Conditional Knockout of PDK1 in Osteoclasts Suppressed Osteoclastogenesis and Ameliorated Prostate Cancer-Induced Osteolysis

Background: The development and maintenance of normal bone tissue is supported by balanced communication between osteoblasts and osteoclasts. The invasion of cancer cells disrupts this balance, leading to osteolysis. As the only bone resorption cells in vivo, osteoclasts play important roles in cancer-induced osteolysis. However, the function of 3-phosphoinositide–dependent protein kinase-1 (PDK1) in osteoclast resorption remains unclear. Methods: In our study, we used a receptor activator of nuclear factor-kappa B (RANK) promoter ‐ driven Cre ‐ LoxP system to conditionally delete the PDK1 gene in osteoclasts in mice. We investigated the impact of Osteoclast ‐ specic knockout of PDK1 on prostate cancer-induced osteolysis. Bone marrow-derived macrophage cells (BMMs) were extracted and induced to differentiate osteoclasts in vitro to examine the function of PDK1 in osteoclasts. Results: In this study, we found that PDK1 conditional knockout (cKO) mice exhibited smaller body sizes when contrasted with the wild-type (WT) mice. Moreover, deletion of PDK1 in osteoclasts ameliorated osteolysis and reduced bone resorption markers in the murine model of prostate cancer-stimulated osteolysis. In vivo, we discovered that osteoclast ‐ specic knockout of PDK1 suppressed RANKL-stimulated bone resorption function, osteoclastogenesis, and osteoclast-specic gene expression (Ctsk, TRAP, MMP-9, NFATc1). Western blot analyses of RANKL-induced signaling pathways showed that conditional knockout of PDK1 in osteoclasts inhibited the early nuclear factor κB (NF-κB) activation, which consequently suppressed the downstream induction of NFATc1. Conclusion: These ndings demonstrated that PDK1 performs an instrumental function in osteoclastogenesis and prostate cancer-induced osteolysis by modulating the PDK1/AKT/NF-κB signaling pathway.

importance to further develop effective drugs to reduce the incidence of SREs associated with bone metastasis in cancer.
An osteoclast is known to be a multinucleated and large cell, which is differentiated from macrophage/monocyte lineage cells by the receptor activator of nuclear factor-κB ligand (RANKL) and macrophage colony-stimulating factor (M-CSF) [9] . The binding of RANK to RANKL activates the nuclear factor of activated T cell cytoplasmic 1 (NFATc1), a key modulator of the formation of osteoclast, further inducing the osteoclast-associated gene expression, including matrix metalloproteinase-9 (MMP-9), cathepsin K (Ctsk), and tartrate-resistant acid phosphatase (TRAP) [10,11] .
The NF-κB signaling pathway is a critical pathway in the differentiation of osteoclast. Many research reports have demonstrated that suppression of the NF-κB signaling pathway inhibits osteoclast formation and function [12][13][14] . The 3-phosphoinositide-dependent protein kinase 1 (PDK1) gene was rst recognized as an essential upstream lipid kinase of the phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT) cascade in insulin signal transduction. Activation of AKT through phosphorylation triggers a cascade that further activates AKT downstream factors [15] . Most studies on PDK1 have focused on its relation to tumors [16] .
Osteoclasts, as the only bone resorptive cells in vivo, play important roles in osteolysis induced by tumors. Tumor cells in the bone microenvironment secrete several cytokines that trigger osteoclast activity, which in turn increases various lymphocytokines and growth factors that stimulate the tumor cells proliferation [17] . Interestingly, recent research reports have demonstrated that F2r reacts to RANKL activation and impedes osteoclastogenesis by suppressing both the F2r-NF-κB and F2r-AKT signaling pathways [18] . Moreover, research has shown that stachydrine inhibiting osteoclastogenesis via AKT signaling prevents LPS-induced bone loss [19] . Since PDK1 is an upstream activating element of AKT, we hypothesized that PDK1 would affect bone resorption and the formation of osteoclast through the AKT/NF-κB pathway.
Generation and identi cation of PDK1-cKO (RANK Cre . PDK1 ox/ ox ) mice PDK1 ox/ ox and RANK Cre mice were designed by GemPharmatech (Nanjing, China). Cre recombinase expression was regulated via RANK promoter transcriptional control. PDK1 ox/ ox and RANK Cre mice were mated to produce RANK Cre . PDK1 ox/+ mice. RANK Cre . PDK1 ox/+ and PDK1 ox/ ox were mice mated to produce PDK1-cKO (RANK Cre . PDK1 ox/ ox ) and wild-type (WT) (PDK1 ox/ ox ) mice (Fig. 1A). Mice were kept in individually ventilated cages (temperature: 22-26°C; humidity: 50-60%; dark/light: 12/12 hours) with free to eat standard feed and freshwater. Mice were numbered with ear tags 2 weeks after birth. Rat tails were cut into 1 mm segments for PCR. DNA was acquired utilizing a TIANamp genomic DNA extraction kit (TIANGEN Biotech, Beijing, China) as per the instructions stipulated by the manufacturer. Primers were provided by Shanghai Sangon Biological Engineering Technology (Shanghai, China) ( Table  1). The PCR reaction system was as follows: 25 μL 2× Tag MasterMix, 2 μL primer F, 2 μL primer F, 1 μL template, and 20 μL ddH 2 O. The PCR reaction conditions were set according to the 2× Tag MasterMix instructions (Shanghai Sangon Biological Engineering Technology, Shanghai, China). The PCR products were used for horizontal agarose gel electrophoresis (Fig. 1B). Mice were weighed weekly and weights were recorded. All of the animal experimentations were undertaken in compliance with the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2020). Mice were anesthetized via of 2% sevo urane in a special container, CO 2 was then injected into the container at a rate that replaced 25% of container volume per minute. When it was con rmed that the mice were dead, the CO 2 was turned off. The execution dates ranged from November 2019 to May 2021.
Alizarin Red and Alcian Blue staining to visualize mice skeletons At 8 weeks, WT and PDK1-cKO mice were euthanized, placed on a foam board in a prone position, and photographed utilizing an X-ray imager (Faxitron MX20/DC2; voltage: 5.0 KV; time of exposure: 6.0 s).
Mice were observed for skeletal deformities and variation and then dissected after photography. The main viscera were dissociated to observe whether there was any variation. The whole skeleton was carefully dissociated and muscle tissue was removed as much as possible, and then the skeleton was xed in 95% alcohol for 3 d, which was then digested in acetone for 48 h to remove excess adipose tissue. The skeleton was stained with pre-con gured staining solution for 5 d (75% alcohol: 0.3% Alcian Blue: 0.1% Alizarin Red: glacial acetic acid = 1:1:1:17). After staining, the skeleton was transferred to 1% KOH and soaked for 48 hours, and then placed to different solutions for 24 hours (glycerol: 1% KOH = 1:4; glycerol: 1% KOH = 1:1; glycerol: 1% KOH = 4:1). After the muscle tissue was transparent, the staining results were observed.

Murine model of prostate cancer-induced osteolysis
Anesthetization of eight-week-old PDK1-cKO and WT mice (n = 6) was done by inhalation of sevo urane and preoperative IP injection of penicillin (200 U/g) to avoid infections. The needle was injected into the proximal tibia head of the right lower extremity with a microsyringe, then rotated 2-3 mm. A 10 μL cell suspension containing 5×10 5 RM-1 was slowly injected, then the syringe was removed, the site was disinfected with iodophor, and mice were placed back into their cages after waking up. After 2 weeks, the tibia of the right lower limb was xed in 4% paraformaldehyde for micro-CT detection.

Detection of bone conversion markers
Eight-week-old mice venous blood was harvested and centrifugated for 20 minutes at 1600 rpm. Bone formation markers (PINP, BGP) and bone resorption markers (TRAC-5b, CTX-I) were detected using an ELISA kit (Wuhan Huamei Biological Engineering Co., Ltd., Wuhan, China) as per the instructions stipulated by the manufacturer.
Micro-CT scanning and 3D reconstruction of the tibia Two weeks after the murine model of prostate cancer-induced osteolysis was established, the right tibia was taken for scanning and 3D reconstruction was conducted using a micro-CT scanner (SkyScan1072)

TRAP and HE staining of the tibial bone
After micro-CT scanning, the tibial bones were decalci ed in 10 percent ethylenediaminetetraacetic acid (EDTA) for a fortnight at a temperature of 4°C ensued by continuous dehydration in 40 percent, 75 percent, and 95 percent ethanol for 60 minutes, followed by two dehydrations in 100 percent ethanol with each for 30 min. Tissue specimens were cleared in xylene for 15 minutes before being in ltrated with para n for 3 hours. The sequential segments were split into 5 μm slices, followed by staining with HE and TRAP, and subsequently visualized utilizing an inverted light microscope (Nikon Eclipse TS100, Tokyo, Japan).

Bone marrow-derived macrophage cell extraction and osteoclast differentiation
Eight-week-old PDK1-cKO and WT mice were euthanized. The intact femur and tibia were isolated after disinfection in 75% alcohol for 5 min. The ends of the femur and tibia were cut off, the α-MEM was drained with a 1 mL syringe, and the bone marrow cavity was rinsed 3 times. The collected rinse solution was sieved with a 200 μm sterile lter, then the supernatant was discarded and subjected to suspension in α-MEM comprising 10 percent FBS and M-CSF (25 ng/mL). The medium was transferred to a T-75 culture ask and subjected to culturing at a temperature of 37°C with a CO 2 concentration of 5%. The α-MEM comprising 10 percent FBS and M-CSF (25 ng/mL) was replaced at a frequency of once every two days. After 4 days, several bone marrow-derived macrophage cells (BMMs) were observed, which were digested and suspended in α-MEM with 10 percent FBS and M-CSF (25 ng/mL). BMMs were kept in 96well plates (8000/well), then subsequently placed in a cell incubator for culturing. After 24 h, the α-MEM comprising 10 percent FBS, RANKL (100 ng/mL), and M-CSF (25 ng/mL) was used to replace the medium. Roughly 6 d after RANKL induction, multinucleated osteoclasts were observed under a microscope. After discarding the medium, rinse operation was down two times using PBS, and then 4% paraformaldehyde was added for xation at room temperature (RT) for 30 min. The xative was subsequently discarded and washing routine was conducted twice with PBS, and TRAP dye was added at RT for 1 h. Then, the TRAP dye was discarded, followed by two-time rinse using PBS and air seasoning. Finally, pictures were taken and the number of osteoblasts was calculated.
BMMs proliferation/viability assay BMMs of PDK1-cKO and WT mice were plated in 96-well plates (6000/well) and subjected to culturing in α-MEM comprising M-CSF (25 ng/mL). Then, after 48hours, each well was added 10 μL CCK-8, followed by incubation at a temperature of 37°C for 2 hours, and values recording of OD at 450 nm utilizing a microplate reader. The proliferation activity of PDK1-cKO and WT mice was statistically analyzed according to the OD value.

Podosome actin belt formation assay
BMMs were put in 96-well plates (8×10 3 /well) to induce mature osteoclasts (same induction process as described above). When mature osteoclasts were observed, cells were gently rinsed twice using 1× PBS, xed using percent paraformaldehyde at RT for 10 minutes, and rinsed thrice using 1× PBS; then cells were permeated using 0.1 percent Triton X-100 in PBS for 5 min at RT, blocked using percent BSA in PBS for 30 minutes, rinsed 2 times with 0.2 percent BSA, diluted with rhodamine-conjugated phalloidin in 0.2 percent BSA (1:100), incubated for 1 hour at RT, rinsed 4 times using 0.2 percent BSA and 4 times using 1× PBS, and stained using DAPI for 5 minutes. Lastly, photos were taken under a uorescence microscope (Life Technologies, Carlsbad, CA, USA). The data were evaluated utilizing ImageJ (NIH, Bethesda, Maryland, USA).

Hydroxyapatite resorption assay
BMMs were placed in 6-well plates (2×10 4 /well) and induced with α-MEM comprising 10 percent FBS, RANKL (100 ng/mL), and M-CSF (25 ng/mL). When the round-like preosteoclasts were observed under a microscope, they were digested with 0.25% trypsin and placed in a 96-well hydroxyapatite-coated bone absorption plate (2000/well). After 3 d of culturing, the hydroxyapatite coating was absorbed into a transparent area with irregular shape and the culture medium was sucked out. Wells were washed with 1× PBS 3 times and washed with 5% sodium hypochlorite solution for 10 minutes to eliminate the remaining adherent cells. Finally, hydroxyapatite-coated bone absorption plates were washed using 1× PBS thrice, visualized using an inverted microscope, and photographed after air drying. The absorption area was measured using ImageJ (NIH, Bethesda, Maryland, USA).
Quantitative real-time PCR (qPCR) The total RNA from PDK1-cKO and WT osteoclasts was obtained utilizing a TaKaRa MiniBEST universal RNA extraction kit (Takara Bio Inc, Kyodo, Japan) as per the protocol stipulated by the manufacturer. The reverse transcription of RNA into cDNA was performed. The acquired cDNA was utilized as a template for qPCR, which was carried out on an ABI Prism 7500 system (Thermo Fisher Scienti c, Waltham, MA). The PCR cycling setting was as illustrated below: 95°C for 30 seconds; 40 cycles at 95°C for 5 seconds; and 60°C for 34 seconds. Table 1 lists the primers utilized in this research. The relative gene expression was evaluated utilizing the 2 -ΔΔCt method.

Western blot analysis and protein extraction
To investigate the impacts of conditional knockout of the PDK1 gene on the early activation of RANKL signaling pathways, BMMs were treated with serum-free starvation for 3 hours and subsequently triggered with RANKL (100 ng/mL) for 5, 10, 20, 30, or 60 min. To evaluate the effects of conditional knockout of PDK1 on the late RANKL activated signal cascade, BMMs were triggered with RANKL (100 ng/mL) for 0, 2, 4, or 6 d; WT mice were used as the control. Total cellular proteins (TCPs) were obtained utilizing a TCP extraction kit (Sigma-Aldrich, St. Louis, MO, USA) as per the protocols stipulated by the manufacturer. TCPs were isolated by 10 percent SDS-PAGE gel and loaded onto nitrocellulose membranes (Thermo Fisher Scienti c, Shanghai, China). Blocking of the membranes was done with 5 percent skim milk in 1× TBST for 1 hour at RT, followed by incubation using primary antibodies (for the dilution ratio, refer to the reagent instructions) for 15 h at 4°C, and three-time rinse with 1× TBST.
Afterward, membranes were subjected to incubation with IRDye uorescent-labeled secondary antibodies at RT for 1 h. The corresponding protein bands were imaged using an LI-COR Odyssey SA Infrared Imaging Scanner. Densitometric analyses were measured using ImageJ (NIH, Bethesda, Maryland, USA).

Statistical analyses
All of the data are presented as the mean ± standard deviation (SD). All the experimentions conducted in this study were replicated thrice unless otherwise noted. Statistical differences were determined using SPSS v22.0 (SPSS Inc., Chicago, IL, USA). To compare the two groups, an unpaired Student's t-test was employed. The signi cance threshold was P < 0.05.

Effects on mice phenotypes after PDK1 deletion in osteoclasts
Previous studies have shown that the whole-body PDK1 knockout gene in mice contributes to early embryonic death [20] . Therefore, in this study, we established a mouse model for the conditional PDK1 knockout in osteoclast cells. To evaluate the effects on the growth and development of mouse bone after PDK1 deletion in osteoclasts, we monitored skeleton size and body weight. Results revealed that, when compared to WT mice, PDK1-cKO mice had smaller skeleton development and lower body weight after 5 weeks and onwards. The difference in body weight between the 2 groups was signi cant (P < 0.05) and no malformations were detected in the bone or vital organs of PDK1-cKO mice ( Fig. 2A-G). To determine the functional level of osteoblasts and osteoclasts in vivo, we detected the contents of bone turnover markers in 8-week-old mice's serum. Results revealed that bone resorption markers (TRAC5b, CTX-1) of PDK1-cKO mice were signi cantly reduced when compared to WT mice (P < 0.01). However, there were no considerable variations between the contents of bone formation markers (P1NP, BGP) (Fig. 2H, P > 0.05).
Osteoclast-speci c knockout of PDK1 ameliorated prostate cancer-stimulated osteolysis in vivo To evaluate the impacts of osteoclast-speci c knockout of PDK1 on prostate cancer-stimulated osteolysis, we established a mice model of prostate cancer-induced osteolysis. Results revealed that there was no considerable variation in tumor mass between WT and PDK1-cKO mice (Fig. 3A, B, P > 0.05). However, osteolysis in PDK1-cKO mice was reduced when contrasted with WT mice (Fig. 3E). Additional analysis of bone parameters showed that when compared to WT mice, BV/TV, Tb. Th, Tb. N, and Conndens. signi cantly increased (Fig. 3C, D, F, G, P < 0.05), whereas Tb. Sp and SMI reduced (Fig. 3I, J, P > 0.05). In the bone tissue sections, the number of trabeculae in PDK1-cKO mice was higher than in WT mice after HE staining; this result is consistent with the bone parameter analysis. Moreover, the amount of TRAP-positive osteoclasts in PDK1-cKO mice was lesser as opposed to that in WT mice after TRAP staining (Fig. 3H).

Osteoclast-speci c knockout of PDK1 suppressed RANKL-induced osteoclastogenesis, podosome belt formation, and bone resorption function in vitro
To explore the impacts of osteoclast-speci c knockout of PDK1 on the proliferation activity of BMMs, after 48 h of M-CSF stimulation, we examined the proliferative activity of BMMs. Results revealed that there was no considerable variation between WT and PDK1-cKO mice (Fig. 4A, P > 0.05). To further explore the effects of osteoclast-speci c knockout of PDK1 on RANKL-induced osteoclastogenesis, on the sixth day of RANKL induction, TRAP staining was applied to mature osteoclasts. Results revealed that conditional knockout of PDK1 in osteoclasts inhibited the differentiation of RANKL-induced osteoclasts (Fig. 4B). The count of TRAP-positive multinucleated cells (n ≥ 3) in PDK1-cKO mice was lesser as opposed to that in WT mice (Fig. 4C, P < 0.001). To examine morphological alterations and podosome belt formation, mature osteoclasts were subjected to staining with rhodamine-phalloidin. When compared to WT mice, there were smaller osteoclasts with fewer nuclei in PDK1-cKO mice (P < 0.001), suggesting that the knockout of PDK1 in osteoclasts inhibited precursor cell fusion (Fig. 4D-F).
Given that osteoclast-speci c knockout of PDK1 impaired the formation of podosome actin belt, which is a precondition for osteoclast function, we postulated that osteoclast-speci c knockout of PDK1 would also inhibit osteoclast bone resorption. We used hydroxyapatite-coated bone absorption plates to explore whether the deletion of PDK1 in osteoclasts would have an effect on osteoclast resorption function. Results revealed a smaller resorption area after PDK1 deletion in osteoclasts (Fig. 4G, H, P < 0.001). These ndings indicated that the PDK1 deletion in osteoclasts can effectively inhibit osteoclast bone resorption function.
Conditional PDK1 knockout in osteoclasts inhibited osteoclast-speci c gene expression Upon the stimulation of osteoclast differentiation during BMM differentiation in osteoclasts, several osteoclast-speci c genes (Ctsk, TRAP, MMP-9, NFATc1) are upregulated in BMMs [21,22] . To further explore the underlying mechanisms, the osteoclast-speci c genes expression was examined at the mRNA level. Results revealed that, when compared to WT mice, the relative expression levels of Ctsk, TRAP, MMP-9, and NFATc1 were considerably inhibited in PDK1-cKO mice (Fig. 5A, P < 0.01).
Deletion of PDK1 in osteoclasts suppressed the RANKL-stimulated NF-κB signaling pathway The NF-κB pathway is the main signaling pathway triggered during osteoclast formation [23] . To investigate the mechanisms that underly the suppression impacts of the osteoclast-speci c knockout of PDK1 on early osteoclastogenesis, the NF-κB signaling pathways in osteoclasts were identi ed by Western blot analysis. Seeding of the BMMs was done on 6-well plates and cultured over the night to enable cells to attach to the wall. Then, the cells were activated by RANKL for 0, 5, 10, 30, or 60 min, TCPs were extracted, and the expressions of p-AKT, AKT, P65, p-P65, IκBα, and p-IκBα were examined. Results revealed that when compared to WT mice, p-AKT/AKT, p-I κBα/IκBα, and p-P65/P65 declined in PDK1-cKO mice (Fig. 5B, C, E, F,). To better examine the mechanisms that underly the suppression effects of osteoclast-speci c knockout of PDK1 on late osteoclastogenesis, TCPs were extracted after BMMs were induced by RANKL for 0, 2, 4, or 6 d. Results revealed that, when compared to WT mice, PDK1, Ctsk, RANK, and NFATc1 protein expression was considerably inhibited in PDK1-cKO mice (Fig. 5D, G-J).

Discussion
Bone metastasis is clinically di cult to treat due to pain, increased SREs, decreased quality of life, and decreased overall survival, and there is currently no effective treatment [24] . Approved therapeutic agents for treating bone metastasis in the 2019 National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology focus on the treatment of obvious pain and SREs. These drugs mainly include bisphosphonates, which inhibit the bone resorptive function of osteoclasts by triggering apoptosis, and denosumab, which is a RANKL-speci c inhibitor that inhibits osteoclast activation and development, reduces bone resorption and increases bone mineral density [25] . Although phosphates are now widely used in clinical practice, there is a risk of osteonecrosis and potential esophageal tumors after long-term use [26] . Therefore, identifying novel bone-protective mechanisms is a pertinent area of research. Activated osteoclasts play key roles in tumor-related bone destruction through bone resorption [27] . RANKL produced by tumor cells stimulates osteoclast precursor cells to differentiate osteoclasts [28,29] , which thereby activates osteoclasts, further providing a suitable bone microenvironment for tumor growth [30,31] . Thus, the inhibition of osteoclasts is an important research direction for reducing SREs.
The present study demonstrated that the osteoclast-speci c knockout of PDK1 ameliorated prostate cancer-induced osteolysis and reduced bone resorption markers in the blood in vivo. There was no considerable alteration in tumor mass; however, the activation of osteoclasts stopped the dormancy of tumor cells and promoted their growth [30,32] . A possible explanation for this result is that the inhibition of osteoclasts was mainly manifested due to certain conditions within the bone microenvironment. The proliferation of tumor cells resulted in tumor tissue covering the tibia, then the diameter of the tumor far exceeded the diameter of the tibia, and osteoclasts were con ned to bone tissue; thus, the inhibition of osteoclasts did not change tumor weight. In this research, we found that the conditional knockout of PDK1 in osteoclasts in vivo led to smaller body size of mice. We used a RANK promoter-driven Cre-LoxP system to conditionally delete the PDK1 gene in osteoclasts. When cells expressed RANK, Cre recombinase knocked out PDK1. However, RANK is not speci c to the expression of osteoclasts and other cells are expressed in a small amount [33,34] . Therefore, the small size of mice is likely due to the knockout of PDK1 in other cells. Another surprising nding was that that the TRAP-positive osteoclast-like cells in tumor tissue were observed after histological TRAP-staining. This result is likely due to the scattered tumor-associated macrophages that were differentiated into osteoclasts under the activation of RANKL secreted by tumor cells, which is consistent with the observation of osteoclast-like cells in the soft tissue of leiomyossarcoma by Gibbons et al. [35] .
In this study, it was veri ed in vitro that the osteoclast-speci c knockout of PDK1 suppressed RANKLstimulated osteoclastogenesis, osteoclast-speci c gene expression, and bone resorption function but the proliferation of BMMs was not affected. This result may be due to the RANK promoter-driven Cre-LoxP system that conditionally deleted the PDK1 gene in osteoclasts. When cells expressed RANK, Cre recombinase knocked out PDK1; without RANKL stimulation, BMMs rarely expressed RANK. At this time, PDK1 was not knocked out or was rarely knocked out in osteoclasts. Further investigation of the molecular mechanisms demonstrated that the deletion of PDK1 in osteoclasts inhibited osteoclastogenesis via the RANKL-stimulated NF-κB signaling pathway (Fig. 6).
Osteoclasts are the only cells derived from macrophage/monocyte lineage cells with bone absorption functions in the body [36] . The proliferation, differentiation, and activation of osteoclasts require the participation of M-CSF and RANKL. MCS-F stimulates BMMs that become osteoclast precursors, and the RANKL binding to its receptor, RANK, stimulates the differentiation of osteoclast precursors into osteoclasts [37] . The NF-κB signaling pathway is an essential pathway for the differentiation of osteoclast. RANK signals recruit tumor necrosis factor receptor-associated factor 6 (TRAF6), which activates the mitogen-activated protein kinases (MAPKs), NF-κB and activator protein-1 (AP-1), which activate NF-κB that induces NFATc1, a key osteoclastogenesis regulator [38,39] . PDK1 expression is dysregulated in many cancer types and is an interesting and unexplored target for cancer therapy [15,40] . The PDK1 protein activates the PI3K-AKT pathway. Previous studies demonstrated that AKT actives the NF-κB signaling pathway in tumor cells [41,42] . AKT has also been found to enhance osteoclast formation and osteolysis in osteolysis-related diseases [43] . Our study demonstrated that PDK1 can further activate the NF-κB signaling pathway in osteoclasts by activating AKT. These results suggested that targeting the PDK1 gene in osteoclasts might be a good treatment approach for prostate cancer-related osteolysis. In our animal model, instead of injecting tumor cells into the circulatory system, we injected tumor cells speci cally into the tibia's bone marrow cavity. Although this did not mimic distant metastasis in tumor cells, it directly led to bone damage. We did not further verify the above experimental results through PDK1-speci c inhibitors, as there is no PDK1-speci c inhibitor for only osteoclasts on the market. Moreover, we were concerned that PDK1-speci c inhibitors would affect tumor cells at the time of intervention and that we could not properly evaluate the function of PDK1 in osteoclasts.
To summarize, this research illustrated that the conditional knockout of PDK1 in osteoclasts ameliorated prostate cancer-induced osteolysis effectively by suppressing RANKL-stimulated bone resorption and osteoclastogenesis.

Declarations
The National Natural Science Foundation of China (grant No.: 81860402), "139" Plan for training highlevel medical backbone talents in Guangxi and the 22nd Batch of Guangxi Ten Hundred Thousand Talents Project funded this research.

Availability of data and materials
The corresponding author can provide datasets created in this work upon reasonable request.