The plant microbiota, defined here as the community of bacteria, fungi, archaea, viruses, and other microscopic organisms that live on (i.e., epiphytically) or in (i.e., endophytically) plant tissues (Hardoim et al., 2015), confer many services as well as disservices to their hosts, including disease development and defense (Busby et al., 2016), protection against herbivory (Shikano et al., 2017), tolerance of abiotic stress (Rodriguez et al., 2004), and aid in nutrient uptake (Christian et al., 2019). These microbial communities associate with all plant tissues (Hardoim et al., 2015), including seeds (Mundt and Hinkle, 1976; Ganley and Newcombe, 2006; U’Ren et al., 2009; Hodgson et al., 2014; Barret et al., 2015). Seeds play a major role in plant communities as agents of dispersal, genetic diversity, and regeneration (Fenner and Thompson, 2005), and they have significant economic and social value through agriculture (Nabhan, 2012). Seeds also are a major bottleneck in natural plant populations, as they face heightened mortality from abiotic stressors, pests, pathogens, and predators (Bever et al., 2015). As the initial source of inoculum in a plant’s life cycle, seed microbes are can be transmitted across plant generations and have lifelong impacts (Barret et al., 2015; Abdelfattah et al., 2021). Consequently, understanding how seeds acquire and interact with their microbiota, for example, via priority effects (Fukami, 2015; Ridout et al., 2019; Johnston-Monje et al., 2021) or according to the Primary Symbiont Hypothesis (Newcombe et al., 2018), has implications for improving seed health, seedling establishment, and plant community structure. Previous work on seed microbiota has primarily taken a pattern-based approach to studying assembly processes (e.g., Rezki et al., 2018). Such an approach uses culturing (Ganley and Newcombe, 2006; U’Ren et al., 2009; Heitmann et al., 2021) and/or next-generation sequencing (e.g., Barret et al., 2015; Rezki et al., 2018; Prado et al., 2020; Bergmann and Busby, 2021; Fort et al., 2021) to compare, contrast, and correlate patterns in microbial community composition, diversity, and species co-occurrences. Typically, however, these community data provide limited (i.e., mostly indirect) insights into processes such as dispersal, microbe-plant interactions, and microbe-microbe interactions. Given that seed microbial communities are highly variable across individual plants, plant species, and locations (Simonin et al., 2021), such pattern-based data cannot always be used to predict assembly outcomes. Moreover, such studies often consider how these assembly processes occur at a single spatial scale (e.g., between sites or plant populations; Klaedtke et al., 2016; Adam et al., 2018; Chartrel et al., 2021). We hypothesize that a mechanistic, multi-scale approach would provide a more complete understanding of how microbial communities assemble in seeds, with the field of metacommunity ecology providing a theoretical framework for such an approach.

Metacommunity theory accounts for the interaction between ecological processes and habitat heterogeneity across spatio-temporal scales to impact community patterns (Leibold and Chase, 2018). This emphasis on multiple scales and heterogeneity can help explain the main drivers of community assembly and patterns of biodiversity and co-occurrence (Leibold and Chase, 2018). Plant-associated microbial communities vary widely across environmental gradients (Barge et al., 2019) and host genetics (Wagner et al., 2016) from the levels of tissues (Cregger et al., 2018) to populations (Bergmann and Busby, 2021). As such, treating individual plants as heterogeneous habitats for microorganisms that are embedded in a larger, heterogeneous landscape of multiple plants representing different species provides a new approach to observing, testing, and modeling drivers of microbial community variation (Figure 1). However, the study of microbiota through a metacommunity lens is still relatively new, both for animals (Miller et al., 2018) and plants (Christian et al., 2015; Borer et al., 2016), and the plant seed represents a relatively understudied microbiome in this context.

In this review, we address how mechanisms of seed microbial community assembly have been studied at different spatial micro-, meso-, and macro-scales (Figure 1), and advocate for a metacommunity-based approach to seed microbiology in future work. For this review, we use the definition of community assembly from Fukami (2010): “the construction and maintenance of local communities through sequential, repeated immigration of species from the regional species pool.” Additionally, most studies that we cover in our review will be focused on fungi and bacteria (Table 1). We acknowledge that archaea, viruses, and protists are frequent members of plant-associated microbial communities (Trivedi et al., 2020), many plant viruses are seed-transmitted (Sastry, 2013), and viruses can play a major role in the diversity and function of soil microbial communities (Albright et al., 2022). However, the ecological roles of these microbes in plant microbial communities, including those of seeds, are still largely unknown. As such, we cannot speak on their contributions to seed microbiota assembly here and recommend new research on these microbes in seeds. We will first summarize the modes of microbial acquisition into seeds, and how metacommunity ecology frames this assembly process. We then discuss studies of seed microbiome assembly which examine the processes of filtering, species interactions, dispersal, and ecological drift. We specifically highlight studies that address assembly processes during seed development and maturation, as these stages are understudied compared to seed dormancy and germination, and they are likely the source of microbes that persist between plant generations (Chesneau et al., 2020). Lastly, we suggest future lines of research to gain a more mechanistic, scale-explicit understanding of seed microbiome assembly.

Plant seeds are generally composed of three tissues: a seed coat which provides physical protection (Belmonte et al., 2013), an embryo which is the precursor to the seedling and is made up of an immature root, a stem, and one or more embryonic leaves (Boesewinkel and Bouman, 1995; Bewley et al., 2012; Bewley and Nonogaki, 2017), and an endosperm which typically consists of carbohydrates and proteins and provides nutrition for the embryo during germination and growth before photosynthesis can occur (Berger, 1999). Seed development involves three stages (Belmonte et al., 2013; Bewley and Nonogaki, 2017). Following fertilization by pollen, the egg cells divide and differentiate into the embryo and endosperm tissues, in a process called histodifferentiation (Bewley and Nonogaki, 2017). Next, the cells expand and mature with reduced division, and seed mass increases during this filling stage, as nutrient reserves are deposited into the endosperm (Bewley and Nonogaki, 2017). After this, nutrient accumulation declines, and the seed goes into maturation drying and loses about 10%–15% moisture content before it is ready to be dispersed (Angelovici et al., 2010; Bewley and Nonogaki, 2017).

During seed development, microbes may enter the seed tissues via three distinct routes of transmission: vertical, floral, and horizontal (Maude, 1996). Vertical transmission involves microbes traveling from other organs (e.g., stems, roots) of the mother plant to the developing embryo. Such transmission is cited as an ecologically important way for plants to inherit beneficial microbes across generations (Barret et al., 2015) and for seed-associated pathogens to disperse (Darsonval et al., 2008; Darrasse et al., 2010, 2018; Barret et al., 2016). Vertical transmission has long been observed in grasses, which are hosts to clavicipitaceous fungal endophytes (i.e., fungi in the family Clavicipitaceae; Porras-Alfaro and Bayman, 2011) such as Epichloe (Schardl et al., 2004; Truyens et al., 2015). Vertical seed transmission has also been observed for non-clavicipitaceous endophytes in Setaria viridis (Rodríguez et al., 2020), Triticum (Vujanovic et al., 2019), Quercus (Fort et al., 2021), and other plants (Newcombe et al., 2018). Floral transmission of microbes into seeds has been studied extensively for pathogens such as Monilinia vaccinii-corymbosi in blueberry (Ngugi and Scherm, 2004) and Acidovorax citrulli in watermelon (Walcott et al., 2003; Lessl et al., 2007; Dutta et al., 2012, 2015). However, flower-to-seed transmission has also been observed for commensal and beneficial bacteria, for example in Brassica napus (Prado et al., 2020). The microbial contributions of the vertical and floral transmission pathways are likely to vary based on a plant species’ pollination mode (Chesneau et al., 2020). Horizontal transmission is the acquisition of seed microbes from the environment, either prior to or after the maturation of the seed as it is still attached to the mother plant (Deckert et al., 2019) or as matured seed disperse and becomes colonized from sources such as air (suggested in Gandolfi et al., 2013), water (Crocker et al., 2016), animals (Correia et al., 2019; Lash et al., 2020), soil (Fort et al., 2021), and other seeds in storage (suggested in Bergmann and Busby, 2021). Seed dormancy and germination are likely to represent a very active period of such horizontal transmission, as soil microbes interact with seed exudates and pre-existing microorganisms on and within the seed (Nelson, 1990; Ofek et al., 2011).

The metacommunity concept was formally described by Leibold et al. (2004), who defined metacommunities as sets of local communities that are interconnected by dispersal. This definition arose out of a need to better account for spatio-temporal scales in ecological studies (Leibold et al., 2004), and also included the impacts of dispersal and habitat heterogeneity on community patterns (Leibold and Chase, 2018). Since it was first described, metacommunity theory has adopted Vellend (2010, 2016) synthesis that community assembly and composition are driven by four categories of processes: (1) abiotic and host filtering, (2) species interactions, (3) dispersal, and (4) ecological drift (Leibold and Chase, 2018). Categories 1 and 2 (filters and interactions) represent a deterministic or niche-based process of selection where differences in fitness between taxa, species, or guilds lead to differences in their abundances (Vellend, 2010). Dispersal is the stochastic (or neutral, chance-based) process by which taxa move between local communities (Vellend, 2010). Finally, drift is the stochastic fluctuation in species abundances, often due to chance birth, death, and migration events (i.e., demographic stochasticity; Leibold and Chase, 2018).

Framing plant microbiomes as metacommunities provides an integrated view of the drivers of their composition, function, and evolution, and of the impacts of these drivers on host health (Mihaljevic, 2012). Traditional metacommunity ecology states that filtering and species interactions occur at the local scale (i.e., less than one square meter to several square kilometers; Cornell and Lawton, 1992), while dispersal and drift occur at the regional scale (i.e., many square kilometers; Cornell and Lawton, 1992; Leibold and Chase, 2018). However, categorizing processes as “local” or “regional” is relative to the community that is being studied, and depends on the scales of interest and on defining the boundaries between a local community and a regional metacommunity. For plant microbiota, including those associated with seeds, the terms “local” and “regional” are contextual because microbes primarily behave at very small scales (i.e., the micrometer scale; Leveau et al., 2018; Leveau, 2019), although they can be affected by much larger scale factors (e.g., plant genotype or climate gradients across kilometers; Dini-Andreote et al., 2020). Furthermore, microbes can be ubiquitous across habitats at multiple scales, blurring the boundaries between patches of local communities in the landscape of interest (Mony et al., 2020). As we apply the first principles of metacommunity ecology to plant and seed microbiology below, we will therefore use three categories of spatial scale: macro- (i.e., meters to kilometers, or within and among sampling sites), meso- (i.e., centimeters to meters, or within and among plant organs to individuals), and micro- (i.e., nano-to centimeters, or within and among plant organs) scales. Integrating the study of assembly processes across these three scales should give a more complete picture of how microbial communities are assembled, and how emergent community patterns occur at individual scales (Ricklefs, 1987; Shade et al., 2018).

At its core, metacommunity ecology emphasizes not only how the processes described above play out individually, but also how they interact with each other to produce emergent community patterns across scales. In plant microbiome research, the interaction between abiotic and host filters, also known as genotype-by-environment (GxE) interactions, has been of growing interest because it provides a more holistic explanation for microbiome variation (Wagner et al., 2016). Such an explanation can be applied to seed microbial communities, which may vary with seed nutrient profiles, osmotic stress, and water availability. However, as previously mentioned, GxE studies on plant microbiota face a scale problem where genotype and environment become synonymous at the micro-scale. Taking a plant trait-based approach to these studies may make the role of these effects more clear, and can connect micro-and macro-scales via host local adaptation.

While not emphasized as much as GxE interactions, the interaction between dispersal and filtering is also important during seed microbiome assembly. At the micro-scale, variation in the plant surface landscape (e.g., water availability and topography of the surface) can create differences in dispersal limitation between taxa. Doan et al. (2020) demonstrated this interaction on synthetic leaf surfaces, finding that surface water acted as a conduit for bacterial dispersal. This effect may also be present in floral stigmas, which are highly heterogeneous landscapes (Spinelli et al., 2005). Indeed, in their work on transmission of the pathogen A. citrulli from watermelon flowers and fruit to seeds, Dutta et al. (2012) found that inoculum from the flower dispersed more frequently and ended up in deeper seed tissues (e.g., endosperm, embryo) than inoculum from the fruit. While these examples suggest that heterogeneity in the plant landscape impacts dispersal limitation to seeds, more studies are needed.

Dispersal also intersects with species interactions, most clearly through historical contingency or priority effects (Fukami, 2015). In this phenomenon, the arrival order of community members dictates assembly outcomes, typically with an advantage to taxa that arrive first (Fukami, 2015). Priority effects can occur either through niche preemption, where the first colonizers fill all available niches, or by niche modification, where the first colonizers alter the environment and its resulting niches (Fukami, 2015). These effects are often cited as important in seed communities because they have few members (Bergmann and Busby, 2021). However, priority effect experiments in plant microbiota have typically been done in leaf (Leopold and Busby, 2020) and wood (Hiscox et al., 2015; Leopold et al., 2017) communities (Maignien et al., 2014). As such, there is a need to understand the role of priority effects in seed communities.

An exciting new approach for studying the multiple, interactive processes of dispersal, filtering, drift, and species interaction is with Joint Species Distribution Models (JSDMs), which extend single-species distributions to community-level dynamics (Ovaskainen et al., 2017). Leibold et al. (2020) used these models in tandem with variation partitioning to explain the internal structure of simulated metacommunities. They found that this approach was a promising way to connect metacommunity pattern data to multiple assembly processes (Leibold et al., 2020). In the seed microbiology literature, Fort et al. (2021) used JSDMs to infer how maternal filtering and abiotic filtering contributed to seed mycobiome (fungal microbiome) variation in Q. petraea seeds (i.e., acorns). They found that fungal guild (e.g., pathogen, saprotroph, etc.) influenced which taxa varied with abiotic filters, with elevation selecting saprotrophs and seed specialists, and all taxon co-occurrences were positive associations (Fort et al., 2021). While JSDMs were not used in a metacommunity context for this study, and they are limited in their omission of abundance data, these models provide an integrative approach for looking at seed microbiome assembly.

Multiple tools exist for exploring and exposing the effects and interactions of filtering, species interactions, dispersal, and drift on microbial community assembly of individual seeds at multiple spatial scales. However, future work can do a better job of integrating and connecting metacommunity ecology models to traditional seed microbial ecology studies at micro-, meso-and macro-scales. One technical challenge of taking this approach pertains to the interrogation of microbial communities in individual seeds. Culture-based studies of individual seeds report low isolation frequencies, with most seeds containing zero or one microbial taxon (Newcombe et al., 2018, 2022). Additionally, most sequence-based studies to date pool seeds by fields or other groupings (Rezki et al., 2018; Kim et al., 2020; Mascot-Gómez et al., 2021; Morales Moreira et al., 2021; Wassermann et al., 2022). As exceptions, Bergmann and Busby (2021) and Fort et al. (2021) sequenced fungi from individual tree seeds and found that sequencing depth was fairly high. However, the tree species in these studies produce large seeds; sequence-based detection of microbiota might be more difficult in small-seeded species (e.g., Arabidopsis). Additionally, it is often difficult or impossible to treat individual seeds as independent since experimental treatments or predictors are often applied at the fruit or plant level. To resolve these issues, future work could focus on species where seeds are fertilized independently (Helianthus annuus, Quercus sp.), or one seed per fruit/plant could be sampled for large-seeded species. Alternatively, seeds could be pooled at the fruit or plant levels, since these are the levels where treatments are often applied and they sufficiently capture the variation in seed microbiota (Bintarti et al., 2022) while still allowing for a metacommunity approach at the meso-and macro-scales. The appropriate level of pooling should be selected based on the transmission pathway of interest (i.e., fruit level for floral transmission, plant level for vertical transmission; Bintarti et al., 2022). Finally, seeds of large-fruited species could be pooled by parts of the fruit/pod for spatially explicit sampling at the meso-and micro-scales. These scale-explicit pooling approaches, along with the use of additional methods at the meso-and micro-scales (e.g., microscopy of individual seeds, inoculation experiments with synthetic communities), will allow for characterization of microbiota at or near the individual seed level while mitigating issues of low DNA amounts and cross-contamination.

At the micro-scale, there are many opportunities to take a traits-based approach to host filtering of seed microbiota. Experiments can go beyond studying if plant traits have an effect to testing what these effects are (e.g., changes in microbial colonization rates, fitness, dispersal, and species interactions). These experiments could also take a microbial trait-based approach to host filtering (and other processes) and identify the genes required for successful transmission, which are still largely unknown (Chesneau et al., 2020). This could provide valuable insights into the genes required for transmission across the different pathways (Chesneau et al., 2020). Furthermore, metagenomic analyses across plants, populations, species, etc., could determine if these transmission-associated genes are common across metacommunities. Such information could show if there is functional conservation across microbial communities, even if they are taxonomically variable. Finally, micro-scale experiments can also test how microbial community assembly is impacted by the interplay between deterministic and stochastic processes.

In addition to these tests of microbial and plant trait impacts, experiments testing the role of dispersal in seed microbial community assembly among spatial scales should be conducted. At the micro-scale, experiments using synthetic microbial communities on stigmas with varying chemistry and topography can demonstrate how dispersal and selection occur between flowers and seeds, and what the role is of plant genetics and microbial adaptations. At the macro-scale, pollinator exclusion experiments similar to those in Prado et al. (2020) could be conducted across sites in natural landscapes. By using sites at varying distances and connectivity levels from each other, and analyzing both within- and among-site seed microbial community variation, one may obtain new information about how pollinators and patch connectivity impact multi-scale dispersal ability. These proposed studies would elucidate how dispersal contributes to metacommunity assembly among spatial scales.

Along with these single-process studies, we envision studying the interactions between processes through both observational and experimental studies. As JSDMs continue to be refined to model nested and continuous metacommunities, they will provide a way to analyze seed microbiome patterns and their associated assembly processes that is more sophisticated than previous modeling approaches. Additionally, priority effect experiments conducted at multiple points in the seed life cycle (e.g., pollination, dispersal, germination) may reveal how historical contingencies impact seed microbiome assembly throughout the seed life cycle. Such experiments would also test the Primary Symbiont Hypothesis (Newcombe et al., 2018), which argues that seed communities are dominated by a single microbe with significant functional consequences for the plant.

Finally, questions will need to be asked about seed microbiome assembly that go beyond just testing for spatial mechanisms. Primary among these questions is: what fitness benefit does transmission into seeds provide to microbes and their host plants? Such a question gets at the eco-evolutionary dynamics in these microbial metacommunities, which can have long-term consequences for both microbes and plants. Because microbial communities behave and evolve at shorter time-scales than macro-organisms (Nemergut et al., 2013), it is feasible to design simple experiments testing how microbes evolve in response to plant defenses, nutrient availability, and micromorphology. Such eco-evolutionary studies may have applications in understanding microbial community shifts with crop domestication (Kim et al., 2020). Additionally, both microbes and seeds have dormant stages, which can impact metacommunity dynamics through tradeoffs with dispersal and delayed responses to environmental conditions (Wisnoski et al., 2019). The role of dormancy in seed and plant microbial metacommunity assembly has yet to be explored, so there is much room to study how dormancy impacts these systems over longer temporal scales. Finally, a hot topic in plant microbiome research is how to modify plant microbial communities for climate resilience and other beneficial traits (Busby et al., 2017; Mitter et al., 2017). However, the impacts of climate change-associated disturbances on plant microbiomes have been limited to pattern-based studies in leaves and roots (e.g., Kivlin et al., 2013). As such, more work can be done on how disturbances alter seed microbiome assembly processes and outcomes.

Seed microbial community assembly, much like the assembly of microbial communities associated with other plant parts, is the result of complex interactions between multi-scale processes. Metacommunity ecology provides a conceptual framework for identifying these processes, and provides new statistical and theoretical tools for testing their interactions. We advocate for the spatially explicit, multi-scale study of seed microbial community assembly, with emphasis on the effects of plant topography and chemistry on micro-scale dispersal and persistence, the role of pollinators and seed dispersers on macro-scale dispersal, and the interactions between processes. These new avenues of study will provide a more generalizable understanding of seed microbiome assembly, with potential applications in plant conservation and sustainable agriculture.

GB wrote and revised the manuscript and created the manuscript figure and tables. JL revised the manuscript and associated tables. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor KM declared a shared affiliation with the authors at the time of review.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.