Establishment of a miniaturized optically accessible microsystems platform technology

In this study, we detail for the first time the design, development, and assessment of a microfluidic osteogenesis-on-a-chip device that incorporates 3D bone cell cultures and dynamic microfluidic conditions as stand alone platform technology operated on a conventional microscope set-up. The initial development strategy was defined to result in (i) a technical solution for the establishment of a miniaturized autonomous incubator with microfluidic cultivating inserts that can be easily fabricated according to requirements and can be designed for a variety of applications, and (ii) the optimization of multi-plexed 3D bone spheroid models for validation of the positive effect of dynamic cultivation to support the differentiation of 3D osteoblast cultures while reducing the disadvantages of spheroid-based cultivation systems such as necrotic core formation.

We prototyped a miniaturized incubator with autonomous CO2 control and integrated thermoregulation (Fig. 1). The incubator was composed of a cover, the body of the chamber, a PDMS microfluidic insert and an anchoring ring (Fig. 1A, A´). Here, we tested several microfluidic inserts for cell cultivation to optimise the design and conditions for 3D cell formation and subsequent cultivation including three distinct microfluidic strategies termed microchannel, microwell and micropillar approach. Initially, a primitive microchannel platform was fabricated (Fig. 1B). The microchannel platform was used to test the physical stability and function of our designed miniaturized incubator system as well as the effect of a dynamic conditions on conventional 2D murine pre-osteoblast cell cultures to act as controls for different platforms for 3D bone formation. Microwell platforms (Fig. 1C) are well-established platforms for 3D cell culture fabrication and cultivation, but the media flow is restricted just to the top part of the microwell, while spheroids inside microcavities may be subjected to irregular and uncontrollable fluid flows. To improve flow profiles and to potentially approximate the basic design towards a niche that resembles trabecular bone, we designed a micropillar platform (see Supp. Chap. 3, 5) and a platform combining microwells with micropillars (Fig. 1D). The whole microfluidic assembly was connected for fluid flow adjustment to a syringe pump and placed on a microscope table for online monitoring (Fig. 1E, E´). To ensure a reproducible cell culture environment outside conventional incubators, the platforms CO2 concentration and temperature were precisely monitored and regulated by an autonomous system (Fig. 1F).

Fig. 1

Design of the miniaturized, microfluidic, and optically accessible chamber. A Assembly parts with holes for inserts (1–5). A´ Top view of cultivation chamber with cross section of PDMS microwell platform. Microwell platform include (6.) media flow part and (7.) microwell part. The microfluidic inserts are anchored by a ring and four screws (8.). B-D Different PDMS microfluidics inside the culture chamber. E, E´ Real prototype of the miniaturized autonomous chamber with control of temperature and CO2 concentration placed on a microscope table for continual observation. F Detailed image of the mounted microfluidic spheroid platform comprising a microfluidic insert, a silicon heating belt, a pt100 temperature sensor and a CO2 sensor for precise control of the cultivation environment

Murine pre-osteoblasts cultivated under dynamic conditions exhibited increased mineralization and more dendritic morphology

Osteoblast differentiation is related to increased extracellular matrix mineralization. To validate the positive effect of our on-chip planar cultivation approach as outlined in the previous chapter, the cells were initially seeded at a concentration of 105 cells/ml and cultivated inside microchannel units of the PDMS insert module. After 4 and 10 days of cultivation, the cells were fixed and stained with Alizarin red to reveal the level of mineralization. The cells cultivated in the microcultivation chamber exhibited opticaly higher levels of Ca2+ deposition within 4 days of cultivation (Fig. 2A-C). The level of mineralized extracellular matrix under static conditions was lower than that under dynamic conditions (Fig. 2C) in both proliferation (Fig. 2A) and differentiation media (Fig. 2B). The intensity of Ca2+ staining also increased during cultivation (Fig. 2D-F) showing a gradual increase in the mineralization of osteoblastic cells when cultivated on-chip under static conditions by threefold (284 ± 97%) and dynamic differentiation media by fifteen-fold (1600 ± 217%) when compared to cells cultured in static conditions using only proliferating media (100 ± 79%) (Fig. 2J). Moreover, cultivation under dynamic conditions led to the formation of micromass condensation spots, which is known to further boost osteogenic mineralization in the condensation core of conventional bone models, also highlighting the capacity of MC3T3-E1 cells to form such enhanced 3D cell culture characteristics under dynamic conditions (white arrow indicators shown in Fig. 2F).

Fig. 2

Impact of dynamic culture conditions on the osteogenic differentiation of 2D on-chip cultures of MC3T2-E1 cells. Osteogenic differentiation capacity (Alizarin red staining) of cells cultivated in the 2D microfluidic biochip system for 4 (A-C) and 10 days (D-F). Significant increased Ca2+ intake was observed in the cells cultivated in differentiation media in static (E, 284% ± 97, p = 0.0029) and dynamic (F, 1600% ± 217, p < 0.0001) conditions using image analysis. Cells cultivated under dynamic conditions also generated micromass formations (F-arrows). G-I Cell spreading analysis using fluorescence staining (blue: Hoechst 33,342; green: F-actin). Cultivation of the cells under dynamic conditions for 4 days led to dendrite formation (I; white arrows) and exhibited a significantly higher aspect ratio (length vs. width depicted in K). This ratio was established as 1.9 ± 0.5, p = 0.9355 for proliferating cells, 1.9 ± 0.5, p < 0.0001 for cells cultivated in differentiation media and 3.6 ± 1.5 p < 0.0001 for cells cultivated in dynamic conditions (K). Scale bars = 100 μm (brightfield images) and 20 μm (fluorescence images)

The so-called osteoblast-osteocyte switch is related to changes in cell morphology, polarization, gene expression, and extracellular matrix deposition as well as occurrence of cell dendrites or pseudopodia during differentiation [5]. To observe such early changes, we next stained the actin filaments after 4 days of cultivation under dynamic conditions and again compared these intricate structural differences to morphological characteristics of the static cultivation controls. Cells were either cultivated in static conditions using proliferation (Fig. 2G) or differentiation media (Fig. 2H) in triplicate for 4 days. Dynamic cultivation was performed in a miniaturized chamber with autonomous CO2 and temperature regulation in three independent microchannels. After initial seeding and adhesion under static conditions, the overnight differentiation medium was pumped at a flow rate of 2 µl/min to each individual cultivation unit for the next 3 days. The results of this experiment suggest that the cells cultivated under dynamic conditions exhibited longer protrusions and displayed thinner morphology in contrast to cells cultivated in static conditions (Fig. 2I, white arrows). Additional image analysis revealed a significantly higher aspect ratio (length vs. width) compared to the cells cultivated under static conditions (Fig. 2K). This effect of media flow force on osteoblast cell spreading is in line with previous studies [6].

The on-chip dynamic 3D culture environment can better induce osteoblast differentiation

To test whether the differentiation capacity can be further enhanced, we next established a microfluidic cultivation protocol for generating bone spheroids. The formation and maturation of MC3T3-E1 cells as spheroids in an advanced nonperfused multiwell platform was optimized and benchmarked against frequently used hanging drop cultivation approaches, that represent the most established technique to investigate primary cell cultures in a advanced 3D environment. To investigate the osteogenic differentiation capacity of MC3T3-E1 spheroids under potentially more reproducible culture conditions, we next fabricated a PDMS microwell platform and investigated the effect of microwell spheroid generation on osteogenic differentiation capacity using qPCR analysis (see Fig. 3A-C). A major advantage of this approach is that in contrast to manual handling with hanging drops and hydrogel-based protocols that are highly variable, spheroids inside a miniaturized platform are less prone to user variability as well as mechanical damage during media replacement while providing multiple technical spheroid replicates at a time from a single injection volume. In detail, we developed a PDMS microwell platform containing 64 microwells with a diameter of 400 μm, a space of 100 μm, and a height of 400 μm. The multi-spheroid inserts were coated with a Pluronic F-127 as anti-adhesive surface coating to prevent cell adhesion and anchored by force to 48-well plates or Petri dishes. The cell suspension was placed in 30 µl drops on the surface of the microwell at initial cell concentrations of 1 × 106, 5 × 106, and 1 × 107 cells per milliliter (Fig. 3C-E). The different concentrations led to the formation of spheroids of diverse sizes and decreased gradually in mass during the cultivation period (Fig. 3D). At the initial cell seeding concentration of 1 × 107 cells per milliliter, the initial size of 144 μm (± 17) at day 1 decreased to 115 μm (± 13 μm difference) and 98 μm (± 11 μm difference) for day 2 and 3, respectively. The initial seeding concentration 5 × 106 led to the formation of 84.0 μm (± 15) spheroids with the decreasing size during 3 day cultivation to 66 μm (± 13 μm difference) after 2 and 59 μm (± 9 μm difference) 3 days of cultivation. At the lowest concentration 1 × 106, the size of the spheroids was 49 μm (± 11) during the first day of cultivation, 39 μm (± 7 μm difference) the second day and 37 μm (± 8 μm difference) the third day of cultivation. Even though a linear spheroid size decrease was observed also for lower seeding densities, the overall shrinking effect was slightly reduced for the lowest seeding density compared to the highest with 25% and 32%, respectively.

Fig. 3

A-C Spheroid formation using the PDMS microwell platform and (D) analysis of spheroid diameters at different initial seeding densities during 3 days of cultivation (n = 60). E Actin-normalized mRNA expression levels of osteoprogenitor, preosteoblast, and osteoblast markers in the monolayer and spheroids were measured by qPCR. n ≥ 3, * p < 0.05, ** p = 0.0019, *** p = 0.0008, **** p < 0.0001. Scale bar = 100 μm, microwell spheroids were captured under 40x magnification

Next, as shown in Fig. 3E, the gene expression levels were examined via qPCR for osteoprogenitor, preosteoblast, and osteoblast markers comparing the monolayer and microwell spheroids of the 1 × 107 cells per milliliter seeding density. Our results indicated an overall improvement of osteogenic gene expression with multiplexed bone spheroids with expression of RUNX2 (254 ± 34%, p < 0.0001), ALP (188 ± 76%, p = 0.0316), COL1a1 (90 ± 15%, ns), OCN 486 ± 220%, p = 0.0019), OSX (544 ± 300%, p = 0.0280), PHEX (295 ± 56%, p = 0.0008), DMP1 (827 ± 102%, p < 0.0001) and E11 (577 ± 358%, p = 0.0375). Notably, the expression levels of all the osteogenic marker genes with the exception of COL1a1 were increased indicating at an improved osteogenic environment in our microwell platform.

The next set of experiments aimed to study the influence of the media flow as potent mechanobiological stimulus on the bone spheroids to further enchance the osteogenic environment. Consequently, we designed a fabricated microfluidic platform fitting into a miniaturized cultivation chamber (see also details in Fig. 1B). First, we tested the microfluidic system consisting of three individual channels with 60 microwells. Microwells with a diameter of 400 μm and a depth of 400 μm were situated in three lines with 100 μm space between each other, and the whole system was connected to the syringe pump (see also Fig. 1C). Cells at a concentration of 1 × 107 cells per milliliter were introduced into the system by cell injection and placed into an incubator overnight. Then, the media flow was set to 2 µl/min, and the cells were cultivated for the next 3 days. The initial spheroids size generated in the closed microwell system was smaller (116 μm ± 23) than the spheroids formed in the open wells (154 μm ± 21) (Fig. 4G). Similarly like in the previous results, the spheroids decreased their mass during cultivation (Fig. 4G). We observed a decreased diameter of spheroids in the order from open microwells (99 μm ± 15), closed microwell platform (95 μm ± 14) to dynamic microwell platform (89 μm ± 21) after next 3 days of cultivation (Fig. 4A, B). The necrotic core formation of these spheroids was visualized by propidium iodide mainly in the centre of the spheroids (Fig. 4A´, B´) and did not show significant differences between dynamic and static conditions (Fig. 4H). As brief example to demonstrate the potential of murine bone spheroid model in regenerative medicine, fusion of spheroids to adult bone was investigated as shown in the supplementary materials (see Supp. Chap. 6). Tissue fusion is a crucial process in embryonic development and also plays an important role in the integration of tissue engineered constructs for regenerative strategies [7]. Also, a decrease of fusion capacity can be correlated with model immaturity [8]. Even though this experimental approach presents an interesting side aspect that highlights the fitness of our murine pre-osteoblasts, subsequent benchmarking activities of our optimized multi-plexed 3D spheroid systems focused on more basic analysis parameters that related to the bone physiological niche rather than regenerative capacity of spheroids.

Even though 3D microwell systems can be used potentially for a variety of applications in biomedical research, they evident lack of media perfusion is a limitation for many bone-on-chip studies, that require a highly dynamic tissue environment. Consequently, we designed and fabricated a micropillar platform forming a flow-through circular area to form bone models under physiologically more relevant conditions found in well-vascularized human bone. While pre-osteoblasts that were grown on flat micropillar chambers under comparable culture conditions as in the microwell system, this particular design that lacked microvacities did not promote any spheroid formation with cell growth on the entire biochip surface (Fig. 4C, D). Nonetheless, a positive effect of fluid flow on both viability (Fig. 4C´, D´) as well as mineralization staining analyzing Ca2+ intake was observable and comparable to that in the simple channel system (Fig. 4D), that lacked the more complex µpillar array. Next, we combined the microcavities as architectural cue for facilitated spheroid formation with µpillars that should simulate trabecular structures of bone tissue in a convergent approach shown in Fig. 4E, F. The depth of the microwells was designed to be only 30 μm to induce cell condensation but to present the effect of media force on 3D cell cultures, and the micropillar array prevents spheroids from being washed away from the surrounding area. The shallow microwells were sufficient for spheroid formation (Fig. 4E, F), but spheroids generated in these structures were smaller than spheroids formed in a closed microwell platform (i.e., 75 μm ± 15 versus 95 μm ± 14 as shown in Fig. 4G). On the other hand, the spheroids cultivated in the microwell-pillar structured dynamic chip showed decreased loss of spheroid mass during cultivation in contrast to static one, 75 μm ± 15 in static and 86 μm ± 18 in dynamic conditions (Fig. 4G), which was also evaluated by increased viability compared to that under static conditions (Fig. 4H). The ratio of live/dead cells in static and dynamic conditions was 1.9 (± 0.5) and 2.1 (± 0.6) for the microwell platform and 1.6 (± 0.3) and 4.2 (± 0.5) for micropillar chip respectively. These results are well correlated with the faster media exchange in microwell-pillar bone-on-a-chip platform than in static conditions or deep microwells (Supp. Chap. 6). Interestingly, a few spheroids that formed in the vicinity of the surrounding pillar structured readily fused and overgrew two pillars. This interaction strongly resembled the previous experiments on shperoid fusion (see Fig. 4E, F in comparison with Fig. 3A,B). Overall, we could established a combined microwell-micropillar system for multiplex bone spheroid formation and cultivation under dynamic conditions that represent a more suitable arrangement mimicking in vivo bone systems aiming and toxicology and drug testing with benefits of our platform approach being highlighted in Fig. 4I.

Fig. 4

Comparison of microfluidic cultivation platforms after 4 days of cultivation. Spheroids cultivated in the microwell (A, B) platform in the cultivation chamber were smaller than spheroids cultivated in the open microwell platform (G). Analysis of mineralization (C, D) and viability (C´, D´) of MCT3T3-E1 cells in micropillar microfluidics, which was not sufficient for spheroid formation. (E, F) Spheroid formation in the microwell-pillar platform. The spheroids cultivated inside the microwell-pillar platform exhibited an altered decrease in sphere size (G), and propidium iodide staining showed the positive effect of dynamic cultivation on spheroids (H). I Summarizing scheme. G n = 120, *** p<0005, **** p<00001. (H) n = 30, **** p<00001. Scale bar = 160 μm

To follow in more details the effect of the newly established platform design on osteogenic differentiation, we compared the mRNA expression of spheroids cultivated in the static and dynamic microwell-pillar platform. The analysis of the advanced cultivation routine was conducted under similar conditions throughout four days of cultivation comprising 1 day of static and 3 days of dynamic culture at a flow rate 2 µl/min. Surprisingly, most of the osteogenic differentiation markers were lower in dynamic conditions than static controls as shown in Fig. 5. The expressions were normalised to spheroids cultivated in static conditions. In detail, we observed expressiosn of RUNX2 (62 ± 5%, ns), ALP (59 ± 19%, p = 0.0459), COL1a1 (84 ± 13%, ns), OCN (12 ± 7, p < 0.0001), OSX (46 ± 19%, p = 0.0078), PHEX (66 ± 16, ns), DMP1 (26 ± 12, *** p = 0.003) and E11 (52 ± 10%, ns), which indicated the murine models sensitivity to fluid flow applications.

Fig. 5

ACTAB-normalized mRNA expression levels of osteoprogenitor, preosteoblast, and osteoblast markers in the monolayer and spheroids were measured by qPCR. n ≥ 3, * p = 0.0459, ** p = 0.0078, *** p = 0.0003, **** p < 0.0001

To understand the apparent sensitivity of our model in more detail, we next used a less integrated mobile bone-on-a-chip platform capable of cultivate in a commercial incubator (Fig. 6A-C). We hypothesized that the attenuating effects of fluid flow on murine 3D biochip models were due to potential fluidic over-stimulation due to change of fluid distributions of the µpillar design. Due to the decrease of cross-sectional area by each and every pillar structure, the regional flow speed thus shear applied to the individual spheroids are being potentiated, which was tested in more detail for microcavity and micropillar chips. The residence time of a defined bolus of 1% Alcian blue solution, which was subsequently pushed through the culture region and finally washed out was investigated using ImageJ using a rather fast syringe pump set to 50 µl per minute. A complete washout cycle took approx. 300 s for the microwell design (Fig. 6D, E), and a 10-fold faster washout of the indicator dye of approx. 30 s in the microwell–pillar platform (Fig. 6F, G). This corresponds to calculated mean velocities of 15.3 and 153 μm/s and fluid shear of 0.005 dyn/cm2 and 0.05 dyn/cm2, respectively. These results suggest that the introduction of an array of µpillars can be used to elevate regional fluid shear to potentially modulate bone physiology aspects of spheroid models. We observed that fluid shear increase by 10-fold improves bone spheroid viability while it attenuated osteogenic potential of the murine bone 3D model. Overall, this can potentially point out at a certain oversensitivity of MCT3T3-E1-based murine spheroids for fluid actuation.

Fig. 6

Schemes (A , B) and photographic image (C) of the mobile dynamic microfluidic culturing platforms suitable for cultivation inside an incubator connected to a linear setup (C). D , F Flow profile visualization and analysis of E microwell and G microwell-pillar bone-on-a-chip systems

Pilot study on the dynamic multiplex spheroid platform as unique opportunity for the study of human 3D bone models

As a final proof-of-concept, we prepared dynamic human bone spheroids-on-a-chip analyses again for microwell and microwell-pillar designs against 2D cultures to see whether the attenuating effects fluid flow on human-bone-derived spheroids derived from knee joint spongiosa were similar to the results of the murine MCT3T3-E1 spheroids, that were used for the optimization and characterization of the proposed microfluidic platform. To initially produce human baseline data on static bone models, we first established primary human osteoblasts from bones obtained after orthopaedic knee surgery (Fig. 7A) using a frequently used enzymatic isolation approach to generate human primary cultures of bone-derived cells (hBDCs). As shown in Fig. 7A-C, hBDCs readily migrated from bone fragments after collagenase II digestion (Fig. 7B, C, arrows) and after expansion, the cells were successfully seeded on the multiplexed spheroid microwell platforms (Fig. 7D). Even though we anticipated that human cells take longer to form spheroids due to donor age and differences in cell doubling capacity, we increased the initial concentrations and to our surprise the hBDC-derived spheroids were forming spheroids overnight (Fig. 7E-G). Notably, similar concentration dependent size decrease dynamics with regard to the previously analyzed murine counterparts were again observable. In brief, the re-optimized concentration of 4 × 107 cells/ml led to the formation of the spheroids in the initial size of 149 ± 34 μm condensing to to 82 ± 22 μm after 3 days of cultivation (Fig. 7H). Reflecting on the condensation dynamics of the murine spheroids of initially 144 μm to 98 μm within a 3-day cultivation peroid, a 4-fold increase of hBDC seeding concentrations could effectively compensate for differences in cell growth and cell activity, and in turn produce spheroids of comparable size and maturation dynamics. For a more detailed look at specific bone mRNA markers, a reduced set of osteogenic markers were analyzed for human 3D bone spheroid cultures in comparison conventional 2D culture. As expected, not only the formation dynamics but also mRNA response patterns were improved similarly to the static murine spheroid models. Figure 7I shows that the mRNA expression levels of human-spongiosa tissue derived 3D spheroids displayed a significant increase in most of the bone markers including RUNX2 (1229 ± 170%, p < 0.0001), COL1a1 (216 ± 42%, p = 0.0145), and OCN (305 ± 37%, p < 0.0001).

Fig. 7

Generation of primary human bone spheroids. A-D Visualisation of the method for osteoblast isolation from human bone fragments. E-G Spheroid formation using the PDMS microwell platform cultivated for 3 days. H Analysis of spheroid diameters at different initial seeding densities during 3 days of cultivation (n  = 150). I Actin-normalized mRNA expression levels of osteogenic markers in the monolayer and spheroids were measured by qPCR. n  ≥ 3, * p  = 0.0145, **** p  < 0.0001. Microwell spheroids were captured under 100x magnification

As a final set of experiments, the response of the primary human bone model to a more dynamic bone-tissue-like cultivation environment was investigated using gradually increasing complexities from open microwells up to fully perfused bone-on-a-chip platforms (Fig. 8A) including static approaches such as the open microwell platform, closed microwell and microwell-pillar designs, as well as more dynamic conditions tested with the microwell-pillar platform with tubes (Fig. 8D) at a flow rate of 2 µl/min through day 4. In brief, the hBDC-derived spheroids for a optimized initial seeding density showed comparable results to the murine models with a decrease of the spheroid size to around 70–100 μm (Fig. 8L) while again improving the viability of the spheroidal bone construct when activating the more physiological dynamic flow conditions. While human spheroids cultivated in the open platforms were the highest in size with a diameter of 63 ± 23 μm after 3 days (Fig. 8D, L), static closed microwell (45 ± 17 μm; Fig. 8E, L) and microwell-pillar platforms (34 ± 12 μm; Fig. 8F, L) showed a slight decrease of spheroid diameters independent of dynamic conditions while significantly improving spheroid viability expressed by calcein/PI ratio by more than 2-fold of the open microwells. Overall, these re-characterization results for the hBDC derived bone spheroid models indicate that human primary cells show higher sensitivity for the static cultivation approaches but retain an improved performance when applying dynamic protocols using fluid perfusion in combination with micropillars.

Fig. 8

Comparison of microfluidic cultivation of human bone-cell-derived spheroids during a total of 4 days of cultivation. Spheroids cultivated in the open microwells (A) exhibited significantly higher size compared to closed platforms (B, C, D). L Graphical analysis of spheroid mass. The closed microwells (B) produced significantly higher spheroids than static closed microwell-pillar platforms (C). There was no significant difference between the dynamic microwell-pillar platform (D) and the closed microwells (B). E-H Propidium iodide staining showed the positive effect of dynamic cultivation on spheroids viability (M). I, J, K Photos of the used platforms. J Relative mRNA expression levels of osteogenic markers of pooled spheroids using the 3D cell culture platform after 3 days of cultivation under dynamic conditions in comparison to static on-chip cultivation regimes. L n = 120, **** p < 0.0001. M n = 70, **** p < 0.0001. J Data are expressed as mean ± SD relative to beta-actin housekeeping gene for n ≥ 2 (Two-way ANOVA with Tukey’s post-hoc test *** p = 0.0002, **** p < 0.0001). Platforms for spheroid mass measurement were captured under 100x magnification. PI staining was analysed for 60x and 100x magnification objectives

As a final effort to demonstrate the successful development of the microfluidic platform technology for not only murine but more importantly human bone models, the gene expression response to the dynamic cultivation protocol again was compared to static biochip cultivation routines to investigate potentially beneficial effects of the media flow on osteogenic markers including ALP, RUNX2, COL1a1 and OCN in the proposed advanced spheroid array platform (Fig. 8J). Notably, the level of ALP (224 ± 67%, p = 0.0002) and COL1a1 (452 ± 86%, p < 0.0001), OCN (204 ± 14%, p = 0.0043) was significantly higher in dynamic conditions than in static ones. On the other hand, the dynamic cultivation did not affect the expression of RUNX2.