The theater management model of plant memory

Abstract

The existence of a memory in plants raises several fundamental questions. What might be the function of a plant memory? How might it work? Which molecular mechanisms might be responsible? Here, we sketch out the landscape of plant memory with particular reference to the concepts of functioning-dependent structures and competitive coherence. We illustrate how these concepts might be relevant with reference to the metaphor of a traveling, avant-garde theater company and we suggest how using a program that simulates competitive coherence might help answer some of the questions about plant memory.

Keywords:

• calcium
• cell wall
• cold shock
• chromatin
• phosphorylation
• hyperstructure
• wind

Introduction

The operation of a memory in plants was observed at the beginning of the 1980's.¹ Since then, several other examples of plant memory have been described. Moreover, 2 different aspects of plant memory have been distinguished.² After exposure to the first stimulus or stimuli, one aspect of memory is proposed to entail the plant modifying the pathway that transduces those stimuli and responding immediately; this modification affects the way the plant responds to exposure to the stimulus on subsequent occasions. In this aspect of memory the initiation of the response to the stimulus is immediate. The other aspect of memory is proposed to entail the plant storing information and recalling that information at a later time. In this aspect of memory the response to the stimulus using the memory is delayed.

The existence of a plant memory raises some fundamental questions.³ How and where does a plant encode information from the environment? How does the plant reconcile what appear to be very different aspects of memory? More specifically, what molecular mechanisms might a plant use to store information without immediately using it? And what mechanisms might it use to recall it? How does a plant integrate the environmental information it has stored along with its own capacities to respond, in a final commitment to a growth strategy? In trying to answer these questions, 2 concepts that have been developed recently may be useful.

One of these concepts is competitive coherence which has been implemented in an artificial learning program.⁴ In competitive coherence, the overall operation of the memory depends on a competition between biological elements (such as genes and proteins) leading to their selection for membership of an active subset of elements from the vast set available to an organism.^5,6 This competition is based on the 2 patterns of discrete links possessed by each element, where such links include those between a transcription factor and its target genes, a protein kinase and its target, 2 enzymes in the same structure etc. One of the patterns of links, the Now links, connects those elements that are active at the same time as one another. The other pattern of links, the Next links, connects those elements that are active at one time with those that are active at the following time(s). Competitive selection between these patterns of elements for the inclusion of an element in the active subset might therefore help a plant find a coherent solution to (1) the need for a plant to have a phenotype that is consistent with its internal and external milieus at the present time with (2) its need for a phenotype that is consistent with its milieus at previous times. An advantage of the concept of competitive coherence from the botanist's point of view is that the operation – or absence of operation – of competitive coherence on elements stored within an organism may be one of the criteria needed “for distinguishing mere traces of incidents from true accessible (and actively accessed) memories, which also have to be stored.”³

The other concept is that of functioning-dependent structures, FDSs, which are assembled (or are disassembled) in response to their activity, such as the metabolizing of a sugar.⁷ This activity reflects the environment, and hence the FDSs and their bigger relatives, the functioning-dependent hyperstructures, constitute a measure of the plant's response to the environment.⁸ Given the activity-dependent interaction of enzymes with the cytoskeleton, the concept of FDSs (which is scale-free) can be taken further to include the enzyme-decorated cytoskeleton as a metabolic sensor.⁹

Here, we bring together ideas about FDSs and competitive coherence in a new, unifying, approach to plant memory. We discuss both concepts in the framework of the metaphor of a theater as previously used in the case of an influential model of consciousness.¹⁰ In our version of this metaphor (which does not involve consciousness in any way), plant memory can be viewed as a traveling theater in which the actors must be chosen and the play itself adapted in response to feedback from an audience representative of the different audiences who will pay to see the performance.

The Theater Management Model of Plant Memory

Our model of plant memory is the metaphor of a peripatic, avant-garde theater. Management of this theater entails casting to fill different roles in a play, rehearsing and adapting this play to the tastes of the audience, and, finally, performing the play. The casting process requires a cast of actors – a subset of genes, macromolecules and ions – to be chosen from a large number of unemployed candidates – unexpressed genes, unsynthesized or inactive macromolecules, and ions at ineffective concentrations. This casting continues through rehearsals. The play is dynamic and interactive: the plot, roles and cast change in response to feedback from a non-paying, invited audience, who represent the environment. The group of actors onstage at any one time changes in a meaningful sequence. An actor can be either onstage playing a role or waiting backstage (or in the wings) ready to come on.

Competitive coherence–choosing the cast

The casting process is described by competitive coherence, which is a learning strategy for choosing a subset of elements to determine the state of the system from a larger set of inactive elements.^4,6 In the context of the theatrical model of plant memory, competitive coherence allows the management to select actors to be on onstage based on 2 sets of requirements. The first requirement is that the actors have a coherent relationship to the present scene; this corresponds to the requirement for memory to be immediate. The second requirement is that these actors have a coherent relationship to the actors present in the preceding scene; this corresponds to the requirement for memory to be sequential in order to take into account temporal changes. Taking the 2 requirements together, the metaphor translates into the developing plant storing a pattern of activity of elements (genes, macromolecules, ions etc.) that is coherent with both its present internal and external milieus and its preceding internal and external milieus.

The elements of memory–the actors

There are 2 classes of elements in our model. One class comprises functioning-dependent structures, FDSs (see below) in the wall, plasma membrane and cytoplasm which are mainly responsible for the immediate aspect of memory. These FDSs capture – indeed constitute – the state of the cell at any one time. Such FDSs include those involved in metabolism and signaling and thus may comprise both enzymes and cytoskeletal proteins. The second class comprises certain genes in the nucleus as well as the polymers that constitute the cell wall; the position and modification status of these genes along with the composition and patterns of the wall polymers reflect the history of the cell and hence the sequential aspect of memory. The genes in this second class may include those involved in sensing temperature, mechanical stimuli, signals from competitors etc.

The mechanisms for creating memories–the decisions of the director

The candidate mechanisms responsible for the immediate aspect of memory include the processes determining the dynamics of FDSs, which are assembled (or are disassembled) in response to their activity, such as the metabolizing of a sugar; this activity reflects the immediate response of the plant to the environment. These FDSs exist in the cytoplasm (but may extend to the wall) and the laying down of the immediate aspect of memory may be based on the interactions of these FDS with one another, with the wall, with the plasma membrane and with the cytoskeleton. Note that FDSs can exist at several levels of organization and, for example, include not only structures within the organelles but also the structures made by organelles in association with the cytoskeleton (as well as multicellular structures made by assemblies of cells, etc.); such FDSs may be stabilised by the post-translational modification of their enzyme and cytoskeletal constituents.

The candidate mechanisms that may be responsible for the inclusion of an element such as a gene in the sequential aspect of memory include: (1) the methylation and demethylation patterns of the DNA and nucleosomal histones and (2) the binding of transcription factors to the gene. Certain of these mechanisms also favor the selection of a coherent set of memory (alias STO) genes because, for example, transcription factors may themselves interact with one another to increase the chances of the genes to which they bind becoming memory genes. The temporal coherence of the successive patterns of memory genes is achieved by coupling the selection of these genes to the expression of the genes controlling circadian and other rhythms.

Laying down the memory, that is, choosing the elements, is a learning process that involves reward and punishment. A part-time or unemployed actor who has just played a scene is more likely (than one who has never been called on) to hang around backstage and therefore be called on again to play another scene. A protein or gene that has been recruited temporarily to the memory to be expressed may undergo a modification that increases its chance of staying available in the ensemble of the memory (i.e., the equivalent of onstage and backstage); in general, modifications to elements, such as the methylation of genes or the phosphorylation of proteins, can increase or decrease the probability that these elements become part of the memory. An actor who has just come offstage from playing in one scene is likely to be followed by an actor who plays in the chronologically next scene in the play (and the first actor may phone this second actor to remind him that his scene is likely to be next). Here, the metaphor means that the genes that have just been expressed (or proteins that are recruited to an FDS) are modified to increase or decrease the probability that the following subset of genes (or proteins) will be expressed (or recruited) again.

Competition for inclusion in the memory means that the plant can combine the coherence of the subset of FDSs assembled and genes expressed in response to a specific environment at a particular time (i.e., immediate memory) with the temporal coherence of the subsets of FDSs assembled and genes expressed as an environment changes over time (i.e., sequential memory). Candidate mechanisms for the competition between candidate elements in the immediate and sequential aspects of memory include: (1) the coupling of the chromatin – via the nucleoskeleton and the cytoskeleton – to the cytoplasmic FDSs (which may even extend to the wall) and (2) the redistribution of calcium via the condensation and decondensation of calcium onto and from the chromatin and FDSs that constitute the memory.^11-13

Commitment–the first night

Once the play has been rehearsed, the first performance is given. Thereafter it is very difficult to change the play. This is also the case for plant memory. Once the plant has sampled and integrated its environment (along with its own capacities), its memory is used to determine a particular regime of differentiation. The timing of the commitment to this regime is a compromise between maximising its sampling of the environment and maximising the time during which it can exploit its environment optimally. The mechanism that determines the timing of the end of the laying down memory and its final recall into the phenotype is based on the plant interpretation of annual rhythms such as those involving plant hormones.

Examples of the Immediate Aspect of Plant Memory

In Arabidopsis thaliana, cold shock and hyperosmosis result in a transient increase in cytosolic calcium; in the former case, cold pre-treatments reduce this increase¹⁴ while a hyperosmotic-stress pre-treatment heightens it.¹⁵ In Nicotiana plumbaginifolia seedlings, a wind stimulus also results in a transient increase in cytosolic calcium but repeated wind stimuli lead to a reduction in the amplitude of the corresponding calcium transients.¹⁶ Other examples include the summing of electrical stimuli in the Venus flytrap,¹⁷ the interpretation of gravitropic stimulation in graminean coleoptiles¹⁸ and the adaptation of the phosphate uptake system by the history of phosphate levels in Anabaena variabilis.¹⁹ More recently, immediate memory has been shown in case of the rapid closure of the leaves of Mimosa pudica.²⁰ This leaf-folding is a defensive response to predation that can be elicited by mechanical disturbances such as those resulting from water dropping on the leaves. Mimosa rapidly learns to suppress this response when it is non-adaptive, as in the case of a series of drops that are not associated with predation, and such suppression or habituation can last without reinforcement for at least a month.²⁰

Examples of the Sequential Aspect of Plant Memory

Pricking the cotyledons of Bidens pilosa seedlings shortly after germination,²¹ followed days later by transfer to a nutrient medium with a very low concentration of mineral ions results in a reduction of the hypocotyl growth, a reduction that does not occur in a standard nutrient medium.²¹ Removal of the terminal bud (“decapitation”) of Bidens pilosa seedlings in certain conditions of light and mineral concentrations allows one of the cotyledonary buds to start growing before the other; the selection of the bud that grows first is random unless one of the cotyledons is pricked, in which case the other bud grows first; the information leading to the preferential growth of the bud at the axil of the non-pricked cotyledon is stored and is only revealed if the plants are subjected to a particular treatment, such as a change of temperature, at a time that can be days later.²² Abiotic stimuli such as the manipulation of flax seedlings followed at a later time by a temporary depletion (e.g., for one day) of calcium in the nutrient medium leads to a production of epidermal meristems in the hypocotyl; without either the abiotic stimulus or the calcium depletion, the meristem production does not occur.²³ The degree of leaf-folding of Mimosa (see above) is less in strong light possibly because the greater risk of predation is outweighed by the benefit of increased photosynthesis²⁰ but also perhaps because, in the theater management model, light intensity is central to the circadian and other rhythms needed for the interpretation of information.

The Implication of the Cell Wall in Plant Memory

The cell wall is the principal interface between the cell and its environment which includes of course other cells. It might be argued that it contains a record of events that could be accessed via, for example, a system of mechano-transduction. In the theater management metaphor, the cell could be likened to the stage scenery or even the seating arrangement. Some FDSs are likely to span the cytoplasm and to include regions of the plasma membrane and cell wall. Consistent with this, firstly, direct physical contacts between integral plasma membrane proteins, the wall and the cortical cytoskeleton have been reported (for a review, see²⁴) and, secondly, large scale reorganisations of proteins occur following environmental stimulation. In Arabidopsis, for example, the ammonium transporter AMT1;3 root cells forms clusters in the plasma membrane in response to potentially toxic levels of ammonium²⁵ while abscissic acid signaling involves the recruitment of to a membrane domain of the anion transporter SLAH3 and its regulatory calcium-dependent protein kinase.²⁶ Interpretation of the enviroment via changes to the cell wall is also revealed in the adaptation in the composition of the walls of tobacco pollen tubes in response to sucrose concentration whereby the deposition of methyl-esterified and acid pectins in addition to callose/cellulose occurs in a temporal series.²⁷ Wall composition depends on the location and activity of enzymes such as callose synthase, cellulose synthase and sucrose synthase. In turn, the location of callose synthase and cellulose synthase depends on actin filaments and endomembranes, cellulose and also, for callose synthase, microtubules; in the case of sucrose synthase, its binding to actin filaments is conditional on the concentration of its substrate,²⁸ as expected for an FDS (see below).

The Implication of Calcium in Plant Memory

In most cases, plants rapidly react to stimuli by raising transiently the concentration of free calcium in their cells.²⁹ Following the stimulus, propagation of information throughout the plant may involve an action potential that depends on calcium, as in the response of Mimosa to being touched.³⁰ Transient elevation of cytosolic calcium sets off a series of processes such as the opening of transmembrane ion channels, post-translational modifications of some proteins and the expression of certain genes. This chain of processes leads to the final response of the plant, which may involve leaf movements, morphogenetic changes and metabolic modifications. Moreover, the kinetics and magnitude of the transient increase of cytosolic calcium (which are different for the different stimuli) are proposed to orient the system toward a response appropriate to the particular stimulus that has been perceived.^15,31-33 Consistent with this importance of calcium, plant memory is perturbed by treatment with pharmacological agents that are known to affect cytosolic calcium and hence any transient increase of cytosolic calcium that might follow stimulus perception.^16,34

With its multiple bridging roles, calcium might increase the probability of colocalisation of specific proteins, nucleic acids and lipids and hence contribute to the formation and the stabilization of FDSs. The condensation of calcium onto charged membranes and linear polymers and might also underpin the organization of FDSs and chromatin and their putative action in memory. In many of our experiments, changes in the level of calcium are responsible for triggering what we have termed the recall of stored information and, significantly, calcium variations are central to diurnal rhythms in plants^35-39 and possibly to seasonal ones.^40-42 This is consistent with such variations playing a role in the sequential aspect of memory.

Implication of Functioning-Dependent Structures or Hyperstructures in Memory

It is conceivable that the entire cytoskeleton with its associated enzymes acts as an FDS that can integrate intracellular and extracellular information.⁹ Such metabolic sensing would put the enzyme-decorated cytoskeleton in a strong position to be a central player in memory. In higher plant cells, MAPs play a major role in the dynamics of the MT network⁴³ and the association of certain enzymes with the cytoskeleton contributes to these dynamics. The cytoskeleton binds some enzymes when they are active, that is, catalyzing their reactions, while it binds others when they are inactive. In plants, protein–protein interactions have been found between actin and enzymes that include cytosolic aldolase, 3 GAPDH isoforms and 2 enolase isoforms, as well as between tubulin and enzymes that include aldolase, GAPDH and sucrose synthase.⁴⁴ Dynamic interactions between microfilaments and MTs also occur in A. thaliana.⁴⁵ In line with the processing of information that is central to memory, in maize, the presence of sucrose is required for the association of sucrose synthase with microfilaments in vitro and probably in vivo.⁷

FDSs can also include those found within mitochondria and chloroplasts and, at a higher level, those between the organelles and the cytoskeleton. In chloroplasts, association between glyceraldehyde-3-phosphate and phosphoribulokinase leads to the latter's activation which persists even after the enzymes separate.⁴⁶ An FDS comprising the Krebs Cycle enzymes might form via increased affinities of enzymes for one another in the presence of substrates and/or calcium and, in line with a role for calcium in FDS-mediated memory, the activities of 3 Krebs cycle dehydrogenases – pyruvate, isocitrate, and α-ketoglutarate dehydrogenase – are modulated by calcium (for references see⁴⁷). The information concerning increasing demands for ATP, which is coded in calcium transients, is stored in the mitochondria of Hela cells and is proposed to involve changes in the activity of mitochondrial enzymes⁴⁷ which may in turn promote FDS assembly. Ilv5p, a mitochondrial protein that catalyzes the synthesis of branched chain amino acids, has been implicated in the formation of mtDNA nucleoids in S. cerevisiae.⁴⁸ Also in this organism, a subunit of α-ketoglutarate dehydrogenase, Kgd2p, is one of 20 proteins found by cross-linking to be bound to mtDNA while double mutants affected in this protein and another DNA-binding protein, Abf2p, produce cells lacking mitochondria⁴⁹ and, for other references, see.⁵⁰

In our hypothesis, plant memory is also based on changes in the organization of the chromatin during the learning period. Chromosome territories, for example, can be regarded as FDSs. If FDSs in the nucleus (as well as in the wall plus cytoplasm) do indeed constitute the molecular basis of memory, there should be systems that link them. This is the case: in animals and plants, the Sad1/UNC-84 (SUN) domain proteins are part of a complex that bridges the nuclear envelope to connect cytoskeletal elements to the nucleoskeleton and chromatin.⁵¹ These complexes allow transmission of cellular signals to the nucleus and are essential in various cell functions such as the movement of the nucleus and movement within it. SUN domain proteins such as AtSUN1 and AtSUN2 are also found in plants.⁵²

Discussion

In our vision of plant development, the early stages of growth entail the plant receiving, processing and storing diverse information from the environment for use at later stages of development. This information comprises the temporal dynamics of factors that therefore include temperature, light, rainfall, wind, nutrients, competitors, predators and pests. To optimise its chances of survival, growth and reproduction, the plant must integrate these factors together with its own capacities. In other words, a plant must learn and a memory is therefore essential.

There are 2 opposed learning strategies that a plant might adopt based on the immediate and delayed aspects of memory. One strategy would be the equivalent of reconnaissance by fire in which one military force fires on possible enemy positions to provoke a reaction and learn about them. In this strategy, the plant would grow immediately and would try to alter its particular pattern of growth during the acquisition of new information; this would have the advantage of the plant getting started but the disadvantage of the high probability of the selected growth pattern being far from optimal. The alternative strategy would be for the plant to learn about its environment for an entire year (or even longer!) before committing itself to a particular pattern of growth; this would have the advantage of a high probability of the pattern being right but the disadvantage that the plant would have left the field wide open to its competitors (which would change the environment that needed to be learnt). An attractive compromise solution would be for a plant to grow initially in a reversible pattern before committing itself to a growth pattern relatively early in the growing season.

In a computer simulation of the competitive coherence model of learning, competition between the equivalent of the processes responsible for the immediate and sequential aspects of memory (Now and Next processes, respectively) for inclusion in an active subset of elements is fundamental to learning. We have argued here that the Now and Next processes have their counterparts in plant memory. If so, it would mean that it is essential for the plant to respond to its environment in order for it to learn. It would therefore mean that the same molecular mechanisms would have to be involved in the immediate and sequential aspects of memory because the underlying processes would be competing for the same final mechanistic space. With this reasoning, if FDSs in the cytoplasm and nucleus (plus probably the wall) are involved in memory, both types of FDS are involved in each of the processes.

An apparently different possibility would be if the immediate and sequential aspects of memory were to depend on, for example, cytoplasmic and nuclear FDSs, respectively. Here, the competition at each step of learning by the plant might entail finding coherent cellular solutions to the problem of integrating FDSs in the cytoplasm and nucleus. This is aesthetically attractive. Moreover, a shift in the importance given to structures in, say, the nucleus, might underpin the commitment step.

The feasibility of the above possibilities might be tested using a program that simulates competitive coherence. The role of cyclically activated proteins and cyclically expressed genes might be tested using elements in the program that are activated cyclically (i.e., selected cyclically for membership of the active subset of elements). The value of separating cytoplasmic and nuclear FDSs might be tested by dividing the elements of the program into 2 classes and by treating them differently. Such differential treatment could include a change in the relative weighting given to the elements such that, at the time representing commitment, the Next process (representing the sequential aspect of memory) takes on a greater importance in determining the behavior of the system.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank an anonymous reviewer for many helpful suggestions and, in particular, the inclusion of the wall in the model.