3.2. Reduction of SZ2080 Polymer’s Auto-Fluorescence
Polymers used for TPP, are usually a mixture of monomers, oligomers and photo-initiators (PIs), where the PIs promote crosslinking by radical polymerization [
27]. The PIs, such as Irgacure 369 in the case of SZ2080, express strong auto-fluorescence even in the final developed TPP structures, which can hinder the use of such structures in fluorescence microscopy. Reduction of scaffold auto-fluorescence is therefore important, and it has been reported that chemical treatment of scaffold structures either before cell seeding [
28] or directly mixed into the polymer before fabrication [
27], can pose an efficient way to reduce autofluorescence while maintaining bio-compatibility after the treatment.
Here, we show an alternative, and a non-chemical way for the reduction of the scaffold’s auto-fluorescence to a level where parallel fluorescence microscopy of cells is feasible, and no changes to the bio-compatibly of the structures are observed. The strong auto-fluorescence of the SZ2080 polymer, see
Figure 4a, is reduced down to only 22% of its starting intensity using non-invasive UV exposure, see
Figure 4b. For the UV-light treatment, the 365 nm emission of a Mercury arc lamp, was selected spectrally by filters and focused through a 10× objective for wide-field illumination in a microscope setup. The sample was exposed to the UV-light treatment from the top continuously for several hours. The integrated fluorescence of the scaffold, in subsequent fluorescence image scans, shows that the average intensity is decreasing with the duration of the UV-treatment, and the trend follows an exponential decrease in the tested time interval, as observable in the log-scale representation of the data shown in
Figure 4b.
While the fluorescence intensity of the scaffolds allows for easy non-invasive optical visualization of the 3D architecture of the fabricated structures, it may pose a challenge to perform simultaneous cellular bioimaging. If the scaffold fluorescence intensity is too strong and saturates the detector, especially at the region of interest, namely the cell –scaffold interface, vital information is lost. Via the presented UV-light treatment it is possible to tune the auto-fluorescence to the desired level for cell studies, where the lowered auto-fluorescence of the scaffold still can be used as a position marker to identify cell position with respect to the scaffold.
3.3. Analysis of Scaffold Influence on Cell Packing
Here we present a 3D hyperspectral image analysis to study how the designed scaffold structure influences the packing of the A549 cells. We can use the lowered remaining auto-fluorescence contribution of the scaffold as position markers. We are able to identify and remove contributions stemming from the scaffold from the acquired image stack taken on scaffolds incubated with the double-stained A549 cells with Hoechst 33342, labeling the nucleus and CF488 labeling the cytoplasm. The 3D confocal fluorescence image stack recorded under 405 nm excitation is shown in
Figure 5a after applying an intensity thresholding that reveals that even after 150 min of UV-light treatment, the scaffold’s auto-fluorescence is considerable stronger than the fluorescence contribution from the stained A549 cells. To be able to observe the cells the scaffold areas had to be leveled to saturation. To separate the fluorescence contributions from scaffolds, with labeled nuclei and cytoplasm, respectively, hyperspectral image stacks were recorded at 405 nm excitation and at 488 nm. Next, three region of interests (ROIs) were selected that represent (i) the auto-fluorescent scaffold, (ii) the labeled cells (in case of 405 nm excitation revealing the cell nuclei) and in case of 488 nm excitation revealing the cytoplasm) and (iii) the background, respectively for both stacks. From the selected ROIs the spectral contributions in the 32 detection channels (nuclei, excitation 405 nm), see
Figure 5b and the 23 channels (cytoplasm, excitation 488 nm), see
Figure 5c, are determined. The auto-fluorescence of the scaffold in both figures, are still strong and show a broad spectral contribution throughout all 32 and 23 detection channels, respectively.
Via the selected ROIs and the spectral linear-unmixing applied to the two image stacks we are able to identify and separate the contribution from the scaffold (as extracted from the 405 nm excitation stack), see
Figure 5d, the cell nuclei, see
Figure 5e and the cytoplasm region see
Figure 5f, in the confocal Z-stack. A representative single confocal image slice of each of the three contributions is shown in
Figure 5g–i, where it is clearly observable that the long scaffold walls have been removed from the images. The representative slides show a scaffold area of 520 × 630 µm which is used in the analysis and corresponds to the field of view of the 10× microscope objective.
Further zooming into the extracted separated 3D image stacks allows a more detailed evaluation of the efficiency of spectral separation and impressive accuracy of the applied spectral linear unmixing can be appreciated, where successful separation into scaffold structure revealing the different niche sizes, the cell nuclei and cytoplasm regions are clearly visible (see
Figure 5j–l).
With the extracted cellular contents information from the spectral unmixing, it is now possible to analyze the influence of scaffold morphology, i.e., wall separation and niche size on cell distribution across the scaffold. In this first analysis of a static condition, where the emphasis is on the separation of fluorescence signals from the different cell parts and the auto-fluorescence of the scaffold, no data is extracted regarding the proliferation of the cells inside the scaffold area.
First, we analyze the number of nuclei and cytoplasm material per unit area found in the scaffold, as defined by the two scaffold parameters. Secondly, we analyze how the nuclei and cytoplasm material are distributed inside the chambers, i.e., if the cells are either in contact with the wall or located in the free space between the walls again as a function of the same two scaffold parameters.
In
Figure 6a,b, the total amount, defined by pixels intensity of nuclei and cytoplasm fluorescent material, respectively, is summed in each chamber over the full height of the scaffold, and normalized to the area of the smallest chamber size of the scaffold, i.e., 10 × 20 = 200 µm
2. For both cellular components, a similar monotonously decreasing trend of material presence as a function of wall separation is observed for all niche sizes. For both cases, the nuclei and the cytoplasm, the curves converges towards a common value for the different types of cellular material, which is independent of the niche size.
In
Figure 6a, when the wall separation is 20 µm, similar to the average cell size the niche size has a large influence on the cell packing, dropping by more than 35% when the niche size goes from a sub-cellular size of 10 µm to a larger than the cell size of 30 µm, and the drop is even larger, for the larger niche sizes. If the wall separation is increased by just 5 µm to 25 µm, a little above the average cell size, the general cell packing drops. This suggests that the influence of the niche size decreases, as seen by the smaller spread of values for a 25 µm, as compared to the 20 µm wall separation. This tendency continues for larger wall separations, until the convergence point, where it seems the niche sizes have no more influence on the cell packing. The same observations are made for the cytoplasm material part in
Figure 6b, where a slightly smaller drop in material per unit area is seen as a function of increasing scaffold dimension parameters.
This clear tendency of a less dense cell packing as the scaffold increases in both cases indicates that a larger influence on the cell arrangement by the scaffold is achieved when the cells are sensing the scaffold from all sides, i.e., when both niche size and wall separation are comparable to the average cell size. If one or both parameters gets larger than the average cell size the scaffold influence starts to diminish.
This is further explored in the following. We calculate for a given chamber size the ratio of the total cellular material compared to the amount of the cellular material found inside the niche, as shown in
Figure 6c for the nuclei, and in
Figure 6d for the cytoplasm contribution. In the analysis, the material found inside the niches, which is considered to be within a range up to 7.5 µm away from the walls, and further corresponds to roughly half the average cell size, is to be considered in contact with the walls. For each niche size there are 4 wall separations, which determine the chamber size, and therefore the ratio between the total area and the area covered by the niches. These ratios are 0.75, 0.6, 0.43 and 0.27 for wall separations of 20, 25, 35 and 55 µm respectively, and have been added to
Figure 6c,d. In both figures, data points for an isotropic distribution have been added to guide the eye. For an isotropic distribution of cellular material in the chambers, and therefore a minimal influence of the scaffold’s presence on the cellular arrangement, the extract data points should match or be close to these curves.
For the nuclei contribution shown in
Figure 6c, we observe that for a 10 µm niche size the ratio of nuclei material in contact with the nearly follows the predicted values for an isotropic distribution. This would indicate that the scaffold does not influence the nuclei packing significantly, which could be contributed to reaching a saturation of the number of cells able to be packed into the small volume of the 10 µm niche, even for larger wall separations. The three other niche sizes follow a general trend, which is different from the behavior observed for the 10 µm niche size. For the 20 µm and 25 µm wall separations the ratio of cells in the chambers compared to those within the niches, in proximity to the walls follow the predicted isotropic distribution values for the three niche sizes, while for larger wall distances the ratio is larger than the predicted values. This would indicate that the scaffold promotes a higher degree of nuclei packing in the niches even for large scaffold parameters giving rise to a non-isotropic cell distribution in the chambers. This could be promoted by the larger wall separation allowing for increased cell motility, and will be analyzed and discussed in the next section.
For a similar analysis on the cytoplasm part of the cells, seen in
Figure 6d, the niche size of 10 µm, again follows the isotropic distribution behavior for all wall separations. For the three other niche sizes, a relatively constant ratio of cytoplasm localization in chambers versus within niches is observed as a function of the wall separation, which indicates that a high percentage of the cytoplasm part of the cells is in contact with the scaffold, even higher than for the nuclei part of the same cells.
3.4. Observation of 3D Cellular Vertical Growth over Time as a Function of Wall Separation
In the time-lapse studies, the scaffold is seeded with a lower total cell count than in the previous experiments, both to simplify tracking of the individual cells, as well as to quantify the effect of the scaffold from an initial situation, while limiting any scaffold saturation effects. We apply the same analysis tool, spectral linear unmixing, with a focus on the fluorescent contribution from the H33342 stained cell nuclei, to define the position of individual cells. Furthermore, the scaffold is separated into two parts in this analysis, a bottom part consisting of the first 14 µm in the z-direction, the average size of the A549 cell, and a top part with the remaining 10 µm of the scaffold height, as seen depictured in
Figure 7a. From each of the 6 time points, a data set of the individual cell positions, identified by the cell nuclei, is recorded in both the top and bottom part of the scaffold. In the first part of the analysis, we only look at the influence of the wall separation. The influence of the niches is not taken into account, since the cell count for each niche size in the initial seeding stage, is too low for such statistical analysis.
Analysis of data for the first 14 µm of the scaffold close to the sample surface show minimal change in the cell count per unit area over time, as a function of the wall separation, even when compared to the flat 2D control surface, see
Figure A2a in
Appendix A. In the analysis, the unit area is chosen to be the area between two walls with a 20 µm separation. For the number of cells located inside the niches as compared to the total number of cells, no significant variations are observed again as a function of time, as depicted in
Figure A2b in
Appendix A.
The same analysis is performed on the 10 µm spanning top part of the scaffold, and similar non-changing behavior is found, with a few exceptions and modifications, see
Figure A3a,b) in
Appendix A. First, the number of cells in the control area, is by default low, since only non-attached floating cells in the growth media are counted in this region. Secondly, the overall number of cells at the top of the scaffold is lower, as compared to the bottom part, and cells are nearly all located inside the niches due to the lack of the 2D glass surface between the walls. Based on the extracted data from both the bottom and top of the scaffold, where the cell number does not change as a function of time, we do not consider proliferation as an influence in further analysis.
Analyzing the ratio of cells inside the niches as compared to the total number of cells, for both the top and bottom parts of the scaffold in the time-lapse study, is shown in
Figure 7b. For the bottom 14 µm of the scaffold and wall separations of 20, 25 and 35 µm, the ratio of cells attached to the niches is between 0.74 and 0.97, as seen in the column in
Table 1. While for a wall separation of 55 µm, only about half of the counted cells are attached to the scaffold, indicating that the wall separation has become large enough so that the A549 cells do not feel the immediate influence of the scaffold. For the 10 µm top part of the scaffold, wall separations of 20 and 25 µm, still have cell attachment ratios between 0.73 and 1, while for 35 µm separation the ratio has fallen to between 0.59 and 0.73, as seen in the second column in
Table 1. For a separation of 55 µm, the ratio is again about half of the cells. This indicates that without the planar 2D surface between the walls, the cells stay floating in the medium, without attaching, if the cells do not come into direct contact with the wall. Comparing the values for the scaffold influence on the cell nuclei distribution in the low population scheme,
Figure 7b, to the extracted nuclei part in
Figure 6c for the full scaffold height in the strongly populated scaffold from
Figure 5, as seen in the third column in
Table 1, a general trend is observed.
For all wall separation distances the values for the low cell density case, have a higher ratio, indicating a stronger scaffold influence on the cells. As the cell number grows and the niches starts to fill, the influence of the scaffold lowers, reaching a saturation effect where the niche and chamber distribution is the same, as seen for the 10 µm niches in
Figure 6c.
The constant number of cells between two walls for the 6 time points, for both regions of the scaffold, indicates that there is no net in- or out-flux of cells to the scaffold, and a limited vertical cell movement in the analyzed time period. The limited vertical movement could be due to the design of the scaffold, where the low height of the scaffold prevents a clear separation of the vertical movement between the two parts. This could be improved with a taller scaffold structure. Furthermore, the designed tight openings both to the outside of the scaffold and between chambers, as well as the wall structure itself, limit cell motility and only allow movement along the wall direction and not freely throughout the entire structure array. Analysis of this horizontal cell movement, as a function of time in the initial phase for both the top and bottom part of the scaffold, will be discussed further in the next section.
3.5. Analysis of Cell Movement Compared to Scaffold Wall Separation and Niche Size
Based on the results from time-lapse studies described in
Section 3.4, where no changes in the total cell number over time and any vertical movement was observed, a more complex analysis of the individual horizontal cell motility is required to conclude the influence of the scaffold parameters. In our time-lapse study, each time point is spaced more than an hour apart, making tracking of the individual cells difficult. We, therefore, apply a custom written Matlab algorithm that, based on the nuclei position at each time point in the time lapse estimates the mobility of a cell, and from this, quantifies the horizontal movement in the two vertical sections in the scaffold area.
The algorithm starts by locating each cell, by the nuclei position in the first time point, t = 0 of the image series, and will, in each subsequent time point in the series, based on the single input parameter N, try to find the position of that cell in the next time point.
In each time point, a tracked nucleus is marked with a colored circle, orange in the first time point, etc., as seen in the tracking example for 6 time points in
Figure 8a. In the current analysis when the algorithm has gone through the image sequence once, it again goes through the data set without already tracked cells and tries to map additional cells. If more than the two iterations of the algorithm used in the presented data is needed it is easily configured in the Matlab code. Furthermore, in our analysis we require that a tracked cell should be present in five sequential time points, which is based on the result from the previous section, where the cell count does not change over time. If the image sequence consists of more time points, or when a large number of cells are leaving the measured area, this number should be set accordingly in the Matlab code. After all possible cells have been tracked, the algorithm returns, based on the total movement of the tracked cells, a color gradient map of cell motility which is overlaid with the outline showing fold structure. A blue color, arrow 1 in
Figure 8a, indicates stationary or low motility while large cell motility is indicated by a red color in the gradient map, see arrow 2 in
Figure 8a). The lowered auto fluorescence of the scaffold, and the use of spectral linear unmixing makes the scaffold walls, and niches easily identifiable markers in
Figure 8a for correlating cell motility to specific regions of the scaffold.
In the algorithm, the value of the parameter N defines a radius where a single cell nucleus has to be found in the next time point, thereby defining the maximum distance a cell can move between time points. N is given in pixels, and in our experimental setup with a 10× microscope objective, the conversion factor is a pixel = 1.66 µm. The algorithm selects a new cell position in the next time step, by choosing the cell closest to the current position if multiple cells are found inside a circle with a radius of N. If no cell is found inside N the cell will not be tracked. This also happens if the value of N is set to a low value compared to the motility of the cells monitored, whereby the cells have migrated beyond the radius of the circle defined by N. If the value N is set too high there is an increased risk that the algorithm will identify the wrong cell in the next time step, which also happens for a high concentration of cells in the scaffold area. The influence of the value of N for the current data set, can be seen in the extracted data in
Table A1 in
Appendix A and for the top and bottom part of the scaffold in
Figure A4 and
Figure A5, also in
Appendix A. From data summarized in
Table A1, one can observe that for N = 2 the highest total movement extracted is only 25% of the maximum allowed distance, while for N = 5 and N = 7 the movement increases to approx. 40% while for N = 10 and above it settles at approx. 30%. This is visible in
Figure A4 and
Figure A5, which represent the results of a mathematical analysis of cell center movements identified within confocal images collected at 6 different time points over an interval of max. 5 h 25 min. N is the number of pixels that define an allowed movement kernel from time point to time point. A gradient map is used to represent identified cell movement distances in dependence of the movement kernel size N, via a cell tracker Matlab code, see
Supplementary Materials. We superimpose the scaffold structure extracted by thresholding from the original confocal image on top of the gradient maps, as well as the various cell positions identified at the different time points (colored circles), as also done in
Figure 8. ROIs (white dotted lines) were added that highlight areas of high cell movement, identified for the N = 10 gradient map and remaining similar even for higher movement kernel sizes (copied into the other gradient maps for reference).
We, therefore, apply N = 10 for further analyses, which is a value of 16.6 µm comparable to the average cell size, and slightly larger than the average cell motility of 10 µm/h found for A549 cells [
29]. This value will optimize the number of cells analyzed since no two cells can occupy the same space in the scaffold, and give an individual cell motility range between 0 and 100 µm, in the time-lapse study.
Figure 8b,c shows the gradient map for the bottom and top of the scaffold, respectively, for a value of N = 10 created by the algorithm, where both figures are scaled to a maximum total distance of 30 µm, the highest value found in the analysis.
The gradient map for the bottom part of the scaffold in
Figure 8b, shows several regions of high motility i.e., orange to red regions [20–30 µm], which are characterized by large free spaces with flat 2D surfaces for unrestricted movement. The free spaces are either in the form of the planar 2D control region at the top of the figure or where the wall separation is 55 µm in combination with a niche size larger than the average cell size, i.e., 30, 50 and 70 µm. Other high motility regions are found where the niche sizes are large 50 and 70 µm, but also in regions where a large number of cells are located in a small region. At the other end of the motility spectrum, where the cells become very stationary seen as blue areas in
Figure 8b, is where at least one scaffold parameter is equal to or below the average cell diameter.
A similar analysis is made for the top part of the scaffold, as seen in
Figure 8c, where less horizontal cell motility is observed, indicating that the cells are more stationary when attached to the scaffold and there is no planar 2D surface to support them. Only when the wall separation becomes larger than the average cell size a few cells are found in the space between the walls.
The current analysis is limited to mapping cell movement in the XY-plane and does not extend to movement in the vertical Z-direction. For the scaffold in the present study with a small height, this limitation is not crucial while for extended scaffolds in Z-dimension, this movement would have to be taken into account.