Tumor Necrosis Factor Receptor Superfamily Member 19 (TNFRSF19) Regulates Differentiation Fate of Human Mesenchymal (Stromal) Stem Cells through Canonical Wnt Signaling and C/EBP*

Mechanisms controlling human multipotent mesenchymal (stromal) stem cell (hMSC) differentiation into osteoblasts or adipocytes are poorly understood. We have previously demonstrated that Wnt signaling in hMSC enhanced osteoblast differentiation and inhibited adipogenesis by comparing two hMSC cell lines overexpressing mutated forms of the Wnt co-receptor LRP5: T253I (hMSC-LRP5T253) and T244M (hMSC-LRP5T244) conducting high and low level of Wnt signaling, respectively. To explore the underlying molecular mechanisms, we compared gene expression profiles of hMSC-LRP5T253 and hMSC-LRP5T244 treated with Wnt3a using whole genome expression microarrays and found that TNFRSF19 is differentially up-regulated between the two cells lines. Bioinformatic analysis and dual luciferase assay of its promoter revealed that TNFRSF19 transcript 2 (TNFRSF19.2) is a target of canonical Wnt signaling. Knocking down TNFRSF19 in hMSC-LRP5T253 cells decreased Wnt3a-induced osteoblast differentiation marker alkaline phosphate activity and its overexpression in hMSC-LRP5T244 cells increased alkaline phosphate activity. In addition, TNFRSF19 was negatively regulated by adipogenic transcription factor CCAAT/enhancer-binding proteins (C/EBP). Knocking down TNFRSF19 in hMSC-LRP5T253 cells or its overexpression in hMSC-LRP5T244 cells significantly increased or decreased adipogenesis, respectively. In conclusion, we revealed a novel function of TNFRSF19 as a factor mediating differentiation signals that determine the hMSC differentiating fate into osteoblasts or adipocytes.

tion, migration, polarity, and gene expression (1). Wnt/␤-catenin or the canonical Wnt signaling pathway is initiated by binding of Wnt ligands to their cognate membrane receptors including Frizzled receptor and low-density lipoprotein receptor-related proteins 5 and 6 (LRP5/6) 3 leading to formation of a stabilized cytosolic ␤-catenin, which translocates to the nucleus and binds to T-cell factor/lymphoid enhancer binding factor (TCF/LEF) binding sites in promoters of target genes (2). Identification of both activated and inactivated mutations in LRP5 as the cause of high bone mass phenotype (also known as osteopetrosis type I) or low bone mass (known as osteoporosis pseudoglioma syndrome), respectively (3)(4)(5)(6)(7), suggested that Wnt signaling in bone cells is responsible for such altered bone mass phenotypes (8).
A large body of evidence has demonstrated that canonical Wnt signaling is important for bone cell biology. Evidence from animal models with modification of different Wnt components has demonstrated a regulatory function of Wnt signaling in bone development and in postnatal bone homeostasis (9). Wnt signaling is thought to be not only important in osteoblastic cell proliferation, differentiation, synthesis of bone matrix, and osteoclast differentiation, but also in osteocytes transmitting signals of mechanical loading to cells on the bone surface (10,11). The role of LRP5 as a Wnt co-receptor (12,13), its involvement in transduction of canonical Wnt signaling (3,14,15), the resistance of activated LRP5 mutations to Wnt antagonists SOST or DKK1 inhibition (16 -18) or disrupted interaction with LRP5/6 chaperone Mesd (19), as well as the impaired canonical Wnt signaling in inactivated mutations of LRP5 (4), have provided strong evidence for the involvement of LRP5 as a Wnt co-receptor in osteoblastic bone biology.
We have previously reported that patients with activated mutation T253I of LRP5 was responsible for the high bone mass * This work was supported in part by grants from the Danish Medical phenotype in a large family in Denmark. Iliac bone biopsies from these patients exhibited increased bone mass and decreased adipose tissue mass in bone marrow suggesting enhanced osteoblastogenesis and inhibition of adipogenesis of human mesenchymal (multipotent stromal) stem cells (hMSC) in bone marrow (15). To establish a model for this human disease, we have established an hMSC cell line overexpressing the LRP5 T253I mutation (named hMSC-LRP5 T253 ) as well as another hMSC cell line overexpressing an inactivated mutation of LRP5 known to cause osteoporosis pseudoglioma syndrome, T244M (named hMSC-LRP5 T244 ). We have also demonstrated that hMSC-LRP5 T253 and hMSC-LRP5 T244 transmit high and low Wnt signaling, respectively. In addition, in vitro and in vivo studies have demonstrated that hMSC-LRP5 T253 cells exhibit enhanced osteogenic differentiation and impaired adipogenic differentiation. The hMSC-LRP5 T244 cells exhibited the opposite phenotype (15). These experiments suggested that canonical Wnt signaling mediated through LRP5 targets MSC and determines their differentiation fate into osteoblasts or adipocytes. This notion has been supported by other studies employing Wnt10b as a canonical Wnt ligand (20,21).
To further identify downstream molecules and/or signaling pathways responsible for cell fate determination of hMSC in this model, we compared global gene expression patterns of hMSC-LRP5 T253 and hMSC-LRP5 T244 cells treated with Wnt3a. We found that tumor necrosis factor receptor superfamily member 19 (TNFRSF19) is regulated by both canonical Wnt signaling and C/EBP. Modulation of TNFRSF19 expression resulted in significant changes in hMSC differentiation ability to the osteoblastic and adipocytic lineage suggesting an important role as a mediator in regulation of differentiation fate of hMSC.
Protein synthesis inhibitor cycloheximide (CHX, Fluka) was dissolved in dimethyl sulfoxide and stored at Ϫ80°C. To inhibit protein synthesis, 1 g/ml of CHX was added into culture medium for 8 h.
Microarray Analysis-hMSC-LRP5 T253 and hMSC-LRP5 T244 cells were treated with 50% control CM or Wnt3a-CM in triplicates for 1 or 24 h and total RNA was prepared by RNeasy mini kit (Qiagen). Synthesis of biotinylated cRNA was performed with 500 ng of total RNA of each sample using the Illumina TotalPrep RNA Amplification Kit (Ambion) according to the manufacturer's instructions. The quantification and quality analysis of total RNA and amplified cRNA were performed on a ND-1000 (NanoDrop Technologies, Wilmington, DE). Size distribution of the extracted total RNA and the amplified cRNA were checked with the Agilent RNA 6000 Nano LabChip kit and 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). HumanRef-8 v2 BeadChips from Illumina (Illumina, San Diego, CA) were used to generate expression profiles of more than 22,000 well annotated RefSeq transcripts following the manufacturer's recommended protocols.
Bioinformatic Analysis-The processing of the microarray expression profiles and identification of differentially expressed genes were performed by Beadstudio software provided by Illumina. We used DAVID (the Data base for Annotation, Visualization and Integrated Discovery) to cluster differentially expressed genes into relevant signaling pathways.
Promoter sequences of selected genes including all of transcripts were retrieved from the Transcriptional Regulatory Element data base (TRED) and DataBase of Transcriptional Start Sites (DBTSS). Promoter analysis was performed using online software TFsitescan and TFSEARCH. Besides, the consensus sequence of the TCF/LEF binding element (TBE, CTTTGWW) was also screened manually.
Site-directed Mutagenesis-Site-directed mutagenesis of TBE by deleting CTTTG was performed by the QuikChange II site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. The mutations were confirmed by sequencing.
Transfection, Retrovirus Production, and Overexpressing of TNFRSF19-For transient overexpressing of TNFRSF19, 3 g of retrovirus vector pBABEpuro or pBABE-TNFRSF19.2 were transfected into the Phoenix A packaging cell line and retrovirus containing medium were collected for infection as above.
ALP Quantitation-ALP activity was measured by using p-nitrophenyl phosphate (Fluka) as substrate and normalized against the cell number. Briefly, cell number was quantitated by adding CellTiter-Blue reagent (Promega) to culture medium, incubating at 37°C for 1 h, and measuring fluorescent intensity (560 EX /590 EM ). Cells were then washed with phosphate-buffered saline and Tris-buffered saline, fixed in 3.7% formaldehyde, 90% ethanol for 30 s at room temperature, incubated with substrate (1 mg/ml of p-nitrophenyl phosphate in 50 mM NaHCO 3 , pH 9.6, and 1 mM MgCl 2 ) at 37°C for 20 min, and the absorbance measured at 405 nm (22).
Cytochemical Staining-Oil-red O staining were performed 7 days after differentiation as described (15).
Statistical Analysis-Statistical testing was determined by Student's t test and p Ͻ 0.05 was considered significant.

TNFRSF19 Is Up-regulated by Canonical Wnt
Signaling-To explore the mechanisms underlying cellular phenotypes of hMSC-LRP5 T253 and hMSC-LRP5 T244 cells, we treated the cells with control CM or Wnt3a-CM for 1 or 24 h and compared their global gene expression profiles by Illumina microarray system. Upon Wnt3a stimulation for 1 h, the gene expression profiles of hMSC-LRP5 T253 and hMSC-LRP5 T244 cells did not exhibit significant changes (data not shown). However, upon Wnt3a stimulation of hMSC-LRP5 T244 cells for 24 h, 131 Illumina target IDs representing 124 non-redundant genes were changed at least 2-fold, 73 of were up-regulated and 51 were down-regulated (supplemental Table S1). In hMSC-LRP5 T253 , 176 Illumina target IDs representing 167 non-redundant genes were changed 2-fold including 94 up-regulated and 73 down-regulated non-redundant genes (supplemental Table S2). By clustering significantly regulated genes into KEGG signaling pathways using the online DAVID program, we found similar signaling pathways affected upon Wnt3a stimulation in both cell lines including cytokine-cytokine receptor interaction, transforming growth factor-␤ signaling, and the Wnt signaling pathway (supplemental Tables S1 and S2). The microarray data reported here was deposited in Gene Expression Omnibus (GEO) with accession number GSE16186.
To detect Wnt target genes, we identified genes differentially changed upon Wnt3a treatment in hMSC-LRP5 T253 and hMSC-LRP5 T244 . As shown in Fig. 1A, the x and y axis represent fold-induction (log 2 ratio) of each Illumina target ID by Wnt3a in hMSC-LRP5 T244 and hMSC-LRP5 T253 , respectively. Most of genes were changed in a similar direction, except there were 15 genes including BMP2, DKK2, NKD1, and TNFRSF19 (here it is TNFRSF19.1) at least 2-fold differentially up-or down-regu-lated between the two cell lines (Fig. 1A, inset). To further identify novel canonical Wnt target genes, we retrieved the promoter sequences from all of transcript variants of these 15 genes as well as from 21 genes up-regulated at least 3-fold by Wnt3a in hMSC-LRP5 T253 (supplemental Table S2) and searched for TBE (CTTTGWW) within 1 kb upstream and 200 bp downstream of transcription start site defined by the reference mRNA sequence in GenBank TM . One of the genes identified was TNFRSF19 that has two different transcripts: TNFRSF19.1 and TNFRSF19.2. We found that the promoter of TNFRSF19.2 (19.2p) but not the promoter of TNFRSF19.1 (19.1p) has six canonical TBE within the defined promoter region (Fig. 2). The microarray data demonstrated that the TNFRSF19.2 expression level was higher in hMSC-LRP5 T253 cells than in hMSC-LRP5 T244 under basal conditions (signal density 2141 versus 960) and after Wnt3a stimulation (signal density 4985 versus 2891) (Fig. 1B). Interestingly, TNFRSF19.1 was differentially up-regulated by Wnt3a in hMSC-LRP5 T253 (2.27-fold) and exhibited no change in hMSC-LRP5 T244 cells (Fig. 1, A and B) but its promoter (19.1p) does not contain any   1C). TNFRSF19.2 was up-regulated by Wnt3a in both hMSC-LRP5 T244 and hMSC-LRP5 T253 cells (2.5versus 2.3-fold, respectively). However, the expression level in hMSC-LRP5 T253 cells was higher than hMSC-LRP5 T244 as observed in microarray data. The up-regulation of TNFRSF19.1 by Wnt3a was also confirmed by real-time RT-PCR (Fig. 1C).
Promoters of TNFRSF19 Transcripts Have Different Compositions-In depth bioinformatic analysis of 19.1p and 19.2p promoters revealed that both promoters have different cisregulatory elements. As shown in Fig. 2, 19.1p does not contain any TBE but have three C/EBP binding sites (Ϫ1048 to Ϫ1035 bp, Ϫ1010 to Ϫ998 bp, and Ϫ408 to Ϫ395 bp), which bind key adipogenic transcription factors C/EBPs. Besides, 19.1p is a TATA-less promoter with higher G ϩ C content (64%, but 42% in 19.2p) and six GC boxes (Ϫ731 to Ϫ725 bp, Ϫ254 to Ϫ248 bp, Ϫ237 to Ϫ231 bp, Ϫ207 to Ϫ201 bp, Ϫ164 to Ϫ158 bp, and ϩ153 to ϩ159 bp), which are typical features of a housekeeping gene promoter (23). The 19.2p contains six TBE (Ϫ916 to Ϫ910 bp, Ϫ656 to Ϫ650 bp, Ϫ612 to Ϫ606 bp, ϩ33 to ϩ39 bp, ϩ70 to 76 bp, and ϩ134 to 140 bp) including three TBE located in the first exon, four C/EBP binding sites (Ϫ840 to Ϫ827 bp, Ϫ687 to Ϫ675 bp, Ϫ68 to Ϫ55 bp, and ϩ202 to 215 bp), and one TATA box located in the first exon (ϩ51 to ϩ57 bp) (Fig. 2). Existence of TBE and C/EBP binding sites in 19.1p and 19.2p suggested TNFRSF19 may be involved in transmitting different signals to cell differentiation to osteoblasts or adipocytes. Based on these observations, we further examined the role of TNFRSF19 in hMSC differentiation.
TNFRSF19.2 but Not TNFRSF19.1 Is a Canonical Wnt Target Transcript-To further confirm that TNFRSF19 is regulated directly by Wnt signaling, we cloned 19.1p (1328 bp, Ϫ1149 to ϩ178 bp) and 19.2p (1318 bp, Ϫ975 to ϩ342 bp) into the promoterless firefly luciferase reporter vector pGL3-basic and the activities of the promoters were determined by dual luciferase assay. Without Wnt activation, 19.1p has a much higher basal activity than 19.2p (Fig. 3A) suggesting that TNFRSF19.1 is con- stitutively expressed by hMSC. We examined responsiveness of promoters to Wnt activation by co-transfection of 293T cells with a stabilized form of ␤-catenin (S33Y) (24). S33Y increased promoter activities of 19.2p to a maximal of 12-44-fold in a dose-dependent manner but the 19.1p promoter was not responsive (Fig. 3B). Furthermore, ␤-catenin-mediated transactivation of 19.2p could be further enhanced by co-transfection with wild type TCF4 and blocked by co-transfection with dnTCF4, which lacks of ␤-catenin binding domain (25) (Fig.  3C). The 19.1p did not respond to either wild type TCF4 or dnTCF4 (Fig. 3C). To determine whether TBE sites are essential for Wnt activation, we deleted each of the TBE sites in 19.2p. The single deletion of the distal three TBE sites (Fig. 2, T1, T2, or T3) did not affect the responsiveness of 19.2p to Wnt activation. On the other hand, single deletion of the proximal three TBE sites (Fig. 2, T4, T5, or T6) decreased the responsiveness of 19.2p to Wnt activation by 18 to 45%. Multiple TBE deletions further decreased 19.2p responsiveness to Wnt activation by 67% (Fig. 3D).
We further evaluated whether Wnt3a up-regulated TNFRSF19 expression requires protein synthesis. hMSC-LRP5 T253 cells were treated with control CM or Wnt3a-CM in the presence of either dimethyl sulfoxide as control or the protein synthesis inhibitor CHX (1 g/ml) for 8 h. Real-time RT-PCR analysis indicated that both TNFRSF19.1 and -19.2 were up-regulated by Wnt3a in the absence of CHX. In the presence of Wnt3a and CHX, expression of TNFRSF19.2 but not TNFRSF19.1 was stimulated (Fig. 3E), suggesting that TNFRSF19.2 but not TNFRSF19.1 is transcriptionally regulated by canonical Wnt signaling.
Expression of TNFRSF19 Is Inhibited by C/EBP-Because TNFRSF19.1 and TNFRSF19.2 promoters contain several C/EBP binding sites and expression of the transcription factor C/EBP has been shown to initiate adipogenesis by transactivating adipocyte-specific genes (26), we examined the regulation of TNFRSF19 expression by C/EBP. Co-transfection of 293T cells with C/EBP␣ and C/EBP␦ significantly decreased 19.1p and 19.2 activities in a dose-dependent manner (Fig. 4, A and B) and the inhibitory effects were more pronounced for 19.1p than 19.2p. On the other hand, C/EBP␤ did not inhibit 19.1p activity but only inhibited 19.2p (Fig. 4C). To further confirm that these C/EBP binding sites are functional elements, we performed site mutagenesis of these sites. Unfortunately, C/EBP mutant constructs displayed extremely low promoter activity that is 10% of the wild type promoter activity, which resulted in non-reproducible results (not shown).
Our data suggested that TNFRSF19 is repressed by C/EBP when adipogenesis is initiated. Therefore, up-regulation of  MAY

TNFRSF19.2 Functions Downstream of Canonical Wnt
Signaling to Regulate ALP Activity-We have previously demonstrated that Wnt activation in hMSC-LRP5 T253 but not in hMSC-LRP5 T244 induced osteoblast differentiation as demonstrated by up-regulation of the osteoblast differentiation marker ALP (15). To examine whether up-regulation of TNFRSF19 by Wnt3a is related to increased ALP activity, we first examined the time course expression of ALP and TNFRSF19 upon Wnt3a treatment. Compared with control CM, Wnt3a-induced ALP expression after 24 h (2.6-fold) and ALP levels continued to increase after 72 (3.3-fold) and 168 h (5.5-fold) (Fig. 5A). Wnt3a-induced up-regulation of TNFRSF19.2 was earlier than ALP and first observed at 8 h (3.4-fold) and increased continuously during Wnt3a treatment (6.7-, 7.9-, and 11-fold at 24, 72, and 168 h, respectively).

TNFRSF19.1 was also up-regulated by Wnt3a treatment but its expression reached a maximal level at 24 h and was maintained until 168 h.
Next, we overexpressed TNFRSF19.2 in hMSC-LRP5 T244 and hMSC-LRP5 T253 cells and measured ALP activity. As shown in Fig. 5B, ALP activity was increased significantly by Wnt3a compared with control CM in both hMSC-LRP5 T253 and hMSC-LRP5 T244 cells (p Ͻ 0.05), but hMSC-LRP5 T253 had higher ALP activity than hMSC-LRP5 T244 cells upon Wnt3a stimulation (p Ͻ 0.05), which corroborated our previous data (15). In the absence of Wnt3a, overexpression of TNFRSF19.2 in hMSC-LRP5 T253 and hMSC-LRP5 T244 cells significantly increased ALP activity compared with cells infected with the empty vector pBABE (p Ͻ 0.01). More importantly, TNFRSF19.2 overexpression in hMSC-LRP5 T253 and hMSC-LRP5 T244 cells increased ALP activity to the equivalent level induced by Wnt3a (p Ͼ 0.05) (Fig. 5B). Overexpression of TNFRSF19.2 in hMSC-LRP5 T244 cells followed by Wnt3a treat- ment further increased ALP activity compared with hMSC-LRP5 T244 cell overexpression of TNFRSF19.2 alone (p Ͻ 0.01). With or without Wnt3a, overexpression of TNFRSF19.2 in hMSC-LRP5 T244 cells increased ALP activity to an equivalent level in hMSC-LRP5 T253 cells, suggesting that TNFRSF19.2 can restore the osteoblastic differentiation capacity of hMSC-LRP5 T244 . Combined overexpression of TNFRSF19.2 and Wnt3a treatment in hMSC-LRP5 T253 cells did not further increase the ALP activity possibly due to saturation effects.
Finally, we knocked down TNFRSF19 in hMSC-LRP5 T253 cells using lentivirus-based shRNA. hMSC-LRP5 T244 cells were not included in this experiment due to their very low ALP expression upon Wnt3a stimulation. Knockdown efficiency was first determined using the MISSION shTNFRSF19 set (TRCN0000058883-TRCN0000058887, Sigma), which targets both TNFRSF19.1 and TNFRSF19.2. We found all of shTNFRSF19 inhibited at least 70% Wnt3a-induced up-regulation of TNFRSF19 transcripts at day 3 (data now shown). We chose TRCN0000058884 (sh19-2), which had the lowest knock down efficiency and TRCN0000058886 (sh19-4) and TRCN0000058887 (sh19-5), which had the highest knockdown efficiency for further experiments. Wnt3a-induced ALP activity was blocked by sh19-4 and sh19-5 (p Ͻ 0.05) but not affected by sh19-2 compared with scramble shRNA (shCtrl) (Fig. 5C). Combined with promoter analysis results, TNFRSF19.2 seems to be a transcript regulated by canonical Wnt signaling to mediate Wnt3a-induced ALP activity.
TNFRSF19.2 Is a Negative Regulator of Adipocyte Differentiation-Because C/EBP proteins are transcription factors inducing adipocytic differentiation, their inhibitory effects on TNFRSF19 expression suggested that TNFRSF19 may function as a regulator of adipogenesis. We first examined the time course expression of C/EBP and TNFRSF19 during in vitro adipogenesis. hMSC-LRP5 T253 and hMSC-LRP5 T244 cells were treated with AIM for 1 or 7 days. The expression levels of C/EBP and TNFRSF19 in AIM was normalized against the expression levels in normal medium at the corresponding day. C/EBP␣ was increased 8.5-fold in hMSC-LRP5 T244 cells and 6.2-fold in hMSC-LRP5 T253 cells at day 1. At day 7, C/EBP␣ expression levels were severalfold higher in hMSC-LRP5 T244 cells (270fold) compared with hMSC-LRP5 T253 cells (93-fold) (Fig. 6A) corroborating our previous findings of enhanced adipogenesis in hMSC-LRP5 T244 cells (15). Similarly C/EBP␤ and C/EBP␦ were up-regulated in both hMSC-LRP5 T253 and hMSC-LRP5 T244 cells with higher levels at day 1 compared with day 7 (Fig. 6A). In contrast to up-regulation of C/EBP, TNFRSF19.1 was down-regulated more than 40% and TNFRSF19.2 was decreased about 60% in both hMSC-LRP5 T244 and hMSC-LRP5 T253 cells at day 1 (Fig. 6B). At day 7, both transcripts were expressed higher in hMSC-LRP5 T253 cells supporting the reverse relationship between TNFRSF19 expression and adipocyte differentiation capacity. In addition, expression of TNFRSF19.1 and TNFRSF19.2 in hMSC-LRP5 T244 and hMSC-LRP5 T253 cells increased close to or above the expression levels in control cells (Fig. 6B) despite the continuous up-regulation of C/EBP␣ but accompanied by reduced up-regulation of C/EBP␤ and C/EBP␦ at day 7.
We then examined the effects of either overexpression of TNFRSF19.2 in hMSC-LRP5 T244 or knocking down TNFRSF19 in hMSC-LRP5 T253 on adipogenic differentiation, respectively. As shown in Fig. 6C, ectopic TNFRSF19.2 was expressed in hMSC-LRP5 T244 cells infected with pBABE-TNFRSF19.2 retrovirus compared with non-infected cells and cells infected with empty vector pBABE virus. Correspondingly, the expression level of adipogenic markers including C/EBP␣, PPAR␥2, aP2, and APM1 were significantly reduced (p Ͻ 0.05) (Fig. 6C). On the other hand, knocking down TNFRSF19 in hMSC-LRP5 T253 cells significantly increased its adipogenic differentiation evidenced by enhanced formation of a higher number of lipid-filled mature adipocytes (Fig. 6D). Real-time RT-PCR confirmed that sh19-4 has significantly higher knocking down efficiency compared with shCtrl (p Ͻ 0.05) up to 7-8 days post-infection and resulted in significant up-regulation of all examined adipogenic markers (p Ͻ 0.05) (Fig. 6D).
TNFRSF19.1 and TNFRSF19.2 Have Different Functions in hMSC Differentiation-Due to high similarities between coding sequences of both transcripts, we designed siRNA targeting transcript-specific 3Ј-regions to distinguish the function of the two transcripts in hMSC differentiation. Previous studies have demonstrated that siRNAs targeting the 3Ј-untranslated region induce gene knock-down (27). As shown in Fig. 7A, si19.2 targeting TNFRSF19.2 decreased ALP activity but had no effect on adipogenesis. In contrast, si19.1 targeting TNFRSF19.1 significantly enhanced expression of adipogenic markers but did not affect ALP activity. si19.1 and si19.2 has about 50% knocking down efficiency for each target transcript (data not shown).

DISCUSSION
TNFRSF proteins are type I transmembrane glycoproteins with a cysteine-rich extracellular ligand binding domain, a single transmembrane region, and a diverse cytoplasmic tail. TNFRSF play an important role in regulating diverse biological activities including key aspects of immune modulation (28). More than 22 TNFRSF have been identified and subgrouped according to their specific cytoplasmic domains that act as docking sites for signaling molecules, e.g. members containing cell death domain (e.g. TNFRSF1A, TNFRSF6, and TNFRSF12) mediate apoptosis and those containing TNFR-associated factor (TRAF) binding motif (e.g. TNFRSF1B and TNFRSF5) promote cell survival (28 -30). Some members of TNFRSF, e.g. osteoprotegerin (TNFRSF11B) and RANK (TNFRSF11A), are the main regulators of osteoclast differentiation and function (31,32) and thus TNFRSF is known to play an important role in bone biology.
Tnfrsf19 also known as toxicity and JNK inducer (Taj) or Troy is a newly identified member of TNFRSF (33)(34)(35). Gene expression studies revealed that Tnfrsf19 is highly expressed in brain and epithelium tissues during embryonic development (35,36). In the postnatal organism, it is expressed in the brain and hair follicle (33,35). In accordance with its expression profiles, Tnfrsf19 has been reported to be involved in brain and hair follicle development. During axon regeneration, the inhibitory molecules in central nervous system myelin signals through the NgR1-LINGO-p75 complex to inhibit successful axon regeneration. Observations that Tnfrsf19 were broadly expressed in postnatal and adult neurons compared with limited expression of p75 to certain types of neurons and its temporal expression during development, as well as Tnfrsf19, can replace p75 to form the NgR1-LINGO-Tnfrsf19 complex transducing signals from myelin offered a novel mechanisms of axon regeneration failure and a possible clinical application (37,38). In addition, Tnfrsf19 exhibits sequence homology and an overlapped expression pattern with Edar, which is another TNFRSF member important in development of ectodermal organs. Studies in Tnfrsf19-deficient mice have suggested a potentially redundant role between Tnfrsf19 and Edar in hair follicle development (39,40). Tnfrsf19 is absent in the spleen and thymus, which are major lymphoid tissues, suggesting that it does not play a role in the immune system (33,35). TNFRSF19 was also reported to interact with TRAF family members to activate JNK or NF-B pathways (34,35) and overexpressing of Tnfrsf19 induce cell death (34,41), although it does not contain cell death domain in its cytoplasmic region.
Tnfrsf19 is also expressed in mesenchymal tissues (39) but its role in mesodermal development is poorly studied. Here, we reported that human TNFRSF19 functions downstream of canonical Wnt signaling and C/EBP proteins to regulate differentiation fate of hMSC. The human TNFRSF19 gene has two transcripts: TNFRSF19.1 is 1485 bp in length and encodes a 423-amino acid polypeptide and TNFRSF19.2 is 4283 bp in length and encodes a 417-amino acid polypeptide. There is a structural homology between TNFRSF19.1 and TNFRSF19.2 except at the end of cytoplasmic tail where TNFRSF19.2 contains a major TRAF2-binding consensus sequence, (P/S/A/T)X(Q/E)E (SLQE at amino acid 413-416), which is absent in TNFRSF19.1. Mouse Tnfrsf19 has only one transcript encoding 416 amino acids with a major TRAF2-binding sequence, TLQE, in the cytoplasmic tail at position 276 -279. The TRAF binding sequence is important for activating downstream JNK or NF-B pathways (29) and its absence in TNFRSF19.1 suggests that both transcripts may have different biological functions. The position of the TLQE site is not conserved between mouse Tnfrsf19 (position 276 -279) and human TNFRSF19.2 (position 413-416) and thus its differential function in the two species as a binding site of TRAF family members has yet to be determined.
The differences between human and mouse TNFRSF19 gene structure were further revealed by promoter analysis. Compared with human TNFRSF19 promoters, mouse Tnfrsf19 promoter (Ϫ1000 to 200 bp according to Refseq NM_013869) contains two TBE but no C/EBP binding sites, GC box, and TATA box, suggesting that the human gene has acquired novel functions during evolution. We observed that TNFRSF19.1 was upregulated differentially by Wnt3a as evidenced by microarray data in the absence of TBE in the promoter. We conducted further analysis for the presence of the CTTTGWW sequence within 3 kb upstream of transcription start sites of 19.1p and 19.2p, and still did not identify TBE in 19.1p but several more in 19.2p (not shown). Up-regulation of TNFRSF19.1 may be mediated through an indirect effect of Wnt3a. It is also plausible that some sequences like CTTGWW or CTTTGXX may also respond to Wnt3a activation but need further confirmation.
In hMSC, canonical Wnt signaling enhance osteoblast differentiation as evidenced by increased ALP production (15). However, late osteoblast differentiation markers, e.g. osteocalcin, did not respond to Wnt3a treatment in our cell model possibly due to inhibition of the osteocalcin promoter (42), and down-regulation of Wnt signaling was reported to be required for late stage osteoblast differentiation (43). This may explain why previous studies employed ALP induction by canonical Wnt signaling as an indicator of osteogenesis (44,45). In this study, we observed that this effect is mediated by up-regulation of TNFRSF19.2 evidenced by overexpression of TNFRSF19.2 alone or combined with Wnt3a treatmentrescued ALP activity in hMSC-LRP5 T244 to the equivalent level of hMSC-LRP5 T253 . On the other hand, canonical Wnt signaling inhibited adipogenesis, suggesting that this signaling pathway may maintain an inactivated state so that TNFRSF19 is impossible to be up-regulated by canonical Wnt signaling during adipogenesis. We observed that C/EBP proteins inhibited TNFRSF19 expression suggesting that the basal expression of TNFRSF19 was required to be repressed at early stages of adipogenesis. In support of this notion, TNFRSF19 expression was reversely correlated with the expression of C/EBP. In our study, we chose different cell lines for overexpressing and knockdown experiments based on differences in the differentiation potential of hMSC-LRP5 T244 and hMSC-LRP5 T253 as reported previously (15), as well as the expression profile of TNFRSF19 transcripts in the two cell lines and the promoter analysis data. Unfortunately, we were not able to obtain an open reading frame of TNFRSF19.1 by PCR due to the very short untranslated region and biased nucleotide distribution. Besides, it was not possible to obtain specific shRNA targeting TNFRSF19.1 or  TNFRSF19.2 due to the nearly identical coding sequence of the two transcripts. Thus, the detailed functional role of TNFRSF19.1 in MSC proliferation and in transferring C/EBP signals to adipocyte differentiation need to be further investigated.
The microarray data revealed significant down-regulation of other members of TNFRSF in hMSC by Wnt signaling (supplemental Tables S1 and S2). TNFRSF11B, also called osteoprotegerin, is an osteoblast-secreted decoy receptor that functions as a negative regulator of bone resorption (46). TNFRSF1B also called p75 was down-regulated only in hMSC-LRP5 T253 cells. Genetic polymorphisms in TNFRSF1B have been reported to be possible determinants of bone mass (47)(48)(49)(50) possibly through epistatic interaction with the collagen type 1, ␣1 (COL1A1) gene (51). Due to the biological functions of TNFRSF19 reported here, it is interesting to examine the role of TNFRSF19 in osteoblast-osteoclast differentiation.
A reciprocal relationship between bone mass and adipose tissue mass in the bone marrow exists in a number of physiological (e.g. aging) and pathophysiological conditions (e.g. osteoporosis, glucocorticoid, or glitazone therapy) (52). Because osteoblasts and adipocytes arise from MSC in the bone marrow microenvironment, understanding the "switching mechanisms" between osteoblast versus adipocyte differentiation can provide novel therapeutic targets for enhancing bone formation (52,53). Previous studies have identified some factors that play a role in the switching mechanism. TAZ functions as a transcriptional modulator to stimulate MSC osteoblast cell differentiation and simultaneously block the differentiation to adipocytes through direct interaction with transcription factors Runx2 and PPAR␥ (54). On the other hand, PPAR␥ enhances adipocyte and inhibits osteoblast differentiation in vitro and in vivo (55). We have previously demonstrated that Wnt signaling in hMSC enhances osteoblast and inhibits adipocyte differentiation. In support of this, Wnt10b have been demonstrated to promote bone formation and inhibit adipogenesis (20,21). Interestingly, TAZ has been reported to be regulated by canonical Wnt signaling (54). We propose here that TNFRSF19 is a possible novel factor mediating the "switch" of osteoblast versus adipocyte differentiation of MSC. The different cis-regulatory elements found in promoters of human TNFRSF19 transcripts, presence or absence of TRAF binding sequence in the two transcripts of TNFRSF19, expression profiles of TNFRSF19 transcripts at basal level and during MSC differentiation, as well as overexpression and knockdown experiments suggest that the major function of TNFRSF19.1 and TNFRSF19.2 is to mediate different differentiation signals driving hMSC to osteoblastic or adipocytic lineages (Fig. 7). The relative importance of these two transcripts and their signaling mechanisms in mediating the biological effects on MSC biology remain to be determined.