www.sciencemag.org/content/351/6269/169/suppl/DC1
Supplementary Materials for
Accurate concentration control of mitochondria and nucleoids
Rishi Jajoo, Yoonseok Jung, Dann Huh, Matheus P. Viana, Susanne M. Rafelski,
Michael Springer, Johan Paulsson*
*Corresponding author. E-mail: [email protected]
Published 8 January 2016, Science 351, 169 (2016)
DOI: 10.1126/science.aaa8714
This PDF file includes:
Materials and Methods
Supplementary Text
Figs. S1 to S16
Table S1
References
Materials and Methods
Nucleoid Counting Assay
We developed a novel assay for counting individual mitochondrial nucleoids in
single dividing cells: live S. pombe stained by Sybr Green I (SGI) showed clear puncta
resembling mitochondrial nucleoids, had no detectable signal in the nucleus and had
minimal autofluorescence (Fig. S4A). These SGI puncta co-localized with a
mitochondrial matrix-targted mCherry fluorophore, and were not observed in cells
lacking mtDNA ( ρ 0 , Fig. S5B). To verify that SGI puncta were nucleoids, we co-stained
cells with a previously used nucleoid stain, DAPI (28) and saw a close correspondence
between individual nucleoids identified by each method. Using a semi-automated image
analysis method to count the number of nucleoids stained (see below), we also found that
the average number of nucleoids detected by each method was nearly identical (Fig.
S4B). This procedure produces two estimates for each individual cell, and we found that,
on average, the low estimate was approximately 92% of the high estimate. To account for
the fact that both methods then slightly undercount nucleoids, we performed a more
rigorous statistical analysis showing that each nucleoid is independently detected with an
efficiency of about 84% with either stain (Fig. S6).
Estimation of detection efficiency from DAPI and SGI staining
To estimate the staining efficiency of DAPI and SGI, each cell as assumed to
have some number of nucleoids, N, and each stain was assumed to detect individual
nucleoids with a probability, p. This probability was assumed to be the same for each
stain as the slope of the best fit line between them was 1.0. For each cell, the average
number of spots detected via DAPI and SGI was used as an estimate for Np. All cells
with the same number of average detected nucleoids were grouped and the individual
samples were treated as binomial process with N trials, each with p chance of success.
The variance of this process is Np(1-p) and p was computed by assuming that the average
number of detected nucleoids for the group was Np. The result from each of these groups
of cells was averaged with weight proportional to the number of samples and resulted in
the estimated 84% detection efficiency.
Estimation of detection efficiency from EdU staining
To further validate SGI staining we compared it to cells where nucleoids were
labeled with the nucleotide analog 5-ethynyl-2'-deoxyuridine (EdU), followed by ``click"
ligation of Alexa-488 azide (see below). EdU labels nucleoids reliably (16) but only in
fixed cells after a cell cycle arrest (Fig. S7A) and thus can only be used for assay
validation purposes. In a matched population of cells, SGI stained approximately 84% of
nucleoids stained by EdU (Fig. S7B), agreeing with our previous estimate of labeling
efficiency. SGI staining is therefore an accurate method for quantifying nucleoid numbers
in single cells, but has two significant limitations: the experimental procedure disrupts the
native localization of mitochondria and nucleoids, and nucleoids do not remain labeled if
cells resume growth. Therefore, this assay cannot be used to monitor nucleoid position or
number over time or in space. However, it does allow us to count the number of
nucleoids in each daughter of dividing cells, and when combined with judiciously chosen
2
cell division mutants, can be used to identify key properties of the nucleoids segregation
process.
Correction for Detection Efficiency of Nucleoids
The observed segregation errors indicate that nucleoids partition to daughter cells
more accurately than a binomial model However, we wanted to ensure that this result
held true even though we detected only ~80-90% of nucleoids. Huh and Paulsson showed
that unbiased incomplete detection would make the data look more like binomial
segregation than it actually is (7) Because we observed better than binomial segregation
without any corrections, this would imply nucleoids are actually even more accurately
segregated when the imperfect detection is accounted for.
To quantitatively correct for detection error of nucleoids, we use equation S81 of
Huh and Paulsson (7) If l and r represent the number of nucleoids detected in the left and
right cells respectively, L and R are the number actually in the left and right cells, and u is
the efficiency of detection, then the conversion between the average squared error
observed and the average squared error actually present is
1⎡
(l − r)2 +(u −1) l + r ⎤ = (L − R)2
2⎣
⎦
u
The result of correcting the observed data in this way is shown in Figure S11. It
shows that the nucleoid segregation is likely even more accurate than we observed. We
plotted the uncorrected data in Figure 2 of the main text as a conservative estimate of the
error in segregation.
Estimation of concentration errors during segregation
Nucleoid partitioning may in fact be significantly more accurate than it appears in
our experiments. The measured relative nucleoid concentration noise between sister cells
is 18%. This is a slight over-estimate due to the under-counting statistics explained
above. In addition, we estimated the cytoplasmic volume of the cell assuming it is
occupied only by nuclei constituting exactly 11% of the cell volume. In reality, the size
of the nucleus varies (29) and other structures such as vacuoles, ER and golgi etc.
displace cytoplasm and introduce noise into the estimate of the cytoplasmic volume.
Since we find that mitochondria segregate in proportion to the available volume (e.g. the
fit to ``equal concentration" is better when we exclude the volume of the nucleus), we are
likely overestimating the amount of concentration noise introduced at cell division.
For comparison, a mechanism that divides exactly half of the nucleoids in each
daughter cell would have concentration errors of 10% between cells. Such a high error
from a ``perfect" mechanisms arises because daughter cells differ in size on average by
5% and when there are an odd number of nucleoids in a cell, they must be divided
unevenly. Therefore, accounting for errors in measurement might indicate that the
segregation of mitochondrial nucleoids is actually more accurate than a ``perfect"
mechanism.
3
On the other hand, if nucleoids segregated by a binomial mechanism with
inheritance proportional to the cell volume of each daughter cell, the concentration error
of nucleoids between daughter cells would be 29%.
Sybr Green I and DAPI staining
All cultures were grown in YES media with 4% glucose at 30° C before staining
experiments. Overnight cultures were diluted approximately 100 to 1000-fold, allowed at
least 2 doublings and grown to OD600 ~0.1. Approximately 1mL of culture was spun at
3000g for 1 minute and washed with 1x PBS. Volume was brought up to 100 uL (PBS)
and 0.5 uL of Sybr Green I (Life Technologies, S-7567) (undiluted from stock) and/or
DAPI (Sigma-Aldrich, D9542) (5 ug/uL) was added and contents were gently mixed.
Cells were then immediately spun (as above) and washed with PBS twice before imaging.
EdU Staining
For EdU incorporation into strains yFS284 and RJP028, strains were grown from
overnight cultures at 27° C and then diluted approximately 100x and grown until OD600
~0.1. at 27° C (roughly 2 doublings). Cells were then spun down (as above) and
resuspended in media at 37° C and incubated at 37° C for 30 minutes before EdU was
added to 10 uM final concentration. Cells were then further grown at 37° C for 90
minutes and either fixed with ethanol immediately (see below) or 37% formaldehyde was
added to cultures to 3.7% final concentration and shaken at 37° C for 30 minutes. Cells
were then spun down (as above) and washed in PBS twice.
Ethanol fixation, either alone or after formaldehyde fixation, was performed by
spinning down cells (as above) and replacing PBS with approximately 1 mL 70%
ethanol. Cells were vortexed until pellet dissolved. Cells were then spun down (as above)
and washed twice with PBS, taking care to dissolve the pellet each time. EdU was ligated
to Alexa-488 azide (Life Technologies) essentially as described by Hua and Kearsey (30)
except that PBS buffer was used in place of TBS buffer. Note that formaldeyhyde
fixation was required to preserve mCherry signal and ethanol fixation was required to
perform the ``click" reaction.
Agarose pad for imaging was created as described previously (31) except that two
layers of frame seals (Biorad, SLF-1201) were placed on microscope slide to contain
slab. Agarose pad was made from 2% low fluoresence agarose (Biorad, Catalog \#1613100) in PBS (for spot counting) or YES media (for time-lapse movies). Cells were
allowed to dry on agar pad for approximately 10 minute before a No 1.5 coverslip
(VWR) was placed on top.
Microscopy
Microscopy was performed on a Nikon Eclipse Ti inverted microscope equipped
with an Orca R2 (Hamamatsu) camera, a 100x Plan Apo oil objective (NA 1.40, Nikon),
an automated stage (H117, Prior Scientific), and a Lumen 200 Pro metal arc lamp
illumination system (Prior Scientific). Image acquisition was performed using
microManager or microManager controlled by custom Matlab scripts. For time-lapse
4
imaging the microscope was encased in a custom-built incubator maintained at 30°C
throughout the experiment. The following filter cubes (Semrock) were used for image
acquisition: DAPI (excitation 390/40, dichroic 405, emission 452/45), GFP (excitation
472/30, dichroic 495, emission 520/35), YFP (excitation 500/24, dichroic 520, emission
542/27) and mCherry (excitation 562/40, dichroic 593, emission 641/75).
For spot detection, images for mCherry (mCherry filter) and SGI (YFP filter) and
DAPI (DAPI filter) were taken in Z-stacks with 0.25 um spacing in a 6 um range with
200-300 ms exposure time. Brightfield images were taken at a single Z position that
clearly showed the cell wall and any septum with 100ms exposure.
For timecourses, images were taken in mCherry and GFP channels in Z-stacks
with 1.0 um spacing over 4 or 5 um range and 50 ms or 40 ms exposure in each channel.
The percent of mitochondria in the smaller daughter was determined as the fraction of the
total fluorescence in the mCherry channel that was in the smaller daughter. The
fluorescence for each cell half was determined by first subtracting background and then
adding pixel values from all Z-planes. A minority of cells with obvious mitochondrial
photodamage (permanently balled, fragmented or aggregated mitochondria) were
excluded from analysis.
Image Analysis
To count nucleoids in dividing cells, dividing cells were first identified by their
straight and complete septa in bright field images and their outlines and septa traced
using the microbeTracker (32) software for Matlab. Cells with incomplete or rounded
septa were excluded from analysis as they had not divided or had been divided for more
than 15 minutes (D. Huh, personal communication).
To identify nucleoids, custom Matlab (Mathworks, version 2013a) code was
written to sharpen spots using a 3-dimensional unsharp mask (blurring ellipsoid was 3
pixels by 3 pixels by 1 pixel (xyz)). The sharpened image was then thresholded on the
first local maximum of the entropy distribution of pixel intensities. The Matlab
morphological commands bwmorph:clean, bwmorph:fill and imopen (with disk of radius
1) were then used on each slice. Connected areas in the 3D image (using the 6 neighbor
definition) with fewer than 20 pixels were then eliminated. All other connected pixels
were provisionally considered nucleoids. These identifications were confirmed by
viewing the sharpened maximum projection of the image and manually correcting any
missed or erroneously detected spots using the spotFinderM feature of microbeTracker.
As spots had a broad range of intensities, the maximum projection of a log transformed
version of the image was also checked for any missed spots and manually corrected. All
code is available upon request.
MitoGraph
Mitograph software was used with the appropriate settings based the xy pixel size
and the distance between images in the z plane. The surface-volume calculation was used
5
to estimate the volume of mitochondria in the cell. Cells where mitochondria were not
correctly identified were excluded from data analysis.
Cell volume estimation
To estimate the available cytoplasmic volume in each cell half at division the
nucleus was estimated to occupy 11% of the cytoplasm before division and equal sized
nuclei were transported to each cell half, regardless of the sizes of the cells. Previous data
report between 8% (29) and 12% (33) of the cell is composed of the nucleus. Using an
average of these two estimates and that 10% of the cell is composed of cell wall (33) then
roughly 11% of the cytoplasmic volume is occupied by the nucleus. Correcting for the
occupied volume made the mitochondrial and nucleoid segregation fit the ``equal
concentration" model even more accurately.
Relative absolute concentration differences between cells
To compute the concentration differences between daughter cells we used the
following formula which normalizes the difference in concentration by the concentration
of the cell before division. If l and r represent the number of nucleoids detected in the left
and right cells respectively, Vl and Vr are the cytoplasmic volumes of the left and right
cells, then the normalized concentration difference between the two cells is
l /Vl − r /Vr
(l + r)/(Vl +Vr )
Computing Distances between nucleoids
To calculate the distances between two nucleoids in the mitochondrial network,
first, nucleoids were identified as previously described and computationally overlayed
onto the network. Nucleoids were assigned to the point on the network with the nearest
Euclidean distance. For this analysis mitochondria were considered 1-dimensional objects
and two nucleoids were considered on the same edge if they were connected by
unbranched mitochondria. Distances between nucleoids on unconnected mitochondria
were not computed. Distances between nucleoids where the mitochondria between them
was branched were also not computed. Only segments greater than 1um were considered
as these are unlikely to be spurious artifacts of the image analysis. As the diffraction limit
for the Alexa-488 was was 0.227 um, any nucleoids closer together than this distance
were excluded.
For the computational ``random" model of nucleoid placement, 200 simulations
were run exactly matching the number of nucleoids observed and the lengths of
mitochondrial segments for each cell. Nucleoids were placed at a random position in
unbranched mitochondrial segments longer than 1um match the analysis performed on
the actual nucleoid data. The number of spots for each segment was Poisson distributed
with average proportional to the length of the segment and the total number of nucleoids.
To match the detection of the real spots, any simulated nucleoids closer together than
0.227 um were excluded from the simulated data as well.
6
Growth rate measurements
To measure growth rates for single cells a microfluidic device was constructed as
in (34) but enlarged for pombe and modified to allow longer channels with even feeding
to all cells.
The imaging conditions and microscopy were similar to (34) with some
modifications. Cells were grown to OD600 ~ 0.4~0.6 in SC+PMG media at 32C, then
loaded into the device. The device was connected to syringe filled with SC+PMG media,
which was pumped by syringe pump at a speed of 20 uL/min, then was mounted onto a
fluorescence microscope equipped with a custom incubator set at 32C. After waiting ~1
day to allow cells to adjust to a new environment, GFP, RFP, and bright field images
were taken at appropriate intervals. Multiple Z planes were taken to cover 3~4um thick
fission yeast. The typical experiment lasted for 3 days. Images were segmented and
tracked with a custom written code in Matlab, and then manually verified. The conditions
of growth were exceptionally uniform and had a coefficient of variation of only 2.76%
over time, 2.45% between positions and 4.15% over the depth of the growth channel.
Plasmid construction
Plasmid pRJ06 was constructed using isothermal assembly of the PCR product of
primers RJ-036 (5'-CAG GTG CCT TCG CTT TTC TTT AAG CAA GAG AAT TGT
CGA GAT GGC CTC CAC TCG TGT CCT-3') and RJ-037 (5'-GAT GAT GGC CAT
GTT ATC CTC CTC GCC CTT GCT CAC CAT GGA AGA GTA GGC GCG CTT
CTG-3') on plasmid pYES-mtGFP (gift of B. Westermann (35) and primers RJ-034 (5'ATG GTG AGC AAG GGC GAG G-3') and RJP-035 (5'-CTC GAC AAT TCT CTT
GCT TAA AGA AAA GCG AAG G-3') on plasmid pDH92 (gift of D. Huh,
unpublished). This plasmid was then cut with NotI and transformed into strain DH0 to
make RJP005.
Plasmid pRJ26 was made from isothermal assembly of the the PCR product of
primers RJ-037 and RJ-094 (5'-AAC ACG GGA TCC CCG GAT CC-3') on plasmid
pRJ07 (unpublished) and primers RJ-034 and RJ-117 (5'-AAT TCC TGC AGG ATC
CGG GGA TCC CGT GTT ATA TTA CCC TGT TAT CCC TAG CGG ATC TGC-3')
on pDH62 (gift of D. Huh, unpublished). This plasmid was cut with MfeI and
transformed into yFS284 to make RJP028.
Strain construction
PRP43-msfGFP translational fusions (for RJP041 and RJP042) were constructed
by colony PCR of YJ014 with primers RJ-157 (5'-AGT GGA TTT TTC ATG CAA GTT
GCC-3') and RJ-158 (5'-CAT ATT TTT GGC ATA AAG CTG CAC G-3') (~3.5 kB) and
subsequently transformed into RJP005 to make RJP041 and RJP025 to make RJP042
with URA selection.
YJ014 was constructed by transforming the PCR product of msfGFP-ura4 (gift of
D. Landgraf) flanked by ~300bp homologous sequences for prp43 following the protocol
provided by Bahler et al. (36) The two homologous sequences H1/H2 were amplified
7
from purified DNA from DH0 using a pair of primers YJ014-f1 (5'-AGT GGA TTT TTC
ATG CAA GTT GCC-3') and YJ014-r1 (5'-CCA GCA CCA GCA CCT GCT CCA CGT
CGA GCG TTC TTT TTT GAT CTA G-3'), and a pair of YJ014-f2 (5'-GTT TAA ACG
AGC TCG AAT TCA TCG AAA TCT AAT TTA CTG CTC GGT GAA TTA CAA
ATA T-3') and YJ014-r2 (5'-CAT ATT TTT GGC ATA AAG CTG CAC G-3'),
respectively. And then, msfGFP-ura4 flanked by the homologous sequences were
amplified from pYJ003 and H1/H2 using primers of YJ014-f1 and YJ014-r2.
Strains and plasmids are available upon request.
Supplementary Text
Correspondence with previous measurements
As a further control of our nucleoid detection methods, we also compared our
average measurements of nucleoid abundances to previous studies. In our assay newborn
cells each inherited an average of 15.4 nucleoids or 18.3 accounting for the detection
efficiency. Using previous estimates of the number of mtDNA genomes in newly divided
S. Pombe cells at 100 genomes (37) this implies that each nucleoid contains roughly 5
genomes on average, within the range reported in other organisms by.
Detailed Explanation of CDF plots
In the CDFs of relative partitioning error of nucleoid segregation, the percent of
total nucleoids in each cell half was subtracted from the percent of cytoplasmic volume in
that cell half. The absolute value of these errors for both cell halves was summed to make
the relative nucleoid segregation error for that cell. To calculate the errors for a model of
randomly placed nucleoids in mitochondria, a binomial distribution of nucleoids in each
cell half was created using the actual number of nucleoids and the actual split in the
amount of mitochondria that was segregated to each cell half.
In Fig. S14, the CDFs of relative error of WT nucleoid segregation were
computed as above. The errors for simulated spacings were created by simulating
nucleoids on a single 1-d mitochondria according the the distributions in Figure 3C. The
number of nucleoids and relative segregation of mitochondria was taken from the data of
actual cells. 100 distributions each for 168 wild-type cells were performed assuming
linear mitochondria that segregated as the observed wild-type cells did and randomly
spaced nucleoids or nucleoids with observed spacing were then computationally placed in
these mitochondria.
Mechanism of mitochondrial re-equilibration after initial separation
As shown in the main text, the mitochondria re-equilibrate throughout the cell
before division. To further understand the mechanism behind this re-equilibration after
initially segregating with the nucleus, we observed cells in which both the mitochondria
and tubulin were tagged with fluorescent proteins (MYP101 cells). We observed that the
mitochondria moved to the cell poles with the ends of the spindle but with a less clear
phenotype than cells without atb1-GFP. This may be an effect from the disruptive GFP
8
tag even though it was transiently induced in addition to the untagged copy.
Mitochondria were released from the poles before or during spindle dis-assembly in
roughly half the cells. Thus spindle dis-assembly does not appear to be necessary for
release of mitochondria from the poles.
In cells for which we can clearly score a marked release from the poles, 13 out of
15 released their mitochondria from the poles after or during the formation of new
microtubules independent of the spindle. In addition mitochondria immediately colocated
with these new microtubules, suggesting that the formation of new microtubules just
before division enable re-equilibration of mitochondria.
We also watched the process of mitochondrial re-equilibration in strains in which
the Mmb1 protein had been deleted (RJP044). In all cells observed (n=13), the
mitochondria still transiently localized to the poles of cells, indicating that Mmb1p is not
required for mitochondria to associate with the poles. Instructively, we observed one cell
in which all the mitochondria was present on one side of the cell and did not move to the
other side even after the nuclei had been separated (Fig S3B). Combined these
observations suggest that mitochondria associate with the poles independent of Mmb1p
and are re-equilibrate when the microtubule network is reformed. Full re-equilibration of
mitochondria throughout the cell seems to require Mmb1p so that mitochondria can
associate with these newly formed microtubules that span both halves of the cell just
before division.
Yeast strains
All strains were based on the S. Pombe strain FWP172 (a gift of F. Winston)
except for PHP14 (a gift of T. Fox). Deletions were constructed using standard PCR
mediated techniques or PCR techniques based on the S. Pombe deletion collection (38).
0 min
5 min
10 min 15 min 20 min 25 min 30 min 35 min 40 min
Brightfield
-10 min -5 min
pom1∆
Nucleus Mito
pom1Δ1um
Fig. S1
Timelapse images of a dividing pom1Δ cell (RJP042) in brightfield and with
mitochondrial matrix-targted mCherry (magenta) and GFP labeled nuclei (cyan). Arrows
at 5 min indicate initial separation of mitochondria at the same time as nuclei and then
9
reformation of a continuous mitochondrial network at 10 min but before division (40
min). Scale bars indicate 1um.
10
0 min
5 min
10 min
15 min
20 min
25 min
30 min
35 min
40 min
45 min
Fig. S2
Timelapse maximum projection images of a cell (MYP101) with the Cox4 leader peptide
fused to DsRed (magenta) showing the mitochondria and atb1-GFP (green) showing the
microtubules and spindle. Cells were grown in Edinburgh Minimal Media overnight to
induce nmt1 promoters. The time lapse shows that the mitochondrial network reforms as
the spindle disintegrates and individual microtubules are formed (see images at 40 min
and 45 min).
11
A
B
0 min
5 min
10 min
15 min
20 min
25 min
30 min
35 min
0 min
5 min
20 min
25 min
10 min
30 min
15 min
35 min
Fig. S3
(A) and (B) Timelapse maximum projection images of a mmb1Δ cell (RJP044) with both
the mitochondria (magenta) and the nucleus (cyan) labeled. The cell in (B) divides with
no mitochondria the bottom daughter cell.
12
A
1 um
Syber Green I
DAPI
Merge
Overlay
Nucleoids detected by Sybr Green I
B
35
30
25
20
15
10
10
15
20
25
30
Nucleoids detected by DAPI
35
Fig. S4
Quantification of mitochondrial nucleoids with fluorescent dyes (A) Sample images of an
S. Pombe cell stained with Sybr Green I (green) and DAPI (magenta). The outline of the
cell is provided for reference. Note that the overlay between SGI and DAPI spots is not
exact and that slight offsets are seen in the overlay because the spots moved slightly
between acquisition of each image. Scale bar is 1um. (B) The number of nucleoids
detected with SGI or DAPI in each cell is plotted. The area of each point is proportional
to the number of samples observed. The dashed gray line indicates equality between the
two methods. (n=116 cells (RJP005)). To detect and count spots, automated image
analysis followed by manual correction was performed (see methods).
13
B
A
1 um
1 um
Fig. S5
Sybr Green I staining is specific to mitochondrial nucleoids. (A) RJP005 imaged with
Top: Mitochondrially localized mCherry (magenta), Middle: Sybr Green I staining
(green). Bottom: overlay. Note that SGI staining was used to determine the number of
nucleoids in each cell but the location and morphology of the mitochondria and nucleoids
were disturbed. (B) ρ 0 cells (PHP14) stained with Sybr Green I (green) and
autofluoresence in mCherry channel (red). Only nucleus and no mitochondrial nucleoids
are visible.
14
25
Frequency
20
15
10
5
0
0.7
0.8
0.9
1
Detection Efficiency
Fig. S6
Histogram of detection efficiencies calculated by using the variability in detection
between SGI and DAPI and assuming a binomial model of nucleoid detection. (See
methods for details on calculation.)
15
A
150
Number of Nucleoids Detected
B
EdU
SGI
100
50
0.7
0
100
150
200
250
0.8
0.9
Staining Efficiency
300
3
Cell Volume [um ]
350
1
400
Fig. S7
EdU staining provides orthogonal method to verify Sybr Green I staining efficiency. (A)
A yeast strain (ySF284) was arrested in G2, allowed to incorporated EdU, fixed and
Alexa-488 azide was ligated to EdU with ``click" chemistry (see methods). Left:
Brightfield image of cell, Right: Alexa-488 channel visualizing nucleoids. (B) The
number of nucleoids detected by SGI and EdU plotted against cell volume. Thick lines
are running average of 20 cells. Inset: histogram of calculated efficiency of SGI vs EdU.
To calculate efficiency, the running average of nucleoids identified by SGI was compared
to the running average of nucleoids identified by EdU. For each point in the running
average of SGI, the cell volume was noted and a matching cell volume from the running
average of EdU cells was chosen for comparison. If there was no exact match in cell
volume, a linear interpolation from the two points closest in volume was used. (n=77
cells for EdU and n=108 cells for SGI).
16
B
Mitochondrial volume
Nucleoids
Best Fit Line
1
0.75
0
C
Fraction of mito and nucleoids in daughter
0.5
0.25
0.25
0.25
0.5
0.75
Fraction of cytoplasm in daughter
1
0
0
0.25
1
0.75
0.5
0.75
1
0.75
1
Fraction of cytoplasm in daughter
D
WT+50 uM mvdivi-1
pom1Δ
1
0.75
0.5
0.25
0
1
0.75
0.5
0
wee1Δ
Fraction of mito and nucleoids in daughter
Fraction of mito and nucleoids in daughter
Wild-type
Fraction of mito and nucleoids in daughter
A
0.5
0.25
0
0.25
0.5
0.75
Fraction of cytoplasm in daughter
1
0
0
0.25
0.5
Fraction of cytoplasm in daughter
Fig. S8
Analysis of mitochondrial volume and nucleoid distribution to daughter cells for (A) WT
(172 cells for mitochondrial volume, 421 cells for nucleoids) (B) wee1Δ (77 cells for
mitochondrial volume, 132 cells for nucleoids), (C) cells treated with the mitochondrial
fission inhibitor mdivi-1 (128 cells with mitochondrial volume) and (D) pom1Δ cells as
shown in Fig 1D (nucleoids, 102 cells, mitochondrial volume, 52 cells).
17
B
wild-type
3
Actual Error/Expected Binomial Error
Actual Error/Expected Binomial Error
A
2.5
2.5
2
2
1.5
1.5
1
1
0.5
0.5
0
0.1
C
Cumulative probability
wild-type
3
0.2
0.3
0.4
Size of Daughter/Size of Mother
0.5
wild-type
0
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Fraction of Mitochondria in one daughter
pom1Δ
wee1Δ
1
0.5
0
Mitochondria
Nucleoids
Binomial given cytoplasm
0
0.25
0.5
0
0.25
0.5
Relative partitioning error
0
0.25
0.5
Fig. S9
(A) Nucleoid segregation error normalized to a binomial model taking into account cell
division asymmetry plotted against the degree of asymmetric cell division in wild-type
cells. (B) Nucleoid segregation error normalized to a binomial model based on
mitochondrial segregation plotted against the degree of asymmetric mitochondrial
division in wild-type cells. For (A) and (B), only the daughter with the smaller fraction is
shown; the green points are individual cells and the black point is their average; error
bars indicate s.e.m. (C) Cumulative distribution functions (CDFs) of the relative
partitioning errors of mitochondrial volume (magenta) and nucleoid partitioning (green)
compared to cytoplasmic partitioning for wild-type (RJP005, n=168 cells), pom1Δ
(RJP025, n=52 cells), and wee1Δ (RJP029, n=77 cells). Mitochondria and nucleoids
were more accurately partitioned than the binomial model would predict (t-test,
Nucleoids: wild-type: p<10-15, pom1Δ: p<10-3, and wee1Δ : p<10-3); mitochondria: wildtype: p<10-27, pom1Δ: p<10-11, and wee1Δ : p<10-2).
18
B
wild-type
1
0.5
0
Binomial considering
cell division
Mitochondria
Nucleoids
0
0.25
0.5
pom1∆
Cumulative probability
Cumulative probability
A
0
0.25
0.5
Relative partitioning error
Cumulative probability
Cumulative probability
0.5
0
0
0
D
wee1∆
1
0.5
0.25
0.5
Relative partitioning error
Relative partitioning error
C
1
mmb1∆
1
0.5
0
0
0.25
0.5
Relative partitioning error
Fig. S10
Cumulative distribution functions (CDFs) of the relative partitioning errors but with total
fluorescence in the mcherry channel instead of extracted mitochondrial volume. Nucleoid
partitioning (green) is also shown.compared to cytoplasmic partitioning for (A) wild-type
(RJP005, n=168 cells), (B) pom1Δ (RJP025, n=102 cells), (C) wee1Δ (RJP029, n=132
cells) and (D) mmb1Δ (RJP029, n=112 cells). Mitochondria and nucleoids were more
accurately partitioned than the binomial model would predict in all cells except mmb1Δ
cells (t-test, mitochondria: wild-type: p<10-23, pom1Δ: p<10-7, wee1Δ : p<10-5 and
mmb1Δ p=0.90).
19
25
on
rn
lo
15
Observed
Corrected
e
20
Al
ΔNucleoids in daughter cells (RMS)
30
10
Binomia
5
0
l
Perfect
10
20
30
40
50
Number of nucleoids in cell before division
Fig. S11
Correcting for random undercounting of nucleoids shows that wild-type cells partition
even more accurately than observed. A running average (50 points) of the root mean
square (RMS) error between wild-type daughter cells is plotted against the total number
of nucleoids in the cell before division (blue). The corrected version (red) shifts to the
right to correct for 84% undercounting of nucleoids. The line moves down because
random undercounting is a binomial process and undoing this, shifts the trend away from
the binomial line. See methods and Huh and Paulsson (22) for details.
20
mmb1∆
A
ΔNucleoids in daughter cells
30
25
1 um
ne
20
All
or
no
15
10
5
Binomia
Perfect
0
10
Cumulative probability
B
l
20
30
40
50
Number of nucleoids in cell before division
mmb1∆
1
Mitochondria
Nucleoids
Binomial Model considering
mitochondrial volume partition
Binomial Model considering
cytoplasmic volume partition
0.5
0
0
0.25
0.5
Relative partitioning error
Fig. S12
(A) Analysis of nucleoid segregation to daughter cells for mmb1Δ cells (n=277) with size
of point representing the number of cells observed. 20% of cells segregated perfectly,
44% segregated worse than binomial and the rest fell between perfect and binomial. The
running average of 50 cells plotted as a line. (B) CDFs of partitioning errors for
mitochondrial volume and nucleoid segregation in mmb1Δ cells and a binomial models of
considering cytoplasmic volume partition and considering mitochondrial volume
partitioning. The mean nucleoid segregation error for mmb1Δ cells is larger than a
binomial model only considering its cytoplasmic segregation but not significantly so (ttest, p=0.035). However, it is significantly less than a binomial model considering its
mitochondrial volume partition (t-test, p=.0015)
21
3
2.5
0.5
0
Actual Error/Expected Binomial Error
Cumulative probability
1
Mitochondria - wild-type
Mitochondria - mmb1∆
Nucleoids - mmb1∆
Binomial + cell volume- mmb1∆
Binomial + mitochondria- mmb1∆
0
0.25
0.5
2
1.5
1
0.5
0
0.4
0.45
Size of Daughter/Size of Mother
0.5
Relative partitioning error
Fig. S13
(A) Cumulative probability distributions of mmb1Δ strains with wildtype and binomial
models for comparison.``Binomial + cell volume" refers to a model where the nucleoids
are independently assorted to daughter cells in proportion to their cell volumes.
``Binomial + mitochondria" refers to a model where the nucleoids are independently
assorted to daughter cells in proportion to their mitochondrial volumes. (B) The ratio of
the actual segregation error to the average expected binomial error vs. the relative size of
the daughter cell is plotted for each dividing cell in an mmb1Δ strain. The smaller of the
two sister cells was chosen to plot. The average and s.e.m. of these cells is also plotted.
22
Cumulative probability
1
0.5
0
0
Observed segregation errors
Errors given observed spacing
Errors given random spacing
0.2
0.4
Relative partitioning error
Fig. S14
CDF of relative nucleoid segregation errors for wild-type cells, the observed nucleoid
spacing (blue) and random spacing (gray). The random spacing model segregates
nucleoids with more error than wild-type segregation (t-test, p<10-6). Modeling the
observed spacing creates segregation that is not distinguishable from the observed wildtype segregation (t test, p=0.02).
23
WT +50 uM mdivi-1
30
∆ Nucleoids in daughter cells
A
25
All
o
rN
on
e
20
15
10
Binomial
5
0
Perfect
10
20
30
40
50
Number of nucleoids in cell before division
1
Cumulative probability
B
0.5
0
mdivi-1 Mitochondrial volume
mdivi-1 Nucleoids
WT Mitochondrial volume
WT nucleoids
Binomial Model
0
0.25
Relative partitioning error
0.5
Fig. S15
(A) Absolute error of nucleoid segregation in cells treated with 50 uM mdivi-1 for greater
than 4 hours in YES media. Size of points indicates number of cells (128 cells total). Blue
line is running average for 25 cells. (B) CDFs of partitioning errors for mitochondrial
volume and nucleoid segregation in WT cells treated with mdivi-1 cells in comparison
with untreated cells and a binomial model. The mean partitioning errors for cells treated
with mdivi-1 were not significantly different than the untreated cells for both nucleoids
and mitochondrial volume.
24
2
B
1.1
1
Normalized growth rate
Normalized growth rate
A
1.5
Wildtype
mmb1Δ
mmb1Δ (no dead cells)
0.9
0.8
0.5
0.7
0.6
1
Wildtype
mmb1Δ
0
0
C
0.2
0.4
0.6
0.8
1
1.2
1.4
Normalized concentration of mitochondria
1.6
0
0.5
1
1.5
Normalized concentration of mitochondria
0.2
Frequency
0.15
Wildtype
mmb1Δ
0.1
0.05
0
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Normalized concentration of mitochondria
2
Fig. S16
(A) The normalized growth rate of cells is plotted against the normalized concentration
of mitochondria for both WT and mmb1Δ cells. Each point is an average of 200 data
points and error bars are s.e.m. as in Figure 4C of the main text. (B) Actual data points
for part (A) and for Fig. 4D of the main text. (C) A histogram of the normalized
concentration of mitochondria WT and mmb1Δ strains. In all panels for this figure and
figure 4D of main text the growth rates are normalized to the mean of all growing cells of
that strain. The actual doubling times for WT cells was 189 min and for mmb1Δ cells was
198 minutes (for cells with finite division times). The two distributions of mitochondria
are further normalized against each other using the difference in mean mitochondrial
concentration from the MitoGraph calculations: WT had 7.4% more mitochondria than
the mmb1Δ strain on average.
25
2
Table S1.
Strains used in this study.
Name
DH0
yFS284
PHP14
DH60
RJP005
RJP016
RJP025
DH132
YJ053
DH128
DH134
YJ72
RJP028
RJP029
RJP029
MYP101 (RJP037)
RJP041
RJP042
YJ014
RJP044
Genotype
ade6::ade6+ ura4-D18 leu1-32 ade6-m210 hura4-D18 his7-366 cdc25-22 leu1::pFS181(leu1+
adh1:hENT1) pJL218(his7+ adh1:tk) (integrated at
random position)
ade6-M216, leu1-32, ptp1-1, ρ0
leu1::Padh1-GST-NES-mCherry-leu4+ leu1-32
ura4-D18 hura4-D18 leu1-32:pRJ06(leu+:adh1pr:mtmCherry)
ura4-D18 leu1-32:pRJ06(leu+:adh1pr:mtmCherry)
mmb1∆::kanMX4
ura4-D18 leu1-32:pRJ06 (leu+:adh1pr:mtmCherry),
pom1∆::kanMX
qcr7-mGFPmut3
leu1::Padh1-GST-NES-mCherry-leu1+ leu1-32
ura4-D18 hqcr7-mGFPmut3
leu1::Padh1-GST-NES-mCherry-leu1+ leu1-32
ura4-D18 h- mmb1∆::kanMX4
isu1-mGFPmut3
leu1::Padh1-GST-NES-mCherry-leu1+ leu1-32
ura4-D18 hsod1-mGFPmut3
leu1::Padh1-GST-NES-mCherry-leu1+ leu1-32
ura4-D18 hshm2-mGFPmut3
leu1::Padh1-GST-NES-mCherry-leu1+ leu1-32
ura4-D18 hura4-D18 his7-366 cdc25-22 leu1::pFS181(leu1+
adh1:hENT1) pJL218 integrated (his7+ adh1:tk),
pRJ26 integrated (pADH1-MLS-mcherry)
ura4-D18 leu+:pRJ06 (leu+:adh1pr:mtmCherry),
wee1∆::KanMX6
ura4-D18 leu+:pRJ06 (leu+:adh1pr:mtmCherry),
wee1∆::KanMX6
h+ nmt1-preCox4-DsRed::leu1+
nmt1-atb1-GFP::LEU2 ade6-M216 ura4-D18 leu1-32
ura4-D18 leu1-32:pRJ06 (leu+:adh1pr:mtmCherry),
prp43:msfGFP:URA4
ura4-D18 leu1-32:pRJ06 (leu+:adh1pr:mtmCherry),
pom1∆::kanMX, prp43:msfGFP:URA4
prp43-msfGFP-ura4+,
leu1::Padh1-GST-NES-mCherry-leu4+ leu1-32
ura4-D18 hura4-D18 leu+:pRJ06 (leu+:adh1pr:mtmCherry),
13
prp43:mGFPmut3:URA4,
mmb1::kanMX4
Source
D. Huh
N. Rhind
T. Fox
D. Huh
This Study
This Study
This Study
This Study
This Study
This Study
This Study
This Study
This Study
This Study
This Study
I. Tolic
This Study
This Study
This Study
This Study
26
References
1. Z. Liu, R. A. Butow, Mitochondrial retrograde signaling. Annu. Rev. Genet. 40, 159–185
(2006). Medline doi:10.1146/annurev.genet.40.110405.090613
2. S. Hoppins, J. Nunnari, Mitochondrial dynamics and apoptosis—the ER connection. Science
337, 1052–1054 (2012). Medline doi:10.1126/science.1224709
3. M. R. Duchen, Mitochondria in health and disease: Perspectives on a new mitochondrial
biology. Mol. Aspects Med. 25, 365–451 (2004). Medline
doi:10.1016/j.mam.2004.03.001
4. S. Mukherji, E. K. O’Shea, Mechanisms of organelle biogenesis govern stochastic fluctuations
in organelle abundance. eLife 3, e02678 (2014). Medline doi:10.7554/eLife.02678
5. Y. H. M. Chan, W. F. Marshall, How cells know the size of their organelles. Science 337,
1186–1189 (2012). Medline doi:10.1126/science.1223539
6. S. M. Rafelski, M. P. Viana, Y. Zhang, Y. H. Chan, K. S. Thorn, P. Yam, J. C. Fung, H. Li, L.
F. Costa, W. F. Marshall, Mitochondrial network size scaling in budding yeast. Science
338, 822–824 (2012). Medline doi:10.1126/science.1225720
7. M. P. Yaffe, N. Stuurman, R. D. Vale, Mitochondrial positioning in fission yeast is driven by
association with dynamic microtubules and mitotic spindle poles. Proc. Natl. Acad. Sci.
U.S.A. 100, 11424–11428 (2003). Medline doi:10.1073/pnas.1534703100
8. N. Krüger, I. M. Tolić-Nørrelykke, Association of mitochondria with spindle poles facilitates
spindle alignment. Curr. Biol. 18, R646–R647 (2008). Medline
doi:10.1016/j.cub.2008.06.069
9. E. C. Garner, C. S. Campbell, D. B. Weibel, R. D. Mullins, Reconstitution of DNA
segregation driven by assembly of a prokaryotic actin homolog. Science 315, 1270–1274
(2007). Medline doi:10.1126/science.1138527
10. M. P. Viana, S. Lim, S. M. Rafelski, Quantifying mitochondrial content in living cells
(Elsevier, 2015).
11. C. Briggs, M. Jones, SYBR Green I-induced fluorescence in cultured immune cells: A
comparison with Acridine Orange. Acta Histochem. 107, 301–312 (2005). Medline
doi:10.1016/j.acthis.2005.06.010
12. D. F. Savage, B. Afonso, A. H. Chen, P. A. Silver, Spatially ordered dynamics of the
bacterial carbon fixation machinery. Science 327, 1258–1261 (2010). Medline
doi:10.1126/science.1186090
13. J. Tauber, A. Dlasková, J. Šantorová, K. Smolková, L. Alán, T. Špaček, L. Plecitá-Hlavatá,
M. Jabůrek, P. Ježek, Distribution of mitochondrial nucleoids upon mitochondrial
network fragmentation and network reintegration in HEPG2 cells. Int. J. Biochem. Cell
Biol. 45, 593–603 (2013). Medline doi:10.1016/j.biocel.2012.11.019
14. F. J. Iborra, H. Kimura, P. R. Cook, The functional organization of mitochondrial genomes in
human cells. BMC Biol. 2, 9 (2004). Medline doi:10.1186/1741-7007-2-9
15. C. Osman, T. R. Noriega, V. Okreglak, J. C. Fung, P. Walter, Integrity of the yeast
mitochondrial genome, but not its distribution and inheritance, relies on mitochondrial
fission and fusion. Proc. Natl. Acad. Sci. U.S.A. 112, E947–E956 (2015). Medline
doi:10.1073/pnas.1501737112
16. C. Fu, D. Jain, J. Costa, G. Velve-Casquillas, P. T. Tran, mmb1p binds mitochondria to
dynamic microtubules. Curr. Biol. 21, 1431–1439 (2011). Medline
doi:10.1016/j.cub.2011.07.013
17. M. P. Yaffe, D. Harata, F. Verde, M. Eddison, T. Toda, P. Nurse, Microtubules mediate
mitochondrial distribution in fission yeast. Proc. Natl. Acad. Sci. U.S.A. 93, 11664–11668
(1996). Medline doi:10.1073/pnas.93.21.11664
18. S. Sivakumar, M. Porter-Goff, P. K. Patel, K. Benoit, N. Rhind, In vivo labeling of fission
yeast DNA with thymidine and thymidine analogs. Methods 33, 213–219 (2004).
Medline doi:10.1016/j.ymeth.2003.11.016
19. A. Salic, T. J. Mitchison, A chemical method for fast and sensitive detection of DNA
synthesis in vivo. Proc. Natl. Acad. Sci. U.S.A. 105, 2415–2420 (2008). Medline
doi:10.1073/pnas.0712168105
20. D. Landgraf, B. Okumus, P. Chien, T. A. Baker, J. Paulsson, Segregation of molecules at cell
division reveals native protein localization. Nat. Methods 9, 480–482 (2012). Medline
doi:10.1038/nmeth.1955
21. S. Tal, J. Paulsson, Evaluating quantitative methods for measuring plasmid copy numbers in
single cells. Plasmid 67, 167–173 (2012). Medline doi:10.1016/j.plasmid.2012.01.004
22. S. Taheri-Araghi, S. Bradde, J. T. Sauls, N. S. Hill, P. A. Levin, J. Paulsson, M. Vergassola,
S. Jun, Cell-size control and homeostasis in bacteria. Curr. Biol. 25, 385–391 (2015).
Medline doi:10.1016/j.cub.2014.12.009
23. I. Lestas, G. Vinnicombe, J. Paulsson, Fundamental limits on the suppression of molecular
fluctuations. Nature 467, 174–178 (2010). Medline doi:10.1038/nature09333
24. J. Paulsson, M. Ehrenberg, Noise in a minimal regulatory network: Plasmid copy number
control. Q. Rev. Biophys. 34, 1–59 (2001). Medline doi:10.1017/S0033583501003663
25. E. M. Ozbudak, M. Thattai, I. Kurtser, A. D. Grossman, A. van Oudenaarden, Regulation of
noise in the expression of a single gene. Nat. Genet. 31, 69–73 (2002). Medline
doi:10.1038/ng869
26. N. Rosenfeld, J. W. Young, U. Alon, P. S. Swain, M. B. Elowitz, Gene regulation at the
single-cell level. Science 307, 1962–1965 (2005). Medline doi:10.1126/science.1106914
27. D. Huh, J. Paulsson, Random partitioning of molecules at cell division. Proc. Natl. Acad. Sci.
U.S.A. 108, 15004–15009 (2011). Medline doi:10.1073/pnas.1013171108
28. P. Haffter, T. D. Fox, Nuclear mutations in the petite-negative yeast Schizosaccharomyces
pombe allow growth of cells lacking mitochondrial DNA. Genetics 131, 255–260 (1992).
Medline
29. F. R. Neumann, P. Nurse, Nuclear size control in fission yeast. J. Cell Biol. 179, 593–600
(2007). Medline doi:10.1083/jcb.200708054
30. H. Hua, S. E. Kearsey, Monitoring DNA replication in fission yeast by incorporation of 5ethynyl-2′-deoxyuridine. Nucleic Acids Res. 39, e60–e60 (2011). Medline
doi:10.1093/nar/gkr063
31. S. O. Skinner, L. A. Sepúlveda, H. Xu, I. Golding, Measuring mRNA copy number in
individual Escherichia coli cells using single-molecule fluorescent in situ hybridization.
Nat. Protoc. 8, 1100–1113 (2013). Medline doi:10.1038/nprot.2013.066
32. O. Sliusarenko, J. Heinritz, T. Emonet, C. Jacobs-Wagner, High-throughput, subpixel
precision analysis of bacterial morphogenesis and intracellular spatio-temporal dynamics.
Mol. Microbiol. 80, 612–627 (2011). Medline doi:10.1111/j.1365-2958.2011.07579.x
33. J.-Q. Wu, T. D. Pollard, Counting cytokinesis proteins globally and locally in fission yeast.
Science 310, 310–314 (2005). Medline doi:10.1126/science.1113230
34. T. M. Norman, N. D. Lord, J. Paulsson, R. Losick, Memory and modularity in cell-fate
decision making. Nature 503, 481–486 (2013). Medline doi:10.1038/nature12804
35. B. Westermann, W. Neupert, Mitochondria-targeted green fluorescent proteins: Convenient
tools for the study of organelle biogenesis in Saccharomyces cerevisiae. Yeast 16, 1421–
1427 (2000). Medline doi:10.1002/1097-0061(200011)16:15<1421::AIDYEA624>3.0.CO;2-U
36. J. Bähler, J. Q. Wu, M. S. Longtine, N. G. Shah, A. McKenzie 3rd, A. B. Steever, A. Wach,
P. Philippsen, J. R. Pringle, Heterologous modules for efficient and versatile PCR-based
gene targeting in Schizosaccharomyces pombe. Yeast 14, 943–951 (1998). Medline
doi:10.1002/(SICI)1097-0061(199807)14:10<943::AID-YEA292>3.0.CO;2-Y
37. Z. Chu, J. Li, M. Eshaghi, R. K. Karuturi, K. Lin, J. Liu, Adaptive expression responses in
the Pol-γ null strain of S. pombe depleted of mitochondrial genome. BMC Genomics 8,
323–11 (2007). Medline doi:10.1186/1471-2164-8-323
38. D.-U. Kim, J. Hayles, D. Kim, V. Wood, H. O. Park, M. Won, H. S. Yoo, T. Duhig, M. Nam,
G. Palmer, S. Han, L. Jeffery, S. T. Baek, H. Lee, Y. S. Shim, M. Lee, L. Kim, K. S. Heo,
E. J. Noh, A. R. Lee, Y. J. Jang, K. S. Chung, S. J. Choi, J. Y. Park, Y. Park, H. M. Kim,
S. K. Park, H. J. Park, E. J. Kang, H. B. Kim, H. S. Kang, H. M. Park, K. Kim, K. Song,
K. B. Song, P. Nurse, K. L. Hoe, Analysis of a genome-wide set of gene deletions in the
fission yeast Schizosaccharomyces pombe. Nat. Biotechnol. 28, 617–623 (2010). Medline
doi:10.1038/nbt.1628