Murine Oncostatin M Acts via Leukemia Inhibitory Factor Receptor to Phosphorylate Signal Transducer and Activator of Transcription 3 (STAT3) but Not STAT1, an Effect That Protects Bone Mass

Abstract

Oncostatin M (OSM) and leukemia inhibitory factor (LIF) are IL-6 family members with a wide range of biological functions. Human OSM (hOSM) and murine LIF (mLIF) act in mouse cells via a LIF receptor (LIFR)-glycoprotein 130 (gp130) heterodimer. In contrast, murine OSM (mOSM) signals mainly via an OSM receptor (OSMR)-gp130 heterodimer and binds with only very low affinity to mLIFR. hOSM and mLIF stimulate bone remodeling by both reducing osteocytic sclerostin and up-regulating the pro-osteoclastic factor receptor activator of NF-κB ligand (RANKL) in osteoblasts. In the absence of OSMR, mOSM still strongly suppressed sclerostin and stimulated bone formation but did not induce RANKL, suggesting that intracellular signaling activated by the low affinity interaction of mOSM with mLIFR is different from the downstream effects when mLIF or hOSM interacts with the same receptor. Both STAT1 and STAT3 were activated by mOSM in wild type cells or by mLIF/hOSM in wild type and Osmr^−/− cells. In contrast, in Osmr^−/− primary osteocyte-like cells stimulated with mOSM (therefore acting through mLIFR), microarray expression profiling and Western blotting analysis identified preferential phosphorylation of STAT3 and induction of its target genes but not of STAT1 and its target genes; this correlated with reduced phosphorylation of both gp130 and LIFR. In a mouse model of spontaneous osteopenia caused by hyperactivation of STAT1/3 signaling downstream of gp130 (gp130^Y757F/Y757F), STAT1 deletion rescued the osteopenic phenotype, indicating a beneficial effect of promoting STAT3 signaling over STAT1 downstream of gp130 in this low bone mass condition, and this may have therapeutic value.

Introduction

Oncostatin M (OSM)² is a member of the IL-6 cytokine superfamily with a diverse range of activities (1). It was first identified as a secreted product of macrophage-like cells that inhibited proliferation of melanoma-, neuroblastoma-, and lung cancer-derived cell lines (2), and although it suppresses breast cancer cell proliferation (3), it has been suggested to support metastasis of breast cancer in the skeleton due to pro-osteoclastic activities (4) and to stimulate Kaposi's sarcoma (5). Studies in knock-out mice have shown that OSM supports hematopoiesis (6) and bone formation at the expense of adipogenesis (7), but it has also been implicated in pulmonary tissue fibrosis (8), cardiac disease and repair (9), prostate cancer (10), asthma (11), periodontal disease (12), and both rheumatoid and osteoarthritis (13).

The diverse roles of OSM may be explained by its ability to act through two distinct signaling complexes. Human OSM (hOSM) signals by binding first to the ubiquitously expressed glycoprotein 130 co-receptor α subunit (gp130) and then recruiting, with equal affinity, either the hOSM receptor (hOSMR) or the human leukemia inhibitory factor receptor (hLIFR) β subunits (14), which may then activate distinct intracellular signaling pathways (1). Murine OSM (mOSM) behaves differently with the mOSM-gp130 complex binding with high affinity to mOSMR and very low affinity to mLIFR (15).

In the skeleton, OSM is a paracrine factor expressed by osteoblasts (7), osteocytes (7), and macrophages (16–19). OSM stimulates the formation of bone-resorbing osteoclasts indirectly by acting on osteoblasts (20) to promote their expression of receptor activator of NF-κB ligand (RANKL) (21). OSM also promotes osteoblast differentiation at the expense of adipogenesis (7, 22) and stimulates bone formation in vivo (7) at least in part by acting on osteocytes (23) (matrix-embedded osteoblasts) and inhibiting their production of the Wnt antagonist sclerostin (7, 23). These actions mediated by OSMR are required for normal levels of bone formation and resorption in juvenile and adult skeletons (7) and for normal skeletal response to parathyroid hormone (24), a therapeutic agent used to reduce fracture risk in osteoporosis. Surprisingly, in the absence of OSMR, mOSM can signal through LIFR to suppress sclerostin expression and stimulate bone formation (7). Unlike LIF and hOSM, when mOSM acted through mLIFR, it did not significantly stimulate RANKL production (7). This suggested a novel pathway that could increase bone mass by stimulating bone formation without promoting bone resorption.

The mechanism by which mOSM-mLIFR activates downstream signals different from those activated by canonical murine LIF (mLIF)-mLIFR signaling is unknown. This study sought to (a) identify unique downstream signaling pathways of mOSM-mLIFR compared with hOSM-mLIFR and mLIF-mLIFR, (b) determine mechanisms by which such distinct signaling pathways could be activated, and (c) ascertain the benefit of promoting the mOSM-mLIFR-specific pathway in low bone mass.

Results

Identification of Downstream Gene Signatures by Microarray

Microarray analysis indicated that the expression changes caused by mLIF and hOSM in Osmr^−/− primary calvarial osteoblasts (defined as mLIF-mLIFR effects and hOSM-mLIFR effects, respectively) were very highly correlated (Fig. 1A), supporting previous observations that hOSM acts entirely through LIFR in murine cells (25). A similar correlation between mLIF and hOSM responses was observed in WT (data not shown). For this reason, mLIF-mLIFR and hOSM-mLIFR responses were averaged for all subsequent analyses.

FIGURE 1.

Identification of the distinct mOSM-LIFR gene signature. A, scatterplot of log -fold changes of gene expression in 17-day differentiated Osmr^−/− primary calvarial osteoblasts treated with mLIF or hOSM for 6 h. The genas biological correlation between the log -fold change profiles is 99.7%. B, scatterplot of log₂ -fold expression changes at 1 h in 17-day-differentiated Osmr^−/− primary calvarial osteoblasts stimulated by mOSM versus control (y axis) and by the mean of hOSM and mLIF versus control (x axis). Genes highlighted in cyan are highly up-regulated by both mOSM and hOSM/mLIF. The black dashed diagonal line shows the least square regression through the origin of y on x with slope 0.367. The cyan dashed line is chosen to have slope 0.275, which is 75% of the least square slope. Gene symbols are shown for selected genes with large -fold changes. C, scatterplot of log₂ -fold expression changes at 6 h in Osmr^−/− osteoblasts. Genes highlighted in cyan are highly up-regulated by both mOSM and hOSM/mLIF. The black dashed diagonal line shows the least square regression through the origin of y on x with slope 0.4. The cyan dashed line is chosen to have slope 0.3, which is 75% of the least square slope. Gene symbols are shown for selected genes with large -fold changes. D and E, real time PCR validation on three independent cultures of wild type and Osmr^−/− cells treated with mOSM, hOSM, or mLIF showing Socs1 (at 1 h), Oasl2 (at 1 h), and Gbp2 (at 6 h) regulated by hOSM, mLIF, and mOSM in wild type cells (black bars), but not by mOSM in Osmr^−/− cells (white bars) (D). Also shown are regulation of Socs3 (at 1 h), Bcl3 (at 6 h), and Sbno2 (at 6 h) by hOSM, mLIFR, and mOSM in Osmr^−/− cells (white bars) (E). Values are means from three independent cultures, each performed in triplicate; error bars represent S.E. F, Venn diagram showing number of genes (bold text) and selected targets differentially expressed between mLIF/hOSM-treated and control in wild type cells (mLIF-mLIFR/hOSM-mLIFR) versus those regulated by mOSM treatment in wild type cells but not regulated by mOSM in Osmr^−/− cells (mOSM-mOSMR). The subset of (hOSM/mLIF)-LIFR genes highly up-regulated by mOSM in Osmr^−/− cells (mOSM-mLIFR) identified by the analysis in B and C are highlighted in the circle outlined with a dashed line. Ctrl, control.

No unique set of genes was differentially expressed in Osmr^−/− cells treated with mOSM (i.e. mOSM-mLIFR targets) compared with mLIF- and hOSM-treated WT or Osmr^−/− cells. Instead, mOSM-LIFR-regulated target genes formed a subset of all those regulated by the canonical pathway ((mLIF/hOSM)-LIFR targets; the slash represents a possible substitution (either hOSM or mLIF)). To identify the distinctive subset of gene activation retained by mOSM in Osmr^−/− cells (i.e. the genes regulated by mOSM-mLIFR), we first determined the positive targets of canonical LIFR signaling with a strong mOSM response by selecting genes up-regulated at least 2-fold by hOSM/mLIF in Osmr^−/− cells with an mOSM -fold change above the slope of the line of regression between the differential expression of hOSM/mLIF versus control and mOSM versus control in Osmr-null cells. This allowed us to account for the difference in magnitude of effect between hOSM/mLIF and mOSM in Osmr^−/− cells. This revealed that STAT3-responsive genes were up-regulated by mOSM in Osmr^−/− cells at both 1 and 6 h (Fig. 1, B and C, and Table 1). At 1 h of treatment, eight of the top 10 genes disproportionately up-regulated by mOSM-LIFR were Socs3 (26), Cxcl1 (27), Zfp36 (28), Mt2 (29), Sbno2 (30), Bcl3 (31), Junb (32), and Fos (33) (Fig. 1B). The only “non-STAT3 targets” in this top 10 were Cish, a STAT5-responsive gene (34), and Rasl11a, a putative small Ras GTPase (35). At 6 h of treatment, the five most strongly regulated probes, Bcl3, Mmp13 (36), Socs3, and Mt2 (Fig. 1C and Table 1), were also STAT3 targets. We confirmed strong enrichment of STAT3 target genes in those retained by mOSM signaling in Osmr^−/− cells by comparison with the available STAT3 gene sets from the Molecular Signatures Database (MSigDB); significant enrichment was observed from a number of STAT3 target data sets at 1 h, including DAUER_STAT3_ TARGETS_UP (five of 66 genes, p = 3.5E−14), ST_STAT3_PATHWAY (two of 15 genes, p = 7.7E−07), and BAKER_HEMATOPOIESIS_STAT3_TARGETS (two of 16 genes, p = 8.8E−07), and at 6 h, including DAUER_STAT3_TARGETS_UP (four of 66 genes, p = 8.9E−10) and NUMATA_CSF3_SIGNALING_ VIA_STAT3 (two of 23 genes, p = 9.1E−06).

TABLE 1.

Top genes up-regulated by mOSM-LIFR

The top 10 probes highlighted in cyan in Fig. 1, C and D, are ordered by magnitude of response to mOSM in Osmr^−/− osteoblasts after 1 and 6 h of treatment, respectively (only nine genes were identified at 6 h).

Time	Rank	Gene name	Probe ID	log2FC			Ratio log2FC mOSM:mLIF/hOSM
				mOSM	mLIF	hOSM
1 h	1	Socs3	1570594	2.523	4.013	4.107	0.62
	2	Cish	4200017	1.38	2.821	2.913	0.48
	3	Cxcl1	3610882	1.273	3.563	3.312	0.37
	4	Zfp36	6220026	1.239	3.084	3.033	0.41
	5	Mt2	2190196	1.225	3.084	3.114	0.40
	6	Sbno2	3360086	1.116	2.247	2.056	0.52
	7	Rasl11a	2850521	1.107	1.468	1.908	0.66
	8	Bcl3	4070020	1.066	1.588	1.758	0.64
	9	Junb	3120014	1.043	2.348	2.241	0.45
	10	Fos	510368	1.04	2.738	2.603	0.39
6 h	1	Bcl3	4070020	1.827	2.615	2.662	0.69
	2	Mmp13	5690131	1.822	2.294	2.66	0.74
	3	Socs3	1570594	1.764	3.178	3.188	0.55
	4	Mmp13	2490131	1.727	1.938	2.441	0.79
	5	Mt2	2190196	1.694	4.576	4.031	0.39
	6	Artn	240608	1.555	2.671	2.522	0.60
	7	Casp4	5290017	1.551	3.255	3.236	0.48
	8	Lrg1	290097	1.495	3.462	3.294	0.44
	9	Tnfsf11	2480255	1.473	2.605	2.603	0.57

In contrast, in the top 10 genes most highly regulated by the canonical (hOSM/mLIF)-LIFR pathway that were not regulated by mOSM-LIFR were known STAT1/IFNγ-responsive genes (Fig. 1, B and C, and Table 2). At 1 h, the top nine known genes regulated were all STAT1/IFNγ targets: Mx2 (37), Irgm1 (38), Ccl2 (39), Serpina3f (40), Socs1 (41), Cxcl10 (39), Ifi47 (42), Trim35 (42), and Cxcl9 (43). At 6 h, the top 10 known probes were all STAT1/IFNγ-responsive genes: Gbp2, Gbp3 (44), Gbp5 (45), Cxcl9 (43), Oasl2 (46), Serpina3f (40), Igtp (47), Psmb9 (48), Dhx58 (49), and Oas1g (49). A subset of genes was confirmed by quantitative real time PCR (qPCR) in a further three independent sets of cultures from Osmr^−/− and WT cells (Fig. 1, D and E). This indicates that the interaction of mOSM with LIFR, unlike canonical mLIF/LIFR signaling, did not induce STAT1-responsive genes but retained the ability to induce STAT3 target genes, albeit at a lower level.

TABLE 2.

Top probes regulated by (hOSM/mLIF)-LIFR but lost by mOSM-LIFR

The top probes ordered by mLIF -fold change lying below the cyan dashed lines in Fig. 1, C and D, respectively, are shown. At 1 h, the selected genes have log₂FC response to mOSM <27.5% of that in hOSM/mLIF. At 6 h, the selected genes have log₂FC response to mOSM <30% of that in hOSM/mLIF.

Time	Rank	Gene name	Probe ID	log2FC			Ratio log2FC mOSM:mLIF/hOSM	p value (mOSM vs. mLIF/hOSM)
				mOSM	mLIF	hOSM
1 h	1	Mx2	520278	0.112	3.445	3.221	0.03	9.00E − 10
	2	Irgm1	5820608	0.191	2.995	2.904	0.06	2.36E − 11
	3	Ccl2	110112	0.331	2.933	3.143	0.11	8.08E − 08
	4	Serpina3f	4540082	0.483	3.163	3.186	0.15	2.17E − 07
	5	Socs1	5670497	0.271	2.749	2.702	0.10	1.77E − 12
	6	Cxcl10	3140209	−0.182	2.440	1.967	−0.08	8.07E − 06
	7	Ifi47	1090139	0.331	2.617	2.508	0.13	2.28E − 09
	8	Trim35	5560762	0.055	2.425	2.093	0.02	6.85E − 11
	9	5530401N12Rik	3190575	0.209	2.347	2.137	0.09	6.51E − 11
	10	Irgm1	7380221	−0.004	1.748	2.148	0.00	6.28E − 05
	11	Cxcl9	10598	0.270	2.182	2.129	0.13	5.05E − 09
6 h	1	Gbp2	6860600	0.219	3.577	3.713	0.06	7.37E − 11
	2	Gbp2	1400041	0.295	3.590	3.511	0.08	5.00E − 11
	3	Cxcl9	10598	0.250	3.408	2.992	0.08	2.61E − 14
	4	Oasl2	4830010	0.209	3.227	3.012	0.07	3.71E − 15
	5	Serpina3f	4540082	−0.010	2.623	2.844	0.00	1.47E − 07
	6	AA467197	1690475	0.403	3.183	3.100	0.13	6.59E − 08
	7	Gbp3	1690475	0.192	2.910	2.720	0.07	5.47E − 15
	8	Gbp3	60553	0.257	2.839	2.697	0.09	1.33E − 11
	9	Igtp	6290037	−0.032	2.572	2.371	−0.01	1.58E − 09
	10	Psmb9	2450064	−0.399	2.040	1.972	−0.20	1.07E − 16
	11	Dhx58	6330592	−0.541	2.016	1.497	−0.31	1.40E − 12
	12	Gbp5	5900520	0.187	2.518	2.329	0.08	2.18E − 07
	13	Oasl2	7050014	−0.074	2.219	1.851	−0.04	8.33E − 12
	14	D14Ertd668e	520053	0.302	2.395	2.39	−2.09	2.97E − 11
	15	Gbp2	2100068	0.046	2.206	1.826	−1.97	5.65E − 07
	16	Oas1g	3890328	−0.281	1.917	1.438	−1.96	1.65E − 09

Tnfsf11 (the gene for RANKL) was significantly stimulated downstream of mOSM-LIFR in contrast to our earlier PCR-based analysis. As we reported previously, Tnfsf11 was not up-regulated by mOSM in Osmr^−/− cells at 1 h after treatment (7), but at 6 h after treatment Tnfsf11 was up-regulated by mOSM 2.8-fold. This response was 9.6-fold lower than the response observed to mOSM in WT cells (p = 2.43 E-14) and less than the response in Osmr^−/− cells treated with hOSM/mLIF (Tnfsf11 was increased 6.1-fold for both hOSM and mLIF).

Our hypothesis was that osteoblastic genes promoting osteoblast differentiation would be regulated through mOSM-LIFR, whereas genes supporting osteoclast formation (catabolic) would be regulated by mOSM through OSMR. To identify those genes regulated by only mOSM through OSMR (mOSM-OSMR) and not by mOSM through LIFR, we determined genes differentially expressed in WT cells treated with mOSM versus control that were not differentially expressed in Osmr^−/− cells treated with mOSM versus control; mOSM-OSMR targets numbered 2669 genes (supplemental Table 1). This gene set was compared with those differentially expressed in mLIF/hOSM-treated WT cells (mLIF/hOSM:LIFR targets) which identified 172 regulated genes in common between mLIF/hOSM:LIFR and mOSM:OSMR, and 2497 genes unique for mOSM:OSMR (Fig. 1F, supplemental Table 1) were identified. mOSM-OSMR target genes included some previously identified to mediate actions of OSM in cartilage degradation (Adamts4) and osteoclastogenesis (Il33) (50, 51) (rank 47, log₂ -fold change (log₂FC) 2.96, p = 9.17E−13). Although previously we could detect no osteoblastic response to IL-31 (7, 24), IL-31α receptor (Il31ra) mRNA was detected and found to be elevated by mOSM-OSMR.

The comparison between target genes revealed that mOSM-OSMR does not stimulate a purely catabolic gene set. Some genes known to suppress osteoblast differentiation and promote adipogenesis, including Zfp521 and Zfp467 (52–54), were down-regulated through mOSM in wild type cells (rank, log₂FC, and p values: Zfp521, 69, −1.6, p = 2.77E−13; Zfp467, 602, −1.65, p = 8.39E−8). Indeed, genes supportive of osteoblast commitment and differentiation, including Sost, Sp7, Ifitm5, Zfp521, Zfp467, Mef2c, and Cebpd, were regulated through multiple pathways as were genes supportive of osteoclast differentiation or precursor migration, such as Tnfsf11, Cxcl1, Wnt16, and Il33 (Fig. 1F). This indicated that there is no purely anabolic gene response to mOSM-LIFR compared with (hOSM/mLIF)-LIFR or mOSM-mOSMR.

Confirmation of LIFR-dependent STAT3 Induction

Western blotting analysis showed that mOSM, hOSM, and mLIF treatment of WT primary calvarial osteoblasts stimulated STAT1 and STAT3 phosphorylation (Fig. 2, A and B). Although Osmr^−/− cells responded to hOSM and mLIF with phosphorylation of the lower molecular weight form of STAT1, no STAT1 phosphorylation was detected in response to mOSM (Fig. 2A). STAT3 phosphorylation at tyrosine 705 was robustly stimulated by mOSM, hOSM, and mLIF in both WT cells and Osmr^−/− cells at 15 and 30 min (15 min shown in Fig. 2B), indicating an intact STAT3 phosphorylation response in the absence of OSMR. To confirm that the mOSM-induced increase in STAT3 phosphorylation was mediated by LIFR, we tested the response when cells were pretreated with a specific LIFR antagonist (LA) (55). This blocked the induction of STAT3 phosphorylation by mOSM, hOSM, and mLIF in Osmr^−/− cells (Fig. 2C), showing that STAT3 phosphorylation in the absence of OSMR was mediated by LIFR. In wild type cells, LA blocked the effects of hOSM and mLIF but not mOSM, indicating that mOSM acts preferentially through OSMR.

FIGURE 2.

mOSM-LIFR signaling stimulates STAT3, but not STAT1, phosphorylation. A and B, Western blots showing STAT1 Tyr-701 (A) and STAT3 Tyr-705 (B) phosphorylation compared with total STAT1, total STAT3, and pan-actin 15 min after treatment with medium alone, mOSM, hOSM, and mLIF (50 ng/ml) in WT and Osmr^−/− primary calvarial osteoblasts. C, Western blots showing the effect of LIFR antagonist pretreatment (2.5 ng/ml, 30 min) on STAT3 (Tyr-705) phosphorylation in WT and Osmr^−/− primary calvarial osteoblasts treated with mOSM, hOSM, or mLIF (50 ng/ml) for 15 min. Western blots are representative of three independent experiments.

Use of a specific STAT3 inhibitor (ML116) (56) validated that the Socs3 gene response to mOSM-LIFR was STAT3-dependent (Fig. 3, A and B). In WT cells, Socs3 mRNA levels induced by mOSM and mLIF were inhibited only by the highest dose of ML116 (10 μg/ml), whereas the increased Socs3 mRNA levels in Osmr^−/− cells treated with mOSM were completely blocked by even the lowest dose of ML116 (Fig. 3A). Phosphorylation of STAT3 in response to mOSM in Osmr^−/− cells was reduced by ML116 treatment as was the response to mLIF in both wild type and Osmr^−/− cells (Fig. 3, B and C).

FIGURE 3.

mOSM induction of Socs3 is mediated by STAT3 activation. Socs3 mRNA levels (A) and STAT3 phosphorylation (Tyr-705) (B and C) in wild type and Osmr^−/− primary calvarial osteoblasts differentiated for 17 days and treated with mOSM (50 ng/ml) or mLIF (50 ng/ml) for 1 h (A) or 15 min (B) in the absence and presence of the STAT3 inhibitor ML116 (0.1, 1, and 10 μg/ml) are shown. PCR data are means of three independent biological replicates, each performed in triplicate; error bars represent S.E. (effect of cytokine: *, p < 0.05; **, p < 0.01 versus untreated cells of the same genotype; effect of ML116: +, p < 0.05; +++, p < 0.001 versus cells of the same genotype treated with the same cytokine but not ML116). The Western blot (B) is representative of three independent experiments, which were each measured relative to total STAT3 levels and normalized to the mOSM response observed in each experiment and shown as means; error bars represent S.E. (C).

Because STAT3 phosphorylation at serine 727 and tyrosine 705 may result in different downstream activities (57), we sought to determine whether mOSM-LIFR influences only a subset of mLIF-LIFR-responsive genes because it phosphorylates only tyrosine 705 (Tyr-705) and not serine 727 (Ser-727). Phosflow analysis indicated STAT3 phosphorylation occured at both Tyr-705 and Ser-727 in response to mOSM in Osmr^−/− primary calvarial osteoblasts (Fig. 4). Levels of phosphorylation were also comparable with those induced by mLIF and hOSM in Osmr^−/− and WT primary calvarial osteoblasts (Fig. 4). This suggests that the difference in genes transcribed in response to mOSM-LIFR versus mLIF-LIFR does not relate to a difference in STAT3 phosphorylation patterns.

FIGURE 4.

mOSM phosphorylates STAT3 on Tyr-705 and Ser-727 downstream of the LIFR. Phosphorylation of STAT3 Tyr-705 and Ser-727 in 22–24-day WT and Osmr^−/− primary calvarial osteoblasts is shown. A and B, histograms depicting a shift in MFI when WT (A) or Osmr^−/− (B) cells were stained for AF488 phospho-STAT3 Tyr-705, AF647 phospho-STAT3 Ser-727, and phycoerythrin total STAT3 protein. C and E, quantification of MFI for phospho-STAT3 Tyr-705 (C), phospho-STAT3 Ser-727 (D), and total STAT3 (E) in 22–24-day-differentiated osteoblasts after 15-min stimulation with mOSM, hOSM, or mLIF (50 ng/ml). Graphed data are mean -fold change versus untreated of three independent experiments; error bars represent S.E. Two-way ANOVA indicated significant treatment effects for all three cytokines at both phosphorylation sites but no significant interaction between treatment and genotype of the cells. Ctrl, control.

To determine whether differences in gp130 and LIFR phosphorylation could be responsible for the bias toward STAT3 signaling of mOSM-LIFR versus mLIF-LIFR, we assessed phosphorylation of these receptor subunits in response to each ligand (Fig. 5, A–C). In WT cells, mOSM treatment resulted in robust OSMR and gp130 phosphorylation (Fig. 5, A and B). OSMR was not phosphorylated in response to mLIF (Fig. 5B) and was not detected in Osmr^−/− cells as expected (Fig. 5B). LIFR phosphorylation was robustly detected in response to mLIF in WT and Osmr^−/− cells (Fig. 5C). Although mLIF and hOSM both induced phosphorylation of gp130 in Osmr^−/− cells, gp130 phosphorylation was barely detected in response to mOSM and in some experiments could not be detected at all (Fig. 5A). In addition, LIFR phosphorylation in response to mOSM in Osmr^−/− cells was also only barely detected (Fig. 5C). This suggests that, in Osmr^−/− cells, although mLIF induces phosphorylation of both gp130 and LIFR, resulting in phosphorylation of both STAT3 and STAT1, the ability of mOSM to activate STAT3 without STAT1 signaling may relate to reduced phosphorylation of gp130 and LIFR (Fig. 5D).

FIGURE 5.

mOSM phosphorylation of gp130 and LIFR is minimal when signaling via LIFR. A–C, immunoprecipitation (IP) and detection of phosphorylated gp130 (A), OSMR (B), and LIFR (C) in response to 50 ng/ml mOSM and mLIF in murine WT and Osmr^−/− cells 15 min after commencement of treatment. Micrographs are representative of three independent experiments; the lower set of bands in the OSMR blot represents nonspecific binding to show the presence of protein in lysates from Osmr^−/− cells. D, model for receptor and STAT phosphorylation in OSMR- and LIFR-mediated activities of mOSM and mLIF in WT and Osmr^−/− cells.

Assessing Possible Therapeutic Benefit in the Skeleton

To assess whether shifting the balance of signaling toward STAT3 rather than STAT1 might promote bone formation, we analyzed the bone phenotype of 12-week-old male global Stat1^−/− mice but observed no significant difference in trabecular bone volume (Fig. 6A), trabecular thickness (Fig. 6B), trabecular number (Fig. 6C), or trabecular separation (Fig. 6D). A significant increase in periosteal circumference without any change in femoral length was observed, suggesting elevated periosteal bone formation when STAT3 signaling is greater than STAT1 (mean periosteal circumference ±S.E.: WT, 7.50 ± 0.14; Stat1^−/−, 8.05 ± 0.12; p = 0.0042; mean femoral length ±S.E.: WT, 13.33 ± 0.13; Stat1^−/−, 13.48 ± 0.09).

FIGURE 6.

Reducing STAT1 relative to STAT3 signaling rescues gp130-dependent osteopenia. A–E, trabecular bone volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), and trabecular separation (Tb.Sp) measured in the distal femur of 12-week-old male wild type (wt) and Stat1^−/− mice (n = 9–10). E–H, trabecular bone volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), and trabecular separation (Tb.Sp) measured in the proximal tibiae of 12–14-week-old wild type (wt), gp130^Y757F/Y757F, gp130^Y757F/Y757F/Stat1^−/−, and gp130^Y757F/Y757F/Stat3^+/− mice (n = 4–6/group). Values are means; error bars represent S.E. (*, p < 0.05; **, p < 0.01; ***, p < 0.001 versus wild type; +, p < 0.05, +++, p < 0.001 versus gp130^Y757F/Y757F). I, representative Von Kossa stained proximal tibial sections from wild type, gp130^Y757F/Y757F, gp130^Y757F/Y757F/Stat1^−/−, and gp130^Y757F/Y757F/Stat3^+/− mice. Black, bone; scale bar, 200 μm.

To further explore the implications of biased STAT3 signaling in a disease setting, we used a previously described mouse model of osteopenia caused by hyperactivation of STAT1/3 signaling downstream of gp130 by deletion of the binding site for the negative regulators SOCS3 and SHP2 (gp130^Y757F/Y757F) (58). When these mice were crossed with Stat1^−/− mice so that only STAT3 signaling downstream of gp130 was hyperactivated (36), the gp130^Y757F/Y757F osteopenic phenotype was rescued; trabecular bone volume, trabecular number, trabecular thickness, and trabecular separation all returned to normal (Fig. 6, E–I). This was not reproduced when gp130^Y757F/Y757F mice were crossed with Stat3^+/− mice, a cross that normalizes STAT3 signaling downstream of gp130 (36). Osteoclast and osteoblast numbers remained significantly elevated in gp130^Y757F/Y757F/Stat1^−/− mice but not in gp130^Y757F/Y757F/Stat3^+/−. Numbers of osteoblasts/bone perimeter (mean ± S.E.) were as follows: wild type, 9.1 ± 1.7; gp130^Y757F/Y757F, 24.0 ± 0.7*; gp130^Y757F/Y757F/Stat1^−/−, 29.3 ± 4.1*; gp130^Y757F/Y757F/Stat3^+/−, 11.9 ± 5.7 (*, p < 0.05). Numbers of osteoclasts/bone perimeter (mean ± S.E.) were as follows: wild type, 2.23 ± 0.45; gp130^Y757F/Y757F, 3.97 ± 0.37*; gp130^Y757F/Y757F/Stat1^−/−, 4.31 ± 0.71*; gp130^Y757F/Y757F/Stat3^+/−, 2.38 ± 1.10 (*, p < 0.05). These results indicate a beneficial effect of promoting STAT3 signaling over STAT1 downstream of gp130 in this low bone mass condition.

Discussion

In this study, we report a full set of gene responses to mOSM in the absence of OSMR indicating that mOSM, which we have previously shown to promote bone formation independently of OSMR (7), regulates multiple gene targets downstream of mLIFR. Unlike canonical signaling of mLIF and hOSM where both STAT1 and STAT3 are phosphorylated downstream of mLIFR, mOSM induces phosphorylation of STAT3 but not STAT1 and stimulates a STAT3-responsive subset of genes downstream of mLIFR. This may be explained by formation of a signaling complex with very low levels of gp130 and LIFR phosphorylation (Fig. 5B). Supporting an anabolic action of this pathway, favoring STAT3 over STAT1 signaling downstream of gp130 protected against osteopenia in a mouse with hyper-responsiveness to IL-6 family cytokines (gp130^Y575F/Y757F). This suggests such a pathway could be exploited in skeletal pathologies where IL-6 family signaling is elevated, such as inflammatory or metastatic bone loss.

Reduced trabecular bone volume due to hyperactivation of STAT1/3 signaling downstream of gp130 in the gp130^Y575F/Y757F mice has been reported previously to result from a high level of bone remodeling (58). Although both bone resorption and bone formation are elevated, the level of resorption outstrips that of formation, resulting in an osteopenic phenotype, a similar etiology to bone loss associated with estrogen deficiency (60). Elevated signaling by gp130-dependent cytokines has been postulated not only to cause osteoporosis after menopause (61) but also to play a role in focal and systemic bone loss that results from colitis (62), inflammatory arthritis (63, 64), multiple myeloma (65), Gorham-Stout disease (66), Paget's disease (67), and breast cancer metastasis to bone (68). The osteopenic phenotype of gp130^Y575F/Y757F mice was rescued by blocking STAT1 signaling. Although bone remodeling remained high, the balance was shifted to favor bone formation, and bone mass was protected. STAT1 signaling blockade may therefore provide a therapeutic benefit in patients with bone loss related to elevated IL-6 family cytokine signaling. The greater level of periosteal bone formation in Stat1^−/− mice may relate to the loss of STAT1-mediated suppression of periosteal development; STAT1 is expressed at high levels in the developing perichondrium (69). The limited benefit of STAT1 deletion without hyperactivation of STAT3 (i.e. only periosteal bone formation was increased in Stat1^−/− mice) suggests that this approach might not benefit all forms of bone fragility.

When IL-6 family cytokines form a receptor-ligand complex, it is phosphorylation of the intracellular domain of the receptor subunits, both gp130 (70, 71) and LIFR (72), that provides docking sites for STAT proteins. These are subsequently phosphorylated by JAKs on specific residues allowing homo- or heterodimerization and translocation to the nucleus to modify transcription. This study identifies that the low affinity interaction of mOSM with mLIFR results in transcription of a subset of genes downstream of canonical (mLIF/hOSM)-mLIFR signaling. These are known STAT3-responsive genes, including Socs3 (26), Bcl3 (31), Junb (32), Zfp36 (28), Rgs16 (28), Cxcl1 (27), Mt2 (29), Sbno2 (30), Mmp13 (36), and c-fos (33). In contrast, the mLIF/hOSM-responsive genes that did not respond to mOSM in the absence of OSMR included STAT1 target genes: Mx2 (37), Irgm1 (38), Ccl2 (39), Serpina3f (40), Socs1 (41), Cxcl10 (39), Ifi47 (42), Trim35 (42) Cxcl9 (43), Gbp3 (44), Gbp5 (45), Cxcl9 (43), Oasl2 (46), Serpina3f (40), Igtp (47), Psmb9 (48), Dhx58 (49), and Oas1g (49). Thus, STAT1 target gene responses to canonical mLIFR signaling were lost when mOSM interacted with mLIFR. This pattern of STAT3, rather than STAT1, target gene transcription appeared to result from an ability of mOSM-mLIFR to phosphorylate STAT3 at both Ser-727 and Tyr-705 but not to phosphorylate STAT1 in response to mOSM in Osmr^−/− cells.

The only gene that was not a classical STAT3 target in the top 10 genes was Cish, which is usually regarded as a STAT5-responsive gene (34). Previously we did not detect STAT5 phosphorylation in response to mOSM in Osmr^−/− osteoblasts (7), and other STAT5 target genes, such as Pim1 and Bcl2 (73), were not in the mOSM-LIFR gene set, suggesting that Cish transcription may occur in response to a broader range of STAT proteins.

The question of how mOSM can act through gp130-LIFR to regulate only a subset of those genes regulated by mLIF/hOSM through the same receptor complex is common to all receptors with multiple ligands that produce different biological effects within a single cell type. These include not only LIFR and OSMR (1, 74, 75) but also type I interferons (76). In the case of mOSM, our data suggest that impaired gp130 and LIFR phosphorylation may be responsible for the bias toward STAT3 signaling, but how receptor phosphorylation level is controlled remains unknown. It may relate to altered binding conformation, binding affinity, or binding stability compared with mLIF/hOSM. This could be better understood by structural studies as achieved with the binding of mLIF to LIFR (75). No structural data yet have defined the binding of mOSM or hOSM to either OSMR or LIFR; it remains structurally unclear why hOSM can bind to both LIFR and OSMR, whereas mOSM binds with higher affinity to OSMR (1). One possible model is that mOSM engages LIFR and gp130 in a manner that leads to incomplete transactivation of JAKs so that the STAT3 docking site on LIFR is phosphorylated, but STAT1 and SOCS3 binding sites are not. This would lead to selective STAT3 signaling and possibly its enhancement by lack of SOCS3 feedback even at low affinity interactions.

STAT3 docking sites on LIFRβ are phosphorylated at Tyr-765, Tyr-812, Tyr-904, and Tyr-914 in mouse (Tyr-981, Tyr-1001, and Tyr-1028 in human), whereas STAT3 docking sites on gp130 are phosphorylated at Tyr-767, Tyr-814, Tyr-905, and Tyr-915 (75). Any one or a combination of these may be modified. In addition, the single SOCS3 binding site on gp130 is phosphorylated at Tyr-757 in mouse (Tyr-759 in human), and a similar SOCS3 binding site on LIFRβ may also be phosphorylated (75). Identifying the specific differences in phosphorylation will require an extensive analysis of mutant receptors.

The focus of mOSM-LIFR action on STAT3 signaling is a similar effect on gene expression as that described previously for hOSM-hLIFR in human breast cancer cell lines (3). In that work, a LIF antagonist was used to distinguish shared and distinct signaling pathways of hOSM-hOSMR and hOSM-hLIFR. They reported that hOSM action through both hOSMR and hLIFR resulted in changes in genes downstream of STAT3 and STAT1 (e.g. Socs3 and Irf1), but regulation of STAT1 genes was lost when hOSM interacted with hLIFR. The similarities we observe suggest that mOSM in mouse cells may interact with mLIFR in a manner similar to hOSM-hLIFR in human cells. This suggests that human analogues that mimic the anabolic action of mOSM-STAT3 signaling could be developed when more information about the structure of the interactions of mOSM-mLIFR and hOSM-hLIFR becomes available.

This study was conducted in primary calvarial osteoblasts because genetic deletion of OSMR in these cells allowed us to identify mOSM-dependent regulation of sclerostin (7), an inhibitor of bone formation, in the absence of OSMR. Whether mOSM is capable of initiating STAT3 phosphorylation in other OSMR-deficient cell types is not known. We could detect no STAT3 phosphorylation response to mOSM in Ba/F3 cells expressing mgp130 and mLIFR. This suggests that accessory factors that promote mOSM-LIFR complex formation may be present in calvarial osteoblasts. One possibility is sortilin, which has been shown to promote ciliary neurotrophic factor and LIF signaling (77) and which we have detected by qPCR in primary calvarial osteoblasts (data not shown). However, addition of sortilin to mgp130/mLIFR-positive Ba/F3 cells did not promote STAT3 phosphorylation.

This study began with the aim of identifying the pathway through which mOSM-mLIFR is able to promote bone formation without stimulating bone resorption. Although in our earlier work we observed that mOSM could promote bone formation in the absence of OSMR, and its effect on promoting RANKL (Tnfsf11) transcription was not detectable in OSMR-deficient cells, the present study detected a delayed, but still significant, increase in Tnfsf11 mRNA in Osmr^−/− cells in response to mOSM. Although our hypothesis that mOSM-LIFR might promote an exclusively “anabolic” gene signature whereas mOSM-OSMR promoted an exclusively “catabolic” gene signature was not upheld, the rescue of the gp130^Y757F/Y757F osteopenic phenotype suggests that an elevated ratio of STAT3:STAT1 responses may be more anabolic than a balanced ratio or one in favor of STAT1. This may be particularly relevant when STAT3 is hyperactivated as in inflammation- or cancer-induced bone loss. This is supported by an earlier report that STAT1 inhibition promoted bone formation in fracture healing (78), the greater periosteal circumference we observed in Stat1^−/− mice, and high alkaline phosphatase activity in cultured osteoblasts with constitutive STAT3 activation (16).

To conclude, this study identifies that mOSM acting through the mLIFR phosphorylates STAT3 but not STAT1, resulting in specific regulation of STAT3-responsive genes, thereby activating a more specific intracellular signaling pathway than those induced by hOSM or hLIF acting through the same receptor. Our data suggest that this specificity relates to reduced gp130 and LIFR phosphorylation by mOSM-mLIFR. Targeted activation of STAT3 signaling downstream of gp130 without promoting STAT1 signaling provides a protective effect in the skeleton, suggesting that mimics of the mOSM-mLIFR interaction or of this pattern of STAT3-specific activation could provide benefit for skeletal fragility.

Experimental Procedures

Cell Culture and Microarray

Osmr^−/− mice (6) were obtained from Prof. Atsushi Miyajima (The University of Tokyo), backcrossed onto C57BL/6 for six generations, and maintained as a heterozygous breeding colony at our institute for >5 years. Primary calvarial osteoblasts were generated as described previously (79) from homozygous litters of Osmr^−/− and wild type cousin-bred mice. On three occasions, WT and Osmr^−/− cells were defrosted, expanded, and differentiated for 17 days in osteoblast differentiation medium. At 17 days, when these cells express elevated levels of sclerostin, they were serum-starved overnight and then treated with 10 ng/ml hOSM (R&D Systems), mOSM (R&D Systems), or mLIF (Chemicon). Cell digests were collected at 1 and 6 h after treatment commenced. mRNA was prepared as described previously (7) and hybridized to Illumina MouseWG-6 V2 BeadChips. Raw probe intensities for both regular and control probes were exported from Illumina's Genome Studio software.

Bioinformatics analysis was conducted using the limma software package (80). The raw intensity values were background-corrected and normalized using the neqc function, which performs “normexp” background correction and quantile normalization using parameters estimated from the control probes (81). Probes were filtered as not expressed if they failed to achieve a detection p value of 0.05 for at least three samples. The Illumina manifest file (MouseWG-6_V2_0_R1_11278593_A) was used for probe annotation. Gene symbol aliases were converted to current official gene symbols using the Bioconductor organism package org.Mm.eg.db. Linear models were used to test for expression differences between treatments while adjusting for batch effects associated with the date on which the cell cultures were performed. Differential expression between treatments was assessed using empirical Bayes moderated t statistics (82), allowing for an intensity-dependent trend in the standard errors. The mLIF and hOSM responses were combined and compared with others using appropriate linear model contrasts. The false discovery rate was controlled at less than 0.05 using the method of Benjamini and Hochberg (59). Biological correlations between log -fold change profiles were estimated by limma's genas function, which estimates genuine associations between profiles, adjusting for any technical component of the correlation. Lists of STAT3 target genes were obtained from the Molecular Signatures Database v5.1 (83), and enrichment was tested using limma's kegga function. The microarray data are available as Gene Expression Omnibus (GEO) series GSE83418.

Phosflow and Western Blotting

Samples for Phosflow were generated by culturing mouse primary calvarial osteoblasts from C57BL/6 mice for 22–24 days in differentiation medium on three separate occasions. Cells were seeded and grown to 80–90% confluence in culture in complete medium and serum-starved overnight in αMEM supplemented with 2% heat-inactivated fetal bovine serum (FBS) + 50 μg/ml ascorbate. The following day cells were stimulated with 50 ng/ml mOSM, hOSM, or mLIF for 15 min in low serum conditions. Cells were rinsed twice with PBS, fixed in 1% formalin at 37 °C for 10 min, and permeabilized in 90% methanol for 30 min on ice. After overnight storage at −80 °C, cells were washed twice in Stain Buffer (PBS plus 1% FBS, 0.09% sodium azide, 0.5 mm EDTA) and stained with the following antibodies: AF647 pSTAT3 Ser-727, AF488 pSTAT3 Tyr-705, and phycoerythrin total STAT3 (all BD Biosciences). Cells were washed, resuspended in Stain Buffer, and analyzed on an LSR Fortessa (BD Biosciences). Mean fluorescence intensity (MFI) was determined for all three biological replicates using FlowJo software. Gating was based on fluorescent signal above background such that all signals were included for cytokine induction of pSTAT3 Tyr-705 and total STAT3; signals in the secondary peak were included for cytokine induction of pSTAT3 Ser-727.

Whole cell lysates were prepared for Western blotting analysis from wild type or Osmr^−/− primary calvarial osteoblasts maintained in normal growth medium (αMEM + 10% FBS) (JRH Biosciences batch numbers 074, 414, and 379) or under differentiating conditions (αMEM + 15% FBS (JRH Biosciences batch numbers 074, 414, and 379; Assay Matrix catalogue number A137A11) and 50 μg/ml ascorbate (Sigma)) for 21 days in 10-cm² dishes. Cultures were serum-starved in 2% FBS overnight prior to cytokine treatment. Cells were treated with mOSM (R&D Systems catalogue number 495-MO, lot number TX0313011), hOSM (R&D Systems catalogue number 295-OM, lot number DY0808111), or mLIF (Merck Millipore catalogue number LIF2010, lot number 2561112) at 50 ng/ml for 15 or 30 min. LA (55) was used at 2.5 μg/ml for 30 min prior to the addition of cytokines; STAT3 inhibitor (ML116) (56) was used for 1 h prior to the addition of cytokines at 0.1, 1, or 10 μg/ml. Cells were then washed twice with ice-cold PBS before cell lysates were prepared by shaking at 4 °C with modified radioimmune precipitation assay buffer (84), proteinase inhibitor (Sigma), and phosphatase inhibitor (Dako) before lysates were centrifuged. Protein concentration was determined according to the manufacturer's protocol (Pierce). 20–50 μg of protein were loaded onto 4–12% gradient gels (Invitrogen) under reducing conditions for electrophoresis before being transferred to nitrocellulose membranes (iBlot, Invitrogen) and probed with pSTAT1(Tyr-701) (Cell Signaling Technology catalogue number 9171, lot number 8), STAT1 (Cell Signaling Technologies catalogue number 9175, lot number 13), pSTAT3(Ser-727) (Cell Signaling Technology catalogue number 9136, lot number 13), pSTAT3(Tyr-705) (Cell Signaling Technology catalogue number 9131, lot number 30), STAT3 (Cell Signaling Technology catalogue number 9139, lot number 8), pan-actin Ab-5 (Neomarkers catalogue number 1295-P0, lot number 1295P1501N). Protein bands were detected with ECL chemiluminescence (GE Healthcare catalogue number RPN2209) and film (Fujifilm catalogue number 4741019289) according to the manufacturers' instructions. ImageJ was used to measure pixel intensity of phospho-STAT3 relative to total STAT3 and corrected for relative response magnitude within each experiment by expressing as a percentage of the level of STAT3 induction reached by mOSM.

Immunoprecipitation was performed by dividing the volume of lysate collected equally into three, one lysate per antibody. Immunoprecipitation was carried out using 30 μl of Protein A-Sepharose fast flow beads (GE Healthcare catalogue number 17-1279-01, lot number 10168456) and 2 μg of either gp130 (M-20) antibody (Santa Cruz Biotechnology catalogue number sc-656, lot number D1515), OSMRβ (C-7) antibody (Santa Cruz Biotechnology catalogue number sc-376380, lot number L2711), or LIFR (C-19) antibody (Santa Cruz Biotechnology catalogue number sc-659, lot number F12115), rotating overnight at 4 °C. Postimmunoprecipitation the beads were washed three times in PBS and once in ice-cold modified radioimmune precipitation assay buffer. Loading buffer was added, and the beads were heated at 60 °C for 10 min prior to being loaded onto 4–12% gradient gels (Invitrogen) under reducing conditions for electrophoresis before being transferred to nitrocellulose membranes (iBlot) and probed with pan-phosphorylation antibody 4G10 (Merck Millipore catalogue number 05-321, lot number 2593884). Protein bands were detected with ECL chemiluminescence and film according to the manufacturers' instructions.

Quantitative Real Time PCR

cDNA synthesis of 1 μg of DNase-treated (Ambion) RNA was performed using AffinityScript (Agilent Technologies catalogue number 600559) according to the manufacturer's instructions. Stock cDNA was diluted 1:5–1:10, and qPCR was performed using either an in-house master mix or 10× AmpliTaq Gold with Brilliant II SYBR Green Mastermix (Agilent Technologies catalogue number 600828). Primers used (Table 3) were designed using NCBI Primer Blast or PrimerQuest. Samples were dispensed into an optically clear 96-well plate (4Titude, Thermo Scientific catalogue number 4ti-0750) and run on a Stratagene Mx3000P (Agilent Technologies) with two-step cycling conditions (95 °C for 10 min followed by 95 °C for 30 s and 60 °C for 1 min) for 40 cycles followed by a dissociation step (95 °C for 1 min, 55 °C for 30 s, and 95 °C for 30 s). Postrun samples were analyzed using Stratagene software MxPro and reported using linear Δ threshold cycle (ΔCt) values normalized to β₂-microglobulin (B2m) or hypoxanthine phosphoribosyltransferase 1 (Hprt1). mRNA levels of these two housekeeping genes were not modified by the cytokine or antagonist treatments used.

TABLE 3.

Primers used for quantitative real time PCR analysis

F, forward; R, reverse.

Gene	Primer sequence	GenBankTM accession no.
B2m	F 5′-TTCACCCCCACTGAGACTGAT-3′	NM_009735.3 (88)
	R 5′-GTCTTGGGCTCGGCCATA-3′
Bcl3	F 5′-CCTTTGATGCCCATTTACTCTACCC-3′	NM_033601
	R 5′-AGCGGCTATGTTATTCTGGACCAC-3′
Gbp2	F 5′-CAGCTCGTTGCTCAGACTTG-3′	NM_010260
	R 5′-TTGCTGCCTCTGTGAGTGAC-3′
Hprt1	F 5′-TGATTAGCGATGATGAACCAG-3′	NM_013556 (89)
	R 5′-AGAGGGCCACAATGTGATG-3′
Oasl2	F 5′-GAAGAGCAGGCGAGAAACC-3′	NM_01185
	R 5′-TGGAAATAGGTCCCAAAGCA-3′
Sbno2	F 5′-AGAGGTGCTGGATGAGAA-3′	NM_183426
	R 5′-GCTGCTGCCATCACTAAT-3′
Socs1	F 5′-GTGGTTGTGGAGGGTGAGAT-3′	NM_00989 (90)
	R 5′-CCCAGACACAAGCTGCTACA-3′
Socs3	F 5′-TGAGCGTCAAGACCCAGTCG-3′	NM_007707 (90)
	R 5′-CACAGTCGAAGCGGGGAACT-3′

Breeding and Analysis of Stat1-null and gp130^Y757F/Y757F Mice

Mice homozygous for the gp130^Y757F/Y757F knock-in mutation and their osteopenic phenotype have been described previously (58) as have the earlier crosses of these mice with the Stat1^−/− and Stat3^+/− mice (36). Stat1^−/− mice (85) were derived from the same colony, bred from heterozygous parents, and compared with littermate wild type mice. For the cross with gp130^Y757F/Y757F mice, mice were from mixed litters including gp130^Y757F/Y757F/Stat1^−/− and gp130^Y757F/Y757F littermates and gp130^Y757F/Y757F/Stat3^+/− and gp130^Y757F/Y757F littermates; wild type mice were from the same breeding colony. All mice were bred on a mixed 129/C57BL/6 background. All mice were bred and maintained in the same facility as their controls; tissues were collected for analysis at 12–14 weeks of age. Micro-CT analysis of Stat1^−/− mice was performed using a SkyScan 1076 system (Bruker micro-CT, Kontich, Belgium) on distal femoral specimens from 12-week-old mice as described previously (86). Briefly, images were acquired using the following settings: 9-μm voxel resolution; 0.5-mm aluminum filter; 50-kV voltage; 100-μA current; 2600-ms exposure time; rotation, 0.5°; frame averaging, 1. Images were reconstructed and analyzed using SkyScan software NRecon (version 1.6.8.0), DataViewer (version 1.4.4), and CT Analyzer (version 1.11.8.0). The femoral trabecular analysis region of interest was determined by identifying the intracondylar notch at the distal end of the femur and calculating 15% of the total femur length toward the femoral midshaft where we then analyzed a region of interest of 15% of the total femur length; a threshold of 43–255 was used to define trabecular bone. Tibial samples from gp130^Y757F/Y757F mice were analyzed by histomorphometry in the proximal secondary spongiosa using the OsteoMeasure system (Osteometrics, Decatur GA) after embedding in methylmethacrylate resin as described previously (87). Microscope images were collected on a Leica DMRB microscope coupled to an Olympus DP72 camera and CellSens software under a 4.0 Plan objective.

Author Contributions

E. C. W. prepared samples for microarray and together with R. W. J. carried out Western blotting, PCR, and Phosflow analysis. Y. H. and G. K. S. carried out microarray analysis. H. J. B. carried out micro-CT analysis of the Stat1^−/− mice. I. J. P. carried out histomorphometric analysis of the gp130^Y757F/Y757F mice and their crosses, which were provided by B. J. J. N. A. S. conceived the project and developed the ideas with J.-G. Z., B. J. J., G. K. S., R. W. J., and N. A. N. N. A. S. wrote the manuscript with contributions from all co-authors. All authors reviewed the results and approved the final version of the manuscript.