A novel reporter allele for monitoring Dll4 expression within the embryonic and adult mouse

ABSTRACT Canonical Notch signaling requires the presence of a membrane bound ligand and a corresponding transmembrane Notch receptor. Receptor engagement induces multiple proteolytic cleavage events culminating in the nuclear accumulation of the Notch intracellular domain and its binding to a transcriptional co-factor to mediate gene expression. Notch signaling networks are essential regulators of vascular patterning and angiogenesis, as well as myriad other biological processes. Delta-like 4 (Dll4) encodes the earliest Notch ligand detected in arterial cells, and is enriched in sprouting endothelial tip cells. Dll4 expression has often been inferred by proxy using a lacZ knockin reporter allele. This is problematic, as a single copy of Dll4 is haploinsufficient. Additionally, Notch activity regulates Dll4 transcription, making it unclear whether these reporter lines accurately reflect Dll4 expression. Accordingly, precisely defining Dll4 expression is essential for determining its role in development and disease. To address these limitations, we generated a novel BAC transgenic allele with a nuclear-localized β-galactosidase reporter (Dll4-BAC-nlacZ). Through a comparative analysis, we show the BAC line overcomes previous issues of haploinsufficiency, it recapitulates Dll4 expression in vivo, and allows superior visualization and imaging. As such, this novel Dll4 reporter is an important addition to the growing Notch toolkit.

Specification of endothelium into arteries and veins involves a cascade of signaling events that begin during embryogenesis (Coultas et al., 2005;Fish and Wythe, 2015;Gale and Yancopoulos, 1999). Current models propose that Sonic Hedgehog activation of the receptor Smoothened induces Vascular endothelial growth factor (Vegf ) transcription (Coultas et al., 2010;Lawson et al., 2002;Vokes et al., 2004). In turn, VEGF activation of VEGF-Receptor 2 (VEGFR2), which is required for arteriovenous specification in the early embryo (Shalaby et al., 1995), initiates expression of Delta-like 4 (Dll4) selectively within arterial endothelial cells. Dll4 encodes a transmembrane ligand for the Notch family of receptors . Notch1, as well as its essential transcriptional co-factor Rbpj-k [also known as CSL, Su(H), CBF] are essential regulators of arteriovenous (AV) patterning in the early vertebrate embryo, as their deletion leads to arteriovenous malformations (AVMs) and embryonic lethality (Krebs et al., 2004(Krebs et al., , 2000.
Dll4 is a critical regulator of vascular morphogenesis, as its loss results in vascular defects and embryonic lethality by embryonic day (E) 10.5 (Duarte et al., 2004;Gale et al., 2004;Krebs et al., 2004). In situ hybridization results show that Dll4 is the earliest Notch ligand detected in arterial precursor cells (aPCs), potentially preceding expression of Notch receptors (Chong et al., 2011;Lindskog et al., 2014;Mailhos et al., 2001;Wythe et al., 2013). Unlike Notch1, Dll4 expression in the dorsal aorta does not require hemodynamic force in the early mouse embryo, and is invariably arterial specific (Chong et al., 2011;Jahnsen et al., 2015). Conversely, Dll4 and Notch gain-of-function manipulations alter arteriovenous patterning and lead to lethality with obvious AV patterning defects in embryos Krebs et al., 2004;Trindade et al., 2008;Wythe et al., 2013), and AVMs in adults (Carlson et al., 2005;Murphy et al., 2014Murphy et al., , 2008. In addition to regulating AV specification, Dll4 function also controls angiogenesis. The dynamic expression of Dll4 within the tip cell, and its repression in the trailing stalk cells that make up a sprouting vessel is controlled by VEGF-VEGFR2 signaling (Gerhardt et al., 2003;Hellström et al., 2007;Lobov et al., 2007). Dll4-Notch signaling acts as a negative feedback regulator of VEGFR2 to establish the proper ratio of tip to stalk cells in the sprouting vasculature. Consequently, loss of Dll4, or Rbpj-k, leads to increased endothelial proliferation and hypersprouting (Jakobsson et al., 2010;Suchting et al., 2007).
Molecular and biochemical methods to query Dll4 expression, such as in situ hybridization, or antibody-based immunostaining, can be time consuming, and yield variable results. Mouse models with a lacZ reporter cassette replacing the translational start site of endogenous Dll4 have been used to visualize Dll4 expression; however, these modifications create a null allele (Duarte et al., 2004;Gale et al., 2004). In the case of Dll4 this is problematic, as these two lines, as well as a third, conventional loss of function allele (Krebs et al., 2004), demonstrated that heterozygous Dll4 mutants displayed incompletely penetrant, lethal haploinsufficiency between E9.5 and E10.5 (Duarte et al., 2004;Gale et al., 2004;Krebs et al., 2004). Outcrossing these lines to different genetic backgrounds reduces the penetrance of this effect, but the ratio of viable offspring remains low (Benedito and Duarte, 2005;Duarte et al., 2004). Furthermore, interpreting Dll4 expression levels in these knockin/knockout reporter mice is complicated due to a positive feedback loop between Dll4 expression and Notch signaling (Caolo et al., 2010). As such, even in viable mutant animals, it is not clear if the knockin reporter faithfully recapitulates Dll4 expression. Precisely defining Dll4 expression in the embryo and adult is central to understanding its role during vascular specification, angiogenesis (Hellström et al., 2007), T-cell development (Koch et al., 2008), and retinogenesis (Luo et al., 2012). Finally, Dll4 may signal to Notch receptors in even more tissues, such as the gut or kidney (Benedito and Duarte, 2005), necessitating an accurate, reliable, and robust method for visualizing its expression domain in vivo.
In the case of Dll4, histochemical detection of β-gal is considered more sensitive than detection of Dll4 mRNA by in situ hybridization (Benedito and Duarte, 2005). To retain this advantage, but overcome the inherit drawbacks of available Dll4 reporter-knockout mouse lines, we generated a transgenic Dll4-BAC-nlacZ reporter line. Herein, we show that this line faithfully recapitulates endogenous Dll4 expression in the embryonic, postnatal, and adult mouse, while avoiding potential confounds associated with disrupted Notch signaling. Furthermore, the signal strength in this model is greater than previous Dll4 reporter lines, and addition of a nuclear localization signal increases cellular resolution. Going forward, this novel tool will facilitate studies of Dll4 expression within the embryonic and adult mouse.

RESULTS
Using recombineering, a nuclear localized lacZ reporter cDNA cassette (nlacZ) was targeted to the start codon of murine Dll4 in a bacterial artificial chromosome (BAC) to generate a Dll4 reporter construct ( Fig. 1A) (Warming et al., 2005). The BAC clone, spanning approximately 81 kb of mouse chromosome 2, contained the entire Dll4 locus, as well as approximately 32.5 kb upstream and 38 kb downstream. The full-length, recombined clone, Dll4-BAC-nlacZ, was linearized and used to create transgenic mice by pronuclear injection. From one round of injections, two successful founders (Dll4-BAC-nlacZ 4336 and Dll4-BAC-nlacZ 4316 ) were identified with germ line transmission of the transgene. We focused our studies on founder Dll4-BAC-nlacZ 4336 , which exhibits more robust β-galactosidase (β-gal) activity (Fig. S1).
To validate our BAC transgenic line, we compared its β-gal activity to that of Dll4 lacZ/+ in the embryonic and postnatal mouse at different developmental time points (Fig. 1B-I). Prior work has suggested Dll4 transcripts are initiated at E8.0 (Benedito and Duarte, 2005); however, using the same allele employed in that study, as well as our novel BAC line, we detect β-gal at E7.75 in the presumptive endocardium of the cardiac crescent, as well as in aortic progenitor cells (aPCs) (Fig. 1B,F), in agreement with previous reports examining endogenous Dll4 transcripts (Benedito and Duarte, 2005;Duarte et al., 2004;Mailhos et al., 2001;Shutter et al., 2000;Wythe et al., 2013). By E8.25, analogous to Dll4 mRNA (Chong et al., 2011;Wythe et al., 2013), lacZ expression was present in the endocardium and sinus venosus, as well as the dorsal aorta (Fig. 1C,G). By E9.5, the dorsal aorta, endocardium, internal carotid artery, hindbrain, intersomitic arterial vessels, and perineural vascular plexus all displayed β-gal activity (Fig. 1D,H). At E10.5, both reporters labelled each of these structures, as well as the retina (Fig. 1E,I), although maintenance of the BAC line on a mixed FVB: C57BL/6 background somewhat obscured β-gal activity in the retina due to endogenous pigmentation. Both reporter lines displayed robust labelling of the vasculature within the embryonic yolk sac at E10.5 ( Fig. 1E′-E″,I′-I″). Magnified views of the yolk sac also showed increased cellular resolution of β-gal activity in the BAC line compared to the knockin reporter (Fig. 1E″,I″). Histological analyses of E9.5 and E10.5 embryos revealed that while β-gal was observed in the dorsal aorta and endocardium of both Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, it was absent from the cardinal vein ( Fig. 1J1-M2), confirming its arterial specificity within the endothelium. Notably, at E10.5, lacZ was expressed within a narrow, ventral stripe of tissue in the neural tube ( presumably V2 interneurons), in agreement with previous reports (Benedito and Duarte, 2005;Mailhos et al., 2001).
Dll4 lacZ/+ mutants can exhibit developmental delay (Duarte et al., 2004). This difference was apparent during later embryogenesis ( Fig. 2A-H), while Dll4-BAC-nlacZ mice were normal in size at all stages examined (Fig. S1). Significantly, the domain of lacZ expression at the level of the wholemount embryo and yolk sac was comparable between the two lines at E12.5 through E18.5 ( Fig. 2). At E14.5, superficial β-gal activity within the skin was evident in both lines (Fig. 2B,F,I,J), but it was not clear if Dll4 reporter activity was restricted to the arterial endothelium, or present within other vessel types, such as the venous vasculature or lymphatic system, as suggested by previous reports (Bernier-Latmani et al., 2015;Niessen et al., 2011). To determine the identity of β-gal-positive cells, skin from the E14.5 forelimb of both genotypes was processed for immunohistochemistry (IHC) using antibodies against β-gal, the endothelial-specific cell surface receptor CD31 (PECAM), the lymphatic vessel-specific antigen Podoplanin, the arterial-specific smooth muscle cell protein alpha smooth muscle actin (SMA), the neuronal-specific marker Tuj1, as well as endogenous Dll4. Confocal microscopy revealed that β-gal-positive cells of either genotype did not colocalize with Podoplanin or Tuj1, but did colocalize with CD31, SMA, and Dll4 ( Fig. 3A1-H6), demonstrating that lacZ expression was restricted to arteries in the embryonic skin.
At these same embryonic stages, the brains ( Fig. 4), hearts (Fig. 5),and lungs (Fig. 5) from the endogenous knockin and BAC transgenic reporter embryos were examined and compared. Within the embryonic brain, the expression domains of Dll4 lacZ/+ and Dll4-BAC-nlacZ were almost indistinguishable from E12.5 through E18.5 at the wholemount level, with signal evident within the vertebral arteries (VA), basilar artery (BA), superior cerebellar arteries (SCA), posterior cerebral arteries (PCA), middle cerebral arteries (Leslie et al., 2007), and anterior cerebral arteries (ACA), as well as their respective branches. Collaterals linking the MCA, ACA, and PCA territories became evident between E14.5 and E18.5, consistent with previous reports (Chalothorn and Faber, 2010). Histological analysis revealed reporter activity throughout the brains in both lines, from the olfactory bulb to the brain stem, and the cortex to the hypothalamus (Fig. 4).
At E12.5, Dll4 lacZ/+ activity was evident at the wholemount level within the great vessels (aorta and pulmonary artery) and the primary plexus that ultimately generates the coronary vasculature of the embryonic heart ( Fig. 5A1-A2). Histological analysis confirmed that β-gal was restricted to the endothelial lining of the great vessels, as well as the endocardium of the atrial and ventricular chambers, the endothelium underlying the epicardium, and vessels within the compact myocardium ( Fig. 5A3-A5). This same pattern of activity was observed in Dll4-BAC-nlacZ hearts ( Fig. 5B1-B5), although labeling of the primary coronary plexus was more robust in the BAC reporter line. At E14.5, in Dll4 lacZ/+ embryos, reporter signal was evident within the great vessels and the atria, as well as the coronary plexus ( Fig. 5C1-C5). Dll4-BAC-nlacZ activity was present in a similar domain, with signal present throughout the endothelium lining the great vessels, the chamber endocardium, and the coronary vessels underlying the compact myocardium ( Fig. 5D1-D5). At E18.5, Dll4 lacZ/+ expression persisted within the aorta, pulmonary artery, and coronary vessels ( Fig. 5E1-E5). Activity was also detected within the freewall myocardium ( Fig. 5E5). At this stage, Dll4-BAC-nlacZ signal was diminished within the endothelium of (H) Intra-littermate body measurements in a Dll4-BAC-nlacZ litter. Data are presented as averages ±s.e.m.; ns, nonsignificant. Comparisons were made by Student's t-test. Noticeable size differences can be observed between genotypes due to heterozygous Dll4 loss of function. (I-J) β-gal IHC on E14.5 skin from (I) Dll4 lacZ/+ or (J) Dll4-BAC-nlacZ embryos. I′ and J′ are magnified views of a respective region shown in corresponding panels I and J. Units depicted are in μm. the pulmonary artery, and was not detected within the aortic root ( Fig. 5F1-F4). β-gal was also detected in both the endocardium and myocardium of the atrial and ventricular chambers (Fig. 5F3,F5).
We next examined reporter activity within the embryonic lung ( Fig. 5G-L). Similar to the heart ( Fig. 5A-F), lungs were noticeably smaller in Dll4 lacZ/+ animals compared to the BAC reporters. In the pseudoglandular stage (E12.5), β-gal activity was detected within the primitive vascular tree of the left and right lobe in Dll4 lacZ/+ animals, with no discernable difference in expression from the rostral to caudal axis, or in any of the lobes (Fig. 5G1-G2). Histological analysis revealed signal within the endothelium of large and small caliber vessels ( Fig. 5G1-G4). This expression pattern was recapitulated in Dll4-BAC-nlacZ animals ( Fig. 5H1-H4). By the canalicular stage (E14.5), β-gal activity was present in narrow, horizontal bands across the ventral side of the trachea in both the knockin and BAC animals, as well as the vascular tree and endothelium, but excluded from smooth muscle ( Fig. 5I1-J4). Interestingly, expression was observedalbeit infrequentlywithin the airway epithelium in both the knockin and the BAC lines ( Fig. 5I4,J4). At the saccular stage (E18.5), β-gal activity within the trachea and vascular tree persisted in both samples, but was more evident in Dll4-BAC-nlacZ animals ( Fig. 5K1-L4). X-gal reactivity was infrequently detected in the airway epithelium of either line, but present within the developing distal alveoli of both reporters, presumably in the capillary endothelium (Fig. 5K4,L4). In both lines, signal was evident in the endothelium of small and medium size vessels, but absent in smooth muscle (Fig. 5K4,L4).
Across all embryonic tissues examined (brain, heart, lung, skin), signal strength and resolution were superior in Dll4-BAC-nlacZ animals compared to Dll4 lacZ/+ mice, particularly in sectioned tissue, where distinct cells could be observed in Dll4-BAC-nlacZ tissue, due to its nuclear localization. Additionally, X-gal staining proceeded more rapidly in Dll4-BAC-nlacZ tissue compared to agematched Dll4 lacZ tissue, in both wholemount and sectioned samples. Overt growth deficits were not observed in embryos derived from either BAC founder line, and viability of either BAC line was close to the expected Mendelian ratio ( Fig. 2H; Fig. S1), unlike Dll4 lacZ/+ animals, which displayed developmental delay (Fig. 2D) and reduced viability (Fig. S1H).
We next surveyed reporter activity in early postnatal and adult tissues. Like the cranial comparisons between Dll4-BAC-nlacZ and Dll4 lacZ/+ embryos, within the postnatal and adult brain β-gal labelled the major cerebral arteries, as well as their branches and collaterals in both lines (Fig. 6). Reporter activity spanned the anterior-posterior and dorsal-ventral axes in both lines. Staining of P1 and P5 brains showed increased vessel density and branching of pial arteries in Dll4-BAC-nlacZ samples (Fig. 6A,B). Staining in the adult brains appeared grossly similar between the two lines at the wholemount level (Fig. 6E,F). Histological analysis revealed activity throughout the olfactory bulb, cerebral cortex, hippocampus, and cerebellum, in an indistinguishable manner between the two reporters ( Fig. 6).
In the P1 postnatal Dll4 lacZ/+ wholemount heart, β-gal was active within the coronary vessels, aorta, and pulmonary artery ( Fig. 7A1-A2). Sections revealed signal within the chamber endocardium ( Fig. 7A3-A5), as well as the endothelial lining of the aorta (Fig. 7A4), the coronary vascular endothelium, and the myocardium (Fig. 7A5). Here, the activity and domain of the Dll4-BAC-nlacZ line differed dramatically from the knockin, in that while expression was also detected (sparsely) within the endothelial lining of the aortic root, it robustly labelled the chamber endocardium, myocardium, and the coronary vasculature ( Fig. 7B1-B5). At P5, Dll4 lacZ/+ drove β-gal within the endothelium of the aorta and pulmonary artery ( Fig. 7C1-C4), the chamber endocardium, myocardium, and coronary vasculature (Fig. 7C5). The BAC reporter marked these same expression domains, but demonstrated elevated β-gal activity within the myocardium compared to the knockin line ( Fig. 7D1-D5). In the adult heart, both lines showed weak signal within the endothelial lining of the aorta, as well as the chamber endocardium, myocardium, and coronary vasculature ( Fig. 7E1-F5). Expression was detected within the epicardium in knockin animals only at P5 (Fig. 7D5), and within BAC reporters only at P1 (Fig. 7B5).
β-gal was present within the trachea in the postnatal and adult lung, at all stages examined, in both lines. At P1 and P5, the endothelium of the small, medium, and large caliber vessels, but not the smooth muscle or airway epithelium, displayed lacZ expression in both Dll4 lacZ/+ and BAC animals. The alveoli were also β-gal positive, with expression in the capillary endothelium ( Fig. 7G1-7J4). This expression pattern perdured in adults, with the only notable difference between the two lines being the extent of activity within the alveoli ( Fig. 7K1-L4).
In the postnatal retina, some notable differences in expression were observed between the two lines. At P1 ( Fig. 8A1-B3), Dll4 lacZ/+ expression within the vasculature was absent ( Fig. 8A1-A3), but signal was present (though minimal) in the vessels of Dll4-BAC-nlacZ animals ( Fig. 8B1-B3). This difference was more pronounced at P5, where signal was virtually absent within the vasculature of Dll4 lacZ/+ animals ( Fig. 8C1-C3), but strong in Dll4-BAC-nlacZ animals ( Fig. 8D1-D3). By P7, lacZ expression was more comparable between the reporter lines, though still diminished in Dll4 lacZ/+ mice compared to the BAC reporter line ( Fig. 8E1-F3). By adulthood, no gross differences were observed in staining between the two alleles ( Fig. 8G1-H3), with labelling throughout the retinal vasculature. In both lines X-gal signal was detectable in the tissue underlying the surface vasculature ( presumably astrocytes) at all stages examined, although this was greatly diminished in the adult retina. Vascular signal in either genotype appeared arterial-specific, and was present within the capillary vasculature in the adult retina ( Fig. 8G1-H3). To determine if lacZ expression was restricted to arteries, adult retinas were immunostained for β-gal, the pan-endothelial marker isolectin B4, and the smooth muscle cell marker SMA (as smooth muscle cells are associated with arteries). In Dll4-BAC-nlacZ retinas, colocalization was observed between all three markers (β-gal, isolectin, and SMA) (Fig. 8I1-I6), suggesting that reporter expression was indeed restricted to the arterial and capillary endothelium. β-gal IHC on adult retinas from knockin reporter animals failed to yield interpretable results (Fig. S2), regardless of fixation method, primary antibody concentration, or length of antibody incubation. This is perhaps attributable to diminished Dll4 expression, which is consistent with tissues processed for X-gal staining, as knockin tissue required longer incubation times to achieve adequate signal compared to BAC reporter samples. Differences in clarity were prominent at all stages examined, with Dll4-BAC-nlacZ retinas displaying better cellular resolution.

DISCUSSION
In the present study, we generated a novel Dll4 reporter and compared its expression to a commonly used Dll4 knockout/lacZ knockin line (Duarte et al., 2004). This unique tool avoids the confounding variable of haploinsufficiency associated with previous Dll4 knockin reporter alleles. In addition, this new line allows for increased resolution of Dll4 expression due to the presence of a nuclear-localized reporter, and it generally recapitulates Dll4 expression in the embryo and adult. The results of our studies with the BAC reporter during embryogenesis mirror previously published data examining Dll4 expression by in situ hybridization (Chong et al., 2011;Mailhos et al., 2001;Shutter et al., 2000;Villa et al., 2001), and are generally concordant with studies of Dll4 knockin reporter alleles (Duarte et al., 2004;Gale et al., 2004;Wythe et al., 2013). Our novel BAC reporter, however, does exhibit important differences in expression compared to the knockin model used for our comparative analyses, and these points will be discussed below on a tissue by tissue basis.
In the developing and adult lung, β-gal was evident within the trachea and endothelium of large and small caliber vessels, as well as the capillaries surrounding the alveoli. Notch signaling in the trachea, as with many other tissues, is known to regulate cell fate decisions, with Notch gain-and loss-of-function manipulations resulting in the failure of proper tracheal branching in Drosophila (Llimargas, 1999;Steneberg et al., 1999). Furthermore, alveologenesis requires Notch signaling in the mouse lung epithelium (Tsao et al., 2016). Dll4 also has a physiological role in allergic inflammatory responses in the airway (Huang et al., 2017). It is possible that Dll4 presented on endothelial cells activates Notch receptors on adjacent cell types to regulate patterning of the trachea and lungs, a notion supported by reporter expression in adult pulmonary tissues in both the BAC and knockin lines.
At E9.5 and E10.5, the arterial cranial vasculature was labelled in both lines (Fig. 1). At E12.5, β-gal was evident in the arterial vasculature of brains from both lines, although MCA labeling was diminished in knockin animals. Previous work reported that loss of Dll4 delayed MCA formation and resulted in hyperbranching (Cristofaro et al., 2013), defects that were observed here as well (Fig. 4). However, knockin brains were often smaller at this stage, and general developmental delay may have caused this defect. The BAC line did not present this phenotype. Brains between the two genotypes were comparable in size at E14.5 and E18.5, but the caliber of vessels feeding into the circle of Willis, as well as the basilar artery, appeared larger in the BAC animals than in the knockins at E18.5. In contrast to previous reports suggesting that Dll4 is excluded from large caliber arteries at later stages (Benedito and Duarte, 2005;Gale et al., 2004), expression was evident in the major cranial arteries of the postnatal and adult brain in both reporters (Fig. 6). Collaterals were well labelled in both lines, but potentially more obvious in the BAC line. Dll4-Notch signaling is essential for embryonic vascular development (Duarte et al., 2004;Gale et al., 2004;Krebs et al., 2004Krebs et al., , 2000Swiatek et al., 1994), and the continued expression of Dll4 in the brain may suggest a role in regulating angiogenesis. Indeed, deletion of Rbpj-k at birth leads to arteriovenous malformations and increased vascular density in the brain at P14, followed shortly thereafter by lethality . By comparison, Rbpj-k loss in the adult mouse produced a mild phenotype in the brain . However, Dll4-Notch signaling has been suggested to modulate angiogenic responses in the brain after ischemic injury (Cristofaro et al., 2013). Given Notch's role in regulating neurogenesis from development through adulthood (  (A1-B5) β-gal activity in E12.5 hearts from (A) Dll4 lacZ/+ or (B) Dll4-BAC-nlacZ mice. A1-A2 and B1-B2 show representative wholemount hearts from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. A3 and B3 show β-gal activity in a representative cross-section through the heart, which is magnified accordingly in panels A4-A5 and B4-B5. β-gal activity is present within coronary plexus (A1,A2,B1,B2), the endocardium of the distal end (A1,B1) and root of the aorta (A4,B4) and pulmonary artery in both lines, as well as the endocardium and subepicardial vasculature (A5,B5), but absent from the epicardium and myocardium. (C1-D5) β-gal activity in E14.5 hearts from (C) Dll4 lacZ/+ or (D) Dll4-BAC-nlacZ mice. C1-C2 and D1-D2 show representative wholemount hearts from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. C3 and D3 show β-gal activity in a representative cross-section through the heart, which is magnified accordingly in panels C4-C5 and D4-D5. β-gal activity is localized to the endocardium of the aorta in both lines (C4,D4), as well as the chamber endocardium (C5,D5), and subepicardial coronary vasculature (C5,D5). β-gal activity was also detected within a small fraction of the myocardium in the BAC reporter line at this stage. (E1-F5) β-gal activity in E18.5 hearts from (E) Dll4 lacZ/+ or (F) Dll4-BAC-nlacZ mice. E1-E2 and F1-F2 show representative wholemount hearts from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. E3 and F3 show β-gal activity in a representative cross-section through the heart, which is magnified accordingly in panels E4-E5 and F4-F5. β-gal activity is localized to the endocardium of the aortic root in Dll4 lacZ/+ animals, but absent from Dll4-BAC-nlacZ mice (E4,F4), and present in both lines within the chamber endocardium and coronary vasculature (E5, F5), and sparsely in the myocardium. Ao, aorta; ec, endocardium; ep, epicardium; IVS, interventricular septum; LA, left atrium; LV, left ventricle; m, myocardium; PA, pulmonary artery; RA, right atrium; RV, right ventricle. (G1-H4) β-gal activity in E12.5 lungs from (G) Dll4 lacZ/+ or (H) Dll4-BAC-nlacZ mice. G1-G2 and H1-H2 show representative wholemount lungs from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. G3 and H3 are representative cross-sections through the lungs, and boxed in areas are magnified in G4 and H4, revealing activity within the endothelium. (I1-J4) β-gal activity in E14.5 lungs from (I) Dll4 lacZ/+ or (J) Dll4-BAC-nlacZ mice. I1-I2 and J1-J2 show representative wholemount lungs from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. I3 and J3 show β-gal activity in a representative cross-section through the lungs, which is magnified accordingly in panels I4 and J4, revealing endothelial-specific activity in both lines. (K1-L4) β-gal activity in E18.5 lungs from (K) Dll4 lacZ/+ or (L) Dll4-BAC-nlacZ mice. K1-K2 and L1-L2 show representative wholemount lungs from Dll4 lacZ/+ and Dll4-BAC-nlacZ mice, respectively, from ventral and dorsal views. K3 and L3 show β-gal activity in a representative cross-section through the lungs, which is magnified accordingly in panels K4 and L4, demonstrating endothelial-specific activity in both lines. D, dorsal; e, endothelium; L, left; R, right; sm, smooth muscle; V, ventral. Units depicted are in μm.
be interesting to discern what role(s), if any, endothelial Dll4-Notch signaling has in embryonic and adult neurogenesis.
In the periphery, nerves and arteries regularly align with one another (Mukouyama et al., 2002), as nerves induce arteriogenesis by secreting VEGF and CXLC12 (Li et al., 2013;Mukouyama et al., 2005). β-gal + cells in the forelimb skin of BAC embryos align with Tuj1 + nerves, are CD31 + , and are encapsulated by SMA + cells, demonstrating they are bona fide arterial endothelial cells (Fig. 3). Notably, endogenous Dll4 displayed a pattern identical to that of the reporters (Fig. 3). As such, it would be interesting to determine if endothelial Dll4 plays a role in nerve-vessel alignment.
Reporter activity within the postnatal and adult retina agree with published reports showing Dll4 transcript and protein expression in the endothelium at P3 (Crist et al., 2017;Hofmann and Luisa Iruela-Arispe, 2007), supporting the concept that Dll4 heterozygosity in the knockin delays reporter expression in the postnatal retina. Vascular lacZ expression was virtually absent until P7 in the knockin, but BAC reporter activity was detectable at low levels in the center of the retina, near the optic nerve, as early as P1. The ultimate impact of delayed Dll4 expression in the postnatal mutant eye (Suchting et al., 2007) may be inconsequential, as vascular patterning in the adult retina was grossly indistinguishable between the two reporter lines (Fig. 8G1-H3). Surprisingly, arterial-specific deletion of Dll4 or Rbpj-k at P10, after the major vascular network has been patterned, does not generate profound vascular remodeling defects in the retina by P28 (nor does deletion at P2 affect vascular structure at P15) (Ehling et al., 2013). However, pan-endothelial deletion of either gene induced vascular defects (Ehling et al., 2013), suggesting a role for Notch signaling in the capillary endothelium and venous tissue, potentially in agreement with reports of capillary and venous Dll4 expression in the retina (Crist et al., 2017;Ehling et al., 2013).
Notably, the BAC allele tended to reveal Dll4 expression at earlier time points compared to the knockin. This may result from increased lacZ expression in the Dll4-BAC-nlacZ 4336 reporter due to transgene copy number, or it may be attributable to normal levels of Notch signaling (unlike in the knockin mutants). In the postnatal retina delayed lacZ expression in the knockin, but not the BAC, suggests that Dll4 heterozygosity, even in a genetic background meant to mitigate the effects of its haploinsufficiency (e.g. CD-1 or FVB), still negatively influences reporter expression. Additionally, the knockin line required significantly longer (by several hours) incubation times for adequate visualization of reporter activity in all tissues examined. In the BAC 4336 line, 30 min was usually more show representative wholemount images of the brain from superior, inferior, and sagittal planes. A4-A9 and B4-B9 show representative coronal sections through the brain, from anterior to posterior. (C1-D9) β-gal activity in the P5 postnatal brain of (C) Dll4 lacZ/+ and (D) Dll4-BAC-nlacZ mice. C1-C3 and D1-D3 show representative wholemount images of the brain from superior, inferior, and sagittal planes. C4-C9 and D4-D9 show representative coronal sections through the brain, from anterior to posterior. (E1-F9) β-gal activity in the adult brain of (E) Dll4 lacZ/+ and (F) Dll4-BAC-nlacZ mice. E1-E3 and F1-F3 show representative wholemount images of the brain from superior, inferior, and sagittal planes. E4-E9 and F4-F9 show representative coronal sections through the brain, from anterior to posterior. ACA, anterior cerebral artery; azACA, azygos of the anterior cerebral artery; AIC, anterior inferior cerebellar artery; BA, basilar artery; ICA, internal carotid artery; MCA, middle cerebral artery; PCA, posterior cerebral artery; SCA, superior cerebellar artery; VA, vertebral artery. Units depicted are in μm. than adequate for robust visualization of β-gal staining and resolution of single endothelial cells. Overall, the BAC reporter generates a more representative pattern of Dll4 expression than the heterozygous knockin reporter.
Multiple studies have suggested that Foxc1/2, as well as β-catenin, mediate the transcriptional induction of Dll4 within the embryonic endothelium (Corada et al., 2010;Hayashi and Kume, 2008;Seo et al., 2006). However, previous work demonstrated that endothelial-specific deletion of β-catenin failed to affect establishment of arteriovenous identity or alter Dll4 expression in the early mouse embryo (Wythe et al., 2013). Additionally, the genomic region [5′ to the transcriptional start site (TSS) of murine Dll4] shown to bind these same transcription factors failed to drive reporter activity in vivo (Wythe et al., 2013), suggesting alternative transcriptional regulators of Dll4. Indeed, we, and others, identified enhancers within the third intron of Dll4, as well as several kilobases upstream of the TSS (-10, and -12, respectively) (Luo et al., 2012;Sacilotto et al., 2013;Wythe et al., 2013). The targeting vector that generated the Dll4 lacZ/+ allele utilized in this study retained the intron 3 enhancer (Duarte et al., 2004). However, another Dll4 knockin/knockout reporter mouse with arterial lacZ expression (Gale et al., 2004) replaced the entire Dll4 locus, suggesting that this region is dispensable for endothelial expression of Dll4 (we are not aware of any study directly comparing β-gal activity between these two mutant lines). Nonetheless, the 81-kb region spanned by the BAC reporter contains each of these in vivo validated genomic elements sufficient to drive Dll4 arterial expression. Future deletion of these, and other conserved regulatory elements, will determine the necessity of these putative enhancers.
In the absence of genetic reporter models, studies must rely on molecular and biochemical methods to assess gene expression in vivo. In situ hybridization using nucleic acid probes is a common method for analyzing gene expression patterns. This technique, however, relies on the quality and fidelity of the probes used to detect mRNA transcripts of interest and, as such, can exhibit high variability from one probe to the next. Furthermore, the utility of this method is limited by riboprobe penetration and cellular resolution in wholemount tissue. Immunostaining presents similar obstacles regarding variability, as often multiple commercial antibodies exist for the same antigen, and several are derived from finite sources (e.g. polyclonal). Our mouse model provides a simple, robust, renewable, and reliable alternative approach to visualize Dll4 expression in vivo. Furthermore, this Dll4 mouse line can be combined with other alleles and genetic backgrounds to elucidate epistatic interactions, without the requirement of having to assess its expression on a confounding heterozygous background, as is the case with current Dll4 reporters. Furthermore, as demonstrated herein, this allele is well-suited for IHC studies due to the restricted nuclear localization of the antigen. Going forward, this new mouse line can be used to determine how Dll4 expression changes in response to gain -or loss-of-function gene manipulations, and whether levels of Dll4 are changed in response to injury, disease, or drug treatment. Collectively, this novel Dll4-BAC-nlacZ reporter mouse line will prove a valuable tool in deciphering the mechanisms underlying Notch signaling, and will provide researchers with a useful reagent for investigating Dll4-associated mechanisms.

Mouse experiments
All mouse protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at Baylor College of Medicine and University of California, San Francisco (UCSF). For all experiments, noon on the day a plug was discovered was considered as E0.5.
loxP GalK ins REV: CCG ATG CAA GTG TGT CGC TGT CGA CGG TGA CCC TAT AGT CGA GGG ACC TAT CAG CAC TGT CCT GCT CCT T Each GalK cassette was sequentially inserted and then replaced by a single, 100 bp oligo that lacks the original loxP511 or loxP sequences, but contains the original plasmid backbone sequence, effectively deleting the loxP511 and loxP sequences and leaving no scar. PCR genotyping for successful replacement of loxP sites was performed with the following primer pairs: loxP511-FWD : GGC AGT TAT TGG TGC CCT TA  loxP511-REV: TTC AAC CCA GTC AGC TCC TT  expected size=353 bp  loxP-FWD: TAG TGA CTG GCG ATC CTG TC  loxP-REV: AAC ATT TTG CGC ACG GTT AT  expected size=396 bp At this point, the resulting plasmid was referred to as ΔloxP-Dll4-BAC. Next, a 5′ homology arm to murine Dll4 was amplified by PCR with the following primers ( pGalK homology in lowercase, unique restriction sites are in italics, Dll4 homology underlined in capitals): FWD: accgggccccccctcgagGTCGACACTGTAGCCACTAGAGGCCTG REV(EcoRV): tgtcaacaggaattcGATATCCATCCCTTGGGGTGTCCTC-TCCAC The resulting fragment was cloned via cold fusion (SBI) into the digested and purified pGalk vector 5′ to the EM7-GalK cassette. After identification of a positive clone and confirmation by DNA sequencing, a 3′ Dll4 homology arm was amplified and then inserted 3′ to the GalK cassette into the SpeI and NotI sites using the following primers: FWD: gacagtgctgaggatccACTAGTACGCCTGCGTCCCGGAGCGCC REV (NotI): tccaccgcggtgGCGGCCGCACCGGCGTGGAGACATT-GCCAAAGG The Dll4 5′ 3′ arm GalK vector was digested, the homology arm Galk fragment purified, and then electroporated into SW102 ΔloxP-Dll4-BAC bacteria for positive selection on M63+galactose plates to isolate a ΔloxP-Dll4-BAC-GalK clone. Concurrently, a codon-optimized nls-lacZ (from Invivogen's pWhere plasmid) was subcloned between the same 5′ and 3′ Dll4 homology arms, into the EcoRI site (5′) and BamHI site (3′). After confirmation by sequencing, Dll4 5′ 3′ arm nls-lacZ-pA, was digested with SalI and NotI to release the targeting fragment, and after purification this element was transformed into electro-competent ΔloxP-Dll4-BAC SW102-GalK bacteria and subjected to negative selection on M63 plates+DOG. The colonies were screened by PCR and the resulting construct, ΔloxP-Dll4-BAC-nlacZ, was confirmed by DNA sequencing.

Generation of transgenic mice
The ΔloxP-Dll4-nlacZ-BAC DNA was purified using the BAC 100 prep kit (Nucleobond) and digested with PI-SceI to linearize the BAC for more efficient transgenesis. A portion was inspected by pulse field electrophoresis to confirm the correct restriction pattern, then the remainder was dialyzed (Spectra/Por Micro DispoDialyzer; 8000 Da molecular weight cutoff, 100 µl) to embryo water (Sigma-Aldrich, W1503), and used for pronuclear injection. Injection of transgenic fragments was performed at the Gladstone Institute. Dll4-BAC-nLacZ #4316 is a weaker, but consistent founder line and Dll4-BAC-nLacZ #4336 is a strong expresser. Dll4-BAC-nLacZ transgenic animals were generated in FVB donor eggs, but subsequently maintained on a mixed FVB:C57BL/6 background. At the time of re-derivation, the Dll4 LacZ/+ allele was re-derived on an ICR background, then re-derived later on an FVB background, both in an effort to minimize haploinsufficient lethality observed on a C57BL/6 background.

lacZ staining
Embryos were harvested at timed intervals. A vaginal plug in the morning indicated E0.5. Appropriately timed-mated, pregnant dams were euthanized by CO 2 , and embryos were carefully dissected away from all internal membranes into cold 1× phosphate buffered saline (PBS). For wholemount processing, embryos were processed as previously described (Wythe et al., 2013). Briefly, following dissection, embryos or their dissected organs (brain, heart, or lungs) were fixed with a formaldehyde/glutaraldehyde solution in 1× PBS (2% formaldehyde, 0.2% glutaraldehyde, 0.02% sodium deoxycholate, 0.01% NP-40). For embryos and organs ≤E8.5, fixation time was 5 min. For embryos and organs ≤E10.5, fixation time was 10 min. For all older embryos and organs, fixation time was 15-20 min. Following fixation, tissues were rinsed briefly in 1× PBS and embryos were placed in permeabilization solution (1× PBS, 0.02% sodium deoxycholate, 0.01% NP-40). Embryos and tissue ≥E10.5 were incubated in permealibilization buffer overnight to allow for sufficient penetration of staining. After permeabilization, embryos were incubated at 37°C in freshlymade, 0.22 µm-filtered X-gal staining solution made in permeabilization buffer (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl 2 , 1 mg/ml X-gal). Whole embryos and organs were incubated in this solution between 3-4 h for Dll4-BAC-nlacZ animals, though often 6-8 h for Dll4 lacZ/+ animals to acquire similar levels in staining. Embryos and organs were then rinsed briefly twice in permeabilization buffer to remove residual staining solution, followed by a longer 20-min wash. This was followed by post-fixation overnight in 4% paraformaldehyde (PFA) at 4°C. The following day, PFA was removed and embryos and organs were washed twice for 10 min each wash in PBST (1× PBS, 0.1% Tween-20). Tissues were then subjected to a serial dehydration with methanol (25% MeOH/PBST, 50% MeOH/PBST, 75% MeOH/PBST, and finally three washes of 100% MeOH for 10 min each wash). Lastly, embryos and organs were washed with 5% H 2 O 2 /95% MeOH for 1 h at room temperature. Larger embryos (≥E12.5) were further washed with 7.5% H 2 O 2 /MeOH for 15 min if any yellowing of the tissue was still present. Tissues were then serially rehydrated in PBST, then stored in 4% PFA until imaging or further processing.
For embryonic and early postnatal brain sections, brains were dissected and rinsed in 1× PBS, followed by overnight fixation in 2% PFA at 4°C. For adult brains, mice were transcardially perfused with 1× PBS followed by 4% PFA before brains were removed. Brains were then transferred into serial sucrose/PBS solutions (10%, 20%, and 30%) and then frozen in optimal cutting temperature (OCT) compound and stored at −80°C. Adult and early postnatal brains were cryosectioned at 40 µm and placed in 2% PFA for 15 min (free-floating). Sections were then rinsed briefly in 1× PBS before being washed twice in permeabilization buffer for 10 min each wash. The tissue was then stained at 37°C in X-gal staining solution for 3 h for Dll4-BAC-nlacZ animals, and 4-5 h for Dll4 lacZ/+ animals. After staining, sections were briefly rinsed several times to eliminate staining solution and mounted using Fluoromount-G mounting media (SouthernBiotech, Birmingham, USA, 0100-01). For embryonic brains, tissue was processed similarly, with the exception that sections were placed directly onto slides after sectioning, rather than using a free-floating method. For embryonic and early postnatal heart and lung sections, tissues were harvested and fixed for 2 h in 2% PFA at 4°C. Tissues were subsequently transferred into serial sucrose/PBS solutions (10%, 20%, and 30%) and then frozen in OCT compound and stored at −80°C. Cryosections were taken at 10 µm and mounted directly to glass slides for processing. X-gal staining was performed as previously stated, but were further processed afterwards for Eosin (Dll4-BAC-nlacZ) or Nuclear Fast Red (Dll4 lacZ/+ ) staining. Slides were submerged in Eosin solution (Thermo Fisher Scientific, 7111) for 3 min before being washed 2× for 3 min each wash in tap water. Slides were then dipped 3× for 30 s in 100% EtOH, followed by 3× for 1 min in xylene. For Nuclear Fast Red staining (Vector Laboratories, Burlingame, USA, H-3403), slides were submerged for 15 min, followed by identical washes in tap water, EtOH, and xylene. Slides were then mounted using Entellan New (Millipore, 107961). For adult heart and lungs, mice were first transcardially perfused with 1× PBS and 2% PFA before identical post-fixation as stated above. Adult lungs were also first infused with 1% low melting point agarose and allowed to solidify prior to post-fixation and processing.

Immunohistochemistry
IHC performed on embryonic limb skin was performed according to Mukouyama et al. (2012). Briefly, forelimbs from E14.5 embryos were removed in ice-cold 1× PBS and subsequently transferred to 4% PFA at 4°C overnight. On the next day, tissue was dehydrated in 100% MeOH and stored at −20°C. Forearm limb skin was gently removed from the underlying tissues and placed in 100% MeOH. Samples were then rehydrated by transferring them into 75%/50%/25% MeOH/1× PBST (1× PBS with 0.2% Triton X-100) for 5 min each step. Samples were then washed twice for 5 min in 1× PBST before putting the tissue in filter-sterilized blocking solution (10% horse serum, 0.5% Triton-X, 1× PBS) for 2 h at room temperature. Blocking solution was then removed and primary antibodies prepared in the same blocking solution were added and left overnight to shake gently at 4°C. Skin samples were then washed in blocking buffer five times for 10 min each wash before adding secondary antibodies diluted in blocking buffer, and allowed to incubate for 1 h at room temperature. Tissues were then washed again in blocking buffer five times for 10 min each wash and mounted on glass slides using Fluoromount-G mounting media (SouthernBiotech, 0100-01) and imaged.

Imaging
Skin and retinas processed for IHC were imaged using a Leica TCS SPE confocal microscope with a 10× objective. All other tissues and embryos were imaged using a Zeiss Axio Zoom.V16 microscope and processed using ZenPro software (Zeiss) and Adobe Photoshop. All images were assembled using Adobe Illustrator.

Body measurements
For intra-and inter-litter embryo body measurements, whole view images of embryos were obtained using a Zeiss Axio Zoom.V16 microscope. Using the length function on the Zeiss software, a line was drawn from the apex of the head to the bottom of the rump (Mu et al., 2008). Measurements were recorded and averaged for each group and comparisons were made by Student's t-test.