, 1565 (2006);
et al.Hal Alper,
Improved Ethanol Tolerance and Production
Engineering Yeast Transcription Machinery for
www.sciencemag.org (this information is current as of November 22, 2009 ):
The following resources related to this article are available online at
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
can be found at:
Supporting Online Material
, 9 of which can be accessed for free:
cites 25 articles
59 article(s) on the ISI Web of Science.
This article has been
11 articles hosted by HighWire Press; see:
This article has been
This article appears in the following
in whole or in part can be found at:
permission to reproduce
of this article or about obtaining
Information about obtaining
registered trademark of AAAS.
is aScience2006 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the Science
on November 22, 2009
Engineering Yeast Transcription
Machinery for Improved Ethanol
Tolerance and Production
Hal Alper,1,3Joel Moxley,1Elke Nevoigt,1,2Gerald R. Fink,3Gregory Stephanopoulos1*
Global transcription machinery engineering (gTME) is an approach for reprogramming gene
transcription to elicit cellular phenotypes important for technological applications. Here we show
the application of gTME to Saccharomyces cerevisiae for improved glucose/ethanol tolerance, a
key trait for many biofuels programs. Mutagenesis of the transcription factor Spt15p and selection
led to dominant mutations that conferred increased tolerance and more efficient glucose
conversion to ethanol. The desired phenotype results from the combined effect of three separate
mutations in the SPT15 gene [serine substituted for phenylalanine (Phe177Ser) and, similarly,
Tyr195His, and Lys218Arg]. Thus, gTME can provide a route to complex phenotypes that are not
readily accessible by traditional methods.
olism. The evolution of such complex traits
requires simultaneous modification in the ex-
pression levels of many genes, which may not
be achievable by sequential multigene modifi-
cations. Furthermore, the identification of genes
requiring perturbation may be largely unan-
ticipated by conventional pathway analysis.
(PCR) mutations] key proteins regulating the
global transcriptome and generates, through
them, a new type of diversity at the transcrip-
This approach has already been demon-
strated by engineering sigma factors in prokary-
otic cells (1), but the increased complexity of
eukaryotic transcription machinery raises the
question of whether gTME can be used to im-
prove traits in more complex organisms. For
example, eukaryotic systems have more
specialization—three RNA polymerase en-
zymes with separate functions, whereas only
one exists in prokaryotes. Moreover, nearly 75
scription factors or coactivators of the RNA
polymerase II (RNA Pol II) system (2), and loss
of function for many of these components is
lethal. Components of the general factor RNA
Pol II transcription factor D (TFIID) include the
he production of desirable compounds
from microbes can often require a com-
TATA-binding protein (SPT15) and 14 other
associated factors (TAFs) that are collectively
thought to be the main DNA binding proteins
regulating promoter specificity in yeast (2–5).
Mutations in a TATA-binding protein have been
shown to change the preference of the three
polymerases and to play an important role in
promoter specificity (6).
Successful fermentations to produce ethanol
using yeast require tolerance to high concen-
trations of both glucose and ethanol. These cel-
lular characteristics are important because very
high gravity (VHG) fermentations, which are
common in the ethanol industry, give rise to
high sugar concentrations (and thus high
osmotic pressure), at the beginning of the
process, and high ethanol concentration at the
end of a batch (7, 8). As with ethanol tolerance
in Escherichia coli, tolerance to ethanol and
glucose mixtures does not seem to be a mono-
genic trait (9). Therefore, traditional methods of
strain improvement have had limited success
beyond the identification of medium supple-
mentations and various chemical protectants
To evaluate the approach of gTME in a
eukaryotic system, two gTME mutant libraries
were created from either SPT15 (which encodes
the TATA-binding protein) or one of the TATA-
binding protein–associated factors, in this case,
TAF25 (15). The yeast screening and selection
was performed in the background of the
standard haploid Saccharomyces cerevisiae
strain BY4741, which contains the endogenous,
unmutated chromosomal copy of SPT15 and
TAF25. As such, thisgenetic screen uses a strain
that expresses both the wild-type and mutated
versions of the protein and, thus, permits the
identification of dominant mutations that lead to
novel functions in the presence of the unaltered
chromosomal gene. These libraries were trans-
formed into yeast and were selected in the pres-
ence of elevated levels of ethanol and glucose.
in the presence of 5% ethanol and 100 g/liter of
glucose, so the stress was increased in the sub-
sequent serial subculturing to 6% ethanol and
120 g/liter of glucose. After the subculturing,
strains were isolated from plates, and plasmids
containing mutant genes were isolated and re-
transformed into a fresh background, then
tested for their capacity to grow in the presence
of elevated glucose and ethanol levels. The best
mutant obtained from each of these two
libraries was assayed in further detail and
The sequence characteristics of these altered
genes conferring the bestproperties (one Spt15p
and one Taf25p) are shown in Fig. 1A. Each of
these mutated genes contained three mutations,
with those of spt15 localized to the second re-
peat element, which consists of a set of b sheets
(5, 16). These specific triple mutations in the
as the taf25-300 and spt15-300 mutations.
The spt15-300 mutant outperformed the
control at all concentrations tested, with the
strain harboring the mutant protein providing up
to 13-fold improvement in growth yield at some
glucose concentrations (Fig. 1B and fig. S1).
The taf25-300 mutant was unable to grow in the
presence of 6% ethanol, consistent with the
observations seen during the enrichment and
selection phase. Despite these increases in
tolerance, the basal growth rate of these mutants
in the absence of ethanol and glucose stress was
similar to that of the control. Furthermore, the
differences in behavior between the spt15-300
mutant and taf25-300 mutant suggest that mu-
tations in genes encoding different members of
theeukaryotic transcriptionmachinery are likely
to elicit different (and unanticipated) phenotypic
The remainder of this study focuses on the
spt15-300 mutant, because this triple mutation
set, in which serine is substituted for phenyl-
alanine (Phe177Ser), and similarly, Tyr195His,
and Lys218Arg (F177S, Y195H, and K218R,
respectively), provided the most desirable phe-
notype with respect to elevated ethanol and
glucose. At ethanol concentrations above 10%,
the spt15-300 mutant exhibited statistically
significantly improved cellular viability (over
the course of 30 hours of culturing) above that
of the control, even at concentrations as high as
20% ethanol by volume (Fig. 2, A and B, and
Transcriptional profiling revealed that the
mutant spt15-300 exhibited differential expres-
sion of hundreds of genes [controlled for false
discovery (17)] in the unstressed condition (0%
ethanol and 20 g/liter glucose) relative to cells
expressing the wild-type SPT15 (18). This anal-
ysis mainlyused the unstressed condition, rather
than the stressed (5% ethanol and 60 g/liter
1Department of Chemical Engineering, Massachusetts Insti-
tute of Technology, Room 56-469, Cambridge, MA 02139,
University of Technology, Seestrasse 13, D-13353 Berlin,
Germany.3Whitehead Institute for Biomedical Research, 9
Cambridge Center, Cambridge, MA 02142, USA.
*To whom correspondence should be addressed. E-mail:
2Department of Microbiology and Genetics, Berlin
VOL 3148 DECEMBER 2006
on November 22, 2009
glucose), because expression ratios were more
reliable under this condition owing to the sim-
ilarity of growth rates, which made gene ex-
pression profiles more comparable (SOM text,
the ethanol and glucose stress had a variable
effect on many of thegenes,and often,the stress
did not further affect many of the genes selected
using unstressed conditions (SOM text, part c).
Although this widespread alteration in tran-
scription is similar to that observed in E. coli
with an altered sigma factor, the majority of the
genes with altered expression are up-regulated,
(SOM text, part b, and fig. S3). The transcrip-
tional reprogramming in the spt15-300 mutant
was quite broad, yet it exhibited some enrich-
ment of certain functional groups such as oxi-
acid metabolism, and electron transport (SOM
text, part b, and fig S4). Unclassified genes or
genes with no known function were also found
with higher levels of expression. An analysis of
promoter-binding sites, as well as a search for
active gene subnetworks using the Cytoscape
(19) framework, failed to show that a particular
pathway or genetic network was predominately
responsible for the observed genetic reprogram-
To determine whether these up-regulated
genes acted individually or as an ensemble to
we examined the effect of individual gene
highly expressed genes in the mutant under
the unstressed conditions of 0% ethanol and
20 g/liter of glucose were selected along with
two additional genes (SOM text, part c, and
tables S2 and S3). The results of the loss-of-
phenotype assay are summarized in Fig. 3A.
They show that deletion of the great majority of
the overexpressed gene targets resulted in a loss
of the capacity of the mutant spt15-300 factor to
impart an increased ethanol and glucose toler-
ance. All tested knockout strains not harboring
the mutant spt15-300 showed normal tolerance
to ethanol and glucose stress, which indicated
05 1015 202530
Relative viable cell count
05 10152025 30
Relative viable cell count
Fig. 2. Cellular viability curves to evaluate the tolerance of the mutant
under ethanol stress. Viability of the spt15-300 mutant strain compared
with the control is measured as a function of time (hours) and expressed as
the relative number of colony-forming units compared with colony count at
0 hours for stationary phase cells treated and incubated in standard
medium in the presence of (A) 12.5% and (B) 15% ethanol by volume.
The spt15-300 mutation confers a significantly enhanced viability at all
concentrations tested above 10% ethanol by volume (fig. S2). Error bars
represent the standard deviation between biological replicate experiments.
Initial cell counts were ~3.5 × 106cells per ml.
Fold Improvement OD
20 40 6080100 120 140
Repeat element 1 Repeat element 1Repeat element 2Repeat element 2
Helix 2 Helix 2Helix 2’Helix 2’
Helix N Helix NHelix 1Helix 1Helix 2Helix 2Helix 3Helix 3 LNLN L1L1 L2L2
Fig. 1. Yeast gTMEmutantswithincreased tolerancetoelevatedethanoland glucoseconcentrations.(A)
aschematic of criticalfunctionalcomponentsof the respective factor (SOM text,part a). (B) Growth yields
of the clones from (A), were assayed in synthetic minimal medium containing elevated levels (6% by
volume) of ethanol and glucose after 20 hours. Under these conditions, the spt15-300 mutant far
exceeded the performance of the taf25-300 mutant. Fold improvements of growth yields are compared
with an isogenic strain that harbors a plasmid-borne, wild-type version of either SPT15 or TAF25.
8 DECEMBER 2006 VOL 314
on November 22, 2009
that, individually, these genes are insufficient to
constitute the normal tolerance to ethanol. Out of
the 14 gene targets assayed, only loss of PHM6
function did not reduce the novel phenotype.
Thus, we hypothesize that each gene encodes a
necessary component of an interconnected net-
work, although there may be some redundancy
of function (SOM text, part c).
Three genes that exhibited the greatest in-
crease in expression level in the spt15-300
mutant were investigated as overexpression
targets in the control strain in a gain-of-function
assay. PHO5, PHM6, and FMP16 were inde-
pendently and constitutively overexpressed
under the control of the TEF promoter, and
transformants were assayed for their capacity to
impart an ethanol- and glucose-tolerance phe-
notype. Overexpression of no single gene
among the consensus, top-candidate genes from
the microarray analysis produced a gain of
phenotype similar to that of the mutant spt15-
300 (Fig. 3B).
We next constructed all possible single- and
double-mutant combinations with the sites
identified in the triple mutant (15). None of
the single or double mutants came even close to
achieving a phenotype similar to that of the
isolated spt15-300 triple mutant (SOM text,
part d, and figs. S6 to S8). One could not
predict the effect of these three mutations by
a “greedy algorithm” search approach or
select these by traditional selection for mu-
tations that cause incremental improvement,
as many of these isolated mutations are in-
dependently relatively neutral in phenotype
fitness. Consequently, such a multiple mutant
is accessible only through a technique that
specifically focuses on the in vitro mutagenesis
of the SPT15 gene followed by a demanding
Genes previously documented as SPT3-
dependent in expression (20, 21) were preferen-
the microarray data, witha Bonferroni-corrected
P value of 1 × 10−12. Furthermore, 7 of the 10
most highly expressed genes in the spt15-300
mutant are SPT3-dependent genes. Genes that
are down-regulated in spt3 mutants were rela-
log2 (OD600with spt15-300 / OD600without spt15-300) at 20 h
0 20 406080 100 120140
Fold Improvement OD
Fig. 3. Gene-knockout and overexpression analysis to probe the
transcriptome-level response elicited by the mutant spt15. (A) Loss-of-
phenotype analysis was performed using 12 of the most highly expressed
genes in this mutant (log2differential gene expression given in parentheses);
two additional genes were chosen for further study (SOM text, part c). The
14genes,respectively,wastestedbycomparing theknockout straincontaining
the spt15-300 mutation on a plasmid to a strain containing the wild-type
SPT15. All gene knockouts, except PHM6, resulted in slight to full loss of
phenotype. Control mutants for all of the gene knockout targets exhibited
similar growth yields. (B) Gene overexpression studies are provided for the top
three candidate genes from the microarray (PHO5, PHM6, and FMP16) and
assayed under 6% ethanol by volume as previously assayed (see also fig. S5).
The overexpression of these genes failed to impart a tolerance phenotype.
Fig. 4. Elucidation and
validation of a mecha-
nism partially mediated
by the SPT3-SAGA com-
spt3 knockout was eval-
uated through the intro-
duction of the spt15-300
mutant and assaying in
the presence of 6% eth-
anol by volume. The in-
capacity of the mutant to
impart the phenotype
illustrates the essentiality
of SPT3 as a part of the
mechanism provided. (B)
The three mutations
(F177S, Y195H, and
K218R) are mapped on
the global transcription machinery molecular mechanism proposed by prior studies, with each of these mutation sites (22–24, 27, 28). Collectively, these
three mutations lead to a mechanism involving Spt3p.
VOL 3148 DECEMBER 2006
on November 22, 2009
tivelyup-regulatedinthespt15-300 mutant.The Download full-text
absence of negative cofactor 2 element (NC2)
repression due to the Y195H mutation (22) may
result in overrepresentation of up-regulated
genes, because part of the negative regulation
of the Spt15p can no longer take place. These
data are consistent with previous work showing
that the spt15-21 mutation [a change from Ser to
Leu or Arg at Phe177(F177L and F177R)]
suppresses an spt3 mutation as the result of an
[part of the Spt-Ada-Gcn5-acetyltransferase
(SAGA) complex] (21, 23, 24). As a further
test of the link between Spt15p and Spt3p, it
was found that an spt15-300 mutant gene was
unable to impart its ethanol- and glucose-
tolerance phenotype to an spt3 knockout strain
From the results of the site-directed muta-
genesis and mechanism depicted in Fig. 4B, it is
conceivable that perturbations to the NC2
complex would also impact the ability of the
spt15-300 mutant to function; however, a null
mutation in one of the genes in this heterodimer
is inviable, which prevents such a follow-up
experiment. Nevertheless, these results further
underscore the importance of all three muta-
tions acting in concert in order to create the
complex phenotype mediated through an
Spt3p-SAGA complex interaction. As a result,
we posit that the mode of action is primarily
a unique protein–protein–DNA interaction
(Spt15-300p mutant–Spt3p–DNA), which leads
to this transcriptional reprogramming of a large
number of genes.
The capacity of the spt15-300 mutant to use
and ferment glucose to ethanol under a variety
of conditions was assayed in simple batch
shake-flask experiments of low and high cell
density under an initial concentration of 20 or
100 g/liter of glucose (SOM text, part e, and
figs. S9 to S11). In each of these cases, the
mutant has growth characteristics superior to
those of the control with a prolonged exponen-
tial growth phase that allows for a higher, more
robust biomass production and a higher ethanol
yield. Specifically, in high–cell density fermen-
tations, with an initial optical density at 600 nm
(OD600) of 15, the mutant’s performance far
exceeds that of the control, with more rapid
utilization of glucose, improved biomass yield,
and higher volumetric ethanol productivity (2
g/liter of ethanol per hour) relative to the con-
trol strain (Table 1). In addition, sugars were
rapidly and fully used at a yield that exceeds
that of the control and approaches the theoret-
ical value when taking into account the amount
of glucose consumed for cell growth.
These results demonstrate the applicability
of gTME to alter cellular eukaryotic pheno-
types. The isolation of dominant mutations
permits the modification of vital functions for
novel tasks, whereas the unmodified allele
carries out the functions critical for viability.
An examination of further modifications of
other transcription factors through gTME could
additionally have the potential for drastically
improving ethanol fermentations and for
improving the prospects of ethanol production.
For the mutants analyzed, altered fermentation
conditions and additional pathway engineering
are likely to further increase ethanol production
(25, 26). Furthermore, the strain used in this
study is a standard laboratory yeast strain, and
this method could be explored in industrial or
isolated yeast exhibiting naturally higher
starting ethanol tolerances. Finally, we note
that the transcription factors modified in this
study have similarity to those in more complex
eukaryotic systems including those of mamma-
lian cells, which raises the possibility of using
this tool to elicit complex phenotypes of both
biotechnological and medical interest in these
systems as well.
References and Notes
1. H. Alper, thesis, Massachusetts Institute of Technology
2. S. Hahn, Nat. Struct. Mol. Biol. 11, 394 (2004).
3. M. Hampsey, Microbiol. Mol. Biol. Rev. 62, 465 (1998).
4. J. L. Kim, D. B. Nikolov, S. K. Burley, Nature 365, 520
5. D. I. Chasman, K. M. Flaherty, P. A. Sharp, R. D. Kornberg,
Proc. Natl. Acad. Sci. U.S.A. 90, 8174 (1993).
6. M. C. Schultz, R. H. Reeder, S. Hahn, Cell 69, 697 (1992).
7. R. Thatipamala, S. Rohani, G. Hill, Biotechnol. Bioeng.
40, 289 (1992).
8. F. W. Bai, L. J. Chen, Z. Zhang, W. A. Anderson,
M. Moo-Young, J. Biotechnol. 110, 287 (2004).
9. F. van Voorst, J. Houghton-Larsen, L. Jonson,
M. C. Kielland-Brandt, A. Brandt, Yeast 23, 351 (2006).
10. Z. P. Cakar, U. O. Seker, C. Tamerler, M. Sonderegger,
U. Sauer, FEMS Yeast Res 5, 569 (2005).
11. K. Furukawa, H. Kitano, H. Mizoguchi, S. Hara, J. Biosci.
Bioeng. 98, 107 (2004).
12. M. Nozawa, T. Takahashi, S. Hara, H. Mizoguchi, J. Biosci.
Bioeng. 93, 288 (2002).
13. Y. Ogawa et al., J. Biosci. Bioeng. 90, 313 (2000).
14. H. Takagi, M. Takaoka, A. Kawaguchi, Y. Kubo, Appl.
Environ. Microbiol. 71, 8656 (2005).
on Science Online.
16. J. H. Geiger, S. Hahn, S. Lee, P. B. Sigler, Science 272,
17. J. D. Storey, R. Tibshirani, Proc. Natl. Acad. Sci. U.S.A.
100, 9440 (2003).
18. For each gene, a P value for differential expression
between the two conditions was calculated by a t test. To
simultaneously test multiple hypotheses, P values were
corrected in a false-discovery rate analysis (17). False-
discovery rates are commonly used for the analysis of
large date sets (such as microarrays), which limits false-
positives, akin to a Bonferroni correction. In this case,
366 genes were found to be significantly differentially
expressed, at a false-discovery rate of 1%.
19. P. Shannon et al., Genome Res. 13, 2498 (2003).
20. K. L. Huisinga, B. F. Pugh, Mol. Cell 13, 573 (2004).
21. T. I. Lee et al., Nature 405, 701 (2000).
22. Y. Cang, D. T. Auble, G. Prelich, EMBO J. 18, 6662
23. H. Kou, J. D. Irvin, K. L. Huisinga, M. Mitra, B. F. Pugh,
Mol. Cell. Biol. 23, 3186 (2003).
24. D. M. Eisenmann, K. M. Arndt, S. L. Ricupero,
J. W. Rooney, F. Winston, Genes Dev. 6, 1319 (1992).
25. T. L. Nissen, M. C. Kielland-Brandt,
J. Nielsen, J. Villadsen, Metab. Eng. 2, 69 (2000).
26. P. Slininger, B. Dien, S. Gorsich, Z. Liu, Appl. Microbiol.
Biotechnol. 10.1007/s00253-006-0435-1 (2005).
27. M. P. Klejman, X. Zhao, F. M. van Schaik, W. Herr,
H. T. Timmers, Nucleic Acids Res. 33, 5426 (2005).
28. D. K. Lee, J. DeJong, S. Hashimoto, M. Horikoshi,
R. G. Roeder, Mol. Cell. Biol. 12, 5189 (1992).
29. G. M. O'Connor, F. Sanchez-Riera, C. L. Cooney,
Biotechnol. Bioeng. 39, 293 (1992).
30. We acknowledge financial support of this work by the
NIH grant GM035010, and Department of Energy grant
DE-FG02-94ER14487. We thank F. Winston for a thoughtful
discussion regarding experimental design. Microarray data
deposited to the Gene Expression Omnibus database under
the accession number GSE5185.
Supporting Online Material
Materials and Methods
Figs. S1 to S11
Tables S1 to S6
30 June 2006; accepted 1 November 2006
Table 1. Fermentation results to evaluate the ethanol production potential of the spt15 mutant.
Cells were cultured in biological replicate in 100 g/liter of glucose with a high inoculum of initial
cell optical density of (OD600) of 15 [~4 g DCW(dry cell weight)/liter]. Fermentation profiles for the
high–cell density fermentation are provided and illustrate the capacity of this mutant to produce
higher productivities of ethanol at the theoretical yield, surpassing the function of the control.
Biomass yield from glucose is from reported values (29). Results represent the average between
biological replicate experiments (SOM text, part e, and figs. S9 to S11). EtOH, ethanol.
Initial DCW (g/liter)
Volumetric productivity (g/liter h−1)
Specific productivity (g/DCW h−1)
Conversion yield calculated between
6 and 21 hours
True EtOH yield accounting for biomass production
(Percentage of 0.41 g/g, which
represents the theoretical maximum)
EtOH produced (g/liter)
1 g glucose
0:5 g DCW)DCW produced (g/liter)
glucose used (g/liter) ?(
8 DECEMBER 2006VOL 314
on November 22, 2009