Dmp1Cre-directed knockdown of PTHrP in murine decidua is associated with increased bone width and a life-long increase in strength specific to male progeny

Parathyroid hormone related-protein (PTHrP) is a pleiotropic regulator of tissue homeostasis. In bone, knockdown in osteocytes by Dmp1Cre-targeted deletion causes osteopenia and impaired strength. We report that this outcome depends on parental genotype. Adult Dmp1Cre.Pthlhf/f mice from homozygous parents (Dmp1Cre.Pthlhf/f(hom)) have stronger bones, with 40% more trabecular bone mass and 30% greater femoral width than controls. At 12 days old, greater bone width was also found in male and female Dmp1Cre.Pthlhf/f(hom) mice, but not in gene-matched mice from heterozygous parents, suggesting a maternal influence before weaning. Milk PTHrP levels were normal, but decidua from mothers of Dmp1Cre.Pthlhf/f(hom) mice were smaller, with low PTHrP levels. Moreover, Dmp1Cre.Pthlhf/f(hom) embryonic bone was more mineralized and wider than control. We conclude that Dmp1Cre leads to gene recombination in decidua, and that decidual PTHrP influences decidual cell maturation and limits embryonic bone growth. This identifies a maternal-derived developmental origin of adult bone strength.


Introduction 39
Bone size and geometry are among the many factors determining bone strength (1). 40 During skeletal development, bone grows in both the longitudinal and radial axes. 41 Longitudinal growth is mediated by chondrocytes at the growth plates, where 42 hypertrophic chondrocytes cease dividing, enlarge and eventually mineralize 43 surrounding matrix (2). Simultaneously, expansion of bone diameter (termed "radial 44 growth") balances bone length and width. Although longitudinal growth has been studied 45 widely, little is known about the signaling pathways orchestrating radial growth (3). 46 Parathyroid hormone-related protein (PTHrP, gene name: Pthlh) is produced by many 47 tissues, and acts locally to maintain their physiological function (4). While global 48 knockout of PTHrP is neonatal lethal and causes widespread skeletal defects including 49 reduced bone length, due largely to PTHrP's role in promoting chondrocyte maturation 50 (5), heterozygous Pthlh deletion causes osteopenia in adult mice (6). Local PTHrP 51 production by bone cells is also required for normal bone formation in adults during 52 bone remodeling. This was established by studies in genetically altered mice; mice with 53 Pthlh knockdown targeted to osteoblasts (Col1(2.3kb)Cre.Pthlh f/f ) (6) or to osteocytes 54 (Dmp1(10kb)Cre) (7) both exhibiting low bone formation and osteopenia in adulthood. 55 Here we report an effect of parental genotype on the bone structure of Dmp1Cre.Pthlh f/f 56 mice. This study arose from an unexpected finding when in follow up of our previous 57 study (7), we sought to assess the effect of Dmp1Cre-targeted knockdown of PTHrP in 58 osteocyte in older mice. For this, we changed our breeding strategy from using 59 heterozygous breeders to homozygous breeders to limit mouse wastage. As in previous 60 studies from our laboratory (8,9), we generated these mice using cousin-bred 61 homozygous breeding pairs. To our surprise, adult male PTHrP-deficient mice generated 62 from homozygous breeders (denoted Dmp1Cre.Pthlh f/f(hom) ) exhibited an opposing 63 phenotype to that of mice used in our previous work, generated from heterozygous 64 breeders (denoted Dmp1Cre.Pthlh f/f(het) ) (7): adult male Dmp1Cre.Pthlh f/f(hom) mice had 65 high trabecular bone mass, and wide long bones, but normal body weight and normal 66 bone length. Since this was a profound and reproducible phenotype, we sought to 67 determine the parental source of the defect in bone structure. 68 Although we previously reported that Dmp1Cre can lead to gene recombination in the 69 mammary gland (7), there was no alteration in milk PTHrP levels in Dmp1Cre.Pthlh f/f(het) 70 dams. However, suckling male and female Dmp1Cre.Pthlh f/f(hom) mice both exhibited the 71 wide bone phenotype. We traced the phenotype back to fetal development and found it 72 was associated with low PTHrP levels in decidua basalis and impaired decidualization in 73 mothers of Dmp1Cre.Pthlh f/f(hom) mice. This implies that PTHrP from the decidua limits 74 bone radial growth in male and female mice, and that this has life-long effects on skeletal 75 size in males, that override the effects of endogenous PTHrP deletion in osteocytes. This 76 has significant implications for bone development, sex-differences in bone growth, and 77 for breeding strategies used with Dmp1Cre-targeted mouse models. 78

Results 79
Adult male Dmp1Cre.Pthlh f/f(hom) mice have a high trabecular bone mass set-point, 80 reached before 14 weeks of age 81 In contrast to our previous experiments showing osteopenia in 12 week old 82 Dmp1Cre.Pthlh f/f(het) mice (7), adult male Dmp1Cre.Pthlh f/f(hom) mice (i.e. mice bred from 83 parents expressing Dmp1Cre and homozygous for the Pthlh f/f genotype, Figure 1A) had 84 greater trabecular bone volume than age-and sex-matched controls ( Figure 1B-G). At 14, 85 16 and 26 weeks of age, male Dmp1Cre.Pthlh f/f(hom) mice had significantly higher 86 trabecular bone volume ( Figure 1B) and trabecular number ( Figure 1C) than age-and 87 sex-matched controls. This phenotype was stable; the proportional difference in 88 trabecular bone volume and number between genotypes was the same at all three time 89 points assessed: i.e. trabecular bone volume and number were at a constant ~140% and 90 ~133% of sex-matched male controls. Trabecular thickness remained unchanged in 91 Dmp1Cre.Pthlh f/f(hom) mice compared to sex-and age-matched controls at all time points 92 ( Figure 1D). Trabecular separation was ~18% lower than controls in male 93 Dmp1Cre.Pthlh f/f(hom) mice, and was statistically significant only at the age of 14 weeks 94 ( Figure 1E). 95 When representative images were generated to show this difference in trabecular bone 96 mass ( Figure 1G), we noted that male Dmp1Cre.Pthlh f/f(hom) mice also had wider bones; 97 when measured in the trabecular region of analysis, metaphyseal periosteal perimeter 98 was significantly higher in male Dmp1Cre.Pthlh f/f(hom) mice at all three time points ( Figure  99 1F), confirming the wider bone phenotype of these mice. When Dmp1Cre.Pthlh f/f(het) mice 100 were bred and aged to 16 and 26 weeks, no significant difference in trabecular structure 101 or metaphyseal bone perimeter was detected at either age between Dmp1Cre.Pthlh f/f(het) 102 mice and their controls (Figure 1 Supplement 1). These indicate that male 103 Dmp1Cre.Pthlh f/f(hom) mice have a different phenotype to Dmp1Cre.Pthlh f/f(het) mice (7), 104 although they have the same genotype, showing that parental genotype influences 105 trabecular bone mass and bone radial growth. 106 The high trabecular bone mass in Dmp1Cre.Pthlh f/f(hom) mice was sex-specific. Female 107 Dmp1Cre.Pthlh f/f(hom) mice showed no significant difference in trabecular bone volume, 108 trabecular number, or metaphyseal periosteal perimeter compared to Dmp1Cre (hom)   Pthlh f/f(hom) distal femoral primary spongiosa analysed by micro-CT in male mice (m) at 14, 16 and 26 weeks of age, and in female mice (f) at 16 and 26 weeks of age. Trabecular bone volume, trabecular number, trabecular thickness, trabecular separation, and metaphyseal periosteal perimeter are shown as mean (dot), interquartile range (box), median (line) and range; n=9-10/group. *p<0.05, **p<0.01, ***p<0.001 compared to sex-and age-matched Dmp1Cre (hom) by two-way ANOVA (16 and 26 weeks old) and Student's t-test (14 weeks old). G: Representative micro-CT images of trabecular bone in the distal femoral primary spongiosa of 26-week old male mice, showing density (scale above images), and raw cross-sectional images of the metaphysis (i), metaphyseal diaphysis (ii) and diaphysis (iii), showing a difference in bone size, and projection of trabecular bone into the lower metaphysis in Dmp1Cre.Pthlh f/f(hom) samples.
To confirm this trabecular phenotype in homozygous-bred mice, trabecular bone 115 structure was studied at a second anatomical region, 5 th lumbar (L5) vertebrae. Similar 116 to long bones, vertebrae of 14 week old male Dmp1Cre.Pthlh f/f(hom) had higher trabecular 117 bone volume and trabecular number than controls (Table 1). Trabecular separation was 118 lower in male Dmp1Cre.Pthlh f/f(hom) mice than controls, and there was no significant 119 difference in trabecular thickness (Table 1). This confirmed the trabecular phenotype 120 was not restricted to a single anatomical location. 121 Although trabecular bone mass was greater in male 14 week old Dmp1Cre.Pthlh f/f(hom) 122 mice, dynamic histomorphometry revealed no difference in any bone formation or 123 resorption parameters compared to controls (Table 2). In addition, no significant 124 difference was detected in serum levels of the bone formation or resorption markers, 125 P1NP and CTX1, of 14 week old male Dmp1Cre.Pthlh f/f(hom) mice compared to controls 126 (Table 2). This, and the similar proportion of elevation in trabecular bone mass at 14, 16 127 and 26 weeks suggest that the high trabecular bone mass arose before 14 weeks of age, 128 and has reached a greater adult set-point for "peak bone mass" than controls. 129 Adult male Dmp1Cre.Pthlh f/f(hom) mice have a wide bone phenotype, reached before 14 130 weeks of age, that leads to greater bone strength 131 Since the metaphyseal bone width was greater in adult male Dmp1Cre.Pthlh f/f(hom) mice 132 than controls, we analysed femoral cortical bone structure in more detail (Figure 2A).

133
Although femoral length was not different between Dmp1Cre.Pthlh f/f(hom) mice and 134 proportion: the ratio of anteroposterior to mediolateral widths was not different in these 138 mice compared to controls at any time point (data not shown). Adult male 139 Dmp1Cre.Pthlh f/f(hom) mice exhibited greater marrow and cortical area, and greater 140 periosteal and endocortical perimeters, compared to age-and sex-matched controls 141 ( Figure 2E-I). Although cortical diameter was greater, this was balanced on the 142 endocortical and periosteal surfaces, as there was no significant difference in cortical 143 thickness ( Figure 2G). As in trabecular bone, the greater cortical bone width phenotype 144 was stable, showing a similar proportional difference compared to controls at all three 145 time points. 146 Female Dmp1Cre.Pthlh f/f(hom) mice showed no significant difference in cortical bone size 147 or shape compared to age-matched Dmp1Cre (hom) mice at 16 or 26 weeks of age ( Figure  148 2). 149 Heterozygous-bred Dmp1Cre.Pthlh f/f(het) mice at 16 and 26 weeks of age did not exhibit 150 any significant difference in anteroposterior or mediolateral femoral width compared to  (hom) and Dmp1Cre (hom) mice at 14, 16 and 26 weeks of age, and female mice (f) at 16 and 26 weeks of age. Anteroposterior (C) and mediolateral (D) width, measured by micro-CT at the midshaft. E:I Femoral marrow area (E), cortical bone area (F), thickness (G), and both endocortical (H) and periosteal (I) perimeter were analysed in cortical ROI by micro-CT . Data are shown as mean (dot), interquartile range (box), median (line) and range; n=9-10/group. *p<0.05, **p<0.01, ***p<0.001 compared to sex-and age-matched Dmp1Cre (hom) by two-way ANOVA (16 and 26 weeks old) and Student's t-test (14 weeks old). Since greater bone width is associated with greater bone strength, we carried out three-160 point bending tests. Femora from 26 week old male Dmp1Cre.Pthlh f/f(hom) mice could 161 withstand higher loads than age-matched Dmp1Cre (hom) controls, reaching a higher 162 ultimate force ( Figure 3A,B) and failure force (Table 3) before breaking. There was no 163 significant difference in ultimate displacement between Dmp1Cre.Pthlh f/f(hom) and control 164 femora ( Figure 3C). When these measurements were corrected for bone size, both 14 and 165 26 week old male Dmp1Cre.Pthlh f/f(hom) femora showed lower ultimate stress and yield 166 stress, compared to controls ( Figure 3D-F). 14 week old male Dmp1Cre.Pthlh f/f(hom) 167 femora had higher ultimate and failure strain than controls (Table 3). 26 week old male 168 Dmp1Cre.Pthlh f/f(hom) femora had lower failure stress ( Figure 3G), and reduced toughness 169 and elastic modulus ( toughness were modified in the male mice, these parameters were not changed in females 174 ( Figure 3 and Table 3), consistent with their loss of the greater bone width with ageing. 175 Surisingly, femora from female Dmp1Cre.Pthlh f/f(hom) mice achieved a greater yield force 176 and displacement compared to age-and sex-matched controls (Table 3), suggesting a 177 higher elastic deformation. When corrected for bone size, 26 week old female 178 Dmp1Cre.Pthlh f/f(hom) femora had higher yield strain and lower elastic modulus (Table 3), 179 suggesting a more flexible material than controls. 180 Femora from 26 week old heterozygous-bred Dmp1Cre.Pthlh f/f(het) mice did not show any 181 significant difference in mechanical properties compared to Dmp1Cre (het) controls ( Figure  182 3 Supplement 1). 183
While no significant differences in cortical dimensions were detected in tibiae from male 198 or female Dmp1Cre.Pthlh f/f(het) mice (from heterozygous breeders), Dmp1Cre.Pthlh f/f(hom) 199 mice (from homozygous breeders) exhibited greater tibial width at 12 days of age ( Figure  200 4B-I) with no difference in tibial length ( Figure 4C). Both male and female 201 Dmp1Cre.Pthlh f/f(hom) mice had significantly greater tibial width in the anteroposterior 202 and mediolateral direction, compared to sex-matched Dmp1Cre (hom) ( Figure 4B,D,E). Male 203 and female Dmp1Cre.Pthlh f/f(hom) mice also showed greater tibial marrow area ( Figure  204 4G), cortical area ( Figure 4H), and periosteal perimeter ( Figure 4I) compared to sex-205 matched cousin-bred Dmp1Cre (hom) controls. Female Dmp1Cre.Pthlh f/f(hom) also had 206 greater cortical thickness than female Dmp1Cre (hom) controls ( Figure 2F). No significant 207 difference was observed in trabecular structure between Dmp1Cre.Pthlh f/f(hom) mice and 208 their Dmp1Cre (hom) sex-matched controls ( Table 4), suggesting that this aspect of the 209 phenotype was secondary to the increase in bone width. 210 Although PTHrP gene recombination was detected in mammary tissue from 211 Dmp1Cre.Pthlh f/f(het) mice (7), milk PTHrP levels, measured either by radioimmunoassay 212 or bioassay, and milk protein levels were not significantly altered in Dmp1Cre.Pthlh f/f(het) 213 mice compared to controls (Table 5). 214 The

Dmp1Cre.Pthlh f/f(hom) wide-bone phenotype exists in utero 215
Since no change in milk PTHrP could explain the phenotype at 12 days of age, we 216 determined whether the phenotype existed in utero by assessing embryonic bone size.

217
Consistent with our observations at 12 days, embryonic Dmp1Cre.Pthlh f/f(hom) femora 218 (E18.5) were wider in both anteroposterior and mediolateral dimensions, and exhibited 219 a higher moment of inertia compared to Dmp1Cre (hom) controls ( Figure (hom) at three different locations shown in panel A: at 20% of the mineralized length distal to the proximal end of the mineralized region (Top 20%), at the midshaft (Mid), and at 20% of the mineralized length proximal to the distal end of the mineralized region (Bottom 20%). Data is shown as mean ± SEM with individual data points, *p<0.05, **p<0.01, and ***p<0.001 compared to controls by two-way ANOVA (B-H).

Low PTHrP levels and modified cell morphology in Dmp1Cre.Pthlh f/f decidua 237
Since we observed increased bone width in utero in Dmp1Cre.Pthlh f/f(hom) mice, and PTHrP 238 is produced by uterus and decidua (11, 12), we sought to determine whether PTHrP 239 expression is modified in placenta or decidua from Dmp1Cre.
with a lack of change in overall bone growth, there were no significant differences in body 241 weight, placental weight, or body to placental weight ratios between Dmp1Cre.Pthlh f/f(hom) 242 embryos and Dmp1Cre (hom) controls ( Figure  PTHrP staining of decidua and placenta showed positive staining for PTHrP in both the 247 decidua and the spongiotrophoblast layer (junctional zone) of the placenta ( Figure 6D).

248
No PTHrP was detected in the placental labyrinth zone. PTHrP staining in decidua from 249 mothers of Dmp1Cre.Pthlh f/f(hom) mice was not as strong as that observed in decidua from 250 mothers of Dmp1Cre (hom) mice ( Figure 6D,E). Quantification revealed a significant 251 reduction in PTHrP staining at all intensities in decidua from mothers of 252 Dmp1Cre.Pthlh f/f(hom) mice, but no change in PTHrP staining frequency in the 253 spongiotrophoblast zone of the adjacent placenta ( Figure 6F). IgG isotype control had 254 minimal intensity in both regions. No alteration in PTHrP staining frequency was 255 observed in in decidua adjacent to Dmp1Cre.Pthlh f/f(het) placenta compared to littermate 256 Dmp1Cre.Pthlh w/w(het) (Figure 6 Supplement 1E). This suggests off-target effects of 257 Dmp1Cre have led to reduced PTHrP protein production by decidual cells. 258 We also examined the morphology of decidua from samples adjacent to 259 Dmp1Cre.Pthlh f/f(hom) placenta. The decidual cells from mothers of Dmp1Cre.Pthlh f/f(hom) 260 embryos appeared more compact than in Dmp1Cre (hom) decidua, suggesting impaired 261 decidual cell maturation ( Figure 6G,H). Total decidual area was significantly less in 262 samples from mothers of Dmp1Cre.Pthlh f/f(hom) embryos than Dmp1Cre (hom) , but the area 263 of the spongiotrophoblast zone was not significantly modified ( Figure 6I)  (hom) and Dmp1Cre (hom) embryos at E17.5; IgG control staining was measured in images of both zones. Decidua (d), spongiotrophoblast (sp) and labyrinth (lb) zones are shown in low power images (D,E). Scale bar = 1 mm. Frequency of PTHrP stained objects segregated by staining intensity in the spongiotrophoblast layer and decidua from Dmp1Cre.Pthlh f/f(hom) and Dmp1Cre (hom) embryos. (F,G) High power images of decidua. Scale bar = 20 micron. (I) Quantitation of total decidual area and spongiotrophoblast area; mean ± SEM with individual data points, *p<0.05, **p<0.01compared to controls by one-way ANOVA.

Discussion 271
This study identifies that off-target effects of Dmp1Cre-mediated recombination led to 272 reduced decidual PTHrP. Reduced PTHrP level in the decidua is associated with increased 273 embryonic bone radial growth and mineralization in utero. This wide bone phenotype is 274 observed in both male and females at 12 days of age, and is sustained until at least 6 275 months of age in male, but not female, skeletons (Figure 7). These effects of reduced 276 decidual PTHrP on bone size, trabecular bone mass, and bone strength dominates over 277 the previously reported effects of reducing endogenous PTHrP in osteocytes, which 278 suppressed bone formation and reduced trabecular bone mass of young adult mice (7). 279 This suggests that locally produced PTHrP is essential for normal decidualization, and 280 through these actions influences embryonic bone growth. This indicates an additional 281 role for PTHrP in maternal physiology. Mice from breeders heterozygous for PTHrP (Dmp1Cre.Pthlh f/w breeders) had normal bone size (length and width) compared to their sex-and age-matched controls, but lower adult trabecular bone mass. Decidual PTHrP may limit fetal skeletal development and radial growth, independent of longitudinal growth. Mothers of Dmp1Cre.Pthlh f/f(hom) mice (which are Dmp1Cre.Pthlh f/f(het) ) had lower levels of decidual PTHrP, leading to wider long bones in Dmp1Cre.Pthlh f/f(hom) progeny. This phenotype was observed not only in embryos and neonatal mice, but also in adult male mice. Adult male mice also showed high trabecular bone mass compared to their sex-matched Dmp1Cre controls. It is very surprising that Dmp1-Cre targeted recombination had an influence on decidua. 318 Although the Dmp1-Cre mouse is widely used as an osteocyte and late-osteoblast 319 conditional knockout mouse, multiple off-target tissues have been reported. These 320 include our previous report of recombination in the mammary gland (7). We and others 321 have shown recombination in skeletal muscle and certain brain cells (7), and reporter 322 genes have also shown Dmp1-Cre expression in preosteoblasts, a subset of bone marrow 323 stromal cells, and gastrointestinal mesenchymal stromal cells (21). To date, there is no 324 report that Dmp1-Cre targets decidua, which is a transient uterine tissue. We previously 325 tested non-pregnant uterus, and found that Dmp1-Cre recombination did not occur (7). 326 The expression of Dmp1-Cre in decidua has major implications for the design and 327 reporting of experiments utilizing Dmp1Cre for gene deletion. However, this clearly 328 depends on the function of the targeted gene. For example, although gp130, and its 329 inhibitor protein SOCS3, are expressed in murine decidua (22, 23), homozygous-bred 330 Dmp1Cre.gp130 f/f mice (8) and Dmp1Cre.Socs3 f/f mice (9) showed phenotypes similar to 331 that of heterozygous-bred mice of the same genotype (9, 24). 332 Although decidua and placenta provide nutrition to promote embryonic and placental 333 weight gain in utero (25) growth. There are two non-mutually-exclusive theories describing how bone radial width 356 is determined: the "mechanostat" theory suggests that bone size and shape are adapt to 357 mechanical strain (32, 33), while the "sizostat" theory suggests a set of genes regulates 358 bone width to reach a pre-programmed setting (34). Although different genomic markers 359 have been correlated with bone size and bone shape (35), no specific genes or molecular 360 pathways have yet been described as major determinants of cortical bone diameter. Our 361 data suggests that decidual PTHrP is a determinant for the cortical width sizostat. 362 Although both male and female Dmp1Cre.Pthlh f/f(hom) mice had wider long bones at 12 363 days of age, this phenotype was retained through to adulthood only in males, suggesting 364 that the mechanisms controlling continued radial growth and bone width are sex-365 dependent. Although placental nutrition has sexually dimorphic effects on embryo 366 growth (36), we did not observe sex differences in this study until after 12 days, again 367 emphasizing that the wide bone phenotype is unlikely to relate to placental nutrition. 368 Post-pubertal sex differences in cortical diameter are common to all mammals (36-45), 369 with females having narrower bones than males, however the molecular mechanisms 370 driving this sexual dimorphism remain largely unknown; this mouse model may 371 therefore shed new light on the mechanisms that contribute to this sexual dimorphism. 372 The retention of this phenotype in males, but not females, suggests that hormonal 373 changes at puberty in females may slow their radial growth. While most studies 374 investigating sexual dimorphism in murine bone width have focused on periosteal 375 growth at the diaphysis, our results suggest that differences between males and females 376 in cortical width might also arise from radial expansion of the growth plate. Estradiol is 377 known to slow longitudinal growth: a previous study has shown that ovariectomy 378 increased tibial length and increased chondrocyte proliferation (46), and 17beta-379 estradiol treatment of 26-day-old female and male rats led to shorter tibial length and an 380 early reduction in growth plate longitudinal width (46). Testosterone also affects 381 chondrocytes: local injection of testosterone into the tibial epiphyseal growth plate of 382 castrated growing male rats significantly increased epiphyseal growth plate length (47). 383 Furthermore, while the perinatal testosterone surge is required for adult bone length, 384 bone width is determined by post-pubertal testosterone (48). The cellular and 385 intracellular pathways by which estradiol and/or testosterone differentially affect 386 growth plate radial growth in control and Dmp1Cre.Pthlh f/f(hom) mice remains to be 387 investigated. 388 In Dmp1Cre.Pthlh f/f(hom) mice, greater cortical width was not associated with greater total 389 bone length. To our knowledge, this is the first evidence of changes in bone diameter 390 independent of cortical thickness, longitudinal growth, and total body weight gain. All 391 previously reported mouse models with changes in bone diameter also showed 392 widespread skeletal development defects such as reduced bone length and width or 393 altered cortical thickness (48-51). For example, mice lacking the endogenous nuclear 394 localization sequence and C-terminus of PTHrP displayed retarded growth with lower 395 body weight and total skeletal size at the age of 2 weeks (50) to age-and sex-matched controls. 406 Although we detected PTHrP recombination in the mammary glands (7), milk PTHrP 407 levels were not significantly modified, and the wide-bone phenotype predated the 408 commencement of suckling, indicating that a change in mammary supply of PTHrP is 409 unlikely to cause the wide-bone phenotype we observed in Dmp1Cre.Pthlh f/f(hom) mice. We 410 had thought that this may have been a possibility since we previously noted Dmp1Cre-411 driven PTHrP recombination in the mammary gland (7), and suckling pups from mice 412 lacking PTHrP in the milk supply (BLG-Cre/PTHrP lox/-) had higher ash calcium content, 413 indicating greater bone mass, compared to controls at day 12 of lactation (10). The 414 normal levels of PTHrP in milk in mothers of Dmp1Cre.Pthlh f/f(hom) mice suggests that the 415 mammary cells expressing Dmp1Cre are not the mammary epithelial cells that secrete 416 PTHrP to the milk (54), and are different to those targeted in the BLG-Cre/PTHrP lox/-417 model (10, 55, 56). Another possibility is that the level of PTHrP recombination was too 418 low in mammary tissues to modify milk PTHrP production. 419 In conclusion, decidual PTHrP limits trabecular bone mass, bone geometry and strength, 420 not only of neonatal mice, but also of adult male mice. Dmp1Cre.Pthlh f/f(hom) embryos had 421 accelerated skeletal development, with more mineralized and wider femora at E18.5.

422
Although this effect was observed in both males and females in neonates, it was retained 423 through to adulthood only in male mice. This indicates that maternal PTHrP limits bone 424 growth, and this has a life-long influence on bone mass, shape and strength in male 425 progeny. 426

Micro-computed tomography (micro-CT) 458
Micro-CT was carried out on samples from E17.5, E18.5, 12 days, 14 and 26 weeks of age. 459 The observer was blinded to genotype and sex of all samples at the time of analysis. 26 460 week old mice were also anaesthetized and scanned by in vivo micro-CT at 16 weeks of 461 age. After collection, embryos were fixed in 95% ethanol for 5 days. Femora of 14 and 26 462 week old mice, and tibiae of 12 day old mice were fixed overnight in 4% 463 paraformaldehyde at 4°C, then stored in 70% ethanol until further analysis. Femoral and 464 tibial morphology and microarchitecture were assessed using the Skyscan 1076 (E18.5, 465 12 days, 14 and 26 weeks of age) or 1276 (E17.5) micro-CT system (Bruker, Aartselaar, 466 Belgium), as described previously (61) with the following modifications. 467 For micro-CT analysis at E17.5 and E18.5, embryos were scanned at 55 kV and 200 mA, 468 and 48 kV and 208 mA, respectively. Projections were acquired over a pixel size of 5µm 469 and 9 µm, respectively. Image slices were then reconstructed by NRecon (Bruker, version 470 1.7.1.0) with beam-hardening correction of 35%, ring artifact correction of 6, smoothing 471 of 1, and defect pixel masking of 50%. The length of mineralized bone was measured in 472 each femur. Femoral cortical structure was analyzed at three sites, based on the extent of 473 mineralised femur: i) 20% of the mineralized length distal to the proximal end of the 474 mineralized region (metaphysis; Top 20%); ii) Midshaft (Mid); iii) 20% of the 475 mineralized length proximal to the distal end of the mineralized region (Bottom 20%).

476
Automatic adaptive thresholding was used for each sample. 477 Tibiae from 12 day old mice were scanned at 37 kV and 228 mA. Regions of interest (ROI) 478 commenced at a distance equal to 30% of the tibial length down the growth plate and an 479 ROI of 10% of the tibial length was analyzed. The lower adaptive threshold limit used for 480 cortical analysis was equivalent to 0.58 g/mm 3 Calcium hydroxyapatite (CaHA). 481 Femora from 14, 16 and 26 week old mice were scanned at 45 kV and 220 mA. For 482 trabecular and cortical analyses, ROI commenced at a distance equal to 7.5% or 30%, 483 respectively, of the total femur length proximal to the distal end of the femur; for each, an 484 ROI of 15% of the total femur length was analyzed. For 14 week old mice, the lower 485 adaptive threshold limits for trabecular and cortical analysis were equivalent to 0.34 486 g/mm 3 and 0.75 g/mm 3 CaHA, respectively. For 16 week old mice, the lower adaptive 487 threshold for trabecular and cortical analysis were equivalent to 0.30 g/mm 3 and 0.64 488 g/mm 3 CaHA, respectively. For 26 week old mice, the lower adaptive threshold for 489 trabecular and cortical analysis were equivalent to 0.33 g/mm 3 and 0.76 g/mm 3 CaHA, 490 respectively. For trabecular analysis in the 5 th lumbar vertebrae (L5), an ROI of half the 491 height of the bone (vertically centered) with a diameter 2/3 the width of the vertebral 492 body was analysed. 493

Histomorphometry 494
Tibiae from 14 week old mice were embedded in methylmethacrylate and sectioned at 5 495 μm thickness for histomorphometric analysis, as previously described (62). The observer 496 was blinded to genotype and sex of all samples during analysis. To determine bone 497 formation rates, calcein was injected intraperitoneally (20 mg/kg) at 7 and 2 days before 498 tissue collection. Sections were stained with Toluidine blue or Xylenol orange, as 499 described (63). Static and dynamic histomorphometry of trabecular bone surfaces was 500 carried out in the secondary spongiosa of the proximal tibia using the OsteoMeasure 501 system (Osteometrics Inc., USA). 502

Three-point bending test 503
Mechanical properties of femora were derived from three-point bending tests using a 504 Bose Biodynamic 5500 Test Instrument (Bose, DE, USA), as described previously (64). 505 The observer was blinded to genotype and sex of all samples during analysis. Once whole-506 bone properties were determined, tissue-level mechanical properties were calculated 507 using micro-CT analysis of the mid-shaft (1). 508

Biochemical assays 509
Cross-linked C-telopeptides of type I collagen (CTX-1) were measured in duplicate with 510 the IDS RatLaps enzyme immunoassay (Abacus, Berkeley, CA, USA) in serum collected 511 from mice fasted overnight. Serum levels of procollagen type 1 N propeptide (P1NP) were 512 measured in duplicate using IDS Rat/Mouse PINP EIA kit (Abacus, Berkeley, CA, USA). 513 To measure milk PTHrP content, amino-terminal PTHrP RIA was carried out as 514 previously described, with a sensitivity of 2 pM (65). Milk was diluted 1:500 in assay 515 buffer prior to measurement of PTHrP. Milk PTHrP levels were also bioassayed as the 516 cAMP generated in response to treatment of UMR106-01 cells, using PTH(1-34)-induced 517 cAMP response as a standard curve (66). Replicate cell cultures in 24-well plates were 518 incubated in cell culture medium with 1 mM isobutylmethylxanthine (IBMX) added. After 519 treatment for 12 mins with 1:8 diluted milk samples, cAMP was measured by removing 520 medium and adding acidified ethanol, drying, reconstituting in assay buffer and cAMP 521 assay as described (67). cAMP was then corrected for total protein content of the milk, 522 measured by Pierce BCA protein assay kit (Thermo Fisher Scientific). For this, milk was 523 diluted 1:400 in PBS and absorbance was measured at OD562nm using the Polarstar 524 Optima+ and a bovine serum albumin standard curve. 525 Embryo skeletal staining 526 Alcian blue and Alizarin red S staining was carried out on E18.5 embryos, as described 527 previously (68). Embryos were fixed in 95% ethanol for 5 days after skin removal. 528 Remnant skin and viscera were dissected as much as possible, followed by defatting in 529 acetone for 2 days. Thereafter, they were stained for 4 days at 40 °C in freshly prepared 530 staining solution: 0.3% Alcian blue in 70% ethanol -1 volume; 0.1% alizarin red S in 95% 531 ethanol -1volume; glacial acetic acid -1 volume; 70% ethanol -17 volumes. After washing 532 in distilled water for 2 hours, they were cleared with 2% potassium hydroxide (KOH) for 533 2 days. Afterwards, they were put in 20% Glycerol in 1% KOH until skeletons were clearly 534 visible, then successively placed into 50%, 80% and 100% glycerol solutions in 1% KOH 535 for 2 days each. Femoral length was determined by measuring the distance between 536 femoral head and distal end through a dissecting microscope, and an average of right and 537 left femur lengths in each embryo was calculated. 538

Immunohistochemistry 539
Immunohistochemistry was carried out as described previously (69, 70) on paraffin-540 embedded placenta/decidua (collected at E17.5) using goat rabbit anti-PTHrP (1:1000, 541 R87, generated against PTHrP(1-14) (71). The observer was blinded to genotype and sex 542 of all samples during analysis. Placental/decidual samples were fixed overnight in 4% 543 paraformaldehyde at 4°C, stored in 70% ethanol, and embedded in paraffin wax until 544 further analysis. Sections (5μm) were taken onto chrome alum-coated slides, dewaxed in 545 Histoclear (National Diagnostics, Atlanta, GA), and rehydrated in graded ethanols. 546 Endogenous peroxidase was blocked for 30 min in 2% H2O2 in methanol.   and range; n=9-10/group. No significant differences associated with genotype were 825 detected (two-way ANOVA). Y-axes are drawn to match those of Figure 1 to allow 826 comparison with Dmp1Cre.Pthlh f/f(hom) mice. Breeding strategy is shown in Figure 1A. 827 range; n=9-10/group. *p<0.05, and **p<0.01 compared to sex-and age-matched 836 Dmp1Cre (het) by two-way ANOVA. Y-axes are drawn to match those of Figure 2 to allow 837 comparison with Dmp1Cre.Pthlh f/f(hom) mice. Breeding strategy is shown in Figure 1A. 838 as mean (dot), interquartile range (box), median (line) and range, n=9-10/group. No 843 significant differences relating to genotype were detected by two-way ANOVA. 844 proximal to the distal end of the mineralized region (Bottom 20%); see Figure 5A. Data 849 is shown as mean ± SEM with individual data points, *p<0.05 and **p<0.01 compared to 850 controls by two-way ANOVA. 851 Dmp1Cre.Pthlh f/f (het) and Dmp1Cre (het) embryos measured in three different regions of 855 femora: 20% of the mineralized length distal to the proximal end of the mineralized 856 region (Top 20%), midshaft (Mid), and 20% of the mineralized length proximal to the 857 distal end of the mineralized region (Bottom 20%). Data is shown as mean ± SEM with 858 individual data points. No significant differences relating to genotype were detected by 859 two-way ANOVA. 860 Frequency of PTHrP stained objects segregated by staining intensity in the 864 spongiotrophoblast layer and decidua from placental/decidual samples from 865 Dmp1Cre.Pthlh f/f(het) and Dmp1Cre (het) embryos. (F,G) Quantitation of total decidual area 866 and spongiotrophoblast area; mean ± SEM with individual data points. No significant 867 differences relating to genotype were detected by two-way ANOVA.