A parallelized, perfused 3D triculture model of leukemia for in vitro drug testing of chemotherapeutics

The magnetic lift was constructed to bear a deep 12-well plate (Greiner Bio-One International GmbH, Frickenhausen, Germany) held by a movable stage (figure 1(A)). At the bottom of the magnet lift 12 disc magnets (Ø 25 mm, height: 15 mm; magnets4you GmbH, Lohr am Main, Germany) were positioned directly under the wells of the deep 12-well plate at its lowest point. Above the moveable stage 12 threaded rods with holders for another 12 disc magnets were installed. The threaded rods were fixed at a height very close to the 12-well plate at its highest moving point. The motor moved with a power supply and the movement of the stage was controlled with a control unit.

Zoom In Zoom Out Reset image size

Magnetic nanoparticles (Chemagen, Baesweiler, Germany) were methacrylated and macroporous 3D magnetic hydrogels suitable to create artificial bone marrow analogs were produced as described previously [17] and as shown schematically in figure 1(C). Briefly, 60 mg methacrylated magnetic nanoparticles and 333 mg poly(ethylene glycol) diacrylate with a molecular weight of 6000 kDa (PEGDA 6000, kindly synthesized by the Soft Matter Synthesis Lab, Institute of Biological Interfaces, Karlsruhe Institute of Technology (KIT), Germany) were diluted in 1 ml saturated NaCl solution. The peptide RGD was added as 20 µM RGDSK-linker-acrylate (kindly provided by Dr Hubert Kalbacher, University of Tübingen, Germany) to complete the hydrogel precursor solution. As a porogen, 600 mg NaCl crystals (40–100 µm diameter) were mixed with 150 mg of the hydrogel precursor solution. The solution was cross-linked by adding 45 µl 10% (w/v) ammonium persulfate (APS; AppliChem, Darmstadt, Germany) and 8 µl N,N,N',N'-tetramethylethylenediamine (TEMED; Merck, Darmstadt, Germany). After 3 d of swelling and salt leaching in water, the gels were dehydrated using ethanol series (50%, 60%, 70%, 80%, 90%, 100% (v/v)), frozen in ethanol for 3 d at −80 °C and lyophilized for 24 h. Photographs of hydrogels after production, swelling and freeze drying are shown in figure 1(B).

The exchange of soluble factors within the macropores of the 3D magnetic hydrogels with the surrounding fluid via diffusion was examined by conductivity measurements using NaCl as the diffusible agent. The exchange of ions in hydrogels using the magnetic lift (dynamic hydrogels) was compared to hydrogels placed in a well plate without movement (static hydrogels). Two hydrogels were soaked in 1 M NaCl solution for 1 h. One hydrogel was transferred to a deep 12-well plate containing 5 or 6 ml of ultrapure water and placed in the magnetic lift. The other hydrogel was placed in a non-moved deep 12-well plate containing the same amount of ultrapure water. The conductivity of the surrounding fluid was measured every three minutes with a conductometer LF 197S (WTW, Weilheim, Germany).

CD34⁺ HSPCs were isolated from umbilical cord blood provided by the DKMS Cord Blood Bank (Dresden, Germany) or from the Cord Blood Bank of the German Red Cross (Mannheim, Germany) with approval by the local ethics committee (Ethik-Kommission bei der Landesärztekammer Baden-Württemberg, B-F-2013-111) and informed consent of the parents. Isolation of HSPCs was performed as described previously [17].

HSPCs isolated from umbilical cord blood were cultured at a density of 2.5 × 10⁵ cells ml⁻¹ for 24 h in Hematopoietic Progenitor Cell (HPC) Expansion Medium XF (PromoCell, Heidelberg, Germany) with 1% Cytokine Mix E (PromoCell) prior to each experiment. KG-1a cells (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany) were maintained in Rosewell Park Memorial Institute (RPMI)-1640 medium (Sigma-Aldrich®, Hamburg, Germany) with 20% (v/v) fetal bovine serum (FBS; Sigma-Aldrich®). RNA sequencing data of KG-1a cells is provided to confirm the leukemic nature of this AML cell line (supplementary figure 1 available online at stacks.iop.org/BF/14/035011/mmedia). MSCs isolated from human bone marrow were kindly provided by Prof. Dr. Karen Bieback (Institute of Transfusion Medicine and Immunology Mannheim, Faculty of Medicine, Heidelberg University; German Red Cross, Blood Donor Service Baden-Württemberg—Hessen, Mannheim, Germany). MSCs were cultured in Dulbecco's Modified Eagle's Medium (high glucose) (DMEM, Sigma-Aldrich®) with 5% (v/v) platelet lysate (PL, PL Bioscience, Aachen Germany) and 2 U ml⁻¹ heparin (PL Bioscience). Only MSCs between passage two and six were used for experiments. KG-1a cells in co-culture with MSCs were cultured in DMEM with 5% (v/v) PL, 2 U ml⁻¹ heparin, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. HSPCs in co-culture with MSCs as well as in triple co-culture with MSCs and KG-1a cells were cultured in HPC medium with 5% (v/v) PL, 2 U ml⁻¹ heparin, 1% (v/v) Cytokine Mix E, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. Culture conditions for all cells were 37 °C and 5% CO₂ in humid atmosphere.

To follow the proliferation in a double co-culture, HSPCs or KG-1a cells were stained with the CellTrace™ Violet (CTV) Cell Proliferation Kit (Thermo Fisher Scientific, Braunschweig, Germany) according to manufacturer's instructions using 5 µM CTV before seeding into 3D magnetic hydrogels for dynamic and 2D cell culture plates for static cultures. For the triple co-culture, HSPCs were stained with CTV and KG-1a cells using the CellTrace™ CFSE Cell Proliferation Kit (Themo Fisher Scientific). KG-1a cells were stained with 4 µM of CFSE equivalent to CTV staining. After culture, the stained cells were analyzed by flow cytometry (BD FACSverse, BD Biosciences, Heidelberg, Germany or MACSQuant® Analyzer 10 Flow Cytometer, Miltenyi Biotec, Bergisch Gladbach, Germany). Further description and exemplary histograms can be found in supplementary figure 2.

The chemotherapeutics imatinib (IMA, Sigma-Aldrich®), 5-fluorouracil (5-FU, Merck) and cyclophosphamide (CPA, Thermo Fisher Scientific) were chosen for experiments based on their different mechanisms of action. Concentrations of 0.45 µM, 0.9 µM and 1.8 µM of IMA, 2.5 µM, 5 µM and 10 µM of 5-FU and 5 mM, 10 mM and 20 mM of CPA were used. Concentrations were chosen to match the previously investigated IC₅₀ values of the compounds, along with their halved and doubled concentrations [18, 32, 33]. IMA and 5-FU were dissolved in DMSO and aliquots of 15 μl were stored at −20 °C. The stock solutions were prepared in a way that the desired concentrations could be used with a final concentration of maximum 0.1% (v/v) DMSO in the media. CPA was dissolved in HPC expansion medium and PBS in a ratio of 19:1. It was stored at 4 °C.

The effectivity of chemotherapeutic treatment of LCs in different concentrations was tested using leukemic KG-1a cells. For this purpose, a 12-well plate (Greiner Bio-One International GmbH) was prepared with 2 ml cultivation medium per well. 2 × 10⁴ cells were resuspended in 50 μl cultivation medium and seeded into each well. After that, the chemotherapeutics and their respective controls were added at the desired concentrations. As control, the solvents of the chemotherapeutics were used in the same amounts as in the samples. After one, four and six days, the cells were analyzed with live/dead staining by flow cytometry and the supernatant was collected and stored at −80 °C for metabolic analysis.

The 3D magnetic hydrogels, used in a dynamic 3D culture in the magnetic lift, were seeded with a double co-culture (HSPCs and MSCs or KG-1a cells and MSCs) or a triple co-culture (HSPCs, KG-1a cells and MSCs). For comparison, a static 3D culture containing the triple co-culture in magnetic hydrogels and 2D static cultures in standard tissue culture plates were used. When further referred to 2D culture, this static culture without movement and magnetic hydrogels is meant.

Before use, the lyophilized 3D magnetic hydrogels were sterilized under UV light for 20 min on each side. Into each hydrogel, a cell suspension of 100 µl containing 1.25 × 10⁵ HSPCs (CTV stained) and/or KG-1a (CFSE stained) cells along with 5 × 10⁵ MSCs was seeded [18]. After 10 min of incubation at room temperature, the soaked hydrogels were transferred into a deep 12-well plate and the wells were slowly filled with 6 ml of cultivation medium. For the static 2D culture, an optimal ratio of 1:4 HSPCs to MSCs was determined before using 0.5 × 10⁴ HSPCs ml⁻¹ [16]. Here, a normal 12-well culture plate with a capacity of 2 ml volume was used. Hence, 1 × 10⁴ HSPCs and/or KG-1a cells and 4 × 10⁴ MSCs were seeded in 2 ml culture media.

The cells were treated with 0.9 µM IMA, 5 µM 5-FU or 10 mM CPA. The solvents of the chemotherapeutics were used as controls. For dynamic cultures, the deep 12-well plates with the cell-laden magnetic hydrogels in the wells were positioned on the stage of the magnetic lift, which was placed in a CO₂ incubator, as performed previously by Rödling et al [17]. The static 2D and 3D cultures were carefully placed next to it. After one, two, five and six days, 500 μl of supernatant were taken from 3D cultures. Additionally, after six days 500 μl of supernatant was taken from the 2D cultures as well. The supernatant was centrifuged for 5 min at 300 × g to remove cell residues and stored at −80 °C until metabolic analyses. The cells were retrieved from the hydrogels as described previously by Rödling et al [34].

Depending on the availability, 2–10 × 10⁴ HSPCs were washed with PBS + 0.1% (v/v) FBS and resuspended in 50 µl of PBS + 0.1% (v/v) FBS. The cells were stained with mouse anti-human CD34-PE (Thermo Fisher Scientific) and mouse IgG1-PE conjugate (Thermo Fisher Scientific) as isotype control as well as mouse anti-human CD45RA-APC (Miltenyi Biotec) and mouse IgG1-APC (isotype control; Miltenyi Biotec) following the manufacturers' instructions. Analysis was carried out using flow cytometry (BD FACSverse, BD Biosciences or MACSQuant® Analyzer 10 Flow Cytometer, Miltenyi Biotec).

After experiments, in which a double co-culture of HSPCs or KG-1a cells together with MSCs was performed, the isolated cells were additionally stained with SytoxAADvanced™ (Thermo Fisher Scientific) and AnnexinV-FITC (BioLegend, San Diego, CA, USA) according to the instructions in order to determine dead and apoptotic cells. After staining, the cells were resuspended in 150 μl of Annexin buffer.

After performing triple co-culture experiments with HSPCs, KG-1a cells and MSCs, the isolated cells were only stained with SytoxAADvanced™ after the immunostaining. This was performed as described previously [17].

On three different days 500 µl of supernatant were taken with a syringe from the 3D cell cultures. In this way, the total volume of initially 6 ml was reduced during the growth period. The concentrations of glucose and lactate were measured using a Vi-cell Metaflex (Beckman Coulter, Krefeld, Germany) according to the manufacturer's instructions.

The cell culture supernatants were filled into concentrators with a molecular cut-off of 10 kDa (Pierce™ Protein Concentrators; Thermo Fisher Scientific) and centrifuged at 10 °C for 20 min at 15 000 × g. An equal volume of all three experiments was taken and interblended. The flow-through was used for analysis of the adenosine and amino acid concentrations via liquid chromatography–tandem mass spectrometry (LC-MS/MS). For the mass spectrometry measurements an API 4000™ quadrupole mass spectrometer with an electrospray ionization source was used (Applied Biosystems/MDS Sciex, Framingham, Massachusetts, USA).

In order to determine the adenosine concentration, the samples were diluted at a ratio of 1:10 in ultrapure water and separated by high-performance liquid chromatography (HPLC) (Agilent 1100, Waldbronn, Germany) using a column with 125 mm × 4 mm (LiChrospher® 100 RP C-18, 5 μm, Merck) and a gradient elution with 0.1% acetic acid (Sigma-Aldrich®) and acetonitrile (Sigma-Aldrich®).

For the determination of amino acid concentrations, the samples were diluted in a ratio of 1:100 or 1:1000 in ultrapure water. The samples were derivatized with butanol prior to determination of the concentrations. The separation was done by HPLC (Agilent 1100) using a column with 100 mm × 2.1 mm (Kinetex® 2.6 µm XB-C18 100 Å, Phenomenex, Aschaffenburg, Germany) and a gradient elution with ultrapure water with 0.1% (v/v) trifluoroacetic acid (TFA) and acetonitrile with 0.1% (v/v) TFA.

After 6 d of culture, 3D cultures were fixed with 2.5% glutaraldehyde (Sigma-Aldrich®) for 10 min, followed by dehydration and freeze-drying as described above in section 2.2. Then, the samples were coated with a 5 nm 80% gold + 20% platinum layer (MED020Coating System BalTec, Balzers, Liechtenstein) and imaged with an scanning electron microscopy (SEM) (XL 30 FEG ESEM, Philips, Amsterdam, Netherlands) using an acceleration of 12 kV.

Determination of amino acids and adenosine was done once with a pool of three independent experiments (n = 1). All other experiments were performed in triple determination (n = 3). For the evaluation of these experiments, the mean value of the obtained data was calculated as well as the standard error. Statistical significances of the obtained data were estimated using a paired Student's t-test (data of sections 3.1, 3.2 and 3.5) with Excel (Microsoft, Dublin, Ireland) and marked with an asterisk * for a p-value ⩽ 0.05 or with 'ns' for a p-value > 0.05 (not significant). The data presented in sections 3.3 and 3.4 were separately analyzed for each chemotherapeutic with a two-factorial ANOVA using the Real Statistics Resource Pack software for Excel (Release 7.6, Copyright (2013–2021), Charles Zaiontz, www.real-statistics.com). Significant differences between the treated samples and their corresponding controls were calculated by a Tukey post hoc test following the ANOVA and marked with an asterisk * for a p-value ⩽ 0.05 or with 'ns' for a p-value > 0.05 (not significant). Additionally, we analyzed separately for each chemotherapeutic in a two-factorial ANOVA whether the effects of the drugs differed significantly between 2D and 3D. A significant interaction term then means that the difference between the chemotherapeutic and its corresponding control was significantly altered between 2D and 3D culture and was marked with a hash #.

To compare cell viability, proliferation and metabolite analyses (see table 1), the respective results were characterized with pluses to indicate an increase in comparison to the control and minuses to mark a decrease. A triple plus or minus (+++ or −−−) indicates a difference that is statistically significant and large, where large means >50% of the control value. When the change was >50% but not statistically different from the respective control or <50% and statistically significant, the result was characterized as double plus or minus (++ or −−). Non-significant differences with changes >30% of the control value are indicated with a single plus or minus. No change is indicated by a ' = '.

A parallelized, perfused culture system using a 3D in vitro bone marrow analog was established for simultaneous drug testing experiments. For this purpose, a contactless controlled movement of cell-laden magnetic 3D scaffolds in their growth media in multiwell plates was applied to ensure perfusion. A magnetic lift was constructed to enable the contactless movement of magnetic hydrogels by applied external magnetic fields. This magnetic lift carried a deep 12-well plate held by a movable stage (figure 1(A)). The top and the bottom of the movable stage were equipped with 12 disc magnets each, creating a magnetic field. This induced a continuous up and down movement of the magnetic hydrogels within the wells of the plate, when the stage with the plate was moved towards the upper or lower magnet panel (supplementary figure 3). The production of the magnetic hydrogel scaffolds, cell seeding and experimental setup as well as photographs of the magnetic scaffolds after production and lyophilization are shown in figures 1(B) and (C).

To characterize the effect of scaffold movement on the exchange of soluble factors of magnetic hydrogels with the surrounding liquid, the hydrogels were soaked with saturated NaCl solution and placed in the wells of a deep well plate filled with 5 or 6 ml water. The increasing conductivity resulting from the diffusion of NaCl from the hydrogels into the surrounding water was measured (supplementary figure 4). Movement of the hydrogels was induced via the magnetic lift in the dynamic setting, which was compared to the static setting using non-moved magnetic hydrogels in culture plates. In the dynamic setting, an accelerated exchange of Na⁺ and Cl⁻ was observed.

In order to develop a 3D model of the leukemic niche for drug testing, LC lines that are responsive to certain chemotherapeutics and resistant against others were required as cellular models. Therefore, the responsiveness of the acute myeloid leukemia (AML) cell lines KG-1a, OCI-AML3 and MOLM-13 to the chemotherapeutics imatinib (IMA), 5-fluorouracil (5-FU) and cyclophosphamide (CPA) in three different concentrations was tested in standard 2D cultures in initial experiments. These particular drugs were chosen because all three of them have different mechanisms of action. IMA is a known tyrosine kinase inhibitor [33, 35, 36], whereas 5-FU is recognized by DNA polymerases and gets incorporated into DNA [18, 37]. CPA alkylates DNA during cell proliferation and forms intra- and interstrand DNA cross-links in addition to DNA-protein cross-links resulting in the inhibition of DNA replication and apoptosis [38]. After one, four and six days, the cells were analyzed with live/dead staining by flow cytometry. CPA treatment of OCI-AML3 and MOLM-13 cells resulted in a reduction of viable cells, which was stronger with higher concentrations and increasing days of culture. The same effect was observed during 5-FU treatment of MOLM-13 cells (supplementary figure 5). KG-1a cells treated with IMA and 5-FU showed viabilities consistently comparable to the survival rates in the controls (figure 1(D)). CPA treatment led to a decrease of viable cells normalized to the control and this effect substantiated with increasing CPA concentrations and time.

All in all, the AML cell line KG-1a showed the highest resistance against chemotherapeutics, followed by the cell line OCI-AML3. Those cell lines were sensitive against all concentrations of CPA treatment whereas only MOLM-13 cells additionally were strongly affected by treatment with 5-FU. For the development of a 3D model of the leukemic niche, the cell line KG-1a was chosen as it fulfilled the criteria to be responsive to some chemotherapeutics (CPA) and resistant against others (IMA, 5FU) in 2D culture.

In order to analyze the effects of chemotherapeutic treatment on healthy HSPCs and LCs in a biomimetic environment, dynamic 3D double co-cultures containing HSPCs or LCs together with MSCs were carried out in the 3D perfused culture system. For this purpose, HSPCs or KG-1a cells were seeded with MSCs in magnetic hydrogels creating artificial healthy and leukemic niches respectively. The chemotherapeutics IMA (0.9 μM), 5-FU (5 μM) and CPA (10 mM) were applied to investigate the influence of the culture mode (static 2D/dynamic 3D) on the effectivity and toxicity of these drugs. Due to limited cell numbers from one donor and high donor-to-donor variability, we focused on the effects of the 3D dynamic system versus 2D static cultures.

The survival and proliferation of healthy HSPCs (figure 2(A)) and leukemic KG-1a cells (figure 2(B)) were analyzed after chemotherapeutic treatment. For this purpose, the cells were isolated from the magnetic hydrogels on day six and analyzed via flow cytometry for live/dead staining to assess survival (figures 2(A) and (B), left) and for CTV intensity (figures 2(A) and (B), right) to mirror the proliferation behavior of the cells. Under IMA treatment, no cytotoxic effects on HSPCs were observed, both in static 2D and in dynamic 3D conditions, in comparison to the untreated controls. Additionally, the proliferation behavior remained unaffected. 5-FU treatment yielded no significant differences in the percentage of viable cells but significantly affected the proliferation of HSPCs in both culture conditions. CPA treatment led to a decrease of HSPC survival in both static 2D and dynamic 3D culture compared to the control. Additionally, this effect was significantly stronger in 3D compared to 2D. Under the same conditions, the proliferation of HSPCs was lower in the static 2D and dynamic 3D cultures than in the corresponding controls as shown by the higher mean fluorescence intensities of the detected CTV signals.

Zoom In Zoom Out Reset image size

Treatment of leukemic KG-1a cells with CPA revealed similar results as for HSPCs (figure 2(B)); a significantly lower percentage of viable cells compared to the control was detected under static 2D and dynamic 3D conditions. In comparison to 2D, a higher percentage of viable cells could be detected in 3D, indicating a significantly higher chemotherapeutic resistance of KG-1a cells in 3D. Similarly, 5-FU treatment led to a significantly reduced cell viability and proliferation in static 2D culture. Irrespective of whether KG-1a cells were cultured under static 2D or dynamic 3D conditions, IMA treatment was not toxic to the cells. In the 3D system, none of the chemotherapeutics had significant effects on the proliferation of KG-1a cells.

To investigate the impact of chemotherapeutic treatment on the cell proliferation, the distribution of viable cells progressing in proliferation was analyzed. For this purpose, the fraction of viable cells was plotted against proliferation progress gates (description in supplementary figure 2), which was determined by CTV distribution (figure 2(C)). Higher gate numbers indicate higher proliferation rates of cells detected within these gates. No changes were observed for HSPCs in either culture condition when they were treated with IMA, while 5-FU and CPA treatment resulted in less-proliferated cells in static 2D as well as in dynamic 3D culture compared to the controls.

In contrast to the HSPCs, two separate peaks representing distinct populations were visible in the CTV analyses of KG-1a cells under certain conditions. While the controls for the dynamic 3D cultures always yielded two peaks, the controls for the static 2D cultures showed a single less-proliferating population. In the static 2D culture under IMA treatment, the largest fraction of viable cells was found in a lower proliferation progress gate compared to the control. However, no changes were observed in the dynamic 3D condition. Treatment with 5-FU resulted in a single peak representing less-proliferated cells for both culture conditions. Interestingly, CPA treatment led to two distinct populations in the 2D culture, one with very less and the other with highly proliferated cells. Similar effects could be seen in the 3D condition where some cells had proliferated less while more cells were found in the higher proliferated population compared to the control.

SEM was performed to visualize the different cell types in the 3D environment and to discover potential effects of the chemotherapeutic treatments on cell morphology (supplementary figure 6). After six days of dynamic 3D double co-culture of HSPCs together with MSCs, IMA and 5-FU treatment changed the appearance of normally round cells with smooth surfaces (as visible in the controls) to a more longish form with rough surfaces and borders. After IMA treatment, the 3D environment created by the magnetic hydrogel and MSCs resulted in a less defined structure where cells and their environment were less distinguishable. In line with the findings of the live/dead staining after CPA treatment, more cell debris was visible in the surrounding of the MSCs and the hydrogel compared to the control, in which many healthy round cells could be observed.

In summary, CPA treatment showed the highest impact on cell behavior in 2D and 3D with lower cell survival and altered proliferation for both, HSPCs and KG-1a cells. Furthermore, chemotherapeutic treatment with 5-FU and CPA led to attenuated HSPC proliferation while for KG-1a cells effects were stronger in 2D than in 3D.

To investigate the impact of chemotherapeutic treatment on the differentiation of HSPCs and KG-1a cells, the expression of the surface markers CD34 and CD45RA was analyzed. After six days of double co-culture, HSPCs (figure 3(A)) and KG-1a cells (figure 3(B)) were isolated from the magnetic hydrogels and stained. Only 5-FU treatment of HSPCs in a dynamic 3D culture resulted in a lower percentage of differentiated CD34⁻/CD45RA⁻ cells compared to the untreated control, although this change was not statistically significant. This was confirmed by the measurement of the percentage of less mature CD34⁺/CD45RA⁻ cells. A higher percentage of these cells in comparison to the control was found after treatment with 5-FU in dynamic 3D culture.

Zoom In Zoom Out Reset image size

In contrast to 2D, 5-FU and CPA treatment significantly reduced the number of CD34⁺/CD45RA⁺ HSPCs in 3D. Treatment of KG-1a cells with IMA and 5-FU had no influence on their differentiation state in comparison to the controls. However, for CPA-treated KG-1a cells, a loss of CD34 and CD45RA expression could be observed. Significantly less KG-1a cells were CD34⁺/CD45RA⁺ after culture with CPA in both culture conditions compared to the controls. These findings were confirmed by the CD34⁻/CD45RA⁻ measurement. CPA treatment resulted in a higher percentage of CD34⁻/CD45RA⁻ KG-1a cells in static 2D and dynamic 3D cultures than in the controls. Taken together, the differentiation of HSPCs was impaired by 5-FU treatment after culture in 3D dynamic conditions resulting in a higher percentage of less mature cells. CPA treatment promoted a change of differentiation marker expression in KG-1a cells cultured under static 2D and dynamic 3D conditions.

During leukemia, LCs coexist with HSPCs and stromal cells in the bone marrow. To mimic this condition, a triple co-culture was established. HSPCs isolated from healthy donors, the AML cell line KG-1a and MSCs were co-cultured in the magnetic hydrogels. SEM revealed that HSPCs occurred as dense but clearly separated round cells within the magnetic hydrogels in both monoculture and double culture with MSCs. In contrast, KG-1a cells formed clusters consisting of a few round cells closely attached to each other (supplementary figure 7 upper panel). MSCs tightly attached to the hydrogel structure were spreading largely, thereby contributing to the matrix within the created artificial niche. These spreading MSCs were also found in the triple co-culture. HSPCs and KG-1a cells could not be distinguished although single roundish cells as well as clusters could be observed (supplementary figure 7 lower panel).

To challenge the triculture leukemic model with chemotherapeutics, 5-FU was chosen as a model drug, because of the observed toxicity on healthy cells and resistance of LCs in dynamic 3D culture. Therefore, 5-FU was well-suited to assess potential toxicity and effectivity by evaluating the sensitivity or resistance of healthy and diseased cells to this drug. Triple co-culture under dynamic 3D conditions was treated with 5-FU (5 µM) and compared to a static 3D culture with unmoved magnetic hydrogels and a traditional static 2D culture. In order to differentiate between HSPCs and KG-1a cells, the cells were stained before seeding with different proliferation tracing dyes (CTV for HSPCs and CFSE for KG-1a cells). In comparison to the controls, the percentage of viable cells after treatment with 5-FU was only reduced for KG-1a cells in the 2D culture. For all other culture conditions no significant differences could be detected (supplementary figure 8).

Regardless of the survival of HSPCs, 5-FU treatment significantly affected their proliferation (figure 4(A)). Higher CTV signals, indicating slower proliferation of HSPCs were recorded for all treated culture conditions in comparison to the controls. A similar trend was observed for KG-1a cells. Looking at the distribution of viable HSPC and KG-1a cells versus progressing proliferation (figure 4(B)), it can be observed that 5-FU treatment led to an increase in less-proliferated cells under all culture conditions compared to the respective controls. When assessing the differentiation of healthy and leukemic hematopoietic cells in triculture via the expression analyses of CD34 and CD45RA, no differences in percentages of early (CD34⁺/CD45RA⁻) or more differentiated (CD34⁺/CD45RA⁺ and CD34⁻/CD45RA⁻) hematopoietic cells could be detected in response to 5-FU treatment (figure 4(C)).

Zoom In Zoom Out Reset image size

All in all, 5-FU treatment of the developed leukemic bone marrow triculture model revealed inhibiting effects on the proliferation of HSPCs in 2D and 3D, which was in tendency also observed for leukemic KG-1a cells. Survival and differentiation of both cell types were largely unaffected.

To analyze the cultures over time, the previous end-point analyses are not suited as they require destruction of the scaffold and final cessation of the experiment. Therefore, the potential of analyzing the medium supernatant for monitoring the culture was assessed.

Measuring amino acid concentrations in the supernatants of double and triple co-cultures did not show any clear effects of chemotherapeutic treatments of static or dynamic culturing conditions (supplementary figure 9, supplementary tables 1–7). Of note, in these analyses we averaged over three independent experiments by pooling the supernatants of these independent experiments. While in this way an averaging is possible, the dispersion of the individual samples cannot be assessed.

Analysis of glucose, lactate and adenosine allowed monitoring the culture state without scaffold destruction. Metabolically active cells consume glucose and produce lactate and adenosine. Hence, supernatant was taken from the chemotherapeutically treated 3D double and triple co-cultures on three different days and lactate, glucose, and adenosine concentrations were determined.

For the dynamic 3D double co-culture of HSPCs with MSCs, higher glucose concentrations indicating lower metabolic activity of the cells were observed compared to the corresponding controls without treatment. Statistically significant differences were obtained for all chemotherapeutics compared to the control. In line with the above findings, lower lactate concentrations were observed after IMA, 5-FU and CPA treatment (figure 5(A)). Additionally, lower adenosine concentrations could be seen when treated with 5-FU and CPA (supplementary figure 10(A)).

Zoom In Zoom Out Reset image size

In the dynamic 3D double co-cultures of KG-1a cells with MSCs, a decreased metabolic activity was reflected by higher glucose and lower lactate and adenosine concentrations for the CPA treatment. However, in contrast to the HSPCs, the KG-1a cells showed a slightly higher metabolic activity when treated with IMA, as was evident from lower glucose as well as higher lactate concentrations (figure 5(A), supplementary figure 10(A)).

The triple co-cultures in both the static and dynamic settings did not show significantly altered glucose and lactate concentrations although a tendency of increased glucose and decreased lactate levels was visible, which could indicate a lower metabolic activity (figure 5(B)). In case of adenosine concentration, no difference was observed between the treated and the untreated samples in the 3D static system. However, in the dynamic setting a strong increase of adenosine level was measurable in the control, but this effect could not be seen when the cells were treated with 5-FU (supplementary figure 10(B)). In triple co-cultures, the analyses of the supernatant draw a direct relation to the overall metabolite concentrations, thus making it difficult to discriminate the influence of 5-FU treatment on the individual hematopoietic cell types. To identify the discrete effects of 5-FU on HSPCs and KG-1a cells, endpoint analyses as performed previously (figure 4, supplementary figure 8) are necessary. Referencing back to these data, the same trend in viability and proliferation was observed for HSPCs and KG-1a cells.

In summary, particular changes in glucose and lactate concentrations over the entire culture period appeared to reflect the results of cell viability and proliferation obtained in the endpoint analyses of the 3D double and triple co-culture experiments (table 1).