Context-Dependent Functional Divergence of the Notch Ligands DLL1 and DLL4 In Vivo

Abstract

Notch signalling is a fundamental pathway that shapes the developing embryo and sustains adult tissues by direct communication between ligand and receptor molecules on adjacent cells. Among the ligands are two Delta paralogues, DLL1 and DLL4, that are conserved in mammals and share a similar structure and sequence. They activate the Notch receptor partly in overlapping expression domains where they fulfil redundant functions in some processes (e.g. maintenance of the crypt cell progenitor pool). In other processes, however, they appear to act differently (e.g. maintenance of foetal arterial identity) raising the questions of how similar DLL1 and DLL4 really are and which mechanism causes the apparent context-dependent divergence. By analysing mice that conditionally overexpress DLL1 or DLL4 from the same genomic locus (Hprt) and mice that express DLL4 instead of DLL1 from the endogenous Dll1 locus (Dll1Dll4ki), we found functional differences that are tissue-specific: while DLL1 and DLL4 act redundantly during the maintenance of retinal progenitors, their function varies in the presomitic mesoderm (PSM) where somites form in a Notch-dependent process. In the anterior PSM, every cell expresses both Notch receptors and ligands, and DLL1 is the only activator of Notch while DLL4 is not endogenously expressed. Transgenic DLL4 cannot replace DLL1 during somitogenesis and in heterozygous Dll1Dll4ki/+ mice, the Dll1Dll4ki allele causes a dominant segmentation phenotype. Testing several aspects of the complex Notch signalling system in vitro, we found that both ligands have a similar trans-activation potential but that only DLL4 is an efficient cis-inhibitor of Notch signalling, causing a reduced net activation of Notch. These differential cis-inhibitory properties are likely to contribute to the functional divergence of DLL1 and DLL4.

Fig 1. Mesodermally expressed CAG:DLL1 but not CAG:DLL4 functionally replaces endogenous DLL1 during somitogenesis.

(A) Structure of unrecombined and recombined (bottom) pMP8.CAG-Stop/Dll vector for integration of Cre-inducible expression constructs into the Hprt locus. 5‘hom and 3‘hom, 5’ and 3’ homology regions from the Hprt gene for homologous recombination; ex (grey boxes), HPRT exons; CAG prom, CAG promoter to drive transgene expression; neo ^r, neomycin phosphotransferase; pA, polyadenylation signal; Dll1/4–Venus, Dll1 or Dll4 ORF–joined to the reporter gene Venus by an internal ribosomal entry site (IRES); hHPRT prom, human HPRT promoter; light/dark grey triangles, loxP/loxM sites (in “flip excision” orientation); “Cre” arrow, Cre-mediated recombination. (B) Venus reporter expression in E8.5 CAG:Dll1 and CAG:Dll4 embryos indicated ubiquitous transgene activation after ZP3:Cre-mediated recombination. As expected, overall fluorescence in female embryos (a,c) was weaker than in male embryos (b,d) due to random X-chromosome inactivation. Numbers of embryos analysed are given in bottom right corner. (C) Quantification of Venus protein (CAG:Dll1 set to one) by Western blot analysis of embryo lysates with anti-GFP antibodies (and anti-β-actin antibodies for normalisation) showed similar expression levels. (D) For direct comparison of DLL protein levels, we also integrated single copies of Dll1 and Dll4 labelled with C-terminal HA-tags following the strategy in (A) using recombined (active) constructs for electroporation of embryonic stem (ES) cells. Western blot analysis of three ES cell clones expressing either of these transgenes using anti-HA antibodies confirmed similar expression levels with expected mild clonal variations (a); means of all three CAG:Dll1-HA and all three CAG:Dll4-HA clones are shown in (b). (E) Cranial-caudal somite patterning visualised by whole mount in situ hybridisation of E9.5 embryos with an Uncx4.1 probe showed an extensive rescue of somitogenesis plus ectopic Notch activation by CAG:DLL1 (b,c) but no appreciable rescue of somitogenesis by CAG:DLL4 (e,f). Insets in Ea-c and Ee-g show higher magnifications of the regions as indicated. Error bars represent standard error of the mean (SEM); ns, not significant; *, P<0.05; **, P<0.01.

Fig 2. Generation of Dll1 ^Dll4ki mice that express Dll4 instead of Dll1 in the endogenous Dll1 domains.

(A) Targeting strategy to insert a Dll4 mini gene into the Dll1 locus. The Dll1 locus contains 11 exons depicted as black boxes (UTRs as white boxes). The targeting construct is comprised of the Dll4 mini gene [Dll4 cDNA from start codon (ATG) in exon 1 to exon 9 (large red box), Dll1 intron 9, Dll4 exon 10 (small red box), Dll1 intron 10 and Dll1 exon 11 that encodes only the terminal valine conserved between Dll1 and Dll4 followed by STOP codon and 3‘UTR], a floxed neo ^r cassette, homology regions for integration between Dll1 start codon and exon 2, and flanking diphtheria toxin genes (DT); insertion of the mini gene is expected to disrupt expression of Dll1. neo ^r is removed by Cre-recombination. The resulting Dll1 ^Dll4ki allele and the Dll1 ^Dll1ki control are shown below (blue boxes, Dll1 mini gene). (B) Heterozygous adult Dll1 ^Dll4ki mice frequently (89%) displayed a kinky tail (arrow in b) but looked otherwise normal. (C) Heterozygous E15.5 Dll1 ^Dll4ki foetuses (c) were indistinguishable from wildtype (wt; a) and homozygous Dll1 ^Dll1ki (b) foetuses while all homozygous Dll1 ^Dll4ki foetuses (d) displayed shortened body axes and large oedemas. (D) Dll1 and Dll4 expression in Dll1 ^Dll4ki and Dll1 ^Dll1ki embryos visualised by whole mount in situ hybridisation of E9.5 embryos of the indicated genotype with a Dll4 ORF, Dll1 ex11 (recognises transcripts from both mini genes) and Dll1 ORF probe confirmed that Dll4ki alleles expressed Dll4 but not Dll1 in Dll1 expression domains (here the PSM, arrowheads). a-c were stained in parallel and colour development was stopped before endogenous Dll4 expression [49] and background became visible. Homozygous Dll1 ^Dll4ki embryos show strong expression in neuroectoderm (white arrow in c; not visible in the weaker staining with Dll1 ex11 probe in f). (E) Northern blot analysis of homozygous Dll1 ^Dll4ki and Dll1 ^Dll1ki E11.5 embryos, 2 μg polyA(+)-RNA loaded per lane, hybridised with 3‘UTR (Dll1 ex11) and β-actin probes; quantification of transgene signals relative to actin is shown at the bottom and indicates similar expression levels. (F) Visualisation of DLL1 and DLL4 expressed in the PSM of homozygous Dll1 ^Dll4ki (a-f) and Dll1 ^Dll1ki E9.5 embryos (g-l) using specific anti-DLL1 and anti-DLL4 antibodies. Co-staining with anti-panCadherin antibodies, which mark the plasma membrane, confirms that transgenic DLL4 and DLL1 predominantly localise to the cell surface (c,l). The lack of DLL1 signal in Dll1 ^Dll4ki (d) and of DLL4 signal in Dll1 ^Dll1ki PSMs (g) confirm the specificity of stainings. Both in anti-DLL4 and anti-DLL1 antibody stainings of PSMs, we observed spots of high signal intensity that may result from accumulation of ligands at these sites and that had also been observed in wildtype PSMs stained with anti-DLL1 antibodies [21]. Scale bars, 10 μm; insets show magnifications of the dotted boxes in c,l.

Fig 3. Homozygous Dll1 ^Dll4ki mice fail to generate proper somites and form reduced skeletal muscle tissue.

Examination of Dll1-dependent (A,B) somitogenesis and (C-F) myogenesis in (a) wildtype, (b) Dll1 ^lacZ/lacZ, (c) Dll1 ^{Dll1ki/Dll1ki}, (d) Dll1 ^Dll4ki/+ and (e) Dll1 ^{Dll4ki/Dll4ki} embryos or foetuses. (A) Uncx4.1 in situ hybridisation of E9.5 embryos. (B) Skeletal preparations of E18.5 foetuses (Dll1 ^lacZ/lacZ foetuses do not survive until E18.5; red arrowheads indicate fused ribs or hemivertebrae in heterozygous Dll1 ^Dll4ki skeletons in d). (C) Myogenin in situ hybridisation to visualise differentiating skeletal muscle cells in myotomes of 17–18 somite stage embryos. (D, E,F) Anti-myosin heavy chain (MHC)-antibody staining of sectioned E15.5 foetuses showing intercostal muscles (D), the diaphragm (E), and muscles in the cross-section of forelimbs (F); black arrowheads indicate examples of muscle tissue, red arrowheads show lack of muscle tissue; asterisks label ribs (D) or bones of the forelimb (F).

Fig 4. DLL4 expressed from the Dll1 locus rescues DLL1 loss-of-function in the retina.

(A) Dll1 null mutant retinas show epithelial disruption with formation of polarised rosettes in which the apical markers N-Cadherin (NCad, a) and ZO-1 (ZO1, b) are abnormally present at the central lumen. Ectopic proliferating progenitors, labelled with PHH3 (b, arrowheads), are located close to the apical lumen of these rosettes. (B) In contrast, the neuroepithelium of homozygous Dll1 ^Dll1ki and Dll1 ^Dll4ki embryos is correctly organised without rosettes, and N-Cadherin shows the normal apical localisation close to the retinal pigmented epithelium (a,b). Mitotic progenitors (PHH3+) are only detected at the apical region of the neuroepithelium (a,b arrowheads). A normal stratification of CHX10+ progenitors and P27+ differentiating neurons is also observed (c,d). (C, D) E13.5 homozygous Dll1 ^Dll1ki and Dll1 ^Dll4ki retinas show no significant difference in the number of ISL1+ RGCs (C) and CRABP+ amacrine cells (D). Cells immunopositive for Islet-1 and Crabp were counted and related to the total number of cells in the retina (DAPI+). Percentages are shown as mean ± SEM; ns, not significant. (E) Expression of DLL4 in homozygous Dll1 ^Dll1ki (a,c) and in homozygous Dll1 ^Dll4ki (b,d) E13.5 retinas as detected by an anti-DLL4 antibody. (c) and (d) are magnifications of (a) and (b), respectively. Endogenous plus transgenic DLL4 is expressed in more cells in Dll1 ^{Dll4ki/Dll4ki} as compared to endogenous DLL4 expression in Dll1 ^{Dll1ki/Dll1ki} while signal strength is similar. Scale bars are 50 μm in (A, B) and 100 μm in (E).

Fig 5. DLL1 and DLL4 trans-activate Notch with similar efficiency, but only DLL4 is an effective cis-inhibitor.

(A) Flag-tagged Dll1 and Dll4 ORFs were inserted into a randomly integrated attP site in CHO^attP cells mediated by ΦC31 site-directed recombination (upper part). Resulting cells were used in Notch-activation assays in combination with HeLa-N1 cells as schematically shown below (DLL1 depicted as blue bar; DLL4, red; NOTCH1, grey; HeLa-N1 cells are encircled in green). (B) Quantification of DLL1-Flag and DLL4-Flag in two independent CHO^attP-DLL1 (B5, C6) and CHO^attP-DLL4 (B5, D3) cell lines by Western blot analysis of cell lysates with anti-Flag and anti-β-actin (for normalisation) antibodies showed similar protein levels. (C) Surface biotinylation assays demonstrated equal surface representation of DLL1 and DLL4 on CHO^attP cells. (D) Notch trans-activation assays by co-culture of HeLa-N1 cells containing an RBP-Jκ:Luciferase reporter with CHO^attP-DLL1 or CHO^attP-DLL4 cells. All DLL1 and DLL4 clones activated Notch similarly, DLL4 being a slightly more efficient activator (compare with similar experiment in S6A and S6G Fig). (E) Notch trans-activation and cis-inhibition assays by culturing HeLa-N1 cells untransfected or transiently transfected with Dll1 or Dll4 expression constructs with or without CHO^attP or CHO^attP-DLL1 cells as indicated (a-c). Co-culture conditions a, b and c correspond to Luciferase measurements a’, b’ and c’, respectively. Results show cis-inhibition by DLL4 but not DLL1; for details see main text. (F) trans-Activation assays (a) without and (b) with NOTCH1 receptor expression in the signal sending CHO cell to test if NOTCH1 cis-inhibits the ligand activity of DLL1 or DLL4. No cis-inhibitory effect on either ligand was observed (columns a‘ and b‘ correspond to assay conditions a and b, respectively). (G) trans-Activation and cis-inhibition assays using chimeric DLL1-DLL4 proteins (G top; depicted as red and blue striped bars in a-c). HeLa-N1 cells were transiently transfected with no or DLL4-DLL1ECD or DLL1-DLL4ECD expression constructs and cultured as indicated (a-c). Under all three conditions, a strong cis-inhibitory activity was detected only for DLL1-DLL4ECD (columns a‘, b’ and c‘ correspond to schemas a, b and c, respectively). Error bars represent SEM; ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001.

Fig 6. Model of Notch signalling in the PSM triggered by DLL1 and ectopic DLL4.

Summary combining our in vivo and in vitro data in three different genetic scenarios (A-C); trans-activation (green arrows) and cis-inhibition (red bars) in cells of the PSM are schematically depicted on the left, representative skeletal preparations to visualise the outcome of somitogenesis are shown on the right; references to Figs. in this paper are given below. (A) In wildtype and Dll1 ^{Dll1ki/Dll1ki} PSMs, endogenous or transgenic DLL1 (D1) trans-activates Notch (N) signalling and results in a regularly segmented axial skeleton. (B) In our in vitro assays, DLL4 (D4) trans-activates Notch with similar efficiency as DLL1 but has an additional strong cis-inhibitory effect on Notch signalling that partially overrides trans-activation. The reduced net Notch activation in Dll1 ^{Dll4ki/Dll4ki} and CAG:Dll4;Dll1 ^loxP/loxP ;T(s):Cre PSMs is insufficient to support normal segmentation. (C) When both DLL1 and DLL4 are expressed (Dll1 ^Dll4ki/+ PSM), cis-inhibition by DLL4 plays a relatively smaller role, the resulting axial skeletons are mostly regular. However, cis-inhibition by DLL4 reduces the robustness of Notch signalling resulting in minor malformations (arrow indicates a misplaced rib), which are consistently seen in Dll1 ^Dll4ki/+ skeletons.