Linking the ‘why’ and ‘how’ of ageing: evidence for somatotropic control of long-term memory function in the pond snail Lymnaea stagnalis

Starting with gestation and ending with death, the life cycles of all species proceed along characteristic timetables with typical patterns of development, growth, maturation, reproduction and characteristic life expectancies that, according to life history theory, are the result of evolutionary optimization of how organisms work around constraints and distribute limiting resources between their various body systems and activities to maximize fitness (Stearns, 1992; Roff, 2002; Flatt and Heyland, 2011). In most species, including our own, this optimized state includes a more or less gradual functional decline during the last stages of life, i.e. most species age. Life history theory and other closely related theories such as the dynamic energy budget theory and the disposable soma theory of ageing treat ageing, either implicitly or explicitly, as a price paid for optimizing the balance of growth, maturation and reproduction at a cost with the maintenance and repair processes required for individual survival (e.g. Kirkwood, 1977, 2011; Kirkwood and Austad, 2000; Kooijman, 2010; Perlman, 2013).

Ageing is a complex, multi-faceted process that manifests itself in many different ways at all levels of biological organization. Despite substantial effort, many aspects of ageing are still not fully understood. One of those puzzles is how the ultimate causes of ageing connect to its proximate exponents and how ultimate evolutionary theories of ageing (the ‘why’ of ageing) translate to proximate mechanisms (the ‘how’ of ageing). In this study, we embarked on putting our previous findings on age-associated memory failure in an invertebrate model system, in the broader evolutionary perspective of life history theory.

Ageing affects all animal tissues and organ systems. Likewise, neurons and brains tend to deteriorate with time, even when they are not affected by disease. Much has been learned about biological ageing of the nervous system. Yet much more needs to be learned before we truly understand all its many intricacies and facets, including, for example, the issue of how neurophysiological changes occurring with age relate to the symptoms of behavioural and cognitive decline manifest in many healthy ageing metazoans, including humans. Using the pond snail Lymnaea stagnalis (Linnaeus 1758), we have gathered biochemical, physiological and behavioural evidence that point towards alterations in (membrane) lipid metabolism as a key factor in the declining neuronal excitability and associative long-term memory (LTM) function that characterize ageing in this model system (Hermann et al., 2007, 2013, 2014; Watson et al., 2012a,b). Here, we investigate how experimental interference with one of the key life history budget control systems of L. stagnalis affects LTM function.

Lymnaea stagnalis provides great opportunities for studying this type of problem: its brain provides great practical advantages in linking whole-animal behavioural (dys)functions to their exact neuronal underpinnings and experimental interference with the animals energy budget control can be carried out in a straightforward by extirpation of the so-called ‘lateral lobes’ (LLs), which are small accessory ganglia controlling the animals somatotropic/gonadotropic balance (Geraerts, 1976a,b, 1992; Roubos et al., 1980). The LLs operate by inhibiting the light green cell (LGC), clusters of neuroendocrine cells located in the cerebral ganglia that produce a set of seven insulin-like peptide homologs (ILPs) called molluscan insulin-like peptides (MIPs; Geraerts, 1976a,b, 1992; Roubos et al., 1980; Geraerts et al., 1992; Smit et al., 1998). In particular, two neuroendocrine neurons, the canopy cells (CCs) located in the LLs, are (based on immunochemical, histochemical and ultrastructural studies) considered to be ectopic specialized LGCs that are implicated in the regulation (i.e. inhibition) of the LGCs (Roubos et al., 1980; Geraerts et al., 1992; Hatakeyama et al., 2000).

ILPs are an evolutionarily ancient class of peptide hormones that play vital role in the regulation of carbohydrate metabolism, growth, cell survival and a great many other body functions in a wide range of animal species. In recent years, ILPs have been increasingly implicated in brain function. For example, in the mammalian brain, ILPs exert beneficial effects on processes such as growth and development, cell survival, adult neurogenesis, synaptic plasticity and cognition (Isgaard et al., 2007; Åberg, 2010; Fernandez and Torres-Alemán, 2012; Chambers et al., 2015). Lymnaea stagnalis appears to form no exception to this rule; several studies report that MIP genes are upregulated during LTM consolidation and that insulin/MIP signalling is necessary for LTM formation (Azami et al., 2006; Murakami, et al., 2013; Mita et al., 2014; Pirger et al., 2014; Kojima et al., 2015). ILPs could, however, also have a ‘dark’ side. They are, paradoxically, implicated with increasing regularity in cell damage and accelerated ageing in many animal model systems and tissues’ and have been linked to aggravation of ageing conditions (Zhao et al., 2004; Cohen and Dillin, 2008; Fernandez and Torres-Alemán, 2012; Gkikas et al., 2014).

Surgical removal of LLs in L. stagnalis releases the LGCs from inhibition, leading to an increase in MIP release, changes in carbohydrate homeostasis and an increased rate of body growth at a cost of reduced female reproductive output, i.e. their extirpation causes substantial shifts in the balance of two of the most important resources of the animal (Geraerts, 1976a,b; Roubos et al., 1980; Geraerts, 1992; Geraerts et al., 1992). Hence, manipulating the LLs provides a unique opportunity with which to explore relationships between the domains of life history optimization and nervous system performance.

Animals were bred and raised in the laboratory under constant and strictly controlled ambient conditions, as previously described (Hermann et al., 2007; Watson et al., 2012a) (12 h:12 h light:dark, ambient temperature 18–19°C). To maintain tight control over the snails aquatic environment, the animals were kept in artificial pond water (pH 7.4–7.6) reconstituted in-house from purified reverse osmosis (RO) water reconditioned to a conductivity of ∼450 µS cm⁻¹ by the addition of Instant Ocean salts (Aquarium Systems, Mentor, OH, USA) at 0.26 g per litre and calcium to a saturating level of ∼60 mg l⁻¹ in the form of calcium carbonate (light powder; EMD analytics, Gibbstown, NJ, USA) to the tanks. In addition, animals had continuous access to sterilized cuttlefish (Sepia officinalis) bone (two or three per tank). Together, these additions chemically recondition the RO source water to that of water within the range of that reported for the natural environments of the species (concentrations for main ion species in mg l⁻¹: Na⁺ 56; K⁺ 2; Ca²⁺ 60; Mg²⁺ 6.7; Cl⁻ 98; SO4²⁻ 14). Animals were fed ad libitum with a standard diet consisting of Romaine lettuce and Aquamax-carniverous Grower 600 trout pellets (Purina Mills, St Louis, MO, USA). For the present study, fully sexually mature snails (age 7–9 months; shell length 2.5–3.0 cm) were taken at random from an age-synchronized population displaying survival characteristics of healthy ageing populations (see Hermann et al., 2014 for further details on population survival assessment).

Both LLs were extirpated using a method modified from Scheibenstock et al. (2002). These lobes can be extirpated without destruction of any other part of the cerebral and other ganglia (see also Geraerts, 1976b). Animals were allowed to recover for at least 10 days before being tested. Untreated (no surgery) and sham-operated animals from the same populations were included as controls. The latter animals were handled in exactly the same way as described above, except that the lateral lobes were kept intact. Sham-operated and LL-extirpated animals were kept in the same tank separated by a perforated screen dividing the tank in half. For identification purposes, each animal was provided a unique code on its shell with an indelible marker. One set of animals was used to monitor survival rate and changes in total wet body mass. A separate set of animals was used to determine the effect of LL extirpation on sensory and motor control, and LTM formation.

Each LL contains a single CC with each cell having an axonal projection in the contralateral median lip nerve (van Minnen et al., 1979; Benjamin et al., 1980). To verify the success of the LL extirpation, the left or right median lip nerve was retrogradely back-labelled with nickel-lysine (for detailed procedures, see Hermann et al., 2000). After back-labelling for 24–36 h, the preparations were developed with rubeanic acid and examined for the absence or presence of CC staining and photographed. All successful backfills should show staining of at least one contralateral LGC underneath the dorsal body. Therefore, preparations without contralateral LGC staining were deemed incomplete and excluded from analysis.

To assess whether LL extirpation alters chemo-sensation and/or motor control, feeding response to sucrose (0.4% w/v; i.e. 11.7 mmol l⁻¹ final concentration in artificial pond water) were tested in LL-extirpated, sham-operated and no-surgery animals as per established protocol (Watson et al., 2012b; Hermann et al., 2013). Tests were performed using 100-ml translucent polystyrene beakers (4.5 cm diameter), filled with 80 ml of artificial pond water that allowed for full submergence of the animals. After transfer into the beakers, the snails were allowed to acclimatize for 15 min before testing commenced. Testing involved counting the number of rasps over three consecutive periods of 2 min, the second period starting with gentle administration of artificial pond water (10 ml) [which is a disturbance stimulus (DS)] and the third period starting with the administration of sucrose solution (0.4% w/v final concentration). The disturbance response was calculated by taking the difference between the number of rasps counted during the second period minus the number counted during the first period (i.e. ΔRasp disturbance=rasps after DS−rasps after pond water). The sucrose response was calculated by taking the difference between the number of rasps counted during the third period from the number counted during the second period (i.e. ΔRasp sucrose=rasps after sucrose−rasps after DS).

LTM performance was assessed 10–12 days after surgery using an established single-day, multi-trial appetitive classical conditioning protocol (Hermann et al., 2007, 2013; Watson et al., 2012b; Beaulieu et al., 2014). Snails were sampled at random and marked for identification purposes with indelible marker. Food was withheld starting 48 h prior to the first pre-training test and for the remainder of the training and post-training testing.

On day 1, prior to behavioural conditioning, snails were individually tested for their natural response to the administration of artificial pond water, the disturbance stimulus (DS) and the conditional stimulus (CS) n-amyl acetate (‘pre-training test’). Tests were performed using 100 ml translucent polystyrene beakers (4.5 cm diameter), filled with 80 ml of artificial pond water (this would provide a water column of ∼4.5 cm). After transfer into the beakers, the snails were allowed to acclimatize for 15 min before testing commenced. Testing involved counting the number of rasps over two consecutive periods of 2 min, the first period starting with gentle administration of the DS (10 ml artificial pond water) and the second period starting with the administration of the CS (10 ml n-amyl acetate in artificial pond water; 4 ppm final concentration). To facilitate observation, the test beakers were elevated on translucent plastic stands surrounded by mirrors. The n-amyl acetate rasping response was calculated by taking the difference between the number of rasps counted during the second period from the number counted during the first period (i.e. ΔRasp=rasps after CS−rasps after DS). To correct for both differences in background rasping activity and potential application artefacts, the pre-training tests were performed in duplicate at least 1 h apart, and behavioural responses were calculated as the average of both ΔRasp (i.e. ΔRasp_pre-test=average ΔRasp_pre-test1 and ΔRasp_pre-test2). After completion of a test, snails were gently rinsed with artificial pond water and returned to their home tanks. A single post-training test was performed on day 3 (22-24 h after the last training) following identical procedures described for the pre-testing above.

Snails were trained in a single-day, multi-trial forward-delay conditioning format. Sucrose dissolved in artificial pond water (to a final concentration of 0.4% w/v) served as the unconditioned stimulus (UCS) and n-amyl acetate (4 ppm final concentration) served as the conditioned stimulus (CS). To control for potential behavioural effects of fluid addition, a disturbance control in which the UCS was paired with the DS (i.e. artificial pond water) was implemented. Snails were randomly assigned to either the CS–UCS (‘conditioned’) or the CS–DS (‘non-conditioned’) group and trained ‘en masse’. Training was performed in 1 l polypropylene beakers (9 cm in diameter) containing 480 ml artificial pond water providing a water column of ∼9 cm. After transfer into the training beakers, the snails were allowed to acclimatize for 60 min. Both ‘control’ and ‘conditioned’ groups received 120 ml of the CS solution, followed 15 s later by 120 ml of the UCS (‘conditioned’ group) or 120 ml of the DS (‘control’ group). After 2 min, the beakers containing the snails were drained and gently rinsed with artificial pond water and the snails were readied for their next training trial by re-placing them in the 1 l polypropylene beakers holding 480 ml of clean artificial pond water. After 11 min and 45 s, the training procedure was repeated. Snails received a total of five training procedures on a single day. The snails were at all times fully submerged during training and testing. Care was taken to ensure that pre-training tests, training and post-training tests commenced at the same time of day for each group and always occurred in the same location. ‘Conditioned’ and ‘non-conditioned’ snails were always tested and trained concurrently.

Our training assay and protocol are direct adaptations from protocols developed by Audesirk et al. (1982) and Alexander et al. (1982, 1984). They differ slightly from related appetitive training protocols commonly used by other research groups (e.g. Kemenes et al., 2011; Marra et al., 2013) in that we fully submerged the animals in training and testing solutions, employed multi-trial rather than single-trial training protocols, and used slightly lower CS (0.0004% n-amyl acetate) and UCS (0.4% sucrose) concentrations than were previously shown to provide optimal behavioural responses under our experimental conditions (see Hermann et al., 2007; Watson et al., 2012b; Hermann et al., 2013; Beaulieu et al., 2014). The use of 0.4% sucrose as an UCS creates a safe margin with the very low UCS concentration (0.067% sucrose) previously reported to induce memory lapses in this learning model system (Marra et al., 2013).

Data were analysed using Statistica version 7.1 (StatSoft, Tulsa, OK, USA), using repeated measures ANOVA for body growth data and a multifactorial ANOVA for LTM and sucrose response data. Compliance with parametric assumptions was confirmed for each data set analysed using ANOVA using both graphical (probability plots applied to raw data and residuals) and analytical techniques (Kolmogorov–Smirnov one-sample test for normality, F-max test). Log-rank (Mantel–Cox) test for survival data and contingency test of backfill success rate were performed in Graphpad Prism version 5.0b (Graphpad Software, La Jolla, CA, USA). Throughout the text, average and data dispersion are expressed as arithmetic means and, unless indicated otherwise, standard error of the mean (s.e.m.). All graphs were made using Graphpad Prism.

Each LL contains a single CC with an axonal projection in the contralateral median lip nerve (van Minnen et al., 1979; Benjamin et al., 1980). Thus, effectiveness of LL extirpation procedures can be evaluated by retrograde nickel-lysine labelling of the contralateral median lip nerve: i.e. no soma labelling of the CC will be observed when extirpation of the LL was successful. Fig. 1A shows an example of a sham-operated (Ai) and LL-extirpated (Aii) backfilled preparation. Despite the labelling in both preparations of a contralateral LGC (arrowhead), which is indicative of a complete backfill of the median lip nerve, staining of the CC is absent in the LL-extirpated CNS (compare the labelled cell circled in Ai with the absence of a labelled cell circled in Aii). Like most other neuronal tracing techniques, retrograde labelling with nickel lysine is not perfect. Although none of the 20 LL-extirpated preparations showed labelling, a low, statistically insignificant, number of false negatives were observed in the no surgery (2 of 15) and sham-operated control preparations (5 of 19; χ²₁=31.76, P<0.0001; Fig. 1B). Thus, we are confident that LL extirpation procedures were effective.

Survival rate of animals was monitored in no surgery, sham-operated and LL-extirpated animals for 4 weeks following surgery. Significant differences in survival were found in the three cohorts (χ²₂=8.512, P=0.0142; Fig. 2A). Whereas no deaths occurred in the no-surgery group, 33% and 34% of the animals in, respectively, the sham-operated and LL-extirpated groups died before post-surgery day 9. Survival of these two groups stabilized after this time point (Fig. 2A). Multiple comparison analysis revealed significant differences in survival between both sham-operated and no-surgery snails (χ²₁=9.423, P=0.0021) and LL-extirpated and no-surgery snails (χ²₁=8.083, P=0.0045). There was no difference in survival between sham-operated and LL-extirpated snails (χ²₁=0.140, P=0.71). Thus, whereas surgery affects survival for 1–2 weeks post-surgery, LL extirpation itself does not appear to have further effects on survival within the first month post-surgery.

In view of earlier findings that LLs control body growth of juvenile sexually immature animals (Geraerts, 1976a,b; Roubos et al., 1980), we examined whether LL extirpation also affected body mass in the older sexually mature animals used here. Before surgery there was no difference in total wet body mass of sham-operated and LL-extirpated animals (Student's t-test: t=0.4532 d.f.=33, P=0.65). On average, sham-operated animals weighed 1.60±0.09 g (n=18), whereas the LL-extirpated animals weighed on average 1.55±0.06 g (n=17). Although both groups showed a steady increase in body mass after surgery, body mass increased significantly faster in the LL-extirpated animals (repeated-measures ANOVA, treatment×time, F_4,132=3.2727, P=0.014; Fig. 2B). Four weeks after surgery the LL-extirpated animals had gained significantly more body mass (21% mass gain relative to pre-surgery mass) than the sham-operated animals (11% mass gain relative to pre-surgery mass; planned comparison, F_1,33=5.707, P=0.02; Fig. 2B). Thus, we conclude that, as previously reported for sexually immature juveniles (Geraerts, 1976a,b), LL extirpation accelerates somatic growth in sexually mature animals. In congruence with the latter studies, we found that the growth-stimulating effect of LL removal does not become apparent until after 3 weeks. Together, these data support the notion that the effects of LL on body mass are mediated by altering the (neurosecretory) activity of the LGCs.

Our main objective was to determine whether LL extirpation alters appetitive LTM formation. This learning model uses sucrose as a UCS and therefore relies on the ability of an animal to detect sucrose and respond with feeding movements to its application (Hermann et al., 2007). To exclude the possibility that surgery affected UCS efficacy, we first examined whether extirpation of the LL altered the unconditioned response (UR) to sucrose application. We found that the behavioural response characteristics of the animals were not affected by surgery (Fig. 3A); UCS application induced robust, statistically indistinguishable, URs in all three test groups (factorial ANOVA F_2,91=0.6559, P=0.52), whereas DS application did not elicit a significant behavioural response in either of the groups (factorial ANOVA F_2,91=0.1341, P=0.87). These findings lead us to conclude that neither surgery alone nor LL extirpation impeded or altered sensory or motor functions underlying the learning paradigm.

Next, we assessed the effect of LL extirpation on LTM formation. No-surgery (n=59), sham-operated (n=34) and LL-extirpated (n=43) animals were subjected to either a paired CS–UCS conditioning protocol or paired CS–DS protocol to control for unconditioned behavioural responses to water disturbance. Regardless of their surgical history, none of the naive animals displayed a conditioned response (CR; ANOVA, F_2,130=2.138, P=0.122; Fig. 3B). Training induced a significant increase in CS-induced rasping in the CS–UCS group of animals but not in the CS–DS control group (ANOVA, training×treatment F_2,130=4.265, P=0.016). Furthermore, the CR of conditioned, LL-extirpated animals was potentiated relative to that of both the conditioned, sham-operated animals (planned comparison, F_1,130=4.512, P=0.035) and conditioned, no-surgery animals (planned comparison, F_1,130=5.795, P=0.017).

The results described above support the conclusion that LL extirpation enhances appetitive LTM formation. Our results have various important implications. First, considering that extirpation of the LLs enhances MIP release by releasing the LGCs from inhibition (Geraerts, 1976a,b, 1992; Roubos et al., 1980; Geraerts et al., 1992; Smit et al., 1998), they argue for the existence of a causal link between learning and memory performance, somatotropic control and ILP neuroendocrinology in L. stagnalis. As such, our findings echo a growing literature linking somatotropic systems to cognition and other brain functions in both vertebrates and invertebrates (Creyghton et al., 2004; Isgaard et al., 2007; Studzinski et al., 2015; Chambers et al., 2015). Second, as the LLs of L. stagnalis arbitrate in the division of resources between growth and reproduction, which are two of the principal items of organismal resource budgets and two of the most important life history variables, our results implicitly corroborate the existence of trade-offs between reproduction and nervous system function, and, intriguingly, suggest that LLs may be one of the key loci that set the pace of neuronal and cognitive ageing in this model system.

The idea that ILPs play important and diverse roles in the fate of the nervous system is supported by various lines of evidence implicating ILPs in aspects of neuronal plasticity and cognitive function in a wide range of invertebrate and vertebrate species, including L. stagnalis (Creyghton et al., 2004; Zhao et al., 2004; Azami et al., 2006; Isgaard et al., 2007; Fernandez and Torres-Alemán, 2012; Murakami et al., 2013; Mita et al., 2014; Pirger et al., 2014; Chambers et al., 2015; Kojima et al., 2015; Studzinksi et al., 2015; Akinola, 2016; Tanabe et al., 2017). Moreover, ILPs are also known to be important regulators of adult neurogenesis, angiogenesis and synaptogenesis, and insulin has been shown to affect synapse remodelling through downstream neurotrophic signalling cascades (Granata et al., 2007; Nelson et al., 2008; Åberg, 2010). On the flipside of the same coin, ILP signalling is frequently implicated in the nervous system's demise. For example, low levels of ILPs and/or insulin resistance are currently seen as a major factor in age-related memory impairment and neurodegenerative diseases (Williamson et al., 2012; Calvo-Ochoa and Arias, 2015; Akinola, 2016), but even normal and high levels of ILPs are increasingly implicated in cell damage and accelerated ageing in many animal model systems and tissues, including the brain (Cohen and Dillin, 2008; Fernandez and Torres-Alemán, 2012). Likewise, the idea of ILP involvement in the regulation of aspects of ageing is, by itself, not a novel one (Sadagurski and White, 2013; Altintas et al., 2016). Yet, the present study adds to what we see as a fascinating novel perspective to this field: by associating the LLs, two tiny accessory CNS lobes that arguably constitute a metabolic control centre in L. stagnalis, with the associative learning abilities of an animal, we have may have identified the embodiment of a nexus of ultimate evolutionary causes of ageing and proximate mechanisms of brain ageing in this model system.