Stem cell modelling of osteoarthritis using an organ-on-a-chip approach

Abstract

Osteoarthritis (OA) is a chronic degenerative joint disease that affects more than 200 million people globally. Despite its high prevalence, treatment efficacy remains low, largely due to the complex nature of the disease and the significant variability in response to medication of individual patients. Both genetic and environmental factors play a major role in disease progression and in how patients respond to various therapies, making personalised treatment strategies crucial for effective disease management. In light of these challenges, there is an urgent need for reliable, objective tools that can assess the response of individual patients to different medications. This would allow clinicians to tailor treatments based on a patient's unique genetic and biological profile, improving outcomes and minimizing unnecessary side effects. Here we are presenting a method, where we are differentiating mesenchymal stem cells (MSCs) into the chondrogenic lineage using a 3D organ-on-a-chip approach. Two sources of MSCs, the infrapatellar fat pad and abdominal adipose tissue are compared using targeted gene expression analysis and morphological assessment. In addition, we assessed how gene expression is changed after artificially inflammatory exposure with medications compared to that in untreated cells. We found that both abdominal adipose and infrapatellar fat pad MSCs were capable of differentiating in the chondrogenic direction however exhibited differences in morphology and gene expression status. These findings suggest that the combination of MSCs and the organ-on-a-chip platform could offer a viable alternative to cartilage biopsy for providing deeper insights into individual genetic susceptibilities related to OA and facilitate the development of personalised treatment strategies, paving the way for more effective management of this chronic and often debilitating condition.

Supplementary Information

The online version contains supplementary material available at 10.1186/s41232-025-00381-6.

Introduction

Osteoarthritis (OA), a debilitating degenerative joint disease, is characterized by progressive damage to joint cartilage, resulting in swelling, stiffness, and pain in the affected joint [31]. Over time, patient mobility and overall well-being is severely impacted, and in serious cases, knee replacement surgery may become necessary [24]. With an expected global prevalence of more than 240 million people, OA poses a substantial health concern [35]. Furthermore, there is evidence suggesting a genetic predisposition to the disease, as it tends to cluster within families and among individuals of the same ethnic background. However, specific single genes responsible for OA etiology have not yet been identified [77]. Addressing OA is crucial not only for alleviating pain and improving mobility but also for enhancing the overall quality of life of affected individuals. Advancing research and understanding of the genetic and environmental factors of this disease will be essential for developing more effective treatments and preventive measures. Additionally, promoting joint health awareness and early intervention strategies can play a significant role in managing this prevalent and impactful condition [53, 74]. One of the significant challenges in managing OA is the limited regenerative potential of joint cartilage. This crucial tissue, which is primarily affected in OA, exhibits only minimal self-repair capabilities in response to damage [67].

Currently, several interventions are available on the market, including Triamcinolone, Diclofenac, Hyaluronic Acid, and platelet-rich plasma (PRP), aimed at slowing down the progression of OA. However, these treatments have individual variations in treatment response and may lead to significant side effects and toxicity when used in the long term [11]. This variability in response makes it challenging for clinicians to define the ideal treatment plan for each patient. To improve the management of OA, further research and innovative approaches are required to identify effective treatments with reduced side effects. Efforts should also focus on developing objective measures to personalize treatment plans based on individual patient characteristics, ultimately providing better outcomes and an improved quality of life for those affected by this debilitating condition.

Organ-on-a-Chip (OoC) technology has revolutionised biomedical research, offering a new dimension for studying complex diseases in a controlled environment [63]. Several OoC models have been developed to study OA, each offering unique advantages for investigating specific aspects of the disease. These models include single-organ chips focused on cartilage, bone, or synovium, as well as multiorgan chips to capture the interplay between various tissues in the joint. By incorporating patient-derived cells, these models can even simulate patient-specific OA conditions, providing a personalized approach to treatment development [2, 3].

Mesenchymal stem cells (MSCs) are adult stem cells found in various tissues, including bone marrow and adipose tissue or the umbilical cord [7, 22]. They possess unique properties that make them ideal candidates for OA research. MSCs have the capacity to differentiate into chondrocytes, the specialized cells responsible for producing and maintaining cartilage [10]. In OoC models, MSCs seeded onto a chip can be directed to differentiate into chondrocyte-like cells under specific conditions, leading to cartilage formation [48].

The ability to study chondrogenesis in controlled OoC environments allows the optimization of conditions for cartilage repair and the identification of factors that promote cartilage regeneration. OoC models that incorporate MSCs allow the creation of more physiologically relevant disease models. By using patient derived MSCs, the OoC model can better represent the genetic and phenotypic characteristics of an individual's OA condition [40]. This personalized approach to disease modelling enables the testing of the efficacy of potential drugs or therapeutic interventions in a patient-specific context, improving the accuracy of drug screening and reducing the need for animal testing [30, 43].

Here, we present an in vitro model, in which we can provide patient-specific information about the response of single patients to medication with minimal intervention [54]. Our study involved the use of MSCs obtained from six different patients. These MSCs were then differentiated into the chondrogenic lineage. For further applications the main advantage of our model is that it eliminates the need for cartilage biopsy, enabling the testing of medications tailored to individual patients. This approach has potential as a more effective and personalized treatment for patients with OA. To assess the effectiveness of the model, we compared MSCs from two different sources: two samples derived from abdominal fat tissue and four samples obtained from the infrapatellar fat pad tissue. This allowed us to evaluate potential differences in their differentiation potential. To replicate the physiological conditions of the joint environment, including 3D tissue architecture, pressure, and nutrient gradients, we developed a microfluidic chip. This innovative chip enabled us to incubate the cells in a 3D environment for several days and evaluate the response to various medications, seeking the best treatment outcome. For our experimentation, we first applied a differentiation protocol to MSCs in 2D and guided them to the chondrogenic lineage within the 3D chip. Additionally, to mimic chronic inflammation, we artificially introduced Interleukin 1-beta and TNF-alpha. After the treatment, we objectively assessed individual treatment responses using biomarkers to evaluate gene expression alterations related to the treatment outcomes.

OoC technology, when combined with MSCs, offers a powerful platform for advancing our understanding of OA and exploring novel treatment strategies. With continued advancements in the field, OoC models have the potential to revolutionize OA research, ultimately leading to improved patient outcomes and a better quality of life for those affected by this debilitating condition.

Materials and methods

Cell isolation

The infrapatellar fat pad or cartilage was isolated during total knee replacement [20]. We obtained ethical approval and informed consent from all patients. The study received approval from the Ethics Committee of Lower Austria (EK 1020/2020, EK 1100/2020) and was conducted in accordance with the Declaration of Helsinki (1969, including any subsequent revisions) of the World Medical Association. The isolation, cultivation, characterization or differentiation (3D) of the MSCs from the infrapatellar fat pad has already been described by [46]. The patients included one woman and three men aged between 42 and 82 years. The numbers 191 (f, age 81), 192 (m, age 46), 193 (m, age 42) and 194 (m, age 82) were assigned for clear identification. Chondrocytes were isolated and cultivated as described by [6].

Adipose-derived MSCs from abdominal fat were purchased from Innoprot (Bizkaia, Spain). The patients included two women aged 26 and 29 years. The numbers 189 (age 26) and 190 (age 29) were assigned for clear identification. Lymphocytes, which served as controls, were isolated from whole blood in Ficoll-Paque^TMPlus (Cytvia, Marlborough, MA, USA) according to the manufacturer’s protocol.

Pre characterization

Prior to their use in the actual experimental procedures, we subjected the MSCs to pre characterization procedures, which are described in the supplementary materials. This included subjecting the MSCs to trilineage differentiation protocols and assessing the expression of key markers, namely, CD73, CD90, CD105, CD11b, CD19, CD34, CD45 and HLA-DR.

Trilineage differentiation

Adipogenic differentiation potential was assessed using Oil Red O staining, as previously described by [38]. Briefly, MSCs were seeded at a density of 2*10⁵ cells/well in 6 well plates and cultured in normal growth medium until they reached 100% confluency for 3–5 days, before differentiation was initiated. Differentiation medium was prepared from serum-free growth medium supplemented with StemXVivo® Adipogenic Supplement (Bio-Techne Ltd., Abingdon, United Kingdom) and 10% FCS, which was changed every 3–4 days. After 21 days, the cells were fixed in 10% formalin for 30 min, rinsed with PBS, and stained with Oil Red O (Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer’s instructions. The fixed cells were washed with distilled water, followed by 60% isopropanol for 2 min, and then incubated with Oil Red O working reagent (3 parts Oil Red O stock solution, 2 parts distilled water) for 5 min. Excess dye was removed with distilled water until the water remained clear. The stained cells were visualized using a phase contrast microscope equipped with a 10X dry objective. Additionally, dye incorporation was quantified by extracting the dye with 100% isopropanol, centrifuging at 14 000 × g for 5 min, and measuring the absorbance of the supernatant in triplicate at 490 nm on a Synergy 2 plate reader (Biotek, VT, USA) (Supplementary Fig. 1A).

The potential for osteogenic differentiation was assessed using Alizarin S staining, as previously described by [46]. Briefly, MSCs were seeded into 6-well plates at a density of 2*10⁵ cells/well and cultured in normal growth medium until they reached 100% confluency over a period of 3–5 days, after which differentiation was initiated. The differentiation medium, which consisted of serum-free growth medium supplemented with 100 nM dexamethasone (Sigma-Aldrich), 50 μg/ml ascorbic acid (Sigma-Aldrich), and 10 mM β-glycerol phosphate (Sigma-Aldrich), was supplemented with 10% FCS and changed twice a week. After 21 days, the cells were rinsed with PBS, fixed in 10% formalin for 30 min, and stained with aqueous 2% Alizarin Red S solution (Sigma-Aldrich), pH 4.3, by covering the cells with 1 ml of dye solution for 30 min at room temperature. Mineralized areas that appeared red were visualized via a light microscope using a 10X dry objective. The stained cultures were then incubated with 500 µl of 10% acetic acid for 30 min, collected with a cell scraper, vortexed, and heated at 85 °C for 10 min. The sample was cooled and centrifuged at 20,000 × g for 15 min. The supernatant was neutralized with 500 µl of 10% ammonium hydroxide, and the absorbance was measured in triplicate at 405 nm using a Synergy 2 plate reader (Biotek) (Supplementary Fig. 1B).

Chondrogenic differentiation was assessed through chondropellet formation, following the methodology described by [42]. In brief, 2.5*10⁵ cells were mixed with chondrogenic medium consisting of DMEM high glucose (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 1% ITS liquid media supplement (Sigma-Aldrich), 100 nM dexamethasone (Sigma-Aldrich), 50 µg/ml ascorbic acid (Sigma-Aldrich), 1% nonessential amino acids (Gibco, Thermo Fisher Scientific), 5 ng/ml TGFβ−3 (PeproTech, Cranbury, NJ, USA), and 4% methylcellulose (Sigma-Aldrich), and centrifuged at 4164 × g for 10 min. The resulting pellets were cultured for 21 days at 37 °C/5% CO₂ in 15 ml tubes with loose caps, with media changes twice a week. The pellets were then recovered from the culture medium, the excess liquid was removed, and they were placed on base molds (Thermo Fisher Scientific) prepared with a layer of frozen tissue matrix (Tissue-Tek® O.C.T.™, Sakura Finetek, CA, USA), and stored at −80 °C for at least 24 h. Cryosectioning was performed on a cryostat device (Cryostar NX70, Thermo Fisher Scientific), and 6 µm thick slices were placed on adhesive glass slides (Thermo Scientific™ SuperFrost Plus™, Thermo Fisher Scientific). The slides were dried and fixed in cold acetone (−20°C) for 10 min. Histologic staining was performed with Alcian Blue 8GX (Sigma-Aldrich) prepared from 1 g Alcian Blue 8GX, 97 ml distilled water, and 3 ml 96% acetic acid (Sigma-Aldrich) for 30 min. The sections were washed in running tap water for 1 min, dehydrated through 95% ethanol and 2 changes of absolute ethanol for 3 min each. The slides were mounted with xylene before a cover glass was added. Sulfated glycosaminoglycans appeared blue and were visualized under a light microscope to determine the diameter of the chondropellets. Stained slices were analyzed via a light microscope equipped with a 4X dry objective (Supplementary Fig. 1C).

Flow cytometry

Cells that adhered to the cell culture dish 48 h postisolation were subjected to flow cytometry using a StemFlow kit (BD, Franklin Lakes, NJ, USA). Despite the manufacturer's recommendation, 10⁵ cells were stained with 4 µl of positive or negative marker and isotype antibody cocktails, and 1 µl of compensation control antibodies was used. Positive markers included CD90, CD105 and CD73 and negative markers included CD34, CD11b, CD19, CD45, and HLA-DR. Data acquisition and analysis were performed using a CytoFlex S device (Beckman Coulter, Brea, CA, USA) and FlowJo software v10.0.7, respectively (Supplementary Fig. 2).

Cell differentiation in 2D and loading

MSCs were seeded in differentiation media, Dulbecco’s Modified Eagle Medium, GlutaMax™ Supplement, 1% Anti-Anti, 2.5% FBS, 1% MEM NEAA (nonessential amino acids) (Gibco, Thermo Fisher Scientific), 100 nM Dexamethasone (Sigma-Aldrich), 50 μg/ml L-Ascorbic acid (Sigma-Aldrich), 1% Insulin-Transferrin-Selenium Premix (Corning®, Sigma-Aldrich), 5 ng/ml Transforming Growth Factor ß (Sigma-Aldrich), in 75 cm² culture flasks (Nunc, Rochester, NY, USA) and cultivated at 37 °C in a humified environment with 5% CO₂ for 3 weeks.

After three weeks of differentiation the MSCs were stained in the culture flasks with CellTracker®Green CMFDA Dye (Thermo Fisher Scientific) according to the manufacturer’s protocol. MSCs were detached with accutase (Corning, Corning, NY, USA), counted and viability was measured with a Countess II FL (Thermo Fisher Scientific). Afterwards the cells were washed and adjusted to a cell concentration of 1*10⁷/ml. Three parts of the cells were mixed with 2 parts of 100 mg/ml fibrinogen (Sigma-Aldrich). Then, 2 U/ml thrombin (Sigma-Aldrich) in 40 mM CaCl₂ (Thermo Fisher Scientific) was added to the cell/fibrinogen mixture and 12 µl was immediately transferred to a chamber of the organ-on-a-chip device (Pregenerate, Vienna, Austria).

The experiment lasted ten days (two days were for the establishment of a three-dimensional structure, eight days were for cell exposure to various treatments in one noninflamed chip and one inflamed chip) to mimic the physiological conditions of a joint, such as intermittent stress and a nutrient gradient [54]. Two chips were loaded per patient. On one chip, the cells were treated with differentiation medium and the medications. In the other chip the cells were treated with differentiation medium and artificial inflammation and the different medications. For artificial inflammation required 4 ng/ml TNF-alpha (Gibco) and 4 ng/ml IL-1β (Gibco) were added to the differentiation media (see pipetting scheme in Supplementary Fig. 3). The following treatments based on the literature were used: 400 µg/ml Hyaluronic acid [17] (Rottapharm Madeus GmbH, Monza, Italy), 60 µg/ml Triamcinolone [15] (Acros Organics, Verona, Italy), 6 ng/ml Diclofenac [58] (Novartis, Basel, Switzerland) and 17.5 ng/ml PDGF [57] (Gibco). The media was changed every 2 days (see Fig. 1).

Fig. 1.

Timeline of the PCR determinations and administration of the various media in 2D and 3D. Abdominal Adipose MSCs (aAMSC) and Infrapatellar fat pad MSCs (IFP-MSC) were seeded in stem cell medium (blue line) for 2–3 weeks to expand and thereafter the MSCs were differentiated (orange line) in 2D for another three weeks. Afterwards, the cells were seeded in 3D and treated in differentiation medium + the medications ± inflammation (yellow line). In parallel, MSCs were grown in differentiation medium (orange line) for further 10 days as control in 2D. The different colored stars indicate that qPCR analysis of the biomarkers and/or CD markers have been done at this time

RT-PCR

After ten days on the chip, the MSCs were harvested with a buffer solution consisting of 1% ß-mercaptoethanol (EMD Millipore Corp., Billerica MA, USA) and lysis buffer (from an in-house high-performance RNA bead isolation kit). The mRNA isolation was performed with a hydrophilic bead-based system (Cytvia) according to the manufacturer’s protocol. cDNA was prepared using the Maxima H minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol.

CD markers, which are listed in Supplementary Table 1 (designed and established previously at Pregenerate GmbH) were analyzed with SYBR Green (Applied Biosystems, Thermo Fisher Scientific) according to the manufacturer’s protocol. Afterwards distinct biomarkers (optimized by company Thermo Fisher Scientific, listed in Supplementary Table 2) were analyzed using Taqman Technology (Thermo Fisher Scientific) according to the manufacturer’s protocol.

Microscopy

Microscopy was performed with an EVOS 7000 M imaging system (Thermo Fisher Scientific), and the results were used to determine the morphology and cell count. CMFDA staining was performed before loading the cells, and images were taken to track the cells on day 1, 3 and 8 after loading into the chip. An EVOS™ LED Cube (Thermo Fisher Scientific) with a wavelength excitation (Ex) of 482/25 and an emission (Em) of 524/24 was used. A maximal intensity projection of a z-stack 1-channel overlay was performed. The EVOS 7000 M imaging system with a 10X EVOS 8.4 objective was used for phase contrast and in the GPF channel for 2D and 3D culture.

Data analysis

Gene expression analysis was performed via the delta Ct method [9]. To determine differences in gene expression between samples, this method was used for the normalization with housekeeping genes including GAPDH [4] and eEf1 [26] for the CD Markers. Two more housekeeping genes namely, YWHAZ and HPRT1 [32], were used for the biomarkers. The results were calculated by the housekeeping gene (ref cq) – gene of interest (mean cq) = △ct and subsequently 2^△ct [9].

For the evaluation of the CD markers, a heatmap was created using GraphPad Prism software 9. For biomarker evaluation, dot plots were created using GraphPad Prism. In all cases 3 technical replicates were performed.

Results

Morphological structures of MSCs from different sources of MSCs in 2D versus 3D

Adipose MSCs (aAMSCs) from 2 patients and infrapatellar fat pad MSCs (IFP-MSCs) from 4 patients were used to examine the morphology in 2D. They were expanded on a cell culture flask,differentiated into chondrocytes and then loaded within a chip for resembling the 3D environment of the knee. Articular chondrocytes (ACs) were grown as control. MSCs spontaneously form spheroids when embedded in a fibrin matrix [12]. To circumvent this effect, the MSCs were first multiplied in cell culture flasks and then differentiated for 24 days to embed individual cells in the matrix.

After 6 days in culture, both aAMSCs and IFP-MSCs exhibited elongated, spindle-shaped cells similar to the morphology of the AC control (Fig. 1A, bottom). Compared with IFP-MSCs, aAMSCs tended to be larger and thicker with a distinct cytoskeleton, but IFP-MSCs, had a finer structure and thin tube networks (Fig. 2A, left). After 24 days in differentiation medium, the aAMSCs spread out and formed large, flattened cells with prominent round nuclei and the IFP-MSCs largely retained their spindle-like structure. (Fig. 2A, right). However, the state of health of the cells was not affected, as more than 90% of the living cells were detected when the cells were harvested and measured with the Countess II FL.

Fig. 2.

Morphology in 2D and 3D. A Phase contrast images of the different cell types in 2D: Abdominal Adipose MSCs (aAMSC), non-differentiated after 6 days in culture with stem cell media on the left and the same sample differentiated for 21 days in differentiation media on the right. Infrapatellar fat pad MSCs (IFP-MSC), non-differentiated after 6 days in culture with stem cell media on the left and the same sample differentiated for 21 days days in differentiation media on the right. Articular chondrocytes (AC) in chondrocyte growth media on the bottom. Microscope: EVOS 7000 M, 10x (B) Phase contrast and Fluorescence images of different cell types: aAMSC loaded in the chip after 1 day (left & middle) and after 8 days (right). IFP-MSC loaded in the chip after 1 day (left & middle) and after 8 days (right). AC loaded in the chip after 1 day (bottom) Microscope: EVOS 7000 M, 10x, maximal intensity projection of a z-stack overlay

After 24 days in differentiation medium, the cells were harvested and loaded onto the chip. Compared with the ACs, both the aAMSCs and IFP-MSCs formed a network in the chip, which remained in a round state. There was no spontaneous spheroid formation and the matrix remained stable until the end of the experiment (Fig. 2B, left). We stained the MSCs with CMFDA to track them. In the middle of Fig. 2B you can see the cells after one day in the chip and in Fig. 2B on the right after 8 days. The cells did not migrate and remained in their original positions. It should also be noted that the cells appear round in the fluorescence image, although the filigree structures of the network cannot be imaged with the current setup of this fluorescence microscope.

CD marker expression of aAMSCs and IFP-MSCs

The MSCs were continuously tracked with the following CD markers using qPCR and are shown in a heatmap: CD29, CD34, CD44, CD45, CD73, CD90, CD79, CD166 and CD105. The analysis was performed in 2D after the expansion phase in stem cell medium and after the growth phase in differentiation medium, as well as after 10 days within the chip (Fig. 1). The heatmap shows the respective gene expression in rows in relation to all analyzed samples with the same gene.

CD29 is a surface marker expressed on MSCs and a cartilage progenitor cell marker [68, 69]. The marker was detected in all MSCs but showed a different expression pattern. IFP-MSCs generally showed higher initial expression in the 2D culture, which decreased over time in the differentiation medium and then increased again in 3D. This effect was not observed in aAMSCs, where it either remained the same or increased in 3D (Fig. 3, first row). CD34 is a positive marker in hematopoietic stem-progenitor cells from human bone marrow [51]. We also observed that CD34 was generally expressed at low levels and not expressed except in 2 IFP-MSC samples in 2D (Fig. 3, second row).

Fig. 3.

Comparison of CD Markers of Adipose MSCs (aAMSC) and infrapatellar fat pad MSCs (IFP-MSC) in 2D and 3D. The colored scale in the lower right corner indicates whether the CD Marker for certain cells is rather expressed (color gradient becomes lighter up to yellow) or rather not expressed (color gradient becomes darker and turns purple). The empty field and the empty field with a cross indicate that the CD marker is expressed very low or not at all or is undetermined. The CD Markers expression is determined from the 2^delta.^Cq method normalized with the mean of the housekeeping genes GAPDH and eEf1

CD44, which is essential for cartilage homeostasis [36], was more highly expressed in IFP-MSCs than in aAMSCs. The expression of this gene was greater in the 2D cells and decreased towards the chondrocyte level upon differentiation (Fig. 3, third row). CD45, a hematopoietic marker [62], was not expressed (Fig. 3, fourth row). The expression of CD73, a MSC marker and regulatory factor in chondrogenic differentiation [47], was the same as that of CD44,it was more highly expressed in IFP-MSCs and decreased in the differentiated cells at chondrocyte level (Fig. 3, fifth row). CD90, which also occurs on chondrocytes [44], was expressed in both IFP-MSCs and in aAMSCs, and here, we observed that the expression level in the differentiation medium was lower (Fig. 4, sixth row). CD79, which exists nearly exclusively on B cells [13], was not expressed (Fig. 3, seventh row). All the MSCs in 3D culture showed greater CD166 expression than did those in 2D culture (Fig. 4, eighth row). CD105 decreased with the differentiation medium and decreased to the level of chondrocytes in 3D culture (Fig. 3, ninth row).

Fig. 4.

Comparison of degenerative gene expression. aAMSCs are shown in the legend as a blue sphere (n = 2) and IFP-MSCs as an orange square (n = 4). Untreated 3D and Inflamed 3D serve as controls on the chips (2 chips were used per sample), so there are 12 data points here; in all others, six data points were plotted when expressed. The abbreviation U stands for the non-inflamed chip and the abbreviation I stands for the inflamed chip. Expression rates resulting from the 2^delta.^Cq method normalized with the mean of the housekeeping genes GAPDH, eEF1, YWHAZ and HPRT1

Degenerative gene expression

In the following, we used a noninflamed chip and an inflamed chip with medications, each with an untreated (differentiation media only) and an inflamed control (Fig. 1). In addition to the chips, we also show the same cells that had grown in 2D before and during 24 days of differentiation (which are referred to as “stem cells 2D” or “differentiated stem cells 2D 24 days” in the figures) or parallel to 34 days in differentiation media (“differentiated stem cells 2D 34 days”).

MMP1 degrades extracellular matrix collagen and contributes to cartilage destruction [25]. Except for 2 IFP-MSC samples, MMP1 was expressed at low levels or not at all. There was a small increase in the 3D environment and greater expression in the inflamed environment. Except for triamcinolone, the medications had only a minor effect in the noninflamed or inflamed environment (Fig. 4A). MMP2 is involved in the degradation of collagens, proteoglycans and fibronectin [23]. Neither the medication nor the environment (2D, 3D) had a major effect on the expression of MMP2 (Fig. 4B).

MMP3 leads to the activation of collagenase in articular cartilage [45]. In 2D, there was some low MMP3 expression which was reduced by the longer growth phase in 2D. An increase in expression was observed in 3D, which was suppressed by triamcinolone. The other medications had no effect in the different environments (Fig. 4C). MMP13 has the ability to cleave type 2 collagens [29]. One IFP-MSC sample showed greater MMP13 expression than the other samples in the inflamed area. For the remaining samples, no change between the noninflamed and inflamed environments was detected, and the medications had little effect (Fig. 4D).

ADAMTS5 contributes to the pathogenesis of OA because it is the major aggrecanase [33]. A higher expression could be measured in 2D, which was reduced with longer incubation or in 3D. In the inflamed environment, the cells were stimulated again, but the medications had no effect (Fig. 4E). Col10 can be used as a marker for the differentiation of hypertrophic chondrocytes [28]. While MSCs cultured in 2D in stem cell medium showed no or minimal Col10 expression, the levels increased once the cells were cultured in a 3D environment. The noninflamed samples treated with the medication showed a slight decrease in expression (Fig. 4F).

Regenerative gene expression

The activities of MMPs are regulated by TIMPs [75]. In general, TIMPS tended to be more highly expressed in aAMSCs than in IFP-MSCs. In any case, there was no difference between 2D or 3D or between the medications (Fig. 5A). Timp2 is involved in the activation of MMP2 [23]. There was an increase in the expression in the differentiation medium in 2D, but there was no difference in the expression between the aAMSCs and IFP-MSCs in 3D or between the medication groups (Fig. 5B). Col1 plays a fundamental role in most connective tissue types [65], and its expression is greater in 3D culture than in 2D culture. However, the medications had no effect (Fig. 5C).

Fig. 5.

Comparison of regenerative gene expression. aAMSCs are shown in the legend as a blue sphere (n = 2) and IFP-MSCs as an orange square (n = 4). Untreated 3D and Inflamed 3D serve as controls on the chips (2 chips were used per sample), so there are 12 data points here; in all others, six data points were plotted when expressed. The abbreviation U stands for the non-inflamed chip and the abbreviation I stands for the inflamed chip. Expression rates resulting from the 2^delta.^Cq method normalized with the mean of the housekeeping genes GAPDH, eEF1, YWHAZ and HPRT1

ACAN, identified as a constituent of the extracellular matrix proteoglycan within articular cartilage [14], is highly expressed within IFP-MSCs. This observation also allows for the discernment of the impact stemming from variations between inflamed and noninflamed environments. The medications had no effect on ACAN expression (Fig. 5D). Sox9, a regulator of chondrogenesis [55], was suppressed in the inflamed environment; similarly, triamcinolone was shown to suppress the expression of SOX9 (Fig. 5E).

Growth factor and cytokine expression

IL-1 A, recognized as a proinflammatory biomolecule [66], exhibited initial expression in 2D culture that diminished over time. Notably, aAMSCs showed higher levels of expression than did IFP-MSCs. This expression pattern was modulated by the administration of triamcinolone, leading to suppression, while PDGF induced an increase within the inflamed milieu (Fig. 6A). The inhibitory action of IL-1RA, which obstructs the interaction of IL-1α and IL-1ß, has been documented [61]. The discernible influence of the inflammatory context was evident in 3D, with triamcinolone effectively curtailing IL-1RA expression in both environments, even reaching negligible levels in the noninflamed setting. Moreover, certain samples exhibited heightened responsiveness to PDGF within the inflamed environment (Fig. 6B).

Fig. 6.

Comparison of cytokines and growth factors. aAMSCs are shown in the legend as a blue sphere (n = 2) and IFP-MSCs as an orange square (n = 4). Untreated 3D and Inflamed 3D serve as controls on the chips (2 chips were used per sample), so there are 12 data points here; in all others, six data points were plotted when expressed. The abbreviation U stands for the non-inflamed chip and the abbreviation I stands for the inflamed chip. Expression rates resulting from the 2^delta.^Cq method normalized with the mean of the housekeeping genes GAPDH, eEF1, YWHAZ and HPRT1

IL-6 plays a pivotal role in modulating hematological and immune responses [19]. The degree of variability in IL-6 expression diminished when transitioning to a 3D culture environment, while a discernible immune response became evident within the inflamed context (Fig. 6C). In parallel, TNF-alpha, a proinflammatory molecule that enhances the secretion of various cytokines [52], was detected in both 2D and 3D samples within the noninflamed environment, with heightened detection levels becoming evident in the inflamed setting (Fig. 6D).

IGF1 is known for its capacity to stimulate the synthesis, proliferation, and survival of the cartilage matrix [72]. IGF1 expression decreased over time in 2D cultures, while it was more prominent in aAMSCs. The administration of medications did not exert any discernible influence (Fig. 6E). TGFß3, acknowledged for its potential to drive chondrogenesis and aid in the restoration of degenerated articular cartilage [73], exhibited a decrease in expression over time in 2D culture, followed by a subsequent increase upon transitioning to 3D culture. However, within the inflamed setting and under the influence of the administered medications, TGFß3 expression once again decreased (Fig. 6F).

Discussion

In this study, we present an innovative model that enables the evaluation of patient-specific medication responses of cartilage differentiated mesenchymal stem cells. Leveraging the OoC technology, we developed a cutting-edge platform for personalized testing of medication efficacy. Our model facilitates the simultaneous testing of four different medications, Hyaluronic acid, Triamcinolone, PDGF (as a standardized representation of PRP), and Diclofenac, along with two control agents, all within a single chip. The focus of our investigation was osteoarthritis, and to achieve comprehensive results from intra- and extra-articular sources, we incorporated two distinct cell origins: the first source comprised mesenchymal stem cells derived from abdominal fat, while the second source consisted of local mesenchymal stem cells obtained from the infrapatellar fat pad. This dual-source approach allowed us to explore the potential differences in medication response based on the origin of the stem cells.

To establish the feasibility and functionality of the platform, we initially expanded the cells in 2D and monitored their growth. The cells exhibited the characteristic typical spindle-like structure during this phase [59]. Although these structures were still evident after the addition of the differentiation medium, the cells became broader and flatter [27]. Moreover, when observing the cells in a 3D environment, contrasting characteristics were observed in comparison to those of chondrocytes. The adipose-derived MSCs demonstrated network-like structures, suggesting their potential angiogenic properties [37]. These MSCs have the capacity to produce various factors that can support angiogenesis and promote vascularization in co—culture scenarios. This feature holds immense promise for enhancing tissue regeneration and repair [70]. To characterize whether the environment influences the expression of characteristic CD markers of MSCs according to the ISCT [18], we demonstrated that MSCs derived from adipose tissue exhibited the expected negative expression of CD45 and CD79 and low expression of CD34 in IFP-MSCs, which occurs in freshly isolated cells and diminishes during cell expansion [41]. We observed some variability in the expression patterns of CD44, CD73, and CD105 in one of the samples. Given that our study involved patients with osteoarthritis, this discrepancy could be attributed to the unique phenotype associated with the disease. The complexities of the osteoarthritic phenotype warrant further investigation to better understand its implications for MSC behaviour and therapeutic potential [16].

The biochemical inflammation induced by the addition of TNF-alpha and IL-1ß, was added as it mimicked OA [54] and we wanted to assess biomarkers in response to inflammation. We aimed to discover which markers were strongly regulated and influenced by medications. The observed upregulation of MMP1 and MMP3 in the inflamed environment aligns with previous studies indicating their crucial roles in tissue remodeling and inflammation [39]. MMP1 and MMP3 are known to degrade various extracellular matrix components, contributing to tissue destruction and inflammation progression [25]. The specific upregulation of these enzymes in the inflamed environment shows their potential as therapeutic targets for modulating inflammatory processes and tissue damage in related pathological conditions. Because of the chondrogenic differentiation of MSCs, MMP2 is expressed at higher levels than other members of the MMP family, and the relatively low doses of IL-1ß and TNF-α do not appear to have an effect on this process. MMP13 upregulation in the 3D environment facilitates the remodeling of the cartilage matrix during chondrogenesis [56]. The elevated expression of ADAMTS5 is consistent with its recognized role as a major aggrecanase involved in cartilage degradation [64]. Col10 serves as a recognized hypertrophic marker, indicating the cessation of hypertrophic cartilage growth concurrent with the initiation of endochondral bone development. However, the expression of Col10 appears to be distinctive to the in vitro chondrogenic differentiation of MSCs, as documented through its observed upregulation. Despite this observed phenomenon, the comprehensive biological significance of these findings remains elusive, underscoring potential implications for the advancement of tissue engineering and regenerative medicine methodologies [28, 60].

The increased expression of TIMP1 in IFP-MSCs indicates a potential role for TIMP1 in tissue homeostasis and regulation. The observed elevated expression of TIMP1 may be linked to a protective mechanism against excessive matrix degradation, supporting tissue integrity and function [1]. In contrast, the lack of effect on TIMP2 in both the inflamed and 3D environments raises intriguing questions about its regulatory mechanisms. TIMP2 is an essential inhibitor of MMP2, and its presence is expected to influence MMP2 activity. The absence of any noticeable impact on MMP2 expression may point to alternate regulatory pathways that govern MMP2 activity in these specific contexts [8]. The higher expression of Col1 in the 3D environment underscores its importance in maintaining tissue architecture and integrity in a three-dimensional culture setting [50]. We also measured the expression of Col2 and, due to the short duration of the experiment, found a low expression that could not be shown in a diagram. The observed downregulation of ACAN and SOX9 in the inflamed environment may indicate compromised chondrogenic potential and impaired cartilage repair capacity [76].

We showed the upregulation of proinflammatory cytokines such as IL-6, TNF-alpha, IL-1 and IL-1RA. These cytokines are known to mediate inflammatory processes and immune cell recruitment [21]. We were also able to simulate the regulation of IGF1 and TGF-ß3, which affect tissue repair and regeneration, since these growth factors play a crucial role in promoting cell proliferation and differentiation [49]. The observed differences in the expression patterns between aAMSCs and IFP-MSCs highlight the heterogeneity of these cell populations. This heterogeneity can be attributed to their distinct tissue origins and functional properties; however, individual responses are also evident, and we cannot rule out the influence of age or other environmental factors on the unique behaviour of each cell population [5]. Adipose-derived MSCs were obtained from patients aged 26–29, whereas IPF MSCs were obtained from patients aged 42–82. Potential age-related variations in proliferation, differentiation potential, and gene expression should be considered when interpreting the findings.

Understanding these variations is crucial for developing targeted therapeutic interventions and personalized treatment strategies. Our organ-on-a-chip standardized approach provides an opportunity to assess the responses of different cell sources and patients on an individual basis.

Except for triamcinolone, the other medications did not have any major effect on differentiated MSCs. Triamcinolone has been widely studied for its suppressive effects on various immune responses and inflammatory processes [34]. When examining its interaction with MSCs, a complex interplay between the medication and cellular behavior emerges in both noninflamed and inflamed environments. It can modulate the secretion of proinflammatory cytokines and influence tissue repair [71].

Our findings establish the groundwork for an innovative platform designed to enable the personalized assessment of medication effectiveness. Initially, our focus was on examining medications at their reported clinical doses, which we would vary and expand; however, we aspire to broaden our scope by incorporating the patient's own PRP into our investigations.

Due to the limited sample availability, we opted for gene expression analysis, as it provides a quantitative measure of the tissue's immediate response to specific medications and allows for the assessment of a broader range of biomarkers compared to protein-level analysis. However, we acknowledge that investigating cellular responses at the protein level would offer deeper insights into the interactions with the medication and potentially open new avenues for understanding these mechanisms.

By utilizing patient-specific stem cells, we aim to advance personalized treatment strategies, ultimately improving outcomes for individuals with osteoarthritis. Despite the small scale of our study, it provided valuable insights into individual cellular responses and highlighted the importance of mesenchymal stem cell (MSC) source selection. Our findings suggest that the tissue origin of MSCs might be playing an important role in establishing future models for durg testing, emphasizing the need to consider different sources for future testing. While our study revealed these important trends, larger sample sizes will be necessary to establish statistical significance.

Moreover, the ability to test multiple medications simultaneously on a single chip enhances efficiency and reduces the need for excessive resources. It is crucial to highlight the significance of our model's potential applications in advancing precision medicine. The ability to predict patient-specific responses to medications could revolutionize the field of osteoarthritis treatment, steering it toward more targeted and effective therapeutic approaches. As we delve deeper into the mechanisms underlying medication response variations, our model opens new avenues for advancing regenerative medicine and promoting better patient outcomes.

Authors’ contributions

The inception of the conceptual framework and project oversight were supervised by MP. NG oversaw the data acquisition and analysis and led the experimental execution. JF assumed responsibility for the data interpretation and initial manuscript composition. JR critically revised the draft for important intellectual content. AO contributed notably to the cellular characterization and differentiation. ADL and SN were responsible for orchestrating and contributing to sample collection. The collaborative efforts of all the authors culminated in a comprehensive manuscript review.

Funding

The material costs attributed to this work were partially funded by a grant from the Austrian FFG.

Data availability

The datasets generated for this study are available upon request to the corresponding author.

Declarations

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by EK 1020/2020 and EK 1100/2020. The patients/participants provided their written informed consent to participate in this study.

Competing interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.