New method for the analysis of flow cytometric data

A numerical method for deriving the fractions of cells in different phases of the cell cycle from a single observed DNA histogram is presented. The observed histogram is regarded as a polluted version (containing allocation errors) of the true histogram. A mathematical model is used to describe the pollution process. A theoretical histogram, representing the true histogram, is constructed so that G1 cells are put into one channel and G2M cells into another; the distribution of S cells in between is approximated with a set of harmonic functions. This theoretical histogram is subsequently disturbed with Gaussian dispersion functions to stimulate the pollution, yielding a predicted histogram. Using a maximum likelihood estimation technique, the model parameters are adjusted iteratively, matching the predicted histogram to the actually observed one. With the final parameter values substituted, the corresponding final theoretical histogram is regarded as a reliable reconstruction of the true histogram. From the latter, the required percentages can be read directly. The advantage of this approach over other mathematical analysis methods is that it allows a wide range of different, continuous distributions for relatively few model parameters (thus featuring flexibility and realism and a diminished risk of encountering computational problems). In addition, estimation errors providing a measure of accuracy can be obtained. To test the method, it was used to analyze various observed histograms from the literature that have been obtained by either simulation or actual flow cytometric measurements. The method appeared to perform well, as compared to the reported results of several other methods of analysis applied to the same data.     

A theory for analysis of cell populations with non-cycling S phase cells

Analyses of cell populations that have been labeled in vivo with analogs of thymidine that are incorporated by cells synthesizing DNA and then monitored over time by bivariate flow cytometry sometimes detect populations of cells that have S phase DNA content but that have not acquired label. Two alternative explanations for the lack of labeling are that either the cells were not exposed to the label or that the cells stopped DNA synthesis and ceased progression through S phase. To help determine which scenario is the more likely, a model has been devised for studying a population of cells that includes the possibility that cells in S phase will cease DNA synthesis. In this model, the initial fraction of unlabeled cells in S phase depends on two rates: the growth rate of the total population and the number of cells that cease progression through S phase per unit time. The model is used to analyze the changing quantities which can be measured by monitoring the population of cells over time and is used to estimate the two rates required to compute the initial fraction of unlabeled S phase cells. Thus, the initial fraction of unlabeled cells can be compared with that predicted by the population dynamics to determine whether one explanation for the failure of some cells to be labeled is preferable to the other, which in turn might offer information about tumor microvascular or cytologic properties.     

Modeling multicellular response to nonuniform distributions of radioactivity: differences in cellular response to self-dose and cross-dose

Radiopharmaceuticals are distributed nonuniformly in tissue. While distributions of radioactivity often appear uniform at the organ level, in fact, microscopic examination reveals that only a fraction of the cells in tissue are labeled. Labeled cells and unlabeled cells often receive different absorbed doses depending on the extent of the nonuniformity and the characteristics of the emitted radiations. The labeled cells receive an absorbed dose from radioactivity within the cell (self-dose) as well as an absorbed dose from radioactivity in surrounding labeled cells (cross-dose). Unlabeled cells receive only a cross-dose. In recent communications, a multicellular cluster model was used to investigate the lethality of microscopic nonuniform distributions of 131I iododeoxyuridine (131IdU). For a given mean absorbed dose to the tissue, the dose response depended on the percentage of cells that were labeled. Specifically, when 1, 10 and 100% of the cells were labeled, a D37 of 6.4, 5.7 and 4.5 Gy, respectively, was observed. The reason for these differences was recently traced to differences in the cellular response to the self- and cross-doses delivered by 131IdU. Systematic isolation of the effects of self-dose resulted in a D37 of 1.2 +/- 0.3 Gy. The cross-dose component yielded a D37 of 6.4 +/- 0.5 Gy. In the present work, the overall survival of multicellular clusters containing 1, 10 and 100% labeled cells is modeled using a semi-empirical approach that uses the mean lethal self- and cross-doses and the fraction of cells labeled. There is excellent agreement between the theoretical model and the experimental data when the surviving fraction is greater than 1%. Therefore, when the distribution of 131I in tissue is nonuniform at the microscopic level, and the cellular response to self- and cross-doses differs, multicellular dosimetry can be used successfully to predict biological response, whereas the mean absorbed dose fails in this regard.     

Free radical-initiated and gap junction-mediated bystander effect due to nonuniform distribution of incorporated radioactivity in a three-dimensional tissue culture model

To investigate the biological effects of nonuniform distribution of radioactivity in mammalian cells, we have developed a novel three-dimensional tissue culture model. Chinese hamster V79 cells were labeled with tritiated thymidine and mixed with unlabeled cells, and multicellular clusters (approximately 1.6 mm in diameter) were formed by gentle centrifugation. The short-range beta particles emitted by (3)H impart only self-irradiation of labeled cells without significant cross-irradiation of unlabeled bystander cells. The clusters were assembled in the absence or presence of 10% dimethyl sulfoxide (DMSO) and/or 100 microM lindane. DMSO is a hydroxyl radical scavenger, whereas lindane is an inhibitor of gap junctional intercellular communication. The clusters were maintained at 10.5 degrees C for 72 h to allow (3)H decays to accumulate and then dismantled, and the cells were plated for colony formation. When 100% of the cells were labeled, the surviving fraction was exponentially dependent on the mean level of radioactivity per labeled cell. A two-component exponential response was observed when either 50 or 10% of the cells were labeled. Though both DMSO and lindane significantly protected the unlabeled or bystander cells when 50 or 10% of the cells were labeled, the effect of lindane was greater than that of DMSO. In both cases, the combined treatment (DMSO + lindane) elicited maximum protection of the bystander cells. These results suggest that the bystander effects caused by nonuniform distributions of radioactivity are affected by the fraction of cells that are labeled. Furthermore, at least a part of these bystander effects are initiated by free radicals and are likely to be mediated by gap junctional intercellular communication.     

A dual fluorescence flow cytometric analysis of bacterial adherence to mammalian host cells

Flow cytometry has provided a powerful tool for analyzing bacteria-host cell associations. Established approaches have used bacteria, labeled either directly with fluorochromes or indirectly with fluorescently conjugated antibodies, to detect these associations. Although useful, these techniques are consistently unable to include all host cells in the analysis while excluding free, aggregated bacteria. This study describes a new flow cytometry method of assessing bacterial adherence to host cells based on direct fluorescent labeling of both bacteria and host cells. Eukaryotic host cells were labeled with PKH-26, a red fluorescent dye, and bacteria were labeled with fluorescein isothiocyanate, a green fluorescent dye. The red host cells were gated and the mean green fluorescence intensity (MFI) of these red cells was determined. We used MFI values obtained from control samples (unlabeled and labeled host cells with unlabeled bacteria) to eliminate contributions due to autofluorescence. The final MFI values represent fluorescence of host cells resulting from the adherent bacteria. Because all red fluorescent cells are analyzed, this method includes all the eukaryotic cells for analysis but excludes all free or aggregated bacteria that are not bound to target cells.     

Precision of DNA flow cytometry in inter-institutional analyses

A Bladder Cancer Flow Cytometry Network study has been carried out to further identify and quantify sources of inter- and intra-laboratory variability. Replicate samples containing four mixtures of peripheral blood lymphocytes and aneuploid cell lines were distributed together with reference standards to six laboratories. The samples were stained for DNA using propidium iodide, with each laboratory using its own staining protocol. Two of each of the four sample types and a reference standard were analyzed by each laboratory on 3 separate days to obtain cellular DNA distributions. DNA index (DI) and hyperdiploid fraction (HDF) were calculated for each histogram using an automated technique. The results showed significant inter- and intra-laboratory differences. Results were evaluated by a two-way analysis of variance to estimate components of the overall variation attributable to individual sources. Error variation was found to be the major component of random variation. Specimen means were also compared for each laboratory. No significant differences were noted in mean DI for similar specimens; however, agreement in HDF between similar specimens was lacking in most laboratories. Prediction intervals were computed to estimate the range of values expected for a single specimen based on the analysis of the previous six. Prediction intervals for DI were quite good while those for HDF were troublesome due to wide variation. The results of these studies indicate that intra- and inter-laboratory variability are high enough that results for a single sample may not be sufficiently precise to allow comparison to results obtained in other laboratories.(ABSTRACT TRUNCATED AT 250 WORDS)     

A method for the selective measurement of the radiosensitivity of quiescent cells in solid tumors--combination of immunofluorescence staining to BrdU and micronucleus assay.

C3H/He mice bearing the SCC VII tumor were irradiated after being given 10 injections of 5-bromo-2'-deoxyuridine (BrdU) to label all proliferating cells in the tumors, and the tumors were then excised and trypsinized. The tumor cell suspensions were incubated with cytochalasin-B (which blocks cytokinesis), and the micronucleus frequency in unlabeled cells was determined using immunofluorescence staining to BrdU. The micronucleus frequency was then used to calculate the surviving fraction of the unlabeled cells, using the regression line relating the micronucleus frequency to the surviving fraction determined separately for the total tumor cell population. Using this technique, a cell survival curve could be determined for the unlabeled cells, which were regarded as the quiescent cells. Assays performed both immediately after and 24 h after irradiation of normally-aerated tumors showed that unlabeled cells were more radioresistant and had a greater capacity for repair of potentially lethal damage than the tumor cell population as a whole. Moreover, when the assay was performed immediately after the irradiation of both normally-aerated and hypoxic tumors, it was found that unlabeled cells had a much higher hypoxic fraction than the tumor cell population as a whole. This appears to be a useful method for determining the responses of quiescent cells in solid tumors to various treatments.     

Two-color fluorescence analysis of individual virions determines the distribution of the copy number of proteins in herpes simplex virus particles

We present a single virion method to determine absolute distributions of copy number in the protein composition of viruses and apply it to herpes simplex virus type 1. Using two-color coincidence fluorescence spectroscopy, we determine the virion-to-virion variability in copy numbers of fluorescently labeled tegument and envelope proteins relative to a capsid protein by analyzing fluorescence intensity ratios for ensembles of individual dual-labeled virions and fitting the resulting histogram of ratios. Using EYFP-tagged capsid protein VP26 as a reference for fluorescence intensity, we are able to calculate the mean and also, for the first time to our knowledge, the variation in numbers of gD, VP16, and VP22 tegument. The measurement of the number of glycoprotein D molecules was in good agreement with independent measurements of average numbers of these glycoproteins in bulk virus preparations, validating the method. The accuracy, straightforward data processing, and high throughput of this technique make it widely applicable to the analysis of the molecular composition of large complexes in general, and it is particularly suited to providing insights into virus structure, assembly, and infectivity.     

I. THEORETICAL OUTLINE OF A NEW METHOD TO ANALYZE TIME SEQUENCES OF DNA HISTOGRAMS

A new mathematical method is presented to analyze a time sequence of DNA distributions taken from perturbed cell populations. The method, called FP_i analysis, consists of plotting the time variation of the fraction of cells in selected DNA contents ‘windows’ of the histogram. It is shown that kinetic information about the flow of cells through the cycle after the perturbation can be estimated from the FP_i curves. An analysis of the method reveals that the method yields accurate results for the instrumental and cytochemical variations obtainable by present technology. The value of the method lies in the fact that the information needed can be obtained directly from the measured DNA distribution, thus bypassing the problems with other methods which estimate the fraction of cells in a given phase directly from a single histogram.     

Comparison of automated and manual techniques for analysis of DNA frequency distributions in bladder washings

Quantitative methods for interpretation of flow cytometry DNA histograms are required for the widespread clinical use of this technology. The usefulness of a histogram analysis technique in this setting requires that it be operator independent, easy to implement in a clinical laboratory, and provide high sensitivity to the desired information. Additionally, the technique must be tolerant of the relatively low signal-to-noise ratios often found in DNA distributions obtained from clinical samples. Among the factors that have been used to assess the malignant potential of tumors are the presence of an aneuploid population, the proportion of hyperdiploid cells, the width of the G1 peak, the DNA index, and the fraction of cells in S. A computer-based method has been developed for extraction of the above-mentioned features from DNA histograms. The program detects peaks in the histogram and uses straight-line fits to the cumulative frequency distribution to define cell population bounds. A test set of 44 histograms compiled from bladder irrigation specimens obtained from patients with a present or past history of transitional cell carcinoma (TCC) was analyzed by five collaborating laboratories forming a Network sponsored by the National Cancer Institute (NCI). This test set was used to evaluate the performance of the computer-based method by comparing results with those of four expert observers. In this preliminary analysis, perfect agreement was found in the detection of aneuploid cell populations by all observers and the computer-based method. Correlation of percent hyperdiploid cell fraction was also excellent.(ABSTRACT TRUNCATED AT 250 WORDS)     

Molecular analysis of thymopentin binding to HLA-DR molecules

Thymopentin (TP5) triggers an immune response by contacting with T cells; however the molecular basis of how TP5 achieves this process remains incompletely understood. According to the main idea of immunomodulation, we suppose that it would be necessary for TP5 to form complex with human class II major histocompatibility complex DR molecules (HLA-DR) before TP5 interacts with T cells. The uptake of TP5 by EBV-transformed B cells expressing HLA-DR molecules and the histogram of fluorescence intensities were observed by using fluorescent- labeled TP5, testifying the direct binding of TP5 to HLA-DR. The binding specificity was confirmed by the inhibition with unlabeled TP5, suggesting the recognition of TP5 by HLA-DR. To confirm the interaction between TP5 and HLA-DR, the complex formation was predicted by using various modeling strategies including six groups of trials with different parameters, alanine substitutions of TP5, and the mutants of HLA-DR. The results demonstrated that TP5 and its alanine substitutions assumed distinct conformations when they bound to HLA-DR. The observation further showed that there was flexibility in how the peptide bound within the binding cleft. Also, the molecular analysis supplemented a newly important discovery to the effect of Val anchor on TP5 binding HLA-DR, and revealed the important effects of Glu11 and Asn62 on the recognition of TP5. These results demonstrated the capability of TP5 to associate with HLA-DR in living antigen presenting cells (APC), thereby providing a new and promising strategy to understand the immunomodulation mechanism induced by TP5 and to design potential immunoregulatory polypeptides.     

Kinetic analysis of single sodium channels from canine cardiac Purkinje cells

Single sodium channel events were recorded from cell-attached patches on single canine cardiac Purkinje cells at 10-13 degrees C. Data from four patches containing two to four channels and one patch with one channel were selected for quantitative analysis. The channels showed prominent reopening behavior at voltages near threshold, and the number of reopenings declined steeply with depolarization. Mean channel open time was a biphasic function of voltage with the maximum value (1-1.5 ms) occurring between -50 and -40 mV and lower values at more and at less hyperpolarized levels. Inactivation without opening was also prominent near threshold, and this occurrence also declined with depolarization. The waiting time distributions and the probability of being open showed voltage and time dependence as expected from whole-cell current studies. The results were analyzed in terms of a five-state Markovian kinetic model using both histogram analysis and a maximum likelihood method to estimate kinetic parameters. The kinetic parameters of the model fits were similar to those of GH3 pituitary cells (Horn, R., and C. A. Vandenberg. 1984. Journal of General Physiology. 84:505-534) and N1E115 neuroblastoma cells (Aldrich, R. W., and C. F. Stevens. Journal of Neuroscience. 7:418-431). Both histogram and maximum likelihood analysis implied that much of the voltage dependence of cardiac Na current is in its activation behavior, with inactivation showing modest voltage dependence.     

Analysis of Competition Binding Assays: Assessment of the Range of Validity of a Commonly Invoked Assumption

Abstract

A common assumption invoked in the analysis of competition binding assays is that the fractional saturation of sites with the unlabeled ligand is given by 1 - (the concentration of bound labeled ligand in the presence of unlabeled ligand)/(the concentration of bound labeled ligand in the absence of unlabeled ligand). This assumption is critically evaluated in the context of several binding models: (a) binding of univalent ligands to multiple classes of equivalent and independent sites, with and without nonspecific binding; (b) cooperative binding of univalent ligands; and (c) binding of multivalent ligands to a single class of univalent acceptors. We show that the conventional assumption is only valid when the labeled ligand is mainly in the free form, occupies a small fraction of the total sites and binds univalently to all sites in an equivalent and independent manner, and when the unlabeled ligand forms l : l complexes with the acceptor sites. When these conditions are satisfied, the conventional assumption is valid even if the unlabeled ligand binds to nonequivalent sites or exhibits cooperativity. Finally, we apply the theory derived for case (a) above to the binding of fluoresceinated epidermal growth factor to A431 cells and demonstrate that the analysis of data obtained from both conventional and competition assays provides information which is difficult, if not impossible, to obtain from either assay alone.     

Measurement variability in DNA flow cytometry of replicate samples

A Bladder Cancer Flow Cytometry Network study has been carried out aimed at identification of the sources of inter- and intralaboratory variability. Replicate "cocktail" samples containing a mixture of peripheral blood lymphocytes and an aneuploid cell line and samples of peripheral blood lymphocytes serving as a DNA reference standard were distributed to five network laboratories. The samples were stained for DNA using propidium iodide, with each laboratory using its own staining protocol. Sets of these samples were analyzed by flow cytometry to obtain cellular DNA distributions. DNA index and hyperdiploid fraction were calculated for each histogram using an automated technique. Results were evaluated by analysis of variance to identify sources of variability. Three important sources of variation were found that affect flow cytometry in general and- the transportability of flow cytometry results to routine clinical use in particular. The significant variation among laboratories that is constant across time most probably represents stable differences in instrumentation, instrument set-up, and laboratory techniques. This variation can be compensated for, if it is known and stable, to develop transportable classification criteria. The second type of variation, termed the interaction component, represents differences among laboratories that are not constant across time. Sources of this variation include inconsistency in sample preparation, staining, and analysis. The elimination of this type of variation is required for meaningful comparison of data within and among laboratories and the creation of interlaboratory data-bases. The third type of variation represents pure measurement variability and affects the sensitivity of the technique.     

Modified histogram subtraction technique for analysis of flow cytometry data

Analysis of flow cytometry histogram data by the subjective selection of an integration window can be a tedious and time-consuming task and is often inaccurate. A new method for automated calculation of the percent positive from immunofluorescence histograms is presented. This new method is a modification of the currently used method of channel-by-channel histogram subtraction. Its accuracy is compared to that of the channel-by-channel histogram subtraction method and to another currently used automated method, which selects an integration window by finding the channels that contain the most fluorescent 2% of a control histogram. The new histogram subtraction method is objective, easy to use, and is more accurate than other currently used automated analysis methods. PASCAL source code is given for each method of analysis.     

An innovation in flow cytometry data collection and analysis producing a correlated multiple sample analysis in a single file

The problems associated with rapid analysis and interpretation of data from multicolor immunofluorescence panels have been a formidable barrier to their routine use. Using present flow cytometry concepts, a panel of 11 tubes each containing multiple phenotypic markers or controls requires postdata acquisition manipulation of many multiparameter histogram and listmode files. We have developed a method that compresses all of the information from such a panel into a single listmode data file during run time. A single data file is used to record the entire phenotypic analysis for a particular patient or series within an experiment. This is accomplished by the incorporation of a tube identifier parameter (TIP) as well as the fluorescence and light scatter parameters normally collected. The TIP can then be used for gating discrimination of any tube or set of tubes within a panel. When the TIP is correlated with the PRISM parameter the entire patient phenotypic image can be represented within a single two-parameter histogram we have called a phenogram. This phenogram can be generated in real time, providing on-line preprocessing of a complex multicolor experiment. By examining the image created by the phenogram it is possible to rapidly flag abnormalities such as incorrect gating. This procedure was carried out on an EPICS Elite flow cytometer in its standard configuration with the addition of hardware to provide an input for the TIP.     

Chromatin-bound PCNA as S-phase marker in mononuclear blood cells of patients with acute lymphoblastic leukaemia or multiple myeloma

Objectives: Proliferating cell nuclear antigen (PCNA) has often been used as a marker to aid assessment of tumour growth fraction. This paper addresses the question of whether it can be used as an S‐phase marker, when the non‐chromatin‐bound form of the protein is removed by pepsin treatment. Materials and methods: Cytofluorometric measurements were carried out after immunofluorescence staining of PCNA and counterstaining of DNA. S‐phase fraction was determined with the help of windows on PCNA versus DNA scattergrams, or mathematically from DNA histograms. Results: S‐phase fractions obtained using the two methods correlated well, but did not always agree, exact discrepancies depending on the mathematical model used for histogram analysis. Conclusions: Determination of S‐phase fractions with the help of PCNA immunofluorescence staining is possible, and probably more reliable than calculation of S‐fractions from DNA histograms. It thus offers an alternative to assays involving BrdU labelling in vivo.     

Spatial Analysis of Multi-species Exclusion Processes: Application to Neural Crest Cell Migration in the Embryonic Gut

Hindbrain (vagal) neural crest cells become relatively uniformly distributed along the embryonic intestine during the rostral to caudal colonization wave which forms the enteric nervous system (ENS). When vagal neural crest cells are labeled before migration in avian embryos by in ovo electroporation, the distribution of labeled neural crest cells in the ENS varies vastly. In some cases, the labeled neural crest cells appear evenly distributed and interspersed with unlabeled neural crest cells along the entire intestine. However, in most specimens, labeled cells occur in relatively discrete patches of varying position, area, and cell number. To determine reasons for these differences, we use a discrete cellular automata (CA) model incorporating the underlying cellular processes of neural crest cell movement and proliferation on a growing domain, representing the elongation of the intestine during development. We use multi-species CA agents corresponding to labeled and unlabeled neural crest cells. The spatial distributions of the CA agents are quantified in terms of an index. This investigation suggests that (i) the percentage of the initial neural crest cell population that is labeled and (ii) the ratio of cell proliferation to motility are the two key parameters producing the extreme differences in spatial distributions observed in avian embryos.     

How microtubules get fluorescent speckles.

The dynamics of microtubules in living cells can be seen by fluorescence microscopy when fluorescently labeled tubulin is microinjected into cells, mixing with the cellular tubulin pool and incorporating into microtubules. The subsequent fluorescence distribution along microtubules can appear "speckled" in high-resolution images obtained with a cooled CCD camera (Waterman-Storer and Salmon, 1997. J. Cell Biol. 139:417-434). In this paper we investigate the origins of these fluorescent speckles. In vivo microtubules exhibited a random pattern of speckles for different microtubules and different regions of an individual microtubule. The speckle pattern changed only after microtubule shortening and regrowth. Microtubules assembled from mixtures of labeled and unlabeled pure tubulin in vitro also exhibited fluorescent speckles, demonstrating that cellular factors or organelles do not contribute to the speckle pattern. Speckle contrast (measured as the standard deviation of fluorescence intensity along the microtubule divided by the mean fluorescence intensity) decreased as the fraction of labeled tubulin increased, and it was not altered by the binding of purified brain microtubule-associated proteins. Computer simulation of microtubule assembly with labeled and unlabeled tubulin showed that the speckle patterns can be explained solely by the stochastic nature of tubulin dimer association with a growing end. Speckle patterns can provide fiduciary marks in the microtubule lattice for motility studies or can be used to determine the fraction of labeled tubulin microinjected into living cells.