Considerations of growth factor and material use in bone tissue engineering using biodegradable scaffolds in vitro and in vivo

Materials

Reagents were purchased as follows: ethyl acrylate (Sigma, UK), ELP with statherin sequence (SN_A15) (Technical Proteins Nanobiotechnology, Valladolid, Spain); collagenase (Gibco, UK); human fibronectin and human recombinant BMP-2 (R&D systems, Biotechne, UK); recombinant human BMP-2 (Infuse/InductOS® Bone graft kit, Medtronic, USA); alcian blue 8X, light green SF, orange G 85% pure, paraformaldehyde 96% extra pure, phosphomolybdic acid hydrate 80% (Acros Organics); Picrosirius Red, Van Gieson’s stain, Weigert’s Haematoxylin Parts 1 and 2 (Clintech Ltd, UK); benzoyl peroxide, GMA solution B, JB4 solution A (Polysciences); GoTaq qPCR master mix, Herring sperm DNA, RNeasy mini prep RNA extraction kit (Promega); phosphate buffered saline (PBS), trypsin/ethylenediaminetetraacetic acid (EDTA), Dulbecco’s Modified Eagle Medium (DMEM), Alpha Minimum Essential Medium (αMEM), penicillin–streptomycin (Scientific Laboratory Supplies, SLS); 4-nitrophenol solution 10 nM, acetic acid, acetone, acid fushsin, alizarin red S, alkaline buffer solution, ascorbic acid-2-phosphate, beta-glycerophosphate disodium hydrate salt (βGP), cell lytic M, dexamethasone, fast violet B salts, glycine, histowax, hydrochloric acid, iodoacetamide, ipegal, L-glutamic acid, Naphthol AS-MX phosphate 0.25%, parafilm, PBS (with CaCl₂/MgCl₂), phenyl methyl sulphonyl fluoride, phosphatase substrate, polysorbate 80, ponceau xylidine, silver nitrate, sodium chloride, sodium hydroxide pellets, sucrose, TRIS–EDTA (TE) buffer solution (Merck, UK); Embedding capsule (TAAB Laboratories equipment); alamarBlue™ HS Cell Viability Reagent, 70 µM cell strainer, dibutyl phthalate xylene (DPX), ethidium homodimer-1, fetal calf serum (FCS), fisherbrand grade 01 cellulose general purpose filter paper, Histoclear, isopropanol, methyl benzoate, Quanti-IT™ Picogreen™ ds DNA reagent, Taqman® Reverse Transcription Kit, Vybrant™ CFDA SE Cell Tracer Kit (Thermofisher Scientific, UK); Fast green and sodium thiosulphate (VWR); Lubrithal (Dechra, UK), Isoflurane (Dechra, UK), Buprenorphine (Buprecare® multidose, Animalcare, UK) and Vetasept® sourced from MWI animal health, UK. Uncoated vacutainers, 3-way stopcock and 5/0 PDS II suture (Ethicon, USA) from NHS supply chain. All other consumables and reagents were from Sigma-Aldrich, UK.

Production of PCL trimethacrylate scaffold material

PCL-trimethacrylate of this molecular weight has been synthesised and 3D printed via stereolithography (SLA) previously^8,9,10. Silica gel (40–63 μm; VWR chemicals) was used as a stationary phase. ¹H NMR and ¹³C NMR spectra were recorded on a JEOL 400 NMR spectrometer, with working frequencies of 400 MHz for ¹H nuclei.

Poly(caprolactone) trimethacrylate synthesis

Poly(caprolactone) triol, M_n = 300 Da, (50 g, 0.17 mmol, 1 eq), anhydrous dichloromethane (350 mL) and triethylamine (100 mL, 0.72 mmol, 4.3 eq) were added to a 1 L two-necked round bottom flask. The reaction was placed under a nitrogen atmosphere and then cooled in an ice-water bath for 15 min. A pressure-equalising dropper funnel charged with methacryloyl chloride (65 mL, 0.67 mmol, 4 eq) was attached to the round bottom flask. The methacryloyl chloride was added dropwise over approximately 3 h. The reaction was covered with aluminium foil to protect it from light and allowed to stir and warm to room temperature (RT) overnight. The next day, methanol (50 mL) was added to quench the reaction, which was allowed to stir at RT for 30 min. The reaction mixture was transferred to a separating funnel and washed with 1 M aqueous hydrochloric acid solution (5 × 250 mL), saturated sodium bicarbonate solution (1 × 250 mL) and brine (1 × 250 mL). The organic layer was then dried with anhydrous magnesium sulphate, filtered and concentrated via rotary evaporation. The crude yellow liquid was then purified using a silica plug, with dichloromethane as the eluent. Fractions containing PCL-trimethacrylate were pooled and concentrated via rotary evaporation. The PCL-trimethacrylate was transferred to a brown glass vial and dried using a stream of air (through a plug of CaCl₂) overnight to yield the title compound as a slightly yellow viscous liquid (82.2683 g). The PCL-trimethacrylate was supplemented with 200 ppm (w/w) of 4-methoxyphenol (MEHQ) as an inhibitor (16.34 mg).

¹H NMR (400 MHz, CDCl₃) δ 6.14 – 6.04 (m, 3H), 5.63 – 5.50 (m, 3H), 4.18 – 4.00 (m, 9H), 2.36–2.32 (m, 3H), 1.94 (m, 9H), 1.75 – 1.47 (m, 9H), 1.47 – 1.32 (m, 2H), 1.02 – 0.83 (m, 3H).

The characterisation data agrees well with that previously reported, however, with an improved degree of functionalisation (> 95%).

3D Printing of PCL-trimethacrylate

PCL-trimethacrylate octet-truss scaffolds were designed based on a modified body centre cubic unit cell (diameter = 5 mm, and height = 5 mm, strut diameter = 0.5 mm, surface area 143.4 mm²), denoted PCL-TMA scaffolds. This scaffold design was chosen to mimic the unit cell geometry of larger octet-truss style scaffolds as previously reported by Reznikov et al. but in a format suitable for in vitro studies⁷. The PCL-TMA scaffolds were printed using masked (SLA) 3D printing on a Prusa SL1 or SL1S. The resin was prepared for 3D printing by first dissolving 0.1% (w/w) 2,5-thiophenediylbis(5-tert-butyl-1,3-benzoxazole) (OB +) as a photoabsorber in the PCL-TMA by stirring at RT for 1 h. Finally, 1.0% (w/w) diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide (TPO-L) as a photoinitiator was added to the resin.

After printing, the scaffolds were rinsed with ethanol and removed from the build plate. Scaffolds were sonicated in ethanol (5 × 5 min) and allowed to dry for 15 min at RT. The scaffolds were post-cured using a Formlabs Form Cure for 60 min at RT. After post-curing, the scaffolds were soaked into ethanol overnight at RT on a rocker (100 rpm), rinsed with ethanol (3x) and allowed to dry at RT before being ELP and/or PEA coated and EO sterilised, or left uncoated and EO sterilised.

Scanning Electron Microscopy (SEM)

The PCL-TMA unit cell structures were imaged using a Zeiss Leo Gemini Scanning Electron Microscope fitted with an SESI detector. The samples were coated with a 15 nm chromium coating prior to imaging. All images were taken using 8 kV accelerating voltage.

ATR-FTIR

The PCL-TMA unit cells were measured using ATR-FTIR using a Bruker Alpha II Compact FTIR. A background scan of 128 scans was used. Samples were measured using 128 scans at a resolution of 1 cm⁻¹. The FTIR spectra was smoothed using a Whittaker-Hayes smoothing function, and baseline corrected using an adaptive smoothness parameter penalized least squares method (asPLS).

Compressive testing

The compressive mechanical properties of the scaffolds were measured using a Bose Electroforce TA 3200 equipped with uniaxial load cells. Due to the geometry of the unit cell scaffolds, the effective elastic modulus was measured. A 1 N preload was implemented, with the effective elastic modulus of the PCL-TMA unit cells measured between 1 – 3% strain (n = 5).

Accelerated degradation

To confirm the biodegradation of PCL-TMA an accelerated degradation study using sodium hydroxide was conducted. The unit cells were first washed five times in milli Q water, then twice in absolute ethanol, wiped off with Kimtech precision wipes, then dried in a vacuum desiccator overnight, before the initial weight (m₀) of the cell cylinder was measured. The cell cylinders were then placed in individual 1.5 mL Eppendorf tubes, before 1 mL of 2 M sodium hydroxide was added. The samples were then incubated at 37 °C for 1, 2, 4, 7, 14, or 21 days before they were taken out, washed and dried in a similar manner as described above, then the new weight (m₁) was measured. Three repeats were conducted at each time point.

Coating of the scaffold material

The scaffold coating process and work flow used in this in the study are illustrated in Supplementary Figure 1.

PEA/FN/BMP-2 coating of the scaffold material

PEA coating of materials was performed as previously described¹⁸. Briefly, the scaffolds were treated in air plasma for 5 min before being exposed to monomer plasma. Plasma polymerization was carried out in a custom-made capacitively coupled plasma installation for low-pressure plasma in a 15-L T-shaped reactor made of borosilicate glass and stainless-steel end plates sealed with Viton O-rings. A vacuum was produced by a rotary pump or a scroll pump (both BOC Edwards, UK), with working pressures for the monomer plasma from 0.15 to 0.25 mbar. The plasma was initiated via two capacitively coupled copper band ring electrodes located outside of the reactor chamber and connected to a radiofrequency generator (Coaxial Power System Ltd.). The monomer pressure was controlled via speedivalves (BOC Edwards, UK) and monitored with a Pirani gauge (Kurt J. Lesker). PEA was applied on the material surfaces for 15 min at a total radiofrequency power of 50 W. The samples were sterilized afterwards using ethylene oxide (EO).

The application of the BMP-2 coating was optimised by investigating different diluents and BMP-2 manufacturers to ensure the biological activity of the BMP-2 protein (Supplementary Table 1, Supplementary Figures 2, 3 and 4). For experiments with the PEA/FN/BMP-2, or ELP/PEA/FN/BMP-2 coated PCL-TMA scaffolds, FN and InductOs® BMP-2 were diluted in PBS with calcium chloride (CaCl₂) and magnesium chloride (MgCl₂) added. A maximum of three scaffolds were placed into a non-coated 11 mL vacutainer and 1 mL of FN solution (20 µg/mL) was added. A vacuum was created and after 1 h in the sterile hood at RT, the FN solution was removed. PBS (containing CaCl₂/MgCl₂) was added at 1–2 mL per vacutainer to rinse off non-bound FN and repeated once. The scaffolds were handled with sterile forceps and transferred to new vacutainers. 1 mL of BMP-2 solution (100 ng/mL for in vitro experiments or 5 µg/mL for CAM assay and subcutaneous implantation experiments) was added to each tube. The formation of a vacuum was repeated. After 1 h at RT, the BMP-2 solution was removed and the scaffolds rinsed twice in PBS (containing CaCl₂/MgCl₂) prior to use.

ELP coating of the scaffold material

ELP coating of scaffolds was based on previous published methods¹⁶. In brief, lyophilized ELP powder was dissolved in solvent mixture of DMF/DMSO (at 9/1 ratio) to prepare 5% (w/v) ELP solution followed by addition of hexamethyl diisocyanate (HDI) crosslinker (cross-linker to lysine ratios of 12/1). 3D printed polyamide or PCL-TMA scaffolds were immersed in the ELP solution for 10–15 s and left to dry overnight at RT (22 °C) inside a glovebox (BELLE Technology, UK) maintained at a humidity < 20%. Dry ELP coated scaffolds were washed several times with deionised water (dH₂O) to remove excess HDI and were termed as ELP coated scaffolds¹⁷.

The ELP coating was initially exogenously mineralised in a fluorapatite solution, however, this was found to affect the material property of the PCL-TMA. Therefore, ELP coating without exogenous mineralisation was deemed a suitable coating material due to results confirming that the ELP coating could be mineralised in vitro using surrounding mineralising media components (Supplementary information).

For the ELP/PEA/FN/BMP-2 coated scaffolds, the ELP coating was applied, followed by PEA and subsequent EO sterilisation. Finally, the FN and BMP-2 were adhered to the scaffolds as described in section “PEA/FN/BMP-2 coating of the scaffold material“.

HBMSC isolation and culture

Isolation and culture of HBMSCs

Human bone marrow samples were collected from patients undergoing hip replacement surgery, with prior informed consent from the patients. The methods were performed in accordance with the relevant guidelines and regulations at the University of Southampton. The samples were identified only by sex (male (M) or Female (F)) and age (e.g., F60) to maintain confidentiality, with approval of the University of Southampton’s Ethics and Research Governance Office and the North West-Greater Manchester East Research Ethics Committee (18/NW/0231) for use of the tissue for research. In a class II hood, under sterile conditions, 5–10 mL alpha-Minimum Essential Medium (α-MEM, Lonza, UK) or Dulbecco’s Modified Eagle Medium (DMEM, Lonza, UK) was added to the universal tube of marrow and shaken vigorously to extract the HBMSCs. A 3 mL sterile Pasteur pipette was used to remove the supernatant media/cellular debris mix to a 50 mL falcon tube and washing was repeated until the bone was light pink/white in colour. The cell suspension was centrifuged (272 g Heraeus mega 1.0R centrifuge) for 5 min. The supernatant was removed, the pellet resuspended in α-MEM or DMEM and passed through a 70 µM cell strainer (Fisher Scientific, UK) to remove bone and fat debris. The suspension was centrifuged and the supernatant poured off. The pellet was resuspended in basal media (α-MEM, 10% fetal calf serum (FCS), 1% penicillin–streptomycin (P/S)) and the HBMSCs were cultured in T175 flasks at 37 °C in 5% CO₂/balanced air until approximately 80% confluent. Collagenase (2% solution and or 0.22 IU/mg) was used prior to trypsin solution (1 × concentration (Stock Trypsin/EDTA (10X), includes 1,700,000 UL trypsin 1:250 and 2 g/L Versene® (EDTA)), for passaging and seeding onto scaffolds.

Osteogenic media consisted of α-MEM, 10% FCS, 1% P/S, ascorbate-2-phosphate 50 mM (2 µL/mL), dexamethasone 10 μM (1 µL/mL). All media was changed every 3–4 days.

Cell seeding onto scaffolds

2.5 × 10⁴ Passage 2 (P2) HBMSCs (F79) per scaffold were used for cell viability/alamarBlue™ HS experiments and 5 × 10⁴ Passage 1 (P1) HBMSCs were used for biochemistry (F75) and molecular (M59) experiments (Supplementary Figure 5). Each PCL-TMA octet-truss scaffold was added individually to a 2 mL Eppendorf tube and 500 µL cell suspension (α-MEM, 1% P/S, without FCS) was added. The Eppendorf tubes were placed in a 50 mL falcon tube (6 per tube) for positioning horizontally on the MACSmix™ Tube Rotator enabling a maximum of 24 scaffolds to be seeded per rotator machine. The Eppendorf tubes were fully sealed and therefore the air available to cells was that only within the tubes. The media maintained the normal pink colour throughout cell seeding, indicating the pH of the media was unaltered. FCS was added at approximately 10% (55 µL) to each tube after 3 h. After 24 h, each scaffold was moved to individual wells of a 24 well plate containing 1.5 mL basal or osteogenic media. Culture was performed at 37 °C in 5% CO₂/balanced air, with media changed every 3 days. Each scaffold coating type/media condition was set up in triplicate.

Assessment of cytocompatibility of PCL-TMA and coatings

alamarBlue™ HS Cell Viability assay

For cell viability experiments alamarBlue™ HS Cell Viability Reagent was added to basal media at 10% (v/v) concentration. Fluorescence measurements were taken at day 1 (when the PCL-TMA scaffolds were removed from the Eppendorf tube after 24 h of cell seeding) and on day 14 (each PCL-TMA scaffold was moved to a new 24 well plate to ensure only the cells adhered to the scaffold were quantified). A 1 mL aliquot of media/alamarBlue™ HS mix was added to each well containing a scaffold and to 3 wells with no scaffold as background measurements and incubated for 4 h at 37 °C in 5% CO₂/balanced air. After 4 h, 100 µL samples were taken from each well and plated in triplicate in a black 96 well plate. The fluorescence was measured using the GloMax® Discover Microplate Reader (Promega, UK) at Green 520 nm excitation and 580–640 nm emission and the average background measurement was subtracted from each sample well.

Labelling of live and dead HBMSCs

Vybrant™ CFDA SE Cell Tracer 10 μM and 5 µg/mL Ethidium homodimer-1 in PBS, were used to label live and dead cells respectively on PCL-TMA scaffolds at day 1 and day 14 post alamarBlue™ HS analysis. Media/alamarBlue™ HS was removed and scaffolds washed twice in PBS. A 1 mL aliquot of labelling solution was added to cover the scaffolds/cells and incubated for 15 min at 37 °C in 5% CO₂/balanced air. The labelling solution was removed and replaced with α-MEM/1% P/S/10% FCS and culture continued for 30 min. The samples were imaged under fluorescence microscopy using the FITC filter (excitation 485 nm, emission 515 nm) for live cells and RHODA/TRITC filter (excitation 510–560 nm, emission 590 nm) for dead cells, with a Zeiss Axiovert 200 microscope and Axiovision 4.2 imaging software.

Biochemistry assays for HBMSC differentiation analysis

Alkaline phosphatase specific activity measurement

Placement of scaffolds in basal or osteogenic media after 24 h of cell seeding was determined as day 0. Therefore, day 1 was after 24 h of culture in basal or osteogenic media and so on until the day 7 when the HBMSCs were lysed. HBMSCs attached to PCL-TMA scaffolds were washed twice in PBS and the scaffolds were transferred to individual 2 mL Eppendorf tubes. A 200 µL aliquot of Cell lytic M was added to cover the scaffold and left in contact for 15 min at RT with vortexing performed three times every 5 min. The Eppendorf tubes were stored at −80 °C.

ALP activity was measured using a colourimetric absorption assay. P-nitrophenol phosphate (pNPP) production was measured against standards. A 10 µL aliquot of cell lysate was transferred to a clear 96 well plate and 90 µL of substrate was added. The plate was incubated at 37 °C until a yellow colour change was noted and the time recorded for this change. A 100 µL aliquot of 1M sodium hydroxide solution was added to stop the reaction, prior to reading absorbance on the GloMax® Discover (spectrophotometer) at 405 nm. The same centrifuged cell lysate samples were used as for ALP quantification. PicoGreen™ was diluted 1/200 in TE (1x) buffer and added to all wells, including standards. The plate was left on the benchtop at RT for 5 min prior to the quantity of DNA being measured using a fluorescence assay at blue 475 nm excitation and 500–550 nm emission, in a GloMax® Discover Microplate Reader (Promega, UK). Results expressed as nanogram (ng) of DNA. ALP specific activity was calculated as ALP produced per ng of DNA as nmol pNPP/ng DNA/hr.

Osteogenic gene expression analysis

RNA extraction

RNA extraction was performed using the Promega ReliaPrep™ RNA Cell Miniprep System instructions. Scaffolds were washed in PBS and transferred to 2 mL Eppendorf tubes with BL-TG buffer (250 µL) for 15 min to lyse the cells prior to storage at -80 °C. The tubes were thawed and isopropanol (85 µL) added, prior to brief vortexing. The lysate was transferred to a minicolumn within a collection tube. The column was washed in a series of steps as per the kit instructions and the final RNA eluted with nanopure water (15 µL). RNA quantification and purity measurement were performed using the Nanodrop 2000 spectrophotometer.

Reverse transcription

RNA quantity for conversion to cDNA was standardised for all samples with dilution in nanopure water to a volume of 9.6 µL in a 0.2 mL PCR reaction tube. The reagents of the TaqMan® Reverse Transcription (RT) Kit were mixed in quantities instructed to make RT master mix and 10.4 µL was added to water/RNA mix to give a final 20 µL reaction volume. The tubes were placed in a SimpliAmp thermal cycler (Applied Biosystems, UK) set to 10 min at 25 °C for primer incubation, 30 min at 37 °C for reverse transcription and 5 min at 95 °C for the reverse transcription inactivation and 4 °C to hold samples until retrieval for storage at −20 °C.

Quantitative Polymerase Chain Reaction (qPCR)

Quantitative qPCR was performed using GoTaq PCR master mix. The master solution was made from 10 µL GoTaq PCR master mix, 0.75 µL of forward primer, 0.75 µL of reverse primer and 7.5 µL ultrapure water (Human β-actin gene forward sequence 5’-3’ GGCATCCTCACCCTGAAGTA, reverse sequence AGGTGTGGTGCCAGATTTTC, Human ALP gene forward sequence 5’-3’ GGAACTCCTGACCCTTGACC, reverse sequence 5’-3’ TCCTGTTCAGCTCGTACTGC and Human Collagen1A1 gene forward sequence 5’-3’ CCCTGGAAAGAATGGAGATGAT and reverse sequence 5’-3’ ACTGAAACCTCTGTGTCCCTTCA). The final 20 µL reaction volume was made of 2 µL of cDNA sample and 18 µL of master solution in a 96 well PCR plate, sealed and centrifuged briefly before analysis using 7500 Real Time PCR system (Applied Biosystems, UK). The resulting data were collected using the AB7500 SDS Software (Applied Biosystems, UK). Ct values for each sample were normalised to the housekeeping gene β-actin, after optimisation experiments revealed β-actin to be a reliable housekeeping gene and there was often not sufficient RNA yield to allow multiple housekeeping genes to be used. Relative-expression levels of each target gene were calculated using the ∆∆ C_t method (cycle threshold) method. The uncoated scaffold cultured in basal media was used as an internal relative reference for the other scaffolds and media types. Each sample was a biological triplicate and plated out once each.

Chorioallantoic Membrane assay

Descriptions are extrapolated from Marshall et al.²³. All procedures were performed with prior received ethical approval from the University of Southampton Animal Welfare and Ethics Review Board and in accordance with the guidelines and regulations of the University of Southampton and those stated in the Animals (Scientific Procedures) Act 1986 and using the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. However, a UK Personal Project License (PPL) was not required as the eggs were used only until embryonic day (ED) 14 when the experiment ended. Briefly, fertilised hens (Gallus gallus domesticus) eggs (Medeggs, Henry Stewart & Co., Lincolnshire) were incubated in a humidified (60%), warm (37 °C) incubator (Hatchmaster incubator, Brinsea, UK) at embryonic day (ED) 0. The eggs were incubated in a horizontal position, on a rotating pattern (1 h scheduled rotation), prior to albumin removal at ED 3 (Supplementary Figure 6). The minimum number of eggs required to determine a significant difference (p < 0.05) between groups was calculated with power 80%. Typically, an n = 6 for each condition (to allow predominantly for non-developing eggs) was used.

Scaffold preparation and implantation of materials

All PCL-TMA scaffolds (uncoated, ELP coated and/or PEA coated) were EO sterilised. The PEA scaffolds and ELP/PEA coated scaffolds were coated with FN (20 μg/mL) and BMP-2 (5 μg/mL) solutions immediately prior to use. Scaffolds were implanted at ED 7. Descriptions are extrapolated from Marshall et al.²³. In brief, eggs were candled to check viability and a No. 10 scalpel blade was used to create a 0.5 cm by 0.5 cm window. The white inner shell membrane was peeled away and the materials placed onto the CAM. Parafilm was used to cover the window and labelled tape applied, parallel to sides of the egg, to hold the parafilm in place. The eggs were placed horizontally within an egg incubator at 37 °C and 60% humidity without rotation.

Analysis of results

Samples were harvested at ED 14 of incubation with methodology extrapolated from Marshall et al.²³. Blinding of the assessor was performed. The window was opened digitally and with forceps to image the scaffold/CAM using a stemi 2000-c stereomicroscope (Zeiss, UK), for illustrative, recording purposes only. Quantification of angiogenesis was performed using the Chalkley eyepiece graticule scoring method. Three separate counts were made and the average score was calculated for each egg. Biocompatibility was assessed by counting live, viable and developed chicks and any dead/deformed chicks. Thereafter, the scaffold and 1 cm diameter of surrounding CAM tissue was collected and placed into 2 mL 4% PFA in a 24 well plate for 72 h at 4 °C followed by exchange of PFA for 70% ethanol. Processing of the scaffold/tissue, glycidyl methacrylate (GMA) resin embedding and subsequent histology followed. The chick was euthanised by an approved schedule 1 method at ED 14.

• SEO Powered Content & PR Distribution. Get Amplified Today.
• PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
• PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
• PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
• PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
• Source: https://www.nature.com/articles/s41598-024-75198-3