Grain feeding and the dissemination of acid-resistant Escherichia coli from cattle.
ABSTRACT The gastric stomach of humans is a barrier to food-borne pathogens, but Escherichia coli can survive at pH 2.0 if it is grown under mildly acidic conditions. Cattle are a natural reservoir for pathogenic E. coli, and cattle fed mostly grain had lower colonic pH and more acid-resistant E. coli than cattle fed only hay. On the basis of numbers and survival after acid shock, cattle that were fed grain had 10(6)-fold more acid-resistant E. coli than cattle fed hay, but a brief period of hay feeding decreased the acid-resistant count substantially.
- SourceAvailable from: puyallup.wsu.edu[Show abstract] [Hide abstract]
ABSTRACT: Would feeding hay to cows instead of grain solve or reduce the E. coli O157:H7 problem? The NY Times recently ran an editorial ("Leafy Green Sewage" By Nina Planck, September 21, 2006) that included the statement: "It's [E. coli O157:H7] not found in the intestinal tracts of cattle raised on their natural diet of grass, hay and other fibrous forage. No, O157 thrives in a new — that is, recent in the history of animal diets — biological niche: the unnaturally acidic stomachs of beef and dairy cattle fed on grain, the typical ration on most industrial farms." A number of local news outlets repeated this claim without discussion of the scientific evidence for or against. The goal of this paper is to provide a summary of the evidence. The hypothesized impact of hay feeding on E. coli O157:H7 in bovines seems to have originated in the late 1990s with a research group at Cornell who published an article in Science (Diez-Gonzalez et al, 1998) that has subsequently been widely cited. The senior author wrote several additional papers arguing his hypothesis (Russell et al, 2000a; Russell et al, 2000b), but the key data cited as a test of the hay-feeding hypothesis in cattle appeared to be from the 1998 Science paper. In that study, a total of three (n=3) cows were rotated through different diets. No evidence is presented that any of the experimental cows were colonized with E. coli O157:H7 at any point during the study. Therefore all inferences made by the authors were based on changes in the naturally-occurring mixed populations of E. coli of the normal colonic flora, with the assumption that had E. coli O157:H7 been there it would have behaved similarly as the rest of the mixed populations of E. coli. Some have criticized this assumption on the grounds that E. coli O157:H7 are known to differ in important ways from most of the other E. coli flora—more on this later. It is also important to note that, in a mixed generic population, it is difficult to distinguish between an induced change (i.e., strains becoming more acid resistant—as the authors hypothesized) and simple trait selection (those naturally more acid resistant being more likely to survive). The authors hypothesized that E. coli O157:H7 with induced acid resistance (were it to occur in cows on grain diets) would more easily make it through the gastric barrier of the human stomach and therefore be more likely to cause disease. Only laboratory data were presented to support the possibility that increased acid resistance can be induced in some strains of E. coli O157:H7. In spite of these weaknesses, the finding in the 1998 Science paper of a 1000-fold decline in total E. coli and a 1-million-fold reduction in acid resistant E. coli among their three cows seemed to be promising pilot data in support of their hypothesis. Unfortunately, their findings --and especially their far reaching conclusions regarding the key significance of grain-based feeds to cattle--have not been corroborated by numerous scientific papers from research groups around the world.
- [Show abstract] [Hide abstract]
ABSTRACT: Despite playing a central role in the regulation of ecosystem services, microorganisms are often neglected when evaluating feedbacks between social and ecological systems. A social–ecological framework is a tool for evaluating how social factors affect ecosystems through human actions and how ecological factors in turn affect social systems through ecosystem services. Here, we consider linkages and trade-offs between social and biophysical factors that arise when unique microbial attributes such as complexity, dispersal, and rapid evolution are integrated into a social–ecological framework. Using case studies from food production systems, wastewater treatment facilities, and synthetic biology, we show that unintended dis-services can arise when microbial information is limited or is ignored as a result of socioeconomic policies and practices. In contrast, when knowledge about microorganisms is integrated into a social–ecological framework, we can identify how to best maximize microbial services. New scientific tools used to characterize microbial traits, communities, and functions will enhance our ability to monitor microorganisms in diverse systems. However, communication and collaboration among stakeholders – including policy makers, landowners, resource managers, and scientists – are also needed to foster more effective “micromanagement” of microbial services.Frontiers in Ecology and the Environment 11/2014; 12(9):524-531. · 8.41 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Özet Tüketiciler konvansiyonel olarak üretilen hayvansal ürünlere göre organik olarak üretilenlerin zararlı madde içermedikleri ve böylece daha sağlıklı ve kalitesinin daha iyi olduğuna inanmaktadırlar. Bu durum, muhtemelen organik üretimde hayvanları uygun yetiştirme sistemlerinde büyütülmesi, doğal yemlere serbestçe ulaşabilmesine, sentetik yem katkılarının ve rasyonlarda genetiği değiştirilmiş yem maddelerinin yasaklanması ve dolayısıyla daha sağlıklı ürün elde edileceği düşüncesine dayanmaktadır. Kimi araştırıcılar bu görüşleri doğrulayan düşünceler bildirirken kimileri ise organik ve konvansiyonel üretim metotları arasında önemli kalite farklılıklarının olmadığına ve kaliteye ilişkin beklentinin karşılanmasının zorluğuna dikkat çekmektedirler. Bu çalışma, organik ve konvansiyonel metotlarla elde edilen hayvansal ürünlerin kimyasal içerik, duyusal analizler, mikroorganizma yükü, kimyasal kalıntı ve ağır metaller bakımından karşılaştırılması amacıyla yapılmıştır. Anahtar kelimeler: Organik, konvansiyonel, hayvansal üretim Comparison of some animal product properties obtained from organic and conventional animals Abstract The consumers believes that the organic animal products not contain harmful substances and therefore they have good quality and healthier than conventional products. In this case, probably, organically produced animals grown under suitable breeding system, availability of natural feed sources freely, prohibition of supplementation of synthetic feed additives and genetic modified organism in the animal feeds and therefore healthier products can be obtained. Some of the researches have verified think, however some of them call attention to there are difficulties about meet of expectation in quality and there are no differences in quality properties of organic and conventional production system. This study carried to compare animal products provided in organic and conventional production system in terms of chemical content, sensual analyses, microorganism load, chemical residue and heavy metals. Keywords: organic, conventional, animal production1. Organik Tarım Kongresi, GÜMÜŞHANE; 08/2010
, 1666 (1998);
et al.Francisco Diez-Gonzalez,
from CattleEscherichia coliAcid-Resistant
Grain Feeding and the Dissemination of
www.sciencemag.org (this information is current as of February 4, 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,
, 7 of which can be accessed for free:
cites 12 articles
148 article(s) on the ISI Web of Science.
This article has been
38 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 aScience1998 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 theScience
on February 4, 2009
Grain Feeding and the
Dissemination of Acid-Resistant
Escherichia coli from Cattle
Francisco Diez-Gonzalez, Todd R. Callaway, Menas G. Kizoulis,
James B. Russell*
The gastric stomach of humans is a barrier to food-borne pathogens, but
Cattle are a natural reservoir for pathogenic E. coli, and cattle fed mostly grain
had lower colonic pH and more acid-resistant E. coli than cattle fed only hay.
On the basis of numbers and survival after acid shock, cattle that were fed grain
had 106-fold more acid-resistant E. coli than cattle fed hay, but a brief period
of hay feeding decreased the acid-resistant count substantially.
Foods can be cooked or irradiated to kill
bacteria, but there are ?30 million food-
borne illnesses each year in the United States
(1). A variety of hypotheses have been for-
mulated to explain the increased incidence of
food-borne illness (2, 3). Modern societies
tend to “dine out” more often and consume
more processed food. Modern detection
methods for pathogenic bacteria are more
sensitive, and this sensitivity has heightened
our awareness of the problem (4). Some ex-
perts have suspected a more rapid evolution
of bacterial virulence factors, but this evolu-
tion is poorly understood (5).
Although Escherichia coli is a normal
inhabitant of the gastrointestinal tract, some
strains (for example, O157:H7) produce tox-
ins and are pathogenic (6). Hamburger has
frequently been contaminated with pathogen-
ic E. coli, and vegetables and fruit juices have
also been sources of infection (4). Cattle, a
natural reservoir for pathogenic strains, have
often been implicated in E. coli infection (7).
It is virtually impossible to prevent all fecal
contamination of meat at slaughter, and veg-
etables are sometimes fertilized with cattle
manure. The ability of E. coli to cause food-
borne illness is enhanced by its low infective
dose, and as few as 10 cells can cause infec-
The ability of bacteria to act as food-borne
pathogens depends on their capacity to sur-
vive the low pH of the gastric stomach and to
colonize the intestinal tract of humans (9), but
the role of acid resistance in the dissemina-
tion of pathogenic bacteria has often been
ignored. Pathogenic and nonpathogenic E.
coli cultures develop extreme acid resistance
only when they are grown at mildly acidic
pH. If E. coli is grown at neutral pH, it is acid
sensitive and killed by the low pH of gastric
Since the Second World War, fattening
beef cattle in the United States have been fed
large amounts of grain (starch) and very little
hay (11), but the impact of grain feeding on
acid-resistant E. coli had not been examined.
Many forms of starch pass through the pre-
gastric stomach (rumen) to the intestines
(11), and cattle are deficient in the starch-
degrading enzyme, amylase (12). Starch can
be fermented in the colon, and E. coli fer-
ments maltose, an extracellular degradation
product of starch (13). Starch fermentation in
the colon produces volatile fatty acids (ace-
tate, butyrate, and propionate) that decrease
To determine the potential impact of grain
feeding on E. coli in cattle, we removed colonic
digesta from the rectums of cattle that were fed
hay, grass, and varying amounts of rolled corn.
Digesta were diluted 10-fold with sterile anaer-
obic water and mixed vigorously with a vortex
mixer for 1 min. The pH was measured with a
combination electrode. Coliforms were enu-
merated by visually monitoring turbidity after
serial dilution in lauryl sulfate broth (14). Esch-
erichia coli was determined by screening the
bacteria on the basis of lactose fermentation,
gas production, indole production, the methyl
red reaction, Voges-Proskauer test, and citrate
fermentation (14). Acid shock was performed
by diluting digesta samples 100-fold into Luria
broth that had been adjusted to pH 2.0 (14).
After 1.0 hour at pH 2.0, viable cell numbers
were determined by serially diluting into lauryl
A survey of 61 cattle indicated that grain
supplementation could increase total and ac-
id-resistant E. coli numbers (Table 1). Cattle
fed either hay or fresh grass (pasture) had a
colonic pH greater than 7.0, the total E. coli
count was only 20,000 cells per gram, and
virtually all of these bacteria were killed by
an acid shock that mimicked the pH of gastric
juice. Moderate amounts of grain (60% of dry
matter) did not cause a statistically significant
decrease in pH (P ? 0.05), but the total E.
coli population was 6.3 ? 106viable cells per
gram of digesta. Some of the E. coli were
Division of Biological Sciences, Section of Microbiol-
ogy, Cornell University and Agricultural Research Ser-
vice, U.S. Department of Agriculture, Ithaca, NY
*To whom correspondence should be addressed. E-
Fig. 1. The effect of grain feeding (percentage
of diet dry matter) on (A) the volatile fatty acid
concentration of the rumen and colon and (B)
pH. The error bars indicate standard deviations
of the mean (three animals, four sampling
Table 1. The effect of grain feeding on the colonic pH and E. coli counts of cattle fed various amounts
Total E. coli
60% rolled corn
?80% rolled corn
7.2 ? 0.1†
7.1 ? 0.1†
4.3 ? 0.5†
5.0 ? 0.9†
316.9 ? 0.3†
6.8 ? 0.7‡
4.4 ? 1.1‡
16 5.9 ? 0.6‡
6.9 ? 0.9‡
5.4 ? 0.7§
* Acid-resistant E. coli are those that survived an acid shock (pH 2.0, Luria broth, 1 hour).
with different superscripts are significantly different (P ? 0.05, Student’s t test).
†‡§ Means within a column
R E P O R T S
11 SEPTEMBER 1998VOL 281SCIENCE www.sciencemag.org
on February 4, 2009
killed by acid shock, but the acid-resistant
count was greater than 25,000 viable cells per
gram. When animals were fed more than 80%
grain, the pH was significantly lower (P ?
0.05), and the acid-resistant E. coli count was
250,000 viable cells per gram.
To define more precisely the role of grain
in promoting the growth of acid-resistant E.
coli, we performed highly controlled experi-
ments. Mature, nonlactating Holstein cows
were surgically modified so that ruminal con-
tents could be removed directly from the
rumen (IACUC protocol 95-1-97). Cattle of
similar size (600 kg) were fed every 2 hours
with a rotary feeder (10 kg of dry matter per
day). The feeds used were medium-quality
timothy hay (14% crude protein, 40% neutral
detergent fiber) and a grain mixture [89%
rolled (cracked) corn and 11% soybean
meal]. The diets were 0, 45, and 90% grain
with the remainder being hay.
Samples of digesta were obtained from
the rumen as well as the colon. Ruminal
contents were squeezed through cheesecloth
and purged with oxygen-free carbon dioxide.
Colonic samples were processed as described
above. Samples were centrifuged at 13,000g
for 10 min to remove bacteria and feed par-
ticles, and fermentation acids were analyzed
by high-pressure liquid chromatography. The
total count of anaerobic bacteria was deter-
mined by serially diluting the digesta in a
nonselective medium designed for strictly an-
aerobic bacteria (15). Samples of colonic di-
gesta were processed as described above. E.
coli strains arising from isolated colonies
were obtained from MacConkey’s plates sup-
plemented with sorbitol as an energy source.
E. coli strains isolated from cattle and E. coli
O157:H7 were given an even longer acid
shock (6 hours), and in this case the recovery
medium was Luria broth.
The randomized block design was a 3 ? 3
Latin square (three animals ? three diets)
with 14 days of adaptation and 4 days of
sample collection (total of 54 days). Because
the animals were mature, and the environ-
ment of the barn was carefully controlled, the
effect of time was judged to be inconsequen-
tial. The data were first analyzed by two-way
analysis of variance (diet versus animal), and
the F values indicated that P ? 0.05 in all
cases. Student-Newman-Keuls test (16) was
used to estimate differences among means,
and the variance estimates were pooled (17).
When cattle were fed increasing amounts of
grain, the volatile fatty acid (acetic, propionic,
and butyric) concentration of the rumen did not
increase significantly (P ? 0.05), but the con-
centration in the colon increased ?fourfold
(P ? 0.05) (Fig. 1A). Under these conditions,
ruminal pH remained essentially constant (P ?
0.05), but the pH of the colon decreased (P ?
0.05) when the volatile fatty acids accumulated
(Fig. 1B). Lactic and succinic acids were never
detected in rumen samples, but small amounts
of both acids were observed in colon samples
when 90% grain was fed. Grain supplementa-
tion had little effect on the numbers of anaero-
bic bacteria in the rumen, but the colon count
increased 1000-fold (Fig. 2A). Hay-fed cattle
had less than 105colonic coliforms, but those
fed 90% grain had ?108coliforms per gram of
digesta (Fig. 2B). Only a small fraction of the
ruminal coliforms were E. coli, but virtually all
of the colonic coliforms were identified as E.
coli (Fig. 2C).
Hay-fed cattle had a low concentration of
volatile fatty acids in their colons (Fig. 1B), and
acid shock killed more than 99.99% of the E.
coli (Fig. 3A). When diets were supplemented
with either 45 or 90% grain, acids accumulated,
colonic pH declined (Fig. 1B), and a much
larger percentage of the E. coli survived acid
shock (Fig. 3B). The idea that grain, by pro-
moting acid production in the colon, was regu-
Fig. 2. The effect of grain feeding (percentage
of diet dry matter) on (A) total anaerobes, (B)
coliforms, and (C) coliforms that were identi-
fied as E. coli. The error bars indicate standard
deviations of the mean (three animals, four
Fig. 3. (A) The survival of E. coli in colonic fluid
(percentage of initial count) after acid shock
(pH 2.0, Luria broth, 1 hour), and (B) the effect
of glucose and final pH on the survival of
colonic E. coli isolates (hay versus grain) and E.
coli O157:H7 after acid shock (pH 2.0, Luria
broth 6 hours). When the cultures were culti-
vated overnight in broth containing large
amounts of glucose (10 mg solids/ml) (filled
bars), the final pH was 4.8. Cultures with small
amounts of glucose (0.5 mg/ml) (open col-
umns) produced less acid and the final pH was
6.8. The error bars indicate standard deviations
of the mean (10 strains, two replicates per
Fig. 4. The effect of hay on the total numbers
of colonic E. coli in cattle that had been
consuming the 90% grain diet. (A) Cattle
were switched from 90% grain to hay on day
zero. (B) The numbers of E. coli that were
able to survival acid shock (pH 2.0, Luria
broth, 1 hour). The bars indicate standard
deviations of the mean (three animals, one
replicate per animal, two independent exper-
iments). The dotted lines show the detection
limit of our enumerations.
R E P O R T S
www.sciencemag.orgSCIENCEVOL 281 11 SEPTEMBER 1998
on February 4, 2009
lating acid resistance in vivo, was corroborated
by in vitro experiments. When E. coli strains
isolated from the cattle were grown in the lab-
oratory with a high concentration of glucose,
acetic acid accumulated in the medium, pH
declined, and the cell survival after acid shock
was high (Fig. 3B). If the glucose concentration
of the medium was low, little acid was pro-
duced, and cell survival was extremely low.
Strains isolated from cattle fed forage or grain,
and E. coli O157:H7 (ATCC 43895, CDC EDL
that grain feeding was inducing acid resistance
rather than selecting a different population of E.
About 5% of our E. coli isolates (n ? 155)
were sorbitol negative, a diagnostic trait of
O157:H7 (14), but none of these strains tested
positive for O157:H7 antigens (18). The ab-
sence of E. coli O157:H7 in our cattle is not
surprising. Previous workers have noted that
nonpathogenic E. coli can often outgrow patho-
genic strains, and this point is illustrated by at
least three observations: (i) The percentage of
O157:H7-positive animals in herds directly
linked to outbreaks was less than 2% (19); (ii)
even cattle experimentally inoculated with E.
coli O157:H7 did not shed the bacterium for
long periods of time (20); and (iii) E. coli
O157:H7 numbers can be reduced by giving
animals doses of nonpathogenic E. coli (21).
The finding that grain feeding increased
both the number and acid resistance of E. coli
in cattle could have significant implications
for food safety. Although not all E. coli are
pathogenic, there is always the risk that at
least some cattle will harbor pathogenic
strains. Acid resistance appears to be a factor
in the dissemination (transmission) of E. coli
from cattle to humans. Therefore, it is rea-
sonable to suggest that the induction of acid
resistance could increase the risk of food-
borne illness. Our studies indicated that the
time needed to decrease E. coli numbers was
relatively short (Fig. 4A). Cattle adapted to a
90% grain diet had an acid-resistant E. coli
count greater than 106viable cells per gram.
After change to a hay diet, the viable cell
number immediately declined, and after 5
days the E. coli population was nearly 106-
fold lower (Fig. 4B).
Grain feeding is a practice that promotes
the production and efficiency of cattle pro-
duction, and it is unlikely that American cat-
tle will ever be fed diets consisting only of
hay. However, our studies indicate that cattle
could be given hay for a brief period imme-
diately before slaughter to significantly re-
duce the risk of food-borne E. coli infection.
1. Foodborne Pathogens: Risks and Consequences (Cen-
ter for Agricultural Science and Technology, Task
Force Report number 122, CAST, Ames, IA, 1994).
2. R. V. Tauxe, Emerging Infect. Dis. 3, 425 (1997).
3. J. E. Collins, ibid, p. 471.
4. R. L. Buchanan and M. P. Doyle, Food Technol. 51, 69
5. J. Lederberg, Emerging Infect. Dis. 3, 417 (1997).
6. C. Su and L. J. Brandt, Ann. Intern. Med. 123, 698
7. S. C. Whipp, M. A. Rasmussen, W. C. Cray, J. Am. Vet.
Med. Assoc. 204, 1168 (1994).
8. L. W. Riley et al., N. Engl. J. Med. 308, 681 (1983).
9. J. Gorden and P. L. C. Small, Infect. Immun. 61, 364
10. M. M. Benjamin and A. R. Datta, Appl. Environ. Mi-
crobiol. 61, 1669 (1995).
11. D. R. Waldo, J. Anim. Sci. 37, 1062 (1973).
12. J. R. Russell, A. W. Youngand, N. A. Jorgensen, ibid. 52,
13. E. C. C. Lin, in Escherichia coli and Salmonella Cellular
and Molecular Biology, F. C. Neidhardt et al., Eds.
(American Society for Microbiology, Washington, DC,
ed. 3, 1996), vol. 1, chap. 20.
14. A. D. Hitchins, P. Feng, W. D. Watkins, S. R. Rippey,
L. A. Chandler, in Food Drug Administration Bacterio-
logical Analytical Manual (Association of Official An-
alytical Chemists International, Gaithersburg, MD, ed.
8, 1995), chap. 4.
15. D. R. Caldwell and M. P. Bryant, Appl. Microbiol. 14,
16. R. R. Sokal and F. J. Rohlf, Biometry (Freeman, New
17. R. L. Ott, An Introduction to Statistical Methods and
Data Analysis (Wadsworth, Belmont, CA, 1993).
18. RIM E. coli O157:H7 Latex Test (Remel, Lenexa, KS).
19. G. L. Armstrong, J. Hollingsworth, J. G. Morris Jr.,
Epidemiol. Rev. 18, 29 (1996).
20. W. C. Cray and H. W. Moon, Appl. Environ. Microbiol.
61, 1586 (1995).
21. T. Zhao et al., J. Clin. Microbiol. 36, 641 (1998).
27 March 1998; accepted 3 August 1998
MP1: A MEK Binding Partner
That Enhances Enzymatic
Activation of the MAP Kinase
Hans J. Schaeffer,* Andrew D. Catling,* Scott T. Eblen,
Lara S. Collier,† Anke Krauss,‡ Michael J. Weber§
Signal transduction is controlled both by regulation of enzyme activation and
and anchor proteins. The extracellular signal–regulated kinase (ERK) cascade is
cascades important in the regulation of growth, apoptosis, and differentiation.
A two-hybrid screen was conducted to identify nonenzymatic components of
this signaling cascade that might be important in regulating its activity. A
protein called MP1 (MEK Partner 1) was identified that bound specifically to
cells, MP1 enhanced activation of ERK1 and activation of a reporter driven by
the transcription factor Elk-1. Expression of MP1 in cells increased binding of
ERK1 to MEK1. MP1 apparently functions as an adapter to enhance the effi-
ciency of the MAP kinase cascade.
The MAP kinases ERK1 and ERK2 are com-
ponents of a protein kinase cascade display-
ing evolutionary conservation of protein se-
quence and a three-kinase architecture (1).
ERKs are activated by the MAP kinase ki-
nases MEK1 or MEK2 . MEKs, in turn, are
activated by members of the Raf family. In a
related MAP kinase pathway in Saccharomy-
ces cerevisiae, the pheromone response path-
way, the specificity and efficiency of the
enzymatic components are facilitated by a
nonenzymatic “scaffolding protein,” STE5,
that directly interacts with the signaling en-
zymes of this pathway (2).
Although a scaffold-like component has
not previously been reported for MAP kinase
pathways in higher eukaryotic systems, our
earlier work indicated that efficient signaling
through the Raf-MEK-ERK pathway also ap-
pears to require an additional, unknown com-
ponent (3). MEK1 has a proline-rich se-
quence (PRS) that spans residues 270 to 307,
between kinase subdomains IX and X (4). A
PRS is present in MEK1 and MEK2, but not
in other members of the MAP kinase kinase
family. Downstream signaling by MEK re-
quires the PRS; MEK can be activated by
mutation to display high enzymatic activity
and to transform Rat1 cells, whereas activat-
ed MEK mutants that lack the PRS retain
Department of Microbiology and Cancer Center, Uni-
versity of Virginia Health Sciences Center, Charlottes-
ville, VA 22908, USA.
*These authors contributed equally to this report.
†Present address: Department of Developmental Bi-
ology, Stanford University School of Medicine, Stan-
ford, CA 94305, USA.
‡Present address: European Molecular Biology Labo-
ratory, Developmental Biology, 69012 Heidelberg,
§To whom correspondence should be addressed. E-
R E P O R T S
11 SEPTEMBER 1998 VOL 281SCIENCEwww.sciencemag.org
on February 4, 2009