EOGT enables residual Notch signaling in mouse intestinal cells lacking POFUT1

Notch signaling determines cell fates in mouse intestine. Notch receptors contain multiple epidermal growth factor-like (EGF) repeats modified by O-glycans that regulate Notch signaling. Conditional deletion of protein O-fucosyltransferase 1 (Pofut1) substantially reduces Notch signaling and markedly perturbs lineage development in mouse intestine. However, mice with inactivated Pofut1 are viable, whereas complete elimination of Notch signaling in intestine is lethal. Here we investigate whether residual Notch signaling enabled by EGF-domain-specific O-linked N-acetylglucosamine transferase (Eogt) permits mice conditionally lacking Pofut1 in intestine to survive. Mice globally lacking Eogt alone were grossly unaffected in intestinal development. In contrast, mice lacking both Eogt and Pofut1 died at ~ 28 days after birth with greater loss of body weight, a greater increase in the number of goblet and Paneth cells, and greater downregulation of the Notch target gene Hes1, compared to Pofut1 deletion alone. These data reveal that both O-fucose and O-GlcNAc glycans are fundamental to Notch signaling in the intestine and provide new insights into roles for O-glycans in regulating Notch ligand binding. Finally, EOGT and O-GlcNAc glycans provide residual Notch signaling and support viability in mice lacking Pofut1 in the intestine.


Abbreviations
The small intestine of male and female mammals is comprised of differentiated epithelial lineages derived from Lgr5 hi -expressing and potentially other intestinal stem cells (ISC), which maintain overall mucosal homeostasis

Eogt, Pofut1, Notch receptors and ligands expressed in mouse intestine
Notch signaling is fundamental in regulating the balance of cell interactions necessary for normal development.
Therefore, fine tuning of Notch signaling is essential and is achieved by post-translational modification of Notch receptors by glycosylation [7][8][9] .Figure 1A illustrates consensus sites of predicted glycosylation of mouse NOTCH1 and NOTCH2 by O-fucose initiated by POFUT1, and by O-GlcNAc initiated by EOGT 5,6,22,23 , and the subsequent extension by different enzymes.O-glucose glycans that also decorate Notch receptor EGF repeats are not shown.O-fucose and O-GlcNAc glycans are recognized by Notch ligands whereas O-glucose glycans probably are not 24 .
Roles for Notch signaling have been clearly documented in intestinal stem cell functions 17,18 , but fine tuning in the intestine has not been defined.Single cell RNAseq (scRNAseq) of mouse intestinal epithelium identified complex relationships among expression of Notch1 and Notch2 receptor genes, Delta1 and Delta4 Notch ligand genes and Eogt and Pofut1 glycosyltransferase genes (Fig. 1B,C).While Notch1 and Notch2 were expressed mainly in stem and progenitor cells, the genes encoding Dll1 and Dll4 were predominantly expressed in early goblet cells, with DLL4 also expressed in early enterocytes.Importantly, Pofut1 and Eogt were expressed in the same cell types as Notch receptors and ligands, and also more heterogeneously.We therefore investigated further roles for Notch signaling in intestinal homeostasis by genetic inactivation of Eogt and/or Pofut1 to determine differential impacts on the differentiation of cells along the crypt-villus axis.But first we investigated effects of O-glycan removal on Notch ligand binding in a cell-based assay using CHO cells.

Notch ligand binding to CHO O-glycan mutants
To investigate potential synergism between O-fucose and O-GlcNAc glycans (Fig. 1A), deletion mutants were generated by a CRISPR/Cas9 strategy in Lec1 CHO cells 25 .Lec1 CHO cells do not make hybrid or complex N-glycans and therefore provide a simplified glycosylation environment for glycosyltransferase gene deletions affecting O-glycan synthesis 26  www.nature.com/scientificreports/dKO compared to Pofut1 cKO cells had a P value of 0.056 in the unpaired, one-tailed Student's t test (Fig. 2B).
Although not meeting a significance of P < 0.05, the difference suggested potential synergism between O-fucose and O-GlcNAc glycans.This hypothesis was pursued in vivo.

Eogt supports residual Notch signaling in Pofut1 cKO intestine
Intestinal development was investigated in mice homozygous for Eogt inactivation compared to heterozygotes and wild type littermates.There were no differences in body weight, length of small intestine, production of secretory cells, or morphology and dimensions of villi or crypts in Eogt[+/+], Eogt[+/−] and Eogt[−/−] mice (Supplementary Figure S2).To investigate potential alterations at the molecular level, intestinal crypts were prepared by fractionation as described in Methods.The relative enrichment of Lyz1 in crypts (fraction IV) and Fabp2 in villi (fraction I) was determined by qRT-PCR (Supplementary Figure S3A,B).Notch1 and Lgr5 transcripts were reduced by homozygous inactivation of Eogt (Supplementary Figure S3C-G), indicating an impact of the loss of Eogt on Lgr5+ stem cells which express Notch1 (Fig. 1B).However, this did not lead to altered expression of Notch signaling target genes such as Hes1 in Eogt null crypts.Finally, in all morphological, histological and molecular parameters examined, wild type and Eogt[+/−] intestine were indistinguishable, and thus Eogt heterozygous mice were appropriate controls for evaluating effects in mice with compound mutations in both Eogt and Pofut1.
Conditional inactivation of Pofut1 in the mouse intestine greatly inhibits canonical Notch signaling, but the mice are viable and live to ≥ 6 months of age, albeit with significantly altered lineage representation 15 S4).The progeny obtained and those surviving to P28 after birth are shown in Table 1.All potential genotypes were detected at P8 and reflected the expected Mendelian inheritance ratios.However, by P28 all Eogt:Pofut1 dKO pups had died (Chi-squared significance < 0.025).This indicated that Eogt contributed to development of intestinal epithelium in Pofut1 cKO pups.To investigate effects earlier in postnatal development, histopathology of small intestine was assessed at P15.There were no changes in body weight, villi length or crypt depth in small intestine of Eogt KO, Pofut1 cKO or Eogt:Pofut1 dKO pups compared to control mice at P15 (Supplementary Figure S5A-C).The numbers of Alcian Blue-positive goblet cells and eosin-stained Paneth cells in the small intestine of Eogt KO mice were similar compared to controls (Supplementary Figure S5D-G).In contrast, goblet cells significantly increased in crypts and villi of Pofut1 cKO and Eogt:Pofut1 dKO pups (Supplementary Figure S5D-F).Importantly, the number of goblet cells per crypt was significantly higher in Eogt:Pofut1 dKO compared to Pofut1 cKO intestine, indicating a contribution of Eogt to maintaining secretory cell development.Consistent with this, there was also an increase in Paneth cells in Eogt:Pofut1 dKO crypts compared to Pofut1 cKO crypts (Supplementary Figure S5G).Thus, while deletion of Eogt had no apparent effects at P15, cKO of Pofut1 caused a shift to greater secretory cell differentiation which was exacerbated by the concomitant elimination of Eogt in Eogt:Pofut1 dKO pups.
Since Eogt:Pofut1 dKO mice died about 1 month after birth, intestinal development was investigated at P28. Body weight and length of small intestine were significantly decreased in Eogt:Pofut1 dKO compared to Pofut1 cKO mice (Fig. 3A,B).Both goblet and Paneth cells were significantly increased in Eogt:Pofut1 dKO compared to Pofut1 cKO intestine (Fig. 3C-F).Villus length was significantly decreased in Eogt:Pofut1 dKO intestine compared to control, Eogt KO and Pofut1 cKO intestine (Supplementary Figure S6A).Crypt depth, crypt length and crypt width were increased in Pofut1 cKO and Eogt:Pofut1 dKO compared to control or Eogt KO intestine (Supplementary Figure S6B,D).Thus, for all parameters assessed, conditional deletion of Pofut1 from mice lacking Eogt in intestine caused a more severe phenotype than deletion of Pofut1 alone, indicating an important effect of Eogt in complementing canonical Notch signaling.

Notch activation is markedly reduced in Pofut1 cKO and Eogt:Pofut1 dKO crypts
Notch signaling is active at the base of crypts within Notch receptor-expressing ISC supported by ligand-expressing Paneth cells 16 .Cleaved NICD1 (activated NOTCH1) is mainly detected in ISC and transit amplifying (TA) cells located in the crypt base 27 , and our scRNAseq trajectory analysis of Notch receptor and ligand expression are consistent with this (Fig. 1B,C).To investigate Notch signaling status in crypt cells, NOTCH1 cleavage to NICD1 was determined by immunohistochemistry (Fig. 4A) and western blot analysis (Fig. 4B).Crypts of Pofut1 cKO and Eogt:Pofut1 dKO intestine were enlarged due to a significant increase in secretory cells and a concomitant reduction in crypt base columnar ISC.Val1744 antibody detects cleaved NOTCH1 and positive cells were rare to non-existent in both Pofut1 cKO and Eogt:Pofut1 dKO crypts (Fig. 4A).Control sections showed abundant signal for cleaved NOTCH1 in nuclei.Specificity of signals for antibody to NICD1 was confirmed in sections stained with secondary antibody alone (Supplementary Figure S7).Further, western blot analyses showed activated NOTCH1 and NOTCH2 in control crypt cells, but essentially no signal for NICD1 or NICD2 in either Pofut1 cKO or Eogt:Pofut1 dKO crypt lysates (Fig. 4B, Supplementary Figure S8) with no significant differences detectable between the single and double mutant samples.

Notch ligand binding to Eogt KO, Pofut1 cKO and Eogt:Pofut1 dKO ISC
To investigate Notch ligand binding to Notch receptors in ISC, crypts were isolated, single cell suspensions prepared, and analyzed by flow cytometry.Viable, single cells were gated on CD44+CD45−CD24− ISC (Supplementary Figure S9).Cell surface expression of NOTCH1 was determined using antibody to NECD1, and binding of soluble Notch ligands carrying a C-terminal Fc tag was determined using anti-Fc antibody.Three experiments were performed, each including ISC from control, Eogt KO, Pofut1 cKO and Eogt:Pofut1 dKO mice.Representative flow cytometry profiles and histograms of mean fluorescence index (MFI) are shown in Fig. 5. MFI data for mutant ISC were normalized to control ISC in each experiment.The expression of NOTCH1 on ISC cell populations from mutant mice was variable and did not differ significantly among groups (Fig. 5D).Binding of DLL1-Fc but not DLL4-Fc was significantly reduced compared to control in Pofut1 cKO and Eogt:Pofut1 dKO ISC.Furthermore, DLL1-Fc binding to Eogt:Pofut1 dKO ISC was significantly lower than to Pofut1 cKO ISC.

Notch target and Notch pathway gene expression in Eogt:Pofut1 dKO intestine
To investigate the functional impact of EOGT and POFUT1 synergism, expression of Notch signaling target genes, lineage and ISC marker genes and Notch pathway member genes were investigated in crypt cells.
Fractionation produced a similar enrichment of crypt versus villi cells regardless of genotype (Supplementary Figure S4A,B).Hes1 is directly regulated by Notch signaling and is decreased by reduced Notch signaling in the intestine, leading to increased expression of targets such as Math1 15,28 .Consistent with this, Hes1 expression was markedly decreased in Pofut1 cKO crypts, and importantly, was significantly further decreased in Eogt:Pofut1 dKO crypts (Fig. 6A).In parallel, Math1 expression increased in both Pofut1 cKO and Eogt:Pofut1 dKO crypts (Fig. 6A).Furthermore, transcripts of secretory cell marker genes ChgA and Lyz1 were increased in Pofut1 cKO and significantly more so in Eogt:Pofut1 dKO crypts (Fig. 6B) as expected from the increased secretory cells in Pofut1 cKO intestine, and the even greater increase in Eogt:Pofut1 dKO intestine (Fig. 3,E).Muc2 transcripts were also increased, more in Eogt:Pofut1 dKO than Pofut1 cKO crypts (Fig. 6B).Expression of two canonical stem cell marker genes, Lgr5 and Olfm4 were both markedly decreased in single and double mutant crypts (Fig. 6C).
Transcripts of Notch1 were decreased in both Pofut1 cKO and Eogt:Pofut1 dKO crypts, while Notch2 expression was decreased in only Eogt:Pofut1 dKO crypts (Fig. 6D).Jag1 expression was also decreased in both mutants (Fig. 6D).In contrast, Dll1 expression was increased in Eogt:Pofut1 dKO crypts while Dll4.expression was increased in both mutants.Jag2 transcripts were also increased in both of the mutants, and increased to a greater extent in Eogt:Pofut1 dKO compared to Pofut1 cKO crypts (Fig. 6D).Overall, therefore, the expression of marker genes of Notch signaling support the hypothesis that EOGT contributes to canonical, Notch ligand-mediated signaling in maintaining homeostasis of the intestinal mucosa.

Discussion
Here we show that Notch signaling in mouse intestine is fine-tuned by the addition of two types of O-glycan to Notch receptors that modulate the binding of canonical Notch ligands-O-fucose and O-GlcNAc glycans.While the contribution to Notch signaling strength of POFUT1 and O-fucose glycans is much stronger than that of EOGT and O-GlcNAc glycans, the necessity for the latter became clear in Eogt:Pofut1 dKO intestine.Whereas Pofut1:Villin1-Cre cKO mice exhibit severely defective Notch signaling in mouse intestine 15 (and herein), Pofut1 cKO mice live for ≥ 6 months.By contrast, we establish here that Eogt:Pofut1:Villin1-Cre dKO  mice die at ~ 1 month with a more severely Notch signaling defective intestinal phenotype.Nevertheless, these compound mutants live longer than mice lacking both Notch1 and Notch2 or Dll1 and Dll4 which die within days of conditional knockout 17,18 .We propose that limited Notch signaling persists in Eogt:Pofut1 dKO intestine via Notch receptors that are modified by only O-glucose glycans.While O-glucose glycans do not appear to interact directly with Notch ligands 12,13,24 , they are important in regulating Notch receptor trafficking to the cell surface 29,30 .Since anti-NOTCH1 ECD antibodies and soluble Notch ligands DLL1-Fc and DLL4-Fc bound to ISC from Eogt:Pofut1 dKO intestine (Fig. 5), Notch receptors were clearly present at the cell surface.We propose that O-Glucose glycans on Notch receptors support their trafficking in Eogt:Pofut1 dKO intestine, and that elimination of Notch signaling by reduced glycosylation would require the triple knockout of Pofut1, Eogt and Poglut1.
Revealing that Eogt contributes to Notch signaling in mouse intestinal development is important for several reasons.First, to better understand mechanistic bases of regulation of the ubiquitously important Notch signaling pathway.Second, to understand potential pathologies in individuals with the congenital disease of glycosylation EOGT-CDG (Adams-Oliver Syndrome Type 4; AOS4).Those afflicted lack a functional EOGT gene and exhibit a range of limb extremity and scalp developmental defects as well as ocular, vascular and intellectual abnormalities consistent with defective Notch signaling 31,32 .In the mouse embryo, the Eogt gene is first expressed in limb buds and the apical ectodermal ridge before expression in digit condensates at ~ E12.5, consistent with defective development of fingers and toes in EOGT-CDG patients 31 .The involvement of Eogt in mouse intestinal development reported here is a potential indication of the possibility of perturbed intestinal development or function in EOGT-CDG patients.Mutations in EOGT have been associated with 2% of 10,967 colorectal adenocarcinoma patients queried 33 .www.nature.com/scientificreports/Determining how EOGT and the O-GlcNAc glycans fine tune Notch functions in the intestinal environment is challenging.Ideally, NOTCH1 from ISC in the various mouse cohorts would be purified and the nature of the linked O-glycans defined, but this is currently not feasible due to the low concentration of NOTCH1 and NOTCH2 in ISC.The first O-glycan analysis of mouse NOTCH1 from a physiologic source required stimulation of splenic T cells in culture to increase Notch1 expression and obtain sufficient NOTCH1 for mass spectrometric analysis 34 .Roles for the O-glycans on canonical Notch ligands must also be evaluated.Deletion of Rfng is reported to reduce modification of Delta ligands and their ability to stimulate Notch signaling in ISC 16 .However, determining which under-glycosylated partner in a defective Notch receptor-ligand interaction is responsible can only be done when the other partner is fully glycosylated.Thus, proving that Notch ligands require specific O-glycans to induce Notch signaling necessitates conditionally deleting the target glycosylation gene in ligand-expressing cells.Finally, to conclude that Notch signaling is affected by the deletion of a glycosyltransferase requires evidence that the loss generates: (1) a phenotype that mimics loss of Notch pathway member(s); (2) causes alterations in Notch ligand binding; and (3) causes altered expression of Notch signaling target genes.Here we provide this combined evidence, demonstrating that EOGT and the O-GlcNAc glycans it initiates support Notch signaling during intestinal development in the absence of POFUT1 and O-fucose glycans.

Single cell RNAseq bioinformatics
Bioinformatic trajectory analysis of scRNAseq data obtained from Epcam+CD45− C57Bl6/J adult intestinal crypt cells and deposited as GSE188339 was performed as described 35 .

Primers and antibodies
Primer sequences are given in Supplementary Table 1 and antibodies are given in Supplementary Table 2.

Mouse models
Mice with a targeted inactivating mutation in the Eogt gene were developed at Nagoya University and previously described 21 .Pofut1[F/F] mice were also generated previously 38 .Mice expressing Villin-Cre (B6.Cg-Tg(Vil1cre)1000Gum/J, JAX strain 021504) were from Jackson Laboratories, Portland, MA.Cross breeding generated Pofut1 cKO (Eogt[+/−]:Pofut1[F/F]:Villin-Cre), Pofut1:Eogt dKO (Eogt[−/−]Pofut1[F/F]:Villin-Cre) and mutants lacking Villin-Cre.Genomic DNA PCR was used to genotype.Body weight was recorded following euthanasia, small intestine was dissected, length measured, and tissue processed.Experimental protocols were approved by the Albert Einstein Institutional Animal Care and Use Committee (IACUC) under protocol numbers 20170709 and 00001311.All experiments were performed in accordance with the relevant rules and regulations in the approved experimental protocols and this study is reported in accordance with the ARRIVE guidelines (https:// arriv eguid elines.org).Mice were euthanized by asphyxiation in a CO 2 chamber followed by cervical dislocation.

Histopathology and immunohistochemistry
Jejunum was fixed in 10% neutral buffered formalin for 48 h and processed through a graded series of alcohol to prepare paraffin blocks.Sections (5 μm) were stained with hematoxylin and eosin or Alcian Blue.Slides were scanned using a 3D Histech P250 High-Capacity Slide Scanner.CaseViewer 2.4 was used to measure villi length, crypt depth, crypt width and crypt length.QuantCenter software was used to quantitate goblet and Paneth cell staining.For immunohistochemistry, 5 μm sections were dipped into xylene and graded concentrations of ethanol as described (http:// www.abcam.com/ proto cols/).Slides were boiled in sodium citrate for 20 min, incubated with 0.1% Triton X-100 followed by 1.5% H 2 O 2 in Tris-buffered saline (TBS) for 20 min.Antibody incubations were performed at room temperature in a humidified chamber as follows: blocking in 10% FBS with 1% BSA in TBS for 1 h, followed by primary antibody overnight in 1% BSA in TBS, washing with TBS containing 0.025% Triton X-100, and incubation in HRP-conjugated secondary antibody diluted in TBS with 1% BSA for 1 h.Diaminobenzidine peroxidase substrate kit (Vector Laboratories, Burlingame, CA) treatment was followed by counter staining with Hematoxylin and Bluing reagent.Slides were treated with a graded ethanol series and xylene and mounted using Permount.

Quantitative RT-PCR (qRT-PCR)
Total RNA was extracted from ~ 10 7 frozen cells using TRIzol (Thermo Fisher Scientific, Waltham, MA).RNA was dissolved in RNAse-free water and 1 µg estimated using Nanodrop was used to make 20 µl cDNA (Verso cDNA synthesis kit, Thermo Fisher Scientific).cDNA was amplified using the PowerUp SYBR Green master mix (Thermo Fisher Scientific) and 750 nM each primer.Vii7 Real-Time PCR system (Applied Biosystems, Foster City, CA) was used to run qRT-PCR for 40 cycles.Each sample was run in triplicates using 384 wells plate.Gene expression was calculated relative to Gapdh and Hprt by log2 ddCT method.

Western blotting
Frozen cells (~ 10 7 ) were homogenized in 100 μl lysis buffer containing 1% IGEPAL, 1%TX-100, 0.5% Deoxycholate (all from Sigma-Aldrich, St. Louis, MO), and Roche complete™ Protease Inhibitor (Sigma-Aldrich) and incubated on ice for 30 min.After centrifugation at 5000g for 5 min at room temperature, the supernatant in a new tube received 20% glycerol and protein concentration was estimated by Bradford's Dye Reagent Concentrate (BioRad Laboratories, Hercules, CA).For gel electrophoresis, 50-100 µg protein was analyzed by SDS-PAGE, transferred to PVDF membrane and blocked in 5% Non-fat dry milk in Tris-buffered saline 0.05% Tween 20 for one hour at room temperature.Membranes were then cut at the mid-point between the 75 and 50 kDa visible molecular weight markers and then were overnight incubated separately with relevant primary antibodies in blocking buffer at 4 °C.Membrane rinsed with Tris-saline was incubated with HRP-secondary antibody in the same buffer for one hour at room temperature.Enhanced chemiluminescent substrate was used and signals visualized on X-ray film (Thermo Fisher Scientific).

Notch ligand binding to ISC
Washed small intestine pieces were scraped to remove villi, transferred to 20 ml ice-cold 2 mM EDTA and 2 mM glutamine in PBS (Ca/Mg free) and incubated on ice for 20 min.Tissue was then transferred to fresh 20 ml PBS (Ca/Mg free) with 2 mM glutamine and shaken by hand for 30 s.This process was repeated another four times.The last four washes were filtered through a 70 μm strainer into 1% BSA/PBS (Ca/Mg free)-coated 50 ml falcon tubes.Filtrates were spun at 1500 rpm for 10 min at 4 °C to pellet crypts.Single cells were prepared in 3 ml enzyme-free dissociation buffer (Gibco, Thermo Fisher Scientific, Waltham, MA) incubated in a 37 °C water bath for 10 min, with pipetting every 30 s, then 7 ml of alpha MEM was added, cells were filtered through a 40 μm strainer, pelleted and resuspended in Zombie NIR dye in PBS (1:7000; Biolegend).After 30 min rocking at 4 °C, cells were washed with PBS, fixed in 4% paraformaldehyde (PFA; Emsdiasum, Hatfield, PA) in PBS (Ca/Mg free) 15 min at 4 °C with rocking, washed 3 times with cold FBB and stored in FBB at 4 °C.Flow cytometry performed within 1-6 days investigated NOTCH1 expression using anti-NECD1 antibody AF5267 and Notch ligand binding for DLL1-Fc (R&D Systems, Inc., Minneapolis, MN) and DLL4-Fc (Aro Biosystems, Newark, DE).Briefly, ~ 10 6 fixed cells were resuspended in 40 μl CD16/CD32 Fc blocker in FBB (1:40) and incubated 15 min on ice.Antibodies in FBB to CD45 (1:400), CD44 (1:800), CD24 (1:800), CD166 (1:800) and GRP78 (1:800) with anti-NOTCH1 or 2 μg of Notch ligand-Fc were added to cells in FcR block and incubated 30 min on ice.Cells with no anti-NOTCH1 or ligand-Fc determined background.Cells were washed and incubated with secondary antibody (1:600 rhodamine Red-X conjugated donkey anti-sheep IgG or 1:100 Fcspecific anti-IgG-Dylight 405) for 30 min on ice, washed, resuspended in 300 μl FBB, filtered into flow tubes and 300,000 events were recorded in a CytekTM Aurora flow cytometer, analysis by FlowJo software (BD Biosciences).

Statistical analysis
Data are represented as mean ± SEM.GraphPad Prism 9.0.1 was used to perform one-way ANOVA followed by Tukey's multiple comparisons tests or unpaired two-tailed Student's t tests with Welch's correction as noted.Observed and expected allele inheritance was analyzed at ~ P8 and of survival at ~ P28 by the Chi-squared test.

Figure 1 .
Figure 1.O-glycans on NOTCH receptors and expression of Eogt and Pofut1 in mouse intestine.(A) Diagram of mouse NOTCH1 and NOTCH2 ECDs showing O-fucose and O-GlcNAc glycans predicted at EGF repeats that contain the appropriate amino acid consensus sequence: C 2 XXXXS/TC 3 for O-fucose and C 5 XX(G/S/P) (Y/F/T)T/SGXXC 6 for O-GlcNAc glycans (indicated by the respective initiating sugar).O-glucose glycans at a third consensus site are not shown.NRR Notch regulatory region.(B) Eogt and Pofut1 expression in mouse intestinal cells by scRNAseq.The trajectory analysis of the combined data from intestinal cells of 3 mice is shown.The definition of each intestinal cell type is given in Abbreviations.(C) Transcript levels of candidate genes expressed in each cell type were quantitated for 3 mice.