Experience teaches plants to learn faster and forget slower in environments where it matters
- First Online:
- Cite this article as:
- Gagliano, M., Renton, M., Depczynski, M. et al. Oecologia (2014) 175: 63. doi:10.1007/s00442-013-2873-7
- 4.5k Downloads
The nervous system of animals serves the acquisition, memorization and recollection of information. Like animals, plants also acquire a huge amount of information from their environment, yet their capacity to memorize and organize learned behavioral responses has not been demonstrated. In Mimosa pudica—the sensitive plant—the defensive leaf-folding behaviour in response to repeated physical disturbance exhibits clear habituation, suggesting some elementary form of learning. Applying the theory and the analytical methods usually employed in animal learning research, we show that leaf-folding habituation is more pronounced and persistent for plants growing in energetically costly environments. Astonishingly, Mimosa can display the learned response even when left undisturbed in a more favourable environment for a month. This relatively long-lasting learned behavioural change as a result of previous experience matches the persistence of habituation effects observed in many animals.
KeywordsBehaviourEcological trade-offsInformationAnti-predator responsesLearningMemory
By definition, learning is a process for acquiring memories by which adaptive changes in an organism’s behaviour arise as a result of experience (Okano et al. 2000). Because learning is not directly observable (i.e. we cannot directly access the traces past experiences have left in the memory of an individual), it can only be detected operationally when an observed behaviour has changed due to a past experience that the individual itself remembers (Kawecki 2010). Recognizing and quantifying such behavioral changes has proven to be no easy task in human and animal subjects, let alone in plants whose ability to behave is still an underappreciated phenomenon (Karban 2008). Even so, researchers have attempted to experimentally test (Applewhite 1972; Thellier et al. 1982, 2000) as well as conceptually validate [through mathematical models (Demongeot et al. 2000; Bose and Karmakar 2008; Inoue 2008)] the idea that plants can learn and remember, since Sir J. C. Bose (1858–1937) first proposed this over a century ago. What these studies have shown is that plants are indeed capable of memory function (see Trewavas 2003; Thellier and Lüttge 2013); in all cases, the behaviour of the different plants included an ability to store and recall biological information in order to produce a specific (and somewhat predictable) developmental (Karban and Niiho 1995; Goodrich and Tweedie 2002; Reyes et al. 2002), physiological (Baldwin and Schmelz 1996; Ruuhola et al. 2007), morphological (Dostál 1967; Thellier et al. 1982, 2000) or epigenetic response (Sung and Amasino 2004; Molinier et al. 2006; Kinoshita and Jacobsen 2012 and related articles in special focus issue). Defined by a fixed ‘blueprint’, such induced adaptive responses are extremely valuable at adapting the organism to its environment and, in fact, are part of an innate behavioral repertoire of an organism. However, this leaves open the question of whether plants, like animals, can show learned behavioral responses (i.e. a behaviour that an individual develops by being taught).
The rapid closure of Mimosa’s delicate leaves in response to mechanical disturbance is believed to be a defence tactic to reduce predation risk (Eisner 1981; Braam 2005; Fromm and Lautner 2007) and this leaf-folding behaviour may be adaptive (Cahill et al. 2013). Moreover, leaf closure is known to be sensitive to light levels and specifically, Mimosa plants are more likely to risk predation when foraging at lower light intensities (Jensen et al. 2011). In this study, we subjected plants to repeated treatments that initially caused leaves to close but the response to which was unnecessary and non-adaptive, to test whether plants could learn that these stimuli should be ignored. Specifically in our first set of experiments, we examined whether Mimosa plants grown under low- (LL) and high-light (HL) conditions differ in their rate of learning about an environmental stimulus and retaining what is learned. Based on the predictions by Jensen et al. (2011) that predation risk (i.e. showing visible leaves) should be traded off with energetic gain (i.e. opportunity for photosynthesis with open leaves), we expected that individuals in the low-light environment would be the faster learners and more likely to retain the learned behaviour for longer. In a second set of experiments, we manipulated the light conditions individuals experienced over time, thereby changing the trade-off between predation risk and energetic reward. We expected that individuals originally grown under the HL environment and later experiencing LL conditions would become faster learners, while those grown under the LL environment and transferred to higher light conditions would display a reduced response.
Materials and methods
Experimental set-up and design
All experiments were done in a 5.30-m2 growth room with three compartments, separated from each other by sheets of black plastic. All training trials were conducted in the middle compartment, whereas all experimental plants were held in one of the side compartments, fitted with fluorescent lights to produce LL (90 μmol m−2 s−1) and HL (230 μmol m−2 s−1) conditions respectively. Light intensity was measured directly above the plants in each compartment prior to the commencement of the experiments. In a first series of experiments, we tested whether light conditions influenced how Mimosa’s leaf-folding reflex habituates and how long habituation persists (i.e. short-term; after plants were left undisturbed for 6 days). In the second series of experiments, we examined long-term habituation of the reflex by re-testing 26 previously trained plants that were left undisturbed for 1 month and assessing whether Mimosa’s habituated response, like in animals (Sztarker and Tomsic 2011), can be evoked only if the plant is tested in the same environment where it was trained. All similarly sized (6–8 cm tall) plants were obtained from a local commercial nursery (Cooperative Legnaia, Firenze) and directly transferred to the experimental growth room at the Department of Plant, Soil and Environmental Science at the University of Firenze. Here, plants were grown individually in 10-cm round plastic pots with a standard mixture (1:1, volume/volume) of loamy soil and organic compost (Neuhaus N2). For the entire study, we maintained the health and growth of all plants by fertilizing with 1/5-strength Hoagland’s solution (Hoagland and Arnon 1950) and watering as needed (i.e. plants did not express any visible signs of nutrient or water stress). Throughout the study, all plants were exposed to identical 12-h light:12-h dark cycle conditions, 60–70 % ambient humidity and 21–24 °C temperature profiles (air temperature in the compartments housing the plants was recorded using an EasyLog EL-USB 2 data-logger).
Habituation training of Mimosa’s leaf-folding reflex
Consistent with the hypothesis of a trade-off between predation risk and energetic demand, we examined the effects of HL and LL conditions on the habituation training of the defensive reflex of this plant. Plants in individual pots were randomly assigned to either HL or LL conditions and then left undisturbed for 5 days until their scheduled training day. Plants were trained using a custom-designed controlled drop system (Fig. 1c) for administering a standardised stimulus (i.e. a 15-cm fall or drop) that successfully elicits the leaf-folding reflex. The response of individual plants was estimated by averaging the response of three randomly selected leaves and measured by how much these leaves re-opened (if at all) after a train of drops. Specifically, each leaf was carefully measured tip-to-tip using digital callipers; its response was then quantified as the maximum leaf breadth (mm) measured immediately at the end of a train of drops (Fig. 1b) relative to the undisturbed pre-stimulus maximum breadth (Fig. 1a). Both before and after a train, the three leaves from each plant were measured in quick succession by the same individual observer.
We first tested whether Mimosa is able to remember a stimulus based on a brief, one-off experience by administering one drop only in the morning and then again 8 h later to a group of 16 naïve plants (n = 8 per light treatment). Next, we trained a group of 56 naïve plants (n = 28 per light treatment) by administering seven consecutive trains of 60 drops each at either 5- or 10-s inter-stimulus intervals, delivered at increasingly longer inter-train intervals but all within a single day.
To test whether the trained response in Mimosa was due to learning rather than caused by other processes such as fatigue or sensory adaptation (see Rankin et al. 2009), we conducted a standard dishabituation test by assessing the plants’ response following the presentation of a novel stimulus. Dishabituation occurs when a novel stimulus [generally, more intense within a given stimulus modality (Rankin et al. 2009; Kenzer et al. 2013)] is briefly presented and elicits the recovery of the original behavioural response (in Mimosa’s case, leaf closure). The recovered response does not necessarily have to return to the initial levels observed before habituation (Wiel and Weeks 1996), but may be less than or equal to baseline levels (Thompson 2009) or those seen immediately prior to the presentation of the dishabituation stimulus (Petrinovich and Widaman 1984).
In this study, we tested for dishabituation by placing individual plants in a close-fitting foam container attached to a shaker plate (Heidolph, Titramax 1000) and shaking them at 250 r.p.m. for 5 s (Dis-hab treatment). This stimulus was chosen on the basis that it did not differ significantly in modality from the habituating stimulus (Grissom and Bhatnagar 2009), but was nevertheless novel to the plant, and hence able to elicit the rapid leaf-folding reflex as effectively as the brief, one-off experience of the habituating stimulus when delivered during the initial test described above.
Short-term memory of the habituated leaf-folding response
To examine the short-term effects of 1-day training on the retention of the habituated response, all trained plants from both light treatments were re-tested 6 days later by observing the leaf-folding response as they received one train of 60 drops.
Long-term memory in Mimosa exposed to a new environment
To test the effects of a changing environment, we conducted a second series of experiments where 13 plants (including five control plants) previously habituated to LL conditions were transferred to the HL treatment and vice versa, and then re-tested 28 days later by undergoing the complete training (i.e. 7 consecutive trains of 60 drops delivered throughout 1 day).
Statistical design and analyses
Analyses were conducted in R using the core and lme4 packages (Bates et al. 2011) to fit and compare linear mixed-effects models, the appropriate modern technique for dealing with this kind of nested, repeated measures data. For each analysis, we first fitted a full model and then conducted step-wise simplification to determine which terms were significant in the standard way (Crawley 2007). Specifically, at each step a simplified model with one term dropped from the previous best model (starting with the full model) was compared to the previous best model using both Akaike’s information criterion (AIC) and a likelihood ratio χ2-test (see Electronic Supplementary Material for more details). AIC values were computed for each of the candidate models and the model with the lowest AIC value was selected in the standard way as the best model of the observed data (Burnham and Anderson 2002).
Habituation training of Mimosa
Response of the habituated leaf-folding reflex to a novel stimulus
We found that the habituated response returned to original baseline levels when the novel stimulus—a shake at 250 r.p.m. for 5 s—was applied (i.e. dishabituation; Fig. 2; differences between +6 h and DIS-hab, P < 0.0001, χ2 = 350 for both environments). This is an important result because it rules out the alternative explanation for a decrease in re-opening response due to exhaustion of energy or other resources. Moreover, it was interesting to find that the habituated response is only provisionally overridden [as observed in animal studies (Grissom and Bhatnagar 2009)] and instead can be fully elicited again when the original familiar drop stimulus is re-presented 10 min later (Fig. 2; differences between DIS-hab and +10 min post-DIS-hab, P < 0.0001, χ2 = 401, and no difference between +6 h and +10 min post-DIS-hab in both environments, P = 0.19, χ2 = 1.7). The maintenance of the habituated response despite dishabituation confirms that the response decrease cannot be attributed to fatigue or sensory adaptation; instead it denotes the selective nature of habituation, an active learning process whereby plants (like animals) perceive an innocuous stimulus but choose not to respond to it while still remaining responsive to the surrounding environment.
Short-term memory of the habituated leaf-folding response
Long-term memory in Mimosa
This not only excludes the possibility that this is simply rapid re-habituation, it demonstrates an increase in the tendency of individuals to take a specific action in response to a known stimulus based on prior learning.
Most probably the oldest form of learning, habituation can be observed across a wide range of organisms. Clearly being able to generalize a learned response is of considerable adaptive value, as long as the response is stimulus-specific enough to allow the organism to keep reacting to unknown, potentially dangerous, stimuli (Rankin et al. 2009). Our results indicate that this is the case for Mimosa; in fact, all experimental plants irrespective of their light treatment respond promptly by closing their leaves when a novel dishabituating stimulus is presented (Fig. 4), hence exhibiting stimulus specificity. Concomitantly, Mimosa can acquire an enduring memory of a past event, whereby the plant recognizes and generalizes the learned stimulus even when the context in which the memory was generated has changed (considering that each plant was immediately returned to its specific light environment in between each training session, the ‘context’ a plant experienced during a whole training day would be best described by the combined ‘holding’ and ‘testing’ environment). In other words, Mimosa’s proficiency in responding appropriately to a known stimulus within a new context does not require brand-new learning but the capacity to ‘remember’ (i.e. recognize that a present cue has been previously encountered) and adjust slightly. It seems, then, that the trade-off between stimulus generalization and stimulus specificity is an essential feature of how plants (akin to animals) respond to their environment. As recently pointed out in the animal literature (Sztarker and Tomsic 2011), the final decision of how an organism will behave is based on the similarity between the known stimulus and the new one, assessed along a perceptual gradient that is relevant to the individual species.
In this study, we have demonstrated the acquisition and expression of a long-lasting memory for a learned behaviour in this plant, but we have left unanswered the fundamental question of the mechanisms underlying this deceptively simple form of learning, the biological basis of which we still do not fully understand in any organism [but see work on invertebrates, such as crayfish (Krasne and Teshiba 1995); sea slugs (Glanzman 2009); and nematodes (Giles and Rankin 2009)]. The current hypothesis for neural organisms (primarily simpler invertebrate animals) proposes that there are a number of cellular mechanisms that are differentially activated to mediate short- and long-term habituation (Rankin et al. 2009) and specifically, the long-term effects involve voltage-gated ion channels, neurotransmitters and changes in protein synthesis [in Aplysia californica (Esdin et al. 2010)]. Interestingly, the physiological and cellular mechanisms of leaf movements associated with fast responses to environmental stimuli in plants like Mimosa have been known for some time (reviewed by Moran 2007), revealing that these plants employ sensory and motor pathways (Shepherd 2012) with the direct involvement of voltage-gated ion channels (Fleurat-Lessard et al. 1997), neurotransmitters (Roshchina 2001) and changes in protein synthesis [e.g. aquaporin (Uehlein and Kaldenhoff 2008)], all of which may also be utilized during the learning process. Hence, we propose that plants such as Mimosa offer a unique opportunity for investigating the biology of learning and memory, notwithstanding the current belief that plant behaviour is a collection of simple, automata-like routines. After all, considerable progress in this research field over the last few decades has been made possible by the study of invertebrates as model species, a choice originally based on the assumption that these were simple animals performing simple tasks [yet, they have turned out to be far more complicated than we assumed (reviewed by Tomsic et al. 2009)].
So how do plants learn and remember?
Plants may lack brains and neural tissues but they do possess a sophisticated Ca-signalling network in their cells (Yang and Poovaiah 2003) similar to those underlying memory processes in animals (Perisse et al. 2009). Specifically, intracellular Ca ([Ca2+]i) signals are known to regulate a large variety of functions in all biological organisms (Berridge et al. 2000), including memory processing and formation of memory imprints of past events ranging from minutes to generations through gene expression (Perisse et al. 2009; Gális et al. 2009). Interestingly in animals, fluctuations in [Ca2+]i during learning seem to be essential in priming the organism for the formation of long-term memory, without affecting short-term memory (Perisse et al. 2009; Bauer et al. 2002). In plants, this same [Ca2+]i system is already known to contribute to the formation of stress imprints (Conrath et al. 2001) and may be responsible for the long-term memory we observed in Mimosa. As a matter of fact, in both animals and plants, fluctuations in [Ca2+]i levels are directly linked to stimulus–response coupling through changes in the concentration of small molecules and proteins, including calmodulin (CaM). What is really interesting about this Ca2+ sensor is that CaM is one of the most conserved Ca2+-binding proteins in eukaryotes [>70 % sequence alignment similarity between animals and plants, albeit plants have a much larger repertoire of genes coding for CaM-target proteins (Yang and Poovaiah 2003)]. The Ca2+/CaM signalling system controls the expression of genes whether their transcription occurs in the neurons of a big-brained animal (Limback-Stokin et al. 2004) or in the root apex cells of a plant (Kim et al. 2009). This might be the starting point from which a minute, yet measurable Ca2+/CaM signal acting on voltage-gated ion channels (Halling et al. 2005) and their production of electrical waves (Yellen 1998), culminates in the formation of memories and expression of the most remarkable behaviours in animals [e.g. Aplysia (Esdin et al. 2010)] and plants alike.
Beside the Ca2+ model, there is another possible hypothesis to explain memory mechanisms in plants at the cellular level. It is now known that many non-neural cells are capable of processing information via ion flows generated by ion channels/pumps and propagated by cell:cell junctions [e.g. cardiac tissue (Chakravarthy and Ghosh 1997); bone cells (Turner et al. 2002)]. Simulating neural network-like dynamics, stable bioelectrical gradients among non-excitable cells could store information, memories and exhibit many other properties usually reserved for brains (Tseng and Levin 2013). Plants are no exception, especially when we consider that they have, in fact, well-developed pathways for the effective transmission of information via electrical signals (Volkov et al. 2008).
Additionally at the molecular level, the role of epigenetic reprogramming has been increasingly identified as a promising candidate mechanism for learning and memory storage in plants (Thellier and Lüttge 2013) and more generally, across all taxa (Ginsburg and Jablonka 2009; Ledón-Rettig et al. 2013). In principle, behavioral responses can be modified or refined through the integration of environmental experiences (i.e. learning) via changes in chromatin status [modulated by acetylation and methylation of DNA and histone proteins (Allis et al. 2007)]. In fact, it is well documented that epigenetic variations can be induced by external stimuli (Boyko and Kovalchuk 2008; Alvarez et al. 2010; Yaish et al. 2011) and more specifically, that changes in chromatin methylation patterns are strongly produced following environmental stress (e.g. Molinier et al. 2006; Chinnusamy and Zhu 2009; Verhoeven et al. 2010; Ding et al. 2012) and retained as a source of ‘stress memory’ [e.g. vernalization, whereby the flowering of plants requires priming by previous exposure to chilling (Sung and Amasino 2004); induced plant defences, whereby an anti-herbivore or anti-pathogen response is primed by environmental cues that reliably indicate an increased probability of attacks prior to their occurrence (Conrath 2009); acquired tolerance to drought stress, whereby exposure of young plants to a mild salt treatment primes enhanced drought tolerance in adult plants (Han and Wagner 2013)]. Yet, it should be cautiously noted here that these stress-induced changes in plant response resulting from previous experiences involve a priming input that modifies chromatin patterns but does not lead to an immediate phenotypic response [i.e. memory with delayed output (Ginsburg and Jablonka 2009)]. These are quite different from habituated responses like the one shown by the Mimosa plants in this study, where an input to the gene may produce a behavioral output that operates as a negative regulator of the gene, imposing an inhibitory epigenetic marking which, in turn, leads to a smaller behavioral response when recurring stimulations of the same kind activate that gene [i.e. memory of inhibitory modifications and recall (Ginsburg and Jablonka 2009)]. Secondly, changes in chromatin marks seem to always require severe stress conditions and prolonged exposure to such conditions (reviewed by Pecinka and Mittelsten Scheid 2012), opening an important question as to whether epigenetic variations mediate memory and learning processes when organisms are not under stress. Obviously, different circumstances and environmental conditions affect the behavioral actions an organism takes in different ways; in turn, different behaviours during a learning experience can lead to functionally significant differences in the information acquired as a result of the experience [i.e. a learned response develops within the lifetime of an individual based on sensory feedback (Kawecki 2010)]. While any environmental factor can become a stressor, stress per se [i.e. the organism fails to respond adequately to unfavourable conditions (Pecinka and Mittelsten Scheid 2012)] is induced only when the intensity of the conditions is too high or too low (i.e. beyond the relative sensitivity and threshold value of the organism). By creating situations in which individuals had an opportunity to acquire useful information without being exposed to such extreme conditions, our study suggests that it is worth investigating the epigenetic reprogramming more broadly to gain a complete view on its potential role as a likely molecular mechanism underlying learning and memory functions across living systems.
Once the domain of psychologists, research on learning and memory has extended its reach beyond the study of human behaviour to include several animal species and more recently, even machines. The recent emergence of interdisciplinary approaches like cognitive ecology, which focuses on the functional roles of learning in nature, and how different species obtain and make use of information about their environment to survive (Dukas 2004), have revealed the exceptional importance that the ability to learn plays in biological evolution as a whole (Kawecki 2010). Because of the very fact that much of the advances in learning research come from humans and animals, the acquisition and use of information through the learning process is implicitly accepted to be contingent on neuronal processes (or artificial neural networks modelled on their biological counterparts, in the case of machines), a view that inevitably excludes non-neural organisms such as plants from the behavioral realm of learning, memory and decision-making (but see Trewavas 2003). What we have shown here, however, leads to one clear, albeit quite different, conclusion: the process of remembering may not require the conventional neural networks and pathways of animals; brains and neurons are just one possible, undeniably sophisticated, solution, but they may not be a necessary requirement for learning.
We thank Elisa Azzarello and Elisa Masi for assistance with setting up the light environments, and Leigh Simmons, Joseph Tomkins, Anthony Trewavas, Daniel Robert for valuable comments on the manuscript. This study was supported by Research Fellowships from the University of Western Australia and the Australian Research Council to M. G. and research funding from European Commission to S. M.