Celastrol treatment protects against acute ischemic stroke-induced brain injury by promoting an IL-33/ST2 axis-mediated microglia/macrophage M2 polarization
Acute ischemic stroke (AIS) is the most common type of cerebrovascular disease and is a leading cause of disability and death worldwide. Recently, a study suggested that transformation of microglia from the pro-inflammatory M1 state to the anti-inflammatory and tissue-reparative M2 phenotype may be an effective therapeutic strategy for ischemic stroke. Celastrol, a traditional oriental medicine, may have anti-inflammatory and neuroprotective effects. However, the underlying mechanisms remain unknown.
We first determined the expression levels of inflammatory factors in patients and rodent models associated with AIS; we then determined the anti-inflammatory effects of celastrol in AIS, both in vivo and in vitro, using animal models of middle cerebral artery occlusion (MCAO) and cell models of oxygen-glucose deprivation (OGD) treatment with or without celastrol, respectively.
The results indicated that expression of both inflammatory (interleukin (IL)-1β, IL-6, and tumor necrosis factor (TNF)-α) cytokines, as well as the anti-inflammatory cytokine, IL-33, and IL-10, were increased following AIS in patients and in animal models. Furthermore, in vitro experiments confirmed that celastrol treatment decreased inflammatory cytokine expression induced by OGD through an IL-33/ST2 axis-mediated M2 microglia/macrophage polarization. Finally, celastrol is protected against ischemic-induced nerve injury, both in vivo and in vitro.
Taken together, these data suggest that celastrol post-treatment reduces ischemic stroke-induced brain damage, suggesting celastrol may represent a novel potent pharmacological therapy.
KeywordsCelastrol ST2 IL-33 M2 microglia/macrophage polarization Acute ischemic stroke
Acute ischemic stroke
Cell counting kit-8
Enzyme-Linked Immuno Sorbent Assay
Erythrocyte sedimentation rate
High-density lipoprotein cholesterol
Low-density lipoprotein cholesterol
Middle cerebral artery occlusion
Red blood cells
Sodium dodecyl sulfate-polyacrylamide gel electrophoresis
Tumor necrosis factor-α
2, 3, 5-Triphenyltetrazolium chloride
White blood cells
Stroke remains one of the leading causes of death and disability worldwide  with nearly 80% of cases characterized as ischemic. Cerebral ischemia results in an inflammatory response which plays an important role in subsequent neuronal death. Very early thrombolysis and mechanical thrombectomy have been shown to improve outcome after ischemic stroke; nevertheless, additional therapies are urgently needed, as most patients with ischemic stroke receive insufficient benefits from existing therapies . Recently, post-ischemic inflammation, which lasts nearly a week after stroke onset, has been considered as a potential target in widening the therapeutic time frame [3, 4]. In addition, uncovering the mechanisms associated with this inflammatory response may provide preventative therapies against subsequent neuronal cell death in ischemia-related events, including AIS [5, 6].
Increasing evidence indicates that microglia and infiltrated macrophages play critical roles in regulating the immune and inflammatory responses after brain injuries . Microglia/macrophages are known to have different phenotypes with distinct functions during the course of ischemic brain injury . M2 microglia protect neighboring cells by removing cell debris and releasing trophic factors for brain repair, while chronically activated M1 microglia exacerbate brain injury by producing neurotoxic substances, although they participate in clearing cell debris at early stages after stroke . Microglia/macrophage phenotype polarization is likely dependent on activation status, and balancing this polarization is a promising therapeutic strategy for stroke treatment.
Expression of interleukin (IL)-33, a member of the IL-1 cytokine family, is considered to be a warning sign , although the role of IL-33 in cerebral ischemic infarction remains controversial. A recent study found that IL-33 expression was increased in acute ischemic stroke (AIS) patients compared with healthy controls . IL-33 expression in brain tissue also improved ischemic-induced brain injury by promoting M2 phenotype polarization and suppressing Th17 responses in mice . Another recent study found that IL-33/growth stimulation expressed gene 2 (ST2) signaling activated beneficial M2 macrophage polarization after ischemia and subsequently reduced neuronal cell death. ST2 (growth stimulation expressed gene 2) is a member of the IL-1 receptor superfamily and is expressed on cardiomyocytes, as well as a large variety of immune cells, including M2 macrophages .
Celastrol is a quinone methide triterpene isolated from the root extracts of Tripterygium wilfordii (thunder god vine) and Celastrus regelii . Increasing evidence suggests that celastrol possesses anti-inflammatory and anti-oxidant activities [15, 16]. Indeed, it was shown that celastrol had a protective effect against ischemic injury in a rat cerebral ischemia model .
This study aimed to determine whether celastrol has protective effects against AIS-induced injury. We assessed the expression of inflammatory factors in patients with AIS and also analyzed the effects of celastrol on microglial polarization and the inflammatory response, both in vitro and in a rat model of ischemic stroke. The results indicate that celastrol does have protective effects against AIS-induced injury and that these effects are related to an IL-33/ST2 axis-mediated microglia/macrophage M2 polarization.
Patients and clinical characteristics
Sixty first-ever AIS patients (34 females and 26 males; average age, 62.2 ± 8.2 years) were recruited from January 1, 2017, to September 31, 2017, at The Pudong New Area Gongli Hospital, Shanghai, China. All patients were diagnosed as having an acute cerebral infarction according to the World Health Organization criteria  and had symptoms within 72 h. Age- and sex-matched healthy individuals were also selected from the Pudong New Area Gongli Hospital (32 males and 28 females; average age, 58.4 ± 6.9 years). Body weight and height were recorded for subsequent calculation of body mass index (BMI) using a standard formula: BMI (kg/m2) = body weight (kg)/height (m2). Stroke severity was assessed using the National Institutes of Health Stroke Scale.
Blood samples (20 ml) were taken from the cubital vein within 24 h of symptom onset and used for further analysis. Whole blood was used for assessment of white blood cells (WBC), red blood cells (RBC), platelet (PLT) count, hemoglobin (Hb) concentration, and erythrocyte sedimentation rate (ESR).
Serum concentrations of total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL), and high-density lipoprotein cholesterol (HDL) were analyzed using Randox kits (Randox Laboratories Ltd., Crumlin, UK) with an automated RX Imola biochemical analyzer (Randox Laboratories, Ardmore, UK); and the resulting atherogenic index was calculated using a standard formula: AI = (TC-HDLC)/HDLC (AI, atherogenic index; TC, total cholesterol; HDLC, high-density lipoprotein cholesterol). The serum expression levels of the inflammatory factors, interleukin 33 (IL-33), interleukin 6 (IL-6), interleukin 1β (IL-1β), and tumor necrosis factor-α (TNF-α), interleukin 10 (IL-10), were measured with commercially available Enzyme-Linked Immuno Sorbent Assay (ELISA) kits (Sen-Xiong Company, Shanghai, China).
Animals and ethics statement
Male Sprague–Dawley rats (230–280 g, 36 rats) were purchased from Shanghai Sippr Bk Laboratory Animals Co., Ltd., (Shanghai, China). All rats were allowed free access to food and water under controlled conditions (12/12 h light/dark cycle with humidity of 60 ± 5%, and a temperature of 22 ± 3 °C). All animals were treated in accordance with the Guide for the Care and Use of Laboratory Animals, and all experiments were approved and performed according to the guidelines of the Ethics Committee of Pudong New Area Gongli Hospital. All surgical procedures were performed under anesthesia, and every effort was made to minimize suffering. Rats were anesthetized by intraperitoneal injection of pentobarbital sodium (30 mg/kg).
Murine models of middle cerebral artery occlusion
Animals were anesthetized by intraperitoneal injection of sodium pentobarbital. Body temperature was monitored and maintained at 36.5 °C to 37.5 °C. A modified model of middle cerebral artery occlusion (MCAO) was used to induce permanent focal ischemia, as previously described . Briefly, the right middle cerebral artery (MCA) was occluded by inserting a monofilament nylon suture (diameter 0.24–0.26 mm) with a heat-rounded tip into the internal carotid artery, which was advanced further until it closed the origin of the MCA. Sham-operated rats underwent the same surgical procedure without insertion of the filament. No mortalities of rats were observed.
Cerebral ischemia and sham-operated animals were randomly assigned to either vehicle (12 rats) or celastrol groups (24 rats). Twelve rats from celastrol groups were administered with 1 mg/kg celastrol dissolved in 0.9% NaCl and 1% dimethylsulfoxide (DMSO) (InvivoGen, San Diego, CA, USA) immediately and 1 day post-operation through intraperitoneal injection. Rats were re-anesthetized and sacrificed 72 h (control: MCAO: MCAO+celastrol = 6: 6: 6) or 10 days (control: MCAO: MCAO+celastrol = 6: 6: 6) post-MCAO. Behavioral tests were performed 3, 5, 7, and 10 days after MCAO.
Measurement of infarct volume
Measurements of infarct volumes were performed as previously described . Briefly, infarct volume was determined with 2, 3, 5-triphenyltetrazolium chloride (TTC) 72 h post-MCAO (n = 6 in each group). Brain tissue was sliced into thick sections (1-mm-thick coronal sections) and stained with a 2% solution of TTC for 20 min at 37 °C, followed by fixation with 4% paraformaldehyde. TTC-stained sections were imaged and analyzed using Image Pro-Plus 5.1 analysis system (Media Cybernetics, NY, NY, USA). Lesion volumes were calculated using the following formula: [total infarct volume − (the volume of intact ipsilateral hemisphere–the volume of intact contralateral hemisphere)]/contralateral hemisphere volume × 100%.
Sensorimotor functional recovery after stroke was measured before MCAO and also 3, 5, 7, and 10 days post-procedure. All behavioral tests were performed by an investigator blinded to the experimental groups. The rotarod (IITC Life Science, NY, NY, USA) test was performed to determine sensorimotor coordination. Briefly, rats were placed on an accelerated rotating rod with an increasing speed from 4 to 120 rpm within 5 min. Rats were tested three times daily with an intermission of 5 min, and latency to fall off the rotating rod was recorded. The data were expressed as mean values from three trials. The adhesive removal test was also employed . In brief, a rat was placed in a cage for 1 min and an adhesive tape (50 mm2) was applied to the distal radial region of the right forelimb as a tactile stimulus. The time to contact and the time to remove the tape were both recorded. Each animal was tested three times with a cutoff time of 120 s per trial. The data are presented as the mean time to contact and the mean time to remove the tape on each testing day.
Cells or post-fixated brain slices with 4 μm thickness were incubated with ST2 (cat. nos. 6131; 1:200; mouse original), Iba-1 (cat. nos. 6756; 1:400), CD206 (cat. nos. 53206; 1:400), CD16 (cat. nos: 6641; 1:400), and NeuN (cat. nos. 6745; 1:400) antibodies (all from InvivoGen, San Diego, CA, USA) overnight at 4 °C, then incubated with conjugated secondary antibody for 1 h at room temperature in the dark. After several washes with phosphate-buffered saline (PBS), the slides were incubated with DAPI for 3 min and then mounted in glycerol. Slides were then visualized using fluorescence microscopy (excitation range from 490 to 495 nm and optimal emission range from 515 to 525 nm), and ten random fields in each section were analyzed. Then further statistical analysis in each independent experiment was performed.
Primary neuron-microglia co-cultures and treatments
Primary neurons and microglia were prepared from the brains of mixed-sex embryonic day 17 Sprague–Dawley rats, as previously described . The co-culture system was established by inserting a transwell chamber (Millipore, Bedford, MA, USA) with 3 μm pores into 12-well plates (Costar, NY, USA). Primary neurons were seeded (Costar) at a density of 1 × 103 cells/well, and an equal amount of microglia were seeded in the transwell chamber. The chamber was inserted into the wells of the plate to allow communication between the two cell types. Co-culture systems were subjected to 3 h oxygen-glucose deprivation (OGD) or control treatment, after which 12.5, 25, 50, or 100 ng/mL IL-33 (Enzo Life Science, NY, USA) or PBS were added to transwell chambers for 24 h. In addition, celastrol (0.25, 0.5, 1, or 2 μM) or PBS were added 3 h after OGD or control treatment for 24 h to assess possible neuroprotective effects. Neurons were subsequently collected for cell viability assays.
Primary microglia-enriched cultures and treatments
Primary microglia-enriched cultures were prepared from whole brains of 1-day old Sprague–Dawley rat pups as previously described . Microglia were transfected with a ST2 interference vector before pretreatment with OGD for 3 h, then treated with 50 ng/mL IL-33, 1 μM celastrol, or PBS for 24 h. Cells were collected for microglial polarization analysis using immunofluorescent staining.
Primary neuronal cultures and treatments
Primary neuronal cultures were prepared from 1-day old Sprague–Dawley rat pups as previously described . Neurons underwent OGD for 3 h after which they were treated with different concentrations of IL-33 (12.5, 25, 50, or 100 ng/mL), celastrol (0.25, 0.5, 1, and 2 μM) or PBS for 24 h. Cell viability was assessed with the CCK8 assay.
ELISA for soluble inflammatory cytokines
The expression of inflammatory factors IL-33, IL-6, IL-10, IL-1β, and TNF-α in cell supernatants or serum were measured using commercially available ELISA kits (Sen-Xiong Company). In accordance with the manufacturer’s instructions, supernatants were stored at − 80 °C before measurements and both standards and samples were run in triplicate. The optical density (OD)450 was calculated by subtracting the background, and standard curves were plotted.
Cell viability assay
Cell counting kit-8 (CCK8) was used to assess cell viability. Cells (1 × 104) were seeded into 96-well plates, and grown overnight and treated as described previously in this section. Medium was removed, and the cells were washed three times with PBS. DMEM (90 μL) and CCK8 (10 μL) were subsequently added to each well and incubated for 1.5 h at 37 °C, and a microplate reader was used to measure the OD450.
Flow cytometry was used to determine the percentage of apoptotic cells. Apoptotic cells were differentiated from viable or necrotic cells by the combined application of annexin V (AV)-FITC (Roche, Indianapolis, IN, USA) and propidium iodide (PI) (Roche). Cells were washed twice and adjusted to a concentration of 1 × 106 cells/mL with cold D-Hanks buffer. AV-FITC (10 μL) and PI (10 μL) were added to 100 μL of cell suspension and incubated for 15 min at room temperature in the dark. Finally, 400 μL of binding buffer was added to each sample without washing and analyzed with flow cytometry. Each experiment was performed at least in triplicate.
To detect apoptotic neurons, post-fixated brain slices (4 μm thickness) were measured with TUNEL detection kit (Roche) according to the manufacturer’s instructions. The nuclei were stained with DAPI, and TUNEL staining was then assessed. Nuclei that were double-labeled with DAPI and TUNEL were considered apoptotic.
Western blotting analysis
Cells were harvested and lysed in Triton X-100 lysis buffer containing 25 mM Tris–HCl (pH 7.5), 137 mM NaCl, 2.7 mM KCl, 1% Triton X-100, and protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA) for 30 min at 4 °C. Purified proteins were then detected by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using a 12% separating gel and a 5% stacking gel followed by staining with Coomassie blue to determine the purity of the enzyme preparations. Western blotting analysis was performed as follows: the proteins were transferred from SDS-PAGE gels onto nitrocellulose membranes and blocked in TBS-T buffer (50 mM Tris–HCl pH 7.5, 150 mM NaCl and 0.2% Tween-20) for 1 h at room temperature with 5% nonfat milk. Membranes were then probed with primary antibodies, including ST2 and GAPDH, followed by incubation in respective secondary antibodies (Santa Cruz Biotechnologies Inc., Santa Cruz, CA, USA). Membranes were then incubated with Lumi-PhosTM WB (Pierce, Rockford, IL, USA) chemiluminescent substrate for 5 min. Protein expression levels were measured by densitometry using ImageJ 1.48u4 software (National Institutes of Health, NY, USA), and the data were normalized against the corresponding loading control (GAPDH).
RNA extraction and qRT-PCR
Total RNA was extracted from cells using TRIzol reagent (Sigma-Aldrich) according to the manufacturer’s instructions. Total RNA was eluted with RNase-free water and stored at − 80 °C. RNA concentrations were determined with Epoch spectrophotometry. Quantification of ST2 and endogenous control mRNA U6 was performed using TaqMan assays with the supplied assay-specific RT primers (Applied Biosystems, Foster City, CA, USA). The data were analyzed using the 2−ΔΔCt method.
Construction and transfection of siRNAs
Specific siRNA against ST2—with the target sequence: 5′-CACTGGGAACTTATGTTGAATGG-3′—was synthesized by GenScript (Piscataway, NJ, USA). Microglia cells were dispersed into collagen I-coated six-well plates at a density of 5 × 104 cells/well and cultured overnight under normal conditions. The following day, cells were transfected with siRNA in serum-free medium using oligofectamine reagent (Invitrogen) for 48 h at 37 °C with 5% CO2. Cells were then collected for further analysis.
Data are presented as the mean ± standard deviation (SD). Statistical analysis was performed using one-way ANOVA with post hoc tests for comparisons between two groups using GraphPad Prism 5.0 software (GraphPad Software, Inc., La Jolla, CA, USA). P values ≤ 0.05 indicate statistically significant differences.
Serum IL-33 levels and inflammatory factor levels are increased in stroke patients
Clinicopathological characteristics of AIS and healthy controls
58.4 ± 6.9
62.2 ± 8.2
78.9 ± 6.3
81.3 ± 7.8
1.7 ± 0.1
1.7 ± 0.1
26.8 ± 3.1
27.1 ± 3.4
12.6 ± 2.4
Risk factors (no)
Hypertension, n (%)
CHD, n (%)
6.1 ± 1.6
10.4 ± 7.2***
4.8 ± 0.3
4.9 ± 0.5
139.3 ± 8.9
144.6 ± 11.8
236.3 ± 72.6
249.6 ± 68.6
5.6 ± 2.3
19.2 ± 8.8***
5.3 ± 0.4
4.9 ± 0.5
1.2 ± 0.12
1.4 ± 0.13
1.0 ± 0.11
1.04 ± 0.08
2.95 ± 0.82
2.97 ± 0.72
2.7 ± 0.3
3.3 ± 0.4*
165 ± 25.6
283 ± 54.8***
23.5 ± 3.2
69.6 ± 3.4***
57.4 ± 6.8
122.8 ± 11.6***
29.5 ± 4.3
82.5 ± 7.9***
32.3 ± 5.2
68.4 ± 6.2***
Celastrol treatment prevents ischemic-induced brain injury in a rodent MCAO model
Celastrol treatment protects against cerebral ischemia-induced neurological dysfunction
Celastrol treatment promotes ST2/IL-33 activation in microglia and promotes M2 microglial polarization
ST2 signaling in microglia is essential for the M2 phenotype in microglia/macrophages
Celastrol treatment protects against OGD-induced neuronal apoptosis and inflammatory factor expression
Taken together, these in vitro studies strongly suggest that activation of the IL-33/ST2 axis in microglia promotes polarization toward the M2 phenotype and protects neighboring neurons from ischemic injury.
Celastrol is a natural compound whose therapeutic potential has been reported in many diseases, including cancer, diabetes, and neurodegenerative disorders [16, 22, 23]. A growing number of reports suggest that celastrol protects against stroke-induced brain damage , and in agreement, here, we report that celastrol treatment not only ameliorated infarct volumes but also decreased the inflammatory response after ischemic stroke.
It is well known that inflammation response play key roles in ischemia-induced nerve injury. A series of studies were conducted to investigate anti-inflammatory effects of a therapy against ischemic damage [24, 25]. Previous research showed that microglia are divided into activated M1 and M2 microglia, according to their phenotypes and polarization . Specifically, activated M1 microglia secreted pro-inflammatory cytokines and were potentially harmful, whereas activated M2 microglia served important roles in repair and plasticity [8, 27].
Recent studies have shown that IL-33 is released from CNS cells rapidly after injury and contributes to the activation of immune responses in lesion areas . Previous work showed that IL-33/ST2 signal activation promoted microglial polarization toward the M2 phenotype, while at the same time, actively inhibited the expression of M1-mediated cytokines . Interestingly, our result shows that IL-33 is increased along with other pro-inflammatory cytokines in serum of patients after stroke. Similarly, rat IL-33 axis is also increased after ischemia, and a further increase induced by celastrol induces M2 microglia phenotype. The anti-inflammatory effects of celastrol on microglial polarization and IL-33/ST2 signal activation remain unclear. Nevertheless, we showed IL-33 promoted microglial ST2 expression, which suppressed neuronal damage by promoting anti-inflammatory cytokine IL-10 expression after induced microglial polarization toward the M2 phenotype. Downregulation of ST2 promoted microglial M1 polarization even when high levels of IL-33 were present. Finally, celastrol treatment increased IL-33 expression and suppressed ischemia-induced inflammatory factor expression by promoting IL-33/ST2-mediated M2 microglial polarization. Therefore, we speculate that the protective effects of celastrol in AIS-induced brain injury are related to an IL-33/ST2 axis-mediated microglia/macrophage M2 polarization. Previous studies have implicated IL-33/ST2 axis in downregulation of TLR signaling . The activation of TLR signaling can shift macrophage polarization toward the M1 phenotype . But if TLR signaling is involved in IL-33/ST2 axis-mediated macrophage polarization, regulation is still unclear. So, in future work, we will generate a ST2 knockout rat model to confirm the above finding and further elucidate the related mechanism of IL-33/ST2-mediated M2 microglial polarization.
In conclusion, we found an increase in serum IL-33 and inflammatory factor levels in AIS patients. Celastrol treatment suppressed AIS-induced pathological nerve injury and sensorimotor dysfunctions by inhibiting inflammatory factor expression. In addition, celastrol markedly reduced ischemia-induced neuronal apoptosis and inflammatory factor expression, which was relative to IL-33/ST2-mediated M2 polarization. Taken together, these data show that celastrol is a promising new agent for preventing AIS-induced brain injury.
This study was supported by the Foundation of Discipline Leader in Health Systems of the Pudong New District (no. PWRd2014-02 for Limin Xu, PWRd2014-09 for Mei Jiang), the Shanghai Sailing Program (no. 15YF1410800 for Fanfan Cao), the National Natural Science Foundation of China (no. 81400793 for Ying Wang, 81201029 for Mei jiang), Science and Technology Development Fund of Pudong New District Minsheng Scientific Research (Medical and Health) Project (no. PKJ2017-Y24 for Mei Jiang), the Shanghai Pudong District Science and Technology Innovation Project (no. PKJ2013-Y03 for Ying Wang), the Shanghai Health and Family Planning Commission General Program (no. 201540395 for Ying Wang), and the Key Specialty Construction Project of Shanghai Municipal Commission of Health and Family Planning (no. ZK2015B16). Shanghai Traditional Chinese Medicine Content Construction Innovation Project (no. ZY3-CCCX-3-7001 for Zhang).
Availability of data and materials
MJin, FC, and LX designed the studies and prepared the manuscript with comments from all authors. MJiang, XH, FX, and YW performed all the experiments and analyzed the data. DZ and XH carried out all of the experiments and revised the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
The study was approved by the Ethics Committee of the Pudong New Area Gongli Hospital.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 1.Huuskonen MT, Tuo QZ, Loppi S, Dhungana H, Korhonen P, McInnes LE, Donnelly PS, Grubman A, Wojciechowski S, Lejavova K, et al. The copper bis (thiosemicarbazone) complex Cu (II) (atsm) is protective against cerebral ischemia through modulation of the inflammatory milieu. Neurotherapeutics. 2017;14:519–32.CrossRefPubMedGoogle Scholar
- 15.Venkatesha SH, Dudics S, Astry B, Moudgil KD. Control of autoimmune inflammation by celastrol, a natural triterpenoid. Pathog Dis. 2016;74 https://doi.org/10.1093/femspd/ftw059.
- 21.Hu X, Zhang D, Pang H, Caudle WM, Li Y, Gao H, Liu Y, Qian L, Wilson B, Di Monte DA, et al. Macrophage antigen complex-1 mediates reactive microgliosis and progressive dopaminergic neurodegeneration in the MPTP model of Parkinson’s disease. J Immunol. 2008;181:7194–204.CrossRefPubMedPubMedCentralGoogle Scholar
- 22.Yu X, Ruan X, Zhang J, Zhao Q. Celastrol induces cell apoptosis and inhibits the expression of the AML1-ETO/C-KIT Oncoprotein in t (8;21) leukemia. Molecules. 2016;21 https://doi.org/10.3390/molecules21050574.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.