Upregulation of eIF4E, but not other translation initiation factors, in dendritic spines during memory formation

Abstract

Local translation can provide a rapid, spatially targeted supply of new proteins in distal dendrites to support synaptic changes that underlie learning. Learning and memory are especially sensitive to manipulations of translational control mechanisms, particularly those that target the initiation step, and translation initiation at synapses could be a means of maintaining synapse specificity during plasticity. Initiation predominantly occurs via recruitment of ribosomes to the 5' mRNA cap by complexes of eukaryotic initiation factors (eIFs), and the interaction between eIF4E and eIF4G1 is a particularly important target of translational control pathways. Pharmacological inhibition of eIF4E-eIF4G1 binding impairs formation of memory for aversive Pavlovian conditioning as well as the accompanying increase in polyribosomes in the heads of dendritic spines in the lateral amygdala (LA). This is consistent with a role for initiation at synapses in memory formation, but whether eIFs are even present near synapses is unknown. To determine whether dendritic spines contain eIFs and whether eIF distribution is affected by learning, we combined immunolabeling with serial section transmission electron microscopy (ssTEM) volume reconstructions of LA dendrites after Pavlovian conditioning. Labeling for eIF4E, eIF4G1, and eIF2α-another key target of regulation-occurred in roughly half of dendritic spines, but learning effects were only found for eIF4E, which was upregulated in the heads of dendritic spines. Our results support the possibility of regulated translation initiation as a means of synapse-specific protein targeting during learning and are consistent with the model of eIF4E availability as a central point of control.

Keywords: amygdala; dendritic spine; electron microscopy; eukaryotic initiation factor; fear conditioning; learning and memory; polyribosome.

Figure 1.

Methods. a) Experimental workflow. b) Freezing to tones during the training session by the rats used for ssTEM. c) Lateral amygdala section immunolabeled for eIF4E and embedded for EM, showing location of sampling for ssTEM (star). d,e) Brightfield image of LA cells labeled for eIF4E (d) and eIF4G1 (e) in EM sample blocks. f,g) Electron micrographs of cell bodies showing labeling for eIF4E (f) and eIF4G1 (g) on rough endoplasmic reticulum (white arrows) and on cytoplasmic polyribosomes (black arrows). h) Montage of three sections along edge of tissue labeled for eIF4E. i) Dendrite from eIF4E labeled tissue reconstructed by ssTEM. Synapses shown in red. Scale = 250 mm in (c); 20 mm in (d,e); 500 nm in (f,g).

Figure 2.

Distribution of immunolabel for eIF4E and eIF4G1. a,b) Immunolabel (arrows) for eIF4E (a) and eIF4G1 (b) in the heads of dendritic spines forming asymmetric synapses (arrowheads). c,d) Dendrites (d) with immunolabel (arrows) for eIF4E in a spine neck (c) and a spine base (d). Another dendrite (d) containing label in its shaft is visible in the upper right corner of (d). Scale = 500nm. e) The frequency of spines labeled for eIF4E, but not eIF4G, increased after learning, with a corresponding decrease in unlabeled spines. f,g) There were no correlations between the frequency of labeled and unlabeled spines for eIF4E (f) or eIF4G1 (g). h) The frequency of spine heads and necks labeled for eIF4E, but not eIF4G1, increased after learning. i) The frequency of spines with eIF4E label in one or multiple locations increased with learning. j) Among spines with only one labeled location, only eIF4E-labeled spine heads increased. k) Among spines with multiple labeled locations, spines with eIF4E in the head and neck were increased. * p < 0.05

Figure 3.

Comparison between eIF4E, eIF4G1, and eIF2α distribution. a) Brightfield image of eIF2α immunolabel in LA tissue prepared for EM. b) EM image of eIF2α immunolabeling in a dendritic spine head (arrow) forming an asymmetric synapse (arrowhead). Scale = 500nm. c) A higher percentage of dendritic spines are labeled for eIF4E, but not eIF4G1 or eIF2α, after learning. d–f) A higher percentage of dendritic spines heads and necks contain eIF4E (d), but not eIF4G1 (e) or eIF2α (f), after learning. * p < 0.05 Scale = 20mm in (a); 500 nm in (b).

Figure 4.

Synapse size. a) EMs showing measurement of postsynaptic density length (red) on spine heads containing label for eIF4E (arrows). Scale = 500 nm. b–e) Spine frequency binned by synapse size for spines with (b) or without (c) eIF4E labeling and with (d) or without (e) eIF4G1 labeling. f–h) Frequency of spines binned by synapse size with eIF4E labeling in the head (f), neck (g), or base (h). * p < 0.05

Figure 5.

Spine apparatus. a) Six non-consecutive serial EM images through a large spine forming an asymmetric synapse (white arrowheads) and containing a spine apparatus (black arrowheads) and immunolabel for eIF4E (arrows). Section numbers are shown at lower left. An astrocytic process (star) containing immunolabel is visible in sections 14–16. Scale = 500nm. b) Two consecutive serial EM images of a spine without a spine apparatus labeled for eIF4E. c) 3D reconstructions of the spines shown in (a) and (b) with spine apparatus in yellow, synapse in red, and label in brown. d,e) Frequency of spines with and without eIF4E (d) or eIF4G1 (e) labeling broken down by the presence of a spine apparatus. f,g) Spine frequency by label location and presence of a spine apparatus for eIF4E (f) and eIF4G1 (g). h) Histogram showing the synapse size distribution for all spines in the dataset by training group and presence of a spine apparatus. i) Frequency of spines with a spine apparatus and eIF4E immunolabel, binned by synapse size. * p < 0.05; # significant interaction with subject.

Figure 6.

Labeling in dendritic shafts. a) Two dendrites (d) with patches of cytoplasm containing immunolabeling for eIF4E outlined in green. b) Average percent of dendritic shaft volume containing immunolabel for eIF4E or eIF4G1. c,d) Plots of percent shaft volume with label for eIF4E (d) and eIF4G1 (d) versus diameter for each dendrite. e–h) Plots of percent shaft volume with label for eIF4E (e,g) and eIF4G1 (f,h) versus frequency of unlabeled spines (e–f) and labeled spines (g–h).

Figure 7.

Comparison with polyribosome distribution. a) EM of a polyribosome (arrow) in the head of a dendritic spine forming a synapse (arrowhead). Scale = 500 nm. B) Percentage of all spines labeled for eIFs or containing polyribosomes. c–f) Effects of learning on the frequency of spines with immunolabel for eIF4G1 or eIF4E, or with polyribosomes after infusion of vehicle or 4EGI-1. Data are broken down by location of label in the spine head (c,d) or base/neck (e,f) and the presence (c,e) or absence (d,f) of a spine apparatus. Labeling data in (c) and (d) replotted from Figure 5f,g, and all polyribosome data replotted from Ostroff et al. (2017). * p < 0.05; # significant interaction with subject; † effect of 4EGI-1

Figure 8.

Summary of learning effects on eIF4E labeling, cap-dependent (4EGI-1 sensitive) and cap-independent (4EGI-1 insensitive) polyribosomes. Effects on eIF4E from this study, effects on polyribosomes from Ostroff et al. (2017).