Loss of Elmo2 leads to carotid artery aneurysm and embryonic lethality
In order to understand the in vivo function of Elmo2, global knockout mice (Elmo2−/−), obtained after ubiquitous Cre-mediated recombination of a “knockout-first” allele (Supplementary Fig. 1a and b), were compared to control littermates (Elmo2+/+) at different embryonic stages. Efficient deletion of Elmo2-encoded transcripts and protein were confirmed by RT-qPCR and Western blot, respectively (Supplementary Fig. 1c and d). The first macroscopic evidence of deleterious phenotypic alterations (Fig. 1a) was detected in E12.5 Elmo2−/− embryos, which show small vascular lesions in the head and neck region. These lesions worsened during the following days of development, giving rise to subcutaneous edema and severe hemorrhages in the cervical region by E13.5 and E14.5. No surviving embryos were obtained beyond E15/E15.5.
Taking advantage of β-galactosidase expression from the targeted allele, heterozygous (Elmo2+/−) embryos, which develop normally, were assessed by X-Gal staining at E12.5. This approach revealed expression of Elmo2 in different structures including the dorsal aorta, laryngotracheal groove, vagus nerve, sympathetic chain ganglia, trachea, esophagus and pharyngeal arch arteries (Fig. 1b and Supplementary Fig. 1e). In line with this expression pattern, the histological analysis of transverse sections from the cervical region of mutant embryos revealed a severe dilation of the third pharyngeal arch artery (3rd PAA) at E12.5 and of the carotid arteries at E13.5 (Fig. 1c-e and Supplementary Fig. 1f).
The complex and dynamic morphogenetic program that shapes hierarchical blood vessel organization in the trunk and cervical area (Supplementary Fig. 1g) prompted us to analyze the three-dimensional organization of the vascular tree in control and Elmo2−/− embryos from E11.5 to E13.5 using whole-mount staining and light-sheet microscopy. This analysis showed that the first vascular defect in global knockout embryos, namely the dilation of the 3rd PAA, arises at E12.5. This vessel further remodels giving rise to the common carotid arteries38, which are severely dilated resulting in fusiform aneurysm formation in E13.5 mutants (Fig. 1g). It is worth noting that these defects in vascular shape and diameter are rather specific to the 3rd PAA and only mildly affect other major vessels, including neighboring pharyngeal arch arteries (Supplementary Fig. 2a-e).
Loss of Elmo2 leads to alterations in endothelial and vascular smooth muscle cells
To characterize the vascular defects in Elmo2−/− embryos in greater detail, the 3rd PAA and carotid arteries were analyzed by immunofluorescence staining and high-resolution confocal microscopy. This revealed discontinuities in the endothelial lining (Supplementary Fig. 3a) and significant changes in the size and morphology of ECs and their nuclei (Fig. 2a-c) in mutant embryos relative to littermate controls. In addition, defects in the polarized expression of the luminal marker Podocalyxin were observed (Fig. 2d) as well as ectopic expression of the VSMC markers α-smooth muscle actin (αSMA) and SM22α in ECs (Fig. 2e and f). Despite normal and domain-specific expression of markers for lymphatic vessels (Prox1) and arteries (SOX17) (Fig. 1e and 2a), clusters of ECs in the dilated carotid arteries of Elmo2−/− embryos express detectable levels of Endomucin, a marker that is normally absent from the arterial endothelium (Fig. 2g and h).
VSMC contractility is an important regulator of vascular tone in the adult organism but also in the embryo39,40. We therefore analyzed VSMCs around the 3rd PAA and carotid arteries by immunostaining against proteins associated with the contractile phenotype. This revealed that the expression levels of SM22α, αSMA and Calponin1 are comparable between control and Elmo2−/− embryos (Supplementary Fig. 3b-g), arguing against major defects in VSMC abundance and differentiation. Likewise, no overt differences were detected for Nestin (Supplementary Fig. 3f), whose expression has been associated to the synthetic phenotype of VSMCs41. The staining intensity and distribution pattern of phospho-myosin light chain 2, a functional marker of VSMC contractility, also does not show obvious changes in the wall of mutant carotid arteries relative to littermate controls (Supplementary Fig. 3c).
Although the expression of VSMC identity and differentiation markers appears unaffected, super-resolution confocal microscopy revealed that the normal alignment of αSMA+ bundles with respect to the longest axis of the underlying endothelium is severely compromised in Elmo2−/− embryos (Fig. 2i-k). Whereas αSMA+ fibers are oriented perpendicular to elongated ECs and thereby to the direction of blood flow in the control E12.5 3rd PAA, both EC elongation and the orientation of αSMA+ bundles are disorganized after loss of Elmo2. Furthermore, αSMA and SM22α immunosignals, which are strongly concentrated near the subendothelial basement membrane in VSMCs of control carotid arteries, have lost their normal polarization and multiple peaks of high staining intensity can be detected throughout the Elmo2−/− vessel wall (Fig. 2l and m). The same analysis also confirmed the abnormal expression of VSMC markers in the Elmo2−/− carotid artery endothelium (Fig. 2m).
Next, we assessed whether significant changes in EC or VSMC proliferation are associated with the vessel enlargement in Elmo2−/− mutants. To this end, we labelled mitotic cells in vivo by injection of 4-Ethynyl-2’-deoxyuridine (EdU) into pregnant females. This approach revealed strong increases in the absolute number of EdU+ ECs and VSMCs in the 3rd PAA of E12.5 mutant embryos. However, these increases are no longer statistically significant after normalization to vessel perimeter and may therefore reflect the dilation of Elmo2−/− carotid arteries (Supplementary Fig. 4a-d). Altogether, these results establish that ELMO2 is required for the normal development of the 3rd PAA and common carotid artery during embryogenesis.
Transcriptomic analysis of Elmo2 mutants at single cell resolution
To gain insight into the molecular changes resulting from the inactivation of Elmo2, we performed single cell RNA-sequencing (scRNA-seq) of the 3rd PAA and surrounding mesenchyme dissected from control (Elmo2+/+), heterozygous (Elmo2+/−) and mutant homozygous (Elmo2−/−) E12.5 embryos. Integrated analysis of the transcriptome from these samples allowed identification of seven major cell populations with distinct expression signatures and enrichment of specific markers (Fig. 3a and Supplementary Fig. 5a). The most abundant cell type is the mesenchymal stromal cell (MSC) population, which represents more than 70% of all cells. MSCs are followed by endothelial, immune and muscle cells, which are found in similar proportions and together represent ~ 20% of total cells. The remaining cell types are mostly erythrocytes, neurons and epithelial cells (Supplementary Fig. 5b). As expected, Elmo2 transcript expression is proportionally reduced in heterozygotes and is below the detection threshold in Elmo2−/− samples (Fig. 3b). Among the members of the Elmo family, Elmo2 has the highest expression, followed by Elmo1 and Elmo3 (Supplementary Fig. 5c), which do not show significant compensatory upregulation upon deletion of Elmo2 (Supplementary Fig. 5d). Furthermore, expression of Elmo2 is rather homogeneous across the different cell clusters with the highest level found in neurons and the lowest in the erythroid lineage (Fig. 3c).
Considering that the phenotypic changes in Elmo2−/− mutants affect mostly the vascular compartment, the control and homozygous mutant EC population in our scRNA-seq data was subclustered for deeper analysis. Three main subsets with distinct markers were identified in a two-dimensional (2D) Uniform Manifold Approximation and Projection (UMAP) representation, namely venous, arterial and lymphatic ECs (Fig. 3d and Supplementary Fig. 5e). Interestingly, color-labelling of cells corresponding to the control (Elmo2+/+) or knockout (Elmo2−/−) samples within the subclustered EC dataset, highlighted an area characterized by overrepresentation of mutant cells in a specific 2D spatial location within the arterial subcluster (Fig. 3e). In addition, differential gene expression analysis (DEG) allowed the identification of de-regulated genes in Elmo2−/− cells relative to control. Interestingly, when a stringent selection criterion for highly statistically significant values is used (p-adjusted < 1− 10), only a single gene (Elmo2) is downregulated. In contrast, 38 genes are upregulated, 9 of them with a log2 fold change above 2 (Fig. 3f). Notably, all these upregulated genes are either exclusively or primordially expressed in the mutant cell hotspot within the arterial subcluster (Fig. 3g). Among the upregulated genes, Acta2 and Tagln were previously identified during our histology analysis because of their ectopic expression in arterial ECs of mutant embryos (Fig. 2e and Supplementary Fig. 3b).
Next, we followed a similar approach for the identification of subpopulations within the mesenchymal stromal cells (MSCs), which are a source of VSMCs during development42,43. Both the UMAP representation (Fig. 3h) and marker analysis (Supplementary Fig. 5f) indicate that the differences between the 8 identified MSC subtypes are less defined than those found during the EC subclustering, potentially reflecting ongoing differentiation and incomplete terminal phenotypic specification. Likewise, cellular distribution of control and Elmo2 mutant cells within the MSC UMAP plot is rather homogeneous (Fig. 3i) and only a few genes were found to be de-regulated (log2 fold change > 2 or <-2) when a cut-off for highly statistically significant differences (p-adjusted < 1− 10) is applied (Fig. 3j). Using this criteria, 4 downregulated (Elmo2, Hoxb6, Car2 and Capn11) and 3 upregulated genes (Cnmd, Matn1 and Acan) were identified, without clear functional relationships among them. A similar profile of very limited or non-significant changes in gene expression was found for the other cell populations in our scRNA-seq data (Fig. 3k).
With the aim of gaining a broader understanding of biological processes potentially affected by gene expression changes in Elmo2−/− ECs and MSCs, a gene set enrichment analysis including all de-regulated genes with a (less stringent) p-adjusted cut-off value of 0.01 and a log2 fold change > 0.5 or < -0.5 was performed. From the top gene ontology terms found (Supplementary Fig. 5g) there is no explicit relation to blood vessel development either in the ECs or MSCs population, yet different aspects related to extracellular matrix organization are highlighted for both cell types. Thus, unexpectedly, the analysis of the sc-RNAseq data reveals rather limited changes in gene expression and provides no clear explanation for the dramatic changes in the mutant common carotid arteries.
Inactivation of Elmo2 in endothelial and smooth muscle cells
For cell type-specific loss-of-function experiments, a conditional (loxP-flanked) allele of Elmo2 (Supplementary Fig. 1a) was established and validated by breeding it to homozygosity in a PGK-Cre+/T background44. As expected, ubiquitous expression of constitutively acting Cre led to widespread Elmo2 inactivation and phenocopied the vascular defects seen Elmo2−/− embryos generated with the “knockout-first” approach (Supplementary Fig. 6a-c). Next, we generated EC-specific mutants by interbreeding of mice carrying the floxed Elmo2 allele and Tek-Cre transgenic animals45. Analysis with the R26-mTmG Cre-reporter46 confirmed successful Tek-Cre-mediated recombination in the embryonic endothelium (Supplementary Fig. 7a). However, EC-specific Elmo2 mutants (Elmo2ΔEC) showed no observable defects in vascular development (Supplementary Fig. 7b). In particular, the size and morphology of Elmo2ΔEC carotid arteries (Supplementary Fig. 7b-c) or the expression of known markers of EC or VSMC (Supplementary Fig. 7d) are indistinguishable from control littermates. These results argue that the vascular malformations observed in global Elmo2 knockout embryos are not caused by cell-autonomous defects in the endothelium.
Next, we conducted genetic experiments to address whether Elmo2 is required in VSMCs. To this end, a transgene expressing constitutive Cre under the transcriptional control of Transgelin (SM22α) (Tagln-Cre47) was introduced into the Elmo2 conditional (floxed) background and embryos at specific developmental stages were collected. Unexpectedly, the resulting smooth muscle cell-specific knockout embryos (Elmo2ΔSMC) were macroscopically indistinguishable from control littermates at E13.5 (Fig. 4a). Immunostaining-assisted analysis of histological sections revealed subtle, yet statistically significant dilation of the carotid arteries at E13.5 and E15.5 (Fig. 4b-c and Supplementary Fig. 8a). Despite carotid artery dilation, other relevant features of the global knockout phenotype were not reproduced after Tagln-Cre-mediated deletion of Elmo2. In particular, there were no signs of aneurysm formation, disorganized VSMC alignment, or altered endothelial morphology, polarity and gene expression (Fig. 4d and Supplementary Fig. 8b).
To rule out that the failure to reproduce the global knockout phenotype is caused by suboptimal Elmo2 deletion in VSMCs, the recombination efficiency of Tagln-Cre in E13.5 embryos was assessed with the R26-mTmG Cre-reporter. Analysis of GFP expression as surrogate marker of recombination clearly showed that the vast majority of VSMCs and mesenchymal cells around the carotid arteries (Fig. 4e) and neighboring vessels (Supplementary Fig. 8c) are efficiently targeted by the Tagln-Cre. In summary, these results argue that Elmo2 deletion in VSMCs by means of a Transgelin-driven constitutive Cre-recombinase is not able to fully phenocopy the effects elicited upon global gene inactivation.
Neural crest-specific deletion of Elmo2 phenocopies the global Elmo2 KO
Neural crest cells contribute to many craniofacial tissues and are an important source of mural cells in the 3rd PAA and thereby the common carotid arteries37,48. We chose Wnt1-Cre2 transgenic mice49 to study NCCs and their progeny, which would also address potential roles early in mural cell differentiation, whereas Tagln-Cre targets more differentiated VSMCs.
In order to assess whether Wnt1-Cre2 allows targeting of neural crest-derived VSMCs progenitors and compare the recombination timing with that of Tagln-Cre, lineage tracing analysis using the R26-mTmG reporter allele were carried out for both lines. Notably, Wnt1-Cre2-mediated labelling allows detection of a large number of GFP+ cells that cluster around the vascular plexus giving rise to the 3rd PAA by E9.5, whereas Tagln-Cre-mediated recombination is restricted to the heart at the same stage (Supplementary Fig. 9a). In line with this result, expression of the VSMC markers αSMA and SM22α is limited to the heart and cannot be detected in mural cells around the blood vessels in the branchial arches (Supplementary Fig. 9b). Expression of SM22α in mural cells or the 3rd PAA is first detected at E10.5, which coincides with the emergence of Tagln-Cre-labelled GFP+ cells in in this structure (Supplementary Fig. 9c). In contrast, at this timepoint, a much larger number of Wnt1-Cre2-traced cells wrap around the 3rd PAA forming a surrounding layer that consists of both SM22α+ VSMCs and yet undifferentiated cells (Supplementary Fig. 9c). At E11.5, recombination with both Cre lines generates robust perivascular GFP labeling in the relevant region (Supplementary Fig. 9d).
We further analyzed the recombination pattern elicited by Wnt1-Cre2 in E13.5 embryos with special interest to the carotid arteries and neighboring vascular structures. Wnt1-Cre2-mediated recombination targeted VSMCs around the carotid arteries with high efficacy but, as expected, spared the mesoderm-derived smooth muscle surrounding the vertebral arteries, dorsal aorta, and jugular veins, as well as the endothelial lining of blood vessels (Fig. 5a and Supplementary Fig. 10a-c). Taken together, these results prove that Wnt1-Cre2 efficiently targets NCC-derived VSMCs and their progenitors in the early mouse embryo.
Remarkably, neural crest-specific mutants (Elmo2ΔNCC), generated by interbreeding of the Elmo2 conditional line with Wnt1-Cre2, reproduce the vascular defects observed in the global knockout model. Macroscopic observation of E13.5 Elmo2ΔNCC embryos revealed severe hemorrhaging in the cervical region and dorsal edema (Fig. 5b) as well as massive dilation of the carotid arteries with aneurysm formation (Fig. 5c and d). Moreover, the defects in the hierarchical expression pattern of αSMA and SM22α found in Elmo2−/− embryos are also present upon neural crest-specific deletion (Fig. 5e), as well as the disorganized alignment of αSMA+ actin bundles with respect to the ECs’ longest axis (Fig. 5f and g).
Strikingly, phenotypic changes observed in the ECs of the global knockout, such as retained expression of Endomucin in arterial territories, discontinuous endothelial lining and ectopic expression of mesenchymal markers are also seen upon neural crest-specific Elmo2 deletion (Fig. 5h and i). Given that Wnt1-Cre2-driven recombination spares the endothelium, this result is further evidence that the EC defects in Elmo2 mutants are secondary and probably a consequence of the severe vessel dilation.
Since neural crest cells give rise to parts of the autonomic nervous system and might therefore control vascular tone through VSMC innervation50, we bred the Elmo2 conditional knockout mice with the TH-IRES-Cre line51, which directs Cre recombination to catecholaminergic sympathetic neurons. Efficient and precise targeting of the paravertebral sympathetic ganglia is confirmed by GFP expression in the R26-mTmG Cre reporter background (Supplementary Fig. 10d). Nevertheless, Elmo2 deletion in these structures did not induce relevant phenotypic alterations and the resulting mutants survived to term and were obtained slightly above the expected ratio at birth (27.3% instead of 25%). Altogether, these results indicate that Elmo2 is essential for vascular diameter control in the developing carotid artery by regulating the properties of NCC-derived VSMC progenitors.
ELMO2 controls contractile ability and actin dynamics of human VSMCs in vitro
To gain insight into the cellular function of ELMO2, we next conducted experiments in cultured VSMCs. We opted for human brain vascular smooth muscle cells (HBVSMCs) because these cells, just like those wrapping around the 3rd PAA and carotid artery, are of neural crest origin52. Moreover, HBVSMCs express ~ 1000-fold higher levels of ELMO2 compared to ELMO1 (Supplementary Fig. 11a).
Silencing RNA (siRNA)-mediated knockdown (KD) of ELMO2 in HBVSMCs significantly decreased transcript abundance already by 24h after treatment and this effect was maintained over several days (Fig. 6a). At the protein level, significant reduction of ELMO2 was obvious at 48h after siRNA treatment, with the highest depletion achieved at 72-96h (Fig. 6b), suggesting a slow protein turnover rate. In vitro, siELMO2-treatment induced a compensatory ~ 2-fold increase in ELMO1 transcription, which coincides with the timepoints of highest ELMO2 protein depletion (Supplementary Fig. 11b), yet this upregulation may be of limited functional significance given the much higher endogenous expression of ELMO2 in HBVSMCs.
Similar to our in vivo observations, siELMO2-treated cells did not show relevant changes in the expression of known VSMC markers relative to siControl cells (Supplementary Fig. 11c-d). Likewise, the overall abundance of previously described ELMO2 interaction partners or downstream effectors relevant for cell adhesion, extracellular matrix binding and cell contractility were unchanged at the protein level (Supplementary Fig. 11e). The subcellular localization of integrin-linked kinase, a well described interactor of ELMO2, and the phosphorylation of myosin light chain 2, a key regulator of cell contractility, were comparable in KD and control cells (Supplementary Fig. 11d and f). On the contrary, analysis of F-actin by phalloidin staining (Supplementary Fig. 11d) revealed a slight reduction in the intensity and abundance of stress fibers in siELMO2 HBVSMCs, whereas cortical actin appeared unaffected.
Next, we analyzed the functional performance of siELMO2 HBVSMCs in assays requiring active remodeling of the cytoskeleton. Loss of ELMO2 reduced cell attachment and spreading in Collagen I-coated culture plates at early timepoints (10min to 2h) relative to siControl cells. These defects were no longer detectable at 6h after seeding, when both the area and number of cells attached are indistinguishable between the KD and control conditions (Fig. 6c-d and Supplementary Fig. 11g). Time-lapse video recordings from live-imaging co-culture experiments confirmed the delayed adhesion of siELMO2 cells (Supplementary Fig. 11h-i). Contrary to control cells, which extend filopodia-like cytoplasmic projections, spherical protrusions (blebs) were continuously formed and retracted in the membrane of KD cells (Fig. 6e and Supplementary Movie 1a and b). Moreover, culture of siELMO2 HBVSMCs in 3D fibrin hydrogels lead to a similar reduction in cell spreading and cellular area relative to siControl cells, consistent with the results seen in 2D experiments (Fig. 6f and g).
In order to test the contractile capacity of VSMCs in vitro, we treated co-cultured control and KD cells with carbachol, a cholinergic agonist that increases cytoplasmic calcium levels and stimulates the RhoA/ROCK (Rho-associated kinase) pathway53. Live-imaging analysis showed significantly impaired contractility of ELMO2 KD cells, which were not able to efficiently retract their cellular projections upon carbachol treatment (Fig. 6h-i and Supplementary Fig. 11j). Further verifying this observation, 3D collagen gel contraction assays confirmed that siELMO2-treatment drastically impaired the contractile capacity of HBVSMCs (Fig. 6j-k). These results point out to an important role for ELMO2 in the regulation of HBVSMC actin dynamics in the context of cell adhesion, spreading and contraction.
With the aim of uncovering the reasons behind the deficient regulation of actin dynamics after ELMO2 downregulation, the ratio of globular (G)-actin to filamentous (F)-actin was determined as a readout of actin polymerization. F-actin abundance was found to be significantly decreased in siELMO2 HBVSMCs relative to siControl cells (Fig. 6l), which showed a higher F-actin to G-actin ratio (Fig. 6m). These changes in the functional state of the actin cytoskeleton may be, at least in part, a consequence of reduced active Rac1 in ELMO2 KD cells, as shown by a G-LISA-based activation assay (Fig. 6n).
The reduced abundance of F-actin prompted us to test the impact of pharmacological actin filament stabilization. For this, siControl and siELMO2 HBVSMCs were analyzed after treatment with jasplakinolide (JAS), a cyclic peptide known to bind and stabilize filamentous actin in vitro54. Notably, JAS allowed efficient formation of filopodia-like structures and improved adhesion and spreading of siELMO2 cells to an extent that makes them undistinguishable from control cells (Fig. 6o-p and Supplementary Fig. 11k). Likewise, siELMO2 cells recovered their contractile ability after JAS treatment in the collagen gel contraction assays (Fig. 6q-r). Altogether, these data indicate that promoting actin polymerization and stabilization restores functional features of ELMO2 deficient HBVSMCs.
Timing of global Elmo2 deletion is determinant for aneurysm formation
Our in vitro data point out to contractility defects as the most likely cause for the vascular dilation phenotype in vivo. Yet, the most severe phenotypic alterations are only triggered upon Elmo2 deletion in neural crest derived progenitors (Wnt1-Cre2-mediated recombination) and do not reach the same extent when contractile VSMCs are targeted (through Tagln-Cre). In this regard, the early onset of Wnt1-Cre2-mediated recombination in the 3rd PAA (Supplementary Fig. 9) and the mild phenotype of Tagln-Cre-generated Elmo2 mutants raise the possibility that the gene product is required during an early stage of vessel wall assembly.
In order to directly assess this hypothesis, we bred the Elmo2 conditional knockout model with mice expressing an inducible recombinase (CreERT2) under control of the ubiquitously expressed Rosa26 locus (R26-CreERT2 55). Pregnant dams were treated with 4-hydroxytamoxifen (4-OHT) at defined stages of embryonic development in order to resemble the recombination timing achieved with Wnt1-Cre2 and Tagln-Cre lines.
4-OHT treatment at E8.5 and E9.5, which coincides with the timing of Wnt1-Cre2 activity, induced severe vascular defects including carotid artery aneurysm by E13.5 (Fig. 7a-c). These defects are identical to those observed upon global or NCC-specific Elmo2 inactivation and correspond to a complete depletion of ELMO2 protein (Fig. 7d). In contrast, 4-OHT administration from E10.5 to E11.5, mimicking the later recombination by Tagln-Cre, did not lead to aneurysm formation by E13.5 but induced a milder dilation of the carotid arteries (Fig. 7e-g). Given that the short time period between 4-OHT treatment and analysis in this treatment regime results in incomplete depletion of ELMO2 protein (Fig. 7h), we tested whether a longer waiting interval could increase the severity of the resulting phenotype. Interestingly, 4-OHT administration from E10.5 to E11.5 with embryo collection at E15.5 (Fig. 7i-k) did not aggravate the arterial dilation despite efficient ELMO2 reduction (Fig. 7l).
These results indicate that ELMO2 function is essential in the 3rd PAA in the early embryo but no longer required for carotid artery development in the second half of gestation. The early emergence of these defects in global and NCC-specific mutants establishes that Elmo2 is directly involved in the regulation of vascular morphogenesis, which argues that vascular defects are probably not secondary to skeletal overgrowth in VMOS patients. Furthermore, it is striking that ELMO2 is specifically required in the vessels supplying the mandible and maxilla, which raises the possibility that the defects in these skeletal elements are a consequence of the vascular dilation.