Sustained Notch2 signaling in osteoblasts, but not in osteoclasts, is linked to osteopenia in a mouse model of Hajdu-Cheney syndrome

Individuals with Hajdu-Cheney syndrome (HCS) present with osteoporosis, and HCS is associated with NOTCH2 mutations causing deletions of the proline-, glutamic acid-, serine-, and threonine-rich (PEST) domain that are predicted to enhance NOTCH2 stability and cause gain-of-function. Previously, we demonstrated that mice harboring Notch2 mutations analogous to those in HCS (Notch2HCS) are severely osteopenic because of enhanced bone resorption. We attributed this phenotype to osteoclastic sensitization to the receptor activator of nuclear factor-κB ligand and increased osteoblastic tumor necrosis factor superfamily member 11 (Tnfsf11) expression. Here, to determine the individual contributions of osteoclasts and osteoblasts to HCS osteopenia, we created a conditional-by-inversion (Notch2COIN) model in which Cre recombination generates a Notch2ΔPEST allele expressing a Notch2 mutant lacking the PEST domain. Germ line Notch2COIN inversion phenocopied the Notch2HCS mutant, validating the model. To activate Notch2 in osteoclasts or osteoblasts, Notch2COIN mice were bred with mice expressing Cre from the Lyz2 or the BGLAP promoter, respectively. These crosses created experimental mice harboring a Notch2ΔPEST allele in Cre-expressing cells and control littermates expressing a wild-type Notch2 transcript. Notch2COIN inversion in Lyz2-expressing cells had no skeletal consequences and did not affect the capacity of bone marrow macrophages to form osteoclasts in vitro. In contrast, Notch2COIN inversion in osteoblasts led to generalized osteopenia associated with enhanced bone resorption in the cancellous bone compartment and with suppressed endocortical mineral apposition rate. Accordingly, Notch2 activation in osteoblast-enriched cultures from Notch2COIN mice induced Tnfsf11 expression. In conclusion, introduction of the HCS mutation in osteoblasts, but not in osteoclasts, causes osteopenia.

Notch signaling plays a fundamental role in cell fate determination (1). Interactions of the four Notch receptors with cog-nate ligands of the Jagged and Delta-like families lead to the proteolytic cleavage of the receptor and the release of the Notch intracellular domain (NICD) 2 from the cellular membrane (2). Subsequently, the NICD translocates to the nucleus and forms a complex with recombination signal-binding protein for the immunoglobulin J region (Rbpj), mastermind-like (Maml), and additional DNA-associated proteins to elicit a transcriptional response (3). These events result in the induction of Notch target genes, such as Hes1, Hey1, Hey2, and HeyL (4). Although this signaling mechanism is shared by the Notch paralogs, each receptor has distinct functions (5). The reason appears to be related to the differential cellular pattern of expression of the receptors, structural differences between the paralogs, and interactions of the individual NICDs with Rbpj (6 -8).
Hajdu-Cheney syndrome (HCS) is a rare and devastating disease with multiple systemic manifestations, including osteopo-rosis, short stature, craniofacial deformities, and acroosteolysis (20,21). The condition is associated with mutations in exon 34 of NOTCH2 that create a premature stop codon immediately upstream of the sequences coding for the proline-(P), glutamic acid-(E), serine-(S), and (T) threonine-rich (PEST) domain (22)(23)(24)(25)(26). The latter is required for the proteasomal degradation of NOTCH2, so that the mutations are predicted to lead to the translation of a stable NOTCH2 protein with sustained activity. Recently, we established a murine model of HCS by introducing the mutation found in a subject with severe osteoporosis into the mouse genome. The mutant, termed Notch2HCS, expresses a Notch2 protein of 2318 amino acids that lacks the PEST domain. Heterozygous Notch2HCS mice exhibit Notch2 gain-of-function and generalized osteopenia secondary to enhanced bone resorption, which was ascribed to the sensitization of osteoclast precursors to Rankl and increased Tnfsf11 expression in osteoblasts (27). However, the individual contribution of cells of the osteoclast and osteoblast lineages to the osteopenic phenotype of Notch2HCS mice remains to be determined.
In this study, a conditional by inversion (COIN) approach was utilized to create a conditional mouse model of HCS (Notch2 COIN ) (28,29). This system was designed to introduce a premature STOP codon in exon 34 of Notch2 following Cremediated recombination leading to the translation of a truncated Notch2 protein, thus mimicking the genetic defect associated with HCS. To study the consequences of the Notch2 truncation in specific skeletal cell lineages, Notch2 conditional mice were crossed with appropriate Cre drivers to introduce the mutation in cells of the osteoclast (Lyz2 Cre ) or osteoblast (BGLAP-Cre) lineages. Mutant and control mice were examined for skeletal phenotypic changes by microcomputed tomography (CT) and bone histomorphometry, and the potential mechanisms of Notch2 action were explored.

Generation of a conditional HCS mouse model
To induce the HCS mutation in selected cell populations, Notch2 COIN mice were created by inserting an artificial COIN intron into exon 34 of the murine Notch2 locus (Fig. 1A). As a result, exon 34 was split into two exons at a position corresponding to lysine 2384, which is upstream of the PEST domain and downstream of the domains required for the transcriptional activation of Notch2 (NCBI protein database NP035058; Fig. 1B). The COIN module is composed of a gene trap-like lox66_HA-egfp-polyA_lox71 cassette encoding for a hemagglutinin (HA)-internal ribosome entry site and enhanced green fluorescent protein (eGFP) and placed in the antisense strand. The cassette is preceded by a 3Ј-splice region derived from the second intron of rabbit HBB2 and followed by the polyadenyl-  34. The sequence of the insertion site of the COIN module is in lowercase, and gray and white boxes indicate the coding sequence and the 3Ј-UTR (r␤glpA), respectively. C, structure of exon 34 and of the targeting construct correctly integrated. From 5Ј to 3Ј: lox71 (L71), rabbit ␤-globin polyadenylation signal, eGFP-coding sequence, internal ribosome entry site (Gtx ires) 5 , human influenza hemagglutinin (HA) tag coding sequence, 3Ј-splice region from the second intron of the rabbit ␤-globin gene (white curved arrow), and lox66 (L66) that constitute the COIN module and a flippase (FLP) recognition site (FRT)-flanked neo cassette downstream of the human UBp promoter (FRT-hUBp-neo-FRT). Removal of the neo cassette by FLP recombination is indicated (gray dotted lines). D, representation of the silent COIN module in the antisense orientation and of the splicing event (black dotted lines) that excises the COIN module from the nascent transcript, allowing expression of a wild-type Notch2 mRNA and protein. E, generation of the Notch2 ⌬PEST allele by Cre recombinase-mediated permanent inversion of the COIN module, and illustration of the splicing event (black dotted lines) that occurs during the maturation of the Notch2 ⌬PEST transcript. The latter is translated into a Notch2 mutant lacking the PEST domain. The position of the silent lox72 (L72) sequence and of the wild-type loxP site created by Cre recombination of L71 and L66 is indicated. Images are scaled either in kilobase (kb) or bp.

Notch2 activation in osteoblasts causes osteopenia
ation region of the same gene. The COIN element contains a neo selection cassette downstream of the UBp promoter and the EM7 prokaryotic promoter and upstream of the polyadenylation region of Pgk1 flanked by flippase (FLP) recognition target (FRT) sequences (Frt-neo-Pgk1polyA-Frt) (Fig. 1C) (30 -32). Prior to Cre recombination, the COIN module is removed by splicing of the precursor mRNA to generate a Notch2 COIN transcript that is indistinguishable from the Notch2 WT mRNA (Fig. 1D). In the presence of Cre recombinase, which recognizes the lox71 and lox66 mutant sites in a mirror image configuration, the lox66_HA-egfp-polyA_lox71 cassette is brought into the sense strand, causing the irreversible conversion of the COIN allele. The resulting allele encodes for a bicistronic message that is translated into an HA-tagged Notch2 mutant truncated at lysine 2384 and thereby lacking the PEST domain and eGFP (Fig. 1E). This allele was termed Notch2 ⌬PEST .
To ensure skeletal equivalency of the Notch2 COIN and Notch2 WT alleles, the microarchitecture of the distal femur in 1-month-old Notch2 COIN/COIN male and female mice and wild-type C57BL/6J controls of the same sex and age was analyzed. Cancellous bone volume and cortical thickness were not different between Notch2 COIN/COIN mice and controls, demonstrating that homozygosity for the Notch2 COIN allele has no appreciable effect on femoral microarchitecture (data not shown).

Inversion of the Notch2 COIN allele in the germ line causes osteopenia
To validate the Notch2 COIN mouse as a model of HCS, the skeletal phenotype of Notch2 ⌬PEST/WT mice created by inversion of the COIN module in the germ line was determined. To this end, Notch2 COIN/WT male mice were crossed with Hprt-Cre female mice to create Notch2 ⌬PEST/WT mice; these were crossed with wild-type mice to create Notch2 ⌬PEST/WT heterozygous and control wild-type littermates. COIN inversion was documented by the presence of the Notch2 ⌬PEST allele in DNA from tails of Notch2 ⌬PEST/WT mice, and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis of total RNA from tibiae confirmed the expression of the Notch2 ⌬PEST transcript in mutant mice but not in control littermates (Fig. 2, A and B). Notch2 WT transcript levels were ϳ50% lower in Notch2 ⌬PEST/WT mice than in wild-type littermates, and this is consistent with a systemic heterozygous Notch2 ⌬PEST inversion and comparable expression levels of the Notch2 ⌬PEST and Notch2 WT alleles (Fig. 2B).
One-month-old germ line Notch2 ⌬PEST/WT male mice appeared normal, albeit a small reduction (ϳ5%; p Ͻ 0.05) in femoral length was noted. Analysis of the distal femur by CT revealed that, compared with sex-matched littermate controls, Notch2 ⌬PEST/WT male mice had a 50% decrease in trabecular bone volume secondary to a reduced number and thickness of trabeculae. Connectivity density was lower, and structure model index (SMI) was higher in Notch2 ⌬PEST/WT mice than in controls, indicating a prevalence of rod-like trabeculae (Table 1 and Fig. 2C). Notch2 ⌬PEST/WT mice had a thin and porous cortical bone, and their femurs were smaller than those from controls, because total area, bone area, and periosteal as well as endocortical perimeters were reduced (Table 1 and Fig.  2C). These results mirror the phenotype reported for global Notch2HCS mutants and validate the Notch2 COIN mouse as a model to study the contribution of selected cell lineages One-month-old male Notch2 ⌬PEST/WT mutants (black bars; Notch2 ⌬PEST ) were compared with wild-type littermate controls (white bars) of the same sex. A, DNA was extracted from tail, and Notch2 COIN inversion was documented by gel electrophoresis of PCR products obtained with primers specific for the Notch2 ⌬PEST allele. Arrows indicate the position of the 250-bp amplicon. B, total RNA was extracted from tibiae, and expression of the Notch2 ⌬PEST and Notch2 WT mRNA was measured by qRT-PCR. Transcript levels are reported as copy number corrected for Rpl38 mRNA levels; data for Notch2 WT were normalized to corrected expression in control. Values are means Ϯ S.D.; n ϭ 4 for control, n ϭ 5 for Notch2 ⌬PEST , all biological replicates. Two technical replicates were used for each qPCR. *, significantly different between control and Notch2 ⌬PEST , p Ͻ 0.05 by t test. C, representative CT images of femoral proximal trabecular bone and midshaft cortical bones of control and Notch2 ⌬PEST mice; complete data set in Table 1.

Table 1
Femoral microarchitecture assessed by CT of 1-month-old Notch2 ⌬PEST/WT (Notch2 ⌬PEST ) mice and sex-matched wild-type littermates (control) CT was performed at the femoral distal end for trabecular or midshaft for cortical bone. Values are means Ϯ S.D.

Inversion of the Notch2 COIN allele in the osteoclast lineage does not cause a skeletal phenotype
To establish whether the osteopenic phenotype of the Notch2HCS mutants is secondary to direct effects in cells of the osteoclast lineage, the Notch2 COIN allele was introduced into Lyz2 Cre/WT heterozygous mice. Subsequently, Lyz2 Cre/WT ; Notch2 COIN/COIN mice were crossed with Notch2 COIN/COIN mice for the creation of Lyz2 Cre/WT ;Notch2 ⌬PEST/⌬PEST experimental mice and Notch2 COIN/COIN littermate controls. In an alternative mating scheme, the Notch2 ⌬PEST inversion was carried out in the context of Lyz2 Cre homozygosity. To this end, Lyz2 Cre/Cre ;Notch2 COIN/WT mice were crossed with Lyz2 Cre/Cre ; Notch2 COIN/WT mice for the creation of Lyz2 Cre/Cre Notch2 ⌬PEST/⌬PEST experimental and Lyz2 Cre/Cre ;Notch2 WT/WT control mice. In preliminary studies, we documented that 1and 4-month-old Lyz2 Cre and 1-month-old Lyz2 Cre/Cre mice did not have a skeletal phenotype as determined by CT of distal femurs, when compared with wild-type controls (data not shown). COIN inversion was demonstrated in cultures of bone marrow-derived macrophages (BMMs) from 1-month-old Lyz2 Cre/WT ;Notch2 ⌬PEST/⌬PEST and Lyz2 Cre/Cre ;Notch2 ⌬PEST/⌬PEST mice, and expression of the Notch2 ⌬PEST mRNA was detected in total RNA from their parietal bones (Fig. 3, A, B, D, and E). These results demonstrate that the Hajdu-Cheney mutation was introduced and transcribed in Lyz2-expressing cells. Femoral microarchitecture of male and female at 1-or 4month-old Lyz2 Cre/WT ;Notch2 ⌬PEST/⌬PEST mice or 1-month-old Lyz2 Cre/Cre ;Notch2 ⌬PEST/⌬PEST mice was not different from that of wild-type sex-matched littermate controls (Tables 2  and 3). In addition, BMM cultures from either Lyz2 Cre/WT ; Notch2 ⌬PEST/⌬PEST or Lyz2 Cre/Cre ;Notch2 ⌬PEST/⌬PEST mice and control littermates formed a similar number of osteoclasts in vitro (Fig. 3

, C and F).
These results demonstrate that the induction of a dual Notch2 mutant allele in cells of the osteoclastic lineage has no skeletal consequences and that the osteopenic phenotype of the global Notch2HCS mutant mice should be attributed to an effect in alternate cells (27).

Inversion of the Notch2 COIN allele in osteoblasts causes osteopenia
To determine whether the osteopenia observed in mice carrying the HCS mutation is driven by an effect in cells of the osteoblastic lineage, the Notch2 ⌬PEST mutation was created in Bglap-expressing cells. For this purpose, BGLAP-Cre ϩ/Ϫ ; Notch2 COIN/COIN and Notch2 COIN/COIN mice were crossed to create BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice and littermate Notch2 COIN/COIN controls. As reported previously, BGLAP-Cre transgenics do not have a skeletal phenotype when compared with wild-type mice (15). Inversion of the COIN allele was detected in DNA from parietal bones of BGLAP-Cre; Notch2 ⌬PEST/⌬PEST mice at 1 and 4 months of age but not in littermate controls (Fig. 4A). Accordingly, the Notch2 ⌬PEST transcript was detected only in bones from BGLAP-Cre; Notch2 ⌬PEST/⌬PEST mice, documenting the induction of the HCS mutation in cells that express BGLAP. The presence of the Notch2 ⌬PEST mRNA was associated with increased transcript levels for Hey1, Hey2, and HeyL, demonstrating increased Notch2 signaling (Fig. 4B).
The general appearance, weight, and femoral length of 1-and 4-month-old BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice were not different from those of control sex-matched littermates ( Fig.  5A). At 1 month of age, CT revealed cancellous and cortical bone osteopenia in BGLAP-Cre;Notch2 ⌬PEST/⌬PEST female but not male mice. BGLAP-Cre/Rpl38 copy number was (mean Ϯ S.D.; n ϭ 5-6) 1.6 Ϯ 0.7 in male and 3.5 Ϯ 1.2 (p Ͻ 0.05) in female littermates, possibly explaining the absence of a phenotype in BGLAP-Cre;Notch2 ⌬PEST/⌬PEST male mice. One month old BGLAP-Cre;Notch2 ⌬PEST/⌬PEST female mice had an ϳ50% reduction in cancellous bone volume secondary to a reduced number of trabeculae and connectivity density, associated with increased SMI, indicating a prevalence of rod-like over platelike trabeculae. Cortical bone thickness and bone area were

Notch2 activation in osteoblasts causes osteopenia
decreased in female mutant mice, and cortical bone was porous ( Fig. 5B and Table 4). At 4 months of age, the skeletal phenotype of BGLAP-Cre;Notch2 ⌬PEST/⌬PEST female mice was less pronounced, and cancellous bone volume/total volume was 30% lower than in control littermates (p Ͻ 0.071). A modest cortical osteopenia with cortical thinning and increased porosity was observed in BGLAP-Cre;Notch2 ⌬PEST/⌬PEST 4-month-old mice of both sexes ( Fig. 6B and Table 4).
Cancellous bone histomorphometry of the distal femur of 1-month-old female BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice confirmed the decreased bone volume/tissue volume secondary to a reduced number of trabeculae. Eroded surface and osteoclast numbers were increased, whereas the numbers of osteoblasts and bone formation rates were not different from control littermates ( Table 5). Cortical bone histomorphometry revealed a suppressed endocortical mineral  apposition rate in BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice ( Table 6).

Inversion of the Notch2 COIN allele in osteoblasts induces Tnfsf11
To determine the mechanisms responsible for the skeletal phenotype of the BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice, osteoblast-enriched cells were obtained from the parietal bones of Notch2 COIN/COIN newborns. Cultures were infected with an adenoviral vector expressing Cre recombinase under the control of the cytomegalovirus (CMV) promoter, and parallel cultures infected with an adenoviral vector where the CMV promoter governs GFP expression (Ad-CMV-GFP) served as controls. Ad-CMV-Cre, but not Ad-CMV-GFP, infection led to the inversion of the COIN module and expression of the Notch2 ⌬PEST mRNA associated with induction of Hey1 and HeyL, demonstrating activation of Notch signaling (Fig. 6, A  and B). In accordance with the enhanced bone resorption  ; in males at 1 month of age n ϭ 7 for control, n ϭ 12 for Notch2 ⌬PEST , and at 4 months of age n ϭ 6 for control, n ϭ 6 for Notch2 ⌬PEST ; in females at 1 month of age n ϭ 7 for control, n ϭ 7 for Notch2 ⌬PEST , and at 4 months of age n ϭ 5 for control, n ϭ 6 for Notch2 ⌬PEST , all biological replicates. B, representative CT images of femoral proximal trabecular bone and midshaft. Complete data set in Table 4.

Discussion
In this study, the individual contributions of the osteoclast and osteoblast lineages to the bone loss observed in Notch2HCS mutant mice were explored by the conditional introduction of the HCS genetic defect in selected cell lineages. The mutations associated with the disease occur within exon 34 of NOTCH2, and conditional insertion of a premature STOP codon in the homologous region of the murine Notch2 locus was achieved by the creation of a COIN allele. The COIN module can be introduced directly into coding exons without disrupting the expression or function of the targeted allele, a goal that cannot be accomplished with traditional Cre-loxP approaches (28). Absence of an appreciable phenotype in Notch2 COIN/COIN mice documented the skeletal equivalency of the wild-type and engineered Notch2 alleles prior to Cre-mediated inversion. The Notch2 ⌬PEST mutants generated by germ line inversion of the COIN module expressed the Notch2 ⌬PEST transcript and exhibited a 50% reduction in wild-type Notch2 mRNA, indicating comparable expression levels of maternal and paternal Notch2. Notch2 ⌬PEST germ line mice exhibited generalized osteopenia and reduced bone size and length, phenocopying global Notch2HCS mutants. These results validated the COIN strat-egy and confirmed that generalized expression of a Notch2 mutant lacking the PEST domain causes bone loss (27). Although these findings should be extrapolated with caution to the human condition, they support the concept that de novo or inherited dominant NOTCH2 gain-of-function mutations are responsible for the bone loss in subjects with HCS (33).
Selective introduction of the HCS mutation in osteoblasts, but not in cells of the myeloid lineage, led to generalized bone loss. The reduction in cancellous bone volume was observed only in female mice and was more pronounced in younger BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice. The bone loss was attributed to enhanced bone resorption uncoupled from a boneforming response and suppressed endocortical bone formation. These features are consistent with the skeletal phenotype of global Notch2HCS mutants and demonstrate that a direct effect in osteoblasts is largely responsible for the osteopenia associated with the HCS mutation in mice (27). Absence of a phenotype in Lyz2 Cre/WT ;Notch2 ⌬PEST/⌬PEST and Lyz2 Cre/Cre ; Notch2 ⌬PEST/⌬PEST mice is congruent with the observation that the Notch2 deletion in Lyz2-expressing cells has no consequences on skeletal homeostasis (18). These results indicate that either the activation or inactivation of Notch2 in myeloid cells in vivo has no skeletal consequences and that the effect of Notch2 on bone resorption is secondary to its actions on alter-

Notch2 activation in osteoblasts causes osteopenia
nate cells (17,27,34). However, the in vivo observations are in contrast with in vitro studies demonstrating that Notch2 enhances osteoclastogenesis directly and as a result bone resorption (17). This would suggest that the overall effect of Notch2 in osteoclastogenesis is complex and derived from its actions in various cellular lineages. In agreement with previous work demonstrating increased expression of Tnfs11 in bone extracts from Notch2HCS mutant mice, Notch2 ⌬PEST/⌬PEST osteoblasts expressed increased levels of Tnfs11 mRNA suggesting that osteoblast-derived Rankl is responsible for the enhanced bone resorption in vivo in HCS mutant mice. These findings are in agreement with those in a subject with HCS and severe osteoporosis who was reported to present with elevated levels of circulating RANKL (27,35). However, a limitation of this work was the inability to detect Rankl protein by Western blot analysis in either control or Notch2 ⌬PEST/⌬PEST osteoblasts. This is possibly related to low levels of Rankl expression and the lack of available antibodies with sufficient sensitivity to detect murine Rankl in osteoblasts. There was an absence of a bone-forming response to the increased bone resorption implying that Notch2 inhibits bone formation. Moreover, Notch2 gain-of-function suppresses endocortical mineral apposition rate, an effect that possibly contributes to the cortical osteopenic phenotype. The role of Notch2 as an inhibitor of bone formation is supported by previous studies demonstrating that deletion of Notch2 in Runx2expressing cells increases trabecular bone volume due to enhanced osteoblast differentiation and activity (18). Further support for an inhibitory role of Notch2 on bone formation is derived from studies showing that the dual inactivation of Notch1 and Notch2 in cells of the osteoblastic lineage increases bone mass (36,37).
It is important to mention that some discrepancies exist between the phenotypes of the BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice and of the global Notch2HCS mutants (27). The osteoblast-selective mutation did not affect femoral length, and this was expected because the BGLAP-Cre transgene is not expressed in chondrocytes, cells that govern longitudinal bone growth. Direct inhibitory effects of Notch2 on endochondral bone formation are accountable for the reduced femoral length of the Notch2HCS mutants (38,39). Cancellous bone osteopenia was detected only in female BGLAP-Cre;Notch2 ⌬PEST/⌬PEST mice, although both sexes were affected by the global Notch2HCS mutation (27). These sex-related differences may be secondary to the more pronounced expression of the BGLAP-Cre transgene in female than in male mice. Alternatively, a higher rate of bone remodeling in young female than in male mice, a known attribute of the C57BL/6 genetic background, might have sensitized female mice to a greater activation of Notch2 in osteoblasts (40,41). The cortical bone

Notch2 activation in osteoblasts causes osteopenia
osteopenia was milder in BGLAP-Cre;Notch2 ⌬PEST/⌬PEST than in the Notch2HCS mice, and low expression of the BGLAP-Cre transgene during embryonic skeletal development might account for the less pronounced phenotype of the conditional mice (42). It is of interest that the BGLAP-Cre; Notch2 ⌬PEST/⌬PEST mice did not display the increase in endocortical bone resorption observed in the global Notch2HCS mutants. This difference may also account for the modest cortical bone phenotype of the conditional mice and suggests that the presence of the HCS mutation in both osteoclasts and osteoblasts might be necessary to recapitulate the cortical bone-resorptive phenotype and osteopenia of the Notch2HCS mouse (27). The conditional HCS model described in this study reaffirmed that Notch2, like Notch1, increases the transcript levels of Hey1, Hey2, and HeyL, thereby confirming that both paralogs are able to activate Rbpj-mediated Notch signaling in skeletal cells. The increase in mRNA levels for the Notch target genes reflects activation of the Notch canonical pathway but does not imply that Hey proteins mediate the effects of Notch2 in bone. In fact, either generalized or skeletal misexpression of Heys has a small impact on skeletal microarchitecture (43)(44)(45)(46). The current observations also indicate that Notch1 and Notch2 have distinct skeletal functions because Notch1 induces osteoprotegerin and inhibits bone resorption, whereas Notch2 induces Rankl and stimulates the resorptive event.
In conclusion, osteoblast expression of a Notch2 mutant lacking the PEST domain causes osteopenia in mice.

Creation of the Notch2 COIN mouse
The targeting vector containing the COIN element was electroporated into embryonic stem (ES) cells, and the cassette was used for the selection of G418-resistant cells from 129SvJ/ C57BL/6J embryos at the Gene Targeting and Transgenic Facility of UConn Health. Targeted clones were verified by longrange PCR of genomic DNA. Correct integration of the 5Ј-homology arm was tested with forward F1 5Ј-GGGAGGT-GCTTACCGACCTCTC-3Ј and reverse R1 5Ј-CACCCT-GAAAACTTTGCCCCCTCC-3Ј primers followed by nested forward F2 5Ј-CTGTTCTTGGATACCGAGGTACAC-3Ј and reverse R2 5Ј-CAATCAAGGGTCCCCAAACTCAC-3Ј primers. Proper integration of the 3Ј-homology arm was ensured with forward F3 5Ј-CCAAAACCCGGCGCGGAGGC-CATGC-3Ј and reverse R3 5Ј-CACTTGAGAGCAAGGCTG-CAAGGC-3Ј primers followed by nested forward F4 5Ј-CCTTCTTCTCTTTCCTACAGTACCCC-3Ј and reverse R4 5Ј-GGTGCAAGGGCAGGAGATCAACAG-3Ј primers (all primers from Integrated DNA Technologies, IDT, Coralville, IA). Positive ES clones were used for morula aggregations and the creation of chimeras, and the Frt-neo-Pgk1polyA-Frt cassette was removed by FLP recombination following crosses of male chimeras with mice expressing FLP under the control of the Rosa26 promoter (Rosa26 FLP ; The Jackson Laboratory, Bar Harbor, ME) (47,48). Excision of the cassette was verified by PCR in ear punches of F1 pups, and the Rosa26 FLP allele segregated by breeding with C57BL/6J wild-type mice.
Correct integration of the COIN module into the Notch2 locus was confirmed in the progeny by loss of wild-type allele assay.

Induction of the HCS mutation in the germ line, osteoclasts, or osteoblasts
To test whether the Notch2 COIN and Notch2 WT alleles are functionally equivalent, the skeletal phenotype of Notch2 COIN/COIN mice was compared with the phenotype of wild-type C57BL/6J controls of the same age and sex. To achieve systemic inversion of the Notch2 COIN allele, F1 heterozygous Notch2 COIN/WT male mice were bred with female mice expressing Cre under the control of the Hprt promoter (Hprt Cre ) (49). This resulted in the germ line inversion of the COIN module and consequent creation of mice heterozygous for the Notch2 ⌬PEST allele (Notch2 ⌬PEST/WT ). The latter were crossed with wild-type C57BL/6J mice to generate Notch2 ⌬PEST/WT experimental and wild-type control cohorts.
C57BL/6J mice where the Cre coding sequence was inserted into the endogenous Lyz2 locus (Lyz2 Cre ; The Jackson Laboratory) were used to express Cre recombinase in cells of the myeloid lineage (50,51

Microcomputed tomography
Femoral microarchitecture was determined using a microcomputed tomography instrument (Scanco CT 40; Scanco Medical AG, Bassersdorf, Switzerland), which was calibrated periodically using a phantom provided by the manufacturer (41,52). Femurs were scanned in 70% ethanol at high resolution, energy level of 55 peak kV, intensity of 145 A, and inte-

Notch2 activation in osteoblasts causes osteopenia
gration time of 200 ms. A total of 100 slices at midshaft and 160 slices at the distal metaphysis was acquired at an isotropic voxel size of 216 m 3 and a slice thickness of 6 m and chosen for analysis. Trabecular bone volume fraction (bone volume/total volume) and microarchitecture were evaluated starting ϳ1.0-mm proximal from the femoral condyles. Contours were manually drawn every 10 slices, a few voxels away from the endocortical boundary, to define the region of interest for analysis, whereas the remaining slice contours were iterated automatically. Total volume, bone volume, bone volume fraction, trabecular thickness, trabecular number, connectivity density, SMI, and material density were measured in trabecular regions using a Gaussian filter ( ϭ 0.8) and user-defined thresholds (41,52). For analysis of cortical bone, contours were iterated across 100 slices along the cortical shell of the femoral midshaft, excluding the marrow cavity. Analysis of bone volume/total volume, porosity, cortical thickness, total cross-sectional and cortical bone area, periosteal and endosteal perimeter, and material density were conducted using a Gaussian filter ( ϭ 0.8, support ϭ 1) with operator-defined thresholds.

Bone histomorphometric analysis
Bone histomorphometry was carried out in 1-month-old mice injected with 20 mg/kg calcein and 50 mg/kg demeclocycline at a 2-day interval and sacrificed 2 days after demeclocycline administration. Femurs were dissected, fixed in 70% ethanol, and embedded in methyl methacrylate. For cancellous bone analysis, bones were sectioned at a thickness of 5 m along the sagittal plane on a Microm microtome (Richards-Allan Scientific, Kalamazoo, MI) and stained with 0.1% toluidine blue. Static and dynamic parameters of bone morphometry were measured in a defined area between 0.35 and 2.16 mm from the growth plate at a magnification of ϫ100 using an OsteoMeasure morphometry system (Osteometrics, Atlanta, GA). Stained sections were used to draw the bone and to measure trabecular separation, number, and thickness, osteoid and eroded surface, as well as to count osteoblast and osteoclast surface and number. Mineralizing surface per bone surface and mineral apposition rate were measured on unstained sections visualized under UV light and a triple diamidino-2-phenylindole/fluorescein/Texas Red set long pass filter, and bone formation rate was calculated.
For cortical histomorphometry, femurs were embedded in methyl methacrylate and cut through the mid-diaphysis along the transverse plane with an EXAKT Precision Saw, ground using an EXAKT 400 CS Micro Grinding System (Exakt Technologies, Oklahoma City, OK), and surface-polished to a thickness of ϳ15 m (Alizee Pathology, Baltimore, MD). Parameters of cortical bone morphometry were measured at a magnification of ϫ400 using OsteoMeasureXP software (Osteometrix). Stained sections were used to draw the cortical bone, marrow space, and cell surfaces, as well as to measure osteoblasts and osteoclasts along the endocortical surface. Mineral apposition rate was measured in unstained sections under UV light, using a triple diamidino-2-phenylindole/fluorescein/Texas Red set long pass filter. Terminology and units used for cancellous and cortical bone histomorphometry are those recommended by the Histomorphometry Nomenclature Committee of the American Society for Bone and Mineral Research (53,54).

Culture of BMMs and osteoclast formation
To obtain BMMs, the marrow was removed by flushing with a 26-gauge needle, and erythrocytes were lysed in 150 mM NH 4 Cl, 10 mM KHCO 3 , and 0.1 mM EDTA (pH 7.4). Cells were centrifuged, and the sediment was suspended in ␣-minimum essential medium (␣-MEM) in the presence of 10% fetal bovine serum (FBS; both from Thermo Fisher Scientific, Waltham, MA) and recombinant human M-CSF at 30 ng/ml. M-CSF cDNA and expression vector were obtained from D. Fremont (St. Louis, MO), and M-CSF was purified as reported previously (55). Cells were seeded at a density of 300,000 cells/cm 2 and cultured for 3-4 days. Inversion of the COIN module was documented by PCR of genomic DNA using primers specific for the Notch2 ⌬PEST allele (Table 7). For osteoclast formation, cells were collected following treatment with 0.05% trypsin/EDTA for 5 min and seeded at a density of 47,000 cells/cm 2 in ␣-MEM

Notch2 activation in osteoblasts causes osteopenia
with 10% FBS, M-CSF at 30 ng/ml, and recombinant murine Rankl at 10 ng/ml. Rankl cDNA and expression vector were obtained from M. Glogauer (Toronto, Canada), and GSTtagged Rankl was expressed and purified as described (56). Cultures were carried out until formation of multinucleated tartrate-resistant acid phosphatase (Trap)-positive cells. Trap enzyme histochemistry was conducted using a commercial kit (Sigma), in accordance with manufacturer's instructions. Trappositive cells containing more than three nuclei were considered osteoclasts.

Osteoblast-enriched cell cultures
The parietal bones of 3-5-day-old Notch2 COIN/COIN mice were exposed to 1.2 units/ml Liberase TM TL (Sigma) for 20 min at 37°C, and cells were extracted in five consecutive reactions (57). Cells from the last three digestions were pooled and seeded at a density of 10,000 cells/cm 2 , as described (40). Osteoblast-enriched cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with non-essential amino acids (both from Thermo Fisher Scientific), 20 mM HEPES, 100 g/ml ascorbic acid (both from Sigma), and 10% heat-inactivated FBS (Atlanta Biologicals, Norcross, GA) in a humidified 5% CO 2 incubator at 37°C. To induce inversion of the COIN allele, cells were infected with Ad-CMV-Cre, and parallel cultures infected with Ad-CMV-GFP (both from Vector Biolabs, Philadelphia, PA) served as controls (58). To this end, sub-confluent osteoblast-enriched cells were transferred to culture medium containing 2% heat-inactivated FBS for 1 h and exposed overnight to 100 multiplicity of infection of replication-defective recombinant adenoviruses. Cells were allowed to recover for 24 h in DMEM containing 10% heat-inactivated FBS and then seeded at a density of 22,000 cells/cm 2 . Confluent cultures were exposed to medium supplemented with 5 mM ␤-glycerophosphate (Sigma) to induce osteoblast maturation. To document inversion of the COIN module, the presence of the Notch2 ⌬PEST allele was determined by PCR in genomic DNA using specific primers (Table 7).

RNA integrity and qRT-PCR
Total RNA was extracted from osteoblast-enriched cells with the RNeasy kit (Qiagen, Valencia, CA) and from homogenized bones with the micro RNeasy kit (Qiagen), in accordance with manufacturer's instructions. The integrity of the RNA was assessed by microfluidic electrophoresis on an Experion system (Bio-Rad), and only RNA with a quality indicator number equal to or higher than 7.0 was used for subsequent analysis (59,60). Equal amounts of RNA were reverse-transcribed using the iScript RT-PCR kit (Bio-Rad) and amplified in the presence of specific primers (all primers from IDT; Table 8) with the iQ SYBR Green Supermix (Bio-Rad) at 60°C for 35 cycles. Transcript copy number was estimated by comparison with a serial dilution of cDNA for Hes1 (from American Type Culture Collection, ATCC; Manassas, VA), Hey1 or Hey2 (T. Iso, Los Angeles, CA), HeyL (D. Srivastava, Dallas, TX), or Tnfsf11 (Source BioScience, Nottingham, UK) (61)(62)(63)(64).
To monitor for the efficiency of the COIN inversion, primers designed to amplify a sequence of the Notch2 transcript coding for the PEST domain were used (Table 8). These primers allow detection by qRT-PCR of the transcripts for Notch2 WT and Notch2 COIN but not for Notch2 ⌬PEST , because the latter lacks the sequences coding for the PEST domain. Notch2 WT and Notch2 COIN copy numbers were measured by comparing with a serial dilution of Notch2 cDNA (Thermo Fisher Scientific). Notch2 ⌬PEST transcripts were detected with primers that generate an amplicon straddling the artificial splice junction generated within exon 34 of the targeted Notch2 locus upon inversion of the COIN module (Table 8). Primers are specific for the Notch2 ⌬PEST mRNA and do not recognize the wildtype Notch2 transcript or the Notch2 COIN mRNA prior to the COIN inversion. Notch2 ⌬PEST copy number was estimated by comparison with a serial dilution of an ϳ200 bp synthetic DNA template (IDT) cloned into pcDNA3.1(Ϫ) (Thermo Fisher Scientific) by isothermal single reaction assembly using commercially available reagents (New England Biolabs, Ipswich, MA) (65).

Notch2 activation in osteoblasts causes osteopenia
Amplification reactions were conducted in CFX96 qRT-PCR detection systems (Bio-Rad), and fluorescence was monitored during every PCR cycle at the annealing step. Data are expressed as copy number corrected for Rpl38 expression estimated by comparison with a serial dilution of Rpl38 (ATCC) (66).

Statistics
Data are expressed as means Ϯ S.D. Statistical differences were determined by Student's t test or two-way analysis of variance with Holm-Šídák post hoc analysis for pairwise or multiple comparisons, respectively.
Author contributions-S. Z. designed research studies, conducted experiments, analyzed data, and wrote the manuscript. J. Y. conducted experiments and analyzed data. A. S. conducted experiments and analyzed data. L. S. conducted the analysis of skeletal phenotypes. C. S. and A. N. E. designed and created the Notch2COIN targeting construct. E. C. designed research studies, analyzed data, and wrote the manuscript.