Single-minded 2 is required for left-right asymmetric stomach morphogenesis

Vertebrate anatomy is characterized by left-right (LR) asymmetries in multiple organs of the cardiovascular, respiratory, digestive and lymphatic systems (Grimes and Burdine, 2017), many of which are derived from or closely associated with, the primitive gut tube (Grzymkowski et al., 2020). Examples of gut tube-derived asymmetries include the differentially lobed left and right lungs, or the disparate proportions of the left and right halves of the liver. However, in most species, the first region of the gut tube to manifest LR asymmetry is the stomach, which curves sharply leftward. The curvature process is driven by asymmetries in cell rearrangement that thin and expand the layers of the left stomach wall (Davis et al., 2017); differential expansion of the contralateral walls of the stomach tube thus impels the organ to curve. Although this process is dependent on global embryonic LR patterning cues, including expression of Pitx2 on the left side of the foregut (Grzymkowski et al., 2020), the side-specific morphogenetic programs that distinguish the left versus right sides of most asymmetric organs, including the stomach, remain largely unknown. Such knowledge is vital for understanding the evolution of digestive organ morphology and function, as well as for illuminating the etiology of laterality-related birth defects in gut tube-derived tissues in diverse organ systems.

To identify factors likely to drive stomach curvature, we took advantage of Budgett's frog embryos (Lepidobatrachus laevis); their extra-large size enables facile bisection of early organs not possible in other vertebrate models (Amin et al., 2015). We performed transcriptome profiling of the left and right halves of the developing stomach at successive phases of morphogenesis: before morphological asymmetry [Gosner stage (GS) 18], and early (GS19), mid- (GS20) and late (GS21) curvature (Fig. 1A; Gosner, 1960).

Differential expression analyses revealed 544 genes with significantly (q-value≤0.05) left or right-enriched expression during at least one phase of curvature morphogenesis (Table S1). The temporal laterality of 278 of these genes (with q-values ≤0.01) is highlighted in Fig. 1B, with more left-enriched (179) than right-enriched (99) genes identified (Fig. 1C). Our list of left-enriched genes includes the transcription factor pitx2, which is known to be expressed on the left side of the stomach and other organs (Davis et al., 2017; Grzymkowski et al., 2020). As expected for this master regulator of LR organogenesis, pitx2 was asymmetrically expressed at all phases of stomach curvature morphogenesis (Fig. 1C,D).

Surprisingly, only one other gene also displayed consistent asymmetric (left-sided) expression across all four stages of curvature (Fig. 1C,D): the highly conserved Per/Arnt/Sim bHLH transcription factor, single-minded 2 (sim2). The prominence of sim2 in our dataset is noteworthy as this gene is a vertebrate homolog of the Drosophila single-minded (sim) gene previously found to be required for LR asymmetrical looping of the fly midgut – a structure functionally equivalent to the vertebrate stomach and intestine (Maeda et al., 2007).

Unlike pitx2, which is expressed in both inner endodermal (future epithelial lining) and outer mesodermal (future smooth muscle) layers of the left stomach wall (Fig. 1E,F), sim2 is expressed exclusively in the endoderm layer (Fig. 1G), as revealed by RNA in situ hybridization (ISH) in the common laboratory frog, Xenopus. The gene is also expressed bilaterally in craniofacial structures (e.g. Fig. S1A-D), pronephric tubules (Fig. S1E), lungs (Fig. S1F) and in a narrow domain of epithelium comprising the dorsal midline of the foregut (asterisks, Fig. 1G and Fig. S1E,F). Below, we describe the functional validation of sim2 as a new organ-specific regulator of asymmetric morphogenesis. Further analysis of other left- and right-enriched genes in our LR transcriptome dataset (Table S1) will be reported elsewhere.

To confirm that the left-sided expression of sim2 is dependent on LR patterning cues, we exposed Xenopus embryos to a selective TGF-β receptor inhibitor (SB505124). This compound disrupts the Nodal signaling in the left lateral plate mesoderm (LPM) that is necessary to establish global LR asymmetry (Dush et al., 2011), consequently inhibiting left-sided pitx2 expression in the left LPM and ultimately leading to random stomach laterality, including straightened and reversed curvatures (Davis et al., 2017).

As expected, the expression of a pan-stomach marker (sox2) was unaffected by SB505124 exposure (compare Fig. 2A and D). However, compared with DMSO controls (Fig. 2B), the expression of pitx2 was reduced or eliminated in stomachs with SB505124-induced straightened or reversed curvatures (Fig. 2E). sim2 expression in the left stomach wall of SB505124-treated embryos was likewise reduced or eliminated (Fig. 2F) compared with controls (Fig. 2C), paralleling perturbed pitx2 expression. However, the midline (asterisks, Fig. 2F) and bilateral pronephric (arrowheads, Fig. 2F) expression domains of sim2 remained unaffected by SB505124.

Left-sided sim2 expression is evident as early as Nieuwkoop and Faber stage (NF) 32 (Nieuwkoop and Faber, 1994) (Fig. S1A,B), shortly after pitx2 expression is established in the left LPM, and overlaps identically with the endodermal domain of pitx2 in the left stomach wall (Fig. 1E-G). To test whether sim2 expression is regulated by Pitx2, we targeted a pitx2 morpholino (MO) to the left side of the developing stomach, a manipulation that we previously showed knocks down Pitx2 translation and perturbs stomach curvature (Davis et al., 2017). In 94.9% (n=46) of embryos injected with a control-MO, sim2 expression was left-sided as usual (Fig. 2G). However, sim2 expression was barely detectable in 66.1% of pitx2-MO-injected embryos (n=26; P<0.01; one-way ANOVA; Fig. 2I), suggesting that Pitx2 activity is indeed required for proper expression of sim2 in the left endoderm.

In the converse experiment, we ectopically expressed Pitx2 on the right side of the stomach by microinjecting mRNA encoding a hormone-inducible version of Pitx2 (pitx2-GR; Davis et al., 2017); exposure to dexamethasone was then used to activate ectopic Pitx2 before stomach curvature morphogenesis (Davis et al., 2017). Although aberrant sim2 expression was rarely seen in ethanol-exposed controls (4.9%; n=65; Fig. 2H), dexamethasone activation of ectopic Pitx2 activity led to expansion and/or bilateral expression of sim2 on both sides of the stomach in 52.4% of cases (n=71; P<0.01; Fig. 2J). Together, these pharmacological perturbation and epistasis results suggest that Pitx2 is both necessary and sufficient for asymmetrical sim2 expression.

To determine whether Sim2 is required for proper stomach curvature, we performed CRISPR-mediated editing of the genomic sequence encoding the HLH domain of sim2 in Xenopus embryos (Fig. 3A-H′). Although control F0 embryos (Fig. 3A,C) rarely (5.7%; n=35) exhibited aberrant stomach curvature, 60.1% (n=58; P<0.01) of sim2 gRNA+Cas9-injected F0 embryos developed straightened stomachs (Fig. 3B,D), correlating with the induction of predicted deleterious indels in the Xenopus sim2 genomic region (Fig. S2). This result confirms that Sim2 activity is required for proper execution of stomach curvature.

We previously showed that thinning and expansion of the left stomach wall is driven by radial intercalary rearrangement of left endoderm cells, which leads to a significant LR difference in stomach wall thickness (Davis et al., 2017). As sim2 is expressed in only the endoderm layer of the stomach, we hypothesized that it may play a role in execution of these cell rearrangements. Indeed, the LR difference in stomach wall thickness seen in control embryos (R/L ratio=1.31) was found to be significantly perturbed in sim2 CRISPants (R/L ratio=0.79; P<0.05). Moreover, in control stomachs, the left wall exhibited layers of radially aligned nuclei that have accumulated near the basement membrane after rearranging (Fig. 3E,G,G′). In contrast, endoderm cells in the straightened sim2-CRISPant stomachs remained distributed more broadly and randomly across the width of the left stomach wall, consistent with a failure to undergo the requisite rearrangements (Fig. 3F,H,H′).

To more specifically elucidate the cellular roles of Sim2 in stomach morphogenesis, we localized a sim2-MO to the left stomach endoderm (along with GFP mRNA as a lineage tracer for identifying Sim2-deficient cells). Although only 6.6% (n=151) of control-MO injected embryos (Fig. 3I,K,O) had abnormal stomach curvature, 84.5% (n=155) of sim2-MO-injected stomachs failed to curve (P<0.01; Fig. 3J,L,P). A second sim2-MO had a similar effect (Fig. S3A). This phenotype correlated with translation knockdown (Fig. S3B) and could be partially rescued by co-injection of exogenous MO-resistant sim2 mRNA (Fig. S3C), confirming the specificity of the result. (Importantly, targeting sim2-MO to the right endoderm had no effect on stomach curvature; Fig. S3D.)

As observed in sim2 CRISPants, the R/L ratio of the width of the stomach epithelium was reduced from 1.51 in controls to 1.01 in sim2 morphants (P<0.001), indicating failure of the process that normally thins the left wall with respect to the right. Furthermore, whereas control-MO injected endoderm cells formed radially aligned layers in the left stomach wall (Fig. 3M-M″), Sim2-deficient endoderm cells remained randomly distributed (Fig. 3N-N″). As expected for radially intercalating cell populations, endoderm cells injected with control-MO exhibited columnar shapes (Fig. 3M,M′,Q) and parallel arrays of microtubules (MTs) aligned perpendicular to the basement membrane (Fig. 3Q′). In contrast, Sim2-deficient endoderm cells appeared to be round (Fig. 3N,N′,R), with misaligned, randomly-oriented MTs (Fig. 3R′). Importantly, sim2-MO injected cells remained proliferative (Fig. S4A,B), suggesting that the abnormal architecture of Sim2-deficient epithelia is not due to decreased viability.

Control cells also displayed robust apicolateral localization of cell-cell adhesion markers, including β-catenin (Fig. 3M′) and E-cadherin (Fig. 3S,S′), as necessary to maintain tissue integrity during epithelial remodeling (Dush and Nascone-Yoder, 2013). In contrast, Sim2-deficient cells exhibited irregularly distributed β-catenin (Fig. 3N′) and E-cadherin (Fig. 3T,T′). These abnormal cell properties were observed at stages before overt curvature (Fig. S4C-J), confirming they precede, and are likely causal to, the stomach morphological defects. Interestingly, β-catenin was sometimes localized to the nucleus in sim2-MO-injected cells (Fig. S4H,I), a condition that has been associated with epithelial plasticity (e.g. Kim et al., 2019b); this observation is consistent with previous results suggesting that Sim2 function may be mediated by β-catenin-TCF signaling (Chen et al., 2014). At later stages, GFP-labeled apoptotic cells were occasionally found in the lumen of sim2-MO injected stomachs (asterisks, Fig. 3R,R′), suggesting the changes in Sim2-deficient cell properties had rendered them unable to intercalate and ultimately excluded them from the epithelial layer. All cellular phenotypes were rescuable by co-injection of exogenous MO-resistant sim2 mRNA (Fig. S4K-V), confirming that these effects are specific to Sim2 function. Taken together, our results indicate that Sim2 acts in the left stomach endoderm to govern the properties necessary for cells to radially rearrange and expand the left wall during curvature.

This study comprises the first genome-wide profiling of the left versus right halves of a developing organ as it acquires morphological LR asymmetry, revealing sim2 as a new LR gene. Our results reveal that, as in flies, single-minded genes also orchestrate asymmetric digestive organ morphogenesis in frogs, suggesting an intriguing cooptive convergence in this bHLH/PAS gene function in insect and amphibian lineages.

In both the fly and frog, sim/sim2 expression is confined to the epithelial layer of the gut tube, and our findings are consistent with the known roles of single-minded genes in regulating epithelial plasticity in other contexts. For example, in Drosophila, sim controls the delamination of neuroectoderm cells (Bossing and Technau, 1994; Thomas et al., 1988) and in the mammary epithelium, Sim2s (a short isoform) regulates duct and gland formation (Laffin et al., 2008). As in the stomach, Sim2s deficiency in the mammary epithelium perturbs cell polarity and decreases E-cadherin, contributing to metastatic epithelial-mesenchymal transitions (EMT) in breast cancer (Micalizzi et al., 2010). In other cancers, Sim2s is upregulated, repressing the expression of differentiation genes (e.g. Aleman et al., 2005; DeYoung et al., 2003a; Halvorsen et al., 2007; Lu et al., 2011). The ability of single-minded genes to modulate epithelial plasticity in a context-dependent manner may be advantageous in the developing stomach wall, where cell rearrangement (i.e. tissue expansion) and differentiation (i.e. gastric epithelial maturation) must be dynamically balanced (Davis et al., 2017).

The effects of sim2 deficiency are similar to those previously reported for loss of pitx2 function (Davis et al., 2017). However, in contrast to Pitx2, Sim2 is insufficient to drive stomach curvature itself, i.e., overexpression of sim2 on the right side has no effect on stomach morphology (Fig. S5), suggesting it may exert its effects only in the context of other bHLH/PAS dimerization partners or other cues also present only in the left stomach wall. It is notable that several potential direct targets of Sim2 (Letourneau et al., 2015) are LR asymmetrically expressed in the stomach. Interestingly, such molecules include vimentin, an intermediate filament protein involved in epithelial-mesenchymal plasticity, and Zic2, a zinc-finger molecule required for multiple earlier phases of LR asymmetry development (Dykes et al., 2018; Barratt et al., 2013).

In the fly, sim is expressed in the midline of both the CNS and gut tube and is a master regulator of midline development (Nambu et al., 1991), consistent with its association with pioneer factors and binding of super-enhancers (Letourneau et al., 2015). In vertebrates, Sim2 is also expressed in the midline of the brain (Marion et al., 2005) and, as shown here, in the dorsal midline of the foregut. It is well known that midline structures (e.g. notochord) are crucial for the acquisition and maintenance of the LR body axis, acting as physical and/or chemical barriers to isolate the left versus right sides of the early embryo (e.g. Bisgrove et al., 2000; Danos and Yost, 1996; Przemeck et al., 2003). The expression of sim2 in the midline of the foregut could serve a similar purpose within the stomach itself, segregating the distinct signaling and morphogenetic events occurring in the contralateral walls of a single organ.

In humans, SIM2 maps within the Down syndrome (DS)-critical region of chromosome 21 (Shamblott et al., 2002). In mammalian embryos, SIM2 is expressed in the brain and craniofacial structures affected in DS (Dahmane et al., 1995; Fan et al., 1996; Yamaki et al., 1996) and the triplication of SIM2, a transcriptional repressor (Moffett et al., 1997), may contribute to DS phenotypes (Chrast et al., 2000; Ema et al., 1999). SIM2 is also expressed in the human embryo stomach (Rachidi et al., 2005), although whether it is left-sided is unknown, and whether single-minded genes are expressed asymmetrically in the mouse embryo has not yet been determined. Nonetheless, we note that individuals with DS exhibit an increased incidence of upper GI conditions also observed in patients with LR anomalies, such as gastric inlet/outlet obstructions (e.g. duodenal atresia, esophageal atresia; Benjamin et al., 1996; Cleves et al., 2007; Heinke et al., 2020; Stoll et al., 2015), annular pancreas, diaphragmatic hernia (Morris et al., 2014) and gastroesophageal reflux disease (GERD; Holmes, 2014; Macchini et al., 2011). Our discovery that a DS-related gene is expressed LR asymmetrically suggests the intriguing possibility that altered dosage (i.e. ectopic or overexpression) of LR genes may contribute to phenotypes seen in non-laterality-related birth defects.

Animals were housed in the North Carolina State University College of Veterinary Medicine vivarium in accordance with Institutional Animal Care and Use Committee (IACUC) regulations. Xenopus laevis in vitro fertilization, embryo culture and staging (according to Nieuwkoop and Faber; Nieuwkoop and Faber, 1994) were performed according to established methods (Sive et al., 2000). L. laevis embryos were collected from natural matings as previously described (Amin et al., 2015) and were staged according to Gosner (Gosner, 1960). Embryos were raised at 23°C or 28°C and anesthetized according to standard procedure (Sive et al., 2000) before phenotyping or further processing for ISH and immunohistochemical analyses.

The left and right stomach walls of L. laevis were manually dissected from anesthetized embryos at GS18, GS19, GS20 and GS21 with sharpened forceps. At each stage, the anterior portion of the embryo (including the head, heart and pharyngeal region of the foregut) and the posterior portion of the embryo (immediately caudal to the liver) were removed and discarded, leaving only a transverse slice of tissue containing the stomach region of the foregut. The dorsal portion of the slice (containing the notochord, somites and pronephros) was discarded, leaving only the stomach, liver and ventral pancreas. The left and right halves of the stomach tube, i.e., the tissues on either side of the slit-like central lumen, surrounded by the outer epidermis, were then easily bisected and collected in Trizol (Ambion). Tissues from multiple embryos were pooled such that each replicate consisted of ten left or ten right half-stomachs; three pooled replicates were collected for each stage. The tissues were stored at −80°C until RNA isolation, which was accomplished by standard Trizol extraction followed by lithium chloride precipitation. Integrity and quantity of RNA yield was confirmed on a Thermo Fisher Scientific Nanodrop 1000 and an Agilent 2100 Bioanalyzer. RNA-seq was performed by Novogene with a poly-A selected non-stranded library prepared with NEBNext^® Ultra™ RNA Library Prep Kit (New England BioLabs). An Illumina Novaseq 6000 was used for sequencing (PE150; Q30≥80%).

Raw reads were trimmed for quality and removal of sequencing adaptors (NEBNext^® Ultra™ RNA Library Prep Kit) using Trimmomatic version 0.36 (Bolger et al., 2014). Trimmed, paired-end reads for each stage and replicate were independently aligned to the soft-masked L. laevis genome (provided by Austin Mudd and Daniel Rokhsar, University of California, Berkeley, USA) using HiSat2 version 2.1.0 (Kim et al., 2019a). The aligned reads were then sorted with samtools, and soft-clipping was removed with custom bash scripts (Li et al., 2009). Gene models were defined by cufflinks and merged into a single gene-transfer file (gtf format) with cuffmerge (Trapnell et al., 2012). A fasta file of the merged transcriptome was produced using gffread (Pertea and Pertea, 2020).

Differential expression analysis was performed by CuffDiff2 version 2.2.1 (Trapnell et al., 2013). The resulting files were then analyzed using the R package, cummeRbund version 2.30.0 (http://bioconductor.org/packages/devel/bioc/html/cummeRbund.html). Human gene symbols were assigned to the L. laevis genes using the top hit found by querying the L. laevis transcriptome with BLASTx against human protein sequences from Ensembl (Altschul et al., 1990; Zerbino et al., 2018).

Left or right-enriched genes were defined as having a log2 right/left ratio with a q-value ≤0.05 during at least one phase of stomach morphogenesis. Genes with no read counts in at least one sample were removed from further analysis; fifteen genes that reversed expression from significantly left-enriched to significantly right-enriched (or vice versa) during stomach morphogenesis were also excluded, as such expression patterns tended to be generated by outliers among the biological replicates. The R packages ‘VennDiagram’ and ‘Heatmap2’ were used to create figures (Chen, 2016; https://cran.r-project.org/web/packages/gplots/index.html) with only highly significant genes (q-value≤0.01).

Whole mount ISH was performed on Xenopus embryos as previously described (Dush and Nascone-Yoder, 2013). Embryos were fixed in MEMFA [100 mM MOPS (pH 7.4), 2 mM EGTA, 1 mM MgSO₄, 3.7% (v/v) formaldehyde] for 3 h, or 4% paraformaldehyde (PFA) overnight, at room temperature before being gradually dehydrated in methanol and stored at −20°C. Digoxigenin-labeled riboprobes were synthesized from linearized plasmids containing X. laevis Sox2, Sim2 or Pitx2 coding sequences using standard methodology (Sive et al., 2007).

ISH on tissue sections was performed as previously described (Butler et al., 2001) with the following modifications. Upon rehydration, sections were re-fixed with MEMFA for 30 min. After overnight probe hybridization, slides were washed with saline-sodium phosphate-EDTA, pH 7.4 (SSPE) before being treated with 20 µg/ml RNase A in 4× SSPE for 30 min at 37°C. Slides were blocked with 0.5% blocking reagent (Roche, 11096176001) in Buffer 1 (Butler et al., 2001) for 2 h at room temperature before incubating with anti-DIG-AP Fab fragments (Roche, 11093274910, 1:5000) overnight at 4°C. At the conclusion of the color reaction, slides were fixed for 15 min with 4% PFA before being washed with PBS plus 0.1% (v/v) Tween 20 (PBT), dehydrated to Xylene, mounted with Permount and cover-slipped.

Pharmacological dosing of Xenopus embryos with SB505124 (Sigma-Aldrich, S4696) was conducted as previously described (Davis et al., 2017). Briefly, SB505124 was diluted to 5 µM in Xenopus culture medium (0.1× MMR); an equal volume of DMSO (solvent) was used as control. Embryos were treated from stage NF19/20 through to collection stage.

Microinjection of loss-of-function reagents (i.e. gRNAs/Cas9 reagents and MO oligonucleotides; see Supplementary Materials and Methods, ‘Morpholino sequences’, for sequences) in Xenopus embryos was performed at the one-cell stage (300 pg gRNA), or at the eight-cell stage (8 ng MO), targeting a blastomere fated to contribute to the left or right side of the foregut, as previously described (Davis et al., 2017). Cas9 mRNA and sim2 gRNA were transcribed in vitro as previously described (Guo et al., 2014); Cas9 mRNA was injected at 200 pg/nl. The MO-resistant version of sim2 mRNA includes seven nucleotide changes that preserve the wild-type Sim2 amino acid sequence (see ‘Morpholino-resistant sim2 mRNA’ in Supplementary Materials and Methods).

For gain-of-function studies, a dexamethasone-inducible synthetic mRNA encoding Xenopus Pitx2c (Pitx2-GR) was injected at the eight-cell stage, targeting a blastomere fated to contribute to the right side of the foregut, as previously described (Davis et al., 2017). Pitx2-GR mRNA injected embryos were subsequently exposed to dexamethasone (10 µM) or ethanol (solvent control) from stage 19/20. In all microinjection experiments, synthetic GFP and/or RFP mRNA, transcribed in vitro using the mMessage mMachine kit (Thermo Fisher Scientific), was co-injected as a lineage tracer to validate proper tissue targeting.

Immunohistochemistry was performed on transverse stomach sections of control and experimental Xenopus embryos, as previously described (Davis et al., 2017). Embryos were fixed by washing eight times with Dent's fixative (80% methanol/20% DMSO) before overnight storage at −20°C. Processing for cryosectioning was carried out as previously described (Dush and Nascone-Yoder, 2013). Slides were post-fixed with 4% PFA [100 mM Hepes (pH 7.4), 100 mM NaCl, 4% PFA] for 2 min, washed with PBT, and blocked for 1 h. Immunohistochemical staining was performed overnight at 4°C with blocking buffer containing primary antibodies (see Supplementary Materials and Methods, ‘List of Antibodies’, for the list of primary and secondary antibodies used). Sections were washed with PBT, stained with secondary antibodies as previously described (Davis et al., 2017; Dush and Nascone-Yoder, 2013), and slides were again washed with PBT and PBS before staining with Topro-3 (Thermo Fisher Scientific, T3604, 1:1000) in PBS for 5 min. Final slides were washed with PBS and mounted with Prolong Gold (Thermo Fisher Scientific), cured overnight in the dark, and visualized with a Leica DM 2500 confocal microscope.

Morphometric measurements of the width of the left and right stomach walls were performed using ImageJ software (National Institutes of Health). For ISH and immunohistochemistry data, multiple sections (anterior to posterior) of at least three different embryos were analyzed for each experimental condition. All perturbation experiments were performed at least in triplicate (total n=35-155 embryos per condition, with similar results across independent trials). Statistical significance was analyzed by one-way ANOVA with post-hoc Tukey HSD.

We are indebted to the North Carolina State University (NCSU) Genomic Sciences Laboratory, Betsy Scholl (NCSU Bioinformatics Consulting and Service Core) and Dereje Jima (NCSU Center for Human Health and the Environment) for bioinformatics consulting, Alex Dornburg (The University of North Carolina at Charlotte) for assistance with phylogenetics, and members of the Nascone-Yoder lab for helpful comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199265