Advantages and limitations of using cell viability assays for 3D bioprinted constructs

Bioprinting shows promise for bioengineered scaffolds and three-dimensional (3D) disease models, but assessing the viability of embedded cells is challenging. Conventional assays are limited by the technical problems that derive from using multi-layered bioink matrices dispersing cells in three dimensions. In this study, we tested bioprinted osteogenic bioinks as a model system. Alginate- or gelatin-based bioinks were loaded with/without ceramic microparticles and osteogenic cells (bone tumor cells, with or without normal bone cells). Despite demonstrating 80%–90% viability through manual counting and live/dead staining, this was time-consuming and operator-dependent. Moreover, for the alginate-bioprinted scaffold, cell spheroids could not be distinguished from single cells. The indirect assay (alamarBlue), was faster but less accurate than live/dead staining due to dependence on hydrogel permeability. Automated confocal microscope acquisition and cell counting of live/dead staining was more reproducible, reliable, faster, efficient, and avoided overestimates compared to manual cell counting by optical microscopy. Finally, for 1.2 mm thick 3D bioprints, dual-photon confocal scanning with vital staining greatly improved the precision of the evaluation of cell distribution and viability and cell–cell interactions through the z-axis. In summary, automated confocal microscopy and cell counting provided superior accuracy for the assessment of cell viability and interactions in 3D bioprinted models compared to most commonly and currently used techniques.


Introduction
Three-dimensional (3D) bioprinting shows promise for fabricating sophisticated living tissue constructs with well-defined architectures and tunable mechanics.The technology can combine biological components with natural or synthetic biomaterials [1].In orthopedics, 3D printed cell-laden scaffolds have enabled significant innovations for engineering soft and hard tissues [2,3], creating complex in vitro models [4,5], including for drug screenings [6,7].Developing such constructs requires reliable and reproducible assessment of cell viability.During bioprinting optimization, evaluating cell viability is key to assessing cytotoxic effects of bioprinting processing and define ideal ink formulations.Similarly, for drug screening or for studying cell functioning in live-tissue-like structures, precisely tracking cell numbers over time is crucial.In the latter case, non-invasive methods are largely appreciated to avoid affecting biological processes during assay conduction.
Transitioning cell viability assays from 2D to 3D is challenging, as 3D cultures differ significantly from 2D, especially with embedded cells.While standardized 2D methods like metabolic activity (e.g.tetrazolium salts and alamarBlue assays) and DNA quantification (e.g.Hoechst 33 258 and PicoGreen) [8], are routine, assessing cell viability in 3D is less precise.Viability evaluation relies largely on live/dead staining and visual quantification with fluorescence microscope, which only approximate cell numbers in 3D cultures.Live/dead staining is popular for its ease of use, scaffold permeation, and direct 3D mapping with minimal protocol adjustments [9][10][11][12].However, scientists struggle to visualize and count single cells embedded in opaque, multi-layered materials, or after intense proliferation, using conventional microscopy.Staining and counting of bioprinted cells can be unreliable due to several factors: (1) dye permeability depends on ink porosity and composition; (2) scaffold geometry affects accessibility; (3) cell detection and counting accuracy are method-dependent and operator-dependent; and (4) scaffold structures can limit fluorescence imaging.For example, adding opaque nanoparticles to transparent hydrogels obstructs signal detection from embedded cells [13].Many bioinks now contain micro/nanoparticles to better mimic tissues and increase processability, such as alginate/gelatin with calcium phosphate nanocrystals for bone models [14][15][16][17][18][19].However, this increase opacity, challenging cell visualization.
In conclusion, appropriate cell viability tests must be chosen based on material, purpose, cost and time.Here, we compared standard and confocal optical microscopy with manual and automated quantification to assess post-printing viability in commercial cell-laden hydrogels.Our comparative analysis establishes a new standard for evaluating cell viability in bioprinted constructs.

Materials
Two commercial bioinks were selected: alginate-RGD and GelXA BONE (CellInk).They differ in composition and ceramic micro-/nanoparticles content.Alginate-RGD comprises RGD motif (L-Arginine Glycine L-Aspartic Acid)-functionalized, highly purified sodium alginate that crosslinks with divalent cations to improve cell attachment.It contains hydrated cellulose nanofibrils used as a viscosity modifier.GelXA BONE contains gelatin methacrylate (GelMA) plus alginate and a ceramic fraction of tricalcium phosphate and hydroxyapatite microand nanoparticles to mimic bone properties [20].A 50 mM CaCl 2 solution (CellInk) enabled ionic crosslinking.

Bioprinting
We used an extrusion bioprinter (CellInk BioXTM) and two bioinks: alginate-RGD, commonly used to mimic extracellular matrix (ECM), and ceramic nanoparticles-loaded GelMA (GelXA BONE) to recapitulate bone matrix [20].RGD stands to Arginylglycylaspartic acid, the most common peptide motif responsible for cell adhesion to the ECM.For mono-cultures, 1 × 10 7 Saos-2 cells ml −1 were mixed 1:10 cell suspension volume: bioink volume.For cocultures, GFP-Saos-2 cells and Vybrant™ DiI-stained MLO-Y4 or AD-MSC (ThermoFisher Scientific) were mixed 1:3 in alginate-RGD at 1 × 10 7 cells ml −1 (total cell density after mixing).GelXA BONE was sonicated before cell mixing to optimize ceramic particles distribution.After loading into a 22G nozzle (internal diameter 410 µm), both the inks were printed at 37 • C on 24-well plates and crosslinked with CaCl 2 for 2 min (for more details see supplementary information).Alginate-RGD and GelXA BONE allowed accurate printing, but GelXA BONE required higher pressures and caused nozzle clogging from particle settling, preventing proper extrusion (supplementary information).After crosslinking, constructs were washed with PBS (phosphate-buffered saline) buffer, also to eliminate any residual CaCl 2 , and then covered in complete IMDM (Iscove's Modified Dulbecco Medium), and incubated (37 • C, 5% CO 2 ) for 14 d with medium changed every 48 h.

Live/dead assay by fluorescence microscopy
Cell viability was determined on day 1-3-7-14-21 using the Live/Dead™ Cell Imaging Kit (Thermo Fisher Scientific).Briefly, samples were washed with PBS, and stained for 15 min at 37 • C with shaking, washed again, and imaged by fluorescence microscopy.For each sample, live (green) or dead (red) cells were counted in 5 random fields (20× lens).Viability percentage was calculated as live cells/total cells.Experiments were performed in triplicate twice.

AlamarBlue assay
On day 1, cell viability was indirectly assessed using alamarBlue assay versus live/dead staining.The medium was replaced with a 10% v/v alamarBlue solution (Invitrogen) in fresh medium and incubated at 37 • C for 3.5 h.Cell-free hydrogels served as background control.Fluorescence was measured on a microplate reader (Infinite F200 PRO, TECAN) at 535/590 nm excitation/emission.The assay was performed in six replicates.

Statistical analysis
Statistical analyses were performed using GraphPad Prism version 5.00.Due to the low number of observations, non-normal distribution was assumed.The Mann-Whitney U test was used to compare two groups.AlamarBlue and live/dead assay comparison was performed by using the Graph Pad tool for the comparison of linear regression slopes (two-tailed), equivalent to an analysis of covariance (ANOVA).The two-way ANOVA test was used to compare manual live/dead assay evaluation by three different operators.Data were expressed as mean ± standard error of the mean (SEM).Only p-values < 0.05 were considered statistically significant.

Analyzing the performance of live/lead assay by fluorescence microscopy
Firstly, we observed that Alginate-RGD and GelXA BONE constructs maintained their shape until day 14, although GelXA BONE degraded due to progressive GelMA dissolution.We then performed the live/dead assay at day 7 (figures 1(A) and (B)) to evaluate cell distribution by fluorescent microscope and establish the bioprinting parameters (supplementary information), but only for the optically transparent Alginate-RGD bioink since the optically dense inorganic particles in GelXA BONE rendered it opaque.The Alginate-RGD construct showed homogeneous cell distribution (supplementary information).Both inks had approximately 90% post-printing cell viability.However, by day 14, cell aggregates within the construct hindered imaging and precise counting by optical microscopy.Furthermore, manual quantification of live/dead cells using fluorescence microscope was time-consuming with significant operatordependent errors, also affected by underestimation.Three independent operators had significantly different counts for the alginate samples at 72 h (figure 1(C), p = 0.0034), clearly demonstrating that accuracy was heavily reliant on individual technique and interpretation, highlighting limitations.

Analyzing the performance of alamarBlue assay
We then performed the alamarBlues indirect assay as an alternative to the live/dead assay and compared the resultant data.We followed the experimental protocol outlined in figure 2(A) and compared the slopes of both viability tests over time, using both the Alginate-RGD and GelXA Bone-based constructs.Statistical analysis demonstrated that for alginate, the alamarBlue assay could replace the live/dead assay, although this was not the case for GelXA bone, suggesting a different trend of live/dead assay in respect to alamarBlue assay, for this specific material.Specifically, considering the overall slopes of ala-marBlue and live/dead assays as identical, there is a 69.71% and 10.77% chance of randomly selecting data points with slopes as different as those observed for GelXA BONE and alginate, respectively (figure 2(B)).Additionally, the GelXA BONE slope exhibited a considerably higher signal than that detected with Alginate-RGD.

Advantages and limitations of confocal microscopy
Through confocal microscopy analysis of the bioprinted construct, we detected individual cells throughout the depth of the sample, although this was less feasible for opaque and autofluorescent materials such as GelXA compared to alginate (figure 3(A)).Automated object counting modules can be integrated with confocal microscopy analysis (figure 3(B), blue mask).When comparing the results obtained from fluorescence and confocal microscopy analyses only for RGD-alginate samples, we found that the number of viable cells was significantly lower by confocal microscopy at 24 h, suggesting that the use When comparing different acquisition and counting methods on the same sample, combining confocal microscopy with automated counting appeared highly advantageous due to its precision, counting 1670 fields on average, compared to only 5 fields with optical microscope.Furthermore, confocal microscopy with automated counting required loss acquisition and counting time (table 1).The confocal microscope speed could also be reduced without increasing error by augmenting the distance between scanning steps below cell size, ensuring all cells were counted (table 1).
Using dual-photon confocal microscopy, it is possible to evaluate cell spatial distribution along the z-axis in relation to bioprinted construct height (figure 3(D)).However, this assessment is only possible for materials with permissive optical densities, even at millimeter depths with good resolution and intensity.This is not achievable for materials containing dense particles like hydroxyapatite in GelXA Bone (figure 3(D)).Automated counting can still be performed in this case, compared to optical microscopy.Again, we observed overestimation of viable cell in alginate-RGD based constructs by optical microscopy, with significant differences at 24 h (figure 3(E)).

Quantification of live/dead cells in spheroids/clusters and drug screening
The confocal microscope is crucial for counting bioprinted constructs cultured long-term with proliferating cells (figures 4(A) and (B)), or for bioprinted spheroids.In both cases, optical counting often cannot distinguish individual cells from clusters or accurately detect t dead cells within cluster.As seen in figure 4(B), the object diameter to be counted in construct with single cells or clusters is very different, but this difference may not be discernible by fluorescence microscopy, while noticeable by automated counting or measurement systems integrated into software.Notably, clusters formed in GelXA BONE were greater and more numerous than in alginate-RGD (figure 4(A)), suggesting higher proliferative capacity in GelXA BONE, confirming ala-marBlue data (figure 2(B)).In this case, even preformatted algorithms for live/dead cell counting may be inadequate for live/dead staining, resulting in gross errors as clusters contain both green and red signals.Here, confocal microscopy can precisely measure green and red volumes by defining specific thresholds (figure 4(C)).We provided an example of how more accurate viability analysis for drug screening by confocal microscopy distinguishing green (viable) and red (dead) volumes with thresholding (figure 4(C)) differs from quantifying total spheroid diameter/volume.We treated bioprinted Saos-2 spheroids of 14 d with DXR at 0-2-10 µM, and quantified changes in total spheroid volume, corresponding to diameter changes, measurable by conventional microscopy and live/dead volume ratios (figure 4(E)).The former showed no difference, while the latter revealed significant dose-dependent inhibition of live cells.The observed effect is likely attributable to dead cells becoming trapped within the hydrogel, unlike in 2D cultures where they are washed away.As such, spheroid diameter does not necessarily represent temporal changes in viable cell volume, whereas precisely determining the ratio of live to dead cell volume in spheroids provides more informative data.

Use of live staining for the study of tumor-stroma interactions in bioprinted constructs
Confocal microscopy with fluorescent viability stains like calcein AM or GFP can accurately evaluate interactions between two cell types in tissues, as in living organisms.We used GFP as an example viability marker due to its short half-life, making it reliable for assessing living cells.We co-bioprinted GFP-Saos-2 cells with AD-MSC or with MLO-Y4 osteocytes (figure 5(A)).We pre-labeled MSCs or osteocytes with a red tracer.
First, we observed homogeneous distribution of both cell types throughout the bioprinted construct using dual-photon confocal microscopy (figure 5(B)).This simple experiment revealed completely different behavior of non-transformed cells near living osteosarcoma cells.AD-MSCs localized near tumor cells but without embracing them to form mixed clusters (figure 5(A), enlarged squares).In contrast, osteocytes created numerous cell-to-cell contacts with osteosarcoma cells, eventually forming homogeneous mixed cell groups.

Discussion
Recent developments in tissue engineering and in vitro tissue modeling have enabled growing cells within 3D bioprinted cultures that closely resemble complex 3D environment and tissue structures.However, evaluating behavior and viability of cells within these constructs can be challenging.The live/dead assay allows quantitative analysis within scaffolds/materials by microscopy and is the most widely used method to screen cell viability in bioprinting [22], since it has an easy and fast approach.To conduct a comparative analysis of quantification methods for estimating post-3D printing viability of cell-laden commercial hydrogels with different chemical compositions, we selected alginate and gelatin methacrylate.These are the most commonly used bioprinting inks, with different characteristics in permeability, printability, and autofluorescence that may interact differently with cell staining and observation.Additionally, we considered enrichment of gelatin-based bioink, with ceramic micro-and nano-particles, often used to mimic bone.However, due to their high specific surface area and opaque nature, they may interfere with staining and/or cause an undesired background fluorescent noise.We found that GelXA BONE was optically opaque, hindering cells for the analysis by fluorescent microscope and, by day 14, the formation of cell aggregates preventing single cell counting.Furthermore, manual quantification was timeconsuming, for both bioinks, with significant operator errors and underestimation.Accuracy depended on individual technique and interpretation, highlighting limitations of cell counting in bioprinted samples, when combining live/dead assay and manual counting by fluorescence microscope.Indirect metabolic tests, like ATP assay, tetrazolium reduction assay (MTT), or alamarBlue assay, could offer a solution, as they are faster, cost-effective, and, may not require sacrificing the sample at each endpoint, like for ala-marBlue, allowing culture monitoring over time.Moreover, they are not influenced by the presence of opaque structures within the construct.In this study we analyzed, as an example, alamarBlue, which uses a redox indicator [23].We selected alamarBlue as it has been frequently used for in vitro 3D models [22,24,25].
We compared it to live/dead assay over time.For alginate, alamarBlue could replace live/dead but not for GelXA BONE.We found a large discrepancy between the assays.This may be due to cell cluster formation causing underestimation in live/dead manual counting, unlikely alamarBlue which measures metabolic activity unaffected by cell cluster formation.Moreover, we saw a clear divergence in alamarBlue slope trends between GelXA BONE and alginate-RGD.This could be from higher proliferation in GelXA Bone, or different bioinks properties causing background signals or limiting reagents/metabolites diffusion.Material porosity and density can significantly affect entry and exit of ala-marBlue substrates and products, causing different metabolic activity levels not correlating linearly [8,25].Finally, indirect tests often use extracted culture supernatant.However, culture medium volume of 3D cell-scaffold construct at the end-points varies greatly, especially because evaporation rates can vary across compartments due to factors like edge proximity and, especially for small volumes, these percentage differences may be considerable, substantially impacting substrate concentration and detection reaction kinetics.Furthermore, the supernatant may not represent the inner environment if a metabolite gradient exists between construct interior and exterior.Therefore, although a more precise quantification method would ensure accuracy, for bioprinted materials, the live/dead direct assay appears superior.More precisely quantification of live-dead assay may be reached by using confocal microscopy which, indeed, enables precise spatial visualization and localization of cells, even along the z-axis.Confocal microscopy can optically section fluorescent signals, discriminating between cells at different heights and spheroids from single cells, as well as observing cells beyond the surface.As a confirmation, we found that confocal microscopy with automated counting provides superior accuracy and faster performance for quantifying viable cells in bioprinted constructs compared to standard fluorescence microscopy.As an example, a recent bioprinted osteosarcoma model for drug screening used indirect markers like Ki-67 and qualitative confocal analysis [26].Also in such cases, automated confocal image analysis or multifocal microscopy may enable direct evaluation of drug effects on viability, greatly improving results reliability.
For anti-cancer drug screening on 3D bioprinted models, live/dead staining is often analyzed by counting green cells/spheroids or measuring diameter.However, both can be non-representative.This is likely because dead tumor cells remain trapped within the hydrogel unlike 2D cultures, erroneously constituting spheroid volume.This behavior reflects in vivo tumors containing necrotic areas after chemotherapy.The presence of these area are an important prognostic factor for osteosarcoma [27], and are thus crucial for modeling tumor response to drugs and in these cases, confocal microscopy analysis is essential.
Finally, confocal microscopy with cells expressing GFP, which has a remarkable short half-like [28], is another reliable viability assay.It can track interactions between different live cells in living organisms using bioprinted models reproducing tissue heterogeneity.To analyze the advantages of confocal microscopy for this assay and model, we used a bioprinted osteosarcoma model.We chose AD-MSCs and osteocytes as representative stromal cells in the bone microenvironment, where they modulate tumors progression [29][30][31][32].Reconstructing their 3D context in vitro may clarify the role of intercellular interactions for drug resistance and help in identifying new targets.Confocal microscopy precisely localized fluorescent signals from living cells, highlighted distinct behaviors of MSCs and osteocytes with osteosarcoma cells.

Conclusions
The live/dead assay for bioprinted constructs using standard fluorescent microscopy is time-consuming with high inter-operator variability and lack of precision.Indirect assays are quite reliable but bioink material characteristics can affect them.Precision of information on cell distribution and viable cell fraction with the live/dead assay improves greatly using confocal microscopy and automatic counting, especially for studying clustered viable cells, drug screening, and cell-cell interactions.For this reason, especially for studies in which it is critical to accurately determine the number of live cells in the bioprinted construct or to compare different treatments, analysis by confocal microscopy is necessary.Admittedly, confocal microscopes are not currently accessible to all research laboratories.However, as technology advances and the cost of basic instrumentation decreases, such instruments will become increasingly available and used by the research community.

Figure 1 .
Figure 1.Live/dead assay at the fluorescent microscope.Live-dead assay of bioprinted RGD-alginate (A) and GelXA Bone (B) constructs with Saos-2 cells at different time-points as observed by fluorescence microscopy (representative images in the left panel, scale bar 50 µM, and graphs of the quantitative analysis in the right panel, mean ± SEM).(C)Statistical analysis of inter-operator variability for the manual quantitative evaluations of the acquired images of live/dead assay shown in panel A, at day 1 (mean ± SEM, n = 5, two-way ANOVA test, * * p < 0.0034 when we consider the comparison between operators as a factor, whereas it was not significant when we considered the comparison between the two samples as a factor).

Figure 3 .
Figure 3. Live/dead by confocal microscopy with automatic quantification.(A) Confocal microscope acquisition of bioprinted constructs 24 h from seeding with two bioinks with different optical features (RGD-alginate and GelXA bone), thereby allowing the visualization of cells at different depth (scale bar 100 µM).In the left, merge of all xy-scans that we performed in the z-axis of the same sample; in the middle, xy, xz, yz projections of a representative scan, at a specific xy-, and z-point, as enlighten by the red dotted line, among those used for the merged image; in the right, volume render of the same scan sections used in the merged images; (B) automatic cell count by image analysis software after confocal microscope acquisition (on the right panel the mask applied for the cell counting process by the software on a single xy scan is shown) (scale bar 100 µM); results of the automatic quantification compared to the manual quantification of RGD-alginate based constructs analyzed by standard confocal microscopy (C) or dual-photon confocal microscopy (E) (Mann Whitney U test, 0.01 < * p < 0.05, 0.005 < * * p < 0.01, mean ± SEM, n = 10).In (D) the volume renders of the dual-photon confocal scan performed on the two constructs are also shown.

Figure 4 .
Figure 4. Cell quantification of cell clusters embedded in bioinks after live/dead assay and standard confocal microscope acquisition.(A) Cell clusters in two different bioinks at 14 d of cell cultures from bioprinting (scale bar 100 µM); (B) representative images of single cells seeded in RGD-alginate at 3 and 14 d after cell culturing from bioprinting (the diameters of two representative cell clusters are also shown); (C) masks of the image analysis software that recognizes green and red volumes and edges in 3D objects (scale bar 100 µM); (D) representative images of cell clusters in the bioprinted constructs after treatment with different DXR concentrations (0-2-10 µM) (scale bar 50 µm); (E) graphs of the spheroids diameters and live/dead volume fractions of cell clusters treated with different DXR concentrations (mean ± SEM, n = 10, Mann Whitney U test, n = XX, * p < 0.05).

Figure 5 .
Figure 5.Using bioprinted constructs and vital cell marker to study tumor-host interactions.(A) Representative confocal images of a single scan (FITC, TRITC, and transmitted light differential interference contrast images using the argon-ion laser set at 488 nanometers) of bioprinted models of tumor-stroma co-cultures (osteosarcoma GFP-Saos-2 cells in green and normal MSC or osteocytes in red).In the lower panel, representative merged images of the scans of the entire bioprinted scaffold of green and red signals of the same constructs shown in the upper panel.Zooms are shown in the rectangles (scale bar, 50 µM).(B) Dual photon microscope acquisition of the constructs shown in panel A: volume renders images.At the bottom, zooms of single cell clusters are shown.

Table 1 .
Comparison of time required and number of image fields acquired with different microscopy techniques and settings.