Aging, mortality and the fast growth trade-off of Schizosaccharomyces pombe

Replicative aging has been demonstrated in asymmetrically dividing unicellular organisms, seemingly caused by unequal damage partitioning. Although asymmetric segregation and inheritance of potential aging factors also occurs in symmetrically dividing species, it nevertheless remains controversial whether this results in aging. Based on large-scale single-cell lineage data obtained by time-lapse microscopy with a microfluidic device, in this report, we demonstrate the absence of replicative aging in old-pole cell lineages of Schizosaccharomyces pombe cultured under constant favorable conditions. By monitoring more than 1,500 cell lineages in seven different culture conditions, we showed that both cell division and death rates are remarkably constant for at least 50–80 generations. Our measurements revealed that the death rate per cellular generation increases with division rate, pointing to a physiological trade-off with fast growth under balanced growth conditions. We also observed the formation and inheritance of Hsp104-associated protein aggregates, which are a potential aging factor in old-pole cell lineages, and found that these aggregates exhibited a tendency to preferentially remain at the old-poles for several generations. However, the aggregates were eventually segregated from old-pole cells upon cell division and probabilistically allocated to new-pole cells. The quantity and inheritance of protein aggregates increased neither cellular generation time nor cell death initiation rates. Furthermore, our results revealed that unusually large amounts of protein aggregates induced by oxidative stress exposure did not result in aging; old-pole cells resumed normal growth upon stress removal, despite the fact that most of them inherited significant quantities of aggregates. These results collectively indicate that protein aggregates are not a major determinant of cell fate in S. pombe, and thus cannot be an appropriate molecular marker or index for replicative aging under both favorable and stressful environmental conditions.


Introduction
for studying aging and growth in E. coli [7], except that the dimensions of the internal local alterations in culture environments. The behaviors of cells destined for death were 141 heterogeneous, but could be broadly categorized into three types: Type I (swollen), 142 Type II (hyper-elongated), or Type III (shrunken). Approximately 80% of the death 143 events were categorized as Type I, and in almost all of these cases, siblings in the same 144 observation channel synchronously died (Fig. S4D, E, and Movie S2). These death rates estimated from the decay curves were small, in the order of 10 -5 per minute 164 (or 10 -2 per generation). We noticed that our standard fluorescence imaging conditions 165 induced weak photo-damage [31]. Consequently, the estimated death rates were slightly 166 higher than of those obtained by bright field imaging alone, but the constancy of the 167 death rates was unaltered ( Fig. S4B and S4C). Trade-off between reproduction and survival in balanced growth conditions 170 We next investigated how cellular division and death rates might be interrelated. As 171 presented in Fig. 3A, we found that the death rate increased linearly with the division 172 and segregation of protein aggregates in an old-pole cell. Once formed at an old-pole 205 end, the (major) aggregate grew and tended to remain at the pole for many generations, 206 but it occasionally migrated toward the new-pole end, and was subsequently segregated 207 to the new-pole cell (Movie S4), which is qualitatively consistent with an earlier report 208 [18]. The distribution of aggregate inheritance duration, which is defined as the time 209 interval between two successive "born-clean events" in units of generation, had a peak 210 at four generations with an extended tail to the right and spreading over more than 40 211 generations (Fig. 4D). The tail can be approximately fitted by an exponential curve with 212 a decay rate of λ = 0.13 (generation -1 ), suggesting that the segregation of protein 213 aggregate to a new-pole cell is a random process that occurs once in every 1/λ = 7.8  and 2C), heterogeneity in each cell cycle length might be related to protein aggregation. 220 We quantified the load of protein aggregation using two metrics: 1) aggregate amount; 221 and 2) aggregate age, the latter being defined as elapsed time (in units of generation) 222 since the last birth without aggregate inheritance (indicated by "Born clean" bars in Fig.   223 4C). The former evaluates the current load of aggregation, whereas the latter evaluates 224 the burden of possessing the aggregate for prolonged periods. We first simply plotted  To analyze such relations in greater detail, we partitioned the data points in Fig. 5A 228 and 5B into three classes (low, middle, and high) according to the aggregation metrics 229 (aggregate amount or aggregation age), and compared the generation time distributions 230 among the classes ( Fig. 5C and 5D). The distributions were essentially identical among 231 the classes for both aggregation indices, which strongly indicates that cell cycle length 232 is unaffected by protein aggregation.  Due to the occurrence of accelerated accumulation in many extinct lineages before 248 deaths, the distribution of aggregates at the death points was shifted toward greater 249 values (Fig. 6D). However, the distribution of the amount at the kink points was very 250 close to that for the total population (Fig. 6D), which suggests that large aggregate 251 amounts are not required for initiating the process of dying. To reinforce this 252 observation, we counted the numbers of cells that exhibited deaths and kinks for given 253 ranges of aggregate amount (Fig. 6E), and evaluated the probability of death (Fig. 6F) 254 and of starting accelerated accumulation (Fig. 6G). The results showed that the 255 probability of commencing accelerated accumulation did not increase with the 256 aggregate amount, although that of observation of cell deaths was elevated, which again 257 suggests that the aggregate amount is not causative of initiation of the dying processes. 258 We also investigated whether retention of protein aggregates increased the 259 probabilities of death and of starting accelerated accumulation by counting the numbers 260 of cells that showed deaths and kinks for given aggregation age (Fig. 6H). We found 261 that both probabilities were nearly constant, irrespective of aggregation age ( Fig. 6I and 262 6J). The death probabilities for aggregation age < 3 generations were slightly lower than 263 the total death probability (1.15 × 10 -2 per generation) (Fig. 6I) possibly due to an 264 identified lag of few generations before death, after the onset of accelerated 265 accumulation (Fig. 6C). Indeed, the probabilities of commencing accelerated 266 accumulation were equally high for these small aggregation-age generations (Fig. 6J).

267
Overall, our data suggest that Hsp104-associated protein aggregation is unlikely to play 268 a major role in initiating the dying process. Our results, however, do not exclude the 269 possibility that rapid accumulation of protein aggregate might accelerate completion of 270 the dying processes post-onset. It has been suggested that fission yeast ages upon stress treatment, and inheritance 274 of large protein aggregate results in increased death probability [11]. To see if these 275 aging phenotypes are also observed in our system, we transiently treated cells with 276 hydrogen peroxide (H 2 O 2 ), a commonly-used oxidative stressor, and monitored cell 277 division/death kinetics along with protein aggregation dynamics. As expected, cells 278 immediately ceased to divide upon stress treatment (Fig. 7A, around t = 6,000 min).

279
After removal of the stress, there was a lag (around t = 6,000-6,500 min) before cells 280 resumed dividing. Strikingly, once cells started to grow again, the division rate was 281 almost the same as that previously observed under unstressed conditions (Fig. 7A). The 282 survival curve in Fig. 7B revealed an increase in death rate upon stress treatment (~10% 283 cells died during an hour of exposure to hydrogen peroxide). Although the recovery was 284 slower than division rate, the death rate also returned to the normal level seen in the 285 unstressed condition. We did not observe the progressive increase of generation time 286 after stress removal, one of the hallmarks of replicative aging; marked increase of 287 generation time was seen only in the first generation after stress removal (Fig. 7C). 288 These results suggest that the apparent deterioration in cellular growth/death is a 289 transient response to the stress, and not a manifestation of aging. 290 We next asked how protein aggregation dynamics are related to the stress response.

291
As reported earlier, we observed that oxidative stress enhanced protein aggregation, and 292 many lineages accumulated aggregate to high levels not attainable in normal conditions 293 (Fig. 7D). When the cells re-entered division cycles after stress removal, the large 294 amount of aggregate persisted (and even continued to grow in some cases). Strikingly, 295 even such significant amounts of aggregate did not affect generation time (Fig. 7E). In 296 non-stressed conditions, the amount of aggregate did not exceed 400 (× 10 3 a.u.) in 90% 297 of extinct lineages, whereas approximately 40% of the lineages that survived the stress 298 treatment until the end of measurement experienced more aggregation than 400 (× 10 3 299 a.u.) (Fig. 7F). These results further support that the absolute amount of protein 300 aggregate does not determine growth kinetics, nor cell fates.     Protein aggregation and cellular growth/death 390 A common perception is that protein aggregate accumulates during the aging 391 process or the stress response, and cells die catastrophically when the aggregation load 392 exceeds the cellular capacity. Our data, however, indicated that Hsp104-associated 393 protein aggregate is also formed in aging-free cell lineages ( Fig. 4 and Movie S4). We 394 showed that neither the aggregate amount nor the retention time affected the generation 395 time (Fig. 5). In addition, we demonstrated that cells transiently exposed to oxidative 396 stress could promptly resume normal growth, even in the presence of unusually large 397 amounts of protein aggregate induced by such stress (Fig. 7). The commonly-observed  The results also suggest that retention of the aggregates did not elevate cell death  Trade-off between reproduction and survival 425 We found that as the cell division rate elevated, the death rate increased in a linear 426 fashion (Fig. 3A). Although a simple extrapolation of the linear trend predicts 427 immortality of single cells when the division rate is below r min , we have not been able to 428 experimentally achieve stable growth with such a low division rate under our current 429 measurement setup. The slow growth of cells with a division rate close to, or even 430 smaller than, r min could be achieved by applying stressors, such as high/low 431 temperatures, nutrient limitation, drug exposure, or harsh chemical conditions (e.g., 432 extreme redox environments, high/low osmolarity). However, stress responses would 433 render internal cellular states different from those in non-stressed conditions. We 434 speculate that this linear trend is a hallmark of a balanced growth state, rather than a 435 universal constraint on cell division and death rates in any environment.  phenotype of DNA damage response [54], was observed in only ~20% cases (Fig. S4E).   The PDMS base and curing agent (Sylgard 184) were mixed at a ratio of 10:1, 509 poured onto the SU-8 mold in a container, and de-gassed using a vacuum desiccator.

510
Curing was performed at 65°C overnight. The device was peeled from the mold,

Long-term time-lapse experiments 551
For long-term time-lapse measurements, 10 mL of a log-phase culture of yeast cells 552 at 28-34°C in YE containing 3% glucose or EMM containing 2% glucose was 553 concentrated 50 fold by centrifugation and injected into the microfluidic device using a 554 1 mL syringe (Terumo). Cells were loaded into the observation channels by gravity, 555 simply slanting the device. The loading procedure typically took a couple of hours, 556 during which time the cells often entered into an early stationary phase in response to 557 the highly crowded environment, resulting in a time lag before stable growth was 558 achieved. The device was supplied with appropriate medium supplemented with a low 559 concentration (10 µg/mL) of ampicillin sodium (Wako) to minimize the risk of bacterial 560 contamination. Note that ampicillin has no effect on fission yeast growth. The flow rate 561 was 10-15 mL/h. For transient oxidative stress treatment, the medium was changed to 562 YE containing 2 mM hydrogen peroxide for 1 hour, then switched back to YE. 563 We used a Nikon Ti-E microscope with a thermostat chamber (TIZHB, Tokai Hit),   Fig. 1-3), or 5 min (for the experiments described in Fig. 4-9). Exposure times were 570 400 ms (for mVenus), 200 ms (for GFP), 100 ms (for mCherry), and 10 ms (for bright 571 field). 574 The acquired fluorescence images were converted into binary images using a 575 custom-written OpenCV program. The binary images were used to identify cellular 576 regions or ROIs, and lineage tracking (relating ROIs along lineages) was performed 577 using a customized ImageJ macro. Transition between ROIs along each lineage was 578 analyzed to mark cell division points where an ROI area suddenly decreased more than 579 1.5 fold. To mark cell death points, two criteria were employed: 1) if there was no 580 division during a 360-min window, then the beginning of the window was defined as a 581 death point, and 2) if there was a profound (more than 1.75 fold) decrease in 582 fluorescence during a 30-min time window, then the beginning of the window was 583 defined as a death point. We confirmed that the decay curve of surviving cell lineages 584 obtained using these death criteria quantitatively concurred with that obtained by 585 manual image inspection (Fig. S4A). In the data set used in Fig. 6-9, we examined all 586 of the cell-size trajectories by eye and manually marked death points so as to ensure 587 confidence in the data.

Single-cell lineage tracking and division/death points identification
Equivalently, (Eq. 1) can be rewritten as is the survival function (complementary cumulative 607 distribution) of ! ! , which represents the probability of a newly divided cell 608 remaining undivided until age !. One can show that ! ! = ! ! ! ! !", and then 609 ! ! ! = ! ! ! ! becomes a probability density function, and can be interpreted as 610 "age distribution" along cell lineages. Therefore, Eq. 2 can be also expressed as We numerically estimate based on a discretized version of Eq. 3, i.e., where is the generation time and is the number of samples.

629
ranges were shown as error bars in Fig. 3A and 3B.
To evaluate errors in the death rate estimations, we produced simulated decay 631 curves of the surviving fraction using parameters (death rate, initial cell number, and Statistical evaluation of death (or death onset) probability 639 We first estimated death probability per generation p 0 to be 1.15 × 10 -2 from the 640 survival curve. In Fig. 6F and I, death probability p for each aggregate amount or 641 aggregation age was then tested using a binomial test for the two-tailed null hypothesis implemented binomial testing at the significance level = 0.05 ( Fig. 6G and 6J).