Dissection of genetic regulation of compound inflorescence development in Medicago truncatula

During vegetative growth, the plant shoot apical meristem (SAM) initiates leaf primordia and axillary shoot meristems on meristem flanks. Upon sensing inducible environmental and/or developmental cues, plants transition from vegetative to reproductive growth stages and the SAM transforms into an inflorescence meristem. Depending on inflorescence meristem activities of when and where floral meristems are initiated, a wide variety of inflorescence structures arise in the plant kingdom. Based on the determinacy of primary inflorescence meristem (I₁), two types of inflorescences are defined: (1) determinate inflorescence, in which I₁ meristem terminates as a floral meristem or secondary inflorescence meristem (I₂); and (2) indeterminate inflorescence, in which I₁ indefinitely grows to produce lateral floral meristems or I₂. Based on inflorescence architectures, three major types of inflorescences are defined: raceme, cyme and panicle (Prusinkiewicz et al., 2007; Fig. S1). I₂ activities further enrich the complexity of inflorescence architectures by reiterating I₁, leading to compound inflorescences, which evolved from simple inflorescences (Stebbins, 1973; Coen and Nugent, 1994; Ma, 1998; Park et al., 2014; Teo et al., 2014).

Plant inflorescence architectures are essential for reproductive success in nature and crop productivity in agriculture. The development of inflorescence architectures is tightly controlled by genetic regulatory networks. Prusinkiewicz et al. postulated a transient model, in which the newly formed meristem is in a transient status (immature) and obtains its identity (mature) to form a flower or a branch depending on its vegetative level (Prusinkiewicz et al., 2007). The model can address the evolution and development of different inflorescence architectures, though the key regulators are different among raceme, cyme and panicle inflorescences (Prusinkiewicz et al., 2007; Lippman et al., 2008; Thompson and Hake, 2009; Park et al., 2014).

In the model species Arabidopsis and Antirrhinum, simple raceme inflorescences are developed. TERMINAL FLOWER1 (TFL1)/CENTRORADIALIS (CEN), LEAFY (LFY)/FLORICAULA (FLO) and APETALA1 (AP1)/SQUAMOSA are key players in inflorescence development (Ratcliffe et al., 1998; Liljegren et al., 1999; Davies et al., 2006; Denay et al., 2017). Arabidopsis TFL1 is closely related to the florigen gene FLOWERING LOCUS T and is upregulated upon floral transition in the apical meristem to repress the expression of floral genes, therefore maintaining the indeterminacy of apical inflorescence meristem. Mutation of TFL1 transforms an indeterminate inflorescence apical meristem to a determinate floral meristem (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992; Bradley et al., 1996, 1997; Blazquez et al., 1997). AP1 and LFY are floral meristem identity genes and play synergistic roles in maintaining floral meristem identity (Irish and Sussex, 1990; Huijser et al., 1992; Mandel et al., 1992; Gustafson-Brown et al., 1994; Moyroud et al., 2009, 2010). Double mutation of LFY and AP1 enhances inflorescence phenotypes, developing inflorescence shoots instead of flowers in Arabidopsis (Huala and Sussex, 1992; Weigel et al., 1992).

Based on extensive genetic and molecular studies in Arabidopsis, a genetic regulatory model of inflorescence development has been proposed. In this model, TFL1 is expressed in the center of apical meristems and TFL1 moves to outer layers of apical meristems to repress the expression of floral meristem identity genes AP1 and LFY, thereby maintaining the indeterminacy of apical meristems (Pidkowich et al., 1999; Ratcliffe et al., 1999; Wagner et al., 1999; Hempel et al., 2000; Blázquez et al., 2006; Conti and Bradley, 2007). This model defines the development of simple raceme inflorescences.

Some legume species are characterized by having a compound type of inflorescence development. In these species, I₁ laterally produces I₂ meristems and each I₂ meristem further develops into one to three lateral flowers before it terminates as a stub (Singer et al., 1999; Benlloch et al., 2003). PROFERATING INFLORESCENCE MERISTEM (PIM) and UNIFOLIATA (UNI)/SINGLE LEAFLET1 (SGL1) are orthologs of AP1 and LFY, respectively, in legumes. PIM has conserved functions in floral meristem identification. The pim mutants exhibit both higher-order secondary inflorescences and abnormal flowers in pea and M. truncatula (Berbel et al., 2001; Taylor et al., 2002; Benlloch et al., 2006). UNI/SGL1 plays a specific role in compound leaf regulation, in addition to its conserved function in flower development. The uni/sgl1 mutants transform compound leaves into simple leaves and produce cauliflower-like floral structures (Hofer et al., 1997; Wang et al., 2008). There are three TFL1 orthologs in pea: DETERMINATE(DET)/PsTFL1a, PsTFL1b and LATE FLOWERING (LF)/PsTFL1c. LF/PsTFL1c controls floral transition, whereas DET/PsTFL1a maintains indeterminacy of apical meristems, indicating divergent functions of different TFL1 orthologs in two distinct developmental processes in pea (Foucher et al., 2003).

By studying the non-flowering mutant vegetative1 (veg1) in pea, Berbel et al. demonstrated that PsFULc is crucial for secondary inflorescence development (Berbel et al., 2012). Therefore, they proposed a modified model in which DET/PsTFL1a maintains I₁ indeterminacy and PIM/AP1 defines the floral meristem identity, whereas VEG1/PsFULc specifies I₂ identity, a new regulatory layer added between DET and PIM to accommodate compound inflorescence development (Benlloch et al., 2007, 2015).

Medicago truncatula is a model legume species with compound leaves and compound inflorescences like pea. Upon floral transition, the shoot apical meristem transforms into I₁. In addition to continuously producing vegetative organs, the I₁ meristem laterally initiates I₂, which produces one bract, one to three flowers and a spike with an elongated petiole. Each floral meristem sequentially develops five sepals and four common primordia, which further differentiate into five petals, 10 stamens and one carpel (Benlloch et al., 2003). By taking advantage of Tnt1 mutant population in M. truncatula, we obtained Tnt1 insertion mutants for MtTFL1, MtFULc, SGL1 and MtAP1. By characterizing mutant phenotypes and analyzing gene expression, we established the genetic relationship of the four genes in regulating compound inflorescence development in M. truncatula.

MtTFL1 (Medtr7g104460), which shares 94% identity in amino acids with PsTFL1a, was retrieved from the M. truncatula genome database as the TFL1 orthologous gene, with three other TFL1 homologous genes. Three insertion lines of MtTFL1 were identified from the Tnt1 insertion population by BLAST searching the flanking sequence tag (FST) database (Fig. S2A). Mutant mttfl1 plants showed no defects during vegetative growth and no difference with wild-type plants in flowering time (Table S1). After floral transition, mttfl1 mutants developed one or two wild-type-like inflorescences, followed by one or two defective inflorescences with shortened pedicels, and then the I₁ meristem abruptly terminated as abnormal flowers subtended with or without defective leaves (Fig. 1A-D). The wild-type-like inflorescences consisted of one to two normal fertile flowers, whereas the defective and terminated inflorescences usually produced abnormal and sterile flowers with short pedicles, loose floral structures, and reduced numbers of petals and stamens (Fig. S2C,D). After termination of the apical meristem, the upmost axillary meristem reiterated the I₁ developmental pattern. Lost dominancy of the primary shoot changed the growth pattern of inflorescence shoots from monopodial to sympodial, resulting in extensively branched mttfl1 plants.

MtFULc (Medtr7g016630) was identified in the Medicago genome database as the ortholog of VEGETATIVE 1 (VEG1)/PsFULc in pea, along with the homologous genes MtFULa and MtFULb (Berbel et al., 2012; Jaudal et al., 2015). By searching the Tnt1 FST database, we obtained one line, mtfulc-1, with a Tnt1 insertion in exon 2, and two lines, mtfulc-2 and mtfulc-3, with Tnt1 insertions in intron 4 and 5, respectively (Fig. S2B). At the vegetative stage, mtfulc-1 plants showed no morphological difference with heterozygous or wild-type plants (Fig. 1E). However, when heterozygous or wild-type plants transitioned to the floral developmental stage, mtfulc-1 mutants continued vegetative growth and produced no flowers. Upon close examination, a small bud was observed in the axillary position beside the outgrowing lateral shoot (Fig. 1F), indicating that the floral transition does occur in mtfulc-1 mutants. Instead of producing one secondary inflorescence and one lateral vegetative branch at the axillary position as in wild type, mtfulc-1 mutants developed two vegetative shoots, which became more apparent after two to three node development (Fig. 1G,H; Fig. S2E). The mtfulc-1 mutants remained vegetative, showed no sign of senescence and never produced flowers, resulting in greater biomass production (Fig. S2F). These data indicated that MtFULc is required for the I₂ identity. In contrast to mtfulc-1, mtfulc-2 and mtfulc-3 did not show any noticeable phenotypes (data not shown).

As described above, mutation of MtTFL1 leads to determination of the primary inflorescence, whereas mutation of MtFULc blocks the floral development of secondary inflorescence, indicating significant roles of MtTFL1 and MtFULc in I₁ and I₂ meristem identity, respectively. To understand the genetic regulation of MtTFL1 and MtFULc during inflorescence development, we generated the mttfl1 mtfulc double mutant by crossing mttfl1-3 with MtFULc/mtfulc-1. The double mutant and single mutant mttfl1 plants developed flowers at a similar time. Instead of developing compound inflorescences, which consist of one to two flowers, one bract and one spike in wild type, the double mutant developed single flowers with short pedicles in leaf axils (Fig. 1I-K, Fig. S2G). After flower formation at two to three nodes, the I₁ meristem abruptly terminated and developed a single flower without bract and spike subtended (Fig. 1J, Fig. S2H). Some double mutant flowers developed into pods with very short pedicles (Fig. 1L), which were in contrast to the long pedicle pods in mttfl1 and wild-type plants (Fig. 1M). Similar to the single mutant mttfl1, the shoot growth pattern in the double mutant changed from monopodial to sympodial after apical meristem termination (Fig. 1N,O). These results indicated that simultaneous loss of MtTFL1 and MtFULc transforms compound inflorescences into simple flowers, indicating that MtFULc is crucial for I₂ meristem development, and that MtTFL1 is epistatic to MtFULc in the regulation of inflorescence development.

It has been reported that mutations of MtAP1 (Medtr8g066260, MtPIM) and SGL1 (Medtr3g098560) affect floral development in M. truncatula (Hofer et al., 1997; Benlloch et al., 2006; Wang et al., 2008). One mutant line (NF11426) with abnormal inflorescences was identified from a forward screen of the Tnt1 insertion population. Genetic analysis showed that the phenotype was caused by a Tnt1 insertion in exon 2 (at 3538 bp from ATG) of MtAP1. Mutation of MtAP1 had no effects on vegetative growth and primary inflorescence development but did affect floral development (Fig. S3A). Instead of producing one or two flowers as in wild type, the mtap1 mutant developed proliferating mixtures of abnormal flowers, vegetative shoots or leaf-like structures at the end of pedicels (Fig. 2A,B). At the early flowering stage, mtap1 developed vegetative shoots instead of flowers (Fig. S3B). Gradually, more meristems gained floral identity and produced abnormal flowers with fused sepals and petals, and fewer stamens (Fig. S3C,D). At late stages, some flowers produced fertile pods (Fig. 2C, Fig. S3E). Scanning electron microscopy (SEM) showed that I₂ meristems in mtap1 did not acquire floral meristem identity for further floral organ development. Instead, the meristems reiterated bracts and I₂-like meristems and spikes (Fig. 2F,J).

Mutant sgl1 exhibited simple leaves and cauliflower-like structures consisting of reiterated sepals and carpels (Fig. 2B). Close-up observation of the mutant revealed that its I₂ meristems produced normal-appearing floral meristems that initiate sepal primordia, common primordia and carpel primordium (Fig. 2G). The sepal and carpel primordia developed into sepals and carpel-like structures. The common primordia, however, did not differentiate into petal and stamen primordia; instead, they acted as floral meristems to reiterate floral primordium initiation, resulting in cauliflower-like structures (Fig. 2K, Fig. S3F). Normal flowers were never observed on sgl1 plants.

It has been reported that LFY plays a key role in floral meristem identity but mtap1 only partially loses floral meristem identity. To further elucidate the function of SGL1 in floral development, we generated the double mutant mtap1 sgl1 by crossing heterozygous plants of MtAP1 and SGL1. The double mutant showed additive morphological phenotypes of mtap1 and sgl1, with simple leaves and branched secondary inflorescences (Fig. 2B,D,E). SEM observation revealed that I₂ meristems in the double mutant gave rise to bracts and floral meristems, and the floral meristems repeated the formation of bracts and meristems. No floral organs were observed (Fig. 2H,L; Fig. S3G). These results indicated that the double mutant completely loses floral meristem identity and the floral meristems act as I₂ meristems.

In summary, mutations of MtTFL1, MtFULc, MtAP1 and SGL1 affected the identity acquisition of I₁ meristem, I₂ meristem, floral meristem and common primordia, respectively. SGL1 played a synergistic role with MtAP1 in floral meristem identity.

To better dissect genetic regulation of inflorescence development genes, we analyzed the temporal and spatial expression patterns of MtTFL1, MtFULc, MtAP1 and SGL1 using quantitative real-time PCR (qRT-PCR) and in situ hybridization. Shoot apices of wild-type plants were sampled at 10, 17, 24 and 30 days after germination. The expression of MtAP1, SGL1, MtFULc and MtTFL1 were very low at day 10, increased significantly at day 17 and increased further at day 24. At day 30, floral buds were visible at shoot apices and the expression of these genes decreased, but were maintained at relatively high levels (Fig. 3A). No temporal expression differences were detected among the four genes in the current sampling intervals; they are all upregulated upon floral transition.

The spatial expression patterns of MtTFL1, MtFULc, MtAP1 and SGL1 in wild-type inflorescence apices were analyzed using in situ hybridization. MtTFL1 was highly expressed in the central part of I₁, but was not detected in I₂ and floral primordia (Fig. 4A-C). MtTFL1 was also expressed in the center of axillary meristems (Fig. 4D). MtFULc was highly expressed in I₂ meristems, but not detectable in I₁ (Fig. 4E). MtFULc was not expressed in newly arising floral meristems and floral organs (Fig. 4F-G), but was detectable in I₂ meristems of axillary shoots (Fig. 4H) and inflorescence stems (Fig. 4F). Similar to the results observed by Benlloch et al. (2006), MtAP1 was first detected at the side top of I₂ and then in the floral meristem and bract primordia (Fig. 4I,J). Later, MtAP1 was confined to the area where sepal primordia and petal primordia will arise (Fig. 4K). During floral organ development, MtAP1 was detectable in sepals and petals, but not in stamen primordia and carpel primordia (Fig. 4I,L). At late floral development stages, MtAP1 was expressed only in petals (Fig. 4L). SGL1 was detected in the rachis of compound leaves, and a low expression level of SGL1 was also detected in I₁, I₂ and leaf primordia (Fig. 4M). High SGL1 expression was detected in floral meristems and floral organ primordia (Fig. 4N-P). In floral organs, SGL1 was mainly expressed in petals and developing carpels (Fig. 4Q).

In summary, the expression of MtTFL1, MtFULc, MtAP1 and SGL1 was all upregulated upon floral transition. MtTFL1, MtFULc and MtAP1 were specifically expressed in I₁, I₂ and floral meristem, respectively; SGL1 showed a broad expression pattern in leaf primordia, floral meristem and all floral organ primordia. The expression domains of each gene were coherent to the inflorescence phenotypes when the corresponding gene was impaired.

To further decipher the genetic regulatory network in inflorescence development, we investigated the expression levels and patterns of the four genes in inflorescence apices of mttfl1, mtfulc, mtap1, sgl1 and mtap1sgl1 mutants by qRT-PCR and in situ hybridization.

In mttfl1, the expression level of MtFULc was slightly decreased, whereas the expression of MtAP1 and SGL1 was similar to that of wild-type-like plants (Fig. 3B). In mtfulc, the expression of MtTFL1 was increased by approximately twofold. The expression of MtAP1 was very low or undetectable and the expression level of SGL1 was also lower than that of wild-type-like plants (Fig. 3B), which was in agreement with the no-flower phenotype of mtfulc. In mtap1, no significant expression change was observed for MtTFL1, but the expression of MtFULc and SGL1 was increased (Fig. 3C). In sgl1, the expression of MtTFL1 and MtFULc showed no significant difference from wild type, whereas the expression of MtAP1 was dramatically increased compared with that of control plants (Fig. 3C). In the double mutant mtap1 sgl1, no significant expression changes were detected for MtTFL1, but the expression of MtFULc was increased, which was similar to that in the mtap1 single mutant (Fig. 3C).

The spatial expression of these genes in mutants was further examined by in situ hybridization. First, the expression domains of MtFULc, MtAP1 and SGL1 were explored in mttfl1. Before the I₁ meristem was terminated, the expression patterns of the three genes were similar to those in wild type (Fig. 5A,B,E,H). Besides the normal expression domain in I₂ meristems (the same as in wild type), a weak signal of MtFULc was detected in the I₁ meristem upon I₁ termination (Fig. 5C). Similarly, MtAP1 was detected in I₂ meristems in addition to expression domains in floral meristems, sepal and petal primordia (Fig. 5F,G). Strong SGL1 signals were detected in all meristems, including axillary meristems and floral organ primordia (Fig. 5H-J).

Next, we examined the expression of MtTFL1 in mtfulc. Besides the normal expression in I₁ meristems, MtTFL1 was also detected in defective secondary inflorescence (I′₂) meristems and axillary meristems in mtfulc plants (Fig. 5K-M). Fig. 5N showed that MtTFL1 was detected in both I₁ and I′₂ meristems in two consecutive sections.

Furthermore, the expression domains of MtTFL1, MtFULc and SGL1 were examined in mtap1. MtTFL1 showed normal expression patterns in I₁ and axillary meristems, but not in I₂ meristems. It was not detected in the defective floral (FM′) meristems (Fig. 6A,B), but was detected in some secondary lateral meristems in proliferating inflorescences (Fig. 6C,D), which was in agreement with the phenotype of vegetative shoot formation in mtap1. MtFULc was normally detected in I₂ meristems, but the expression was maintained in initiated meristems and reiterating meristems (Fig. 6E,F). MtFULc was also detectable in floral primordia and inflorescence stems but absent in developing floral organs (Fig. 6F,G); sometimes MtFULc was detectable in secondary lateral meristems of proliferating inflorescences (Fig. 6H). SGL1 was detected in most reiterating meristems and floral organ primordia in mtap1 (Fig. 6I-K).

We also investigated the expression of MtTFL1, MtFULc and MtAP1 in sgl1. MtTFL1 and MtFULc showed similar expression domains as in wild type and no ectopic expression in reiterated floral meristems (data not shown). MtAP1 was expressed in floral meristems, sepal primordia and common primordia as in wild type (Fig. 6M,N). However, unlike in wild type, in which the expression of MtAP1 was restrained in petal primordia but absent in stamen primordia during common primordium differentiation (Fig. 4I,K,L), MtAP1 was detected in common primordia reiterated meristems of sgl1 plants (Fig. 6L,N,P). MtAP1 was absent in developing sepals and carpel-like structures (Fig. 6L,O,P).

Finally, we examined the expression domains of MtTFL1 and MtFULc in mtap1 sgl1. MtTFL1 was detected in the center of the I₁ meristem and axillary meristems as in wild type (Fig. 7A,B). It was barely detectable in reiterated meristems, but was detected in some secondary lateral meristems (Fig. 7C,D). MtFULc showed normal expression in I₂ meristems as in wild type (Fig. 7E), but was maintained in the reiterated floral meristems, including bract primordia (Fig. 7F,G). In the double mutant, MtTFL1 and MtFULc showed wild-type-like expression patterns in some lateral meristems, which were consistent with the observed vegetative shoot formation from the secondary inflorescence (Fig. 7D,H).

In summary, MtTFL1 was ectopically expressed in I₂ meristems in mtfulc. MtFULc showed ectopic expression in floral meristems in mtap1 and mtap1 sgl1. MtAP1 and SGL1 were ectopically expressed in reiterated meristems in the reciprocal mutants. MtFULc, MtAP1 and SGL1 were ectopically expressed in determinate apical meristems of mttfl1, indicating a repression loop in the sequential control of inflorescence meristem identity.

The global gene expression patterns in inflorescence apices of mtap1, sgl1 and mtap1sgl1 were compared using microarray analysis. Eight-hundred and twenty-eight genes (probesets with gene ID) showed significant expression changes between the mutants and wild-type-like plants (Table S2). Among them, 246 genes were up- or downregulated (with two-fold changes) in the three mutants. However, when the expression levels of individual samples were examined, 445 genes showed a similar expression change trend at various levels (less than twofold) in the three mutants, including 435 genes that were downregulated. Seventy-four genes were regulated only in mtap1, of which 64 were downregulated. Sixty-seven genes were regulated only in sgl1, most of them were upregulated (Fig. S4, Table S2).

We specifically analyzed all MADS-box genes and some known floral development-related genes in the microarray results (Table S3). Sixteen MADS-box and three other genes were regulated, including three class A genes, three class B genes, six AGAMOUS-like genes to class C and three SEP-like genes to class E (Table 1). As demonstrated above, MtAP1 and SGL1 were upregulated reciprocally in mtap1 and sgl1. The other two class A genes were also upregulated in sgl1 but were not significantly changed in mtap1 and mtap1 sgl1. All class B and class C genes were downregulated in the three mutants, but class B genes were dramatically downregulated in sgl1. Class E genes were also downregulated in mtap1 and mtap1 sgl1, but were not significantly changed in sgl1. Noticeably, SHORT VEGETATIVE PHASE, a flowering repressor, was significantly downregulated in all three mutants. Therefore, in agreement with the phenotypes of defective floral organs in all three mutants, the majority of the impaired genes, including most floral organ identity MADS-box genes, were downregulated.

Taking all the results together, a series of schematic diagrams has been drawn to illustrate the inflorescence phenotypes of individual mutants and the associated gene expression patterns (Fig. 8A). Furthermore, the effects of gene mutation on inflorescence development are summarized in one simple scheme, and the genetic regulatory network of inflorescence architecture has been established in M. truncatula (Fig. 8B).

Flowering involves two sequential events: floral transition and floral meristem identity determination. TFL1 genes play essential roles in both events by repressing floral transition and maintaining inflorescence meristem indeterminacy (Shannon and Meeks-Wagner, 1991). The function of TFL1 in maintaining inflorescence meristem indeterminacy is well conserved in orthologous genes (Liu et al., 2010). The expression pattern of MtTFL1 and the formation of determinate I₁ meristems in mttfl1 plants indicate that MtTFL1 shares the conserved function in maintaining primary inflorescence meristem indeterminacy, but not in regulating flowering time in M. truncatula. The flowering time may be regulated by other MtTFL-like genes. The consistency of the ectopic expression of MtTFL1 and the vegetative shoot growth of mtfulc and mtap1 mutants further confirm the function of MtTFL1 in inflorescence meristem indeterminacy. The ectopic expression of MtFULc, MtAP1 and SGL1 in the apical meristem of mttfl1 and mttfl1 mtfulc suggests that MtTFL1 represses both MtFULc and the floral meristem identity genes MtAP1 and SGL1 in I₁ meristems. The conclusion is in agreement with the genetic regulation models proposing that TFL1/CEN plays antagonistic roles with floral meristem identity genes to maintain dominancy in primary inflorescence development (Fig. 8A,B) (Bradley et al., 1997; Blázquez et al., 2006; Benlloch et al., 2015).

Although it does not affect flowering time, mutation of MtTFL1 does result in abrupt termination of inflorescence meristems with shortened internodes and pedicles, and degenerative compound leaves and flowers, indicating that MtTFL1 is involved in the regulation of meristem maturation. The process shares great similarity with the precocious maturation of sympodial growth, which is caused by the mutation of the TFL1 orthologous genes SELF PRUNING (SP) in tomato or FASCICULATE (FA) in wild pepper (Pnueli et al., 1998; Elitzur et al., 2009). The evidence is in agreement with the suggested common mechanism by which TFL1 genes regulate the progression rate of shoot apices through different developmental phases, which integrates the functions of TFL1 orthologs across species with different inflorescence architectures (Ratcliffe et al., 1998).

Plant growth habit, or branching pattern, determines shoot architecture. There are two basic growth patterns: monopodial growth, as in Arabidopsis, in which the primary apical meristem remains indeterminate and develops a main axis; and sympodial growth, as in Solanaceous plants, in which the primary apical meristem terminates and further development is reiterated by upmost axillary meristems. In pea and M. truncatula, inflorescence shoots display monopodial growth. Mutation of MtTFL1, however, changes the growth pattern of inflorescence shoots from monopodial to sympodial in both mttfl1 and mttfl1 mtfulc, resulting in short and extensively branched shoot architecture. However, mutation of DET does not change the growth pattern of pea, thus no extensively branched shoots are observed (Singer et al., 1999). Tomato and pepper plants have typical sympodial growth during the reproductive stage. SP in tomato and FA in pepper are not expressed in shoot apical meristem, but are expressed in sympodial inflorescence meristems and auxiliary meristems, and decrease with the meristem outgrowth (Thouet et al., 2008; Elitzur et al., 2009). Mutations of SP or FA result in the precocious maturation of sympodial growth without affecting the inflorescence development in Solanaceous (Pnueli et al., 1998; Elitzur et al., 2009). The similarity of inflorescence shoot architectures between mutants sp or fa and mttfl1 or mttfl1 mtfulc indicates that TFL1 genes may be also involved in the genetic regulation of shoot growth patterns and that controlling the expression pattern of TFL1 could be a very useful tool for manipulating crop shoot architecture.

FUL belongs to the AP1/SQUA gene family. FUL genes show diverse expression patterns and functions in different species (Immink et al., 1999; Litt and Irish, 2003; Pabón-Mora et al., 2012; Jaudal et al., 2015). FUL plays significant roles in carpel/fruit development, and is also involved in flowering time regulation, floral meristem identity, cauline leaf development and meristem determinacy in Arabidopsis (Gu et al., 1998; Ferrándiz et al., 1999, 2000; Melzer et al., 2008). Even though FUL shows transient specific expression in inflorescence meristems, mutation of FUL has no visible impact on the inflorescence architecture in Arabidopsis. Only during the development of compound inflorescences does VEG1/PsFULc acquire significant functions in the identity of secondary inflorescence meristems in pea (Berbel et al., 2012). The specific expression pattern and no-flower phenotype in mtfulc confirm that MtFULc shares similar functions in I₂ meristem identity with VEG1. Development of normal pods in mttfl1 mtfulc indicates that MtFULc may not share the function of FUL in carpel and fruit development in Arabidopsis (Gu et al., 1998; Ferrándiz et al., 1999).

Compared with the compound inflorescence in mttfl1, mttfl1 mtfulc produces simple flowers, indicating that MtFULc is essential for compound inflorescence development. The expanded expression domain of MtFULc in proliferating inflorescences of mtap1 and mtap1 sgl1 further supports its role in I₂ meristem identity. On the other hand, it may also indicate the correlation between the expansion of MtFULc spatiotemporal expression and the increase of inflorescence complexity (Figs 2B-E, 6F and 7C). Inflorescence architecture is controlled by the rate of meristem maturation and the inflorescence complexity can be increased by delaying floral meristem commitment. This notion is supported by the effect of manipulating TFL1 and LFY expression on inflorescence structures in Arabidopsis, and is shown by the effect of COMPOUND INFLORESCENCE(S) and ANATHA on the cyme inflorescence complexity in tomato (Bradley et al., 1997; Prusinkiewicz et al., 2007; Lippman et al., 2008; Park et al., 2012). Results from the present study support the notion that MtFULc, in coordination with MtTFL1 and MtAP1, regulates inflorescence complexity by controlling the commitment of secondary inflorescence meristems. It also supports the theory that regulation of floral meristem commitment is a common mechanism for defining inflorescence complexity, but the key regulators are different in different inflorescence architectures (Bradley et al., 1997; Park et al., 2012, 2014). Whether simple flowers can be changed to compound inflorescence by extending FUL expression in the inflorescence meristems in Arabidopsis is an interesting topic that will shed light on further understanding the formation of compound racemose inflorescence.

Mutant mtfulc shows transformation of I₂ to lateral I₁-like meristems and no-flower phenotype. MtTFL1 is ectopically expressed in I₁-like meristems, which, on the one hand, indicates that MtTFL1 expression is repressed by MtFULc in I₂ meristems, and on the other hand, further confirms the function of MtTFL1 in inflorescence meristem indeterminacy. The no-flower phenotype is in agreement with the low expression level of floral meristem identity genes in mtfulc, which is in contrast to the dramatically increased expression upon floral transition in wild-type plants. Does mutation of MtFULc directly block the upregulation of floral meristem genes and result in no flower formation? Or does ectopic expression of MtTFL1 inhibit the expression of floral meristem genes to prevent flower initiation? If the first explanation were true, MtTFL1, MtFULc and MtAP1/SGL1 would act in a linear mode and the double mutant mttfl1 mtfulc would show a similar no-flower phenotype to mtfulc. In fact, the double mutant displays flower formation and a determinate primary inflorescence meristem, which is similar to mttfl1 mutant. This observation suggests that mutation of MtFULc releases only the inhibition of MtTFL1 expression in I₂ meristems, but does not directly affect the expression of MtAP1/SGL1 to prevent floral development. However, in mutant mttfl1, MtAP1 is still absent in I₂ meristems, even though the repression of MtTFL1 is abolished, indicating that MtFULc does repress the MtAP1 expression in I₂ meristems. Therefore, the reciprocal repression between MtTFL1 and MtFULc is similar to that of PsTFL1 and VEG1 in pea (Benlloch et al., 2007, 2015; Berbel et al., 2012). Both MtTFL1 and MtFULc repress the expression of MtAP1/SGL1, and MtTFL1 is epistatic to MtFULc in the regulation of floral development, partially maintaining the antagonistic mode between TFL1 and AP1/LFY in Arabidopsis

The specified expression domain of MtAP1 and the conversion of floral meristems into secondary inflorescences, leaf-like bracts and sepals, and defective petals in the mtap1 mutant, confirm its conserved function with orthologous genes in floral meristem and floral organ identity (Mandel et al., 1992; Taylor et al., 2002; Benlloch et al., 2006). The complete conversion of floral meristems into I₁ in the basal part and the absence of wild-type-like secondary inflorescence even at late stages in the mtap1 mutant indicate that MtAP1 is also involved in secondary inflorescence development. The complementary expression domains between MtAP1 and MtFULc in wild-type plants and the ectopic expression of MtFULc in mtap1 suggest that MtAP1 represses MtFULc to acquire floral meristem identity, which agrees with studies in pea and Arabidopsis (Ferrándiz et al., 2000; Berbel et al., 2012). In mtap1 and mtap1 sgl1, some I₁-like inflorescence shoots are produced at early reproductive stages and the expression of MtTFL1 is detectable in some secondary lateral meristems, indicating that some of the meristems regain I₁-like meristem identity. Secondary I₁-like meristem formation is not fully understood. However, we speculate that MtFULc represses MtTFL1 in I₂ meristems, whereas MtAP1 represses MtTFL1 expression in floral meristems, which is possibly reminiscent of the reciprocal repression between TFL1 and AP1 in Arabidopsis (Liljegren et al., 1999).

LFY/FLO/UNI is another major floral meristem player in plants. Mutation of LFY/FLO causes complete or partial conversion of floral meristems into inflorescence meristems (Coen et al., 1990; Schultz and Haughn, 1991). UNI is involved in the regulation of I₁ meristem identity in pea as uni mutant plants terminate with the formation of a stub at the apical meristem (Singer et al., 1999; Berbel et al., 2012). Unlike lfy/flo/uni, sgl1 never shows the replacement of flowers with inflorescences or termination of I₁ meristem; instead, it exhibits reversion of common primordia into floral meristems without further differentiation of petal and stamen primordia. MtAP1 maintains its expression in reversed and reiterated meristems, and the expression of class B function genes is dramatically reduced in sgl1, indicating that SGL1 is required for petal/stamen specification from common primordia by repressing MtAP1 expression and upregulating class B function genes (Wagner et al., 1999). However, the double mutant mtap1 sgl1 enhances mtap1 phenotypes and leads to the complete conversion of floral meristems into I₂-like meristems. This observation suggests that SGL1 has synergistic function with MtAP1 in floral meristem identity. The large number of commonly regulated genes among mtap1, sgl1 and mtap1 sgl1 further indicate the conserved functions of MtAP1 and SGL1 in floral meristem identity and floral organ development in M. truncatula.

Besides the conserved functions in flower development, SGL1/UNI plays significant roles in compound leaf development. Mutation of SGL1/UNI leads to simple leaf formation in some legume species (Hofer et al., 1997; Wang et al., 2008). Additionally, mutation of RFL, the ortholog of LFY in rice, leads to simplification of panicles in rice (Rao et al., 2008). These results indicate that LFY genes are required for maintaining meristem transient indeterminacy to form determinate structures, i.e. leaflets and panicles (Hofer et al., 1997; Rao et al., 2008; Wang et al., 2008). The formation of common primordia is a unique transition phase during flower development in legumes. The process resembles the formation of compound leaves from leaf primordium to leaflets. The mutation of SGL1 results in the reversion of common primordia to floral meristems. Therefore, one of the SGL1/UNI functions in flower development is to promote or maintain a transient phase of indeterminate growth, as suggested in compound leaf development (Hofer et al., 1997).

In conclusion, we deciphered the genetic regulation of inflorescence development in M. truncatula via comprehensive and systematic expression and genetic analyses of four key genes. Our data show that inflorescence architecture is controlled by spatiotemporal expression of MtTFL1, MtFULc, MtAP1 and SGL1 through reciprocal repression, which is not well defined in pea. The genetic regulatory network involving the four genes shares similarity with the modified model in pea, but also shows specificity in compound inflorescence development in M. truncatula. The study establishes M. truncatula as an excellent genetic model for understanding compound inflorescence development in related important crop species.

Seeds of Tnt1 insertion lines for MtTFL1 (NF7897, NF17535 and NF18476), MtFULc (NF10716, NF15285 and NF18126), SGL1 (NF0740) and MtAP1 (NF11426) were requested from Noble Research Institute. The insertion homozygosity of individual genes was identified by PCR with combinations of gene-specific primers and Tnt1 primers. PCR-based reverse screening for additional insertion lines for MtFULc was carried out as previously described (Cheng et al., 2014), though no additional line was obtained. For obtaining double mutant mttfl1 mtfulc, mttfl1-3 (NF18476) flowers were cross-pollinated with heterozygous MtFULc flowers (NF10716); F1 plants with Tnt1 insertions in both MtTFL1 and MtFULc were identified by PCR genotyping, and F2 seeds were harvested from confirmed F1 plants. The mttfl1 mtfulc double mutant plants were identified from the F2 segregating progenies by PCR genotyping. To obtain double mutant mtap1 sgl1, heterozygous MtAP1 flowers were cross-pollinated with heterozygous SGL1 flowers. The F2 seeds were collected from F1 plants and germinated for F2 segregating progenies. The double mutants from the F2 segregation population were confirmed by PCR genotyping. The primer sequences of each gene for genotyping are listed in Table S4.

Seeds of M. truncatula wild type (R108) and Tnt1 insertion mutants were scarified with concentrated sulfuric acid for 8 min, rinsed with water and put on filter paper for 7-10 days at 4°C. Germinated seeds were transferred into one-gallon pots with Metro-Mix 350 (Scotts) composite soil and grown at 16 h/8 h day/night cycle at 150 μE/m²/s light intensity, 22°C/18°C day/night temperature and 70% humidity for sampling, phenotype observation and pod collection.

Inflorescence shoot apices from six-week-old wild-type and mutant plants were fixed and the in situ hybridization was performed as described by Zhou et al. (2010). cDNA fragments (400-600 bp) from non-conserved regions of MtTFL1, MtFULc, MtAP1 and SGL1 were used as the probes for in situ hybridization. The primer pairs of each probe are listed in Table S4.

Inflorescence shoot apices from six-week-old wild-type and mutant plants were dissected and fixed in 4% paraformaldehyde in 0.1 M PBS (pH 7.2) for 12 h at 4°C. After rinsing with PBS for 3 h, fixed samples were dehydrated in an ethanol series and critical point dried. The samples were mounted on metal stubs and sputter-coated with gold, observed under a Zeiss DSM-960A SEM (Carl Zeiss) at an accelerating voltage of 5 kV.

Shoot apices of wild type were collected at day 10, 17, 24 and 30 days after germination for expression analysis during floral transition. Inflorescence apices from six-week-old wild-type and mutant plants were sampled. Three biological replicates were used for each sample. Total RNA was extracted using Tri-Reagent (Gibco-BRL Life Technologies) and treated with Turbo DNase I (Ambion). For RT-PCR and qRT-PCR, 3 µg of total RNA were used for reverse transcription using SuperScript III Reverse Transcriptase (Invitrogen) with the oligo (dT)²⁰ primer. cDNA (2 µl diluted 1:20) was used as a template. Three technical replicates were carried out for each template. Gene-specific primers, which span non-conserved regions and are used for RT-PCR and qRT-PCR, are listed in Table S4. All qRT-PCR was carried out using a 7900HT Fast Real-Time PCR System (Applied Biosystems), and the data were analyzed using SDS 2.2.1 software (Applied Biosystems). The transcript levels were determined by relative quantification using the M. truncatula ubiquitin-conjugating enzyme gene (Medtr3g110110) as the internal reference.

For transcriptome analysis, three biological replicates of the inflorescence shoots were sampled from six-week-old wild-type-like plants (including wild-type and heterozygous plants in the same segregating progeny), mtap1, sgl1 and mtap1 sgl1 homozygous plants in F2 segregation progenies. Total RNA was isolated using RNeasy Plant Mini Kit (Qiagen). Purified RNA was treated with Turbo DNase I (Ambion). Total RNA (500 ng) from each sample was used for labeling with the GeneChip WT PLUS Reagent Kit (WT PLUS Kit) (Affymetrix) based on the manufacturer's recommendation. Hybridization and scanning for microarray analysis were conducted according to the manufacturer's instructions (GeneChip Medicago Transcriptome, Affymetrix). Data normalization between chips was conducted using RMA (Robust Multichip Average) (Irizarry et al., 2003). Presence/absence calls for each probeset were obtained using dCHIP (Li and Wong, 2003). Gene selections based on the associative t-test were made using Metlab (MathWorks) (Dozmorov and Centola, 2003). A selection threshold of two for transcript ratios and a Bonferroni-corrected P value threshold of 2.19202E-06 were used.

The authors hank Dr Elison Blancaflor for his critical reading of the manuscript and Kuihua Zhang for plant care.