kSHREC ‘Delta’ reflects the shape of kinetochore rather than intrakinetochore tension

Distance between fluorescent spots formed by various kinetochore proteins (‘Delta’) is proposed to reflect the level of intrakinetochore tension (IKT). However, larger-scale changes in the kinetochore architecture may also affect Delta. To test this possibility, we measure Delta in long kinetochores of Indian muntjac (IM) whose shape, size, and orientation are discernable in conventional light microscopy. We find that architecture of IM kinetochores and the value of Delta change minimally when microtubule-mediated forces are suppressed by Taxol. In contrast, large decreases of Delta observed in Taxol-treated human cells coincide with prominent changes in length and shape of the kinetochore. We also find that inner and outer kinetochore proteins intermix within a common spatial compartment instead of forming separate thin layers. These observations, supported by computational modelling, suggest that changes in Delta reflect changes in the kinetochore shape rather than the level of IKT.


Introduction 26
Segregation of chromosomes during cell division (mitosis) depends upon 'kinetochores ', 27 macromolecular assemblies located at the centromere of each chromosome. Kinetochores 28 perform two principal functions: they generate the force that propels chromosomes and produce 29 a checkpoint signal that delays progression through mitosis until all chromosomes attach to 30 spindle microtubules. Molecular composition of the kinetochore is complex, comprising over a 31 hundred of various proteins [1]. Further, as the cell progresses through mitosis, molecular 32 composition of the kinetochore and its architecture change in response to various types of 33 interactions with spindle microtubules. These adaptive changes in size and shape of the 34 kinetochores ensure that microtubule attachments form rapidly yet with a low number of errors 35 [2,3]. Thus, revealing mechanisms that govern kinetochore architecture is of a significant 36 interest. 37 Due to its small size in most mammalian cells (~300 nm), the shape of the kinetochore 38 or the distribution of its components cannot be directly delineated in conventional light 39 microscopy (LM). A popular approach to overcoming this limitation is based on measuring the 40 'Delta', a distance between centroids of fluorescent spots formed by various proteins, visualized 41 in different colors within the same kinetochore. Measurements of Delta lay the foundation of a 42 nanometer-scale map that attributes various proteins to thin layers orthogonal to the inner-outer 43 (from the centromere towards attached microtubules) axis of the kinetochore [4,5]. Delta 44 between the proteins at the base of the kinetochore (CenpA) and those in the microtubule 45 binding domain (Hec1) has been shown to decrease ~30% when microtubule dynamics are 46 suppressed by Taxol [4,6]. The decrease was interpreted as a manifestation of changes in the 47 physical separation between the inner and outer layers, which in turn led to the concept of 'intra-48 kinetochore tension' and the notion that stretching the kinetochore is necessary for the 49 satisfaction of the spindle assembly checkpoint (SAC) [7]. This attractive hypothesis; however, 50 remains debatable for two main reasons. First, there exists a high degree of variability in Delta 4 measurements conducted via different techniques and in different laboratories. While some 52 report that Delta (CenpA-Hec1) decreases by ~30 nm when human cells are treated with Taxol 53 [4], others observe a lesser decrease [8] or no statistically significant change in separation of 54 the same kinetochore proteins under similar experimental conditions [9, 10]. These 55 discrepancies likely arise from the alternative measuring techniques, particularly various 56 approaches to compensating chromatic aberration, inevitable in LM [4,5,8,10,11]. Second, 57 interpretation of Delta as a metric for physical distances between molecules is obfuscated by 58 the malleable shape of the kinetochore. While literal interpretation of Delta is proven for 'single-59 molecule high-resolution colocalization' (SHREC) analyses of individual molecules [12,13], 60 applicability of this approach to the kinetochore (SHREC on kinetochores, kSHREC [4]) was 61 originally justified by the layered-disc appearance of kinetochores in Electron Microscopy (EM) 62 [14]. More recent observations of significant alterations in the size and shape of the kinetochore 63 in response to various types of interactions with microtubules [2,3,15,16] challenge this 64 justification [8]. The ongoing debate on the importance of intrakinetochore tension (IKT) 65 prompted us to evaluate relative contributions of distances between kinetochore layers vs. 66 changes in shape and orientation of the kinetochores on the centromere towards the value of 67 Delta. To this end, we applied kSHREC analysis to the compound kinetochores of Indian 68 muntjac (IM) that formed by a lateral fusion of typical mammalian centromeres during evolution 69 of this species. IM kinetochores comprise the conventional thin trilaminar plate (75 nm); 70 however, the length of the plate exceeds 1.5 μm instead of ~0.3 μm observed in most 71 mammalian cells. The increased length makes the shape and orientation of the kinetochore 72 discernable in LM. Comparative kSHREC analyses in IM vs. human cells suggest that the 73 decrease of Delta observed in Taxol-treated human cells primarily reflects changes in the length 74 of curved kinetochore plates rather that the level of IKT. This prompts re-evaluation of the role 75 ascribed to IKT in the control of mitotic progression. 76

Advantages of Indian muntjac kinetochores for kSHREC analysis 78
The low number of chromosomes in Indian muntjac (IM) arose from the tandem fusion of 79 8 LM/immuno-EM analyses suggest that Hec1 molecules are spread in the direction of the 151 attached microtubules for greater than 200 nm in human cells [8]. A unique advantage of IM 152 kinetochore is that the width of the spatial domains occupied by various kinetochore proteins 153 can be directly measured as the Full Width at Half Maximum (FWHM) of the fluorescence peak 154 in line scans orthogonal to the orientation of the plate ( Figure 3A,B). 155 We find that FWHM of the Hec1 distribution within IM kinetochores during metaphase is 156 233±20 nm after GA fixation and 225±22 nm after PFA fixation ( Figure 3C). These values are 157 significantly larger than FWHM of line scans across individual microtubules (170±7 nm after GA 158 and 189±13 nm after PFA, Figure 3C), visualized with the same fluorophore and under identical 159 optical conditions. EM firmly establishes that the diameter of a microtubule in GA-fixed cells is 160 ~25 nm [25], which implies that FWHM of line scans across a microtubule is determined by the 161 diffraction-limited resolution of the optical system. Thus, the width of Hec1 distribution that 162 exceeds FWHM of diffraction-limited profile by 63 nm, reflects the true physical width of the 163 layer occupied by Hec1 within the kinetochore. Importantly, FWHM values are significantly more 164 variable among Hec1 than among microtubule profiles ( Figure 3C). This increased variability 165 supports the notion that the width of microtubule profiles is determined by the optics while the 166 Hec1 profiles reflect natural fluctuations in the organization of the kinetochore plate. We also 167 notice that the width of microtubule profiles in PFA-fixed samples is greater than after GA 168 fixation. This increase likely reflects disintegration of microtubule structure that results in spatial 169 redistribution of tubulin. Indeed, microtubules are not structurally detectable in EM on PFA-fixed 170 samples [25,26]. 171 Measurements in metaphase cells treated with Taxol demonstrate that abrogation of 172 centromere tension increases FWHM of Hec1 to 251±27 nm after GA and 241±30 nm after PFA 173 fixation. Complete depolymerization of microtubules with high concentration of nocodazole 174 results in a more prominent increase of Hec1 FWHM to 278±35 nm (GA) and 273±39 nm (PFA) 175 ( Figure 3C).

9
CenpA-GFP also localizes within a layer of measurable width. Because resolution is 177 proportional to the wavelength, FWHM is less for a diffraction-limited peak formed by a green 178 vs. red fluorophore. On the microscope used in this study, FWHM of microtubules visualized 179 with a green fluorophore (GFP or Alexa488) is 155±6 nm after GA and 172±11 nm after PFA 180 fixation ( Figure 3C). FWHM of CenpA-GFP peaks is significantly wider, measuring 208±18 nm 181 after GA fixation and 202±19 nm after PFA. The width of CenpA-GFP layer increases slightly to 182 216±24 nm (GA) and 210±26 nm (PFA) in Taxol-treated cells, and further increases to 224±24 183 nm (GA) and 222±30 nm (PFA) when microtubules are completely depolymerized with 184 nocodazole ( Figure 3C). 185 FWHM measurements demonstrate that both inner (CenpA) and outer (Hec1) 186 kinetochore components reside within layers whose widths are significantly larger than the 187 diffraction limit of resolution and therefore the widths of these layers are measurable in LM. 188 Further, CenpA and Hec1 layers are approximately twofold larger than the distance between the 189 centers of the layers occupied by these proteins (i.e., Delta). These dimensions imply that 190 approximately half of CenpA and Hec1 molecules are spatially intermixed within the same 191 compartment, which in turn means that Delta does not accurately reflect the typical distance 192 between molecules within IM kinetochores. 193

CenpA and Hec1 spatially overlap within kinetochores in human cells. 194
Our observation that the width of the layers formed by inner and outer proteins within 195 compound kinetochores of IM exceeds 200 nm raises a question of whether a similar 196 architecture exists in human kinetochores. The underlying assumption in Delta measurements is 197 that kinetochore proteins form negligibly thin layers within an ~300-nm long plate [4]. A corollary 198 of this assumption is that FWHM of the kinetochore spots in LM should reflect the length of the 199 plate in one direction and be diffraction limited in the orthogonal direction. 200 To explore whether fluorescently labeled kinetochores in human cells resemble the 201 shape and dimensions assumed in kSHREC analyses, we constructed a computational 202 simulation in which the inner and outer kinetochore layers are modeled as a specified number of 203 'molecules' (points) randomly distributed within a 3-D volume of specified shape. This 204 distribution of molecules is convolved (blurred) with a 3-D Gaussian filter which mimics the point 205 spread function (PSF) of a light microscope. The blurred image is then scaled down to match 206 dimensions of voxels in a typical LM volume recorded on a CCD camera at specified Z-steps. 207 Fitting a Gaussian function to these simulated 3-D images of kinetochores is then used to 208 determine coordinates and FWHM of the peaks ( Figure 4A). 209 To establish PSF parameters, we recorded 3-D volumes of multi-color 100-nm 210 fluorescent beads under the same optical conditions as kinetochores ( Figure 4B Figure 4C). In contrast to this prediction, kinetochores appear as more 219 rounded and FWHM of both red and green spots is significantly larger than the diffraction limit, 220 which indicates a noticeable overlap between the inner and outer domains ( Figure 4D - Figure  221 supplement 1, also see [8]). This appearance implies that the width of kinetochore layers is 222 similar to the length of the plate. Indeed, computational modelling predicts that for ~300 long 223 kinetochore plates the width of both inner and outer layers need to be ~250 nm to match the 224 typical appearance of human kinetochores ( Figure 4E). This value is similar to the widths of 225 kinetochore layers measured in the compound IM kinetochores ( Figure 3C).
Another feature inconsistent with the notion that proteins form relatively thin layers in 227 human kinetochores is the orientation of the longer axis, particularly for the outer kinetochore 228 components (Hec1, Figure 4  Canonical interpretation of Delta as metrics for the distance between layers of molecules 238 assumes that the overall shape and dimensions of the kinetochore plate remain constant. 239 Previous computational analyses suggest that for inflexible layers Delta accurately reflects the 240 distance even if the layers are slanted or jagged [4]. However, for curved layers, the value of 241 Delta is smaller than the physical distance between the layers and the magnitude of this 242 difference is greater for curvatures with smaller radii and/or longer plates ( Figure 5A, also see 243 reference [8]). Systematic analysis of the kinetochore plate length and curvature in EM 244 preparations has not been reported, which prompted us to assess these parameters in serial  Figure 5B'). This approach yields more uniform results than direct fitting of the plate with a 257 circular arc. The latter produces large residual errors due to minute irregularities in the shape of 258 the plate, particularly when the plate is short (our unpublished observation). We find that a 259 typical outer kinetochore plate in untreated cells is encompassed by a 270±47-nm by 61±18-nm 260 rectangle while Taxol-treated kinetochores require a 350±57-nm by 110±39-nm bounding box 261 ( Figure 5D). Assuming a perfectly circular arc, these dimensions suggest that typical 262 kinetochore curvatures are ~180 nm for the untreated and ~194 nm for Taxol-treated cells 263 ( Figure 5D). The difference between these values is not statistically significant. 264 Thus, EM analyses suggest that a typical kinetochore plate is an arc with the curvature 265 radius of ~190 nm. Taxol does not significantly change the curvature, but it prominently 266 increases the length of the plate. To more clearly understand how these geometric features of 267 the kinetochore plate influence the results of FWHM analyses, we employed a computational 268 model. The model makes two major predictions. First, to achieve ~145-nm Delta observed in 269 untreated RPE1 cells, kinetochore layers need to be separated by ~155 nm ( Figure 5E). 270 Second, elongation of the plate with fixed curvature leads to a decrease in Delta from 145 nm to 271 125 nm ( Figure 5F). The model also predicts LM appearance of kinetochore spots that 272 resembles appearance of real kinetochores in untreated and Taxol-treated RPE1 cells (  distances between the spots formed by various kinetochore proteins (i.e., Delta) laid the 282 foundation of a linear map that ascribed various proteins to specific layers within the kinetochore 283 and interpreted Delta as the distance that separates these layers [4]. Essential for this 284 interpretation of Delta was the assumption that kinetochores are flat thin plates whose shape 285 and dimensions remain constant upon Taxol treatment. Here we test this assumption and find 286 that a typical kinetochore plate in human cells is curved, which implies that Delta 287 underestimates the distance between the layers. We also find that the length of curved 288 kinetochore plates increases in Taxol-treated cells. This type of architectural reorganization 289 inevitably decreases the value of Delta even if the distance separating inner and outer 290 kinetochore layers remains constant. Thus, our data demonstrate that Delta is not directly 291 proportional to IKT. 292 Our observation of a significant plate curvature is consistent with early EM descriptions 293 of kinetochores as crescents with the convex side facing away from the centromere [reviewed in 294 28]. Indeed, kinetochore plates in human cells cannot be truly flat because they protrude from 295 the surface of a cylindrical centromere, whose relaxed radius is less than 400 nm. A force 296 pulling kinetochores outwards is expected to increase the plate's curvature and our EM results 297 suggest that the typical radius of kinetochore plates visualized via conventional EM is less than 298 200 nm ( Figure 5). At this curvature Delta would underestimate the distance between negligibly 299 negligibly thin, and the widths of the layers are not constant -they increase under conditions 301 that decrease Delta (Figure 3, also see [8]). These complex architectural changes further 302 complicate quantitative interpretation of Delta. 303 FWHM measurements of Hec1 and CenpA spots in human cells also suggest a 304 significant spatial overlap between the inner and outer proteins. Our data demonstrate that 305 human kinetochores are much larger than diffraction-limited spots in all directions, i.e., the width 306 of the kinetochore is as large as its length. This observation is supported by computational 307 simulations with experimentally-determined parameters of PSF. These simulations predict that  Figure 2 and [4]) render the first possibility unlikely. Therefore, Delta in human cells appears to reflect 327 primarily larger scale deformations of malleable outer plate rather than elastic stretching within 328 the plate (i.e., IKT). Our interpretation gains additional support from observations of different 329 Delta values in sister kinetochores [8,27,29]. It is impossible for linearly connected elastic 330 elements to simultaneously experience different levels of tension. In contrast, changes in the 331 shape of kinetochore plate are expected to be independent for sister kinetochores and 332 prominent differences in the shape of sister kinetochores is directly observed in EM ( reports [4,5,8,10]. In part, this lack of consensus may be due to different approaches to Delta 342 measurements employed in different laboratories and there is a thoughtful debate on the 343 conditions necessary for accurate localization of fluorescent spots in cells [5,11]. However, a 344 larger issue is whether the size of kinetochore is sufficiently small and whether the shape of this 345 organelle is sufficiently invariant to allow unequivocal interpretation of Delta as the distance 346 between various molecules. This issue has not been considered since the introduction of 347 Cells were concurrently permeabilized and fixed in PEM buffer (100 mM Pipes, pH 6.9, 2.5 mM 374 EGTA, and 5 mM MgCl2, pH 6.9) supplemented with 1% Triton X-100 and 1% glutaraldehyde 375 (G5882; Sigma-Aldrich) or 1% Triton X-100 and 3. Immunostained cells were embedded in non-solidifying media containing 90% glycerol, 391 10% 1M Tris-Cl, pH 8.5, and 1 mg/ml PPDA. Wide-field fluorescence images were obtained on 392 a Nikon TE2000E2 microscope equipped with 100x 1.49 NA PlanApo TIRF lens and LED 393 illuminator (CoolLED PE 4000). Images were captured with an Andor Zyla 4.2 camera at 43-nm 394 X-Y pixels and 200-nm Z steps. The system was controlled by Nikon NIS-Elements Advanced 395 Research software (Nikon instruments). All images were deconvolved with SoftWorRx 5.0 396 deconvolution software (Applied Precision), and objective lens-specific in house-recorded point-397 spread functions. Deconvolution method was set to "Conservative", noise level to "Medium" for 398 kinetochores and "Low" for 100-nm beads, and the process ran for 10 iterations. 399 Serial section electron microscopy 400 IM and RPE1 cells were fixed with 2.5% glutaraldehyde (G5882; Sigma-Aldrich) in PBS, pH7.4 401 for 30 min, rinsed with PBS (3 X 5 min), and post-fixed with 2% OsO4 in dH2O for 60 min at 4°C. 402 The coverslips were then rinsed in dH2O, treated with 0.25% tannic acid for 20 min, and stained with 2% uranyl acetate for 60 min. Dehydration was achieved by a series of ethanol solutions 404 (30-50-70-80%-96%, 10 min in each solution) followed by acetone (10 min peakfit-m). Profiles with fit errors larger than 3% were discarded (less than 3% of all profiles). 420 Interkinetochore distance (IKD) was calculated by subtracting the distance between the 421 centers of CENP-A or Hec1 peaks. Delta values were calculated by subtracting the distance 422 between the centers of CENP-A peaks from the distance between the centers of Ndc80/Hec1 423 peaks and dividing the result by 2. In this approach, potential errors due to chromatic aberration 424 are automatically negated by the opposite orientation of sister kinetochores. However, the same 425 (average) value of Delta is assigned to both sister kinetochores. 426 shifting one color channel so that the global centers of mass for the red and green channels 435 coincide. Delta is then calculated directly as the distance between the red and green centroids 436 within the same kinetochore. 437

Measurement of the outer plate length and curvature 438
Length of the kinetochore outer plate was determined by tracing the contours of the plate with 439 the Freehand line tool in ImageJ (National Institutes of Health). To determine curvature of the 440 outer plate, a minimal rectangular box that encompasses the entire plate was drawn with the 441 Rectangle tool of ImageJ. For a perfectly circular segment, curvature radius of an arc constraint 442 within a rectangular box is calculated as R = W 2 /8H+H/2 where W is the length of the rectangle 443 that defines the chord, and H is the height of the rectangle that defines the sagitta of the arc. 444

Computational simulation for evaluation LM appearance of kinetochores with various 445
architecture 446 The simulation is made using a Matlab code, and the following parameters are considered. 447 Kinetochores are modeled as two plates, each consisting of fixed number of molecules 448 uniformly distributed randomly within a 3-D volume of specified dimensions and position relative 449 to the other plate with the possibility of overlap. The volume is rectangular except for curvature 450 in either the XY or XZ planes with assumed rotational symmetry with respect to both the Z and 451 Y axes. Two PSFs for "green" and "red" colors are constructed by convolving a single point with 452    Taxol-and  597 nocodazole-treated cells. Increase in CenpA-GFP layer is apparent after PFA but not after GA 598 fixation. Student's t-test p values are less than 10 -4 (****), 10 -5 (*****), or greater than 0.5 (NS). 599        and immunostaining for Hec1 (red). Double arrows denote orientation of the longer axis (in XY) for Hec1 spots. Notice that kinetochore spots are larger than spots formed by 100-nm beads (shown in the bottom row). The shape of kinetochore spots and orientation of the longer axis are variable. Kinetochores on chromosomes 4, 7-9, 11, 27, 30, 33, and 38-42 could not be segmented due to spatial overlap. These kinetochores were not considered in calculation of Delta, FWHM, and orientation of the larger axis. 3-D volume of the entire cell that depicts orientation of the spindle and positions of individual chromosomes/kinetochores is shown in Video 3.