A modifier in the 129S2/SvPasCrl genome is responsible for the viability of Notch1[12f/12f] mice

Background Mouse NOTCH1 carries a highly conserved O-fucose glycan at Thr466 in epidermal growth factor-like repeat 12 (EGF12) of the extracellular domain. O-Fucose at this site has been shown by X-ray crystallography to be recognized by both DLL4 and JAG1 Notch ligands. We previously showed that a Notch1 Thr466Ala mutant exhibits very little ligand-induced NOTCH1 signaling in a reporter assay, whereas a Thr466Ser mutation enables the transfer of O-fucose and reverts the NOTCH1 signaling defect. We subsequently generated a mutant mouse with the Thr466Ala mutation termed Notch1[12f](Notch1tm2Pst). Surprisingly, homozygous Notch1[12f/12f] mutants on a mixed background were viable and fertile. Results We now report that after backcrossing to C57BL/6 J mice for 11–15 generations, few homozygous Notch1[12f/12f] embryos were born. Timed mating showed that embryonic lethality occurred by embryonic day (E) ~E11.5, somewhat delayed compared to mice lacking Notch1 or Pofut1 (the O-fucosyltransferase that adds O-fucose to Notch receptors), which die at ~E9.5. The phenotype of C57BL/6 J Notch1[12f/12f] embryos was milder than mutants affected by loss of a canonical Notch pathway member, but disorganized vasculogenesis in the yolk sac, delayed somitogenesis and development were characteristic. In situ hybridization of Notch target genes Uncx4.1 and Dll3 or western blot analysis of NOTCH1 cleavage did not reveal significant differences at E9.5. However, qRT-PCR of head cDNA showed increased expression of Dll3, Uncx4.1 and Notch1 in E9.5 Notch1[12f/12f] embryos. Sequencing of cDNA from Notch1[12f/12f] embryo heads and Southern analysis showed that the Notch1[12f] locus was intact following backcrossing. We therefore looked for evidence of modifying gene(s) by crossing C57BL/6 J Notch1 [12f/+] mice to 129S2/SvPasCrl mice. Intercrosses of the F1 progeny gave viable F2 Notch1[12f/12f] mice. Conclusion We conclude that the 129S2/SvPasCrl genome contains a dominant modifying gene that rescues the functions of NOTCH1[12f] in signaling. Identification of the modifying gene has the potential to illuminate novel factor(s) that promote Notch signaling when an O-fucose glycan is absent from EGF12 of NOTCH1.


Background
Canonical Notch signaling is induced by cell-cell interactions between a Notch receptor and a Notch ligand [1]. There are four mammalian Notch receptors (NOTCH1 to NOTCH4) and five Notch ligands (DLL1, DLL3, DLL4, JAG1 and JAG2). The extracellular domain (ECD) of Notch receptors comprises 29-36 EGF repeats that may be modified by O-glycans [2][3][4]. O-fucose glycans are initiated by the transfer of fucose from GDP-fucose by POFUT1 at Ser/Thr in the consensus sequence C 2 XXXXS/TC 3 between the second and third cysteines of an EGF repeat [5]. There are 21 O-fucose glycan consensus sequences in the 36 EGF repeats of NOTCH1 and a similar number in the other Notch ECDs. Experiments in Drosophila determined that deletion of EGF11 and EGF12 destroys Notch ligand binding and defined a Notch ligand binding domain [6]. EGF12 of both Drosophila and mammalian Notch carry O-fucose [7,8] and Xray crystallography of co-crystals between fragments of NOTCH1 and DLL4 [9] as well as NOTCH1 and JAG1 [10] revealed that O-fucose on EGF12 plays a key role in binding interactions between NOTCH1 and both ligands.
Conversion of Thr to Ala in the EGF12 O-fucose consensus site of Drosophila Notch precludes the addition of O-fucose to EGF12, and results in embryos that go through neurogenesis but fail to exhibit Notch signaling at the dorsal/ventral boundary of the wing disc [11]. In vitro binding assays revealed that there is a marked loss in the ability of Fringe (which adds a N-acetylglucosamine to Ofucose in EGF repeats) to inhibit the binding of Serrate to Notch lacking O-fucose in EGF12 [11]. However, Fringe markedly enhances the binding of both Delta and Serrate to Drosophila Notch when O-fucose is absent from EGF12 [11]. Thus, when all other O-fucose sites remain intact and can be modified by Ofut1 and Fringe, the loss of O-fucose in EGF12 primarily has an effect on Serrate/ Notch interactions. Interestingly, a recombinant mammalian NOTCH1 fragment containing EGF12 lacking Ofucose does bind to Notch ligands in vitro in the presence of calcium [12].
Notch signaling in mice is very similar to Notch signaling in humans. Since disrupted Notch signaling causes numerous human pathologies [13], defining mechanisms of Notch signaling in mice provides insights into human disease. Introduction of the Thr to Ala mutation (Thr466Ala) in EGF12 of a mouse Notch1 cDNA causes the resulting mutant NOTCH1[12f] to be inactive in a Notch signaling reporter assay [14]. Reversion of the mutation in the mutant cDNA rescues Notch signaling. Introduction of a Ser in place of Ala in the Notch1[12f] cDNA allows the acquisition of O-fucose and the concomitant rescue of Notch signaling. Thus, in cultured mammalian cells, the presence of O-fucose on EGF12 is necessary for Notch signaling. Therefore, it was surprising to find that Notch1[12f/ 12f] mouse homozygotes are viable and fertile [15,16]. The Notch1[12f] allele is hypomorphic, however, since it only partially rescues a compound heterozygous Notch1[12f/lbd] mouse [15]. Notch1[lbd] (Notch1 tm2.1Pst ) is a strong loss-offunction allele lacking 141 amino acids that include the NOTCH1 ligand binding domain [17,18]. In this paper we show that Notch1[12f] becomes a strong loss-of-function allele after backcrossing at least 11 generations onto the C57BL/6 J background, and we reveal the existence of a modifier gene in either the 129129S2/SvPasCrl or C57BL/ 6J genome.

Results
C57BL/6 J Notch1[12f/12f] embryos die before birth The Notch1[12f] allele (Notch1 tm2Pst ) was originally characterized in mice of mixed genetic background. WW6 embryonic stem (ES) cells were engineered to carry a Notch1 point mutation in exon 8 in a floxed targeting construct. WW6 ES cells are~75% 129S2/SvPasCrl,2 0% C57BL/6 J and~5% SJL [19]. Chimeric mice were crossed to C57BL/6 J MeuCre40 mice [20] to obtain the Notch1[12f] allele, and then crossed 5-7 generations to C57BL/6 J mice before heterozygotes were intercrossed and their progeny characterized [15]. This colony was maintained over several years by heterozygote crosses and progeny were obtained at the expected Mendelian frequency. Backcrossing of heterozygotes to C57BL/6 J males and females was subsequently performed in a separate cohort until Notch1[12f/+] heterozygotes of the eleventh backcross generation were intercrossed. When several pairs did not give rise to Notch1[12f/12f] homozygous progeny at the expected frequency ( We also reacquired Notch1[12f/12f] homozygotes from a colony established in Toronto from Notch1[12f/12f] males we sent. The latter were backcrossed 3 times to C57BL/6 J mice and maintained by intercrossing before sending us male homozygotes. These mice were used to independently replicate our published data [15] [21][22][23][24][25]. In C57BL/6 J Notch1[12f/12f] embryos of E9.5 to E11.5, the yolk sac was poorly vascularized with major blood vessels ill-defined compared to control littermates which had a clear vascular tree with prominent blood vessels at each stage (Fig. 2). The overall morphology of wild type and heterozygous embryos was similar to Notch1[12f/12f] mutants at E9.5 (Fig. 3). However, based on comparisons of their Theiler stage (TS) [26], E9.5 Notch1[12f/12f] homozygotes were delayed in development compared to littermates. They had a significant reduction in the total number of somites at E9.5 (Table 3), mutant embryos were relatively anemic and close to TS14-late or TS15-early, whereas littermates were more similar to TS15-late ( Fig. 3 and Additional file 1: Figure S1). By E10.5, mutant embryos were similarly less developed and more anemic than controls ( Fig. 3 and Additional file 2: Figure S2), and by E11.5 dying mutants exhibited a strong Notch signaling-defective phenotype with an enlarged pericardial sac and small heart, severely truncated somites and general regression of most tissues (Fig. 3). In general, overall body size of Notch1[12f/12f] homozygotes was smaller, features were less defined, the heart and mandibular arches were generally smaller and limb buds less developed. Staining with anti-PECAM1 of E10.5 embryos highlights these features (Additional file 3: Figure S3). The staining intensity was somewhat reduced in Notch1[12f/12f] embryos and vascularization appeared slightly more sparse.
To determine whether a mutation of the Notch1[12f] gene might have occurred during backcrossing that could explain the embryonic lethality of Notch1[12f/12f] embryos, cDNA from wild type and E9.5 homozygous mutant embryo heads was prepared. Sequencing of cDNAs obtained from PCR products using 12 overlapping primers that spanned the entire coding region (Additional file 4: To determine if Notch signaling was reduced in the presomitic mesoderm (PSM), in situ hybridization of Uncx4.1 and Dll3 was performed. Previous studies of Notch pathway mutants have shown reduced expression of these genes at E9.5 [21,27,28]. However, we observed similar signals for control and mutant embryos at E9.5 (Fig. 4). Western blot analysis to detect cleaved NOTCH1 intracellular domain (NICD1) as an indicator of Notch1 signaling was performed on the PSM of E10.5  Figure S4). A subset of Notch pathway genes whose expression was previously shown by in situ hybridization to increase in defined brain regions of Notch-signaling defective Pofut1 null embryos [21], was investigated by qRT-PCR of the cDNA from Notch1[12f/12f] embryo heads. At E9.5, Notch1, Jag1 and Dll3 expression were significantly increased in Notch1[12f/12f] compared to wild type embryo transcripts in replicate experiments performed in triplicate using two housekeeping genes (Fig. 5), and in a Notch pathway qRT-PCR array using five housekeeping genes (not shown). In E10.5 embryos, Uncx4.1 was significantly increased based on qRT-PCR analysis with Actb and Gapdh housekeeping genes (Fig. 5).    [29]. It is expected that these genes would derive from the predominant 129S2/SvPasCrl genome of WW6 cells [29].

Discussion
In  [3]. The chance of loci outside the congenic footprint of the Notch1[12f] gene being heterozygous after 15 generations of backcrossing is (1/2) 15-1 or 0.000061 [30]. It is expected that the congenic footprint derives from the 129S2/SvPasCrl genome since it is predominant in WW6 ES cells. The fact that a single intercross between C57BL/6 J/129S2/SvPasCrl F1 heterozygotes produced viable Notch1[12f/12f] pups in the expected Mendelian ratio suggests that a 129S2/SvPasCrl gene (or tightly linked genes) outside the congenic footprint is responsible for the viability of F2 Notch1[12f/12f] homozygotes. The fact that the loss of this modifying gene during backcrossing was not evident after~7 backcrosses, when less than 1% of the genome would be 129S2/ SvPasCrl, is surprising. However, both the viability of Notch1[12f/12f] mice from intercrosses after~7 backcrosses, and the loss of viability after backrossing reached 11 or more generations, were observed in two cohorts maintained separately. This apparent paradox indicates that the "viability" gene may be linked to the Notch1[12f] allele on chromosome 2, though outside the congenic footprint.
The potential function(s) encoded by the 129S2/ SvPasCrl modifying gene are myriad because the Notch signaling pathway is so complex [21][22][23][24][25]. It requires correct synthesis of a large membrane receptor, modification, processing and trafficking through the secretory pathway to the cell surface, stable expression at the cell surface, interaction with canonical ligands and endocytosis into the ligand-expressing cell. While there are only 5 canonical NOTCH1 ligands, there are numerous membrane and secreted proteins that bind to NOTCH1 ECD and may affect canonical Notch signaling [31] [18]. Evidence from electron microscopy suggests that Drosophila Notch and mammalian NOTCH1, respectively, dimerize through their extracellular domains [32]. A modifier could encode a molecule that assists transit of NOTCH1[12f] from the endoplasmic reticulum through Golgi compartments, or provides stability at the plasma membrane, or modifies NOTCH1[12f] through a posttranslational mechanism to interact with Notch ligands. Previously identified modifiers of Notch signaling include Fringe which modulates the strength of Notch signaling via canonical Notch ligands [33], DNER, a non-canonical Notch ligand [34] and molecules of a variety of activities identified by genetic interactions [35][36][37] or modifier screens [38]. An example of mouse background influencing a Notch signaling phenotype is Jag1 mutant heterozygotes that exhibit a more severe inner ear phenotype on a C3H background compared to a C57BL/6 background, and eye dysmorphologies that manifest on a C57BL/6 background but are suppressed on a C3H background [39]. Identifying the mechanism by which NOTCH1[12f] becomes functional will require isolating the modifying gene and showing how the activity it encodes enables NOTCH1[12f] to participate in Notch signaling.
The most straight-forward way to identify a modifier would be a SNP genotyping strategy on a large cohort of 129S2/SvPasCrl/C57BL/6 J Notch1[12f/12f] F2 progeny to locate a region of 129S2/SvPasCrl homology present in all progeny. This strategy was successful in genomic mapping of several recessive mutant genes with phenotypic traits using a small panel of 394 SNPs and a relatively modest number of F2 progeny [40]. However, the regions mapped in that study varied from 17 to 83 Mb. By contrast, the Affymetrix Mouse Diversity Genotyping Array uses 63,000 SNPs with a potential of mapping to a region as small as 4.6 kb. A complementary strategy would be to transfect a range of cell types with the mouse Notch1[12f] cDNA shown previously to be inactive in ligand-induced Notch signaling in CHO cells [14], identify a host cell that allows NOTCH1[12f] to function, and isolate the modifying gene from a cDNA expression library made from that cell. This would be Fig. 5 qRT-PCR of Notch pathway genes expressed by Notch1[12f/ 12f] embryos. RNA from E9.5 and E10.5 embryo and control heads (n = 3 of each) was converted to cDNA and tested in replicate qRT-PCR experiments using Actb and Gapdh housekeeping genes. Values for mutants and controls compared to each housekeeping gene were obtained and fold-difference between mutants and controls (taken as 1.0) was determined. The data were plotted using Prism version 8.0. Significance was determined by unpaired Student's t test, (*) p = 0.07; * p < 0.05; ** p < 0.005 facilitated if the gain-of-function phenotype was easy to screen for or for which to select. While each of these approaches would require considerable effort, it would potentially unearth a novel mechanism by which NOTCH1 with an inactivating point mutation that precludes the addition of O-fucose in the canonical Notch ligand binding domain can become functional. Such a mechanism may be in operation in cell types that do not completely lose Notch signaling when Pofut1 is deleted. Examples include the inner ear [41], hemopoietic stem cells [42], and endocardial cells [43].

Conclusion
We conclude that the viability of Notch1[12f/12f] homozygotes on the mixed genetic background generated by 5-7 backcrosses to C57BL/6 J mice after derivation of the targeted allele in WW6 ES cells, is due to the presence of a gene in the 129S2/SvPasCrl background. This modifier gene allows the production of viable Timed mating was performed to obtain embryos at different stages of gestation. A time of 0.5 days of gestation was taken as the first morning that a vaginal plug was observed in a female who had been placed with a male the evening before. Mice were sacrificed using asphyxiation to effect in a chamber containing CO 2 delivered from a compressed gas cylinder, followed by cervical dislocation. Mice were bred in the animal facilities at the Albert Einstein College of Medicine and at the University Health Network (UHN) in Toronto, and sacrificed using CO 2  After backcrossing 5-7 times to C57BL/6 J, the colony was maintained for several years by intercrossing heterozygotes and Notch1[12f/12f] homozygote progeny were viable and fertile. Following backcrossing to 11 or more generations, however, Notch1[12f/12f] homozygotes generally died during gestation. We analyzed a large number of embryos from this 1st cohort. We then reacquired viable Notch1[12f/12f] mice that we had previously sent to Toronto and which had been backcrossed to C57BL/6 for 3 generations and then intercrossed to derive [12f/ 12f] homozygotes. These mice were crossed once to C57BL/6 J mice at Albert Einstein and then heterozygotes were intercrossed to maintain a colony that generated viable Notch1[12f/12f] mice. Other Notch1[12f/12f] mice from Toronto were backcrossed to C57BL/6 J mice to 15 generations before heterozygotes were intercrossed. This colony was maintained as Notch1[12f/+] heterozygotes because Notch1[12f/12f] homozygotes were not viable. We analyzed numerous embryos from this 2nd cohort. 129S2/SvPasCrl mice were used to cross to the latter Notch1[12f/+] heterozygotes.
Genotyping of progeny was performed by PCR of genomic DNA from toe or yolk sac using forward primer Fwd-CGGAGTGACGGGTATGTAT and reverse primer Rev-GTGTGTGTTGTGGCAGGTTC which span a lox p site between exons 5 and 6 of the mouse Notch1 locus [15]. The Notch1[+] allele product was 458 bp and the Notch1[12f] allele product was 561 bp. Timed mating was performed to determine stage of embryonic death and to generate embryos for analysis. A time of 0.5 days of gestation was taken as the first morning that a vaginal plug was observed in a female who had been placed with a male the evening before. Embryos were staged according to established Theiler staging criteria [26].

Southern blot analysis
Genomic DNA was prepared, digested with HindIII or KpnI, electrophoresed on an agarose gel, transferred to membrane and hybridized to a 32 P-labeled probe as previously described [15,44]. The probe was generated from genomic DNA with forward primer 5'AGT-CACCTGTGGTTCCCTTG and reverse primer 5'AAGGGATGGGGAGAGAGAAA, gel purified and labeled with α-32 P-dCTP using the Prime-It RmT Random Primer hybridization labeling kit (Agilent Technologies). After development, films were scanned and quantitated by NIH ImageJ.

In situ hybridization
In situ hybridization probes were prepared and used to perform in situ hybridization as described previously [21,45]. Anti-PECAM1 antibody (BD Pharminogen, MEC13.3) binding to whole mount embryos was also performed as described previously [18].

qPCR and qRT-PCR
To determine copy number of Notch1[+] and Notch1[12f] alleles a qPCR strategy was employed. A standard number-of-molecules curve was performed with a plasmid containing a 100 bp fragment of the serum albumin precursor gene Alb and used in subsequent experiments to calculate the number of molecules in each sample. Numbers between 1,000 and 1,000,000 were within the linear range. Genomic DNA from wild type, Notch1[12f/+] and Notch1[12f/12f] progeny of Notch1[12f/+] heterozygote crosses was subjected to qPCR using the following primers: Alb Fwd-CAG GTGTCAACCCCAACTCT; Rev-CCACACAAGGCAG TCTCTGA; Notch1[+] Fwd-GGTTCTTGTGAGCTAT TAGCAGCT; Rev-AAGATGAGCTCCCTGCTCTG; Notch1[12F] Fwd-TCCAGTTGGAGTCAGCACAG; Rev-CAGCCCAACCTGATCCTCTA. The amount of product from each allele compared to Alb was calculated and the ratio of alleles was determined and compared to a wild type C57BL/6 J genomic DNA standard.
To determine expression of Notch target genes, wild type and Notch1[12f/12f] embryos at E9.5 were genotyped from yolk sac genomic DNA and their heads were dissected for extraction of RNA using the Trizol method (Invitrogen). RNA amount was determined by nanodrop and 250 ng was reverse transcribed using a Verso cDNA kit (Thermo Fischer Scientific cDNA kit). Primers for Notch1 RT-PCR (Additional file 4: Table S1) and for qRT-PCR of Notch target genes (Additional file 6: Table  S2) were designed using NCBI Primer blast. PCR products of overlapping Notch1 primers were generated and sent for sequencing by Macrogen. Primers used in qRT-PCR were shown to give single amplification products in an Applied Biosystems Viia 7 thermal cycler. Real time PCR was performed using 1 μl of the cDNA and Absolute Blue qPCR mix (Thermo Fischer Scientific). PCR conditions were 50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min, 95°C for 15 s and 60°C for 1 min. Gene expression relative to housekeeping genes (Actb, Gapdh) was calculated using the 2 -ddCt method as described [46]. A Notch1 pathway RT 2 Profiler PCR PAMM-259Z array (Qiagen) was also used to determine relative gene expression in control versus Notch1[12f/12f] cDNA preparations from embryonic brain. The array included 5 housekeeping genes.

Statistics
The significance of mouse progeny expected versus observed was determined by Chi-squared test. Numeric data were analyzed for significance using the unpaired Student's t test assuming both populations have the same STDEV for qRT-PCR or using Welch's correction for comparison of the number of somites in Prism, version 8.0, as noted.
a Canada Research Chair, the Sherman Family Chair in Genomic Medicine