C2bbe1 cells, a subclone of the human CaCo-2 cell line, recapitulate human intestinal enterocytes by spontaneously creating polarized monolayers with differentiated brush-borders in culture (Peterson and Mooseker, 1992). For easy genetic manipulation, we created a C2bbe1 cell line with a blasticidin resistance gene surrounded by loxP recombination sites in the AAVS1 safe harbor locus. This system allows for the controlled insertion of doxycycline-inducible transgenes using cre recombinase (Khandelia et al., 2011). This knock-in cell line was created with CRISPR/Cas9, an sgRNA targeting the AAVS1 locus, and a homologous repair donor plasmid containing the loxP cassette (Figure 1—figure supplement 1). Following selection with blasticidin, integration of the cassette was confirmed by PCR (Figure 1—figure supplement 2). This cell line is amenable to quick and efficient cre-mediated cassette switching allowing for stable doxycycline-inducible expression of any RNA or protein of interest (Figure 1—figure supplement 3).

We used our recently developed proximity labeling technique Halo-seq (Engel et al., 2021; Lo et al., 2022) to assay RNA localization across the apicobasal axis of C2bbe1 monolayers (Figure 1B). Halo-seq relies on the expression of a Halo-tagged protein with specific localization to a subcellular compartment of interest. HaloTags are small protein domains that covalently bind to small molecules known as Halo ligands (Los et al., 2008). Through conjugation to a Halo ligand backbone, the photo-activatable singlet oxygen generator dibromofluorescein (DBF) can be specifically recruited to the Halo-tagged protein and therefore share the same subcellular distribution as the fusion protein.

When exposed to green light, DBF generates oxygen radicals that are restricted to a radius of approximately 100 nm around the Halo fusion protein (Jin et al., 1995; Li et al., 2018). RNA bases within this cloud are oxidized by the radicals, making them susceptible to nucleophilic attack by propargylamine, a cell permeable alkyne (Li et al., 2018). Halo-fusion-proximal RNAs are therefore selectively alkynylated in situ.

Following total RNA isolation, alkynylated transcripts are biotinylated using Cu(I)-catalyzed azide-alkyne cycloaddition (CuACC) ‘Click’ chemistry with biotin azide (Hein et al., 2008). Halo-proximal transcripts are therefore selectively biotinylated by this reaction allowing for purification of localized RNAs from total RNA via streptavidin pulldowns. By comparing the abundances of transcripts in total RNA and streptavidin-purified RNA samples using high-throughput sequencing, transcripts that were Halo-proximal can be identified.

To first assay the efficiency with which Halo-seq can enrich localized RNAs in C2bbe1 monolayers, we targeted the cytoplasmic compartment for analysis with Halo-seq. Because differences between the cytoplasmic and nuclear transcriptomes are well characterized (Engel et al., 2021; Zaghlool et al., 2021), we reasoned that this would provide a good benchmark of the technique in C2bbe1 monolayers. We fused a HaloTag domain to a cytoplasmically restricted NF-kappa B subunit, P65. This doxycycline-inducible transgene was integrated into the genome of C2bbe1 cells using cre/loxP recombination. The doxycycline-inducible expression and size of the cytoplasmic-Halo fusion were confirmed by gel electrophoresis with fluorescent Halo ligands (Figure 1—figure supplement 4). We also confirmed the cytoplasmic localization of the Halo fusion protein in situ using fluorescent Halo ligands (Figure 1C). We then visualized the site-specific alkynylation of biomolecules following DBF-mediated oxidation by performing azide-alkyne cycloaddition in situ with Cy3-conjugated azides. The presence of Cy3 labeled molecules required the addition of DBF, indicating that the observed alkynylation was due to the Halo-seq procedure (Figure 1—figure supplement 5). These experiments revealed that both the cytoplasmic-Halo fusion and the molecules it labeled were restricted to the cytoplasm, indicating that the DBF-mediated labeling was spatially specific.

To determine if we could biotinylate alkynated RNA, we performed Halo-seq in differentiated C2bbe1 monolayers expressing the cytoplasmic-Halo. Following in-cell alkynylation and in vitro Click-mediated biotinylation, we visualized biotinylated RNA using RNA dot blots with streptavidin-HRP. We observed strong RNA biotinylation signal that was dependent upon the addition of DBF to cells (Figure 1—figure supplement 6). We then incubated biotinylated RNA samples from DBF-treated and untreated cells with streptavidin-coated beads. For three DBF-treated samples, we recovered an average of 1.6% of the total RNA sample from the streptavidin-coated beads. In contrast, we only recovered an average of 0.05% of the total RNA sample from DBF-untreated samples, further illustrating that RNA biotinylation requires DBF treatment (Figure 1—figure supplement 7).

We then created rRNA-depleted RNAseq libraries from three streptavidin input and enriched RNA samples from DBF-treated cells. Following high-throughput sequencing, we quantified transcript and gene abundances in each sample using salmon and tximport (Patro et al., 2017; Soneson et al., 2015). As expected, samples separated primarily by fraction (input vs. streptavidin pulldown) in a principal component analysis (PCA) plot of transcript expression profiles (Figure 1—figure supplement 8).

To quantify cytoplasmic RNA localization, we created a metric called cytoplasmic bias (CB), which we defined as the log₂ of a transcript’s abundance in the streptavidin-pulldown samples compared to its abundance in the input RNA samples (Figure 1D). We used DESeq2 to identify genes that were significantly enriched in streptavidin-pulldown samples compared to input samples (Love et al., 2014). These represent genes whose transcripts were enriched in the cytoplasm. Using this approach, 1192 genes were identified as significantly cytoplasmically enriched (FDR < 0.01) while 715 genes were significantly depleted and therefore likely enriched in the nucleus (Figure 1E, Supplementary file 1).

We then looked at CB values for lncRNAs specifically. As a class, these transcripts are known to be enriched in the nucleus (Lubelsky and Ulitsky, 2018; Shukla et al., 2018), and therefore should be depleted from the streptavidin-pulldown samples. As expected, lncRNA transcripts had significantly lower CB values, indicating that Halo-seq was faithfully reporting on RNA localization in C2bbe1 cells (Figure 1F). Additionally, we used a custom salmon index to simultaneously quantify intron-containing and fully spliced versions of every transcript. Spliced transcripts are exported to the cytoplasm while intron-containing transcripts are often restricted to the nucleus (Luo and Reed, 1999). Further validating Halo-seq, we observed that intron-containing pre-mRNA transcripts generally had negative CB values, indicating their nuclear enrichment. Conversely, intron-lacking, mature transcripts of the same genes had positive CB values, indicating their cytoplasmic enrichment (Figure 1G). These results provided confidence that Halo-seq identifies localized RNAs in C2bbe1 monolayers.

To identify RNAs localized across the apicobasal axis of enterocytes, we first needed to ensure that we could establish polarized monolayers of C2bbe1 cells. We plated C2bbe1 cells on porous transwell inserts at high confluency and allowed spontaneous differentiation to occur for 7 days (Masuda et al., 2011; Peterson and Mooseker, 1992). This produced highly polarized monolayers with the expected spatially distinct localization of the polarity protein markers Ezrin and the sodium-potassium pump across the apicobasal axis, suggesting mature enterocyte-like organization (Figure 2—figure supplement 1).

To interrogate apicobasal RNA localization using Halo-seq, we created HaloTag fusion proteins that specifically localized to either the apical or basal poles of differentiated cells. In intestinal epithelial cells, podocalyxin-like protein 1 (PODXL) is apically localized while a subunit of the sodium-potassium pump (ATP1A1) is basally localized (Deborde et al., 2008; Meder et al., 2005). Using these membrane proteins, we created transgenes where HaloTags were fused to their N-termini such that the HaloTag domain was inside the cell. These doxycycline-inducible transgenes were integrated into the C2bbe1 genome using the cre/loxP strategy described in Figure 1—figure supplement 3.

We confirmed doxycycline-inducible expression of the full-length HaloTag fusion proteins using gel electrophoresis and fluorescent Halo ligands (Figure 2—figure supplement 2). The subcellular localization of both HaloTag fusions was then assessed using fluorescent Halo ligands. The PODXL fusion (henceforth referred to as apical-Halo) was restricted to the apical region of cells while the ATP1A1 fusion (henceforth referred to as basal-Halo) was restricted to the basal pole (Figure 2A–B). To assess the localization of alkynylated molecules produced by the Halo-seq protocol with these fusions, we visualized alkynylated molecules using in situ Click chemistry with a fluorescent azide as in Figure 1—figure supplement 5. We found that in cells expressing apical-Halo, alkynylated molecules were concentrated toward the apical poles of cells. Conversely, in cells expressing Basal-Halo, alkynylated molecules were concentrated toward the basal pole. In both cases, alkynylation was dependent upon the addition of DBF to cells (Figure 2—figure supplements 3 and 4). These results indicate that Halo-seq can spatially specifically alkynylate biomolecules across the apicobasal axis in epithelial monolayers, giving us confidence in the ability of Halo-seq to isolate RNA molecules that were localized to the apical and basal poles of epithelial cells.

Figure 2 with 13 supplements see all

Download asset Open asset

We then confirmed the ability of the apical-Halo and basal-Halo fusions to facilitate RNA biotinylation using RNA dot blots (Figure 2—figure supplement 5) and purified the resulting products using streptavidin-coated beads. The apical-Halo and basal-Halo fusion proteins produced less biotinylated RNA than the cytoplasmic-Halo fusion. However, for both the apical and basal fusions, we recovered significantly more RNA following streptavidin pulldown from cells treated with DBF than from control untreated cells (Figure 2—figure supplement 6), further indicating the ability of these Halo fusions to mediate selective RNA biotinylation.

We isolated apically and basally localized RNA by performing Halo-seq in triplicate with apical-Halo and basal-Halo. We then quantified the resulting RNA samples using high-throughput sequencing of rRNA-depleted RNAseq libraries, calculating transcript and gene abundances using salmon and tximport (Patro et al., 2017; Soneson et al., 2015). Sequencing replicates of all Halo-seq samples first separate by genotype (PC1xPC2) then by fraction (PC2xPC3) in a PCA of gene expression profiles (Figure 2—figure supplement 7). To identify RNAs that were differentially localized across the apicobasal axis, we calculated an apical bias (AB) value for each gene. We defined this metric as the log₂ ratio of a gene’s abundance in the apical-Halo pulldown samples compared to its abundance in the basal-Halo pulldown samples (Figure 2C, Supplementary file 1). Genes with positive AB values are therefore apically-enriched while those with negative AB values are basally-enriched. AB values and their associated p values were calculated using DESeq2 (Love et al., 2014). With this approach, transcripts from 1053 genes were found to be significantly apically localized (FDR < 0.05) while transcripts from 1200 genes were significantly basally localized (Figure 2D). These results suggest that many RNAs are asymmetrically distributed across the apicobasal axis in epithelial cells.

To preliminarily characterize the apical and basal transcriptomes, we used TopGO (Alexa, 2010) to identify enriched gene ontology (GO) terms associated with the proteins encoded by RNAs localized to each compartment (Figure 2E). The term ‘trans-membrane transport’ was significantly associated with RNAs localized to both compartments (p<0.05), consistent with the fact that both the apical and basal membranes of enterocytes play major roles in nutrient import and export (Kiela and Ghishan, 2016; Snoeck et al., 2005). Unique GO terms associated with apically localized RNAs included ‘cell junctions’ including apical tight junctions, and ‘microvillus membrane’ which is exclusively associated with enterocyte apical membrane (Crawley et al., 2014; Sluysmans et al., 2017). Proteins encoded by basally localized RNAs were enriched (p<0.05) for functions associated with cell adhesion and integrin binding, both of which are important for basal membrane anchoring to the extracellular matrix (Yurchenco, 2011).

RNAs from nuclear-encoded mitochondrial proteins were also found to be basally enriched. Because these RNAs are often localized to mitochondria due to their translation on the mitochondrial surface (Corral-Debrinski et al., 2000; Williams et al., 2014), this would imply that mitochondria themselves may also be basally localized in C2bbe1 monolayers. To test this, we visualized mitochondria using both immunofluorescence and a mitochondrial-specific dye, mitoView. Both methods indicated that the localization of mitochondria was, in fact, basally biased (Figure 2—figure supplement 8). These results gave us confidence in the reliability of the Halo-seq quantifications.

A previous study used laser capture microdissection (LCM) and RNAseq to assess RNA localization across the apicobasal axis in adult mouse intestine tissue slices (Moor et al., 2017). In this study, RNA samples from apical and basal LCM slices were compared to define RNAs differentially distributed across the apicobasal axis of these cells. To assess the quality of our Halo-seq data, we compared our AB values to those defined in this dataset. Using the mouse intestine data, we calculated AB values for all genes and compared them to Halo-seq derived AB values of their human orthologs. We found a modest (R=0.08), yet positive and significant (p=0.003) correlation between these two datasets (Figure 2—figure supplement 9). This correlation overcomes differences in samples (cultured cells vs. tissue), methodology (Halo-seq vs. LCM), and species (human vs. mouse). We found this correlation to be encouraging and further suggestive of the ability of Halo-seq to faithfully report on RNA localization across the apicobasal axis in epithelial cells.

To experimentally validate Halo-seq’s ability to identify transcripts differentially localized across the apicobasal axis, we performed single molecule fluorescent in situ hybridization (smFISH) to visualize endogenous transcripts within C2bbe1 monolayers (Tsanov et al., 2016). We selected one transcript with a positive AB value, COL12A1, and one transcript with a negative AB value, SERPINF2, for validation (Figure 2F, Figure 2—figure supplement 10). Spots corresponding to RNA molecules in smFISH images were detected using FISH-quant (Mueller et al., 2013; Tsanov et al., 2016), and the 3D coordinates of each spot were identified. This allowed calculation of the position of each transcript’s position along the image’s Z axis, which corresponds to the cell’s apicobasal axis. When analyzing smFISH images for each endogenous transcript, we ensured that the number of spots detected per cell was proportional to the expression levels of the transcript as calculated by RNAseq (Figure 2—figure supplements 11 and 12), suggesting that each custom designed smFISH probe set (Tsanov et al., 2016) was faithfully detecting its cognate transcript. By comparing to the localization of an exogenous transcript encoding Firefly luciferase as a nonlocalized control, we found that COL12A1 transcripts were significantly apically localized while SERPINF2 transcripts were basally localized (Figure 2G, Figure 2—figure supplement 13). These results are consistent with their AB values as determined by Halo-seq, giving us additional confidence in the reliability of the Halo-seq dataset.

Using our Halo-seq derived dataset of RNAs localized across the apicobasal axis of enterocytes, we observed that mRNAs encoding ribosomal proteins (RP mRNAs) were, as a class, strongly basally localized (Figure 3A). This is consistent with previous observations made in adult mouse intestines (Moor et al., 2017), as well as with a LCM-based study of RNA localization in Drosophila follicular epithelial cells (Cassella and Ephrussi, 2022; Figure 3—figure supplement 1), further suggesting that our model and technique are faithfully recapitulating phenomena that occur in vivo. While RP mRNAs are known to be basally localized in epithelial cells (Cassella and Ephrussi, 2022; Moor et al., 2017), the mechanisms underlying this localization are unknown.

Figure 3 with 13 supplements see all

Download asset Open asset

Essentially all RP mRNAs share a highly conserved, pyrimidine-rich sequence motif in their 5′ UTR sometimes called a 5′ TOP (terminal oligopyrimidine) motif that likely regulates their translation (Iadevaia et al., 2008; Lahr et al., 2017; Yoshihama et al., 2002). Given that RP mRNAs were strongly basally localized in epithelial cells, we hypothesized that 5′ TOP motifs may be regulating their localization. Similar 5′ UTR motifs called pyrimidine-rich translational elements (PRTEs) have additionally been found to regulate the translation of some non-RP mRNAs (Hsieh et al., 2012). We found that PRTE-containing RNAs were also basally enriched, suggesting that like 5′ TOP motifs, PRTEs may also be regulating RNA localization (Figure 3A). Although there is some inconsistency in the literature, in general, 5′ TOP motifs are defined as being located immediately next to the 5′ cap (Berman et al., 2021) while PRTEs can be more generally localized anywhere within the 5′ UTR. There are conflicting reports as to whether the proximity of a pyrimidine-rich motif to the 5′ cap affects its function (Al-Ashtal et al., 2021; Hong et al., 2017; Lahr et al., 2017).

To determine if the location of pyrimidine-rich elements within 5′ UTRs of mRNAs was connected to RNA localization, we defined the presence and location of pyrimidine-rich elements within all RP and PRTE-containing mRNAs (see Methods). If a UTR contained multiple pyrimidine-rich elements, we used the position of the most 5′ element. We observed no difference in basal localization between RNAs that contained 5′ TOP motifs immediately adjacent to the 5′ cap and those that contained PRTE elements further downstream in the 5′ UTR (Figure 3B). From this, we conclude that there is no relationship between the localization of an RNA and the position of pyrimidine-rich elements within its 5′ UTR. Therefore, we will refer to these elements as pyrimidine-rich regulatory elements (PRREs) as (1) their position within 5′ UTRs seems unimportant and (2) they may regulate more than translation.

Although the above analyses showed a relationship between PRREs and RNA localization, they do not demonstrate causality. To directly test the ability of PRREs to regulate RNA localization, we selected two RP mRNAs, RPL7 and RPS28, that were basally localized in the Halo-seq data (Figure 3—figure supplement 2) and validated their localization using smFISH (Figure 3C, Figure 3—figure supplements 3 and 4, Supplementary file 2). We then placed their entire 5′ UTRs in front of the open reading frame (ORF) of Firefly luciferase to create two reporter transcripts such that the PRREs in these transcripts were approximately 90 nt from the 5′ cap (Figure 3D). As negative controls, we created similar reporter RNAs containing mutated PRREs such that every pyrimidine within the motif was replaced with a purine (Figure 3D). We determined the localization of these reporter RNAs in C2bbe1 intestinal epithelial monolayers using smFISH probes designed to target the Firefly luciferase ORF.

On its own, the Firefly luciferase transcript is an exogenous, non-localized RNA. We found that addition of the PRREs from both RPL7 and RPS28 were sufficient to drive the Firefly luciferase transcript to the basal compartment. Mutated PRREs were not able to drive basal localization (Figure 3E). We then quantified the localization of each PRRE reporter’s transcripts and found that PRRE-containing reporter transcripts were significantly localized to the basal pole of the cells (p<0.01, Wilcoxon rank sum test) while those containing mutant PRREs were not (Figure 3F). To our knowledge, this is the first identified RNA element that is sufficient to drive RNAs to the basal pole of epithelial cells.

5′ TOP motifs have been previously shown to negatively regulate RNA stability (Fonseca et al., 2015). Consistent with this, mutant PRRE reporter transcripts were more abundant than their wildtype counterparts (Figure 3—figure supplement 5). It should be noted again, though, that in these reporters the PRREs are located well into the 5′ UTR, further demonstrating that they need not be immediately adjacent to the 5′ cap to exert their potential regulatory ability.

To more broadly assess the ability of the PRRE motif to regulate RNA localization in intestinal epithelial cells, we investigated the localization of our reporter constructs in a different human intestinal enterocyte cell line, HCA-7 cells. HCA-7 cells create monolayers that are more colon-like than the enterocyte-like C2bbe1 monolayers (Kirkland, 1985). Reporter constructs were integrated into HCA-7 lox cells via cre-lox recombination. We again found that PRRE motifs from RPL7 and RPS28 were both sufficient for basal RNA localization in HCA-7 monolayers while mutated PRRE motifs were not (Figure 3—figure supplement 6). Also, as in the C2bbe1 monolayers, PRRE motifs were associated with decreased transcript abundance compared to their mutated counterparts (Figure 3—figure supplement 7).

To test the ability of PRRE motifs to regulate RNA localization in epithelial cells generally, we integrated our human PRRE reporter constructs into MDCK cells through cre-mediated recombination. MDCK cells are epithelial cells derived from canine kidneys and have been used to study polarity across the apicobasal axis (Balcarova-Ständer et al., 1984). As before, we found that PRRE-containing reporter RNAs were strongly basally localized while mutant PRRE motifs were not able to drive basal RNA localization (Figure 3—figure supplement 8) and that PRREs were associated with decreased transcript abundance (Figure 3—figure supplement 9). PRRE motifs are therefore able to promote basal RNA localization in a variety of epithelial cell types across multiple mammalian species. Furthermore, elements derived from human transcripts are able to regulate RNA localization in canine-derived cells.

RP mRNAs have been found to be localized to cellular projections made by neurons and migrating cancer cells (Dermit et al., 2020; Shigeoka et al., 2019; Taliaferro et al., 2016; Figure 3—figure supplement 1). These observations were made by sequencing RNA fractions collected following mechanical fractionation using microporous membranes where small projections grow through pores restricting cell bodies on top of the membrane (Figure 3G). In these experiments, RNA localization can be described as a localization ratio (LR) comparing the log₂ abundance of a transcript in projection or neurite fraction to its abundance in the soma or cell body fraction (Figure 3H). To identify RNAs robustly localized into neurites, we combined published datasets from over 30 unique experiments where neuronal cells were subjected to mechanical fractionation. This dataset included neuronal samples spanning a range of neuronal cell types from cultured cell lines to primary cortical neurons (Goering et al., 2020; Taliaferro et al., 2016). For each neuronal sample, we calculated an LR value for each gene in each sample. These were then Z-normalized within a sample, and the gene’s LRz value was defined as the gene’s median normalized LR value across all samples. Genes with positive LRz values therefore have RNAs that are neurite-enriched while genes with negative LRz values have RNAs that are cell body enriched. Using this dataset, we confirmed that RP mRNAs are efficiently localized to neurites across dozens of experiments (Figure 3I).

As in the epithelial cells, RNAs previously defined as having a pyrimidine-rich translation element (PRTE) (Hsieh et al., 2012) were also neurite-enriched, and the location of pyrimidine-rich elements with 5′ UTRs was not correlated with RNA localization (Figure 3J). This again indicates that PRREs can likely regulate RNA localization in neurons independent of their proximity to the 5′ cap.

Although RP mRNAs are enriched in neuronal projections, the RNA sequences within the RP mRNAs that regulate this localization are unknown. We reasoned that because PRRE motifs are highly conserved and had the ability to regulate RNA localization in epithelial cells that perhaps they may similarly be able to drive RNA transcripts to neurites. Endogenous mouse Rpl7 and Rps28 transcripts were found to be localized to neurites in the transcriptome-wide subcellular fractionation datasets (Figure 3—figure supplement 10) and by RT-qPCR of CAD cell body and neurite RNA samples (Figure 3K). The PRRE motifs of these mouse RNAs differ from their human orthologs by only a few nucleotides (Figure 3—figure supplement 11).

To test if PRRE elements are sufficient for neurite localization, we integrated and expressed our human PRRE reporter RNAs derived from human RPL7 and RPS28 in CAD mouse neuronal cells. Using smFISH, we found that reporters with wildtype PRRE motifs were efficiently localized to neurites while reporters that contained mutated PRRE motifs or lacked PRRE motifs were not (Figure 3L). We quantified the smFISH data by comparing the number of RNA puncta in the neurite to the number of puncta in the soma for each reporter. We found that PRRE-containing reporters were significantly more enriched in neurites than reporters that contained mutated PRRE motifs (p<0.001, Wilcoxon rank-sum test) (Figure 3M). This effect was driven by a neurite-specific increase in transcript abundance and not due to any cell-wide RNA expression changes (Figure 3—figure supplement 12).

We then confirmed this result using an orthologous technique. We mechanically fractionated reporter-expressing CAD cells into cell body and neurite fractions and quantified reporter RNA in both fractions using qRT-PCR. Consistent with the smFISH results, we found that PRRE-containing reporter RNAs were neurite-enriched while RNAs containing mutated PRREs were significantly less enriched (Figure 3—figure supplement 13).

Human PRRE motifs are therefore sufficient for neurite localization in mouse neuronal cells, suggesting that PRRE-mediated RNA localization mechanisms are conserved across species. Further, given that PRRE motifs regulate RNA localization in intestinal epithelial cells, neuronal cells, and migrating cancer cells (Dermit et al., 2020; Figure 3F and M, Figure 3—figure supplement 1), these results demonstrate that PRRE-mediated RNA localization mechanisms are conserved across cell types with vastly different morphologies.

We next sought to identify RBPs that are required for PRRE-mediated RNA transport in neuronal cells. Multiple members of the highly conserved La superfamily of RBPs (also known as La Related Proteins or LARPs) regulate multiple aspects of RNA metabolism in a variety of cell types (Deragon, 2020; Dock-Bregeon et al., 2021). Two of these, Ssb and Larp1, have been previously found to bind pyrimidine-rich motifs within 5′ UTRs (Dermit et al., 2020; Lahr et al., 2015; Maraia et al., 2017). Given this, we assayed the requirement of Larp family members for PRRE-mediated RNA localization in CAD cells by performing a targeted RNAi screen and monitoring the localization of our PRRE-containing reporters. Importantly, although Larp6 is required for endogenous RP mRNA transport in migrating cancer cell projections (Dermit et al., 2020), it is not expressed in CAD cells and therefore was not included in the screen.

We knocked down Larp family members Ssb, Larp1, Larp4, and Larp7 using siRNA transfection. Knockdown efficiencies assayed by qPCR (Figure 4—figure supplement 1) and immunoblotting (Figure 4A) were between 80% and 95% for each Larp protein assayed. We hypothesized that if a Larp RBP was responsible for PRRE-mediated RNA localization to neurites, we would see a decrease in the neurite localization of PRRE-containing reporters when expression of the Larp RBP was reduced. Following the depletion of Larp1, we observed that wildtype PRRE-containing reporters were significantly less neurite-enriched, demonstrating that Larp1 is required for their efficient transport. This effect was PRRE-specific as the localization of mutant PRRE-containing reporters was unaffected (Figure 4B–C). Larp1 was the only assayed RBP that was required for PRRE-mediated localization as knockdown of the other Larp family members did not consistently affect the localization of wildtype PRRE reporters (Figure 4—figure supplement 2). Larp1 therefore mediates PRRE localization to neurites, and this function is not shared with other Larp family members.

Figure 4 with 11 supplements see all

Download asset Open asset

Figure 4—source data 1

Protein lysates from siRNA treated CAD cells were ran on a GoPAGE Bis-Tris Precast Gel, 4–12% before being transferred to a PVDF membrane via the iBlot dry transfer device.

LARP1 expression was visualized with a mouse monoclonal anti LARP1 antibody.

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data1-v1.zip

Download elife-80040-fig4-data1-v1.zip

Figure 4—source data 2

The ladder from the same blot was imaged.

The ladder is a broad-spectrum protein ladder (Fisher Scientific PI26623).

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data2-v1.zip

Download elife-80040-fig4-data2-v1.zip

Figure 4—source data 3

Protein lysates from siRNA-treated CAD cells were ran on a GoPAGE Bis-Tris Precast Gel, 4–12% before being transferred to a PVDF membrane via the iBlot dry transfer device.

Beta-actin expression was used as a loading control and visualized with a mouse monoclonal anti ACTB antibody.

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data3-v1.zip

Download elife-80040-fig4-data3-v1.zip

Figure 4—source data 4

The ladder from the same blot was imaged.

The ladder is a broad-spectrum protein ladder (Fisher Scientific PI26623).

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data4-v1.zip

Download elife-80040-fig4-data4-v1.zip

Figure 4—source data 5

Protein lysates from C2bbe1 LARP1 Knockout cells rescued with doxycycline-inducible cDNAs were ran on a GoPAGE Bis-Tris Precast Gel, 4–12% before being transferred to a PVDF membrane via the iBlot dry transfer device.

LARP1 expression was visualized with a mouse monoclonal anti-LARP1 antibody.

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data5-v1.zip

Download elife-80040-fig4-data5-v1.zip

Figure 4—source data 6

Protein lysates from C2bbe1 LARP1 Knockout cells rescued with doxycycline-inducible cDNAs were ran on a GoPAGE Bis-Tris Precast Gel, 4–12% before being transferred to a PVDF membrane via the iBlot dry transfer device.

Beta-actin expression was used as a loading control and visualized with a mouse monoclonal anti-ACTB antibody.

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data6-v1.zip

Download elife-80040-fig4-data6-v1.zip

Figure 4—source data 7

The ladder from the same blot was imaged.

The ladder is a broad-spectrum protein ladder (Fisher Scientific PI26623).

https://cdn.elifesciences.org/articles/80040/elife-80040-fig4-data7-v1.zip

Download elife-80040-fig4-data7-v1.zip

During our Larp depletion experiments, we observed that the loss of Larp1 resulted in CAD cells with defects in neurite outgrowth (Figure 4—figure supplement 3). Specifically, the loss of Larp1 expression reduced both the number of cells producing neurites (Figure 4—figure supplement 4) and shortened overall neurite length (Figure 4—figure supplement 5). Larp1 was the only tested Larp protein for which this effect was observed. The depletion of other Larp proteins increased the number of cells producing neurites (Figure 4—figure supplement 4), and in the case of Larp4, also increased overall neurite length (Figure 4—figure supplement 5). Larp1 is therefore both required for transport of PRRE-containing transcripts to neurites and required for proper neurite outgrowth. However, because many Larp proteins, including Larp1, are implicated in a wide range of RNA regulatory modes, we cannot directly ascribe this neurite phenotype to the loss of RP mRNA localization.

Following our knockdown experiments in mouse neuronal cells, we suspected that the pyrimidine-binding protein LARP1 may be responsible for the basal localization of PRRE-containing RNAs in human epithelial monolayers. To test this, we used CRISPR/Cas9 to generate a LARP1 knockout C2bbe1 cell line.

Using two guide RNAs, we targeted exon 2 and exon 19, both of which are included in all major isoforms of LARP1, with the goal of deleting the sequence in between them (Figure 4—figure supplement 6). Because C2bbe1 cells may not be diploid at the LARP1 locus, we expected to identify multiple Cas9-generated LARP1 alleles in a single clonal line. We identified one clonal line in which no wild type LARP1 was detectable. One or more alleles had lost the sequence in between exon 2 and exon 19 and the only remaining allele contained a 53 base pair deletion in exon 2 that resulted in a premature stop codon (Figure 4—figure supplements 6 and 7). The LARP1 RNA produced by this clone is therefore not expected to produce a full length protein product and should be targeted for nonsense-mediated decay (NMD) (Figure 4—figure supplement 6). Consistent with this, the level of LARP1 RNA produced in this clone was less than 3% of that produced in wildtype cells (Figure 4—figure supplement 8).

To reintroduce expression of LARP1 in the selected LARP1 knockout clone, we integrated a full-length LARP1 rescue cDNA under a doxycycline-inducible promoter via cre-mediated recombination. This LARP1 KO-rescue line expressed LARP1 protein only in the presence of doxycycline. As a control, we also created a line in which the knockout was rescued with doxycycline-inducible GFP (Figure 4D).

By expressing our PRRE-containing reporter RNAs in the LARP1 knockout and rescue lines, we found that PRRE-containing reporters were basally localized only when LARP1 was present, indicating that LARP1 is required for their localization (Figure 4E and F). We wondered if the loss of LARP1 expression would also reduce the basal localization of endogenous RPL7 and RPS28 transcripts. Using smFISH probes targeting the endogenous transcripts (Supplementary file 2), we assayed the localization of the endogenous RP RNAs across the apicobasal axis in our wildtype C2bbe1 cells as well as in our LARP1 KO cell line (Figure 4—figure supplements 9 and 10). We found that loss of LARP1 did not change the abundance of endogenous RP transcripts (Figure 4—figure supplement 11). However, the basal localization of the endogenous transcripts was reduced when LARP1 was not present (Figure 4G). Altogether, we found that LARP1 mediates the basal localization of PRRE-containing RNAs in epithelial monolayers and the neurite localization of the same RNAs in neuronal cells.

To better understand how PRRE motifs were regulating the localization of our reporter transcripts, we asked if the location of the motif in the transcript beyond the 5′ UTR changed its ability to regulate RNA localization. Much of the data demonstrating that LARP1 interacts with pyrimidine-rich elements does so with motifs that are close to the 5′ 7-methylguanosine cap (Al-Ashtal et al., 2021; Lahr et al., 2017; Philippe et al., 2018). To test if PRRE motifs could regulate RNA localization when they were in 3′ UTRs, we placed the RPL7 and RPS28 PRRE motifs after the Firefly luciferase ORF, thus creating 3′ PRRE reporter constructs (Figure 5A). We reasoned that if the PRREs must be in 5′ UTRs in order to regulate RNA localization, the 3′ PRRE reporters would display a reduced ability to regulate RNA localization. In C2bbe1 monolayers, we found that 3′ PRRE reporters were significantly less basally localized than their 5′ PRRE counterparts (Figure 5B). We observed similar results in mouse neuronal cells, finding that 3′ PRRE reporters were significantly less efficient at driving RNAs to neurites than their 5′ PRRE counterparts (Figure 5C). Therefore, in order to efficiently regulate RNA localization, PRRE motifs must be in the 5′ UTR of transcripts. This is consistent with LARP1 binding both PRRE motifs and the 5′ 7-methylguanosine cap (Al-Ashtal et al., 2021; Lahr et al., 2017; Philippe et al., 2018).

Figure 5 with 9 supplements see all

Download asset Open asset

Interestingly, the mutant, purine-rich 3′ PRRE reporter transcripts were strongly basally localized in epithelial cells (Figure 5—figure supplements 1 and 2). We found similar results in neuronal cells where the mutant RPL7 3′ PRRE reporter RNA was neurite-localized (Figure 5—figure supplements 3 and 4). Computational prediction of RNA secondary structure revealed that the mutant RPL7 PRRE motif is likely folded into a G-quadruplex conformation (Lorenz et al., 2011 Figure 5—figure supplement 5). G-quadruplex structures within 3′ UTRs have been previously found to target RNAs to neurites (Goering et al., 2020; Subramanian et al., 2011) and may therefore be similarly regulating RNA localization in epithelial cells.

To ask if LARP1 regulates the localization of 3′ PRRE-containing transcripts, we assayed 3′ PRRE localization in LARP1- and GFP-rescued LARP1 KO C2bbe1 monolayers using smFISH. We found that the loss of LARP1 did not affect the localization of 3′ PRRE-containing transcripts (Figure 5D, Figure 5—figure supplements 6 and 7). Similarly, in mouse neuronal cells, siRNA-mediated depletion of Larp1 did not affect the neurite localization of 3′ PRRE-containing transcripts (Figure 5E, Figure 5—figure supplements 8 and 9). These results further strengthen the argument that LARP-1-mediated RNA localization requires a PRRE motif in the 5′ UTR of the transcript.

We had identified an RNA element, PRRE motifs, that regulated RNA localization in both epithelial and neuronal cells. We wondered whether other RNA elements may be similarly able to regulate RNA localization in both cell types. We therefore set out to find other examples of RNAs that were localized in both cell types as well as the sequences that drive their localization.

Net1 RNA encodes a guanine exchange factor and is strongly localized to neurites (Arora et al., 2022a), the basal pole of intestinal epithelial cells (Moor et al., 2017), and to the projections of migrating cancer cells (Moissoglu et al., 2020; Wang et al., 2017). The element responsible for the localization of mouse Net1 RNA to neurites has recently been identified as a GA-rich region in the transcript’s 3′ UTR, and when it is deleted or targeted with morpholinos, the transcript is no longer localized (Arora et al., 2022a; Chrisafis et al., 2020).

To ask if the same RNA sequences in Net1 RNA that drive RNA localization in mouse neuronal cells are also active in human epithelial cells, we created reporter transcripts that contained 3′ UTR sequences derived from mouse Net1. These reporter constructs either contained the full length Net1 3′ UTR, just the identified GA-rich localization element (LE) or the entire Net1 3′ UTR lacking the LE (ΔLE) (Figure 6A). As before, the reporter constructs were integrated into the genomes of C2bbe1 cells, inducibly expressed with doxycycline, and the localization of the resulting transcripts was visualized using smFISH. We found that the full-length Net1 3′ UTR was sufficient to strongly drive RNAs to the basal pole of epithelial cells. The LE displayed significant yet weaker RNA localization activity, and the 3′ UTR lacking the localization element (ΔLE) displayed no basal localization activity (Figure 6B and C, Figure 6—figure supplement 1). These results indicate that the GA-rich LE within the mouse Net1 3′ UTR is both necessary and sufficient to drive RNA to the basal pole of human enterocytes.

Figure 6 with 4 supplements see all

Download asset Open asset

We then assayed the same reporter constructs in mouse neuronal CAD cells using smFISH. As in the human epithelial cells, we found that the full-length Net1 3′ UTR strongly drove transcripts to neurites, the LE had significant but reduced activity, and the ΔLE 3′ UTR displayed no activity (Figure 6D and E). These results were due to a neurite-specific accumulation of reporter transcripts and not an overall difference in transcript expression (Figure 6—figure supplement 2). The GA-rich sequence element within the 3′ UTR of mouse Net1 therefore represents another sequence that regulates RNA localization across cell types, species, and cellular morphologies.

If similar mechanisms were transporting RNAs to the basal pole of epithelial cells and the neurites of neuronal cells, we reasoned that these two subcellular compartments should be enriched for similar RNAs. To test this, we compared LRz and AB values from the neuronal fractionation and epithelial Halo-seq datasets, respectively, transcriptome-wide (Figure 6F). Somewhat surprisingly, we found a reasonably strong and highly significant correlation between these values (R=–0.28, p<2.2e-16). The negative sign of this correlation indicates that neurites and the basal pole of epithelial cells are enriched for similar transcripts. RP mRNAs, highlighted in red, exemplify this relationship (R=–0.46, p=1.5e-11). These correlations are even more striking when considering the fact that the LRz and AB values are derived from experiments that differ greatly in their approaches (mechanical subcellular fractionation vs. proximity labeling), species (mouse vs. human), and cell types (neurons vs. intestinal epithelial cells). These very different cellular compartments are therefore spatial equivalents, at least in terms of RNA localization, and their RNA contents are likely governed by the same regulatory mechanisms.

The genes that fall in the fourth quadrant of Figure 6F are localized basally in enterocytes and to the neurites of neurons. To interrogate the identity of these genes we used GO enrichment analysis (Alexa, 2010). The most enriched GO terms were ‘ribosome’ and ‘mitochondrion’ (Figure 6—figure supplement 3). We have extensively detailed that RP mRNAs are enriched both in neurites and the basal compartment of epithelial cells (Figure 3A and I, Figure 3—figure supplement 1). We have similarly found that mitochondria are enriched toward the basal pole of epithelial cells (Figure 2—figure supplement 8), and mitochondria have previously been reported to be found in neuronal projections (Mandal and Drerup, 2019). Given that many nuclear-encoded RNAs that encode mitochondrial proteins are translated on the surface of the mitochondria, it may be expected that these RNAs are also neurite- and basal-enriched. Consistent with this idea, we found that RNAs that encode components of the electron transport chain (ETC) were both neurite- and basal-enriched according to our sequencing data sets (Figure 6—figure supplement 4). This suggests that similarities in RNA localization between these two cell types can be driven by similar organelle localization.

Despite drastically different morphologies, intestinal epithelial cells and neurons are organized using similar underlying microtubule structures. These polar molecules are organized in such a way that their plus ends are oriented basally in intestinal epithelial cells and out to neurite tips in neurons (Müsch, 2004; Sugioka and Sawa, 2012; Yogev et al., 2016; Figure 7A). Because of the similar microtubule architecture in both cell types and the fact that RNAs are known to be trafficked along microtubules (Cross et al., 2021; Gagnon and Mowry, 2011), we hypothesized that RNAs were being trafficked to neurites and the basal pole of epithelial cells through the action of kinesin motor proteins. To test this hypothesis, we monitored the localization of our reporter RNAs following treatment of cells with the kinesin-1 inhibitor kinesore (Randall et al., 2017).

Figure 7 with 6 supplements see all

Download asset Open asset

Cells expressing the full length Net1 3′ UTR reporter construct were treated for four hours with kinesore or ciliobrevin A, a dynein-1 inhibitor. In C2bbe1 monolayers, we found that kinesin inhibition dramatically reduced the basal localization of reporter RNAs containing the Net1 3′ UTR (Figure 7B and C), consistent with the hypothesis that these RNAs are trafficked along microtubules in epithelial cells. This effect was specific to the Net1 3′ UTR as the localization of reporter transcripts lacking the 3′ UTR was unaffected by kinesin inhibition (Figure 7—figure supplement 1). Interestingly, we found that inhibition of dynein strengthens basal localization of Net1 3′ UTR transcripts. Kinesin and dynein activity are often closely linked, pulling the same cargoes in different directions. When the activity of one motor is inhibited, it can therefore increase the apparent activity of the other (Soppina et al., 2009). In neuronal cells, kinesin inhibition again significantly reduced the neurite localization of the Net1 3′ UTR reporter RNAs, again suggesting they are transported along microtubules (Figure 7D and E). We did not observe any changes in the localization of our control reporter transcript, Firefly luciferase, in response to drug treatments (Figure 7—figure supplements 1 and 2). Additionally, as before, changes in RNA localization were driven by changes in RNA content of the neurite and not any gross changes in expression (Figure 7—figure supplement 3).

We found very similar results with our PRRE reporter transcripts when inhibiting motor proteins. In C2bbe1 monolayers, kinesin inhibition removed the basal localization of both RPL7 and RPS28 PRRE reporter constructs (Figure 7F, Figure 7—figure supplement 4). Kinesin-mediated transport along microtubules was also required for the neurite localization of PRRE-containing transcripts (Figure 7G). Once again, changes in RNA abundance were neurite-specific and were not observed cell-wide (Figure 7—figure supplement 5).

Given that the plus ends of microtubules are associated with both neurites and the basal pole of epithelial cells and that kinesin inhibition reduced RNA localization to both compartments, we conclude that the RNA contents of both locations are determined at least in part through microtubule plus-end directed trafficking mechanisms.

C2bbe1 cells (CRL-2102) were maintained in DMEM with 110 mg/L Sodium Pyruvate (Thermo 11995065) supplemented with 10% FBS (Atlas Biologicals F-0500-D), penicillin-streptomycin (Thermo 15140122) and 10 μg/mL human transferrin (Sigma-Aldrich T8158). HCA-7 cells were maintained in a 1:1 mix of DMEM and F12 (Thermo 11320033) supplemented with 10% FBS (Atlas Biologicals F-0500-D) and penicillin-streptomycin (Thermo 15140122). MDCK cells were maintained in DMEM (Thermo 11965092) supplemented with 10% Equafetal (Atlas Biologicals EF-0500-A) and penicillin-streptomycin (Thermo 15140122).

C2bbe1 LoxP cells were maintained in 20 μg/mL blasticidin (A.G. Scientific B-1247-SOL) while HCA-7 and MDCK lox cells were maintained in 10 μg/mL blasticidin. Following cre-mediated recombination with a puromycin resistance plasmid, all reporter-expressing lines were maintained in 5 μg/mL puromycin (Cayman Chemical Company 13884).

C2bbe1, HCA-7 and MDCK cells were allowed to differentiate for 7 days after plating at 100% confluency in the presence of 2 μg/mL doxycycline (Fisher Scientific AAJ6042203) when inducing transgene and/or reporter transcript expression. Differentiation can occur in dishes, plates or on 0.4 μm transwell inserts (Corning 353090). Media was changed every 3–4 days.

CAD cells were maintained in a 1:1 mix of DMEM and F12 (Thermo 11320033) supplemented with 10% Equafetal (Atlas Biologicals EF-0500-A) and penicillin-streptomycin (Thermo 15140122). CAD lox cells were received as a gift from Eugene Makeyev (Khandelia et al., 2011) and maintained in 5 μg/mL blasticidin (A.G. Scientific B-1247-SOL).

CAD cells were grown in full growth media and induced with 2 μg/mL doxycycline for 48 hr when inducing expression of transgenic constructs. Following doxycycline induction, media was replaced with serum-free media for 48 hr which induces neurite outgrowth. Doxycycline induction was continued through the differentiation period when inducing construct expression. Reporter-expressing CAD lines were maintained in 5 μg/mL puromycin (Cayman Chemical Company 13884).

All cell lines were tested for mycoplasma and found to be negative.

A loxP-flanked blasticidin resistance cassette was integrated into the AAVS1 safe harbor of C2bbe1 and HCA-7 cells via CRISPR/Cas9. RNPs from Synthego with AAVS1 targeting sgRNA (UAGUGGCCCCACUGUGGGGU) were electroporated with the Neon electroporation system (C2bbe1:1600 V 10ms and 3 pulses, HCA-7: 1100 V 20ms 2 pulses). A donor plasmid containing the loxP cassette with homology to the AAVS1 safe harbor was co-electroporated with the RNPs. Following electroporation, cells were incubated at 37 °C for 48 hr before selection with blasticidin (20 μg/mL for C2bbe1, 10 μg/mL for HCA-7) until untransfected negative control cells had died and the newly selected loxP line exhibited normal growth patterns.

Integration of the loxP cassette into the AAVS1 safe harbor was validated by PCR amplifying the junctions of the AAVS1 site upstream and downstream the inserted cassette. Primers were designed such that PCR products are only created when the LoxP cassette is inserted as expected. The parental line was included as a negative control and did not produce junction PCR products.

MDCK loxP cells were created through random integration using the same electroporation (1600 V 10ms 3 pulses) strategy and donor plasmid for human lines. CRISPR/Cas9 RNPs were omitted from this electropporation. Selection of blasticidin resistant clones contained loxP cassettes in undetermined locations.

For the LARP1 knockout line, C2bbe1 loxP cell lines were treated with two sgRNAs from Synthego that targeted exon 2 (GCUGUUCCUAAACAGCGCAA) and exon 19 (GCUGUUCCUAAACAGCGCAA) of LARP1. CRISPR Cas9 RNPs and a GFP plasmid were co-electroporated using the Neon electroporation system (1600V 10ms and 3 pulses) and incubated at 37 °C for 48 hr. GFP positive cells were sorted to single cells using a flow cytometer (MoFlo XDP100) and allowed to expand for 2–4 weeks. Clones were screened for gene deletion by PCR as well as LARP1 mRNA and protein depletion determined by qPCR and western blot analysis with a monoclonal LARP1 antibody (Santa Cruz sc-515873). Lesions at the locus of cutting were identified by PCR using genomic DNA to amplify the locus followed by Sanger sequencing and Synthego ICE analysis.

All loxP-containing cell lines followed the same protocol for cassette switching. Cells were plated in 12-well plates at 1.0–1.5×10⁵ cells per well in full growth media 12–18 hr before transfection. Cells were then co-transfected with 500 ng donor cassette plasmid with 1% (wt/wt) of a Cre-encoding plasmid (pBT140, addgene #27493). Donor plasmids with doxycycline inducible transgenes also contained a constitutively active copy of the tetracycline transactivator. Plasmids were transfected using Lipofectamine LTX reagent (Thermo Fisher Scientific 5338030) and Opti-MEM (Thermo Fisher Scientific 31985070) following the manufacturer’s protocol. Following 24 hr with transfection reagents, the medium was changed and the incubation continued for another 24 hr before addition of 1–5 μg/mL puromycin (VWR 97064–280). Incubation with puromycin continued for several days until untransfected control wells were depleted of cells. The puro-resistant colonies were expanded in a full growth medium supplemented with 5 μg/mL puromycin.

Halo-seq was performed as previously described (Engel et al., 2021; Lo et al., 2022). Details regarding each step of Halo-seq are provided below.

Halo transgene-expressing C2bbe1 cells were seeded at high confluency on 0.4 μm PET transwell inserts (Corning 353090). Nine 6-well transwells were pooled for each replicate sample. Monolayers were differentiated for 7 days and Halo fusion proteins were induced with 0.5 µg/mL doxycycline. Following PBS wash, cells were incubated with 1 µM DBF Halo ligand in HBSS (gift from Rob Spitale) at 37 °C for 15 min. Negative control samples lacking DBF were incubated in HBSS only.

Cells were washed with full growth media for 10 min at 37 °C twice. Immediately following this, cells were incubated in 1 mM propargylamine (Sigma-Aldrich P50900) in HBSS for 5 min at 37 °C. Cells were then moved to the green light chamber containing two 100 W LED flood lights (USTELLAR) in an enclosed dark space. Cells were irradiated with green light for 10 min in the room temperature chamber sandwiched between two flood lights. Cells were collected in Trizol (Ambion) and homogenized by passing lysates through a 20 G needle 20 times. Total RNA was then isolated following Trizol’s manufacturer instructions.

RNA was DNase treated using DNase I (NEB M0303S) for 30 min at 37 °C. RNA was then recovered using Quick RNA columns (Zymo Research R1055) following the sample clean up protocol and eluted in 100 μL water.

Approximately 200 µg of total RNA was used in the Click reaction. The Click reaction contained 10 mM Tris pH 7.5, 2 mM biotin picolyl azide (Sigma-Aldrich 900912), 10 mM sodium ascorbate made fresh (Sigma 11140), 2 mM THPTA (Click Chemistry Tools 1010–100), and 100 µM copper sulfate (Fisher Scientific AC197722500). Click reactions were split into 50 μL reaction volumes and then incubated for 30 min in the dark at 25 °C. Click reactions were pooled and cleaned using a Quick RNA Mini kit (Zymo Research R1055) following the sample clean up protocol and eluted in 100 μL water. Negative control samples where cells were not treated with DBF were also biotinylated using this procedure.

A total of 100 µg of biotinylated RNA at 1 µg/µL was used as the input for streptavidin pulldowns. Different volumes of streptavidin-coated magnetic beads (Pierce PI88816) were used depending on the amount of labeled RNA by each Halo fusion protein. Halo-P65 required 1 μl beads per 2 μg total RNA while apical-Halo and basal-Halo required 1 μl beads per 4 μg RNA. Prior to mixing with RNA, beads were washed three times in B&W buffer (5 mM Tris pH 7.5, 0.5 mM EDTA, 1 M NaCl, 0.1 Tween 20), two times in solution A (0.1 M NaOH, 50 mM NaCl), and one time in solution B (100 mM NaCl). The beads were resuspended with 100 μg RNA and 50 mM NaCl to a total volume of 100 μL. The pulldown reaction was rotated for 2 hr at 4 °C. The beads were then washed three times for 5 min each in B&W buffer with rotation at room temperature.

RNA was then eluted from the beads by resuspending in 50 μL PBS and 150 μL Trizol and incubated for 10 min at 37 °C. The eluted RNA was recovered from this mixture using a DirectZol micro RNA columns (Zymo Research R2062) following the manufacturer’s instructions and eluted in 10 µL of water. The eluted RNA concentration was measured by Qubit high sensitivity RNA kit (Invitrogen Q32855) and +DBF and -DBF samples were compared.

The efficiency of the biotinylation reaction and the streptavidin pulldown were assayed by RNA dot blots. Hybond-N +membrane (GE) was wet with 2 X SSC then allowed to dry for 15 min. Five μg of biotinylated RNA samples or 1% of streptavidin eluted RNA were spotted on the membrane and allowed to dry for another 30 min. To stain for total RNA, the blot was incubated in 1% methylene blue (VWR 470301–814) for 2 min and destained using deionized water. Next, the membrane was blocked using 5% BSA for 30 min and washed three times in PBST (PBS +0.01% Tween). Biotinylated RNA was then detected using streptavidin-HRP (Abcam ab7403) at a dilution of 1–20,000 in 3% BSA by addition to the membrane with rocking overnight at 4 °C. Membranes were washed three times for 10 min each in PBST at room temperature. Streptavidin-HRP was detected using standard HRP chemiluminescent reagents (Advansta) and visualized using chemiluminescent imaging on a Sapphire molecular imager (Azure Biosystems).

Halo-transgene expressing C2bbe1 cells were plated at high confluency on poly-D-lysine-coated coverslips (neuVitro). Monolayers were differentiated for 7 days with 0.5 μg/mL Doxycycline to induce Halo fusion protein expression. Media was changed every 3–4 days. Cells were treated with DBF, propargylamine, and irradiated with green light exactly as described for Halo-seq. Instead of lysing cells with Trizol to recover RNA, cells were washed with PBS 3 times and fixed and permeabilized with 3.7% formaldehyde and 0.3% Triton-X (VWR 80503–490) in PBS for 30 min at room temperature. Following PBS wash, in situ click reactions were performed by incubating cells with 100 µL Click buffer (100 µM copper sulfate, 2 mM THPTA, 10 mM fresh sodium ascorbate, 10 µM Cy3 picolyl azide (Click Chemistry Tools 1178–1)) for 1 hr at 37 °C in the dark. Two negative controls were included: samples with DBF omitted from the in-cell labeling reaction and cells with Cy3 picolyl azide omitted from the in situ Click reaction. Following the Click reaction, the coverslips were washed three times for 5 min each in wash buffer (0.1% Triton, 1 mg/mL BSA in PBS). Next, coverslips were incubated in wash buffer supplemented with 100 ng/mL DAPI for 30 min at 37 °C and then washed twice with wash buffer for 5 min each. The coverslips were mounted with fluoromount G and imaged using a Deltavision Elite widefield fluorescence microscope (GE). DAPI was imaged with 10% transmittance with an exposure of 0.05 seconds. Cy3 azide was visualized in the TRITC channel with 32% transmittance and an exposure of 0.2 s. Z stacks were collected of approximately 51 images, 0.2 µm apart for a total thickness of 10.2 µm.

rRNA-depleted RNAseq libraries were prepared using an RNA HyperPrep Kit (KAPA / Roche). A total of 100 ng of RNA was input into the procedure and fragmented for 3.5 min at 94 °C. Libraries were amplified using 14 PCR cycles.

Libraries were sequenced using paired end sequencing (2x150 bp) on a NovaSeq high-throughput sequencer (Illumina) at the University of Colorado Genomics Core Resource. Between 20 and 40 million read pairs were sequenced for each sample.

Library adapters were removed using cutadapt 2.1 (Martin, 2011). Transcript abundances in RNAseq data were quantified using Salmon 0.11.1 (Patro et al., 2017) and a human genome annotation retrieved from GENCODE (https://www.gencodegenes.org/, GENCODE 28). Gene abundances were then calculated from these transcript abundances using tximport (Soneson et al., 2015), and genes whose abundance in input and streptavidin-pulldown samples were identified using DESeq2 (Love et al., 2014).

To quantify the relative abundances of pre-mRNA (including introns) and mature mRNA (all introns spliced), a custom fasta file was supplied to Salmon which contained two versions of every transcript, one with all introns remaining and one with all introns removed. The custom fasta file was generated using this script: https://github.com/rnabioco/rnaroids/blob/master/src/add_primary_transcripts.py (Riemondy, 2018). Salmon then assigned reads competitively to these transcripts.

C2bbe1 cells were seeded on PDL coated glass coverslips (neuVitro) within 12-well plates at high confluency. Monolayers were allowed to differentiate for 7 days, and media was changed every 3–4 days. Following PBS wash, cells were fixed with 3.7% formaldehyde at room temperature for 15 min. After another PBS wash, cells were permeabilized for 30 min at room temperature with 0.3% Triton-X (VWR 80503–490) in PBS. Primary antibodies with various dilutions (Ezrin, Cell Signaling Technology 3145 S, 1:250; NaK ATPase, DSHB A5-s, 1:50; TOM20, ProteinTech 11802–1-AP, 1:250) were diluted in PBS containing 1% BSA and 0.3% Triton-X for 1 hr at 37 °C. Cells were then rinsed three times with PBS before adding 100 ng/mL DAPI and fluorescent secondary antibody (Cell Signaling Technology 4413 S) diluted 1:500 in PBS containing 1% BSA and 0.3% Triton-X for 1 hr at room temperature. When imaging for mitochondria, 100 ng/mL MitoView Green (Biotium 70,054 T) is added at this point as well. Following PBS wash, cells were mounted onto slides with Fluoromount G (SouthernBiotech 0100–01) and sealed with nail polish. Slides were imaged on a widefield DeltaVision Microscope at 60 X (GE) with consistent laser intensity and exposure times across samples. DAPI was imaged with 10% transmittance with an exposure of 0.25 s. Antibodies are visualized in the TRITC channel with 5–32% transmittance and an exposure of 0.3–0.5 s. Z stacks were collected of approximately 63 images, 0.2 µm apart for a total thickness of 12.6 µm.

To assay RNA localization of endogenous genes, the R script Oligostan (Tsanov et al., 2016) was used to design primary probes for each RNA of interest (Supplementary file 2). Probes were 26–32 nt in length with 40–60% GC content and repetitive sequences were avoided. Primary probes were designed with excess sequence for hybridization to a fluorescence FLAP sequence.

FLAP Y (TTACACTCGGACCTCGTCGACATGCATT) covalently attached to Cy3 on both ends was used for all smiFISH experiments targeting endogenous RNA. 1 X pmol of Primary probes and 1.2 X pmol FLAP sequences were hybridized in NEB buffer 3 in a thermocycler. The reaction was heated to 85 °C for 3 min, cooled to 65 °C for 3 min then held at 25 °C yielding hybridized smiFISH probes specific to the endogenous RNA target.

Customized Stellaris smFISH probes were designed and ordered using the freely available probe designer software by Biosearch Technologies. Probes that were designed across the open reading frame of Firefly luciferase were used to assay localization of Firefly reporter constructs. Probes were resuspended in water to a final concentration of 12.5 μM.

Generally, FISH experiments were performed as previously described (Arora et al., 2022b). CAD cells expressing reporter constructs were induced with 2 μg/mL doxycycline for 48 hr. Cells were then plated on PDL coated glass coverslips (neuVitro) within 12-well plates at approximately 2.5×10⁴ cells per well in full growth media. Cells were allowed to attach for 2 hr before changing to serum depleted media supplemented with 2 μg/mL doxycycline for 48 hr.

C2bbe1 cells expressing reporter constructs were seeded on PDL coated glass coverslips at high density and allowed to differentiate into monolayers for 7 days in full growth media supplemented with 2 μg/mL doxycycline. Media was changed every 3–4 days. Cells were washed once with PBS before being fixed with 3.7% formaldehyde (Fisher Scientific) for 10 min at room temperature. Following 2 PBS washes, cells were permeabilized with 70% Ethanol (VWR) at 4 °C for 6–8 hr or at room temperature for 2 hr. Cells were incubated in smFISH wash buffer (2 x SSC with 10% formamide) at room temperature for 5 min. Per coverslip, 2 μL of Stellaris FISH Probes, or hybridized smiFISH probes were added to 200 μL of hybridization Buffer (10% dextran sulfate, 10% formamide in 2 X SSC). Using a homemade hybridization chamber made from an empty tip box with wet paper towels and parafilm, coverslips were incubated cell side down in the hybridization buffer overnight at 37 °C. Coverslips were washed with wash buffer in fresh 12-well plates cell side up for 30 min at 37 °C in the dark. A total of 100 ng/mL DAPI was diluted in wash buffer and added to the cells in the dark at 37 °C for 30 min. Cells were then washed for 5 min at room temperature with wash buffer.

Coverslips were then mounted onto slides with Fluoromount G and sealed with nail polish. Slides were imaged at 60 X on a widefield DeltaVision Microscope with consistent laser intensity and exposure times across samples. DAPI was imaged with 10% transmittance with an exposure of 0.05 s. FISH probes were visualized in the TRITC channel with 100% transmittance and an exposure of 0.15 s. Z stacks were collected of approximately 68 images, 0.2 µm apart for a total thickness of 13.6 µm.

FISH-quant (Mueller et al., 2013) was used to identify smFISH puncta within CAD neurites and somas through use of its built-in Gaussian filtering and spot detection. All spot thresholding parameters were set to exclude gross outliers. The detection settings were the same for every soma and neurite imaged within an experiment. Approximately ten cells were analyzed for each sample. The mean number of Firefly luciferase transcripts in neurites compared to soma was calculated by FISH-quant mature mRNA summary files. Cell outlines for somas and neurites were manually drawn within the FISH-quant software. Usually cell outlines were visible in the FISH channel with high brightness. Outlines were drawn for proximal neurites that did not overlap with other cell’s neurites, protruded across the field of view, and were easily associated with the correct soma.

Images of epithelial monolayers were analyzed similarly with cell outlines manually drawn and approximated from GFP expression. Despite imaging a monolayer, smFISH spots were only quantified from outlined cells entirely within the field of view with sufficient RNA expression. Instead of using summary files, all detected spot files were used to ensure thresholding included spots from positive controls while excluding spots detected in negative control samples in which the RNA species (e.g. Firefly luciferase) being probed was not expressed.

Firefly luciferase reporter transcript puncta were thresholded until an expected expression of 100–1000 spots per cell was achieved. smFISH puncta counts of different transcripts were expected to approximately reflect relative expression values calculated from RNA sequencing.

Images were collected using a widefield Deltavision Elite microscope at ×60 magnification. All images were deconvoluted by Deltavision’s default software, SoftWorx. Images of monolayers were collected as large stacks (40–70 images) with many slices 0.2 μm apart from each other. FIJI (Schindelin et al., 2012) was used to filter all images in a stack using the default ‘rolling ball’ algorithm before being average or max projected through the XY, XZ, and YZ planes by the XYZ projection plugin (Omer et al., 2018). Average projections were used for all protein signal while max projections were used for smFISH signals. Projections were brightness and contrast matched for all relevant samples. RGB profile lines were drawn and analyzed with RGB profiler (Christophe Laummonerie, 2004).

CAD neurites were also imaged as stacks (20–50 images) with many slices 0.2 μm apart from each other. Images were filtered using the default ‘rolling ball’ algorithm and max projected through the XY.

Genes that were translationally down regulated upon mTOR inhibition (Hsieh et al., 2012) were assayed for pyrimidine-rich stretches (at least 9 pyrimidines in a 10 nucleotide window) across their longest annotated 5′ UTRs. The position of these PRTEs was defined as the most 5′ ’ start position of pyrimidine richness. If the position of the motif was 1, or the first nucleotide in the 5′ UTR, the PRTE could be more accurately defined as a 5’ TOP motif. Because this manuscript demonstrates that these pyrimidine-rich motifs can regulate localization in addition to translation, we refer to these motifs more broadly as pyrimidine-rich regulatory elements (PRREs).

Transcript abundances were calculated by salmon v0.11.1 (Patro et al., 2017) using Gencode genome annotations. Transcript abundances were then collapsed to gene abundances using txImport (Soneson et al., 2015). Localization ratios were calculated for each gene as the log₂ of the ratio of neurite/soma normalized counts for a gene produced by DESeq2 using the salmon/tximport calculated counts (Love et al., 2014). Genes were required to have a minimum of 10 counts in any sample for analysis.

Mouse-specific siRNAs were ordered from IDT for each Larp target. siRNAs were transfected into cells with Lipofectamine RNAiMAX (Thermo Fisher, 13778100). Briefly, 0.5 μL of 10 μM siRNA was combined with 50 μL Optimem media and 1.5 μL RNAiMAX reagent for roughly 1.0×10⁵ cells in one well of a 24-well plate. CAD cells were transfected in the presence of doxycycline to induce expression of reporter constructs. Media was replaced 24 hr after transfection with full growth media supplemented with 2 μg/mL doxycycline. 24 hr following media change, cells were plated on PDL-coated coverslips (if imaging) and after attaching, media was changed again for serum-free media with 2 μg/mL doxycycline to induce neurite outgrowth for 48 hr. Negative control siRNAs from IDT were used to control for effects due to transfection. For all proteins tested, knockdown of expression was confirmed by qPCR normalized to HPRT. For Larp1, immunoblots confirming protein knockdown were performed with mouse monoclonal Larp1 antibody (Santa Cruz sc-515873).

SYBR green (Biorad) mastermix was used to analyze RNA knockdown of Larp siRNAs and LARP1 rescue in C2bbe1 cells. RNA was collected from siRNA transfected CAD cells as detailed above. Total RNA was collected from C2bbe1 monolayers plated at high confluency in 12 well plates. For C2bbe1 cells, media containing 2 μg/mL doxycycline was changed every 3–4 days for 7 days of differentiation. Following differentiation of both cell types, media was removed from cells and replaced with RNA Lysis Buffer (Zymo Research, R1060-1). Cells were lysed at room temperature for 15 min on a rocker. RNA was collected using Quick RNA micro columns (Zymo Research, R1051). The provided on-column DNAse treatment was performed for 15 min at room temperature. 500 ng of purified RNA was used to synthesize cDNA using iScript Reverse Transcriptase Supermix (BioRad, 1708841). cDNA was combined with gene specific primers and iTaq Universal SYBR green master mix (BioRad, 1725122). Gene expression of siRNA targeted Larps was normalized to HPRT before being compared to the expression in negative control siRNA targeted samples.

Expression of Larp1 siRNA treated and CRISPR/Cas9 KO cell lines were determined by immunoblot analysis. Lysates were collected by scraping cells into 100 μL RIPA buffer, incubated on ice for 15 min then sonicated for 30 s. Lysates were stored at –80 °C. Lysates were denatured with sample buffer (Invitrogen NP0008) and boiling at 100 °C for 5–10 min. Denatured lysates were separated by PAGE on 4%–12% Bis-Tris gradient gels along with Flash Protein Ladder (Gel Company FPL-008) using MOPS SDS NuPAGE Running Buffer (Thermo Fisher, NP0001) at 200 V for 1 hr. Gels were transferred to PVDF membranes using an iBlot2 dry transfer device (Thermo Fisher, IB24002). Total protein was assayed by incubation with Ponceau staining (Sigma-Aldrich P7170), which was later destained and imaged on a Sapphire biomolecular imager (Azure Biosystems). Blots were then blocked with agitation in 5% milk powder (Sigma) in PBST (PBS +0.01% Tween) for 1 hr at room temperature. Blots were washed three times, 5 min each in PBST at room temperature with agitation before being incubated in 1:100 LARP1 mouse monoclonal (Santa Cruz) primary antibody in 2% BSA (Research Products International A30075) in PBST overnight at 4 °C with agitation with 1:5000 beta actin mouse monoclonal antibody (Sigma A5441) in 2% milk powder in PBST for 2 hr at room temperature with agitation or with 1:10,000 H3 mouse monoclonal antibody (Abcam ab10799) in 2% milk powder in PBST for 2 hr at room temperature with agitation. Blots were washed three times, 5 min each at room temperature with agitation and then incubated with Anti-mouse IgG HRP-conjugated secondary antibody (Cell signaling 7076 S) in 2% milk powder in PBST and incubated at room temperature for 2 hr. Blots were washed again in PBST and visualized using WesternBright HRP Substrate Kit (Advansta, K-12043-D10) and Sapphire biomolecular imager (Azure Biosystems) set to collect chemiluminescent signal.

Generally, neuronal fractionation experiments were performed as previously described (Arora et al., 2021). To fractionate CAD cells into soma and neurite fractions, CAD cells were plated on porous, transwell membranes that had a pore size of 1.0 micron (Corning 353102). One million CAD cells were plated on each membrane. These membranes fit in one well of a six well plate, and three membranes were combined to comprise one single biological replicate. After allowing the cells to attach, the cells were induced to differentiate into a more neuronal state through the withdrawal of serum. Cells were then differentiated on the membranes for 48 hr.

To fractionate the cells, the membranes were washed one time with PBS. One mL of PBS was added to the top (soma) side of the membrane. The top side of the membrane was then scraped repeatedly and thoroughly with a cell scraper and the soma were removed into an ice-cold 15 mL falcon tube. After scraping, the membranes were removed from their plastic housing with a razor blade and soaked in 500 µL RNA lysis buffer (Zymo R1050) at room temperature for 15 min to make a neurite lysate. During this time, the 3 mL of soma suspension in PBS was spun down and resuspended in 300 μL PBS. Of this soma sample, 100 µL was then carried forward for RNA isolation using the Zymo QuickRNA MicroPrep kit (Zymo R1050). In parallel, RNA was also isolated from the 500 μL neurite lysate using the same kit. Typically, the neurite RNA yield from the six combined membranes was 100–300 ng. The efficiency of the fractionation was assayed by Western blotting using anti-beta-actin (Sigma A5441, 1:5000) and anti-Histone H3 (Abcam ab10799, 1:10000) antibodies. cDNA was created from 100 ng of RNA from each fraction using LunaScript RT supermix kit (NEB #E3010) following the manufacturer’s instructions.

All Firefly luciferase reporter constructs were driven by a bi-directional tet-On promoter that also induces Renilla luciferase expression. Renilla was never modified and acts as an internal control in these experiments. Expression of Firefly and Renilla luciferase transcripts was induced by incubating cells with 2 μg/mL doxycycline for 48 hr. qPCR was performed in BioRad thermal cyclers with taqman mastermix (Life Technologies 4444557) and probes specific to each luciferase but with different fluorophores (VIC and FAM). The relative expression of Firefly and Renilla luciferase was quantified in the same sample using Taqman qPCR then compared between neuronal fractions to calculate a log₂ normalized localization ratio.

Endogenous mouse Rpl7 and Rps28 localization was also assayed by Taqman qPCR. Probes specific to these genes were designed with fluorophores FAM. A non-localized endogenous transcript, Tsc1 (VIC) was used as an internal control. The relative expression of these genes was quantified in the same sample to calculate log₂ normalized localization ratios (LR). The endogenous Rpl7 and Rps28 LRs were compared to Firefly expression also normalized to Tsc1.

All qPCRs were performed with triplicate technical replicates averaged to produce a single biological replicate. qPCR experiments usually have three to four biological replicate samples included.

CAD cells were siRNA treated with various targeting or negative control siRNAs as detailed above. Differentiation occurred in a single well of a 6-well plate. An EVOS imager (Thermo Fisher) at 20 X was utilized to take tiled brightfield images across the entire well. Images were analyzed in FIJI (Schindelin et al., 2012) using high contrast settings and NeuronJ (Meijering, 2004) was used to measure neurite lengths. The same images were used to count cells with and without neurites for qualitative morphological analyses.

Moor et al., 2017, Mouse LCM RNAseq (GSE95416: GSE95276; PRJNA377081).

Dermit et al., 2020, projection localization RNA seq (E-MTAB-8470, E-MTAB-9520, and E-MTAB-9636).

Cassella and Ephrussi, 2022, Drosophila Follicular Cell LCM RNAseq (E-MTAB-9127).

All materials including cell lines and reagents are available upon request.

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

We thank members of the Taliaferro lab for helpful discussions regarding experiments and analyses. We further thank Rytis Prekeris and Peter Dempsey for helpful discussions and advice regarding epithelial cell biology. Portions of the figures in this manuscript were created using Biorender. This work was funded by the National Institutes of Health (R35-GM133385 to JMT), the WM Keck Foundation (JMT) and the RNA Bioscience Initiative at the University of Colorado Anschutz Medical Campus (JMT, RG). This work was further supported by a Predoctoral Training Grant in Molecular Biology NIH-T32-GM008730 to RG.

1. Preprint posted: April 14, 2022 (view preprint)
2. Received: May 6, 2022
3. Accepted: January 24, 2023
4. Version of Record published: March 3, 2023 (version 1)

© 2023, Goering 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.