Noncoding RNA circBtnl1 suppresses self‐renewal of intestinal stem cells via disruption of Atf4 mRNA stability

Abstract

Intestinal stem cells (ISCs) at the crypt base are responsible for the regeneration of the intestinal epithelium. However, how ISC self‐renewal is regulated still remains unclear. Here we identified a circular RNA, circBtnl1, that is highly expressed in ISCs. Loss of circBtnl1 in mice enhanced ISC self‐renewal capacity and epithelial regeneration, without changes in mRNA and protein levels of its parental gene Btnl1. Mechanistically, circBtnl1 and Atf4 mRNA competitively bound the ATP‐dependent RNA helicase Ddx3y to impair the stability of Atf4 mRNA in wild‐type ISCs. Furthermore, ATF4 activated Sox9 transcription by binding to its promoter via a unique motif, to enhance the self‐renewal capacity and epithelial regeneration of ISCs. In contrast, circBtnl1 knockout promoted Atf4 mRNA stability and enhanced ATF4 expression, which caused Sox9 transcription to potentiate ISC stemness. These data indicate that circBtnl1‐mediated Atf4 mRNA decay suppresses Sox9 transcription that negatively modulates self‐renewal maintenance of ISCs.

A newly identified circRNA, CircBtnl1 inhibits the self‐renewal of mouse intestinal stem cells through indirect suppression of Sox9 activity.

Introduction

Rapid turnover and regeneration of the intestinal epithelium in mammals are supported by functionally distinct intestinal stem cells (ISCs) localized in intestinal crypts, which migrate upwards and differentiate into enterocytes, tuft, goblet or enteroendocrine cells (Rodriguez‐Colman et al, 2017; de Sousa & de Sauvage, 2019; Guiu et al, 2019). Lineage‐tracing studies have identified two distinct populations of ISCs: the rapidly dividing crypt base columnar cells marked by Lgr5, and a quiescent population of cells at the +4 cell position (Takeda et al, 2011). The development of intestinal organoids from single adult intestinal stem cell in vitro recapitulates the regenerative capacity of the intestinal epithelium (Lukonin et al, 2020). During regeneration, tissue pattern and homeostasis are restored through numerous signaling pathways, including Wnt (Janda et al, 2017; Yan et al, 2017), Notch (Serra et al, 2019), YAP (Ayyaz et al, 2019), BMP (Martyn et al, 2018; Yum et al, 2021), and Hedgehog (Degirmenci et al, 2018). Although the signaling requirements for maintaining stem cell function and crypt homeostasis have been well studied, the underlying mechanism of adult ISCs maintain their self‐renewal capacity still needs to be further explored.

Circular RNAs (circRNAs) are now widely recognized as a novel subset of endogenous RNAs that generated from exons or introns of parental genes by back‐splicing (Li et al, 2019). Although backsplicing is generally less efficient than linear splicing, circRNAs can accumulate in specific cell types in a temporally regulated manner owing to their high stability (Vo et al, 2019). Some circRNAs are abundant and evolutionarily conserved, and many circRNAs exert important biological functions by acting as microRNA or protein inhibitors (“sponges”), by regulating protein function or by being translated themselves. Cyrano could promote unusually efficient destruction of mature miR‐7, which enables circRNA Cdr1as to accumulate in the brain (Kleaveland et al, 2018). We previously showed that a circRNA cia‐cGAS associates with a DNA sensor cGAS in the nuclei of hematopoietic stem cells (HSC) to block its synthase activity, maintaining HSC hemeostasis (Xia et al, 2018). circPan3 can promote the self‐renewal of intestinal stem cells through IL‐13 produced by niche ILC2s (Zhu et al, 2019). We also demonstrated that circKcnt2 facilitates colitis resolution by inhibiting ILC3 activation (Liu et al, 2020). However, how circRNAs regulate ISC self‐renewal remains elusive.

The activating transcription factor 4 (ATF4) is a basic region leucine zipper transcription factor and plays pivotal roles in physiological responses to stress, including hypoxia (Longchamp et al, 2018), endoplasmic reticulum stress, amino acid deprivation (Kim et al, 2021), oxidation (Shi et al, 2019), and mitochondrial stress (Guo et al, 2020; Tremblay & Haynes, 2020). ATF4 was recently reported to contribute to functional expansion of HSCs during embryonic development. ATF4 deficiency impairs HSC function with aging‐like phenotype and alleviates leukemogenesis by regulating p16Ink4a and HIF1α (Sun et al, 2021). ATF4 was identified as a direct transcriptional target of RUNX1 in HSCs to confer resistance against apoptosis and survival advantage under stress conditions (Matsumura et al, 2020). Deletion of ATF4 significantly impairs hematopoietic development and reduces HSC self‐renewal in fetal liver by transcription control of Angptl3 (Zhao et al, 2015). To date, no studies have examined the functions of ATF4 in ISC biology. Here we identified a novel circRNA circBtnl1 (mmu circ: chr17: 34380961–34382488) that was highly expressed in ISCs. We showed that circBtnl1 knockout enhanced the self‐renewal capacity and epithelial regeneration. CircBtnl1 negatively regulated the self‐renewal maintenance of ISCs via the Ddx3y‐ATF4‐Sox9 axis.

Results

CircBtnl1 is highly expressed in intestinal stem cells

To explore the role of circRNAs in ISCs, we isolated small intestinal crypts from C57BL/6 mice and performed high throughput circRNA‐seq. We then selected top eight highly expressed circRNAs (circOgdh, circRasa2a, circBtnl1, circMed13I, circVapa, circSlc43a2, circCnot6I, circAlg12) in mouse ISCs and validated their high levels of expression in Lgr5‐GFP⁺ ISCs in comparison to Lgr5‐GFP⁻ cells through real‐time PCR (Fig 1A; Appendix Fig S1A). Using cDNA and genomic DNA from intestinal crypts as templates, circRNAs were amplified by divergent primers in cDNA, but not in genomic DNA (Appendix Fig S1B). Head‐to‐tail splice junction sites of these top eight circRNAs were confirmed by PCR and Sanger sequencing using divergent primers (Appendix Fig S1C). We further employed quantitative reverse transcription polymerase chain reaction (qRT‐PCR) to confirm that the eight selected circRNAs were resistant to RNase R (Appendix Fig S1D) and actinomycin D treatment (Appendix Fig S1E)，while the linear RNAs of GAPDH was substantially digested with RNase R and actinomycin D.

Figure 1. CircBtnl1 is highly expressed in intestinal stem cells.

• A
  A heatmap of circRNAs from small intestine crypts of C57BL/6 mice. The top eight highly expressed circRNAs (circOgdh, circRasa2a, circBtnl1, circMed13I, circVapa, circSlc43a2, circCnot6I, circAlg12) are listed.
• B
  Small intestine crypts were isolated from C57BL/6 mice for organoid culture. The top eight upregulated circRNAs were depleted in organoid with a shRNA strategy, followed by organoid formation. Organoid formation ratios are shown as mean ± SD， n = 3 mice. *P < 0.05, **P < 0.01, NS, not significant, by one‐tailed Student's t‐test.
• C
  Organoid images of small intestine (SI) (up) and colon (down) grown under circBtnl1 deletion and control. Scale bars，200 μm.
• D
  One‐week‐old mice were euthanized for longitudinal sections. A global view of the section is shown in the left panel and the indicated tissues are shown to the right. 1, Brain; 2, Heart; 3, Liver; 4, Lung; 5, Stomach 6, SI (Small Intestine). Scale bars, 500 μm (left panel); 100 μm (magnified panels).
• E
  Northern blot analysis of circBtnl1 in different tissues of 2‐month‐old mice, 18S RNA served as a loading control.
• F， G
  RNA fluorescence in situ hybridization of circBtnl1 in small intestine tissues (F) and organoid (G). Green region shows the distribution of circBtnl1 using antisense probe, gray region shows the nuclei staining by DAPI. For (F)，scale bars，5 μum and for (G)，scale bars，20 μm.
• H
  Relative distribution of circBtnl1 in mice ISCs determined by RT‐qPCR in different cell fractions. U1 was served as the nuclear RNA marker and GAPDH was served as the cytoplasmic RNA marker. Data are shown as mean ± SD. Data are representative of at least three independent experiments.

Source data are available online for this figure.

Intestinal organoids recapitulate the ability of intestinal tissue to regenerate and return to homeostasis following damage (Lukonin et al, 2020). We next depleted these eight selected circRNAs individually via using short hairpin RNA (shRNA) in ISCs, and established organoid assays to determine effects of their loss on the stemness of ISC cells. We found that depletion of circBtnl1 most significantly induced organoid formation (Fig 1B). circBtnl1, located on mouse chromosome 17, was formed by back‐splicing of Btnl1 transcripts from exon 4 to exon 8 (E4–E8; Appendix Fig S2A). In contrast, combining circRNA array and circBase data, host gene Btnl1 transcribed four circRNA transcripts (E4–E8; E4–E5; E4–E6 and E4–E7) due to variable cyclization (Appendix Fig S2B). Compared with other cycled transcripts, circBtnl1 transcript (E4–E8) was most abundant in mouse ISCs (Appendix Fig S2C). Many circRNAs exhibit tissue‐specific expression patterns, implying their specific function in various tissues. Next, we evaluated the function of circBtnl1 in ISCs. We found that circBtnl1 knockdown remarkably disrupted organoid formation capacity in SI and colon (Fig 1C). We then tested the expression of circBtnl1 in various tissues in mice by RNA fluorescence in situ hybridization (FISH) assay, we found that circBtnl1 was widely expressed in various tissues, with the small intestine (SI), colon and stomach ranking the top three highly expressed tissues (Fig 1D). These results were further validated by Northern blot with a single transcript (Fig 1E). In addition, circBtnl1 was localized exclusively in the cytoplasm rather than in the nucleus in intestinal crypts (Fig 1F) and small intestine organoids (Fig 1G). These observations were further validated by nuclear‐cytoplasmic separation assay (Fig 1H). Collectively, these results indicate that circBtnl1 is highly expressed in mouse ISCs.

CircBtnl1 knockout promotes the self‐renewal of ISCs

To determine the physiological role of circBtnl1 in the regulation of ISCs, we generated circBtnl1 knockout (circBtnl1 ^−/−) mice by deletion of its intron 8 via CRISPR/Cas9 technology (Appendix Fig S3A). Deletion of 793 bp fragment containing the intron 8 was confirmed by genotyping and PCR in ISCs (Appendix Fig S3B). Meanwhile, the inability of the divergent primer to amplify the circBtnl1 fragment further indicated that circBtnl1 was successfully deleted in mice by PCR (Appendix Fig S3C). As expected, circBtnl1 was completely deleted in the intestine of circBtnl1 ^−/− mice compared with that of littermate circBtnl1 ^+/+ mice (Appendix Fig S3D). circBtnl1 knockout in mice was further validated by Northern blot (Appendix Fig S3E). By contrast, real‐time PCR and immunoblot assays showed no alteration in Btnl1 protein in circBtnl1 ^−/− mice versus circBtnl1 ^+/− and circBtnl1 ^+/+ mice (Appendix Fig S3D and F). We next depleted Btnl1 via using short hairpin RNA (shRNA) in ISCs, and established organoid assays to determine effect of Btnl1 loss on the stemness of ISC cells. Btnl1 deleption were further validated by real‐time PCR (Appendix Fig S3G) and immunoblot assays (Appendix Fig S3H). Of note，deletion of Btnl1 did not alter the organoid formation capacities compared with empty vector. (Appendix Fig S3I). We observed that small intestines and colons of circBtnl1 ^−/− mice had much longer length than those of circBtnl1 ^+/+ mice (Fig 2A; Appendix Fig S4A), but no apparent changes in other organs including brain, heart, lung, liver, spleen, and kidney (Appendix Fig S4B). Moreover, circBtnl1 ^−/− mice showed increased numbers of crypts and cell numbers of crypts and villi in all of three small intestine regions (duodenum, jejunum, and ileum) (Fig 2B; Appendix Fig S4C). We also observed that circBtnl1 ^−/− mice had much higher crypts and villi compared with those of circBtnl1 ^+/+ mice (Appendix Fig S4D). Meanwhile, circBtnl1 ^−/− mice showed increased numbers of crypts in colon (Appendix Fig S4E). In addition, we also detected expansion of small intestinal cells with Ki67 staining to further confirm the proliferation of ISCs (Fig 2C). Consistent with crypt Ki67 staining, ethynyldeoxyuridine (EdU) staining assays demonstrated that circBtnl1 overexpressed intestinal organoids exhibited decreased proliferation and growth capacities compared with cells transfected with empty vector on day D4 and D6 (Fig 2D; Appendix Fig S4F).

Figure 2. Deletion of circBtnl1 promotes the self‐renewal of ISCs.

1. Small intestine images of circBtnl1 ^+/+ and circBtnl1 ^−/− mice. n = 3 mice were used for SI detection. Small intestine length are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test.
2. H&E staining of global view small intestine (left) and all of three enlarged small intestine regions: duodenum, jejunum and ileum from circBtnl1 ^+/+ and circBtnl1 ^−/− mice (right). Scale bars, 50 μm. Number of crypts were calculated as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test in right panel. Three fields were calculated for each group.
3. Representative images of Ki67 staining of in small intestinal tissues from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. For all representative images, at least three independent experiments were performed with similar results. For every microscope field, average Ki67⁺ ISC numbers per crypt are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test in right panel. Scale bar, 50 μm.
4. EdU staining of organoids from the circBtnl1 ^+/+ and circBtnl1 ^−/− mice under circBtnl1 overexpression. circBtnl1 ^+/+ and circBtnl1 ^−/− mice crypts were collected for organoid formation, and 10 μM EdU solution was added to the cultured organoid 4 h before collected. Scale bars, 50 μm.
5. H&E staining of small intestine tissues from indicated mice at different time points (D0–D7) after 8 Gy radiation damage. Number of crypts were calculated as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test in right panel. n = 5 fields were calculated for each group. Scale bar, 50 μm.
6. Quantification of organoid formation from intestinal crypts isolated from circBtnl1 ^+/+ and circBtnl1 ^−/− mice induced by radiation (8 Gy) on D5 and D7. (n = 3 circBtnl1 ^+/+ and circBtnl1 ^−/− mice were used and for each mouse, three separate organoid assays were performed). Crypt area were calculated as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test in right panel. Scale bar，200 μm.
7. Confocal microscopy of ISCs makers Olmf4 in organoids from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. Scale bars, 30 μm. Three independent experiments were performed with similar results, and representative experiments are shown. For every microscope field, average Olmf4⁺ ISC numbers per crypt are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test in right panel in right panel.

Given that ISCs have injury‐induced plasticity and are capable of self‐regeneration, we then examined intestinal regeneration of circBtnl1 ^−/− mice. Five days following gut epithelium injury induced by radiation (8 Gy), circBtnl1 ^−/− mice showed enhanced intestinal regeneration by H&E staining compared with that of littermate circBtnl1 ^+/+ mice (Fig 2E). In addition, circBtnl1 ^−/− ISCs showed enhanced organoid formation capacity and elevated stem cell marker Olfm4 expression after irradiation injury (Fig 2F; Appendix Fig S4G). Of note, the Olmf4 was upregulated in circBtnl1 ^−/− mouse organoids compared with circBtnl1 ^+/+ counterparts (Fig 2G). Altogether, circBtnl1 plays a critical role in the regulation of the stemness of ISCs.

CircBtnl1 associates with Ddx3y in ISCs

Given that circBtnl1 deletion did not alter the mRNA and protein levels of its parental gene, circBtnl1 could modulate the self‐renewal of ISCs through other pathways. To identify binding protein candidates of circBtnl1, we carried out an RNA pulldown and mass spectrometry assay from mouse ISC lysates with biotin‐labeled circBtnl1. Ddx3y, a member of the DEAD‐box RNA helicase family, was identified as an associated candidate protein (Fig 3A; Appendix Fig S5A). Ddx3y is an ATP‐dependent RNA helicase. Moreover, RNA‐FISH and immunofluorescence assay confirmed the colocalization of circBtnl1 with Ddx3y in the cytoplasm of mouse SI crypts (Fig 3B) and ISC organoids (Fig 3C). The interaction of circBtnl1 with Ddx3y was further validated by RNA pulldown and followed by immunoblotting analysis (Fig 3D and E). Through analysis by Uniprot website (https://www.uniprot.org/uniprotkb/O15523/feature‐viewer), there were two RNA‐binding sites in Ddx3y, indicating that Ddx3y can serve as a RNA‐binding partner (Yang et al, 2019) (Appendix Fig S5B). We noticed that the RNA‐binding domain (161–260 nt) of FLAG‐tagged Ddx3y protein was crucial for its interaction with circBtnl1 through in vitro mapping assay (Fig 3F). Consistently, RNA electrical mobility shift assay (EMSA) showed that circBtnl1 strongly interacted with endogenous Ddx3y in ISC lysates or GST‐tagged recombinant Ddx3y protein (Fig 3G). Of note, the exon 4 of the circBtnl1 transcript (HR2) was necessary to bind Ddx3y protein through fragment mapping analysis (Fig 3H; Appendix Fig S5C). These results indicated that circBtnl1 was predominantly localized in the cytoplasm and interacted with Ddx3y protein in mouse ISCs. Next, we evaluated the function of Ddx3y in ISCs. We found that Ddx3y knockdown remarkably disrupted organoid formation capacity (Fig 3I). In contrast, Ddx3y overexpressed organoids displayed increased formation capacity than vector controls (Fig 3I). Furthermore, circBtnl1‐binding domain deleted Ddx3y (oeDdx3y^▵160–260) overexpression abrogated organoid formation even in WT organoids (Fig 3I). In addition, we also detected small intestinal organoids with Ki67 staining to further validate the proliferation of circBtnl1 ^+/+ ISCs (Fig 3J). In parallel, Ddx3y deletion remarkably decreased numbers of Ki67⁺ ISC subpopulations, while Ddx3y overexpression increased ISC proliferation (Fig 3J). Moreover, the inhibitory role of circBtnl1 in organoid formation and proliferation could be abolished by Ddx3y knockdown or overexpression, indicating that circBtnl1 functions through Ddx3y in ISCs (Fig 3J).

Figure 3. CircBtnl1 binds Ddx3y in ISCs.

• A
  Small intestine crypts from WT mice were lysed and incubated with biotin‐labeled and linearized sense (circBtnl1 transcripts), anti‐sense, or Sepharose 4B beads control. Eluted fractions were resolved by SDS–PAGE followed by silver staining and mass spectrometry. Representative protein Ddx3y was shown on the right.
• B, C
  CircBtnl1 was co‐localized with Ddx3y protein in small intestine tissue (B) and organoid (C) by RNA FISH, followed by immunofluorescence staining. Scale bars, 5 μm.
• D
  Immunoblotting analysis of Ddx3y in RNA pull‐down samples by circBtnl1 sense, antisense and control probes in mice ISCs.
• E
  Immunoblotting analysis of Ddx3y in RNA pull‐down eluate by circBtnl1 probes in circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs.
• F
  Domain mapping analysis of circBtnl1‐binding domains of Ddx3y protein. Different domains of Ddx3y protein were incubated with circBtnl1, followed by RNA pulldown assay and Western blot.
• G
  Electrophoretic mobility shift assay (EMSA) using biotin‐labeled circBtnl1 and recombinant Ddx3y.
• H
  Truncated fragments of biotinylated circBtnl1 were incubated with small intestine crypts lysates, followed by RNA pulldown assay and Western blot. Schematic diagram of circBtnl1 truncated fragments is shown in Appendix Fig S5C.
• I
  Organoid formation was conducted with Ddx3y knockdown (Ddx3y KD), Ddx3y overexpression (oeDdx3y) and Ddx3y‐binding domain deletion overexpression (oeDdx3y^{▵160–260)} in circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs. Typical images and organoid formation ratios are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test. Scale bars, 200 μm.
• J
  Confocal microscopy of cell proliferation makers Ki67 in Ddx3y KO and control organoids from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. Scale bars，20 μm. Three independent experiments were performed with similar results, and representative experiments are shown.

Source data are available online for this figure.

RK‐33, one of the Ddx3 inhibitors can specifically bind Ddx3 and the most studied Ddx3 inhibitor in the cancer field (Vellky et al, 2020). The effect of Ddx3 inhibitor on organoid formation was examined using mouse SI organoids treated with RK‐33 at the indicated doses for 48 h. We observed that RK‐33 inhibited organoid cell viability in vitro in a dose‐dependent manner, with IC50 values 5 μM (Appendix Fig S5D). We then examined intestinal regeneration, seven to 12 days following gut epithelium injury induced by radiation (8 Gy), RK‐33 treated mice showed impaired intestinal regeneration compared with that of vehicle treated mice (Appendix Fig S5E). We noticed that there were no apparent differences for the expression of Ddx3y between circBtnl1 ^+/+ and circBtnl1 ^−/− mouse organoids using immunofluorescence assay, excluding the possible role of circBtnl1 in Ddx3y expression (Appendix Fig S5F). Collectively, circBtnl1 plays the role in the regulation of ISCs through direct interaction with Ddx3y.

CircBtnl1 and Atf4 mRNA competitively binds Ddx3y to modulate Atf4 mRNA stability

To further identify target genes of circBtnl1, we performed transcriptome microarray analysis on Lgr5⁺ ISCs sorted from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. Totally, 231 genes were downregulated and 418 genes were upregulated (fold change > 1.5, FDR < 0.05) in circBtnl1 ^−/− ISCs compared with their expression in circBtnl1 ^+/+ ISCs (Fig 4A). Top 10 upregulated transcription factors (Reg4, Fabp2, Atf4, Odc1, Sox9, Kif4, Stat3, Atf3, and Car4) were selected to further verify their expression in circBtnl1 ^−/− ISCs by qRT‐PCR (Fig 4B). In order to determine their roles in the ISC self‐renewal maintenance, we then used shRNA to individually silence these 10 upregulated candidate TF genes in circBtnl1 ^−/− ISCs followed by organoid formation assays. We observed that Atf4 depletion showed most reduced organoid formation (Fig 4C). Our previous results demonstrated that circBtnl1 strongly interacted with Ddx3y in ISCs, we next wanted to know whether Atf4 mRNA could also bind to Ddx3y. We found that Atf4 mRNA was able to interact with Ddx3y (Fig 4D). By contrast, circBtnl1 overexpression impaired the interaction of Atf4 mRNA with Ddx3y (Fig 4E). We next determined whether Ddx3y and circBtnl1 bound to the same locus of Atf4 mRNA. As expected，the interaction of Atf4 mRNA with Ddx3y was further confirmed by immunoblot analysis in a dose dependent manner (Fig 4F), and truncation mapping indicated that Atf4 mRNA interacted with Ddx3y in the same binding site (161–260) as circBtnl1 did (Fig 4G). We noticed that circBtnl1 ^−/− ISCs showed reduced stability of Atf4 mRNA compared with that of circBtnl1 ^+/+ ISCs (Fig 4H and I), suggesting that circBtnl1 could regulate the stability of Atf4 mRNA.

Figure 4. CircBtnl1 inhibits Atf4 expression through facilitation of its mRNA stability.

• A
  Volcano plot analysis of differentially expressed stem factors from circBtnl1 ^+/+ and circBtnl1 ^−/− mice ISCs. Representative genes were shown.
• B
  Small intestine crypt cells were isolated from circBtnl1 ^+/+ and circBtnl1 ^−/− 2‐month‐old mice, followed by RNA extraction and q‐PCR analysis for top 10 upregulated transcription factors. n = 3 independent samples. Results of relative fold changes are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test.
• C
  Small intestine crypts were isolated from C57BL/6 mice for organoid culture. The top 10 upregulated transcription factors were depleted in organoid with a shRNA strategy, followed by organoid formation. Organoid‐formation ratios are shown as mean ± SD, n = 3 mice. *P < 0.05, **P < 0.01, and ***P < 0.001, NS, not significant, by two‐tailed Student's t‐test.
• D
  Immunoblotting analysis of Ddx3y in RNA pull‐down samples by Atf4 sense, antisense and control probes in mice ISCs.
• E
  Immunoblotting analysis of Ddx3y in RNA pull‐down samples by Atf4 and control probes in circBtnl1 ^+/+ and circBtnl1 ^−/− mice ISCs.
• F
  Immunoblotting analysis of Ddx3y in RNA pull‐down samples by Atf4 sense probe and gradient circBtnl1 in mice ISCs.
• G
  Domain mapping analysis of circBtnl1‐binding domains of Ddx3y protein. Different domains of Ddx3y protein were incubated with circBtnl1, followed by Atf4 pulldown assay and Western blot.
• H, I
  RNA blot analysis of Atf4 mRNA in circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs treated with 2 μg/ml actinomycin D for the indicated times by Northern blot analysis (H) and q‐PCR (I). n = 3 independent experiments. Results of relative fold changes are shown as mean ± SD. *P < 0.05, ** P < 0.01, and *** P < 0.001, NS, not significant, by two‐tailed Student's t‐test.
• J
  Confocal microscopy of ATF4 in Ddx3y knockdown and overexpression organoids from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. Scale bars, 20 μm.
• K, L
  Expression levels of Ddx3y and Atf4 in circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs were examined by real‐time PCR (K) and Western blot (L), three independent experiments were performed. Results of relative fold changes are shown as mean ± SD. ** P < 0.01, NS, not significant, by two‐tailed Student's t‐test.
• M
  Organoid formation was conducted with Atf4 knockdown (Atf4 KD) from circBtnl1 ^+/+ and circBtnl1 ^−/− organoid. Typical images and organoid formation ratios (n = 3) are shown. Scale bars, 300 μm. **P < 0.05, NS, not significant, by one‐tailed Student's t‐test.

Source data are available online for this figure.

In order to determine the role of Ddx3y in the self‐renewal maintenance of ISCs, we performed Ddx3y depletion and Ddx3y overexpression in circBtnl1 ^+/+ and circBtnl1 ^−/− organoids. We observed that Ddx3y knockdown in circBtnl1 ^−/− organoids dramatically decreased expression of ATF4 (Fig 4J). By contrast, Ddx3y overexpression in circBtnl1 ^−/− ISCs highly increased the expression of ATF4 (Fig 4J). We noticed that ATF4 was remarkably expressed in circBtnl1 ^−/− mice, but no apparent alteration of Ddx3y levels in circBtnl1 ^−/− mice versus circBtnl1 ^+/+ mice (Fig 4K and L). Of note, Atf4 knockdown remarkably decreased organoid formation of ISCs in circBtnl1 ^+/+ and circBtnl1 ^−/− organoids (Fig 4M). Taken together, circBtnl1 deletion increases Atf4 mRNA stability in a Ddx3y‐dependent manner.

CircBtnl1 knockout facilitates ATF4 onto Sox9 promoter to trigger its transcription

To determine the biological function of ATF4 in the modulation of downstream target genes, we performed Tagmentation (CUT&Tag) assay in HCT‐116 (Human colorectal cancer) cells. Putative ATF4‐binding motifs were predicted in the mouse (up) and human (down) Sox9 promoter by Homer assay (Fig 5A). We analyzed a 2 kb region upstream of the Sox9 gene transcription start site (TSS) and found that ATF4 was enriched at the −100 to −200 bp region of the Sox9 promoter (Fig 5B). In addition, dual luciferase reporter assay also revealed Sox9 promoter (−160 to −120) deletion dramatically reduced ATF4 mediated Sox9 luciferase activity (Fig 5C). Through ChIP assay, we found that circBtnl1 deletion promoted enrichment of H3K4me3 on Sox9 promoter in ISCs (Fig 5D). By contrast, circBtnl1 deletion decreased H3K27me3 enrichment on Sox9 promoter in ISCs (Fig 5E). To assess density of nucleosomes in Sox9 promoter, we performed DNase I accessibility assay. We observed that Sox9 promoter in circBtnl1 ^−/− ISCs was more susceptible to DNase I digestion (Fig 5F). Consistently, circBtnl1 deficiency caused elevated transcription of Sox9 mRNA by nuclear run‐on assay (Fig 5G). Next, we carried out immunofluorescence assays to measure the Sox9 protein levels in Atf4 depleted organoids from circBtnl1 ^+/+ and circBtnl1 ^−/− mice compared with vector control. We found that Atf4 knockdown in circBtnl1 ^−/− organoids remarkably decreased Sox9 expression (Fig 5H). By contrast, circBtnl1 overexpression in circBtnl1 ^−/− organoids could suppress Sox9 expression (Fig 5I). More importantly, we noticed that Ddx3y knockdown in circBtnl1 ^−/− organoids also reduced Sox9 expression as Atf4 knockdown did, but Sox9 knockdown in circBtnl1 ^−/− organoids did not affect Atf4 expression (Fig 5J). Finally, Ddx3y overexpression in circBtnl1 ^−/− organoids was able to enhance Sox9 expression (Fig 5J). These data indicated that Sox9 acted as a downstream TF of ATF4 in the regulation of ISC self‐renewal maintenance. Taken together, circBtnl1 deficiency facilitates ATF4 expression to enhance Sox9 transcription in ISCs.

Figure 5. CircBtnl1 deficiency promotes ATF4 onto Sox9 promoter to trigger its expression.

• A
  Putative ATF4‐binding motifs are predicted in the mouse Sox9 promoter (up) by Homer assay. ATF4‐binding motif CACAATGCCCC were shown in mouse and human Sox9 promoter.
• B
  ChIP was performed to identify binding regions of ATF4 on Sox9 promoter in ISCs, followed by qPCR. IgG enrichment served as a control. Results are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test. Data are representative of at least three independent experiments.
• C
  Luciferase reporter assays were performed in Sox9 promoter‐depletion (−160 to −120) and control cells. Data are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test. Data are representative of at least three independent experiments.
• D, E
  ChIP assay was performed using anti‐H3K4me3 (D) or anti‐H3K27me3 (E) with circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs lysates. **P < 0.01 by two‐tailed Student's t‐test. Data are representative of at least three independent experiments.
• F
  DNase I accessibility assay was conducted using circBtnl1 ^+/+ and circBtnl1 ^−/− ISCs. n = 3 independent samples. **P < 0.01 by two‐tailed Student's t‐test.
• G
  circBtnl1 ^+/+ or circBtnl1 ^−/− ISCs were subjected to nuclear run‐on assay, followed by detection of Sox9 transcription through qPCR analysis. Results of relative fold changes are shown as mean ± SD. ** P < 0.01 by two‐tailed Student's t‐test.
• H
  Confocal microscopy of Sox9 in Atf4 knockdown (Atf4 KD) organoids from circBtnl1 ^+/+ and circBtnl1 ^−/− mice. Scale bars, 30 μm.
• I
  Crypts were isolated from circBtnl1 ^+/+ and circBtnl1 ^−/− mice for organoid culture, and the circBtnl1 overexpression organoid was conducted from circBtnl1 ^−/− organoid. RNA was extracted from the organoid and the expression of Sox9 was analyzed by RT‐PCR. Results of relative fold changes are shown as mean ± SD. **P < 0.01 by two‐tailed Student's t‐test. Data are representative of at least three independent experiments.
• J
  Immunoblotting analysis of ATF4 and Sox9 in organoids from Ddx3y overexpression (oe Ddx3y) and Atf4 knockdown (Atf4 KD) as indicated in circBtnl1 ^+/+ and circBtnl1 ^−/−. Results are representative of two independent experiments.

Source data are available online for this figure.

Sox9 knockout impairs the self‐renewal maintenance of ISCs

To further verify the physiological role of Sox9 in the self‐renewal maintenance of ISCs, we generated Sox9 deficient mice (circBtnl1 ^+/+; sgSox9 and circBtnl1 ^−/−; sgSox9) through CRISPR/Cas9‐mediated genome editing (Appendix Fig S6). Sox9 was dramatically reduced in circBtnl1 ^+/+; sgSox9 and circBtnl1 ^−/−; sgSox9 mice, meanwhile ISC marker Olfm4 was consequently decreased in these two mouse strains (Fig 6A). As expected, Sox9 deletion caused remarkably decreased numbers of Ki67⁺ ISC subpopulations in small intestines (Fig 6B) and colons (Fig 6C) in circBtnl1 ^+/+ mice. Moreover, Sox9 deletion showed reduced numbers of crypts in small intestine tissues in both circBtnl1 ^+/+ and circBtnl1 ^−/− mice (Fig 6D). We then examined intestinal regeneration of Sox9‐deficient mice. Five days following gut epithelium injury induced by radiation (8 Gy), Sox9 deletion displayed impaired intestinal regeneration in both circBtnl1 ^+/+ and circBtnl1 ^−/− mice (Fig 6E). Consistently, we found that Sox9 depletion in circBtnl1 ^+/+ and circBtnl1 ^−/− organoids dramatically impaired organoid formation (Fig 6F). Collectively, these data indicate that Sox9 participates in the circBtnl1‐mediated self‐renewal regulation of ISCs.

Figure 6. Sox9‐mediated signaling is required for ISC maintenance.

• A
  Immunoblotting analysis of Sox9 and Olfm4 (stem cell maker) in small intestine crypts from circBtnl1 ^+/+; sgSox9 and circBtnl1 ^−/−; sgSox9 mice, vector served as control. Results are representative of two independent experiments.
• B, C
  Representative images of Ki67 staining of in small intestine (B) and colon (C) from circBtnl1 ^+/+ mice. For all representative images, at least three independent experiments were performed with similar results. Scale bar, 30 μm.
• D
  H&E staining of small intestinal tissues from indicated mice. Two‐month‐old mice were used and n = 3 fields were taken for each group. Scale bar, 100 μm. Number of crypts were calculated as mean ± SD. *P < 0.05 and **P < 0.01 by two‐tailed Student's t‐test in right panel.
• E
  H&E staining of intestinal tissues from indicated mice at different time points after 8 Gy of radiation damage. Two‐month‐old mice were used and n = 3 fields were taken for each group. Number of crypts were calculated as mean ± SD. *P < 0.05 and **P < 0.01 by two‐tailed Student's t‐test in right panel. Scale bar, 50 μm.
• F
  Representative organoid images of circBtnl1 ^+/+; sgSox9 and circBtnl1 ^−/−; sgSox9 mice, sgCtrl served as control. Organoid formation ratios (n = 3) are shown. Organoid formation ratios are shown as mean ± SD. *P < 0.05 by two‐tailed Student's t‐test. Scale bars, 200 μm.

Source data are available online for this figure.

Discussion

Under homeostatic conditions, the intestinal epithelium harbors remarkable self‐renewal capacity that is driven by ISCs residing within specialized niches at the crypt base (Guiu et al, 2019; Flanagan et al, 2021; Zhu et al, 2022). ISCs sense stemness signaling and initiate their stemness TFs to support them for self‐renewal and epithelial regeneration (Zhu et al, 2018; Yum et al, 2021). However, the regulation of self‐renewal capacity of ISCs is still less defined. In this study, we showed that circBtnl1 was highly expressed in ISCs. CircBtnl1 knockout enhanced the self‐renewal capacity and epithelial regeneration. CircBtnl1 knockout did not alter mRNA and protein levels of its parental gene Btnl1. Mechanistically, circBtnl1 bound to Ddx3y to impair the stability of Atf4 mRNA in normal ISCs. Furthermore, ATF4 as an upstream TF activated Sox9 transcription by binding onto its promoter via a unique motif. Sox9 promoted the self‐renewal capacity and epithelial regeneration of ISCs. By contrast, CircBtnl1 knockout promoted Atf4 mRNA stability to lead to Sox9 transcription, resulting in enhancement of ISC stemness.

CircRNAs form a closed continuous loop with covalently joined 3′ RNA and 5′ RNA, as a new type of single‐stranded RNA (Slack & Chinnaiyan, 2019). Owing to the flexibility of RNAs and complementary sequences on RNAs, circRNAs can be composed of exon, intron, or even exon–intron sequences. Although back splicing is generally less efficient than linear splicing, circRNAs can accumulate in specific cell types in a temporally regulated manner due to their high stability. Similar to other noncoding RNAs, most circRNAs exhibit strong tissue‐specific expression patterns, implying their unique functions in various tissues. However, how circRNAs modulate ISC biology is still needs to be further explored. In this study, we found that circBtnl1 was highly expressed in the small intestine and colon. CircBtnl1 was localized exclusively in the cytoplasm rather than in the nucleus of ISCs. CircBtnl1 deletion enhanced the self‐renewal capacity and epithelial regeneration. Of note, circBtnl1 deletion did not alter the mRNA and protein levels of its parental gene Btnl1 in ISCs. In addition, we observed that circBtnl1 ^−/− mice displayed more cell numbers of goblet cells, TA cells, Paneth cells and enteroendocrine cells compared with those of circBtnl1+/+ mice. Based on these observations, circBtnl1 could affect ISC differentiation besides regulation of their self‐renewal capacity.

CircRNAs are now widely recognized as a novel subset of endogenous RNAs that interact with RNA‐binding proteins (RBPs) to enhance or suppress their functions (Conn et al, 2015). RBPs are key regulators of multiple post‐transcriptional events, including RNA splicing, stability, transport and translation (Tay et al, 2014). circCTNNB1 binds to DDX3 and increases its interaction with transcription factor YY1, resulting in enhanced transactivation of YY1 target promoters (Yang et al, 2019). DDX3 is a DEAD‐box RNA helicase that regulates translation (Guenther et al, 2018), which is encoded by the X‐ and Y‐ linked paralogs DDX3X and DDX3Y (Samir et al, 2019). DDX3Y mRNA is broadly expressed across various tissues, but activation of a testis‐specific distal promoter produces a transcript isoform with a distinct 5′ leader that also contributes to its translational control. Although the role for Ddx3y has not been reported in ISCs, its role in stem cell maintenance and differentiation is beginning to be unraveled in lower organisms. In this study, we showed that circBtnl1 bound to Ddx3y and colocalization of circBtnl1 and Ddx3y in the cytoplasm of ISCs. The binding of circBtnl1 with Ddx3y blocks the binding of Atf4 mRNA to enhance Atf4 mRNA decay in ISCs.

Transcription factors play central roles in cell fate determination. Here we identified ATF4 and Sox9 as two functional TFs in ISC maintenance. ATF4 is a basic region‐leucine zipper transcription factor that is widely expressed in many tissues and cells. It has been reported that ATF4 functions as a stress response factor and a developmental regulator. ATF4 plays a critical role in definitive hematopoiesis at the stem cell level, such as de novo generation, migration, amplification, and maintenance of HSCs. However, how ATF4 regulates the self‐renewal of ISCs remains elusive. Here we showed that circBtnl1 and Ddx3y could competitively bind Atf4 mRNA. The binding of Ddx3y with Atf4 mRNA enhanced the stability of Atf4 mRNA, leading to ATF4 expression for initiation of Sox9 transcription. Sry‐box containing (Sox) factors are transcription factors with broad regulatory roles in stem cell maintenance and differentiation in many adult tissues. Early insights into the function of Sox factors have involved cell fate determination during development (Sato et al, 2011), although recent findings reveal its crucial role in establishing and maintaining stem and progenitor cell pools (Mamidi et al, 2018; Ransom et al, 2018; Morita et al, 2021; Willnow et al, 2021). A common feature of both the Wnt and Notch pathways is considered to be modulated by Sox genes, indicating a critical role for Sox transcription factors in normal intestinal epithelium homeostasis. Sox9 is characterized by the presence of a conserved HMG DNA‐binding domain, which is a potent regulator of cell fate decisions and stem cell maintenance (Xian et al, 2017; van Gastel et al, 2020). In the adult intestine, Sox9 is highly expressed in stem cells of the intestinal crypts where it is regulated by the canonical WNT/β‐catenin/T cell factor (WNT/β‐catenin/TCF) pathway (Shyer et al, 2015). In this study, we found that ATF4 activated Sox9 promoter through a unique binding motif to trigger its transcription. We showed that Sox9 deletion remarkably decreased numbers of ISCs and epithelial regeneration, suggesting that Sox9 enhances the maintenance of ISCs.

In summary, we identified and characterized an important circRNA, circBtnl1, which could physically interact with Ddx3y to impair Atf4 mRNA stability and inhibited its translation. CircBtnl1 mediated downregulation of ATF4 suppressed Sox9 transcription that negatively modulated the self‐renewal maintenance of ISCs.

Materials and Methods

Antibodies and reagents

Anti‐ATF4 (Cat# 11815), Anti‐Olfm4 (Cat# 14369), Anti‐H3K27ac (Cat# 8173), and anti‐H3K27me3 (Cat# 9733) were purchased from Cell Signaling Technology (Danvers, USA). Anti‐Sox9 (Cat# ab185230), Anti‐Ddx3y (Cat# ab196663), Anti‐GFP (Cat# ab183735), anti‐Ki67 (Cat# ab15580), and F‐actin staining kit (Cat# ab1112127) were all obtained from Abcam. Anti‐β‐actin (Cat# A1978) and anti‐Flag (Cat# F1804) antibodies were from Sigma‐Aldrich. Alexa‐594, Alexa‐488 and Alexa‐647 conjugated anti‐rabbit and anti‐mouse secondary antibodies were purchased from Invitrogen. The Dual Luciferase Reporter Gene Assay Kit (Cat# RG027) was purchased from Beyotime. Biotin RNA Labeling Mix (Cat# 11685597910) and T7 RNA polymerase (Cat# 10881767001) were from Roche. The Light Shift Chemiluminescent RNA EMSA kit (Cat# 20158) and Chemiluminescent Nucleic Acid Detection Module (Cat# 89880) were from Thermo Scientific. Paraformaldehyde (PFA) and 4′, 6‐diamidino‐2‐phenylindole (DAPI) were from Sigma‐Aldrich. DNase I was purchased from Roche Molecular Biochemicals (Basel, Switzerland).

Generation of circBtnl1 ^−/− knockout mice by CRISPR/Cas9 technology

For generation of circBtnl1 knockout (circBtnl1 ^−/−) mice, CRISPR‐mediated single‐stranded oligodeoxynucleotides donors were synthesized as previously described (Zhu et al, 2019). About 250 zygotes from C57BL/6 mice were injected with sgRNAs and subsequently transferred to the uterus of pseudo‐pregnant ICR females from which viable founder mice were obtained. Genomic DNA mutation was identified by PCR screening and DNA sequencing, followed by western blotting or northern blotting. sgRNA sequences are listed in Appendix Table S1. All mice were back‐crossed into the C57BL/6 genetic background for at least 10 generations. Both male and female mice ranging from 2 to 4 months old in age were used. Generally, we used at least five mice per genotype in each experiment. None of the animals was excluded from the experiment, and the animals used were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. For Cre‐induction, mice were intraperitoneally injected with 100 ml tamoxifen in sunflower oil at 20 mg/ml for 5 consecutive days. For irradiation, mice received a single dose of abdominal X‐ray radiation (8 Gy) and were then analyzed at different time points. Rosa26‐cas9‐P2A‐EGFP and Lgr5‐EGFP‐IRES‐creERT21 mice were obtained from Jackson Laboratory. All animal studies were performed in accordance with the relevant guidelines and under the approval of the Institutional Committee of Institute of Biophysics, Chinese Academy of Sciences.

CircRNAs sequencing and microarray assay

For circRNAs sequencing, RNA was isolated with Trizol method from small intestine crypts, and purified with RNase Mini kit (Y5‐74104，QIAGEN). Then circRNA were sequenced using circRNA Array (BGI Tech Company, China) was performed. For identification of circRNA downstream target genes, 2 × 10⁵ circBtnl1 ^+/+ and circBtnl1 ^−/− Lgr5‐GFP⁺ ISCs were collected from small intestines by FACS sorting and were detected by confocal microscopy (purity over 95%). Total RNA was extracted using a standard RNA‐extraction protocol and then mRNA microarray assay analysis was performed by BGI Tech Company.

CRISPR/Cas9 knockout system

Sox9‐deletion cells were established using CRISPR/Cas9 approaches provided by Zhang's lab. All sgRNAs were designed by online CRISPR Design Tool (http://crispr.mit.edu/). In brief, sgRNAs were cloned into LentiCRISPRv2 (puro, catalog 52,961, Addgene). We used LentiCRISPRv2, pVSVg (catalog 8,454, Addgene), and psPAX2 (catalog 12,260, Addgene) plasmids to produce CRISPR‐Cas9 lentivirus. sgRNA sequences are listed in Appendix Table S1. We generated lentivirus in 293 T cells with lipo3000 for 2 days. After the supernatant was filtered with 0.45 μm strainer and mixed with equal volume fresh DMEM (10% FBS), organoid cells were infected for 12 h, followed by puromycin selection. The organoid cells were propagated and confirmed by DNA electrophoresis and sequencing. The correctly targeted clones were obtained and used for subsequent experiments.

shRNA knockdown system

Silencing of indicated genes was performed by short hairpin RNA as before. shRNAs of indicated genes were designed on online RNAi designer (Thermo Fisher Scientific). 2–3 shRNAs of each target gene were selected and cloned into pLVshRNA‐EGFP(2A) Puro or pLVshRNA‐Puro lentivirus vector, and the primers are listed in Appendix Table S5. Lentiviral vectors were co‐transfected with packaging plasmids psPAX2 and pMD2G into HEK293T cells. Infectious lentiviruses were harvested at 48 and 72 h after transfection and filtered through 0.45‐μm filters. Recombinant lentiviruses were concentrated by virus precipitation solution (Genstar). For gene silencing in organoid cells, organoids were cultured for 7 days. Organoids were broken as described in organoid passaging step. Then cells were resuspended with 500 ml complete organoid media and mixed with lentivirus solution adding 6 mg/ml polybrene (Sigma‐Aldrich) and transferred into 24‐well plates for centrifugation at 32°C, 600 g, for 1 h. Cells were cultured at 37°C for 6 h. Then cells were centrifuged to remove virus and resuspended with Matrigel for organoid culture. After 3 days, EGFP of organoids was detected with fluorescence microscope and 1 mg/ml puromycin was added into the media for selection. Organoids were passaged and maintained and gene silencing efficiency was analyzed by real‐time qPCR, and the primers are listed in Appendix Table S2.

Isolation of intestinal crypts and organoid culture

Intestinal crypts were isolated and cultured as previously described with minor modifications (Zhu et al, 2019). Briefly, mouse intestine was cut longitudinally and washed three times with cold PBS. Villi were carefully scraped away and small pieces (5 mm) of intestine were incubated in collagenase solution (DMEM/F12 medium, containing 0.1% type I collagenase (Invitrogen), 100 units/ml penicillin, 0.1 mg/ml streptomycin and 10 mM HEPES). Every 10 min, the samples were mixed vigorously by a 1,000 μl pipette and the separation of the crypts was checked with a microscope, until more than 50% of crypts were separated from tissue fragments (about 30 min). Then the samples were filtered through a 70 μm cell strainer, followed by 80 g centrifugation for 5 min. After one wash, the crypts were enriched in the precipitant. Then the crypts were embedded in Matrigel (BD Biosciences) and seeded on 24‐well plate. After polymerization, crypt culture ENR medium (Advanced DMEM/F12 supplemented with Penicillin/Streptomycin), Gluta MAX‐I, N2, B27 and N‐acetylcysteine (Sigma‐Aldrich), EGF (50 ng/ml, Invitrogen), Noggin (100 ng/ml, R&D), and R‐spondin1 (500 ng/ml, R&D) was added and refreshed every 2 days. For passaging, the organoids embedded in Matrigel in each well were directly suspended in 1 ml cold PBS after removal of medium and were pelleted by centrifugation (3 min at 300–400 g), and then pipetted repeatedly. The pelleted organoids were embedded in fresh Matrigel and seeded on plate followed by addition of culture medium as indicated in the figure legends.

EdU treatment

For EdU treatment, 10 μM EdU solution was added to the cultured organoid 4 h before collected, and EdU was detected according to the manufacturer's protocol (YF® 594 Click‐iT EdU Imaging Kit, Cat# YF594, BIORIGIN).

Real‐time quantitative PCR

Total RNA was extracted from crypts or organoid cells with standard Trizol methods. For circRNAs detection, RNase R (3 U/mg, Epicenter) digestion was undertaken at 37°C for 15 min. Reverse transcription and real‐time PCR were performed using the 5X All‐In‐One RT Master Mix (Applied Biological Materials Inc.), Super Real Premix Plus (SYBR Green) (Tiangen), and primer sets targeting the conjunction sequence of circRNA (Appendix Table S3). The results of transcript levels were analyzed by the 2ct method.

Northern blot

Northern blot was performed as previously described (Zhu et al, 2019). Total RNA was extracted from crypts or organoid cells with standard Trizol methods, and then subjected to electrophoresis on 2% denaturing agarose gel with 1% formaldehyde for 1.5 h. Samples were transferred to positively charged NC membranes with 20 × SSC buffer. After UV cross‐linking (265 nm ultraviolet with energy of 200,000 μJ/cm²) and prehybridization, membranes were incubated with biotin‐labeled probes at 65°C for 16–20 h. After washed with washing buffer, biotin signals were detected with Chemiluminescent Nucleic Acid Detection Module according to the manufacturer's instructions. For detecting circRNAs only, inverse complementary sequences of junction sequences were used for probes.

Biotin‐labeled RNA pulldown and mass spectrometry assay

The 5′‐monophosphorylated linear or random probe of circBtnl1 was in vitro transcribed using biotin RNA labeling mix (Roche), and T7 RNA polymerase (Roche). The lysates of 2 × 10⁷ crypt cells were incubated with 3 mg of biotin‐labeled probe, and treated with 60 uL of Streptavidin Magnetic Beads (Biolabs) for 4 h. The retrieved protein was detected by Western blot or mass spectrometry analysis (LTQ Orbitrap XL, Thermo).

Fluorescence in situ hybridization

For circBtnl1 FISH, fluorescence‐conjugated circBtnl1 probes (targeting the junction sequence of circBtnl1, sequence: 5′‐TAAACCAGGGTACAGCAAGG‐3′) were generated according to the protocols of Biosearch Technologies. Organoids were seeded on Chambers (43300‐774, Lab‐Tek), and fixed with 4% paraformaldehyde for 30 min, and then permeated by 1% triton X‐100 for 30 min. For FISH and double‐FISH assays, samples were treated in a non‐denaturing condition, followed by hybridization with probe sets. The indicated antibodies were then added after RNA hybridization for co‐localization of RNA and indicated proteins. Nuclei were counterstained using 4′, 6‐diamidino‐2‐phenylindole (DAPI). Confocal microscopy (Nikon A1R⁺) was performed for observation.

Immunofluorescence staining

Organoids were fixed by 4% paraformaldehyde (PFA) for 30 min, and then permeated by 1% triton X‐100 for 30 min. After blocking with 10% donkey serum for 30 min, primary antibodies were added and incubated overnight at 4°C. After washing with PBS, fluorescence‐conjugated secondary antibodies were added and incubated at RT for 1 h. After sealing, confocal microscopy (Nikon A1R⁺) was performed for observation.

DNase I accessibility assay

Nuclei were isolated from organoid using the Nuclei Isolating Kit (NUC101, Sigma‐Aldrich) according to the manufacturer's protocol. Nuclei were resuspended in 200 μl of DNase I digestion buffer (1 mM EDTA, 0.1 mM EGTA, 5% sucrose, 1 mM MgCl₂, 0.5 mM CaCl₂). Two equal aliquots of 100 μl nuclei were treated with the indicated units of DNase I (Sigma, USA) and incubated at 37°C for 5 min. Reactions were stopped by 2× DNase I stop buffer (20 mM Tris pH 8.0, 4 mM EDTA, 2 mM EGTA). DNA was extracted and analyzed by qPCR，and the primers are listed in Appendix Table S4.

Nuclear and cytoplasmic fraction isolation

Cytoplasmic and nuclear RNA was extracted using Cytoplasmic & Nuclear RNA Purification Kit (Norgen Biotek Corp.). Briefly, the cells were suspended and lysed with cell fraction buffer and then centrifuged at low speed to separate the nuclear fraction from the cytoplasmic fraction. Subsequently, the cytoplasmic fraction was carefully aspirated away from the nuclear pellet, and the cell disruption buffer was added to the nuclear pellet.

RNA electrophoretic mobility shift assay (EMSA)

For RNA EMSA, biotin‐labeled RNA probes were obtained by in vitro transcription assay with Biotin RNA Labeling Mix (Roche). Recombinant Ddx3y protein was obtained by purification (constructed into plasmids pGEX‐6P‐1 and 3xFlag). Probes and recombinant proteins were incubated in binding buffer and mobility shift assay was performed using native gel electrophoresis, biotin signals were detected according to the Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific).

Cell viability assay

Organoids were seeded at in 24‐well plates. Viability was detected following 24 and 48 h incubation with different dose of RK33 (0, 0.626, 1.25, 2.5, 5, 10, and 20 μM). Cells were lysed with Cell Titer‐Glo Lumines‐ cent Cell Viability Assay Reagent (Promega, Madison, WI, USA), and luminescence was quantified using the BioTek Synergy HTX Multi‐mode Micro Plate Reader (Winooski, VT, USA).

Immunoblotting

Crypts or Organoids were harvested, washed with cold PBS twice, and lysed with RIPA buffer (Cell Signaling). Cell lysates were maintained on ice and then centrifuged at 14,000 g for 20 min at 4°C. The protein concentration of each supernatant fraction was determined using the Bio‐Rad Bradford assay (Hercules, CA, USA). After electrophoresis, the proteins were transferred onto a nitrocellulose membrane (Bio‐Rad) and immunoblotted with primary and secondary antibodies. The membranes were visualized using the electro chemiluminescent detection reagent (Pierce Bio‐technology, Rockford, IL, USA).

Chromatin immunoprecipitation (ChIP) assay

ChIP was performed according to the standard protocol (Upstate Biotechnology, Inc.). Briefly, intestinal organoids were fixed in 1% formaldehyde for 10 min at 37°C, and then cracked by SDS lysis buffer for 10 min on ice, followed by ultrasonic treatment to shear DNA into fragments between 200 and 500 bp. The samples were pre ‐ cleared with salmon sperm DNA/protein agarose beads for 0.5 h in rotor at 4°C, and then incubated with the anti‐ATF4 antibodies overnight. The enrichments after elution were analyzed by qPCR. Primer sequences are shown in Appendix Table S4.

Statistics

For statistical analysis, data were analyzed using GraphPad Prism 8.0, and the two‐tailed unpaired Student's t‐test was used for the analysis of significance. P < 0.05 was considered significant (*P < 0.05, **P < 0.01, and ***P < 0.001), and P > 0.05 was considered non‐significant (NS).

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix S1

Source Data for Appendix

Source Data for Figure 1

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

The EMBO Journal (2023) 42: e112039

Data availability

CircRNAs sequencing data that support the findings of this study have been deposited in the Gene Expression Omnibus under accession code GSE218560 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE218560); CircBtnl1 Knockout sequencing data are available from the Gene Expression Omnibus database under accession code GSE218136 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE218136); CUT&Tag data are available from the Gene Expression Omnibus database under accession code GSE218559 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE218559).