^{1}

^{2}

^{3}

^{4}

^{1}

^{3}

^{5}

^{1}

^{6}

^{*}

The authors have declared that no competing interests exist.

The author(s) have made the following declarations about their contributions: Conceived and designed the experiments: MC SL SA IT. Performed the experiments: MC SL. Analyzed the data: MC SL SA TG IT. Contributed reagents/materials/analysis tools: MC SL SA TG IT. Wrote the paper: MC SL SA TG IT. Have an equal contribution as first authors: MC SL.

Fusion of harmful aggregated proteins into larger clumps increases the asymmetry of segregation of damage at cell division, favoring the production of rejuvenated cells.

Asymmetric segregation of damaged proteins at cell division generates a cell that retains damage and a clean cell that supports population survival. In cells that divide asymmetrically, such as

During their lifetime, cells accumulate damage that is inherited by the daughter cells when the mother cell divides. The amount of inherited damage determines how long the daughter cell will live and how fast it will age. We have discovered fusion of protein aggregates as a new strategy that cells use to apportion damage asymmetrically during division. By combining live-cell imaging with a mathematical model, we show that fission yeast cells divide the damage equally between the two daughter cells, but only as long as the amount of damage is low and harmless. However, when the cells are stressed and the damage accumulates to higher levels, the aggregated proteins fuse into a single clump, which is then inherited by one daughter cell, while the other cell is born clean. This form of damage control may be a universal survival strategy for a range of cell types, including stem cells, germ cells, and cancer cells.

A dividing cell can deal with damaged material in two different ways. First, the damaged material can be segregated asymmetrically during division, such that it is concentrated in one of the two newborn daughter cells, while the other cell is born clean. The damage is then removed from the population when the cell retaining the damaged material dies. Second, in phases of rapid growth, damaged material can be segregated randomly, leading to less asymmetric segregation of damage between daughters. In this case, accumulation of damage within any cell is prevented by rapid divisions that dilute the damaged material.

Protein aggregates are a type of damaged material, composed of insoluble and dense protein particles

We have recently shown that the symmetrically dividing fission yeast

Here we study the mechanism underlying the transition from symmetric to asymmetric aggregate segregation. By combining

We monitored protein aggregates using the Hsp104 disaggregase, a chaperone that binds and separates aggregated proteins

(A) Aggregate movement and fusion in control cells and after cytoskeleton depolymerization (Latrunculin B and MBC); kymographs (space-time plots, right). The fusion of two aggregates (arrows at 0′) is marked by an asterisk. (B) Mean squared displacement (MSD) of aggregates in control cells grouped by area (see labels) and a weighted fit ^{2}^{2} +^{2}, linear_{(2–5 µm2)} = 0.996, nonlinear_{(2–5 µm2)} = 0.969). (C) Aggregate segregation (|n_{1}−n_{2}|) at division in wild type and Δ

To study aggregate dynamics during the cell cycle, we followed Hsp104-associated aggregates with wide-field fluorescence microscopy (

We next studied how aggregates are segregated between cells at division. Because aggregates nucleate and move randomly, we hypothesized that sister cells arising from a morphologically symmetrical division inherit the same number of aggregates on average. Indeed, the aggregates did not segregate specifically to a cell inheriting the new or the old pole (

Based on our experimental observations, we developed a stochastic aggregation model (

(A) The model. Smallest size aggregates (gray) are generated (_{0} (

Three key processes operate on size distributions of aggregates in each of the two compartments of a cell (

Generation and fusion of aggregates within compartments were simulated with a stochastic aggregation algorithm

The experimentally measured size distribution of aggregates shows that small aggregates are found more frequently than large ones (

The parameterized model predicts a pattern of aggregate segregation at cell division by aggregate number that is between completely symmetric segregation, where the difference in the aggregate number is the minimal possible, and fully random segregation, where each aggregate can segregate to either of the two newborn cells, corresponding to the model without compartmentalization. The experimentally measured segregation pattern closely matches that predicted by the model, thereby confirming prediction iv (

If the average aggregate amount formed per cell cycle is substantially less than the amount which affects cell growth (death threshold “

We tested the effect of a range of aggregate levels on segregation dynamics and on cell viability. To increase the aggregate amount, we used stress conditions such as oxidative stress (H_{2}O_{2}) and transient or continuous heat stress (T = 40°C) (

(A) Aggregate segregation after stress (Hsp104-GFP, see labels). Thin lines encircle cells; scale bar, 1 µm. (B) Aggregate nucleation and fusion during the first cell cycle after stress. (C, Left) Three distinct optimal segregation regimes (1–3) that maximize the number of surviving cells depending on the aggregate amount in the mother cell at division (a): (1) any segregation asymmetry when the total aggregate amount is below the death threshold (_{2}O_{2} for 1 h (“H_{2}O_{2}”), 40°C for 30 min (“Heat”), continuous growth at 40°C (“Continuous heat”), and the model (see labels). Grey areas numbered 1–3 are the optimal segregation regions (see scheme and text). The data are mean ± SEM; number of cells from >3 independent experiments are given in graphs. See also

To test whether the transition to asymmetric segregation could be reproduced theoretically, we introduced stress into the model, using the parameters fitted for control conditions. We raised the aggregate generation rate

To understand which segregation modes maximize daughter cell survival for a given total aggregate amount, we model the effect of the segregation asymmetry on cell survival by assuming that, as observed experimentally

The model predicts that fusion facilitates asymmetric segregation in response to different levels of stress, where high asymmetry is optimal (

(A) Quantification of the percentage of contact events between two aggregates which resulted in fusion. Nucleation (B), fusion (C), and fission (D) events per cell cycle in the wild-type strain and in the strains in which Hsp16, Hsp40, or Hsp70 was deleted (statistical difference between wild type and mutants: **

The model predicts that reducing fusion decreases segregation asymmetry (

The decrease in fusion efficiency was specific to Hsp16 deletion, as deleting Hsp40 or Hsp70, molecular chaperones that participate in protein disaggregation

We proceeded to test the prediction of the model in the strain deleted for Hsp16. We observed that decrease in fusion resulted in a decrease in the segregation asymmetry of aggregate amount (

As predicted by the model without fusion, we observed in the experiments a ∼50% decrease in the fraction of aggregate-free cells in Δhsp16 compared to wild-type cells (

In this study, we show that fusion of aggregated proteins into a single large unit is sufficient to establish asymmetric segregation of damage, thereby generating a cell clean of aggregates. Below we explore how fusion compares to other mechanisms described to establish asymmetric segregation at cell division, and how fusion might represent a general strategy for asymmetric segregation of cellular components.

We have demonstrated that the symmetrically dividing cells of

In response to increased aggregate nucleation, two distinct mechanisms—stochastic movement and chaperone-mediated fusion of aggregates—combine to generate a single large unit of damage, which has to be segregated asymmetrically, resulting in the birth of a damage-free cell (

Due to the geometry of cell division in

How do protein aggregate dynamics and segregation in

While in

Fusing a number of molecules/components in a cell represents an opportunity to segregate asymmetrically. In mathematical terms, fusion increases the difference between the number of aggregates inherited by daughter cells at segregation. While low numbers of a component that is randomly segregated at division assures a higher variability in individual cells in the population, the formation of a unitary component assures a complete asymmetry in segregation that might be important when minimizing damage or maximizing resources. Fusion might also be a mechanism to establish asymmetry in the localization of aggregated functional molecules within the cell

Cells were grown as described before _{2}O_{2} (Sigma-Aldrich, Hannover, Germany) followed by growth at T = 30°C (70% of cells undergo mitosis,

Cells were imaged in a DeltaVision core microscope, with a motorized XYZ stage (AppliedPrecision, USA). An Olympus UPlanSApo 100× 1.4 NA Oil (R.I. 1.516) immersion objective was used (Olympus, Tokyo, Japan). The illumination was provided by a LED (transmitted light) and Lumicore solid-state illuminator (SSI-Lumencore, fluorescence), and the images were acquired with a Cool Snap HQ2 camera (Photometrics, Tucson, AZ, USA) and the SoftWorx software (AppliedPrecision, USA), using 2×2 pixel binning, to minimize light exposure (pixel size = 0.1288 µm). For long-term time lapse imaging, Z-stacks for 6–12 nonoverlapping imaging areas in the sample were acquired every 10 min (total time = 20 h) and in short time-lapses every minute (total time = 1–3 h). For single Z-stacks cells were imaged with exposure = 0.05–0.20 s, 2%–50% transmission, depending on the protein and fluorescent label. As a control for photo-toxicity, cell cycle duration and protein aggregate number were measured and found similar in the presence and absence of continuous illumination.

To quantify the total number of aggregates and to visualize small fast-moving aggregates and fusion events, we used highly inclined and laminated optical sheet microscopy (HILO)

Protein aggregates and subcellular structures were imaged simultaneously to test for co-localization and coordinated movement using bright field, a complementary set of fluorescent proteins (GFP, RFP, or mCherry) and dyes (Phalloidin and FM-464). We labeled protein aggregates indirectly with Hsp104-GFP or Hsp104-mCherry. Bright field was used to directly visualize cell poles and the division plane. Actin was indirectly labeled

Hsp104 interacts with aggregated proteins in

(EPS)

Protein aggregates move by diffusion, are not associated with the actin or microtubule cytoskeleton, and nucleate and fuse when the cytoskeleton is absent. (A) Aggregate nucleation (white stars) occurs in the cellular compartments on either side of the nucleus (old pole corresponds to the larger sister cell; new pole corresponds to the smaller sister cell). (B, Left) Average aggregate nucleation in each cellular compartment per cell cycle and average number of aggregates segregated to each sister cell and (Right) segregation of the aggregate in cells that contained only a single aggregate, to old or new pole cells. The ^{2}/s. Fitting with a nonlinear equation (^{2}^{2} +^{2}_{(linear, 0–0.1 µm2)} = 0.925 r^{2}_{(nonlinear, 0–0.1 µm2)} = 0.851). (F) Small aggregates (^{2}^{2} +^{2}_{(linear, Lat.B, 2–5 µm2)} = 0.950, r^{2}_{(nonlinear, Lat.B, 2–5 µm2)} = 0.947; adjusted r^{2}_{(linear, MBC, 2–5 µm2)} = 0.947, r^{2}_{(nonlinear, MBC, 2–5 µm2)} = 0.477). (J, Left) Actin depolymerization after Latrunculin B and (Right) microtubule depolymerization after MBC treatment. The actin cytoskeleton was also disrupted upon heat stress (see labels). (K) Quantification of nucleation and fusion events in the absence of the actin or microtubule cytoskeleton (see labels). Data are shown as mean ± SEM; number of cells are given in the graphs. Thin lines encircle cells; scale bars, 1 µm.

(EPS)

Sensitivity test of the model parameters. (A) Parameters of the model. Data are shown as mean ± SEM; number of cells are given in the graphs. The sensitivity of two key model outputs, (B) the fraction of cells born clean at division 3 after stress, and (C) the average number of aggregates per cell immediately after stress, to variations in the parameters indicated. Sensitivity is calculated as (% change in output/% change in parameter).

(EPS)

Dynamics of individual protein aggregates after stress is similar to favorable conditions. (A) Aggregate movement after stress. Fusion events (cross) are shown in the kymograph. (B) MSD of aggregates after stress grouped by size as a function of Δt (for control, see ^{2}^{2} +^{2}_{(linear, 2–5 µm2)} = 0.964, r^{2}_{(nonlinear, 2–5 µm2)} = 0.661). Similarly to the control situation, aggregates move by diffusion after stress. (C) Quantification of co-localization of actin (GFP-CHD, green, strain MC193, _{Oxidative stress} = 67 cells, n_{Heat stress} = 108 cells, _{0}), in the second division after stress (Div._{2}), and in the control population (model and experiments, see legend). (H) Number of fusion events during the first cell cycle after stress is plotted against the number of aggregates present in the cell immediately following stress from the experiment (_{1}−_{2}|, as a function of the number of aggregates at division (_{1}+_{2}), where _{1} and _{2} are the numbers of aggregates in the sister cells, in the experiment and the model. (J) Percentage of cells born without stress-induced aggregates, after a fixed number of divisions after stress. (K) Hsp104-GFP–labeled protein aggregates in wild-type cells and mutants under favorable growth conditions (see labels). (L) Aggregate amount per cell, in the wild-type cells and Hsp16, Hsp40, and Hsp70 deletion mutants. (M) Aggregate number distribution, per cell, in the wild-type cells and Hsp16, Hsp40, and Hsp70 deletion mutants (

(EPS)

Aggregates nucleate, fuse, and grow in the same cytoplasmic compartment during the cell cycle. Nucleation events are shown by the appearance of puncta (Hsp104-GFP, black) and fusion events occur by the merging of two puncta. Aggregates do not cross over from the cytoplasmic space on one side of the nucleus to the other during the cell cycle. The strain used for imaging was MC19 (Table S1). On the left, a bright-field image of the cells and on the right a maximum intensity projection of a z-stack of 10 images, acquired every minute. Movie is displayed at 7 fps. Time is shown in minutes; scale bar, 2 µm. (avi, 0.7 MB).

(AVI)

Aggregates move by diffusion in the cytoplasm. The movement of Hsp104-GFP–labeled aggregates (black dots) in short (Left, TIRF) and long (Right, conventional wide-field) time scales is shown. Fast moving small aggregates are visible, while large aggregates move slower, which is indicative of diffusive movement. The strain used for imaging was MC19 (Table S1). The movie on the left is a maximum intensity projection of five single plane TIRF images acquired at 200 fps. The movie on the right is a maximum intensity projection of a z-stack of 10 images, acquired every minute. Movies are displayed at 7 fps. Time is shown in seconds; scale bars, 2 µm (avi, 0.7 MB).

(AVI)

Aggregate segregation symmetry depends on the morphological symmetry of cell division. Hsp104-GFP–labeled aggregates segregate at division in cells that divide off-center (

(AVI)

Supporting text for experimental and theoretical procedures. (1) Supporting experimental procedures: this section contains the specific details of the experimental methods. (2) Supporting theoretical procedures: this section contains the mathematical description of the model for aggregation and aggregate segregation, including the relevant equations. A full list containing the genotype of the strains is available in Table S1.

(DOCX)

We thank J. Bähler, K. Sawin, G. Rödel, S. Vogel, and J. Glover for strains and reagents; C. Bicho, D. White, J. Peychl, B. Schroth-Diez, N. Maghelli, A. Krull, V. Ananthanarayanan, D. Ramunno-Johnson, B. Borgonovo, D. Drechsel, and I. Šarić for technical support; and W. Zachariae, J. Matos, T. Franzmann, N. Pavin, G. Salbreux, M. Glunčić, J. Shorter, and the members of the Tolić group for discussions and comments on the manuscript.

arbitrary units

green fluorescent protein

heat shock protein

mean squared displacement