DEP dots demonstrate robust, repeatable aggregation

Dot electrodes (Fig. 1) use negative (repulsive) DEP to contain the cells, by repelling them from the edges of circular patterns in the electrode and forcing them into aggregates15. Negative DEP occurs at specific frequency ranges, determined by the electrical properties of cells and medium. Optimal frequencies were established for each cell type used in the formation of aggregates using a DEPtech 3DEP reader (Labtech, Heathfield, UK) as described previously15,22,29. This approach allowed for cell specific aggregation protocols and control of construct size and geometry.

Figure 1

Schematic of the DEP dot array. The removable gasket connects the ITO ground electrode to the gold array and also serves as the reservoir in which the CPM and cell mixture is contained. The height of the gasket is approximately 280 µm.

In order to assess the repeatability (n = 100) of aggregate formation in the polyethylene glycol diacrylate (PEG-DA) system, aggregate size and cell number were measured for different cell types with different properties and diameters, and dot radii between 100 and 150 µm. Examples of these aggregates are shown in Fig. 2. The number of cells per aggregate were measured, showing no significant difference (p > 0.5) to their respective means, for all cell types. When aggregate size was compared to the electrode size used for aggregation, all cell lines demonstrated similar (p > 0.1) aggregate size to dot size ratios (Table 1) with aggregates covering just under half of the dot diameter. Since the cell lines used were of similar radii (~ 10 µm), and resulted in similar aggregate size, this demonstrates that the platform can reliably produce aggregates of uniform size. Interestingly, though the aggregates were of similar relative size, as were the cells, the number of cells per aggregate varied, suggesting that changes in cell morphology allowed for significantly different packing densities.

Figure 2

DEP DOT Electrode repeatability across cell types. The following cell types were encapsulated in PEG-DA (A) yeast, (B) HL-1 (C) K562, and (D) HeLa. The resultant peeled gel is shown (E) with a thickness of approximately 280 µm and width of approximately 5 mm. The peeled gels, like that shown, were placed in well plates for cell culture and subsequent drug treatment.

Table 1 Aggregate characterisation (n = 100) for the various cell types.

We also established the UV exposure during cross-linking did not negatively affect cell viability by monitoring viability with Trypan Blue (as described in Abdallat et al.15). There was no significant difference (p > 0.1) in viability values, with viabilities recorded in ranges between 92–100% for HeLa cells and 94–97% for yeast cells when comparing 2D cell culture methods and cells in 3D aggregates.

DEP-Dots can form aggregates in multiple gelling agents

Since UV light and the photo-initiator 2,2-dimethoxy-2-phenylacetophenone (DMPA) can be hazardous to certain cells, it is beneficial for a 3D cell construction technique to be sufficiently flexible to allow the use of alternative gelling agents which are activated by different means; similarly, a user may wish to form a gel that (unlike PEG-DA) can be subsequently broken down to allow recovery of the aggregated cells. To this end we investigated multiple approaches to gel formation using the DEP-Dots.

First, in order to avoid the use of UV exposure we investigated a more biocompatible system using PEG-DA and the blue light initiators triethanolamine (TEA) and eosin-Y27 (Fig. 3A). Cells were observed to form patterns similar to those formed using the UV initiator, as described in the previous section. However, the blue light initiator required a longer curing time (over 10 min under blue light). Aggregate relative size and cell number was consistent with our previous PEG-DA experiments.

Figure 3

The DEP DOT Electrode system is consistent in cell aggregation using various gel systems. (A) K562 cells in PEG-DA cross linked under blue light with initiators TEA and eosin-Y. (B) Yeast cells in 6.75% w/v collagen, and (C) Yeast cells in 25% PuraMatrix.

Second, we investigated gels with different functional characteristics. A distinct disadvantage of PEG-based systems is that once cells are encapsulated, they remain in the gel permanently. This does not allow for post treatment analysis of the individual cells, or layers of these cells. One particularly useful tool would be to characterize changes in the cells pre- and post-encapsulation, for example to allow analysis of cell behaviour using flow cytometry. We have previously shown these electrophysiological properties can predict drug IC5022; if cells could be obtained post-encapsulation a similar comparison could be performed.

To this end, other hydrogel options were then investigated for aggregate dissociation. Collagen hydrogels (Fig. 3B) are widely used for scaffolding and tissue engineering. However, this is a challenging gel system for use in DEP, due to the composition; the solution is highly conductive and the larger amounts necessary for a durable gel (for complete removal from the chip) increase the viscosity, rendering movement by DEP force difficult. We were able to minimize these effects within the dot array at sufficiently low gel content, and determined 6.75 w/v% with a solution conductivity of 875 mS/m achieved adequate negative DEP response, and produced a sufficiently strong gel to be peeled from the chip intact. This gel system was temperature cured, so samples were kept warmed to 40 °C, aggregated with DEP, and then placed in an incubator set to 20 °C for 10 min. These remained intact, and the cells remained viable for at least 1 week after gel formation.

PuraMatrix (Fig. 3C) was also investigated, with concentrations of 25% yielding adequate durability of the gel. PuraMatrix self-assembles when exposed to physiological levels of salt (such as culture media) into nanofibers on a scale similar to the extracellular matrix. Like collagen, PuraMatrix can be dissociated, and whilst fibre density and pore size can be more tightly controlled (a crucial criteria for drug/molecular studies), at 25% concentration necessary for encapsulation. It is the most expensive option considered here.

Drug efficacy of 2D versus 3D varies in suspension and adherent cells

Since the primary function of most 3D cell cultures is the assessment of efficacy of new pharmaceutical interventions, it is important to understand how, and why, this response differs from a 2D culture. To this end, two drugs in wide clinical use, were investigated to benchmark their response in 3D against their 2D model.

Vinblastine is a chemo-therapeutic agent that interferes with a cell’s ability to undergo mitosis by binding to tubulin and inhibiting microtubule production causing mitotic arrest and, ultimately, cell death30. Amphotericin B (AmB) is an anti-fungal agent that acts by stopping the production of ergosterol which causes channels to open in the fungal cell membrane, compromising the cell and ultimately leading to cell death31. Both Vinblastine and Amphotericin B are small molecule drugs of similar molecular weight (909.05 Da and 924.079 Da, respectively) and as such can be assumed to diffuse similarly through hydrogels.

First, the effect of vinblastine on the survival of patterned and encapsulated HeLa aggregates was investigated (Fig. 4A,B). Statistical analysis indicated that the control sample for monolayer and for aggregates had no significant difference in their cell viability. The viability of monolayer cells exposed to a drug concentration of 11 μM dropped by an average of 62% after 3 h of drug incubation, compared to an average of 13% for aggregates.

Figure 4

(A) HeLa cells aggregated prior to drug treatment and (B) viability of HeLa cells in aggregates assessed with live/dead assay. (C) Average HeLa response curves for monolayers and aggregates 48 h post-treatment with vinblastine. (D) Average yeast response curves for monolayers and aggregates 48 h post-treatment with Amphotericin B (E) Drug concentration of drug (Vinblastine/AmB) in the middle of gel layer as a fraction of the applied concentration cmax, calculated from Eq. (3) with Dg = 3.3 × 10–10 m2/s and L = 300 µm. (F) Geometry of a single DEP array dot with cellular aggregate, L = 300 µm and d = 300 µm and the aggregate is modelled as a half sphere of diameter 0.46d. (G) Drug concentration of drug vinblastine in DEP gels as fraction of applied concentration cmax in a section through the centre of the cellular aggregate. Drug concentrations obtained by solving Eq. (3) in (i) a cellular aggregate region with Dc = 1.9 × 10–12 m2/s, k = 0.01 s−1 and (ii) a gel region with Dg = 3.3 × 10–10 m2/s and k = 0 s−1.

When viability was measured at 48 h post treatment, it was observed to have dropped further, reaching a mean viability of 18% in 2D monolayer and 71% in 3D cell aggregates. Control samples for both 2D and 3D remained above 90% viable. The mean viability for the experiments showed that the viability for 2D monolayer dropped immediately after drug incubation, and the drug had a prolonged effect which decreased viability further 48 h post drug removal. The 3D aggregates did not show the same immediate drug effect, as viability only decreased to 78%; however, they did continue to decrease at 48 h post treatment (Fig. 4C).

This was then compared with the result of a second test; the action of AmB on yeast cells in cell aggregates. Since yeast cells are unicellular organisms which grow in suspension (in their natural shape) in isolation and do not normally interact with adjacent cells, they provide a suitable control to assess whether the process of cell resistance in clusters is due to simple blocking of cells against diffusional processes by cell packing, or a more complex process whereby cell morphology and inter-cell communication allows the cells to better adapt to the presence of the toxin. Control samples for the 2D (cells suspended in YPD broth and 3D aggregates demonstrated viability above 96%. These were also used to establish estimated growth curves of yeast over a 48 h period (due the more rapid doubling time compared to human cell lines). The treated cells started demonstrating significant viability decrease (p < 0.01) for the 2D treatment at 16 µg/ml reaching a viability below 50% at 160 µg/ml. The 3D treated cells followed a similar viability curve remaining higher than the 2D treated cells, though only significantly so (p < 0.001) for the 160 µg/ml treatment. Again, a 48 h post-treatment follow-up was conducted to examine the long-term effects of the drug treatment in 2D and 3D culture (Fig. 4D). Control samples for 2D demonstrated an 82% viability whilst the 3D samples were still 96% viable. With each increase in AmB viability decreased with the 160 µg/ml treated cells exhibiting viability of 7% and 29% for 2D and 3D, respectively.

The effectiveness of standard drug treatment of adherent cells (which naturally interact with surrounding cells in vivo) was shown to differ from that in yeast cells (which only act as independent unicellular organisms) when grown in similar conditions. The results obtained from HeLa 2D HeLa monolayers and 3D aggregates were found to be significantly different (p < 0.05) at all drug concentrations. However, the difference in drug effectiveness in suspension versus 3D culture in yeast only showed a 16% change between the different culture conditions. Taken together, this study confirms that both suspension and adherent cells do respond to drug treatments differently in a 3D environment when compared to their 2D counterpart, as has been previously demonstrated32. However, this also raises questions about the reason for this change in behaviour.

In order to explore this further, we developed a model of molecule diffusion through the aggregates. The drug concentrations within the entire DEP dot array can be calculated from the diffusion equation20. Given that the lateral extent of the array (5 mm) is significantly larger than its thickness (300 µm) it is sufficient to consider the transport of the drug through the thickness alone so that we solve:

$$frac{partial c}{{partial t}} = D_{g} frac{{partial^{2} c}}{{partial z^{2} }},$$


where (D_{g}) is the diffusion coeffcient specific to the drug and gel. The boundary and initial conditions are (c = c_{max}) at (z = 0), (L) and (c = 0) at (t = 0), where (c_{max}) is the concentration of the medium in which the array is submerged and (L) is the gel thickness, which we have taken as 300 µm throughout. The solution of Eq. (2) with these initial and boundary conditions may be obtained by the method of separation of variables as:

$$frac{c}{{c_{max} }} = 1 – mathop sum limits_{k = 0}^{infty } frac{4}{{pi left( {2k + 1} right)}}sin left( {frac{{pi left( {2k + 1} right)z}}{L}} right)e^{{ – frac{{pi^{2} left( {2k + 1} right)^{2} D_{g} }}{{L^{2} }}t}} .$$


A range of diffusivities have been reported in the literature33,34,35, and choosing the value corresponding to the slowest diffusion (Dg = 3.3 × 10–10 m2/s27), we plot in Fig. 4E the concentration in the middle of the DEP array. We find that the drug saturates the prepared samples within minutes. This is significantly shorter than the experimental timescales, so that the gels may be considered to be at a uniform concentration equal to the applied concentration cmax, so that diffusion through the gel is not a limiting factor.

Turning now to a model of an individual aggregates (Fig. 4F) within each dot, we can consider each dot as a cylinder of gel with a dome-like cell aggregate at its base. The diameter of the aggregate has been measured as 0.46d (Table 1); in these simulations we used d = 300 µm (the largest value measured) and (L) = 300 µm.

Within each dot we solved the full three-dimensional diffusion equation:

$$frac{partial c}{{partial t}} = { }Dnabla^{2} c – fleft( c right),$$


where (fleft( c right)) is a term accounting for removal of drug from the system. Within the inert encapsulating gel, we take (fleft( c right) = 0) and the diffusion constant (D = D_{g}). Within the aggregates, we allow for removal of the drug from the system by cellular processes by incorporating a positive (fleft( c right)) and take (D = D_{c} < D_{g}) as a modified diffusion coefficient for the cellular aggregate36. The boundary conditions are as before that (c = c_{max}) at the top and bottom surfaces (z = 0), (L), complemented by a no-flux condition at the outer edge of the cylinder so that drug may not leave the dot through the sides. Continuity of concentration at the interface between gel and aggregate is always assumed. The initial condition is that (c = 0) at (t = 0).

Whilst different drug responses are often observed in 3D aggregates when compared to cells in 2D culture, it is a matter of conjecture as to whether these are due to the packed outer cells reducing the diffusion to the inner cells, or due to changes in the drug response of the individual cells when grown in 3D. Our model suggests that without any modification to drug transport within the cellular aggregates, the inner cells would be rapidly exposed to any applied small molecule drug. Indeed, our model suggests that, in the absence of an additional protection mechanism such as cellular uptake (i.e. if (left( c right) = 0)), uniform high concentration is achieved in time ~ O(r2/Dc). For the small aggregates studied here, this is relatively short; simulations based on the reported diffusivities for e.g. Vinblastine in cellular aggregates37 show that over 90% of the cells experience (c_{max}) within ~ 7 min. Even considering the reduced effective diffusivities that have been reported in three-dimensional tissues33,34,35,38 for a range of substances including vinblastine, oxygen, sodium fluorescein and dextrans, this is insufficient to prevent the chemicals reaching the center of the aggregate within a timescale short in comparison with the study length.

In order to account for the observed reduced effectiveness of Vinblastine in 3D we incorporated the loss term (fleft( c right)) in Eq. (3) when solving the diffusion equation in the aggregate. Many different functional forms for (fleft( c right)) are commonly used including constant39, linear40 and hyperbolic41. However, the data for HeLa response to vinblastine in Fig. 4C shows a relatively weak dependence on (c_{max}), suggesting a good first-order approximation is (fleft( c right) = – kc); in this case the applied concentration may be scaled out of the solution by setting (c = c/c_{max}).

This model provided a solution which demonstrated limited drug penetration to the centre of the aggregate, with the depth of drug penetration determined by the diffusive length scale parameter (α = √(Dc/k)). We solved Eq. (3) (aggregate) to arrive at a steady-state solution using COMSOL (COMSOL Multiphysics v.5.2, Stockholm, Sweden). The results for vinblastine applied to HeLa aggregates (with (D_{g} = 3.3 times 10^{ – 10}) m2/s, (D_{c} = 1.9 times 10^{ – 12}) m2/s, and (k = 0.01) s−1) are shown in Fig. 4G; this shows a region of low drug concentration in the centre of the aggregate which is effectively ‘protected’ by the outer layer of cells removing drug from the system by a combination of factors. Solving the full time-dependent model, we found that this steady-state solution is effectively achieved within five minutes. However, from the model alone it is impossible to distinguish between the protective effects of reducing Dc or increasing cellular absorption k as they only appear in ratio in the effective diffusive length scale.

Since the tightly-packed yeast cells would present similar simple inhibitory barriers to drug diffusion in 3D to those seen in the HeLa model, we propose that this suggests that diffusion in 3D is not the primary reason for the change in HeLa behaviour, and that (as in the situation described elsewhere36) the primary reason for differences in cell behaviour is due indeed to cell–cell interaction and cytoplasmic changes that allow the cell to better mitigate the action of the drug in this case. In Fig. 5 HeLa cells are shown in their 2D monolayer state (Fig. 5A) in which cell attachment and actin activity can be observed, in the 3D aggregate similar cell attachment can be seen when comparing treated (non-viable) cells (Fig. 5B) to healthy cells (Fig. 5C). Compared to constructing aggregates formed spontaneously or by culturing them on treated surfaces, the hydrogel system represents a structure more like the original tissue in terms of having a polymer surrounding cells, which serves as a barrier that can represent blood (growth medium with dissolved drug) and extracellular matrix (hydrogel). Clearly this is significant in the development of new pharmaceuticals, particularly in the use of the IC50 model, where the clinical relevance of cell toxicity in vivo based on cell viability in vitro is clearly to be called into question.

Figure 5

(A) HeLa cells grown in monolayer on a standard culture flask, (B) HeLa cells aggregated and 48 h post treated with 11 µM of Vinblastine and (C) HeLa cells aggregated and cultured with no treatment. From (B) it is visible that the treated cells lack the cell–cell connections shown in (C) of the untreated cells.

Measuring electrophysiological changes post 3D encapsulation

Previous work23 suggested that cells grown in 3D differed in their electrophysiology from those grown in 2D culture. In order to conduct a more rigorous study into the effect of DEP-based 3D cell culture on cells, we investigated the properties of yeast, K562, and HeLa cells after culture. Briefly, trypsin was added to both the 2D and 3D cell cultures for the same amount of time (this varied by a few minutes per sample, but was kept constant between the 2D and 3D replicates). Once the gels were dissociated, cells were resuspended in 10 mS/m DEP buffer, sonicated and analysed in the 3DEP reader (Labtech, Heathfield, UK)22,29,42. Cellular properties of cells grown in 2D for 24 h and 48 h were compared to those grown in 3D for 24 h and 48 h. The results are summarised in Fig. 6.

Figure 6

The three electrohpysiological parameters determined through the single-shell model analysis in 3DEP. Membrane capacitance, Membrane conductance, and cytoplasmic conductivity were determined for yeast (A), K562 (B) and HeLa (C) cells grown in their appropriate 2D models (suspension or monolayer on a flask surface) and cells removed from their 3D model formed by the DEP dot electrode using PuraMatrix gel system. Values on the graphs are given as the mean (n = 5) calculated parameter determined by fitting DEP spectra data of the cells to a single shell model ± SEM p < 0.01.

Analysis of the dielectric properties of these cells revealed significant (p < 0.0001) changes in the membrane capacitance for HeLa cells post 3D encapsulation compared to the 2D model (Fig. 6C) whereas for both suspension cell types, this parameter did not change significantly. Membrane conductance for both yeast and K562 demonstrated a significant change (p < 0.001) due to 3D encapsulation (Fig. 6A,B) although for yeast this effect was only observed at 24 h. The decrease in membrane conductance on the K562 cell line, with no changes in other parameters observed, could be an early indicator of changes to cell functionality2. In common with previous studies of cells following 3D culture24, variation in the membrane capacitance of the adherent HeLa cell types was the most significant change in electrophysiology post 3D encapsulation; there were no similar changes observed in the two suspension cell lines. Since no changes to cell radius were observed (n = 100 cells measured with p > 0.05 between the two groups) the change observed in the membrane capacitance of HeLa cells from 2 to 3D suggests changes to the membrane morphology have occurred.

The mean difference between each of the properties was investigated to determine whether changes in electrophysiology between 2 and 3D were significantly different between cell types (Fig. 7). HeLa cells were found to differ between 2 and 3D in ways which differed significantly from the other two cell types (p < 0.0001 in most cases). The change in both membrane capacitance and membrane conductivity from 2 to 3D between K562 and HeLa cells was consistently significant (p < 0.0001 in most cases, p < 0.01 in Fig. 7A); interestingly, the K562 cells differed in properties in a manner similar to yeast cells (another suspension cell), rather than HeLa (another mammalian cell). Also of note is that whilst differences are still observed between 2 and 3D cells after 48 h, these differences are smaller than those observed after 24 h. It is possible that changes may be due to trypsinisation, though a study43 of DEP response of cells to various detachment methods suggested that this does not have a significant effect on K562 cells.

Figure 7

The mean difference between the 2D and 3D cell properties were determined and compared across cell types. Values given are the mean differences calculated between the means of 2D and 3D (Fig. 6) + SEM.

Could changes in membrane capacitance be responsible for the differences found in the 2D versus 3D drug response of HeLa cells? It is notable that the morphology change only occurs in adherent cells, and may reflect an integration of the cell layer—possibly through gap junctions—that might allow the cells to mitigate the effects of drugs. However, this would need to be investigated further.

This also demonstrates the flexibility of gel culture systems for applications that involve dissociation of the gel and aggregate for further analysis, such as analysing drug diffusion through layers of aggregate to study the necrotic core of a tumour model.


Source link

Leave a Reply

Your email address will not be published. Required fields are marked *