1. Neuroscience
Download icon

A molecular mechanism underlying gustatory memory trace for an association in the insular cortex

  1. Chinnakkaruppan Adaikkan
  2. Kobi Rosenblum Is a corresponding author
  1. University of Haifa, Israel
Short Report
Cited
6
Views
1,294
Comments
0
Cite as: eLife 2015;4:e07582 doi: 10.7554/eLife.07582

Abstract

Events separated in time are associatively learned in trace conditioning, recruiting more neuronal circuits and molecular mechanisms than in delay conditioning. However, it remains unknown whether a given sensory memory trace is being maintained as a unitary item to associate. Here, we used conditioned taste aversion learning in the rat model, wherein animals associate a novel taste with visceral nausea, and demonstrate that there are two parallel memory traces of a novel taste: a short-duration robust trace, lasting approximately 3 hr, and a parallel long-duration weak one, lasting up to 8 hr, and dependent on the strong trace for its formation. Moreover, only the early robust trace is maintained by a NMDAR-dependent CaMKII- AMPAR pathway in the insular cortex. These findings suggest that a memory trace undergoes rapid modifications, and that the mechanisms underlying trace associative learning differ when items in the memory are experienced at different time points.

https://doi.org/10.7554/eLife.07582.001

eLife digest

The survival of animals, including us humans, depends on the ability to discriminate good food from bad. We would prefer eating a given taste if it did not cause any negative feelings after eating it for the first time; however, we would avoid eating that specific taste if it caused any digestive discomfort. This ability to connect sensory events that happen close in time is called associative learning.

One longstanding theory of associative learning suggests that if the neurons that are activated by a taste fire at the same time as those that control nausea, the connections between the two groups of neurons are strengthened. This helps that particular taste to become associated with the feeling of illness. Animals can also link events that are separated in time – for example, they can become averse to a food even when its ill effects are felt several hours after eating it. An important question is how a new event (such as a new food) is internally represented and maintained for a certain time so that it associates with a response (sickness) that occurs much later.

One method used to investigate associative learning is to feed rats a new food, and then later make them feel nauseous to measure how much this causes them to avoid the food in the future. The gustatory cortex is the part of the brain responsible for perceiving taste.

Chinnakkaruppan and Rosenblum now use this experimental method to investigate the molecular mechanisms in the gustatory cortex that enable the internal representation (or memory trace) of a new taste to be associated with an unwell feeling that occurs much later.

The results of the experiments show that rats will avoid food with a certain flavor if they feel unwell within eight hours of eating it. However, the response of the rats differs depending on when the rat becomes ill. Underpinning these behaviors is the formation of two parallel internal representations of the new taste: a short-term, robust trace that lasts for three hours; and a parallel, longer lasting, weaker trace that lasts for eight hours to associate the taste with its outcome. The weaker, longer-lasting memory trace only forms if the shorter, stronger trace also occurs.

Chinnakkaruppan and Rosenblum found that forming the shorter, stronger memory requires the activity of a signaling pathway in the gustatory cortex that involves biochemical molecules called NMDAR-CaMKII-GluA1. These molecules can increase the strength of signaling between neurons and are already implicated in learning and memory. The next challenge is to put this newly identified molecular mechanism within the relevant neural circuit in the gustatory cortex.

https://doi.org/10.7554/eLife.07582.002

Introduction

Association between events in time and space is a major mechanism for all animals, including humans, to learn about the world, which can potentially change their behavior in future circumstances. The most influential books concerning learning theory published early in the 20th century were Animal Intelligence (1911), wherein Thorndike proposed that the rate of learning diminishes as the interval between response and discomfort/satisfaction is increased (Thorndike, 1911), and Conditioned Reflexes (1927), wherein Pavlov highlighted that a connection is formed in the nervous system, not between a conditioned stimulus (CS) and unconditioned stimulus (US) separated in time, but between the sensory aftereffects, i.e., trace of the CS, and the US (Pavlov, 1927). Thus, in essence, these theories mention that temporal contiguity has a vital influence on associative learning.

The prevailing dogma for cellular mechanisms underlying associative learning is based on modifications of Hebb’s pioneering cell assembly theory (1949) which posits that ‘cells which fire together, wire together', and thus the firing cells epitomize the internal representations of the two sensory events related in time (Hebb, 1949; Miltner et al., 1999; Nabavi et al., 2014; Yiu et al., 2014; Johansen et al., 2014). Indeed, it has been shown that at least in certain cases, selective facilitation of densely connected neurocircuits through excitability and/or synaptic plasticity, as a result of an experience, constitutes the basis for learning and memory (Nabavi et al., 2014; Yiu et al., 2014; Johansen et al., 2014).

Most of these studies have examined the cellular and molecular mechanisms underlying trace associative learning processes using classical conditioning paradigms in which the association takes place within the timescale of msec to sec (Johansen et al., 2014; Kitamura et al., 2014; Beylin et al., 2001). However, evolution has produced other type(s) of associative learning, wherein the CS and US can be experienced several minutes to hours apart. For example, the inter-time-interval (ITI) between the CS (novel taste) and the US (visceral information) in the conditioned taste aversion (CTA) paradigm can last for hours (Garcia et al., 1955; 1985; Kalat and Rozin, 1971; 1973; Chambers, 1990; Rosenblum et al., 1993; 1997; Yamamoto et al., 1995; Hashikawa et al., 2013; Adaikkan and Rosenblum, 2012; Stern et al., 2013; Inberg et al., 2013; Chinnakkaruppan et al., 2014; Parkes et al., 2014; Sano et al., 2014). Here, we attempted to understand the mechanisms that enable the taste memory trace to be associated with visceral information after such a long time.

Results and discussion

In order to evaluate the effect of the ITI between taste and malaise in CTA learning (ITI-CTA) quantitatively and qualitatively, we conducted a behavioral experiment, in which rats were presented with a novel taste (0.1% saccharin; CS) and later were i.p. injected with lithium chloride (0.14 M LiCl; US) at increasing time points ranging between 1 hr and 20 hr (Figure 1A). In agreement with previous findings, we found that CTA was acquired after separating the taste and malaise stimuli for up to 8 hr but not 20 hr (Figure 1B, Figure 1—source data 1 and Figure 1—figure supplement 1) (Kalat and Rozin, 1971; Kalat and Rozin, 1973; Rozin and Kalat, 1971; Revusky, 1971; Gutiérrez et al., 2003; Koh et al., 2009; Chinnakkaruppan et al., 2014). Interestingly, non-parametric cluster analysis revealed that there are two different clusters among 1–8 hr ITI-CTAs: 1–3 hr (hereinafter short-trace) and 4–8 hr (hereinafter long-trace) ITI-CTAs (Figure 1B).

Figure 1 with 1 supplement see all
Temporal boundaries of taste memory trace for the association with malaise in CTA.

(A) Schematic diagram of the experimental design. CS, 0.1% saccharin and US, 0.15M LiCl. (B) Box-whisker plots showing Test1 results (memory). 1–8 hr ITI-CTA groups but not 20 hr ITI-CTA group showed significantly higher aversion index (AI; a measure of CTA memory) compared to control group. There are two different clusters among 1–8 hr ITI-CTA groups: short-trace (1–3 hr ITI-CTAs) and long-trace (4–8 hr ITI-CTAs). (C) 1–3 hr ITI-CTA groups exhibited a similar pattern of extinction and reached AI similar to control group by Test7 whereas 4–8 hr ITI-CTA groups extinguished CTA memory by Test4. Dashed line denotes the mean AI of the control group upon Test1. (D) Box-whisker plots showing test1 results following ITI-CTA conditioning with weak US (0.025M LiCl). 1 hr, 2 hr, 3 hr, 4 hr but not 5 hr ITI-CTA groups were significantly different from the control group. (E) Top: schematic representation of the CTA conditioning and reverse conditioning trials. In reverse conditioning the novel taste was presented 1 h after the LiCl injection. Bottom: reverse conditioning group did not show CTA. Data in B and D are median and quartile range between 25% to 75%. Data in C and E are mean ± SEM, *p <0.05, **p <0.01, ***p <0.001. n ≥ 5. See also Figure 1—source data 1 and Figure 1—figure supplement 1.

https://doi.org/10.7554/eLife.07582.003

Consistent with the retrieval results, long-trace (4–8 hr) CTA groups exhibited a similar faster CTA extinction pattern, whereas short-trace (1–3 hr) CTAs exhibited a slower extinction. The long-trace CTAs were not different from an unpaired control group by extinction day4, in comparison with day7 aversion index (AI) in short-trace CTAs (Figure 1C and Figure 1—source data 1). Reverse conditioning did not lead to taste aversion (Figure 1E), which indicates the importance of temporal order between taste and nausea in CTA learning. The strength of an association can be affected by ITI between CS and US or by the strength of US. We thus conducted an ITI-CTA with a weak US (0.025M LiCl) to test whether weaker CTA learning can be acquired when an association is made during the short- (1–3 hr) but not the long-trace (4–8 hr) intervals. We found that CTA was formed following the short-trace intervals, and interestingly, also following a 4 hr but not 5 hr trace interval. Together these results suggest that there is a shift in the associability of the taste memory trace at around 3–4 hr after exposure to the taste stimulus (Figure 1D).

The long ITI between CS and US in CTA is reasonable from physiological, ecological, and evolutionary perspectives. The minutes to hours process of digestion followed by absorption may dictate the animal to associate between the taste consumed and its delayed physiological consequences (for more information see reviews, Gal-Ben-Ari and Rosenblum, 2012; Kong and Singh, 2008).

It is possible that depending on the ITI there are two taste memory traces (as indicated by the behavioral readout): the first one is strong, but lasts for approximately 3 hr (short-trace), whereas the second trace is weak, but lasts for up to 8 hr (long-trace). This observation is further supported by the previous demonstration that micro-injection of the protein synthesis inhibitor, anisomycin, into the insular cortex (IC) within 3 hr but not 4 hr after the taste inhibits associative taste memory in the latent inhibition of CTA paradigm (Merhav and Rosenblum, 2008). It is therefore plausible to hypothesize that the underlying biological mechanism(s) of taste-nausea association may differ between short- and long-trace CTAs. Next, we set out to test this hypothesis.

Given ample evidence that novel taste experience impacts the phospho-proteome in the IC (see reviews, Adaikkan and Rosenblum, 2012; Gal-Ben-Ari and Rosenblum, 2012) and the idea that calcium calmodulin-dependent protein kinase II (CaMKII) has the potential to store information (Lisman, 2014 and the references therein), we aimed to test the possibility that CaMKIIα, through auto-phosphorylation (at T286; which can act as a molecular positive feedback loop), maintains the taste memory trace in the IC. Moreover, we were encouraged by the previous report that showed that novel taste consumption increases CaMKIIα expression in the IC (Belelovsky et al., 2005). Therefore, we followed up this finding and expanded it by measuring the phosphorylation/expression levels of CaMKIIα in the IC for up to 8 hr after the consumption of novel taste. We subjected IC samples of rats exposed to either water or novel taste solution (0.1% saccharin) to biochemical fractionation (Stern et al., 2013), and examined the phosphorylation and expression levels of CaMKIIα in the crude synaptosomal fraction (P2-fraction) by Western-blotting analysis (Figure 2A–C and Figure 2—figure supplement 1). We replicated the finding from the previous report (Figure 2—figure supplement 2), that 15 min after novel taste consumption, CaMKIIα expression is increased in the IC. Interestingly, we observed that the T286 phosphorylation of CaMKIIα (pT286CaMKIIα) was increased in the P2-fraction 30 min following novel taste consumption, persisting for up to 3 hr (1 and 3 hr), but not 5 or 8 hr afterwards (Figure 2D,E and Figure 2—figure supplement 1,2).

Figure 2 with 3 supplements see all
Novel taste experience induces CaMKIIα phosphorylation in the IC in an NMDAR-dependent manner.

(A) Experimental design depicting biochemical fractionation from the IC after behavioral training. (B) Representative immunoblots of marker proteins for different fractions. (C) Rats were sacrificed at the indicated time points after exposure to either water or novel taste solution. (D) Representative pT286CaMKIIα, pT305CaMKIIα, and total CaMKIIα immunoblots from P2-fractions of water (W) and novel taste (N) groups are shown. (E) Novel taste groups showed increased pT286CaMKIIα levels in the IC synaptosomal fraction at 0.5, 1, and 3 hr compared to water controls. 5 and 8 hr groups showed no difference in pT286CaMKIIα levels between water and novel taste groups. (F,G) There was no difference in (F) pT305CaMKIIα and (G) total CaMKIIα levels between water and novel taste groups at any time point. (H) Schematic representation of experimental design. Rats were injected with saline or MK801 30 min before they were exposed to water or novel taste, and 1 hr later were sacrificed and IC was extracted. (I) Novel taste group injected with saline showed increased pT286CaMKIIα compared to water group. MK801-injected groups which received novel taste did not differ from the water group but expressed significantly less phosphorylated CaMKIIα (pT286) than novel taste group injected with saline. (J,K) There was no difference in (J) pT305CaMKIIα and (K) total CaMKIIα levels between any groups. Data are mean ± SEM, *p<0.05, **p<0.01. n ≥ 6. See also Figure 2—source data 1 and Figure 2—figure supplement 13.

https://doi.org/10.7554/eLife.07582.006

There was no significant difference in T305 phosphorylation of CaMKIIα (pT305CaMKIIα; a phosphorylation site which prevents calcium/calmodulin activation of CaMKII) between water control and novel taste groups at any time point (Figure 2D–G and Figure 2—figure supplement 1,2). Intriguingly, the temporal dynamics of pT286CaMKIIα in the IC after the novel taste experience corresponds to the short-trace timescale, indicating a possible link between CaMKIIα and short-trace CTA learning.

It is possible that NMDAR activation is upstream to CaMKIIα phosphorylation, because activation of NMDAR in the IC is crucial for taste-malaise association and NMDAR regulates CaMKIIα phosphorylation (Rosenblum et al., 1997; Ferreira et al., 2002; Sanhueza et al., 2011; Halt et al., 2012; Parkes et al., 2014). Therefore, we examined if NMDAR activation is necessary for novel taste-dependent CaMKIIα activation. Indeed, novel taste experience-induced phosphorylation of T286CaMKIIα is NMDAR-dependent, since i.p. injection of the NMDAR antagonist, MK801 (0.2 mg/kg b.w), 30 min before novel taste learning precluded phosphorylation of T286CaMKIIα in the IC (Figure 2H,I and Figure 2—figure supplement 3). pT305CaMKIIα and total CaMKIIα levels did not differ among the water group injected with saline and the novel taste groups injected with either saline or MK801 (Figure 2J,K and Figure 2—figure supplement 3).

Taste memory trace can persist for an association, and at the very same time can undergo sensory information processing to form incidental taste learning (Chinnakkaruppan et al., 2014; Rosenberg et al., 2014). Therefore, the correlation between the consumption of novel taste and NMDAR-dependent T286CaMKIIα phosphorylation can serve the taste memory trace for an association, and/or incidental taste memory. In order to dissociate between these possibilities, we investigated the role of NMDAR and CaMKIIα in non-associative incidental and associative CTA learning by micro-infusing the NMDAR antagonist APV (10 μg/1 μl/hemisphere) or CaMKIIα inhibitor TatCN21 (0.3 nM/1 μl/ hemisphere (Buard et al., 2010)) bilaterally into the IC. Consistent with the literature, our data reveal that NDMAR in the IC is dispensable for incidental taste learning but necessary for CTA learning (Figure 3—figure supplement 1) (Rosenblum et al., 1997; Barki-Harrington et al., 2009; Parkes et al., 2014). We also found that CaMKIIα in the IC is dispensable for incidental taste learning but necessary for CTA learning (Figure 3—figure supplement 1,2).

Given that NMDAR-dependent T286CaMKIIα phosphorylation in the IC is required specifically for CTA, we sought to test whether NMDAR-CaMKIIα signaling in the IC maintains the taste memory trace for the association with the US by infusing the respective antagonist or inhibitor into the IC at various time points between the CS and US. We micro-injected the NMDAR antagonist into the IC 25 min from the beginning of the taste consumption in 1 hr ITI-CTA conditioning with a weak US and observed an attenuated CTA memory, consistent with a previous report in which a strong US was administered (Rosenblum et al., 1997; but also see Ferreira et al., 2002) (Figure 3—figure supplement 3A). Intriguingly, we did not observe any effect when we made a similar manipulation in the IC 4 hr after the taste consumption in 5 hr ITI-CTA conditioning with a strong US (Figure 3—figure supplement 3B; all the remaining experiments were done with a strong US). Next, we micro-injected CaMKIIα inhibitor TatCN21 or Tat control (Tatcont) into the IC 25 min from the beginning of the taste consumption in short-trace 3 hr ITI-CTA conditioning, and observed an attenuated CTA memory (Figure 3B and Figure 3—source data 1). We micro-injected TatCN21 into the IC 25 min (i.e. short-trace timescale) from the beginning of the taste consumption in the long-trace 5 h ITI-CTA conditioning, and also observed an attenuated CTA memory (Figure 3C). However, consistent with the temporal dynamics of CaMKIIα phosphorylation in the IC, CaMKIIα inhibition in the IC 4 h after the taste consumption in the long-trace 5 hr ITI-CTA conditioning had no effect on CTA memory (Figure 3D).

Figure 3 with 4 supplements see all
The requirement of CaMKIIα in the IC for associative learning of CTA is a function of time.

(A) Outline of the experimental design. (B) Infusion of CaMKIIα inhibitor TatCN21 into the IC 25 min after the taste consumption in 3 hr ITI-CTA conditioning attenuated the CTA memory. (C) Infusion of CaMKIIα inhibitor TatCN21 into the IC 25 min after the taste consumption in 5 hr ITI-CTA conditioning attenuated the CTA memory. (D) Infusion of CaMKIIα inhibitor TatCN21 into the IC 4 hr after the taste consumption in 5 hr ITI-CTA conditioning had no effect on CTA memory. (E) TatCN21 micro-infusion 2 hr after the consumption of taste in 3 hr ITI-CTA training did not affect the CTA memory. Upper panel in B, C, D, and E depict the conditioning trial. The syringe represents microinfusion. (F) A series of coronal sections from a representative rat brain, showing the cannula placement in the rostro-caudal planes (lower panel) and the corresponding coronal sections of rat brain atlas images (upper panel). Abbreviations; AI-agranular insular cortex, DI-disgranular insular cortex, GI-granular insular cortex. Data are mean ± SEM, *p <0.05. n ≥ 11. See also Figure 3—source data 1 and Figure 3—figure supplement 14.

https://doi.org/10.7554/eLife.07582.011

These results indicate that novel taste consumption induces NMDAR-dependent T286CaMKIIα phosphorylation in the IC, which is necessary for maintaining the short taste memory trace for the association with visceral information. If this is the case, then inhibiting CaMKII 2 hr after novel taste consumption in 3 hr ITI-CTA should still impede the CS-US association. To answer this, we performed a 3 hr ITI-CTA experiment in which we micro-infused CaMKIIα inhibitor TatCN21 2 hr after the taste consumption, and administered the LiCl 1 hr later. We observed a small non-significant effect (Figure 3E). It is thus more likely that T286CaMKIIα phosphorylation in the IC is necessary for the development of the taste memory trace for the association and downstream target/s play a vital role in maintaining the taste memory trace for the CS-US association.

CaMKII-dependent modulation of GluA1-containing AMPA receptors and GluN2B-containing NMDARs has been implicated in different forms of learning, memory and synaptic plasticity, and can induce changes in synaptic strength (Hayashi et al., 2000; Whitlock et al., 2006; Sanhueza et al., 2011). Therefore, we investigated whether pT286CaMKIIα modulates GluA1 and/or GluN2B phosphorylation/expression in the IC after novel taste experience. Indeed, experiencing novel taste increased total GluA1 but not pS831GluA1 or pS1303GluN2B in the synaptosomal fraction in the IC 1 hr later (Figure 4A and Figure 4—figure supplement 1). Linear regression analysis revealed a positive correlation between pT286CaMKIIα and GluA1 expression in animals sampling novel taste (Figure 4B and Figure 4—figure supplement 1).

Figure 4 with 5 supplements see all
The requirement of CaMKIIα-dependent GluA1 expression in the IC for the associative process of CTA is a function of time.

(A) 1 hr after novel taste consumption total GluA1 but not pS831GluA1 was increased in the P2-fraction. Upper panel shows the representative immunoblots. (B) pT286CaMKIIα was positively correlated with GluA1 levels in the novel taste group but not in the water group. (C) Upper panel depicts the experimental design. Novel taste-dependent increased GluA1 expression in the IC was precluded by TatCN21 microinjection into the IC. (D) CNQX microinjection into the IC 1 hr after the taste consumption in 3 hr ITI-CTA conditioning attenuated the CTA memory. (E) CNQX infusion into the IC 2 hr after the consumption of taste in 3 hr ITI-CTA training attenuated CTA memory. (F) CNQX microinjection into the IC 1 hr after taste consumption in 5 hr ITI-CTA conditioning attenuated the CTA memory. (G) CNQX application 4 hr after taste consumption in 5 hr ITI-CTA conditioning had no effect on CTA memory. Upper panels in D, E, F, and G depict the conditioning trial. Data are mean ± SEM, *p<0.05, **p<0.01. n ≥ 9. See also Figure 4—source data 1 and Figure 4—figure supplement 15.

https://doi.org/10.7554/eLife.07582.017

To examine whether pT286CaMKIIα is required for the increased GluA1 expression in the P2-fraction after the taste consumption, we micro-infused CaMKIIα inhibitor TatCN21 into the IC 30 min before novel taste exposure and measured GluA1 expression 1 hr later. TatCN21 application reduced novel taste experience-dependent GluA1 expression in the P2-fraction (Figure 4C and Figure 4—figure supplement 2). Moreover, NMDAR antagonist MK801 injection 30 min before the novel taste also reduced GluA1 expression (Figure 4—figure supplement 3), indicating that NMDAR- and CaMKIIα-dependent increased synaptic expression of GluA1 in the IC mediates short-trace taste memory for the association with the US.

If indeed GluA1 in the IC mediates the taste memory trace for the association, we hypothesized that pharmacological inhibition of AMPAR following taste experience would interfere with the association with the US. Microinjection of AMPAR antagonist, CNQX, (1 μl/hemisphere; 3 nM/μl) (Tse et al., 2011) into the IC 1 hr after the taste consumption in short-trace 3 hr ITI-CTA conditioning attenuated the CTA memory (Figure 4D). Interestingly, CNQX micro-injection into the IC 2 hr after the taste consumption in 3 hr ITI-CTA learning also attenuated CTA memory (Figure 4E). Furthermore, microinjection of CNQX into the IC 1 hr (short-trace timescale) after the taste consumption in long-trace 5 hr ITI-CTA conditioning also attenuated the CTA memory (Figure 4F). However, in accordance with the timescale of short-trace and CaMKIIα-GluA1 activation, we found that CNQX micro-injection 4 hr after novel taste consumption in 5 hr ITI-CTA had no effect on CTA (Figure 4G).

It is intriguing that the inhibition of CaMKIIα or AMPAR in the IC during short-trace timescale attenuated short-trace CTA memory. In addition, inhibition of the CaMKIIα-GluA1 pathway during short-trace timescale also attenuated long-trace CTA. However, long-trace CTA was not affected when CaMKIIα-GluA1 pathway in the IC was left intact for 3 hr. Together, these data reveal that there are two parallel taste memory traces which are subserved by different mechanisms: one that is robust but decays quickly, and another which is weak but lasts longer.

It is important to note that NMDAR-CaMKIIα-AMPAR in the IC are not necessary for incidental taste learning and memory, whereas muscarinic receptors are critical for incidental taste learning (Ferreira et al., 2002; Parkes et al., 2014). It is interesting that different molecular mechanisms take place in the same cortex, the insular cortex, to mediate associative and non-associative taste learning and memory.

We should emphasize that different pharmacological manipulation (APV, TatCN21, CNQX) in the IC when applied within 3 hr after the CS presentation only modifies but not completely erases the CTA memory. It is also noteworthy that several previous reports demonstrate that pharmacological perturbations in the IC during CTA conditioning only partially disrupt CTA memory, and that the disrupted CTA appears similar to long-trace CTA memory (Rosenblum et al., 1993; Rosenblum et al., 1997; Berman et al., 2000; Eisenberg et al., 2003; Gutiérrez et al., 2003; Barki-Harrington et al., 2009; Inberg et al., 2013; Stern et al., 2013; Parkes et al., 2014). Thus, it is possible that over time, there is a transformation from the taste memory trace dominance (i.e. short-trace) and IC dependency to multiple memory traces (i.e. during long-trace) due to the multi-channeled background experience, e,g., time, space, food digestion-dependent body physiology, etc. Such transformation, which involves multiple memory traces, for instance, episodic memory trace (how long ago the taste was consumed and under what context?), may result in a wider distribution and processing of the taste memory trace with the passage of time (Kalat and Rozin, 1971; Revusky, 1971; Kalat and Rozin, 1973; Koh et al., 2009; Chinnakkaruppan et al., 2014).

Since the initial demonstration of CTA by Garcia (1955), several studies have proposed that the reduced aversion index followed by the increased interval between the taste and malaise in CTA conditioning could be attributed to (i) the learned safety about the taste, (ii) accumulating background interference, and (iii) taste memory trace decay (Kalat and Rozin, 1971; Rozin and Kalat, 1971; Revusky, 1971; Kalat and Rozin, 1973; Gutiérrez et al., 2003; Koh et al., 2009; Chinnakkaruppan et al., 2014). First, in support of the learned safety theory, and in line with the previous reports, we found that animals exhibit higher preference to the taste hours following the first-time experience (Figure 3—figure supplement 4) (Kalat and Rozin, 1973; Gutiérrez et al., 2003). Second, with regards to background interference theory, we propose that the multi-channeled background information can be viewed as episodic components and that, for example, the hippocampus is critical to assimilate the episodic component in CTA (Chinnakkaruppan et al., 2014; Koh et al., 2009). Third, in line with the trace decay theory, our findings demonstrate that the robust short-trace fades within 3 hr in the IC but the long-trace, which is weak, lasts longer. Moreover, the non-linear decay of the short trace suggests that it dominates over the long trace, so that and when the former fades, the latter is revealed (Figure 4—figure supplement 5). We propose a multiple memory trace theory, wherein we suggest that CTA learning may engage multiple memory systems to take part in the long-trace associative process and that after experiencing the taste, the IC plays a critical role during short-trace (Figure 4—figure supplement 5).

Although multiple lines of evidence support that NMDAR-CaMKII-AMPAR signaling in the IC plays a crucial role in maintaining the short-trace for the association with the US, we do not rule out the possibility that other molecular mechanisms in the IC may participate as well in the long-trace CTA. For instance, protein acetylation, phosphorylation, and induction of proteins were observed in the IC many hours following novel taste consumption, and it is possible that these correlative molecular changes (proteins synthesis and epigenetic regulation) may contribute to the long-trace CTA (Swank and Sweatt, 2001; Yefet et al., 2006; Elkobi et al., 2008; for reviews see Gal-Ben-Ari and Rosenblum, 2012).

Conceptually, on the one hand, our short-trace CTA complies with the commonly held assumption of neuroscience theories of associative learning that convergence of CS and US information onto particular cells/circuits leads to changes in synaptic strength at the synapses mediating the CS input to those commonly activated cells/circuit and that it underlies association formation, i.e. Hebbian plasticity (NMDAR-CaMKII-AMPARs in the IC probably in co-ordination with the amygdala; Yasoshima et al., 2000; Ferreira et al., 2005; Hashikawa et al., 2013). On the other hand, the long-trace CTA challenges this assumption, and it is possible that because of the multi-channeled background information with an increasing time interval as discussed above, homeostatic plasticity may play a crucial role to encode multiple memory traces via coordinating several brain structures such as the IC, amygdala, hippocampus, prefrontal cortices in the long-trace CTA (Vitureira et al., 2012; Turrigiano, 2012; Pozo and Goda, 2010). Overall, our data suggest that the neural mechanisms of associative learning can be an assimilation of multiple memory traces when the two relevant experiences are separated in time.

Materials and methods

Subjects

The experimental subjects were rats (Rattus Norvegicus; Wistar Hola and Sprague Dawley), obtained from Harlan (Rehovot, Israel). They were maintained at the University of Haifa in a temperature controlled (22–24°C) animal core facility under a 12 hr light/12 hr dark cycle (light phase 7:00–19:00). Experiments were conducted at least 7 days after the acclimatization to the facility when the body weight of the rats was 220-–350 g. Rats were group housed (4–6 rats) in home cages with food and water ad libitum, and were individually housed before the start of the experiments. All the experiments were conducted during the light phase.

Behavioral procedures

Conditioned taste aversion

Rats were habituated to get their daily water ration (regular drinking tap water) once a day for 20 min from two pipettes, each containing 10ml of water for three days. On the conditioning (fourth) day, they were allowed to drink 0.1% sodium saccharin (Sigma-Aldrich, Rehovot, Israel; prepared in tap water) solution instead of water from similar pipettes for 20 min, and 1, 2, 3, 4, 5, 6, 7, 8, or 20 hr later from the beginning of drinking were injected with lithium chloride (LiCl; Sigma-Aldrich, Israel; 0.15M (for strong CTA) or 0.025M (for weak CTA) prepared in double distilled water; injected i.p. 2% b.w.) for 1, 2, 3, 4, 5, 6, 7, 8, or 20 hr inter-time interval (ITI)-CTA training, respectively. They were given 20 min access to water on days 5 and 6. On day 7 rats were subjected to a more sensitive multiple choice test situation in which two pipettes with 10ml each of saccharin taste solution and two with 10ml each of tap water were presented. The order of the pipettes was counter-balanced and the volume of fluid consumed from each pipette was recorded. The rats were tested again similarly in 24 hr intervals to study memory extinction. The behavioral data are expressed in terms of aversion index, defined as the volume of water consumed divided by the total fluid consumed (water [ml]/water [ml] + taste [ml]).

Reverse conditioning

Singly housed rats underwent a 3 day water restriction training session in which once a day for 20 min they were offered 20 ml of water from two pipettes, each containing 10 ml. On the conditioning day (fourth day), rats received an i.p injection of LiCl (0.14 M; 2% b.w.) and 1 hr later were offered 20 ml of saccharin for 20 min from 2 pipettes each containing 10 ml. 72 hr post-conditioning the rats were tested similarly as mentioned above and the behavioral data are expressed in terms of aversion Index.

Incidental taste learning

Singly housed rats underwent a 3 day water restriction training session in which once a day for 20 min they were offered 20 ml of water from two pipettes, each containing 10 ml. On the fourth day, the rats received saccharin solution from 2 pipettes 10 ml each for 20 min. The rats were given 20 min access to water on days 5 and 6 and were tested in a multiple choice test involving two pipettes of water and two of saccharin on day7. The order of the pipettes was counter-balanced and the volume of fluid consumed from each pipette was recorded. The rats were tested again similarly in 24 hr intervals to study the attenuation of taste neophobia. The behavioral data are expressed in terms of aversion index.

Taste learning and immunoblot analysis

Taste learning and tissue preparation: rats were trained to drink from pipettes for 20 min per day for three days, and on the fourth day they received either water (control) or novel taste (saccharin) for 20 min. They were sacrificed 0.25, 0.5, 1, 3, 5, or 8 hr later. Brains were quickly excised and snap frozen in liquid nitrogen and transferred to -80°C. Insular cortex was extracted from cryostat sections at -20°C. Sub-cellular fractionation was performed according to (Stern et al., 2013). Briefly, brain tissues were homogenized (H) in ice-cold homogenization buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM EGTA, 320 mM sucrose (all from Sigma-Aldrich, Rehovot, Israel), 1X protease inhibitor mixture (Sigma-Aldrich Rehovot, Israel or Thermo Scientific, Rockford, USA); and 1X phosphatase inhibitor mixture (Sigma-Aldrich or Thermo Scientific). The homogenates were sonicated, kept on ice for 20 min and then centrifuged at 1000 g for 8 min at 4°C to isolate nuclei and large debris (P1). The supernatant (S1) was centrifuged at 12,000 g for 30 min at 4°C in a conventional Eppendorf centrifuge to obtain a crude synaptosomal fraction (P2), and subsequently lysed hypo-osmotically (7 mM HEPES buffer, pH 7.5) and centrifuged at 25,000 g in a Beckman Coulter ultracentrifuge at 4°C for 23 min, to pellet a synaptosomal membrane fraction (LP1). The resulting supernatant (LS1) was centrifuged at 165,000 g for 2 hr at 4°C to obtain a synaptic vesicle enriched fraction (LP2). The supernatant (S2) obtained from the fraction P2 was centrifuged at 165,000 g for 2 hr to obtain a cytosolic fraction (S3) and a light membrane fraction (pellet; P3). Protein quantity was determined with the BCA Protein Assay Kit (GE Healthcare).

Protein Samples were prepared in SDS sample buffer, subjected to 7.5% or 4–20% gradient gel (Bio-Rad pre-cast gels) SDS-PAGE (electrophoresed on Bio-Rad PAGE apparatus) and Western blot analysis. Each sample was loaded with the same amount of total protein (7–10 μg; according to antibody linearity). After transfer to a 0.2 μm pore size nitrocellulose membrane, the blots were blocked with 4% bovine serum albumin (BSA) in tris-buffered saline plus 0.5% tween-20 (TBST) at room temperature for 1 hr. They were then incubated overnight with the suitable primary antibodies: CaMKIIα (1:25,000; Santa Cruz Biotechnology, SCBT), pT286CaMKIIα (1:1000; SCBT), pT305CaMKIIα (1:1000; Novus biological), GluN2B (1:1000; SCBT), GluA1 (1:1000; Abcam or SCBT), pS831GluA1 (1:5000; Epitomics), PSD-95 (1:1000; SCBT), HMGB1 (1:10,000; Abcam), eIF2α (1:1000; Cell Signaling Technology), pY1472GluN2B (1:1000; a gift from Prof. Nakazawa), pS1303GluN2B (1:10,000; Abcam), β-actin (1:3000; SCBT) and β-tubulin (1:30,000; Sigma). The blots were then subjected to three 5 min washing steps in TBST, after which they were incubated with the corresponding HRP-conjugated secondary antibodies: goat anti-rabbit (IgG), goat anti-mouse (IgG) or rabbit anti-goat (IgG) (1:10,000; Millipore Bioscience Research Reagents) for 1 hr at room temperature followed by three 10 min washing steps with TBST. Immunodetection was performed with the enhanced-chemiluminescence EZ-ECL kit (Biological Industries, Israel). The immunoblots were quantified with a CCD camera and Quantity One software (Bio-Rad). Each immunoblot was measured relative to the background and normalized to the endogenous controls (β-tubulin). Phosphorylation levels were calculated as the ratio between the readings from the antibody directed against the phosphoproteins and those from the antibody directed against the phosphorylation state-independent forms of the proteins.

Pharmacology

Surgery

The rats were anesthetized with equithesin (2.12% [w/v] MgSO4, 10% [v/v] ethanol, 39.1% [v/v] 1,2-propranolol, 0.98% [w/v] sodium pentobarbital, and 4.2% [w/v] chloral hydrate) at 0.3 ml per 100 g body weight (Stern et al., 2013; Merhav and Rosenblum, 2008). They were then restrained in a stereotactic apparatus (Stoelting) and implanted bilaterally with 10 ± 0.02 mm, 23gauge stainless steel guide cannulae above the IC. Coordinates with reference to Bregma were: A/P = +1.2 mm, L/M = ± 5.3 mm, D/V = -5.2 mm. The cannulae were fixed in position with acrylic dental cement and secured with two skull screws. Animals were allowed 5–7 days to recover from the surgery before undergoing the experimental manipulations.

Microinjection

Rats were habituated and the quality of the guiding cannula was checked 24 hr before the microinjections. Rats were brought to the micro-injection room 5–10 min before the microinjection and after careful preparation a 28-gauge, 11.2 ± 0.02 mm long injection cannula, extending 1.2 mm from the tip of the guide cannula, was carefully inserted. The injection cannula was connected via PE20 tubing (backfilled with saline) to a Hamilton micro-syringe, driven by a microinjection pump (Harvard) that operated at a rate of 1 μl/min for 1 min. Following injection, the injection cannula was left in place for an additional 1 min before withdrawal to minimize dragging of injected liquid back along the injection track.

Drugs

Stock TatCN21 and Tatcontrol peptides (a kind gift from Dr. Ulrich Bayer, Department of Pharmacology, University of Colorado School of Medicine, Aurora, Colorado, USA) were prepared in double distilled water to a concentration of 5 mM, aliquoted, and stored at -20°C. Before the microinjection TatCN21 and Tatcont were further diluted in saline to a final concentration of 0.3 nM/μl and injected. CNQX disodium salt (6-cyano-7-nitroquinoxaline-2,3-dione disodium, Tocris, Bristol, UK) was prepared in saline to a concentration of 3 mM, aliquoted and stored in -20°C. CNQX was freshly prepared for most of the experiments or was used within 2 weeks of preparation. Before micro-injection CNQX was thawed at 40°C and injected. APV (2R-amino-5-phosphonovaleric acid, Sigma-Aldrich) was prepared in saline to a final concentration of 10 μg/μl, aliquoted, and stored at -20°C. Before microinjection APV was thawed at room temperature and injected.

Histology

At the end of the behavioral experiments, the rats were sacrificed. The brains were removed, frozen with dry ice and kept at -80°C. Randomly chosen rats were micro-injected into the IC with 0.5 or 1 μl/min cresyl violet 30 min before sacrifice. Coronal 50-μm sections were cut in a cryostat, stained with cresyl violet, and examined with a computerized Olympus microscope camera to verify cannula placement.

Statistical analysis

All grouped data are presented as mean ± SEM. Comparisons between data of two independent groups were analyzed by unpaired Student's t test and the differences between the variances of groups were corrected following Levene’s test for equality of variances. Multiple group comparisons were assessed using one way analysis of variance (ANOVA) and repeated-measures ANOVA. Follow-up analyses were conducted using Fisher’s least significant difference, Bonferoni, and independent sample t-tests, when significant main effects or interactions were detected. Non-parametric Kruskal-Wallis test, Friedman's test, and Mann Whitney U test were conducted when the data were non-normally distributed. Pearson’s correlation was conducted to analyze the association between GluA1 and pT286CaMKIIα. All of the comparisons were conducted using two-tailed tests of significance. The null hypothesis was rejected at the p <0.05 level. Data analysis was performed using SPSS-version 19.

References

  1. 1
  2. 2
  3. 3
  4. 4
    The role of identified neurotransmitter systems in the response of insular cortex to unfamiliar taste: activation of ERK1-2 and formation of a memory trace
    1. DE Berman
    2. S Hazvi
    3. V Neduva
    4. Y Dudai
    (2000)
    The Journal of Neuroscience  20:7017–7023.
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
    Conditioned aversion to saccharin resulting from exposure to gamma radiation
    1. J Garcia
    2. DJ Kimeldorf
    3. RA Koelling
    (1955)
    Science 122:157–158.
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    Organization of Behavior
    1. DO Hebb
    (1949)
    New York: Wiley.
  21. 21
  22. 22
  23. 23
    Role of interference in taste-aversion learning
    1. JW Kalat
    2. P Rozin
    (1971)
    Journal of Comparative and Physiological Psychology 77:53–58.
    https://doi.org/10.1037/h0031585
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
    Conditioned reflexes: an investigation of the physiological activity of the cerebral cortex
    1. IP Pavlov
    (1927)
    Oxford, UK: Oxford University Press.
  34. 34
  35. 35
  36. 36
  37. 37
    NMDA receptor and the tyrosine phosphorylation of its 2B subunit in taste learning in the rat insular cortex
    1. K Rosenblum
    2. DE Berman
    3. S Hazvi
    4. R Lamprecht
    5. Y Dudai
    (1997)
    The Journal of Neuroscience 17:5129–5135.
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    Increased histone acetyltransferase and lysine acetyltransferase activity and biphasic activation of the ERK/RSK cascade in insular cortex during novel taste learning
    1. MW Swank
    2. JD Sweatt
    (2001)
    The Journal of Neuroscience 21:3383–3391.
  44. 44
    Animal intelligence
    1. EL Thorndike
    (1911)
    New York, NY: Macmillan.
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52

Decision letter

  1. Christian Rosenmund
    Reviewing Editor; Charité, Universitätsmedizin Berlin, Germany

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for choosing to send your work entitled "A Molecular Mechanism Underlying Gustatory Memory Trace for an Association in the Insular Cortex" for consideration at eLife. Your full submission has been evaluated by a Senior Editor and two peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the decision was reached after discussions between the reviewers. Based on our discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

After careful discussion among the reviewers, we found that your paper contained interesting findings, but the evidence for some key conclusions is too thin and requires a lot more work. We also felt that the rational for the two parallel memory traces are not clear. We therefore came to the conclusion that the manuscript is not suitable for eLife at this time. We hope that the comments of the reviewers are helpful in further improving the manuscript.

Reviewer #1: The authors investigate a special form of memory, conditioned taste aversion (CTA) which can be often learned in a single trial. They first describe the temporal dependence of convergence of CS and US information for successful establishment of CTA and discover two components with different dwell time and strength, the first lasting app. 3 hours, with robust retentions, while US applied later than 3 hours and until 8 hours showed significant less learning. Subsequent combination of behavioral testing with biochemical analysis showed that after novel taste experience, CaMK2 phosphorylation is increased for up to 3 hours that CaMKII activity is required in IC and that GluA1 receptor number increases.

A major claim of the authors is that the drop in learning efficacy after 3 hours signifies the presence of two parallel mechanisms of CTA acquisition, a claim that is also supported by a previous study from the same group (Merhav & Kornblum, 2008) that shows that protein synthesis inhibitors have a stronger impact on CTA shorter than 3 hours. However, this claim receives overall weak support, as retention of the CS signal may simply be decaying nonlinear with time; or alternatively, the underlying processes linked to novel taste acquisition and induction of CTA crosses below a threshold for efficient learning acquisition after app. 3 hours.

The authors state that short-memory trace is required to generate the weak memory trace which lasts longer, yet the authors claim that the two memory traces are parallel. This is confusing, as when one trace requires the other one, they are by definition not parallel processes.

There are also several issues about the experimental design, in particular appropriate control experiments. For example, the authors show that MK-801 pre injection obliterated the CS impact, and the authors claim this to be a sign that NMDA receptors are required for CS induced CTA. Unfortunately, the result does not allow to distinguish whether NMDA receptors are required for learning or whether it is required for sensory perception. MK-801 as a general blocker of NMDA receptor function may affect many other aspects of CS learning including perception of taste, general sensory processing. Similar concerns relate to manipulations used in Figure 4 using CNQX.

Reviewer #2: In this paper, Chinnakkaruppan & Rosenblum used behavioral, biochemical and pharmacological approaches to characterize two different, and potentially parallel, taste memory traces: a short-trace taste conditioning (1-3 h between taste and malaise) leading to strong conditioned taste aversion (CTA) and a long-trace taste conditioning (4-8 h between taste and malaise) leading to weak CTA. In particular, the authors provide an original demonstration that the short-trace CTA critically dependent on CaMKII-GluA1 pathway in the insular cortex whereas the long-trace CTA seems to be partially independent of these processes. The authors then propose that there are two parallel taste memory traces, a stronger one dependent on insular cortex and another weaker one independent of the insular cortex.

In my opinion, the rationale for the study is clearly presented, the experiments are well designed, and the data are appropriately analyzed. I have some major comments meant to encourage the authors to clarify their interpretations of the two parallel memory traces. If these concerns can be addressed the work could yield an important paper.

I encourage the authors to elaborate on the existence of the two parallel memory traces. In particular I suggest several points:

1) Short-trace: in order to clearly demonstrate that the 1-3 h taste trace is dependent on CaMKII-GluA1 pathway in insular cortex, the authors should evaluate the impact of intra-insular infusion of TatCN21 and CNQX 2h after saccharin consumption and 1h before LiCl (ITI 3h). If their assumption is right these treatments should still impair 3h ITI-CTA.

2) Long-trace: the authors clearly demonstrate that the long-trace (> 3h ITI) is not dependent on CaMKII-GluA1 pathway in insular cortex. However, as they have previously shown that novel taste is able to enhance protein expression (C/EBPbeta) in insular cortex long after its consumption (i.e. 18h, Merhav et al. 2006), they should consider that the long-trace CTA could still depend on molecular mechanism in insular cortex (but not on CaMKII-GluA1 pathway).

3) Long-trace: the authors propose that the taste memory trace results in a wider distribution with the passage of time and that there is competition across multiples memory systems. Therefore they should propose a network of brain areas able to sustain the long-trace different from the network sustaining short-trace.

[Editors’ note: the authors appealed against the decision and were later encouraged to resubmit.]

Thank you for choosing to send your work entitled "A Molecular Mechanism Underlying Gustatory Memory Trace for an Association in the Insular Cortex" for consideration at eLife. Your submission has now been considered by a Senior editor, the original Reviewing editor, and an additional reviewer, and we are prepared to consider a revised submission with no guarantees of acceptance. The third reviewer's full comments are included below. Before you submit your revised paper please include a thorough explanation of the model and provide additional data.

Reviewer #3:

In this article, Chinnakkaruppan & Rosenblum raise an interesting question regarding the nature of memory trace formation following conditioned taste aversion, a special form of single trial associative learning where rats can acquire a memory for a specific taste (CS) associated with malaise (US) up to several hours following the presentation of the CS. They suggest the possible existence of two parallel taste memory traces: one which is strong but limited in time and lasts up to 3 hours and a second one weaker but longer in duration and lasts up to 8 hours from CS presentation. Furthermore, by applying biochemical and pharmacological techniques they suggest the involvement of specific molecular mechanism mediated by CamKII-GluA1 pathway that underlies the strong and shorter memory type but not of weak and longer memory type. Importantly, their results help to explain previous results from the same and other groups regarding the partial disruption of CTA memory by some pharmacological interventions. The possibility of parallel memory traces, which are dependent on separate molecular mechanisms, is exciting. The presented data is highly interesting with a clear potential to make it into eLife.

However, I have several concerns regarding the conclusions of the authors about the two parallel memory traces, which should be addressed before the article can be accepted for publication.

1) The authors should at the very least hypothesize regarding the possible molecular mechanisms that might underlie the long but weak memory trace formation (e.g. NMDAR dependent), or even better test their hypothesis by applying the relevant pharmacological intervention.

2) Related to this issue, it would be important to examine the effect of intra-insular infusion of MK-801 at a time interval longer than 3 hours to examine the possible effect of NMDAR on the weak memory trace.

3) Although the authors nicely demonstrate the temporal effects of the CTA (CS + US) acquisition traces, there is no quantification of the impact of the novel taste (CS) and malaise (US) on the formation of these memory traces. The authors could apply a weak CTA learning by reducing the LiCl concentration to address this question. Alternatively, they could induce weak learning by using latent inhibition, where the CS is familiarized before association. They should test: (a) whether this weaker learning can be acquired when association is made during the short (1-3 hrs) interval and long (4-8 hrs) interval and (b) is it sensitive to the CaMKII-GluA1 proposed mechanism or perhaps to a NMDAR dependent one.

https://doi.org/10.7554/eLife.07582.025

Author response

Reviewer 2 was extremely positive towards our manuscript and made some excellent suggestions on experimental and editorial additions that we are happy to comply with. Given this, we were surprised that the manuscript was rejected based on the opinion of Reviewer 1. As detailed in our point-by-point response below, data which address most, if not all, of the concerns that Reviewer 1 raised was already present in the supplemental data portion of the submitted paper. We apologize if this was not made clear in our initial submission and are happy to address these semantic shortcomings in the revised version.

Reviewer #1:

[…] A major claim of the authors is that the drop in learning efficacy after 3 hours signifies the presence of two parallel mechanisms of CTA acquisition, a claim that is also supported by a previous study from the same group (Merhav & Kornblum, 2008) that shows that protein synthesis inhibitors have a stronger impact on CTA shorter than 3 hours. However, this claim receives overall weak support, as retention of the CS signal may simply be decaying nonlinear with time; or alternatively, the underlying processes linked to novel taste acquisition and induction of CTA crosses below a threshold for efficient learning acquisition after app. 3 hours. We would like to emphasize that three different lines of evidence (behavioural, Figure 1; biochemical, Figures 2 and 4; and pharmacological, Figures 3 and 4) support the possibility of two parallel taste memory traces.

The reviewer commented that ‘retention of the CS signal may simply be decaying nonlinear with time’. It is clearly a possibility; however, our experiments and the respective statistical analyses did not support this possibility, but instead support the parallel memory traces hypothesis.

First, non-parametric one way ANOVA (Kruskal-Wallis) and non-parametric repeated measures ANOVA (Friedman’s) revealed that the aversion index measure following 1h, 2h, and 3h ITI-CTAs significantly differs from that of 4h-8h ITI-CTAs (Figure 1B, C and Figure 1—source data 1). The follow-up unbiased cluster analysis revealed that 1h-8h ITI-CTAs fall into two different clusters; 1h,2h, and 3h ITI-CTAs as small cluster, and 4h-8h ITI-CTAs as large cluster and they are statistically different from each other (Mann Whitney U test; Figure 1—source data 1). Moreover, as we have presented in the revised manuscript (Figure 1D), the associability of the CS with the weak US completely drops at around 4h-5h.

Secondly, in striking opposite to the reviewer’s comment, in Figure 3—figure supplement 4 we have demonstrated (consistent with the previous reports, Kalat and Rozin, 1973; Gutierrez et al., 2003) that the retention of the CS signal gets stronger with time.

The reviewer again presented an alternative comment ‘the underlying processes linked to novel taste acquisition and induction of CTA crosses below a threshold for efficient learning acquisition after app. 3 hours’. This is counter-intuitive to what we have presented in the manuscript: that the underlying process (NMDAR-CaMKII-GluA1 pathway) for formation of CTA with upto a 3h ITI is different from those who have an ITI >3h (4h-8h ITI CTA).

The authors state that short-memory trace is required to generate the weak memory trace which lasts longer, yet the authors claim that the two memory traces are parallel. This is confusing, as when one trace requires the other one, they are by definition not parallel processes.

We see the point of the reviewer. We would like to clarify this point with a graphical representation (Author response image 1).

Author response image 1

(A) Immediately after ingesting a novel taste a robust taste memory trace is generated and it lasts for about 3h. (B) concurrently, the weak taste memory trace is also generated and it lasts longer, for about 8h. (C) As presented in the manuscript that “inhibition of the CaMKII-GluA1 pathway during short-trace timescale attenuated long-trace CTA, however, long-trace CTA is not affected when CaMKII-GluA1 pathway in the IC was left intact for 3h” and that we suggest that there are two parallel taste memory traces; one that is robust but decays quickly and the second which is weak but lasts longer.

https://doi.org/10.7554/eLife.07582.024

Furthermore, as we mentioned in the submitted manuscript “that several reports including ours demonstrate that pharmacological perturbations (before the acquisition or during the short-trace timescale) in the IC during CTA conditioning only partially disrupt CTA memory, and that the disrupted CTA appears similar to long-trace CTA memory (Rosenblum et al., 1993; Rosenblum et al., 1997; Berman et al., 2000; Eisenberg et al., 2003; Gutiérrez et al., 2003; Barki-Harrington et al., 2009; Inberg et al., 2013; Stern et al., 2013; Parkes et al., 2014)”. Therefore, we reason that the long (weak)- taste memory trace overlaps with the short-trace (CaMKII-GluA1 in the insular cortex) during short-trace timescale.

As mentioned by reviewer 2 ‘they should propose a network of brain areas able to sustain the long-trace different from the network sustaining short-trace’, it is possible that amygdala together with the PFC, or the brain-stem gustatory circuit such as parabrachial nucleus (PBN) and nucleus of solitary tract (NTS) could sustain the long-taste memory trace (Parabuchi and Netser 2014; Carter et al., 2015). These possibilities can be discussed in the updated/revised version of the manuscript.

There are also several issues about the experimental design, in particular appropriate control experiments. For example, the authors show that MK-801 pre injection obliterated the CS impact, and the authors claim this to be a sign that NMDA receptors are required for CS induced CTA. Unfortunately, the result does not allow to distinguish whether NMDA receptors are required for learning or whether it is required for sensory perception. MK-801 as a general blocker of NMDA receptor function may affect many other aspects of CS learning including perception of taste, general sensory processing. Similar concerns relate to manipulations used in Figure 4 using CNQX.

We believe that these comments are not valid. It is well demonstrated that NMDAR in the insular cortex is necessary for the associative CTA learning but dispensable for taste perception and incidental taste learning (Rosenblum et al., 1997; Gutierrez et al., 2003; Parkes et al., 2014). Moreover, we replicated these findings and they are included in the manuscript (Figure 3—figure supplement 1).

We have also observed that CNQX micro-injection did not have any effect on the taste perception and incidental taste learning (Figure 4—figure supplement 4).

Reviewer #2:

[…] In my opinion, the rationale for the study is clearly presented, the experiments are well designed, and the data are appropriately analyzed. I have some major comments meant to encourage the authors to clarify their interpretations of the two parallel memory traces. If these concerns can be addressed the work could yield an important paper.

We thank the reviewer for the positive comments and encouragement.

I encourage the authors to elaborate on the existence of the two parallel memory traces. In particular I suggest several points:

1) Short-trace: in order to clearly demonstrate that the 1-3 h taste trace is dependent on CaMKII-GluA1 pathway in insular cortex, the authors should evaluate the impact of intra-insular infusion of TatCN21 and CNQX 2h after saccharin consumption and 1h before LiCl (ITI 3h). If their assumption is right these treatments should still impair 3h ITI-CTA.

We have addressed this constructive comment in the revised manuscript.

As suggested by the reviewer, we have performed two 3h ITI-CTA experiments in which we micro-infused CaMKII inhibitor TatCN21 or AMPAR antagonist CNQX 2h after the taste consumption and administered the LiCl 1h later. Indeed, when we performed this experiment with CNQX, we observed a significant effect of CNQX on 3h ITI-CTA memory (Figure 4E). However, we observed a small non-significant trend when we used the TatCN21.

We would like to reiterate three main points: 1) CaMKII-dependent GluA1 induction occurs as early as 1h after the consumption of novel taste. 2) There is a persistent activation of CaMKII for upto 3h after the consumption of a novel taste. 3) The above two new pharmacological experiments, as well as our previous results, demonstrate differential time effects between CaMKII and GluA1. Taken together, we suggest that it is the downstream of CaMKII (i.e. GluA1) that mediates the trace longer than 1h for the association with the US. This interpretation is discussed in the revised manuscript (Results and Discussion).

2) Long-trace: the authors clearly demonstrate that the long-trace (> 3h ITI) is not dependent on CaMKII-GluA1 pathway in insular cortex. However, as they have previously shown that novel taste is able to enhance protein expression (C/EBPbeta) in insular cortex long after its consumption (i.e. 18h, Merhav et al. 2006), they should consider that the long-trace CTA could still depend on molecular mechanism in insular cortex (but not on CaMKII-GluA1 pathway).

We agree with the reviewer on this point and presented a detailed explanation in the revised manuscript (Results and Discussion, eighteenth paragraph).

3) Long-trace: the authors propose that the taste memory trace results in a wider distribution with the passage of time and that there is competition across multiples memory systems. Therefore they should propose a network of brain areas able to sustain the long-trace different from the network sustaining short-trace.

In the last paragraph of the Results and Discussion, we elaborated on the neuronal network that might underlie long-trace CTA.

[Editors’ note: the authors appealed against the decision and were later encouraged to resubmit.]

Thank you for choosing to send your work entitled "A Molecular Mechanism Underlying Gustatory Memory Trace for an Association in the Insular Cortex" for consideration at eLife. Your submission has now been considered by a Senior editor, the original Reviewing editor, and an additional reviewer, and we are prepared to consider a revised submission with no guarantees of acceptance. The third reviewer's full comments are included below. Before you submit your revised paper please include a thorough explanation of the model and provide additional data.

In addition to the detailed point-by-point response below, we would like to highlight several key additions and changes we have made that we feel would result in a successful resubmission:

As demanded by Reviewer #1 we have expanded our analyses and discussion of the role of NMDAR in the insular cortex (IC) in taste perception (presented in Figure 3—figure supplement 1). We also provide a result (Figure 4—figure supplement 4) that further supports that AMPAR in the IC is dispensable for taste perception. In addition, we provide detailed explanation of the parallel taste memory trace model (Figure 4—figure supplement 5).

As requested by Reviewer #2 we have performed an additional experiments with a 3h inter-time interval of conditioned taste aversion (ITI-CTA) using CaMKII inhibitor TatCN21 (now Figure 3E) and AMPA/Kainate receptor antagonist CNQX (now Figure 4E).

As requested by Reviewer #3 we have performed an additional standard ITI-CTA experiment with a weak unconditioned stimulus (US) to provide a greater insight into the shift in the associability of the taste memory trace that happens at around 4h after the consumption of a novel taste (Figure 1D). In addition, we have conducted two pharmacological experiments in which we micro-infused an NMDAR antagonist into the IC 30min or 4h after taste consumption in a 1h ITI-CTA with a weak US or a 5h ITI-CTA with a strong US, respectively (Figure 3—figure supplement 3).

At the request of all the reviewers we have substantially expanded our discussion concerning the molecular and circuit mechanisms that might underlie long-trace memory and parallel memory trace model (Figure 4—figure supplement 5).

Reviewer #1:

A major claim of the authors is that the drop in learning efficacy after 3 hours signifies the presence of two parallel mechanisms of CTA acquisition, a claim that is also supported by a previous study from the same group (Merhav & Kornblum, 2008) that shows that protein synthesis inhibitors have a stronger impact on CTA shorter than 3 hours. However, this claim receives overall weak support, as retention of the CS signal may simply be decaying nonlinear with time; or alternatively, the underlying processes linked to novel taste acquisition and induction of CTA crosses below a threshold for efficient learning acquisition after app. 3 hours.

We would like to expand upon our previous responses to Reviewer #1. We would like to emphasize that three different lines of evidence (behavioural, Figure 1; biochemical, Figures 2 and 4; and pharmacological, Figures 3 and 4) support the possibility of two parallel taste memory traces.

The reviewer commented that ‘retention of the CS signal may simply be decaying nonlinear with time’. It is clearly a possibility; however, our experiments and the respective statistical analyses did not support this possibility, but instead support the parallel memory traces hypothesis.

First, non-parametric one way ANOVA (Kruskal-Wallis) and non-parametric repeated measures ANOVA (Friedman’s) revealed that the aversion index measure following 1h, 2h, and 3h ITI-CTAs significantly differs from that of 4h-8h ITI-CTAs (Figure 1B, 1C, and Figure 1—source data 1). The follow-up unbiased cluster analysis revealed that 1h-8h ITI-CTAs fall into two different clusters; 1h,2h, and 3h ITI-CTAs as small cluster, and 4h-8h ITI-CTAs as large cluster and they are statistically different from each other (Mann Whitney U test; Figure 1—source data 1). Moreover, as we have presented in the revised manuscript (Figure 1D), the associability of the CS with the weak US completely drops at around 4h-5h.

Secondly, in striking opposite to the reviewer’s comment, in Figure 3—figure supplement 4 we have demonstrated (consistent with the previous reports, Kalat and Rozin, 1973; Gutierrez et al., 2003) that the retention of the CS signal gets stronger with time.

The reviewer again presented an alternative comment ‘the underlying processes linked to novel taste acquisition and induction of CTA crosses below a threshold for efficient learning acquisition after app. 3 hours’. This is counter-intuitive to what we have presented in the manuscript: that the underlying process (NMDAR-CaMKII-GluA1 pathway) for formation of CTA with upto a 3h ITI is different from those who have an ITI >3h (4h-8h ITI CTA).

The authors state that short-memory trace is required to generate the weak memory trace which lasts longer, yet the authors claim that the two memory traces are parallel. This is confusing, as when one trace requires the other one, they are by definition not parallel processes.

We see the point of the reviewer. We would like to clarify this point with the following graphical representation (please see Figure 4—figure supplement 5).

There are also several issues about the experimental design, in particular appropriate control experiments. For example, the authors show that MK-801 pre injection obliterated the CS impact, and the authors claim this to be a sign that NMDA receptors are required for CS induced CTA. Unfortunately, the result does not allow to distinguish whether NMDA receptors are required for learning or whether it is required for sensory perception. MK-801 as a general blocker of NMDA receptor function may affect many other aspects of CS learning including perception of taste, general sensory processing. Similar concerns relate to manipulations used in Figure 4 using CNQX.

Reviewer #2:

Please see the previous responses to Reviewer #2.

Reviewer #3:

[…] I have several concerns regarding the conclusions of the authors about the two parallel memory traces, which should be addressed before the article can be accepted for publication. 1) The authors should at the very least hypothesize regarding the possible molecular mechanisms that might underlie the long but weak memory trace formation (e.g. NMDAR dependent), or even better test their hypothesis by applying the relevant pharmacological intervention.

In order to directly address the reviewer’s concerns regarding the possible molecular mechanisms (NMDAR dependent) that might underlie the long but weak memory trace formation, we conducted an additional experiment in which we injected NMDA receptor antagonist APV into the insular cortex 4h after novel taste consumption in a 5h ITI-CTA experiment and found no effect on CTA memory (Figure 3—figure supplement 3). The result of this experiment indicates that long-trace taste memory is not mediated by the NMDA receptor in the IC.

We strongly feel that the data we provide in the manuscript allows us to confidently argue that the NMDA receptor dependent CaMKII-GluA1 signaling in the insular cortex mediates the short-trace CTA memory formation. Moreover, recently we have reported that the NMDAR dependent function of the hippocampus in mice is crucial when there is a long temporal gap between taste and malaise (Chinnakkaruppan et al., 2014). And also, a previous study reported that a hippocampus lesion strongly interferes with the long- but not short-trace CTA in rats (Koh et al., 2009).

Taken together, we think that the dominance of the gustatory trace in the IC weakens with time while other components in CTA learning and the respective brain circuit (for instance, the necessity of NMDA receptors in the hippocampal dentate gyrus) may play a more dominant role with time.

As also suggested by the Reviewer #2, we have presented possible molecular and circuit mechanisms that might underlie the long-trace CTA in the Discussion part of the revised manuscript (last paragraph).

2) Related to this issue, it would be important to examine the effect of intra-insular infusion of MK-801 at a time interval longer than 3 hours to examine the possible effect of NMDAR on the weak memory trace.

Please refer our comment above.

3) Although the authors nicely demonstrate the temporal effects of the CTA (CS + US) acquisition traces, there is no quantification of the impact of the novel taste (CS) and malaise (US) on the formation of these memory traces. The authors could apply a weak CTA learning by reducing the LiCl concentration to address this question. Alternatively, they could induce weak learning by using latent inhibition, where the CS is familiarized before association. They should test: (a) whether this weaker learning can be acquired when association is made during the short (1-3 hrs) interval and long (4-8 hrs) interval and (b) is it sensitive to the CaMKII-GluA1 proposed mechanism or perhaps to a NMDAR dependent one.

These comments are very interesting and we have addressed them in the revised manuscript.

First, we performed an ITI-CTA experiment with the weaker US (reduced LiCl concentration; 0.025M LiCl compared to strong 0.15M LiCl, 2% b.w.) and found that the CTA was formed following the interval pertaining to the short-trace, and interestingly, also following a 4h but not 5h interval, suggesting that there is a shift in the associability at around 4h- 5h after exposure to the taste stimulus. We have included this result in the revised Figure 1D and discussed it in the text accordingly (Results and Discussion, second paragraph).

Second, as shown previously with the strong US (Rosenblum et al., 1997), we found an attenuated CTA when the NMDAR antagonist, APV, was micro-injected 25min after novel taste consumption in a 1h ITI-CTA experiment with the weaker US (Revised Figure 3—figure supplement 3). This result, together with a result presented in our response to major comment 1, suggests the dependency of the short-trace but not long-trace CTA on NMDAR in the IC.

https://doi.org/10.7554/eLife.07582.026

Article and author information

Author details

  1. Chinnakkaruppan Adaikkan

    Sagol Department of Neurobiology, University of Haifa, Haifa, Israel
    Contribution
    CA, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist
  2. Kobi Rosenblum

    1. Sagol Department of Neurobiology, University of Haifa, Haifa, Israel
    2. Center for Gene Manipulation in the Brain, University of Haifa, Haifa, Israel
    Contribution
    KR, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    kobir@psy.haifa.ac.il
    Competing interests
    The authors declare that no competing interests exist

Funding

Israeli Science Foundation Legacy Heritage (1315/09)

  • Kobi Rosenblum

German-Israeli Foundation for Scientific Research and Development (RO3971/1-1)

  • Kobi Rosenblum

Israel Science Foundation (1003/12)

  • Kobi Rosenblum

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Ulrich Bayer for providing us TatCN21 and Tatcont peptides, Dr. Takanobu Nakazawa for pY1472GluN2B antibody, and Dr. Thomas McHugh, Dr. Genela Morris, and members of KR lab for helpful discussions. This work was supported by the Israel Science Foundation (1003/12), Morasha (Israeli Science Foundation Legacy Heritage 1315/09), and German-Israeli Foundation DIP (RO3971/1-1) for KR.

Ethics

Animal experimentation: The procedures were approved by the University of Haifa ethics committee for animal research and were in accordance with the NIH guidelines for the ethical treatment of animals. All of the animals were handled according to the Haifa University animal care and use committee. All surgery was preformed under sodium pentobarbital anesthesia, and every effort was made to minimize suffering.

Reviewing Editor

  1. Christian Rosenmund, Reviewing Editor, Charité, Universitätsmedizin Berlin, Germany

Publication history

  1. Received: March 19, 2015
  2. Accepted: October 8, 2015
  3. Accepted Manuscript published: October 9, 2015 (version 1)
  4. Version of Record published: December 11, 2015 (version 2)

Copyright

© 2015, Adaikkan et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,294
    Page views
  • 367
    Downloads
  • 6
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Genes and Chromosomes
    Wahid A Mulla et al.
    Research Article Updated