Abstract

   Osteoblasts play an important role in the regulation of bone
   homeostasis throughout life. Thus, the damage of osteoblasts can lead
   to serious skeletal diseases, highlighting the urgent need for novel
   pharmacological targets. This study introduces chemical genetics
   strategy by using small molecule forskolin (FSK) as a probe to explore
   the druggable targets for osteoporosis. Here, this work reveals that
   transglutaminase 2 (TGM2) served as a major cellular target of FSK to
   obviously induce osteoblast differentiation. Then, this work identifies
   a previously undisclosed allosteric site in the catalytic core of TGM2.
   In particular, FSK formed multiple hydrogen bonds in a saddle‐like
   domain to induce an “open” conformation of the β‐sandwich domain in
   TGM2, thereby promoting the substrate protein crosslinks by
   incorporating polyamine. Furthermore, this work finds that TGM2
   interacted with several mitochondrial homeostasis‐associated proteins
   to improve mitochondrial dynamics and ATP production for osteoblast
   differentiation. Finally, this work observes that FSK effectively
   ameliorated osteoporosis in the ovariectomy mice model. Taken together,
   these findings show a previously undescribed pharmacological allosteric
   site on TGM2 for osteoporosis treatment, and also provide an available
   chemical tool for interrogating TGM2 biology and developing bone
   anabolic agent.

   Keywords: allosteric activation, forskolin, mitochondria, osteoblast
   differentiation, transglutaminase 2
     __________________________________________________________________

   In this study, transglutaminase 2 (TGM2) is identified as a target of
   forskolin to promote osteoblast differentiation. Forskolin induces an
   “open” conformation of TGM2 on a previously undisclosed allosteric
   site, which promotes mitochondrial fusion and ATP production for
   osteoblast differentiation. These data provide an available chemical
   tool for interrogating TGM2 and developing anti‐osteoporosis agent.

   graphic file with name ADVS-10-2206533-g002.jpg

1. Introduction

   Bone reconstruction is a hot research field in regenerative medicine.
   The balance between bone resorption and bone formation is crucial for
   maintaining bone mass and systemic mineral homeostasis.^[ [48]^1 ^]
   Accumulating evidence suggests that the mechanism of bone metabolic
   regulation is extremely complex, and abnormalities may cause bone
   metabolic disorders.^[ [49]^2 ^]

   Osteogenic processes are tightly regulated through complex networks
   activated by distinct osteogenic signaling molecules.^[ [50]^3 ^]
   Transglutaminase 2 (TGM2), is a multifunctional enzyme implicated in
   the pathogenesis of a number of human diseases, including celiac,
   neurological, and renal diseases, and various forms of cancer.^[ [51]^4
   ^] Recently, the biological functions of TGM2 have been extensively
   studied in matrix maturation and mineralization processes of bone
   matrix.^[ [52]^5 ^] For example, TGM2 is observed to be highly
   responsible for osteoblast‐like cell differentiation and osteogenic
   potential.^[ [53]^6 ^] Moreover, TGM2 knockout results in lower bone
   mass and an increased number of osteoclasts.^[ [54]^7 ^]
   Mechanistically, TGM2 contributes to the crosslinking of both intra‐
   and extracellular proteins by promoting the formation of ε‐(γ‐glutamyl)
   lysine bonds, further leading to the expression of osteogenic genes as
   well as bone formation and repair.^[ [55]^8 ^] Thus, TGM2 targeting may
   represent a promising strategy to treat osteoporosis by screening
   specific small molecules to regulate TGM2 biology. However, very few
   candidates are reported to target TGM2 for osteoblast differentiation
   and bone reconstruction.

   Chemical genetics is a promising strategy for drug‐target prediction.^[
   [56]^9 ^] Natural products possess diversified biological activities
   and have become a key resource for drug development.^[ [57]^10 ^] Thus,
   it is worth discovering naturally derived osteogenic compounds to
   explore their pharmacological targets, thereby clarifying novel
   osteoblast differentiation mechanisms. Forskolin (FSK) is a unique
   diterpenoid from the medicinal plant Coleus forskohlii.^[ [58]^11 ^]
   FSK has been previously demonstrated to induce the differentiation of
   adipose‐derived stem cells into Schwann cell‐like cells.^[ [59]^12 ^]
   Meanwhile, FSK also stimulates osteogenic differentiation, suggesting a
   potential therapeutic effect on bone diseases such as osteoporosis.^[
   [60]^13 ^]

   In this study, we synthesized a photoaffinity probe through introducing
   a diazirine photocrosslinker and a click chemistry handle in FSK. Then,
   chemical genetics revealed that TGM2 functioned as a direct target
   protein of FSK by a previously undisclosed allosteric activation
   mechanism via inducing an “open” conformation of the N‐terminal
   β‐sandwich domain. This resulted in an improvement in mitochondrial
   energy‐generating processes for osteoblast differentiation. Taken
   together, we elucidate a druggable hotspot in TGM2 for promoting
   osteoblast differentiation, and also highlight a previously unknown
   allosteric activation role for TGM2 as well as potential translational
   value in the development of bone anabolic agent.

2. Results

2.1. FSK Promotes the Osteoblastic Differentiation of MC3T3‐E1 Cells

   Osteoblast differentiation is an effective strategy to promote bone
   formation. Thus, we investigated the ability of FSK to promote
   osteogenic differentiation in pre‐osteoblastic MC3T3‐E1 cells. Alkaline
   phosphatase (ALP) is a crucial osteoblastic differentiation biomarker
   that is highly associated with osteogenesis.^[ [61]^14 ^] In MC3T3‐E1
   cells, we observed that FSK (0.5, 1, and 5 µm) significantly increased
   ALP activity (Figure [62]1A) during 7 days of treatment. Moreover, the
   mineralization assay by Alizarin red staining indicated that FSK (0.5,
   1, and 5 µm) obviously increased the mineralization node area and the
   absorbance of the alizarin red dye in MC3T3‐E1 cells after 21 days of
   treatment (Figure [63]1B,C). Furthermore, we confirmed these
   observations in primary osteoblasts from mouse skulls. Consistent with
   MC3T3‐E1 cells, FSK (0.5, 1, and 5 µm) significantly increased ALP
   activity (Figure [64]1D) and mineralization nodes (Figure [65]1E,F) in
   primary osteoblasts. Meanwhile, quantitative real‐time PCR (Rt‐PCR)
   demonstrated that FSK (0.5, 1, and 5 µm) upregulated the expression of
   several osteogenesis markers, including Ocn, Ranx2, Alp, and Col1a1 ^[
   [66]^15 ^] (Figure [67]1G–J). In addition, FSK slightly increased the
   viability of MC3T3‐E1 cells, indicating no obvious cytotoxicity (Figure
   [68]S5A, Supporting Information). Taken together, these results reveal
   that FSK promotes osteoblastic differentiation and accelerates bone
   mineralization.

Figure 1.

   Figure 1
   [69]Open in a new tab

   FSK promotes the osteoblastic differentiation of MC3T3‐E1 cells and
   primary osteoblasts. A) FSK increased the ALP activity of MC3T3‐E1
   cells. B,C) FSK upregulated the mineralization of MC3T3‐E1 cells by
   Alizarin red staining and quantification. D) FSK increased the ALP
   activity of primary osteoblasts. E,F) FSK upregulated the
   mineralization of primary osteoblasts by Alizarin red staining and
   quantification. G‐J) FSK upregulated the expression of osteoblasts
   differentiation marker genes in MC3T3‐E1 cells. Data are expressed as
   mean ± SEM for 3–6 individual experiments. ***p < 0.001, **p < 0.01, *p
   < 0.05 versus control group.

2.2. TGM2 Serves as a Cellular Target of FSK

   To identify the cellular target of FSK, we synthesized a FSK molecular
   probe (AD‐FSK) by introducing a difunctional side chain (AD) with a
   diazirine and an alkyne (Figure [70]2A). Diazirine was used for
   photoaffinity labeling of target proteins in situ by forming covalent
   bonds. Meanwhile, the alkyne was conjugated with the reporter
   biotin‐PEG‐azide or 5‐TAMRA‐azide via a click reaction. The chemical
   structure of AD‐FSK was verified by ^1HNMR, ^13CNMR spectra and
   high‐resolution mass spectrometry (Figures [71]S1–S3, Supporting
   Information), which suggested that the difunctional side chain was
   attached to the hydroxyl on the C1 site in FSK skeleton (Figure [72]S4,
   Supporting Information). To verify the biological activity of AD‐FSK,
   ALP activity was tested by AD‐FSK treatment and the result demonstrated
   that AD‐FSK significantly upregulated ALP activity in MC3T3‐E1 cells
   (Figure [73]2B). Meanwhile, AD‐FSK showed a mineralization promotion
   effect similar to that of FSK in MC3T3‐E1 cells (Figure [74]2C,D).
   Therefore, these results indicate that AD‐FSK possesses similar
   biological activity to FSK and is suitable for target identification.

Figure 2.

   Figure 2
   [75]Open in a new tab

   TGM2 serves as a cellular target of FSK. A) Schematic representation of
   the AD‐FSK synthesis. B,C) AD‐FSK upregulated the mineralization of
   MC3T3‐E1 cells by Alizarin red staining and quantification. D) AD‐FSK
   increased the ALP activity of primary osteoblasts. E) Volcano plot of
   the cellular targets of FSK. F) SPR analysis of FSK‐binding to TGM2. G)
   Pull‐down analysis of FSK‐binding to TGM2 in MC3T3‐E1 cells. H) FSK
   promoted resistance of TGM2 to different temperature gradients (CETSA).
   I) FSK promoted resistance of TGM2 to proteases (DARTS). Data are
   expressed as mean ± SEM for 3–4 individual experiments. ***p < 0.001,
   **p < 0.01, *p < 0.05 versus control group.

   Next, we conducted in‐gel fluorescence analysis to determine the
   protein labeling capacity of AD‐FSK. MC3T3‐E1 cells were incubated with
   AD‐FSK and exposed to UV irradiation. The cell lysates were then
   conjugated to 5‐TAMRA‐azide via click reaction. In‐gel fluorescence
   scanning showed that the fluorescence intensity of labeled proteins was
   significantly blocked with an excess amount of FSK for competition
   (Figure [76]S5B, Supporting Information), indicating that AD‐FSK can be
   used to capture the protein targets in cells. Next, MC3T3‐E1 cells were
   treated with AD‐FSK in the presence or absence of FSK. After UV
   irradiation, cell lysates were prepared and biotin‐PEG‐azide was
   conjugated to AD‐FSK via a click reaction. Then, the AD‐FSK‐captured
   proteins were pulled down by streptavidin beads for liquid
   chromatography tandem‐mass spectrometry (LC‒MS/MS) analysis. A total of
   821 proteins were identified, and transglutaminase 2 (TGM2) possessed
   the highest significance of differences under the competition of FSK
   (Figure [77]2E). In particular, previous reports indicated that TGM2 is
   highly associated with bone formation and repair;^[ [78]^16 ^] thus, we
   speculated that TGM2 may serve as a cellular target of FSK to promote
   osteogenic differentiation. Furthermore, surface plasmon resonance
   (SPR) analysis indicated that FSK directly bound to TGM2 (K [D] =
   3.1 µm, Figure [79]2F). We also observed that TGM2 was pulled down by
   AD‐FSK, which was obviously blocked by an excess amount of FSK for
   competition (Figure [80]2G). Meanwhile, a cellular thermal shift assay
   (CETSA)^[ [81]^17 ^] revealed that FSK enhanced the thermal stability
   of TGM2 (Figure [82]2H). Drug affinity responsive target stability
   (DARTS)^[ [83]^18 ^] experiments indicated that FSK stabilized TGM2 by
   increasing resistance to proteolysis (Figure [84]2I). Collectively,
   these results demonstrate that FSK directly binds to TGM2 in MC3T3‐E1
   cells.

2.3. FSK Binds to TGM2 in the Catalytic Core Domain

   To determine the FSK‐binding region in TGM2, TGM2 protein was incubated
   with FSK, followed by mild proteolysis. Given that the FSK‐binding
   region was more stabilized, the residual peptides (non‐enzymatic
   hydrolysis) protected by FSK were then identified by LC‒MS/MS analysis.
   As a result, we determined 11 potential FSK‐protecting peptides, 8 of
   which were located in the catalytic core domain of TGM2. Of note, the
   3D model of TGM2 showed that these peptides were spatially adjacent
   (Figure [85]3A), indicating that FSK‐binding region was mainly located
   in the TGM2 catalytic core domain. To further corroborate these
   observations, we established four TGM2 truncations based on different
   domain deletions^[ [86]^19 ^] (Figure [87]3B). SPR analysis
   demonstrated that FSK did not bind to C‐terminal β‐barrel domain
   truncation (TGM2‐B) (K [D] > 100 µm). Meanwhile, the binding ability of
   FSK with N‐terminal β‐sandwich domain truncation (TGM2‐S) (K [D] =
   78 µm), catalytic core domain fusion with C‐terminal β‐barrel domain
   (TGM2‐CB) (K [D] = 22 µm), and catalytic core domain fusion with
   N‐terminal β‐sandwich domain (TGM2‐SC) (K [D] = 5.8 µm) increased
   successively (Figure [88]3C,F). Thus, FSK showed a strong binding
   ability to TGM2‐SC and a moderate binding ability to TGM2‐CB and
   TGM2‐S, indicating that FSK may prefer to bind to TGM2 in the catalytic
   core domain. Thus, these experiments provided valuable evidence to
   further explore the FSK‐binding sites by exact analytical method.

Figure 3.

   Figure 3
   [89]Open in a new tab

   FSK binds to TGM2 in the catalytic core domain. A) Mapping of FSK
   potential binding peptides to TGM2. B) Diagram of the establishment of
   four TGM2 truncations. C–F) SPR analysis of the binding ability of FSK
   with TGM2 truncations. G) Schematic diagram of capturing photoaffinity
   labeling peptide by AD‐FSK. H) LC‐MS/MS analysis of photoaffinity
   labeling peptides by AD‐FSK in TGM2. I) Docking model of FSK into the
   structure of TGM2 (PDB: 1KV3). Hydrogen bonds are shown as yellow
   dashed lines (N229, T368). J,K) SPR analysis for the binding of FSK
   with TGM2 N229A and TGM2 T368A mutants.

   Next, to identify the detailed FSK‐binding sites in TGM2, AD‐FSK was
   incubated with TGM2 protein for UV irradiation. Then, AD‐FSK was
   covalently crosslinked with TGM2 at the binding sites by photoaffinity
   diazirine (Figure [90]3G). The AD‐FSK‐TGM2 complex was further
   determined by LC‒MS/MS, and we found that AD‐FSK covalently modified
   the peptide (residues 355–369) at glutamine (Q) 355 and alanine (A) 356
   (Figure [91]3H). As the previous description, AD‐FSK possessed similar
   biological activity with FSK (Figure [92]2B–D) and FSK decreased TGM2
   pulled down by AD‐FSK for competition (Figure [93]2G), indicating both
   FSK and AD‐FSK interacted with TGM2 in the same sites.

   To map the binding reactions between FSK and TGM2, we docked FSK into
   the structure of TGM2 (PDB code: 1KV3)^[ [94]^20 ^] around residues
   355–369. In the docking 3D model (Figure [95]3I), two hydrogen bonds
   were formed between FSK and asparagine (N) 229 (2.8 Å) or threonine (T)
   368 (3.3 Å). Notably, both of these residues were located in the
   catalytic core domain. Furthermore, N229 and T368 were mutated to
   alanine (A). SPR analysis demonstrated that FSK showed a weaker ability
   to bind to TGM2 N229A (K [D] = 26 µm, Figure [96]3J) or TGM2 T368A (K
   [D] = 64 µm, Figure [97]3K) than TGM2 wild type (K [D] = 3.1 µm).
   Therefore, these results further support that FSK binds to TGM2 by
   forming hydrogen bonds with N229 and T368 in the catalytic core domain.

2.4. FSK Activates TGM2 via Allosterically Inducing an “Open” Conformation

   To explore the effect of FSK on TGM2 biological activity, we tested the
   transglutaminase activity of TGM2.^[ [98]^21 ^] As a result, we found
   that FSK obviously increased the content of labeled proteins for
   transamidation reaction catalyzed by TGM2 with an EC[50] of 2.2 µm
   (Figure [99]4A, Figure [100]S6A, Supporting Information). Moreover, we
   tested the transglutaminase activity in different concentration
   substrate 5‐(biotinamido) pentylamine (BPA). We found the
   substrate‐activity curve was sigmate and FSK treatment made the curve
   move to the upper left, suggesting that FSK probably induced TGM2
   activation via an allosteric mechanism (Figure [101]4B, Figure
   [102]S7B, Supporting Information). Next, to confirm the potential
   allosteric effect of FSK on TGM2 structure, we performed tryptophan
   fluorescence scanning of TGM2. We observed that TGM2 fluorescence
   intensity was significantly decreased by FSK (Figure [103]4C),
   indicating an obvious conformational change. Additionally, molecular
   dynamics simulation (MDS) analysis was conducted on both “open” and
   “close” TGM2 confirmations. We found that FSK induced the conformation
   change of TGM2 “close” structure (PDB:1kv3) change (Gibbs free energy
   was −8.65 kcal mol^−1) in the N‐terminal β‐sandwich domain,
   particularly in residues 38–42, 89–102 and 117–121. Meanwhile, the
   N‐terminal β‐sandwich domain moved away from the transglutaminase
   catalytic pocket to promote the formation of more exposed domains
   (Figure [104]4D). It is well known that the catalytic pocket of TGM2,
   comprising D358, H335, C277,^[ [105]^22 ^] is buried inside in “close”
   conformation and exposed to the surface in “open” conformation.^[
   [106]^23 ^] Then, MDS analysis between FSK and “open” conformation
   (PDB:2q3z) (Gibbs free energy was −16.38 kcal mol^−1) indicated that
   FSK did not further induce obvious conformation change (Figure
   [107]S6F, Supporting Information). Next, SPR analysis showed that the K
   [D] of FSK to “open” conformation (86 µm) is higher than that (0.72 µm)
   in “close” conformation (Figure [108]S6G, Supporting Information).
   Therefore, these observations indicated that FSK was inclined to bind
   to TGM2 in “close” conformation as a transition state, thereby
   dynamically inducing the “close” conformation change into the “open”
   conformation of TGM2.

Figure 4.

   Figure 4
   [109]Open in a new tab

   FSK activates TGM2 via allosterically inducing an “open” conformation.
   A) FSK activated transglutaminase activity of TGM2 (EC[50] = 2.2 µm).
   B) Transglutaminase activity of TGM2 in different concentration
   substrate BPA. C) Tryptophan fluorescence scanning analysis of TGM2. D)
   3D model of TGM2 allosteric regulation by molecular dynamics
   simulation. The allosteric N‐terminal β‐sandwich domain is emphasized
   by the dotted rectangle. E) FlAsH‐based FRET assay in cells with or
   without FSK treatment (n = 14 cells). F) Schematic diagram of FRET
   assay and FSK‐induced TGM2 “open” conformation. E) FSK allosterically
   regulated TGM2 conformation by HDXMS. Peptides with increased deuterium
   uptake ratio after FSK treatment are highlighted in blue.

   To verify the existence of the “close‐to‐open” conformation change, we
   designed a TGM2 fluorescence resonance energy transfer (FRET) plasmid
   containing enhanced cyan fluorescent protein (ECFP) on the N‐terminal
   β‐sandwich domain and a 12‐amino acid peptide (FLNCCPGCCMEP) on the
   C‐terminal β‐barrel2 domain. Fluorescein arsenical hairpin binder
   (FlAsH) can specifically bind on the 12‐amino acid peptide. We found
   that FSK treatment significantly reduced the FRET ratio of FlAsH to
   ECFP (Figure [110]4E), indicating that the distance between N‐terminal
   β‐sandwich domain and C‐terminal β‐barrel2 domain was significantly
   increased by FSK. Furthermore, hydrogen deuterium exchange mass
   spectrometry (HDXMS) was performed to support the allosteric regulation
   mechanism. As a result, FSK treatment increased the hydrogen/deuterium
   exchange levels of seven specific peptides (peptides 60–69, 167–174,
   179–185, 221–229, 309–326, 355–363, and 622–628, Figure 4G, Figure
   [111]S6C‐E, Supporting Information). Noteworthy, peptides 221–229 and
   355–363 are closed to the catalytic pocket (red residue), and they
   showed higher H/D exchange rates upon FSK treatment (Figure [112]4G),
   indicating that TGM2 catalytic pocket was induced to be exposed.
   Besides, peptides 60–69 and 622–628 with higher H/D rates also
   suggested these areas were exposed on TGM2 surface during allosteric
   progress (Figure [113]4G).

2.5. Effects of TGM2 Knockdown on FSK‐Induced MC3T3‐E1 Cell Differentiation

   To test the potential role of TGM2 in MC3T3‐E1 cell differentiation by
   FSK, TGM2 was knocked down by shTGM2 lentivirus transfection (Figure
   [114]S7H, Supporting Information). In MC3T3‐E1 cells, TGM2 knockdown
   significantly inhibited ALP activity that was promoted by FSK (Figure
   [115]5A). Moreover, Alizarin red staining demonstrated that the area of
   FSK‐mediated mineralization nodes and the absorbance of the alizarin
   red dye extracted from the mineralization nodes were not significantly
   changed in TGM2 knockdown MC3T3‐E1 cells (Figure [116]5B,C). Meanwhile,
   TGM2 knockdown inhibited FSK‐induced Col1a1, Ocn, Ranx2, and Alp genes
   expression (Figure [117]5D–G, Supporting Information). Thus, these
   results confirm that FSK accelerates osteogenic differentiation by
   targeting TGM2.

Figure 5.

   Figure 5
   [118]Open in a new tab

   Effects of TGM2 knockdown on FSK‐induced MC3T3‐E1 cell differentiation.
   A) Knockdown of TGM2 reversed the FSK‐promoted ALP activity in MC3T3‐E1
   cells. B,C) Knockdown of TGM2 reduced the promotion of FSK on
   mineralization in MC3T3‐E1 cells. D–G) TGM2 knockdown inhibited the
   promotion of FSK on the expression of osteogenesis marker genes
   including Col1a1, Ocn, Ranx2 and Alp in MC3T3‐E1 cells. Data are
   expressed as mean ± SEM for 3–6 individual experiments. ***p < 0.001,
   **p < 0.01, *p < 0.05.

   Interestingly, it was reported that FSK activated adenylate cyclase at
   picomolar range to increase cAMP content,^[ [119]^24 ^] indicating that
   FSK may promote osteoblast differentiation by targeting adenylate
   cyclase.^[ [120]^25 ^] Here, CETSA analysis suggested that FSK showed
   weak binding capacity with adenylate cyclase than TGM2 in MC3T3 cells
   (Figure [121]S7I, Supporting Information). Meanwhile, no adenylate
   cyclase signal was detected in photocrosslinking analysis in MC3T3
   cells (Figure [122]2E), indicating that FSK may show varying
   targets‐binding abilities in different situations. To further explore
   the role of adenylate cyclase in FSK‐mediated osteoblast
   differentiation, MC3T3‐E1 cells were treated with adenylate cyclase
   inhibitor SQ22536^[ [123]^26 ^] to block adenylate cyclase function. We
   found that FSK still significantly promoted ALP activity (Figure
   [124]S7A, Supporting Information), mineralization nodes number (Figure
   [125]S7B,C, Supporting Information), and the expression of osteoblast
   differentiation genes (Figure [126]S7D–G, Supporting Information) in
   the presence of SQ22536, indicating that adenylate cyclase was only
   partly involved in FSK‐mediated osteoblast differentiation.

2.6. Mitochondrial Homeostasis Contributes to TGM2‐Mediated MC3T3‐E1 Cell
Differentiation

   To investigate the potential biological mechanism of TGM2‐mediated
   MC3T3‐E1 cell differentiation, ascorbate peroxidase 2 (APEX2)^[
   [127]^27 ^] was performed to evaluate the interaction proteins with
   TGM2 (Figure [128]6A). SILAC‐labeled cells were transfected with
   TGM2‐APEX2 fusion plasmid. After FSK treatment, proximity‐dependent
   biotin labeling of proteins via APEX2 was enriched with streptavidin
   beads, followed by LC‒MS/MS analysis (Figure [129]6B). Then, we
   identified a total of 101 TGM2‐interacting proteins in volcano plots
   (Figure [130]6C). Red dots represented the interacting proteins
   upregulated (high H/L ratio) or downregulated (low H/L ratio) by FSK
   treatment with significant differences. Particularly, we observed that
   an amount of TGM2 interaction proteins were involved in mitochondrial
   signaling to regulate energy metabolism. Moreover, KEGG analysis
   (Figure [131]6D) indicated that a variety of pathways including
   propanoate metabolism, central carbon metabolism, pyruvate metabolism,
   thyroid hormone, gap junction and RNA degradation were markedly altered
   by FSK treatment. Thus, we hypothesized that FSK activated TGM2 to
   promote energy metabolism in mitochondria to drive osteogenesis. To
   this end, we detected the mitochondrial morphology in MC3T3‐E1 cells by
   Mito‐Tracker staining. As a result, The mean mitochondrial branches
   became longer from 0.822 to 1.258 µm (p < 0.05) in the presence of 5 µm
   FSK in MC3T3‐E1 cells (Figure [132]6E,F). Moreover, Rt‐PCR analysis
   revealed that the mitochondrial DNA (mtDNA) content in MC3T3‐E1 cells
   was significantly upregulated by FSK treatment (Figure [133]6G).
   Meanwhile. FSK induced ATP production in MC3T3‐E1 cells
   (Figure [134]6H). Furthermore, the mitochondrial morphology was
   detected with TGM2 knockdown. As shown in Figure [135]6I,J, TGM2
   knockdown obviously blocked the mitochondrial morphology change
   mediated by FSK in MC3T3‐E1 cells. Meanwhile, TGM2 knockdown also
   inhibited FSK‐upregulated mitochondrial biomarker gene expression
   including Pgc‐1α ^[ [136]^28 ^] and Atp5b ^[ [137]^29
   ^](Figure [138]6K,L). Besides, FSK promoted TGM2 to colocalize with
   mitochondria (Figure [139]S8A, Supporting Information). Collectively,
   these results indicate that FSK promotes energy metabolism for
   osteogenic differentiation by upregulating mitochondrial function in
   MC3T3‐E1 cells.

Figure 6.

   Figure 6
   [140]Open in a new tab

   Mitochondrial homeostasis contributes to TGM2‐mediated MC3T3‐E1 cell
   differentiation. A) Schematic diagram of the APEX2‐mediated proximity
   labeling strategy. B) Flow chart of TGM2‐interacting protein
   identification. C) Volcano plots of TGM2‐interacting proteins by FSK
   treatment. Red dots represent the proteins with significant
   differences. D) Network diagram of pathway enrichment analysis of
   TGM2‐interacting proteins. E,F) FSK promoted mitochondrial branch
   length in MC3T3‐E1 cells. White arrows indicate the representative
   region. G) FSK upregulated the mtDNA content in MC3T3‐E1 cells. H) FSK
   increased ATP content in MC3T3‐E1 cells. I,J) Knockdown of TGM2
   reversed the FSK‐mediated increase in mitochondria branch length. White
   arrows indicate the representative region. K,L) The relative expression
   of the mitochondrial biomarker genes Pgc‐1α and Atp5b with TGM2
   knockdown. Data are expressed as mean ± SEM for 3–6 individual
   experiments. ***p < 0.001, **p < 0.01, *p < 0.05.

   To explore the promotion effect of adenylate cyclase on mitochondrial
   homeostasis, adenylate cyclase inhibitor SQ22536 was treated in
   MC3T3‐E1 cells. It was found that FSK still significantly promoted the
   mean mitochondrial branch length (Figure [141]S8D,E, Supporting
   Information), content of mtDNA (Figure [142]S8G, Supporting
   Information), relative expression of mitochondrial marker genes (Figure
   [143]S8B,C, Supporting Information), and ATP concentration (Figure
   [144]S8F, Supporting Information) in the presence of SQ22536,
   indicating that FSK may promote mitochondrial energy metabolism via an
   independent mechanism rather than cAMP pathway.

2.7. FSK Alleviates Osteoporosis in the Ovariectomy Mice Model

   To reveal that FSK may promote the osteoblastic differentiation in vivo
   for translational medicine study, we established the
   ovariectomy‐induced osteoporosis mice model (OVX), which were treated
   by intraperitoneal injection once 2 days with FSK and alendronate
   sodium (ALN), a clinical anti‐osteoporosis agent for 16 weeks (Figure
   [145]7A). The dose of FSK and ALN were designed according to previous
   reports.^[ [146]^30 ^]

Figure 7.

   Figure 7
   [147]Open in a new tab

   FSK alleviates osteoporosis in the ovariectomy mice model. A)
   Intraperitoneal injection FSK or ALN once every 2 days for 16 weeks
   after ovariectomy. B) Representative micro‐CT and reconstructed 3D
   images of femora in ovariectomy osteoporosis mice on week 16. C–G)
   Corresponding measurements of bone mineral density (BMD), relative bone
   volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th),
   and trabecular separation (Tb.Sp). H,I) Representative H&E staining
   images showing the osteoblast number. Black arrows indicate
   osteoblasts. J,K) Representative OCN immunohistochemical images showing
   the number of mature osteoblasts. Black arrows indicate OCN positive
   region. L,M) Representative COX4 immunohistochemical images showing
   cell energy metabolism. Black arrows indicate COX4 positive region.
   Data are expressed as mean ± SEM for 4–6 individual experiments. ***p <
   0.001, **p < 0.01, *p < 0.05, ^# p < 0.05, ^## p < 0.01, ^### p <
   0.001.

   By micro‐CT analyzing, we found that bone mineral density, relative
   bone volume, trabecular number, and trabecular thickness significantly
   decreased and trabecular separation increased than sham group,
   suggesting a typical osteoporosis symptom. Importantly, FSK
   significantly improved bone mineral density, relative bone volume,
   trabecular number, trabecular thickness, and reduced trabecular
   separation (Figure [148]7B–G). Furthermore, Micro‐CT analysis showed a
   thinner cortical bone in OVX mice than sham group, and FSK
   administration alleviated OVX‐induced bone loss (Figure [149]S9A,B,
   Supporting Information). In addition, H&E staining of femora slides
   showed that FSK reversed the decrease of osteoblast number in bone
   surface (Figure [150]7H,I). Moreover, the osteocalcin (OCN) level, a
   marker of mature osteoblast, was significantly increased upon FSK
   treatment (Figure [151]7J,K), suggesting that FSK improved osteoporosis
   by promoting osteoblast differentiation.

   To explore bone formation in mice, the procollagen I N‐terminal
   propeptide (PINP) concentration in serum was detected by ELISA. As a
   result, the PINP concentration decreased significantly of OVX mice than
   sham group (p < 0.001) and was effectively recovered treated with FSK
   (p < 0.001) (Figure [152]S9G, Supporting Information). Similarly, FSK
   treatment significantly increased the PINP positive region in OVX mice,
   indicating obvious promotion effect of FSK for bone formation (Figure
   [153]S9C, Supporting Information). Indicating obvious promotion effect
   of FSK for bone formation. Meanwhile, cathepsin K (CTSK) positive
   osteoclasts were increased in OVX mice than sham group but there was no
   significant difference on the expression of CTSK with FSK treatment in
   the bone marrow of OVX mice (Figure [154]S9D, Supporting Information),
   which demonstrated that osteoblasts but not osteoclasts were
   responsible for the anti‐osteoporosis effect of FSK.

   We also observed that a mitochondria marker cytochrome C oxidase 4
   (COX4) in bone marrow noticeably increased with FSK treatment
   (Figure [155]7L,M). These results suggested that FSK promoted
   mitochondrial energy metabolism in bone marrow for osteoblast
   differentiation. Furthermore, FSK treatment significantly promoted the
   expression of mitofusin 2 (MFN2) and citrate synthase (CS) in bone
   marrow, indicating FSK improved energy metabolism by inducing
   mitochondrial fusion and enhancing tricarboxylic acid cycle. (Figure
   [156]S10A,B, Supporting Information). Moreover, no obvious change on
   p‐CREB expression with FSK treatment, indicating that FSK may tend to
   mainly mediate TGM2 function for activating downstream
   mitochondria‐dependent anabolic response rather than cAMP signal
   (Figure [157]S10C, Supporting Information).

   Of note, although it was reported that TGM2 activation was associated
   with the development of cancer,^[ [158]^31 ^] celiac disease^[ [159]^32
   ^] and neurodegenerative diseases,^[ [160]^33 ^] no obvious tumors,
   diarrhea, and cognitive behavioral impairments were found in mice
   during FSK administration. Moreover, H&E staining of livers and kidneys
   in mice were performed and there was no obvious pathological change in
   cytomorphology (Figure [161]S9E,F, Supporting Information). These
   results confirmed that FSK‐mediated TGM2 activation did not produce
   serious side effects.

3. Discussion

   Osteoporosis is the most prevalent bone disease. Osteoblast malfunction
   induces bone formation abnormalities, resulting in low bone mass and
   fragile fractures in bone diseases.^[ [162]^34 ^] Therefore, the
   development of novel drugs that induce osteoblast differentiation and
   maturation may help to maintain bone homeostasis and control skeletal
   development, formation and remodeling.^[ [163]^35 ^] However, there is
   a serious lack of effective therapeutic targets for this pathological
   process. In this study, through a chemical genetic study using FSK as a
   chemical probe, we identified a unique allosteric site on TGM2 for
   inducing its enzyme activity that promotes osteoblast differentiation.

   Currently, TGM2 inhibitors have been widely reported to show anticancer
   effects such as GK921^[ [164]^36 ^] However, the specific TGM2
   activator is still unknown. Here, we revealed FSK as a selective TGM2
   activator to promote osteogenesis. Mechanistically, FSK exhibited an
   allosteric effect on TGM2 by inducing a previously unexpected “open”
   conformation. We speculated that FSK‐mediated “open” conformation may
   form an efficient substrate‐recognition platform on TGM2, thereby
   leading to accelerated capture and process of substrate proteins. Of
   note, we further elucidated a pharmacological binding pocket of FSK at
   a cave with N229 and T368, which was located in a saddle‐like surface.
   To our knowledge, this is the first reported allosteric activation site
   on TGM2 for inducing osteoblast differentiation, and can be used for
   further lead compound screening to develop novel TGM2 activators.

   As a natural‐derived small molecule, FSK may interact multiple target
   proteins to exert its biological activities. Therefore, FSK may
   potentially promote osteoblast differentiation via multiple cellular
   targets such as TGM2 and adenylate cyclase. In the present study, we
   preferred to exploring new druggable targets that contribute osteoblast
   differentiation via using FSK as a chemical probe. Here, TGM2 was
   identified as a crucial cellular target of FSK for osteoblast
   differentiation, which provides a potential direction to deplore new
   drug candidate for osteoporosis.

   Mitochondria play a crucial role in bone metabolism and bone diseases
   by regulating the biology of bone tissue cells.^[ [165]^37 ^]
   Increasing evidence shows that mitochondria provide a place for ATP
   production to drive osteoblast differentiation.^[ [166]^34 , [167]^38
   ^] Here, we showed that TGM2 knockdown significantly reversed
   FSK‐induced mitochondrial fusion as well as ATP production, indicating
   that TGM2 played a potential role in mitochondrial function
   maintenance. Moreover, we employed APEX2‐based proteomics to identify
   several mitochondrial dynamics or energy metabolism‐associated
   functional proteins including HADHA, SLC25A6 and ATP5C. Previous
   studies have demonstrated that these proteins mainly contribute to
   mitochondrial β‐oxidation,^[ [168]^39 ^] ATP/ADP transfer^[ [169]^40 ^]
   and ATP synthesis,^[ [170]^41 ^] thereby indicating a potential
   biological relationship between TGM2 and mitochondrial energy
   synthesis. Therefore, pharmacological targeting of the
   TGM2/mitochondrial signaling pathway may serve as a promising strategy
   to lead compound discovery for bone metabolism regulation.

4. Conclusion

   In this study, we identified TGM2 as the direct cellular target of FSK
   to induce osteoblast differentiation. In particular, we revealed an
   allosteric activation site in the catalytic core of TGM2. FSK induced
   an unexpected “open” confirmation of the N‐terminal β‐sandwich domain,
   which accelerated the substrate protein crosslinks. We also
   systematically explored the interactome of TGM2 for interrogating the
   mitochondrial energy generation mechanism in osteoblast differentiation
   and proved anti‐osteoporosis effect of FSK in vivo. Therefore, these
   findings provide a fresh perspective for exploring TGM2 biology in
   osteoblast differentiation and enable new opportunities for innovative
   drug development for osteoporosis.

5. Experimental section

   Supporting Materials and Methods, Supporting Data are provided in
   Supporting Information.

Conflict of Interest

   The authors declare no conflict of interest.

Supporting information

   Supporting Information
   [171]Click here for additional data file.^ (1.8MB, pdf)

Acknowledgements