HSPB7 Protected the Osteogenic Differentiation of Mesenchymal Stem Cells From TNF-α Through Chaperone-mediated Autophagy/beta-catenin Pathway

uorescent puncta triplicates.


Introduction
In the United States, around 8 million people fracture a bone each year, around 5%-20% of these result in impaired or delayed healing, thereby requiring therapeutic intervention eventually, and causing an important clinical and economic issue (Kovach et al., 2015).
An in ammatory response is necessary for the bone regeneration process after injury. A large number of proin ammatory and chemotactic cytokines including tumor necrose factor (TNF)-α, interleukin (IL)-1β, IL-6 and C-C motif chemokine ligand (CCL)2 are released in the early in ammatory phase. Evidence has shown that these mediators not only initiate the repair process by stimulating angiogenesis, but also attract mesenchymal stem cells (MSCs) to the repair region and promote their osteogenic differentiation.
Human bone marrow mesenchymal stem cells (BMSCs), also known as multipotent mesenchymal stromal cells, have self-renewal and multipotent differentiation abilities (Hang et  However, in ammation is a double-edged sword. Prolonged or uncontrolled in ammation has been shown to have destructive effects on bone healing (Takayanagi, 2009). some chronic in ammation conditions, such as bacterial infection and system immunity, cause impaired fracture healing, which is TNF-α is a critical in ammatory factor in bone fracture healing. On one hand, in the acute in ammatory response phase 24 h after trauma, TNF-α concentration peaks, and is maintained above baseline for 72 h. TNF-α has been shown to recruit necessary cells such as MSCs for fracture healing, and have shown a pivotal role in fracture healing using TNF-α receptor-de cient mice (Cho et (Shakoori et al., 1992). Among these sHSPs, HSPB7 is one of the least characterized and is highly expressed in the heart, HSPB7 gene mutation is strongly associated with heart diseases (Krief et al., 1999;Mercer et al., 2018). Within the HSPB family of molecular chaperons, HSPB7 is the most potent inhibitor of polyQ protein aggregation (Vos et al., 2010).
There are several types of autophagic pathways in mammals, namely macroautophagy, microautophagy and chaperone-mediated autophagy (CMA) (Kaushik and Cuervo, 2018). In contrast to the other two autophagy pathways, CMA only degrades proteins that contain a speci c KFERQ-like motif in their amino acid sequences (Dice, 1990). Chaperoe HSCA8 and co-chaperones recognize KFERQ-like motif proteins, which are then delivered to lysosomes by interacting with the receptor lysosome-associated membrane protein type 2A (LAMP2A), the substrate is then degraded in the lysosomal lumen (Cuervo and Wong, 2014).
In this study, we provide the rst evidence that a heat shock protein family composed by HSPB7, HSPH1 and HSCA8 may be a potential therapeutic target for impaired fracture healing in osteomyelitis patients induced by excess of TNF-α through heat shock protein family network.

TNF-α different treatment times have opposing effects on BMSC osteogenesis
We performed a Cell Counting Kit (CCK)-8 assay to determine the in uence of TNF-α at different concentrations (0-100 ng/mL) on the proliferation of BMSCs during osteogenic differentiation, we found no signi cant differences on cell number between different TNF-α concentration groups in the rst 6 days ( Fig. 2a) To determine the effect of TNF-α on the osteogenesis of BMSCs, BMSCs were treated with different concentrations of TNF-α during osteogenic differentiation. In the early days (1-2d), TNF-α (1-100 ng/mL) promoted the expression of osteo-speci c proteins such as collagen type 1 alpha 1 (Col1A1), runt-related transcription factor 2 (Runx2) and SP7/Osterix (SP7/Osx). However, these osteo-speci c proteins were downregulated when BMSCs were treated with TNF-α (1-100 ng/mL) more than 3 d (4-5d) during osteogenesis differentiation ( Supplementary Fig. 1).
Triplicates of Os4_T and Os4_C samples were collected and subjected to mRNA microarray analysis to identify differentially expressed genes (DEGs).
Finally, a total of 890 DEGs were identi ed in the Os4_T samples. Of these genes, 399 were up-regulated and 491 were down-regulated in the Os4_T group compared with the Os4_C group (Fig. 1a). A heatmap was generated using hierarchical cluster analysis to show the expression of 20 DEGs which changed most signi cantly (Fig. 1b). Gene Ontology (GO) enrichment analysis was performed and we found that over 200 DEGs were involved in protein binding (Fig. 1d).
From these DEGs, we selected 10 up-regulated and 10 down-regulated genes based on the signi cance and GO enrichment analysis results. These genes were validated by quantitative real-time PCR (qPCR) (Fig. 1c).
HSPB7 rescued the impaired osteogenesis of BMSC induced by TNF-α Among these selected genes, HSPB7 expression was reduced signi cantly in the Os4_T group. HSPB7 is mainly involved in protein binding and thus was selected as a potential therapeutic target. Western bolt assay has shown the decreased expression of HSPB7 in BMSCs treated with TNF-α for 4 days during osteogenesis ( Fig. 2b-c).
From osteomyelitis patients and normal patients, the qPCR results showed that the mRNA level of HSPB7 was downregulated in osteomyelitis patients compared to that of normal patients, immuno uorescence (IF) showed that HSPB7 protein expression was signi cantly reduced in the bone peripheral muscle tissue of osteomyelitis patients ( Supplementary Fig. 2b) To clarify the protective role of HSPB7, we overexpressed HSPB7 in BMSCs by lentiviral vector transfection. Using IF, we chose 40 as the optimal multiplicity of infection (MOI), we performed western blot analyses to detect the expression of the osteo-speci c proteins analyzed above, including Col1A1, Runx2, Osteoponitn (OPN) and SP7.
Compared with the control group, 4 days TNF-α (100ng/mL) treatment signi cantly reduced the protein expression of Col1A1, Runx2, OPN and SP7, and the inhibitory effect of TNF-α was signi cantly reduced by HSPB7 Overexpression (OE). And HSPB7 overexpression also promoted the protein expression of Col1A1, OPN and SP7 compared with control group (Fig. 2d-e). Alizarin Red S (ARS) and ALP staining results showed that HSPB7 overexpression did not change the calcium deposit level nor ALP activity induced by TNF-α ( Fig. 2g-h).
HSPB7 protects the impaired osteogenesis induced by TNFα via the CMA pathway To determine the mechanism underlying the protective role of HSPB7, we sought to identify its protein binding partners. Coomassie blue staining and silver staining con rmed that HSPB7 interacted with several proteins (Supplementary Fig. 3a-b), and a mass spectrometry assay was performed to identify these interactions. Mass spectrum results showed that HSPB7 interacted with heat shock protein 105 (HSPH1), which has been found to be in a complex with heat shock cognate A8 (HSCA8) and accelerates the ATP hydrolysis of HSCA8 ATPase (Hatayama et al., 1998;Yamagishi et al., 2000).
We then performed in vitro and in silico experiments to determine the nature of the interactions among HSPB7, HSCA8 and HSPH1. In vitro co-immunoprecipitation (CoIP) experiments performed using exogenous recombinant proteins showed that HSPH1 interacted with HSPB7 and HSCA8 (Fig. 3a, Supplementary Fig. 3c), but HSPB7 was not able to interact with HSCA8 (data didn't show here). When HSPB7, HSCA8 and HSPH1 were subjected to CoIP together, HSPH1 would preferentially interact with HSPB7, however, HSPB7 was not able to interact with HSPH1 which interacted with HSCA8 ( Fig. 3a, Supplementary Fig. 3c).
We then performed in vivo CoIP experiments to con rm the interaction between HSPB7 and HSPH1 ( Fig.  3b, Supplementary Fig. 3d). Since HSCA8 is one of the important components of chaperone-mediated autophagy, we performed an IF assay to detect the activity of chaperone-mediated autophagy in BMSCs treated with TNF-α for 4 d during osteogenesis. The results showed that TNF-α decreased the CMA activity, but HSPB7 overexpression signi cantly enhanced CMA activity inhibited by TNF-α (Fig. 3c).
Western blot assay showed that one day treatment of TNF-α (1-100 ng/mL) promoted the expression of LC3B and decreased the expression of SQSTM1/p62 on the rst day in BMSCs of osteogenesis, but enhanced the expression of SQSTM1/p62 signi cantly when BMSCs were treated for four days (Fig. 3de). Additionally, we detected the activity of macroautophagy through Ad-GFP-LC3B and Ad-GFP-p62 infection, nding that macroautophagy in ux decreased when treated with TNF-α for 4 days in BMSC during osteogenesis. HSPB7 overexpression didn't change the state of macroautophagy ux ( Fig. 3f-g).
HSPB7 OE MBMSCs protected fracture healing impaired by TNF-α in a murine tibial fracture model A murine tibial fracture model was used to investigate the protective effects of HSPB7 on TNF-α impaired osteogenesis in vivo. TNF-α (100 ng/mL) was locally administered on the day of surgery and injected repeatedly on days 2, 4, and 6 after surgery, resulting in impaired fracture healing. Impaired fractured healing was rescued by the addition of HSPB7 overexpressing (OE) lentiviral particles. Animals that received HSPB7 overexpressing lentiviral particles (TNF-α-B7 OE group) showed increased expression of RUNX2 and SP7 compared to the group received GFP lentiviral particles (TNF-α-GFP) (Fig. 5a-d). The number of Type H (CD31 hi /Emcn hi ) endothelial cells, which are important for bone formation, (Kusumbe et al., 2014) also increased signi cantly in TNF-α-B7 OE group compared to the TNF-α-GFP group (Fig. 5ef).
The fracture healing effect was also assessed by staining with the Masson's trichrome stain, Safranin O and fast green stain, which indicated the remodeling of the mineralized callus of the fracture healing samples at day 14 ( Fig. 6a-b).
To con rm this observation quantitatively, fracture healing samples were collected at day 28 and analyzed by microcomputed tomography (microCT). The quantitative analysis of the micro-CT data further validated the histological results that HSPB7 protects the impaired fracture healing induced by TNF-α ( Fig. 6c-e).

Discussion
Prolonged TNF-α stimulation impaires osteogenic differentiation of MSCs which is essential for fracture healing. Herein, we offer the rst evidence that HSBP7 serves as a novel therapy target rescue the impaired osteogenesis of MSCs induced by TNF-α, wherein, chaperone-mediated autophagy plays an important role.
TNF-α was con rmed to have a major role in the in ammatory response during fracture healing, by recruiting MSCs and by promoting their differentiation. Previous studies have shown that treatment times of TNF-α have opposite effects on the osteogenesis of BMSCs. Here, we provide a general view of the effect of TNF-α on the osteogenesis of BMSCs over different periods of time. In the rst early days (1-2d), TNF-α promoted the expression of osteo-related proteins of BMSCs during osteogenesis, however, the opposite effect was observed after 3 d. Interestingly, TNF-α effectively enhanced calcium deposits formation and ALP activity regardless of the treatment time.
The mRNA microarray showed that endogenous HSPB7 expression was reduced during osteogenesis in a TNF-α dose-dependent manner. In sHSP family, HSPB1 and HSPB8 have been found to play an important role in bone metabolism. Unlike other sHSP members, HSPB7 is highly expressed in the heart and is one of the least studied. HSPB7 is also expressed in the skeletal muscle, however, no studies have explored the effect of HSPB7 in the skeletal system. Here, we found that HSPB7 effectively rescued the impaired osteogenesis of BMSCs caused by TNF-α treatment, meanwhile, it did not in uence the promoting effect of TNF-α on the calcium deposits and ALP activity. In vivo, murine tibial fracture model shown that HSPB7 has a protective effect of on TNF-α impaired osteogenesis. Thus, HSPB7 may be a therapeutic target for impaired fracture healing caused by overdose TNF-α in disease including osteomyelitis and systemic lupus erythematosus.
In human atherosclerotic vascular smooth cells, TNF-α promotes macroautophagy-related genes including microtubule-associated protein 1 light chain 3 (MAPLC-3) and Beclin-1 (Jia et al., 2006). Activation of the autophagy pathway by TNF-α reduced the production of reactive oxygen species (ROS) (Baregamian et al., 2009;Sivaprasad and Basu, 2008). TNF-α induced necroptosis in L929 cells, while, autophagy induced by TNF-α suppressed this necroptosis process by blocking the p38-NF-kB pathway (Ge et al., 2018;Ye et al., 2011). The evidence mentioned above con rmed that autophagy played an important role of in the protection of TNF-α. In our study, we found that TNF-α promoted macroautophagy of BMSCs during osteogenesis in early days, however, macroautohagy ux decreased when treated with TNF-α for 4 d. HSPB7 interacted with HSPH1 which suppressed the ATPase activity of HSCA8. Inhibition of chaperone-mediated autophagy not only reduced the osteogenesis ability of BMSCs, but also suppressed the rescue effect by depleting HSPH1 on the impaired osteogenesis induced by TNF-α.
However, this study has some limitations that should be considered. First, CMA activity deletion inhibited ARS composites and ALP activity, which demonstrated that TNF-α promoted ARS composites and ALP activity in a CMA-independent way, which should be further investigated. Second, in our study, we found that HSPB7 protected the impaired fracture healing induced by TNF-α via the CMA pathway, however, limited by laboratory conditions, we did not develop gene-editing animal models to con rm our results. Finally, CMA promotes cell survival through degrading KFERQ-like motif proteins which are present in about 30% of cytosolic soluble proteins and providing alternative source of amino acids (Cuervo, 2010;Cuervo et al., 1995), in our study, we did not clear whether CMA degrades proteins that inhibited the osteogenesis of BMSCs or CMA protects the osteogenesis of BMSCs only by providing alternative amino acids.

Conclusion
Taken together, these ndings indicate that a heat shock protein family network including HSPB7, HSPH1 and HSCA8 protects the impaired osteogenesis induced by TNF-α via the CMA/β-catenin pathway (Fig. 7). And in vivo, HSPB7 overexpression lentiviral particles effectively protects the impaired fracture healing induced by TNF-α in a mouse tibia fracture healing model.

Methods
Reagents hBMSCs were purchased from Cyagen Biosciences (Guangzhou, China). hBMSC growth medium were purchased from Cyagen Biosciences. Cells cultured in a atmosphere of 5% CO2 at 37°C. Osteogenic induction medium was prepared according to previous methods (Hang et  Cell proliferation assay Cells were cultured in 96-well plate at a density of 5000 cells per well, CCK-8 (Beyotime) of 10% was added into wells and incubated with cells for 2h at 37°C. Then the cell proliferation was measured by a microplate reader at the absorbance of 450 nm (BioTek, ELX808).

Protein interaction assay
For in vitro immunoprecipitation, recombinant proteins were incubated within lysis buffer contained protease inhibitors and phosphatase inhibitors for 20 minutes at RT. Then adding anti-rabbit HSPB7 or anti-rabbit HSPH1 incubated for 90 min at RT, rabbit IgG (A7016, Beyotime) was used as a negative control. Then the mixture was incubated with protein G gammabind plus sepharose (17088601, GE life, SWEDEN) for 2h at RT. protein complex were centrifuged and washed with lysis buffer 3 times, suspended in 2X loading buffer and boiling for 5 min at 100°C. Finally, proteins were detected by western blotting.
For in vivo co-immunoprecipitation assay, cells were lysed in lysis buffer contained protease inhibitors and phosphatase inhibitors for 30 min, centrifuged and the supernatant were cleared with protein G for 10 min at 4°C. Then the cell lysates were incubated with primary antibodies and rabbit IgG which as a negative control overnight (16 h) at 4°C. Then the antibodies were captured by protein G gammabind plus sepharose (17088601, GE life, SWEDEN) overnight at 4°C. Protein complex were centrifuged and washed with lysis buffer 3 times, suspended in 2X loading buffer and boiling for 5 min at 100°C. Finally, proteins were detected by western blotting, coomassie blue staining and silver staining.

Protein mass spectrometry analysis
Proteins interacted with HSPB7 were freeze dried and enzymolysised by Trypsinbuffer for 16-18h at 37℃. Enzymolysis products were separated by high performance liquid chromatography (HPLC) according to the manufacturer's instructions. Then the enzymolysis products were analyzed by QExactive mass spectrometer (ThermoFisher). Finally, the RawFile of mass spectrum were analyzed by MaxQuant 1.5.5.1 software.

RT-qPCR
Total RNA from cells or human bone tissues were extracted with RNA isolation reagent (Takara) and RNA was quanti ed by spectrophotometer at 260 nm wave length (NanoDrop 2000; ThermoFisher). Total RNA was then reverse-transcribed with the Double-Strand cDNA Synthesis reagent (Takara) and cDNA (2 ul) was used to quanti cation using SYBR Green PCR Master Mix reagent (Takara), and nally detected by a StepOnePlus System (Applied Biosystems). The reaction conditions were as previous described.

Western blot analysis
Proteins was extracted from cells by lysing in RIPA buffer supplemented with protease inhibitors and phosphatase inhibitors (Beyotime). After protein quanti cation, cell lysates were suspended in 5X loading buffer and then equal amounts of proteins were subjected to migration on 10%-12% polyacrylamide gels, then transferred to a polyvinylidene uoride membrane (Millipore, Shanghai, China). The membranes were blocked by non-fat milk (5%) for 60 min and then incubated with primary antibodies for 12h-16h at 4°C. After, membranes were washed with Tris-buffered saline-Tween (TBST) 3 times (10 min each) and incubated with HRP-conjugated secondary antibodies (Beyotime) for 1 h at RT. Finally, proteins were revealed by chemiluminescence reagents (Millipore) and the signal intensity was detected using Bio-Rad XRS system (Bio-Rad).

ALP staining
Cells were passaged to 12-well plates and cultured with osteogenic induction medium when cells reached at 70% con uence. Osteogenic induction medium cultured for 5 days, cells were washed by PBS 3 times, and were xed in 4% paraformaldehyde for 20min-30min at RT. Washed by PBS 3 times, cells were stained by the ALP Color Development Kit (Beyotime). Finally, the ALP activity quanti cation was performed as previous described.

Alizarin Red staining
Cells were passaged to 12-well plates and cultured with osteogenic induction medium when cells reached at 70% con uence. Osteogenic induction medium cultured for 15 days, cells were washed by PBS 3 times and were xed in 4% paraformaldehyde for 20min-30min at RT. Washed by PBS for 3 times, calcium deposition was stained by Alizarin Red staining (Cyagen) for 5min-10 min at RT, and then was rinsed by PBS 3 times. The ARS quanti cation was performed as previous described.

siRNA transfection
Cells were seeded in plates and transfected with siRNA when cells reached 30-50% con uence. Transfection process was according to the manufacturer's instructions. Transfection e ciency was veri ed by qPCR.

Adenovirus transfection
Adenovirus expressing the GFP-LC3B (Ad-GFP-LC3B) and GFP-P62 (Ad-GFP-P62) were purchased from Beyotime. Brie y, cells were cultured in 12-well plates and transfected with adenovirus when cells reached at 70% con uence according to the manufacturer's instructions. Finally, signals were visualized by a uorescence microscope (Leica) and quanti ed by Image J software.

Lentiviral overexpression HSPB7
Human and murine Lentiviral overexpression GFP-HSPB7 (GFP-B7 OE ) particles and the GFP control group (GFP-NC) were purchased from Cyagen Biosciences. A multiplicity of infection (MOI) which was used as the optimal amount of virus dose was con rmed by IF assay. 50-60% con uent cells were incubated with lentiviral particles at a MOI of 40 and 4 µg/ml polybrene. 12 h later, the culture medium was changed and the transfected cells were used in next experiments.
Immuno uorescence assay for CMA-active lysosomes CMA-active lysosomes were identi ed by the colocalization of LAMP-2A and HSCA8, procedures were performed according to the standard immuno uorescence protocol described by Cuervo et. (Kaushik and Cuervo, 2009). And signals were detected by a confocal microscopy (Leica TCS SP8, USA) and quanti ed by Image J software.

Animals
All C57/B16 mice (male, 12 weeks) were purchased from SLAC Laboratory Animal Co (Shanghai, China). All of the animal experiments were approved by the Institutional Animal Care and Use Committee of the 2nd A liated Hospital, Zhejiang University (No: 2020 − 1035). All mice were randomly divided into control and experimental groups.
Sample sizes used at 6-8 animals per experimental procedure Mice-fracture model Mice-fracture model was modi ed on the basis of previous method (Harry et al., 2008). Brie y, mice anesthetized by intraperitoneally injection of 0.3% pentobarbital sodium (30 mg/kg body). Exposed the right lower limb, made an incision lateral between the middle of tibia tuberosity and crest. A 0.38-mm diameter intramedullary xation pin was then inserted into the tibia's medullary canal at the level of the tibia tuberosity for xation. Separated the soft tissue carefully and stripped the periosteum above the crest of tibia. Then an osteotomy was created above the crest of tibia. The same leg was used in each group.
For in vivo study of HSPB7 protecitve effects, animals were divided into three groups. 20ul volume of TNF-α (100ng/ml) or PBS was injected at the fracture site locally on days 0, 2, 4 and 6. GFP-B7 OE or GFP-NC lentiviral particles were injected at day 6. 2 and 4 weeks after surgery, limbs were harvested by lethal intraperitoneal injection of 0.1ml sodium pentobarbitone (200mg/ml) for next experiments.

Histology
Following harvest, 2 weeks of samples were xed by 10% paraformaldehyde for 36h at 4°C and then were decalci ed by 0.5M ethylene diaminetetra acetic acid (EDTA, Beyotime) for 3 days at 4°C. Specimens were then embedded in para n and sectioned at a 5um thickness. Serial sections were depara nized and then stained with Safranin O and Fast Green, Masson's Trichrome according to the standard procedures.

Radiographic analysis
Following harvest, 4 weeks of samples were send to make a microcomputed tomography (µCT) evaluation. Each tibia was scanned using µCT-100 imaging system (Scanco Medical, Switzerland), operation parameters were according to the previous report(Glass et al., 2011).
Then sections were blocked with 2% BSA for 20 min at RT. Washed by PBS 3 times and incubated with primary antibody for 1h at RT. Washed by PBS 3 times and incubated with second antibodies in a wet box for 1h at RT. Then sections were washed by PBS 3 times and stained with DAPI. Target protein was observed under a uorescence microscope (Leica) and quanti ed by Image J software.

Statistical analysis
All experiments were performed at least in triplicate, data are presented as means ± SD, statistical signi cance between two groups was determined by Student's t test, one-way ANOVA or Bonferroni's post- The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Consent for publication
All authors agree to publish this manuscript.  HSPB7 overexpression rescued the impaired osteogenesis of BMSC induced by TNF-α. A) CCK-8 assay showed that no signi cant differences on cell number between different TNF-α concentration groups in the rst 6 days; B-C) HSPB7 expression decreased in BMSCs treated with TNF-α(0-100ng/mL) for 4 days during osteogenesis; D-E) Relative expression of osteo-speci c proteins (RUNX2, COL1A1, OPN and SP7) on day 4 of osteogenesis, Protein expression levels were normalized to that of Tublin. OE: HSPB7 overexpression, NC: HSPB7 overexpression control; F-G) Alizarin Red S (ARS) and alkaline phosphatase (ALP) activity staining. ALP activity staining was performed when cells were cultured for 5 days, ARS staining was performed when cells were cultured for 15 days. Scale bars, 500 um. Data are expressed as mean±SD. Assays were performed in triplicates. *, P<0.05, **, P<0.01, ***, P<0.001 compared with the control group.    GFP-B7OE lentiviral particles protected fracture healing impaired by TNF-α in a murine tibial fracture model. A-B) TNF-α (100ng/mL) resulted in impaired fracture healing, which was rescued by the addition of GFP-B7OE lentiviral particles. The fracture healing effect was assessed by staining with the Masson's trichrome stain, Safranin O and fast green stain. Scale bars, 500 um. C-E) Microcomputed tomography analysis for bone healing. Scale bars, 500 um. GFP-NC: GFP expressed lentiviral particles (Negative Control); GFP-TNF-α: GFP expressed lentiviral particles with TNF-α; B7OE-TNF-α: HSPB7 overexpressed lentiviral particles with TNF-α. Data are expressed as mean±SD. Sample sizes used at minimum 6 animals per experimental procedure. *, P<0.05, **, P<0.01, ***, P<0.001 compared with the control group.

Figure 7
A heat shock protein family network including HSPB7, HSPH1 and HSCA8 protects the impaired osteogenesis induced by TNF-α via the CMA/β-catenin pathway.