Exploratory longitudinal cohort studyStudy design

We prospectively collected serum samples of patients with advanced solid cancer treated with ICIs for up to 6 months or until disease progression and assessed markers of bone turnover and bone metabolism.

Study population

From July 2020 to June 2021, nine patients were recruited from the Division of Pneumology and Internal Medicine 2 of the University Hospital Krems. All subjects had to meet the following inclusion criteria: age ≥18 years, advanced cancer, and treatment with PD1i or PD-L1i as monotherapy. Prior systemic treatment was permitted if completed at least 3 months before ICI initiation. Patients were excluded if they had evidence of bone metastases assessed per bone scintigraphy or fluorodeoxyglucose positron emission tomography scan, had a diagnosis of osteoporosis or primary hyperparathyroidism, or had received bone-modifying agents (bisphosphonates or denosumab).

Serum sample collection and analysis

Serum samples were collected in the morning in fasting condition prior to ICI administration at prespecified timepoints: baseline, month 1, month 2, month 4, and month 6 of ICI treatment. All biospecimens were processed within 1 hour of collection and stored at −80°C before assessment with commercially available ELISA kits. Standard clinical laboratory tests were performed to evaluate C-terminal telopeptide (CTX) (marker of bone resorption), procollagen type I N-terminal propeptide (PINP) and osteocalcin (OCN) (markers of bone formation), and parathyroid hormone (PTH), 25-hydroxy vitamin D (25(OH) Vit D), and calcium (markers of bone metabolism).

Statistical analysis

Descriptive statistics were provided for demographic and laboratory variables. Data are presented as median and range. Differences in concentrations of markers of bone turnover and metabolism between baseline and month 1 or month 4 measurements were evaluated using the Wilcoxon signed-rank test. P≤0.05 indicated statistical significance. Effect size r was calculated using the formula r=Z/√N and classified as large (r ≥0.50), moderate (r ≥0.3), or small (r ≥0.1).8 Statistical analyses were performed using SPSS V.28 (SPSS, Chicago, IL, USA), and plots were generated using R (V.4.3.2).9

In vitro and ex vivo studiesMedia and reagents

Cell culture medium α-Modified Essential Medium+GlutaMAX (α-MEM), L-glutamine, and fetal bovine serum (FBS) were purchased from Gibco, Life Technologies Limited, UK. Dulbecco phosphate-buffered saline and media supplements (including penicillin/streptomycin, β-glycerol phosphate, ascorbic acid, and dexamethasone) were purchased from Sigma Aldrich, USA. The cytokines human macrophage colony-stimulating factor (M-CSF) and human receptor activator of nuclear factor kappa B ligand (RANKL) were obtained from Peprotech, UK. The PD1i pembrolizumab and the PD-L1i atezolizumab were purchased from Selleck Chemicals, TX, USA; as an antibody control, we used the human IgG1 from Biolegend, San Diego, CA, USA (cat. number 403501). In line with most in vitro studies,10 11 experiments were performed with ICI concentrations of 0.05–1 µg/mL, which reflect the poor biodistribution of ICIs in the bone.12–14

Cell culture

OCs were generated from healthy donor peripheral blood mononuclear cells (HD-PBMCs) as previously described.15 HD-PBMCs were seeded into 6-well or 96-well plates at the density of 2×106 cells/cm2. After 2 hours, non-adherent cells were removed to enrich for OC progenitors, which were then cultured in α-MEM containing 10% FBS, 1% penicillin-streptomycin, RANKL 50 ng/mL, and M-CSF 50 ng/mL. OBs were differentiated from healthy donor mesenchymal stem cells (hMSCs, LONZA, Switzerland) as previously described.16 Briefly, hMSCs were seeded in 96-well plates at a density of 6×103 cells/cm2 and cultured in osteogenic media, consisting of α-MEM with 20% FBS, 1% penicillin/streptomycin, 2,5% L-glutamine, 2.16 mg/mL β-glycerol phosphate, 0.05 mg/mL ascorbic acid, and 10 nM dexamethasone.

Dynamic 3D model of BMU

To generate the dynamic 3D bone model, OC progenitors and hMSCs were mixed in 1:2 ratio (2.5×104 and 5×104, respectively) and seeded onto scaffolds of poly-ε-caprolactone (PCL) (3D BioTek, NJ, USA). After overnight incubation, the cell-loaded scaffolds were transferred into disposable High Aspect Ratio Vessels of a RCCS-8 bioreactor (Synthecon Inc, TX, USA) and cultured at a rotation speed of 24–26 rpm in the absence (undifferentiated scaffolds: HD-PBMC/hMSC) or presence of osteogenic medium (differentiated scaffolds: OC/OB).17 18 After 3 weeks, scaffolds were assessed for cell viability and OB differentiation or fixed in paraformaldehyde and processed for imaging studies.

Viability assay

Bone cells or scaffolds were pulsed with alamarBlue (Invitrogen, Germany) at a concentration of 1:100 for 4 hours at 37°C, according to the manufacturer’s instructions. Absorbance was detected at 570 nm and 600 nm using the Tecan plate reader (Infinite M Nano+, Tecan, Switzerland).

Immunohistochemistry and immunofluorescence

For immunohistochemistry analysis, bone cells or scaffolds were stained using chromogenic substrates for tartrate-resistant acid phosphatase (TRAP) and/or alkaline phosphatase (ALP) (Takara Bio, France) according to the manufacturer’s instructions. Scaffolds were processed for cryosections at a thickness of 6 µm and imaged with the Leica DMIL LED microscope (Vienna, Austria). For immunofluorescence, scaffolds were incubated with antibodies against cathepsin K, DAPI (Santa Cruz Biotechnology, TX, USA), and phalloidin (Cell Signaling Technology, MA, USA). Images were taken with the Leica SP8 confocal microscope (Vienna, Austria). Fiji software was used to quantify TRAP+cells and cathepsin K+ areas.19

Resorption pit assay

OC function was assayed by measuring resorption pit areas on dentine slices. Briefly, OCs were differentiated on dentin slices (Immunodiagnostic Systems Holdings Ltd, Boldon Business Park, UK) for 3 weeks. After gently scraping off adherent cells with 0.1% Triton X-100, dentine slices were stained with 1% toluidine solution and resorption pits quantified by light microscopy using the Fiji software.19

ALP expression and alizarin red staining

OB differentiation was assessed by measuring ALP activity with the p-nitrophenyl phosphate substrate (Sigma-Aldrich, Schnelldorf, Germany) according to the manufacturer’s protocol. OB function was evaluated by alizarin red (Sigma-Aldrich) staining of mineralized areas. Briefly, the dye was extracted with 10% acetic acid and read at 405 nm in a Tecan plate reader (Infinite M Nano+).

Western blot analysis

Cells were lysed in radioimmune precipitation assay buffer (ThermoFisher Scientific, MA, USA). Lysates were separated and transferred to nitrocellulose membranes (Bio-Rad, CA, USA). Primary antibodies against pSTAT3/STAT3, pAKT/AKT, pJNK/JNK (Cell Signaling Technology), PD-L1, cathepsin K, pERK, ERK, actin, and vinculin (Santa Cruz Biotechnology) were used. Blots were imaged on a ChemiDoc system (BioRad, CA, USA). Densitometric analysis was performed using the Fiji software.19

Quantitative real-time PCR

Total mRNA was isolated using the RNeasy Kit (Qiagen, Germany), according to the manufacturer’s protocol. cDNA was synthesized with iScript cDNA kit (BioRad) and processed by quantitative real-time PCR (qPCR) using SYBR Green Master Mix (BioRad) on a LightCycler 96 (Roche Diagnostics GmbH, Germany). Transcript levels were normalized to GAPDH. Primers for GAPDH, PD-L1, and NFATc1 were purchased from Eurofins Scientific (Luxembourg). The primer sequences were as follows: GAPDH (forward, 5’- TCGCTGTTGAAGTCAGAGGAGA-3’; reverse, 5’-GTCTTCACCACCATGGAGAAGG-3’); PD-L1 (forward, 5’-TGCAGGGCATTCCAGAAAGA-3’; reverse, 5’-TAGGTCCTTGGGAACCGTGA-3’); NFATc1 (forward, 5’-CACCAAAGTCCTGGAGATCCCA-3’; reverse, 5’-TTCTTCCTCCCGATGTCCGTCT- 3’).

Liquid chromatography-mass spectrometry analysis

For the profiling of the secretome of the dynamic 3D model, scaffolds were incubated in starving media (0.1% FBS) for 24 hours and supernatant processed and stored at −20°C until further use, as previously described.20 A total of 12 individual proteomic analyses were performed corresponding to starving media and supernatant of HD-PBMC/hMSC, OC/OB, OC/OB+IgG, OC/OB+PD1i, and OC/OB+PD-L1i at day 21 of differentiation using 2 different healthy donors for HD-PBMCs and hMSCs. The analysis of secreted proteins was performed with a nano-Ultra High Performance Liquid Chromatography (UltiMate 3000) coupled to an Orbitrap Eclipse Tribrid mass spectrometer (both Thermo Fisher Scientific, Vienna, Austria) applying data-independent acquisition (DIA) method with settings described in Colleselli et al.21 To deepen the analysis, a pool of all samples was created and used for gas phase fractionation (GPF).22 For protein identification and quantification, Spectronaut (17.7.230531.55965) was used in direct DIA mode running the BGS Factory settings using the human (Uniprot V.10.2021, 20 386 entries) and bovine (Uniprot V.06.2022, 6391 entries) protein database to predict the spectral library. GPF measurements were used as Library Extension Runs to maximize protein identification. Protein lists and quantities were exported and used for further statistical evaluations. The mass spectrometry data are deposited at the ProteomeXchange Consortium via the PRIDE partner repository23 with the dataset identifier PXD047604.

Bioinformatic analysis

To minimize false identification of residual FBS-derived proteins as human proteins, we excluded all proteins classified as bovine. In addition, proteins identified in fractions of cell-free starving medium were considered serum-derived contaminants and were removed. Bioinformatic analysis of the proteomics data was done with Perseus software V.2.0.10.0.24 Only proteins with two valid values in at least one condition were included, resulting in a total of 757 proteins identified from all samples. After filtering, log2 transformed values were normalized by subtraction of the median. The identified proteins were assessed for the presence of a signal peptide or their extracellular localization based on gene ontology (GO) cellular compartment annotations using UniProt database. We classified proteins as uniquely expressed in a specific sample if the expression level was greater than zero in both biological replicates and not expressed if the expression levels were zero in all replicates.

To assess differential protein abundance between conditions, the dataset was filtered to include only proteins identified in all groups, and any missing values were imputed using the normal distribution with a width of 0.3 and a down-shift of 1.8. Group comparisons were made using a two-sample t-test with p<0.05, unless otherwise specified. GO (biological process and molecular function) and reactome pathway analysis of the differentially expressed proteins were performed using DAVID (V.2021)25 against the Homo sapiens background using a count threshold of 2 and an EASE value of 0.1.

Statistical analysis

Each experiment was performed in triplicate using primary cells from at least three different donors. Results are displayed as the mean±SE unless otherwise specified. For statistical comparison of in vitro experiments, we used two-tailed Student’s t-test. P<0.05 indicates statistical significance. Venn diagrams, enrichment bubble plot, and heat map were plotted using http://www.bioinformatics.com.cn/srplot, an online platform for data analysis and visualization.