Limits...
Dynamic epitope expression from static cytometry data: principles and reproducibility.

Jacobberger JW, Avva J, Sreenath SN, Weis MC, Stefan T - PLoS ONE (2012)

Bottom Line: The resulting 5 dimensional data were analyzed as a series of bivariate plots to isolate the data as segments of an N-dimensional "worm" through the data space.Very precise, correlated expression profiles for important cell cycle regulating and regulated proteins and their modifications can be produced, limited only by the number of available high-quality antibodies.These profiles can be assembled into large information libraries for calibration and validation of mathematical models.

View Article: PubMed Central - PubMed

Affiliation: Case Comprehensive Cancer Center, Case Western Reserve University, Cleveland, Ohio, United States of America. jwj@case.edu

ABSTRACT

Background: An imprecise quantitative sense for the oscillating levels of proteins and their modifications, interactions, and translocations as a function of the cell cycle is fundamentally important for a cartoon/narrative understanding for how the cell cycle works. Mathematical modeling of the same cartoon/narrative models would be greatly enhanced by an open-ended methodology providing precise quantification of many proteins and their modifications, etc. Here we present methodology that fulfills these features.

Methodology: Multiparametric flow cytometry was performed on Molt4 cells to measure cyclins A2 and B1, phospho-S10-histone H3, DNA content, and light scatter (cell size). The resulting 5 dimensional data were analyzed as a series of bivariate plots to isolate the data as segments of an N-dimensional "worm" through the data space. Sequential, unidirectional regions of the data were used to assemble expression profiles for each parameter as a function of cell frequency.

Results: Analysis of synthesized data in which the true values where known validated the approach. Triplicate experiments demonstrated exceptional reproducibility. Comparison of three triplicate experiments stained by two methods (single cyclin or dual cyclin measurements with common DNA and phospho-histone H3 measurements) supported the feasibility of combining an unlimited number of epitopes through this methodology. The sequential degradations of cyclin A2 followed by cyclin B1 followed by de-phosphorylation of histone H3 were precisely mapped. Finally, a two phase expression rate during interphase for each cyclin was robustly identified.

Conclusions: Very precise, correlated expression profiles for important cell cycle regulating and regulated proteins and their modifications can be produced, limited only by the number of available high-quality antibodies. These profiles can be assembled into large information libraries for calibration and validation of mathematical models.

Show MeSH
Information details: cyclin A2 and PHH3 expression.Molt 4 cells were stained and analyzed as in Figures 3, 4, and 5. A: brackets show are of cyclin A2 decrease in early mitosis; small arrow points to “new born” cells; large arrow points to large cluster in early mitosis, largely composes of prophase cells. B: gray zone highlights the data derived from bracketed segments in A; small arrow points to data derived from the large mitotic cluster in A; added lines indicate expected expression profile if more data could be obtain within the large cluster in A.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3275612&req=5

pone-0030870-g006: Information details: cyclin A2 and PHH3 expression.Molt 4 cells were stained and analyzed as in Figures 3, 4, and 5. A: brackets show are of cyclin A2 decrease in early mitosis; small arrow points to “new born” cells; large arrow points to large cluster in early mitosis, largely composes of prophase cells. B: gray zone highlights the data derived from bracketed segments in A; small arrow points to data derived from the large mitotic cluster in A; added lines indicate expected expression profile if more data could be obtain within the large cluster in A.

Mentions: Expression profile derivation for cyclin A2 and PHH3 can be performed as in Figure 4. The approach is to create regions that enclose statistically irreducible clusters (generally, at the ends and corners of the data trajectory) and divide the “stretches” between these clusters into regions with a significant number of events and a goal of providing as many regions within the stretch as is practical. This is valid provided that a cell's residence in any “state” described by a segmented region is dependent on having existed in a previous “state” described by an adjacent region at an immediately earlier time, and that all previous regions are to one side of the immediate region - i.e., the “states” and regions are contiguous, ordered, and unidirectional. This requirement can be satisfied either from direct experimentation - e.g., [4], prior knowledge, or inferred logic. The directionality for cyclin A2 versus PHH3 is shown by arrows in Figure 3B. The shape of any bivariate region is conceptually straightforward. The goal is to create adjacent, contiguous “sides” for tri- or tetragons that are perpendicular to a line that follows the local slope of the data. The other two sides are trivial, enclose the data, but could be thought of as parallel to the local slope of the data. This approach is an attempt to assign cells to a level of expression that approximates two-dimensional Gaussian fitting – i.e., account for appropriate variation in two dimensions. The second rule is to bound data at the ends of “stretches” by regions that are not further reducible. For example, R7 in Figure 4A, which encloses G1 cells, cannot be made smaller without creating a center statistic for expression that maps variation rather than expression, since that region includes variation that is not further resolvable. This is often not a difficult decision point, since the frequency information, in this case denoted by contour lines, provides information that can be used to determine the best placement of the region boundary. Equally, when the data turn at a right angle, the width of the data in each arm can direct the height and width of the region at the corner. For example, the center and width of the data region bounded by R42 and R44 can guide the size of R43 (Figure 4B). In practice, segmenting orthogonal data is very straight forward, whereas data with curvature is problematic because of current software limitations. When data are not orthogonal (e.g., the data captured by regions, R15 through R24, Figure 4A), the intent is to follow the two dimensional peak range (i.e., the backbone), distributing the data within a region in a manner perpendicular to the slope. However, we do not know of current software packages that provide the ability to create the desired region shape or eliminate region-to-region overlap with any assurance. For the data presented in Figures 4, 5, 6, 7, to ensure non-overlap and comprehensive inclusion, we calculated the region vertices in a spreadsheet program, then directly modified the “wlx” file (an “XML” descriptor file) in WinList.


Dynamic epitope expression from static cytometry data: principles and reproducibility.

Jacobberger JW, Avva J, Sreenath SN, Weis MC, Stefan T - PLoS ONE (2012)

Information details: cyclin A2 and PHH3 expression.Molt 4 cells were stained and analyzed as in Figures 3, 4, and 5. A: brackets show are of cyclin A2 decrease in early mitosis; small arrow points to “new born” cells; large arrow points to large cluster in early mitosis, largely composes of prophase cells. B: gray zone highlights the data derived from bracketed segments in A; small arrow points to data derived from the large mitotic cluster in A; added lines indicate expected expression profile if more data could be obtain within the large cluster in A.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3275612&req=5

pone-0030870-g006: Information details: cyclin A2 and PHH3 expression.Molt 4 cells were stained and analyzed as in Figures 3, 4, and 5. A: brackets show are of cyclin A2 decrease in early mitosis; small arrow points to “new born” cells; large arrow points to large cluster in early mitosis, largely composes of prophase cells. B: gray zone highlights the data derived from bracketed segments in A; small arrow points to data derived from the large mitotic cluster in A; added lines indicate expected expression profile if more data could be obtain within the large cluster in A.
Mentions: Expression profile derivation for cyclin A2 and PHH3 can be performed as in Figure 4. The approach is to create regions that enclose statistically irreducible clusters (generally, at the ends and corners of the data trajectory) and divide the “stretches” between these clusters into regions with a significant number of events and a goal of providing as many regions within the stretch as is practical. This is valid provided that a cell's residence in any “state” described by a segmented region is dependent on having existed in a previous “state” described by an adjacent region at an immediately earlier time, and that all previous regions are to one side of the immediate region - i.e., the “states” and regions are contiguous, ordered, and unidirectional. This requirement can be satisfied either from direct experimentation - e.g., [4], prior knowledge, or inferred logic. The directionality for cyclin A2 versus PHH3 is shown by arrows in Figure 3B. The shape of any bivariate region is conceptually straightforward. The goal is to create adjacent, contiguous “sides” for tri- or tetragons that are perpendicular to a line that follows the local slope of the data. The other two sides are trivial, enclose the data, but could be thought of as parallel to the local slope of the data. This approach is an attempt to assign cells to a level of expression that approximates two-dimensional Gaussian fitting – i.e., account for appropriate variation in two dimensions. The second rule is to bound data at the ends of “stretches” by regions that are not further reducible. For example, R7 in Figure 4A, which encloses G1 cells, cannot be made smaller without creating a center statistic for expression that maps variation rather than expression, since that region includes variation that is not further resolvable. This is often not a difficult decision point, since the frequency information, in this case denoted by contour lines, provides information that can be used to determine the best placement of the region boundary. Equally, when the data turn at a right angle, the width of the data in each arm can direct the height and width of the region at the corner. For example, the center and width of the data region bounded by R42 and R44 can guide the size of R43 (Figure 4B). In practice, segmenting orthogonal data is very straight forward, whereas data with curvature is problematic because of current software limitations. When data are not orthogonal (e.g., the data captured by regions, R15 through R24, Figure 4A), the intent is to follow the two dimensional peak range (i.e., the backbone), distributing the data within a region in a manner perpendicular to the slope. However, we do not know of current software packages that provide the ability to create the desired region shape or eliminate region-to-region overlap with any assurance. For the data presented in Figures 4, 5, 6, 7, to ensure non-overlap and comprehensive inclusion, we calculated the region vertices in a spreadsheet program, then directly modified the “wlx” file (an “XML” descriptor file) in WinList.

Bottom Line: The resulting 5 dimensional data were analyzed as a series of bivariate plots to isolate the data as segments of an N-dimensional "worm" through the data space.Very precise, correlated expression profiles for important cell cycle regulating and regulated proteins and their modifications can be produced, limited only by the number of available high-quality antibodies.These profiles can be assembled into large information libraries for calibration and validation of mathematical models.

View Article: PubMed Central - PubMed

Affiliation: Case Comprehensive Cancer Center, Case Western Reserve University, Cleveland, Ohio, United States of America. jwj@case.edu

ABSTRACT

Background: An imprecise quantitative sense for the oscillating levels of proteins and their modifications, interactions, and translocations as a function of the cell cycle is fundamentally important for a cartoon/narrative understanding for how the cell cycle works. Mathematical modeling of the same cartoon/narrative models would be greatly enhanced by an open-ended methodology providing precise quantification of many proteins and their modifications, etc. Here we present methodology that fulfills these features.

Methodology: Multiparametric flow cytometry was performed on Molt4 cells to measure cyclins A2 and B1, phospho-S10-histone H3, DNA content, and light scatter (cell size). The resulting 5 dimensional data were analyzed as a series of bivariate plots to isolate the data as segments of an N-dimensional "worm" through the data space. Sequential, unidirectional regions of the data were used to assemble expression profiles for each parameter as a function of cell frequency.

Results: Analysis of synthesized data in which the true values where known validated the approach. Triplicate experiments demonstrated exceptional reproducibility. Comparison of three triplicate experiments stained by two methods (single cyclin or dual cyclin measurements with common DNA and phospho-histone H3 measurements) supported the feasibility of combining an unlimited number of epitopes through this methodology. The sequential degradations of cyclin A2 followed by cyclin B1 followed by de-phosphorylation of histone H3 were precisely mapped. Finally, a two phase expression rate during interphase for each cyclin was robustly identified.

Conclusions: Very precise, correlated expression profiles for important cell cycle regulating and regulated proteins and their modifications can be produced, limited only by the number of available high-quality antibodies. These profiles can be assembled into large information libraries for calibration and validation of mathematical models.

Show MeSH