Optimization of a Clinically Relevant Chemical-Mechanical Tissue Dissociation Workflow for Single-Cell Analysis

Abstract

Introduction

While single-cell analysis technology has flourished, obtaining single cells from complex tissues continues to be a challenge. Current methods require multiple steps and several hours of processing. This study investigates chemical and mechanical methods for clinically relevant preparation of single-cell suspension from frozen biopsy cores of complex tissues. The developed protocol can be completed in 15 min.

Methods

Frozen bovine liver biopsy cores were normalized by weight, dimension, and calculated cellular composition. Various chemical reagents were tested for their capability to dissociate the tissue via confocal microscopy, hemocytometry and quantitative flow cytometry. Images were processed using ImageJ. Quantitative flow cytometry with gating analysis was also used for the analysis of dissociation. Physical modeling simulations were conducted in COMSOL Multiphysics.

Results

A rapid method for tissue dissociation was developed for single-cell analysis techniques. The results of this study show that a combination of 1% type-1 collagenase and pronase or hyaluronidase in 100 U/µL HBSS solution is the most effective at dissociating 2.5 mm thawed bovine liver biopsy cores in 15 min, with dissociation efficiency of 37-42% and viability >90% as verified using live MDA-MB-231 cancer cells. Cellular dissociation is significantly improved by adding a controlled mechanical force during the chemical process, to dissociate 93 ± 8% of the entire tissue into single cells.

Conclusions

Understanding cellular dissociation in ex vivo tissues is essential to the development of clinically relevant dissociation workflows. Controlled mechanical force in combination with chemical treatment produces high quality tissue dissociation. This research is relevant to the understanding and assessment of tissue dissociation and the establishment of an automated preparatory workflow for single cell diagnostics.

Supplementary Information

The online version of this article (10.1007/s12195-021-00667-y) contains supplementary material, which is available to authorized users.

Introduction

Tissue biopsy cores are often taken for the diagnosis and confirmation of many different types of cancer.7,24 While liquid biopsy and in vivo optical techniques have also been developed for detection, they have some limitations, meaning that follow-up confirmation by tissue biopsy is still the standard of care when it is physiologically possible.10,19 Molecular characterization in tissue biopsies has historically been achieved by sequencing the heterogeneous profile of the tissue in an attempt to screen for specific markers, which has previously limited the technique.23 Ensemble measurement has numerous disadvantages mainly relating to high levels of noise produced by variability in the sample.

It can be challenging to reliably determine the presence and proportion of various markers as diseased cells are present at much lower concentrations than healthy cells, which drown out the signal.32 Additionally, even within the population of cancerous cells, rare markers can be challenging to detect in heterogeneous sequencing approaches for this same reason. This poses a significant issue in determining the treatment and prognosis of cancer patients, as it is these rare cancer cell mutations that commonly drive metastasis.29 Therefore, isolating individual cells for cancer diagnostics is becoming an area of increasing importance for overcoming this issue.

Because of this, a Next-Generation Sequencing (NGS) approach known as Single Cell Sequencing (SCS) has been developed in order to determine the genetic profiles of individual cells. SCS can be used to provide the genetic (DNA) or expression (RNA) profiles of individual cells by analyzing them one by one.27 Through this technique, entire populations of cancerous cells from a tissue biopsy can be examined and categorized. While SCS has solved many of the problems of heterogeneous sequencing, getting the starting material required for SCS - perfectly dissociated single cells - is a challenge in and of itself.1,2,8

Dissociating complex tissues into single cells while maintaining cellular integrity and stable genetic expression is still a topic that is poorly understood. Multiple benchtop, bioreactor, and other preparation procedures have been developed and discussed elsewhere.4,22 However, many of these approaches do not have the efficiency that is needed to be clinically relevant for samples of heterogeneous tissue. As a result, tissue is rarely dissolved fully by treatments that are purely chemical in nature.9

Selectively dissociating the outer layers of a biopsy core is not sufficient for understanding the complexity of an entire tissue sample. For this reason, in clinical environments, researchers are becoming increasingly reliant on instruments such as the GentleMACS dissociator Tumor Dissociation kit, which combines mechanical agitation with chemical dissociation (Miltenyi Biotec).

However, this standard technique has several drawbacks, such as high cost ($567 for only 25 samples), and a significant reduction in cellular viability.18 Cell death can result from disruption of membrane integrity during mechanical processing. Futhermore, many tissue biopsy cores are frozen at − 20 °C at the very least, potentially damaging cellular integrity and making full tissue dissociation a more challenging process.3

Previous literature on the subject of chemical dissociation of tissue is mainly focused on in vitro models, which are known to be less complex and easier to dissociate fully than ex vivo tissues, limiting their clinical relevance.6,9 Furthermore, studies that do address complex tissues rarely examine dissociation from the perspective of obtaining single target cells from a diverse profile of chemical reagents, illustrating the need for such research.

Additionally, few studies analyze chemical/mechanical processing in conjunction. Those available do not incorporate a discussion of mechanical forces or physical modeling approaches to improve dissociation. However, it is critical to understand the biomechanics of mechanically integrated dissociation protocols in order to increase dissociation while preserving cellular integrity.

In this work, we investigate the effects of chemical and mechanical factors on the dissociation of complex tissues, using bovine liver tissue biopsy cores as a model. We contextualize the results with an explanation of the chemical and mechanical rationale for applications to clinically relevant cancer diagnostics.

Finally, we discuss the synthesis of these results and their implications for developing a sample processing workflow for single-cell marker and sequencing analysis in cancer diagnosis and beyond. This workflow is a competitive and simple preparatory method that can be easily translated to most laboratories. It has the potential to reduce costs, decrease dissociation time, and enable high-throughput, regional analysis of single cells by Single-Cell Sequencing (SCS).

Methods

Liver tissue biopsy

Bovine liver tissue was purchased from a local grocery store (Stop and Shop, Providence, RI, US). From there, the bulk tissue was frozen in a − 20 °C freezer. A day before specimens were to be analyzed; the tissue was removed and biopsied by using mechanical coring instrument of 2.5 mm diameter and 2 mm length (World Precision Instruments, Limited Reuse Biopsy Punch) (Fig. 1).

Figure 1.

Overview of the protocol used in the experiment consisting of dissociation and analysis steps. The tissue is first biopsied from the organ, placed into a 1.5 mL Eppendorf tube, confirmed to be standardized for weight and dimensions, and then frozen for preservation. Subsequently, the tissue can be thawed and chemically dissociated to produce individual cells. The dissociation efficacy into individual cells was measured by two separate methods in this experiment: hemocytometry and flow cytometry. Hemocytometry tests require pipetting a sample from the pre-diluted tube, then adding trypan blue and inhibitor to an established point.

The biopsied tissue cores were then added to pre-weighed 1.5 mL DNA Lo-Bind Eppendorf tubes (Eppendorf). These tubes were placed on a balance, and the mass was recorded to an uncertainty of 0.00005 g (Sartorious). Afterwards, the tubes containing the tissue samples were again placed in the − 20 °C freezer. Then, 24 h before the dissociation tests, the tube containing the tissue was placed in a 4 C refrigerator to be thawed while kept cool enough to prevent tissue deterioration.

Tissue cores were then placed on test tube racks and allowed to come to room temperature for an hour. They were then re-weighed and then used in chemical/mechanical dissociation trials. For mechanical tests, the same workflow was used but followed the transfer of the biopsy core within a flat bottom 96 well plate (ThermoFisher Scientific).

The weights of the tissue biopsy cores had a narrow range of around 10 mg. Statistical tests were performed using MATLAB software to ensure standardization of both weight and dimension for all samples included in the analysis (MathWorks). The data for tissue weight was normally distributed, and any outlying samples were discarded. The weight differences of individual tubes and tissue samples were taken into account, with the amount of added reagent adjusted depending on sample mass, as recommended by ThermoFisher Scientific (Fig. 1).

Culture of MDA-MB-231 Cells

Highly metastatic “triple negative” breast adenocarcinoma cells (MDA-MB-231) were subcultured for use in order to assess viability of live cancer cells exposed to the dissociation protocol and thereby determine its utility in cancer cell retrieval and culture, a routine procedure in cancer tissue biopsy. The MDA-MB-231 cells were cultured in media consisting of Corning DMEM with l-glutamine, 4.5 g/L glucose, and sodium pyruvate supplemented with 10% fetal bovine serum (GE Healthcare) and 1% pen/strep.

Partial passage was used to elute three-dimensional spheroidal cellular clumps, which were tested in the dissociation workflow for the purpose of examining viability of live cells only. These cells were not used to assess ex vivo dissociation efficacy due to their limited complexity in comparison to ex vivo tissues.

Chemical Dissociation Protocol

Chemical Repertoire

The chemical dissociation protocol was tested on biopsy cores of ex vivo bovine liver tissue in order to assess the ability of chemical reagents to dissolve complex tissues and thus discern a chemical cocktail to be used in the workflow. The chemicals that were tested included: trypsin-EDTA at a concentration of 0.5%, Liver Digest collagenase-dispase media, collagenase 1%, collagenase 1% and hyaluronidase 1% (Sigma-Aldrich), collagenase 1% and pronase 1% (Sigma-Aldrich), TryPLE Express reagent, 0.5 M EDTA 0.8%. All reagents were from ThermoFisher Scientific unless noted otherwise.

Protease reagents were prepared by performing a basic units per volume calculation using the activity of the given protease in U/mg, the desired total solution volume, and the desired concentration of the final solution: Concentration = Activity x (Mass / Volume). This returned a weight of dry protease which was then measured out precisely using the aforementioned scale and added to solution, which was then vortexed until thoroughly mixed.

Reagent concentration and the combination of different reagents was determined using the Worthington Tissue Dissociation references and several historically established protocols for liver tissue dissociation as guidelines.33 The dissociation media was applied in a 1 mg per 10 µL ratio, as recommended by ThermoFisher Scientific. Freshly prepared reagents were used to prevent variability in results caused by degradation of enzymes frozen in storage over long periods of time. All dissociation trials were performed at optimum temperatures for respective chemical activity (25 or 37 °C) to control the effect of temperature.

General Dissociation Protocol

At time intervals of 0, 5, 10, and 15 min, 5 µL of the solution was pipetted from the center of each tube twice to create replicates which could later be cross compared. Thirty seconds before these intervals, the suspension was very gently stirred with a pipette for even dispersion of cells. This enabled the quantification of cells suspended into the dissociative reagent over time. From there, a 1:10 dilution factor dilution was performed, in which the 5 µL of the solution was pipetted into 45 µL of an inactivating agent, such as EDTA or media containing FBS (ThermoFisher Scientific) for the respective enzyme. The dilution was then gently pipetted 3x to mix. This resulted in the final samples for flow cytometry analysis.

For the hemocytometry and microscopy trials, an additional step was performed after this. 5 µL of the aforementioned dilution was pipetted into a tube containing 5 µL of an inactivating agent (either media containing FBS, EDTA, or PBS as a control if no inactivation was required) and 10 µL of trypan blue stain (ThermoFisher Scientific). This solution was gently pipetted 3x to mix and imaged on a reusable hemocytometer (Hausser Scientific).

Second-Round Dissociation

Certain trials involved reagents that were incompatible. Consequently, a two-step sequential protocol was used to test the combination. For example, trypsin and collagenase were not capable of digesting tissue simultaneously as one is inhibited by and another activated by the presence of calcium ions. At the end of these chemical trials, the remaining tissue in the tube was collected, quantified, and subjected to further dissociation by the following process: The tube was placed in a centrifuge for 5 min at 1500 RPM in order to pellet the suspended cells out of solution. The remaining supernatant was then removed by pipette. The tissue, collected at the bottom of the tube, was then weighed, and the weight recorded.

Subsequently, the tissue was subjected to another round of chemical dissociation with the second reagent, again using the 1 mg per 10 µL ratio of tissue to chemical dissociation media recommended by the supplier (ThermoFisher). This second round of dissociation was recorded in 5-minute intervals from time points 20 min to 45 min (Table 1).

Table 1.

A summary table of different reagents involved in the study, as well as their mechanism, use, and interactions that are relevant to the understanding of the problem of vivo tissue dissociation.ex.

Cell Dissociation Reagent	Mechanism of Action	Organs	Interactions	Refs
Collagenase	Attacks X-Gly bond in collagenase helix	Variable; most common type used for epithelial, liver, lung, fat, and adrenal tissue cell	Requires calcium ions for activity. Presence of EDTA inhibits proteolysis.	20,33
Trypsin	Attacks Lys-Arg bonds	General purpose, primarily for cell culture cell dissociation	EDTA present in stock solution to chelate metallic ions.	12,17,33
TrypLE	Attacks Lys-Arg bond, similar to trypsin	Marketed as trypsin substitute.	EDTA present in stock solution to chelate metallic ions.	26,33
Hyaluronidase	Bonds between 2-acetoamido-2-deoxy-beta-D-glucose and D-glucuronate, as seen in hyaluronic acid.	Connective tissues, such as liver and kidney. Almost always used in conjunction with a protease like collagenase.	Often combined with Collagenase.	25,33
Dispase	Non-polar amino acid bonds, nonspecific mix.	Primarily used to dissociate neural tissue. Is a mixture of proteases useful for this application.	Often combined with Collagenase.	13,33
EDTA	Chelation agent; removes Ca++ and Mg++ on cell surfaces to lower cell adhesion due to cadherins and integrins.	Not widely used for cell dissociation alone. Certain publications recommend EDTA solutions over enzymatic processes for flow cytometry.	Often found with trypsin/trypLE - chelation action allows for better penetration of protease enzymes to dissociate cells.	33

Mechanical Dissociation Protocol

The mechanical dissociation protocol was completed using a simple heated orbital plate shaker to induce agitation (New Brunswick Scientific, Innova 4080). All of the parameters and key variables were kept constant between the chemical and mechanical protocols to limit variability, including temperature, tissue dimensions, reagent properties, and size scales.

The workflow was tested in both tubes and 96-well plates initially to observe dissociation efficacy. Improved preliminary results were observed in 96-well plates, which enable multiplexing of samples. The mechanical trials were thus conducted using a flat bottom conventional 96-well plate.

Another parameter that was optimized was rates per minute. Experimental results and COMSOL modeling was used to set the selected RPM used in the experiment from the possible RPM allowed by the instrument (25-500 RPM).

Flow Cytometry

A FACS AriaIIIu Flow Cytometer and Cell Sorter (Brown University, Flow Cytometry and Sorting Facility) was used to show the percentage of cell aggregates vs. single cells, and to exclude non-cellular debris. Flow cytometry graphs were gated using size gating with flow cytometry beads of 6 µm and 32-38 µm diameter to create lower and upper limits for cell sizes (ThermoFisher; Microspheres). A target region was also established for hepatocyte cells based on pre-established settings in the flow cytometer. Hoechst 33258 dye was used as an additional confirmation method to discern cells from non-cellular materials and detritus. Dyes were applied following concentration schema specified by ThermoFisher Scientific at 2 µg/mL.

Liver Tissue Modeling

Two established models for estimating the number of cells within liver tissue samples were cross-referenced and compared to data from fully dissociated tissue sections in order to create a numerical methodology to probe dissociation efficacy. The first model was derived from histological liver tissue sections and was calculated using dimensions of a tissue section.11 The second model calculated results from tissue weight.34

Both studies were applied to create a new model for expected total numbers of cells, as well as numbers of specific cell types within the tissue. The new model was compared to 10 fully dissociated tissue sections to the bovine liver tissue model and found to have effective predictive capacity for dissociation.

The retrieval of target cells can be predicted using the established composition of four distinct cell types within the liver (Fig. 2). The established liver cellular composition was used as a basis for estimating percent retrieval of hepatocytic target cells within the study.30

Figure 2.

Overview of the underlying basis of the liver tissue model used to determine cellular dissociation. The model relies on the characteristics of the dimensions and weight of biopsy cores to determine the approximate total number of cells. Furthermore, the known relative cellular composition of liver tissue can be used to calculate the expected number and percent retrieval of an individual cell type, such as hepatocytes as done in this study.

Quantification of Cells by Confocal Microscopy

Hemocytometry: Trypan Blue Stain

Trypan blue staining was used for the purpose of counting cells from images taken in light microscopy. Images were captured on the Nikon Eclipse TE2000-U microscope using Pylon Viewer Software (Basler Web). In the liver tissue samples from an excised organ, all cells were expected to be stained with trypan blue, as they would all be dead.

Hemocytometry is a common cell counting technique in which a chamber glass slide is used for the counting of trypan blue stained cells.21 Slides were prepared using a hemocytometer and loaded with 10 µL of solution which consisted of 5 µL sample of dissociated cells from the 10x inactivating dilution, 5 µL inactivating agent and 10 µL trypan blue. In trypsin-EDTA solutions, the inactivating agent that was used was media containing FBS. For collagenase solutions, EDTA was used as an inactivator. In the other samples that did not require inactivation or are inactivated by dilution (ex. TrypLE), PBS was used instead.

Images were taken on the microscope and then analyzed using the National Institutes of Health FIJI ImageJ analysis software, as described below. After cells were counted, they were quantified using existing equations, in which the dilution factor was calculated to be 4 (Supplementary Information). Disposable slides were used for the Countess automated hemocytometer, while reusable slides for the manual hemocytometer were cleaned after each sample using 70% ethanol in distilled water.

Viability Assay: DRAQ7 & 33258 Hoechst Stain

An Olympus FV3000 confocal microscope (Brown University Leduc Bioimaging Facility) was used to assess viability and membrane integrity of Triple Negative Breast Cancer MDA-MB-231 after being treated with the preferred dissociation protocol. Hoechst 33258 is a supravital cell-permeable bisbenzimidazole dye that can bind to both live and dead or fixed cells.15 DRAQ7, an anthracycline derivative, is useful for long-term cell death monitoring in real time by entering damaged cells and binding nuclear DNA.35 DRAQ7 does not penetrate the plasma membrane or influence susceptibility to death – it only enters when the membrane integrity of the cell has been compromised, making it an ideal marker for membrane disruption and cell death produced by the tissue dissociation protocol.35 These two dyes are co-stained for fluorescent microscopy and flow cytometry “live/dead” analyses.

Image Analysis Platform

The ImageJ - FIJI image analysis software was used for the purposes of cell counting from confocal microscopy images (National Institutes of Health). A workflow was developed for visual processing. The FIJI workflow consisted of the following. A size gate was set, similarly to in flow cytometry, by using the known size of lines on the hemocytometer to convert pixels to µms.

Known cell size information was used to discern red blood cells and other debris from single cells and cell clumps. Other particles that may have shown up in the images, such as dust or cellular fragments were excluded based on size, shape, and circularity using the processing workflow. Visual debris was removed from the image before executing cell counting. From that point, a threshold was set, highlighting all particles of interest. With the analyze particles function, cells can be quantified in terms of number, distribution, and size.

In order to ensure the relevance of this workflow in discerning single cells, it was tested using freshly passaged, fully dissociated MDA-MB-231 cells and fresh liver tissue as a positive control.

This same imaging analysis workflow was used both for Hemocytometry image processing and viability microscopy images. The viability images required a simple additional step of discerning between different fluorophores by setting fluorescence thresholds before counting.

COMSOL Multiphysics Computational Fluid Dynamics Simulation

COMSOL Multiphysics Software was used in order to predict optimum mixing parameters using the orbital plate shaker. A 96-well-plate replicate with individual wells was created in Fusion360 Software. The well was filled with 200 µL of fluid with assumed water-like properties. An orbital mixing force of varying RPM was applied about the well and fluid velocity mapped. Comprehensive processing parameters and equations are detailed in the Supplementary Information.

Results

The experimental study consisted of dissociation of ex vivo bovine liver tissue biopsy using various methods, and validation of the efficacy of dissociation using a few available techniques. Hemocytometry is considered to be the simplest technique for cell counting, as it can be done with relatively simple instrumentation in many labs. Although hemocytometers are most often applied to the counting of red and white blood cells, they have been used across a variety of other applications, including cancer cell study.5,21 Two different hemocytometry systems were compared: The Countess Automated Hemocytometer and a manual hemocytometer with ImageJ image processing software.

Flow cytometry size gating analysis was also used to quantify single cells and cell clumps in this way. Flow cytometry is known to be a more sophisticated technique for cell counting and was found to be the most reproducible technique for analysis. Additionally, an optimum dissociation media was determined that performed well across all trials. From there, the effect of added mechanical agitation on the time required for dissociation and dissociation efficacy was examined, using a heated plate shaker.

Lastly, viability studies were conducted that used live cells from the MDA-MB-231 Triple Negative Breast Cancer cell line in order to better understand the effects of the protocol on membrane integrity and other factors that may affect cell death and produce altered expression. The DRAQ7/Hoechst 33258 live dead stain enabled examination of the effects of the optimum chemical/mechanical protocol on cellular viability.

Manual vs. Automated Hemocytometry

Commercially available automated hemocytometers are considered practical and more technically feasible technologies for clinical in-lab cell enumeration when compared to traditional hemocytometry workflows. This is because they do not require an added step of image processing and can give a much quicker time-to-result. The Countess automated hemocytometer eliminates the need for labor spent on manual data processing and inefficiencies thus associated. The instrument enables the researcher to quantify cells on site, without effort. Unfortunately, we have found that reproducibility with this technique was significantly lower when compared to the manual hemocytometry workflow (Fig. S1). This contextualizes why the technique is used frequently for clinical cell culture research but is less frequently associated with the pipeline of cancer diagnostics, which often uses flow cytometry instead.31

In manual hemocytometry, ImageJ returned a cell count, which was used to calculate the total number of cells using the hemocytometry equations (Supplementary Information). This was a more time intensive process but was still effective with a total time to result of under 10 min. Within the quantitative hemocytometry workflow, image analysis software was successful at setting up a size gate, similarly to that established in flow cytometry. The ImageJ analysis approach was consistent with low technical error and high reproducibility of results and exceeded the performance of the automated hemocytometer by ~20%, with a p value < 0.05. This method was favored over automated hemocytometry in ex vivo tissue dissociation for this study and cancer diagnostics at large because of its improved capability to cope with variable cell counts and detritus (non-cellular material such as connective tissue) on the slide.

Hemocytometry images have been used in many previous research studies to quantify the efficacy of tissue dissociation or cell retrieval.5,21 However, despite the increase in technical reproducibility that was found with the ImageJ processing approach, this method remains limited in practical utility. It is best used for getting a rough idea of the efficacy of tissue dissociation and producing quick, qualitative images. From the hemocytometry visualization and ImageJ analysis, clear differences can be observed between the control non-dissociative reagent, EDTA, when compared to actively dissociating reagents, such as 0.5% trypsin (Fig. 3).

Figure 3.

Hemocytometry images show qualitative dissociation patterns. 10x magnification. Images on the left of each set (a, c, e, g) are the direct images; images on the right (b, d, f, h) are intermediates in the ImageJ analysis process that contrast cells and blank media. Images A and B represent a sample incubated in 0.5% trypsin at 2 min; (c) and (d) Represent the same conditions but at 15 min. EF and GH sets represent similar processes for a control volume of EDTA, at 2 and 15 minute timepoints. Image scale: large gridlines = 200 µm, small gridlines = 50 µm.

When analyzing the images quantitatively, large standard deviations were found within replicates of the same chemical treatment and time, making robust quantitative analysis challenging. It is possible that this could also have been the result of differences in the dispersion of cells throughout the sample due to poor suspension. When pipetting a small volume out for analysis, the location of the pipette tip (which was standardized to the bottom near the wall) and dispersal of the sample become variables that continuously introduce error, irrespective of how controlled the technique is. This was eliminated by flow cytometry, which was able to analyze the entire dilution sample.

Although processing with ImageJ was shown to have a greater reproducibility in results, there remained high observable variability between different regions of the hemocytometry slide. This, in addition to the error introduced by cell dispersal and pipetting within the sample tube, made it impractical to use the hemocytometry workflow with ImageJ processing for quantification of cell dissociation.

This high intra-slide variability was due to the heterogeneity of the fluid and the capillary force, which pushes cells to the outer corners of the slide. This results in an uneven distribution of cells and ineffective quantitative estimates of cell counts. Making the cell containing solution monodisperse and loading it evenly onto the hemocytometer continues to present a problem. This has been addressed from the perspective of incorporating nanointerstice flow in a similar device to the Countess and other standard hemocytometry slides, as described by Kim et al.14 Nantointerstic flow incorporation creates a strong driving force with a sub-hydrophobic contact angle. Similar effects can be created by using hard plastics (polymethylmethacrylate, polystyrene, polycarbonate).

The capillarity is the dimensionless driving force without the nanointerstices expressed in terms of the dynamic contact angle of the surface, $\epsilon$ is the aspect ratio of the channel (h x w-1). The capillarity is an innate component of the overall nondimensional force which pushes the liquid-air interface towards the end of the slide during the process of loading. Capillarity and nanointerstice flow are responsible for the disproportionate movement of cells in the solution towards the edges of the slide.

Systems that use this loading technique consistently report lower counts due to this phenomenon. When nanointerstices are added, as is common for many hemocytometry slides, the surface energy is increased by a given geometric term which represents pressure differential at the meniscus and the NIs. While helping to encourage effective loading, the use of NIs on hemocytometer surfaces creates a driving force which propels fluid to the edge of the slide. Furthermore, high driving forces unevenly distribute the particles within the liquid.

This effect on cell loading was confirmed by looking at recorded video data of cell motion across the hemocytometry slides as they are filled in real time. Because the capillarity resulted in an observed distortion of results and inaccurate quantification, flow cytometry was used to analyze and quantify cells, while hemocytometry was used to qualify dissociation and visualize cells. Additionally, As the hemocytometer’s total surface area exceeds what can be captured by a microscope, it is also possible that a sizeable number of cells are lost from the data. Another possible solution could be to increase the interrogation area to the entire slide, but this is likely to introduce additional problems.

Flow Cytometry

Using the size gate and target region, as well as side scatter information, it was possible to bin particles within the sample into three different overall categories: non-cellular debris and red blood cells, single cells derived from the tissue, and large cellular aggregates. The size-controlled particles were used to sort by size, shape and complexity. From this, information on total cell count, total number of single cells, number of target cells, and percent dissociation can be determined. It was also possible to discern the total number of hepatocytes using the hepatocyte gate, which selected for single isolated hepatocyte cells.

The flow cytometry data illustrated similar trends that were observed using hemocytometry across various time points and chemical conditions in a more quantitative and reliable manner due to the elimination of slide dispersion variability. There was a high level of consistency between manually set target particle gates for single cells from the tissue and what were inferred to be single cells by the bead-based size gating process. This shows that using known gates for certain cell types or FSC/SCS based gating with beads works well for retrieving targets of interest, even with an incredibly complex sample which contains detritus from the dissociated tissue. Furthermore, it’s possible to cross reference data from multiple approaches in a single run.

Flow cytometry graphs revealed that, as one might expect from the dissociation of complex tissues, large quantities of non-cellular debris are present within the sample. By using size gating, established gating settings for hepatocytes, and fluorescent nuclei staining, it was possible to purify the solution in order to focus on the target cells of interest. This enabled quantitative analysis of the dissociated tissues (Fig. 4).

Figure 4.

Flow cytometry establishes target gates and quantifies dissociation across chemical treatment and time. (a) shows the particles detected after 0 min of tissue in 0.5% trypsin; (b) shows the sample at 15 min.(c) shows 1% collagenase/pronase at 0 min. (d) is the same treatment at 15 min. (e) shows 0 min in EDTA. (f) shows 15 min in EDTA. (g) was a suspension prepared with microparticles to verify the accuracy and applicability of Flow cytometry sorting to our purposes.

Extracted Data Analysis

The data obtained from the individual flow cytometry samples was further scrutinized using a meta-analysis technique for several samples in order to infer differences resulting from different chemical treatments. The optimum effects were observed after a 15-minute endpoint by measuring at points throughout the time course. From this analysis, 1% collagenase and pronase as well as 1% collagenase and hyaluronidase can be seen to be the most successful media for dissociating the tissues, as evidenced by their higher dissociated cell counts (Figs. 5, 6) and percent dissociation (Fig. 6).

Figure 5.

Efficacy of different chemical dissociation protocols as evidenced by the unadjusted total flow cytometry counts at the final, 15 minute timepoint. The total counts of cells the flow cytometer evidence the comparative efficacy of each dissociation protocol for cellular dissociation of biopsy cores. Not all cells or particles are considered to be dissociated single cells within the target range, but the size gating system is effective at discerning this. N=10. P values: ** = 0.05, *** = 0.01.

Figure 6.

Efficacy of different chemical dissociation protocols at 5 minute intervals over a 15 minute time course. The analysis examines: (a) total cell counts, (b) single cell counts, (c) and dissociation efficacy (percent dissociation). The metrics were obtained by using established gating for target cells as well as size gating with particles of controlled size to establish upper and lower limits for single cells. Furthermore, percent dissociation was determined by using the calculated expected total number of cells from the tissue model. All A/B group comparisons were significant at a P value of 0.05 with the exception of 10X TrypLE and collagenase, significant at 0.1. N=10.

It was also easy to observe that all of the collagenase-based dissociation media is significantly more effective in the dissociation of bovine liver tissue. This is contrary to the results produced by other studies, which use in vivo tissues as a model, and often find that trypsin-EDTA based reagents, including TrypLE, are the most effective for dissociating these tissues. 9

The effects of the various chemical reagents were later analyzed over 5-minute intervals from 0-15 min in order to gain an understanding of the kinetic profile for the dissociation process and inter-chemical differences. This multi-part analysis examined total particle count, cell count, number of single cells, and number of target cells (Figs. 6, 7). Throughout the analysis, collagenase-based reagents continued to outperform trypsin-based reagents, including the highly concentrated 10X TrypLE reagent.

Figure 7.

Added mechanical force increases tissue dissociation over time. The mechanical force that was used was the optimum orbital plate shaking condition, as predicted by theoretical calculations and physical modeling. The results are statistically significant with a P value = 0.05 at 5 min and 0.01 at 10 and 15 min. N=10

Percent dissociation of the tissue to single cells was also examined, and all of the treatments were relatively effective in single cell dissociation when compared to the control, EDTA. While the control dissociation of single cells when compared to cell aggregates was around 40%, it was found that the addition of active dissociative agents helped to increase this to above 80% in most cases. Interestingly, trypsin-based reagents were found to be slightly more successful for dissociating cell clumps into single cells when compared to collagenase-based reagents in these results. This suggests that, while collagenase is significantly more effective for breaking up the biopsy core into suspension, later treatment with trypsin-based reagent could help to further dissociate individual cells.

The percent dissociation of tissue to cells was determined to reach a maximum of 42% at 15 min during this time course, in the 1% collagenase and hyaluronidase treatment, with 1% collagenase and pronase close by at 38% (Fig. 6). These results confirmed visual observations that much of the tissue remained undissociated at the end of purely chemical trials. Although this purely chemical dissociation was relatively low, it is significantly lower for trypsin-based reagents. Furthermore, when complexed with mechanical force from a heated plate shaker, the percent tissue dissociation was significantly improved (Fig. 7).

Automation and High Throughput

In order to streamline multiple experimental workflows to increase throughput and automation, 96 well plates are used instead of tubes. 96 well plates are easy to use and relevant to clinical and industrial laboratories, as they can be complexed easily with automated pipetting systems (PerkinElmer JANUS Liquid Handler) and are standard in most high throughput systems. Thus, after placing the tissue biopsy cores into individual wells and thawing, chemical dissociation and analysis can all be automated on the plate with liquid handling systems, creating an entirely “hands free” dissociation and processing workflow.

This mechanism makes it easy to perform numerous follow up experiments on the dissociated cells. Flow cytometry and microscopy experiments can follow for validation. SCS and other single cell analyses can be complexed with the workflow, expediting the time to result.

Another advantage of JANUS and other similar liquid handling systems in tissue dissociation protocols is the presence of heated plate shaker systems. Previous research addressing the interaction of chemical and mechanical forces in dissociation have generally not addressed the simultaneous activity of both factors, but have aimed to incorporate shaking and centrifugation steps at room temperature, which will affect the chemical kinetics of many dissociation reagents from trypsin to collagenase and beyond (Table 1).

When the best chemical treatment (1% collagenase and pronase) is complexed with mechanical plate shaking, there is a substantial increase in the total number of dissociated cells by the 15-minute timepoint (Fig. 7). At this time point, there was a dissociation efficacy of 92 ± 8%, based on the established tissue model. This is a significant improvement from a dissociation efficacy below 40% in the purely chemical dissociation trials. The two groups were compared across timepoints using a one-tailed t-test for two independent means, yielding a result that was significant at p < 0.05 for 5 min, and p < 0.01 for 10 and 15 min.

By the end of the 15-minute dissociation trial with this mechanical plate shaker, tissue samples were essentially fully dissociated and appeared visually homogenized. Visual and weight analysis revealed few if any remaining solid tissue chunks in the solution. In contrast, digestion with enzymes alone will not fully dissociate the samples, even when allowed to sit for hours or days, as further experiments showed.

Optimal mixing in a plate shaker can be defined by Equation 1, as well as other equations detailed in the Supplementary Information. When mixing fluids in an orbital plate shaker, the geometry of the wells as well as the volume of liquid within them will impact results. When volumes inside wells decrease, the liquid experiences increased surface tension due to increasing surface area/total volume. Small volumes, therefore, do not require a large orbit, but a higher mixing speed than small samples. In order to exceed the surface tension, the centrifugal force must surpass the frequency represented by Eq. (1).

$$n_{min}=\sqrt{\frac{\sigma D_W}{4\pi V_F\rho d_0}}$$ 1

In which n is the mixing frequency, σ is the surface tension, D_w is the diameter of the well, V_F is the fill volume, ρ is the fluid density and d₀ is the amplitude (Q Instruments). In the flat bottom 96 well plate, the fill volume is 340 µL, with recommended working volumes of 200-300 µL. An optimum fill volume of 200 µL was determined for best results with agitation (Supplementary Information). It was calculated that an effective mixing speed range for the plate shaking system is 200-500 RPM, depending on the amount of reagent added and the specific properties of the reagent of interest such as viscosity (SI). This was further assessed in COMSOL and supported by experimental results.

Target Cell Retrieval

The capacity of this workflow for the retrieval of individual target cells from complex tissues was examined using the flow cytometry size and complexity gate for hepatocytes. From there, the isolation of hepatocytes was compared to the established tissue composition model, in order to assess the efficacy of the protocol for the dissociation and retrieval of target cells. In supplemented collagenase tissues that were approximately fully dissociated by the 15-minute time point, the total number of isolated hepatocytes either met or even slightly exceeded the calculated number of expected target cells (Fig. 8).

Figure 8.

Effective target cell retrieval via size gating flow cytometry. A represents the numbers of total isolated hepatocytes from the treatment. B represents the percentage recovery when compared to the total number of hepatocytes theoretically expected to be recovered from the treatment, as estimated using the aforementioned tissue model. P values: *** = 0.01, ** = 0.05, * = 0.1. N=10

Overall, the synthesized modeling approach fit the experimental results. The model was able to be used to predict the approximate number of hepatocytes and determine the efficacy of tissue dissociation, as well as cell retrieval. The slightly higher experimental hepatocyte counts could show a slight divergence from the applied model in bovine tissue. Additionally, this could also represent rare instances of 2-3 cell aggregates of smaller cells that are perceived by the flow cytometer as hepatocytes, although this is unlikely due to complexity and fluorescence read outs.

Viability Assay

Although the main model for this study was ex vivo bovine liver tissue, Triple Negative Breast Cancer MDA-MB-231 cells were subjected to these dissociation conditions and then tested for viability in order to assess the applicability of the protocol in clinical cancer cell retrieval. Many cancer researchers are interested in an approach that allows isolation of single viable cells from tissue biopsies for drug screening. However, most of the techniques that are commonly used to obtain genetic material from biopsies, such as tissue homogenization, obscure the possibility for downstream single cell analysis due to disruption of cellular integrity during the homogenization process. Furthermore, when membrane integrity and cell death occur, expression profiles can change.

Microscopy analysis using the Olympus FV3000 Confocal Microscope and DRAQ7 / Hoechst staining illustrated a high level of viability of these cells after being treated with the aforementioned protocol. Images from different samples showed similar trends in viability. Specifically, there was consistently greater than 90% viability of the MDA-MB-231 cells across all images, with results of 95 ± 4% indicating that disruption of cellular integrity may not be present in cancer cells isolated from biopsy cores with this protocol (Fig. 9)

Fig. 9.

Viability and membrane integrity are preserved in MDA-MB-231 breast cancer cells exposed to the dissociation protocol. Live cells were used in order to determine the effect of the protocol on membrane integrity. (a) Confocal microscopy image of cells taken in imaging dish. (b) Image processing cell counting generated in ImageJ

Other studies investigating the effect of dissociation on cell viability have shown decreased viability using metallic screen, sheer force slide, stomacher and plunger screen dissociation with 70-80% viability without culture.16 Specialized commercial instruments for tissue dissociation into cellular suspension, such as the GentleMACS dissociator often have similar or lower viabilities when compared to this technique, with company reports of 81% viability.18

Conclusion

In conclusion, this paper examined the combination of chemical and mechanical treatments to dissociate individual cells from clinically relevant frozen biopsy core samples in order to gain a better understanding of cell-cell dissociation from a biochemical and physical standpoint.

Hemocytometry was found to be a simple qualitative method for the examination of dissociation patterns. Manual hemocytometry with image analysis proved to be a better technique for ex vivo cell isolation studies. Flow cytometry with size gating was found to be a more practical and quantitative method for the assessment of dissociation into single cells. Furthermore, flow cytometry proved to be an excellent method in terms of its target cell retrieval capability. This suggests that flow cytometry can be used both to study tissue dissociation, as done in this paper, and as a pre-processing step before single cell sequencing and analysis.

This research provides helpful insight to contextualize and shape future protocols for improved ex vivo tissue dissociation, a process that is rarely studied. Furthermore, most available protocols rely on fresh tissue, or use tissue that has been dissociated while fresh (10X Genomics, etc.), frozen tissue samples are neglected, despite making up a large proportion of the tissues that are used for sequencing today. Consequently, understanding cell-cell dissociation in frozen tissues will help to develop dissociation protocols that are reflective of conditions in clinical laboratories.

While this research represents an important step to quantifying dissociation of single cells from complex tissue samples, future work must be done in order to address other important questions. The applicability of this workflow is likely to hold for most soft tissues, which are the tissue type that is most frequently dissociated in practice. However, the success of this protocol across different tissue and cell types, especially fibrous, crosslinked, and necrotic or diseased tissues, should be investigated further. Different flow cytometry and other microfluidic sorting parameters can be tested, including cell-specific tagging. Additionally, the accurate assessment of expression profiles would be a valuable addition as RNA can be altered by freezing and chemical exposure.3,28 Although preliminary research suggests that brief plate shaking will not alter genetic expression, future research must be conducted in order to confirm this.

Supplementary Information

Below is the link to the electronic supplementary material.