Genetic Structure and Molecular Mechanisms Underlying the Formation of Tassel, Anther, and Pollen in the Male Inflorescence of Maize (Zea mays L.)

Maize tassel is the male reproductive organ which is located at the plant’s apex; both its morphological structure and fertility have a profound impact on maize grain yield. More than 40 functional genes regulating the complex tassel traits have been cloned up to now. However, the detailed molecular mechanisms underlying the whole process, from male inflorescence meristem initiation to tassel morphogenesis, are seldom discussed. Here, we summarize the male inflorescence developmental genes and construct a molecular regulatory network to further reveal the molecular mechanisms underlying tassel-trait formation in maize. Meanwhile, as one of the most frequently studied quantitative traits, hundreds of quantitative trait loci (QTLs) and thousands of quantitative trait nucleotides (QTNs) related to tassel morphology have been identified so far. To reveal the genetic structure of tassel traits, we constructed a consensus physical map for tassel traits by summarizing the genetic studies conducted over the past 20 years, and identified 97 hotspot intervals (HSIs) that can be repeatedly mapped in different labs, which will be helpful for marker-assisted selection (MAS) in improving maize yield as well as for providing theoretical guidance in the subsequent identification of the functional genes modulating tassel morphology. In addition, maize is one of the most successful crops in utilizing heterosis; mining of the genic male sterility (GMS) genes is crucial in developing biotechnology-based male-sterility (BMS) systems for seed production and hybrid breeding. In maize, more than 30 GMS genes have been isolated and characterized, and at least 15 GMS genes have been promptly validated by CRISPR/Cas9 mutagenesis within the past two years. We thus summarize the maize GMS genes and further update the molecular regulatory networks underlying male fertility in maize. Taken together, the identified HSIs, genes and molecular mechanisms underlying tassel morphological structure and male fertility are useful for guiding the subsequent cloning of functional genes and for molecular design breeding in maize. Finally, the strategies concerning efficient and rapid isolation of genes controlling tassel morphological structure and male fertility and their application in maize molecular breeding are also discussed.


Introduction
Maize (Zea mays L.) is an important staple crop which has been widely used as food, animal feed and industrial raw materials [1]. According to FAOSTAT (https://www.fao.

Introduction
Maize (Zea mays L.) is an important staple crop which has been widely used as food, animal feed and industrial raw materials [1]. According to FAOSTAT (https://www.fao.org/faostat, accessed on 16 February 2022), maize has been the second most cultivated crop after wheat, but its total production and grain yield per unit area are highest among the main cereal crops, such as wheat and rice. Worldwide, 1162.4 million tons (Mt) of maize was produced in 2020, with an average productivity of 5.75 t/ha (FAO-STAT, 2020). However, the demand for maize in the developing world is increasing [2] and there is still an urgent need to improve maize yield.
Maize belongs to the Poaceae family, which contains many important crops such as sorghum, wheat and rice [3]. In contrast to these crop species, maize is monoecious with two types of unisexual inflorescences, i.e., the terminal staminate inflorescence (called tassel), which directly derives from the shoot apical meristem (SAM) that formed during embryogenesis, and the lateral pistillate inflorescence (called ear), formed in the axil of one or more leaves [4]. In fact, maize flowers are initially bisexual in the floral organ stage, with degeneration of the pistil primordia in tassel florets and staminal primordia in ear florets occurring thereafter [5,6]. As a result, the tassel and ear become male and female inflorescences, respectively [5,6]. The mature tassel consists of several basal branches and a central spike, which together bear the spikelets with two florets. The complex morphology of the mature tassel is usually features a number of sub-traits, the most researched of which are tassel length (TL), central spike length (SL), tassel branch angle (TBA), branch length (BL), tassel weight (TW) and tassel branch number (TBN) (Figure 1) [7][8][9]. Several studies have indicated that maize was domesticated from teosinte (Z. mays subsp. parviglumis) in the Balsas region of southwest Mexico about 9000 years ago [10,11]. Multiple morphological differences have accumulated in maize after domestication, [12,13], such as the tassel traits, and it is now one of the most important morphological characteristics for taxonomical comparisons between them [14]. The most evident difference in tassel traits is that the tassels of cultivated maize usually possess fewer branches, Several studies have indicated that maize was domesticated from teosinte (Z. mays subsp. parviglumis) in the Balsas region of southwest Mexico about 9000 years ago [10,11]. Multiple morphological differences have accumulated in maize after domestication [12,13], such as the tassel traits, and it is now one of the most important morphological characteristics for taxonomical comparisons between them [14]. The most evident difference in tassel traits is that the tassels of cultivated maize usually possess fewer branches, whereas the tassels of teosinte are highly branched and have more than 50 branches [9]. A larger male inflorescence would decrease the photosynthetic efficiency because of the shading effect on upper leaves and it would also compete for photosynthates with the ear, so the size of the tassel is thought to be negatively correlated with the biomass and grain yield of maize [15]. Consequently, the tassel size tends to be reduced during the process of domestication and selection [16]. For example, a significant decrease in male inflorescent-related traits, such as TBN, from landraces to temperate-adapted lines was observed [16]. Duvick and Cassman [17] found that the TW of Pioneer hybrids decreased about 36% from 1967 to 1991. By examining 350 inbred lines from multiple eras of germplasm in both the USA and China, Wang et al. [18] revealed that breeders preferred to select varieties with smaller tassel size, which showed a reduction in TBN. Maize is one of the most successful crops in utilizing heterosis, and the hybrids are widely used in maize production. This reduction is possibly the result of selecting hybrids with moderately small tassels for enhancing maize grain yield [19]. However, tassels will become very small if this trend continues, eventually resulting in a difficulty maintaining and propagating maize inbred lines for hybrid seed production. In practice, breeders usually select male parents with a relatively larger tassel and plenty of pollen grains released over a desired period of time, and female parents with a smaller tassel and a relatively larger ear so that more energy is used for kernel development [8].
In hybrid seed production, male parents with sufficient quantities of fertile pollen grains released over a desirable period are preferred. Plant anther and pollen development is a complicated process governed by the synergistic expression of multiple genes, and any disturbance of them may result in male sterility [38]. In rice, cytoplasmic male sterility (CMS) lines caused by both nuclear and mitochondrial genes or photoperiod and thermosensitive genic male-sterile (P/TGMS) lines caused by nuclear genes have been widely utilized since the 1970s in China [39,40]. However, due to the inseparable inheritance of disease susceptibility and male sterility, the T-type CMS lines were no longer widely used commercially in maize hybrid seed production since the epidemic of southern corn leaf blight in USA [41]. Recently, the biotechnology-based male-sterility (BMS) systems by using genic male sterility (GMS) genes have been developed and showed great potential in grain yield enhancement by improving the efficiency of hybrid seed production and heterosis application [42][43][44]. Therefore, understanding the regulation network of anther and pollen development, deciphering the mechanisms of male sterility and creating new GMS lines are meaningful for maize breeding and production.
Plant anthers arise from the floral meristem, and studies have shown that the developmental process of maize anther and pollen are highly consistent with those of rice and Arabidopsis [43,45,46]. Each maize anther contains four identical locules; each locule has four layers of somatic tissues, i.e., epidermis, endothecium, middle layer and tapetum, from outermost to the innermost layer [43,45,46]. The tapetum plays many crucial functions, including secretion of nutrients to support gametophyte development and synthesis of most pollen exine components [43,[45][46][47]. Consequently, many characterized GMS mutants in maize exhibited tapetal defects [43,44,48,49]. Pollen development comprises three major developmental stages, that is, sporogenesis, post-meiotic development of free microspores and microspore mitosis [50,51]. The pollen wall, which is comprised of an exine and intine, also has a complex structure [52,53]. Multiple genes participating in pollen wall development have been reported [43,44,48,49]. Transcriptome analysis showed that both anther and pollen development involve transcription of tens of thousands of genes in maize [53][54][55][56], most of which may play specific functional roles during anther and pollen development. To date, hundreds of GMS mutants have been identified and characterized and more than 30 GMS genes have been cloned in maize [43,44,48,49], which has greatly enhanced our understanding of the cellular, genetic and molecular mechanisms of plant anther and pollen development.
Some excellent reviews have summarized the molecular control of grass inflorescence development [57][58][59], the potential regulatory pathways modulating inflorescence architecture in maize [60], and specific aspects of the maize GMS gene form [43,44,48,49]. However, very few of them have focused on the genetic structure and molecular mechanisms underlying the formation of tassel, anther and pollen in the male inflorescence of maize. In this review, we summarized the existing literature concerning the cytological, molecular and genetic basis of maize tassel morphology and male fertility, and constructed the molecular networks underlying tassel morphology and male fertility for increasing maize grain yield by genetically manipulating the tassel and fertility traits.

Cellular Developmental Process of Maize Male Inflorescence
Maize inflorescences develop in a strictly organized manner and form different types of meristems. Compared with the ear, the tassel has distinctive and highly varied branching inflorescences, due to the diverse activities of its axillary meristems (AMs) that form varied lateral organs ( Figure 1). The SAM is a stem cell niche located at the plant apical pole that depends on meristem activity for continuous proliferation and initiation of organs throughout plant development. Three zones of cells, the central zone (CZ), the organizing center (OC) and the peripheral zone (PZ), are recognized in the SAM and the latterly initiated meristems [61,62]. The CZ possesses multipotent stem cells that proliferate continually to supply a primitive cell population to the PZ, which is the starting location for leaf primordia and the AMs that subsequently generate vegetative branches or inflorescences. The OC is located below the CZ and regulates stem cell proliferation in the CZ as well as stem cell differentiation in the PZ.
In response to endogenous and exogenous signals, SAM converts into the inflorescence meristem (IM), beginning the stages of reproductive growth [63]. First, the base of the IM initiates varying numbers of primordia, called branch meristems (BMs), the quantity and rate of which are important factors for the morphology of the tassel. However, the initiation process is absent from the ear development. Then, the peripheral regions of BM and IM form transient spikelet-pair meristems (SPMs), which are also considered indeterminate meristems due to the presence of stem cells in SPM. Each SPM generates a pair of spikelet meristems (SMs), which continue to develop and produce a glume primordium to form glume that wraps the flower organs; then, the upper-floret meristem (UFM) and lowerfloret meristem (LFM) successively initiate, which in turn develop as upper and lower florets, respectively [4,64,65]. Each FM then forms a floral organ including three anthers, two lodicules and a palea. Gynoecial development arrests in male inflorescence, and male gametes are produced within the anthers ( Figure 1) [4,64,65].
Maize anthers are comprised of the sporophyte tissue and gametophyte, which are radially symmetrical. The male gametophyte is highly specialized in maize as well as other angiosperm plants to two-or three-celled pollen grains, and its development is relatively uniform [66,67]. The connective tissue that unites the four anther locules, the vascular bundle and the anther wall together show a butterfly shape in the transverse plane and make up the sporophyte portion ( Figure 1). Thus, the development and cell composition of sporophyte tissues are substantially more complicated than those of gametophytic microspores [67]. The developmental process of the maize anther is divided into 14 stages, which include four major developmental phases, i.e., phase I (archesporial (AR) cell specification), phase II (anther somatic cell division), phase III (tapetum development and pollen mother cell (PMC) formation) and phase IV (mature pollen formation and anther dehiscence) ( Figure 2) [43,68]. The AR cell is the origination of male gametes in the plant germline [69]. Phase I includes anther developmental stages 1 and 2. Stamen primordium initiates and forms L1, L2 and L3 layers of cells at stage 1. Then, the formation of anther primordia: the stamen primordia differentiate to form sub-rectangular anther primordia, at which point the L2 layer of cells forms AR cells at the center of each corner. Phase II includes stages 3-5. Primary parietal cells develop and two layers of secondary parietal cells (SPL) form during stage 3 and 4. Sporogenous cells (Sp), the middle layer and the tapetum formation mainly take place at stage 5. Phase III includes stages 6-9. At this phase, microspore mother cells form a tetrad through meiosis I and II. Additionally, the middle layer becomes ribbon-like and progressively vanishes from stage 7, the cytoplasm of the tapetum is concentrated and the coloration deepens at stage 8a, and is further condensed and vacuolated at stage 8b. At stage 9, the callose outside tetrads is degenerated and microspores are released, tapetal cells become condensed and the middle layer vanishes. Phase IV includes stages 10-14. At stage 10, microspores are round and vacuolated, and tapetum degeneration occurs. At stage 11, the starch starts accumulating inside the microspore which shows falcate shape, and the tapetum has almost completely degraded into cellular debris. At stages 12 and 13, starch granules are continually accumulated into microspores, and mature pollen grains are gradually produced in the locule. Anthers release pollen grains at stage 14.
stages, which include four major developmental phases, i.e., phase I (archesporial (AR) cell specification), phase II (anther somatic cell division), phase III (tapetum development and pollen mother cell (PMC) formation) and phase IV (mature pollen formation and anther dehiscence) ( Figure 2) [43,68]. The AR cell is the origination of male gametes in the plant germline [69]. Phase I includes anther developmental stages 1 and 2. Stamen primordium initiates and forms L1, L2 and L3 layers of cells at stage 1. Then, the formation of anther primordia: the stamen primordia differentiate to form sub-rectangular anther primordia, at which point the L2 layer of cells forms AR cells at the center of each corner. Phase II includes stages 3-5. Primary parietal cells develop and two layers of secondary parietal cells (SPL) form during stage 3 and 4. Sporogenous cells (Sp), the middle layer and the tapetum formation mainly take place at stage 5. Phase III includes stages 6-9. At this phase, microspore mother cells form a tetrad through meiosis I and II. Additionally, the middle layer becomes ribbon-like and progressively vanishes from stage 7, the cytoplasm of the tapetum is concentrated and the coloration deepens at stage 8a, and is further condensed and vacuolated at stage 8b. At stage 9, the callose outside tetrads is degenerated and microspores are released, tapetal cells become condensed and the middle layer vanishes. Phase IV includes stages 10-14. At stage 10, microspores are round and vacuolated, and tapetum degeneration occurs. At stage 11, the starch starts accumulating inside the microspore which shows falcate shape, and the tapetum has almost completely degraded into cellular debris. At stages 12 and 13, starch granules are continually accumulated into microspores, and mature pollen grains are gradually produced in the locule. Anthers release pollen grains at stage 14.

Genes and the Molecular Regulation Mechanisms of Male Inflorescence Development and Tassel Morphology
As mentioned above, the tassel is one of the most complex organs in maize. The processes of its initiation, development and morphology formation are well-orchestrated and controlled by multiple genes [7,9,21]. Most of the known genes controlling tassel development were identified from mutants through transposon tagging or positional cloning, which provided valuable information on how a single gene affected tassel development.
Here, we summarize all the cloned genes regulating tassel development and morphology (Table 1). Furthermore, their molecular mechanisms are also proposed based on current knowledge ( Figure 3). cesses of its initiation, development and morphology formation are well-orchestrated and controlled by multiple genes [7,9,21]. Most of the known genes controlling tassel development were identified from mutants through transposon tagging or positional cloning, which provided valuable information on how a single gene affected tassel development.

Meristem Activity and Maintenance
Accurate control of plant stem cell proliferation is essential for the reproducible and continuous development of plant organs. Stem cell homeostasis is regulated by a deeply conserved negative feedback signaling between the CLAVATA (CLV) and WUSCHEL (WUS) pathways throughout plant life ( Figure 3). WUS encodes a homeodomain transcription factor (TF), which is expressed in the OC and moves to the CZ to promote stem cell proliferation and repress differentiation, and its mutation causes a loss of SAM [116,117]. In the CLV signaling pathway, which controls stem cell differentiation, CLV3 and the related CLV3/EMBRYO-SURROUNDING REGION (CLE) members are small peptide ligands whose expressions are induced by WUS in the CZ; the peptide of CLV3 is perceived by receptor-like kinase CLV1 and related CLV1-family members or by receptor-like protein CLV2 combined with a receptor kinase CORYNE (CRN), and the signal is then transmitted to restrict the expression of WUS in OC cells [118][119][120][121]. The mutation of CLV1, CLV2 or CLV3 results in increased IM size, as well as increased numbers of floral organs [116,122]. Receptors for CLV3 have been further expanded with additional studies, such as members of the BARELY ANY MERISTEM (BAM) family and CLAVATA3 INSENSITIVE RECEPTOR KINASES (CIK) family, which indicate that the regulation of stem cell fate by CLV3 signal transduction is completed in plants [123][124][125].
The components of the CLV-WUS regulatory pathway are conservative in different plant species. ZmWUS1 and ZmWUS2 are predicted as the WUS orthologs by phylogenetic analysis in maize [77]. Barren inflorescence3 (Bif3) is a dominant mutant which possesses a tandem duplicated copy of ZmWUS1, and its inflorescences appear defects in meristem initiation and maintenance, indicating ZmWUS1 also plays a crucial role in meristem size regulation in maize [126]. ZmCLE7, ZmCLE14 and ZmFON2-LIKE CLE PROTEIN1 (ZmFCP1) are CLV3 orthologs, and FASCIATED EAR2(FEA2) and FEA3 are maize orthologs of CLV2 respectively [72,127]. FEA2 can transmit signals from ZmFCP1 and ZmCLE7 through ZmCRN and the α subunit of maize heterotrimeric G protein COMPACT PLANT2 (CT2/Gα), but ZmCLE7 and ZmFCP1 transmit signals specifically via CT2 and ZmCRN, respectively [72,127]. Further information shows that G protein β subunit (ZmGB1) controls the meristem size in maize and functions together with CT2 during inflorescence development [76,128]. FEA3 transmits the ZmFCP1 signal, but genetic analysis indicates that FEA3 functions in a distinct pathway with ZmCRN and CT2 [76,128].
THICK TASSEL DWARF1 (TD1) is the ortholog of CLV1. The td1 mutant shows an increase in the spikelet density of maize tassel, but a synergistic phenotype appears in the td1;fea2 double mutant compared to that of either single mutant, indicating that TD1 and FEA2 function in different pathways [74].
In addition, KNOTTED 1-Like Homeobox (KNOX) proteins also play crucial roles in meristem activity and maintenance [85]. KNOX and BELL1-like homeobox (BLH) proteins belong to the three-amino acid loop extension (TALE) homeodomain superfamily, which interacts to produce functional heterodimers [129]. Maize KNOTTED 1 (KN1) is the first described KNOX gene in plants. BLH12 and BLH14 are cofactors of KN1, which interact with KN1; the blh12;14 double mutant fails to maintain AMs and has highly branched tassels, which is similar to the kn1 loss-of-function mutant [86,129]. KN1 cistrome shows that KN1 binds to thousands of loci, and the direct targets are strongly enriched for key regulators involved in the cytokinin, gibberellic acid (GA) and auxin signaling pathways [86]. KN1 can directly activate and bind to ga2ox1, which encodes an enzyme that inactivates GA, and the KN1 binding site of ga2ox1 genes is also present in different gramineous plants, indicating that the local regulation of GA levels by KNOX proteins is conserved in gramineous plants [87]. However, KN1 has numerous targets that need to be further explored in order to better understand its important role in meristem maintenance.
FEA4 is orthologous to the Arabidopsis PERIANTHIA, which encodes a basic region leucine zipper (bZIP) TF. The most significant phenotypes of fea4 are a significantly thickened tassel and much higher spikelet density than the wild-type (WT) plants [73,130,131]. Phenotypic analysis of fea4;fea2 double mutant suggests that FEA4 regulates meristem size in a different mechanism to the CLV-WUS pathway [73,130,131]. MALE STERILE CONVERTED ANTHER1 (MSCA1) is a maize CC-type glutaredoxin that can modify FEA4 activity through adjustment of the redox (reduction/oxidation) state and direct interaction with FEA4, thus controlling the meristem size [73,130,131]. Redox signaling also plays an important role in controlling meristem growth by regulating related genes such as WUS [132]. Thus, modification of the Redox signaling pathway should be a potential way to manipulate plant meristem activity.

Axillary Meristem Initiation and Formation in the Reproductive Phase
After the vegetative development shifts to reproductive, the radially symmetric SAM is progressively disrupted with laminar initiation of the AM, which is tightly associated with auxin accumulation in the IM peripheral zone. Auxin plays a key role in the initiation of AM ( Figure 3). VANISHING TASSEL2 (VT2) encodes a co-ortholog of Arabidopsis TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS1 (TAA1), which converts tryptophan to indole-3-pyruvic acid [97]. SPARSE INFLORESCENCE1 (SPI1) is a flavin monooxygenase with a similarity to Arabidopsis YUCCA (YUC) genes, which also work in the auxin biosynthesis pathway [91]. Loss of function of VT2 or SPI1 leads to defective initiation of the axillary meristem in maize [97]. Degradation of AUXIN/INDOLE-3-ACETIC ACID (Aux/IAA) proteins is a crucial step for auxin signaling in plants [133]. BARREN INFLORESCENCE1 (BIF1) and BIF4 encode Aux/IAA proteins in maize; initiation of AMs in the inflorescence of bif1 and bif4 mutants is inhibited, and as the result the numbers of branches and spikelets of mature tassels is largely reduced [80,81]. Upon the polar transportation of auxin to the PZ region of the SAM, BIF1 and BIF4 are degraded, which activates the expression of AUXIN RESPONSE FACTOR (ARF) TFs to promote the expression of BARREN STALK1(BA1) [81]. BA1 encodes a bHLH TF which regulates the initiation of AMs, and the ba1 mutant produces plants without reproductive AMs [79]. BARREN STALK FASTIGIATE1 (BAF1) encodes an AT hook TF that can transcriptionally regulate BA1 expression, and both the TBN and TBA are decreased in the baf1 mutant [113]. BIF2 encodes a PINOID-like serine/threonine protein kinase, which can phosphorylate BA1 and the auxin transport protein ZmPINFORMED1a (ZmPIN1a) [83,84,134]. BIF2 positively regulates auxin transport [135], and the tassel branches, spikelets, and florets are reduced in bif2 mutants [82]. The ba1;bif2 double mutant exhibits slight additive effects when compared to the single mutant, indicating that BIF2 and BA1 play overlapping and distinct roles in inflorescence development [83,84,134].
BA2 is the co-ortholog of rice gene LAX2, encoding a protein that heterodimerizes and co-localizes with BA1 in the nucleus [64,136]. The ba2 mutant is similar to ba1, indicating ba2 also plays a role in reproductive AM development [64]. However, double mutant analysis showed that BA2 functions in parallel with BAF1 and BIF2 in reproductive AM development [64].
Combining the above information, auxin biosynthesis and signaling play a central role in the generation of AMs. However, genetic interaction tests for BA1, BA2, BAF1 and BIF2 suggest that more than one regulatory pathway may exists during this developmental process in maize, and much more efforts should be made for further elucidation of the molecular mechanism regulating AM initiation.

Regulation of Meristem Fate
The initiation and development of each type of meristem are tightly regulated during inflorescence development. The main IM first initiates a number of BMs that form long branches of the tassel in normal maize tassels. It was proposed that Unbranchded2 (UB2) and UB3, which encode SQUAMOSA promoter binding (SBP)-box TFs, control the rate of axillary primordia initiation [95]. The mutation of UB2 or UB3 results in significant reduction in TBN [95]. Another SBP-box TF, TASSELSHEATH4 (TSH4), is also required in branch meristem initiation and maintenance; the tsh4 mutant has reduced TBN and presents long bract leaves in place of tassel branches [93,95]. The phenotype of the ub2;ub3 double mutant is enhanced by the loss of function of TSH4, indicating the functional redundancy of these three genes [93,95]. Expression analysis showed all three genes were expressed complementarily to miR156, indicating these SBP-box TFs are targets of miR156 [95].
LIGULELESS2 (LG2) encodes a bZIP protein and was initially reported to regulate leaf angle by narrowing the ligular region during maize leaf primordia development [137]. In the reproductive stage, the lg2 mutant shows a reduced number of TBN compared to its WT siblings [88]. This indicates LG2 plays an important role in the transition from the vegetative stage to the reproductive stage in the shoot apex. However, the molecular mechanism of LG2 regulating the fate of branch meristems remains unclear.
The branching and fate of SPM are also significantly regulated by the RAMOSA pathway [138,139], in which the known genes RAMOSA1 (RA1), RA2 and RA3 encode different types of proteins (Figure 3). RA1 and RA2 encode a Cys2-His2 zinc-finger protein and a LOB-domain-containing TF, respectively [104]. RA3 encodes a functional trehalose-6phosphate phosphatase (TPP) and regulates inflorescence branching by mediating sugar signals [105]. The phenotypes of these mutants similarly give rise to a highly branched tassel, indicating the potential regulatory role of RAMOSA genes on meristem determinacy. Further molecular and genetic analysis show that RA1 acts downstream of both RA2 and RA3 [104,105]. Genetic analysis indicates that TSH4 is epistatic to the three RA genes, and expression analysis shows TSH4 has a mutual negative regulatory relationship with RA2 but not with RA1 or RA3 [93]. Similar to Arabidopsis TOPLESS protein, RAMOSA ENHANCER LOCUS2 (REL2) encodes a transcriptional repressor. The mutation of REL2 enhances the phenotype of ra1 and ra2, and REL2 can physically interact with RA1, but not with RA2 and RA3, which was validated by both BiFC and two yeast hybrid assays [106]. The paired spikelets in specific Andropogoneae plants, including maize, are thought to be an evolutionary trait [140]. The Suppressor of sessile spikelets1 (SOS1) mutation resulted in the generation of a single spikelet in the inflorescence, and reduced TBN [140]. Although the SOS1 gene has not been cloned yet, investigation of the interaction between SOS1 with RAMOSA genes indicated that SOS1 functions downstream of RA1, and SOS1 may also have a positive regulatory role with RA1, which forms a negative feedback loop in controlling SPM determinacy [140]. However, despite the progress made in genetic and molecular studies of RAMOSA genes regulating the SPM fate, it is still unclear how this important pathway works.
The identity and determinacy of the SM are crucial for the biological function of the male inflorescence (Figure 1). Branched silkless1 (BD1) encodes an ethylene-responsive element-binding factor (ERF) TF. In the bd1 mutant, the identity of the spikelet meristem is altered and produces a series of spikelets indeterminately in the tassel [99]. Further genetic and expression analysis showed that BD1 negatively regulates TSH4 protein accumulation [93,95].
Plant miRNAs are endogenous short-length RNAs that post-transcriptionally repress gene expression and play crucial roles in regulating plant vegetative and reproductive developmental processes [141]. RNA endonuclease DICER-LIKE1 (DCL1) is the core component for miRNA processing during plant miRNA biogenesis; mutations of Arabidopsis and rice DCL1 genes have led to obvious developmental defects [142,143]. SPMs of the fuzzy tassel (fzt) mutant produce more than two SMs and SMs produce extra FMs, which resemble the tassel seed4 (ts4) mutant [110]. Positional cloning showed that fzt encodes a mutation in a dcl1 homolog while ts4 is caused by a dysfunction mutation in miR172e [110,111]. Expression of miR172e is substantially reduced in dcl1-fzt mutants, suggesting miR172e is possibly regulated by DCL1 and that miRNAs are required for meristem determinacy. Furthermore, TS4 targets the indeterminate spikelet1 (IDS1) and the sister of indeterminate spikelet1 (SID1), both of which encode APETALA2 (AP2)-like proteins that are required for IM branching, SM determinacy and FM initiation [101,111].
Taken together, the determinacy of the meristem fate is a much more complicated process that involves different types of genes. However, interactions of the molecules among different pathways still remain largely unknown, and it will be interesting to test the relationship between miRNA and RAMOSA pathways, considering that mutual negative regulation between RA2 and TSH4 has been detected [105].

Regulation of TBA
TBA is one of the main traits in tassel morphology; a larger TBA increases the dispersed area of the pollen but affects photosynthesis because of the shading on the leaves [15]. One reason for the angle of the spikes is the presence of the pulvinus that lies between the main axis and the branch of the spikes; in addition, the degree of softness and stiffness of the spikes can cause variations in the branch angle [112,114]. The maize gene BRANCH ANGLE DEFECTIVE (BAD1) has been shown to encode a plant-specific TCP TF and to be expressed in the tips of leaf primordia, SPM, SM and husk leaves [112]. In the bad1 mutant, the size of the pulvinus is decreased, affecting TBA. Interestingly, RA2, which affects the identity of SM, and REL2, which is the enhancer of RA1 and RA2 of the RAMOSA pathway, also have impact on TBA [104,106]. BAD1 expression and protein accumulation have a large reduction in ra2 mutant inflorescences, suggesting RA2 functions positively upstream of BAD1 in controlling TBA in maize. LIGULELESS1 (LG1), which was reported to initially regulate the leaf angle [144], was also found to regulate TBA, whereas the expression pattern of BAD1 has no influence on the lg1 mutant, implying LG1 may function parallelly with BAD1 [112].
BAF1 participates in AM formation, and the primary branches of baf1 tassels are more upright and shorter compared with those of WT plants [113]. In the baf1 mutant, a reduced pulvinus was clearly observed in the base of lateral branches, while the pulvinus of the rel2 mutant showed an increase in lignification but nearly identical size and shape compared with WT plants, indicating different mechanisms underlying TBA formation exist in maize [106].

Genetic Basis of Tassel Traits Revealed by Quantitative Genetics Methodologies
Most of the tassel traits, including TL, SL, TBA, TW, TBN, total branch length (TBL) and others, can be quantified simply, inexpensively and precisely, making maize tassel architecture one of the systems susceptible to QTL research [8]. With recent advances in molecular biology techniques, many tassel architecture-related QTLs have been mapped in the maize genome. Berke and Rocheford [7] first reported the QTLs impacting tassel traits, including TBA, TBN and TW, and detected six, three and seven associated QTLs, respectively. Latterly, Mickelson et al. [20] identified three TBA and six TBN QTLs using a population composed by recombinant inbred lines (RILs). Upadyayula et al. [21] then detected QTLs that influence developmentally relevant patterns of tassel architecture in a BC 1 S 1 population which derived from self-pollinated BC 1 plants. Recently, by using 866 maize-teosinte RILs and 19,838 SNP markers, a high-resolution QTL mapping for five tassel morphological traits has been conducted, and in total 72 QTLs were identified with many known maize inflorescence development related genes significantly enriched in the mapped regions [9]. In classical linkage mapping, bi-parental segregating populations were typically used, so only a limited number of genetic loci explaining partial phenotypic variation could be identified [7,8,[20][21][22][23]. The strategy of GWAS benefiting from high genetic diversity and historical accumulation of recombination events in natural populations, leading to the rapid decay of linkage disequilibrium, has been recently widely used [145]. For example, Yang et al. [29] conducted GWAS for 17 agronomic traits using 513 inbred lines, and a different number of loci were detected by the methods of the Anderson-Darling (A-D) test and mixed linear models (MLM), respectively. However, both QTL mapping and GWAS have their own merits and demerits [145][146][147]. In order to increase the accuracy of genetic mapping, a combination of GWAS and linkage mapping was recently applied to detect common quantitative trait nucleotides (QTNs) in order to uncover the genetic architecture of maize tassel traits. QTL mapping and GWAS for four tassel traits using a maize nested association mapping (NAM) population were conducted by Brown et al. [33], which identified 26-39 QTL and 241-325 SNP associations with four tassel traits. By using NAM populations from both the USA and China, and a global association panel, Wu et al. [28] dissected the genetic basis of TL and TBN, and identified a total of 125 QTLs and 965 QTNs, respectively. Wang et al. [27] evaluated TBN using 359 inbred lines and 273 doubled haploid lines; a total of 19 QTLs and 295 QTNs were identified by QTL mapping and GWAS, respectively, and one physical overlap between QTNs and QTLs was further detected.
The above QTLs/QTNs have given valuable insights into the genetic components governing maize tassel variation. Here, we summarize tassel-trait QTLs and QTNs according to the genetic studies conducted over the past 20 years, in order to construct a consensus map revealing the genetic architecture of tassel-related traits. All the positional information was normalized to the B73 reference genome version 4 (RefGen_v4.0, https://www.maizegdb.org/, accessed on 16 February 2022), and a total of 586 QTLs and 1816 QTNs for seven tassel traits were collected. The chromosomal distribution of these QTLs and QTNs was firstly examined, revealing that the number of both QTLs and QTNs on chromosome 1 is the most abundant ( Figure S1), implying that chromosome 1 may contain more tassel-trait related genes. Consistent with this assumption, relatively larger numbers of genes have been cloned from chromosome 1 ( Table 1). For a specific trait, TBN, for example, the number of both QTLs and QTNs on chromosome 2 is higher than on other chromosomes ( Figure S1), indicating TBN-related genes are enriched on chromosome 2, or some of the genetic loci can be repeatedly identified from different populations or under different environments. The repeatedly mapped loci should be more important for further gene cloning and molecular breeding. For the QTLs, the intervals that could be mapped by three or more independent studies and QTLs with intervals <20 Mb were defined as QTL-hotspot intervals (QTL-HSIs). For the QTNs, the regions containing four or more significant SNPs in a 1 Mb interval by using the sliding window analysis method were defined as QTN-hotspot intervals (QTN-HSIs). The ability of the HSI to anchor to the same physical interval by using different genetic mapping populations and methods provides strong evidence that the HSI is important in regulating the target traits. Such an association would be more convincing if the HSI also contains known genes regulating male inflorescence development.
Meanwhile, a total of six common HSIs (CHSIs) were found between the TBN QTN-HSIs and TBN QTL-HSIs. LG1 and LG2, which are involved in inflorescence development, are located in the CHSIs chr2: 3. 42 [9]. In addition, the CHSIs without known genes are possibly the most important targets for cloning new tassel-trait genes. LG2 still have the possibility of containing new candidate genes, as tassel traits have highly polygenic genetic architectures [9]. In addition, the CHSIs without known genes are possibly the most important targets for cloning new tassel-trait genes.

Application of the Quantitative Genetic Variation Regulating Tassel Traits in Maize Breeding
Many studies have shown that the majority of the identified loci have minor impacts on tassel traits, and therefore cannot be simply and directly employed in molecular breeding. However, there has been some progress in the application of the quantitative genetic variations in maize tassel-trait improvement recently. For example, one QTN (Q Dtbn1 ) for TBN was detected by GWAS located in the 5 -upstream of an F-box/kelch-repeat protein coding gene; the TBN was significantly reduced in F-box gene over-expression and the hybrid lines, indicating that Q Dtbn1 negatively influences TBN via a dominant model, and making Q Dtbn1 particularly valuable in maize hybrid breeding [89].
Fine-tuning of the inflorescence meristem-related genes provides another strategy for their utility in maize breeding. FEA2 is a core component in the CLV-WUS regulatory pathway and controls stem cell differentiation; the fea2 mutant showed higher spikelet density in the tassel central spike and a shorter, wider and flattened ear compared with normal plants [71]. A series of weak alleles of FEA2 were generated by ethyl methane sulfonate (EMS) mutagenesis and phenotypic analysis showed that the IMs of these fea2 weak mutants were substantially larger than WT siblings. Further yield parameter measurement confirmed that these weak alleles of FEA2 can increase kernel row number (KRN) without compromise in ear length or causing the fasciation that occurs in the fea2 loss-of-function mutant, which is detrimental to yield [127]. Interestingly, a series of weak promoter alleles of ZmCLE7 and two null alleles of ZmCLE1E5, which is a partially redundant compensator for ZmCLE7, were generated through CRISPR/Cas9 genome editing [148]. Similar to the fea2 weak mutants, the IMs of these types of mutants were significantly larger than those of their controls [148]. Many yield-related traits, including KRN, cob diameter, ear diameter, ear weight, kernel depth and grain yield per ear were significantly increased, and notably, unlike the Zmcle7 null allele, the ears of these mutants were not fasciated [148]. One of the important phenomena is that the tassel length was slightly affected in Zmcle1e5 mutants [148], implying a potential for manipulating inflorescence development-related genes for tassel-trait improvement. Considering that the null mutation of most of those genes results in serious defects in tassel development, the generation of weak alleles through CRISPR-based/EMS mutagenesis or by exploitation of compensation and redundancy mechanisms should also be more applicable in tassel-trait modification [148].

The Cloned Maize GMS Genes and Their Roles in Anther and Pollen Development
Plant anther and male gamete development is a fairly complicated process, and abnormalities occurring at any stage of anther development may cause male sterility [38]. Studies have shown that male fertility in maize also has a highly polygenic genetic architecture. More than 200 maize GMS mutants have been characterized so far [149], while only 34 GMS genes were successfully cloned [150], which can be classified into four categories according to their functional properties ( Table 2).
It is noteworthy that 15 new GMS genes have recently been identified and validated using the CRISPR/Cas9 system [68,150,151], showing the power of gene editing technologies in maize genomics and breeding. Here, a molecular regulatory network for maize male fertility is summarized and updated by considering the functional properties and sequential expression patterns of the reported maize GMS genes ( Figure 5).

MAC1
Small secreted protein S6 Suppresses excess AR proliferation, triggers periclinal division of subepidermal cells

GMS Genes in Phase I (Archesporial Cell Specification)
The AR cell is the origination of gametes in the plant germline [69]; relatively few functional genes at this early phase of anther development were identified in maize (Figure 5; Table 2). ZmMs22/MSCA1 encodes a CC-type glutaredoxin (GRX) and regulates the   AmACOS5-2 Acyl-CoA Synthetase 5 -

MAC1
Small secreted protein S6 Suppresses excess AR proliferation, triggers periclinal division of subepidermal cells

GMS Genes in Phase I (Archesporial Cell Specification)
The AR cell is the origination of gametes in the plant germline [69]; relatively few functional genes at this early phase of anther development were identified in maize ( Figure 5; Table 2). ZmMs22/MSCA1 encodes a CC-type glutaredoxin (GRX) and regulates the initiation of AR cells [201]. In the msca1 mutant, AR cells fail to progress and instead differentiate into vasculature, but reductive treatments can rescue AR cell specification in msca1 anthers, indicating that hypoxia is crucial for germ cell fate [201]. Rice MICROSPORELESS1 (MIL1) and Arabidopsis ROXY1/2 encode close orthologs of MSCA1, with the male-sterility phenotypes of their mutants similar to msca1 [199,202]. In Arabidopsis, ROXY1/2 directly interacts with TGACG (TGA) motif-binding proteins TGA9 and TGA10, two bZIP TFs that are expressed throughout early anther primordia, and the tga9/tga10 double mutant shows male sterility, suggesting TGA9 and TGA10 work redundantly in anther development [153]. In maize, direct interaction between ZmMs22/MSCA1 and TGA type TF FEA4 was verified, but whether ZmMs22/MSCA1 interacts with the TGA9 and TGA10 orthologs has not been confirmed. Three homologs of TGA9, ZmTGA9-1, -2, and -3, were found in the maize genome, and all of them are expressed in early anther primordia [68]. The CRISPR/Cas9mediated gene mutagenesis verified that neither of the single or double-gene mutants show male sterility; only the ZmTGA9-1/2/3 triple-gene mutants lose the male sterility and fail to exude pollen grains, indicating the functional redundancy of the three genes in maize [68].

GMS Genes in Phase II (Anther Somatic Cell Division)
At maize anther developmental phase II, at least five GMS genes patriciate in anther somatic cell division (Figure 5), i.e., MULTIPLE ARCHESPORIAL CELLES 1 (MAC1), OUTER CELL LAYER 4 (OCL4), ZmMs23, ZmbHLH122 and ZmTGA10. MAC1 encodes a small secreted protein, which is possibly preferentially expressed in AR cells [200]. The proliferation of AR cells in the mac1 mutant is more rapid than WT, and as a result, the mac1 mutant displays extra AR cells in anthers and is completely sterile [200]. OCL4 encodes a HD-ZIP IV TF and is specifically expressed in young organ primordia and the epidermis of shoot meristems [152]. The ocl4 mutant shows partial male sterility, and its anthers exhibit an additional cell layer with endothecium cellular characteristics, which is probably the cause of the poor anthesis and dehiscence observed at maturity [152]. Analysis of the mac1/ocl4 double mutant demonstrates that OCL4 and MAC1 act parallelly in controlling the proliferation of subepidermal cells [200]. The maize TGA10 ortholog ZmTGA10 expresses before stage 6 during anther development; the pollen of knockout mutants of ZmTGA10 by CRISPR/Cas9 technology look nearly normal, but the anthers fail to dehisce [68]. Both ZmMs23 and ZmbHLH122 encode bHLH TFs, and play an important role in the differentiation of anther somatic wall layers [68,155,159]. In Zmms23 mutant anthers, the tapetum divides into two layers, forming an anther with five wall layers [155,159,203]. The Zmbhlh122 mutant displays complete male sterility, similar to Zmms23, but whether the defects of the tapetum happen during anther wall development has not yet been determined [68].

GMS Genes in Phase III (Tapetum Development and PMC Formation)
The tapetum, which occupies the innermost layer of the anther wall and surrounds the developing microspores, plays sequential and essential roles in ferrying nutrients to microspores, remodeling the extracellular callose covering the PMCs or tetrads, and secreting exine precursors onto maturing pollen grains [155,159], so that mutation of many GMS genes exhibit tapetal defects [149]. Over half of the cloned maize GMS genes function mainly in this phase, and a relatively complex molecular network, involving transcription regulation, lipid and sugar metabolism and others, was revealed ( Figure 5).
In Arabidopsis, a DYT1-TDF1-AMS-MS188-MS1 regulatory network controlling tapetum development and pollen formation has been validated [162,204]. Although the corresponding TFs (ZmMS32, ZmMS9, ZmbHLH51, ZmMYB80 and ZmMS7) are found to be encoded by GMS genes, whether a similar genetic pathway and regulatory network exist in maize needs to be investigated. Nan et al. [68,150,151] compared the mutant phenotypes of four bHLH genes, i.e., ZmMs32, ZmbHLH51, ZmMs23 and ZmbHLH122, and found that all four bhlh mutants have an aberrant tapetum. Combining several recent studies, a more complicated cascade of bHLH-regulated pathways was established ( Figure 5) [68,151,155,159]. Protein-protein interaction assays showed that ZmMS23 interacts with both ZmMS32 and ZmbHLH51, ZmbHLH122 interacts with both ZmMS32 and ZmbHLH51, and ZmbHLH51 forms homodimers. Supported further by cytological observation and gene expression analysis in each mutant, the order of essential actions is ZmMS23, ZmMS32, ZmbHLH122, and then ZmbHLH51 [68,151,155,159]. ZmbHLH122 directly activates ZmbHLH51 and interacts with ZmbHLH51 [68], but it appears that ZmbHLH122 is not a direct target of ZmMS23 or ZmMS32 [155,159].
Maize ZmMs9 is the homolog of Arabidopsis TAPETAL DEVELOPMENT AND FUNC-TION 1 (TDF1) [162]. In Arabidopsis, TDF1 (a MYB TF) works downstream of DYSFUNC-TIONAL TAPETUM 1 (DYT1) (a bHLH TF), the ortholog of ZmMS32 [205], and bHLH proteins interact with DYT1 and act downstream of TDF1 to activate the target genes through feed-forward and positive feedback loops [206,207]. We thus propose that ZmMS9 works between ZmMS32 and ZmbHLH122 ( Figure 5); nevertheless, this needs to be confirmed via experimental data. ZmMYB80 is the ortholog of AtMYB80 [208]; the regulatory relationship between ZmMYB80 and ZmMS7 has not been validated, but mutation of each gene leads to male sterility in maize [68,169,170]. ZmMs7 and ZmPHD11 encode PHD-finger TFs, and the former regulates genes involved in exine formation and tapetum development as a transcriptional activator [68,169,170]. ZmMS7 can directly interact with nuclear factor Y (NF-Y) subunits NF-YA6 and NF-YC9/12/15, and then the complex directly activates expression of ZmMT2C, the homolog of OsMT2b, both of which are involved in the tapetal cell programmed cell death (PCD) of rice and maize anthers [68,169,170,209].
The second major group of GMS genes in phase III involves lipid metabolism ( Figure 5; Table 2). Pollen exine (sporopollenin and tryphine), anther subcellular organelle membranes' wax, and cutin are mainly composed of lipidic substances, and any disturbance of lipid metabolism can result in microspore abortion and GMS [210], suggesting the importance of lipid metabolism in anther and pollen development. However, lipid biosynthesis and transport pathways involve many steps, most of which remain uncovered in maize [48,185].
The anther cutin and wax biosynthetic pathways share some common steps and functional genes with the sporopollenin biosynthetic pathway [48,185]. In addition, some proteins such as sn-2 Glycerol-3-Phosphate Acyltransferase (GPAT) are mainly involved in anther cutin and wax biosynthesis [48,185]. ZmMs33/ZmGPAT6 catalyzes the biosynthesis of glycolipids and phospholipids at the first step of the glycerolipid synthetic pathway [48,175]. The Zmms33 mutant displays an abnormal anther cuticle and endothecium chloroplast, in which starch granules are continuously filled at early developmental stages [175].
Zmms44 encodes a lipid transfer protein (LTP), which transport lipidic precursors for anther cuticle and exine formation; variation at the signal peptide cleavage site disables the secretion ability of Zmms44 protein from tapetal cells into the locule, resulting in dominant male sterility [179]. ZmMs2/ZmABCG26 encodes an ATP-binding cassette (ABC) transporter involved in the transport of sporopollenin and/or tryphine [185,187]. The ZmMS2/ZmABCG26 protein is localized in the plasma membrane, chloroplast/plastid and ER. The anthers of its mutants are shrunken without viable pollen grain produced [185,187].
There is possibly a multidimensional regulatory network among different types of GMS genes involving maize anther and pollen development. By examining the differentially expressed (DE) genes in ms23, ms32, bhlh51 and bhlh122 mutants, many lipid metabolism-related GMS genes, including ZmMs25, ZmMs26, ZmMs30, ZmMs33 and Zmms44, are DE in one or more of these mutants [151]. In addition, dozens of potential sugar metabolism-related GMS genes are predicted in the maize genome through bioinformatic analysis [49], but only one GMS gene, ZmMs8, has been cloned [185,187]. ZmMs8 encodes 1,3-galactosyltransferase in maize, and is DE in bHLH TF mutants [151]. Thus, the lipid and sugar metabolism-related DE genes are possibly regulated by bHLH TFs. Similarly, the MYB TF ZmMYB80 directly regulates expression of ZmMs25 ( Figure 5) [184][185][186]. However, the more detailed regulatory relationship between GMS TFs and lipid and sugar metabolism-related GMS genes needs to be further investigated, which is essential for the understanding of the molecular mechanism underlying maize anther and pollen formation.
Multi-types of regulatory mechanisms of miRNAs with lipid metabolism-related GMS genes or TF GMS genes have been predicted or identified [48,185,211]. For example, ZmABCG26 is a potential target of zma-miR164h-5p revealed by sequence analysis and expression pattern comparison between them, as well as by gene editing [185,187,211].

GMS Genes in Phase IV (Mature Pollen Formation and Anther Dehiscence)
Some TFs have been reported to work in this developmental phase [68]. ZmMYB33-1/-2 encode MYB family TFs and act redundantly, as only the double-gene mutant generated by CRISPR/Cas9 technology exhibits complete male sterility [68]. INDETERMINATE GAME-TOPHYTE1 (IG1), ZmLBD10 and ZmLBD27 encode LOB domain family TFs, and the ig1 mutant shows sporophytic male sterility, failing to shed pollen or extrude anthers [68,171]. However, ZmLBD10 and ZmLBD27 function redundantly, as viable pollen formation was only significantly affected by their combination in maize [68,171].
It is noteworthy that the functional stages of GMS genes in regulating anther and pollen development are not always completely consistent with their expression patterns; some GMS genes may function later than their peak expression stage, while others that act at previous stages may also function at later stages, particularly with two expression peaks [48,185].
Taken together, transcriptional regulation, lipid and sugar metabolism and redox homeostasis have been demonstrated to be involved in maize anther and pollen development ( Figure 5), and much progress in the cognizance of maize anther and pollen development has been made by dissecting the GMS mutants. Accordingly, we updated the molecular regulatory networks ( Figure 5) governing male fertility in maize by combining recent discoveries with our previous work [43,48,68,150,151,170,171,175,185,186]. However, the detailed molecular mechanisms remain largely uncharacterized. More experimental data is thus required to construct an integral molecular regulatory network underlying maize anther and pollen development.

Application of GMS Genes in Maize Heterosis Utilization
The identification of more GMS genes not only boosts our understanding of molecular mechanisms of male sterility but also facilitates the development and utilization of BMS systems in hybrid seed production and the efficient application of heterosis [43]. The utility of GMS genes for developing BMS systems has been recently reported in maize. Some highly efficient hybrid seed production systems that exploit transgene technology but generate a non-transgenic final product have been developed due to the public concern about transgenic plants [43]. ZmMs45 encodes a strictosidine synthase and its mutant shows complete male sterility [185,187]. The "Seed Production Technology" (SPT) using transgenic construct-driven non-transgenic seed systems was developed by DuPont Pioneer based on the ZmMs45 restorer gene and ms45 mutant, and the system has been commercially utilized in the USA since 2012 for maize hybrid seed production [42]. The SPT transgenic maintainer line possesses a homozygous Zmms45 allele, and a SPT construct containing a WT ZmMs45 gene for fertility restoration, a ZmAA gene to disrupt normal pollen formation and a DsRed2 gene for seed sorting (Table 3). Consequently, the produced pollen grains of maintainer line do not carry the SPT transgenes, and half of the self-pollinated seeds are transgenic but can be separated depending on the fluorescence of the DsRed2 protein by mechanical color sorting [42].
However, transgene transmission through pollen in the SPT maintainer transformants was reported up to 0.518% [42]. New technologies are needed to further decrease the rate of transgene transmission. To this end, the Multi-Control Sterility (MCS) system, by transforming a single MCS construct into maize ms7-6007, ms30-6028 or ms33-6038 GMS mutants, was developed in our laboratory [169,174,178]. Compared with the SPT, two additional functional modules, the Dam and Bar genes, were appended to the MCS construct (Table 3); however, the transgene transmission rate was greatly decreased to lower the risk of transgene flow in commercial maize hybrid seed production [169,174,178].
Dominant genic male-sterility (DGMS) systems have been also developed and tested for seed production and hybrid breeding in maize [44]. Zmms44 encodes a LTP and ZmMs44 is a dominant GMS mutant; the Zmms44-based DGMS system was created by expressing an artificial miRNA that silences Zmms44 expression [179]. Most importantly, the kernels per ear of ZmMs44 sterile plants increased at an average of 9.6% compared with WT plants, due to the reduced tassel growth partitioning more nitrogen to the ear, and the hybrids carrying the ZmMs44 allele showed significant yield increase when nitrogen is limited, indicating that the Zmms44-based DGMS system can not only reduce maize hybrid seed production costs but also provide a new strategy for increasing nitrogen use efficiency in maize [179]. Recently, a ZmMs7 DGMS system has been developed in our laboratory, where ZmMs7 was driven by the p5126 promoter coupling with a red fluorescence seed-sorting module in a single construct [170]. The p5126-ZmMs7 DGMS lines show dominant male sterility, whereas their female fertility and vegetative growth are normal [170]. Compared with other BMS systems, this DGMS system is simpler and more efficient, and can be broadly applied even for plants without applicable GMS genes [170].

Conclusions and Perspectives
Tassel morphology and male fertility, as two main aspects of male inflorescence development, are tightly correlated with maize yield. Over the past decades, more than 40 genes involving male inflorescence development have been identified. We further constructed a molecular regulatory network, and found that the CLV-WUS pathway, KN1 pathway and Redox signaling pathway play vital roles in inflorescence meristem activity and maintenance; auxin biosynthesis, transport and signaling transduction mainly participate in axillary meristem initiation and development, while RAMOSA pathway and some miRNA associated proteins regulate meristem fate and tassel morphogenesis (Figure 3). However, the precise optimization of tassel structure using these gene resources is challenging, as null alleles of many inflorescence meristem development-related genes have serious phenotypes [148]. Recently, weak alleles of CLE genes have been quickly generated by modifying the promoter sequences via CRISPR/Cas9 genome editing, which significantly improved multiple maize grain yield-related traits [148], suggesting that inflorescence meristem development related gene resources combining CRISPR/Cas9 technology have enormous potential in crop enhancement. It is interesting to generate different alleles of other male inflorescence development-related genes by CRISPR/Cas9 technology to test their breeding utilities. Considering many countries have permitted the release of gene editing crops [212], the generated elite alleles can be thus directly applied for crop improvement.
By integrating the results of genetic mapping conducted over the past 20 years, we constructed a consensus physical map for tassel traits and extracted 97 HSIs (Figure 4), which lays the foundation for subsequent candidate gene cloning, and provides valuable targets for MAS in improving maize yield via tassel morphology modification. Application of the quantitative genetic variations in Q Dtbn1 locus for tassel-trait improvement has been reported [89]. However, GWAS studies aimed at the discovery of more tassel-trait QTNs will still be the main task in the future. With the fast development of DNA sequencing technologies, our capacity to dissect genomes far exceeds the ability to measure phenotypes of plant traits. Lacking precise, high-throughput and labor-saving phenotyping technologies is one of the bottlenecks for the genetic dissection of maize tassel traits. Recently, many efforts have been made to address this imbalance. A variety of imaging approaches, such as hyperspectral imaging, have been developed, and unmanned aerial vehicles (UAVs) are also applied in crop phenotyping [213,214]. Traditional QTL mapping and GWAS methods have their own merits and demerits, with newer approaches needed to deal with the demerits. Machine learning (ML) using iterations where the computer tries to discover the patterns hidden in the current data that are then used to predict future data have shown advantages in increasing predictive power, and the use of ML approaches has become common for GWAS in recent years [215]. We suggest that high-throughput phenotyping combined with deep ML should be quickly used for the genetic dissection of tassel traits. Last but not least, MAS has not been widely used in maize tassel trait improvement; functional and linked molecular markers should be developed based on the cloned genes and HSIs mined here, which will accelerate modifications of maize tassel traits.
Maize has developed a monoecious and diclinous flowering system during species differentiation, making it an ideal model for studying inflorescence development. Heterosis has been especially successfully utilized in maize breeding for decades and male fertility control has a great significance in maize production. Four types of GMS genes, including TF, lipid-and sugar-related, as well as the unclassified type, were summarized in this study (Table 2), and some GMS genes work redundantly. Such redundancy is difficult to detect by traditional mutagenesis and map-based cloning, but CRISPR/Cas9-mediated gene mutagenesis provided a new opportunity. It is worth noting that 15 of the 34 maize GMS genes were validated via CRISPR/Cas9 technology within the last two years [68,150,151,216]. We thus constructed and updated the working model regulating maize anther and pollen development ( Figure 5). A large number of studies have demonstrated that the GMS genes tend to be specifically expressed in the anther [49,211,216]; thus, we should focus on the rapid isolation of anther-specific genes and then validate their functions by CRISPR/Cas9 mutagenesis, which is crucial for thoroughly unraveling the molecular regulatory network underlying maize anther and pollen development. Recently, different kinds of BMS systems in maize, including SPT, MCS and DGMS, have been established [42][43][44]169,170,174,178,179]; the newly identified GMS genes can be tested for their potential in increasing maize yield via the utility of BMS systems.

Conflicts of Interest:
The authors declare no conflict of interest. Beijing Solidwill Sci-Tech Co., Ltd. has the Beijing Engineering Laboratory and Beijing International Science and Technology Cooperation Base, which provided the research platform, instruments and funds for this study, so the company is one of the signed units of this paper.