Revisiting the origin and identity specification of the spikelet: A structural innovation in grasses (Poaceae)

Abstract

Spikelets are highly specialized and short-lived branches and function as a constitutional unit of the complex grass inflorescences. A series of genetic, genomic, and developmental studies across different clades of the family have called for and permitted a synthesis on the regulation and evolution of spikelets, and hence inflorescence diversity. Here, we have revisited the identity specification of a spikelet, focusing on the diagnostic features of a spikelet from morphological, developmental, and molecular perspectives. Particularly, recent studies on a collection of barley (Hordeum vulgare L.), wheat (Triticum spp.), and rice (Oryza sativa L.) mutants have highlighted a set of transcription factors that are important in the control of spikelet identity and the patterning of floral parts of a spikelet. In addition, we have endeavored to clarify some puzzling issues on the (in)determinacy and modifications of spikelets over the course of evolution. Meanwhile, genomes of two sister taxa of the remaining grass species have again demonstrated the importance of genome duplication and subsequent gene losses on the evolution of spikelets. Accordingly, we argue that changes in the orthologs of spikelet-related genes could be critical for the development and evolution of the spikelet, an evolutionary innovation in the grass family. Likewise, the conceptual discussions on the regulation of a fundamental unit of compound inflorescences could be translated into other organismal groups where compound structures are similarly formed, permitting a comparative perspective on the control of biological complexity.

The identity specification and evolution of the spikelet is reexamined from morphological, developmental, and molecular perspectives.

The spikelet functions as a fundamental unit to build up the compound inflorescences in the grass family

The fundamental building blocks are key to understanding the design and organization of complex biological systems at different levels. The identification and characterization of the properties pertaining to the fundamental elements are therefore essential for decoding the underlying evolutionary and developmental mechanisms of complex traits. The compound inflorescences of the grass family (Poaceae), including arguably the most important plant species to human civilization because of their ecological and economical importance, exhibit staggering diversity and structural complexity, and hence are a great system to investigate the evolution of biological design and complexity (Kellogg, 2015; Whipple, 2017; Bommert and Whipple, 2018; Koppolu et al., 2022; Zhong and Kong, 2022). The spikelet, a little spike, represents a key evolutionary novelty and is the basic building block whose self-repetitions ultimately form the higher-order organization of the complex and compound inflorescence architectures (Kellogg, 2015; Whipple, 2017; Bommert and Whipple, 2018; Zhong and Kong, 2022). Unlike an individual flower—a basic unit in many flowering plants, the spikelet itself is a specialized inflorescence and comprises one to several small flowers (i.e. florets). However, a spikelet also displays a series of flower-specific features. Particularly, similar to the regulation of an individual flower, a spikelet is indeed found to be specified by discrete genetic factors that control its identity and variation (Jeon et al., 2000; Prasad et al., 2001, 2005; Chuck et al., 2002; Komatsu et al., 2003; Whipple, 2017; Bommert and Whipple, 2018; Poursarebani et al., 2020; Du et al., 2021a; Li et al., 2021a, 2021c; Zhong et al., 2021). Even more intriguingly, as those of floral meristem identity genes in Arabidopsis (Arabidopsis thaliana [L.] Heynh.), changes of spikelet meristem identity genes in grasses also appear to impact inflorescence architectures as suggested by several latest studies (Prusinkiewicz et al., 2007; Poursarebani et al., 2015; Du et al., 2021a; Li et al., 2021a; Ma et al., 2021; Seetharam et al., 2021).

In the grass family, two species-poor subfamilies Anomochlooideae (incl. Anomochloa and Streptochaeta) and Pharoideae (incl. Pharus) are successively sister to the remaining members of the family (Figure 1A; Grass Phylogeny Working Group II, 2012; Soreng et al., 2015, 2022). However, the typical spikelet is found in all grasses but not in Anomochlooideae, the sister clade of all other grass species, during the diversification of the family (Grass Phylogeny Working Group, 2001; Sajo et al., 2008; Kellogg, 2015; Grass Phylogeny Working Group II, 2012). Recent comparative developmental and genomic studies have provided unprecedented insights into the origin and identity specification of a spikelet (Xu et al., 2020; Zhuang et al., 2020; Du et al., 2021a; Li et al., 2021a; Ma et al., 2021; Seetharam et al., 2021; Zhong et al., 2021).

Figure 1.

Spikelets are a structural novelty and a developmentally integrated structure in the grass family. A, A whole-genome duplication event ρ (marked by a black bar) is predicted to have occurred prior to the diversification of the grass family (Poaceae; Seetharam et al., 2021; Ma et al., 2021), and a true spikelet likely evolved subsequently after the divergence of Anomochlooideae (incl. Streptochaeta) from the rest of the grass family (labeled by a green bar). Pseudo-spikelets (red bars) are described in Streptochaeta (D) and bamboos (E), compared with those spikelets in wheat (B) and rice (C) (Soderstrom and Londoño, 1987; Sajo et al., 2008; Preston et al., 2009; Kellogg, 2015). B, A true spikelet in wheat can be diagnosed by having two basal sterile glumes (gl, color green) and paired lemma and palea (color teal) enclosing each floret (fl). Only two florets are depicted in the floral diagram (right). A rachis is the axis of inflorescences bearing spikelets, while a rachilla is the axis in the spikelet that carries florets. C, A modified three-flowered spikelet in rice has a fertile floret and two empty sterile lemmas (sl), subtending by two basal rudimentary glumes (rg; Zhang et al., 2017). D, Pseudo-spikelets in Streptochaeta produce multiple sterile bracts (1–5 and 6–8) (Sajo et al., 2008; Preston et al., 2009); The bracts (1–5, 6–8, and 10–12) are ordered according to a previous study and show different expression patterns of the LHS1/MADS1 ortholog (Preston et al., 2009). Particularly, MADS1 transcripts are found in the inner six bracts (colored in teal, 6–8 and 10–12) but not in the five basal sterile bracts (marked in green, 1–5) (Preston et al., 2009). E, Pseudo-spikelets of a Brazilian bamboo Guadua produce branches in the axils of glumes (Soderstrom and Londoño, 1987). It should be noted the bracts in bamboo inflorescences are not suppressed as found in the majority of grasses; and an adaxial tip-notched prophyll, the first leaf of a branch, is developed, as commonly seen in monocots (Soderstrom and Londoño, 1987). Stamens (st) and gynoecia are labeled in yellow and gray, respectively. Image credits: Ms. Wenyan Ding (rice), Dr. Dexin Kong (maize), Dr. Paula Rudall (Streptochaeta), Yanli Wang (wheat), and Jinshun Zhong (bamboo). Plant images were cropped from the original photographs and this figure was assembled using Inkscape0.92.4 (https://inkscape.org/). Cartoons are made to depict the overall structures but not to scales from different views. Scale bars: 1 mm.

In this review, we summarize and synthesize recent findings on the origin and specification of a spikelet during the diversification of the grass family. Particularly, we revisit and discuss three fundamental questions: (1) what a spikelet is, (2) how a spikelet is defined and specified from the morphological, developmental, and molecular perspectives, and (3) when a spikelet originated from an evolutionary point of view. This review provides a timely discussion on the origin and identity regulation of a spikelet, and how variation in the specification of a spikelet might have contributed to inflorescence diversity and complexity in the grass family. In addition, the synthesis of an empirical example discussed here in grasses on the organization of compound inflorescences could be extrapolated to other compound structures in flowering plants, permitting an evolutionary view on the development of biological complexity.

The spikelet may be diagnosed by the production of distinctive leaf-like glumes

The canonical definition of a spikelet is often based on its functional role or operational features. As defined in the Merriam-Webster English dictionary, a spikelet is “a small or secondary spike” and “one of the small few-flowered bracted spikes that make up the compound inflorescence of a grass.” A typical spikelet bears a series of distichous bracts on a short axis called rachilla (i.e. a spikelet axis), in that the two basal (proximal) bracts are named glumes and the remaining (distal) bracts are considered as lemmas (Figure 1, B and C; Whipple et al., 2007; Sajo et al., 2008; Kellogg, 2015). Glumes are often empty (i.e. sterile) and have no associated axillary meristems (florets or buds) being initiated or developed in their axils. In contrast, each lemma, often paired with an adaxial palea, subtends/encloses a floret within a spikelet (Figure 1, B and C). Therefore, a true spikelet may often be diagnosed by the presence of a pair of basal and empty glumes (Figure 1, A–C; Kellogg, 2015; Zhong et al., 2021).

In some cases, modifications of spikelets, in glumes and lemmas, in particular, are found to confound the diagnosis of spikelet identity in several lineages and crop cultivars. For example, in some bamboos (Bambusoideae), sister lineages of the bulk of the grasses (e.g. Streptochaeta), and wheat (Triticum spp.) cultivars, axillary buds in the axils of the glumes may be initiated and elaborated (one to several) extra sterile bracts may be formed, and/or additional spikelets may be developed within the “normal” spikelets (Soderstrom and Londoño, 1987; Sajo et al., 2008; Kellogg, 2015; Amagai et al., 2017; Dobrovolskaya et al., 2017; Ma et al., 2021; Seetharam et al., 2021). These atypical “spikelets” are often referred to as pseudo-spikelets in Streptochaeta and bamboos (Figure 1, A, D, and E; Soderstrom and Londoño, 1987; Sajo et al., 2008; Kellogg, 2015). In fact, some lemmas may become sterile like glumes that produce no associated axillary meristems in their axils as found in rice (Oryza sativa L.) and barley (Hordeum vulgare L.) intermedium-m (int-m) mutants (Figures 1C and 2A; Kellogg, 2009; Zhang et al., 2017; Zhong et al., 2021). In rice, spikelets are modified and have two pairs of sterile bracts, in that the two basal bracts are considered as rudimentary glumes, and the two leaf-like organs above the glumes are best interpreted as sterile lemmas according to comparative genetic and gene expression analyses (Figure 1, A and C; Prasad et al., 2001, 2005; Kellogg, 2009; Yoshida et al., 2009; Zhang et al., 2017; Xu et al., 2020; Zhuang et al., 2020). It hence appears that comparative analyses of expression patterns of spikelet meristem identity genes would help determine organ identity and homology between pseudo-spikelets and spikelets. However, genetic and expression analyses of spikelet identity genes are limited in species/cultivars carrying pseudo-spikelets, particularly in the sister species of the bulk of grasses and bamboos (Whipple et al., 2007; Preston et al., 2009). This then begs the question how a spikelet is specified at the molecular level, which we will discuss in detail in the next section. We will also come back to the issue of spikelet identity and patterning in the final section with regard to the origin and evolution of spikelets in light of comparative genomic and gene expression analyses.

Figure 2.

Spikelets are regulated by a sets of transcription factors. A, Representative images of developing inflorescences of int-m mutants showing the formation of a terminal spikelet, bearing multiple basal sterile bracts (incl. glumes and sterile lemma [sl]) (left) and only two basal sterile bracts (i.e. glumes) (right). Glumes are highlighted in green false color. B, Transcripts of COM2 are restricted in the axils of the two basal glumes in the developing spikelets in barley wild-type Bowman. In barley int-m mutants, a terminal spikelet is formed upon the transition to reproductive stage (Zhong et al., 2021). COM2 is expressed in the axils of the two basal glumes, as well as sterile lemmas of the terminal spikelets in barley int-m mutants, but not in the developing floret meristem. C, In contrast, MADS1 transcripts are detected in lemmas and paleae (Zhong et al., 2021). D–F, In addition, LOG1, KN1, and APO1 orthologs are transcribed in the emerging floret meristems (Zhong et al., 2021). G, A schematic diagram of a hypothetical spikelet showing expression patterns of genes involved in the specification of spikelet meristem, floret meristem, and floral organs. The results were compiled largely from the work on rice, maize, wheat, and barley (Chuck et al., 1998, 2002; Komatsu et al., 2003; Prasad et al., 2005; Yoshida et al., 2009; Zhang et al., 2017; Li et al., 2021a; Zhong et al., 2021; Kuzay et al., 2022). Particularly, NSG1/ LRG1 is restricted in rudimentary glumes (rg) and sterile lemmas (sl) in rice (Xu et al., 2020; Zhuang et al., 2020). Orthologs of BD1/FZP/COM2 exhibit conserved expression patterns in the axils of glumes (gl) and/or sterile lemmas (sl), and the transcripts of G1 are specifically confined to sterile lemmas (sl; Chuck et al., 2002; Komatsu et al., 2003; Yoshida et al., 2009). MADS1 and IDS1/INT-M/AP2-5A orthologs are expressed in lemmas and paleae but not in glumes (Prasad et al., 2005; Zhong et al., 2021). A YABBY gene DROOPING LEAF (e.g. rice DL, and maize DRL1, DRL2) and an AGL6-like MADS-box gene rice MFO1/maize BDE are found in the developing lemmas/pistils and paleae/lodicules in various grasses, respectively (Yamaguchi et al., 2004; Ohmori et al., 2009, 2011; Reinheimer and Kellogg, 2009; Thompson et al., 2009). In addition, LF1, KN1, LOG1, and APO1 are transcribed in floret meristems, likely promoting the formation of fertile floret meristems. Collectively, the expression patterns of these genes could be used to diagnose the identity of spikelets and floral organs, and thereby are informative for comparative analyses to understand the evolution of spikelets and inflorescence architectures in grasses. Note: A variety of dots of different color and size depict the domains but not the exact ranges of expression patterns of genes that are important for spikelet identity and lateral organ patterning. Images of in situ hybridization were cropped from the original photographs with the adjustment of brightness and contrast using GIMP version 2.10.30 (https://www.gimp.org/) and this figure was assembled using Inkscape version 0.92.4 (https://inkscape.org/).

Spikelet meristem identity is controlled by discrete transcription factors

Several recent studies have highlighted that a spikelet indeed exhibits some flower-specific characteristics from the evolutionary and developmental perspectives (Li et al., 2019, 2021a; Du et al., 2021a; Yang et al., 2021; Zhong et al., 2021). For instance, a spikelet is the fundamental unit of grass compound inflorescences—similar to an individual flower as the basic self-repeating unit in many flowering plants. Intriguingly, homologous genes, such as SQUAMOSA/APETALA1(AP1)-clade genes, including Arabidopsis AP1 and FRUITFUL (FUL), and wheat VERNALIZATION1 (VRN1) are recruited in eudicot and grass species to control meristem identities of a flower and a spikelet (or an inflorescence-like structure), despite some lineage-specific gene duplications and modifications (Table 1; Kaufmann et al., 2010; Kobayashi et al., 2012; Bommert and Whipple, 2018; Li et al., 2019; Yang et al., 2021; Zhong and Kong, 2022). In addition, orthologs of PANICLE PHYTOMER2/MADS34 and LEAFY HULL STERILE1 (LHS1)/MADS1/Zea mays MADS8 (ZMM8) in rice, barley, and maize (Z.mays L.) contribute to spikelet meristem initiation, whose loss-of-function mutations lead to an increased number of inflorescence branches, hinting at partial losses of spikelet meristem identity (Gao et al., 2010; Kobayashi et al., 2010, 2012; Du et al., 2021b; Li et al., 2021a). MADS34 and MADS1 encode SEPALLATA (SEP)-like MADS-box transcription factors whose Arabidopsis homologs are key components of the floral quartet model (Table 1; Cacharrón et al., 1999; Honma and Goto, 2001; Prasad et al., 2001; Theißen and Saedler, 2001; Malcomber and Kellogg, 2004; Gao et al., 2010; Kobayashi et al., 2010; Zhang et al., 2012; Du et al., 2021b; Li et al., 2021a).

Table 1.

Genes that are important for the specification of spikelet identity and lateral organs of a spikelet

Arabidopsis	Rice	Maize	Wheat/Barley	Function
AP2-PUCHI	FZP	BD1	BH (WFZP)/COM2	Spikelet meristem identity and floral organ identity
AP1	OsMADS14	Z. mays MADS4,15 (ZMM4, ZMM15)	VRN1
FUL	OsMADS15	ZAP1, ZMM3	FUL2
	OsMADS18	–	FUL3
SEP4	LEAFY HULL STERILE1 /OsMADS1	ZMM8 , ZMM14	SEP1-2/HvMADS1
	OsMADS34	ZMM31, ZMM24	SEP1-6/HvMADS34
CKX7 *	OsCKX3	ZmCKX3	TaCKX3/HvCKX3
CKX2,3,4*	OsCKX2/Gn1a	ZmCKX2	TaCKX2a,b/HvCKX2a,b**
AP2	OsIDS1	IDS1 ,	Q /INT-M	Meristem determinacy of inflorescence and spikelet meristems
	SUPERNUMERARY BRACT	SISTER OF IDS1	–
REVOLUTA	LATERAL FLORET1 /OsHB1	Rolled leaf1 (RLD1), RLD2	TaHB1/HvHB1	Rice three-flowered spikelets; maize leaf adaxial/abaxial patterning (Juarez et al., 2004)
ZINC FINGER PROTEIN 1	NSG1/LRG1	ZmLRG1a, ZmLRG1b	TaLRG1/HvLRG1	Rice rudimentary glume and sterile lemmas
CRABS CLAW	DL	DRL1, DRL2	TaDL/HvDL	Lemmas and pistils
AGL6	MFO1/OsMADS6	BDE/ZAG3, ZAG5	TaAGL6/HvAGL6	Paleae and lodicules
UFO	APO1	ZmAPO1a,b	WAPO1 /HvAPO1	Inflorescence meristem activity
LFY	APO2 /RICE LFY	ZmLFY1, ZmLFY2	TaLFY/HvLFY
LIGHT-DEPENDENT SHORT HYPOCOTYLS1	TAWAWA1	ZmTAW1a,1b	TaTAW1/HvTAW1
	G1	ZmG1	TaG1/HvG1	Sterile lemma identity

Notes: Genes in bold have been functionally characterized in respective species, and relevant findings are discussed in the main text.

Asterisks (*) mark the closest paralog(s) of OsCKX2 or HvCKX3 in Arabidopsis.

^**Two different genes of CKX2 (e.g. HORVU.MOREX.r2.3HG0202950.1 and HORVU.MOREX.r2.3HG0202940.1 in barley, and AET3Gv20309400.5, AET3Gv20308400.5 in Aegilops tauschii) are found to be sister to OsCKX2/Gn1a in Triticeae genomes; however, it remains unclear which HvCKX2 is differentially expressed in vrs4 mutants (Koppolu et al., 2013).

Moreover, some distinct transcription factors are required to regulate spikelet meristem identity in grasses, such as an ERF-type AP2 transcription factor, including orthologs of maize BRANCHED SILKLESS1 (BD1), rice FRIZZY PANICLE (FZP), Brachypodium (Brachypodium distachyon (L.) P.Beauv.) MORE SPIKELETS1, barley COMPOSITUM2 (COM2) and wheat BRANCHED HEADS (BHs or Wheat FZP [WFZP]) (Table 1; Prasad et al., 2001, 2005; Chuck et al., 2002; Komatsu et al., 2003; Derbyshire and Byrne, 2013; Poursarebani et al., 2015; Bommert and Whipple, 2018; Zhong et al., 2021). The loss-of-function mutations in BD1/FZP/COM2 orthologs convert some spikelets into inflorescence-like branches in maize, rice, Brachypodium, barley, and wheat, leading to production of additional branches and more (pseudo)-spikelets in their inflorescences (Chuck et al., 2002; Komatsu et al., 2003; Derbyshire and Byrne, 2013; Poursarebani et al., 2015; Li et al., 2021c). Furthermore, wheat bh mutations result in the conversion of the terminal spikelet meristem into an inflorescence meristem, supporting a role of BH/WFZP on the spikelet identity specification (Poursarebani et al., 2015).

However, temporal and spatial expression patterns of meristem identity genes are not completely overlapped and they are not entirely linked with the timing or positions of emerging spikelets. Particularly, the AP1-clade and MADS34 genes in rice and barley are in fact actively transcribed prior to the initiation of spikelet meristems in various types of meristems, including vegetative shoot apical meristems, inflorescence meristems, and/or branch (triple spikelet) meristems, hinting at broad roles of AP1-clade and MADS34 genes on the patterning of different meristems (Kobayashi et al., 2012; Zhong et al., 2021). In contrast, transcripts of MADS1 and BD1/FZP/COM2 are localized in distinct (non-overlapping) domains of an emerging spikelet (Figure 2B;Prasad et al., 2005; Du et al., 2021a; Li et al., 2021a; Zhong et al., 2021). Particularly, orthologs of BD1/FZP/COM2 are transcribed in the axils of glumes, whereas the transcripts of MADS1 orthologs are specifically restricted in spikelet meristems, lemmas, and paleae (Figure 2, B, C, and G; Chuck et al., 2002; Komatsu et al., 2003; Poursarebani et al., 2020; Du et al., 2021a; Li et al., 2021a, 2021c; Zhong et al., 2021). BD1/FZP/COM2 proteins control the specification of spikelet meristems and are hypothesized to inhibit the initiation of axillary meristems in the axils of glumes (Komatsu et al., 2003; Dobrovolskaya et al., 2015; Poursarebani et al., 2015; Dobrovolskaya et al., 2017; Bai et al., 2017; Du et al., 2021a; Li et al., 2021c). The transcriptional or translational repression of BD1/FZP/COM2 prolongs the production of primary branches/spikelets, ultimately leading to an increased number of branches/spikelets, or the production of spikelets within spikelets (Komatsu et al., 2003; Dobrovolskaya et al., 2015, 2017; Poursarebani et al., 2015; Bai et al., 2017; Huang et al., 2018; Du et al., 2021a; Li et al., 2021c; Chen et al., 2022). Mechanistically, wheat BH/WFZP proteins appear to directly activate the expression of VRN1-A and HOMEOBOX4-A, thereby promoting the acquisition of spikelet meristem identity (Li et al., 2021c). Furthermore, MADS1 directly promotes the transcription of a CYTOKININ OXIDASE3 (CKX3) gene, encoding an enzyme that degrades bioactive cytokinins, whose transcripts are co-localized with those of MADS1 in spikelet meristems (Li et al., 2021a). The loss-of-function mutations of MADS1, and hence the de-activation of CKX3, lead to local accumulation of bioactive cytokinins, thereby promoting indeterminate growth of axillary meristems, which often develop into spikelets in the wild-type, ultimately resulting in the formation of lateral branches (Li et al., 2021a). Indeed, like rice fzp mutants, null alleles of rice OsCKX2 (a QTL GRAIN NUMBER 1A [GN1A]) result in more primary branches and hence increased number of spikelets, hinting at the crucial role of cytokinin oxidases in spikelet identity specification (Table 1; Ashikari et al., 2005; Bai et al., 2017). Likewise, in barley, decreased transcription of a HvCKX2 in six-rowed spike4 (vrs4) mutants and elevated levels of HvCKX3 transcripts in int-m mutants are associated with respective indeterminate and determinate growth in vrs4 and int-m mutants, highlighting their roles in specifying spikelet identity (Table 1; Koppolu et al., 2013; Zhong et al., 2021).

Barley int-m mutants initiate a terminal spikelet meristem early in the reproductive development, immediately after the transition of a vegetative shoot apical meristem to an inflorescence meristem, thereby providing a gain-of-function perspective on the identity specification of a spikelet meristem (Zhong et al., 2021). INT-M, encoding an AP2 transcription factor, is orthologous to maize INDETERMINATE SPIKELET1 (IDS1), SISTER OF IDS1 (SID1), and wheat Q gene (AP2-5A) (Chuck et al., 1998; Debernardi et al., 2017; Zhong et al., 2021). In barley, INT-M promotes the maintenance and indeterminacy of an inflorescence meristem, likely via promoting the expression of meristem identity genes and suppressing spikelet differentiation cues (Zhong et al., 2021). Notably, barley MADS1 and COM2 are precociously transcribed around the tip of a developing inflorescence upon the transition into double ridge (reproductive) stage in barley int-m mutants, coinciding with the initiation of the terminal spikelet meristem (Figure 2, B and C and Table 1; Zhong et al., 2021). In other words, this observation strongly supports that the expression patterns of MADS1 and BD1/FZP/COM2 are temporally and spatially linked with the establishment of a spikelet meristem (Figure 2, B and C; Zhong et al., 2021).

The findings from barley int-m, and other grass mutants also suggest that proteins that maintain activities of inflorescence meristems delay the initiation of spikelet meristems. Particularly, barley INT-M prolongs inflorescence meristem activity, inhibiting the conversion of an inflorescence meristem into a terminal spikelet meristem (Zhong et al., 2021). INT-M likely suppresses the expression of MADS1 and COM2, such that in int-m mutants MADS1 and COM2 are both precociously and ectopically expressed around the tip of an inflorescence, leading to the formation for a terminal spikelet (Figure 2, B and C; Zhong et al., 2021). Indeed, orthologs of INT-M in grasses, including maize IDS1 and SID1, rice IDS1 and SUPERNUMERARY BRACT (SNB), and wheat Q (mircoRNA172-resistant AP2-5A) are functionally conserved in prolonging inflorescence meristem activity (Chuck et al., 1998; Lee et al., 2006; Lee and An, 2012; Debernardi et al., 2017; Zhong et al., 2021). Consistently, in Arabidopsis, AP2 transcription factor, independent of its role in flowering time regulation and floral organ identity, promotes inflorescence meristem activity (Würschum et al., 2006; Balanzà et al., 2018; Sang et al., 2022). In addition, in rice, TAWAMA1 (TAW1), encoding an ALOG (Arabidopsis LHS1 and Oryza G1) transcription factor, promotes inflorescence meristem activity via inducing the expression of two SVP-like MADS-box genes MADS22 and MADS55 and hence delays the establishment of spikelet meristem identity (Yoshida et al., 2013). Likewise, in rice, ABERRANT PANICLE ORGANIZATION loci (APO1 and APO2), forming a core UNUSUAL FLORAL ORGANS (UFO)-LEAFY (LFY) module, promote inflorescence meristem identity, and mutations in either locus result in shorter inflorescences bearing fewer primary branches and spikelets (Table 1; Ikeda et al., 2005, 2007; Rao et al., 2008; Ikeda-Kawakatsu et al., 2009; Ikeda-Kawakatsu et al., 2012). Furthermore, APO1 may be conserved in prolonging inflorescence meristem activity among grasses and it should be noted that APO1 is also transcribed in floret meristems, hinting at its moonlighting function in subsequent floral patterning/differentiation (Zhong et al., 2021; Kuzay et al., 2022).

Therefore, these case studies have collectively highlighted a set of transcription factors that temporally and spatially coordinate the specification of spikelet meristem identity, including those that promote spikelet meristem initiation, such as AP1-clade MADS-box proteins, MADS34, MADS1, and an ERF-type AP2 transcription factor BD1/FZP/COM2, and those that inhibit spikelet meristem establishment, such as INT-M/IDS1/Q, TAW1, APO1, and APO2 (Table 1).

Spikelets are short-lived and generally determinate, but produce a varied number of florets among different species

The spikelet, as a compound structure, exhibits some flower-specific and inflorescence-like features, leading to confounding issues regarding its “determinacy or indeterminacy.” The definition of “determinacy” differs between traditional morphologists and developmental biologists, highlighting its qualitative and quantitative features (Kellogg, 2015; Bommert and Whipple, 2018). Particularly, a determinate axis is defined qualitatively and ended by a flower or flower-like structure as defined by classical morphologists. Whereas developmental biologists emphasize the quantitative aspect of a determinate structure wherein a fixed number of lateral organs are produced. That is, like a flower, the spikelet, a short branch as a basic unit of grass inflorescences, is finite and hence determinate. While by morphological definition, spikelets of the majority of grass species are indeterminate as florets within a spikelet are mostly lateral (Chuck et al., 1998; Kellogg, 2015; Zhong et al., 2021). Therefore, these two distinct definitions regarding the determinacy of a spikelet have caused inevitable confusions in different communities. In fact, neither morphological nor developmental definition of determinacy is completely satisfactory. For instance, classical morphologists and developmental biologists have agreed that spikelets in barley and maize are determinate as they are described as producing a terminal floret in their spikelets and a fixed number of florets per spikelet. However, the morphological definition appears to be inaccurate as studies have shown that florets of barley and maize spikelets seem to be lateral (Chuck et al., 1998; Zhong et al., 2021). Furthermore, due to the fact that a spikelet is also a modified inflorescence, spikelets may produce a variable number of florets per spikelet among different species (Whipple, 2017). For instance, spikelets in barley are single-flowered, whereas spikelets in wheat and Brachypodium have two-to-several florets and the number of florets per spikelet varies among species/lines (Derbyshire and Byrne, 2013; Sakuma et al., 2019; Zhong et al., 2021). Hence, the states of spikelet meristems in barley and wheat/Brachypodium are unlikely identical, but determinate and indeterminate, respectively, as spikelet meristems stop making new cells after a fixed number of primordia in barley and are capable of producing variable number of lateral flower meristems in wheat/Brachypodium (Laudencia-Chingcuanco and Hake, 2002; Derbyshire and Byrne, 2013; Sakuma et al., 2019; Zhong et al., 2021). The situation gets even more complex and perplexing in the cases of pseudo-spikelets in some bamboos, as the pseudo-spikelets reiteratively produce several rounds of branches from the axils of bracts, which bear lateral florets (Kellogg, 2015). Therefore, spikelets are compound structures and function as basic constitutional units to make up complex inflorescences, thereby harboring both flower-specific and inflorescence-like characteristics. Accordingly, context is needed when talking about the (in)determinacy of a spikelet to precisely discuss its relevant intrinsic properties.

As a compound inflorescence-like structure, the number of florets/grains per spikelet varies considerably among species. In many cases, florets are largely lateral in spikelets, the number of florets and grains per spikelet is therefore controlled by the number of lateral floret meristems emerged in the flanking regions of spikelet meristems and the fertility of initiated lateral florets (Chuck et al., 1998; Kellogg, 2015; Zhang et al., 2017; Sakuma et al., 2019; Zhong et al., 2021; Zhong and Kong, 2022).

The spikelet originated early in the diversification of the grass family

A true spikelet is found in all grasses but not in the subfamily of Anomochlooideae, which is sister to all the other members of the grass family (Figure 1A;Grass Phylogeny Working Group, 2001; Sajo et al., 2008; Grass Phylogeny Working Group II, 2012; Kellogg, 2015; Soreng et al., 2015, 2022). Unlike a typical spikelet having two basal sterile bracts and paired inner lemma and palea, a “pseudo-spikelet” or “spikelet-equivalent,” a flower-bearing structure in Streptochaeta (Anomochlooideae), possesses about eight basal sterile bracts 1-8 and three inner fertile bracts 10–12 enclosing stamens and carpels (Figure 1, A and D; Sajo et al., 2008; Seetharam et al., 2021). It thus becomes critical to accurately determine the identity of these sterile bracts.

As we have discussed in the previous section, MADS1 and BD1/FZP/COM2 genes are two hallmarks for the initiation of spikelet meristems. Previous analyses have shown that the MADS1 ortholog in Streptochaeta is transcribed in the inner three sterile bracts 6–8, and the three fertile bracts 10–12 but not in the sterile bracts 1–5, suggesting that the outer five sterile bracts 1–5 are glume-like and inner three sterile bracts 6–8 are lemma/palea-like in Streptochaeta (Figure 1D;Preston et al., 2009). Indeed, spikelets of Streptochaeta partly resemble those of rice and barley int-m mutants, bearing multiple basal sterile bracts (Zhang et al., 2017; Zhong et al., 2021). A series of studies in rice have identified a number of responsible genes that contribute to the lateral organ identity specification of the unique rice three-flowered spikelet, bearing two basal rudimentary glumes and two successive sterile lemmas (Figure 1, A and C; Yoshida et al., 2009; Zhang et al., 2017; Xu et al., 2020; Zhuang et al., 2020). Particularly, LONG STERILE LEMMA1 (G1) encodes an ALOG transcription factor and is specifically localized in sterile lemmas and basal region of paleae, but not in the two basal rudimentary glumes or lemmas (Yoshida et al., 2009). In contrast, LHS1/OsMADS1 and OsIDS1/SNB (orthologs of IDS1/INT-M) are expressed in lemmas and paleae, but not in sterile lemmas or rudimentary glumes in rice (Table 1; Malcomber and Kellogg, 2004; Prasad et al., 2005; Lee and An, 2012). In addition, a YABBY gene DROOPING LEAF (DL) and an AGL6-like MADS-BOX MOSAIC FLORAL ORGANS1 (MFO1) are found in distinct parts of a rice spikelet, in the developing lemmas/pistils and paleae/lodicules, respectively (Yamaguchi et al., 2004; Ohmori et al., 2009, 2011; Toriba and Hirano, 2014). In short, in rice, G1 controls the identity of sterile lemmas, LHS1/OsMADS1 and OsIDS1/SNB regulate the development of lemmas and paleae, and DL and MFO1 mark the identity of lemma and palea, respectively (Figure 2G and Table 1). Furthermore, NONSTOP GLUMES1 (NSG1)/LACKING RUDIMENTARY GLUME1 (LRG1), a C2H2 zinc-finger transcription repressor, is transcribed in rudimentary glumes and sterile lemmas and likely recruits the co-repressor TOPLESS-RELATED PROTEIN to suppress the expression of LHS1/OsMADS1, DL, and MFO1, thereby regulating the identities of lateral organs of a rice spikelet (Xu et al., 2020; Zhuang et al., 2020). In addition, in rice, LATERAL FLORET1 (LF1) promotes the initiation/outgrowth of axillary buds in the two sterile lemmas, whose gain-of-function mutations in the microRNA165/166 target site lead to the production of two additional fertile lateral florets developing from the axils of the two sterile lemmas (i.e. three-flowered spikelet) (Table 1; Zhang et al., 2017). LF1 encodes a class III homeodomain-leucine zipper transcription factor and is strongly expressed in the emerging floret meristems (Zhang et al., 2017).

Studies in several disparate grass species have highlighted that these genes exhibit a high degree of conservation in their expression patterns, despite some modifications. For instance, the expression patterns of IDS1/OsIDS1/INT-M/AP2-5A are highly conserved in lemmas and/or paleae in maize, rice, barley, and wheat (Chuck et al., 1998; Lee and An, 2012; Zhong et al., 2021). Maize DROOPING LEAF1 (DRL1) and DRL2, orthologs of rice DL, seem to play conserved function in the regulation of leaf, spikelet, and carpel development (Strable et al., 2017; Strable and Vollbrecht, 2019). Moreover, AGL6-like genes, such as maize BEARDED-EAR (BDE), are likely conserved in the patterning of paleae in grasses (Reinheimer and Kellogg, 2009; Thompson et al., 2009). Likewise, transcripts of BD1/FZP/COM2 orthologs are uniquely restricted in the axils of the two basal glumes, inhibiting the initiation/outgrowth of axillary buds in maize, rice, and barley (Figure 2B;Chuck et al., 2002; Komatsu et al., 2003; Poursarebani et al., 2015).

In addition, mRNAs of LONELY GUY1 (LOG1, a cytokinin biosynthesis gene), KNOTTED1 (KN1), and APO1 orthologs are detected in the developing floret meristems that are initiated from the axils of lemmas, but not in the sterile lemmas in barley int-m, rice, and maize ids1 mutants (Figure 2, D–F; Vollbrecht et al., 1991; Chuck et al., 1998; Kurakawa et al., 2007; Lee and An, 2012; Zhong et al., 2021). Indeed, in rice, LF1 regulates the initiation/outgrowth of axillary buds in the axils of sterile lemmas likely via the regulation of the rice KN1 ortholog OSH1 (Zhang et al., 2017). Therefore, we argue that the combination of these tissue-specific marker genes could be used to help interpret the organ identity of leafy bracts of flower-bearing structures in the family of Poaceae, and hence help trace the developmental and evolutionary origin of spikelets (Figure 2G).

Despite having identical names, the pseudo-spikelets in Anomochlooideae and those in woody bamboos are not structurally equivalent (Figure 1, D and E). The pseudo-spikelets in the early-divergent Anomochlooideae have multiple basal sterile bracts, while pseudo-spikelets in woody bamboos produce branch-like structures in the axils of bracts (glumes or lemmas) with newly formed bracts, wherein additional branch-like structures may again emerge from the axils of the new bracts (Soderstrom and Londoño, 1987; Kellogg, 2015). Therefore, the pseudo-spikelets in bamboos cannot be simply considered as “reversals” or reminiscent of the flower-bearing structures in the sister taxa of the rest of grass family (Figure 1, A, D, and E).

In addition, genomes of two early-diverging grass species, Streptochaeta angustifolia (Anomochlooideae) and Pharus latifolius (Pharoideae) have been sequenced, providing an unprecedented view on the formation and evolution of spikelets (Ma et al., 2021; Seetharam et al., 2021). As previously predicted, many floral meristem identity genes, such as MADS-box and AP2 transcription factors, duplicated prior to the diversification of the grass family, concurrent with a genome doubling event ρ (Preston and Kellogg, 2006; Ma et al., 2021; Seetharam et al., 2021). Gene and genome duplications, along with subsequent copy number variation, sub- or neo-functionalization, might have been important to potentiate the origin and evolution of spikelets (Preston and Kellogg, 2006; Ma et al., 2021; Seetharam et al., 2021). The phylogenetic analyses of AP2 genes indicate that Streptochaeta might have lost the ortholog of IDS1/INT-M (Seetharam et al., 2021). Intriguingly, null mutations in INT-M in barley lead to the production of a terminal spikelet that occasionally bears multiple basal sterile bracts, superficially resembling those in Streptochaeta (Figures 1 and 2; Zhong et al., 2021). In barley, INT-M proteins appear to repress the ectopic expression of MADS1 or MADS3/58, thereby inhibiting the formation of a terminal spikelet in the species (Zhong et al., 2021). In addition, INT-M/IDS1/AP2-5A proteins also influence the fertility of florets in maize, barley, and wheat, leading to production of sterile lemmas in spikelets (Chuck et al., 1998; Debernardi et al., 2017; Zhong et al., 2021). Furthermore, we here showed that COM2 transcripts are specifically restricted in the axils of these sterile bracts in int-m mutants using in situ hybridization (Figure 2B), which supports the hypothesis that BD1/FZP/COM2 proteins inhibit the initiation/outgrowth of the axillary buds in sterile bracts (Chuck et al., 2002; Komatsu et al., 2003; Poursarebani et al., 2015; Zhong et al., 2021). Similarly, orthologs of G1 and NSG1/LRG1 could be informative to determine the identities of glumes and/or sterile lemmas, and DL and AGL6/MFO1/BDE could be used as makers for lemmas and paleae, respectively (Figure 2G;Yamaguchi et al., 2004; Ohmori et al., 2009; Thompson et al., 2009; Yoshida et al., 2009; Xu et al., 2020; Zhuang et al., 2020). To sum up, we hypothesize that changes in these tissue-specific spikelet-related genes are of potential importance to infer the evolutionary origin and understand developmental regulation and modifications of spikelets in the grass family (Figures 1A and 2G).

Concluding remarks and future perspectives

The spikelet is a highly specialized and developmentally integrated structure, and hence is often considered as a structural novelty during the diversification of the grass family. Indeed, a spikelet exhibits some unique features of both a flower and an inflorescence, which have led to challenges in precisely determining their identity specification and tracking their evolutionary history. In this review, we have synthesized the latest comparative genetic, genomic, and developmental findings on the specification and evolution of spikelets. Particularly, a set of transcription factors coordinately control the spikelet meristem specification and floral organ identity. Intriguingly, several laser-capture microdissection and single-cell RNA sequencing studies have identified a number of tissue-specific transcripts and genes that might have contributed to the patterning of various types of sequentially emerged meristems during inflorescence development (Satterlee et al., 2020; Thiel et al., 2021; Xu et al., 2021; Yang et al., 2022). In addition, increasing number of fine-scale dissection of protein-protein interactions might also shed insights on how MADS-box transcription factors and other interacting proteins together shape the emergence of many different types of meristems and distinct floral parts that ultimately result in highly complex and diverse grass inflorescence structures (Abraham-Juárez et al., 2020; Li et al., 2021b). Moreover, comparative analyses of expression patterns of floral identity genes in the phylogenetically distinct grass species would help determine how florets cluster and become a coherent and developmentally integrated unit, and how spikelets could be modified in various different lineages (e.g. bamboos) over the course of evolution. Also, a comprehensive understanding of the regulation of a spikelet might allow precise genetic modification(s) to increase the number of fertile florets/spikelets, and hence grain yields (see “Outstanding questions”).

Advances.

• Functional analyses in barley MADS1, wheat WFZP, and various rice and maize mutants have highlighted some important transcription factors that contribute to spikelet identity specification and floral organ patterning of a spikelet.

• Results from barley int-m, wheat ap2-5A, and several rice mutants have demonstrated that the specification and identity of leaf-like bracts could be critical for the understanding of the development and evolution of spikelets.

• Genomic studies of S. angustifolia and P. latifolius, successive sister taxa of the remaining grass species, have shown that modifications of floral patterning genes might be important for the evolution of spikelets.

• Microdissection-based RNA-seq in barley and single-cell RNA-seq of maize shoot apical meristems have identified additional key players during grass inflorescence and spikelet development.

Outstanding questions.

• How do different transcription factors coordinate the sequential production of distinct meristems and spikelet specification throughout development?

• How are the distinct cell types of developing shoot apical meristems specified over time?

• How does INT-M regulate cell-type specification?

• How do changes in the genes that play a role in spikelet patterning contribute to the origin and variation of spikelets?

• How may changes in the floral identity genes shape the diversity and variation of inflorescences and spikelets of bamboos?

 

J.Z. conceived and designed the project. Y.W., X.B., and J.Z. acquired the funding for the project and wrote the manuscript.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions) is: Jinshun Zhong (jinshun.zhong@scau.edu.cn).