RNA degradation eliminates developmental transcripts during murine embryonic stem cell differentiation via CAPRIN1-XRN2

Abstract

Embryonic stem cells (ESCs) are self-renewing and pluripotent. In recent years, factors that control pluripotency, mostly nuclear, have been identified. To identify non-nuclear regulators of ESCs, we screened an endogenously labeled fluorescent fusion-protein library in mouse ESCs. One of the more compelling hits was the cell-cycle-associated protein 1 (CAPRIN1). CAPRIN1 knockout had little effect in ESCs, but it significantly altered differentiation and gene expression programs. Using RIP-seq and SLAM-seq, we found that CAPRIN1 associates with, and promotes the degradation of, thousands of RNA transcripts. CAPRIN1 interactome identified XRN2 as the likely ribonuclease. Upon early ESC differentiation, XRN2 is located in the nucleus and colocalizes with CAPRIN1 in small RNA granules in a CAPRIN1-dependent manner. We propose that CAPRIN1 regulates an RNA degradation pathway operating during early ESC differentiation, thus eliminating undesired spuriously transcribed transcripts in ESCs.

Keywords: CAPRIN1; RNA degradation; RNA stability; XRN2; differentiation; embryonic stem cells; pluripotency; stress granules.

Graphical abstract

Figure 1

CAPRIN1 changes expression and localization during early ESC differentiation (A and B) Endogenously tagged CAPRIN1-YFP in undifferentiated mouse ESCs (A) and after 4 days of retinoic acid-induced differentiation (B). Scale bars, 10 μm. (C) CAPRIN1-YFP (green) colocalizes with CAPRIN1 antibodies (red). Right panel: merge. Scale bars, 10 μm. (D) Western blot for OCT4 (top), CAPRIN1 (middle), and GAPDH (bottom) in undifferentiated ESCs (left) and after 3-day (middle) or 5-day (right) retinoic acid treatment. The results from one of the three independent experiments. (E) Relative expression (using real-time RT-PCR) of Caprin1 mRNA in undifferentiated (left) and RA-induced (right) ESCs, n = 3. (F) Same as (E) in undifferentiated (left) and 3-day-old (middle) and 5-day-old embryoid bodies (right), n = 3. (G) CAPRIN1-YFP localization in interphase (left), mitosis (middle), and immediately after mitosis (right). Scale bars, 5 μm. (H) CAPRIN1-YFP ESCs (green, left) stained with anti-CAPRIN1 antibodies (middle, red) following 1 h treatment with sodium arsenite (200 μM, 2 h). Scale bars, 5 μm. (I) Immunofluorescence of non-stressed CAPRIN1-YFP ESCs (left, green) using TIA1 antibodies (middle, red). Right: merge. Arrows point to the CAPRIN1 cytoplasmic aggregates, which show no TIA1 staining. Scale bars, 5 μm.

Figure 2

Aberrant differentiation of Caprin1-KO ESCs (A) Western blot for CAPRIN1 in WT and two Caprin1-KO clones (KO2; KO4) ESCs (top) and RA-induced ESCs for 4 days (bottom). The results from one of the three independent experiments. (B) Relative expression (RT-PCR) of Caprin1 mRNA in the WT (blue) and KO4 (gray) ESCs, n = 3. (C) Relative expression (RT-PCR) of Nanog, Oct4, Sox2, and Klf4 in WT (blue) and KO4 (gray) ESCs. (None of the changes are statistically significant, U test), n = 3. (D) Proliferation rates in WT (blue) and KO4 (red) ESCs. Shown are averages of four independent experiments. Error bars indicate standard deviations. Changes are significant (p < 0.005, U test). (E) Relative expression (RT-PCR) of Oct4, Nodal, Nestin, and Sox17 in WT (blue) and KO4 (gray) ESCs after 4 days of RA-induced differentiation, n = 3. (F) Number of beating cardiomyocytes differentiated from WT (left), KO4 (middle), and KO-Caprin1-add-back clones (right). (G and H) Volcano plot of RNA-seq comparing WT and Caprin1-KO cells in undifferentiated ESCs (G) and in 4-day RA-induced cells (H). (I) Volcano plots of RNA-seq results of three independent experiments comparing either WT versus Caprin1-KO ESCs (left) or WT versus Xrn2-KD ESCs (right). (J) Venn diagram of the differentially expressed genes between WT versus Caprin1-KO ESCs and WT versus Xrn2-KD ESCs. (K) GO analysis of the downregulated genes shared between the teratomas formed by the Caprin1-KO ESCs and the Xrn2-KD ESCs. Brain-related genes are dramatically enriched.

Figure 3

CAPRIN1 interacts with developmental transcripts in early differentiating ESCs (A) Gene ontology (GO) enrichment analysis of CAPRIN1-interacting RNA transcripts specific for undifferentiated ESCs, identified by RIP-seq. Categories are sorted according to fold enrichment over genomic average. All categories are significant (FDR < 0.005). (B) Same as (A) for RA-induced ESCs. (C) Validation of RIP-seq by RIP-RT-PCR for selected transcripts in undifferentiated (blue) and RA-induced (orange) ESCs. IgG (gray) was used as control. (^∗p < 0.05, U test), n = 3. (D) Motif enrichment analysis for CAPRIN1. All RIP-seq data were analyzed for enriched motifs. The motif was identified specifically in undifferentiated ESCs, suggesting a mechanistic switch for CAPRIN1 interaction with RNAs during differentiation. (E) Secondary structure enrichment analyses of CAPRIN1-bound RNAs in undifferentiated ESCs (blue); RA-induced ESCs (yellow), both (green) and a background dataset (red). Only undifferentiated ESCs are highly enriched for CAPRIN1-interacting RNAs with secondary structure (p < 2 × 10⁻¹⁶).

Figure 4

CAPRIN1 is involved in global RNA stability (A and B) Cumulative fraction analysis of SLAM-seq in WT (cyan) and Caprin1-KO (red) undifferentiated ESCs (A) and in RA-induced ESCs (B). p << 10⁻¹⁰, n = 3. (C and D) RNA half-life in WT (y axis) versus Caprin1-KO (x axis) in undifferentiated ESCs (C) and RA-induced ESCs (D). (E–G) RNA decay rates (1/h; log₂) for 3,132 fitted genes in cells treated with α-amanitin for 8 h. (E) wild-type (x axis) versus Caprin1-KO2 (y axis), (F) wild-type (x axis) versus Caprin1-KO4 (y axis), and (G) Caprin1-KO2 (x axis) versus Caprin1-KO4 (y axis). Color scale indicates the density of data (yellow = high, blue = low). (H–J) RNA expression levels (TPM; log₂) before α-amanitin treatment (0 h) for 3,132 fitted genes in cells treated with α-amanitin for 8 h. (H) Wild-type (x axis) versus Caprin1-KO2 (y axis), (I) wild-type (x axis) versus Caprin1-KO4 (y axis), and (J) Caprin1-KO2 (x axis) versus Caprin1-KO4 (y axis). Color scale indicates the density of data (yellow, high; blue, low).

Figure 5

CAPRIN1 interacts with XRN2 (A) LC-MS/MS of CAPRIN1-interactome. (B) GO analysis of CAPRIN1-interactions in RA-induced ESCs. (C) Venn diagram of CAPRIN1-interacting partners in ESCs (blue) and RA-induced ESCs (pink). (D) CoIP of CAPRIN1 and XRN2. Shown are cytoplasmic (Cyto) and nuclear (Nuc) fractions of non-stressed (-ARS) and arsenite-treated (ARS, 200 μM, 2 h) ESCs pulled-down using CAPRIN1 antibodies (top) and blotted using XRN2 antibodies (second row from top). α-Tubulin (α-tub, third row) and histone H3 (H3, fourth row) were used to validate enrichment of cytoplasmic and nuclear fractions, respectively. Heavy- and light-chain IgGs are shown below. The results from one of the three independent experiments.

Figure 6

CAPRIN1 is required for XRN2 relocalization into SGs (A) Immunofluorescence (IF) for XRN2 (red) in non-stressed ESCs (top) and in sodium arsenite-treated ESCs (bottom). Blue: DAPI; green: CAPRIN1-YFP; right: merge. Scale bars, 10 μm. (B) Same as (A) in non-stressed RA-induced ESCs. XRN2 can be found in small cytoplasmic foci (top panels) or dispersed in the cytoplasm (bottom panels). (C) CAPRIN1 is required for XRN2 localization in SGs. Shown are IF images for XRN2 (red) in stressed (ARS-treated) WT (top) and stressed Caprin1-KO ESCs (bottom). (D) CAPRIN1 is not entirely essential for the formation of SGs. Shown are IF images of XRN2 (red, top) and G3BP1 (green, bottom) in stressed (ARS-treated) Caprin1-KO ESCs. SGs are formed, although to a lesser extent, but XRN2 remains nuclear. (E) Quantification of SG formation in Caprin1-KO ESCs. G3BP1-positive SGs were quantified in WT (blue) and Caprin1-KO (gray) ESCs following arsenite treatment. ^∗p < 0.015, t test, n = 90 (WT) and n = 94 (Caprin1-KO).