Multiparameter Cell Cycle Analysis

James W. Jacobberger, R. Michael Sramkoski, Tammy Stefan, and Philip G. Woost

Abstract

Cell cycle cytometry and analysis are essential tools for studying cells of model organisms and natural populations (e.g., bone marrow). Methods have not changed much for many years. The simplest and most common protocol is DNA content analysis, which is extensively published and reviewed. The next most common protocol, 5-bromo-2-deoxyuridine S phase labeling detected by specific antibodies, is also well published and reviewed. More recently, S phase labeling using 50-ethynyl-20-deoxyuridine incorporation and a chemical reaction to label substituted DNA has been established as a basic, reliable protocol. Multiple antibody labeling to detect epitopes on cell cycle regulated proteins, which is what this chapter is about, is
the most complex of these cytometric cell cycle assays, requiring knowledge of the chemistry of fixation, the biochemistry of antibody-antigen reactions, and spectral compensation. However, because this knowledge is relatively well presented methodologically in many papers and reviews, this chapter will present a minimal Methods section for one mammalian cell type and an extended Notes section, focusing on aspects that are problematic or not well described in the literature. Most of the presented work involves how to segment the data to produce a complete, progressive, and compartmentalized cell cycle analysis from early G1 to late mitosis (telophase). A more recent development, using fluorescent proteins fused with proteins or peptides that are degraded by ubiquitination during specific periods of the cell cycle, termed “Fucci” (fluorescent, ubiquitination-based cell cycle indicators) provide an analysis similar in concept to multiple antibody labeling, except in this case cells can be analyzed while living and transgenic organisms can be created to perform cell cycle analysis ex or in vivo (Sakaue-Sawano et al., Cell 132:487–498, 2007). This technology will not be discussed.

Key words Cell division cycle, Cell proliferation, Mitosis, Mitotic states, Cell states, Antibodies, Monoclonal antibodies, Intracellular antigens, Fixation, Immunochemical staining, Immunofluorescence

1 Introduction

This chapter is narrow in focus, providing methods we have used with human cell lines, dispersed tissue, blood, and bone marrow; however, the methods should apply to most animal cells. The large fraction of multiparametric human cell cycle studies justifies this focus (Table 1). In discussion and references, we did not actively exclude rodent studies but found no compelling reason to cite Teresa S. Hawley and Robert G. Hawley (eds.), Flow Cytometry Protocols, Methods in Molecular Biology, vol. 1678, DOI 10.1007/978-1-4939-7346-0_11, © Springer Science+Business Media LLC 2018
203.

1.1 Definition

Here, we define multiparametric cell cycle analysis as measurements and computation aimed at defining or identifying cell cycle com- partments, phases, or states and/or quantitative cell cycle phase/ state-related expression of biomolecules. This is independent of the means by which a biomolecule is labeled or detected, but method- ologically, we present only the combination of DNA-binding dyes and antibodies labeled with fluorescent small molecules or proteins. Technically, a multiparametric analysis wherein a marker is used to define or isolate a cell type and an additional, single parameter aimed at identifying phases or phase-related expression is not multi- parametric analysis (e.g., cytokeratin and DNA content). However, in our evaluation of the literature, a large number of publications that we counted as multiparametric cell cycle papers were exactly just that. The idea that we are adhering to here is that within our narrow definition, the intent of multiple parameters is to impart more information about the cell cycle, whereas in assays of cell type (e.g., differentiation markers) and a single-cell cycle parameter, the cell cycle information is one dimensional. Also, we have not pur- posefully or comprehensively included studies that mainly use cell size, light scatter, and morphology or imaging topological features as parameters in the multiparametric cell cycle analysis class. The reasoning here is almost the same. Cell size and light scatter mea- surements are most often utilitarian—used to clean the data (e.g., eliminate debris, subtract background). They have rarely been used to impart cell cycle information—most likely because in mamma- lian cells, the measurements are too broadly distributed and over- lapping to be of much use. Imaging feature analysis should be included, because the cell cycle-related translocation of biomole- cules is a common and important feature of cell cycle regulation. However, imaging cytometry of the cell cycle is a very understudied area, and most publications do not demonstrate the kind of added information that we are requiring for inclusion.

1.2 History For a brief and personal history of cytometry and the cell cycle see ref. 5. Even more briefly, we recapitulate. The transition from autoradiography (see Note 1) to DNA content measurements by flow cytometry was a transition from laborious, time-consuming (days/weeks) effort between experiment and data to measure- ments, data, and analysis within minutes. Multiparametric work began with microfluorimetry of DNA and RNA in cells stained with acridine orange (AO), developed by Darzynkiewicz and col- leagues, that produced a set of protocols that segment the cell cycle into seven compartments (G0, G0T, G1A, G1B, S, G2, and M). The discovery of AO and Hoechst dye fluorescence quenching by DNA with halogenated pyrimidines and later the ability to detect the halogenated epitopes with monoclonal antibodies rendered cell kinetics (measuring phase transition times) through easily imple- mented, robust, relatively low-labor procedures yielding precise data. These assays were necessarily bivariate with anti-BrdU or anti-IdU coupled with DNA content. More generally, Jacobberger et al. [6] and Clevenger et al. [7] introduced high-quality intracel- lular antigen detection by flow cytometry, dependent on monoclo- nal antibodies, and coupled with DNA content, which led to multiparametric cell cycle analyses of the type presented here (see Note 2). Using similar approaches, the Darzynkiewicz group even- tually defined 8-cell cycle compartments or states (see Note 3) based on multiple bivariate analyses of various cyclin (see Note 4) expression patterns. Using the staining scheme we present here, which uses five parameters, at least 13 states can be defined in a single assay with an unprecedented ability to measure the frequency of cells within states that exist for minutes (e.g., late mitosis) or hours (e.g., G1, S). The supporting references for unreferenced statements in this paragraph can be found in [5].

1.3 Rationale

When we started this work, it was driven by an interest in complex cytometric data and a desire to explore the information that could be derived from the complex patterns of clustered events (see Note 5). The cell cycle was a good system with which to address this interest because even though a proliferating cell population con- tains cells with different histories, the gene expression programs within those histories are repeated with approximately the same dynamics, and thus from a data point of view, data visualizations for each historical cycle can be represented by a single unified, generic pattern (e.g., see Fig. 1 and [10]). Compared to a system where changes occur in successive periods, analytical complexity is simpli- fied for the cell cycle. We worked on this problem from 1993 through 2012. The journey has been rewarding in that we now have the backbone of an extensible system (presented here) that “solves” these complex patterns in terms of correlated simple single-parameter expression profiles over one cell cycle period [11, 12]. This opens up multiparameter cell cycle analysis to a number of parameters limited only by the available antibodies (and cell numbers) using multiple samples and mathematically correlating the results through cell cycle time. 1 Bivariate cell cycle expression patterns and terms. (a) Pattern for an epitope that is not expressed in early G1 then expressed exponentially through the cell cycle at a constant rate until M, then degraded in M back to background levels. Red lines have been placed where the pattern can be segmented meaningfully: (1) the boundary between non-expression and the beginning of expression in G1, (2) the G1/S boundary, (3) the S/ G2 boundary, and (4) the first detectable cells in which the epitope decreases through degradation. This hypothetical pattern can be observed when cyclin B1 is measured [8]. The gray dotted gridlines are at 1 (background), 50, 100. Degradation begins at 100. (b) Expression pattern for a “housekeeping” epitope— expressed from a “steady state” level of 50 at the beginning of G1 to 100 just before cell division. Red lines are drawn at (1) an objective “early” G1 (e.g., G1A for rRNA [9]) where expression is less than that expressed in late G1/early S, and at (2) G1/S and (3) S/G2. These patterns were generated by creating an expression profile with 2000 successive calculated (synthetic) time points that were then replaced by a value drawn from a normal Gaussian distribution with mean equal to the profile value and a standard deviation calculated from a coefficient of variation of 10%. Microsoft Excel and GraphPad Prism 7.0 were used. The terms “Cluster” and “Transitions” point to regions where data accumulate (Clusters), which reflect periods wherein at least one of the parameters is expressed at equilibrium levels (synthesis and degradation are balanced), or where data reflect periods of net synthesis or degradation in one or both parameters (Transitions).

1.3.1 The Relationship Between Parameters and Information (I)

Our effort was entirely exploratory with a single idea—that more parameters equate to more information. Of course, this is true in a trivial sense (see Note 6) because correlating any measurement with the DNA content of an asynchronously cycling cell population delivers the cell cycle-related expression of that parameter in some detail and with precision. In a nontrivial sense, if a parameter is functionally cell cycle related—i.e., if it plays a role in organizing and regulating the rates of transition through various phases or states within the cell cycle, then (when co-measured with another functional variable) there should be a limited set of complex multi- variate patterns to which that “parameter vs. parameter” pattern belongs, and that pattern should produce information about the cell cycle of the cell population under study. This nontrivial infor- mation is a property of data clusters (“stable” states) and “transi- tion” regions between clusters (Fig. 1a). Relatively uninformative cell cycle-independent epitopes (see Note 7), e.g., “housekeeping class” epitopes, should belong to a more restricted pattern set, increasing approximately from ~1 to ~2 as a function of one cell cycle period (Fig. 1b). It is not hard to imagine that the expression of two “housekeeping” epitopes will be highly corre- lated and uninformative. On the other hand, two parameters that have regulating or organizing functions (i.e., cell cycle-related epi- topes (CCRE)) will most likely demonstrate phase/state-specific changes in expression and create a complex pattern when plotted versus each other, or more simply versus DNA content, which is a marker for an organizing activity. Plotting an unknown epitope versus DNA content can identify CCREs. In 1989, Jacobberger presented a sketch of hypothetical bivariate expression patterns (epitope vs. DNA) based on limited experience, limited literature, and imagination [13]. Eleven patterns were drawn with nine con- sidered informative. A “housekeeping” pattern and “negative con- trol” (autofluorescence + nonspecific probe binding) were considered uninformative. In that chapter, real data were shown that were similar to three of the hypothetical patterns. Since then, we have observed or noted from the literature bivariate patterns that essentially recapitulate the ideas behind all of those hypotheti- cal CCRE patterns. Figure 2 is an updated version of this work. This figure, based on synthesized hypothetical data for a cell popu- lation with a 22 h cell cycle time and expression, ranges from 50 to 100 or 0.1 to 100 arbitrary units and backgrounds of 10 to 20 units. Each bivariate plot shows DNA-phase specific expression for a set where the expressed parameter of newly divided cells is at half the level at the cytokinetic stage (top two rows) or phase- specific expression from and “off” state to “on” and back to “off” (bottom two rows). Eight of these 13 patterns have been observed. This figure shows that bivariate plots of parameters vs. DNA are informative for CCREs. It also suggests strongly that more para- meters will provide more substantive information. For example, the (top, left) pattern represents data synthesized from an expression profile for a parameter synthesized to a high level in G1 and then maintained at that level to the end of the cycle. However, it can only be labeled “G1” because we know the profile. A parameter that begins the cycle high, then degrades to half that level, then re- synthesizes back to the high level—all in G1, would look the same. This uncertainty can be resolved by adding one or more para- meters. As an example, Fig. 3 shows a plot of G1* vs. G1. Two definitive clusters are identified (ovals) and two transition regions that can only represent an increase and then a decrease in expression (this statement is predicated on having access to the G1 and G1* vs. DNA plots of Fig. 2).

This chapter will make this point again, explicitly, on mitotic states with real data (see Subheading 3.5 on Analysis). Thus, this section demonstrates that a bivariate analysis equates to more information than a univariate analysis. When one of the parameters is DNA content, the least information imparted is a quick guess at the phase specificity of expression and expression rates through those phases (G0/G1, S, G2 + M). Of course, the quest is that the pattern imparts more information than that. For example, the expression could be “unscheduled,” which could point to an abnormally regulated process in pre-cancer or cancerous cells (see Note 8) [14, 15]. Cell states: Multiparametric patterns of CCREs form linked clustered data that can be defined as cell cycle states and the cycle can be modeled as a series of these “finite” states. Figure 4 shows this idea, and it has been conceptually elucidated for mitosis by Pines and Roeder [21], which is essentially that a progression of mitotic states can be defined by the periodic activity of dominant regulatory enzymes. In our approach, cell states are defined by periods of correlated and uncorrelated dynamic expression. Although this view of the cell cycle is appealing (at least to us), in practice it has little use other than as a framework to give us a “short-hand” language with which to discuss various cells states. For example, prophase can be described as having two distinct states based on the expression of phospho-S10-histone H3 and cyclin A2 [1, 16]. This language could play a distinct role in, e.g., a sophisticated platform for measuring the action of drugs on the cell cycle by evaluating the cell frequencies within specific states as funcThe relationship between parameter number and cell states: We have started to ask the question whether the amount of information correlates with parameter number by counting cell states (unpub- lished), and so far we have not observed a simple relationship. For example, with the parameters DNA (peak signal), DNA (integrated signal), light scatter, cyclin A2, cyclin B1, and phospho-histone H3, we can define at least 15 distinct and easily measured cell compart- ments (Fig. 4). Hypothetically, adding a single additional
Relationship of cytometric data to a compartment or state model of the cell cycle. Upper left: DNA content data for asynchronous, exponentially growing cultures constitute a bi-modal histogram with three regions, a G1 peak, a G2 + M peak, and between the two, an S phase component that can be thought of as composed of a series of Gaussian distributions equivalent to the G1 and G2 peaks, with successive means and constant CV. The regions correspond to the three phases of the cell cycle, originally defined by DNA labeling and autoradiography. Lower left: DNA content vs. a mitotic marker resolve G2 + M into G2 and M. PHH3 labels all mitotic cells, from the earliest states of chromosome condensation in early prophase to telophase and cytokinetic cells [16, 17]; therefore, there is ~100% correlation between this marker and morphologically defined mitosis. Bivariate analysis commonly uses a mitotic region to separate M from interphase and DNA content analysis to estimate G1, S, and G2 fractions.

A more correct analysis would be to apply the same principles used in DNA content analysis to single-parameter PHH3 histograms gated on G2 + M cells, or even better—two parameter bivariate multiple Gaussian fitting to deconvolve the two-parameter histogram. A third option, and likely the best option would be to employ probability state modeling [18]. Right: Compartment analysis, using regions and Boolean logic as explained in this chapter, invokes a “Finite State”-like model and analysis rather than the classical phase model and analysis. In this analysis, the color-coded cell states are occupied by some frequency of cells, proportional to the time spent in the state, and rate constants that define the rates of transition from state to state [19]. There are at least two points places for uncertainty—the transition from G1a to G1b and the transition from G2 to P2 [20]. Cells have options to move in both directions, which complicates analysis and interpretation parameter, Ki67, would add only one additional state, early in the cycle (G0), and adding PCNA would be expected to increase that by only one more (see Note 9). Nevertheless, once we have defined an unambiguous path from the start to the end of the cell cycle (demarcated by cell division), we can analyze any additional param- eter as a cell cycle expression profile [11, 12], unless it is expressed or has significant changes in expression within 1 of our 15 back- bone compartments (Fig. 4).

1.3.2 The Relationship Between Parameter Number and Information (II)

Thus, at a minimum, one additional parameter equals one unit of additional information. It seems clear, given that the cell cycle is organized and regulated by thousands of genes, regulatory RNAs, protein/molecule interactions, and protein modifications, that the number of cell states and thus, the amount of additional informa- tion past the expression profile of each parameter will be limited— i.e., a good bit of the measurable units will be redundant from a correlated-data point-of-view. Beyond that, the most basic (or perhaps easiest to define) units of useful information constitute a normal/abnormal call. After that, we get into non-quantitative or hard-to-quantify territory. Nevertheless, we hope that we have made the point that the original poorly thought out idea that more parameters equal more information is a sound idea. Whether more parameters equal more useful information is still a valid question (see Subheading 1.4), although we hope that the examples in Figs. 2, 3, and 4 and the text of this chapter convince the reader to lean toward “yes.”

1.4 Current State of Multiparameter Cell Cycle Analysis

We have looked for evidence that this approach—measuring several parameters within a cell cycle context to generate complex, highly informative cell cycle information, is currently and routinely applied in any area of cell-based science, and we are empty handed. We could have missed rare, high-quality papers, but whether we have or not, it seems clear that “multiparameter cell cycle analysis” does not currently have the research momentum that we believed/ believe it could and should have. To support the above statement, we present Table 1, which shows cell cycle/cytometry-related publication data from a series of PubMed searches. Our purpose was to evaluate the impact of “multiparameter cell cycle analysis.” About 8% of “cell cycle [MeSH]” papers were classified by “cell cycle [MeSH]” AND “cytometry.” This is a reasonable impact for cytometry. Of these publications, 69% concern human or mouse cells. A search on (cell cycle [MeSH] & cytometry & animal & human & multiparameter) accounts for almost all “multiparameter,” cytometry, cell cycle papers. However, 234 papers are an under-representation. For example, for the work from our lab, 4 papers are returned by the search, whereas 18 papers qualify as multiparametric cell cycle studies. An examination of 144 papers from this search, covering the dates 1997 to the present, demonstrates that most are bivariate analyses and many included in this list are of the type that detect cell type and then perform univariate cell cycle analysis. Further, 43 papers (17%) are from the Darzynkiewicz group, and most of the technologically appealing papers present the technology rather than new biological information. Therefore, within some limits, Table 1 (and an examination of the papers that are detected) illustrates that “multiparameter cell cycle analysis” is a niche area, included in a very small fraction of cell cycle studies, mostly about human cells. Figure 5 puts the niche idea within a historical context and suggests that interest in this niche area peaked some time ago. This state-of-affair exists, ironically, when the tools to develop these assays have never been more extensive and powerful. The tools are instruments, probes, and analytical approaches. Instruments with unprecedented multiparameter capability are commercially avail- able; monoclonal antibodies of high specificity to a large number of protein or function-identifying epitopes are widely available; using fluorescent proteins tags of multiple colors is commonplace; and finally, cytometric analysis and access to software packages continues to grow. On the plus side, this search was performed in Feb 2016 and updated in Mar 2017. There was a 12% increase in the number of papers satisfying the search “Cell Cycle AND Cyto- metry AND Multiparameter” (Row 20, Table 1).

1.5 Future of Multiparameter Cell Cycle Analysis

The areas where cytometry-based multiparametric cell cycle analysis should prove useful are: (1) evaluation of pre-cancer/cancer versus non-cancer cells, (2) healthy versus non-healthy systems, e.g., the immune system or hematopoiesis, (3) mathematical modeling of the cell cycle, (4) drug development, and (5) pharmacodynamic monitoring of patients in clinical trials and medical practice.

1.6 Common Research Objectives for “Cell Cycle Analysis”

There are five major reasons for cell cycle analysis by cytometry. These are to quantify: (1) the fraction of proliferating cells, (2) the histories of proliferating cells, (3) cell cycle phase or state fractions, (4) phase/state transit times, and (5) the cell cycle-related expres- sion of parameters. In the same order, examples of each modality are: (1) stimulated lymphocyte assays [22, 23] and proliferation antigen detection [24, 25], (2) dye dilution analysis [26], (3) S phase fraction analysis of tumors and hematopoietic malignancies [27, 28] and DNA/cyclin/mitotic marker studies [1, 17, 25, 29–32], (4) cell kinetic studies [33–35], and (5) SV40 large T antigen, p53, cyclin, and modified histone expression studies [11, 12, 36–38]. The latter category reaches its full potential in work from us and collaborators in which parameter expression drives cell cycle models [39, 40].

1.7 Methodology
1.7.1 DNA Content Coupled with Antibodies

Simple cell cycle analysis consists of DNA content measurements. By itself, this is insufficient for measuring the fraction of proliferat- ing cells or complete phase fraction analysis. Quiescent or G0 cells have the same DNA content as G1 cells, and, when assaying cell populations that contain non-proliferating and proliferating cells, the best that can be achieved by DNA content measurements is to obtain a proliferation index by quantifying the S phase or S + G2 + M phase fractions. A better assay is to label cells by continuous incubation with halogenated pyrimidines over some time period beyond the expected length of the G1 phase (thus, all cycling cells will be labeled), and then detect the labeled cells with antibodies to substituted DNA in cells that are co-stained for DNA content. This does not solve problems within the 4C (C, the genome complement) fraction that can be composed of G2, M, and endoreduplicating or binucleate G1 cells from a subpopulation cycling from 4C 8C. These fractions can be resolved by the methods presented in this chapter, which are also antigen plus DNA content measurements.

1.7.2 Limitations of Antibodies

The limitations of single-parameter DNA content measurements were overcome over a 20-year period from 1985 through 2005. During this time, the value of monoclonal antibodies as specific, quantitative probes for epitope expression in fixed cells and tissues became clear. And with these probes, the entire cell cycle could be quantitatively evaluated and additional new cell states were defined. Epitope availability: The entire approach rests on one caveat that is difficult to validate—that is, that the probe detects an unbi- ased fraction of the epitope. Antibody assays are always performed as a function of something else. If the fraction of epitope that is available for reaction with the antibody is changing as a function of “something else,” then the assay is subject to misinterpretation, unless that changing availability can be evaluated or measured. In lieu of direct validation, the caveat is supported by the large body of work that leads to the same answers whether the assay is done in tissue, whole cells, fractionated cells, or cell lysates. The level of possible “masking” (a description of unavailable epitope) follows the order tissue > whole cells > lysates; therefore, the agreement between, e.g., histology, cytometry, and western blots is a relatively powerful validation. We have done several correlative studies between cytometry and western blotting [37, 38, 41].

Antibody specificity and affinity: After the caveat of uniform epitope availability, the next weakness of the approach is the varia- bility in antibody specificity and affinity. Before cytometric or microscopic assays can be relied on, the antibodies involved need to be validated. There are many published examples wherein this was not done. For a critical analysis of one group of widely used antibodies to the p53 protein, see Bonsing et al. [42]. Antibody validation: Methods for validating an antibody are not standard and not established. As we have stated previously [1], many commercial antibody catalog sites still do not present validat- ing data or the validating data is of incomplete or poor quality. For the companies that do, the data are rarely if ever quantitative (see Note 10), and therefore, almost always visual, always anecdotal, and never rigorous (means and standard deviations), and not uni- versally applied—i.e., they sell some with and some without evi- dence of validation. Perhaps a common method of screening monoclonal antibodies is by ELISA using the purified antigen. This may be fine for producing antibodies that work for immuno- blotting electrophoretic gels that get around cross reactions via peptide fractionation, but for cell-based assays in which specific and nonspecific reactions are either integrated (as in flow cytome- try) or only crudely differentiable (i.e., by low-resolution localiza- tion), it is not sufficient for validation. In cytometric assays, the quality of a reaction is defined by the antigen-antibody avidity versus nonspecific binding and cross-reactive binding to other cell constituents. It is avidity that matters because the increased proba- bility that one or the other antigen combining site will be bound significantly reduces the effective antibody off rate, even though most of the antibody is bound monovalently at any one time. Because each sample type comes with its own set of unique poten- tial background issues, especially for heterogeneous samples like blood and bone marrow, antibodies should be validated and stain- ing should be optimized with these samples. Obviously, this level of assay-specific validation and optimization would be difficult and costly for commercial enterprises; therefore, it is left to the investi- gator to validate antibodies when using them in previously untested circumstances.

Our approach to validation is the following. First, we preferen- tially choose antibodies from companies that we trust and that provide some evidence for specificity. Generally, at a minimum, that means a western blot of whole cell lysates using positive and negative cells with the full molecular weight range displayed. If the epitope is localized in the cell, and/or modulated by drug treat- ment, evidence of correct localization and/or modulation will lead us to choose one antibody over another. Second, after choosing an antibody, we perform both immuno-blot and flow cytometric titers with negative and positive cell samples. If a negative cell source is unavailable, we default to a negative control with an isotype control and/or secondary antibody (see Note 11). We design titrations with sufficient data to generate a curve so that we can evaluate signal to noise (e.g., see ref. 43, 44). Third, we try to obtain a biological test—siRNA knockdown, gene transfer, virus infection, cytokine stimulation, drug treatment, etc. Fourth, we either per- form fluorescence microscopy or laser scanning cytometry [45] to make some check on localization. The working concentrations are defined by the cytometric titration. If our endpoint analysis is laser scanning cytometry, then we re-titer by twofold dilutions around the concentration determined by flow cytometry. This is because we work with higher staining volumes, and volume matters to the signal-to-noise ratio [46].

1.7.3 Fixation

Two general fixation classes: Unlike cell surface immunophenotyp- ing, most (if not all) of the interesting epitopes for cell cycle analysis are inside the cell. This means cells have to be stabilized (fixed)— proteases, nucleases, transporters, channels, and other active mole- cules need to be inactivated, and the cell needs to be made perme- able to large molecules. This has been reviewed extensively [13, 47–55]. Briefly, there are two basic modalities. The first uses dena- turation and begins with formaldehyde fixation sufficient to stabi- lize cells, which is followed by alcoholic dehydration. For epitopes that are sensitive to formaldehyde, the formaldehyde step can be omitted because the dehydration process denatures and inactivates all the activities we worry about. The second modality is non- denaturing, using formaldehyde followed by nonionic or zwitter- ionic detergent, or saponins. In this case, the formaldehyde is not dispensable because it also serves to cross-link molecules, creating a matrix through which large molecules diffuse slowly—thus, allow- ing staining and measurement of even soluble epitopes. Nothing used in this latter process efficiently denatures large molecules, and therefore, this is an approximately native state system. In both processes, some molecules are extracted—either completely or par- tially, and some are displaced (e.g., see ref. 7). If the target epitope is involved in tight binding to other molecules, denaturation may be required to “unmask” the epitope (e.g., phospho-Y694-Stat5 requires alcohol fixation [43, 56, 57]). If enough formaldehyde is used for a long enough time, it is possible to make penetration of antibodies difficult, and it is possible to promote “masking” relative to cells fixed with less formaldehyde. In all of this, there are many variables that can be adjusted. For example, different salts can be used to differentially extract molecules during the fixation step (e.g., see ref. 58). Another example is to permeabilize with deter- gent first and then fix with formaldehyde and alcohol [25]. This removes loosely bound proteins and other molecules. Use of the latter protocol gives a pattern of staining of PCNA versus DNA that identifies S phase better than DNA staining alone. Two studies using this approach are the exploration of new cell cycle states defined by correlated analysis of Ki-67 and PCNA [59] and corre- lated analysis of Mcm-6 and PCNA [60].

Recent advances in fixation: In recent years, there have been two notable advances in the development of fixation and permea- bilization methods. The first is described in a paper by Chow et al. [61] that identifies an alcoholic denaturing procedure that will work on whole blood or bone marrow, leaving light scatter patterns and surface staining intact enough to identify subpopulations by standard immunophenotyping procedures. The second has a simi- lar goal and is described in a patent awarded to Keith Shults and uses heat as the denaturant [62, 63]. A third effort, while not fixation/permeabilization development per se, is valuable. Krutzik and Nolan did a careful analysis of several fixation/permeabilization variants and arrived at formaldehyde followed by methanol as the overall best general approach to phospho-epitopes [64].

1.7.4 Recent Analytical Advances

There is a group of recent papers that deconvolve the complex clustered patterns that one encounters in multiparametric cell cycle analyses [11, 12, 65]. These papers present methods to derive a set of single-parameter “expression over cell cycle time” profiles from multiparametric cell cycle data that completely describe the complex multi-variate histogram patterns that represent all views of the data. Thus, the complex data patterns, arising from the corre- lated nature of cytometric single-cell data, are completely solved by these approaches. There are two features of this work that are critically important. The first is that the expression profiles are exactly equivalent to the outputs of cell cycle mathematical models based on systems of ordinary differential equations, and therefore presents an almost ideal platform to guide modeling [40]. And second, by deconvolving a set of multiparametric analyses, per- formed on aliquots of the same cell population, the expression relationship over the cell cycle between two epitopes that have not been measured together can be determined. This means that an open ended study of a very large number (only limited by the number of available cells, the number of available antibodies, dol- lars, manpower, and time) can be performed, potentially enabling development of highly complex mathematical models.

1.8 List of Key Publications

Table 2 lists a set of papers that anyone new to multiparametric cell cycle analysis should read. There are many, many papers and these were chosen by one of us (jwj) because they cover the major relevant areas. We tried to list only original research papers, but some areas are too broad and cannot be covered with a single research paper, and therefore reviews are listed. However, these are reviews with liberal presentations of data to present ideas or methods. We did not try to cite the first papers to report on a specific subject. Rather, these represent well-written, clearly pre- sented work on the subject. Two of the subjects need some expla- nation. The first is paraffin-embedded tissue. This is a large area with a large number of papers, most of which are single-parameter DNA analyses. The single listed paper presented an advance that substituted collagenase for pepsin, was carefully done, used two parameters (although the cell cycle analysis was single parameter), and is a good introduction to the subject. The second area is “Drugs.” This is also a very large and complex area and many more papers could have been listed. The paper by Kurose et al. [82] presents some signaling data, reflecting DNA repair pathway activities, coupled with DNA content as a basic means of examining drugs that affect the cell cycle by inducing DNA damage, invoking checkpoints, and leading to outcomes that include apoptosis and survival. It is rather easy to think of more complexity from an analytical point of view, but is rather hard to implement. This paper is a good introduction to this complex, and in our opinion underworked, area. This list is a good start, and following the research groups in the list and the papers they cite would constitute a thorough education in multiparametric cell cycle analysis.

2 Materials

In the following sections, a single complete protocol (including antibodies) will be presented to completely resolve the major cell cycle phases, compartments, and states. Analytical steps both dur- ing and after data acquisition are also included. A major alternative is to use detergent permeabilization before or after fixation with formaldehyde. These alternatives are available in the referenced literature. This chapter focuses on cell cycle phase/compartment/ state-analysis, which is the difficult part to master.

2.1 Cell Culture 1. MOLT-4 cells (ACC 362 from DSMZ, Braunschweig, Germany; or CRL-1582 from ATCC, Manassas, VA) is an easily grown human T-cell leukemia line with a single-stem line, low 4C 8C subpopulation (see Note 12), and published records of cyclins E, A2, and B1 expression consistent with wild-type, healthy-cell cyclin regulation [14, 15]. Cells are grown in suspension. K562 cells (ACC 10, DSMZ; CCL-243, ATCC) are a human Chronic Myelogenous Leukemia (CML) cell line

2.2 Biochemicals and Reagents

1. 16% Formaldehyde, methanol-free, Ultra Pure (Polysciences, Inc., Warrington, PA) (see Note 14).
2. Methanol (MeOH), spectrophotometric grade, >99%.
3. Phosphate-buffered saline (PBS): 150 mM NaCl, 10 mM phosphate, Na counter ion, pH 7.4. Prepare solution in deio- nized water and pass through 0.2 μm filter.
4. PBS-BSA: 2% (wt/vol) bovine serum albumin—Fraction V in PBS.
5. Antibodies reactive with: phospho-S10-histone H3-A488 (A488 Alexa Fluor 488; #9708; Cell Signaling Technology, Waverly, MA; see Note 15), cyclin A2-PE (PE Phycoerythrin; Beckman Coulter, Brea, CA; see Note 16); cyclin B1-A647 (clone GNS1, #554176; A647 Alexa Fluor 647; BD Bios- ciences, San Jose, CA; see Note 16).
6. 40,6-Diamidino-2-phenylindole (DAPI) solution: Prepare
1 mg/mL stock solution in double distilled water. Dilute working solution to 1 μg/mL in PBS. 2.3 Labware 1. Adjustable pipettors and pipet tips (1–20 and 100–1000 μL);
glass or disposable pipettes (1, 5, 10, 25 mL).
2. Microfuge tubes: Snap-cap tubes for assays that will be pro- cessed within a few days to weeks. If fixed samples are to be stored for weeks to months, then tubes with rubber o-rings and screw-cap tops are required (otherwise, the samples will dry out). Tubes should be polypropylene.
3. Tissue culture dishes, flasks, or multi-well plates: Any size or type provided they hold more than 2 mL of media.
2.4 Instruments 1. Humidified CO2 incubator.
2. Biosafety hood Class II.
3. Electrical impedance particle counting instrument (Beckman Coulter) or hemocytometer (Hausser Scientific, Horsham, PA). We use a Sceptor™ (Millipore, Billerica, MA).
4. Phase contrast inverted microscope with 10 , 20 , and 40 lenses (Olympus, Nikon, Leica, or Zeiss).
5. Microfuge: Variable speed, swinging bucket rotor, set at a low speed (see Note 17).
6. Suction device for removing supernatants. We use house vac- uum hooked to a side arm flask with pasteur pipette hooked to rubber tubing hooked to a glass tube through a cork in the top of the side arm flask.
7. Flow cytometer with ultraviolet (UV) and/or violet, blue, and red lasers at a minimum, but add a green or yellow laser if you can, it will make the separation of A488 and fluorescein iso- thiocyanate (FITC) signals from PE very easy. The data pre- sented in Fig. 6 were acquired on a LSR II (BD Biosciences) equipped with UV (355 nm), violet (405 nm), blue (488 nm), and red (633 nm) lasers. The data presented in Figs. 7, 8, 9, 10, 11, 12, 13 and 14 were acquired on an Attune NxT (Thermo Life Sciences/Invitrogen, Carlsbad, CA) equipped with violet (405 nm), blue (488 nm), yellow (561 nm), and red (637 nm) lasers. The filters were set up to detect DAPI, A488, PE, and A647.

3 Methods

3.1 Culture and Cell Preparation

1. In the examples, cells were growing exponentially as a suspen- sion. MOLT-4 cells do not adhere to the dish, but many hematopoietic cells lines do adhere loosely. Gentle pipetting can be used to remove these cells (see Note 18).
2. To obtain an exponential culture, MOLT-4 or K562 cells should be serially passaged with splits of 1–2 every 2–3 days if cells are approaching their upper density of 1–2 106/mL. Cultures should be split before reaching 2 106/mL; other- wise, cells will begin to die as the culture becomes dense.
3. Examine with a phase contrast-inverted microscope. Cells should be free floating single cells. To obtain an even cell suspension, repeatedly pipette the culture. Pipet cells into a large test tube (15 or 50 mL) for cell counts (necessary if pooling flasks or dishes for large numbers of cells) and/or volume adjustments (see step 4).
4. Count the cells. Adjust cell concentration to 2 106/mL either by adding media or centrifuge to concentrate and then add media for correct density.
5. Dispense aliquots of 2 106 cells to 1.5 mL microfuge tubes. This works for antigens that are not labile during processing. If working with labile epitopes (e.g., some phospho epitopes), then formaldehyde can be added directly to the tissue culture

3.2 Fixation 1. Move to cold room. All the lab procedures after cells are
removed from the incubator and counted are in a cold room at 4 ◦C. Pipettors, pipet tips, PBS, PBS-BSA reside at 4 ◦C in the cold room (see Note 19). 2. Pellet (centrifuge for 30 s to 1 min) and then wash with 1 mL of PBS. Resuspend in 50 μL of PBS (see Note 20).
3. Add 450 μL of MeOH (stored at —20 ◦C). At this point samples can be stored at —20 to —80 ◦C (see Note 21).3.3 Staining 1. Centrifuge cells from fixative and aspirate supernatant. Wash
cells twice with 1 mL of PBS and then wash once with 0.5 mL of PBS-BSA. The second, PBS-BSA wash is to begin the block- ing process.
2. Resuspend pellet in 50 μL of PBS-BSA containing 0.125 μg anti-cyclin A2-PE, 0.06 μg anti-cyclin B1-A647, and 0.0125 μg anti-phospho-S10-histone H3-A488 (see Note 22).
3. Incubate at 37 ◦C for 90 min (see Note 23).
4. Cool to 4 ◦C and then wash three times at 15 min per wash (in cold room) with 0.5 mL of PBS-BSA (see Note 24).
5. Resuspend in 0.5 mL of DAPI solution.

4 Notes

1. Photographic film exposed to 3H–thymidine labeled cells led to the modern cell cycle model with four phases—Gap1 (G1), S (synthesis), Gap2 (G2), and M.
2. At the time, the two sources of immunological probes were anti-sera and relatively new monoclonal antibodies. See Jacob- berger et al. [6] for data relevant to the problems with anti-sera. The problems were: unknown levels of immunoglobulin in pre-immune serum, unknown, low fractions of specific anti- body in immune serum, and high nonspecific binding to cells fixed and permeabilized with organic solvents.
3. Compartments, phases, stages, and states are used interchange- ably in the literature. We will generally follow that pattern and use them interchangeably; however, multiparametric cell cycle analysis has a need to more precisely define the compartments defined in our parameter-based models. We will leave that problem for another time, but we will employ a not-too-strict hierarchy in terminology. “Compartment” is the most general term, used for any well-defined and measurable class of cells within a dynamic system. “Phase” will be used for the most commonly accepted terms to define the ordered periods within the mammalian cell cycle (interphase, G1, S, G2, and M). “Stage” will be restricted to morphologically defined mitotic classes (prophase, prometaphase, metaphase, telophase), and “state” will refer cell classes that are based on dominant enzyme/molecular complex activities. For example, in this chapter, we will define two states in S phase, imaginatively termed S1 and S2 based on the rates of accumulation of cyclin A2 relative to cyclin B1. We will also use some terms that in our view have no broader classification: e.g., G0, mitosis, and cyto- kinetic cells—the first defined by an absence of a cell cycle program and the second defined by a process. However, the “phases” M and CK are synonomous with mitosis and cytokinesis.
4. Cyclin class of cell cycle regulating genes/proteins with cyclins D1, D2, D3, E1, E2, A1, A2, B1, B2, and B3 being the most studied mammalian cyclins.
5. That is, we did not have a hypothesis, but we did have an idea— not very original or profound, but it was based on the success- ful work of Darzynkiewicz and colleagues with acridine orange and our success using monoclonal antibodies reactive with intracellular antigens. The drive was to understand the mea- surement system through exploration rather than deep think- ing and hypothesis creation—a kind of bottom-up approach
rather than top-down, and the idea was the notion that more (parameters) is better.
6. That is, measuring the G1, S, G2 + M distribution of any parameter produces expression information about that param- eter and therefore at least one “unit” of information.
7. From this point, we will use “epitopes” to refer to anything measured as a function of DNA content because our most common mode of parameter detection is by antibody.
8. Unscheduled expression of cyclins has been described [14] and is defined as patterns of expression wherein the cyclin is expressed higher than it would be in normal cells relative to the next most earliest expression state. For example, high expression of cyclin B1 in G1 would be “unscheduled.” Villiard et al. went to great lengths to show that cyclin B1 was expressed in G1 cells of human T-cell lymphoma lines but either not expressed or expressed at a lower level in normal human T cells [84]. These levels were small but significant. We have measured cyclin B1 in G1 in all cells that we have examined with GNS1-A647. We have not observed what we would con- sider large variation in cyclin A2 and B1 expression patterns in human solid tumor cell lines, hTert-immortalized human cell lines, K562, MOLT-4, and normal human T cells. However, we have not tested the cell lines displaying unscheduled expres- sion presented by Gong et al. [14] for cyclin B1. The features of the MOLT-4 patterns shown here describe the cyclin A2 pat- tern relatively universally in our experience. For cyclin B1, we see essentially two patterns, one with higher G1 expression coupled with larger variance in the early S phase cells in one pattern relative to the other. The first pattern is typified by K562 cells and normal T cells, and the second pattern is typified by MOLT-4 cells. For some examples of the magnitude of expression differences for cyclin B1 for interphase, see Frisa and Jacobberger [38]. For both of these proteins, there is variation in expression to be sure, but these features: (1) cyclin B1 expression starting before cyclin A2, (2) ~linear synthesis of cyclin A2 and nonlinear synthesis of cyclin B1, and (3) degra- dation of cyclin B1 after cyclin A2, seems to be universal in our experience. In view of our experience, the unscheduled expres- sion of cyclin B1 appears to be largely confined to the differing overall levels of expression and different levels of expression in G1 (and perhaps differing fractions of cells expressing at the higher levels) rather than different patterns of expression. We have not explored unscheduled expression under conditions of growth imbalance, which has been studied by Gong et al. [15] for MOLT-4 cells. In this case, each of the cyclins D3, E, A, and B1 displayed unscheduled expression as defined above in cul- tures synchronized with mimosine or double thymidine block. The abnormalities in cyclin A expression were especially striking after mimosine treatment.
9. This is based on an expected offset for the onset of synthesis of both proteins. In this example, we have ignored possible addi- tional mitotic states based on possible differential degradation of the two proteins.
10. Single positive and negative cytometry histograms may repre- sent quantitative data, but they do not constitute a quantitative analysis.
11. With the advent of gene editing (e.g., Crispr/Cas9), settling for this faulty approach may soon be a thing of the past. It will be increasingly standard practice to create negative cell line controls from parental positive cell lines for validation. These same controls can serve as assay staining standards after the assay is completely developed. For many proteins/epitopes, such cells are currently commercially available for a reasonable cost (www.horizondiscovery.com).
12. The total 4C 8C cycling population for the culture repre- sented by the data set used in Fig. 6 was 0.7%, and the con- taminating fraction of 4C G1 and early S phase cells was ~2% of the stem line G2 + M fraction.
13. For MOLT-4 cells, DSMZ suggests 20% fetal bovine serum. These cells are derived from a leukemia and are not very sensi- tive to serum concentration and grow well in 10% FBS.
14. First, no one fixes cells with paraformaldehyde, which is an insoluble polymer of formaldehyde. For detailed discussions, search the Purdue University Cytometry Laboratories (PUCL) web site (http://www.cyto.purdue.edu/) on paraformalde- hyde. Second, this product comes sealed in glass under nitro- gen. Third, the reason to use this high-quality product is that it is supplied at a known concentration (formaldehyde produced from hydrolyzed paraformaldehyde produces a product with an unknown concentration); also, the “ultrapurity” helps to ensure that the least “autofluorescence” is induced in fixed cells.
15. We have used most of the phospho-S10-histone H3 antibodies (pH 3) from Cell Signaling Technology (CST). The unconju- gated rabbit polyclonal (#9701) works very well and provides flexibility for secondary reagents. The same is true for the conjugated versions (#9704, #9708). The mouse monoclonal (clone 6G3) works well but does not appear to be available in a conjugated form (previously, it was). The rabbit polyclonal was available in a biotinylated form, which was very useful when other rabbit antibodies were being used in the assay. The rabbit monoclonal is currently available in biotinylated form (#3642). When using the A488 version with anti-cyclin A2-PE, the spectral overlap of A488 to PE is problematic if a blue laser is used to excite PE. Despite compensation, this seriously degrades the cyclin A2 signal by increasing variance after com- pensation. It is useful to minimize this by titering the conju- gated form of PHH3 to a minimum acceptable signal and substituting unconjugated antibody to keep the concentration at optimum.
16. We previously obtained the cyclin A2-PE antibody as a gift from Vince Shankey at Beckman Coulter, but now purchase it from the same company. An equivalent cyclin A antibody is available from BD Biosciences (clone BF683). We have com- pared the two antibodies in unconjugated form and they per- form equally well. We obtained the GNS1 clone in a bulk form and conjugated it ourselves using the Alexa Fluor conjugating kits or the active dye from Thermo Life Sciences/Invitrogen (Carlsbad, CA).
17. We prefer a Fisher Scientific Micro-centrifuge (Model 59A). Ours finally stopped working after 26 years old. As far as we can tell, this model is no longer made. Some are available from used equipment suppliers and eBay.com. Beckman Coulter sells a 22R microfuge and an S241.5 swinging bucket rotor that look like they will work. Any microfuge or larger format table-top centrifuge can be used. The value of the 59A is that it can be set to spin at very low speed in increments of minutes, or manually turned on and off to spin for seconds. The combination of low speed and swing out rotor is that the cells are pelleted gently and are easily resuspended. This helps for cell recovery and reducing clumping, which is an advantage when working with alcohol-fixed cells.
18. Adherent cells can be trypsinized and pipetted to a single-cell suspension without affecting the cell cycle-related distribution of the cyclins. We have performed PAGE and western blotting on cells before and after trypsinization and observed no reason to be concerned about effects of trypsinization (unpublished).
19. The value of a cold room is that MeOH fixed cells stick to plastic surfaces and cold inhibits this. Working in the cold room is ideal (except for human comfort). If one is not available, this can be done working at room temperature or on ice with room temperature equipment.
20. This is for one to two million cells. We generally scale up if we go higher—e.g., ten million cells are resuspended in 250 μL of PBS and fixed (next step) with 2.25 mL of MeOH. We have not investigated what the scaling factor actually is.
21. Fixed cells in MeOH can be stored for years. However, there is loss of reactivity over time. We believe that loss of epitope reactivity and loss of masking come to equilibrium at about one half the available epitope. For some antigens we have checked, the patterns of expression are the same (e.g., see ref. 29). This is further evidence that the epitopes are exposed in an unbiased manner. They can be stored for days to weeks without concern. Storing at —80 ◦C will retard epitope loss.
22. Cell Signaling Technology does not print the antibody concen- tration or amount in most data sheets for their products. We phone to get the information. The way we optimize is to determine the antibody to target ratio—the target in this case being the amount of antigen in a positive cell line, averaged over two million cells and the antibody defined in micrograms. The goal is to react the cells and the antibody in the smallest feasible volume (highest concentration). This produces opti- mal staining in terms of signal-to-noise ratio [46]. When man- ufacturers sell us antibodies at dilute concentrations, they enforce reaction volumes that are not optimal.
23. A longer staining time (90 min) produces a better result (higher signal-to-noise ratio). For routine purposes, 30 min is sufficient to achieve good results. Nonspecific antibody dif- fuses into MeOH fixed cells to equilibrium in 5 min. Specific antibody achieves approximate equilibrium staining (reaches an asymptotic cusp) in 15 min [13, 52]. Staining can also be done at 4 ◦C or room temperature. The reason we stain at
37 ◦C is that antibodies develop in vivo at body temperature (39 ◦C for mice and rabbits).
24. Three washes are better than two, but two are sufficient for routine purposes.
25. See the chapter by Peter Rabinovitch or books by Alice Givan or Howard Shapiro for descriptions of how doublet discrimi- nation works [86–88]. Essentially, identify the G1 cells, then set a quadrilateral gate with sufficient width to enclose the G1 cells, widening it out as a function of intensity to ~2× the
width of the G1 population at 4C and ~4× at 8C, etc.
26. It is not necessary to orthogonalize the PHH3 vs. DNA data, especially with complex equations. Most investigators would be satisfied with a linear subtraction. However, this transform did improve the analysis and made it easier to set gates, and the data were more orthogonal than a linear subtraction.
27. For the kinetic argument for these properties, see ref. 17.
28. If doublets cannot be discriminated on PHH3 vs. DNA, then the following logic works: (R1 & R2) OR R4. This works because mitotic cells are rare and doublet mitotic cells are even rarer. This logic was used in analyses of the data presented in Figs. 6 and 17.
29. The figure shows two G2 gates (Fig. 10b). The first gate set by eye produced a value of 5.9% G2 cells. Modeling the DNA distribution (Fig. 10d) with ModFit LT (Verity Software House) produced a value of 5.3% G2 cells. The region (R9) was then adjusted to the second setting to produce 5.3% within R9. This same logic could not be applied to the G1 phase fraction that is either over-estimated by DNA content model- ing or under-estimated by cyclin A2 expression gating. We use cyclin A2 gating because cyclin A2 gating correlates with BrdU incorporation.
30. The separation of nuclei by peak vs. integral analysis is not likely to perfectly correlate with metaphase vs. telophase. The begin- ning of visual chromosome separation and 2 standard devia- tions below the high DNA-peak distribution mode may not align. Further, it is clear that some metaphase cells have depleted cyclin B1 and some telophase cells have residual cyclin B1, so there is not a perfect correlation at the biological level between the biomarkers and the mitotic stage. Integrating mitotic states based on biochemistry and stages based on mor- phology into a unified analysis would be an important advance in this area.

Acknowledgments

This work was supported by grants from NCI, R01CA73413 to JWJ and P30CA43703 to Stan Gerson, which supports the Cyto- metry and Imaging Microscopy Core facility in which the cytome- try is performed. Additional thanks go to Vince Shankey (retired), Chuck Goolsby (retired), David Hedley (Ontario Cancer Insti- tute), Stan Shackney (deceased), and Elena Holden (Compucyte) for advice and comments on this work for many years; Sue Chow (Ontario Cancer Institute) on fixation and staining, and Bruce Bagwell, Ben Hunsberger, and Chris Bray (Verity Software House) for many innovations in WinList and ModFit that enable or facilitate the analytical part of the work.

References

1. Jacobberger JW, Sramkoski RM, Stefan T (2011) Multiparameter cell cycle analysis. Methods Mol Biol 699:229–249. doi:10. 1007/978-1-61737-950-5_11
2. Patterson JO, Swaffer M, Filby A (2015) An imaging flow cytometry-based approach to analyse the fission yeast cell cycle in fixed cells. Methods 82:74–84. doi:10.1016/j.ymeth. 2015.04.026
3. Calvert ME, Lannigan JA, Pemberton LF (2008) Optimization of yeast cell cycle analysis and morphological characterization by multi- spectral imaging flow cytometry. Cytometry A 73(9):825–833. doi:10.1002/cyto.a.20609
4. Blasi T, Hennig H, Summers HD, Theis FJ, Cerveira J, Patterson JO, Davies D, Filby A, Carpenter AE, Rees P (2016) Label-free cell cycle analysis for high-throughput imaging flow cytometry. Nat Commun 7:10256. doi:10.1038/ncomms10256
5. Darzynkiewicz Z, Crissman H, Jacobberger JW (2004) Cytometry of the cell cycle: cycling through history. Cytometry A 58(1):21–32
6. Jacobberger JW, Fogleman D, Lehman JM (1986) Analysis of intracellular antigens by flow cytometry. Cytometry 7(4):356–364
7. Clevenger CV, Bauer KD, Epstein AL (1985) A method for simultaneous nuclear immunofluo- rescence and DNA content quantitation using monoclonal antibodies and flow cytometry. Cytometry 6(3):208–214
8. Sherwood SW, Rush DF, Kung AL, Schimke RT (1994) Cyclin B1 expression in HeLa S3 cells studied by flow cytometry. Exp Cell Res 211(2):275–281. doi:10.1006/excr.1994. 1087
9. Darzynkiewicz Z, Traganos F (1990) Multipa- rameter flow cytometry in studies of the cell cycle. In: Melamed MR, Lindmo T, Mendel- sohn ML (eds) Flow cytometry and sorting, 2nd edn. Wiley-Liss, Inc., New York, p 824
10. Filby A, Perucha E, Summers H, Rees P, Chana P, Heck S, Lord GM, Davies D (2011) An imaging flow cytometric method for measuring cell division history and molecular symmetry during mitosis. Cytometry A 79(7):496–506. doi:10.1002/cyto.a.21091
11. Avva J, Weis MC, Sramkoski RM, Sreenath SN, Jacobberger JW (2012) Dynamic expression profiles from static cytometry data: component fitting and conversion to relative, “same scale” values. PLoS One 7(7):e38275. doi:10.1371/ journal.pone.0038275
12. Jacobberger JW, Avva J, Sreenath SN, Weis MC, Stefan T (2012) Dynamic epitope expres- sion from static cytometry data: principles and reproducibility. PLoS One 7(2):e30870. doi:10.1371/journal.pone.0030870
13. Jacobberger JW (1989) Cell cycle expression of nuclear proteins. In: Yen A (ed) Flow cytome- try: advanced research and applications, vol 1. CRC Press, Boca Raton, pp 305–326
14. Gong J, Ardelt B, Traganos F, Darzynkiewicz Z (1994) Unscheduled expression of cyclin B1 and cyclin E in several leukemic and solid tumor cell lines. Cancer Res 54 (16):4285–4288
15. Gong J, Traganos F, Darzynkiewicz Z (1995) Growth imbalance and altered expression of cyclins B1, a, E, and D3 in MOLT-4 cells syn- chronized in the cell cycle by inhibitors of DNA replication. Cell Growth Differ 6 (11):1485–1493
16. Stefan T, Jacobberger JW (2011) Laser scan- ning cytometry of mitosis: state and stage anal- ysis. Methods Cell Biol 102:341–372. doi:10. 1016/B978-0-12-374912-3.00014-6
17. Jacobberger JW, Frisa PS, Sramkoski RM, Ste- fan T, Shults KE, Soni DV (2008) A new bio- marker for mitotic cells. Cytometry A 73 (1):5–15
18. Bagwell CB, Hunsberger BC, Herbert DJ, Munson ME, Hill BL, Bray CM, Preffer FI (2015) Probability state modeling theory. Cytometry A 87(7):646–660. doi:10.1002/ cyto.a.22687
19. Kafri R, Levy J, Ginzberg MB, Oh S, Lahav G, Kirschner MW (2013) Dynamics extracted from fixed cells reveal feedback linking cell growth to cell cycle. Nature 494 (7438):480–483. doi:10.1038/nature11897
20. Potapova TA, Sivakumar S, Flynn JN, Li R, Gorbsky GJ (2011) Mitotic progression becomes irreversible in prometaphase and col- lapses when Wee1 and Cdc25 are inhibited. Mol Biol Cell 22(8):1191–1206. doi:10. 1091/mbc.E10-07-0599
21. Pines J, Rieder CL (2001) Re-staging mitosis: a contemporary view of mitotic progression. Nat Cell Biol 3(1):E3–E6. doi:10.1038/ 35050676
22. Crissman HA, Tobey RA (1974) Cell-cycle analysis in 20 minutes. Science 184 (143):1297–1298
23. Fattorossi A, Battaglia A, Ferlini C (2001) Lymphocyte activation associated antigens. In: Darzynkiewicz Z, Crissman HA, Robinson JP (eds) Cytometry, Methods in cell biology, vol 63, 3rd Part A edn. Academic, San Diego, p 614
24. Endl E, Hollmann C, Gerdes J (2001) Anti- bodies against the Ki-67 protein: assessment of the growth fraction and tools for cell cycle analysis. In: Darzynkiewicz Z, Crissman HA, Robinson JP (eds) Cytometry, Methods in cell biology, vol 63, 3rd, Part A edn. Academic, San Diego, p 614
25. Larsen JK, Landberg G, Roos G (2001) Detec- tion of proliferating cell nuclear antigen. In: Darzynkiewicz Z, Crissman HA, Robinson JP (eds) Cytometry, Methods in cell biology, vol 63, 3rd, Part A edn. Academic, San Diego, p 614
26. Lyons AB, Hasbold J, Hodgkin PD (2001) Flow cytometric analysis of cell division history using dilution of carboxyfluoroscein diacetate succinimidyl ester, a stably integrated fluores- cent probe. In: Darzynkiewicz Z, Crissman HA, Robinson JP (eds) Cytometry, Methods in cell biology, vol 63, 3rd, Part A edn. Aca- demic, San Diego, p 614
27. Braylan RC, Duque RE, Hedley DW, Friedlan- der ML, Shankey TV, Bauer KD, Visscher DW, Crissman JD, Taylor SG, Shapiro DN, Look

TA, Beckmann E, Mazenet R, Weinberg DS (1993) Section C. Applications in Clinical Oncology. In: Bauer KD, Duque RE, Shankey TV (eds) Clinical Flow Cytometry. 1st edn. Williams & Wilkins, Baltimore, pp 203–372
28. Hedley DW, Shankey TV, Wheeless LL (1993) DNA cytometry consensus conference. Cyto- metry 14(5):471
29. Sramkoski RM, Wormsley SW, Bolton WE, Crumpler DC, Jacobberger JW (1999) Simul- taneous detection of cyclin B1, p105, and DNA content provides complete cell cycle phase fraction analysis of cells that endoredu- plicate. Cytometry 35(3):274–283
30. Soni DV, Sramkoski RM, Lam M, Stefan T, Jacobberger JW (2008) Cyclin B1 is rate limit- ing but not essential for mitotic entry and pro- gression in mammalian somatic cells. Cell Cycle 7(9):1285–1300. 5711 [pii]
31. Juan G, Li X, Darzynkiewicz Z (1997) Corre- lation between DNA replication and expression of cyclins a and B1 in individual MOLT-4 cells. Cancer Res 57(5):803–807
32. Huang Y, Sramkoski RM, Jacobberger JW (2013) The kinetics of G2 and M transitions regulated by B cyclins. PLoS One 8(12): e80861. doi:10.1371/journal.pone.0080861
33. Zhang D, Jacobberger JW (1996) TGF-beta 1 perturbation of the fibroblast cell cycle during exponential growth: switching between nega- tive and positive regulation. Cell Prolif 29 (6):289–307
34. Sladek TL, Jacobberger JW (1992) Simian virus 40 large T-antigen expression decreases the G1 and increases the G2 + M cell cycle phase durations in exponentially growing cells. J Virol 66(2):1059–1065
35. DiSalvo CV, Zhang D, Jacobberger JW (1995) Regulation of NIH-3T3 cell G1 phase transit by serum during exponential growth. Cell Pro- lif 28(9):511–524
36. Jacobberger JW, Sramkoski RM, Wormsley SB, Bolton WE (1999) Estimation of kinetic cell- cycle-related gene expression in G1 and G2 phases from immunofluorescence flow cytome- try data. Cytometry 35(3):284–289
37. Frisa PS, Lanford RE, Jacobberger JW (2000) Molecular quantification of cell cycle-related gene expression at the protein level. Cytometry 39(1):79–89
38. Frisa PS, Jacobberger JW (2009) Cell cycle- related cyclin B1 quantification. PLoS One 4 (9):e7064
39. Singhania R, Sramkoski RM, Jacobberger JW, Tyson JJ (2011) A hybrid model of mammalian cell cycle regulation. PLoS Comput Biol 7(2):

e1001077. doi:10.1371/journal.pcbi. 1001077
40. Weis MC, Avva J, Jacobberger JW, Sreenath SN (2014) A data-driven, mathematical model of mammalian cell cycle regulation. PLoS One 9 (5):e97130. doi:10.1371/journal.pone. 0097130
41. Frisa PS, Jacobberger JW (2002) Cell density related gene expression: SV40 large T antigen levels in immortalized astrocyte lines. BMC Cell Biol 3:10
42. Bonsing BA, Corver WE, Gorsira MC, van Vliet M, Oud PS, Cornelisse CJ, Fleuren GJ (1997) Specificity of seven monoclonal antibo- dies against p53 evaluated with western blot- ting, immunohistochemistry, confocal laser scanning microscopy, and flow cytometry. Cytometry 28(1):11–24
43. Jacobberger JW, Sramkoski RM, Frisa PS, Ye PP, Gottlieb MA, Hedley DW, Shankey TV, Smith BL, Paniagua M, Goolsby CL (2003) Immunoreactivity of Stat5 phosphorylated on tyrosine as a cell-based measure of Bcr/Abl kinase activity. Cytometry A 54(2):75–88
44. Jacobberger JW, Sramkoski RM, Zhang D, Zumstein LA, Doerksen LD, Merritt JA, Wright SA, Shults KE (1999) Bivariate analysis of the p53 pathway to evaluate ad-p53 gene therapy efficacy. Cytometry 38(5):201–213
45. Kamentsky LA, Kamentsky LD (1991) Microscope-based multiparameter laser scan- ning cytometer yielding data comparable to flow cytometry data. Cytometry 12 (5):381–387
46. Srivastava P, Sladek TL, Goodman MN, Jacob- berger JW (1992) Streptavidin-based quantita- tive staining of intracellular antigens for flow cytometric analysis. Cytometry 13
(7):711–721. doi:10.1002/cyto.990130707
47. Bauer KD, Jacobberger JW (1994) Analysis of intracellular proteins. Methods Cell Biol 41:351–376
48. Camplejohn RS (1994) The measurement of intracellular antigens and DNA by multipara- metric flow cytometry. J Microsc 176(Pt 1):1–7
49. Clevenger CV, Shankey TV (1993) Cytochem- istry II: immunofluorescence measurement of intracellular antigens. In: Bauer KD, Duque RE, Shankey TV (eds) Clinical flow cytometry, 1st edn. Williams & Wilkins, Baltimore, pp 157–175
50. Jacobberger JW (1991) Intracellular antigen staining: quantitative immunofluorescence. Methods 2:207–218
51. Jacobberger JW (2000) Flow cytometric analy- sis of intracellular protein epitopes. In: Stewart

CA, Nicholson JKA (eds) Immunophenotyp- ing. Cytometric cellular analysis. Wiley-Liss, Inc., New York, pp 361–405
52. Jacobberger JW (2001) Stoichiometry of immunocytochemical staining reactions. Methods Cell Biol 63:271–298
53. Jacobberger JW, Hedley DW (2001) Intracel- lular measures of signalling pathways. In: McCarthy DA, Macey MG (eds) Cytometric analysis of cell phenotype and function. Cam- bridge University Press, Cambridge
54. Koester SK, Bolton WE (2000) Intracellular markers. J Immunol Methods 243 (1–2):99–106
55. Koester SK, Bolton WE (2001) Strategies for cell permeabilization and fixation in detecting surface and intracellular antigens. Methods Cell Biol 63:253–268
56. Woost PG, Solchaga LA, Meyerson HJ, Shan- key TV, Goolsby CL, Jacobberger JW (2011) High-resolution kinetics of cytokine signaling in human CD34/CD117-positive cells in unfractionated bone marrow. Blood 117(15): e131–e141. doi:10.1182/blood-2010-10- 316224
57. Marvin J, Swaminathan S, Kraker G, Chadburn A, Jacobberger J, Goolsby C (2011) Normal bone marrow signal-transduction profiles: a requisite for enhanced detection of signaling dysregulations in AML. Blood 117(15): e120–e130. doi:10.1182/blood-2010-10- 316026
58. Bruno S, Gorczyca W, Darzynkiewicz Z (1992) Effect of ionic strength in immunocytochemi- cal detection of the proliferation associated nuclear antigens p120, PCNA, and the protein reacting with Ki-67 antibody. Cytometry 13 (5):496–501
59. Landberg G, Tan EM, Roos G (1990) Flow cytometric multiparameter analysis of prolifer- ating cell nuclear antigen/cyclin and Ki-67 antigen: a new view of the cell cycle. Exp Cell Res 187(1):111–118
60. Frisa PS, Jacobberger JW (2010) Cytometry of chromatin bound Mcm6 and PCNA identifies two states in G1 that are separated functionally by the G1 restriction point. BMC Cell Biol 11:26. doi:10.1186/1471-2121-11-26
61. Chow S, Hedley D, Grom P, Magari R, Jacob- berger JW, Shankey TV (2005) Whole blood fixation and permeabilization protocol with red blood cell lysis for flow cytometry of intracellu- lar phosphorylated epitopes in leukocyte sub- populations. Cytometry A 67(1):4–17
62. Shults KE, Flye LA (2008) Cell fixation and use in phospho-proteome screening. United States Patent

63. Shults KE, Flye LA, Green L, Daly T, Manro JR, Lahn M (2009) Patient-derived actute myleoid leukemia (AML) bone marrow cells display distinct intracellular kinase phosphory- lation patterns. J Cancer Manage Res 2009 (1):1–11
64. Krutzik PO, Nolan GP (2003) Intracellular phospho-protein staining techniques for flow cytometry: monitoring single cell signaling events. Cytometry A 55(2):61–70
65. Avva J, Weis MC, Soebiyanto RP, Jacobberger JW, Sreenath SN (2011) CytoSys: a tool for extracting cell-cycle-related expression dynam- ics from static data. Methods Mol Biol 717:171–193. doi:10.1007/978-1-61779- 024-9_10
66. Darzynkiewicz Z, Traganos F, Melamed MR (1980) New cell cycle compartments identified by multiparameter flow cytometry. Cytometry 1(2):98–108. doi:10.1002/cyto.990010203
67. Traganos F, Darzynkiewicz Z, Melamed MR (1982) The ratio of RNA to total nucleic acid content as a quantitative measure of unbal- anced cell growth. Cytometry 2(4):212–218. doi:10.1002/cyto.990020403
68. Mann GJ, Dyne M, Musgrove EA (1987) Immunofluorescent quantification of ribonu- cleotide reductase M1 subunit and correlation with DNA content by flow cytometry. Cyto- metry 8(5):509–517. doi:10.1002/cyto. 990080512
69. Landberg G, Roos G (1991) Antibodies to proliferating cell nuclear antigen as S-phase probes in flow cytometric cell cycle analysis. Cancer Res 51(17):4570–4574
70. Darzynkiewicz Z, Gong J, Juan G, Ardelt B, Traganos F (1996) Cytometry of cyclin pro- teins. Cytometry 25(1):1–13
71. Juan G, Traganos F, James WM, Ray JM, Roberge M, Sauve DM, Anderson H, Darzyn- kiewicz Z (1998) Histone H3 phosphorylation and expression of cyclins a and B1 measured in individual cells during their progression through G2 and mitosis. Cytometry 32 (2):71–77
72. Poot M, Schmitt H, Seyschab H, Koehler J, Chen U, Kaempf U, Kubbies M, Schindler D, Rabinovitch PS, Hoehn H (1989) Continuous bromodeoxyuridine labeling and bivariate ethi- dium bromide/Hoechst flow cytometry in cell kinetics. Cytometry 10(2):222–226. doi:10. 1002/cyto.990100215
73. Pinto M, Azzam EI, Howell RW (2006) Bystander responses in three-dimensional cul- tures containing radiolabelled and unlabelled human cells. Radiat Prot Dosim 122 (1–4):252–255. doi:10.1093/rpd/ncl460

74. White RA, Terry NH (1992) A quantitative method for evaluating bivariate flow cytometric data obtained using monoclonal antibodies to bromodeoxyuridine. Cytometry 13
(5):490–495. doi:10.1002/cyto.990130507
75. Terry NH, White RA, Meistrich ML, Calkins DP (1991) Evaluation of flow cytometric methods for determining population potential doubling times using cultured cells. Cytometry 12(3):234–241. doi:10.1002/cyto. 990120305
76. Landberg G, Roos G (1992) Flow cytometric analysis of proliferation associated nuclear anti- gens using washless staining of unfixed cells. Cytometry 13(3):230–240. doi:10.1002/ cyto.990130304
77. Darzynkiewicz Z, Zhao H, Zhang S, Lee MY, Lee EY, Zhang Z (2015) Initiation and termi- nation of DNA replication during S phase in relation to cyclins D1, E and a, p21WAF1, Cdt1 and the p12 subunit of DNA polymerase delta revealed in individual cells by cytometry. Oncotarget 6(14):11735–11750. doi:10. 18632/oncotarget.4149
78. Tomasoni D, Lupi M, Brikci FB, Ubezio P (2003) Timing the changes of cyclin E cell content in G1 in exponentially growing cells. Exp Cell Res 288(1):158–167
79. Yanagisawa M, Dolbeare F, Todoroki T, Gray JW (1985) Cell cycle analysis using numerical simulation of bivariate DNA/bromodeoxyuri- dine distributions. Cytometry 6(6):550–562. doi:10.1002/cyto.990060609
80. Li B, Zhao H, Rybak P, Dobrucki JW, Darzyn- kiewicz Z, Kimmel M (2014) Different rates of DNA replication at early versus late S-phase sections: multiscale modeling of stochastic events related to DNA content/EdU (5- ethynyl-2’deoxyuridine) incorporation distri- butions. Cytometry A 85(9):785–797. doi:10.1002/cyto.a.22484
81. Glogovac JK, Porter PL, Banker DE, Rabino- vitch PS (1996) Cytokeratin labeling of breast cancer cells extracted from paraffin-embedded tissue for bivariate flow cytometric analysis. Cytometry 24(3):260–267. doi:10.1002/( SICI)1097-0320(19960701)24:3<260::AID-
CYTO9>3.0.CO;2-L
82. Kurose A, Tanaka T, Huang X, Halicka HD, Traganos F, Dai W, Darzynkiewicz Z (2005) Assessment of ATM phosphorylation on Ser- 1981 induced by DNA topoisomerase I and II inhibitors in relation to Ser-139-histone H2AX phosphorylation, cell cycle phase, and apopto- sis. Cytometry A 68(1):1–9. doi:10.1002/ cyto.a.20186
83. Gerashchenko BI, Hino M, Hosoya H (2000) Enrichment for late-telophase cell populations using flow cytometry. Cytometry 41 (2):148–149
84. Viallard JF, Lacombe F, Dupouy M, Ferry H, Belloc F, Reiffers J (2000) Different expression profiles of human cyclin B1 in normal PHA- stimulated T lymphocytes and leukemic T cells. Cytometry 39(2):117–125
85. Bagwell CB (1993) Theoretical aspects of flow cytometry data analysis. In: Bauer KD, Duque RE, Shankey TV (eds) Clinical flow cytometry, 1st edn. Williams & Wilkins, Baltimore, pp 41–61
86. Givan AG (2001) Flow cytometry first princi- ples, 2nd edn. Wiley-Liss, Inc., New York
87. Rabinovitch PS (1993) Practical considerations for DNA content and MPI-0479605 cell cycle analysis. In: Bauer KD, Duque RE, Shankey TV (eds) Clin- ical flow cytometry, 1st edn. Williams & Wilk- ins, Baltimore, pp 117–142
88. Shapiro HM (2003) Practical flow cytometry, 4th edn. Wiley, Hoboken