CENP-C/H/I/K/M/T/W/N/L and hMis12 but not CENP-S/X participate in complex formation in the nucleoplasm of living human interphase cells outside centromeres

Kinetochore proteins assemble onto centromeric chromatin and regulate DNA segregation during cell division. The inner kinetochore proteins bind centromeres while most outer kinetochore proteins assemble at centromeres during mitosis, connecting the complex to microtubules. Here, we measured the co-migration between protein pairs of the constitutive centromere associated network (CCAN) and hMis12 complexes by fluorescence cross-correlation spectroscopy (FCCS) in the nucleoplasm outside centromeres in living human interphase cells. FCCS is a method that can tell if in living cells two differently fluorescently labelled molecules migrate independently, or co-migrate and thus are part of one and the same soluble complex. We also determined the apparent dissociation constants (Kd) of the hetero-dimers CENP-T/W and CENP-S/X. We measured co-migration between CENP-K and CENP-T as well as between CENP-M and CENP-T but not between CENP-T/W and CENP-S/X. Furthermore, CENP-C co-migrated with CENP-H, and CENP-K with CENP-N as well as with CENP-L. Thus, in the nucleoplasm outside centromeres, a large fraction of the CENP-H/I/K/M proteins interact with CENP-C, CENP-N/L and CENP-T/W but not with CENP-S/X. Our FCCS analysis of the Mis12 complex showed that hMis12, Nsl1, Dsn1 and Nnf1 also form a complex outside centromeres of which at least hMis12 associated with the CENP-C/H/I/K/M/T/W/N/L complex.


Introduction
Chromosome segregation is executed by a conserved molecular machinery which contains a large number of subunits and recruits many additional regulatory proteins (recently reviewed in [1][2][3][4][5]). A multi-protein complex, the "kinetochore", assembles onto centromeric chromatin [6][7][8][9][10][11][12][13][14][15]. During mitosis, the kinetochore mediates the interaction between DNA and the mitotic spindle [16][17][18][19][20][21][22][23][24][25][26][27][28] ( [90]. FCCS analysis can identify if two differently labeled molecules move together (in a complex) or independently, indicating direct or indirect protein-protein interaction. Fluorescence correlation spectroscopy (FCS) analyses diffusing particles (summarized in [117]) by monitoring the fluorescence fluctuations created when molecules diffuse through a diffraction limited confocal detection volume. Photons emitted from the fluorescent particles are counted continuously over time. While the fluctuation amplitude depends on particle concentration and brightness, its frequency contains information on the diffusion times of the fluorescent particles. For quantitative evaluation, the fluctuation frequency is correlated with a time-shifted replica of itself (autocorrelation). The amplitude of the autocorrelation curve is inversely proportional to the average number of fluorescent molecules in the confocal volume. By monitoring the diffusion coefficient FCS can determine molecular association and binding affinities. However, the diffusion coefficient is inversely proportional to the hydrodynamic radius of a particle and thus weakly dependent on its overall mass. Therefore, it is difficult to determine the association of similar sized particles. This limitation can be circumvented by fluorescence cross-correlation spectroscopy (FCCS). When two differently labeled molecules are measured in the same volume and correlated by cross-correlation, the analysis can tell if both molecules move together (in a complex) or independently, enabling the deduction of dissociation constants (K d ) of binding reactions in vivo [117].
Here, we extend our FCCS studies in living human cells [90,97] to a number of other CCAN and hMis12 complex proteins. We found that in the nucleoplasm outside kinetochores CENP-T co-migrates with CENP-K and CENP-M. The in vivo interactions between CENP-T and CENP-W as well as between CENP-S and CENP-X are studied in detail. We found that a large fraction of the CCAN proteins CENP-C/H/I/K/M/T/W/N/L but not CENP-S/X interact with one another in vivo outside centromeres, supporting the findings of Weir et al. [15].

Cell lines and transfection
HEp-2 and U2OS cells were cultured in DMEM with 10% FCS as previously described [79,119]. For the dual color fluorescence cross-correlation spectroscopy (DC-FCCS) experiments, appropriate constructs were transfected into HEp-2 and U2OS cells (ATCC, Manassas, VA, USA) by electroporation and interphase cells were assayed by DC-FCCS after 24-48 hrs.
For single-wavelength fluorescence cross-correlation spectroscopy (SW-FCCS), HEK293 cells (ATCC) were cultivated in Dulbecco's modified Eagles's medium DMEM (Hyclone, GE Healthcare, UK) supplemented with 10% FBS (Hyclone fetal bovine serum, GE Healthcare, UK) and 1% PS, penicillin G and streptomycin (PAA, Austria) at 37˚C in 5% CO 2 atmosphere. Appropriate plasmids were transfected into the cells by electroporation (Neon R Transfection system, Life Technologies, Carlsbad, CA). Around 90% confluent cells were washed twice with 5 ml 1×PBS (phosphate buffer saline), treated with 0.5 ml Trypsin for 1 min and re-suspended in 5 ml DMEM culture medium. Cells (1×10 6 ) were centrifuged for 3 min, re-suspended in R buffer and transfected according to the manufacturer's protocol. The transfected cells were seeded in DMEM and 10% FBS culture medium on glass coverslips (MatTEK, Ashland, US) and kept at 37˚C in 5% CO 2 atmosphere overnight. Before starting the experiments, the cells were washed with 1×PBS and maintained in 1 ml 1×PBS.
Olympus FV300 confocal microscope with a 1.2 NA 60x water immersion objective (Upla-nApo, Olympus, Japan). EGFP-and mCherry-labelled proteins were excited by a 20 μW Argon ion 514 nm laser line (Melles Griot, Albuquerque, NM, USA). The emission passed a 150 μm pinhole, split by a 560 DCLP emission dichroic mirror (Omega Optical, Brattleboro, VT, USA) followed by the band-pass filters (545AF35 and 615DF45, respectively, Omega Optical, Brattleboro, VT, USA) to the two avalanche photodiodes (SPCM-AQR-14; PerkinElmer, Canada). Auto-and cross-correlation curves were generated by a hardware correlator (Flex02-01D, www.correlator.com, Bridgewater, NJ, USA). The experiments were conducted at 37˚C (TempContro 37-2, Pecon, Erbach, Germany) equipped with an objective heating ring (TC-124A, Warner Instruments, Hamden, CT, USA). Each single-point measurement recorded 3 times 25 sec runs in the nucleoplasm. The obtained auto-and cross-correlation functions (ACF and CCF) were fitted with a self-written software in Igor Pro 6.22A (WaveMetrics, Lake Oswego, OR, USA) using a 3-dimensional 1-particle 1-triplet model [124]. The concentration calculations and dimer fraction quantification was conducted by a self-written program in Mathematica 10 (Wolfram Research, Champaign, IL, USA). For the calculation of apparent dissociation constants (K d ) the data were corrected for the probability of mCherry to be fluorescent (p r ) and for the differences in the effective observation volumes at different wavelength. The detailed protocols for the correction have been published previously [124]. EGFP was assumed to be fully fluorescent (p g = 1). The green observation volume V G was 0.54 fl as determined by calibration of the system with Atto488. The red observation volume V R was determined by the ratio of diffusion times of mCherry/EGFP (τ D,R /τ D,G ) from the measurement of a tandem EGFP-mCherry construct. The ratio of τ D,R /τ D,G was 1.1, which translated into a V R / The overlap of these two observation volumes gave an effective cross-correlation volume V x of 0.58 femto-litre (fl). These parameters resulted in a p r = 0.6, implying that slightly more than half of all mCherry fluoresce. The fraction of non-fluorescent mCherry proteins limits the maximum cross-correlation that can be achieved. 16 CCAN proteins, forming the sub-complexes CENP-T/W/S/X, CENP-H/I/K/M, CENP-L/N and CENP-P/O/R/Q/U, assemble at the centromere during all phases of the cell cycle. We studied if these proteins and the four proteins of the hMis12 complex form pre-complexes in the nucleoplasm outside centromeres.

Results and discussion
Recently we analysed protein interactions between proteins of the CENP-P/O/R/Q/U complex and their interactions with other CCAN proteins [90]. By Yeast-2-Hybrid (Y2H) analysis of these human proteins in S. cerevisiae we could only detect a weak interaction of CENP-R with itself. However, using a mammalian three-hybrid (F3H) [125,126] assay for protein-protein interactions in living human cells at non-centromeric chromatin locations, we detected the interactions between CENP-O and CENP-P as well as CENP-R, between CENP-U and CENP-P, CENP-R and CENP-Q and between CENP-Q and CENP-R. When analysing the diffusible proteins of the CENP-P/O/R/Q/U complex individually in the nucleoplasm of human cells by dual wavelengths fluorescence cross-correlation spectroscopy (DC-FCCS), we observed co-migration between CENP-O and CENP-P (40-50%) and between CENP-Q and CENP-R (16-20%) (see Table 1) but not between other complex members [90], suggesting that the CENP-P/O/R/Q/U proteins only partially pre-form sub-complexes but not a fivefold complex in the nucleoplasm outside centromeres.
Here we analysed potential protein-protein interactions among further CCAN proteins in the nucleoplasm of living human interphase cells outside centromeres. We double-transfected  . The original data behind "means" and "SD" are listed in S1 Table. Since cross-correlation values under-estimate the factual percentage of co-migrating molecules approximately by a factor two, due to various effects including non-fluorescent proteins and differences in the size of the diffraction limited volume for the two measured wavelengths, the values measured with DC-FCCS had to be corrected (see main text and reference [124]). Ã ) [83] # ) [90] human HEp-2 cells so that they express two CCAN proteins, terminally tagged with EGFP in the one and mCherry in the other case. By DC-FCCS we then measured the mobility of both tagged proteins in a very small confocal volume (ca. 1 fl, about half the size of E. coli) and, by cross-correlation of the fluorescence signal, we determined if in the focal volume the two proteins co-migrate which would indicate both proteins being part of a di-or multimeric complex [90,117,[127][128][129]. FCCS possesses single-molecule sensitivity and can be conducted at concentrations in the nanomolar range. Tagged proteins were introduced by transient transfection and exhibited centromeric targeting independently of the N-or C-terminal site of fluorescent protein fusion. Before data acquisition, the selected focal volume was bleached, destroying the fluorescence of immobile fluorophores. Thus, after bleaching, only mobile fluorophores (like those tagged to CENPs) can contribute to the DC-FCCS signal. Furthermore, cross talk was minimized by microscope adjustment. Experiments with control fluorochromes in monomeric (EGFP and mRFP) or fused (mRFP-EGFP) forms determined the dynamic range of this system. Co-diffusion generally was detected as described in detail by Bacia & Schwille [120,121]. First we analyzed EGFP and mRFP as single molecules (negative control, Fig 1A). The count rates of EGFP and mRFP over time, reflecting the fluctuations of the analyzed fluorescent molecules, are displayed in insert b). With the measured signals, autocorrelation analyses were performed (measured signals are compared with itself at later time points and analyzed for repeated motifs) resulting in the FCS-or autocorrelation-curves G (τ) for EGFP (green) and mRFP (red) with calculated autocorrelations (AC) for EGFP (AC EGFP = 1.151) and mRFP (AC mRFP = 1.057). By subtraction of 1 from the AC values, the corresponding amplitudes of the autocorrelation curves A(AC EGFP ) = 0.151 and A(AC mRFP ) = 0.057 are obtained. In addition, cross-correlation analyses were performed (measured signals of the green channel were compared with measured signals of the red channel at identical time points and analyzed for identic motifs), resulting in the FCCS-or cross-correlation-curve G (τ) (black). In this example, the calculated cross-correlation (CC) of 1.001 and the amplitude A(CC) = 0.001 indicate 0% co-migration of EGFP with mRFP. In all experiments with this pair of single proteins, the measured cross-correlation values never indicated co-diffusion. For the mRFP-EGFP fusion protein (positive control, Fig 1B) we observed a CC = 1.029, an AC EGFP = 1.064 and an AC mRFP = 1.142. The corresponding amplitudes of the autocorrelation and cross-correlation curves are A(AC EGFP ) = 0.064, A (AC mRFP ) = 0.142. and A(CC) = 0.029. A(CC), relative to the diffusion related amplitude of one of the autocorrelation curves (A(AC EGFP ) in this case), is a measure of binding or dynamic co-localization indicating 45% co-migration. The A(AC)s correspond to the reciprocals of the average numbers of particles in the detection volume, resulting here in 15.6 particles EGFP (N EGFP ) and 7 particles mCherry (N mCherry ). The A(CC), however, is directly proportional to the concentration of the co-migrating molecules (N EGFP-mCherry ). The N EGFP-mCherry can be calculated according to the ratio A(CC)/A(AC mCherry ) = N EGFP-mCherry /N EGFP resulting in about 3.2 co-migrating molecules ((N EGFP-mCherry = 3.2). In addition, N EGFP-mCherry (= 3.2) relative to the number of analyzed molecules of one of the two fluorophores (in this case N mCherry = 7) also indicated 45% co-migration. In general, we obtained CC-values between 40 and 55% for mRFP-EGFP and EGFP-mCherry fusion proteins [90,97]. These results are in agreement with results of Kohl & Schwille [130]. For such fusion proteins, 100% cross-correlation should be observed. Since cross-correlation values obtained with these protein pairs rate the factual percentage of co-migrating molecules too low by approximately a factor of two, we corrected the values measured with DC-FCCS by this factor. The lower values of 40-55% is due to at least four effects [124]: (i) differences in observation volumes for fluorescent proteins of different emission wavelengths and displacement of these observation volumes due to chromatic aberrations, (ii) the existence of dark fluorescent proteins due to maturation problems [130] and photo-bleaching, (iii) Förster resonance energy transfer (FRET) between the labels changing the amplitudes of auto-and cross-correlations, and finally (iv), endogenous proteins might interfere with the analysed biomolecular interactions. The dominant reason seems to be that roughly only about half of all mCherry fluoresce (see Materials & Methods). Thus, crosscorrelation values obtained from DC-FCCS experiments under-estimate the percentage of comigrating molecules. In order to avoid these effects, we carry out single wavelength fluorescence cross-correlation (SW-FCCS) experiments (see below). In addition, the fluorescent tag itself, in particular when being tied by a short linker, might interfere with protein-protein interaction, for example by being attached at or close to the binding site, introducing sterical hindrance. Therefore, when detecting no protein-protein interaction, in most cases additional cross-correlation experiments were carried out with (i) proteins labeled at the other terminus and/or (ii) introducing a long protein linker between EGFP and CENP, reducing or excluding the effects of sterical hindrance by the tagged fluorescent protein.

CENP-H/I/K/M complex
As a further CCAN sub-complex next to CENP-P/O/R/Q/U [90], we analysed the potential pairwise interactions between proteins of the CENP-H/I/K/M sub-complex [35,92] in order to determine if CENP-H/I/K/M proteins form hetero-dimers or -multimers in the nucleoplasm outside centromeres. By Y2H we only detected a single weak interaction between CENP-H and -K. In double-transfected interphase HEp-2 cells, we analysed various combinations of EGFP and mCherry tagged CENP-H/I/K/M proteins ( Table 1). Measurements of EGFP-(l)-CENP-M and mCherry-(s)-CENP-K in the nucleoplasm (CENP-M was fused to EGFP via a long linker, whereas CENP-K and mCherry were connected by a short linker, see Materials & Methods) showed individual autocorrelation values of 1.101 (EGFP) and 1.130 (mCherry). Cross-correlation measurements resulted in a CC = 1.050, indicating that in this cell 49% of the molecules were co-migrating ( Fig 1C, Table 1). In total, from 21 FCCS measurements carried out in the nucleoplasm of 13 cells, 19 measurements in 12 cells indicated that outside kinetochores (37±9)% of EGFP-(l)-CENP-M and mCherry-(s)-CENP-K were co-diffusing and thus co-resident in a single complex. Since cross-correlation values under-estimate the percentage of co-migrating molecules (see above) roughly by the factor two, hetero-dimerisation of EGFP-(l)-CENP-M and mCherry-(s)-CENP-K is expected to be twofold higher than the calculated (37±9)%; we estimate a corrected value of about 56-92%. When measuring the crosscorrelation between EGFP-(s)-CENP-I and mCherry-(s)-CENP-M as well as between CENP-I-(s)-EGFP and CENP-M-(s)-mCherry (all fusions with short linkers), we found no cross-correlation (Table 1). We speculated that the fluorescent tag might interfere with protein-protein interaction (see above). In order to test this hypothesis, we introduced a long protein linker between EGFP and CENP-M and repeated the cross-correlation experiment, now between mCherry-(s)-CENP-I and EGFP-(l)-CENP-M. In this case, when EGFP is fused to CENP-M by a long linker, we found (24±9)% of the molecules co-migrating (corrected value 30-66%) ( Table 1). In further experiments we measured the cross-correlation, and found comigration, between EGFP-(s)-CENP-H and CENP-I-(s)-mCherry, between EGFP-(s)-CENP-H and mCherry-(s)-CENP-K and between EGFP-(s)-CENP-K and mCherry-(s)-CENP-I (Table 1). By FCCS in the nucleoplasm outside centromeres, we estimate pairwise co-migration of CENP-H/I/K/M proteins in the range 30 to >90% (corrected values, considering the under-estimation discussed above; see Fig 1B and 1C). This suggests that in the nucleoplasm of interphase cells outside centromeres, most of the CENP-H/I/K/M proteins (30 to >90%) form a multimeric complex.
Our results are in agreement with earlier observations. CENP-H [67,131,132] and CENP-K [35,68,100,133] form a tight dimer when co-expressed in and purified from insect cells [92]. This stabilisation is functionally relevant since CENP-M and CENP-I are mutually required for kinetochore localization. Thus, these proteins form a quaternary complex [12,13] which, as indicated by our data here, partially forms already in the nucleoplasm outside centromeres (see Fig 5).

CENP-C
CENP-C is required for the incorporation and stabilisation of CENP-A nucleosomes [63,81], has a central CCAN forming function [12] and recruits the CENP-T/W/S/X and CENP-H/I/ K/M complexes to centromeres. CENP-C binds the CENP-L/N and CENP-H/I/K/M complexes directly, with CENP-H and CENP-K being the main determinants of this interaction [12][13][14][15]92]. By DC-FCCS we therefore measured co-migration between EGFP-(s)-CENP-H and CENP-C tagged at either end (mCherry-(s)-CENP-C and CENP-C-(s)-mCherry). We found co-migration between CENP-H and CENP-C in the range of 16 to 66% (corrected, Table 1). This indicates that in the nucleoplasm outside centromeres, CENP-C interacts with CENP-H (either directly or indirectly), and suggests that there CENP-C is associated with the CENP-H/I/K/M complex, consistent with [15] (see Fig 5).

CENP-N/L complex
Next we studied the CENP-N/L complex. CENP-N associates with CENP-A nucleosomes and directly binds CENP-L [12,15,75,79,80]. By DC-FCCS we identified a cross-correlation between EGFP-(l)-CENP-L and mCherry-(s)-CENP-N ((23±9)% co-migration, corrected value 28-64%) ( Table 1), indicating that most of the CENP-N and CENP-L proteins comigrate. Thus, our data show that to a considerable amount CENP-N and CENP-L already interact in the nucleoplasm outside centromeres. CENP-K and CENP-N interact with one another [75], probably inducing an interaction between the complexes CENP-L/N and CENP-H/I/K/M [12,15]. We therefore measured DC-FCCS between EGFP-(l)-CENP-L and mCherry-(s)-CENP-K and between CENP-K-(s)-EGFP and mCherry-(s)-CENP-N, and found co-migration between 26 and 78% (corrected, see Table 1). The results indicate that in the nucleoplasm about one to two thirds of the available CENP-N and CENP-L proteins (most of them complexed together, see above) associated with proteins of the CENP-H/I/K/M complex. This is consistent with the recent observation that CENP-N seems to be associated with CENP-H/I/K/M, however not always [96]. Interestingly, DC-FCCS between EGFP-(s)-CENP-N and mCherry-(s)-CENP-K showed no co-migration (Table 1). The CENP-K N-terminal, but not the C-terminal tag, fused by a short linker, seems to interfere with binding to CENP-N. In summary, in the nucleoplasm outside centromeres, the CENP-L/N complex is partly associated with the CENP-H/I/K/M complex (see Fig 5).

CENP-T/W/S/X complex
We then studied the interactions between CENP-T/W/S/X proteins [91] in the nucleoplasm not only by DC-FCCS (as above) but also in more detail by quantitative SW-FCCS [124,[135][136][137]. This approach allows the estimation of effective dissociation constants. In principle, the ratio of the auto-(ACF) and cross-(CCF) correlation amplitudes can vary between zero (no binding) and one (complete binding (assuming 1:1 stoichiometry)). Factors affecting the correlation amplitudes and their ratios include background fluorescence (which in our DC-FCCS experiments was eliminated by bleaching the focal volume before data acquisition) and cross talk [138][139][140] (which was minimized by microscope adjustment). Nevertheless, the largest values for the CCF/ACF amplitude ratios for fused fluorescent proteins (like EGFP-mRFP or EGFP-mCherry) were in the range of~0.5 (see Fig 1, Table 1), due to differences in observation volumes for different fluorescent proteins, the existence of dark fluorescent proteins and photo-bleaching, Förster resonance energy transfer (FRET), and endogenous protein interference with the analysed biomolecular interactions (see above). Recently, we established procedures for correcting these effects [124].
Single wavelength FCCS (SW-FCCS) overcomes the difficulty of aligning two lasers to the same spot [135][136][137]. Further quantitative corrections allowed us to accurately determine the ACF and CCF amplitudes and thus obtain correct concentrations of the interacting molecules so that we can deduce K d values. K d values from in vivo experiments usually are effective dissociation constants since in many cases interactions in cells are not simple pure binary reactions but can be affected by the presence of other biomolecules. Even unspecific background reactions of low binding strength can influence the measured K d if the concentration of the unspecific binding partner is high. Nevertheless, in the absence of detailed knowledge of the composition and all interactions in a cell, this effective K d is a relevant parameter.
First, by SW-FCCS we analyzed mCherry and EGFP expressed in interphase HEK293 cells as single proteins (negative control) or expressed as the fused hetero-dimer mCherry-EGFP (positive control). The diffusion time τ D of mCherry-EGFP was slightly higher than those of the single proteins ( Table 2). The uncorrected amount of co-migration was (8.6±6.7)% for the single proteins (defining the background level) and (47±14)% for the hetero-dimer (Table 2), defining the maximally expected value.
Recently, by DC-FCCS we detected a cross-correlation between EGFP-(s)-CENP-X and mCherry-(s)-CENP-S in the nucleoplasm, suggesting that 60-100% (corrected value) of the molecules co-migrate and thus co-reside in a single complex (Table 1) [97]. By contrast, the FCCS analysis of CENP-S association with CENP-T revealed no detectable soluble complex containing these two proteins, consistent with our Y2H experiments which also did not detect an interaction between CENP-T and CENP-S. We concluded that, outside centromeres and ectopic chromatin sites, to a large extent CENP-S and CENP-X exist in a diffusible complex independent of CENP-T [97]. Here, we repeated this experiment, now fusing mCherry to CENP-S by a long linker. By DC-FCCS we measured a cross-correlation between EGFP-(s)-CENP-X and mCherry-(l)-CENP-S ((31±7)% co-migration, corrected value 48-76%) ( Table 1), confirming our previous results. Then, we exchanged the N-terminal tags and studied EGFP-(s)-CENP-S and mCherry-(s)-CENP-X by SW-FCCS. In 101 measurements in 36 cells, both tagged proteins showed similar diffusion times and high uncorrected co-migration of (42±12)% ( Table 2). The fusion proteins moved slower than the single fluorescent proteins. From our data we deduced an effective dissociation constant K d of (264±37) nM from the scatter plot and the similar value 287 nM (SD ln 0.5) from the histogram analysis (Fig 2,  Table 2). When correcting these data according to [124], we obtained (58±15)% as the relative amount of complex formation (Table 2). Thus, in agreement with our DC-FCCS results, most  1. The interacting proteins have both, a) a cross-correlation above the negative control and, b) similar diffusion coefficients for red-and green-labeled molecules.
2. The non-interacting proteins have a) cross-correlations close to or at the negative control and b) different diffusion coefficients for the red-and green-labeled molecules.
3. The proteins fulfill both conditions (as in 1), they have a) cross-correlation above the negative control and, b) similar diffusion coefficients for red-and green-labeled molecules. However, their cross-correlation is only slightly different from the negative control and thus their interaction cannot be ascertained under the measurement conditions. The original data behind "means" and "SD" are listed in S2 of the CENP-S and CENP-X molecules form a complex in the nucleoplasm of interphase cells outside centromeres.
In vitro, in the absence of CENP-T/W, CENP-S/X can assemble into a tetramer [91]. CENP-T/W or CENP-S/X complexes alone can induce negative supercoils into DNA similar to canonical histones, while the CENP-T/W/S/X complex induces positive supercoils [98].
Here we observed that the diffusion times of EGFP-(s)-CENP-X and mCherry-(s)-CENP-S are slower than those of the single proteins mRFP and EGFP but clearly faster than the diffusion time of mCherry-H4. Potentially, this suggests that in the nucleoplasm EGFP-(s)-CENP-X and mCherry-(s)-CENP-S are not or only to a small extent incorporated into chromatin structures. A freely diffusible CENP-S/X complex would support its non-centromeric functional activities at DNA repair sites [141,142].
CENP-T binds DNA in vitro and in vivo [70,98]. Thus, in the chromatin environment of the nucleoplasm, single CENP-T and CENP-T/W complexes envisage multiple DNA binding sites. This disturbs our binding analysis and complicates the interpretation of the measured effective K d for the interaction between CENP-T and CENP-W. We therefore constructed two separate CENP-T domains: the CENP-T mutants CENP-T ΔC (aas 1-395) and CENP-T ΔN (aas 396-561). We found no co-migration between the N-terminal domain of CENP-T (EGFP-(s)-CENP-T 1-395 ) and CENP-W (mCherry-(s)-CENP-W) ( Table 2), indicating that the C-terminal domain CENP-T 396-561 , containing the histone-fold domain (aas 376-561), interacts with CENP-W, supporting results by [70]. Next, by SW-FCCS we analysed the interaction between the C-terminal domain of CENP-T (EGFP-(s)-CENP-T 396-561 ) and mCherry-(s)-CENP-W. We found similar mobility in the nucleoplasm and clear co-migration of (24±17)% (corrected value) with an effective K d of (1.3±0.2) μM (scatter plot) and 944 nM (SD ln 1.9) (histogram) (corrected values, Fig 3, Table 2). Compared to wildtype, we thus found modified binding Pre-complex formation of kinetochore components outside centromeres between CENP-W and CENP-T ΔN , with a shifted effective K d . Taken together, for both complexes, CENP-T/W and CENP-S/X, in the nucleoplasm we measured apparent dissociation constants in the 1 μM range. At the centromere, we expect stronger binding to the kinetochore multi-protein complex due to additional protein-protein contacts by CENP proteins.
By DC-FCCS, in the nucleoplasm, we found no cross-correlation between EGFP-(s)-CENP-T and mCherry-(s)-CENP-S, between EGFP-(s)-CENP-T and CENP-S-(s)-mCherry, between EGFP-(s)-CENP-T and mCherry-(l)-CENP-S and between EGFP-(l)-CENP-T and mCherry-(l)-CENP-S (Table 1), independent of the linker lengths of the fusion proteins. Analysis of EGFP-(s)-CENP-T and mCherry-(s)-CENP-S in the nucleoplasm showed autocorrelation values AC EGFP = 1.322 and AC mCherry = 1.106. Cross-correlation measurements resulted in a CC = 1.001, indicating that the molecules were not co-migrating (Fig 4A,  Table 1). The localization of EGFP-(s)-CENP-T and mCherry-(s)-CENP-S in the nucleus of a living human HEp-2 cells, used for FCCS analysis, is shown in Fig 4B. Compared to centromeres, the fluorescence intensity in the nucleoplasm is rather weak. For the analyzed centromere in spot 1 the ratios of nucleoplasmic to centromeric fluorescence intensities was 1:43 for EGFP-(s)-CENP-T and 1:33 for mCherry-(s)-CENP-S. A(AC EGFP ) = 0.322 and A(AC mCherry ) = 0.106 correspond to the reciprocals of the average numbers of particles in the detection volumes. Accordingly 3.1 particles EGFP-(s)-CENP-T and 9.1 particles mCherry-(s)-CENP-S were found in the analyzed confocal detection volume. When knowing the effective detection volume, the concentrations of the samples can be deduced. Rüttinger et al. [143] described a linear dependence of the number of particles detected by FCS with the concentrations of the analyzed samples in a range between 0.5-100 nM. From these results, under their experimental conditions, they determined an effective detection volume of (1.0±0.1) fl for Atto-655 in H 2 O. Under our experimental conditions, with a purified EGFP solution with known concentration, for EGFP we determined an effective detection volume of roughly 0.8 fl. Due to the above described problems with insufficient maturation and stability of mRFP and mCherry, we could not experimentally determine the effective detection volume for these proteins. Due to the longer emission wavelengths we estimated them to be about 30% larger than the detection volume of EGFP. Thus, for calculations with mCherry we used an effective detection volume of 1.04 fl. Determining the effective detection volume of the fusion proteins in living cells is more complex due to different local compositions and viscosity, and additional interacting components, influencing the mobility of the proteins. Therefore, for DC-FCCS experiments the intracellular concentrations were only estimated. Using the effective detections volumes measured in solution (see above), nucleoplasmic protein concentrations were deduced to be 6 nM for EGFP-(s)-CENP-T and 14 nM for mCherry-(s)-CENP-S.
We neither found co-migration between EGFP-(s)-CENP-T and mCherry-(s)-CENP-X nor between EGFP-(l)-CENP-W and CENP-X-(l)-mCherry (Table 1). Using different linker lengths and a switched tag for CENP-X, we confirmed the latter result by SW-FCCS. Furthermore, we found no co-migration between EGFP-(s)-CENP-S and mCherry-(s)-CENP-W ( Table 2). In SW-FCCS, for mCherry-(s)-CENP-T/EGFP-(s)-CENP-X and EGFP-(s)-CENP-S/mCherry-(s)-CENP-T the cross-correlations are slightly above the negative control value, and the diffusion times in both channels are close ( Table 2). According to these data, we cannot rule out possibly very weak interactions between these two protein pairs in the nucleoplasm. In summary, our results indicate that, consistent with earlier observations [23, 89,91], CENP-T and CENP-W as well as CENP-S and CENP-X form two separate hetero-dimeric complexes which, in the nucleoplasm outside centromeres, practically do not co-migrate as a joint larger complex (see Fig 5).
Human CENP-T binds to Spc24/25 of the human Ndc80 complex [21][22][23][24]84]. This interaction is influenced by CENP-T phosphorylation [21,22]: CENP-T is phosphorylated on multiple cyclin dependent kinase (CDK) consensus sites in G2 of HeLa cells and remains phosphorylated until anaphase. In vivo, CENP-T phosphorylation is important since its elimination prevents Ndc80 localization and causes defects in chromosome segregation [83]. However, also the un-phosphorylated human CENP-T 76-106 N-terminal fragment binds to the human Spc24 137-197 /Spc25 129-224 complex (with a K d of 645 nM) while phospho-mimetic human CENP-T shows an increased K d of 150 nM [84]. In the nucleoplasm of interphase cells, Fig 5. Protein-protein co-migration of CCAN and Mis12 proteins. The degree of co-migration (corrected) in the nucleoplasm outside centromeres of human interphase cells is color-coded (red dashed arrows: no or hardly detectable co-migration, green arrow: 5-30%, blue arrows: 30-60%, black arrows: above 60% co-migration). Please note: these green, blue or black arrows do not necessarily indicate direct protein-protein interaction; the labeled proteins comigrate in a complex: their interaction might be either direct or mediated by a third (or more) protein.
by SW-FCCS we analysed EGFP-(s)-Spc24 and mCherry-(s)-CENP-T and detected a very weak cross-correlation; we deduced a K d of 1.8 μM (Table 2). EGFP-(s)-Spc25 and mCherry-(s)-CENP-T 1-395 showed an even weaker cross-correlation with a K d of (3.0±0.9) μM (scatter plot) and 4.3 μM (SD ln 0.9) ( Table 2). These values are consistent with in vitro results [24,84]. That these weak cross-correlations slightly above background level might indeed indicate comigration and complex formation is supported by our observation that in both cases the fusion proteins move with similar diffusion times.
CENP-T/W interacts with the CENP-H/I/K/M complex, an interaction mediated by CENP-H/K [12,13,15,21,23,70,92]. In vitro, CENP-T/W was found to bind CENP-K [83]. By DC-FCCS we studied CENP-K-(s)-EGFP and CENP-T-(s)-mCherry and found (31±13)% comigration (corrected value 36-88%). Also for EGFP-(s)-CENP-M and CENP-T-(s)-mCherry we measured 20-32% (corrected value) co-migration (Table 1). A co-migration with CENP-M was not detected when replacing full-length CENP-T by its N-terminal domain CENP-T ΔC (CENP-T 1-395 -(s)-mCherry). Thus, the co-migration of CENP-T with CENP-M is caused by its C-terminal domain (in agreement with yeast data [23]). Our data suggest that, in the nucleoplasm outside kinetochores, the proteins of the CENP-H/I/K/M complex form a complex with the CENP-T/W hetero-complex (see Fig 5). The measured pairwise co-migrations do not necessarily indicate a direct protein-protein interaction between the two analysed proteins: the detected co-migration might be mediated by a third protein (or more) in the nucleoplasm which probably would be a CENP protein. In the nucleoplasm, the hetero-dimer CENP-S/X does not bind to the CENP-C/H/I/K/M/T/W proteins ( Fig 5); it becomes part of the CCAN kinetochore complex only at centromeres. Recently we observed by fluorescence-three-hybrid studies (F3H) that in vivo CENP-K is able to recruit CENP-O and CENP-U to non-centromeric chromatin sites, and that CENP-L can recruit CENP-R [90]. We thus speculate that at least also CENP-O/U/R of the CENP-P/O/R/Q/U proteins bind to the CENP-C/H/I/K/M/T/ W/N/L proteins in vivo in the nucleoplasm outside centromeres.

hMis12 complex
Finally, by DC-FCCS we studied the proteins hMis12, Nsl1, Dsn1 and Nnf1 of the hMis12 complex [77,85,144,145]. We measured the cross-correlation between these four proteins and found co-migration for all six protein pairs in the range between about 20 and 90% (corrected values) ( Table 1), indicating that a large fraction of these proteins in the nucleoplasm are part of complexes, either of hetero-dimers or -multimers (Fig 5). Our results are in agreement with published conclusions that the four proteins of the hMis12 complex form a quaternary complex [25,26,144,145]. Here we find that, in the nucleoplasm outside centromeres, a larger part of these proteins aggregate to hetero-dimers and hetero-multimers (Fig 5). In vitro, the hMis12 complex directly interacts with the CCAN, in particular with CENP-C, CENP-T, CENP-K and CENP-N [21,22,24,25,106]. Indeed, FRAP showed dynamic binding to centromeres with fast protein exchange in interphase cells [116] indicating weak binding. Here we showed that 30-44% of hMis12 of the Mis12 complex proteins co-migrated with CENP-K and CENP-T in vivo in the nucleoplasm outside centromeres (Table 1), indicating interaction of hMis12 (and probably the whole Mis12 complex) with the CENP-C/H/I/K/M/T/W/N/L complex. In mitosis, hMis12 stably binds to centromeres [116]. This stable interaction requires protein phosphorylation which in vivo takes place only in mitosis [29].

Conclusion
In this work, by FCCS we measured the co-migration of CCAN and hMis12 complex protein pairs in the nucleoplasm outside centromeres in living human interphase cells. Since cross-correlation values obtained from DC-FCCS experiments under-estimate the percentage of comigrating molecules, for some protein pairs we carried out SW-FCCS experiments which correct for these effects. Furthermore, the fluorescent tags might introduce sterical hindrance, preventing the detection of protein-protein interaction. Therefore, we labeled proteins at both termini and, in cases we detected no co-migration, introduced longer protein linkers between tag and CENP. For the hetero-dimerisation of CENP-T/W and CENP-S/X, we also determined apparent dissociation constants (K d ). We detected co-migration between CENP-K and CENP-T as well as between CENP-M and CENP-T but not between CENP-T/W and CENP-S/ X. Furthermore, CENP-C co-migrated with CENP-H, and CENP-K with CENP-N as well as with CENP-L. Thus, in the nucleoplasm outside centromeres, the CENP-H/I/K/M proteins interact with CENP-C, CENP-N/L and CENP-T/W but not with CENP-S/X. Probably these interactions take place in the nucleoplasm independent of modifications specifically occurring at centromeres. Nevertheless, we assume that at the centromere cell cycle and location specific protein modification of at least some of the proteins will enhance their interactions with other CCAN proteins. Furthermore, at the centromere CCAN protein-protein interactions will be further stabilized by multiple protein-protein binding [146]. Our FCCS analysis of the hMis12 complex showed that also these four proteins form a diffusing complex outside centromeres with at least hMis12 interacting with members of the CENP-C/H/I/K/M/T/W/N/L complex. Overall we show here that human CCAN and hMis12 kinetochore proteins assemble in the nucleoplasm outside centromeres, potentially forming dynamic multimeric complexes.
Supporting information S1 Table. Cross-correlation of protein pairs in the nucleoplasm by DC-FCCS. Original Data from which "means" and "SD" values were calculated and presented in Table 1. Table. Cross-correlation of protein pairs in the nucleoplasm by SW-FCCS. Original Data from which "means" and "SD" values were calculated and presented in Table 2