Contact: joanna.jia@wecistanche.com / WhatsApp: 008618081934791

Pls click here to Part 2

I-Fang Wang,1,2 Yihan Wang,2 Yi-Hua Yang,1,3,4 Guo-Jen Huang,5,6 Kuen-Jer Tsai,3,4,* and Che-Kun James Shen1,2,7,*

1Graduate Institute of Neural Regenerative Medicine, College of Medical Science and Technology, Taipei Medical University, Taipei 11031, Taiwan

2Institute of Molecular Biology, Academia Sinica, Taipei 11529, Taiwan

3Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan 70403, Taiwan

4Research Center of Clinical Medicine, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan 70403, Taiwan

5Department and Graduate Institute of Biomedical Sciences, College of Medicine, Chang Gung University, Taoyuan 33302, Taiwan

6Neuroscience Research Center, Chang Gung Memorial Hospital, Linkou 33302, Taiwan

7Lead contact

*Correspondence: kjtsai@mail.ncku.edu.tw (K.-J.T.), ckshen@tmu.edu.tw (C.-K.J.S.)

https://doi.org/10.1016/j.celrep.2021.109477

Cistanche can improve memory

SUMMARY

Phenotypic variation is a fundamental prerequisite for cell and organism evolution by natural selection. Whereas the role of stochastic gene expression in phenotypic diversity of genetically identical cells is well studied, not much is known regarding the relationship between stochastic gene expression and individual behavioral variation in animals. We demonstrate that a specific miRNA (miR-466f-3p) is upregulated in the hippocampus of a portion of individual inbred mice upon a Morris water maze task. Significantly, miR- 466f-3p positively regulates the neuron morphology, function and spatial learning, and memory capability of mice. Mechanistically, miR-466f-3p represses translation of MEF2A, a negative regulator of learning/memory. Finally, we show that varied upregulation of hippocampal miR-466f-3p results from randomized phosphorylation of hippocampal cyclic AMP (cAMP)-response element-binding (CREB) in individuals. This finding of modulation of spatial learning and memory via a randomized hippocampal signaling axis upon neuronal stimulation represents a demonstration of how variation in tissue gene expression leads to varied animal behavior.

INTRODUCTION

Phenotypic variation among individuals endows an evolutionary advantage since it provides the population diversity on which natural selection acts (Pavlicev et al., 2011). Genetic backgrounds and environmental factors contribute to generating such natural variation (Bendesky and Bargmann, 2011). Experimentally, inbred mice have often been used to study the molecular and cellular basis of average phenotypes and behaviors so to minimize the effects of genetic differences among the tested individuals (Casellas, 2011). However, variation of tissue transcript abundance has been shown between individual isogenic mice. Genes exhibiting highly variable expression levels are often associated with immune function, stress responses, and hormonal regulation, i.e., processes sensitive to environmental cues (Vedell et al., 2011). Some of these studies have also shown that differences in gene expression or cellular signaling, with or without the influence of environmental factors such as maternal licking, the nutrient provision in utero, or stress-induced resilience versus susceptibility (Bale, 2015; Danchin et al., 2011; Lorsch

et al., 2019; Pedersen et al., 2011), can lead to differences in phenotypes and behaviors (Casellas, 2011; Locke et al., 2015; Loos et al., 2015; Oey et al., 2015).

Spatial learning and memory formation controlled by the brain can be separated into two systems: (1) egocentric navigation employing self-movement and internal cues and (2) allocentric navigation primarily involving the hippocampus and nearby brain structures that is stimulated by distal cues outside of the organisms (Ekstrom et al., 2014). The capability of spatial learning and memory allows most animal species to progressively adjust their behaviors to adapt in spatially and temporally variable environments, which is critical for their survival. This can be assessed in laboratory animals by applying behavioral paradigms such as the Morris water maze (MWM) (Vorhees and Williams, 2014). Notably, genetically identical inbred rodents have been shown to exhibit phenotypic variation in spatial learning and memory (Tsai et al., 2002), but the underlying mechanisms of this behavioral variation remained unknown.

The formation of new memories is a complex process requiring activity-dependentgenetranscription, new protein synthesis, and

Figure 1. Correlation between the variation of spatial learning and memory capability and miR-466f-3p induction

(A) MWM task of wild-type inbred C57BL/6J mice. Left panel: MWM performances of GLN mice (dots) and PLN mice (squares). Of a total of 289 mice tested, 180 werecategorizedasGLNand109asPLN.Rightpanel:probetestoftheMWMtask(n=47and30ineachgroup).Meandurationsspentineachofthefourquadrants and the platform region are compared in the histogram. P, place of the missing platform; T, target zone without P; R, right zone; O, opposite zone; L, left zone. (B) RT-qPCR analysis of miRNA expression. Relative expression levels of six miRNAs located within the miR-466-669 cluster and miR-132-3p, as a brain-en- riched positive control, were analyzed and normalized with internal control U6 snRNA from GLN and PLN mouse hippocampus (n = 9–18 per group). I, miR-466f- 3p; II, miR-466g; III, miR-466i-3p; IV, miR-467b-3p; V, miR-467f; VI, miR-669f; VII, miR-132-3p.

finely tuned specific neuronal networking as memory engrams to make new neuronal connections and sustained plasticity (Asok et al., 2019). Hippocampal engrams, which are sparse populations of neurons in the dentate gyrus (DG), represent ensembles of neurons displaying increased activity after memory formation. While the DG engram neurons exhibit a highly distinct pattern of gene expression (Rao-Ruiz et al., 2019), the engram-specific molecular mechanisms underlying memory consolidation remain largely unknown. Various factors and signaling pathways are known to either positively or negatively regulate learning and memory processes have been identified (Abraham et al., 2019; Humeau and Choquet, 2019). Among them, the activated form of cyclic AMP (cAMP)-response element-binding protein (CREB) stimulates gene transcription in response to activity-dependent in-creases of intracellular Ca2+ in neurons (Kandel, 2012). On the other hand, the camp/PKApathwayrepressesmyocyteenhancer factor 2 (MEF2) transcriptional activity by preventing nuclear export of its co-repressor, HDAC5, and nuclear import of co-activator NFAT (Belfield et al., 2006). Furthermore, unlike CREB, MEF2A activity constrains memory formation (Cole et al., 2012). Apart from the protein factors, microRNAs (miRNAs) are also involved in regulating neuronal functions, including learning and memory (McNeill and Van Vactor, 2012). miRNAs are small (22 nt)non-codingRNAsthatprimarilyactaspost-transcriptional regulators of gene expression via sequence-specific base-pairing with their recognition sites in the 30 untranslated regions (30 UTRs) of specific mRNAs (Daugaard and Hansen, 2017). miRNAs participate in a range of signaling pathways in post-mitotic neurons, and they mediate activity-dependent cellular processes, including dendritic growth and branching, synapse formation, and maturation. By regulating gene expression, they also play important roles in long-lasting forms of synaptic plasticity that underlie memory formation, retrieval, and consolidation (Chen and Shen, 2013; Wang et al., 2012b).

In the following, we present evidence for a causative link between stochastic tissue gene expression and individual behavioral variation in animals. In particular, we show that the stochastic activation of a hippocampal CREB-pCREB / miR- 466f-3p-MEF2A axis modulates the individual variation of the spatial learning and memory capability in inbred mice.

RESULTS

Individual variation of spatial learning and memory capability is correlated with miR-466f-3p induction from a specific miRNA cluster

To identify miRNAs involved in the regulation of learning and memory formation, we used the MWM task to distinguish inbred wild-type C57BL/6J mice with good or poor learning and memory capability (Figure 1A). We defined mice that completed the task within 30 s in the last (6th) session as ‘‘good learners’’ (GLN), whereas mice that could not find the platform within 30 s in the last session were considered ‘‘poor learners’’ (PLN). As shown in Figure 1A (left panel), 62% of the mice belonged to the GLN group (180 of 289 mice), exhibiting a marked reduction in escape latency from 98.1 s (1st session) to 21.8 s (6th session). In contrast, the escape latency of the remaining 38% of the mice, the PLN group (109 of 289 mice), was only moderately reduced between the first and sixth sessions (from 111.8 s in the 1stsession to 82.1 s in the 6th session). The probe test, indicating that the mice had indeed learned the task over the six sessions, also showed that GLN mice stayed longer in the platform region than PLN mice (Figure 1A, left pair of bars in the right panel).

Next, we analyzed RNAs from the hippocampus of GLN and PLN mice by miRNA microarray hybridization (n = 4 each group). In general, we observed relatively minor differences in the expression profiles of hippocampal miRNAs from GLN and PLN mice. However, we noted that the top 10 miRNAs for which expression levels were higher in GLN mice than PLN mice (data not shown) were all derived from the same rodent-specific miRNA cluster, miR-466-669, located in intron 10 of the mSfmbt2 gene (see below). Based on reverse transcription-quantitative polymerase chain reaction (RT-qPCR) analysis the expression levels of selected miRNAs in the cluster, we chose miR-466f-3p for further investigation (Figure 1B). Average expression of this miRNA was 1.5-fold higher in the hippocampus of GLN mice relative to PLN mice. Notably, average levels of hippocampal miR-466f-3p were similar among the PLN group, home cage (HC) control group, and swimming control group (Figure 1C), thereby excluding exercise-related effects during swimming and further supporting the idea that miR-466f-3p was induced during spatial learning and memory formation.

In Figure 1C, the data reveal that expression levels of hippocampal miR-466f-3p of over 42% of GLN mice were at least 1.5-fold higher than the average expression level of HC control mice, whereas the levels in 15% of PLN mice were half that of HC mice. Approximately 25% of GLN mice and 55% of PLN mice had similar levels of miR-466f-3p (0.8- to 1.2-fold relative to HC). Pearson correlation scatterplot showing a positive correlation between miR-466f-3p levels and percentage duration on the platform during the probe test is presented in Figure 1D. We also performed miR-466f-3p and U6 small nuclear RNA (snRNA) in situ hybridization (ISH) of the brain slices in both GLN and PLN mice (Figure 1E). Statistical analysis of the miR- 466f-3p signals via ISH confirmed that induction of miR-466f- 3p was indeed greater in the DG of GLN mice relative to that of

PLN mice, whereas no difference was found when the U6 snRNA signals were compared (right histogram, Figure 1E). Although miR-466-3p signals were not equal among the individual granule cells in DG, they still increased significantly and ubiquitously in the hippocampus of GLN mice compared to PLN mice. There- fore, the miR-466f-3p induction did not happen only in a small portion of cells, e.g., the engram neurons. These data indicate that upregulation of miR-466f-3p during the MWM task is closely associated with better spatial learning and memory capability among a significant proportion of GLN mice.

We also analyzed the average expression levels of several brain-specific miRNAs by RT-qPCR. Among them, miR-132-3p was upregulated in mouse hippocampus after the MWM task, but we did not observe significant differences between the GLN and PLN groups (bar pair VII, Figure 1B). Expression levels of miR-335-5p and miR-22 were similar among the GLN, PLN, and HC groups (Figure S1A). Thus, unlike miR-466f-3p, hippocampal expression of these miRNAs during MWM training is not correlated with the spatial learning and memory ability of the mice.

miR-466f-3p upregulation promotes neurite outgrowth and dendritic spine formation

To confirm that miR-466f-3p was indeed expressed in neurons, we performed miRNA ISH combined with immunofluorescence (IF) staining of NeuN and MAP2 in primary hippocampal neurons. The results showed that similar to other miRNAs (Cohen et al., 2011; Thomas et al., 2017), miR-466f-3p was mainly expressed in the neuronal soma region, with some additional signals in the dendrites (Figure S2A). To understand the molecular and cellular basis of the association of upregulated miR-466f-3p with learning and memory formation, we first transiently overexpressed miR-466f-3p together with a dsRed fusion polypeptide under the control of the ubiquitin promoter from the pFUGW- miR-466f-3p-dsRed plasmid into DIV10 primary hippocampal neurons (Figure S2B). Through miRNA ISH, we confirmed that the miR-466f-3p signal is stronger in the miR-466f-3p overexpression group relative to vector control or mutant miR-466f-3p group (arrows, Figure S2B). In parallel, we also established a platform to examine the effect of miR-466f-3p loss of function by inhibition miR-466f-3p using miR-sponge located in the 30 UTR of EGFP mRNA expressed from pFUGW-miR-sponge- EGFP plasmid (Figure S2C). The EGFP reporter in the sponge plasmid served as both an indicator of transfection efficiency and a sensor of cellular miRNA activity (Kluiver et al., 2012). In a significant portion of the transfected HEK293T cells, the EGFP signal decreased upon co-expression with wild-type miR-466f-3p, but not with mutant miR-466f-3p, due to inhibition of EGFP mRNA translation by binding of miR-466f-3p to miR- sponge in the 30 UTR (compare the left four images and two his- program, Figure S2C). In contrast, we did not observe decreased EGFP signal upon co-expression of the control sponge encoding eight copies of a scrambled sequence from pFUGW-SCR- sponge-EGFP plasmid with either miR-466f-3p or its mutant (compare the right four images and two histograms).

As shown in Figure 2A, ectopic expression of miR-466f-3p re- resulted in morphological changes of the neurons, most notably enhanced neurite outgrowth relative to control neurons transfected with the dsRed-expressing vector pFUGW-dsRed or mutant miR-466f-3p-expressing plasmid pFUGW-mut-miR- 466f-3p-dsRed. Quantification data showed that the average dendritic branch number per neuron was not significantly different between the miR-466f-3p-overexpressing group (bar II) and vector or mutant control (bars I and III) (right upper histogram, Figure 2A). However, the mean total dendrite length and the mean length of primary dendrites increased upon miR-466f-3p overexpression (compare bar II to bars I and III, right lower histogram of Figure 2A). In parallel, we inhibited endogenous miR-466f-3p using miR-sponge. As seen, there was no significant difference in dendritic branch number between miR-466f-3p-inhibition and control groups (compare bar IV to bars I and V, right upper histo- gram, Figure 2A), but the miR-466f-3p-inhibition group displayed a notable reduction in mean total dendrite length as well as mean lengths of primary and secondary dendrites relative to other groups (compare bar IV to bars I and V, right lower histogram of Figure 2A). Furthermore, IF staining demonstrated that miR- 466f-3p overexpression, but not inhibition, also increased the density of dendritic spines (Figure 2B, upper panels and lower left histogram). We also performed co-staining for postsynaptic density protein 95 (PSD-95) and counted the density of colocalized dendritic spines to establish exact numbers of excitatory synapses (Figure 2B, lower right histogram), which revealed that miR-466f-3p overexpression increased the density of PSD-95-positive spines compared to other control groups. How- ever, miR-466f-3p inhibition did not significantly alter the density of PSD-95-positive spines with respect to that of control groups (Figure 2B, lower right histogram). Together, these data indicate that in the absence of neuronal stimulation, miR-466f-3p inhibition alone does not affect the densities of excitatory synapses or total dendritic spines.