Evolution of the clock from yeast to man by period-doubling folds in the cellular oscillator.

Analysis of genome-wide oscillations in transcription reveals that the cell is an oscillator and an attractor and that the maintenance of a stable phenotype requires that maximums in expression in clusters of transcripts must be poised at antipodal phases around the steady state-this is the dynamic architecture of phenotype. Plots of the path through concentration phase space taken by all of the transcripts of Saccharomyces cerevisiae yield a simple three-dimensional surface. How this surface might change as period lengthens or as a cell differentiates is at the center of current work. We have shown that changes in gene expression in response to mutation or perturbation by drugs occur through a folding or unfolding of the surface described by this circle of transcripts and we suggest that the path from this 40-minute oscillation to the cell cycle and circadian rhythms takes place through a series of period-two or period-three bifurcations. These foldings in the surface of the putative attractor result in an increasingly dense set of nested trajectories in the concentrations of message and protein. Evolutionary advantage might accrue to an organism that could change period by changes in just one or a few genes as day length increased from 4 hours in the prebiotic Earth, through 8 hours during the expansion of photoautotrophs, to the present 24 hours.

Application of Bayesian decomposition for analysing microarray data.

MOTIVATION: Microarray and gene chip technology provide high throughput tools for measuring gene expression levels in a variety of circumstances, including cellular response to drug treatment, cellular growth and development, tumorigenesis, among many other processes. In order to interpret the large data sets generated in experiments, data analysis techniques that consider biological knowledge during analysis will be extremely useful. We present here results showing the application of such a tool to expression data from yeast cell cycle experiments. RESULTS: Originally developed for spectroscopic analysis, Bayesian Decomposition (BD) includes two features which make it useful for microarray data analysis: the ability to assign genes to multiple coexpression groups and the ability to encode biological knowledge into the system. Here we demonstrate the ability of the algorithm to provide insight into the yeast cell cycle, including identification of five temporal patterns tied to cell cycle phases as well as the identification of a pattern tied to an approximately 40 min cell cycle oscillator. The genes are simultaneously assigned to the patterns, including partial assignment to multiple patterns when this is required to explain the expression profile. AVAILABILITY: The application is available free to academic users under a material transfer agreement. Go to http://bioinformatics.fccc.edu/ for more details.

Genome wide oscillations in expression. Wavelet analysis of time series data from yeast expression arrays uncovers the dynamic architecture of phenotype.

A reanalysis of expression arrays in yeast cells synchronized by alpha factor blockade or through the use of temperature sensitive mutants uncovered a genome wide pattern of oscillations in mRNA concentrations. Using wavelet decomposition as a signal processing technique and enhancement strategies borrowed from image processing, noise and trends in the Stanford yeast cell cycle data were partitioned away from time series profiles to uncover genome-wide oscillations in expression. These oscillations which were typically of cell cycle or half cell cycle duration, 40 and 80 minutes in the Stanford data set suggest that there are large-scale temporal structures and high frequency oscillations in mRNA levels through the cell cycle. Wavelet decomposition, which acts like a band pass filter bank, was used to determine where most of the power appeared in the decomposition. The approximately 40-min oscillation is mirrored in continuous chemostat cultures. In these cultures, metabolic synchrony involving an unknown proportion of the transcriptome can be monitored by measurement of oxygen consumption and can be sustained for weeks. These 40-min oscillations are stable and precise with coefficients of variation less than 1% for both period and amplitude. The hypothesis that high and low amplitude oscillations are a ubiquitous property of the genetic regulatory circuitry was supported by the observation of period doubling bifurcations in the distribution of population doubling times in yeast.

Tuning in the transcriptome: basins of attraction in the yeast cell cycle.

Image processing techniques and wavelet analyses have been applied to the yeast cell cycle expression microchip data to reveal large-scale temporally coherent structures and high frequency oscillations in mRNA levels through the cycle. Because transitions in expression frequently occur in phase, they appear as peaks or troughs in colour maps and contour plots of expression levels. Although apparent in the untreated data, these transitions were identified and enhanced by convolution of a Laplacian kernel with the expression arrays of the first 4096 genes. Transitions associated with maximum up- or down-regulation of mRNA levels appear as bands at 30-40 min intervals through two cell cycles. Time-frequency analyses using wavelet transforms support these visualization techniques and lead to the conclusion that, with respect to gene expression, the dominant period is not the cell cycle (90-120 min) but, more commonly, the higher frequency 30-40 minute submultiple of the cycle period.

Dynamic architecture of the yeast cell cycle uncovered by wavelet decomposition of expression microarray data.

Wavelet analysis has been applied to yeast cell cycle expression microchip data to reveal large-scale temporal structures and ubiquitous oscillations in mRNA levels. Discrete intervals in time within the cell cycle when expression levels changed were visualized as contour maps in which points of transition in gene expression among all 6178 genes were plotted as a function of cell cycle time. Time-frequency analysis using wavelet transforms supported the direct visualization and led to the conclusion that the predominant period is not the cell cycle but a higher frequency, 40 min, submultiple of the cycle. Each of the 6178 gene expression profiles was dissected by wavelet decomposition into all permitted frequencies from the Nyquist limit to roughly twice the cell cycle length. Transitions associated with maximum up- or down-regulation of mRNA levels appear as bands at circa 40-min intervals, half the length of the cycle, through two cell cycles. More than two thirds of the genes, including many of the cyclins, showed this half-cycle periodicity. Gene expression and events within the yeast cell cycle may be regulated by an attractor whose fundamental period is an emergent property of dynamic interactions within the yeast transcriptome.

Phase I trial of intraperitoneal iododeoxyuridine with and without intravenous high-dose folinic acid in the treatment of advanced malignancies primarily confined to the peritoneal cavity: flow cytometric and pharmacokinetic analysis.

In this Phase I study, the maximally tolerated doses (MTDs) of i.p. iododeoxyuridine (IdUrd) alone and in combination with i.v. calcium leucovorin (LV) were determined. The pharmacokinetics and pharmacological advantage of IdUrd were evaluated, and flow cytometric analysis allowed examination of the extent of incorporation of IdUrd into tumor cells with and without the addition of i.v. LV. Thirty-nine patients with advanced neoplasms primarily confined to the peritoneal space were enrolled in a dose-escalation trial using 4-h dwells of IdUrd administered i.p. daily for 4 days with and without an i.v. infusion of LV 500 mg/m2/day for 4.5 days. Twenty-three patients received single-agent therapy, and 13 patients received i.p. IdUrd in combination with i.v. LV. The MTD of single-agent IdUrd administered on this schedule was 4125 mg/m2/day for 4 days; and that of the IdUrd in combination was 3438 mg/m2/day. Dose-limiting toxicities were myelosuppression and stomatitis. During the period of the dwell, the peritoneal AUC (area under the curve) of IdUrd exceeded the plasma AUC of IdUrd by one or two orders of magnitude in all patients at all doses tested; there was a possible effect of LV on peritoneal AUC. The geometric mean pharmacological advantage (AUCperitoneal/ AUCplasma) was 181 at 625 mg/m2/day and 90 at 4538 mg/m2/day. Flow cytometric analysis suggests saturation of IdUrd measured in DNA at the 2500-3125 mg/m2 dose level, without an increase after the addition of LV. Twelve patients received 4-12 courses of therapy. One patient with recurrent ovarian cancer who received 16 courses of therapy experienced complete resolution of her ascites, near normalization of CA-125 levels, and improved quality of life; two patients with high-risk tumors receiving "adjuvant" therapy are disease-free at 3 and 6 years after treatment; other patients experienced transient clearing of ascites. The recommended Phase II dose of i.p. IdUrd using a 4-h dwell daily for 4 days is 3750 mg/m2/day alone or 3125 mg/m2/day in combination with continuous i.v. LV at 500 mg/m2/day for 4.5 days. Although flow cytometric data suggest that DNA incorporation of IdUrd is not affected by the addition of LV, the cytotoxicity of the combination regimen may be increased due to LV-enhanced, IdUrd-related inhibition of thymidylate synthase. For this reason, we recommend that efficacy studies of the combination continue in parallel with studies of IdUrd alone.

Distinct circadian time structures characterize myeloid and erythroid progenitor and multipotential cell clonogenicity as well as marrow precursor proliferation dynamics.

Circadian differences in the susceptibility of the marrow to the effects of radiation, myelotoxic drugs, and growth factors suggest that hematopoietic processes vary significant throughout each day. One mechanism possibly responsible for the differing degrees of marrow damage sustained from a fixed dose of a cytotoxic agent at different times of day is the circadian organization of cell cycle events. Previous circadian rhythm-oriented studies of proliferation using unfractionated marrow have reported seemingly contradictory peak and nadir times of day. Marrow represents a heterogeneous population of stem cells and various hematopoietic progenitors whose proliferation and differentiation are controlled by both common and unique factors. Therefore, we examined lineage-specific circadian marrow proliferative dynamics for evidence of parasynchronous circadian DNA synthesis. Cell cycle phase was determined using flow cytometry with both propidium iodide staining and 5-bromo-2'-deoxyuridine (BrdU) incorporation concurrently with cell culture-based determinations of lineage-specific progenitor numbers in the same marrow samples. Although no clear circadian (24-hour) rhythm characterized unfractionated marrow DNA synthesis, both erythroid- and myeloid-enriched subpopulations demonstrated distinct circadian patterns with respect to the percentage of cells incorporating BrdU, with up to 50% differences throughout each day. Interestingly, these circadian rhythms in erythroid and myeloid progenitor cell DNA synthesis are entirely different from one another. The lineage-specific circadian patterns in the fraction of cells undergoing DNA synthesis are, in part, paralleled by up to eightfold larger circadian differences in erythroid and myeloid colony numbers. Multipotential colony numbers likewise vary throughout the day, with a unique pattern of their own. The predominant period length of daily rhythms in colony numbers and their amplitudes differ as a function of the stage of progenitor commitment. Multipotent and early progenitor colony numbers each exhibit 24-hour rhythms, with three- to fivefold daily peak-trough differences, whereas later progenitor colony numbers exhibit two peaks per day (12-hour rhythms) with twofold peak-trough differences throughout each day. In vivo erythropoietin (Epo) administration enhances daily rhythms in erythroid colony numbers by increasing their amplitudes while leaving their circadian shapes virtually unchanged. The increment in erythroid colony numbers after Epo administration varies up to 16-fold with the time of day of treatment. In summary, we have defined distinctly different lineage-dependent circadian patterns of marrow progenitor numbers and proliferating cells. We can infer from these data that the circadian timing of administration of physical, chemical, or biologic agents, whose bioactivity toward marrow precursors depends on the cell cycle phase of its presentation, can be expected to affect this activity predictably and significantly. These results may have practical applications in improving stem and progenitor cell yields by optimal circadian timing of growth factor administration and harvest.

Replication synchrony-PCR: a sampling-time-independent assay for replication synchrony in human tissues and tumors in situ.

Replication synchrony within a cell population can be demonstrated by pulse-labeling followed by PCR amplification of immunoprecipitated 5-iodo-2'-deoxyuridine (IdUrd)-labeled DNA from cells of otherwise indeterminant kinetic stages. This replication synchrony-PCR approach may be valuable in understanding the dynamics of human normal tissue or solid tumor replication in situ where access for repeated sampling is severely limited. IdUrd labeling provides a sampling-time-independent method for assessing the replicative status of a cell population at the time when the label was presented. Using genes whose time of replication in S phase is already known, the presence of a cell in early or late S phase can be determined and a qualitative measure made of replication synchrony in the population. This approach was evaluated in synchronous and random cultures of Ej cells using the early replicating PGK-1 gene to identify cells in early S phase at the time of labeling and the late replicating factor IX gene to identify cells that were in late S phase. To test the feasibility of clinical application of this technique, human tumor cells from patients with advanced cancers, given IdUrd therapeutically at specified times of the day, were evaluated. In some patients, replication synchrony-PCR provided evidence of parasynchronous DNA replication in tumor cells. This technique could be appended to existing clinical studies in which BrdUrd or IdUrd is being given to patients either diagnostically or therapeutically.

Flow cytometric evaluation of ovarian cancer.

BACKGROUND: Despite recent advances in the treatment of ovarian cancer, the long-term prognosis for patients with this malignancy appears to depend more on tumor prognostic factors than on treatment regimens. The traditionally used prognostic factors are often subjective and, currently, have not been sufficient to determine individual patient prognosis. METHODS: Newer techniques of quantitative cytologic testing, including flow cytometry, facilitate the objective evaluation of tumor cell heterogeneity and the identification of additional prognostic factors. RESULTS: There is good evidence, mainly from retrospective studies, that DNA ploidy is a valuable prognostic indicator in patients with both early-stage and late-stage ovarian cancer. Most of the recent flow cytometric studies have identified ploidy as an independent prognostic factor, with aneuploidy predicting a significantly shorter survival time, even in patients with borderline malignant tumors. Flow cytometric determination of cell cycle information (e.g., S-phase fraction or proliferative index) may represent additional prognostic information and may be used to predict the early tumor response to treatment. CONCLUSIONS: Although additional prospective studies are needed to establish the exact value of flow cytometric evaluation for ovarian cancer and other gynecologic malignancies, there is little doubt that the prognostic value of this information will influence clinical management of patients with these malignancies in the near future.

Autogenous formation of spiral waves by coupled chaotic attractors.

When large arrays of strange attractors are coupled diffusively through one of the variables, chaotic systems become periodic and form large archimedean spirals or concentric bands. This observation may have importance for many applications in the field of deterministic chaos and seems particularly relevant to the question of the formal temporal structure of the biological clock in metazoan organisms. In particular, although individual cellular oscillators, as manifested in the cell cycle, may have deep basins of attraction and appear to be more or less periodic, we suggest that cells oscillate with chaotic dynamics in the ultradian domain. Only when large aggregates of these cells are tightly coupled can a precise circadian clock emerge. For changing coupling strength or parameter values, period increase occurs through quantal or integral multiple increments of the fundamental. All calculations were implemented on a 386AT, using a Mercury MC6400 floating point processor.

Circadian gating of S phase in human ovarian cancer.

Cell proliferation in 30 patients with ovarian cancer was analyzed using flow cytometry to determine changes in the percentage of cells in S phase. By this measure, proliferation in tumor cells appears to follow a cyclical pattern of peaks and troughs that is out of phase with the circadian rhythm in proliferation of normal tissues. In round-the-clock monitoring of replication stages in tumor cells recovered from i.p. lavage fluid in postsurgery patients, peaks of tumor and nontumor cell DNA synthesis commonly occurred at different times of day. When patients were grouped so that only tumor cell proliferation was being measured, a highly significant 24-h rhythm nearly 12 h out of phase with nontumor cell proliferation was found. This peak in the percentage of S-phase cells occurs most commonly in mid- to late morning and appears to offer an opportunity for timing chemotherapy to coincide with high tumor cell vulnerability and low toxicity to normal tissue.

Tumor DNA content as a prognostic feature in advanced epithelial ovarian carcinoma.

Tumor DNA content (ploidy) was determined in 84 patients with epithelial ovarian carcinoma. Stage II-IV. A total of 251 DNA histograms generated by flow cytometry on cells derived from paraffin-embedded specimens were analyzed retrospectively. Of the 84 patients, 44 had tumors which were aneuploid, whereas 33 had diploid, and 7 had tetraploid tumors. Cox regression analysis revealed that age (P less than 0.001), stage (P less than 0.001), and ploidy (P less than 0.001) were independent prognostic features. The median survival time was 19 months and 48 months, respectively, in aneuploid and euploid tumors (P less than 0.001). The size of residual after surgery lost its significance when corrected for stage. Multivariate analysis in Stage III tumors revealed that ploidy was the most important prognostic factor (P less than 0.001) followed by age (P less than 0.025). A remarkable stability of cellular DNA content was found when the primary tumor was compared to the following groups: (1) various metastatic specimens from the primary operation in the same patient; (2) specimens analyzed sequentially from primary, secondary, and tertiary exploratory laparotomy; and (3) peritoneal washings before and after intraperitoneal chemotherapy.

Isolation of viable mouse primordial germ cells by antibody-directed flow sorting.

The germ line represents a cell type of unique interest in mammals because it retains complete genotypic totipotency while undergoing significant phenotypic differentiation. Analysis of the mechanism that underlies the maintenance of this totipotency requires the ability to isolate and study all stages of the germ cell lineage. The primordial germ cells (PGC) are the earliest identifiable germ cells in the embryo. It has not previously been possible to isolate PGC in sufficient numbers and purity to facilitate biochemical and/or molecular analysis. We report here that the use of a monoclonal antibody in combination with flow cytometry does permit the isolation of reasonably large and pure yields of viable mouse PGC.

Circadian and ultradian rhythms of proliferation in human ovarian cancer.

Synchronous waves of proliferation in tumor cells taken from patients with ovarian cancer were observed using flow cytometry to measure the fraction of cells undergoing DNA replication and displaying tumor-cell-specific immunofluorescence. When saline washings of the abdominal cavity were analyzed at 2-4 hr intervals round-the-clock, the percentage of cells in the chromosome replication cycle (S + G2 percentage) showed 12-hr and often higher frequency rhythms in proliferation. These higher frequency rhythms in DNA replication show a relatively constant phase relationship to the patient's circadian clock with peak proliferation occurring most commonly at 10 a.m. to 12 noon and again at 10 p.m. This proliferation rhythm is therefore partially out of phase with the 24-hr rhythms in proliferation seen in normal cells. The findings on human cancer reveal a fundamental difference in the temporal organization of normal and tumor cell growth that should be exploited for therapeutic benefit.

Automatic digital image analysis for identification of mitotic cells in synchronous mammalian cell cultures.

Mitotic frequency in a synchronous culture of mammalian cells was determined fully automatically and in real time using low-intensity phase-contrast microscopy and a newvicon video camera connected to an EyeCom III image processor. Image samples, at a frequency of one per minute for 50 hours, were analyzed by first extracting the high-frequency picture components, then thresholding and probing for annular objects indicative of putative mitotic cells. Both the extraction of high-frequency components and the recognition of rings of varying radii and discontinuities employed novel algorithms. Spatial and temporal relationships between annuli were examined to discern the occurrences of mitoses, and such events were recorded in a computer data file. At present, the automatic analysis is suited for random cell proliferation rate measurements or cell cycle studies. The automatic identification of mitotic cells as described here provides a measure of the average proliferative activity of the cell population as a whole and eliminates more than eight hours of manual review per time-lapse video recording.

Quasi-exponential generation time distributions from a limit cycle oscillator.

In spite of the apparently random behaviour and the often exponential distribution of generation times expressed in cell populations, there is evidence for rather precise timekeeping in the cell cycle. In experiments using time-lapse video-tape microscopy, we have noted that cell generation times are often not distributed smoothly but in many cases seem to cluster at roughly 4 hr intervals. Phase shift responses following application of heat shock, ionizing radiation or serum pulses in each case show a pattern which is repeated twice in cells with an 8-9 hr modal generation time. We describe here a cell cycle model with an independent cellular clock controlling cell cycle events which accounts for the phase response data, while also reconciling the stochastic and periodic behaviour characteristic of animal cells.

Cell-cycle-dependent expression of human melanoma membrane antigen analyzed by flow cytometry.

Knowledge of tumor antigenic expression is crucial to the design of therapeutic strategy. A murine monoclonal antibody (BE4) against a human melanoma membrane antigen, was used to study the in vitro expression of this antigen. By membrane immunofluorescence, BE4 reacted against 5 of 8 melanoma lines as compared to zero of 13 other cell populations. Using flow cytometry, the antigenic M14 CEM melanoma cells consisted of 40% to 60% of the total cell population. Dual-parameter measurements of DNA content and membrane antigen demonstrated that the nonantigenic cells were predominantly in G0/G1 phase, whereas the antigenic cells were distributed throughout the cell cycle. Within one passage, the sorted and recultured nonantigenic population demonstrated a similar proportion of antigenic cells as the unsorted original population. It was concluded that the expression of human melanoma antigen was cell-cycle-dependent. Understanding factors that turn off the expression of antigen in G0/G1 phase may lead to better immunotherapeutic strategies.