Subcytoplasmic location of translation controls protein output

Highlights

• •The majority of genes generate transcripts with biased subcytoplasmic localization
• •Gene architecture and RNA-binding proteins influence cytoplasmic mRNA localization
• •mRNAs that encode transcription factors are strongly enriched in TIS granules
• •A change in mRNA localization changes protein abundance

Summary

The cytoplasm is highly compartmentalized, but the extent and consequences of subcytoplasmic mRNA localization in non-polarized cells are largely unknown. We determined mRNA enrichment in TIS granules (TGs) and the rough endoplasmic reticulum (ER) through particle sorting and isolated cytosolic mRNAs by digitonin extraction. When focusing on genes that encode non-membrane proteins, we observed that 52% have transcripts enriched in specific compartments. Compartment enrichment correlates with a combinatorial code based on mRNA length, exon length, and 3′ UTR-bound RNA-binding proteins. Compartment-biased mRNAs differ in the functional classes of their encoded proteins: TG-enriched mRNAs encode low-abundance proteins with strong enrichment of transcription factors, whereas ER-enriched mRNAs encode large and highly expressed proteins. Compartment localization is an important determinant of mRNA and protein abundance, which is supported by reporter experiments showing that redirecting cytosolic mRNAs to the ER increases their protein expression. In summary, the cytoplasm is functionally compartmentalized by local translation environments.

Introduction

In polarized cells such as neurons, intestinal epithelial cells, or cells of the early fly embryo, the majority of mRNAs have a distinct spatial localization pattern.^1,2,3,4,5 mRNA localization enables the local control of protein production.^6,7,8 In non-polarized cells, mRNA localization has primarily been studied for membrane proteins.^9,10,11,12 Whereas the rough endoplasmic reticulum (ER) is established as a major site of local protein synthesis for membrane and secretory proteins,^9,10,13 the cytoplasm is compartmentalized by additional membrane-bound and membraneless organelles.^14,15,16,17 These compartments may enable the generation of unique biochemical translation environments, which have been suggested to be crucial for protein interaction partner selection during protein synthesis.^16,18,19,20 However, it is currently largely unknown whether the location of protein synthesis also matters for protein output.

TIS granules (TGs) represent one such unique translation compartment, which promotes the co-translational formation of protein complexes. Both endogenous and overexpressed TIS11B promote the formation of specific protein complexes when mRNAs are translated in TGs.^16,19 TGs are formed by the RNA-binding protein (RBP) TIS11B, together with its bound mRNAs. TIS11B mRNA is ubiquitously expressed, suggesting that TGs are widespread. TGs are present under steady-state cultivation conditions and form a network-like structure that is intertwined with the rough ER.^16,21 To investigate the broader biological significance of TGs, we determined the mRNAs enriched in TGs, the neighboring rough ER, and the surrounding cytosol.

Because TIS11B protein is present in cells in two states (Figure 1A), as soluble cytosolic protein and as phase-separated TG network,^16,21 we decided to use fluorescent particle sorting²³ to identify TG-enriched mRNAs. We also applied fluorescent particle sorting to isolate ER-enriched mRNAs and extracted cytosolic mRNAs using digitonin. We analyzed genes that encode non-membrane proteins and found more than 3,600 that have transcripts enriched in one of the three compartments. mRNAs enriched in each compartment share similar mRNA architectures, which differ strongly between compartments. Compartment-enriched mRNAs also differed significantly in production and degradation rates as well as in the functional classes and expression levels of their encoded proteins. TIS11B knockout (KO) and reporter experiments support a model by which a combinatorial code based on mRNA architecture features, together with 3′ UTR-bound RBPs TIS11B, TIA1/L1, and LARP4B, correlated with the compartment-biased mRNA localization pattern. Intriguingly, we observed that redirecting cytosolic mRNAs to the ER controls protein expression, which indicates that protein abundance is regulated by the location of translation in the cytoplasm.

Results

Approach to determine subcytoplasmic mRNA localization

We set out to identify mRNAs that are localized in non-polarized human HEK293T cells under steady-state cultivation conditions. We focused on three major unenclosed cytoplasmic compartments—TGs, a condensate network formed by the RBP TIS11B, the cytosolic surface of the ER, and the soluble part of the cytoplasm known as the cytosol (Figure 1B). For simplicity, we consider here the sum of the three compartments as the universe of cytoplasmic mRNAs.

To identify TG-enriched (TG+) and ER-enriched (ER+) mRNAs, we performed fluorescent particle sorting followed by RNA sequencing (RNA-seq). To label TGs and rough ER, we co-transfected cells with mCherry-TIS11B and GFP-SEC61B, respectively. After flow-cytometry-based sorting of fluorescent particles, we used confocal microscopy and western blot analysis to assess the purity of the particles (Figures S1A–S1E). We modestly overexpressed mCherry-TIS11B compared with its endogenous levels (Figure S1C), which resulted in approximately 30% of cells forming TGs. This amount was chosen because 25%–30% of HEK293T cells form TGs from endogenous TIS11B. ER particles did not contain mCherry-TIS11B. TG particles contained GFP-SEC61B, but they contained 13-fold more mCherry-TIS11B than ER particles (Figures 1C and S1C–S1E). As TGs are defined by the presence of TIS11B,¹⁶ we reasoned that the strong overrepresentation of TIS11B in TG particles would allow us to identify relative enrichments of mRNAs between the compartments.

To isolate cytosolic mRNAs, we used digitonin extraction.²⁴ The extracted cytosol was not contaminated by nuclei or the ER, but it contained TIS11B, which was expected because soluble TIS11B is known to be present in the cytosol (Figure S1C). We performed RNA-seq on biological replicate samples to determine the mRNA composition in the three fractions and focused our analysis on protein-coding mRNAs (Figure S1F).

mRNAs that encode membrane or secretory proteins largely localize to the ER membrane

We investigated whether the relative mRNA transcript distribution differs across the three compartments. For each gene, we determined a compartment-specific localization score (LS). This score is calculated using the reads per kilobase per million mapped reads (RPKM) value obtained in each of the three compartments, respectively, and dividing it by the sum of the RPKM values in all three compartments. Thus, each gene is assigned three LSs that correspond to the fraction of its transcripts localizing to each of the three compartments: TGs, the ER, and the cytosol.

First, we focused on mRNAs that encode membrane or secretory proteins, which are known to be translated on the ER. In line with previous analyses, we find preferential partitioning of mRNAs encoding membrane/secretory proteins in the ER samples (Figure S1G). To validate our compartment isolation method, we compared it with datasets from three alternative isolation methods. We consider 69% (N = 1,476) of membrane/secretory proteins to be enriched on the ER (Figure S1H; Table S1) and we detected between 80% and 90% overlap between our data and previous methods (Figures S1I and S1J). These results strongly support the validity of our purification strategy for mRNAs that encode membrane/secretory proteins.

Half of the genes that encode non-membrane proteins have a biased cytoplasmic transcript distribution

For mRNAs that encode non-membrane proteins, we observed that their LSs were more evenly distributed across the compartments (Figure S1G). To identify absolute differences in mRNA distribution, the relative size of each compartment needs to be considered. However, this parameter is currently unknown.

Therefore, we instead calculated the relative enrichment of mRNAs within each compartment. We considered an mRNA to be compartment-enriched if its mean LS across biological replicates was at least 1.25-fold higher than the median LS of the compartment samples (Figures 1D–1F). Based on this criterion, we identified 1,246 TG+ mRNAs, 919 ER+ mRNAs, and 1,481 mRNAs enriched in the cytosol (CY+), which were non-overlapping (Figures 1D–1F; Table S1). The remaining 3,369 mRNAs were not enriched in a single compartment and were considered to have an unbiased localization pattern (Figures 1D–1F and S1K). The distribution of LSs of TG+, ER+, or CY+ mRNAs is significantly different from the LSs of mRNAs with unbiased localization patterns (Figures 1D–1F). Because LSs across the compartments sum to 1, an mRNA enriched in one compartment is relatively de-enriched in the other two (Figure S1L). Based on this strategy, 52% of genes that encode non-membrane proteins have transcripts that are significantly enriched in one of the three subcytoplasmic compartments in steady-state conditions.

As a recent study also analyzed the relative distribution of mRNA transcripts across subcellular compartments, we compared our data with their results.²⁵ Although their dataset was generated by density gradient centrifugation in a different cell line, the two datasets strongly agreed in a qualitative and quantitative manner (Figure S1M), suggesting that our isolation method as well as our strategy to define compartment-enriched mRNAs are valid. As non-membrane protein-encoding mRNAs with biased transcript distributions in the cytoplasm have not been systematically characterized, we focused all subsequent analyses on mRNAs that encode non-membrane proteins.

Validation of compartment-enriched mRNAs by smRNA-FISH

We further validated the mRNAs designated as compartment-enriched by performing single-molecule RNA-fluorescence in situ hybridization (smRNA-FISH) on endogenous mRNAs.²⁶

 Candidates for validation were primarily chosen based on their respective LSs, and most ranked in the top 10% of their respective compartments (Table S2). To distinguish between TG+ and ER+ mRNAs, we performed smRNA-FISH together with co-transfection of blue fluorescent protein (BFP)-TIS11B and GFP-SEC61B to simultaneously visualize mRNA puncta, TGs, and the rough ER (Figures 1G, 1H, and S2A–S2G). We considered an mRNA to have an unbiased localization pattern if its transcript distribution correlated with compartment size. As proxy for relative compartment size, we used the areas of the maximum projection of the fluorescent signals for each compartment and compared them to the whole-cell area. For unbiased mRNAs, we expect that 11% of transcripts localize to TGs and 29% of transcripts localize to the ER (Figures 1I and 1J).

For 3/3 TG+ mRNAs, we observed a significant enrichment of mRNA puncta in TGs but not on the ER (Figures 1G–1J, S2A, S2B, S2H, and S2I). For the five ER+ mRNAs tested, the mRNA puncta of 4/5 mRNAs were significantly enriched on the ER and, for all five, we observed a 2- to 4-fold higher fraction of mRNA puncta that colocalized with the ER compared with TGs (Figures 1H–1K and S2C–S2I)….