Integration of Magnetic Bead-Based Cell Selection into Complex Isolations.

Magnetic bead-based analyte capture has emerged as a ubiquitous method in cell isolation, enabling the highly specific capture of target populations through simple magnetic manipulation. To date, no "one-size fits all" magnetic bead has been widely adopted leading to an overwhelming number of commercial beads. Ultimately, the ideal bead is one that not only facilitates cell isolation but also proves compatible with the widest range of downstream applications and analytic endpoints. Despite the diverse offering of sizes, coatings, and conjugation chemistries, few studies exist to benchmark the performance characteristics of different commercially available beads; importantly, these bead characteristics ultimately determine the ability of a bead to integrate into the user's assay. In this report, we evaluate bead-based cell isolation considerations, approaches, and results across a subset of commercially available magnetic beads (Dynabeads FlowComps, Dynabeads CELLection, GE Healthcare Sera-Mag SpeedBeads streptavidin-blocked magnetic particles, Dynabeads M-270s, Dynabeads M-280s) to compare and contrast both capture-specific traits (i.e., purity, capture efficacy, and contaminant isolations) and endpoint compatibility (i.e., protein localization, fluorescence imaging, and nucleic acid extraction). We identify specific advantages and contexts of use in which distinct bead products may facilitate experimental goals and integrate into downstream applications.

Introduction

Cell isolation provides a foundation for both clinical assays and basic biological research. The isolation of a subset of cells from a large, diverse population enables the enrichment of a specific population, unmasking the isolated population for continued analyses. In clinical assays—whether a tissue biopsy or blood draw—cell isolation is a critical step as patient-derived samples yield a complex mixture consisting of a broad spectrum of cell types, matrices, and biological factors. Cell isolation is required to (1) access a target population hidden within the sample and (2) assess specific (and often rare) analytes contained within target cells (i.e., RNA, DNA, and protein).[1,2] Without isolation, the noise introduced by contaminating populations impairs detection of the target-specific markers needed to inform clinical care. As assays continue to delve deeper into the interrogation of specific target populations—such as circulating fetal cells,[3,4] circulating tumor cells (CTCs),[5,6] and stem cells[7]—cell isolation processes will become essential and drive the development of commercial cell isolation products. Reflective of the ubiquitous nature of cell isolation in biologic studies, the current estimated market value (over 3.5 billion USD in 2016) is predicted to reach over 7.8 billion USD by 2021.[8]

Traditional approaches to tackle cell isolation, which purifies or extracts the intended population, have centered on filtration, centrifugation, sedimentation, and adherence. Filtration enables cell sorting based on size, largely performed by selecting or excluding populations using mesh filters of a specific pore size.[9,10] Centrifugation and sedimentation enables sorting based on cell density, often aided by density gradients to subdivide subtle density differences across populations.[11,12] Adherence relies on differential cellular interactions with specific substrates over a specified timeframe.[13] Although all are relatively simple and easy to scale, these methods are quickly limiting when cells lack significant, differential cell size, density, or adhesion, requiring new approaches to cell isolation.

Solvingthe limitations of density- and size-based cell sorting is an emerging and quickly growing field, magnetic bead cell isolation. Magnetic bead isolation has found widespread use in biological assays and applications[14−16] utilizing small (nanometer- or micrometer-sized), magnetically responsive beads to manipulate a biological target. A wide variety of magnetic beads with a diverse offering of surface chemistries are commercially available enabling easy manipulation of proteins,[17,18] nucleic acid,[19−21] and whole cells,[22−25] providing a powerful isolation tool.[26] For cell isolation, magnetic beads can be combined with a diverse offering of commercially available antibodies specific to cell surface proteins to enable the targeting of nearly any cell population.

Although magnetic beads are widely developed with well-characterized physical traits and magnetic properties,[27,28] limited literature exists directly comparing multiple bead types within the same biological context to benchmark performance (i.e., capture efficacy and nonspecific binding) and impact on common downstream endpoints (e.g., fluorescent staining of proteins to quantify localization, nucleic acid extraction, and cell culture) across bead types. Here, we evaluate five common cell isolation magnetic beads (Table S1)—Dynabeads M-270 Epoxy, Dynabeads M-280 Streptavidin, CELLection Biotin Binder, FlowComp Dynabeads, and Sera-Mag SpeedBeads streptavidin-blocked magnetic particles—to highlight the tradeoffs and considerations in integrating cell isolation magnetic beads into biologic assays. These particular beads were selected to provide a range of capabilities that may be attractive to users, such as cell release—CELLection, FlowComp; biotin-based antibody conjugation for flexibility in cell capture—M-280, CELLection, FlowComp, Sera-Mag; batch conjugation of antibody to bead—M-270s; and advertised low nonspecific binding—Sera-Mag, M-270s. On the basis of these reported favorable cell capture attributes, these commercially available magnetic beads were chosen for comparison. Beads were characterized in the context of EpCAM-specific (epithelial cell adhesion molecule) cell capture. EpCAM is a cell isolation and identification marker for epithelial cells, including CTCs.[29−33] CTCs are rare tumor cells, which are shed from a tumor lesion and enter the bloodstream. If captured, CTCs have the potential to provide insight into cancer and were thus selected as a representative rare cell population. To characterize the capture of EpCAM-positive cells, we evaluated the capture of cell lines with differential EpCAM expression, the release of those cells following capture (for FlowComp, CELLection magnetic beads), and the nonspecific capture of relevant background populations. Furthermore, we assessed the impact of the beads in integrating with standard downstream assays, including cell culture, fluorescent immunohistochemistry, and nucleic acid extraction. By evaluating a variety of magnetic bead types across a spectrum of molecular biologic assays, we aim to highlight the strengths, weaknesses, tradeoffs, and considerations when integrating beads into a cell isolation protocol.

Results and Discussion

Basic Magnetic Bead Characterization: Antibody Binding and Surface Density

Cell capture exists as a balance between the frequency of antibody–antigen interactions (which can be problematic at low antibody densities) and steric hindrance (which emerges at high antibody densities). To first visualize the antibody density on beads, the bead-bound capture antibody was fluorescently labeled. The fluorescence readout of the antibody density on the magnetic beads was generated from antibody–density curves for each bead type (Figure A). Each magnetic bead demonstrated a saturation point, at which maximal binding was observed. Upon saturation, the addition of more antibody resulted in no further increase in the signal. In translating the generated antibody density curves (FlowComp, M-280) to cell capture (Figure B), similar results are observed with poor capture at low antibody densities (too few antibody–cell interactions). Capture increased with increasing antibody density, until upon surpassing the maximal binding capacity densities identified in Figure A, a decrease in capture was observed between capture at ∼7 and 40 ng antibody per μg bead (FlowComp p = 0.027; M-270 p = 0.047, N = 3). The subtle decrease was likely a result of steric hindrance due to the high, saturated density of the antibody. Notably, at least in the case of LNCaPs (high EpCAM expresser), too low of antibody density was much more detrimental to capture than too high of antibody density. Conceivably, the relationship between the antibody density and capture is dependent on a number of factors including antibody, cell type, antibody presentation, and the size of the magnetic bead and cell. Thus, understanding the balance between these metrics remains important for optimizing the capture of a target population.

Figure 1

Characterization of bead–antibody binding. (A) Magnetic bead fluorescence intensity curves generated by varying densities of a fluorescently labeled anti-EpCAM antibody on beads. A 50% maximal binding capacity for each bead type was then identified (Table S2). Dots represent the average of three technical replicates consisting of 100 beads each (300 beads total); error bars represent the standard deviation of the analyzed population. (B) Impact of the bead’s antibody density on target capture. The 50% maximal antibody binding density for M-280s and FlowComps is denoted by their respective vertical dashed line. Points represent the average of three technical replicates; error bars represent the standard deviation; * denotes p < 0.05, which in this case applies to both M-280 and FlowComp.

On the basis of both the fluorescence titration curve of each antibody and the impact of antibody density on cell capture, an antibody density of fifty percent of maximal capacity was identified for each bead type (Table S2). Because of the different sizes of the beads used (Table S1) and the potential differences in surface roughness (not evaluated), the surface area of each bead type varies. Thus, rather than choosing a set concentration of antibody per milligram of beads (which ignores surface area discrepancies across bead types), the fifty percent maximal capacity was determined for each bead (Figure A) and used in all capture experiments unless otherwise noted. Similarly, the number of beads per milligram was different across bead types. To determine the impact the quantity of beads had on cell capture, bead quantity per sample was titrated using two different EpCAM cell lines, a high-expressing EpCAM line (LNCaPs) combined with either a medium or low EpCAM expresser (22Rv1s or Du145s, respectively) (Figure S1). Across the evaluated conditions, the maximal capture was reached with ∼100 μg of beads (e.g., upon the addition of beads, little-to-no increase in capture was observed); thus for consistency, 100 μg of beads was used per sample (Figure S2). In applying beads to cell isolation, the combination of antibody density and the bead number can greatly impact results; thus in ideal situations, the antibody density and bead number introduced per isolation should be titrated for each application.

Cell Capture

Target cell capture efficiency and purity is one of the most important magnetic bead characteristics. To characterize target capture across each bead type, a low, medium, and high EpCAM-expressing cell line (Du145s, 22Rv1s, and LNCaPs, respectively) was tested (Figure A). Despite the use of an identical antibody to capture with, the capture varied greatly across magnetic bead types, especially in the low-expressing Du145s; CELLection and FlowComp magnetic beads resulted in the lowest capture, whereas M-270s notably captured the largest population of Du145s. Although the identical antibody lot (and conjugated stock with the exception of M-270 beads) was used across all capture experiments, differences in the bead surface (e.g., roughness and curvature due to size differences) or the functionality of the surface could impact how the antibody orientates on the surface of the bead. Antibody orientation would impact the antibody’s potential for successful epitope binding, possibly explaining the variable capture observed. Similarly, how (and where) the antibody is biotinylated could impact antibody–bead performance, highlighting the need to optimize each component of the process for each bead type and new application.

Figure 2

Characterization of cell capture. (A) Capture of EpCAM-expressing cell lines (Du145 = low, 22Rv1 = medium, LNCaP = high) by each bead type. Beads are abbreviated as follows: SM = Sera-Mag, FC = FlowComp, CELL = CELLection. (B) Nonspecific capture of PBMCs by each bead type across varying PBMC inputs. (C) Direct vs indirect capture of Du145s from a PBMC background. (D) Resultant purity of the captured target cells from direct and indirect capture of Du145s. In all plots, the bars represent the technical replicate average (n = 3) with error bars representing the standard deviation.

Although specific capture of the target is critical, for many endpoints (e.g., sequencing), high purity is also required as contaminating populations bias and mask target cell signatures. To determine the potential of contaminants to reduce purity for each magnetic bead type, the nonspecific capture of each bead was estimated by incubating the beads with a mixed background population (PBMCs) (Figure B). Sera-Mag and M-270 beads had the lowest rate of nonspecific capture of PBMCs compared to the almost 10-fold increase in nonspecific binding with M-280s and CELLection beads. The balance between the specific capture (target cells) and nonspecific capture (i.e., background cells) often determines the endpoints available as these metrics will determine both yield and purity, a consideration in identifying an assay-specific bead type.

Both capture efficacy and purity may also be impacted by the bead isolation method used: direct or indirect. Direct cell capture is the most common, wherein the prebound antibody–bead complex is incubated with the cells. In contrast, indirect capture involves incubating the antibody with the cells, followed by bead capture of the antibody-labeled cells. Typically, indirect capture results in higher capture efficacy yet also results in increased contaminants. For indirect capture, the antibodies dispersed in the sample are free to interact and may incidentally bind nonspecifically with contaminant cells at higher frequencies than when attached to the bead in the direct capture; the balance of captured target cells (increased target yields higher purity) and contaminants (increased contaminants yields lower purity) will ultimately determine the resultant impact of indirect capture on purity for each bead type. Indirect versus direct capture of low EpCAM-expressing cells (Du145s) from PBMCs was evaluated for each bead type (except for M-270s, which are limited to direct capture) (Figure C). Additionally, the resultant purity is reported (Figure D). Interestingly, the capture of Du145s was highly variable across bead types. Notably, if direct capture was low (average ∼10% capture), switching to indirect capture had no impact, as demonstrated by the CELLection and FlowComp magnetic beads. With indirect capture, the antibody is first added to the cells to bind (little-to-no variation across conditions); then beads are added and the antibody–cell complex is bound to the beads. On the basis of the physical surface (e.g., roughness), the positioning of streptavidin on the surface of the bead, and the location of biotin(s) on the antibody, the orientation of the bound antibody on the bead can be impacted. Binding of the biotin (or modified biotin) antibody to the streptavidin may result in an antibody orientation, which results in decoupling of the antibody from the EpCAM, resulting in the release of the cell; this may explain why some bead types saw little difference between direct and indirect captures. In contrast, M-280s and Sera-Mag magnetic beads improved capture efficacy when transitioning from direct to indirect capture.

When evaluating the impact of direct and indirect capture on purity (Figure D), the results were surprisingly mixed. Although M-280s gained in capture efficiency using indirect capture (from ∼30 to ∼50%), the overall purity did not change, because of the increased contaminants captured in parallel. Sera-Mags, which also saw gains in capture efficiency (from ∼30 to ∼85%) with indirect capture, saw an increase in the captured contaminants, but overall exhibited an increase in purity with indirect capture. In contrast, CELLection and FlowComp saw little differences in capture efficiency and no notable changes in purity. Although differences in the indirect and direct capture are difficult to predict without experimentation and variable across beads, the gains in the target capture efficacy for M-280 and Sera-Mag beads highlight the potential benefit of evaluating these metrics when evaluating bead types.

Magnetic Bead Release Characterization

To characterize release, the release of fluorescently labeled EpCAM antibodies from each bead type was first characterized, followed by cell release from each bead type. For each of the bead types, release is accomplished through different mechanisms. FlowComp beads accomplish release by introducing d-biotin (B-1595 or B-20656, Thermo Fisher) or d-desthiobiotin (D-20657, Thermo Fisher). These molecules have a higher affinity for (modified) streptavidin on the bead surface and thus displace DSB-X biotin from streptavidin, thereby releasing the antibody and the attached cell from the bead. In contrast, CELLection beads utilize a DNA linker to connect the antibody to the bead; the provided DNase I can cleave the DNA linker attaching the bead to the antibody to release captured cells from the bead. To fluorescently characterize the release, a fluorescent secondary system was used. Beads with different densities of primary antibody (22.5, 0.5, and 0.025 ng antibody per μg PMP) were placed in the release buffer for varying durations, the magnetic beads were then removed from the release buffer, and the remaining fluorescence on the magnetic bead was measured (Figure A,B). Using this system, a decrease in fluorescence intensity corresponds to a release of antibody. Although CELLection (Figure A) demonstrated the most rapid release, both chemistries demonstrated at least 50% release within 5 min. Notably, FlowComp bead release seemed hindered at higher antibody densities, resulting in incomplete release. The delayed release by FlowComp at a high, maximal antibody density could be due to limited access of the release buffer into the tightly packed antibody–bead complex. CELLection may not suffer from this issue as, by using a DNA linker between the bead and antibody, CELLection provides added space between the antibody and bead allowing easier access of the DNase.

Figure 3

Characterization of the release from FlowComp and CELLection beads. (A,B) Release of a fluorescently labeled anti-EpCAM antibody from (A) FlowComp and (B) CELLection beads. Beads were labeled with low, medium, and high levels of antibody and released for the specified time intervals. (A,B) Dots represent the average of three technical replicates with each technical replicate representing the average of 100 beads (total of 300 beads); error bars represent the standard deviation of the technical replicates. (C) Release of 22Rv1s from FlowComp and CELLection beads across time. (D) Capture efficiency of both FlowComp and CELLection beads when used to capture Du145, 22Rv1, and LNCaP cells. (E) Release efficiency of the three cell lines following bead-based capture. (F) Effective capture following the release of the cells. Gray bars represent the population of cells lost during the release process because of the inefficient release. In each plot, bars represent an average of n = 3; error bars represent the standard deviation.

Next, the release of captured cells across different time points was evaluated using 22Rv1s. Cells were captured on magnetic beads and allowed to incubate in the release buffer for varying durations. The released and bead-bound fractions were then used to determine the release efficiency across time (Figure C) (Figure S3). Within 5 min, the maximal release was obtained for each cell line. These data are comparable to the fluorescence data (Figure A,B), where for medium and low antibody densities, substantial release occurred by 1 min, then diminishing gains were observed as time increased. In the context of cells, however, the release was delayed—complete release occurred at 5 min instead of 1 min. As cells are often bound to beads via multiple antibody linkages, multiple linkages must be broken to release the cell; this is likely slowing the process when compared to the fluorescent antibody characterization.

The release characteristics of the low, medium, and high EpCAM-positive cell lines were then characterized. Using the FlowComp (blue bars) and CELLection (yellow bars) beads, we evaluated the best-case capture efficacy of each bead (Figure D). Next, cells were released for 20 min (Figure E). CELLection beads were the most effective at releasing cells, releasing ∼78–88%. The release from the FlowComp beads was considerably lower than the CELLection beads; yet, both bead types resulted in some cell loss during the release because of an unreleased fraction remaining bound to the beads (Figure F). The lack of complete release and discrepancies between release efficiencies of FlowComp and CELLection beads could be due to a number of differences in the release approaches. Release is dependent on ensuring that the bead binds the cell through the antibody as the antibody is ultimately released from the bead. For instance, because of the close proximity of cells and beads, cells may nonspecifically interact with the surface of the bead. As a result, antibody-based release methods become ineffective at the release of the beads as the cells are no longer solely bound via the specific antibody interaction. Additionally, CELLection use a spacer (DNA linker) between the cell and bead. This spacer may both place some additional distance between the target cell and bead (reducing direct bead interactions) as well as enable easier access of the releasing agent (DNase) to its target especially when a number of antibody–EpCAM interactions are likely occurring in a small area (e.g., high-expressing cells, LNCaPs). Although difficult to determine the mechanism(s) impairing release, evaluating different approaches with a relevant target of interest is important for optimal results.

Although magnetic beads enable isolation of a target population, bead removal may be required for optimal compatibility with downstream techniques such as fluorescence microscopy. Releasable bead chemistries enable downstream separation and removal of the magnetic beads following capture, frequently by the dissociation of the antibody and bead. Thus, although release can be advantageous for an assay (i.e., imaging), the benefits may be counter-balanced by a decreased, resultant-captured (and released) target population.

Impact of Magnetic Beads on Imaging and Analysis

Once a cell population of interest is captured, many downstream applications involve fluorescent protein staining, either for verification and identification of the population or for protein localization and expression. In either application, the fluorescent signal from the magnetic beads may ultimately limit the fluorescent stains or channels as well as impact the ability to discern localization or expression. Thus, the fluorescence of each bead type on glass was initially characterized across five different fluorescent channels (Figure A), with the flowing excitation/emission filters (center(range)): 390(22)/440(40), 485(25)/525(30), 560(32)/607(36), 648(20)/684(24), and 740(13)/809(81) (as highlighted in the methods) (Figure S4). Each bead had some autofluorescence; while the intensity varied between bead types and channels, each bead peaked at an emission of (560 nm) (Figure A).

Figure 4

Impact of magnetic beads on downstream fluorescence microscopy readouts. (A) Baseline autofluorescence of magnetic beads imaged on glass across five different filter sets with emission wavelengths listed. Bars represent the average of 200 beads; error bars represent the standard deviation of the analyzed events. (B) Example images of single cells bound to each of the bead types (note: bead coverage of the cell greatly varied cell-to-cell for each bead type from a few beads to complete coverage), demonstrating variable staining patterns as influenced by the presence of cell isolation magnetic beads. (C) Impact of magnetic beads on identifying the LNCaP nuclear area based on Hoechst staining for a nucleus. (D) Identified cellular androgen receptor signal. (F) Ratio of nuclear to cytoplasmic androgen receptor identified in LNCaPs captured with each bead type and compared to bead-free cells. In the box plots, 50 cells were analyzed per condition; the notch represents the 95% confidence interval of the median and the circles are possible outliers. A statistically significant difference with respect to the no-bead group is indicated (*).

To evaluate the potential impact of magnetic beads on the evaluation of both protein expression (staining intensity) and localization (based on a nuclear and cytoplasmic staining), LNCaPs were captured with each magnetic bead type, fixed, and stained with a nuclear stain (Hoechst) as well as antibodies to pan-cytokeratin (Alexa790), EpCAM (PE), and androgen receptor (AR) (Alexa488). The captured cells were then compared to a bead-free population (Figure B). Beads were found to have a variable impact on the identified nuclear area, which conceivably would impact the ability to easily discern nuclear localization of proteins (Figure C). Total calculated cellular AR resulted in statistical difference in every bead type compared to bead-free cells, demonstrating the potential of magnetic beads to modify detected signal per cell, an issue when attempting to identify populations based on expression (or lack of) (Figure D). In all bead types evaluated, cellular AR decreased, likely due to the beads attenuating fluorescent intensity. Alternatively, if a protein was expressed at a very low level (or not at all), the bead autofluorescence might lead to a false quantification of positive signal. Using both the AR signal and localization based on the determined nuclear area, the ratio of AR nuclear localization (nuclear AR to total AR) was calculated (Figure E). Although the localization ratio across bead types seemed less variable than the results for cellular AR signal, certain beads better yielded expression and localization patterns as bead-free cells (Sera-Mag and CELLection).

Magnetic beads impact the evaluation of fluorescent staining for protein expression and localization as well as identification of the nuclear area. One additional variable, bead coverage of the cells, is also likely to influence these results. All bead types appeared to variably cover cells ranging from only a few beads per cell to complete coverage within a single sample, highlighting cell-to-cell heterogeneity. As the number of beads, which bind to a cell is difficult to control and highly variable (as observed by the range of bead coverage of the cells within each bead type used), we assessed the impact of beads bound to cells using the entire population of the captured cells (both highly and sparsely covered cells). Thus, assays relying on endpoint protein localization or cell identification through fluorescent staining should closely evaluate the impact of cell isolation beads, as beads can significantly distort population appearances; distortion likely impacted by the number of beads bound to a cell, a variable difficult to control.

Postcapture Culture of Cell Lines

Following cell isolation, many assays require the user to culture the cells rather than perform a terminal endpoint assay such as intracellular staining. Thus, the viability of cells captured via magnetic beads was evaluated. Anti-EpCAM beads were incubated with ∼5000 22Rv1s or LNCaPs and isolated resulting in a captured bead-bound population. A total of 50 μg of magnetic beads were used for each bead type. Following isolation, the cells and beads were transferred into a 96-well plate and cultured for 3 days. After 3 days of culture, cells viability was assayed (Figure ).

Figure 5

Cell viability following capture and release. (A) The viability of cells (LNCaP, 22Rv1) bound to nonreleaseable beads compared to untouched cells (underwent no magnetic bead isolation) following a 3-day culture (abbreviations: SM = Sera-Mag). (B) The viability of cells bound to (bound) and released from (released) releasable beads (CELLection, FlowComp) following a 3-day culture. Bars represent the average of three technical replicates; error bars represent the standard deviation.

For the nonreleaseable beads, results indicated that M-280 and Sera-Mag beads have no statistical impact on cell viability compared to cell only (no-bead) viability for both cell lines (Figure A). M-270 beads resulted in a decrease in viability relative to the cell only control (p-value < 0.01 for both LNCaPs and 22Rv1s). For the releasable CELLection and FlowComp beads, viability is shown for (1) a no-bead cell only control (none), (2) cultured bead-bound cells (bound), and (3) cells cultured postcapture and post release (released) (Figure B). Overall, the viability across conditions—including released and bead-bound cells—remained high in both LNCaPs and 22Rv1s. For LNCaP and 22Rv1 cells, no significant differences were seen, regardless of the bead type used or the culture condition (i.e., bound or release). Ultimately, viability is likely an artifact of the cell type, cell density, and bead density; nevertheless, these high viability results demonstrate both the promise and potential impact on postcapture culture.

Integration of Magnetic Bead-Based Cell Isolation with Standard Nucleic Acid Extraction Methods

Downstream of cell isolation, many endpoints involve nucleic acid isolation. To characterize the potential impact of each cell isolation magnetic bead type on nucleic acid isolation, both RNA and DNA were evaluated. Ultimately each nucleic acid isolation protocol differs in buffers, nucleic acid binding mechanisms, and the impact of the cell isolation beads; thus, to highlight the variable impact of cell isolation beads, two extraction methods were analyzed for completeness. For both RNA and DNA, a low cell number sample (∼5000 cells) was evaluated using a spin column and a bead-based technique.

For RNA, 50 μg of cell capture magnetic beads were added to each cell sample prior to the addition of any lysis buffer to ensure that the impact of cell isolation beads on the entire RNA isolation process was evaluated. RNA was then isolated with a magnetic bead-based method (Dynabeads mRNA DIRECT) as well as a spin column (Qiagen RNeasy Mini Kit). Following isolation, the eluted RNA (and beads) underwent reverse transcription (RT) into cDNA; the cDNA was quantified by real-time quantitative-PCR (qPCR) (no beads were loaded into the reaction). When cell isolation magnetic beads were integrated into a spin-column isolation, little loss in RNA was detected compared to the cell-only control (Figure A); rather FlowComp magnetic beads resulted in a statistically significant (p-value = 0.032) increase in the detected mRNA (Figure A). Similarly, the bead-based mRNA extraction—the Dynabeads mRNA DIRECT Kit—resulted in no statistical difference in RNA quantified from the cell-only condition or, in the case of CELLection and Sera-Mag (M-270s resulted in an average increase, but was not significant), a significant increase in RNA was detected (Sera-Mag p-value = 0.043; CELLection p-value = 0.012) (Figure B).

Figure 6

Characterization of nucleic acid extraction with cell isolation magnetic beads present. (A,B) Relative fold change in the mRNA transcript (HPRT) detected from LNCaPs. Isolations containing cell isolation beads were compared to no-bead controls for two methods of RNA extraction: (A) spin columns and (B) bead-based extraction. (C,D) Similarly, relative fold change in GAPDH from DNA extracted via (C) spins columns or (D) bead-based extraction. Bars represent the average of three technical replicates; error bars represent the standard deviation; * denotes p < 0.05 and *** denotes p < 0.001; --- indicates the cell only, no-bead control (abbreviations: CELL = CELLection, SM = Sera-Mag, FC = FlowComp).

Similarly for DNA, 50 μg of cell isolation magnetic beads were added to the cells prior to DNA isolation. Both a magnetic bead-based approach (DNA-binding silica bead) and a spin column approach (QIAamp DNA Mini Kit) were used to evaluate the potential impact of cell isolation beads. To quantify the isolated DNA, qPCR was performed for a housekeeping gene (GAPDH). When DNA was isolated by spin columns, no statistical differences were seen in detected DNA yields (Figure C). In comparison, when DNA was isolated by silica beads (Figure D), the DNA yield (via GAPDH) was comparable to the control for M-280, Sera-Mag, and CELLection beads. However, a decreased yield was observed when either FlowComp or M-270s beads were present during the lysis step (p-value of 0.049 and ≪0.01, respectively). Notably, FlowComp beads resulted in some loss (approximately half the DNA yield compared to control), but M-270s resulted in over a 90% decrease in the detected DNA, a very different result compared to the spin column DNA isolations. In this finite test of five bead types, M-270s and FlowComp beads were the only beads that resulted in the loss of DNA, specifically when DNA was isolated using the bead-based DNA isolation protocol.

Across isolation methods in both RNA and DNA (i.e., spin columns vs magnetic bead-based extraction), cell isolation magnetic beads had variable impacts on nucleic acid extraction based on the nucleic acid approach used. In bead-based DNA isolation, cell isolation magnetic beads could significantly hinder yield, yet identical cell isolation beads had no statistical impact on the spin column isolation. Furthermore, bead types did not affect the capture of all nucleic acids alike; a cell isolation bead that seemed to impact DNA yield did not necessarily impact RNA yield (e.g., M-270). Overall, the variable impact of cell isolation beads across nucleic acid extraction methods highlights many of the potential nuances in integrating cell isolation beads into complex cell isolation protocols.

Conclusion

Cell isolation magnetic beads enable the rapid targeting of nearly any cell population, paired with a nearly endless offering of commercial antibodies. Yet, how the isolation magnetic beads perform and how they integrate into downstream endpoints impact their utility to users. Different cell isolation magnetic beads come with trade-offs in their ability to facilitate and integrate into different endpoints of cell isolation protocols. For baseline performance metrics, M-280s facilitated strong target capture enabling the use of both direct and indirect capture approaches. For purity, Sera-Mag and M-270s paired strong capture with low nonspecific binding of a complex background PBMC population. Although FlowComp and CELLection did not perform as well, these beads enabled release, which may be required for different culture applications as well as facilitate precise fluorescent immunohistochemistry endpoints such as protein localization. All cell isolation beads demonstrated compatibility with RNA and DNA extraction; yet results highlighted the method and buffer dependency of these results. This article aims to evaluate the beads in the presented context of EpCAM-specific cell capture, highlighting the range of results obtainable depending on the bead-type utilized. Although this paper attempted to ensure optimal performance across bead types, the attempts to standardize traits (e.g., fifty percent maximal binding capacity and bead number added) could all strongly influence the resultant capture (and release). Similarly, although buffers were standardized across isolations, the buffers and additives [e.g., fetal bovine serum (FBS), bovine serum albumin (BSA), and ethylenediaminetetraacetic acid (EDTA)] could impact performance. Thus, this paper serves the introduction of different bead types and provides insight into downstream users. Ultimately, further investigation is required to better understand the mechanisms behind the observed variation and direct the design of improved magnetic beads for cell applications.

Materials and Methods

Magnetic Beads, Antibody Conjugation, and Binding

Capture experiments used a goat polyclonal anti-EpCAM antibody (Clone AF960) (AF960, R&D Systems) conjugated to the following magnetic beads: Dynabeads M-270 Epoxy beads (14311D, Thermo Fisher), Dynabeads M-280 Streptavidin (11205D, Thermo Fisher), CELLection Biotin Binder Kit (11533D, Thermo Fisher), FlowComp Dynabeads (11061D, Thermo Fisher), and Sera-Mag SpeedBeads streptavidin-blocked magnetic particles (21152104011150, GE Healthcare Life Sciences) (see Supporting Information for more detailed information). An overview of the magnetic beads evaluated is provided in Table S1.

Antibody was batch-conjugated to M-270s, as per the manufacturer’s instructions, using the Dynabeads Antibody Coupling Kit (14311D, Thermo Fisher). Because of the batch conjugation of M-270s, unlike the alternative magnetic bead types, an antibody to bead density could not be easily titrated. Thus, M-270s were conjugated following the manufacturer’s recommendation, at a density of 6 μg antibody per milligram of beads. For all other beads, the antibody was first biotinylated following the manufacturer’s instructions (DSB-X Biotin Protein Labeling Kit D-20655, Thermo Fisher) to facilitate streptavidin–biotin binding of the antibody to the beads. Magnetic beads were washed by placing the beads on a magnetic tube rack (DynaMag Rack, Thermo Fisher), removing the original buffer, and resuspending the beads in an identical volume of buffer [0.1% BSA in Ca2+ and Mg2+-free phosphate-buffered saline (PBS) with 2 mM EDTA]; the beads were again washed prior to use. Separately, the antibody was diluted into an identical volume of buffer, which was combined with the washed beads and tumbled for 30 min using a Labquake rotator (Thermo Fisher) (set to approximately 6 rpm). Following binding, the fluid was removed, and the beads were washed and resuspended in buffer for use.

EpCAM Expression

To more robustly characterize EpCAM-based cell capture, EpCAM expression was assessed for each cell line to differentiate a high, medium, and low EpCAM expresser. Cells were stained with an anti-EpCAM antibody conjugated to phycoerythrin (PE) (Clone VU-1D9) (ab112068, Abcam) (1:100) and Hoechst 33342 (H3570, Thermo Fisher) (20 μg/mL), both diluted in 1× PBS supplemented with 2 mM EDTA and 0.1% BSA. Once stained, cells were washed, resuspended in PBS, and imaged on glass. After imaging, the mean fluorescence (expression) of each cell line was calculated and normalized to the maximum EpCAM-expressing cell line (LNCaP) to more easily compare the relative expression (Figure S1). On the basis of these results, Du145 (low EpCAM), 22Rv1 (medium), and LNCaPs (high) were used in all subsequent experiments. Each of the three cell lines had a similar diameter ranging from ∼15–25 μm, with Du145s being slightly smaller, generally ∼15–20 μm. The cell lines screened for EpCAM expression were all prostate cell lines; prostate cancer has been one cancer type for which EpCAM-based capture has proven clinically relevant in capturing CTCs.[34,35] Additionally, many of the cell lines evaluated have been used in the characterization of CTC capture platforms;[30,33] thus, these cells represent a relevant target, spanning a wide range of EpCAM expression, in the context of EpCAM-based capture of prostate cancer cells.

Cell Isolation and Release

All cell isolations were performed using EXTRACTMAN (EM) (22100000, Gilson), a platform based on the sliding lid for immobilized droplet extraction technology.[36] EM allowed the simultaneous isolation of up to four samples. For direct isolation, 100 μg of antibody-coated magnetic beads (as described above) was incubated with cells (total volume of 475 μL) on a Labquake tumbler (Thermo Fisher) rotating at ∼6 rpm for 30 min at 4 °C. For indirect isolation, the anti-EpCAM antibody was first added to the cell solution (475 μL) and tumbled for 30 min at 4 °C (as specified above); 100 μg of magnetic beads were then added and the solution tumbled for 10 min. After incubation, the entire volume was loaded into the input well of an EM plate (22100008, Gilson). Using EM, the cells were then captured on the EM consumable strip (22100007, Gilson) as the built-in EM magnets were held over the middle of the well for 30 s to enable the collection of the beads, and then the EM handle was moved over a wash well (small wash well, 110 μL), where the lower magnets automatically pulled the beads into the well. Once dropped in the wash well, the EM collection strip was pulled back from the well, the lower magnet was removed, and the cells were mixed three times using a pipette (set to 70 μL). For experiments characterizing nonspecific capture of PBMCs, two additional washes (110 μL) were carried out to improve stringency in deciphering between nonspecific binding (i.e., cells and beads) and basic carryover (i.e., cells caught in residual fluid on the strip). Once mixed, the bead-bound cells were recollected on the consumable strip by leaving the EM handle positioned over the well for 15 s and then moving the handle to the next well. The contents of all wells were then collected and imaged to ensure accurate cell counts.

For experiments involving release, a similar experimental design was followed, except that once washed, the beads and bound cells were dropped into a release well containing either FlowComp Release Buffer (FlowComp Flexi Kit 11061D, Thermo Fisher) or CELLection release buffer (CELLection Biotin Binder Kit 11533D, Thermo Fisher) prepared according to the manufacturer’s instructions (release volume of 110 μL). Once dropped into the well, cells and beads were mixed by a pipette (three mixes; pipette set to 70 μL), allowed to incubate for 20 min, mixed again, and collected. Any nonreleased population of cells was then collected by a magnet and transferred using EM to the final well. All wells were imaged to ensure accurate cell counts.

Fluorescent Staining

Fluorescence characterization of the antibody–bead interaction and antibody–bead density was performed with either antigoat Alexa488 (ab150129, Abcam) or antigoat Alexa555 (ab150130, Abcam) secondary antibodies. In brief, following the binding of the primary antibody to the beads, the diluted secondary antibody (in buffer) was added to the beads for 30 min. The beads were then washed and resuspended in PBS prior to imaging. This fluorescence characterization was used to identify an optimal anti-EpCAM antibody density for each bead type (with the exception of batch-conjugated M-270s). Using the fluorescent secondary antibody, the fluorescent signal on the bead (due to bound anti-EpCAM antibody on the bead surface) was quantified across increasing amounts of primary antibody (Figure A). The resultant intensity curve of fluorescence versus antibody density was then used to identify the 50% maximal binding capacity for each bead type (Table S2). To standardize the antibody function on the surface of the beads for all subsequent experiments unless noted (given the differing surface areas and surface functionalities), the identified 50% maximal antibody density was used.

To first fluorescently characterize release, FlowComps (with bound primary and fluorescent secondary antibodies) were resuspended in FlowComp release buffer (110 μL). At set time points, the beads were collected using the EM handle and removed from the release buffer. The beads were then dropped in a wash buffer and were imaged. The CELLection beads were similarly characterized (utilizing 110 μL CELLection Biotin Binder Kit Release Buffer). The measured bead fluorescence was corrected by subtracting the baseline autofluorescence of the blank-bead incubation with the appropriate secondary antibody as in the experimental conditions.

For all cell line-based capture experiments, cells were prestained with either Calcein, AM (C3100MP, Thermo Fisher) or CellTracker Red CMTPX Dye (CTR, C34552, Thermo Fisher). CTR was utilized when the background cells were present, which were concurrently stained with Calcein, AM to enable identification of each cell type. For viability experiments, a live–dead assay was performed on the cultured populations with Calcein, AM and ethidium homodimer-1 (E1169, Thermo Fisher) at a final concentration of 1 and 20 μg/mL, respectively. Cells were allowed to incubate for 20 min and then imaged using a 10× objective.

To determine the impact of cell isolation beads on fluorescent immunohistochemistry, bead-captured LNCaPs were compared to an untouched population. Bead-bound cells were captured and washed using EM to ensure that bead-free cells were removed from the population. Once isolated, the bead-bound cells and untouched population were incubated in buffer containing anti-EpCAM antibody conjugated to PE (1:100) and Hoechst 33342 (20 μg/mL) for 30 min. The cells were then washed with buffer, fixed for 15 min in 4% paraformaldehyde (P6148, Sigma), and washed again. Following permeabilization (PBS with 1% Tween-20 and 0.05% saponin for 30 min), cells were resuspended in buffer containing an antipan cytokeratin antibody [fluorescein isothiocyanate (FITC)] (35 μg/mL) (ab11214, Abcam) and antiandrogen receptor antibody (1:100) (5153S, cell signaling) (incubated at 4 °C overnight). After washing the cells, goat antirabbit Alexa Fluor 488 (ab150073, Abcam) was added at 10 μg/mL for 1 h in buffer. Samples were washed in buffer and resuspended in PBS prior to imaging on glass.

Imaging and Image Analysis

Samples were imaged on a Nikon Eclipse Ti at 10× magnification (0.33 μm/pixel) (Nikon, USA). Acquisition was performed with one of the following channels and filter sets: 390 × 440 (ex 390/22; em 440/40), 485 × 525 ((ex 485/25; em 525.30), 560 × 607 (ex 560/32; em 607/36), 648 × 684 (ex 648/20; em 684/24), and 790 × 809 (ex 740/13; em 809/81). For capture and viability experiments, images were analyzed using the provided NIS-Elements AR Microscope Imaging Software.

The quantification of fluorescent intensity of beads and cells was performed using custom scripts written in MATLAB version R2016B (Mathworks, Natick, MA). All raw fluorescence images were corrected for the background signal by subtracting the local median within a square-moving window with dimension at least 5 times the diameter of the cell type or bead of interest. Background-subtracted bead images were smoothed using a Gaussian filter (σ = 0.66 μm), and masks of the beads were generated by thresholding (Otsu method) off the autofluorescence of the unstained 485 nm channel. The relative density of EpCAM was quantified by calculating the mean intensity of the EpCAM channel (560) within the masks and dividing this value by the mean autofluorescence intensity.

For experiments investigating the relative EpCAM expression of cell lines, images were normalized to have zero local mean and unit local variance to improve the robustness of segmentation. Masks of the periphery of the cells were created by thresholding the normalized image at 1, and the relative EpCAM expression was quantified as the mean intensity within these masks.

For AR localization experiments in LNCaPs, cell locations were manually marked in the cytokeratin channel (FITC), and masks of the cell were generated by thresholding (Otsu method) followed by morphological reconstruction using the manual markers. Nuclear masks were generated using the same method on the Hoechst channel, and cytoplasmic masks were calculated by subtracting the nuclear region from the cell masks. Relative nuclear and cytoplasmic expression of AR was quantified as the mean signal in these channels within each respective mask, and a nuclear localization metric was defined as the ratio of nuclear to cytoplasmic expression.

Cell Culture

Cells were cultured under sterile culture conditions at 37 °C in 5% CO2. VCaPs (courtesy of Dr. Scott Dehm, University of Minnesota) were cultured in Dulbecco’s modified Eagle’s medium (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin–streptomycin (PS) (Gibco). All other lines—LNCaPs (ATCC), Du145s (courtesy of Dr. Scott Dehm), 22Rv1s (courtesy of Dr. Douglas McNeel, University of Wisconsin–Madison), PC3s (courtesy of Dr. Scott Dehm), and PC3-MM2 (courtesy of Dr. C. Pettaway, MD Anderson Cancer Centre, TX, USA) were cultured in RPMI1640 media (#11875–093, Thermo Fisher Scientific) with 10% FBS and 1% PS. To maintain consistency across experiments, cells were counted, plated at 0.3 × 106 cells per well in a 6-well plate, and cultured for 48 h prior to use.

Blood Processing

PBMCs were isolated from whole blood for use as background cells. The whole blood—collected from healthy donors and treated with K3 EDTA (Biological Specialty Corporation)—was received within 24 h of the blood draw and was processed. Briefly, the whole blood was mixed 1:1 with 1× PBS, overlaid on 15 mL of Ficoll Paque PLUS (17-1440-02, GE Healthcare), and centrifuged following the manufacturer’s instructions. After centrifugation, the buffy coat was removed and diluted in 20 mL wash buffer (1× PBS supplemented with 0.5% BSA and 2 mM EDTA). Cells were then centrifuged (200 rcf, 10 min), pelleted, and resuspended again in 20 mL wash buffer and stored on ice until ready for use. Once ready for use, cells were centrifuged and resuspended as needed.

Nucleic Acid Extraction and Quantification

For RNA, cell samples (including 50 μg of beads) underwent either a spin column RNA extraction kit (AllPrep Spin Columns, Qiagen) or a magnetic bead-based extraction (Dynabeads mRNA DIRECT 61011, Thermo Fisher). For the spin column, the manufacturer’s protocol was followed, eluting into 15 μL of the provided elution buffer. For the magnetic bead-based RNA extraction, 200 μL of provided lysis/binding buffer and 20 μL of oligo(dt) beads were added to the cells. Using a magnetic rack (12321D, Thermo Fisher), RNA was isolated following two 200 μL washes of Wash Buffer B (10 mM Tris–HCl (pH 7.5) (Sigma), 0.15 M LiCl (Sigma), 1 mM EDTA (Sigma)), followed by elution in 15 μL of elution buffer, 10 mM Tris–HCl. The eluted sample (including magnetic beads) was reverse-transcribed using the high-capacity cDNA reverse transcriptase kit (4387406, Thermo Fisher) on a Techne TC-412 Thermal Cycler (37 °C for 1 h; 85 °C for 5 min).

For DNA extraction (including 50 μg of magnetic beads), a spin column (QIAamp DNA Mini Kit 51304, Qiagen) and a silica magnetic bead-based approach were evaluated. For the spin column, the manufacturer’s instructions were followed until elution where a modified elution volume of 15 μL was used. For the silica bead-based isolation approach, cells were lysed in 200 μL RLT Plus (1053393, Qiagen) with 5 μL of Magnesil KF magnetic beads (MD1471, Promega). Following lysis, beads and extracted DNA were washed with Wash Buffer B (above) and eluted in 15 μL of nuclease-free water.

To quantify, DNA or cDNA was mixed with a LightCycler 480 Probes Master Mix (04535286001, Roche) and a Taqman assay for either GAPDH (Hs02786624_g1, LifeTech) (DNA) or HPRT (Hs11501003267_m1, LifeTech) (cDNA). The reaction underwent quantitative PCR on a LightCycler 480 (Roche) thermal cycler (preincubation of 95 °C for 5 min; 45 cycles of 95 °C for 10 s, 60 °C for 30 s, 72 °C for 1 s). The cycle threshold was calculated by the LightCycler software with the second derivative algorithm.

Statistics

AR localization results were analyzed for difference by one-way ANOVA. Posthoc multiple comparisons were performed using a t-test with Bonferroni correction. Statistical significance was defined as p ≤ 0.05/15 = 0.0033.