produce GM crops for boosting yield, we chose the SOC1 MADS-box gene as a target due to its significant role as a major integrator in the plant flowering pathway (Fornara et al. 2010;Lee and Lee 2010;Song and Chen 2018;Song et al. 2013). In this study, we cloned the ZmSOC1 gene to explore the potential of utilizing this gene to increase maize grain production. The ZmSOC1 was transformed into maize Hi-II. BC 1 plants of nine transgenic lines were grown in three experiments to evaluate phenotypic changes of nine traits between transgenic and nontransgenic plants in each BC 1 line population. We also conducted transcriptome analysis of six transgenic and three nontransgenic plants to reveal the DETs driven by ZmSOC1 overexpression. We demonstrated that manipulation of the expression of ZmSOC1 is a powerful approach to increase maize yield.
SOC1 is a key pathway integrator and flower activator Lee and Lee 2010;Moonet al. 2003). In monocots, functional SOC1 orthologues have been identified and demonstrated to be flowering activators (Alter et al. 2016;Lee et al. 2004;Ryuet al. 2009). Functional analysis of maize ZmMADS1 through both overexpression and RNA interference-mediated down-regulation has confirmed that maize ZmMADS1 is a functional SOC1 orthologue of maize (ZmSOC1) (Alter et al. 2016). In the overexpression experiment, transgenic plants (offspring of two lines) overexpressing the ZmSOC1 using the maize ubiquitin promoter showed early flowering, decreased leaf number, and reduced plant height compared to nontransgenic maize plants under greenhouse conditions (Alter et al. 2016). Our data, based on the field evaluation of the BC 2 plants from six lines, also found overexpression of ZmSOC1 driven by 35S promoter was able to promote flowering and reduce plant height and leaf number. In addition, our work demonstrated that ZmSOC1-OX approach can be adopted to enhance maize yield potential.
Orthologues of SOC1 in many plant species have been studied to reveal flowering mechanisms (Lee and Lee 2010). However, efforts using SOC1 orthologues to increase crop yield have not been documented. Regardless of the results from potted plants in this study, the increased grain production per TR plant suggests that ZmSOC1-OX, at least in some transgenic lines, has potential to increase maize grain yield. Additionally, the reduced plant height suggests that TR plants can be planted at a higher density, potentially increasing maize grain yield per acre. More experiments are ongoing to investigate BC 2 plants growing with different planting densities under field conditions. MADS-box transcription factors (TFs) function at almost every aspect of plant reproductive development through a complex protein-protein interaction network (Hugouvieux and Zubieta 2018). In Arabidopsis, CONSTANS (CO) activates SOC1 through FT to promote flowering (Yoo et al. 2005). In this study, comparative transcriptome analysis revealed that ZmSOC1-OX enhanced expression of maize FLOWERING LOCUS T (ZmFT) and maize GIGANTEA (ZmGI) genes. The result is similar to that observed in transgenic rice where rice SOC1 was overexpressed (Lee et al. 2004). In maize, the orthologue of FUL (ZMM28) is the MADS-box gene considered to be a duplication of the maize AP1 gene; it has been successfully used for increasing grain yield through overexpression (Wu et al. 2019). In this study, the overexpressed ZmSOC1 enhanced expression of the downstream AP1 MADS-box gene annotated as ZmAP1, but not the FUL gene (ZMM28), suggesting that ZMM28 is not a direct target at transcript level by overexpressed ZmSOC1 at the developmental stage tested. On the other hand, both ZmSOC1-OX and ZMM28 could be essential at protein levels (Abraham-Juarez et al. 2020). In Arabidopsis, the functions of AP1 and FUL are partially overlapping. The sequence difference in ZmAP1 and ZMM28 may be responsible for their divergence in responding to the overexpressed ZmSOC1 (Kater et al. 2006;McCarthy et al. 2015).
The increased grain yield in the selected maize ZMM28overexpression line DP202216 was attributed to increased early plant vigor and total leaf area (Wu et al. 2019). Similarly, the overexpressed ZmSOC1 also increased plant vigor in this study. Phenotypic variations in the nine BC 1 transgenic lines tested have suggested that there is high potential to identify ideal transgenic lines for high grain yield. Backcrosses of the selected ZmSOC1-OX line to B73 are ongoing to produce new inbred B73 lines containing homozygous ZmSOC1-OX for field trials and commercialization of the ZmSOC1-OX technology. In addition, further studies are still needed to determine transgene insertion position(s), copy number, and the correlation between the ZmSOC1 expression levels and the phenotypic changes in all 16 transgenic lines we produced. Taken together, modulating expression of MADS-box genes (e.g., the ZmSOC1 in this report) opens a new approach to enhance crop yield or change grain quality.

Conclusions
This is the first investigation of using a SOC1 gene to increase the potential for high grain yield. Overexpression of ZmSOC1 accelerated flowering, reduced plant stature, and increased/unreduced grain dry weight of the BC 1 and BC 2 TR plants grown under natural conditions. In addition, the grain of BC2 TR plants showed an increase in the content of starch, simple sugars, and fat. Transcriptome analysis revealed potential genes responding to ZmSOC1-OX. Overall, the results facilitate a better understanding of SOC1regulated growth and flowering in maize. More importantly, modulating expression of SOC1 opens a new and effectual approach to promote flowering and reduce plant height, which may have the potential to enhance crop yield and improve grain quality.