Content uploaded by Douglas Brian Murray
Author content
All content in this area was uploaded by Douglas Brian Murray on Apr 08, 2014
Content may be subject to copyright.
A genomewide oscillation in transcription gates DNA
replication and cell cycle
Robert R. Klevecz*, James Bolen, Gerald Forrest, and Douglas B. Murray
Dynamics Group, Department of Biology, Beckman Research Institute of the City of Hope Medical Center, Duarte, CA 91010
Edited by Steven L. McKnight, University of Texas Southwestern Medical Center, Dallas, TX, and approved November 26, 2003 (received for review
October 7, 2003)
Microarray analysis from a yeast continuous synchrony culture
system shows a genomewide oscillation in transcription. Maxi-
mums in transcript levels occur at three nearly equally spaced
intervals in this ⬇40-min cycle of respiration and reduction. Two
temporal clusters (4,679 of 5,329) are maximally expressed during
the reductive phase of the cycle, whereas a third cluster (650) is
maximally expressed during the respiratory phase. Transcription is
organized functionally into redox-state superclusters with genes
known to be important in respiration or reduction being synthe-
sized in opposite phases of the cycle. The transcriptional cycle gates
synchronous bursts in DNA replication in a constant fraction of the
population at 40-min intervals. Restriction of DNA synthesis to
the reductive phase of the cycle may be an evolutionarily impor-
tant mechanism for reducing oxidative damage to DNA during
replication.
U
ntil recently, dynamic analyses of the regulation of gene
expression were necessarily confined to data from a few
protein products or messages or were approached indirectly by
means of perturbation analysis supported by computer simula-
tions (1–3). Cell structure and the spatial compartmentalization
of metabolic and macromolecular processes are well understood
and have been for some time. In contrast, the temporal orga-
nization of cells, the origins of order and periodicity in pheno-
type, while intriguing several generations of theoretical biolo-
gists (4–6), have only recently become accessible to definitive
experimentation. The impediments to such experimentation
have, in part, been removed by microchip and expression array
technologies (7–11), which allow one to focus not so much on the
proximal function of a few genes but on the search for a global
mechanism capable of organizing a stable phenotype and timing
cellular events.
The gating of cell cycle events such as DNA replication and
cell division by an oscillator that is an integral submultiple of the
cell cycle was first observed as ‘‘quantized’’ generation times in
mammalian cells (12) and subsequently extended to a wide
variety of organisms (13, 14). Recently, evidence has been
obtained in fission and budding yeast that is consistent with the
idea that generation times are quantized, or that there is a
free-running oscillator underlying the cell cycle (15, 16). One
expectation that follows from this observation is that the entry
into S phase or cell division would occur only at times that are
equal to or multiples of the fundamental period. The question of
whether such timing extends to transcription as a whole as
suggested by recent analyses (10, 11, 17), or whether the oscil-
lation will be limited to genes already identified as important in
cell cycle progression, should be examined.
Much of our success in the genetic and molecular genetic
dissection of the genome has been accomplished by setting aside
concerns about the moment-to-moment, or hour-to-hour,
changes in the transcriptome. This facility with genetic dissection
has led to an emphasis on steady-state relationships in the
connectivity among genes. In consequence, it may be surprising
to find evidence for genomewide oscillations. Computational
studies have suggested that periodic gating might be modeled as
a population of coupled attractors, where each attractor element
is taken to represent the dynamic behavior of individual genes or
clusters of coregulated genes (18–21). If the cell is a coupled
system, oscillations already identified in ‘‘cell cycle-regulated’’
genes might imply genomewide oscillations because behavior of
a transcript can be readily tuned or moved to different patterns
of expression internally by interactions with other transcripts, or
externally by cell-to-cell signaling or perturbations. Indeed, it is
difficult in coupled complex systems to maintain some elements
or genes at a constant level while others are oscillating.
Materials and Methods
Media and Culture Conditions. The WT parent strain used through-
out this study was the diploid Saccharomyces cerevisiae IFO 0233.
Details of culturing methods were as described (22–26). Fer-
menters were from B. Braun Biotech, Surrey, U.K. (Model
Biolab; working volume 750 ml). The basic medium consisted of
5g兾liter (NH
4
)
2
SO
4
,2g兾liter KH
2
PO
4
, 0.5 g兾liter MgSO
4
䡠7H
2
O,
0.1 g兾liter CaCl
2
䡠2H
2
O, 0.02 g兾liter FeSO
4
䡠7H
2
O, 0.01 g兾liter
ZnSO
4
䡠 7H
2
O, 0.005 g兾liter CuSO
4
䡠 5H
2
O, 0.001 g兾liter
MnCl
2
䡠4H
2
O,1ml兾liter 70% H
2
SO
4
,and1g兾kg Difco yeast
extract. Glucose medium was supplemented with 22 g兾liter
glucose monohydrate and 1 ml兾liter Sigma Antifoam A. Unless
otherwise stated, the fermenters were operated at an agitation
rate of 750 rpm, an aeration rate of 150 ml䡠min
⫺1
, a temperature
of 30°C, and a pH of 3.4. Dissolved oxygen (DO) levels in the
medium, carbon dioxide production, and oxygen consumption
were measured every 10 s. Cultures were not nutrient limited,
and glucose levels oscillated between 50 and 100
M in each
cycle.
Total RNA Preparation. RNA samples were collected every 4 min
(Fig. 1). Five hundred microliters of yeast sample was pipetted
into 1 ml of ice-cold absolute ethanol. Samples were stored at
⫺72°C overnight and centrifuged at 10,000 ⫻ g for 2 min, and the
RNA was isolated by using glass beads to rupture the cells
according to the small-scale RNA procedure of Brown (27). The
RNA was dissolved in diethyl pyrocarbonate-treated milli-Q
water and DNase treated with DNA-free (Ambion) in buffer.
The final RNA samples were analyzed in a Uvikon spectropho-
tometer (Kontron Instruments, Zurich), giving 260兾280 ratios of
2.0–2.2. Samples were also analyzed on an Agilent (Palto Alto,
CA) 2100 Bioanalyzer, giving 25S兾18S ratios of 1.8–2.0.
Target Preparation兾Processing for Affymetrix GeneChip Analysis.
Purified total RNA samples were processed as recommended in
the Affymetrix GeneChip Expression Analysis Technical Man-
ual (Affymetrix, Santa Clara, CA). RNA samples were adjusted
to a final concentration of 1
g兾
l. Typically, 25–250 ng was
loaded onto an RNA Lab-On-A-Chip (Caliper Technologies.,
This paper was submitted directly (Track II) to the PNAS office.
Abbreviation: DO, dissolved oxygen.
See Commentary on page 1118.
*To whom correspondence should be addressed. E-mail: rklevecz@coh.org.
© 2004 by The National Academy of Sciences of the USA
1200–1205
兩
PNAS
兩
February 3, 2004
兩
vol. 101
兩
no. 5 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0306490101
Mountain View, CA) and analyzed in an Agilent 2100 Bioana-
lyzer. Double-stranded cDNA was synthesized from 10
gof
total RNA by using a SuperScript double-stranded cDNA syn-
thesis kit (Invitrogen) and oligo(dT) primers containing a T7
RNA polymerase promoter. Double-stranded cDNA was used as
a template to generate cRNA by using a BioArray High-Yield
RNA Transcript Labeling kit (Enzo Diagnostics). The biotin-
labeled cRNA was fragmented to 35–200 bases following the
Affymetrix protocol. Ten micrograms of fragmented cRNA was
hybridized to yeast S98 Affymetrix arrays at 45°Cfor16hinan
Affymetrix GeneChip Hybridization Oven 320. The GeneChip
arrays were washed and then stained with streptavidin-
phycoerythrin on an Affymetrix Fluidics Station 400, followed by
scanning on a Hewlett–Packard GeneArray scanner. Results
were quantified and analyzed by using
MICROARRAY SUITE 5.0
software (Affymetrix). EXCEL files were created to permit fur-
ther processing in
MATHCAD (Mathsoft, Cambridge, MA), SIG-
MAPLOT,orMATLAB (Mathworks, Natick, MA). Intensity values
for each of the 6,317 ORFs in the chip were linked to the
Saccharomyces Genome Database site, and both their genetic
and physical map locations were associated with the intensity
values for each gene. The results for all ORFs scored as present
by using the default Affymetrix settings were identified accord-
ing to the original sample number and the phase in the DO
oscillation to which they are mapped for presentation. Further
analysis was performed for any ORF present in at least one
sample in each of the three cycles. Of the ORFs scored as present
by these criteria, 41 were removed from the analysis based on the
recent sequence reanalysis (28).
Northern Analyses. Total RNA (5–7
g) was denatured in glyoxal
and analyzed on a 1.0% agarose gel according to the procedure
of Burnett (29). The RNA was transferred to a Nytran nylon
membrane by using high-efficiency transfer solution (Tel-Test,
Friendswood, TX) in a rapid downward flow turboblotter
(Schleicher & Schuell), washed in 20 mM Tris buffer, pH 8 at
60°C for 10 min to remove glyoxal and baked (1 h, 80°C). The
remaining gel was stained and checked for complete RNA
transfer. Probes were labeled with
32
P dCTP by using the
Prime-It RmT random primer labeling kit (Stratagene). Blots
were prehybridized in QuikHyb solution (Stratagene) at 68°C for
30 min. Hybridization was carried out at 68°C in QuikHyb for 2–3
h. The blots were washed, exposed to a Molecular Dynamics
phosphor screen, and analyzed on a Typhoon 9410 variable mode
imager (Amersham Pharmacia).
Flow Cytometry. Samples were taken from fermenter cultures
(300
l; ⬇1 ⫻ 10
8
cells) and fixed by adding 700
lof95%
ethanol and stored at 4°C. Before staining for flow cytometric
analyses aliquots of 50
l (containing ⬇1 ⫻ 10
7
cells) were
washed twice in 1 ml of PBS, centrifuged at 5,000 g, and
resuspended in PBS. Samples were then treated with RNase (150
g enzyme; 37°C for 2 h) and stained overnight with propidium
iodide (PI) (5
g for ⬇1 ⫻ 10
7
cells). The samples were analyzed
on a MoFlo MLS (Dako Cytomation, Fort Collins, CO) flow
cytometer. Data were acquired by using dual laser excitation.
Light scatter signals (forward and side) were acquired with
excitation from a HeNe laser (Melles Griot, Carlsbad, CA). PI
was excited with 500 mW at 488-nm wavelength from an Innova
90 Argon laser (Coherent Radiation, Santa Clara, CA). The PI
signal was split through a 580DRLP and a 630DRLP dichroic
beam splitting filters and collected with a 640LP filter (Omega
Optical, Brattleboro, VT).
Results
The Benchmark Oscillation in DO. To begin an experimental analysis
of the connectivity relationships among genes, it was essential to
have a very reproducible and precise biological system. Oscilla-
tions in respiratory activity in budding yeast and other organisms
have been known for many years (30, 31). At sufficient cell
densities in batch or continuous cultures, cell-to-cell signaling
synchronizes the culture with respect to oxidative and reductive
functions (22). These changes are mirrored by changes in the
mitochondrial structure and function (23), flux through the citric
acid cycle (24), and redox state of the cell measured by
NAD(P)H fluorescence and glutathione levels (24). Synchroni-
zation of the respiratory oscillation in the population of cells
appears, at a minimum, to involve respiratory inhibition by the
release of small amounts of H
2
S (25) and phase shifts induced by
acetaldehyde (26). Caused presumably by initial condition de-
pendence, the period of the oscillation can vary from 40 to 44
min, but the coefficient of variation in period or amplitude in a
given culture, once the oscillation is established, is ⬍2% (32). In
Fig. 1, the DO level, indicating the respiratory (low DO) and
reductive phases (high DO levels) is shown for one of the
continuous cultures sampled for these experiments, together
with the points in the cycle when samples were taken.
Expression Microarray Analysis. Using Affymetrix chips, the pat-
terns of gene expression for 5,329 expressed genes were followed
by sampling at 4-min intervals through three cycles of the DO
oscillation. The reproducibility of the global pattern of expres-
sion can be seen in the color contour map of expression levels
through the three cycles where orange-red indicates high levels
of expression and blue indicates low levels. Genes are arranged
in Fig. 2 according to their time of maximum expression in the
cycle. The time of maximum gene expression in the cycle was
determined by averaging gene expression in the three replicates
(Fig. 2A), assigning the phase of maximum expression to each
transcript then examining how well this assignment is repro-
duced through the three cycles (Fig. 7, which is published as
supporting information on the PNAS web site). More than 87%
of all expressed genes are expressed maximally in the reductive
Fig. 1. Sampling for expression microarray analysis from continuous syn-
chrony cultures. Respiratory-reductive synchrony is monitored by using DO
levels in the media of aerobic continuous cultures. The sample times are shown
imposed on the DO oscillation. High DO levels are associated with the reduc-
tive phase and low DO levels with the respiratory phase of the cycle. Samples
were taken at 4-min intervals through three complete cycles each from two
independent cultures. A total of 68 RNA samples were isolated and analyzed.
All RNA samples were analyzed for quality by using an Agilent Bioanalyzer
capillary electrophoresis system. Thirty-two samples were selected from the
best RNA profiles and aligned based solely on their phase position on the DO
curve, before Affymetrix expression analysis. Sampling for flow cytometry is
shown as red squares and for RNA as yellow triangles.
Klevecz et al. PNAS
兩
February 3, 2004
兩
vol. 101
兩
no. 5
兩
1201
CELL BIOLOGY SEE COMMENTARY
phase of the cycle. Of these, 2,429 genes are maximally expressed
early in the reductive phase, whereas 2,250 genes are maximally
expressed late in the reductive phase of the cycle. The remaining
650 genes are maximally expressed in the respiratory phase of the
cycle (Fig. 2B). In general, the genes maximally expressed in the
respiratory phase show higher-amplitude oscillations than do
those maximally expressed in the reductive phase. The average
peak-to-trough ratio in the respiratory phase is 2.40, whereas in
the reductive phase it is 1.89. Only 41 transcripts show a ⬎6-fold
change, whereas just 7 show a ⬎20-fold change. It appears that
low-amplitude oscillations in expression are a near universal
property of this system.
Northern Blot Analysis. Probes for six genes were prepared from
yeast cDNA by PCR using primers that amplify the entire coding
regions of the expressed genes (Fig. 3). These genes were chosen
as representative based on their patterns of expression and time
of maximum expression as indicated: early respiratory phase
(STR3), mid respiratory phase (MET17), late respiratory兾early
reductive phase (CRC1), early reductive phase (CYB5), and late
reductive phase (YRO2 and YDR070c). A total of 18 RNA time
series samples taken from the second and third cycles of Fig. 7
were subjected to Northern blot analysis for each of the six genes.
Comparison between the Affymetrix intensity results and the
Northern blots was accomplished by dividing the radioactivity
after background subtraction by the average radioactivity in all
18 samples. This scaling to the average expression of all mea-
Fig. 2. Average expression levels from three cycles of the respiratory oscil-
lation. Color contour (intensity) maps of the expression levels of the 5,329
expressed genes are shown for all 32 RNA samples through three cycles of the
DO oscillation. (A) The average expression level for the three biological
replicates are shown. High levels of expression are orange, and low levels are
blue. Genes were scored as present based on the Affymetrix default settings
and included in the analysis if at least 1 of the 32 was scored as present in each
of the three cycles. Genes are defined according to the most recent assign-
ments (23). Values shown here were scaled by dividing the average expression
level for each gene into each of the time-series samples for that gene.
Transcripts were ordered according to their phase of maximum expression in
the average of the three replicates. This same scaling and ordering was used
in Fig. 7. Samples are identified according to their phase in the cycle (0–360°兾
cycle). Sample phases are shown in reference to the DO curve (thick black line,
B). The reductive phase is taken to be the period of minimal oxygen consump-
tion (maximum DO, yellow background) in the interval between the minimum
DO levels, and the respiratory phase is shown against a blue-green back-
ground. (B) Summary of the results for the time of maximum expression
(red line) for the transcripts of A. Color scale: orange-red ⫽⬎1.6 and dark blue
⫽⬍0.8.
Fig. 3. Comparison of Affymetrix microchip data with Northern blots.
Northern blot analyses of transcripts found to be maximally expressed at
differing points in the redox兾transcriptional cycle were compared with the
results from the microarray analyses. Six clones of representative genes were
examined in 18 time-series samples representing 1.5 cycles of the DO oscilla-
tion and were scaled by dividing each time point by the average of all of the
points. Red lines indicate Northern blot data, and black lines indicate
Affymetrix intensity data.
1202
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0306490101 Klevecz et al.
surements for each transcript is the same scaling as was per-
formed on the Affymetrix fluorescence intensity measures
throughout this study. In Fig. 3, the results of the two assays are
compared. The patterns of expression found in the Affymetrix
microchips are reproduced to a considerable degree by the
results from the Northern blot analysis.
Temporal Organization of Functionally Related Transcripts. Prior
emphasis in microarray analysis has been on paired treated
versus control samples examining those genes showing differ-
ences ⬎ 2-fold, although it has been reported that changes in the
1.3- to 1.4-fold range can be significant (33). Of the 5,329
expressed genes only 161 show oscillations in the range of 1.3- to
1.4-fold peak-to-trough differences through the cycle. These fall
into two classes with 118 showing intensities ⬎2,000 per slide.
Most of these are known highly expressed genes such as actin,
ribosomal proteins (52 of 118), genes of intermediary metabo-
lism such as PGK, and other structural or maintenance proteins.
This low-amplitude oscillation in maintenance genes is consis-
tent with the report of Warrington et al. (34), who found
evidence that maintenance genes in mammalian systems show
oscillations. If the time course of actin (ACT1) is followed, there
is an intensity oscillation of 1,350 fluorescent units superimposed
on a high actin level of 5,862 intensity units within the cycle.
Thus, although the fold change in expression is low, the pattern
of oscillations in ACT1 and the others in the cluster with high
average expression levels is clear. The pattern shows three peaks
of expression, as though these constitutive expressers are tran-
scribed at every permitted phase, whereas expression of other
functionally related groups is limited to one phase. This high
transcript level, low-amplitude behavior, is shown for the cyto-
solic ribosomal proteins in Fig. 4 Upper Right.
To look at the temporal organization of functionally related
transcripts (Fig. 4), we compared transcripts for genes previously
identified as important in respiration with those identified with
respiratory inhibition (reduction). For example, transcripts for
mitochondrial ribosomal proteins are maximally expressed dur-
ing the reductive phase when mitochondrial function is minimal
(23) and the cristae are more clearly defined (orthodox config-
uration). Conversely, transcripts such as those involved in me-
thionine and sulfur metabolism, important in the establishment
of the reductive phase and DNA replication are made early in the
respiratory phase, before their putative maximal function at the
beginning of DNA replication (Fig. 4). As with all transcripts
examined, there is a 4- to 12-min lag between the peak in
transcript level and the expected time of maximum function of
the protein product. This phase anticipation is functionally
significant if protein synthesis is similarly temporally coherent
and occurs with a fixed-phase relationship to transcription. In
support of this idea, the ubiquitin-constituent transcripts of the
proteosome are clustered and maximally expressed in the re-
ductive phase (Fig. 4 and Table 1, which is published as
supporting information on the PNAS web site).
The
␣
-DNA-polymerase兾primase complex is made up of four
subunits, each of which is an independent transcript with non-
contiguous physical map locations. Although it could be other-
wise, one might expect that the biological function would make
simultaneous transcription likely and thus offers an opportunity
to apply a biological rationale to the examination of the patterns
in the transcriptional cycle. The average of the three replicates
of the four subunits are shown together with the SD of intensity
change (Fig. 5). The timing of expression shows the same 4- to
12-min phase anticipation as mentioned above, consistent with
the timing of DNA synthesis from the flow cytometric analysis
of the cell cycle presented below. Similarly, transcripts function-
ally related to sulfur metabolism also show very reproducible
patterns (Fig. 5). As an alternative to the color map of Fig. 2, a
subset of the early reductive-phase transcripts representative of
the range of expression in this group have been coplotted with
the DO oscillation (Fig. 8, which is published as supporting
information on the PNAS web site). Table 1 contains the
expression patterns for the transcripts shown in Figs. 4 and 5.
Flow Cytometry. The relationship between DNA replication and
the transcriptional cycle was studied by flow cytometric analysis.
Fig. 4. Temporal organization of functionally related transcripts. Subsets of
functionally related transcripts were color-mapped as in Fig. 1 to show the
time of expression maximums as described above. The files used to generate
these color maps are available Table 1. For sulfur metabolism transcripts
(Lower Right) the scale is orange-red ⫽⬎2.2 and dark blue ⫽⬍0.8. For all
other panels, orange-red ⫽⬎1.4 and dark blue ⫽⬍0.8.
Fig. 5. Cycle-to-cycle reproducibility in selected functional groupings. Sub-
tracting the minimum expression and scaling the result to an average of 1
determines the change in average intensity for each gene. (Left) Six sulfur-
associated genes (see Table 1) were averaged, and the SD was determined.
(Right) The genes for the four subunits of the
␣
DNA polymerase兾primase
complex were averaged, and the SD was determined. The error bars represent
1 SD.
Klevecz et al. PNAS
兩
February 3, 2004
兩
vol. 101
兩
no. 5
兩
1203
CELL BIOLOGY SEE COMMENTARY
By sampling at 4-min intervals through four cycles of the DO
oscillation (Fig. 6C) and then aligning multiple flow cytometric
DNA histogram analyses with the respiratory oscillation, we
found that DNA replication in these cells begins abruptly at the
end of the respiratory phase, as oxygen consumption decreases
and H
2
S levels rise. To make this synchronous gating of cells into
S phase clear, all 39 of the 1D propidium iodide-DNA histo-
grams (Fig. 6A) were stacked, and the values for S were
determined by removal of G
1
and G
2
by Gaussian deconvolution.
Although ⬍10% of the cells are gated into S phase in any turn
of the respiratory兾transcriptional cycle, this synchronous gating
is so precisely timed that a track representing the movement of
cells from early to late S can be followed in the 3D reconstruc-
tions. In Fig. 6B, the increase in DNA content is shown by the
green track running from left to right and diagonally upward
from early to late S phase, DNA replication continues through-
out the reductive phase, ending just as respiration begins. The
duration of S phase and DNA synthesis in these cells is 24 min,
close to that observed in the most rapidly growing high glucose
cultures. The movement of the synchronous cohort through S
can also be followed in animations viewed at www.talandic.com.
The behavior of DNA replication, the most readily identifiable
cell cycle event in this culture system, is dynamically similar to
the circadian time-scale gating seen in animal model systems (35)
and humans (36) where a correlation between glutathione levels
and circadian gating of cells into S phase has been observed (36).
The role of reactive oxygen intermediates and the cellular
mechanisms preventing damage to single-stranded DNA during
replication have been extensively studied (37, 38). However, the
possibility that temporal organization in the cell has evolved to
prevent or minimize such damage has received little attention. In
the case of RNA synthesis where repair pathways for oxidative
damage are unknown, it is of interest to note that 87% of the
transcripts examined here are maximally expressed during the
reductive phase.
Discussion
Although the respiratory oscillation has been investigated in a
number of laboratories, neither the relationship between tran-
scription and respiratory metabolism, nor its connection to the
cell cycle and to the gating of cells into S phase and mitosis has
previously been explored. Under the growth conditions com-
monly used in this system, the population doubling time is ⬎8h,
and it has been assumed that the ⬇ 40- to 44-min respiratory
oscillation was uncoupled from the cell cycle. It appears rather
that this short period cycle might be the fundamental timekeep-
ing oscillator because transcription and the gating of DNA
replication and the duration of the cell cycle show similar
quantized distributions. In the strain of yeast used in these
studies a series of deletion mutations have been developed that
show DO oscillations with either half or twice the period (20 and
80 min) of the WT (39). It will be of interest to determine
whether these period additions or period doublings of the DO
oscillation and the respiratory cycle (39) are manifested as well
in the period of transcription. A reanalysis of population dou-
bling time data in yeast cultures growing on different media with
differing carbon sources from rather poor to rich found that
population doubling times cluster at multiples of the ⬇40-min
fundamental (32, 40), further urging upon us the view that the
cell cycle is a downstream manifestation of a timekeeping
attractor (17). The spatio-temporal dynamics and cell-to-cell
signaling in these vigorously stirred cultures that gives rise to
precise periodic outputs is the subject of ongoing investigations.
Alter et al. (10, 41) and others (11, 19) in their reanalysis of
the original Stanford cell cycle data using singular value decom-
position or wavelet decomposition have all come to a similar
conclusion: that there is evidence for a genomewide, low-
amplitude oscillation in transcription. Here, using a yeast con-
tinuous culture system exhibiting a very high level of precision
and stability, genomewide oscillations in transcription have been
demonstrated by using Affymetrix chips and replicate sampling
through three cycles of the respiratory oscillation. The long-
standing assumptions that the oscillator is uncoupled from the
cell cycle and that the oscillation is confined to a small subset of
genes intimately connected to energy metabolism is refuted by
the finding of genomewide oscillations in transcription. We find
sufficient evidence to support the claim that housekeeping,
maintenance, or constitutive genes are not expressed at a
constant level through the transcriptional cycle. As a practical
matter, the canonical-treated versus control paradigm is made
suspect when the assumption of steady-state conditions is shown
to be incorrect. Given that the patterns of expression shown here
hold through multiple cycles even when transcripts showing very
low-amplitude expression patterns are considered, the question
is raised whether these oscillatory patterns, unacknowledged in
most expression array experiments, represent an uncontrolled
variable and contribute to the biological variability and hence the
apparent 2-fold limitations to sensitivity.
Fig. 6. Flow cytometric analysis of S-phase gating in the reductive phase of
the cycle. The 1D frequency histogram (A) showing DNA content of the
population for 39 samples taken at 4-min intervals have been stacked (B)in
relationship to the time in the DO oscillation at which they were sampled (C)
and color-mapped to show numbers of cells in early, middle, and late S phase
(B). In the resulting 2D color map, with red indicating more cells and blue
representing fewer cells, the track of cells through S phase appears as a series
of green bands moving left to right and upward on the diagonal. Red ⫽ 350
cells, light green ⫽ 200 cells, and dark blue ⫽ 0 cells.
1204
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0306490101 Klevecz et al.
Continuous culture techniques allow the examination of
genomewide expression in a manner that avoids the pitfalls of
conventional synchrony that is, synchronization-associated per-
turbations to the system, lack of cycle-to-cycle reproducibility,
and rapid decay of blockage-induced synchrony, while permit-
ting sampling to be as frequent and enduring as required. In
other words, we can sample for expression array analysis from a
biologically stable platform as frequently as we can afford to
sample, and we can do so under conditions where the benchmark
oscillation in respiration guides the sampling regimen. Using this
system it should be possible to better define and eliminate
sources of variability in expression arrays.
The most interesting realization to come out of these studies
is that the separation in time between oxidative and reductive
phases propagates throughout the transcriptome and is coordi-
nated with the initiation of DNA replication. This may be an
important strategy evolved in cells to prevent oxidative damage
to DNA during replication. Because transcription is mirrored in
the oscillations in oxidative and reductive genes, there is the
implication that protein synthesis and catabolism must be sim-
ilarly organized in time. Given the totality of this coordinate
expression of such central and yet disparate elements as DNA
replication, the machinery for protein synthesis and degradation,
and mitochondrial ribosomes, it seems unavoidable to expect
that similar temporal organization will be found in mammalian
and other systems.
We thank the Analytical Cytometry Core facility for help in flow
cytometric analysis and J. Denis Heck and the University of California-
Irvine DNA MicroArray Facility for help with the expression array
studies. This work is supported by a Beckman Research Institute grant.
1. Klevecz, R. R. & Ruddle, F. H. (1968) Science 159, 634–636.
2. Kauffman, S. A. & Wille, J. J. (1975) J. Theor. Biol. 55, 47–93.
3. Mackey, M. C. & Glass, L. (1977) Science 197, 287–289.
4. Waddington, C. D. (1957) The Strategy of the Gene (Allen and Unwin, London)
5. Eigen, M. & Schuster, P. (1977) Naturwissenschaften 64, 541–565.
6. Kauffman, S. A. (1993) The Origins of Order (Oxford Univ. Press, Cambridge,
U.K.)
7. Wodicka, L., Dong, H., Mittman, M., Ho, M. H. & Lockhart, D. (1997) Nat.
Biotechnol. 15, 1359–1367.
8. Cho, R. J., Campbell, M. J., Winzeler, E. A., Steinmetz, L., Conway, A.,
Wodicka, L., Wolfsberg, T. G, Gabrielian, A. E., Landsman, D., Lockhart, D. J.
& Davis, R. W. (1998) Mol. Cell. 2, 65–73.
9. Spellman, P. T., Sherlock, G., Zhang, M. Q., Iyer, V. R., Anders, K., Eisen,
M. B., Brown, P. O., Botstein, D. & Futcher, B. (1998) Mol. Biol. Cell. 9,
3273–3297.
10. Alter, O., Brown, P. O. & Botstein, D. (2000) Proc. Natl. Acad. Sci. USA 97,
10101–10106.
11. Klevecz, R. R. (2000) Funct. Integr. Genomics 1, 186–192.
12. Klevecz, R. R. (1976) Proc. Natl. Acad. Sci. USA 73, 4012–4016.
13. Chisholm, S. W. & Costello, J. H. (1980) J. Phycol. 16, 375–383.
14. Homma, K. & Hastings, J. W. (1988) J. Biol. Rhythms 3, 49–58.
15. Haase, S. B. & Reed S. I. (1999) Nature 401, 394–397.
16. Sveiczer, A., Csikasz-Nagy, A., Gyorffy, B., Tyson, J. J. & Novak, B. (2000)
Proc. Natl. Acad. Sci. USA 97, 7865–7870.
17. Rifkin, S. A. & Kim, J. (2002) Bioinformatics 18, 1176–1183.
18. Klevecz, R. R., Kauffman, S. A. & Shymko, R. M. (1984) Int. Rev. Cytol. 86,
97–128.
19. Klevecz, R. R. (1998) Physica D 124, 1–10.
20. Klevecz, R. R., Bolen, J. L. & Dura´n, O. (1992) Int. J. Bifurcation Chaos 2,
941–953.
21. Bolen, J. L., Dura´n, O. & Klevecz, R. R. (1993) Physica D 67, 245–256.
22. Satroutdinov, A. D., Kuriyama, H. & Kobayashi, H. (1992) FEMS Microbiol.
Lett. 77, 261–267.
23. Lloyd, D., Salgado, L. E., Turner, M. P., Suller, M. T. & Murray, D. B. (2002)
Microbiology 148, 3715–3724.
24. Murray, D. B., Engelen, F. A. A., Lloyd, D. & Kuriyama, H. (1999) Microbiology
145, 2739–2745.
25. Sohn, H. Y., Murray, D. B. & Kuriyama, H. (2000) Yeast 16, 1185–1190.
26. Murray, D. B., Klevecz, R. R. & Lloyd, D. (2003) Exp. Cell Res. 287, 10–15.
27. Brown, A. J. P. (1996) in Yeast Protocols, ed. Evans, I. H. (Humana, Totowa,
NJ), pp. 269–276.
28. Kellis, M., Patterson, N., Endrizzi, M., Birren, B. & Lander, E. S. (2003) Nature
423, 241–254.
29. Burnett, W. V. (1997) BioTechniques 22, 668–671.
30. Mochan, E. & Pye, E. K. (1973) Nat. New Biol. 242, 177–179.
31. Poole, R. K., Lloyd, D. & Kemp, R. B. (1973) J. Gen. Microbiol. 77, 209.
32. Klevecz, R. R. & Murray, D. B. (2001) Mol. Biol. Rep. 28, 73–82.
33. Yue, H., Eastman, P. S., Wang, B. B., Minor, J., Doctolero, M. H., Nuttall,
R. L., Stack, R., Becker, J. W., Montgomery, J. R., Vainer, M. & Johnston, R.
(2001) Nucleic Acids Res. 29, E41.
34. Warrington, J. A., Nair, A., Mahadevappa, M. & Tsyganskaya, M. (2000)
Physiol. Genomics 2, 143–147.
35. Potten, C. S., Booth, D., Cragg, N. J., O’Shea, J. A., Tudor, G. L. & Booth, C.
(2002) Cell Prolif. 35, 16–21.
36. Smaaland, R., Abrahamsen, J. F., Svardal, A. M., Lote, K. & Ueland, P. M.
(1992) Br. J. Cancer 66, 39–45.
37. Georgiou, G. (2002) Cell 111, 607–610.
38. Koerkamp, M. G., Rep, M., Bussemaker, H. J., Hardy, G. P., Mul, A.,
Piekarska, K., Szigyarto, C. A., De Mattos, J. M. & Tabak, H. F. (2002) Mol.
Biol. Cell. 13, 2783–2794.
39. Adams, C. A., Kuriyama, H., Lloyd, D. & Murray D. B. (2003) Yeast 20,
463–470.
40. Wheals, A. E., Tyson, C. B. & Lord, P. G. (1979) J. Bacteriol. 138, 92–98.
41. Alter, O. Brown, P. O. & Botstein, D. (2003) Proc. Natl. Acad. Sci. USA 100,
3351–3356.
Klevecz et al. PNAS
兩
February 3, 2004
兩
vol. 101
兩
no. 5
兩
1205
CELL BIOLOGY SEE COMMENTARY