Vertically- and horizontally-transmitted memories - the fading boundaries between regeneration and inheritance in planaria
Following splitting, each fragment obtains a different composition of molecules (e.g. proteins, RNA molecules, gradients of morphogens) (Adell et al., 2010), which influence and guide its subsequent physiology and regeneration. The existence of such gradients and local environments in the worm may contribute to the initial state of the newly-formed fragments. In addition to short-term immediately derived ‘maternal effects’, long-term effects, amplified by positive feedback processes, could perpetuate after ‘maternal’ factors are diluted. It must be noted that such gradients have to be self-scaling, to maintain their instructive pattern within the resulting small fragments (Ben-Zvi and Shilo, 2011; Werner et al., 2015).
Epigenetic mechanisms in planarian neoblasts are currently being explored (Duncan et al., 2015; Hubert et al., 2013; Robb and Sánchez Alvarado, 2014; Rouhana et al., 2014). If different environmental events affect small RNA pools (microRNAs and piRNAs have been described in planaria) or chromatin modifications in a spatially restricted manner, then such epigenetic processes, which in a number of organisms perpetuate transgenerational gene regulation, could mediate asymmetry following fission. Interestingly, as is the case in C. elegans nematodes and in plants, RNA interference (RNAi) works systemically in planaria (Rouhana et al., 2013). Thus, in theory, small RNAs could allow both spatial and temporal spreading of epigenetic memory in planaria.
As neoblasts are influenced by information received from other somatic cells around them (Oviedo and Levin, 2007), the practical meaning is that in planaria a breaching of Weismann's barrier could take place. While neoblasts are thought to drive regeneration, the anatomical outcomes they implement are regulated by gap junctional coupling and neural inputs from other cells (Oviedo et al., 2010); however, it is unclear precisely which elements of patterning information are intrinsic to the stem cell and which are computed by interactions with surrounding cells and the environment. If the neoblasts are indeed influenced by somatic cells while regenerating, somatic cells may be involved in determining the phenotype of the new individual.
Communication of somatic cells with neoblasts could be mediated by multiple mediators (e.g. hormones, small RNAs, ionic signaling). One common solution for coordinating the activity of cell networks is the use of gap junctions (electrical synapses that underlie plasticity in networks, both neural and non-neural) (Palacios-Prado and Bukauskas, 2009; Pereda et al., 2013). Such channels were directly shown to be required for neoblast function (Oviedo and Levin, 2007). Gap junctions are critical for cell-cell communication in embryogenesis (reviewed in Mathews and Levin, 2016) and in patterning disruptions such as cancer (Mesnil et al., 2005; Trosko, 2007; Yamasaki et al., 1999); this is well-conserved, from invertebrates through man, including the regulation of stem cell activity by gap junction-dependent signals (Jäderstad et al., 2010; Todorova et al., 2008; Wolvetang et al., 2007; Wong et al., 2008). Because they determine a cell's resting potential (by allowing electrical inputs from neighboring cells) but are themselves voltage-gated, they implement positive feedback loops that are an ideal mechanism for stabilizing physiological signals as stable memories (Levin, 2014b; Palacios-Prado and Bukauskas, 2009). It is thus no accident that brains capitalize extensively on gap junction-mediated plasticity for learning and memory in the CNS (Allen et al., 2011; Maciunas et al., 2016; Wang and Belousov, 2011; Wu et al., 2011).
Recent work has begun to reveal that patterns of resting potential differences across cell groups in vivo specify aspects of large-scale pattern formation during development and regeneration (reviewed in Levin, 2012, 2014b). Memory in the CNS is thought to involve synaptic plasticity implemented by neurotransmitters, ion channels and gap junctions (electric synapses) (Bailey and Kandel, 2008; He et al., 2014; Pereda et al., 2013). However not only neurons and muscle cells possess these proteins and the ability to communicate electrically (Bates, 2015; Funk, 2013; Sundelacruz et al., 2009). Slow changes in resting potential (not millisecond-rate spiking) regulate proliferation, differentiation, apoptosis and migration in a range of somatic and stem cells (reviewed in Blackiston et al., 2009; Funk, 2015; Sundelacruz et al., 2009). The dynamics of these bioelectric circuits implement signals that trigger or suppress regeneration (Adams et al., 2007; Jenkins et al., 1996; Tseng et al., 2010). In both embryogenesis and regeneration, endogenous spatial gradients of these potentials across tissues and anatomical axes coordinate aspects of large-sale patterning, including stem cell differentiation (Sundelacruz et al., 2008, 2013), size control (Beane et al., 2013; Perathoner et al., 2014), polarity of the left-right (Levin et al., 2002), dorso-ventral (Stern, 1987), and anterior-posterior (Beane et al., 2011) axes, and induction of organs such as eyes (Pai et al., 2012), limbs (Altizer et al., 2001), and brains (Pai et al., 2015), in a range of species from planaria to mammals.
Thus, many tissues (not only the brain) can keep a record of physiological experience in stable modifications of bioelectric circuits that impinge on form and function of the animal. Indeed, physiological circuits consisting of ion channels and electrical synapses have now been shown to underlie long-term cardiac memory, where stable changes of heart beat rhythm to a different pattern can be induced by transient physiological effects (Chakravarthy and Ghosh, 1997; Zoghi, 2004), changes of pancreas response due to patterns of physiological stimuli in type II diabetes (Goel and Mehta, 2013), and bone, where osteogenesis is induced as a long-lasting effect of use-dependent potentiation (Spencer and Genever, 2003; Turner et al., 2002). Even single cells can stably store bioelectric state (induced changes in their resting potential) as intrinsic plasticity commonly studied in neurons (Cervera et al., 2014; Law and Levin, 2015; Levin, 2014a; Williams et al., 2002). However, far more complex memory can be implemented in networks of electrically-active cells by synaptic plasticity; experience-dependent changes in the electrical connectivity (topology) of a tissue and resulting reverberating loops. In many tissues (including the brain), this is in part mediated by gap junctions; electrical synapses that are themselves voltage-sensitive, allowing physiological history to shape future cell interactions (Palacios-Prado and Bukauskas, 2012; Pereda et al., 2013).
We recently tested the ability of gap junctional communication in somatic cell networks to implement somatic memory in planaria (reviewed in Durant et al., 2016) by transiently reducing gap junctional connectivity among cells. This can be accomplished by RNAi targeting 3 distinct Innexin proteins (Oviedo et al., 2010), which resulted in a bipolar two-headed planarian; posterior wounds of middle fragments grew heads instead of tails. The same result can be achieved by a transient (2-day) inhibition of gap junction communication using a blocker such as octanol (Nogi and Levin, 2005). The benefit of this approach is that unlike RNAi, which persists in tissues for long periods of time, octanol leaves planarian tissues within 24 h (as shown by HPLC) (Oviedo et al., 2010).
Remarkably, two-headed worms derived from a brief exposure to octanol immediately after cutting, continue to regenerate as two-headed in future rounds of amputation without the presence of octanol. This may be a result of gap junction connections being stably altered through conventional synaptic plasticity, or whether gap junction connectivity is restored to a normal state after the effect is canalized into another medium (e.g. chromatin modification), or both.
The ability of a transient physiological modulator to stably change the target morphology (the shape to which planarian fragments regenerate upon damage) suggests that at least some aspect of pattern memory is encoded in physiological networks and can be re-written by life events. Related phenotypes have also been produced by altering neurotransmitter pathways (Chan et al., 2014) and voltage-mediated circuits in planaria (Beane et al., 2011; Nogi et al., 2009; Zhang et al., 2011), consistent with a conserved role for bioelectric modules exploited for adaptive, plastic control of cell behavior in the body and organism behavior in the brain (Pezzulo and Levin, 2015).
The planarian brain can form complex associations, such as learning and utilizing a set of context-specific behaviors (Best and Rubinstein, 1962; Halas et al., 1962; Thompson and McConnell, 1955; Umesono and Agata, 2009). Planaria have a true centralized brain (Nakazawa et al., 2003; Sarnat and Netsky, 1985), and brainless fragments exhibit no internally-motivated behavior or complex responses. Long-term memory in a number of vertebrate and invertebrate species has been shown to survive massive brain remodeling and regeneration (reviewed in Blackiston et al., 2015). Controversial experiments conducted with planaria in the 60s and 70s, but also modern experiments that were properly controlled and conducted using fully-automated training and tracking (Shomrat and Levin, 2013), suggest the possibility that some type of neuronally produced memories (an association between food and the haptic characteristic of the plate) can survive decapitation. Specifically, the data show that tail fragments of trained worms can retain information acquired during learning phases of the worm's life (Corning, 1966; McConnell et al., 1959; Shomrat and Levin, 2013). While the mechanisms by which information is encoded, stored, and imprinted on the newly regenerating brain remain to be understood, these data show modifications induced in adulthood can propagate via this animal's most frequent mode of reproduction.
These experiments raise the intriguing possibility that experience-dependent modifications are not limited to the head, but present throughout the animal. As the CNS is known to control not only behavior but also pattern regulation (Kiortsis and Moraitou, 1965; Mondia et al., 2011; Singer, 1952), including in planaria (Oviedo et al., 2010), fragments that inherit distinct portions of the nervous system could exhibit not only varied behavior but potentially different anatomical structure. In some species of planaria, transient changes of bioelectric connectivity, in the absence of mutation or introduction of foreign genes (i.e. despite a constant, normal genomic sequence), induce the formation of head morphology, brain shape and neoblast distribution typical to other extant species of planaria (Emmons-Bell et al., 2015). These data suggest that species-specific anatomical pattern upon regeneration is a function not only of the organism's genomic sequence but also of physiological events impinging on the body. These changes, unlike the induction of the two-head phenotype using gap junction inhibitors, are transient, again supporting the hypothesis that the effects are mediated by epigenetic mechanisms.
The provocative idea, which demands additional study, that certain memories in planaria survive decapitation, presents a useful opportunity for debate. We present a few hypothetical scenarios, not mutually exclusive, that will allow us to ask whether upon fission a planarian that is derived from the head fragment can consider the regenerated fragment that arises from its cut-off tail fragment as ‘my twin’, ‘my sibling’, ‘my child’ or ‘myself’ (Fig. 1).
Case 1: If upon fission and regeneration the two resulting fragments are identical in every aspect (genetically, epigenetically), and if experiential brain-encoded memory is erased (Rilling, 1996), then the two individuals can be considered clones or truly ‘identical twins’.
Case 2: If asymmetric fission non-homogenously establishes epigenetic differences, including in the process of brain development, so that the two planarians have different starting conditions to life, then the two individuals are ‘siblings’, not identical twins.
Case 3: If a memory is specifically acquired in the brain, and if upon beheading the worm that regenerated the tail retains the memory, while the worm which regenerated a new brain does not, then perhaps the birth of the new naïve tissue (e.g. a new brain) is the birth of a new generation. The naïve fragment is the ‘child’ in this case, and the experienced fragment is the ‘parent’.
Case 4: If some neuronally acquired memories can still be maintained in a new worm regenerated from the tail piece of the original worm (Corning, 1966; Shomrat and Levin, 2013), then the underlying mechanisms for transgenerational transmission of memories, if found, could produce two individuals that are true clones in every way, similarly to the situation in Case 1. Moreover, since in this hypothetical scenario the two fragments share their neuronally produced memories, one fragment could consider the other fragment as ‘myself’.
Diverse phyla of animals regenerate body portions after damage (Birnbaum and Sánchez Alvarado, 2008). In the phylum Cnidaria, this capability is the rule rather than the exception, with some of its members possessing remarkable regenerative capabilities. For instance, isolated medusa muscle cells can undergo transdifferentiation and regenerate an entire organism (Schmid and Alder, 1984). This attribute of cnidarians had already been documented more than two centuries ago by Abraham Trembley, who first described the regeneration of Hydra, a capability which had hitherto been supposed unique to plants and fungi (Galliot, 2012). In some cnidarians this regenerative ability has parallels to regeneration in planaria; for example, in Hydra the somatic stem cells that drive regeneration express piRNAs, similarly to planarian neoblasts (Juliano et al., 2013).
In general, even traditional reproduction can be considered an ultimate form of regeneration, where an entire organism is re-created from a single cell of the adult (the egg). In this section we wish to expand the discussion, and consider whether the questions that were raised above in regard to planaria apply to other organisms as well. We have discussed how memory could be transferred between individuals as a result of the blending of boundaries between development and inheritance in asymmetrically dividing animals. Since diverse organisms use different mechanisms to procreate and to store information, it is worthwhile to reflect on the broader definition of the terms ‘generations’ and ‘memory’, and the interaction between these processes.
When cells are grown in culture in the laboratory, the ‘generation time’ of the culture is frequently tracked and different ‘generations’ display different phenotypes, which often accumulate in later ‘generations’ (Merrill, 1998; Niida et al., 1998). In addition to amassing damage (e.g. shortening of telomeres, mutations), when cells divide, whether in a multicellular organism or in unicellular organisms, certain memories can be inherited through mitosis; daughter cells can stably maintain the memory of different cellular activities initiated in the parental cell when the cytoplasm is split in two, through different feedback mechanisms (Campos et al., 2014; Wang et al., 2013). The ability to maintain expression patterns of the parental cells in the daughter cells is a key to development and differentiation (Hobert, 2011). Not all the information is preserved; for example, DNA replication and the ensuing dilution of the histones present a challenge for preservation of chromatin marks (which epigenetic marks are maintained after S phase is still an open question in the field) (Budhavarapu et al., 2013; Lanzuolo et al., 2011; Probst et al., 2009). Histone variants are being removed off replicating DNA, and the new histones are being deposited on the newly synthesized DNA as the replication fork progresses. Which molecules or information enable, in cases when this type of memory is indeed preserved (Gaydos et al., 2014), to decorate the histones of the daughter strands with the same post-translational modifications that were present on the histones of the template DNA? This is a very active field of investigation and there are currently no definitive answers (Campos et al., 2014).
In contrast, re-establishment of DNA methylation patterns on the newly synthesized DNA is fairly well understood (the process depends on the maintenance activity of the DNA methyltransferase, DNMT1) (Kar et al., 2012). Despite the mechanistic ambiguity, it is clear that certain environmental changes can elicit responses that are memorized over cell division; maintenance of acquired properties in a bacterial and yeast population, such as fast responses to different environmental conditions or nutrients, was shown to persist over long periods of time (and thus through generations) (Lambert and Kussell, 2014).
Plants provide a striking example of evolution of organisms which lack a designated population of stem cells that will become germ cells. One of the aspects of plant cell biology that distinguishes between regeneration in plants and planaria is the ability of certain plant cells to dedifferentiate or transdifferentiate in response to environmental cues. There is no single source of cells for new tissues in plants, as apart from meristems (structures consisting of pluripotent stem cells) there are various undifferentiated cell populations in the plant that can propagate and differentiate (Aichinger et al., 2012). Additionally, certain somatic cells may transdifferentiate to grow various plant tissues (Sugimoto et al., 2011). The ability of plant cells to dedifferentiate is under tight regulation of cell-specific gene expression, in the absence of which flower meristems and even embryos may develop spontaneously from somatic tissues (Bowman et al., 1992; Horst et al., 2016; Ikeuchi et al., 2015).
The same processes that regulate the dedifferentiation of somatic cells are part of normal plant growth and development. For instance, the presence of an apical meristem inhibits the development of axillary meristems (Leyser, 2003). The absence of a nearby meristem, either caused by its removal or by the growth of the plant, will reduce this inhibition and allow the development of dormant meristems, or the development of undifferentiated cells into meristems or the formation of new meristems from dedifferentiated somatic tissue (Leyser, 2003).
Whether through a natural or artificial process, cloning can result from injury or detachment of a portion of the plant. However, formally, the definition of ‘generations’ in plants refers to the completion of a ‘life-cycle’, from embryo to adult (Bai et al., 2000; Harada et al., 2001). Since in vegetative reproduction there is no passing through an embryonic stage, does the individual which grows out of the severed part constitutes a new generation? In addition to vegetative reproduction, some plant species reproduce through the formation of plantlets on somatic tissues (Kulka, 2006). This process is defined as asexual reproduction, due to the formation of an embryo. Although the ‘progeny’ is a clone, and there is no germline involved, the embryo can mark a border between generations, due to its position in the plant's life cycle (Bai, 2015). Also in the case of clonal reproduction in plants, clones may differ depending on the fragmented tissue from which it was grown. This type of variation is termed somaclonal variation, and may be caused by genetic or epigenetic differences in the cells from which the clone develops (Wang et al., 2013). The notion that mosaicism can give rise to differences between regenerated parts is schematically described in Fig. 2.
Though it is also possible to clone plants without the mediation of an embryo using its regenerative properties, and despite the fact that this action yields two individuals, this form of cloning is not commonly referred to as asexual reproduction. Plants display a full arsenal of epigenetic mechanisms, including the ones described in relation to planaria, such as histone modifications, DNA methylation, and small RNA-induced RNAi (Dunoyer et al., 2010; Habu et al., 2001; Kaeppler et al., 2000). Moreover, plants have the ability to amplify heritable small RNAs that are used for gene silencing using RNA-dependent RNA polymerases (similarly to C. elegans nematodes) (Rechavi et al., 2011) and small RNAs can also direct DNA methylation in the nucleus. These mechanisms enable preservation of transgenerational epigenetic memory, in addition to maintenance of epigenetic memory after cell division (Castel and Martienssen, 2013). Additionally, as dedicated structures such as plasmodesmata connect different plant cells, diffusible epigenetic markers in somatic cells may affect the stem cells that regenerate, produce embryos or germ cells. It is possible that the mechanisms, which may create variability in planarian clones, could contribute to somaclonal variation. Indeed, in addition to prevalent genetic mosaicism (Gill et al., 1995), some ‘epimutations’ that originate in plant ancestors can become stable over hundreds of generations (Ong-Abdullah et al., 2015).
Since different reproduction processes in plants, as described above, do not require passage through an embryonic step (that defines which individual is the ‘parent’ and which is the ‘child’), the relationship between the two resulting individuals is somewhat ambiguous, and bears many similarities to the relationship between two regenerated planarian fragments.
In sexually reproducing animals, the ‘clear’ conceptual classification of individuals along a lineage to distinct generations is allowed due to the discrete steps of meiosis and fertilization. In mice and humans, extensive erasure of epigenetic information that originated in the parent by germline and embryo ‘reprograming’ takes place (Hajkova, 2011). Reprogramming of DNA methylations, for example, has been shown to be critical for totipotency (Messerschmidt et al., 2014; Surani, 2001). Since reprogramming entails the erasure of ancestral ‘memories’, it might be suggested that reprogramming could serve to define ‘time zero’, when the separation of the new generation from the parent takes place. However, in sexually reproducing animals, for example C. elegans, it is not clear to what degree epigenetic marks undergo ‘reprograming’ (Anava et al., 2014); C. elegans do not methylate cytosines, however some ancestral small RNAs and chromatin modification were explicitly shown to persist in the progeny, for multiple generations (Gaydos et al., 2014; Rechavi et al., 2011, 2014).
It is not yet known which type of memories/reactions can persist across generations in sexually reproducing animals (not even in organisms where this is an intensely studied question, such as C. elegans). Thus, it is not clear in what sense animal pedigrees could be considered to form an epigenetic ‘continuum’ which stretches over time, and to what extent each member in a lineage is a ‘true epigenetic individual’.
It is probable that the degree of ‘epigenetic continuity’ between generations of different animals differ, since different animals appear to diverge in the mechanisms that are at their disposal for maintaining epigenetic memory across generations. For example, no mammals are currently known to share the ability of C. elegans to amplify heritable small RNAs using RNA-dependent small RNAs (Rechavi et al., 2011). The notion of a clear-cut generation is an abstract concept, however, in sexually reproducing animals a new generation can be identified solely based on meiosis and fertilization – the definition should not be based on epigenetic resetting.
We proposed that asymmetric fission might encourage variation between the individuals that regenerate from the fragments. Here we detail experiments that could add support to this hypothesis.
Each fragment has a gene expression pattern that is specific to its morphology; however, when it is removed from an intact worm and forced to regenerate new structures, it must remodel these gene-regulatory events on top of new anatomy (i.e. a trunk fragment containing largely intestine must generate new positional information to specify head and tail regions). The incomplete reprogramming of these markers may lead to their maintenance throughout the complete animal (Thomas and Schötz, 2011). In other words, an organism that regenerates from a tail may be more ‘tail-like’ than one that regenerates from a head. This can be assessed after a single fission event by comparing the gene expression of the resulting whole organisms and those of the specific tissue of the fragment. While the continued success of regeneration over millions of fission events through the history of planaria suggests that such history or enrichment cannot accumulate indefinitely, it is possible that some limited amount of ‘recent’ history of spatial origin is kept. It will be especially interesting to identify persistent molecular or biophysical markers of anatomical (positional) history (Carlson, 1983; Chang et al., 2002) in fragments that originate in different regions of one-headed versus permanently two-headed worms, to decipher the algorithm by which blastema cells of any fragment type decide which structures to generate at each wound surface.