‘Ruminate’ endosperm – why and from where? A commentary on ‘The development of the embryo and ruminate endosperm in an early-divergent angiosperm, Asimina triloba (Annonaceae)’

The developing seed has been likened to a Russian doll, since it comprises many layers of tissues that enclose, nourish and protect the embryo. It may seem surprising that internal tissue boundaries so faithfully follow the outer contours of the growing seed, since tissues are under the control of different genetic actors, perform different functions, and develop at different times and rates. A notable exception is the phenomenon of ‘ruminate endosperm’, a focus of the study of Ferrer-Blanco et al. in this issue of Annals of Botany.

Endosperm is described as ruminate when it is penetrated by shallow to deep lobes of the seed coat and/or nucellus in a mature seed (Fig. 1). The term was coined by Asa Gray in 1879, from the Latin ‘to chew’, since the endosperm of the nutmeg seed (Myristica fragrans) appeared to have been eaten in places by ingrowths of the seed coat (Periasamy, 1990). However, it is not at all clear who initiates and who benefits from ruminations, if at all. Most subsequent botanists have viewed embryo-nourishing tissue as ‘the aggressor’; after all, it is a novel bisexual tissue in angiosperms, and it replaces nucellar ovule tissue during seed growth. Is the ruminate phenotype an emergent outcome of differential development – ‘the struggle of two growing tissues that have been abutting upon one another through their whole period of growth’ (Coulter and Land, 1905) – or is it a product of selection on features of adult morphology? Surprisingly, we still have no answer to the question of its function, if any, but detailed developmental studies in non-monocot/eudicot lineages are providing insights into the developmental and evolutionary origins of ruminations, a closely linked topic.

Fig. 1.

Ruminate embryo-nourishing tissues in mature seeds. (A) Torreya taxifolia (a conifer), in which haploid female gametophyte tissue meets diploid nucellar tissue, without participation of the seed coat in the ruminations (transverse section; from Coulter and Land, 1905). (B) Myristica fragrans (a eumagnoliid), in which triploid endosperm, after replacing diploid nucellar tissue, meets deep, vascularized ruminations of the seed coat (transverse section). (C–E) Austrobaileya scandens (Austrobaileyales), in which diploid endosperm, after replacing diploid nucellar tissue, meets shallow, irregular ruminations of the seed coat (longitudinal section). Note false-coloured rudimentary embryo. (D) Ruminate seed coat surface. (E) Transverse-section of fruit with embedded seeds.

In this issue, Ferro-Blanco et al. study the developmental timing and provisioning of the growing tissues of the ovules and seeds of Asimina triloba, a member of the custard apple family, Annonaceae. All members of this mostly tropical family of eumagnoliids and several close familial relatives (Doyle et al., 2008) have some form of ruminate endosperm. The endosperm of Asimina is ruminate in the mature seed, but as the authors show, ruminations are initiated at the nucellus–integument boundary (both maternal tissues) far from the young and symmetrical young endosperm. It is only after the seed reaches its nearly mature size that endosperm begins to replace nucellar tissue in increasingly deep lobes.

But why the convergent evolution of ruminations? Ruminate endosperm is widespread and occurs in at least 58 families of angiosperms (Bayer and Appel, 1996). For Asimina, Ferrer-Blanco et al. conclude that ‘ruminations increase water permeability’ and ‘the ruminate endosperm could facilitate seed germination by enhancing water absorption’. They may be right, but the seed of Asimina is recalcitrant, has a long period of development within fleshy fruits and is dispersed into a typically moist soil environment in its native range, all of which suggests that desiccation sensitivity is the larger problem. In fact, the long literature on ruminate seeds provides few functional hypotheses to choose from. Below I list several published hypotheses (1–5) and call attention to a few others that might be considered (6–9).

• (1) Increased absorption surface: ruminations act like ‘vili in the rumen of ruminants’ to increase surface area of the boundary between tissue, which (1a) could allow more water, oxygen and/or nutrient transfer from the integument to the developing or mature endosperm. Alternatively, (1b) ruminations could also allow more rapid imbibition of water during seed germination. However, integuments have cuticle and, in fact, Ferrer-Blanco et al. show cuticle on both the early developing outer integument and on the nucellar epidermis within deep ruminations.

• (2) Water storage: ruminations could actually act as a water storage tissue, or reservoir for other compounds (Periasamy, 1990; Bayer and Appel, 1996). In Asimina and many other species, seed coat ruminations consist of small, thicker walled cells, whereas there are large, vacuolate cells in the nucellus and endosperm, which suggests that water storage is in tissues internal to the seed coat ruminations.

• (3) Predator deterrent: phenolics, oils, secondary metabolites and tannins are often found in the testal ruminations, but not in the endosperm (references in Bayer and Appel, 1996). Most seeds discussed here are animal-dispersed and face the possibility of being eaten.

• (4) Emerging seedling protection: the ruminate folds might cover, protect and even transfer nutrients even after emergence of the seedling, since it is a thin folded tissue attached to cotyledons during germination and could stretch as the seedling expands during early establishment (Gottsberger, 2016).

• (5) Mechanical support: ruminations could provide structural support. Fibres are often abundant within testal ruminations deep into the endosperm, as shown in Asimina. Interestingly, many of the ruminate seeds in the ANA grade + Magnoliales are intermediate to large albuminous seeds that are dispersed in fleshy fruits. Ruminations might protect the amorphous, thin-walled, large-celled endosperm tissue from getting crushed, causing seed coat rupture, during fruit consumption and dispersal.

• (6) Harmomegathy: In contrast to hypothesis 5, the testal folds and internal fibres could allow for expansion and contraction of the seed coat without cracking during the long period of time it must survive in fluctuating conditions before germination, as originally suggested in the 1880s for sculptured pollen grains (Katifori et al., 2010). This could be especially true of taxa in which the epidermis of the outer integument initiates or participates in the ingrowths, as occurs in many Annonaceae (Corner, 1949; Svoma, 1998).

• (7) Resource allocation: ruminations are costly to make, but cheap to maintain in long-lived seeds – many of the cells are dead at maturity – and the costs of making and provisioning nucellus and endosperm may be greater. Since much of the final volume of a seed is predetermined by nucellar growth before being replaced by endosperm, endosperm might be ‘overproduced’, relative to the minimal needs of the embryo. Ruminations might be a mechanism for reducing the cost of endosperm without reducing seed size. I know of no study that has measured the relative costs of production and maintenance of ovular tissues from anthesis until seed germination.

• (8) Parent–offspring conflict: kin-conflict theories see many of the variant embryo-nourishing phenotypes as outcomes of a struggle for control of the ovule-to-embryo resource pathway. Are ruminations an example of such genetic conflicts playing out? Both unisexual, gymnosperm and bisexual, angiosperm embryo-nourishing tissues invade and replace the nucellus along their lateral margins, but haustorial outgrowths of chalazal endosperm (near the ovule vascular tissue supply) can be interpreted as overly ‘selfish’ endosperm behaviour (Povilus and Gehring, 2022). Maternal responses to ‘selfish’ embryo-nourishing traits have long been hypothesized – for example, the addition of a second (or more) maternal nucleus to bisexual endosperm, perisperm instead of endosperm as a storage tissue, co-option of the chalazal hypostase (a transfer tissue) to instead block and redirect the chalazal nutrient flow, and parent-of-origin gene imprinting. Integument-derived ruminations might also be a maternal response to overaggressive endosperm. Indeed, Westoby and Rice (1982) interpreted the origin of the integument(s) in a similar light, because cuticle on their surfaces restricts the flow of nutrients across the integument–nucellus–endosperm boundary. One might accept the fact of conflicts of interest but still argue that the evolutionary effects of kin selection are small relative to other effects, such as changes in endosperm ploidy (Donoghue and Scheiner, 1992). After all, ruminations are found in haploid-unisexual, diploid-bisexual and triploid-bisexual endosperms (Fig. 1).

• (9) Differential growth rates: no study has proposed that ruminations are a completely neutral trait, but it is entirely possible that they arise simply as an emergent property of differential growth rates caused by ‘overgrowth of the integuments’ relative to the basipetal direction of growth of the ovule, as suggested for Annonaceae (Corner, 1949).

The study of Ferro-Blanco et al. should be placed among a handful of anatomical studies that bear on the impact of the origin of bisexual endosperms in angiosperms and potential ovular accommodations. Asimina seed development retains many probably plesiomorphic traits, such as a massive chalazal nucellus at the time of fertilization and a rudimentary (small) embryo within a massive and slow-growing embryo-nourishing tissue (all shared with gymnosperms), which is ab initio cellular and bisexual (angiosperm synapomorphies) and triploid (a mesangiosperm synapomorphy). Seeds of woody eumagnoliid and ANA-grade angiosperms, including Asimina, are albuminous, often within fleshy fruits, desiccation-sensitive (high water content), with oils as the main storage compound, and have morphological or morphophysiological (Asimina) dormancy. Ruminate endosperms are not likely to be ancestral in angiosperms, since ruminations arise by a variety of different mechanisms, but initiate from various ovular, not endosperm, tissues and have originated in many unrelated, derived lineages. Furthermore, early Cretaceous fossil seeds lack ruminations (Friis et al., 2015).

If bisexual endosperm was originally a transfer tissue alongside a large nucellar storage tissue (Friedman et al., 2012), ruminations in ancient angiosperm seeds might simply reflect an instability in the transference of storage function from nucellar perisperm to bisexual endosperm. If instead the storage function of bisexual endosperm evolved more directly from a unisexual, female gametophyte ancestor, then ruminations would be better interpreted as a maternal (ovular) innovation. The phylogenetic comparative study of Doyle and LeThomas (1996) strongly suggests that the endosperm–nucellus boundary in mature Annonaceae seeds was ancestrally irregular (an undulating boundary with shallow lobes, also in Degeneriaceae and Eupomataceae) followed by increasingly deeper and more canalized ruminations in more derived genera, such as the highly derived lammelate (transverse-oriented, plate-like) forms in Asimina and Annona. There are at least two other origins of ruminate endosperm outside of monocots and eudicots (Fig. 1). The Myristicaceae seed has deep ruminations similar to those of Asimina, except that the lamellae run longitudinally rather than transversely to the ovule axis and are often vascularized (Periasamy, 1990). Myristicaceae and Annonaceae share an ancient Magnoliid common ancestor, which leaves open the possibility that the simplest irregular form of ruminations could be ancestral, followed by modifications or losses within Magnoliids. The Austrobaileya scandens (Austrobaileyales) seed also has shallow-lobed, irregular ruminations, whereas seeds in its sister clade have no ruminations, but a large pre-fertilization nucellus (Friedman and Bachelier, 2013; Losada, 2023). Irregular Austrobaileya-like endosperm ruminations may represent a general ancestral starting point for more canalized ruminate forms. Thus, it seems that irregular ruminations of the seed coat have probably often arisen as a nearly neutral character, and repeated co-option has followed in both ancient and recent lineages. Co-option to what end, however, remains the question!