Alternative Splicing and Neuritic mRNA Translocation Under Long-Term Neuronal Hypersensitivity

See allHide authors and affiliations

Science  18 Jan 2002:
Vol. 295, Issue 5554, pp. 508-512
DOI: 10.1126/science.1066752


To explore neuronal mechanisms underlying long-term consequences of stress, we studied stress-induced changes in the neuritic translocation of acetylcholinesterase (AChE) splice variants. Under normal conditions, we found the synaptic AChE-S mRNA and protein in neurites. Corticosterone, anticholinesterases, and forced swim, each facilitated a rapid (minutes), yet long-lasting (weeks), shift from AChE-S to the normally rare AChE-R mRNA, promoted AChE-R mRNA translocation into neurites, and induced enzyme secretion. Weeks after stress, electrophysiological measurements in hippocampus slices displayed apparently normal evoked synaptic responses but extreme hypersensitivity to both anticholinesterases and atropine. Our findings suggest that neuronal hypersensitivity under stress involves neuritic replacement of AChE-S with AChE-R.

Traumatic stress is often followed by long-term pathological changes (1, 2). In humans, extreme cases of such changes are clinically recognized as posttraumatic stress disorder (PTSD) (3). Although the immediate response to acute stressful insults has been extensively studied, the molecular mechanisms leading to the long-term neuronal hypersensitivity that is characteristic of PTSD are yet unknown. Stimulus-induced changes in alternative splicing have recently emerged as a major mechanism of neuronal adaptation to stress, contributing to the versatility and complexity of the expression patterns of the human genome (4–6). Another stimulus-induced post-transcriptional process is dendritic mRNA translocation, which has been described for several transcripts (7–12). Because psychological, physical, and chemical stressors all cause neuronal activation and hyperexcitation, dendritic translocation of specific target mRNAs may follow.

Acetylcholinesterase (AChE) modulations provide an appropriate case study for exploring long-term stress effects. Chemical, psychological, and physical stresses all shift splicing from the primary mRNA product that encodes the synaptic membrane AChE-S multimeric protein to the normally rare “readthrough” AChE-R transcript, which yields soluble monomers (13). We thus examined neuronal distributions of the two splice variants, which have characteristic 3′ regions (Fig. 1A). A comprehensive search of the NCBI GenBank and EST databases revealed several AChE-S mRNAs but only a single AChE-R mRNA of rodent brain origin (GenBank accession numberX70141), attesting to the scarcity and/or instability of neuronal AChE-R mRNA under normal conditions. To study changes in gene expression at the subcellular level, we used double-label fluorescence in situ hybridization (FISH) of specific AChE mRNA splice variants (14) and confocal microscope image analysis.

Figure 1

Variant-specific subcellular distribution of AChE mRNAs. (A) Alternative splicing. Shown are the distal enhancer glucocorticoid response element (GRE) and the mouseACHE gene (top), as well as AChE-S mRNA (S) and AChE-R mRNA (R). Linkage of exons 2, 3, and 4 is common to both variants (a). The R transcript includes at its 3′ terminus pseudointron 4 (green) and exon 5; option b yields the S transcript by connecting exon 4 to 6 (red). (B through E) Localization of the S and R transcripts in vitro and in vivo. Shown are percentages ± SD of FISH signals for the S and R transcripts (each totaling 100%) in three parts of the perikaryon and two neurite areas of 10 PC12 cells (B), cultured primary cerebellar neurons (C), pyramidal neurons from a paraffin-embedded slice of the prefrontal cortex (D), and cerebellar neurons from an embedded brain slice of a mouse that had been stressed for four consecutive days (2× 4 min forced swim), and sacrificed 2 weeks later (E). In control mice (D), the fraction of AChE-R mRNA in segments 4+5 was 9 ± 2% in both cortical and cerebellar neurons; for poststress mice, it was 24 ± 7% (E) (P< 0.05). Note nuclear labeling of the R transcript in cultured cerebellar neurons (arrowheads) and the punctated pattern of transcript accumulation in vivo [arrows in (D) and (E)]. Asterisks indicate columns with significant differences between AChE-S and AChE-R mRNAs (P < 0.05).

FISH detection efficiencies likely depend on probe sequences, but subcellular distributions can be reliably compared for single transcripts in different cells and conditions. Cultured PC12 cells (15), primary cultures of mouse cerebellar neurons (16), and pyramidal neurons in paraffin-embedded sections of the prefrontal cortex (17) all displayed a larger fraction of AChE-S mRNA transcripts in neuronal processes than of AChE-R mRNA (Fig. 1, B through D). Also, both cell types displayed nuclear localization of AChE-R but not of AChE-S mRNA (Fig. 1, B and C) (18). To test whether labeling properties prejudiced this conclusion, we reversed the fluorophores on the two probes (Fig. 1, B and C). In paraffin-embedded brain sections from naı̈ve mice, cortical pyramidal neurons presented dispersed AChE-S mRNA throughout the processes, while AChE-R mRNA was localized to the cell body. In addition, the perikaryal cytoplasm exhibited punctated concentrations of AChE-R mRNA alternating with double-labeled regions, whereas neurites had AChE-S mRNA with foci of both transcripts (Fig. 1D). The neurite contents of cultured PC12 cells and developing cerebellar neurons, also known to express ample AChE (19), and of prefrontal cortex neurons in vivo were 22 ± 3, 28 ± 4 and 28 ± 7%, respectively, for AChE-S mRNA but only 10 ± 2, 7 ± 2 and 11 ± 6% for AChE-R mRNA (P < 0.05, two-tailed Student's t-test). AChE-S mRNA, thus, preferentially localized in neurites (except under stress, see Fig. 1E, below).

The human ACHE gene includes a glucocorticoid response element (GRE) about 17 kb upstream from the transcription initiation site (Fig. 1A). In humans, a deletion adjacent to the GRE causes constitutive overexpression and anti-AChE hypersensitivity (20), which suggests a physiologically significant role for glucocorticoids in regulating both neuronal ACHE gene expression and anticholinesterase hypersensitivity. We therefore compared PC12 cells treated for 6 hours with corticosterone (10 μM, 0.1% ethanol) to control cells (0.1% ethanol). Twenty-four hours after treatment, catalytic activity against acetylthiocholine rose by 25 ± 14% (P < 0.05) in PC12 cells. Corticosterone, further, increased by 30 to 50% the levels of both splice variants within 24 hours (P < 0.05) (Fig. 2, A through H). However, the AChE-S mRNA-labeled area remained essentially unchanged under control conditions (distance from nuclear border to limit of mRNA labeling = 37 ± 13 μm) or corticosterone (34 ± 13 μm). In contrast, AChE-R mRNA labeling extended a smaller distance from the nucleus, 25 ± 9 μm under control conditions, increasing to 33 ± 17 μm under corticosterone (P < 0.05).

Figure 2

Corticosterone induction and antisense suppression of AChE mRNA in PC12 cells. Shown are confocal micrographs (A through F), average labeling intensities (G and H) and distributions of distances from the nucleus to the cell border (I and J) of the S (red) and R (green) transcripts in PC12 cells (n = 40) under control conditions [(A) and (B)], or after 3 hours in 10 μM corticosterone [(C) and (D)] or 1.5 nM EN101 [(E) and (F)]. Asterisks indicate columns with significant differences from controls (P < 0.05). Panels (I) and (J) present the numbers of cells that contain AChE-S or AChE-R mRNA versus their maximum distance from the nucleus.

Three-hour incubation of PC12 cells with 1.5 nM EN101, an antisense oligonucleotide that induces preferential degradation of AChE-R mRNA (15), affected the selective, yet limited, suppression of AChE-R mRNA (30 ± 8%, P < 0.05), but left unchanged AChE-S mRNA levels (Fig. 2, E through H). An inversely oriented sequence, INV101, affected neither the AChE-R mRNA level nor its distribution in PC12 cells (21), attesting to the sequence specificity of the antisense effect. The labeled area remained unchanged either with AChE-S or AChE-R cRNA probes, following EN101 treatment, suggesting full EN101 accessibility to the cell body (Fig. 2, I and J).

To test whether EN101 is effective similarly in neurites and perikaryal subcellular sites, we employed cerebellar neurons in primary culture. DIG labeling of AChE-S mRNA was not affected differently by EN101 and INV101, in either cell body or neurites. In contrast, EN101 reduced AChE-R mRNA labeling in the cell body of cerebellar neurons in culture, down to almost 50% of its level in INV101-treated controls (22). This suppression was completely restricted to the perikaryon, with no difference observed for neuritic AChE-R mRNA in antisense-treated cerebellar neurons.

Cytochemical staining of catalytically active AChE (17) revealed intensified AChE activity in the cell bodies of corticosterone-treated cultured cerebellar neurons (Fig. 3, A and C). This suggested that the overexpressed AChE mRNA transcripts were translated to yield active enzyme. Immunocytochemical staining with an antibody targeted against recombinant AChE-S (23) presented apparently similar staining patterns in neurites of control and corticosterone-treated cultured cerebellar neurons (Fig. 3, B and D). This indicated that neurites under stress secrete the hormone-induced soluble AChE-R, with no increase in the synaptic membrane-associated AChE-S.

Figure 3

Increased AChE-R expression and dendritic translocation under hormone, genetic background, sham injection, and chemical stressors. (A through D) Shown are primary cerebellar neurons stained for AChE activity under vehicle treatment [(A), 0.1% ethanol] or corticosterone [(C), 10 μM, 6 hours]. (B) and (D) Higher magnification immunocytochemical AChE-S labeling. (E through J) Confocal field images of hippocampal CA1 neurons from normal FVB/N mice [(E), (G), and (I)] and transgenic animals overexpressing human AChE [(F), (H), and (J)], under control conditions [(E) and (F)], 2 days following cannula implantation and perfusion with artificial cerebrospinal fluid [(G) and (H)], or following injection through this probe of 125 nmol neostigmine [(I) and (J)]. (K and L) Neuritic translocation of detectable AChE-R mRNA labeling (in μm from nucleus) for 30 neurons from at least two animals in each group. Significant differences from noted values are starred (P < 0.0005).

We used FISH detection of the intracellular AChE-R mRNA transcript to assess the expression levels of this variant in vivo. Dorsal hippocampus neurons of naı̈ve FVB/N mice express extremely low levels of AChE-R mRNA under normal conditions (Fig. 3E). Two days following the stress of surgical implantation of a microdialysis cannula (24), the range of neurite labeling increased from 2.0 ± 0.3 to 5.1 ± 1.0 μm from the nucleus (Fig. 3G, P < 0.0005). Injection through the cannula of the AChE inhibitor neostigmine (125 nmol) resulted in a more extensive translocation of AChE-R mRNA within 25 min (to 8.5 ± 1.2 μm from the nucleus, Fig. 3I). In neurites of untreated hypersensitive AChE-S transgenic mice, AChE-R mRNA reached greater distances from the nucleus (9 ± 1 μm) than those of their strain-matched controls (Fig. 3F, P < 0.0005). Sham injection failed to further increase transport (10 ± 1 μm, Fig. 3H), but AChE-R mRNA reached dendrite distances of 15 ± 2 μm under neostigmine (Fig. 3J), significantly longer than either similarly treated nontransgenic animals, or sham injected transgenics (in both cases P < 0.0005; Fig. 3, K and L).

Reported rates of mRNA dendrite transport range from 10 to 20 μm·hour−1 (25) to 300 to 360 μm·hour−1 (26). In our study, assuming a constant rate, AChE-R mRNA traveled a minimum of 8 ± 5 μm·hour−1 in anticholinesterase-treated FVB/N mice, which increased to 14 ± 7 μm·hour−1in similarly treated hAChE-S transgenic mice (P < 0.0005). This rate is consistent with the lower range estimate. The stability of AChE-R mRNA in the face of an antisense agent predicted long-lasting poststress neuritic presence of this transcript in vivo. To test this, we subjected FVB/N mice to 4 consecutive days of forced swim (two 4-min sessions per day). In naı̈ve mice, cerebellar granule neuron processes were loaded with 28 ± 12% of the cellular AChE-S mRNA but only 9 ± 2% of the AChE-R mRNA content (Fig. 1C). Two weeks post-stress, a considerably larger fraction (24 ± 7%) of the stress-increased amount of AChE-R mRNA translocated into neurites and displayed patches of concentrated AChE-R mRNA in both the cell bodies and processes (Fig. 1E). Thus, both the absolute levels and the neurite fraction of AChE-R mRNA increased considerably.

In nontransgenic mice, hippocampal AChE-R mRNA is generally limited to the granular layer (17). AChE-R mRNA levels remained considerably higher than baseline 4 weeks after 4 consecutive days of forced swim or 3-day exposure to very low diisopropylfluorophosphonate (DFP) levels (0.1 mg·kg−1·day−1, i.p., LD50 = 2.5 mg·kg−1). Prestressed or preexposed animals presented high-intensity labeling in the hippocampus CA1 region, dentate gyrus, and dendrite layer (Fig. 4, A through F), predicting modified composition of neuritic AChE variants either after stress or low-level exposure to an anticholinesterase.

Figure 4

Hypersensitivity in hippocampal slices under long-lasting overexpression of AChE-R mRNA. (A throughF) Neuronal overexpression. Shown is a comparison of hippocampal slices from control mice to those stressed 1 month earlier by forced swim [(A) to (D), Fast Red colorimetry] or exposure to DFP [0.1 mg kg−1, three consecutive daily injections, (E) and (F), ELF fluorimetry]. (G through J) Cholinergic hypersensitivity. Shown are extracellular recordings (average ± SD) on the CA1 areas of two to four hippocampus slices from each of three mice, in response to stratum oriens stimulation of slices from controls [(G) and (I)] or 1 month following consecutive daily stress sessions [(H) and (J)]. Above treatment descriptions are shown representative voltage traces. Stimulations (20 s each) were delivered using a bipolar electrode (10 μm insulated tungsten wires, 200 μm apart) placed in the Schaffer collaterals. Applied drug concentrations were: physostigmine, 10 μM; atropine, 10 μM; APV, 50 μM; DNQX, 20 μM.

Released neuronal acetylcholine binds to both pre- and postsynaptic receptors and is assumed to serve as a modulatory neurotransmitter and set the response level of the neuronal network to incoming stimuli (27). This involves electrophysiological mechanisms that are only partly understood, but depend on synaptic hydrolysis of acetylcholine by AChE-S. To test whether the supplementation of neuritic AChE-S with AChE-R compromises the capacity to confront intensified cholinergic stimuli, we stimulated the CA2/CA3 region of the stratum oriens, a region rich in cholinergic fibers, and recorded population field potentials (pfps), which sum the responses of a large number of neurons, in the cell body layer of the CA1 region (17). Recording in hippocampus slices (28) was performed 1 month after exposure to stress or in slices from naı̈ve animals. The anticholinesterase physostigmine induced an increase (42 ± 15%, P < 0.05) in the amplitude of the evoked population spike response of naı̈ve mice (Fig. 4, G and I). This response was 70% reversible following the addition of atropine to the bathing solution, indicating major action through muscarinic receptors. In nonstimulated slices from animals tested 1 month following 4 days of repeated stress, the mean pfp amplitude was similar to that of nonstimulated controls. However, exposure of slices from stressed mice to physostigmine resulted in a 12-fold larger increase over baseline in stimulation-evoked population spike amplitudes than that observed in nonstressed animals. Atropine administration reversibly blocked this response by more than 90%, much more effectively than its capacity to block field potentials in the control brain (Fig. 4H versus 4G). The stronger responses to both stimulators and antagonists following stress spanned the entire range of stimulus intensity (Fig. 4, I and J), similar to the hypersensitivity patterns seen in humans as PTSD. As in control mice, the evoked pfp was blocked by the NMDA and AMPA antagonists, aminophosphonopentanoic acid (APV) and dinitroquinoxalinedione (DNQX), respectively, attesting to the glutamatergic, i.e., excitatory, nature of the hypersensitized synapses. Neuronal hypersensitivity has long been known to follow stressful experiences (29), but the mechanisms leading to it remained unknown. Our current findings suggest a role for stimulus-induced AChE-R overexpression and neuritic translocation in this phenomenon.

Translocation into neuronal processes presumably depends oncis-acting elements and/or secondary structures, primarily within the 3′ untranslated region (26, 30). AChE-R mRNA includes no known neurite-targeting motif; its transport into dendrites may thus be associated with its unique 3′ sequence, or with the stress-induced accumulation of many more nascent AChE-R mRNA transcripts. The normally short half-life of AChE-R mRNA, ∼4 hours (31), also appears to be modified following stress. AChE-R mRNA includes a long U-rich element in the 3′-UTR (positions 13,020 through 13,007 in GenBank accession number AF312033), which may contribute to mRNA destabilization through the binding oftrans-acting proteins (32). Although the importance of this element to the stability of AChE-R mRNA remains to be tested, the differential resistance of neurite AChE-R mRNA to antisense oligonucleotide-mediated degradation may reflect low nuclease levels in neurites, especially of ribonuclease H, which targets DNA-RNA hybrids. AChE mRNA had been shown to be stabilized during neuronal differentiation (33); our observations suggest that this may be due, at least partially, to AChE-R transcripts being sequestered in neurites.

The weeks-long hypersensitivity of prestressed hippocampus to anticholinesterases attributes a role to the hippocampus in the cholinergic component of post-stress sensitization. The recorded field potentials were sensitive to both cholinergic and glutamatergic antagonists, suggesting a long-term change in the interactions between these two transmitter systems. Cholinergic-glutamatergic interactions have been associated with higher brain functions such as long-term potentiation, memory, and behavior (34), all of which are affected by stress. The hypersensitization may also reflect an involvement of neuritic AChE-R in plasticity, due to its capacity to affect cell-cell and cell-matrix interactions (13).

In conclusion, we find that the stress-induced increases in cholinergic neurotransmission (35) promote a long-term replacement of synaptic membrane AChE-S protein by its soluble AChE-R counterpart. The long-term hypersensitization to repeated stimuli shown by cholinergic brain tracts may be due to this substitution.

  • * Present address: Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA.

  • To whom correspondence should be addressed. E-mail: soreq{at}


View Abstract

Stay Connected to Science

Navigate This Article