Regulation of myocardial ketone body metabolism
by the gut microbiota during nutrient deprivation
Peter A. Crawforda,b,1, Jan R. Crowleyb, Nandakumar Sambandamb, Brian D. Mueggea, Elizabeth K. Costelloc,
Micah Hamadyd, Rob Knightc, and Jeffrey I. Gordona
aCenter for Genome Sciences andbDepartment of Medicine, Washington University School of Medicine, St. Louis, MO 63108; and Departments of
cChemistry and Biochemistry anddComputer Science, University of Colorado, Boulder, CO 80309
Edited by Kurt J. Isselbacher, Massachusetts General Hospital, Charlestown, MA, and approved May 19, 2009 (received for review March 3, 2009)
Studies in mice indicate that the gut microbiota promotes energy
harvest and storage from components of the diet when these
components are plentiful. Here we examine how the microbiota
shapes host metabolic and physiologic adaptations to periods of
nutrient deprivation. Germ-free (GF) mice and mice who had
received a gut microbiota transplant from conventionally raised
donors were compared in the fed and fasted states by using
functional genomic, biochemical, and physiologic assays. A 24-h
fast produces a marked change in gut microbial ecology. Short-
chain fatty acids generated from microbial fermentation of avail-
able glycans are maintained at higher levels compared with GF
controls. During fasting, a microbiota-dependent, Ppar?-regulated
increase in hepatic ketogenesis occurs, and myocardial metabolism
is directed to ketone body utilization. Analyses of heart rate,
hydraulic work, and output, mitochondrial morphology, number,
and respiration, plus ketone body, fatty acid, and glucose oxida-
tion in isolated perfused working hearts from GF and colonized
animals (combined with in vivo assessments of myocardial physi-
ology) revealed that the fasted GF heart is able to sustain its
performance by increasing glucose utilization, but heart weight,
measured echocardiographically or as wet mass and normalized to
tibial length or lean body weight, is significantly reduced in both
reversed in GF mice by consumption of a ketogenic diet. Together,
these results illustrate benefits provided by the gut microbiota
during periods of nutrient deprivation, and emphasize the impor-
tance of further exploring the relationship between gut microbes
and cardiovascular health.
energy homeostasis ? gnotobiotic mice ? gut–heart metabolic axis ?
host-microbial mutualism ? isolated perfused working heart
host phylogeny (1, 2). One important function of the gut
microbiota is to break down dietary polysaccharides into end
products, including short-chain fatty acids (SCFAs), that can be
absorbed by the host (3). Without the glycoside hydrolases,
polysaccharide lyases, and other components of microbial fer-
mentation pathways, calories present in various classes of glycans
would be lost to the host.
Studies of germ-free (GF) and colonized mice have revealed
a dynamic relationship between the gut microbiota and host
adiposity. Mice raised in the absence of microbes are leaner than
their conventionally raised (CONV-R), microbe-laden counter-
parts, even though GF animals consume more food (4). Obesity
caused by a null allele in the leptin gene (ob/ob), or by con-
sumption of a high-fat, high-sugar ‘‘Western’’ diet, is associated
with a shift in microbial ecology in the distal gut and an increase
in the proportional representation of genes in the gut microbi-
ome involved in processing dietary carbohydrates into SCFAs (5,
6). Transplantation of a distal gut microbiota from these obese
mice to adult GF recipients causes a greater increase in adiposity
than does a microbiota transplanted from lean donors. More-
over, switching from a Western diet to a reduced-calorie diet
ammals harbor large and diverse communities of gut
microbes, whose compositions are shaped by both diet and
stabilizes or diminishes host adiposity, reverses the changes in
gut microbial ecology and representation of genes involved in
carbohydrate metabolic pathways seen in obese hosts, and is
associated with diminished capacity of the gut microbial com-
munity to promote adiposity in GF recipients (5, 6). Studies of
GF and CONV-R normal and knockout mice have also identi-
fied host genes that mediate the effects of the microbiota on host
energy balance. These genes include Gpr41, a G protein-coupled
receptor produced by intestinal enteroendocrine cells that is
activated by SCFAs (7). Comparative metagenomic studies of
the distal gut (fecal) microbiotas of adult monozygotic and
dizygotic lean and obese twins have documented alterations in
gut microbial ecology and the composition of the microbiome
that are analogous to those seen in obese mice (8).
The fact that increased adiposity is associated with a change
in the microbiota that allows it to be more efficient at harvesting
calories seems somewhat paradoxical: What benefit would such
a relationship have for the host? The capacity of the microbiota
to promptly adjust its configuration in ways that allows it to
increase energy harvest from a diet and promote energy storage
in the host could be beneficial under conditions where there is
only intermittent access to sources of nutrients/energy. A cor-
ollary hypothesis is that when nutrients are no longer available,
the microbiota assumes another state or configuration that
provides a different form or forms of benefit to the host. The
nature of these latter postulated adjustments, and their potential
benefits, are ill-defined. Intriguingly, after withdrawal of nutri-
ents, GF mice die more rapidly than their CONV-R counter-
parts, despite losing weight at approximately the same rate (9).
Because the heart must maintain constant and high levels of
ATP to support its mechanical and electrical functions, it has
evolved the capacity to use different substrates for its energy-
generating pathways, depending on their availability (10–12).
Therefore, we hypothesized that this organ could be a sensitive
reporter of adaptive changes during periods of nutrient depri-
vation that involve the gut microbiota. In the current report, we
use gnotobiotic mice to identify a metabolic cross-talk between
liver and heart that is influenced by the gut microbiota. This
cross-talk, which is apparent during fasting, is centered around
Author contributions: P.A.C. and J.I.G. designed research; P.A.C., J.R.C., N.S., and B.D.M.
performed research; P.A.C., J.R.C., N.S., B.D.M., E.K.C., M.H., R.K., and J.I.G. analyzed data;
and P.A.C. and J.I.G. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
Data deposition: The GeneChip data reported in this paper have been deposited in the
Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no.
GSE14929). The 16S rRNA sequence data reported in this paper have been deposited in the
GenBank database (accession no. SRA008712.2).
1To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/cgi/content/full/
July 7, 2009 ?
vol. 106 ?
no. 27 www.pnas.org?cgi?doi?10.1073?pnas.0902366106
Results and Discussion
Ketogenesis Is Reduced in Fasted GF Mice. Nutrient deprivation in
mammals induces a shift from carbohydrate to fat utilization
(13). Free fatty acids, the products of lipolysis from peripheral
adipose stores, are oxidized in the liver either to acetyl-CoA and
ultimately CO2via the tricarboxylic acid cycle, or partially to
ketone bodies (acetoacetate, D-?-hydroxybutyrate, and ace-
tone). Ketone bodies are used as carbon sources in the brain and
other tissues (14, 15), including the heart where they form an
avidly oxidized substrate pool (16). Therefore, we measured
serum ?-hydroxybutyrate levels in 6–10-week-old male
C57BL/6J GF mice that were fed a standard low-fat, polysac-
charide-rich chow (CARB) diet or fasted for 24 h, and in
similarly aged fed or fasted mice that had received a microbiota
animals 3 weeks before the experiment [transplant recipients are
referred to as ‘‘conventionalized’’ (CONV-D)]. No significant
difference in serum ?-hydroxybutyrate levels was observed
between fed GF and CONV-D mice. Levels were increased by
fasting but were 37% lower in GF animals compared with their
fasted, colonized counterparts (1.76 ? 0.07 mM vs. 2.78 ? 0.27
mM, n ? 10 animals per group, P ? 0.001) (Fig. 1A). There were
no significant differences in serum insulin, glucose, free fatty
acid, or triglyceride levels in fasted or fed GF versus CONV-D
mice (Fig. S1 A–D).
The effect of a 24-h fast on adiposity was determined by
measuring epididymal fat-pad-to-body-weight ratios. Fasting pro-
CONV-D animals, indicating that the capacity to mobilize periph-
eral fat stores is intact even in the absence of a microbiota (Fig. S1
E and F). In the fed state, CONV-D mice have increased hepatic
triglyceride stores compared with GF animals. This difference
between GF and CONV-D animals is dramatically enhanced with
fasting (Fig. S1G), consistent with the notion that although mobi-
of a gut community promotes hepatic triacylglyerol synthesis and
storage as well as ketogenesis.
Peroxisome proliferator-activated receptor alpha (Ppar?) is a
fatty acyl-binding, nuclear receptor transcription factor required
for utilization of fatty acids during fasting in a number of tissues
including liver and heart. Ppar? expression was significantly
higher in the livers of fasting CONV-D animals compared with
GF animals (Fig. 1C). The fasting-induced ketogenic response
was blunted in CONV-D Ppar??/? mice (serum ?-hydroxybu-
tyrate levels were 0.84 ? 0.09 mM vs. 2.78 ? 0.27 mM in
wild-type CONV-D controls, n ? 10 animals per group, P ?
0.001; Fig. 1A). Moreover, unlike wild-type mice, there was no
significant difference in fasting serum ?-hydroxybutyrate levels
between GF and CONV-D Ppar??/? mice (Fig. 1A). There-
fore, we concluded that the observed effect of the gut microbiota
on fasting-induced ketosis involves Ppar?.
The Effects of the Absence of a Gut Microbiota on Hepatic Ketogen-
esis. The liver generates, but does not oxidize, ketone bodies
because it does not express the key ketolytic enzyme, 3-oxoacid
CoA transferase (17). To confirm that fasting-induced hepatic
ketogenesis is reduced in the absence of a gut microbiota, we
measured ?-hydroxybutyrate levels and the expression of several
regulators of the ketogenic response in the livers of fasted and fed
GF and CONV-D mice. Although no significant difference in
hepatic ?-hydroxybutyrate levels were observed between fed GF
and CONV-D wild-type animals, fasted GF mice had 50% lower
concentrations of this ketone body (0.36 ? 0.05 vs. 0.79 ? 0.07
is a Ppar?-inducible target that stimulates ketone body production
in the liver (18, 19). In the fed state, hepatic Fgf21 mRNA levels
Levels rose 3.08 ? 0.8-fold after a 24-h fast in CONV-D mice (P ?
0.001, n ? 10) to a value 2.72 ? 0.7-fold greater than in fasted GF
animals (P ? 0.001, n ? 10; Fig. 1D). Hepatic Fgf21 mRNA
concentrations were at the lower limits of detection by qRT-PCR
in the livers of both fed and fasting CONV-D and GF Ppar??/?
animals, emphasizing the dependence of this gene’s expression on
Ppar?. The mitochondrial form of 3-hydroxy-3-methylglutaryl-
CoA synthase (Hmgcs2) is a critical ketogenic enzyme in the liver
and a Ppar? target (20). Hmgcs2 expression was unchanged by
fasting in GF mice, but was induced 1.96 ? 0.09-fold in CONV-D
animals (P ? 0.05) (Fig. 1E). Together, these observations under-
score the role of Ppar? in regulating the augmentation in hepatic
ketogenesis during fasting that is associated with the presence of a
A colonized mouse has 2 sources for generating acetyl-CoA in
the liver that are not well-represented in a GF animal. One source
is the linkage of acetate, derived from microbial fermentation of
polysaccharides in the gut, to CoA via acetyl-CoA synthetase (EC
126.96.36.199). The other source, as described above, is from oxidation of
fatty acids derived from adipocyte triglyceride stores, which are
larger in CONV-D animals (4). During states of high fatty acid
oxidation, such as those encountered during nutrient deprivation,
acetate generated by fermentation is absorbed from the gut and
24h Fast24h Fast
nmol/mg wet weight
microbiota. (A) Serum ?-hydroxybutyrate levels in GF and CONV-D mice fed a
10 animals per condition. (B) Steady-state ?-hydroxybutyrate levels in liver
extracts.***, P ? 0.001; n ? 6 animals per condition. (C–E) qRT-PCR assays of
liver Ppar?, Fgf21, and Hmgcs2 mRNA levels that are expressed relative to fed
GF controls.*, P ? 0.05;***, P ? 0.001 (n ? 5–10 mice per condition; 2-way
ANOVA with Bonferroni posthoc testing). Note that the difference in Fgf21
mRNA levels observed between GF- and CONV-D-fed animals does not reach
statistical significance by ANOVA test.
Hepatic ketogenesis is enhanced during a 24-h fast by the gut
Crawford et al. PNAS ?
July 7, 2009 ?
vol. 106 ?
no. 27 ?
lipogenesis pathway) or for mevalonate synthesis (21), or it is
oxidized completely in mitochondria after conversion to acetyl-
CoA (22). A fourth fate for hepatic acetyl-CoA is mitochondrial
ketogenesis (see ref. 14 for further discussion).
Cecal acetate levels were 20-fold higher in fed CONV-D mice
compared with GF wild-type mice (126.7 ? 7.5 vs. 6.3 ? 2.7
nmol/mg wet-weight cecal contents, P ? 0.001, n ? 5), a finding
consistent with the fact that the gut microbiota ferments dietary
polysaccharides to SCFAs, including acetate (23). Cecal acetate
concentrations did not change significantly with fasting in GF
animals. Levels in CONV-D mice dropped 3.5 ? 1.3-fold (P ?
0.001) but remained significantly higher than those in fasted GF
animals (36.0 ? 4.3 vs. 10.8 ? 2.0 nmol/mg cecal contents, P ?
0.01, n ? 5). These findings are compatible with the notion that
acetate production, generated through fermentation of exoge-
nous and endogenous polysaccharides by members of the gut
microbiota, increases hepatocellular acetyl-CoA pools and
thereby contributes to the increased hepatic ketogenic response
observed in fasting CONV-D compared with GF animals.
The gut microbiota of mammals, including mice and humans,
is generally dominated by two phyla (divisions) of bacteria: the
Bacteroidetes, whose members possess myriad glycoside hydro-
lases and polysaccharide lyases (www.cazy.org), and the Firmi-
cutes (1). Studies in gnotobiotic mice colonized with saccharo-
lytic Bacteroides have shown that these microbes are able to
adaptively forage complex host glycans present in mucus and on
the surface of gut epithelial cells when animals are deprived of
dietary polysaccharides (24, 25). Therefore, we assessed the
effects of a 24-h fast on gut microbial ecology, including the
representation of the Bacteroidetes, by preparing DNA from
the cecal contents of fasted and fed mice, and performing
multiplex pyrosequencing of amplicons generated by PCR using
primers that flank variable region 2 (V2) of bacterial 16S rRNA
genes (26). Sequence reads were grouped into operational
taxonomic units (OTUs) based on a threshold cutoff of ?97%
identity among the V2 16S rDNA reads (1). A phylogenetic tree
was built by using 1 representative sequence from each OTU,
and employed for UniFrac analysis [this metric uses the tree to
measure the degree of similarity of any 2 communities based on
the proportion of branch length (evolutionary history) that they
share on the tree (27); see SI Text for additional details].
Although a 24-h fast did not result in a significant shift in
bacterial diversity (P ? 0.05; Fig. S2A) or a significant departure
from fed communities in overall community composition (P ?
0.05; Fig. S2B), fasting was associated with a significant increase
in the proportional representation of the Bacteroidetes [from
20.6% (fed) to 42.3% (fasted); P ? 0.01; 2-tailed Student t test
with unequal variance], and a significant diminution in the
Firmicutes [from 77.1% (fed) to 52.6% (fasted); P ? 0.007] (n ?
8 fasted; 13 controls; 2 independent experiments; Fig. S2C). This
rapid and marked expansion in the representation of the Bac-
teroidetes, achieved after a 24-h fast, yields a microbiota whose
configuration contrasts with that observed in mice who are
switched from a CARB to more-calorically dense Western-style
diet, or mice that become obese because they are leptin-deficient:
In both these cases, development of obesity is accompanied by
reduced proportional representation of the Bacteroidetes (6).
Functional Genomics Studies of the Effects of the Absence of a Gut
Microbiota on Cardiac Metabolism. Ingenuity Pathways Analysis
(IPA) of GeneChip datasets generated from the hearts of GF
and colonized animals revealed that the ketone body metabolic
pathway was significantly enriched within the myocardial tran-
scriptome of both CARB-fed CONV-D and GF wild-type mice
compared with their Ppar??/? counterparts. Enrichment was
also observed in CONV-D fed versus fasted wild-type mice but
not in fed versus fasted GF animals (Tables S1 and S2).
To explore more fully the extent to which the microbiota
influences substrate selection by the myocardium under fasting
conditions, we used GeneChip and qRT-PCR assays to compare
the levels of expression of genes involved in ketone body
metabolism, fatty acid oxidation (FAO), and glucose uptake/
oxidation. The results revealed that key FAO-associated genes
increased in CONV-D and GF hearts after a 24-h fast (Fig. S3A),
as did the ketogenic biomarker Hmgcs2 (Fig. S3B). Oxct1
encodes the ketolytic enzyme 3-oxoacid CoA transferase re-
quired for ketone body oxidation. Although there was no
significant difference between Oxct1 expression in fed CONV-D
and GF mice, fasting led to a significantly higher level of this
mRNA in the myocardium of fasted CONV-D compared with
fasted GF animals [P ? 0.001, n ? 5; Fig. S3C; note that
differences between fasted and fed GF vs. CONV-D animals
were not observed in their skeletal (gastrocnemius) muscle].
ketone body oxidation in fasting CONV-D animals. Fasting was
associated with a very modest, albeit statistically significant,
1.50 ? 0.1-fold higher level of expression of Glut1 (which
encodes a critical noninsulin-dependent glucose transporter) in
the hearts of GF animals but produced no significant change in
CONV-D mice (Fig. S3 A and D). These findings, plus the
microbiota-associated enhancement in ketone body delivery to
the fasting myocardium, prompted us to test whether glucose
utilization was also altered under these conditions.
Metabolic and Physiologic Studies of Isolated Perfused Working
Hearts: A Compensatory Shift Toward Glucose Oxidation in the Fasted
Germ-Free Myocardium. To directly assay whether the presence of
a gut microbiota altered substrate selection by actively function-
ing myocardial tissue, we turned to isolated working hearts
prepared from GF and CONV-D mice that had just completed
a 24-h fast (n ? 4 or 5 animals per group). Hearts were excised
immediately after animals were killed, and perfused ex vivo in
the working mode by using a buffer that included labeled and
unlabeled energy substrates whose final concentrations mim-
icked the fasting serum levels of glucose, fatty acids, and ketone
bodies (i.e., 5 mM glucose, 1.2 mM palmitate, and 1.7 mM
Myocardial ketone body oxidation in the isolated working
heart model is known to be proportional to the concentration of
ketone bodies delivered in the perfusate (16, 28–30). By keeping
the input ketone body concentration constant for both perfused
fasted GF and CONV-D hearts, we were able to document that
ketone body oxidation rates remained constant (773.3 ? 157 and
779.5 ? 80 nmol ?-hydroxybutyrate/min/mg dry weight for
fasted GF and CONV-D myocardium, respectively), thereby
allowing us to evaluate the relative roles of glucose versus fatty
acid oxidation on myocardial energetics.
The rate of glucose oxidation was significantly increased in the
isolated working hearts of fasted GF mice (180.6 ? 26.8 vs.
83.8 ? 14.0 nmol/min/mg dry weight (CONV-D), P ? 0.02, n ?
5 animals per group; Fig. 2A), whereas fatty acid oxidation was
not significantly altered (Fig. 2B). Glucose utilization was also
significantly greater in GF hearts in the absence of ?-hydroxy-
butyrate in the perfusion buffer (Fig. 2C). Thus, an increased
glucose utilization phenotype is ‘‘imprinted’’ in the fasted iso-
lated working GF heart and is indicative of how, in the absence
of a microbiota-dependent increase in hepatic ketogenesis, the
heart undergoes a pronounced shift in its metabolism. Stated
another way, because ketone bodies suppress myocardial glucose
utilization (31), with reduced delivery of ketone bodies, the GF
myocardium is able to turn to glucose as a prime energy source.
This increase in glucose utilization was not accompanied by
alterations in steady state levels of phosphorylated Akt or
www.pnas.org?cgi?doi?10.1073?pnas.0902366106 Crawford et al.
phosphorylated AMPK-? in the fasted GF compared with
CONV-D myocardium (Fig. S4 A and B). Glycogen levels were
significantly reduced in the fasting GF versus fasting CONV-D
myocardium (20.9 ? 3.5 vs. 41.0 ? 6.0 nmol/mg protein, n ?
5/group, P ? 0.02), consistent with increased glucose utilization
in the GF state. These results are also consistent with studies
performed in conventionally raised animals where enhanced
?-hydroxybutyrate delivery is known to suppress myocardial
glucose oxidation, and thus increase glycogen storage (32).
Heart rate, cardiac hydraulic work, and cardiac output were
not statistically different between isolated working hearts pre-
pared from fasted GF versus CONV-D mice (Fig. S4 C–E)
despite decreased glucose utilization in CONV-D hearts. There
were no detectable differences in mitochondrial morphology or
number (Fig. S5 A and B) or in the results of in vitro assays of
state 2, state 3, or state 4 mitochondrial respiration (Fig. S5C).
In vivo echocardiographic assessments indicated that heart rate
and fractional shortening decreased to a comparable degree in
both groups of animals after a 24-h fast (Fig. S6 A–C). Thus, in
the absence of a microbiota-associated, fasting-induced shift
in myocardial metabolism toward ketone body utilization, the
(fasting) GF heart is able to sustain its performance, at least in
part, by increasing glucose utilization.
An Effect of the Gut Microbiota on Myocardial Mass. Intriguingly,
heart weight, whether measured echocardiographically or as wet
mass, and whether normalized to tibial length or lean body mass,
was significantly reduced in fasted GF versus CONV-D mice
(Fig. 3 A–C). This difference was also evident in CARB-fed
animals (HW/TL ? 7.28 ? 0.21 in CONV-D vs. 6.65 ? 0.18 in
GF animals; P ? 0.03 according to Student’s t test). IPA analysis
of GeneChip datasets indicated that the increased myocardial
mass observed in fed and fasted CONV-D compared with GF
of signaling pathways that mediate pathological hypertrophy (for
further details see SI Text and Fig. S7).
Endurance Training Produces Physiological Hypertrophy of the Heart.
To determine whether the GF heart is able to manifest the
features of the normal physiological hypertrophic response to
training, GF and CONV-D animals were exercised in swimming
tanks placed in each of their respective gnotobiotic isolators (for
further details concerning these ‘‘gnotatoria’’ and the exercise
regimen, see SI Text). After the 30-day training period, echo-
cardiograms were performed: Heart rate was decreased in GF
animals to a level that was not statistically different from
CONV-D mice (Fig. S7C). However, the hearts of trained GF
mice remained smaller than the hearts of the trained CONV-D
mice (Fig. S7 D and E). IPA-based comparisons of the myocar-
dial transcriptomes of untrained versus trained CARB-fed GF
mice indicated that exercise resulted in gene expression changes
in the IGF-1 and ERK/MAPK signaling cascades, signal trans-
duction pathways that in CONV-D mice are associated with
‘‘physiological hypertrophy’’, plus enrichment for mRNAs en-
coding mediators of oxidative phosphorylation (Fig. S7F; Tables
S3 and S4). However, unlike the transcriptional changes that
occur in the hearts of trained CONV-D mice, trained GF mice
lack significant enrichment in a subset of pathways that includes
ketone body metabolism (Table S3). Together, these findings
suggest that diminished ketone metabolism during exercise may
contribute to the inability of endurance training to rescue the
reduced-myocardial-mass phenotype in GF animals.
To further test the notion that the myocardial metabolic and
morphologic phenotypes observed in fasting GF mice reflect the
consequences of decreased hepatic ketogenesis, untrained GF
and CONV-D mice were placed on a very low carbohydrate,
ketogenic diet (0.4% of calories as CARB, 95% as fat, 4.5% as
protein). A ketogenic diet did not produce a notable change in
the ratio of Bacteroidetes to Firmicutes in the cecal microbiota
compared with CARB-fed, untrained (or trained) controls (Fig.
S2C). However, unlike fasting, a ketogenic diet was associated
with a significant decrease in bacterial diversity (P ? 0.001; Fig.
S2A). Principal coordinates analysis (PCoA), based on UniFrac
distances, also revealed clustering of the gut communities of
ketogenic diet-fed animals and a significant difference in the
overall composition of their cecal microbiota versus mice fed the
CARB diet (P ? 0.001 along PC1, Fig. S2B).
Nonfasting GF animals that consumed the ketogenic diet for
30 days were able to mount a hepatic ketogenic response
equivalent to that seen in their ketogenic diet-fed CONV-D
counterparts (Fig. S8 A–D). Correspondingly, this diet produced
a shift in the transcriptome toward ketone body metabolism in
the hearts of GF animals that was equivalent to CONV-D mice
(see Fig. S8E). Remarkably, the reduction in heart weight
observed in GF animals was completely abrogated (Fig. 3 B and
C). Together, these results support the notion that myocardial
nmol/min/g dry weight
nmol/min/mg dry weight
nmol/min/mg dry weight
p = 0.14
Hearts from fasted WT GF or CONV-D animals were perfused for 60 min in the
butyrate, 5 mM glucose, and 1.2 mM palmitate. (A) Oxidation of D-[14C(U)]-
glucose; n ? 5 animals per group. (B) Oxidation of [9,10]-3H-palmitate; n ? 8
Influence of the gut microbiota on myocardial substrate selection.
24h Fast Ketogenic
LVM/TL index (mg/mm)
per group). (B and C). The reduction in myocardial mass [heart weight (HW)]
normalized to tibial length (B), or to lean body mass [fat-free mass (FFM), as
defined by MRI] (C), that is observed in GF mice fasted for 24 h after being
maintained on a CARB diet, is abrogated in GF animals maintained on a
ketogenic diet.*, P ? 0.05;**, P ? 0.01;***, P ? 0.001 (n ? 8 animals in each
fasting and ketogenic diet group, and 6 animals in each CARB-fed group;
2-way ANOVA with Bonferroni posthoc testing).
Influence of the presence or absence of the gut microbiota on
Crawford et al. PNAS ?
July 7, 2009 ?
vol. 106 ?
no. 27 ?
ketone body metabolism contributes to the myocardial-mass
phenotype observed in GF mice.
We subsequently placed Ppar??/? mice on the ketogenic
diet. However, in keeping with their deficiency in hepatic fatty
a moribund appearance within 2 weeks, and overt hepatic
steatosis at the time of euthanasia. Therefore, we were not able
to use these animals to perform a direct genetic test of the role
played by Ppar? in mediating the effect of the ketogenic diet on
the GF myocardial-mass phenotype.
Prospectus. Although considerable progress has been made in
characterizing the mechanisms regulating fatty acid and glucose
metabolism in normal, diabetic, and failing heart by Ppars and
insulin, substantially less is known about the relationship be-
tween ketone body metabolism and cardiovascular health. The
results presented here indicate that during fasting, the presence
of a gut microbiota is able to enhance the supply of a nutrient
source (ketone bodies) that the heart avidly oxidizes. In the
absence of a microbiota, the relative paucity of ketone bodies is
associated with a compensatory increase in glucose utilization.
As such, the present study expands our view of the benefits
provided by a gut microbiota, both during periods of exposure
to nutrient excess, where harvest and storage of energy is
promoted, and during periods of nutrient deprivation, where
substrate selection in the heart and myocardial mass are mod-
ified through effects on systemic ketone body metabolism.
We anticipate that using metagenomic methods to define the
composition and operations of the gut microbial community will
result in better understanding of interrelationships between diet,
nutritional status, and the determinants of cardiovascular health.
Moreover, comparative metagenomic studies of the gut micro-
biota of a variety of host species may be particularly informative,
e.g., Burmese pythons increase their heart mass by 40% in 48 h
after breaking a long fast (35).
Animals. All protocols involving animals used in this study were approved by
the Washington University Animal Studies Committee. GF wild-type C57BL/6J
mice were maintained in flexible plastic gnotobiotic isolators under a strict
12-h light cycle (lights on at 0600), and fed an autoclaved CARB diet (B & K
Universal) ad libitum. GF C57BL/6J Ppar??/? mice were generated as previ-
ously described (4).
Male GF mice were colonized at 6–10 weeks with a single gavage of cecal
contents that had been harvested from adult CONV-R donors and resus-
pended in PBS (5 mL per donor; 500 ?L of this suspension administered per
recipient). Recipients were subsequently kept in gnotobiotic isolators and
24 h or continued on the CARB diet. A subset of GF and CONV-D mice was fed
an irradiated sterile ketogenic diet (95.1% fat, 4.5% protein, 0.1% carbohy-
drate, Bio-Serv AIN-76A) for 30 days. All animals in all treatment groups were
killed in mid-morning.
16S rRNA-Based Enumeration Studies of the Gut Microbiota. Isolation of DNA
from cecal contents, PCR of the V2 region of bacterial 16S rDNA genes,
multiplex sequencing of the resulting bar-coded amplicons, and data analysis
were performed by using methods described previously (1, 26, 27); see SI Text
for more details.
Metabolite Analyses. ?-hydroxybutyrate was measured in serum or liver ex-
tracts by GC-MS, using DL-3-Hydroxybutyric acid-13C4sodium salt (Sigma) as
extract). Liver extracts were prepared by homogenizing a 100-mg sample of
in 750 ?L of chloroform/methanol (2:1 ratio). Protein concentrations in liver
extracts were measured by using a bicinchoninic acid assay (Pierce) and were
not significantly different among fasted and fed GF and CONV-D animals.
To quantify acetate in cecal contents, 100 nmol sodium acetate-13C2,d3
(Sigma) was added to snap-frozen cecal contents (100 mg).The material was
emulsified by adding 50 ?L of 37% HCl, extracted with diethyl ether, and
derivitized with N-(butyl-dimethyl-silyl)-2,2,2-trifluoro-N-methyl-acetamide
(Sigma). Derivatized samples were analyzed on a Hewlett Packard 6890 gas
chromatograph interfaced to an Agilent 5973 mass spectrometer.
GeneChip Analyses. Mammalian cRNA targets were prepared and hybridized
to Affymetrix MOE 430 2.0 GeneChips. Five hearts were analyzed individually
(5.0) software to an intensity of 500 and subsequently analyzed with dChip
(http://biosun1.harvard.edu/complab/dchip/). Selection criteria for differen-
bound 90% confidence interval) ? 2; (ii) P value (t test) ? 0.05; (iii) 100%
present call in the condition with higher expression; (iv) minimal absolute
intensity ?100; (v) false discovery rate ?1%. Datasets generated in dChip
using these criteria were organized into canonical signaling and metabolic
pathways by using IPA software (www.ingenuity.com). For more information
on the genes used in dChip analysis, see SI Text and Table S5.
qRT-PCR. Total cellular RNA was isolated from snap-frozen heart and liver sam-
ples (4). qRT-PCR assays were performed by using SYBR-green detection with an
Mx3000P Stratagene Thermocycler. Data were normalized to Rpl32 mRNA and
analyzed by using the ??C?method. PCR primers are listed in Table S6.
Isolated Working Heart. Mice received 100 units heparin by i.p. injection and
10 min later were anesthetized with an i.p. injection of 10 mg of sodium
pentobarbital. Hearts were excised and placed in ice-cold Krebs–Henseleit
bicarbonate solution (118 mM NaCl, 25 mM NaHCO3, 4.7 mM KCl, 0.4 mM
KH2PO4, 2.5 mM CaCl2, pH 7.4) supplemented with 5 mM glucose, 1.2 mM
palmitic acid (bound to 3% fatty acid-free BSA), and 1.7 mM D-?-hydroxybu-
tyrate. Hearts were cannulated via the aorta and temporarily perfused, in a
retrograde fashion, by using the Langendorff mode, with continuous bub-
bling of a 95% O2/5% CO2gas mixture into the buffer reservoir to maintain
tissue oxygenation during subsequent left atrial cannulation. Following can-
nulation, the perfusion circuit was changed to the antegrade working mode
at 37 °C. Samples of the perfusate were collected every 10 min for measure-
ment of (i)14CO2(trapped in 1 M hyamine hydroxide solution), generated
from oxidation of glucose or ?-hydroxybutyrate, and (ii)3H2O, released into
the buffer as a result of palmitate oxidation (36).
Measurements of cardiac output, aortic flow, peak systolic pressure, and
heart rate were acquired every 10 min for 10 sec by using inline flow probes
(Transonic Systems, Inc), the MP100 system from AcqKnowledge (BIOPAC
Systems, Inc), and a pressure transducer (TSD 104A, BIOPAC System, Inc).
Cardiac work was calculated as the product of peak systolic pressure and
cardiac output. At the end of each perfusion, the dry weight of the ventricles
was defined after desiccation in a vacuum oven.
Statistical Analysis. The approach used for analysis of GeneChip datasets is
Student’s t test or 2-way ANOVA with Bonferroni posthoc testing, as appro-
priate, by using GraphPad Prism software. Mean values ? SEM are shown for
all bar graph plots.
ACKNOWLEDGMENTS. We thank Maria Karlsson, David O’Donnell, Sabrina
Wagoner, Jill Manchester, Howard Wynder, Sandeep Mahajan, Naveen
Reddy, and Attila Kovacs for superb technical assistance; Jean Schaffer, Dan
Ory, and Clay Semenkovich for critical review of the manuscript; and Laura
Institutes of Health Grants DK073282, DK70977, DK020579, DK56341, and
1. Ley RE, et al. (2008) Evolution of mammals and their gut microbes. Science 320:1647–
2. Ley RE, et al. (2008) Worlds within worlds: Evolution of the vertebrate gut microbiota.
Nat Rev Microbiol 6:776–788.
3. Flint HJ, et al. (2008) Polysaccharide utilization by gut bacteria: Potential for new
insights from genomic analysis. Nat Rev Microbiol 6:121–131.
4. Backhed F, et al. (2004) The gut microbiota as an environmental factor that regulates
fat storage. Proc Natl Acad Sci USA 101:15718–15723.
5. Turnbaugh PJ, et al. (2006) An obesity-associated gut microbiome with increased
capacity for energy harvest. Nature 444:1027–1031.
6. Turnbaugh PJ, Backhed F, Fulton L, Gordon JI (2008) Diet-induced obesity is linked to
marked but reversible alterations in the mouse distal gut microbiome. Cell Host
7. Samuel BS, et al. (2008) Effects of the gut microbiota on host adiposity are modulated
by the short-chain fatty-acid binding G protein-coupled receptor, Gpr41. Proc Natl
Acad Sci USA 105:16767–16772.
www.pnas.org?cgi?doi?10.1073?pnas.0902366106Crawford et al.
8. Turnbaugh PJ, et al. (2009) A core gut microbiome in obese and lean twins. Nature Download full-text
9. Tennant B, Malm OJ, Horowitz RE, Levenson SM (1968) Response of germfree, con-
ventional, conventionalized and E. coli monocontaminated mice to starvation. J Nutr
of balance. J Clin Invest 115:547–555.
11. Lopaschuk GD (2006) Optimizing cardiac fatty acid and glucose metabolism as an
approach to treating heart failure. Semin Cardiothorac Vasc Anesth 10:228–230.
12. Kodde IF, van der Stok J, Smolenski RT, de Jong JW (2007) Metabolic and genetic
regulation of cardiac energy substrate preference. Comp Biochem Physiol A Physiol
13. Cahill GF, Jr, et al. (1966) Hormone–fuel interrelationships during fasting. J Clin Invest
14. McGarry JD, Foster DW (1980) Regulation of hepatic fatty acid oxidation and ketone
body production. Annu Rev Biochem 49:395–420.
15. Cahill GF, Jr (2006) Fuel metabolism in starvation. Annu Rev Nutr 26:1–22.
16. Jeffrey FM, Diczku V, Sherry AD, Malloy CR (1995) Substrate selection in the isolated
working rat heart: Effects of reperfusion, afterload, and concentration. Basic Res
messenger RNA levels of succinyl-coenzyme A (CoA):3-ketoacid CoA transferase and
mitochondrial and cytosolic acetoacetyl-CoA thiolases. Pediatr Res 42:498–502.
18. Badman MK, et al. (2007) Hepatic fibroblast growth factor 21 is regulated by PPARal-
pha and is a key mediator of hepatic lipid metabolism in ketotic states. Cell Metab
19. Inagaki T, et al. (2007) Endocrine regulation of the fasting response by PPARalpha-
mediated induction of fibroblast growth factor 21. Cell Metab 5:415–425.
20. Hegardt FG (1999) Mitochondrial 3-hydroxy-3-methylglutaryl-CoA synthase: A control
enzyme in ketogenesis. Biochem J 338:569–582.
in the oxidation of acetate. J Biol Chem 276:11420–11426.
[1–13C] acetate. Am J Physiol 271:E58–E64.
24. Sonnenburg JL, et al. (2005) Glycan foraging in vivo by an intestine-adapted bacterial
symbiont. Science 307:1955–1959.
transmission of a saccharolytic human gut bacterial symbiont. Cell Host Microbe
26. Hamady M, et al. (2008) Error-correcting barcoded primers for pyrosequencing hun-
dreds of samples in multiplex. Nat Methods 5:235–237.
27. Lozupone C, Knight R (2005) UniFrac: A new phylogenetic method for comparing
microbial communities. Appl Environ Microbiol 71:8228–8235.
28. Yaffe SR, Gold AJ (1979) Effect of prolonged starvation on substrate uptake in the
isolated perfused rat heart. J Nutr 109:2140–2145.
Mol Cell Biochem 110:17–23.
30. Fukao T, Lopaschuk GD, Mitchell GA (2004) Pathways and control of ketone body
metabolism: On the fringe of lipid biochemistry. Prostaglandins Leukot Essent Fatty
31. Pelletier A, Coderre L (2007) Ketone bodies alter dinitrophenol-induced glucose up-
take through AMPK inhibition and oxidative stress generation in adult cardiomyo-
cytes. Am J Physiol 292:E1325–E1332.
32. Goodwin GW, Taegtmeyer H (1994) Metabolic recovery of isolated working rat heart
after brief global ischemia. Am J Physiol 267:H462–H470.
33. Kersten S, et al. (1999) Peroxisome proliferator-activated receptor alpha mediates the
adaptive response to fasting. J Clin Invest 103:1489–1498.
34. Leone TC, Weinheimer CJ, Kelly DP (1999) A critical role for the peroxisome prolifera-
tor-activated receptor alpha (PPARalpha) in the cellular fasting response: The PPARal-
pha-null mouse as a model of fatty acid oxidation disorders. Proc Natl Acad Sci USA
35. Andersen JB, et al. (2005) Physiology: Postprandial cardiac hypertrophy in pythons.
36. Finck BN, et al. (2002) The cardiac phenotype induced by PPARalpha overexpression
mimics that caused by diabetes mellitus. J Clin Invest 109:121–130.
Crawford et al.PNAS ?
July 7, 2009 ?
vol. 106 ?
no. 27 ?