Physical Characterization of Colorectal Cancer Spheroids and Evaluation of NK Cell Infiltration Through a Flow-Based Analysis

To improve pathogenetic studies in cancer development and reliable preclinical testing of anti-cancer treatments, three-dimensional (3D) cultures, including spheroids, have been widely recognized as more physiologically relevant in vitro models of in vivo tumor behavior. Currently, the generation of uniformly sized spheroids is still challenging: different 3D cell culture methods produce heterogeneous populations in dimensions and morphology, that may strongly influence readouts reliability correlated to tumor growth rate or antitumor natural killer (NK) cell-mediated cytotoxicity. In this context, an increasing consensus claims the integration of microfluidic technologies within 3D cell culture, as the physical characterization of tumor spheroids is unavoidably demanded to standardize protocols and assays for in vitro testing. In this paper, we employed a flow-based method specifically conceived to measure weight, size and focused onto mass density values of tumor spheroids. These measurements are combined with confocal and digital imaging of such samples. We tested the spheroids of four colorectal cancer (CRC) cell lines that exhibit statistically relevant differences in their physical characteristics, even though starting from the same cell seeding density. These variations are seemingly cell line-dependent and associated with the number of growing cells and the degree of spheroid compaction as well, supported by different adenosine-triphosphate contents. We also showed that this technology can estimate the NK cell killing efficacy by measuring the weight loss and diameter shrinkage of tumor spheroids, alongside with the commonly used cell viability in vitro test. As the activity of NK cells relies on their infiltration rate, the in vitro sensitivity of CRC spheroids proved to be exposure time- and cell line-dependent with direct correlation to the cell viability reduction. All these functional aspects can be measured by the system and are documented by digital image analysis. In conclusion, this flow-based method potentially paves the way towards standardization of 3D cell cultures and its early adoption in cancer research to test antitumor immune response and set up new immunotherapy strategies.

To improve pathogenetic studies in cancer development and reliable preclinical testing of anti-cancer treatments, three-dimensional (3D) cultures, including spheroids, have been widely recognized as more physiologically relevant in vitro models of in vivo tumor behavior. Currently, the generation of uniformly sized spheroids is still challenging: different 3D cell culture methods produce heterogeneous populations in dimensions and morphology, that may strongly influence readouts reliability correlated to tumor growth rate or antitumor natural killer (NK) cell-mediated cytotoxicity. In this context, an increasing consensus claims the integration of microfluidic technologies within 3D cell culture, as the physical characterization of tumor spheroids is unavoidably demanded to standardize protocols and assays for in vitro testing. In this paper, we employed a flowbased method specifically conceived to measure weight, size and focused onto mass density values of tumor spheroids. These measurements are combined with confocal and digital imaging of such samples. We tested the spheroids of four colorectal cancer (CRC) cell lines that exhibit statistically relevant differences in their physical characteristics, even though starting from the same cell seeding density. These variations are seemingly cell line-dependent and associated with the number of growing cells and the degree of spheroid compaction as well, supported by different adenosine-triphosphate contents. We also showed that this technology can estimate the NK cell killing efficacy by measuring the weight loss and diameter shrinkage of tumor spheroids, alongside with the commonly used cell viability in vitro test. As the activity of NK cells relies on their infiltration rate, the in vitro sensitivity of CRC spheroids proved to be exposure time-and cell line-dependent with direct correlation to the cell viability reduction. All these functional aspects can be measured by the system and are documented by digital image analysis. In conclusion, this flow-based method potentially paves the way towards standardization of 3D cell cultures

INTRODUCTION
In the last years, growing interest has developed in creating and evolving three-dimensional (3D) culture systems to allow more detailed studies of cancer biology and response to therapy (1)(2)(3)(4). Indeed, 2D cultures cannot reproduce all the complex features and cell-to-cell interactions occurring at the site of lesion. On the other hand, animal models do not always reproduce the pathophysiology of human cancers, besides being very expensive and showing a negative environmental impact (5)(6)(7)(8). Several 3D culture systems, including spheroids, have been validated by the European Union Reference Laboratories for Alternatives to Animal Testing (EURL ECVAM) as preclinical models (9-11). These systems possess many advantages over 2D cultures or animal models, such as reproducibility, the high number of replicates, systematic evaluation of multiple parameters and the possibility to set up standardized cocultures with different cell types (12)(13)(14). In this context, we used the spheroid culture system to evaluate some aspects of the antitumor immune response exerted by a subset of T lymphocytes against colorectal cancer (CRC) cells (15). Thus, spheroids can be a reliable 3D culture system that allows the co-culture of cancer cells and effector immunocompetent cells. However, static culture conditions do not allow the evaluation of all the dynamic events occurring during tumor growth, moreover, such systems cannot fully overcome the limited distribution of oxygen or nutrients and waste removal operating in vivo. As a consequence, there is a growing interest for the combination of microfluidic technologies within 3D cultures, including tumor spheroids, to reproduce more faithfully the real tissue microenvironment that undergoes multiple chemical and mechanical challenges, eventually leading to changes in physical parameters (16)(17)(18)(19).
Physical features of a tumor include spatial cell organization, external mechanical stimuli, extracellular matrix architecture and stiffness. Mechanobiological modifications, such as active stretch and tension, are other important functional aspects that may condition tumor cell growth and drug response (18,19). Such microenvironmental aspects should be considered when using 3D culture methods, including multicellular spheroids that take advantage of the natural disposition of several epithelial tumor cells to aggregate (16,19). Thus, the measurement of physical variations of growing tumors in vitro represents a useful tool in cancer biology, both for elucidation of the mechanisms underlying tumor development and spreading and for testing anti-cancer drugs.
In a recent paper, a linear relationship between impedance and cell number was found for some tumor cell lines and referred to proliferation rates (17). Also, cell size can be measured as volume or mass, as an indicator of cell integrity and state; more precisely, the ratio between mass and volume (i.e. mass density) can be used to distinguish between cell populations even when volume and mass do not vary (20). From this viewpoint, mass density may evidence early modifications in cellular composition that precede size and weight changes, such as organelle enlargement underlying the high metabolic rate of neoplastic cells or nuclear segmentation that occurs at the beginning of apoptotic process due, for example, to anti-cancer drugs (21). In more advanced phases of cell proliferation, mass density might also be influenced by cell number, especially during the first cellular duplications that conceivably do not alter weight and volume yet. In principle, since variations in mass density and volume occur during cell growth and are connected to changes in the cell cycle, the measurement of cell weight and mass density can provide a direct evaluation of the biological processes, underlying tumor progression and environmental changes (20,21).
Mass density is usually indirectly monitored by software elaboration (22) of bi-dimensional images to assess the compaction degree of cells inside the spheroids, even if 2D projection might alter the real 3D structure. Some papers report mass density as "tightness" (23) "solidity" (24), "optical density" (25) and "compactness index" (26), each one presenting some measurement variations due to different chosen parameters or image quantification software. Therefore, these parameters are used to assess the effects of a specific treatment in the spheroids, such as a toxicity test, efficacy study, tumor cell resistance, and others (24).
Several methods for determining weight and mass density of cells were established with the development of nanomechanical resonator or electrokinetic microfluidic chip (20,(27)(28)(29)(30). However, these techniques are not intended, nor proven, for reaching the average size of spheroids and organoids. Indeed, the above-mentioned systems are suitable for corpuscles ranging between nanometers to a few micrometers scale, as single cells or small samples. Moreover, data are often collected in one step, limiting sample replication and analysis repetition essential for accurate biometric studies. We have recently developed a precise and rapid technique, that enables the simultaneous determination of weight, size and mass density of sphere-like samples (31).
In the present paper, we adopted this method to further describe the physical intrinsic differences of spheroids generated from four human CRC cell lines. Furthermore, by this system, we analyzed possible variations induced by spheroid interaction with anti-cancer effector lymphocytes: natural killer (NK) cells were chosen due to their reported role in anti-tumor immunity against CRC. As effectors of natural immunity, they work without the need of recognizing tumor associated antigen by specific receptors, such as the T cell receptor, in the HLA-class I context. Thus, their action is efficient, although broadly directed towards a large panel of cancers (32)(33)(34). In this experimental setting, we recorded variations of these parameters depending on interactions between CRC spheroids and antitumor natural killer (NK) cells during the killing process. These changes are dependent on CRC infiltration by NK cells, followed by the elimination of tumor targets.

Cell Cultures and Tumor Spheroid Generation
The HT-29, HCT-15, SW620, and DLD-1 were analyzed by indirect immunofluorescence and flow cytometry for the expression of the epithelial-specific antigen (ESA) and HLA-I with the specific monoclonal antibodies (mAbs) TROP1 (R&D System, Minneapolis, MN, USA) and W6/32 (American Type Culture Collection, Manassas, WA, USA) respectively, followed by Alexafluor647-goat anti-mouse anti-isotype antibody (GAM, ThermoFisher Scientific,) and cytofluorimetric analysis as reported (15). As shown in Supplementary Figure S1, all the cell lines were ESA positive, but only HT-29 and SW620 expressed HLA-I. CRC spheroids were generated as described (15)

Measurement of Weight, Diameter, and Mass Density of Spheroids
The adopted method is based on tracking the sample's motion, when free-falling into a vertical flow-channel while the flow is at rest, to calculate its terminal velocity. This is achieved by using customized software that relies on a specific physical method, combined with detailed statistical analyses (31). Briefly, as shown in Supplementary Figure S3, the device is composed of a fluidiccore chip, equipped with a bright-field imaging setup, a peristaltic pump, a temperature sensor and a flow-circuit for the introduction and elimination of samples. Furthermore, the software assigns a circular reference to each image of the falling sample, allowing the extrapolation of the final radius from the physical calculation, as the average of the maximum radii obtained from each repetition. The terminal velocity is whereas calculated from the vertical position of the falling sample, and for each repetition a linear regression plot is derived. During the entire analysis different internal data control are performed: (i) initial visual screening of the operator; (ii) circular reference assignment; (iii) measurements repetition; (iv) calculation of the average radius; (v) linear regression plot analysis; and (vi) statistical validation with outliers elimination. All these check points insure the conformity of the analyzed samples in term of shape and size. CRC spheroids were fixed with PFA 4% overnight at 4°C, resuspended in 3.5mL of Dulbecco's phosphate-buffered saline (DPBS), 1X w/o Ca 2+ & Mg 2+ (Corning ® Life Sciences) at low concentration (<200 spheroids/mL), transferred in a centrifuge conical tube and then analyzed according to the previous protocol performed by Cristaldi et al. for biological samples (31). A minimum of 10 single spheroids was analyzed for every test condition and values were extrapolated from at least 10 repetitions. The same procedure was applied to CRC spheroids exposed to NK cells.The fixation protocol had no effect on mass density, as no statistical difference was observed in live and fixed spheroids of the 4 CRC cell lines (not shown). In particular, no statistical differences were found in weight and diameter variations in fixed or unfixed HT-29 and HCT-15 spheroids, while weight changes in SW620 (+53%) and DLD-1 (−49%), as well as diameter (+21% in SW620 and −14% in DLD-1). Since variations are concordant (a spheroid with a big diameter, also has a big weight, while a small spheroid in size also presents a little weight), mass density is not influenced: in fact, these spheroids show the same mass density values, as mass density normalizes the values of diameter and weight.
Cell Viability and Cytolytic Assay by Crystal Violet Staining NK cells were obtained from peripheral blood mononuclear cells by negative depletion using RosetteSep NK negative selection kit (StemCell Biotechnology, Vancouver, Canada) as described and cultured for 15 days with interleukin-2 (IL-2, 30IU/ml, Miltenyi Biotech, Milan, Italy) (34). NK cell population used in functional assays was >97% CD3 negative CD56 positive (34) (not shown in this paper). The number of CRC cells in a culture well was determined measuring the ATP spheroid content in that well, referred to the ATP content of the same cell line at a known cell concentration in suspension. The optimal amount of NK cells to detect the cytotoxic effect, determined by crystal violet assay or by image analysis, was 0.75x10 5 cells/well and it corresponded approximately to a 1:1 effector to target (E:T) ratio (15

Cytofluorimetric Analysis of Spheroid Invasion
Spheroids incubated with NK cells as described in paragraph 1.5, were washed to remove free-floating NK cells. Staining with the anti-CD56 mAb C218, followed by Alexafluor488 anti-isotypespecific GAM (34) was then performed to label external NK cells, taking into consideration that mAbs can penetrate some peripheral layers of the spheroid (not shown), thus labeling also NK cells infiltrating spheroid periphery. The spheroids were then dissociated in phosphate buffer without Ca ++ /Mg ++ and the resulting cell suspension was labeled with the anti-CD45 mAb 9.4, followed by Alexafluor647 anti-isotype-specific GAM (34)

Statistical Analysis
The Shapiro-Wilk test was performed on the measured outputs obtained from the biological experiments to analyze the distribution of the dataset based on skewness and/or kurtosis (36). For all the cases that resulted in a non-normal distribution, descriptive statistics box plots (Tukey method plots) were carried out for determining outliers' values. Outliers were identified as individual points for the terminal velocity, mass density, diameter, and weight box plots. The presence of at least one outlier in one of the categories was considered sufficient to remove the related sample from the dataset. For these cases, the Shapiro-Wilk approach was reused to confirm the normal distribution. Data are presented as mean ± SEM or ±SD. Statistical analysis was performed using two-tailed unpaired Student's t-test. The cutoff value of significance is indicated in each figure legend.

Characterization of the Physical Properties of CRC Spheroids
We analyzed tumor spheroids obtained culturing the representative CRC cells lines HT-29, SW620, HCT-15, and DLD-1 in ultra-low attachment plates, as reported (15). Ten single spheroids for each cell line were analyzed, and each spheroid values of mass density, diameter and weight, and the related standard deviation, are extrapolated from 10 repetitions. Experiments were performed on a heterogeneous population of spheroids in terms of dimension, to test the feasibility of our methods in cell aggregates having the different size ranging from 100 to 200µm diameter, as shown in Figure 1A. HCT-15, DLD-1, and SW620 spheroids displayed round shape with a smooth surface, while HT-29 spheroids showed irregular shape with a rough surface ( Figure 1A). CRC spheroid samples were fixed with 4% PFA and analyzed with the flow-based system: as shown in Figure 1B, SW620 and DLD-1 spheroids' weights (ng, left graph) were higher than that of HT-29. Also, their diameter (mm, central graph), calculated automatically from the images acquired during the free-fall motion, were larger than those of HT-29 spheroids. Noteworthy, the measured mass density of the sample was consistently higher in SW620 and DLD-1 than HT-29 spheroids (fg/mm 3 , right graph). As mass density represents a  direct parameter to evaluate the degree of aggregate compaction, data agreed with the preliminary microscopic investigation, where HT-29 cells formed loose aggregates, instead of compact 3D tumor spheres. Weight and diameter of HCT-15 spheroids were comparable to that of SW620 and DLD-1 ones, although much more dispersed due to sample intrinsic heterogeneity ( Figure 1B, left and central graphs); however, their mass density was similar to that of HT-29 and significantly different from that of DLD-1 ( Figure 1B, right graph).

Quantification of Cell Number and ATP Content Within the CRC Spheroids
To better clarify the biophysics of mass density heterogeneity, spheroids of HT-29, SW620, DLD-1, or HCT-15 cell lines underwent nuclear staining with the green Syto16 probe followed by laser scanning confocal microscope analysis. Images were taken at different Z points set every 10mm (Figure 2A), spheroids were identified with the threshold tool of the Image J software in red pseudocolor, while nuclei were evidenced in blue pseudocolor ( Figure 2B) and counted with the multipoint analyze particle tool (30 spheroid/cell line counted on 6 Z points/spheroid). Figure 2C shows in SW620 spheroids a striking higher cell number/mm 2 than in HT-29, DLD-1, and HCT-15 spheroids: the latter two, in turn, contain many more cells than HT-29. Intracellular ATP content was detectable in all spheroids, documenting cell viability ( Figure 2D). The higher ATP content in SW620, and to a lesser extent in DLD-1, than in HT-29 or HCT-15 spheroids, can be referred to the higher cell number, mainly evident in SW620 spheroids ( Figure 2D vs Figure 2C). Since DLD-1 and HCT-15 spheroids contained approximately the same number of cells, differences in ATP content may depend on a different metabolic state. This is also suggested by the finding that ATP values measured in cell suspensions of each cell line display differences among the four cell lines, as shown in Figure 2E.

Cytotoxic Activity of NK Cells on CRC Spheroids Evaluated by Weight, Diameter, and Mass Density Measurement
We further planned to test the variations of physical parameters caused by the antitumor effect of NK cells and occurring during the killing process. Then, SW620, HCT-15, and DLD-1 spheroids were chosen due to their comparable weights,  diameters and shape, although with different mass density. On the contrary, HT29 aggregates exhibited significantly less weight and lower, dispersed values of mass density. Confocal analysis, followed by orthogonal cross-section reconstruction from a set of a z-stack scans, showed that HT-29 spheroids were far from being spherical (example shown in Figure 3A) so that the flowbased system struggled to properly assign them a circular reference and, as a consequence, also the final calculated radius was affected by the samples' irregularity. Conversely, SW620, HCT-15, and DLD-1 3D structures predominantly displayed a round shape in the orthogonal views of the z planes analyzed (z3, z6, and z9 shown in Figures 3B-D), indicating that they can be considered spheroids. HT29 were thereby considered less suitable models for the reliable testing of infiltrating NK cells. CRC tumor spheroids were co-incubated with NK cells at E:T ratio of 1:1 and parallel samples were analyzed for weight, diameter and mass density at 6h and 24h, while cell viability was measured with crystal violet assay. In SW620 and DLD-1 CRC spheroids, co-culture with NK cells led to a statistically significant decrease in weight (Figures 4Aa, Ba) and diameter (Figures 4Ab, Bb). In particular, SW620 weight loss was about 47% at 6h and 23% at 24h, with a decrease in the diameter of 18% and 13% respectively (Figures 4Aa, Ab), while DLD-1 and HCT-15 weight and diameter decreased at 24h by 34% and 13% for DLD-1 (Figures 4Ba, Bb) versus 23% and 7% for HCT-15 (Figures 4Ca, Cb). Notably, these changes were already evident at 6h, when cytotoxicity with the crystal violet assay, that is determined after 6h or 24h followed by further 24h of cell adhesion, was still undetectable or negligible (Figures 4Ad, Bd). Furthermore, a tendency for an increase in the mass density of SW620 and DLD-1 spheroids was observed upon co-incubation with NK cells, with different kinetics, as depicted in Figure 4. The tendency for an increase in SW620 mass density was already evident at 6h (Figure 4Ac, brown boxes and whiskers); conversely, in DLD-1 (Figure 4Bc, light blue boxes and whiskers) spheroids mass density significantly raised at 24h, immediately before the detection of cytotoxicity by crystal violet assay (Figure 4Bd). Weight and diameter of HCT-15 showed a tendency for a decrease only after 24h of coincubation with NK cells (Figures 4Ca and Cb), although cytotoxicity was already evident at 6h (Figure 4Cd); mass density did not vary significantly, conceivably due to the small decrease in weight and diameter (Figure 4Cc vs Figures 4Ca and  Cb). Thus, weight measurement can precisely reveal spheroid variations, due to NK cell cytotoxic activity, even earlier than crystal violet assay and in addition to other information such as diameter modifications and mass density changes.
When the cytolytic assay was performed with CRC cells in suspension, we observed that DLD-1 was less susceptible than SW620 and HCT-15 cell lines to NK cell-mediated lysis (Supplementary Figure S4A); however, this difference was not evident using the three cell lines as adherent targets (Supplementary Figure S4B), suggesting that cancer cell shape might influence the outcome of NK cell anti-tumor activity.

NK Cell Infiltration of CRC Spheroids
We further investigated whether mass density variations in SW620 and DLD-1 spheroids were associated with NK cell infiltration. To this aim, NK cells were labeled with CFSE probe (green) and NK cell entry into the spheroids was   Frontiers in Immunology | www.frontiersin.org December 2020 | Volume 11 | Article 564887 evaluated after 24h of co-culture. These two CRC spheroids were chosen since they underwent significant variations in their mass density upon 24h of interaction with NK cells. The method used for cell count applied the Multipoint and Analyze Particle tool of the Image J software to the ROI defined as single spheroids, as described in Supplementary Figure S2. Data were plotted as the number of infiltrating NK cells/mm 2 . Figure 5A shows SW620 spheroids, surrounded and progressively infiltrated by NK cells (green CFSE + ), documented by the representative images taken at 10 different Z planes. In Figure 5B, NK cells invading DLD-1 spheroids are depicted, displaying an elongated shape that indicates their position inside the spheroid. A significantly higher number of NK cells could infiltrate DLD-1 ( Figures 5C,  D, blue boxes and whiskers) compared to SW620 (Figures 5C,  D, red boxes and whiskers) spheroids, where NK cells accumulated at the periphery ( Figure 5C: mean±SD of NK cells/mm 2 in a single spheroid evaluated at 10 different Z positions; Figure 5D: mean±SD of NK cells/mm 2 infiltrating 20 spheroids evaluated each at 10 different Z positions). These data might explain the differences in mass density detected with the flow-based system and reveal a different mode of action of NK cells to attack tumors, depending on the biological features of tumor spheroids associated with their physical properties. Parallel specimens of spheroids incubated with NK cells were extensively washed, to remove unbound NK cells, and counterstained with the anti-ESA mAb, followed by Alexafluor647-GAM. Samples were analyzed by confocal microscopy with the FluoView 4.3b software. Figure 6 shows three representative z-stack images, out of 10 set every 10mm, of SW620 (A) vs DLD-1 (B) spheroid (CRC cells identified in red as ESA-positive) infiltrated by NK cells (green CFSE + ). DLD-1 or SW620 incubated with NK cells were also included in melted agarose and paraffin embedded for IHC; serial sections were cut and stained with the anti-CD45 mAb to detect NK cells (C and D). Immunofluorescence was analyzed by Image J software on the ROI designed on the inner spheroid perimeter, to exclude NK cells confined in the external spheroid layer, defined on the basis of ESA staining; 6 spheroids/cell line were analyzed at 10 different Z positions and the number of NK cells calculated with the Image J Multipoint Analyze Particle tool. Figure 6E shows that the number  Figure 5, incubated with CFSE-labeled NK cells (green, E:T ratio of 1:1) for 24h, washed and counterstained with the anti-ESA mAb TROP-1, followed by Alexafluor647-GAM (red). Samples were run under the FV500 confocal microscope and analyzed with FluoView 4.3b software (Olympus). Images were taken at different Z planes (Z1-Z10) with a 20x objective NA 0.40: three representative z stack sections are shown. Orthogonal cross-section (indicated as yellow line for x axis, purple for y axis and green for z axis) were reconstructed from z-stack scans (80 µm total thickness). For each panel: a) an example of x-y focal plane of the z reported in the subpanel a; b) side view of the zstack (orthogonal x-z plane) for subpanel a); c) view in the orthogonal y-z plane for subpanel a. Bar in subpanels a: 100µm.  of NK cells inside DLD-1 spheroids was significantly higher than that of NK cells infiltrating SW620 spheroids (p<0.0001). That NK cell infiltration of DLD-1 was more efficient than that of SW620 spheroids was also documented by IHC and digital imaging ( Figure 6C vs Figure 6D). Image J Multipoint Analyse Particle tool was applied on the ROI defined on the visible spheroid perimeter on 10 serial sections. Also in this case, the number of infiltrating NK cells was significantly greater in DLD-1 than in SW620 spheroids ( Figure 6F, p<0.0002).
To further verify the degree of NK cell infiltration of DLD-1 vs SW620 spheroids, staining with the anti-CD56 mAb, followed by Alexafluor488 anti-isotype-specific GAM was performed to label external NK cells. The spheroids were then dissociated and the resulting cell suspension was labeled with the anti-CD45 mAb, followed by Alexafluor647 anti-isotype-specific GAM to identify all NK cells, including those derived from the inner part of tumor spheroids. Thus, CD45 single-positive cells should identify bona fide deeply infiltrating NK cells. Samples were then analyzed by flow cytometry, gating on lymphocytes based on FSC/SSC physical parameters ( Figure 7A, left plots). Of note, the percentage of CD45 + CD56 -NK cells was much higher in cell suspensions derived from DLD-1 than from SW620 spheroids (39.1% vs 12.2% NK1, Figure 7A right plots; 26.6% vs 16% NK2, Figure 7B). These findings are in line with those obtained by confocal microscopy and digital imaging analysis.

DISCUSSION
Alongside with the optimization of 3D cell culture methods, many efforts have been made to develop new technologies for the full characterization of these complex spherical aggregates (3,37,38). As regards the biophysical characterization of size and mass density, although few technical solutions have been presented for single-cell analysis, literature does not cover this aspect for 3D models of hundreds of micrometers in diameter, such as spheroids (30).
To overcome this shortage, microfluidics provides a promising technique due to its peculiar properties, including liquid handling automation, small volumes of reagents/sample required, and cost-effective fabrication. In this work, we have described a flow-based technology and the associated method for the non-invasive and accurate measure of size, weight and mass density. Although both mass and volume are important parameters for a comprehensive physical overview of the samples, mass (i.e. weight) is more fundamentally related to cell growth than is volume, thereby altering the density (39). Volume can change disproportionately to mass, thereby altering a cell's density (39). As previously shown, this system relies on software-driven process automation that remarkably increases the ease of use of the device (30).
Tumor spheroids composed of four different CRC cell lines (HT-29, SW620, DLD-1, and HCT-15) were analyzed with this flow-based method. Interestingly, not all the CRC cell lines tested give rise to spheroids of the same shape and size, under the same experimental conditions, starting from the same number of cells seeded. Indeed, HT-29 cell spheroids appeared like loose cell aggregates where single cells can be still distinguishable. In turn, SW620 and DLD-1 spheroids displayed a smooth surface and oval (SW620) or round (DLD-1 and HCT-15) shape. This might represent a limitation, since physical measures and functional assays were performed on heterogeneous spheroids populations; nevertheless, the standardized culture conditions used allow the formation of spheroids mostly included in the range of 60 to 140 µm for all the CRC cell lines. Of note, these different characteristics can be captured, measured and evaluated by the flow-based system that can evidence distinct weight, diameter and mass density, according to the cell composition of the  tumor-spheres. Although fixation, that allows sample recruitment in multicenter studies, may modify weight and diameter of spheroids in some instances, both parameters vary in the same direction so that mass density is not influenced. These findings disclose that mass density is a more reliable parameter than diameter and weight. Of note, the higher mass density detected in DLD-1 compared with HCT-15, both derived from the same patient but displaying a different karyotype (35), might indicate that the heterogeneity of clones inside the same tumor can be distinguished also on the basis of mass density measurement. In addition, the differences in spheroid mass density can be referred to the spheroid cell number. Indeed, computerized imaging, showed in SW620 spheroids a significantly higher cell number/mm 2 than in HT-29, DLD-1, and HCT-15 spheroid: the latter two, in turn, contain a higher cell number compared to HT-29. Likewise, intracellular ATP content was not only detectable in all spheroids, as a parameter of cell viability and metabolism, but also related to cell number and mass density, determined with the flow-based system. We could also measure the results of effector lymphocyte activity on CRC spheroids. Although controversial for many years, the role of NK cells in anti-tumor immune response is documented (31,32,40,41). In CRC, the contemporary infiltration of T and NK cells is apparently linked to a better prognosis and an anti-cancer response mediated by NK cell activation has been reported in a metastatic CRC patient (42,43). In the present flow-based system, co-culture with NK cells led to a decrease in weight and diameter of CRC spheroids, with different kinetics depending on the cell line. Indeed, SW620 and DLD-1 changes were already evident at 6 h, much earlier than cytotoxicity detection with the crystal violet assay, whereas changes in HCT-15 spheroids were evident only after 24h of coincubation of NK cells. This might depend, at least in part, on the size, being small spheroids more susceptible to NK cell infiltration than large ones. However, the CRC spheroid populations exposed to NK cells displayed similar size variations (diameter of 80-130 µm for SW620, 90-130 µm for DLD-1 and 70-140 µm for HCT-15). Furthermore, the tendency for an increase in mass density observed at first in SW620 and later in DLD-1 spheroids, upon co-incubation with NK cells, conceivably reflects the different kinetics and degree of spheroid infiltration by NK cells. As demonstrated by confocal analysis and computerized imaging, DLD-1 spheroids are invaded by NK cells that reach in great number the inner layers and the center of the tumor by 24 h; conversely, NK lymphocytes remain in the periphery of SW620 spheroids and a few cells reach the center of the tumor mass. This behavior in invasion might explain the differences in mass density detected with the flow-based system in DLD-1 and SW620 spheroids during interaction with NK cells and evidences a different mode of action of NK cells, depending, at least in part, on the physical properties of the tumor. Also in the case of SW620 spheroids, however, the antitumor cytotoxic activity is operating, as revealed by the reduction in weight and diameter. In addition, differences in the expression of molecules involved in tumor cell killing could contribute to cancer cell sensitivity. We reported that all these CRC cell lines expressed adhesion molecules, such as ICAM1, and the ligands of NKG2D and DNAM-1 NK cell-activating receptors (35). The expression of these ligands, and the consequent reactivity of the Fc-NKG2D and Fc-DNAM1 chimeras, is maintained in CRC cell lines involved in tumor spheroid formation; this would indicate that NK cells can use most of their classic receptor-ligand systems to invade CRC spheroids and exert their anti-tumor function. On the other hand, the expression of HLA-I may be relevant to favour NK cell spheroid invasion and killing: indeed at variance with SW620, DLD-1 does not express HLA-I, that can deliver inhibitory signals to NK cells [reviewed in (33)], thus reducing their infiltrating and killing potential. Nevertheless, cancer cell shape and tumor architectural organization seem to be crucial for the outcome of NK cell antitumor activity; indeed the degree of cytolytic activity exerted by NK cells varies using the same tumor targets as single-cell suspensions, as adherent cells or as spheroids.
Since the 3D spheroid system is evocative of the small tumor cell clusters that may occur in the first cancer stages, a precise measurement of weight, size and density variations provide substantial information on disease progression. This is particularly relevant in CRC, where a reliable animal model is still to be defined. Moreover, with this innovative flow-based system we can measure the size and physical properties of a large number of spheroids in several replicates and different experimental conditions; this experimental setting can be useful to test new drugs or therapeutic schemes for their antitumor efficiency. Not only CRC cell lines but also primary tumor cells obtained from bioptic specimens can be used in this system, allowing the assembling of personalized precision medicine.
In conclusion, 3D spheroid models represent a reliable, reproducible and cost-effective solution that allows the evaluation and measurement of the first steps of cancer growth, taking into account the heterogeneity of tumor cells. This experimental setting also allows the evaluation of the degree and kinetics of antitumor effects exerted by immunocompetent cells; also, the system reveals any difference in the sensitivity of CRC cell types to lymphocyte effects. Potentially, the fine-tuning of the physical parameter recording could be useful in the evaluation of anti-cancer drug efficacy, including that of therapeutic antibodies and immunotherapy.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Peripheral blood mononuclear cells (PBMC) were obtained from healthy adult donor's buffy coat upon institutional informed consent signed at the time of donation and EC approval PR163-REG2014 of the Ligurian Regional Ethics Committee. This Committee is placed at the IRCCS Ospedale Policlinico San Martino, 16132, Genoa, Italy. The patients/participants provided their written informed consent to participate in this study.