A retained intron in the 3′‐UTR of Calm3 mRNA mediates its Staufen2‐ and activity‐dependent localization to neuronal dendrites

Abstract

Dendritic localization and hence local mRNA translation contributes to synaptic plasticity in neurons. Staufen2 (Stau2) is a well‐known neuronal double‐stranded RNA‐binding protein (dsRBP) that has been implicated in dendritic mRNA localization. The specificity of Stau2 binding to its target mRNAs remains elusive. Using individual‐nucleotide resolution CLIP (iCLIP), we identified significantly enriched Stau2 binding to the 3′‐UTRs of 356 transcripts. In 28 (7.9%) of those, binding occurred to a retained intron in their 3′‐UTR. The strongest bound 3′‐UTR intron was present in the longest isoform of Calmodulin 3 (Calm3 _L) mRNA. Calm3 _L 3′‐UTR contains six Stau2 crosslink clusters, four of which are in this retained 3′‐UTR intron. The Calm3 _L mRNA localized to neuronal dendrites, while lack of the 3′‐UTR intron impaired its dendritic localization. Importantly, Stau2 mediates this dendritic localization via the 3′‐UTR intron, without affecting its stability. Also, NMDA‐mediated synaptic activity specifically promoted the dendritic mRNA localization of the Calm3 _L isoform, while inhibition of synaptic activity reduced it substantially. Together, our results identify the retained intron as a critical element in recruiting Stau2, which then allows for the localization of Calm3 _L mRNA to distal dendrites.

Introduction

Dendritic mRNA localization enables neurons to alter the synaptic proteome thereby inducing plastic changes at selected synapses 1. In this multi‐step process, a selected set of RNA‐binding proteins (RBPs) assembles mRNAs containing cis‐acting sorting signals into ribonucleoprotein particles (RNPs) that are then transported along the cytoskeleton into dendrites, near synapses 2. Specific regulation of local mRNA translation at synapses in response to synaptic stimuli then allows long‐term synaptic plasticity, the cellular basis for learning and memory. Moreover, several other aspects of mRNA regulation, from nuclear RNA splicing to mRNA stability, play crucial roles in the adaptation of the synaptic proteome that is required to maintain synaptic homeostasis 3. Several studies have recently reported extensive alternative splicing in neurons 4. Furthermore, alternative polyadenylation in neurons leads to a variety of mRNA isoforms of the same transcript differing only in their 3′‐UTR length 5. Together, these phenomena give rise to mRNAs, all expressing the same polypeptide, but harboring additional regulatory elements that recruit the neuronal RBPs necessary for fine post‐transcriptional regulation 3, 4.

Staufen2 (Stau2) is a well‐known neuronal double‐stranded RBP (dsRBP) involved in asymmetric cell division of neural progenitor cells and has been implicated in dendritic RNA localization, in mature hippocampal neurons 6, 7, 8, 9. Previously, we identified a repertoire of physiologically relevant target mRNAs from neuronal Stau2‐containing RNA granules 10. For some of these targets (e.g., Rgs4, Cplx1), Stau2 influences their mRNA stability. However, only 38 of the 1,169 Stau2 targets identified by Stau2 IP (3.2%) show changes in mRNA levels upon Stau2 downregulation 10. Thus, Stau2 function is not restricted to regulation of mRNA stability.

It is unclear how Staufen proteins bind to their target mRNAs with the observed specificity, with several studies coming to different conclusions 11, 12, and even some studies suggesting its binding to be non‐sequence‐specific 8. Hence, we applied individual‐nucleotide resolution CLIP (iCLIP) 13 as this yields information about the direct Stau2 binding to its mRNA targets with higher resolution compared to Stau2 IP microarray experiments previously performed 10. This approach allowed us to uncover significant Stau2 binding to 3′‐UTRs of 356 mRNAs, and 28 of these retained an intron in their 3′‐UTR. The strongest Stau2 binding within a retained intron was seen in the 3′‐UTR of the longest isoform of Calmodulin 3 (Calm3 _L) transcript. Interestingly, Calm3 _L is the top target identified by both iCLIP and Stau2 IP microarray. Furthermore, we showed that Calm3 _L mRNA localized to dendrites in hippocampal neurons and that NMDA‐mediated neuronal activation specifically promoted dendritic localization of the intron‐containing Calm3 _L mRNA. Importantly, neither neuronal activation/silencing nor Stau2 knockdown showed any changes in total Calm3 mRNA levels. We then set out to investigate a direct role of Stau2 in dendritic mRNA localization of Calm3 _L. Finally, we demonstrated that the recruitment of Stau2 to the Calm3 _L mRNA allows its dendritic localization.

Results and Discussion

iCLIP reveals specific binding of Stau2 to Calm3 mRNA via a retained intron

In order to get a mechanistic insight into Stau2 binding to mRNAs, we performed iCLIP experiments. We immunoprecipitated the endogenous Stau2‐RNA complexes from embryonic day 18 (E18) mouse brains by using an antibody against Stau2 (Dataset EV1). IPs using a rabbit pre‐immune serum (PIS) were done in parallel as negative control. We compared our results with the iCLIP data of other two RBPs (TDP‐43 and FUS) that were also produced from E18 mouse brains 14. Here, we identified significant Stau2 binding in 3′‐UTRs of 356 neuronal mRNAs (Table EV1). For 28 of these, the binding sites in 3′‐UTRs were found in the retained introns (Fig 1A; Table EV2). Figure EV1A shows three examples of such mRNAs with Stau2 binding sites within a retained 3′‐UTR intron. Among these, Calm3 mRNA stood out as the top Stau2 mRNA target with a retained 3′‐UTR intron and with 0.24% of all iCLIP tags on mRNAs originating from Calm3. This binding was specific for Calm3 transcripts, but not the other calmodulin orthologs (Calm1, Calm2) 15, that do not contain any crosslink clusters. This was confirmed by IP experiments: Only Calm3 transcripts were highly enriched in the immunoprecipitates of Stau2‐containing RNPs (Fig 1B) 10, but no enrichment was obtained with control PIS IPs. In rat brain, three mRNA isoforms of Calm3 have been reported, which differ in their 3′‐UTR length 15. In primary rat cortical neurons, we only observed the presence of two isoforms (Fig EV1B), the longest of which (Calm3 _L) contains a retained intron in its 3′‐UTR. Our iCLIP analysis revealed that the Calm3 mRNA contains six high‐confidence crosslink clusters in the 3′‐UTR of the Calm3 _L isoform, four of which overlapped with the retained intron (Fig 1C, upper panel). The lower panel in Fig 1C shows the specific positions of these Stau2 crosslink clusters on a predicted structure of the Calm3 _L 3′‐UTR. These positions were located right next to the long‐range predicted RNA duplexes. The cluster with most crosslinking (cluster 2) was located next to a predicted long‐range duplex, which bridged regions of 3′‐UTR that are ± 700 nt apart. This observation is in agreement with the previous finding that long‐range duplexes in 3′‐UTRs of mRNAs are enriched on Stau1 binding sites 16. Importantly, no high‐confidence crosslink clusters were detected in other intronic regions of Calm3 transcript (data not shown).

Figure 1. The intron‐containing Calm3 mRNA isoform interacts with Stau2 and localizes to dendrites.

1. The proportion of cDNAs (out of all cDNAs that mapped to the mouse genome) produced by iCLIP (using E18 mouse brain extracts) from the FUS, TDP‐43, and Stau2 experiments that mapped to different RNA regions (i.e., UTR: untranslated region; CDS: coding sequence) and intergenic regions (i.e., non‐annotated transcripts).
2. Relative values of Calm3, Calm2, and Calm1 mRNA enrichment upon control or anti‐Stau2 IPs from E17.5 rat brains. Pre‐immune serum was used to perform control IPs; n = 3, average + SEM.
3. iCLIP results show that Stau2 specifically binds to the longest Calm3 (Calm3 _L) isoform retaining an intron in its 3′‐UTR (schematic in the middle). Lower panel shows the specific positions of these Stau2 crosslink clusters on a predicted structure of the Calm3 _L mRNA.
4. Representative images of endogenous Calm3 _L mRNA in rat primary hippocampal neurons (DIV15) visualized by a FISH probe directed against the intron (Calm3 intron FISH; red). Magnified insets (40‐μm dendritic sections) below identify MAP2‐positive (box 1; MAP2 in green) and MAP2‐negative (box 2) neuronal processes; arrowheads indicate the FISH signal for Calm3_L mRNA in the dendritic section; nucleus (DAPI; blue). Boxes on the top right show images of bright field (above) and (below) Stau2 (purple) co‐staining with DAPI (cyan); asterisk denotes the soma of the neuron under study; scale bar, 20 μm.
5. Co‐localization of endogenous Stau2 with endogenous Calm3 _L mRNA in the panel represented in (D). Arrowheads indicate dendritic co‐clusters (Calm3 intron FISH; green) and Stau2 immunostaining (purple); inset (top right) shows Stau2 staining in the soma at low exposure. The white perforated line marks the position of the nucleus; scale bar, 20 μm.
6. Quantification of the number of spots of Calm3 _L mRNA per length (40‐μm region) in MAP2‐positive and MAP2‐negative neuronal processes; average + SEM taken from three independent experiments (30 dendrites each), ***P < 0.001, unpaired Mann–Whitney U‐test.
7. Quantification of the cell body intensity normalized to area to measure levels of Calm3 _L mRNA in different stages of in vitro development of rat hippocampal neurons (1, 4, 8, 12, and 15 DIV), n = 3, average + SEM, t‐test; *P < 0.05, **P < 0.01.
8. qRT–PCR experiments to measure the relative levels of Calm3 _L to total Calm3 mRNA in E17.5 or adult rat cortex and 0, 2, 4, and 6–7 DIV rat cortical neurons; n = 3, average + SEM, t‐test; *P < 0.05, **P < 0.01, ***P < 0.001.

Figure EV1. Stau2 binds to the intron‐containing Calm3 mRNA isoform and regulates its expression (related to Fig 1).

1. Snapshots from UCSC genome browser visualization of transcripts on mm10 version of mouse genome that have retained introns annotated within 3'‐UTRs, as defined by UCSC genes or ENCODE. The number of cDNAs that identify each crosslink position is shown in the track “STAU2 crosslinks”. The maximum cDNA number is shown at the left of each track. Crosslinks in RNAs on the plus strand are shown in blue, and the cDNA count value is positive. In genes on the minus strand, crosslinks are shown in orange, and cDNA count value is negative.
2. Northern hybridizations using radiolabeled double‐stranded DNA probes as indicated in scheme; Calm3 ORF in green and intron in blue (left panel). Ribosomal RNAs (rRNAs) were visualized by ethidium bromide staining (right panel). The arrows indicate the position of the long and the spliced isoform of Calm3 mRNA detected with probe Calm3 ORF.
3. qPCR experiments to measure the percentage of Calm3 _L mRNA isoform in total Calm3 mRNA levels in E17.5 or adult rat cortex and 0, 2, 4, and 6–7 DIV rat cortical neurons; n = 3.
4. Representative images of endogenous Calm3 intron FISH (red) to visualize the levels of Calm3 _L mRNA in different stages of in vitro development of rat hippocampal neurons (1, 4, 8, 12, and 15 DIV). Nuclei (DAPI; blue) and anti‐MAP2 immunostaining (green) are also included. Scale bar: 20 μm.

The expression of Calm3 _L isoform increased with the development of in vitro‐cultured hippocampal (Figs 1G and EV1D) and rat cortical neurons (RCN) (Figs 1H and EV1C) and reached maximal levels in mature neurons that have undergone synaptogenesis (stage 5 neurons; see 17). Calm3 _L mRNA localizes mainly in the somato‐dendritic compartment. Also, its localization was restricted to the MAP2‐positive neuronal processes (i.e., dendrites) but not to the MAP2‐negative ones (Fig 1D and F). Importantly, Stau2 co‐localized with the Calm3 _L mRNA endogenously (Fig 1E).

Synaptic activity regulates Calm3 mRNA localization via the retained 3′‐UTR intron in hippocampal neurons

Localization of an mRNA to dendrites can lead to its local translation and hence spatio‐temporal regulation of its function. This localization is known to be influenced by neuronal activity 18. Therefore, we analyzed the effect of neuronal activation and silencing on the observed dendritic localization of endogenous Calm3 _L in rat hippocampal neurons (DIV12). The dendritic localization of Calm3 _L mRNA increased upon NMDA treatment (Fig 2A and C), and it decreased drastically upon neuronal silencing (Fig 2B and D) (see Materials and Methods). Importantly, these treatments did not affect total Calm3 _L mRNA levels as fluorescence in situ hybridization (FISH) signal intensity in the cell body (Fig 2E and F) and qPCR values (Fig EV2D) were not modified. Next, we investigated in detail the role of the 3′‐UTR intron in the dendritic localization of Calm3 transcripts. As the endogenous short isoform of Calm3 (Calm3 _M; lacking the retained exon) cannot be identified specifically due to overlapping sequences with the Calm3 _L, we took advantage of a GFP mRNA reporter assay. We generated several GFP reporters, which contained different Calm3 3′‐UTRs (scheme in Fig EV2E). We analyzed the localization of the GFP reporters in rat hippocampal neurons upon NMDA stimulation or synaptic silencing by FISH against the GFP sequence. The dendritic localization of GFP transcripts containing the intron (Calm3 _L and Calm3 _INT) increased upon neuronal stimulation (Figs 2G and I, and EV2A) and was dramatically reduced upon synaptic silencing (Figs 2H and J, and EV2B). These data showed that the Calm3 intron in the 3′‐UTR is sufficient to confer activity‐dependent changes in GFP mRNA localization. Moreover, neither of these pharmacological treatments altered the total GFP mRNA levels in Calm3 _L and Calm3 _INT reporters (Fig EV2C). Together, these experiments showed that neuronal activity regulated dendritic localization of Calm3 _L mRNA via the 3′‐UTR intron without altering its total levels.

Figure 2. Neuronal activity regulates dendritic localization of Calm3 intron‐containing endogenous and GFP reporter RNAs.

• A, B
  Representative Calm3 intron FISH images of primary rat hippocampal neurons (DIV12) that were either stimulated by NMDA (15 min) (A) or silenced O/N (B) using a standard cocktail containing TTX, CNQX, and AP5 (see Materials and Methods). Mock‐treated neurons serve as the respective controls. Dendrites are visualized with anti‐MAP2 staining (green) and nuclei with DAPI (blue). Insets below show 40‐μm dendritic sections marked by white boxes. Scale bar, 20 μm.
• C, D
  Quantification of dendritic localization experiments shown in panels (A) and (B), respectively. Bars represent the mean number of spots per 40‐μm dendritic section normalized to control + SEM taken from three independent experiments. Selected dendritic regions were at least 2 cell body diameters away from the soma. n ≥ 30 dendrites per condition; *P < 0.05; **P < 0.01; unpaired Mann–Whitney U‐test.
• E, F
  Quantification of fluorescence intensity in the cell body for Calm3 _L mRNA (detected by Calm3 intron FISH) upon neuronal stimulation by NMDA (15 min) (E) or silencing O/N (F) of three different experiments as shown in panels (A) and (B), respectively. Values are normalized to respective controls; n = 3; mean number of spots per 40‐?m dendritic section normalized to control + SEM, t‐test, P > 0.05, n.s. = not significant.
• G, H
  Representative images of primary rat hippocampal neurons (DIV11) expressing TagRFP (purple) together with either GFP‐Calm3 _L (upper rows) or GFP‐Calm3 _M (lower rows) that were stimulated by NMDA (15 min) (G) or silenced O/N (H) using a standard cocktail containing TTX, CNQX, and AP5 (the same as in panels A and B; see Materials and Methods). Bright‐field images are also included. Mock‐treated neurons serve as the respective controls. GFP mRNA is detected using a GFP FISH probe. Insets below each image show 40‐μm dendritic sections marked by white boxes. Scale bar, 20 μm.
• I, J
  Quantification of dendritic localization in panels (G) and (H). Bars represent the mean number of spots per 40‐μm dendritic section normalized to control + SEM taken from three independent experiments. Selected dendritic regions were at least 2 cell body diameters away from the soma. n ≥ 30 dendrites per condition; ***P < 0.001; unpaired Mann–Whitney U‐test.

Figure EV2. Dendritic localization of Calm3 intron‐containing GFP reporter RNAs increases with neuronal stimulation and is reduced by silencing of neurons (related to Fig 2).

• A, B
  Representative images of primary rat hippocampal neurons (DIV11) expressing TagRFP (purple) together with either GFP‐Calm3 _INT (upper row) or GFP (lower row) that were stimulated by NMDA (A) or silenced (B) using a standard cocktail containing TTX, CNQX, and AP5 (see Materials and Methods). Bright‐field images are also included. Mock‐treated neurons serve as appropriate controls for (A) and (B). These data compliment the datasets shown in Fig 2A and B. GFP mRNA is detected using a GFP FISH probe. Scale bar, 20 μm.
• C
  Quantification of fluorescence intensity in the cell body to measure total levels of GFP reporter mRNA (detected by GFP FISH experiments as represented in Figs 2G and H, and EV2A and B) using the indicated Calm3 3′‐UTR constructs with indicated pharmacological treatment; bars represent mean ± SEM values normalized to respective controls, n = 3, t‐test; P > 0.05.
• D
  qPCR to detect the effect of neuronal activation/silencing on total levels of endogenous Calm3 _L or all Calm3 isoforms in rat cortical neurons DIV7; n = 3. Bars represent mean + SEM values of mRNA levels normalized to PP1a, t‐test; P > 0.05.
• E
  Scheme representing the GFP reporter constructs used in this study.

Staufen2‐mediated dendritic localization of Calm3 mRNA via its 3′‐UTR intron

To investigate whether Stau2 directly mediates dendritic localization of intron‐containing transcripts, we evaluated the subcellular localization of the different GFP mRNA reporters when they were co‐expressed with the exogenous 62‐kDa isoform of Stau2 (TagRFP‐Stau2⁶²). TagRFP‐Stau2⁶² expression significantly increased the dendritic localization of the intron‐containing reporter mRNAs (Figs 3A and B, and EV3A and B). The presence of the intron in the reporter constructs (Calm3 _L and Calm3 _INT) was confirmed by qPCR (Fig 3H). Importantly, exogenous TagRFP‐Stau2⁶² expression did not alter the mRNA levels of either Calm3 _L and Calm3 _INT GFP or luciferase reporter constructs (Fig EV3C and D) or endogenous Calm3 _L mRNA (Fig 3F and G). This highlighted the dependence of Calm3 mRNA on the cis‐element (i.e., the Calm3 intron) and the trans‐factor (i.e., Stau2) for its localization. Moreover, the GFP‐Calm3 _INT transcripts co‐localized with TagRFP‐Stau2⁶² as well (Fig 3C and E). This co‐localization was specific for the trans‐factor Stau2 since GFP‐Calm3 _INT transcripts did not co‐localize with another RBP such as TagRFP‐Pum2 (Fig 3D and E; Calm3 mRNA does not harbor any consensus sites for Pum2 binding). Control experiments showed that Pum2 does not alter either the Calm3 _L or Calm3 _INT reporter expression in luciferase expression assays (data not shown).

Figure 3. Stau2 overexpression increases the dendritic localization of Calm3 intron‐containing GFP reporter mRNA .

• A
  Representative images of primary rat hippocampal neurons (DIV12) co‐expressing either GFP‐Calm3 _L (upper rows) or GFP‐Calm3 _M (lower rows) together with either TagRFP (left) or TagRFP‐Stau2⁶² (both in purple) (right). GFP mRNA is detected using a GFP FISH probe (green). Bright‐field images are also shown. Arrowheads in the top right panel indicate co‐localization. Insets below each image show 40‐μm dendritic sections marked by white boxes. Scale bar, 20 μm.
• B
  Quantification of dendritic localization of the different GFP reporters identified by GFP FISH upon co‐transfection together with either TagRFP or TagRFP‐Stau2⁶² as shown in (A). Bars represent the mean number of spots per 40‐μm dendritic section normalized to control + SEM taken from three independent experiments. n ≥ 30 dendrites per condition. **P < 0.01, ***P < 0.001, unpaired Mann–Whitney U‐test.
• C, D
  Representative images of neurons co‐transfected with the GFP‐Calm3 _INT reporter (mRNA identified by GFP FISH) together with either the RBP TagRFP‐Stau2⁶² (C) or the unrelated RBP TagRFP‐Pum2 (D). Bright‐field images are also shown. Insets below each image show 100‐μm dendritic sections marked by white boxes in the main image. Arrowheads indicate dendritic co‐clusters. Scale bar, 20 μm.
• E
  Quantification of the percentage of co‐localization of GFP‐Calm3 _INT total GFP mRNA spots together with TagRFP‐Stau2 or TagRFP‐Pum2 within a 100‐μm dendritic section as shown in panels (C) and (D); mean + SEM, taken from three independent experiments. n ≥ 40 dendrites per condition. Selected dendritic regions were at least 2 cell body diameters away from the soma. ***P < 0.001; unpaired Mann–Whitney U‐test.
• F, G
  Relative mRNA levels of endogenous Calm3 transcripts (total and intron‐retained isoform) upon exogenous TagRFP/TagRFP‐Stau2 expression nucleofected rat cortical neurons DIV1 (F). The total Stau2 mRNA levels in these cells are also shown (G); n = 3. Bars represent mRNA levels mean + SEM normalized to control PP1a mRNA levels; t‐test; *P < 0.05.
• H
  Relative mRNA levels of intron‐retained GFP‐Calm3 _L and GFP‐Calm3 _INT transcripts to total GFP mRNA; graph represents mean ± SEM values normalized to respective controls, n = 3, t‐test; P > 0.05.

Figure EV3. Stau2 overexpression increases the dendritic localization of Calm3 intron‐containing GFP reporter mRNA (related to Fig 3).

• A, B
  Representative images of primary rat hippocampal neurons (DIV11), expressing either GFP‐Calm3 _INT (A) or GFP (B) with co‐expression of either TagRFP (left) or TagRFP‐Stau2⁶² (right) (both in purple). GFP mRNA is detected using a GFP FISH probe (in green); bright‐field images are also shown. These data compliment the datasets shown in Fig 3A and B. Scale bars, 20 μm.
• C
  Quantification of fluorescence intensity in the cell body to measure total levels of GFP reporter mRNA (detected by GFP FISH experiments represented in Figs 3A and EV3A and B) using the indicated Calm3 3′‐UTR constructs with co‐expression of either TagRFP or TagRFP‐Stau2⁶²; bars represent mean values ± SEM normalized to respective controls, n = 3, t‐test; P > 0.05, n.s. = not significant.
• D
  Quantification of Renilla (RL) vs firefly (FL) luciferase mRNA in HeLa cells using the indicated Calm3 3′‐UTR constructs with co‐expression of either TagRFP or TagRFP‐Stau2⁶²; bars represent mean values + SEM normalized to respective controls, n = 3, t‐test; P > 0.05.

Endogenous Stau2 regulates dendritic localization of Calm3 _L mRNA

Staufen forms RNPs that mediate localization of its target mRNAs in the Drosophila oocyte 19, 20. The mammalian homolog Stau2 has been implicated in dendritic mRNA localization 6, 8; however, a precise mechanism and its specificity remains elusive. In order to investigate whether mammalian Stau2 directly regulates the dendritic localization of endogenous Calm3 _L mRNA, we performed FISH in neurons in culture. Here, we observed that downregulation of Stau2 using transiently transfected shRNA (shStau2) 21 in rat hippocampal neurons led to substantial reduction of dendritically localized intron‐retaining Calm3 _L transcripts (Fig 4A and C), while a control shRNA (shControl) did not (Fig EV4A). Importantly, the reduction in dendritic localization of Calm3 _L mRNA was completely rescued when an RNAi‐resistant Stau2 (Stau2^R) 6 was co‐expressed together with shStau2 (Fig 4B and C). Co‐expression of Stau2^R together with an shControl plasmid did not further increase the dendritic localization of Calm3 _L mRNA significantly (Fig EV4B). Interestingly, the localization of the Calm3 _L isoform was mainly restricted to the nucleus in the absence of Stau2 (Fig 4A, inset) and this effect could also be rescued by co‐expression of Stau2^R (Fig 4B, inset). Importantly, the total levels of the Calm3 _L mRNA isoform did not change upon Stau2 downregulation in hippocampal (Fig 4D) or cortical neurons (Fig 4E and F). Stau2 can shuttle between the nucleus and the cytoplasm 22. The nuclear restriction of the Calm3 _L mRNA in the absence of Stau2 further suggests a role for Stau2 in the nucleus. Whether this is linked to its role in dendrites needs further investigation.

Figure 4. Dendritic localization of endogenous intron‐containing Calm3 _L mRNA is regulated by Stau2 expression levels.

• A
  Representative image of cellular localization of endogenous Calm3 _L mRNA (identified by intron FISH) in a rat hippocampal neuron transfected with an shRNA to downregulate Stau2 (shStau2; green). Asterisks indicate Stau2 downregulated neurons (Stau2 levels were assessed by anti‐Stau2 antibody staining; purple). Bright field is also shown. Magnified inset (soma) shows a relative increase in nuclear localization. Magnified box on the right shows a decreased number of dendritic Calm3 _L puncta in a 40‐μm dendritic section marked with a white box and the corresponding phase image. The white perforated line marks the position of the nucleus. Scale bar, 20 μm.
• B
  Representative image of cellular localization of endogenous Calm3 _L mRNA (identified by intron FISH) in a rat hippocampal neuron co‐transfected with an shRNA to downregulate Stau2 (shStau2; green) and an RNAi‐resistant Stau2 rescue construct (Stau2^R). Selected dendritic regions were at least 2 cell body diameters away from the soma. Magnified box on the right shows the dendritic Calm3_L puncta in a 40‐μm dendritic section marked with a white box and the corresponding phase image. Panel on the lower left shows Stau2 immunostaining (purple), and panel on the upper left displays the corresponding bright field. Magnified inset (top right corner) shows the soma. The white perforated line marks the position of the nucleus. Asterisk denotes the soma of the neuron under study. Scale bar, 20 μm.
• C
  Quantification of dendritic Calm3 _L mRNA localization (identified by intron FISH) in the dendrites of neurons as in the experiments shown in (A) and (B). Bars represent the mean number of spots per 40‐μm‐long dendritic region normalized to control + SEM taken from three independent experiments. n ≥ 30 dendrites per condition, ***P < 0.001; unpaired Mann–Whitney U‐test.
• D
  Quantification of fluorescent intensity signal in the cell body to measure total levels of Calm3 _L mRNA [detected by Calm3 intron FISH experiments as represented in panel (A) and (B)]. Bars represent mean values + SEM, normalized to respective controls, n = 3 experiments; t‐test; P > 0.05.
• E, F
  qPCR to measure endogenous levels of Calm3 (E) or Stau2 (F) in rat cortical neurons, where Stau2 was downregulated using an shRNA. n = 3 experiments, mean + SEM, t‐test, *P < 0.05.

Figure EV4. Stau2 downregulation substantially decreases dendritic localization of endogenous intron‐containing Calm3 _L mRNA (related to Fig 4).

• A, B
  Representative images of endogenous Calm3 _L mRNA visualized by a FISH probe directed against the intron (Calm3 intron FISH) in primary rat hippocampal neurons expressing a control shRNA with co‐expression of either TagRFP (A) or a (TagRFP‐tagged) RNAi‐resistant Stau2^R (B). Asterisks indicate shRNA‐expressing neurons. The white perforated line marks the position of the nucleus. Magnified insets on the upper left show cytoplasmic localization of Calm3 _L mRNA. Magnified boxes on the right show dendritic Calm3 _L puncta and the corresponding phase image. Selected dendritic regions were at least 2 cell body diameters away from the soma. Panels on the lower left show Stau2 immunostaining and panels on the upper left the corresponding phase contrast. Scale bar, 20 μm. These data compliment the datasets represented in Fig 4A–C.

In summary, these experiments showed that mammalian Stau2 regulates the dendritic localization of the intron‐containing Calm3 _L isoform in primary neurons without affecting its stability.

Together, our study has several implications. Genome‐wide expression of mRNA with longer 3′‐UTRs increases during development in brain and muscle 23, and such isoforms have an increased probability of localizing to neural projections of hippocampal neurons 24. This is in line with our findings that the long isoform of Calm3 containing the retained intron in its 3′‐UTR is preferentially expressed in mature hippocampal neurons. Furthermore, it is this 3′‐UTR intron that enables Stau2 binding and mediates its dendritic localization. Since this dendritic Calm3 _L mRNA localization is regulated by NMDA receptor activation, it is tempting to speculate that Calm3 _L recruitment enables local protein synthesis at synapses. Importantly, Stau2 recruitment to the retained intron in the Calm3 _L mRNA suggests that Stau2 recognizes RNA structure as an element mediating specificity. While such specific recruitment is clear for RBPs that bind in an mRNA sequence‐dependent manner (in line with our “RNA signature” hypothesis 2), there has been limited evidence for dsRBPs, like Stau2. In addition, the regulation of dendritic mRNA localization of Calm3 _L, without changes in mRNA stability or decay, indicates that Stau2 performs distinct functions on specific targets. Importantly, the binding of Stau2 to 3′‐UTR introns in other 27 mRNA targets ensues an elegant mechanism wherein Stau2 recruitment can be achieved by selective intron retention. This would then render its function regulatable in specific cell types or during developmental stages. This mechanism of intron retention would be of general importance not just for Stau2 but also in the case of other RBPs.

Materials and Methods

Immunoprecipitations and RNA isolation

Stau2 RNP isolation and immunoprecipitation (IP) were performed in triplicate as described 10, 25. Pre‐immune serum was used to perform control IPs. Total RNA was isolated using mirVana™ miRNA isolation kit according to the manufacturer's instructions (Applied Biosystems). RNA was eluted, ethanol‐precipitated, and resuspended in nuclease‐free H₂O and RNA concentration measured using a NanoDrop spectrophotometer (Thermo Scientific). For quantification of mRNA levels in nucleofected rat cortical neurons, we used the QIAshredder and RNeasy kit (Qiagen) for RNA isolation. On‐column DNase (Qiagen) treatment was performed before proceeding to cDNA synthesis. All steps were performed according to the manufacturer's instructions. For Northern blot analysis, total RNA was isolated from DIV11 rat cortical neurons (RCN). For quantification of endogenous Stau2 and Calm3 (Calm3 _L and Calm3 all isoforms) mRNA levels from 0/2/4/6–7 DIV RCN or from E17.5/adult rat cortex, the total RNA was obtained using TRIZOL reagent (Thermo Scientific, 15596018) according to the manufacturer's instructions.

cDNA synthesis and quantitative RT–PCR

cDNA was synthesized from 0.5 to 1 μg DNase‐treated RNA using random primers and Superscript III™ reverse transcriptase (Invitrogen) according to the manufacturer's instructions. For IPs, 0.5 μg of input RNA, 0.5 μg of IP RNA, and an equal volume of pre‐immune IP RNA were used as template. To detect Calm1, Calm2, and Calm3 mRNAs, quantitative reverse transcriptase PCR (qRT–PCR) was performed using the SYBR Green Master Mix (Bio‐Rad) according to the manufacturer's instructions. Primers were optimized to achieve 95–105% efficiency; qRT–PCR data were analyzed using the comparative ΔΔC _T method 26. For cDNA synthesis using total RNA isolated from rat cortical neurons, 2 μg total RNA for each sample was treated with 1 unit of DNase I (Thermo Fisher Scientific) at 37°C for 30 min. DNase‐treated RNA was split in two: 1 μg was used for cDNA synthesis and the rest 1 μg for minus reverse transcriptase (−RT) reactions. Superscript III (#18080093; Thermo Fisher Scientific) was used to perform cDNA synthesis according to the manufacturer's instructions. For detecting mRNA levels by qPCR, a homemade SYBR Green Mix [containing the following components at a final concentration of 1 M betaine (B0300; Sigma), 1× standard Taq buffer (NEB), 16 μM dNTPs (N0447S; NEB), BSA 20 μg/ml (B9000S; NEB), 0.6 U per reaction Hot‐Start Taq DNA polymerase (M0495S; NEB), and 1 μl/ml of 1:100 SYBR Green (20010; Lumiprobe)] (in ddH₂O) was used. Forward and reverse primer pair mix was used 2 μl per reaction from the following stock concentrations: Renilla and firefly luciferase at 3 μM, GFP 5 μM, and Calm3 ORF Fwd/Rev 2.5 μM; Calm3 ORF Fwd/Calm3 intron Rev 4 μM, Stau2 5 μM, pp1a 3 μM, GFP Fwd/Calm3 intron Rev 4 μM, and Renilla luciferase Fwd/Calm3 intron Rev 4 μM. Five microliters of a 1:10 dilution of cDNA was added to total 15 μl reaction, in duplicate for each primer set. –RT and ddH₂O controls were used for each sample. qPCRs were performed in a LightCycler 96 system (Roche) and analyzed using the comparative ΔΔC _T method 27. PP1a mRNA levels were used as internal control for normalization. Primer sets were rigorously validated on dilution series and optimized to achieve 95–105% efficiency before use.

Stau2 iCLIP and analysis

We performed Stau2 iCLIP from E18 (embryonic day 18) mouse brain samples using anti‐Stau2 antibodies 25 or the pre‐immune serum (PIS) as a control according to a protocol described previously 14 with the following modifications. At the RNase digestion step, 20 U of RNase I (Thermo Fisher Scientific, #AM2295) was added to 1 ml of brain lysate. At the IP step, 450 μl of lysate was incubated with 2 μg of anti‐Stau2 antibody for 2 h at 4°C, followed by incubation for 1 h at 4°C on a rotation wheel with 100 μl of protein G beads (Dynabeads, Thermo Fisher Scientific, #10004D). Upon SDS–PAGE and transfer to nitrocellulose, the region corresponding to the molecular weight larger than Stau2 (> 60 kDa) was excised and RNA was extracted from the membrane.

High‐throughput sequencing was done using 50 cycles on Illumina GAII. The sequence reads were processed using the iCount server (http://icount.biolab.si) as described before 14. Briefly, sequence reads are mapped to the mouse genome (mm9/NCBI37) using Bowtie software 28 with the following parameters (‐v 2 ‐m1 ‐a –best –strata) and the mapped reads were collapsed referring the unique molecular identifiers included in the reverse transcription primer. The genomic regions were annotated using Ensembl annotation (V.59). The significant crosslinking clusters (flanking region of 15 nt and FDR < 0.05) were identified by comparing them with randomized control 13, 14, 29. The randomers were registered and the barcodes were removed before mapping the sequences to the genome sequence allowing two mismatches using Bowtie version 0.12.7 (command line: ‐v 2 ‐m 1 ‐a –best –strata). The nucleotide preceding the iCLIP cDNAs mapped by Bowtie was used to define the crosslink sites identified by truncated cDNAs. The method for the randomer evaluation, annotation of genomic segments, and identification of significantly clustered crosslinking events was performed with FDR 0.05 and a maximum spacing of 15 nt, as described earlier 13, such that the positions of crosslink sites were randomized within individual RNA regions (i.e., introns, CDS, and UTRs separately). The replicate iCLIP experiments for the same protein were grouped before performing the analyses. We used Gencode annotation to define the RNA regions and identify 3′‐UTRs containing retained introns.

Calm3 mRNA 3′‐UTR secondary structure prediction

The minimum free energy secondary structure of mouse Calm3 mRNA long 3′‐UTR (TROMER Transcriptome database id: MTR004019.7.453.0) was predicted using the RNAfold program with the default parameters 30.

Primary neuron cultures, transfections, and pharmacological treatments

Embryonic day 17 (E17) hippocampal neurons were isolated from embryos of timed pregnant Sprague Dawley rats (Charles River) as described 6 and transfected using a calcium phosphate protocol 31, 32. For NMDA‐mediated neuronal stimulation, after 24 h of expression of transfected plasmids, hippocampal neurons were treated for 10 s with 100 μM NMDA in Ca²⁺/Mg²⁺‐free PBS and then incubated in B27‐NMEM medium containing 100 μM NMDA for 15 min. Cells were then rinsed with HBSS and fixed. For neuronal silencing, after 6 h of expression of plasmids, cells were washed once with HBSS and then incubated overnight at 37°C in NMEM‐B27 medium containing 100 μM 6‐cyano‐7‐nitroquinoxaline‐2,3‐dione (CNQX), 50 μM 2‐amino‐5‐phosphonopentanoic acid (AP5), and 1 μM tetrodotoxin (TTX). Cells were then washed twice with HBSS and fixed. Mock‐treated cells were used as controls for both treatments. Where indicated, dissociated primary cortical neurons were prepared from rat cortices remaining from hippocampal dissections 10. Rat primary cortical neurons (E17) were transfected using Amaxa Nucleofection (Rat Neuron Nucleofector Kit, Lonza, program O‐003 according to the manufacturer's instructions).

Imaging‐based GFP expression assay in primary neurons

Gene fragments of interest were cloned downstream of the GFP gene into the pEGFP vector under the control of a shorter version of the synapsin promoter. As control, empty GFP reporter plasmid was used. DIV11 primary rat hippocampal neurons were co‐transfected with 1.5 μg of reporter plasmid and 1.5 μg of TagRFP or TagRFP‐Stau2 or TagRFP‐Pum2 plasmid using calcium phosphate and fixed at DIV12 for performing FISH as described below.

FISH and immunocytochemistry

Fluorescence in situ hybridization using tyramide signal amplification was performed as described 33, 34. The following RNA probes were used: Calm3 intron and GFP sense and antisense from a cloned pBluescript II KS⁺ construct described below. Immunocytochemistry was performed as previously described 34. For FISH following Stau2 knockdown, DIV11 primary hippocampal neurons were transfected (co‐transfection of control/Stau2 shRNA with either TagRFP or RNAi‐resistant TagRFP‐Stau2 62‐kDa isoform) using calcium phosphate and fixed at DIV16. For GFP mRNA FISH, DIV11 primary hippocampal neurons were transfected [co‐transfection of GFP constructs containing different Calm3 3′‐UTRs with either TagRFP or TagRFP‐Stau2 (i.e., the 62‐kDa isoform)] using calcium phosphate and fixed at DIV12.

Imaging and statistical data analysis

Images were acquired in Zen acquisition software using an Observer Z1 microscope (both from Zeiss) with a 63× planApo oil immersion objective (1.40 NA) and a CoolSnap HQ2 camera (Olympus). For FISH experiments, z‐stacks of neurons were acquired (50 stacks with optimal step size suggested by the Zen acquisition software ~0.26 nm). Images were then deconvoluted using the Zen deconvolution module. For quantification, imaging and selection of 40‐μm dendritic regions for each image was done. Images were then projected orthogonally. The Analyze particles plugin in the ImageJ software was used for quantifying the number of spots per 40‐μm dendritic region that was at least 2 cell body diameters away from the soma. For cell body intensity quantification from FISH images, the measure function in the ImageJ software was used. The average intensity/μm² of each soma was quantified.

For quantifying co‐localizing events between GFP‐Calm3 _INT and TagRFP‐Stau2⁶², a deconvoluted set of images were selected and manually scored by two independent observers. Percentage of total GFP‐Calm3 _INT particles (in a 100‐μm dendritic region) co‐localizing with TagRFP‐Stau2⁶² were quantified. We used TagRFP‐Pum2 as control in parallel. For quantifying co‐localizing events, a deconvoluted set of z‐stacks was selected and manually scored blind to the experimental conditions (without knowing whether it was TagRFP‐Stau2 or TagRFP‐Pum2) by two independent observers. Only the spots that were in the same plane and had their center of focus co‐localizing were scored as co‐localization events. Individual “blind” scores were then averaged, and the data were presented as percentage of total particles co‐localizing in a 40‐μm dendrite (2 cell body diameters away from the soma).

For all experiments, ≥ 30 dendrites (1 dendrite per neuron)/≥ 30 cell body per set from three independent experiments were selected for quantification. The conditions of experiments were kept blind for the observer until final analysis. GraphPad Prism 7.0 software was used to test the normal distribution of the data (D'Agostino–Pearson omnibus test) and for significance testing before decision of the statistical test to be used. Normalized values were used to determine significant differences with the unpaired Mann–Whitney U‐test for samples with unequal variances. All graphs were plotted in MS Excel.

Antibodies

Primary antibodies: Mouse monoclonal and rabbit polyclonal anti‐Stau2 antibodies (both used at 1:500 dilution) 35 were generated by affinity purification from existing immune sera; polyclonal anti‐GFP antibodies were a gift from Werner Sieghart (CBR, Vienna, Austria) (used at 1:5,000 dilution). The following commercial antibodies were used: rabbit polyclonal anti‐RFP (1:4,000) (Life Technologies, R10367); and monoclonal anti‐MAP2 (1:500) (Sigma‐Aldrich, M4403).

Secondary antibodies: Donkey anti‐mouse A488‐, A555‐, or A647‐conjugated antibodies and donkey anti‐rabbit A555‐ or A647‐conjugated antibodies (all from Life Technologies) were used at 1:1,000 dilution.

Plasmids

shControl (Dharmacon) and shStau2 21 sequences were cloned into the pSuperior + GFP vector system as described. Full‐length Calm3 3′‐UTR was PCR‐amplified from a rat EST plasmid obtained from ImaGenes (IMAGp998L0619945Q) (accession number AF231407), using the primers Calm3_L Fwd and Rev, and then cloned into the psiCHECK2 vector (Promega) as described 10. For cloning the Calm3 intron, the primers Calm3_INT Fwd and Rev were used, and it was cloned into pEGFP‐C2 vector via EcoRI/BamHI and then further sub‐cloned into pBluescript KS⁺ vector via SacI/BamHI (for FISH probes) and into psiCHECK2 (Promega) dual‐luciferase reporter plasmid via SacI/SalI for qPCR on luciferase reporters.

Intermediate Calm3 3′‐UTR (Calm3_M) was PCR‐amplified from a rat EST plasmid obtained from ImaGenes (IRBQp994H052D) (accession number AF231407), using the primers Calm3_M Fwd and Rev, and then cloned into psiCHECK2 plasmid via SalI/NotI. The CMV promoter in pEGFP‐C1 plasmid was replaced by synapsin (Syn) short promoter using the following primers: Syn Fwd and Rev. This construct was then used to generate GFP constructs with Calm3 full‐length 3′‐UTR (GFP‐Calm3 _L) via HindIII/SalI, Calm3 intermediate 3′‐UTR (GFP‐Calm3 _M) via HindIII/SalI, and Calm3 intron (Calm3 _INT) via EcoRI/BamHI. The coding sequence for 62‐kDa isoform of Stau2 was cloned into pTagRFP vector using the following primers: Stau2 Fwd and Rev via XhoI/SacI. The RNAi‐resistant Stau2 (Stau2^R) (in pEGFP‐N1) generated previously 6 was sub‐cloned into pTagRFP‐C. See the “Primers and shRNA sequences” section below.

Northern blot

10–15 μg of total RNA from cortical neurons DIV11 was electrophoresed in agarose–formaldehyde gels (1%), transferred to Nytran membranes (Hybond‐N, Amersham), hybridized following standard procedures 36, and analyzed using PhosphorImager screens in a Typhoon FLA9500 multi‐mode imaging scanner and Fiji software. Fragments of open‐reading frames (ORF) and 3′‐UTR intron were obtained by PCR and cloned into pBluescript KS⁺ plasmid (see primer sequences below). Double‐stranded DNA was obtained afterward by restriction endonuclease digestion and radiolabelled using the random primer PrimeIt kit (Stratagene) and [α‐³²P]dCTP.

Luciferase reporter mRNA expression and RNA isolation

Gene fragments of interest were cloned downstream of the Renilla luciferase (Luc) gene into the psiCHECK2 vector (Promega). As control, empty Luc reporter plasmid was used. HeLa cells (plated in 24‐well plates with 100,000 cells per well) were transfected with 0.1 μg of reporter plasmid and 0.4 μg of TagRFP or TagRFP‐Stau2 plasmid (per well) using Lipofectamine 2000 (Invitrogen). Total RNA was isolated 24 h after transfection. For quantification of luciferase mRNA levels in Lipofectamine‐transfected HeLa cells, we used the QIAshredder and RNeasy kit (Qiagen) for RNA isolation. On‐column DNase (Qiagen) treatment was performed before proceeding to cDNA synthesis. All steps were performed according to the manufacturer's instructions.

Primers and shRNA sequences

Primer name	Species	Primer sequence (5′–3′)
(qPCR) Calm1	Rattus norvegicus	Fwd: TTCCCCCTCTAGAAGAATCAAA
		Rev: CCACCAACCAATACATGCAG
(qPCR) Calm2	Rattus norvegicus	Fwd: AAGGTTCCCCCACTGTCAGA
		Rev: AAGCCACATGCAACATGGTA
(qPCR) Calm3	Rattus norvegicus	Fwd: ACAGCGAGGAGGAGATACGA
		Rev: CATAATTGACCTGGCCGTCT
(qPCR) Firefly luciferase	Photinus pyralis	Fwd: GAGTCTATCCTGCTGCAGCAC
		Rev: CTCGTCCACGAACACCACTC
(qPCR) Renilla luciferase	Renilla reniformis	Fwd: GTCCGGCAAGAGCGGGAATGG
		Rev: ACGTCCACGACACTCTCAGCAT
(qPCR) Calm3L	Rattus norvegicus	Calm3 ORF Fwd: GGAGACGGCCAGGTCAATTATGC
		Calm3 intron Rev: GTCACCCAAAAGAAGGGCAAACC
(qPCR) Pp1a	Rattus norvegicus	Fwd: GTCAACCCCACCGTGTTCTTG
		Rev: CTGCTGTCTTTGGAACTTTG
(qPCR) Stau2	Rattus norvegicus	Fwd: GAACATCTCCTGCTGCTGAAG
		Rev: ATCCTTGCTAAATATTCCAGTTGT
(qPCR) GFP	Aequorea victoria	Fwd: ACCCAGTCCGCCCTGAGCAA
		Rev: GCGGCGGTCACGAACTCCAG
(Cloning) Calm3L	Rattus norvegicus	Fwd: AGGCCCGGGCAGCT
		Rev: GGTAGTCACTGTATTTTATTGGAAAACA
(Cloning) Calm3INT	Rattus norvegicus	Fwd: GAATTCGGGAGCCTCTGC
		Rev: CTGGGCAGGTCCCAGGGATCC
(Cloning) Calm3M	Rattus norvegicus	Fwd: ACTTCAGTCGACAGGCCCGGGCAGCTGGC
		Rev: CTGGTTGCGGCCGCGGTAGTCACTGTATTTTATTGGAAAAC
(Cloning) Stau262	Rattus norvegicus	Fwd: ATGGCAAACCCCAAAGAGAA
		Rev: CTAGATGGCCGACTTTGATTTC
(Cloning) Syn	Rattus norvegicus	Fwd: ATACCCTGTGTCATTCCTTGTT
		Rev: GGTGGCAGCTTGGGGCA
Calm3 ORF (Northern blot)	Rattus norvegicus	Fwd: CATGGCTGACCAGCTGACC
		Rev: CACTTCGCAGTCATCATCTGTAC
Calm3 3′‐UTR intron (Northern blot)	Rattus norvegicus	Fwd: GGTCTCACTGACGCTGTTCT
		Rev: GGCAGAAAGCGATGCCAAGT
shStau2	Rattus norvegicus	GATATGAACCAACCTTCAA
shControl	Rattus norvegicus	GATCCCCCTCCAAAGTTCGATGGTTTTCAAGAGAAACCATCGAACTTTGGAG

Data availability section

Primary data

Stau2‐iCLIP data were submitted to ArrayExpress. It can be accessed through the following link: http://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-5703.

Referenced data

iCLIP data for the RBPs TDP‐43 and FUS from E18 mouse brains are published 14. The Stau2 IP data from E17 rat brains are published as well 10.

Author contributions

JU and MAK conceived the project. TS, YS, JH‐F, SMF‐M, IRM, and JE conducted experiments. All authors analyzed data. TS and MAK wrote the manuscript with feedback from all coauthors. JU and MAK provided resources and supervision.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Dataset EV1

Review Process File

EMBO Reports (2017) 18: 1762–1774