- Email: [email protected]
Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy
Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy Takafumi Fukuda a, Shigeko Kawai-Noma b, Chan-Gi Pack c, Hideki Taguchi a, d, * a
School of Life Science and Technology, Tokyo Institute of Technology, Yokohama, Japan Department of Applied Chemistry and Biotechnology, Chiba University, Chiba, Japan c Asan Institute for Life Sciences, Asan Medical Center, Seoul, 05505, South Korea d Cell Biology Center, Institute of Innovative Research, Tokyo Institute of Technology, Yokohama, Japan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 September 2019 Accepted 17 September 2019 Available online xxx
In the living cells, the majority of proteins does not work alone, but interact with other proteins or other biomolecules to maintain the cellular function, constituting a “protein community”. Previous efforts on mass spectroscopy-based protein interaction networks, interactomes, have provided a picture on the protein community. However, these were static information after cells were disrupted. For a better understanding of the protein community in cells, it is important to know the properties of intracellular dynamics and interactions. Since hydrodynamic size and mobility of proteins are related into such properties, direct measurement of diffusional motion of proteins in single living cells will be helpful for uncovering the properties. Here we completed measurement of the diffusion and homo-oligomeric properties of 369 cytoplasmic GFP-fusion proteins in living yeast Saccharomyces cerevisiae cells using fluorescence correlation spectroscopy (FCS). The large-scale analysis showed that the motions of majority of proteins obeyed a two-component (i.e. slow and fast components) diffusion model. Remarkably, both of the two components diffused more slowly than expected monomeric states. In addition, further analysis suggested that more proteins existed as homo-oligomeric states in living cells than previously expected. Our study, which characterizes the dynamics of proteins in living cells on a large-scale, provided a global view on intracellular protein dynamics to understand the protein community. © 2019 Elsevier Inc. All rights reserved.
Keywords: Budding yeast Fluorescence correlation spectroscopy (FCS) Diffusion coefficient GFP Large-scale analysis Protein dynamics
1. Introduction The inside of cells is crowded with biomolecules, such as proteins, nucleic acids and lipids. In such crowded environment, proteins would interact with other proteins or other biomolecules to perform their own functions, and to maintain cell homeostasis, constituting a protein network, termed here as “protein community”. Considerable efforts have been devoted to characterize the protein community. In a large-scale analysis, mass spectrometrybased methods have been applied to identity a component of the
Abbreviations: GFP, green fluorescent protein; FCS, fluorescence correlation spectroscopy; FRAP, fluorescence recovery after photobleaching; D, diffusion coefficient; CPM, counts per molecule; AOTF, acousto-optical tunable filter; FAF, the fluorescence autocorrelation functions; LSM, laser-scanning microscopy. * Corresponding author. School of Life Science and Technology, Tokyo Institute of Technology, Yokohama, Japan. E-mail address: [email protected] (H. Taguchi).
protein complex, combined with a variety of pull-down assays using either affinity tags or antibodies. Although the datasets on an interaction network and a cellular abundance of proteins are invaluable to describe the protein community, most of them were static states of proteins after cells were disrupted [1,2], where information on weak protein-protein interaction is in general lost. Dynamic aspects of proteins in living cells (such as diffusion and interaction) are crucial to understand the protein community, including an emerging concept, in which some kind of proteins forms droplet, caused by liquid-liquid phase separation, depending on the situation in the cell, in addition to the simple diffusion of proteins inside the cytosol or organelles [3,4]. So far, the dynamics of proteins in the living cells have not been well appreciated on a large-scale. It is important to characterize proteins as dynamic molecules, especially for their interactions, including self-assembly in the cellular milieu. The challenges to understand the molecular basis of protein function and dynamics in living cells have been performed in recent years using fluorescence microscopic techniques such as
https://doi.org/10.1016/j.bbrc.2019.09.066 0006-291X/© 2019 Elsevier Inc. All rights reserved.
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
fluorescence recovery after photobleaching (FRAP), single particle tracking, and fluorescence correlation spectroscopy (FCS) [5,6]. Among them, FCS, which uses confocal microscope setup, has an advantage to measure a broad range of diffusion of proteins labelled with fluorescence molecules in a low fluorescent intensity. FCS is a powerful technique with a single-molecule sensitivity to assess the diffusion of fluorescent proteins, most typically GFP-fused proteins, [7]. FCS detects fluorescence intensity fluctuations caused by Brownian motion of fluorescent probe molecules in a tiny detection volume (~0.2 fL) generated by confocal illumination. Through time correlation analysis of the fluorescence fluctuations, FCS analysis provides a diffusion coefficient (D) of fluorescently labelled proteins, which reflects apparent molecular size, and molecular brightness represented by counts per molecule (CPM), which reflects the tendency of self-assembly (i.e. oligomerization). Although FCS can provide the valuable information for protein dynamics, it is unsuitable for large-scale analysis. By recent efforts based on machine learning, high-throughput FCS system has been attempted; however it is a developing technique [8,9]. Therefore, there were no high-quality dataset on large-scale analysis of yeast proteins. The present study aimed to characterize dynamics of cytoplasmic proteins at endogenous expression levels in yeast living cells on a large-scale using FCS. We performed FCS measurements one by one on yeast GFP clones collection, which is a Saccharomyces cerevisiae collection expressing GFP tagged open reading frames [10]. The dynamics of proteins were quantified by D, which reflects apparent molecular size and CPM which reflects homo-oligomeric state. 2. Materials and methods 2.1. Yeast strains and media The strains used in this study were BY4741 (MATa his3 D1 leu2 D0 met17 D0 ura3 D0) and BY4741 (Yeast-GFP Clone Collection) (MATa his3 D1 leu2 D0 met17 D0 ura3 D0 HIS3MX6). The Yeast-GFP Clone Collection is a collection of Saccharomyces cerevisiae expressing full-length open reading frames containing the GFP (S65T) tag at the C-terminus of each protein [10]. The GFP genes are integrated into the yeast chromosome by homologous recombination and expressed as fusion proteins with the target gene using an endogenous promoter. Standard rich medium (YPD) was used to cultivate the GFP clone collection, and synthetic complete medium (using Difco yeast nitrogen base) lacking leucine (SC-Leu) was used to cultivate yeast cells containing plasmids. SRaf-Leu media contained 2% raffinose instead of glucose. To induce expression of 1 GFP and 2 GFP under the control of GAL1 promoters, galactose was added to final concentrations of 2% and 50 mmmM, respectively. Yeast strains were grown at 30 C. 2.2. Plasmid construction of 1 GFP and 2 GFP and yeast transformation Plasmid expressing monomeric GFP (mGFP) was synthesized with plasmid YCplac111-GAL1p, which is a single copy plasmid. GFP (S65T) was cloned from the Yeast GFP clone collection with SacI, SmaI and EcoRV site and inserted into YCplac111-GAL1p. To construct 2 GFP, GFP (S65T) was excised at Sac I and EcoRV and ligated between the SacI and SmaI restriction sites of another GFP (S65T). The linker between GFP (S65T) was Gly-Asp-Gly-Met-Ala All plasmid constructs for transfection were purified using a plasmid DNA midiprep kit (QIAGEN, Hilden, Germany). For 1 GFP and 2 GFP analysis, BY4741 was transformed with YCplac111GAL1p by lithium acetate method [11].
2.3. Yeast cell homogenization Yeast strains containing YCp-GAL1p were grown to a mid-log phase in SRaf-Leu. After 2% galactose was added, the yeast cultures were incubated for 4 h at 30 C. Cells were collected by centrifugation, broken with glass beads (Sigma) by vortexing for 1 min at 4 C in lysis buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl2, 10 mM KCl, 0.1 mM EDTA, pH 8.0, 1 mM DTT, Complete™ protease inhibitor cocktail EDTA-free (Roche)), and incubated for 2 min on ice. This procedure was repeated for 4e6 cycles. The crude lysates were prepared by centrifugation. 2.4. FCS measurements and quantitative analysis Yeast at log-phase was subjected to FCS measurements for living cells. FCS measurements were all performed at 25 C with ConfoCor 2 (Carl Zeiss) microscope as described [12,13]. GFP fluorescence was excited at 488 nm with a 6.3 mW in total power by adjusting the acousto-optical tunable filter (AOTF) to 0.3%. The fluorescence autocorrelation functions (FAF; G (t)), from which the average residence time (ti) and the absolute number of fluorescent proteins in the detection volume are calculated, are obtained as follows;
G ðtÞ ¼
½IðtÞIðt þ tÞ
(1)
½IðtÞ2
where I (t þ t) is the fluorescence intensity obtained by the single photon counting method in a detection volume at a delay time t (brackets denote ensemble averages). The curve fitting for the multicomponent model is given by:
GðtÞ ¼ 1 þ
1X t 1 t 1 2 1þ 2 yi 1 þ N i ti s ti =
2
(2)
where yi and ti are the fraction and the diffusion time of the component i, respectively, N is the number of fluorescent molecules in the detection volume defined by the beam waist w0 and the axial radius z0, s is the structure parameter representing the ratio of w0 and z0. The detection volume made by w0 and z0 was approximated to that of a cylinder. All FAFs in aqueous solutions were measured three times for 30 s at 5 s intervals. For the intracellular measurements, FAFs were measured 3e5 cells (six times for 10 s/cell), thereby calculating average values. The measurement position was chosen in the confocal image. Because the optical paths of laser-scanning microscopy (LSM) and FCS are not the same, the real position of the FCS measurement was tuned to the position on the LSM images with a cover glass coated with dried rhodamine 6G (Rh6G), which was the protocol provided by the manufacturer [14]. Immediately after each FCS measurement, the cell was again imaged by LSM and checked for a displacement. In cases where a cell appeared to have moved, the measurements were discarded. The detection pinhole for FCS was fixed to a diameter of 70 mm and the emission was recorded through a 505e550 nm band pass filter for measurements on living cells. All measured FAFs were fitted by a software installed on the ConfoCor 2 system using the model equation (eq. (2)). FAFs of monomeric GFP in aqueous solution and lysate were fitted by an one-component model (i ¼ 1). FAFs of 1 GFP and 2 GFP in living cells and of GFP-fusion proteins in living cells were fitted by a twocomponent model (i ¼ 2). The pinhole adjustment of the FCS setup, the structure parameter, and the detection volume were calibrated for 488 nm excitation each day with a Rh6G solution at a concentration of 107 M. Average values of the structure parameter, ranging from 4 to 8, were fixed for the FCS analysis throughout this
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
study. The diffusion time of component i, ti, is related to the translational diffusion constant D of component i by
ti ¼
w2 4Di
(3)
The diffusion of a spherical molecule is related to various physical parameters by the Stokes-Einstein equation as follows.
kB T Di ¼ 6phri
(means ± S.E.), therefore 3.0 times lower than D in the lysate and 4.3 times lower than D in the solution, respectively. This result indicated that the apparent viscosity in the living cells was 4.3 times higher than in solution. Above diffusion coefficients of mGFP in cells, the lysate and the solution were almost consistent with previous analyses using FCS [12,13]. 3.2. Measurements of diffusion coefficients of 369 cytoplasmic proteins in living yeast cells
(4)
where T is the absolute temperature, ri is the hydrodynamic radius of the spherical molecule, h is the fluid-phase viscosity of the solvent, and kB is the Boltzmann constant. Because ti is proportional to viscosity, the relative viscosity (t cell/t lysate) can be easily estimated. The diffusion coefficients of GFP (DGFP) of mGFP and GFP-fusion protein in the solution and cells were calculated from the published diffusion constant of Rh6G, DRh6G (280 mm2/s), and the measured diffusion times of Rh6G (tRh6G) and GFP (tgfp) as follows:
DGFP t ¼ Rh6G DRh6G tGFP
3
(5)
Measured t of each sample was converted to D by eq. (5). The D of the fast-free and slow-restricted diffusion components in the 2component model were represented by Dfast and Dslow, respectively. The corresponding fractional ratio (yfast and yslow) of the diffusional components in the 2-component model was presented by a percentage (%). 2.5. Statistical analysis Values are expressed as means ± S.E. or S.D. Comparison of result was examined with unpaired t-test (between groups). All statistical analyses were performed with BellCurve for Excel (Social Survey Research Information Co., Ltd., Tokyo, Japan). 3. Results 3.1. Evaluation of intracellular environment using monomeric GFP FCS measures fluorescence fluctuation within confocal volume of 0.2 fL to determine FAFs for probe molecules (Fig. 1) [15]. Fitting of mathematical function models of FAF allows estimation of individual diffusion coefficients (Di), which reflect the hydrodynamic size (i.e. molecular weight, MW z r3 z D000 3) of fluorescent molecules. FCS analysis also presents the CPM, which is calculated by dividing averaged fluorescence intensity by the number of molecules in the detection volume. Therefore, CPM reflects the homooligomeric state of the fluorescent proteins expressed in a single GFP strain of yeast. We first quantified dynamics of monomeric mGFP as a control for the diffusion coefficients in living cells, a lysate and a solution using FCS to assess possible effect of intracellular environments. Yeast cells, in which mGFP was highly expressed under the control of GAL1 promoter, were used, and the lysate was subjected to the FCS analysis after homogenization. Purified mGFP was used for the solution analysis. As shown in Fig. 2 and Table 1, the FAF curves of mGFP in the lysate and the solution were well fitted with a one-component diffusion model. The results indicated that mGFP diffuses homogeneously in these environments. D in the lysate and the solution were 52.7 ± 3.0 and 76.2 ± 2.8 mm2/s (means ± S.E.), respectively. On the other hand, the FAF curves of GFP in cells were well fitted with a two-component diffusion model. This result indicated that GFP diffused heterogeneously in living cells. The Dfast in living cells was 17.7 ± 0.6 mm2/s
We measured the diffusion properties of 369 cytoplasmic proteins fused with GFP in budding yeast S. cerevisiae, which were randomly selected from the Yeast-GFP Clone Collection, using FCS (Fig. 3A) (Yeast GFP Fusion Localization Database, https://yeastgfp. yeastgenome.org/). Our measurements covered about 20% of the cytoplasm-localized proteins including ambiguous-localized proteins listed in a database [10]. The distribution of molecular weights of the 369 proteins almost reflected that of all cytoplasmic proteins in yeast (Supplementary Fig. S1). Also, the classification of functional categories on the 369 proteins was widespread (Supplementary Fig. S1). The FAF curves of all proteins were well fitted with the two-component diffusion model. The diffusion coefficients and the correspondent fractions of examined proteins were provided in Fig. 3B (refer also Supplementary Table S1 for raw data), which shows a broad spectrum of diffusion coefficients in sub-proteome members. The values of Dfast and Dslow for 369 proteins range from 2.7 to 14.6 mm2/s (means ± S.D.; 6.4 ± 2.1 mm2/s) and from 0.04 to 0.97 mm2/s (means ± S.D.; 0.30 ± 0.16 mm2/s), respectively. The fraction of fast components was 31.4e99.5% (means ± S.D.; 75.1 ± 13.5%). As a representative example, Hsp104, which has been reported to interact with aggregated proteins, had relatively low fast fraction (means ± S.E.; 52 ± 5%) and Dfast (means ± S.E.; 3.8 ± 0.4 mm2/s) (Supplementary Fig. S2). The distributions of Dfast and Dslow were shown in Fig. 3C. Less than 60% fraction of fast components was observed in 50 proteins (Supplementary Table 1). On the other hand, 40 proteins had fraction of fast component above 90% (Supplementary Table S1). To compare measured D values with theoretical D values calculated from theoretical MW, Dfast and Dslow were plotted against MW (Fig. 3D). Combined with the plot for the fast component fraction against MW (Fig. 3D inset), the plots show that even proteins of similar MW had various diffusion coefficients and the fractions of the components. Importantly, almost of them were diffused slower than calculated theoretical diffusion coefficients (Dtheoretical) in the living cells even fast components (Fig. 3D). The MW estimated from the measured Dfast values for each protein (MWfcs) was divided by their theoretical MW (MWtheoretical), which was calculated with a relation of MW z r3 z D000 3 for a spherical molecule and with the molecular weight of GFP (26.9 kDa) as a reference (Fig. 3E) [16]. Even considering the apparent viscosity in the living cells was 4.3 times higher than in the solution, the distribution of the MWfcs/ MWtheoretical values indicates that more than a hundred of proteins were present with sizes of 10 times larger than their MWtheoretical in the living cells. 3.3. Evaluation of homo oligomeric state of 369 cytoplasmic proteins in yeast living cells FCS also provides the information about the homo-oligomeric states of proteins using CPM (Fig. 1). We first examined whether CPM reflected the self-assembly of the proteins in our experiment system using monomeric GFP (1 GFP) and artificial dimeric GFP (2 GFP), in which two GFP were genetically fused with a linker residue. As shown in Fig. 4A, the CPM value of monomeric GFP was 3.9 ± 0.5 kHz (means ± S.E.) and that of dimer GFP was 5.9 ± 0.3 kHz
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
4
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 1. Overview of fluorescence correlation spectroscopy (FCS) and the parameters.
with high CPM, it can be seen that some proteins have CPM equivalent to the monomer. On the other hand, some proteins, for which there is no information about the homo-oligomeric state in the database, had higher CPM values compared with those for monomeric GFP (Fig. 4B and Supplementary Table S1), suggesting that these proteins might form homo-oligomers in the living cells.
4. Discussion
Fig. 2. Quantitation by fluorescence correlation spectroscopy of monomer GFP under different conditions. The representatives of normalized fluorescence autocorrelation functions of monomeric GFP under each condition. Closed circle, open circle, and cross indicate the data in cells, the lysate, and the solution, respectively.
Table 1 Diffusion coefficients (D) of monomeric GFP under different conditions. Protein
Condition
Dfast (mm2s1) (%)
Dslow (mm2s1) (%)
monomeric GFP
in living cells in lysate in solution
17.7 ± 0.6 (92 ± 1) 52.7 ± 3.0 (100) 76.2 ± 2.8 (100)
0.3 ± 0.1 (8 ± 1) N.D. N.D.
Data are presented as the means ± S.E. (in living cells: N ¼ 24, in lysate: N ¼ 4, in solution: N ¼ 20). N.D. means “not detected”.
(means ± S.E.). There was significant difference between them (p ¼ 0.006). Fig. 4B shows the CPM distribution of 369 cytosolic proteins we examined in this study (Fig. 4B). Based on a database information of protein self-assembly state (Universal Protein Knowledgebase, https://www.uniprot.org/), the data set was classified into monomer, homo-dimer and homo-trimer or higher (homo-oligomer) proteins in the inset of Fig. 4B. For example, Hsp104 which forms homo-hexameric state had relatively high CPM (16.4 kHz) (Fig. 4B inset and Supplementary Table S1). According to the information of the database, while there are proteins
The current study is a novel large-scale analysis of protein dynamics using Yeast-GFP Clone Collection. We quantified the Dfast, Dslow, ratio of these components and CPM of 369 cytoplasmic yeast proteins at endogenous expression level using FCS. By measuring cells one by one, high quality data could be provided. The data obtained from this study suggested that, overall, proteins examined do not exist homogeneously, but heterogeneously in living cells. The most important finding of this study was to provide a dataset on dynamics including diffusion and homo-oligomeric state of proteins in yeast living cells under a normal growth condition. In addition, the analysis of the dataset revealed that most of the proteins are described with a two-component diffusion model, the fast and slow diffusions. First of all, we measured the diffusion dynamics of monomer GFP in yeast living cells. The Dfast in living cells was slower than that in lysate or solution. This result reflected the molecular crowding environment in living cells [17,18]. More than 90% of monomeric GFP diffused fast as a monomeric state. This is expected because GFP is a heterologous protein derived from a jellyfish. However, ~8% of GFP diffused slowly, as was previously reported [19], suggesting that a homologous assembly or an interaction with other proteins/ components in yeast cells. Further studies are needed to determine whether this slow diffusion of GFP is an artifact of fitting or reflects the heterogeneity of GFP in living cells. Subsequently, the diffusional dynamics of 369 cytosolic proteins were quantified using FCS. We calculated Dtheoretical using Dfast of GFP in living cells. As shown in Fig. 3D and E, almost 369 proteins had slower diffusion coefficients than Dtheoretical even in the fast components, indicating that the majority of proteins interacted with other molecules in living cells. In addition, there were no
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
5
Fig. 3. Summary of diffusion coefficients (D) and the ratio of slow components and fast components. (A) Yeast GFP clone collection was used to measure the cellular diffusion constants (D) of hundreds of proteins one by one. (B) The fractions of fast and slow components were plotted against D. The closed and open circles indicate slow and fast components of each protein, respectively. FCS measurements were conducted for 3e5 cells (6 times for 10 s/cell) for each protein, and the average values were provided. (C) The distributions of Dfast and Dslow. (D) Dfast and Dslow were plotted against molecular weights (MW). The solid line shows the Dtheoretical values, which were calculated from DGFP. The fractions of fast components were plotted against MW in the inset. (E) The distribution of MWfcs/MWtheoretical. MWtheoretical was calculated by measured Dfast using eq. (5).
Fig. 4. Counts per molecule (CPM) analysis to estimate oligomeric states of proteins. (A) The CPM values of 1 and 2 GFP (the means ± S.E.) (1 GFP: N ¼ 5, 2 GFP: N ¼ 7, respectively). Experimental data were analyzed with unpaired t-test. (B) The distribution of the CPM values of each protein. Based on database information of protein self-assembly state, the data set was classified into monomer, homo-dimer and homo-trimer or higher (homo-oligomer) proteins in the inset. CPM of Hsp104, known as a homo-hexamer, is shown as a representative.
relation between the cytosolic diffusion coefficients and theoretical molecular weights (Fig. 3D), although molecular weight of molecules is one of the factors determining diffusion coefficients of them in a solution. These results suggest that the diffusion coefficients in cells reflected dynamic intracellular states of each protein, rather than their molecular weights of monomeric states. Interestingly, the evaluation of CPM revealed the homooligomeric state of each protein. For example, Hsp 42, Glt1 and Fas1, which were previously reported that they formed homooligomers, had high CPM of 12.8, 16.8 and 16.4 kHz, respectively (Supplementary Table 1) [20e22]. On the other hand, Mrp8, whose function was unknown according to a database, had relatively high
CPM (8.6 kHz) and low fraction of the fast component (63%). These observations might indicate that Mrp8 existed as a dimer and strongly interacted with other molecules, thereby functioning in the living cells. Therefore, evaluation of the dynamic status of proteins directly in single living cells provided novel insight into protein society. Recent progress in cell biology has revealed that proteins in the cell could phase-separate depending on a variety of cellular stress [3,4]. In order to understand molecular basis of the phase separation, it is important to evaluate not only information on static quantity but also on cellular dynamics of related biomolecules. Actually, previous study using proteome-wide analysis reported
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
6
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
that various stresses affect protein localization and function by proteome-wide analysis [23]. It is our future study to clarify on a large-scale how the dynamics of proteins change under different conditions. Funding This work was supported by MEXT Grant-in-Aid for Scientific Research (Grant No. 19058002, 24113705, 24657070, 26116002 to HT) and Daiichi Sankyo Foundation of Life Science. Declaration of competing interest The authors declare no conflicts of interest associated with this manuscript. Acknowledgement We thank Motomasa Tanaka for providing a portion of Yeast GFP collection, Tatsuya Niwa for technical advice and valuable discussion, Yasushi Sako for the FCS measurement. Transparency document Transparency document related to this article can be found online at https://doi.org/10.1016/j.bbrc.2019.09.066. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrc.2019.09.066. FCS is a method based on fluctuation analysis of fluorescence intensity detected from the tiny confocal area. The amplitude of fluctuation reflects the number of molecules (N) in the detection volume, and the average intensity of fluorescence reflects the abundance of the analyzed protein fused with GFP. Counts per molecule (CPM), which reflects homo-oligomeric state is calculated by dividing the average intensity by N. FCS provides a fluorescence auto-correlation function, allowing the measurement of important biophysical parameters: the translational diffusion time of the molecules through the open volume. Mobility parameter, i.e., diffusion coefficient (D) of molecules reflects the molecular size (MW) of fluorescent species and the apparent viscosity in solution or in a live cell (see also Methods). References [1] S. Ghaemmaghami, W.K. Huh, K. Bower, et al., Global analysis of protein expression in yeast, Nature 425 (2003) 737e741, https://doi.org/10.1038/ nature02046. [2] Y. Ho, A. Gruhler, A. Heilbut, et al., Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry, Nature 415 (2002)
180e183, https://doi.org/10.1038/415180a. [3] S.F. Banani, H.O. Lee, A.A. Hyman, et al., Biomolecular condensates: organizers of cellular biochemistry, Nat. Rev. Mol. Cell Biol. 18 (2017) 285e298, https:// doi.org/10.1038/nrm.2017.7. [4] M. Kato, S.L. McKnight, A solid-state conceptualization of information transfer from gene to message to protein, Annu. Rev. Biochem. 87 (2018) 351e390, https://doi.org/10.1146/annurev-biochem-061516-044700. [5] S. Diekmann, C. Hoischen, Biomolecular dynamics and binding studies in the living cell, Phys. Life Rev. 11 (2014) 1e30, https://doi.org/10.1016/ j.plrev.2013.11.011. [6] S. Veerapathiran, T. Wohland, Fluorescence techniques in developmental biology, J. Biosci. 43 (2018) 541e553. [7] Y. Nomura, T. Nakamura, Z. Feng, et al., Direct quantification of gene expression using fluorescence correlation spectroscopy, Curr. Pharmaceut. Biotechnol. 8 (2007) 286e290. [8] M. Wachsmuth, C. Conrad, J. Bulkescher, et al., High-throughput fluorescence correlation spectroscopy enables analysis of proteome dynamics in living cells, Nat. Biotechnol. 33 (2015) 384e389, https://doi.org/10.1038/nbt.3146. [9] W. Christopher, H. Joseph, M. Will, et al., Fluorescence correlation spectroscopy as tool for high-content-screening in yeast (HCS-FCS), Proc. SPIE (2011) 7905. [10] W.K. Huh, J.V. Falvo, L.C. Gerke, et al., Global analysis of protein localization in budding yeast, Nature 425 (2003) 686e691, https://doi.org/10.1038/ nature02026. [11] D.C. Chen, B.C. Yang, T.T. Kuo, One-step transformation of yeast in stationary phase, Curr. Genet. 21 (1992) 83e84. [12] S. Kawai-Noma, S. Ayano, C.G. Pack, et al., Dynamics of yeast prion aggregates in single living cells, Genes Cells : devoted to molecular & cellular mechanisms 11 (2006) 1085e1096, https://doi.org/10.1111/j.13652443.2006.01004.x. [13] S. Kawai-Noma, C.G. Pack, T. Tsuji, et al., Single mother-daughter pair analysis to clarify the diffusion properties of yeast prion Sup 35 in guanidine-HCltreated [PSI] cells, Genes to cells, devoted to molecular & cellular mechanisms 14 (2009) 1045e1054, https://doi.org/10.1111/j.13652443.2009.01333.x. [14] K. Weisshart, V. Jungel, S.J. Briddon, The LSM 510 META - ConfoCor 2 system: an integrated imaging and spectroscopic platform for single-molecule detection, Curr. Pharmaceut. Biotechnol. 5 (2004) 135e154. [15] Y. Sako, M. Hiroshima, C.G. Pack, et al., Live cell single-molecule detection in systems biology, Wiley interdisciplinary reviews, Systems biology and medicine 4 (2012) 183e192, https://doi.org/10.1002/wsbm.161. [16] C.G. Pack, G. Nishimura, M. Tamura, et al., Analysis of interaction between chaperonin GroEL and its substrate using fluorescence correlation spectroscopy, Cytometry 36 (1999) 247e253. [17] T. Ando, J. Skolnick, Crowding and hydrodynamic interactions likely dominate in vivo macromolecular motion, Proc. Natl. Acad. Sci. U.S.A. 107 (2010) 18457e18462, https://doi.org/10.1073/pnas.1011354107. [18] J.W. Wojcieszyn, R.A. Schlegel, E.S. Wu, et al., Diffusion of injected macromolecules within the cytoplasm of living cells, Proc. Natl. Acad. Sci. U.S.A. 78 (1981) 4407e4410. [19] C.G. Pack, Y. Inoue, T. Higurashi, et al., Heterogeneous interaction network of yeast prions and remodeling factors detected in live cells, BMB reports 50 (2017) 478e483. [20] S.J. Kolodziej, P.A. Penczek, J.P. Schroeter, et al., Structure-function relationships of the Saccharomyces cerevisiae fatty acid synthase. Three-dimensional structure, J. Biol. Chem. 271 (1996) 28422e28429. [21] L. Valenzuela, P. Ballario, C. Aranda, et al., Regulation of expression of GLT1, the gene encoding glutamate synthase in Saccharomyces cerevisiae, J. Bacteriol. 180 (1998) 3533e3540. [22] D. Wotton, K. Freeman, D. Shore, Multimerization of Hsp42p, a novel heat shock protein of Saccharomyces cerevisiae, is dependent on a conserved carboxyl-terminal sequence, J. Biol. Chem. 271 (1996) 2717e2723. [23] K. Lee, M.K. Sung, J. Kim, et al., Proteome-wide remodeling of protein location and function by stress, Proc. Natl. Acad. Sci. U.S.A. 111 (2014) E3157eE3166, https://doi.org/10.1073/pnas.1318881111.
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy Takafumi Fukuda a, Shigeko Kawai-Noma b, Chan-Gi Pack c, Hideki Taguchi a, d, * a
School of Life Science and Technology, Tokyo Institute of Technology, Yokohama, Japan Department of Applied Chemistry and Biotechnology, Chiba University, Chiba, Japan c Asan Institute for Life Sciences, Asan Medical Center, Seoul, 05505, South Korea d Cell Biology Center, Institute of Innovative Research, Tokyo Institute of Technology, Yokohama, Japan b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 September 2019 Accepted 17 September 2019 Available online xxx
In the living cells, the majority of proteins does not work alone, but interact with other proteins or other biomolecules to maintain the cellular function, constituting a “protein community”. Previous efforts on mass spectroscopy-based protein interaction networks, interactomes, have provided a picture on the protein community. However, these were static information after cells were disrupted. For a better understanding of the protein community in cells, it is important to know the properties of intracellular dynamics and interactions. Since hydrodynamic size and mobility of proteins are related into such properties, direct measurement of diffusional motion of proteins in single living cells will be helpful for uncovering the properties. Here we completed measurement of the diffusion and homo-oligomeric properties of 369 cytoplasmic GFP-fusion proteins in living yeast Saccharomyces cerevisiae cells using fluorescence correlation spectroscopy (FCS). The large-scale analysis showed that the motions of majority of proteins obeyed a two-component (i.e. slow and fast components) diffusion model. Remarkably, both of the two components diffused more slowly than expected monomeric states. In addition, further analysis suggested that more proteins existed as homo-oligomeric states in living cells than previously expected. Our study, which characterizes the dynamics of proteins in living cells on a large-scale, provided a global view on intracellular protein dynamics to understand the protein community. © 2019 Elsevier Inc. All rights reserved.
Keywords: Budding yeast Fluorescence correlation spectroscopy (FCS) Diffusion coefficient GFP Large-scale analysis Protein dynamics
1. Introduction The inside of cells is crowded with biomolecules, such as proteins, nucleic acids and lipids. In such crowded environment, proteins would interact with other proteins or other biomolecules to perform their own functions, and to maintain cell homeostasis, constituting a protein network, termed here as “protein community”. Considerable efforts have been devoted to characterize the protein community. In a large-scale analysis, mass spectrometrybased methods have been applied to identity a component of the
Abbreviations: GFP, green fluorescent protein; FCS, fluorescence correlation spectroscopy; FRAP, fluorescence recovery after photobleaching; D, diffusion coefficient; CPM, counts per molecule; AOTF, acousto-optical tunable filter; FAF, the fluorescence autocorrelation functions; LSM, laser-scanning microscopy. * Corresponding author. School of Life Science and Technology, Tokyo Institute of Technology, Yokohama, Japan. E-mail address: [email protected] (H. Taguchi).
protein complex, combined with a variety of pull-down assays using either affinity tags or antibodies. Although the datasets on an interaction network and a cellular abundance of proteins are invaluable to describe the protein community, most of them were static states of proteins after cells were disrupted [1,2], where information on weak protein-protein interaction is in general lost. Dynamic aspects of proteins in living cells (such as diffusion and interaction) are crucial to understand the protein community, including an emerging concept, in which some kind of proteins forms droplet, caused by liquid-liquid phase separation, depending on the situation in the cell, in addition to the simple diffusion of proteins inside the cytosol or organelles [3,4]. So far, the dynamics of proteins in the living cells have not been well appreciated on a large-scale. It is important to characterize proteins as dynamic molecules, especially for their interactions, including self-assembly in the cellular milieu. The challenges to understand the molecular basis of protein function and dynamics in living cells have been performed in recent years using fluorescence microscopic techniques such as
https://doi.org/10.1016/j.bbrc.2019.09.066 0006-291X/© 2019 Elsevier Inc. All rights reserved.
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
fluorescence recovery after photobleaching (FRAP), single particle tracking, and fluorescence correlation spectroscopy (FCS) [5,6]. Among them, FCS, which uses confocal microscope setup, has an advantage to measure a broad range of diffusion of proteins labelled with fluorescence molecules in a low fluorescent intensity. FCS is a powerful technique with a single-molecule sensitivity to assess the diffusion of fluorescent proteins, most typically GFP-fused proteins, [7]. FCS detects fluorescence intensity fluctuations caused by Brownian motion of fluorescent probe molecules in a tiny detection volume (~0.2 fL) generated by confocal illumination. Through time correlation analysis of the fluorescence fluctuations, FCS analysis provides a diffusion coefficient (D) of fluorescently labelled proteins, which reflects apparent molecular size, and molecular brightness represented by counts per molecule (CPM), which reflects the tendency of self-assembly (i.e. oligomerization). Although FCS can provide the valuable information for protein dynamics, it is unsuitable for large-scale analysis. By recent efforts based on machine learning, high-throughput FCS system has been attempted; however it is a developing technique [8,9]. Therefore, there were no high-quality dataset on large-scale analysis of yeast proteins. The present study aimed to characterize dynamics of cytoplasmic proteins at endogenous expression levels in yeast living cells on a large-scale using FCS. We performed FCS measurements one by one on yeast GFP clones collection, which is a Saccharomyces cerevisiae collection expressing GFP tagged open reading frames [10]. The dynamics of proteins were quantified by D, which reflects apparent molecular size and CPM which reflects homo-oligomeric state. 2. Materials and methods 2.1. Yeast strains and media The strains used in this study were BY4741 (MATa his3 D1 leu2 D0 met17 D0 ura3 D0) and BY4741 (Yeast-GFP Clone Collection) (MATa his3 D1 leu2 D0 met17 D0 ura3 D0 HIS3MX6). The Yeast-GFP Clone Collection is a collection of Saccharomyces cerevisiae expressing full-length open reading frames containing the GFP (S65T) tag at the C-terminus of each protein [10]. The GFP genes are integrated into the yeast chromosome by homologous recombination and expressed as fusion proteins with the target gene using an endogenous promoter. Standard rich medium (YPD) was used to cultivate the GFP clone collection, and synthetic complete medium (using Difco yeast nitrogen base) lacking leucine (SC-Leu) was used to cultivate yeast cells containing plasmids. SRaf-Leu media contained 2% raffinose instead of glucose. To induce expression of 1 GFP and 2 GFP under the control of GAL1 promoters, galactose was added to final concentrations of 2% and 50 mmmM, respectively. Yeast strains were grown at 30 C. 2.2. Plasmid construction of 1 GFP and 2 GFP and yeast transformation Plasmid expressing monomeric GFP (mGFP) was synthesized with plasmid YCplac111-GAL1p, which is a single copy plasmid. GFP (S65T) was cloned from the Yeast GFP clone collection with SacI, SmaI and EcoRV site and inserted into YCplac111-GAL1p. To construct 2 GFP, GFP (S65T) was excised at Sac I and EcoRV and ligated between the SacI and SmaI restriction sites of another GFP (S65T). The linker between GFP (S65T) was Gly-Asp-Gly-Met-Ala All plasmid constructs for transfection were purified using a plasmid DNA midiprep kit (QIAGEN, Hilden, Germany). For 1 GFP and 2 GFP analysis, BY4741 was transformed with YCplac111GAL1p by lithium acetate method [11].
2.3. Yeast cell homogenization Yeast strains containing YCp-GAL1p were grown to a mid-log phase in SRaf-Leu. After 2% galactose was added, the yeast cultures were incubated for 4 h at 30 C. Cells were collected by centrifugation, broken with glass beads (Sigma) by vortexing for 1 min at 4 C in lysis buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl2, 10 mM KCl, 0.1 mM EDTA, pH 8.0, 1 mM DTT, Complete™ protease inhibitor cocktail EDTA-free (Roche)), and incubated for 2 min on ice. This procedure was repeated for 4e6 cycles. The crude lysates were prepared by centrifugation. 2.4. FCS measurements and quantitative analysis Yeast at log-phase was subjected to FCS measurements for living cells. FCS measurements were all performed at 25 C with ConfoCor 2 (Carl Zeiss) microscope as described [12,13]. GFP fluorescence was excited at 488 nm with a 6.3 mW in total power by adjusting the acousto-optical tunable filter (AOTF) to 0.3%. The fluorescence autocorrelation functions (FAF; G (t)), from which the average residence time (ti) and the absolute number of fluorescent proteins in the detection volume are calculated, are obtained as follows;
G ðtÞ ¼
½IðtÞIðt þ tÞ
(1)
½IðtÞ2
where I (t þ t) is the fluorescence intensity obtained by the single photon counting method in a detection volume at a delay time t (brackets denote ensemble averages). The curve fitting for the multicomponent model is given by:
GðtÞ ¼ 1 þ
1X t 1 t 1 2 1þ 2 yi 1 þ N i ti s ti =
2
(2)
where yi and ti are the fraction and the diffusion time of the component i, respectively, N is the number of fluorescent molecules in the detection volume defined by the beam waist w0 and the axial radius z0, s is the structure parameter representing the ratio of w0 and z0. The detection volume made by w0 and z0 was approximated to that of a cylinder. All FAFs in aqueous solutions were measured three times for 30 s at 5 s intervals. For the intracellular measurements, FAFs were measured 3e5 cells (six times for 10 s/cell), thereby calculating average values. The measurement position was chosen in the confocal image. Because the optical paths of laser-scanning microscopy (LSM) and FCS are not the same, the real position of the FCS measurement was tuned to the position on the LSM images with a cover glass coated with dried rhodamine 6G (Rh6G), which was the protocol provided by the manufacturer [14]. Immediately after each FCS measurement, the cell was again imaged by LSM and checked for a displacement. In cases where a cell appeared to have moved, the measurements were discarded. The detection pinhole for FCS was fixed to a diameter of 70 mm and the emission was recorded through a 505e550 nm band pass filter for measurements on living cells. All measured FAFs were fitted by a software installed on the ConfoCor 2 system using the model equation (eq. (2)). FAFs of monomeric GFP in aqueous solution and lysate were fitted by an one-component model (i ¼ 1). FAFs of 1 GFP and 2 GFP in living cells and of GFP-fusion proteins in living cells were fitted by a twocomponent model (i ¼ 2). The pinhole adjustment of the FCS setup, the structure parameter, and the detection volume were calibrated for 488 nm excitation each day with a Rh6G solution at a concentration of 107 M. Average values of the structure parameter, ranging from 4 to 8, were fixed for the FCS analysis throughout this
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
study. The diffusion time of component i, ti, is related to the translational diffusion constant D of component i by
ti ¼
w2 4Di
(3)
The diffusion of a spherical molecule is related to various physical parameters by the Stokes-Einstein equation as follows.
kB T Di ¼ 6phri
(means ± S.E.), therefore 3.0 times lower than D in the lysate and 4.3 times lower than D in the solution, respectively. This result indicated that the apparent viscosity in the living cells was 4.3 times higher than in solution. Above diffusion coefficients of mGFP in cells, the lysate and the solution were almost consistent with previous analyses using FCS [12,13]. 3.2. Measurements of diffusion coefficients of 369 cytoplasmic proteins in living yeast cells
(4)
where T is the absolute temperature, ri is the hydrodynamic radius of the spherical molecule, h is the fluid-phase viscosity of the solvent, and kB is the Boltzmann constant. Because ti is proportional to viscosity, the relative viscosity (t cell/t lysate) can be easily estimated. The diffusion coefficients of GFP (DGFP) of mGFP and GFP-fusion protein in the solution and cells were calculated from the published diffusion constant of Rh6G, DRh6G (280 mm2/s), and the measured diffusion times of Rh6G (tRh6G) and GFP (tgfp) as follows:
DGFP t ¼ Rh6G DRh6G tGFP
3
(5)
Measured t of each sample was converted to D by eq. (5). The D of the fast-free and slow-restricted diffusion components in the 2component model were represented by Dfast and Dslow, respectively. The corresponding fractional ratio (yfast and yslow) of the diffusional components in the 2-component model was presented by a percentage (%). 2.5. Statistical analysis Values are expressed as means ± S.E. or S.D. Comparison of result was examined with unpaired t-test (between groups). All statistical analyses were performed with BellCurve for Excel (Social Survey Research Information Co., Ltd., Tokyo, Japan). 3. Results 3.1. Evaluation of intracellular environment using monomeric GFP FCS measures fluorescence fluctuation within confocal volume of 0.2 fL to determine FAFs for probe molecules (Fig. 1) [15]. Fitting of mathematical function models of FAF allows estimation of individual diffusion coefficients (Di), which reflect the hydrodynamic size (i.e. molecular weight, MW z r3 z D000 3) of fluorescent molecules. FCS analysis also presents the CPM, which is calculated by dividing averaged fluorescence intensity by the number of molecules in the detection volume. Therefore, CPM reflects the homooligomeric state of the fluorescent proteins expressed in a single GFP strain of yeast. We first quantified dynamics of monomeric mGFP as a control for the diffusion coefficients in living cells, a lysate and a solution using FCS to assess possible effect of intracellular environments. Yeast cells, in which mGFP was highly expressed under the control of GAL1 promoter, were used, and the lysate was subjected to the FCS analysis after homogenization. Purified mGFP was used for the solution analysis. As shown in Fig. 2 and Table 1, the FAF curves of mGFP in the lysate and the solution were well fitted with a one-component diffusion model. The results indicated that mGFP diffuses homogeneously in these environments. D in the lysate and the solution were 52.7 ± 3.0 and 76.2 ± 2.8 mm2/s (means ± S.E.), respectively. On the other hand, the FAF curves of GFP in cells were well fitted with a two-component diffusion model. This result indicated that GFP diffused heterogeneously in living cells. The Dfast in living cells was 17.7 ± 0.6 mm2/s
We measured the diffusion properties of 369 cytoplasmic proteins fused with GFP in budding yeast S. cerevisiae, which were randomly selected from the Yeast-GFP Clone Collection, using FCS (Fig. 3A) (Yeast GFP Fusion Localization Database, https://yeastgfp. yeastgenome.org/). Our measurements covered about 20% of the cytoplasm-localized proteins including ambiguous-localized proteins listed in a database [10]. The distribution of molecular weights of the 369 proteins almost reflected that of all cytoplasmic proteins in yeast (Supplementary Fig. S1). Also, the classification of functional categories on the 369 proteins was widespread (Supplementary Fig. S1). The FAF curves of all proteins were well fitted with the two-component diffusion model. The diffusion coefficients and the correspondent fractions of examined proteins were provided in Fig. 3B (refer also Supplementary Table S1 for raw data), which shows a broad spectrum of diffusion coefficients in sub-proteome members. The values of Dfast and Dslow for 369 proteins range from 2.7 to 14.6 mm2/s (means ± S.D.; 6.4 ± 2.1 mm2/s) and from 0.04 to 0.97 mm2/s (means ± S.D.; 0.30 ± 0.16 mm2/s), respectively. The fraction of fast components was 31.4e99.5% (means ± S.D.; 75.1 ± 13.5%). As a representative example, Hsp104, which has been reported to interact with aggregated proteins, had relatively low fast fraction (means ± S.E.; 52 ± 5%) and Dfast (means ± S.E.; 3.8 ± 0.4 mm2/s) (Supplementary Fig. S2). The distributions of Dfast and Dslow were shown in Fig. 3C. Less than 60% fraction of fast components was observed in 50 proteins (Supplementary Table 1). On the other hand, 40 proteins had fraction of fast component above 90% (Supplementary Table S1). To compare measured D values with theoretical D values calculated from theoretical MW, Dfast and Dslow were plotted against MW (Fig. 3D). Combined with the plot for the fast component fraction against MW (Fig. 3D inset), the plots show that even proteins of similar MW had various diffusion coefficients and the fractions of the components. Importantly, almost of them were diffused slower than calculated theoretical diffusion coefficients (Dtheoretical) in the living cells even fast components (Fig. 3D). The MW estimated from the measured Dfast values for each protein (MWfcs) was divided by their theoretical MW (MWtheoretical), which was calculated with a relation of MW z r3 z D000 3 for a spherical molecule and with the molecular weight of GFP (26.9 kDa) as a reference (Fig. 3E) [16]. Even considering the apparent viscosity in the living cells was 4.3 times higher than in the solution, the distribution of the MWfcs/ MWtheoretical values indicates that more than a hundred of proteins were present with sizes of 10 times larger than their MWtheoretical in the living cells. 3.3. Evaluation of homo oligomeric state of 369 cytoplasmic proteins in yeast living cells FCS also provides the information about the homo-oligomeric states of proteins using CPM (Fig. 1). We first examined whether CPM reflected the self-assembly of the proteins in our experiment system using monomeric GFP (1 GFP) and artificial dimeric GFP (2 GFP), in which two GFP were genetically fused with a linker residue. As shown in Fig. 4A, the CPM value of monomeric GFP was 3.9 ± 0.5 kHz (means ± S.E.) and that of dimer GFP was 5.9 ± 0.3 kHz
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
4
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
Fig. 1. Overview of fluorescence correlation spectroscopy (FCS) and the parameters.
with high CPM, it can be seen that some proteins have CPM equivalent to the monomer. On the other hand, some proteins, for which there is no information about the homo-oligomeric state in the database, had higher CPM values compared with those for monomeric GFP (Fig. 4B and Supplementary Table S1), suggesting that these proteins might form homo-oligomers in the living cells.
4. Discussion
Fig. 2. Quantitation by fluorescence correlation spectroscopy of monomer GFP under different conditions. The representatives of normalized fluorescence autocorrelation functions of monomeric GFP under each condition. Closed circle, open circle, and cross indicate the data in cells, the lysate, and the solution, respectively.
Table 1 Diffusion coefficients (D) of monomeric GFP under different conditions. Protein
Condition
Dfast (mm2s1) (%)
Dslow (mm2s1) (%)
monomeric GFP
in living cells in lysate in solution
17.7 ± 0.6 (92 ± 1) 52.7 ± 3.0 (100) 76.2 ± 2.8 (100)
0.3 ± 0.1 (8 ± 1) N.D. N.D.
Data are presented as the means ± S.E. (in living cells: N ¼ 24, in lysate: N ¼ 4, in solution: N ¼ 20). N.D. means “not detected”.
(means ± S.E.). There was significant difference between them (p ¼ 0.006). Fig. 4B shows the CPM distribution of 369 cytosolic proteins we examined in this study (Fig. 4B). Based on a database information of protein self-assembly state (Universal Protein Knowledgebase, https://www.uniprot.org/), the data set was classified into monomer, homo-dimer and homo-trimer or higher (homo-oligomer) proteins in the inset of Fig. 4B. For example, Hsp104 which forms homo-hexameric state had relatively high CPM (16.4 kHz) (Fig. 4B inset and Supplementary Table S1). According to the information of the database, while there are proteins
The current study is a novel large-scale analysis of protein dynamics using Yeast-GFP Clone Collection. We quantified the Dfast, Dslow, ratio of these components and CPM of 369 cytoplasmic yeast proteins at endogenous expression level using FCS. By measuring cells one by one, high quality data could be provided. The data obtained from this study suggested that, overall, proteins examined do not exist homogeneously, but heterogeneously in living cells. The most important finding of this study was to provide a dataset on dynamics including diffusion and homo-oligomeric state of proteins in yeast living cells under a normal growth condition. In addition, the analysis of the dataset revealed that most of the proteins are described with a two-component diffusion model, the fast and slow diffusions. First of all, we measured the diffusion dynamics of monomer GFP in yeast living cells. The Dfast in living cells was slower than that in lysate or solution. This result reflected the molecular crowding environment in living cells [17,18]. More than 90% of monomeric GFP diffused fast as a monomeric state. This is expected because GFP is a heterologous protein derived from a jellyfish. However, ~8% of GFP diffused slowly, as was previously reported [19], suggesting that a homologous assembly or an interaction with other proteins/ components in yeast cells. Further studies are needed to determine whether this slow diffusion of GFP is an artifact of fitting or reflects the heterogeneity of GFP in living cells. Subsequently, the diffusional dynamics of 369 cytosolic proteins were quantified using FCS. We calculated Dtheoretical using Dfast of GFP in living cells. As shown in Fig. 3D and E, almost 369 proteins had slower diffusion coefficients than Dtheoretical even in the fast components, indicating that the majority of proteins interacted with other molecules in living cells. In addition, there were no
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
5
Fig. 3. Summary of diffusion coefficients (D) and the ratio of slow components and fast components. (A) Yeast GFP clone collection was used to measure the cellular diffusion constants (D) of hundreds of proteins one by one. (B) The fractions of fast and slow components were plotted against D. The closed and open circles indicate slow and fast components of each protein, respectively. FCS measurements were conducted for 3e5 cells (6 times for 10 s/cell) for each protein, and the average values were provided. (C) The distributions of Dfast and Dslow. (D) Dfast and Dslow were plotted against molecular weights (MW). The solid line shows the Dtheoretical values, which were calculated from DGFP. The fractions of fast components were plotted against MW in the inset. (E) The distribution of MWfcs/MWtheoretical. MWtheoretical was calculated by measured Dfast using eq. (5).
Fig. 4. Counts per molecule (CPM) analysis to estimate oligomeric states of proteins. (A) The CPM values of 1 and 2 GFP (the means ± S.E.) (1 GFP: N ¼ 5, 2 GFP: N ¼ 7, respectively). Experimental data were analyzed with unpaired t-test. (B) The distribution of the CPM values of each protein. Based on database information of protein self-assembly state, the data set was classified into monomer, homo-dimer and homo-trimer or higher (homo-oligomer) proteins in the inset. CPM of Hsp104, known as a homo-hexamer, is shown as a representative.
relation between the cytosolic diffusion coefficients and theoretical molecular weights (Fig. 3D), although molecular weight of molecules is one of the factors determining diffusion coefficients of them in a solution. These results suggest that the diffusion coefficients in cells reflected dynamic intracellular states of each protein, rather than their molecular weights of monomeric states. Interestingly, the evaluation of CPM revealed the homooligomeric state of each protein. For example, Hsp 42, Glt1 and Fas1, which were previously reported that they formed homooligomers, had high CPM of 12.8, 16.8 and 16.4 kHz, respectively (Supplementary Table 1) [20e22]. On the other hand, Mrp8, whose function was unknown according to a database, had relatively high
CPM (8.6 kHz) and low fraction of the fast component (63%). These observations might indicate that Mrp8 existed as a dimer and strongly interacted with other molecules, thereby functioning in the living cells. Therefore, evaluation of the dynamic status of proteins directly in single living cells provided novel insight into protein society. Recent progress in cell biology has revealed that proteins in the cell could phase-separate depending on a variety of cellular stress [3,4]. In order to understand molecular basis of the phase separation, it is important to evaluate not only information on static quantity but also on cellular dynamics of related biomolecules. Actually, previous study using proteome-wide analysis reported
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066
6
T. Fukuda et al. / Biochemical and Biophysical Research Communications xxx (xxxx) xxx
that various stresses affect protein localization and function by proteome-wide analysis [23]. It is our future study to clarify on a large-scale how the dynamics of proteins change under different conditions. Funding This work was supported by MEXT Grant-in-Aid for Scientific Research (Grant No. 19058002, 24113705, 24657070, 26116002 to HT) and Daiichi Sankyo Foundation of Life Science. Declaration of competing interest The authors declare no conflicts of interest associated with this manuscript. Acknowledgement We thank Motomasa Tanaka for providing a portion of Yeast GFP collection, Tatsuya Niwa for technical advice and valuable discussion, Yasushi Sako for the FCS measurement. Transparency document Transparency document related to this article can be found online at https://doi.org/10.1016/j.bbrc.2019.09.066. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrc.2019.09.066. FCS is a method based on fluctuation analysis of fluorescence intensity detected from the tiny confocal area. The amplitude of fluctuation reflects the number of molecules (N) in the detection volume, and the average intensity of fluorescence reflects the abundance of the analyzed protein fused with GFP. Counts per molecule (CPM), which reflects homo-oligomeric state is calculated by dividing the average intensity by N. FCS provides a fluorescence auto-correlation function, allowing the measurement of important biophysical parameters: the translational diffusion time of the molecules through the open volume. Mobility parameter, i.e., diffusion coefficient (D) of molecules reflects the molecular size (MW) of fluorescent species and the apparent viscosity in solution or in a live cell (see also Methods). References [1] S. Ghaemmaghami, W.K. Huh, K. Bower, et al., Global analysis of protein expression in yeast, Nature 425 (2003) 737e741, https://doi.org/10.1038/ nature02046. [2] Y. Ho, A. Gruhler, A. Heilbut, et al., Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry, Nature 415 (2002)
180e183, https://doi.org/10.1038/415180a. [3] S.F. Banani, H.O. Lee, A.A. Hyman, et al., Biomolecular condensates: organizers of cellular biochemistry, Nat. Rev. Mol. Cell Biol. 18 (2017) 285e298, https:// doi.org/10.1038/nrm.2017.7. [4] M. Kato, S.L. McKnight, A solid-state conceptualization of information transfer from gene to message to protein, Annu. Rev. Biochem. 87 (2018) 351e390, https://doi.org/10.1146/annurev-biochem-061516-044700. [5] S. Diekmann, C. Hoischen, Biomolecular dynamics and binding studies in the living cell, Phys. Life Rev. 11 (2014) 1e30, https://doi.org/10.1016/ j.plrev.2013.11.011. [6] S. Veerapathiran, T. Wohland, Fluorescence techniques in developmental biology, J. Biosci. 43 (2018) 541e553. [7] Y. Nomura, T. Nakamura, Z. Feng, et al., Direct quantification of gene expression using fluorescence correlation spectroscopy, Curr. Pharmaceut. Biotechnol. 8 (2007) 286e290. [8] M. Wachsmuth, C. Conrad, J. Bulkescher, et al., High-throughput fluorescence correlation spectroscopy enables analysis of proteome dynamics in living cells, Nat. Biotechnol. 33 (2015) 384e389, https://doi.org/10.1038/nbt.3146. [9] W. Christopher, H. Joseph, M. Will, et al., Fluorescence correlation spectroscopy as tool for high-content-screening in yeast (HCS-FCS), Proc. SPIE (2011) 7905. [10] W.K. Huh, J.V. Falvo, L.C. Gerke, et al., Global analysis of protein localization in budding yeast, Nature 425 (2003) 686e691, https://doi.org/10.1038/ nature02026. [11] D.C. Chen, B.C. Yang, T.T. Kuo, One-step transformation of yeast in stationary phase, Curr. Genet. 21 (1992) 83e84. [12] S. Kawai-Noma, S. Ayano, C.G. Pack, et al., Dynamics of yeast prion aggregates in single living cells, Genes Cells : devoted to molecular & cellular mechanisms 11 (2006) 1085e1096, https://doi.org/10.1111/j.13652443.2006.01004.x. [13] S. Kawai-Noma, C.G. Pack, T. Tsuji, et al., Single mother-daughter pair analysis to clarify the diffusion properties of yeast prion Sup 35 in guanidine-HCltreated [PSI] cells, Genes to cells, devoted to molecular & cellular mechanisms 14 (2009) 1045e1054, https://doi.org/10.1111/j.13652443.2009.01333.x. [14] K. Weisshart, V. Jungel, S.J. Briddon, The LSM 510 META - ConfoCor 2 system: an integrated imaging and spectroscopic platform for single-molecule detection, Curr. Pharmaceut. Biotechnol. 5 (2004) 135e154. [15] Y. Sako, M. Hiroshima, C.G. Pack, et al., Live cell single-molecule detection in systems biology, Wiley interdisciplinary reviews, Systems biology and medicine 4 (2012) 183e192, https://doi.org/10.1002/wsbm.161. [16] C.G. Pack, G. Nishimura, M. Tamura, et al., Analysis of interaction between chaperonin GroEL and its substrate using fluorescence correlation spectroscopy, Cytometry 36 (1999) 247e253. [17] T. Ando, J. Skolnick, Crowding and hydrodynamic interactions likely dominate in vivo macromolecular motion, Proc. Natl. Acad. Sci. U.S.A. 107 (2010) 18457e18462, https://doi.org/10.1073/pnas.1011354107. [18] J.W. Wojcieszyn, R.A. Schlegel, E.S. Wu, et al., Diffusion of injected macromolecules within the cytoplasm of living cells, Proc. Natl. Acad. Sci. U.S.A. 78 (1981) 4407e4410. [19] C.G. Pack, Y. Inoue, T. Higurashi, et al., Heterogeneous interaction network of yeast prions and remodeling factors detected in live cells, BMB reports 50 (2017) 478e483. [20] S.J. Kolodziej, P.A. Penczek, J.P. Schroeter, et al., Structure-function relationships of the Saccharomyces cerevisiae fatty acid synthase. Three-dimensional structure, J. Biol. Chem. 271 (1996) 28422e28429. [21] L. Valenzuela, P. Ballario, C. Aranda, et al., Regulation of expression of GLT1, the gene encoding glutamate synthase in Saccharomyces cerevisiae, J. Bacteriol. 180 (1998) 3533e3540. [22] D. Wotton, K. Freeman, D. Shore, Multimerization of Hsp42p, a novel heat shock protein of Saccharomyces cerevisiae, is dependent on a conserved carboxyl-terminal sequence, J. Biol. Chem. 271 (1996) 2717e2723. [23] K. Lee, M.K. Sung, J. Kim, et al., Proteome-wide remodeling of protein location and function by stress, Proc. Natl. Acad. Sci. U.S.A. 111 (2014) E3157eE3166, https://doi.org/10.1073/pnas.1318881111.
Please cite this article as: T. Fukuda et al., Large-scale analysis of diffusional dynamics of proteins in living yeast cells using fluorescence correlation spectroscopy, Biochemical and Biophysical Research Communications, https://doi.org/10.1016/j.bbrc.2019.09.066