Given that the eGFP and mCherry signals in dtCENP-A cells were not co-localised in the z-direction based on the chromatic shift calculated from fluorescent microspheres, dtCENP-A cells were investigated for calculating the chromatic shift between 525 nm and 615 nm emission wavelengths themselves. However, ÷z as measured in metaphase dtCENP-A cells yielded a broad distribution. To ensure that this wasn’t specific to one experiment and therefore one batch of cells, image stacks of Ø5 metaphase cells
were captured as before over a number of experiments on different days. Similarly,Ø5
fields of interphase dtCENP-A cells were also imaged on each occasion. As discussed previously, eGFP was located using Gaussian MMF and the mCherry signal found within a spherical mask of radius 300 nm centred at the GFP spot. The vector, ÷, was calculated for each kinetochore, again pointing from the eGFP to the mCherry signal. Given that the microscope setup may be subject to drift, the measurements of ÷ were experiment-specific. In order to compare all experiments, each experiment’s distributions in ÷x, ÷y and ÷z were normalised to zero by subtracting their median measurement, ÷i for i œ [x, y, z], for that experiment – these distributions will be referred to as centralised distributions. Each of the centralised distributions of ÷i,
direction (Figure 4.4). This is confirmed by the population standard deviation, ‡i, for iœ[x, y, z], for each distribution, where interphase cells (blue; n = 36,727) yielded smaller‡i than those derived from metaphase cells (gold; n = 6,664) (Figure 4.4). Most significantly, ‡z took values 59.1 nm and 120.0 nm for interphase and metaphase cells, respectively, demonstrating that the broad distribution observed previously was not specific to the experiment, and is indeed a consistent issue with metaphase dtCENP- A cells. However, visual inspection of calculated measurements of ÷ highlighted a potential quality issue, as some eGFP and mCherry spots appeared to be incorrectly located in both interphase and metaphase dtCENP-A cells (e.g. Figure 4.5 A). It was therefore crucial that measures for quality control were enforced in order to minimise the occurrences of these events – these events could even be the cause for the broader distribution in measurements of ÷z in metaphase cells.
In order to investigate events where ÷ has visibly been incorrectly calculated, vec- tors for ÷ in the xy-plane were plotted onto images of interphase cells (Figure 4.5 A, yellow vectors). It became apparent that there were two possible reasons for these incorrect measurements: firstly, some eGFP signals were located nearby another eGFP signal of similar signal, however they were accompanied by mCherry signals with very different intensity signals. Sometimes the result of this was that the same mCherry signal was localised for both eGFP spots and the lower-intensity spot ignored (Fig- ure 4.5 A, yellow vector labelled †). This resulted in erroneous measurement of ÷ for both pairs of eGFP-mCherry markers, as the Gaussian MMF assumed that there were two Gaussians contributing to the brighter signal. To reduce the number of these in- stances, any eGFP spots located within 750 nm in 3D from another eGFP spot were removed from the calculation of chromatic shift (Figure 4.5 B). A second apparent source of erroneous ÷ measurement were extremely low eGFP signals (Figure 4.5 A, yellow vector labelledú). These instances were removed by removal of all eGFP spots
below a certain intensity threshold, so that all eGFP spots with intensity below 25% of that of the image’s maximum eGFP spot intensity were removed from the chromatic
−0200 −100 0 100 200 2 4 6 8 ηz – ηz (nm) x 10–3 −0200 −100 0 100 200 8 10 14 16 18 ηy – ηy (nm) x 10–3 12 2 4 6 −0200 −100 0 100 200 2 4 6 8 10 12 14 ηx – ηx (nm) normalised frequency x 10–3 z xROI σ =43.7 nm51.4 nm y σ =32.9 nm41.7 nm σ =59.1 nm120 nm interphase (n = 36,727) metaphase (n = 6,664)
Figure 4.4 Calculation of inter-marker separation using interphase dtCENP-A cells gives better resolution than metaphase cells.
Histograms of residuals in measured ηx, ηy and ηz, calculated using interphase
(blue; n = 36,727) and metaphase (gold; n = 6,664) dtCENP-A cells. Residuals are calculated relative to the median in each coordinate, i, for each experiment,
ηi, resulting in distributions centred at 0. Values given are distribution standard
−0200 −100 0 100 200 2 4 6 8 10 12 ηz – ηz (nm) x 10–3 −0200 −100 0 100 200 5 10 15 20 25 ηy – ηy (nm) x 10–3 −0200 −100 0 100 200 2 4 6 8 10 12 14 ηx – ηx (nm) normalised frequency x 10–3 z xROI y Distance to nearest neighbour GFP spot all spots cut-off at 750 nm σ =43.7 nm 35.6 nm σ =32.9 nm29.9 nm σ =59.1 nm41.8 nm all spots (n = 36,727) quality controlled (n = 5,922) eGFP intensity (% of maximum intensity spot) 0 50 40 30 20 10 cut-off at 25% all spots
C
eGFP-CENP-A mCherry-CENP-A (z-projection) * †Figure 4.5 Chromatic shift calculation in interphase cells can be improved by removal of interfering and low-brightness spots.
(A) Image of interphase dtCENP-A showing the chromatic shift between eGFP and mCherry signal. White arrows point from eGFP to mCherry signal. Yellow arrows demonstrate false positive measurements: (*) vector originating from a GFP spot of low intensity; (†) vector originating from GFP signal not distinct from a neighbouring GFP spot. Scale bars = 750 nm.
(B) Quality control imposed on GFP signal, so that: (left) GFP spots must be separated by at least 750 nm; and (right) GFP signal intensity must be higher than 25% of the maximum intensity spot in the image.
(C) Histograms of residuals in measured ηx, ηy and ηz, calculated using all raw data (blue; n = 36,727) and quality-controlled data (gold; n = 5,922) from inter- phase dtCENP-A cells. Residuals are calculated relative to the median in each coordinate, i, for each experiment, ηi, resulting in distributions centred at 0. Values given are standard deviation, σ, of the distribution.
shift calculation (Figure 4.5 B). Indeed, centralised distributions of ÷i in interphase cells, foriœ[x, y, z], were improved after quality control, where standard deviations of the centralised distributions, ‡i, decreased by 14.9%, 9.1% and 29.3% for i œ [x, y, z], respectively (Figure 4.5 C). Quality control therefore proved incredibly important in ensuring accurate determination of the chromatic shift using dtCENP-A cells, yielding its biggest improvement in the z-direction. I hypothesised then that this method of quality control could also help improve the standard deviations of centralised distri- butions of ÷ in metaphase dtCENP-A cells, and therefore the accuracy of the median measurement of ÷.
In order to check initially whether restricting nearest-neighbour distance and eGFP intensity in ÷ measurements from metaphase dtCENP-A cells would improve the ac- curacy of their measurement, vectors of ÷ in the xy-plane were plotted on images of metaphase dtCENP-A cells (Figure 4.6 A, yellow vectors). Like interphase cells, er- roneous measurements appeared to occur as a consequence of both the localisation of a neighbouring mCherry signal rather than its own counterpart (Figure 4.6 A, yellow vector labelled †), and also vectors originating from eGFP spots with low intensity
(Figure 4.6 A, yellow vector labelled ú). To check whether removal of these erroneous
measurements could also decrease the ÷i (iœ[x, y, z]) distribution’s variance, and per- haps rectify the large uncertainty in z-directional÷ measurements, quality control was imposed on metaphase dtCENP-A cell data. The criteria were slightly loosened as the amount of metaphase cell data was much smaller than for interphase cells (n = 6,664 vs. 36,727, respectively): eGFP spots had to be at least separated by 750 nm and eGFP spots with an intensity smaller than 33% of the maximum eGFP spot inten- sity were removed (Figure 4.6 B). Comparing centralised distributions in ÷i between metaphase cells before and after these measures of quality control yielded smaller stan- dard deviations in ÷ inx, y and z, where‡i decreased by 30.7%, 13.7% and 29.2% for i œ [x, y, z], respectively (Figure 4.6 C). However, despite the decrease in population standard deviation measured for ÷z, the distribution was still broad whilst appearing to contain multiple peaks. Therefore, even after quality control, z-directional mea-
standard deviation (Figure 4.6 C, right). In contrast, population standard deviations of chromatic shift, ÷i (i = [x, y, z]), from interphase dtCENP-A cells were improved after quality control to within almost half a pixel in each coordinate (Figure 4.5 C).
This difference may be a result of the difference between cellular environment within interphase and metaphase cells – metaphase cells contain a well-organised spindle and condensed DNA in chromosomes, whereas interphase cells contain a nuclear envelope, outside which is the main population of microtubules. Therefore, light may expe- rience chromatic shift to a different extent in metaphase as compared to interphase cells. On the other hand, the average axial chromatic shift (÷x and ÷y) calculated in both interphase and metaphase cells for a single experiment were consistent to within measurement accuracy (Figure 4.3). The only difference measured between each cell type was in ÷z. I therefore wanted to utilise the superior precision in measurement of chromatic shift in interphase cells to apply to each experiment in § 5. Firstly, I sought to optimise the precision of measurements of chromatic shift in interphase dtCENP-A cells.