Long noncoding RNA Malat1 inhibits Tead3-Nfatc1–mediated osteoclastogenesis to suppress osteoporosis and bone metastasis

MALAT1, one of the few highly conserved nuclear long noncoding RNAs (IncRNAs), is abundantly expressed in normal tissues. Previously, targeted inactivation and genetic rescue experiments identified MALAT1 as a suppressor of breast cancer lung metastasis. On the other hand, Malat1-knockout mice are viable and develop normally. On a quest to discover new roles of MALAT1 in physiological and pathological processes, we found that this lncRNA is downregulated during osteoclastogenesis in humans and mice. Notably, Malat1 deficiency in mice promotes osteoporosis and bone metastasis, which can be rescued by genetic add-back of Malat1. Mechanistically, Malat1 binds to Tead3 protein, a macrophage-osteoclast–specific Tead family member, blocking Tead3 from binding and activating Nfatc1, a master regulator of osteoclastogenesis, which results in the inhibition of Nfatc1-mediated gene transcription and osteoclast differentiation. Altogether, these findings identify Malat1 as a lncRNA that suppresses osteoporosis and bone metastasis.


Introduction
Osteoporosis, characterized by decreased bone mineral density (BMD), increased bone fragility, and susceptibility to fracture, reflects an imbalance in which osteoclastic bone resorption exceeds osteoblastic bone formation 1, 2 and is a potential contributor to acceleration of bone metastasis [3][4][5] .
Primary osteoporosis occurs during the aging process, particularly in postmenopausal women 6 .
Secondary osteoporosis has the same outcome as primary osteoporosis but is caused by certain medical conditions or medications 7 . In either condition, excessive osteoclastogenesis plays a key role and provides opportunities for therapeutic intervention.
Osteoclasts, a class of multinucleated giant cells (MGCs) that originate from the monocyte/macrophage lineage, are responsible for the resorption of bone matrix and minerals 8, 9 . Osteoclastogenesis is initiated by macrophage colony-stimulating factor (M-CSF) and receptor activator of nuclear factor-κB (RANK) ligand (RANKL), which induce the expression of osteoclast markers, such as cathepsin K (CTSK) and acid phosphatase 5 (ACP5, also known as TRAP), followed by maturation of osteoclast precursors and cell-cell fusion 10 . As a master regulator of osteoclastogenesis, nuclear factor of activated T cells 1 (NFATC1) is induced by RANKL, which in turn forms a complex with other transcription factors 11 and activates the transcription of its own coding gene as well as other genes involved in osteoclast adhesion, cell fusion, and bone resorption [12][13][14] .
Long noncoding RNAs (lncRNAs), transcripts that are longer than 200 nucleotides and are not translated into proteins, function through binding to DNA, other RNA, and proteins 15,16 . LncRNAs usually have low evolutionary conservation. One of the few exceptions, MALAT1, is a highly conserved nuclear lncRNA that is abundantly expressed in many tissues 17 . MALAT1 has been shown to modulate alternative pre-mRNA splicing based on siRNA knockdown results from cultured cell lines 18 . In 2012, three groups reported that Malat1-knockout mice showed no obvious phenotypic differences compared with wild-type mice under physiological conditions, and loss of Malat1 in mice did not affect alternative pre-mRNA splicing [19][20][21] . On the other hand, recent animal studies suggested that Malat1 has important functions under pathological conditions. For instance, through targeted inactivation, restoration (genetic rescue), and overexpression of Malat1 in mouse models, we found that Malat1 suppresses breast cancer lung metastasis through binding and inactivation of the Tead family of transcription factors 22 . Moreover, Malat1-null mice exhibited enhanced antiviral responses, suggesting that Malat1 may suppress antiviral innate immunity 23 . In addition, when fed a high-fat diet, Apoe -/mice transplanted with Apoe -/-;Malat1 -/bone marrow showed higher atherosclerotic plaque burden in the aorta and increased hematopoietic progenitor cells and their progeny, suggesting that Malat1 may regulate hematopoietic cells 24 .
Recent genome-wide association studies (GWAS) showed that single nucleotide polymorphisms (SNPs) are associated with osteoporosis 1,25 . Interestingly, one such analysis identified an SNP (rs202070768) at the MALAT1 locus that was associated with low BMD 26 . However, functional evidence of MALAT1 alterations having a role in low BMD and osteoporosis is lacking. In the present study, by using genetically engineered mouse models, we identified Malat1 as a negative regulator of osteoporosis and bone metastasis. Mechanistically, Malat1 binds and sequesters Tead3, blocking Tead3 from interacting with and activating Nfatc1. Consequently, loss of Malat1 derepresses Tead3, which in turn promotes Nfatc1-mediated osteoclast differentiation. Importantly, the osteoporosis and bone metastasis phenotypes of Malat1-deficient mice can be rescued by genetic add-back of Malat1. Taken together, our findings reveal that Malat1 inhibits Tead3-Nfatc1-mediated osteoclastogenesis to suppress osteoporosis and bone metastasis.

MALAT1 is downregulated during osteoclast differentiation in humans and mice
Hematopoietic stem cells (HSCs) undergo self-renewal and differentiation in the bone marrow.
During a hierarchical differentiation process, HSCs turn into multipotent progenitors (MPPs), which then differentiate into oligopotent progenitors, including common myeloid progenitors (CMPs) and common lymphoid progenitors (CLPs) 27 . Recent reports of Malat1 having a role in regulating hematopoietic cells under pathological conditions 23, 24 prompted us to analyze MALAT1 expression during differentiation of HSCs by using publicly available high-throughput sequencing datasets. Interestingly, in both humans and mice, MALAT1 was expressed at higher levels in HSCs than in MPPs or CMPs (Extended Data Fig. 1a-d).
CMPs can differentiate into monocytes and macrophages, which are the precursors of osteoclasts 28 .
We analyzed gene expression during the differentiation of human placental CD14 + macrophages into multinucleated giant cells (MGCs) 29 (Fig. 1a), in which osteoclasts are the major cell population 8 . Compared with CD14 + macrophages, MGCs showed elevated expression of osteoclast markers, including NFATC1, CTSK, DCSTAMP, ATP6V0D2, ATP6V0E2, and ATP6V0A1 (Fig. 1b, c). In contrast, MALAT1 was significantly downregulated in MGCs relative to CD14 + macrophages (Fig. 1a-c). Consistent with the functions of osteoclasts, gene set enrichment analysis (GSEA) indicated that genes upregulated in MGCs compared with CD14 + macrophages were related to collagen organization, extracellular structure remodeling, and skeletal development ( Fig. 1d and Supplementary Table 1). To further validate the downregulation of Malat1 during the differentiation of macrophages into osteoclasts, we treated a mouse macrophage/pre-osteoclast cell line, RAW264.7, with soluble RANKL to induce osteoclast differentiation 30 . After this treatment, markers of osteoclasts, including Nfatc1, Ctsk, and Trap5, were upregulated in a time-dependent manner ( Fig. 1e-g), whereas Malat1 expression levels were markedly decreased (Fig. 1h). Taken together, these results reveal MALAT1 as a lncRNA that is downregulated during osteoclastogenesis in humans and mice.
Consistent with previous reports 31, 32 , we observed that LPS treatment alone was insufficient to initiate osteoclast differentiation; instead, when RAW264.7 cells were pretreated with RANKL, LPS treatment promoted osteoclastogenesis, as gauged by staining for tartrate-resistant acid phosphatase (TRAP), a widely used marker of osteoclasts (Fig. 1i, j), and upregulated the expression of Nfatc1 and Ctsk (Fig. 1k). TNF-α, on the other hand, can induce osteoclast differentiation in both RANKL-dependent and RANKL-independent manners 34,35 . Indeed, we found that treating RAW264.7 cells with TNF-α induced osteoclastogenesis and the expression of Nfatc1 and Ctsk, either with or without RANKL pretreatment (Fig. 1i, l, m). Notably, Malat1 was substantially downregulated during LPS-or TNF-α-induced osteoclast differentiation (Fig. 1k, m).
These findings suggest that Malat1 may be involved in osteoclastogenesis in response to various stimuli.

Genetic models reveal that Malat1 protects against osteoporosis and bone metastasis
To study the role of Malat1 in osteoclastogenesis and osteoporosis in vivo, we used a Malat1knockout mouse model (Malat1 -/-) described in our previous study, in which a transcriptional terminator was inserted downstream of the transcriptional start site of Malat1, causing the loss of Malat1 RNA expression without altering expression levels of Malat1's adjacent genes 22 . Also, we previously engineered mice with targeted transgenic Malat1 expression from the ROSA26 locus (Malat1 Tg/T g ), which enabled us to conduct genetic rescue studies in Malat1 -/mice by generating Malat1 -/-;Malat1 Tg/Tg animals 22 .
By performing microcomputed tomographic (μCT) analysis of the femurs of 6-month-old animals, we found that both male and female Malat1 -/mice had much less bone mass than age-and sexmatched Malat1 +/+ mice; importantly, this osteoporotic phenotype was rescued in Malat1 -/-;Malat1 Tg/T g mice (Fig. 2a, b). Quantification of the µCT data revealed that compared with Malat1 +/+ mice, the trabecular bone density (Fig. 2c, d) and these reductions were largely reversed by genetic restoration of Malat1 expression (Fig. 2c, d and Extended Data Fig. 2a-f). Staining for TRAP revealed a significant increase in osteoclasts in the femurs of Malat1 -/mice compared with Malat1 +/+ mice, and this increase was reversed in Malat1 -/-;Malat1 Tg/Tg mice (Fig. 2e). Quantification of femoral osteoclasts showed that relative to Next, to determine the role of Malat1 in modulating pathological bone loss, we used a wellestablished mouse model of inflammatory bone resorption, which involves the injection of lipopolysaccharides (LPS) into the subcutaneous space over the calvarial bones 36 . As gauged by µCT imaging, administration of LPS to 8-week-old Malat1 -/mice resulted in significantly aggravated erosions on the surface of the calvarial bones, compared with Malat1 +/+ or Malat1 -/-;Malat1 Tg/T g mice (Fig. 2h, i). TRAP staining and quantification revealed that after LPS injection, Malat1 -/mice had higher osteoclast numbers per bone perimeter (Oc.N/B.Pm, Fig. 2j, k) and more osteoclast surface per bone surface (Oc.S/BS, Fig. 2j, l) than either Malat1 +/+ or Malat1 -/-;Malat1 Tg/T g mice. Collectively, these findings indicate that Malat1 deficiency promotes osteoporosis under both physiological and inflammatory conditions. Untreated osteoporosis is associated with accelerated progression of bone metastasis in cancer patients [3][4][5] . Drugs for osteoporosis therapy, such as bisphosphonates that inhibit osteoclastmediated bone resorption, have been used to treat bone diseases including bone metastases 37 . To determine whether Malat1 in the host confers protection from bone metastases, we injected luciferase-labeled B16F1 melanoma cells into the tibiae of 6-month-old Malat1 +/+ , Malat1 -/-, or Malat1 -/-;Malat1 Tg/Tg mice, and we found that bone metastases were markedly exacerbated by Malat1 loss in the hosts, a phenotype that was rescued by Malat1 re-expression, as gauged by bioluminescent imaging of live animals (Extended Data Fig. 2h, i) and dissected bones (Fig. 2m, n), as well as gross examination of visible bone lesions (Fig. 2o). These results indicate that Malat1 has a non-tumor cell-autonomous role in bone metastasis suppression.
Because bone homeostasis is maintained by osteoclastic bone resorption and osteoblastic bone formation, we next determined whether Malat1 modulates the number and differentiation potential of osteoblasts. To this end, we stained bone sections with toluidine blue 38 (Extended Data Fig.   3a), which revealed no significant difference in the numbers of osteoblasts per bone perimeter (N.Ob/B.Pm) among Malat1 +/+ , Malat1 -/-, and Malat1 -/-;Malat1 Tg/Tg mice (Extended Data Fig. 3a,  b). Further, we isolated mouse mesenchymal stem cells (MSCs) from these three mouse lines and cultured them in osteogenic differentiation medium 39 ; we observed comparable osteogenic differentiation in all three groups, as gauged by calcium mineralization (via alizarin red staining, Extended Data Fig. 3c) and alkaline phosphatase (ALP, Extended Data Fig. 3d, e). Moreover, we found no significant difference in bone formation rates among Malat1 +/+ , Malat1 -/-, and Malat1 -/-;Malat1 Tg/T g mice, as gauged by dynamic histomorphometry measurements through sequential labeling with calcein, a fluorescent chromophore that binds to calcified skeletal structures 40,41 (Extended Data Fig. 3f, g). Taken together, our results suggest that Malat1 inhibits osteoclast differentiation and protects against osteoporosis and bone metastasis without affecting osteoblastic bone formation.

Malat1 deficiency promotes osteoclastogenesis through the activation of Nfatc1
Because the Malat1 -/and Malat1 Tg/Tg animals used in our study are whole-body knockout and transgenic mice, the osteoporotic phenotype observed above may or may not be a direct effect of Malat1 loss in osteoclast precursors. To address this issue, we isolated primary bone marrowderived macrophages (BMMs) from Malat1 +/+ , Malat1 -/-, and Malat1 -/-;Malat1 Tg/Tg mice, and then treated these osteoclast precursors with M-CSF and RANKL for 4-6 days to induce their differentiation into osteoclasts. Genetic ablation and restoration of Malat1 expression in BMMs were confirmed by qPCR (Fig. 3a). After M-CSF-and RANKL-induced differentiation, we detected osteoclasts by TRAP staining (Fig. 3b), finding that knockout of Malat1 led to a prominent increase in the number of TRAP-positive multinucleated osteoclasts, and that reexpression of Malat1 reversed the observed induction of osteoclastogenesis (Fig. 3b). The mRNA levels of osteoclast markers Ctsk and Trap5 were much higher in Malat1 -/cells than in Malat1 +/+ and Malat1 -/-;Malat1 Tg/T g cells after RANKL treatment (Fig. 3c, d).
We also used CRISPR interference (CRISPRi) to knock down Malat1 in cell lines. Eleven single guide RNAs (sgRNAs) targeting mouse Malat1 were tested by using the mouse B16F1 cell line (Extended Data Fig. 4a). sgRNA-2 and sgRNA-3 were chosen to knock down Malat1 in RAW264.7 cells, which was validated by qPCR (Fig. 3e), and the two resulting Malat1knockdown stable cell lines were named Malat1 KD1 and Malat1 KD2 . After RANKL-induced differentiation, both Malat1 KD1 and Malat1 KD2 cells gave rise to more TRAP-positive multinucleated osteoclasts than the control RAW264.7 cells (Fig. 3f). Fluorescent staining of Factin rings (microfilament structures that are characteristic of mature osteoclasts 42, 43 ) and nuclei, by phalloidin and DAPI, respectively, revealed that Malat1 KD1 and Malat1 KD2 cells had higher numbers of nuclei per osteoclast than the control cells (Fig. 3g). Moreover, Ctsk and Trap5 mRNA levels were upregulated by knockdown of Malat1 in RANKL-treated RAW264.7 cells (Fig. 3h, i).
Collectively, the results from primary BMMs and RAW264.7 cells suggest that Malat1 deficiency in osteoclast precursors promotes RANKL-induced osteoclastogenesis.
Upon binding to the RANK receptor, RANKL stimulates multiple signaling cascades, including nuclear factor-κB (NF-κB) signaling, mitogen-activated protein kinases (MAPK) signaling, and activator protein-1 (AP-1, whose major components are c-Jun and c-Fos proteins) signaling, leading to activation of downstream transcription factors, such as Nfatc1, Mitf, and Creb 14 . To understand how Malat1 inhibits osteoclastogenesis, we first stimulated BMMs with RANKL for short periods (5-60 minutes) and examined the phosphorylation events in the signaling pathways mentioned above, finding no substantial difference in the phosphorylation levels of p65 (also known as RelA), Erk1/2, Jnk, or c-Jun among the BMMs isolated from Malat1 +/+ , Malat1 -/-, and Malat1 -/-;Malat1 Tg/Tg mice (Extended Data Fig. 4b). Thus, Malat1 loss does not affect the early kinase signaling events during RANKL-induced osteoclast differentiation.
Chromatin immunoprecipitation-qPCR assays revealed that after RANKL treatment, Malat1knockdown RAW264.7 cells showed more occupancy of these two regions by Nfatc1 than the control RAW264.7 cells (Fig. 3o, p). Importantly, knockdown of Nfatc1 in Malat1-depleted RAW264.7 cells reversed the induction of osteoclastogenesis and Ctsk expression upon stimulation with RANKL (Fig. 3q, r and Extended Data Fig. 4e). Taken together, these results suggest that Malat1 loss promotes osteoclast differentiation through Nfatc1.

Malat1 binds to Tead3 to inhibit Nfatc1 activity and osteoclastogenesis
How does Malat1 regulate Nfatc1? The binding of Nfatc1 to other proteins can lead to either activation or inhibition of the transcriptional activity of Nfatc1 47 , while lncRNAs often exert their functions by interacting with proteins, and this mode of action has been demonstrated for Malat1 15,17,22,48 . We speculated that Malat1 might regulate the Nfatc1 auto-amplification loop by interacting with Nfatc1 and/or its binding proteins, and thus we searched a database of protein-protein interactions, Mentha (http://mentha.uniroma2.it/index.php). Of all potential NFATC1-interacting proteins (Extended Data Fig. 5a), TEAD was of particular interest, because our previous chromatin isolation by RNA purification coupled to mass spectrometry (ChIRP-MS) analysis captured an endogenous Malat1-Tead interaction in mouse tissues, which was validated by ChIRP-Western, RNA pulldown, and RNA immunoprecipitation assays 22 . Hence, we hypothesized that Malat1 might regulate Nfatc1 through Tead.
To determine the role of Tead in Malat1-mediated regulation of Nfatc1, we first performed RNA immunoprecipitation assays, finding that Malat1 was enriched in pan-Tead immunoprecipitates from RAW264.7 cells (Fig. 4a), which validated the interaction between Malat1 and Tead in these osteoclast precursors. We next examined the protein levels of the four Tead family members (Tead1-4) in BMMs and RAW264.7 cells along with several other mouse cell lines. Interestingly, Tead1 and the Tead co-activator Yap were undetectable in BMMs and RAW264.7 cells but were abundantly expressed in the mouse melanoma cell line B16F1, mouse embryonic fibroblasts (MEF), mesenchymal stem cells (MSC), and the mouse fibroblast cell line L929 (Fig. 4b). In contrast, Tead3 showed a relatively specific expression pattern in BMMs and RAW264.7 cells (Fig. 4b).
To determine whether Tead3 directly binds to Malat1, we performed RNA pulldown assays with six non-overlapping biotinylated fragments of Malat1 (P1-P6; 1.1-1.2 kb each) generated by in vitro transcription 22 , and we found that all six Malat1 fragments, but not an unrelated nuclear RNA U1, bound to Tead3 protein (Fig. 4c), suggesting that the Tead3-binding sites may be distributed diffusely on Malat1. Interestingly, RANKL treatment of RAW264.7 cells upregulated Tead3, but not other Tead family members (Fig. 4d). Further, co-immunoprecipitation (co-IP) assays revealed that Nfatc1 protein interacted with Tead3 protein (Fig. 4e) but not Yap protein (Extended Data Fig. 5b).
After validating the interaction of Tead3 with Malat1 and Nfatc1, we sought to determine whether Malat1 modulates the binding of Tead3 to Nfatc1. To this end, we generated MALAT1-knockout HEK293T cells (Extended Data Fig. 6a, b) and transfected these cells with Tead3 and Nfatc1.
Co-IP assays showed that Malat1 loss significantly increased the interaction between Tead3 and Nfatc1 ( Fig. 4f, g). To further corroborate this result, we re-expressed Malat1 in MALAT1knockout HEK293T cells (Extended Data Fig. 6c), finding that restoring Malat1 expression reduced the Tead3-Nfatc1 interaction (Fig. 4h, i). These results suggest that Malat1 may bind and sequester Tead3, thereby blocking Tead3 from associating with Nfatc1.  (Fig. 4j). Tead3 protein mainly consists of two domains: an N-terminal DBD (also known as the TEA domain) and a C-terminal YAP-binding domain 49 . Accordingly, we generated truncation mutants of Nfatc1 and Tead3 (Fig. 4j) and performed co-IP assays, finding that both the N-terminal region (containing a TAD and the NHR domain) and the central DBD, but not the C-terminal TAD of Nfact1, could bind Tead3 (Fig. 4k). In addition, co-IP assays using truncated Tead3 mutants and full-length Nfatc1demonstrated that the TEA domain of Tead3, but not the YAP-binding domain, was responsible for interaction with Nfatc1 (Fig. 4l).
We further examined whether Malat1 and Tead3 modulate Nfatc1's transcriptional activity by using a luciferase reporter containing either tandem Nfatc1-binding sites or the Ctsk promoter, and we found that overexpression of Tead3 indeed enhanced the transcriptional activity of Nfatc1 (Fig.   4m, n). We then transfected Tead3 into wild-type, MALAT1-knockout, and Malat1-restored HEK293T cells, finding that Tead3 overexpression led to higher dose-dependent increases in Nfatc1 activity in MALAT1-knockout cells compared with either wild-type or Malat1-rescued cells (Fig. 4o, p). These results lend support to a model in which Malat1 loss derepresses Tead3, which in turn binds and activates Nfatc1. Finally, to determine Tead3's role in osteoclastogenesis, we knocked down Tead3 in RAW264.7 cells (Extended Data Fig. 7a), finding that Tead3 depletion attenuated RANKL-induced osteoclast differentiation (Extended Data Fig. 7b) and

Discussion
This study identified Malat1 as an osteoporosis-suppressing and bone metastasis-inhibiting lncRNA that is downregulated during RANKL-triggered osteoclastogenesis. RANKL stimulates multiple signaling pathways, most of which (such as MAPK and NF-κB pathways) can also be activated by other factors, and yet RANKL is indispensable and irreplaceable in osteoclastogenesis 14 , which could be explained by Nfatc1's role as a specific master regulator of osteoclast differentiation. As a transcriptional factor of its own coding gene and other osteoclastspecific genes, the binding of Nfatc1 to other nuclear proteins can lead to synergistic activation of gene transcription 47 , as exemplified by the AP-1 transcription factor complex, which interacts with Nfatc1 to boost the transcriptional activity of Nfatc1 50 . Here, we identified Tead3 as a macrophageosteoclast-specific Tead family member and a binding partner of Nfatc1, and our data suggest a model ( Fig. 4s) in which Malat1 binds and sequesters Tead3, blocking Tead3 from associating with Nfatc1 and inducing the transcription of Nfatc1 target genes, including Nfatc1 itself and Ctsk.
In response to RANKL stimulation, downregulation of Malat1 releases Tead3, thereby enhancing both the Tead3-Nfatc1 interaction as well as the transcription factor occupancy of Nfatc1 target genes, which leads to activation of Nfatc1-mediated gene transcription and osteoclast differentiation.
In addition to hyperactivation of osteoclastic bone resorption, suppression of osteoblastic bone formation can also contribute to low BMD and osteoporosis. Several previous publications reported that MALAT1 promotes osteoblast differentiation by acting as a competing endogenous RNA (ceRNA) to microRNAs (miRNAs), i.e., a "miRNA sponge", based on MALAT1 shRNA or siRNA knockdown in cell culture 51-55 . How a nuclear lncRNA could bind miRNAs is unclear.
Considering the potential pitfalls in using RNAi, large genomic deletion (MALAT1, a single-exon gene, is ~7 kb in mice and ~8 kb in humans), promoter deletion, and RNase H-dependent antisense oligonucleotide approaches to deplete nuclear lncRNAs 15,16,56, 57 58 , we used different methods for loss-of-function analyses of Malat1 in vitro and in vivo, including CRISPRi, double gRNAmediated focal deletion in the 5′ region (without affecting the promoter), and insertional inactivation, along with genetic rescue experiments. In the present study, loss of Malat1 in preosteoclasts (including RAW264.7 cells and primary BMMs from Malat1 -/mice) promoted osteoclastogenesis, a phenotype that could be reversed by restoration of Malat1 expression. On the other hand, the results from Malat1 +/+ , Malat1 -/-, and Malat1 -/-;Malat1 Tg/Tg mice, as well as osteoblast differentiation assays of MSCs isolated from these animals, showed no evidence for the regulation of osteoblastogenesis by Malat1 (Extended Data Fig. 3a-g).
Via its C-terminal YAP-binding domain, the TEAD family of transcription factors binds to the transcriptional co-activator YAP to turn on the expression of TEAD-YAP target genes 59 . In doing so, TEAD proteins and YAP are involved in several processes, including organ growth, regeneration, tumor progression, and metastasis 59 . Potentially druggable sites in the proteinprotein interaction between YAP and TEAD, as well as a highly conserved palmitoylation pocket in TEADs, have been identified and exploited for drug development 60 . However, whether the TEAD family can function in a YAP-independent manner is elusive. In this study, Tead3, but not other Tead family members, exhibited a specific expression pattern in macrophages/preosteoclasts, whereas Tead1, Yap, and classic Yap-Tead targets were barely detectable in these cells ( Fig. 4b and data not shown). We further found that Tead3 binds and activates Nfatc1 via its Nterminal TEA domain (but not its C-terminal YAP-binding domain), which is required for RANKL-induced osteoclastogenesis, thus revealing a previously undescribed, non-canonical function of Tead that is mediated by Nfatc1 and is controlled by Malat1 lncRNA. Our findings suggest the therapeutic potential of developing agents that disrupt the TEAD3-NFATC1 interaction for treating osteoporosis and bone metastasis.
Future studies should address the following issues: first, we found that Malat1 expression is downregulated during osteoclast differentiation; yet, how this lncRNA is regulated by proosteoclastogenic factors under physiological and pathological conditions is unknown. Second, our study identified a Malat1−Tead3−Nfatc1 axis that regulates osteoclastogenesis, but it is possible that additional binding partners of Malat1 and Tead3 could also be involved in osteoclast differentiation. Third, our previous study 22 and the present study collectively demonstrate that Malat1 binds to Tead to inactivate Yap-Tead's pro-metastatic function in lung-metastatic breast cancer cells and to inhibit Nfatc1-Tead3's pro-osteoclastogenic function in pre-osteoclasts,

Genetically engineered mouse models
All animal studies were performed in accordance with a protocol approved by the Institutional to produce Malat1 -/-;Malat1 Tg/T g mice. All mice described here were on a C57BL/6 background.
Primers for PCR genotyping were listed in our previous paper 22 .

Lipopolysaccharides (LPS)-induced inflammatory osteoporosis model
The procedure for inflammation-induced bone destruction was performed as previously described 36 . Briefly, 8-week-old female mice were injected above the calvarium with 12.5 mg/kg of LPS (Sigma, L4391) or vehicle (phosphate buffered saline, PBS). After 6 days, calvariae were collected and analyzed by micro-computed tomography (µCT), followed by embedding, sectioning, and TRAP staining. Authentication Core.

Osteoclast differentiation
Osteoclast differentiation from bone marrow-derived macrophages (BMMs) was induced as described previously 61 . Briefly, femurs, tibiae, and iliac bones were removed from mice after euthanasia. Small incisions were made at both the proximal and distal ends of the bones, and the bones were placed in a sterile tube and centrifuged at 10,000 g at room temperature for 15 s. After  (300 µg/mL). qPCR was used to verify Malat1 re-expression.

RNA interference
The siRNA Universal Negative Control (Sigma, SIC001) and siRNAs targeting mouse Tead3 or

Cytoplasmic-nuclear fractionation
Control and Malat1-knockdown RAW264.7 cells were plated in 6-cm dishes. At 12 hours after seeding, the cells were treated with soluble RANKL (50 ng/mL) for 3 days. Nuclear and cytoplasmic proteins were fractionated by using the NE-PER Nuclear and Cytoplasmic Extraction Kit (ThermoFisher Scientific, 78833) according to the manufacturer's protocol. After protein extraction, Western blot analysis was performed to detect Nfatc1 protein in the cytoplasmic and nuclear fractions. Gapdh and Lamin B1 were used as markers of the cytoplasm and the nucleus, respectively.

Protein pulldown and immunoprecipitation
HEK293FT cells were transfected with SFB (a triple-epitope tag containing S-protein, FLAG, and streptavidin-binding peptide)-tagged Nfatc1 (full-length or truncation mutants) and MYC-tagged

RNA extraction, cDNA synthesis, and quantitative PCR (qPCR)
Total RNA from cells was extracted by using a TRIzol reagent (Invitrogen, 15596026) or a PureLink RNA Mini Kit (Invitrogen, 12183018A). cDNA was synthesized from 1 µg of total RNA by using an iScript cDNA Synthesis Kit (Bio-Rad, 1708891). Real-time PCR and data collection were done with SYBR Green Supermix (Bio-Rad, 1725124) on a CFX96 instrument (Bio-Rad).
Data were normalized to Actb, Gapdh, or U6. The primer sequences are listed in Supplementary   Table 4.

RNA pulldown assay
Full-length mouse Malat1 (NR_002847) was divided into six non-overlapping pieces (P1-P6, 1. Purified RNA samples were used for cDNA synthesis, followed by qPCR analysis with the primers listed in Supplementary Table 4. U6 was used as a negative control. The results are presented as fold enrichment (normalized to IgG).
The mouse Ctsk promoter region was PCR-amplified (PCR primers are listed in Supplementary instructions.

Bioinformatic analysis
The (http://www.ncbi.nlm.nih.gov/geo/). After the probe ID was converted to the gene symbol based on the annotation of the platform, the data were normalized and an expression matrix was obtained.
The differentially expressed genes (DEGs) were identified by limma package in R language with the following cut-off values: |log 2 (fold change)| > 1 and P value < 0.001. The heatmap and volcano map of the DEGs were created by R language, and the DEGs were subjected to Gene Set Enrichment Analysis (GSEA) through clusterProfiler and enrichplot package in R 64 . Gene sets with an adjusted P value of less than 0.05 were considered significantly enriched and are listed in Supplementary Table 1.

Statistical analysis
Except for the animal experiments, all experiments were repeated two to three times. Statistical analyses were done with GraphPad Prism 9.0. Unless otherwise noted, data are presented as mean ± s.e.m., and a two-tailed unpaired t-test was used to compare two groups of independent samples.
Statistical methods used for RNA-seq analysis and Expression Atlas data analysis are described above. P < 0.05 was considered statistically significant. Competing interests. The authors declare no competing interests.
Correspondence and requests for materials should be addressed to L.M. (lma4@mdanderson.org).   Statistical significance in c, d, f, g, i, k, l, and n was determined by a two-tailed unpaired t-test.
Error bars are s.e.m.