Figure 5.
Plk4 Levels Can Continue to Oscillate and Promote Centriole Duplication Even When the CCO is Perturbed
(A) Graph shows the Plk4-NG oscillations in an embryo injected with dsRNA against cyclin A-B-B3; the schema above the graph illustrates the experimental protocol. The nuclei in this embryo arrest in interphase, but centrioles go through an additional round of division—centriole separation (CS)—accompanied by a Plk4 oscillation. See Figure S2B for additional examples; n = 30 centrioles (mean) per embryo.
(B) Graphs show the raw (black lines) and filtered (red lines) fluorescence intensity data of 3 individual “fertile” centrioles and 3 individual “sterile” centrioles within the same cyclin-depleted embryo. The fertile centrioles duplicate (black dotted lines), and these events were often closely associated with computed Plk4 oscillation peaks (red dotted lines) (see STAR Methods for further details of the peak calling methodology).
(C) An unbiased computational analysis of all 45 fertile centrioles in 3 embryos reveals that >80% of the computationally detected Plk4 oscillation peaks occur within 5 min of an experimentally observed duplication event. A simulation with randomly distributed centriole duplication events and Plk4 oscillation peaks showed a mean time separation of 10.5 min (data not shown).
(D) Venn diagram shows how, using a 5-min window, the oscillation peaks can be used to predict duplication events with both high precision and high recall (40/49 Plk4 oscillation peaks are associated with a duplication event, and 40/52 duplication events are associated with a Plk4 oscillation peak).
(E) Graph shows the ability of Plk4 oscillation peaks to “retrieve” centriole duplication events across all peak prominences. All detected oscillation peaks were ranked in order of their peak prominence from high to low (black dots) and assigned uniquely to a duplication event if within a 5-min time window. The graph then plots the precision and recall values if the threshold for calling a peak were set as the peak prominence value of each peak (in descending order). Below the detected peak that is associated with a peak prominence threshold of 0.12, the precision dramatically drops, suggesting the existence of a minimum peak amplitude for centriole duplication. At this threshold, precision and recall are jointly optimized. Note, if there were no overall correlation between Plk4 peaks and a duplication event, the integrated area under the curve across all peak prominences or average precision (AP) for the 5-min time window (AP5min) would be ~50% (given by # duplications/(# duplications + # peaks)); so the score of ~75% indicates a meaningful correlation.
(F) Graph shows the correlation between the time of the computationally determined Plk4 peaks and their respective experimentally observed duplication events. Correlation strength was examined using Pearson’s correlation coefficient (r < 0.40 weak; 0.40 < r < 0.60 moderate; r > 0.60 strong); significance of correlation was determined by the p value (p < 0.05) (see STAR Methods for a full description of this analysis).
See also Figures S2 and S8.