Abstract

The gastrointestinal tract and the microbes colonizing it form a complex ecosystem that has various effects on the well-being of the host. In addition to acute infections, the composition of the gastrointestinal microbiota has been suspected to influence the etiopathogenesis of many chronic diseases, such as rheumatoid arthritis and inflammatory bowel diseases. It has been suggested that the bacterial colonization of the gastrointestinal tract is genetically determined. Using gas–liquid chromatography of bacterial cellular fatty acids we show in this study that modulation of the microbiota by a course of antibiotics is followed by regeneration of the murine intestinal flora depending on the genotype of the host. The mice used in our study were acclimatized to identical living conditions before treatment with ciprofloxacin and clindamycin for 1 week via drinking water. Within a few days of finishing the antibiotic course, the cellular fatty acid profiles of fecal samples resembled those of the pre-course community, showing a considerable indigenous recovery potential. Colonization of the gastrointestinal tract appeared to be genetically regulated since differences in communities between the mouse strains were observed. Our results are in harmony with earlier observations, indicating that the gut community is not established by chance and that it is influenced by host-derived factors.

Introduction

The gastrointestinal tract is a complex ecosystem containing a large number of resident microorganisms. Large bowel contents of mammals are estimated to contain 1012 bacterial cells per gram [1,2]. Interactions with these microbes are important to the well-being of the host. However, the factors influencing the establishment and consistency of the gut communities are still poorly understood. This is partly due to limitations of methods such as traditional plate culturing and species detection based on biochemical reactions. Plate culturing allows detection of bacteria only as colony-forming units, which underestimate true viable cell counts [3]. Consequently methods that are independent of bacterial growth have been developed. Gas–liquid chromatography (GLC) of bacterial cellular fatty acids (CFAs) has successfully been used to analyze gut communities in ecological studies [4,5]. GLC does not identify single bacterial species from multi-species samples but rapidly and accurately recognizes changes in bacterial populations. Methods based on detection of DNA profiles of fecal samples have are also useful and are increasingly used in ecological studies. Denaturing-gradient gel electrophoresis (DGGE) and temperature-gradient gel electrophoresis of bacterial ribosomal RNA fragments have in previous studies proved their effectiveness in the analysis of fecal flora [6–9]. In these studies fecal communities are observed to be stable over time. Proportions of bacterial species in human intestinal communities have been found to alter significantly at infancy but tend to stabilize at a certain age [10]. In mice the sex does not have an effect on the composition of community, while ageing seems to change communities slowly [11]. Changes in diet, disease and medication are known to alter the stability of the gut communities rapidly [5,12,13]. However, communities seem to have an inborn tendency to recover its original composition after the modulating factor has been removed.

In our previous studies, we have observed fecal communities to be very similar within the inbred mice of the same age living in identical conditions [11,14]. Fecal communities of congenic mouse strains differing only in major histocompatibility complex (MHC)-genotype were found to differ significantly when compared to each other. Thus, host genetics seems to have an effect on the composition of murine intestinal communities. Recently, the effect of genotype on intestinal communities has been observed also in humans by using DGGE to study the 16S rRNA gene-profiles of fecal samples [8]. 16S rRNA profiles were more homogeneous in monozygotic than in dizygotic twins. Homogeneity was lower when samples from unrelated persons were compared to each other. However, in human studies the living conditions cannot be standardized and the influence of unique life style factors on communities are difficult to exclude. In our current study, we wanted to find out whether the murine gut community recovers to its original state after modulation with antibiotics and whether the host-derived factors determine the fecal bacterial composition after it has been dramatically changed.

Materials and methods

Mice

Five-week-old female mice from eight strains were purchased from The Jackson Laboratory (Bar Harbor, Maine, USA). The strains were C57BL/10J, B10.A(4R), B10.A(5R), B10.HTG, C57BL/6J, B6.C, A/J and A.TL and detailed information on genotypes is provided in Table 1. Five to seven animals of each strain were included. The mice were born by natural births and weaned at the age of 3–4 weeks. All baby mice belonging to a particular strain were not from the same mother but were collected from two or more mothers that had given birth on the same date. After weaning, the mice were fed with NIH Rat and Mouse/Auto 6F-diet (Purina Mills, St. Louis, MI, USA) packed individually and transported to Finland. They arrived at our laboratory at the age of five weeks. In the test animal facility the mice were housed singly in the same room in Macrolon I cages to eliminate environmental bacterial exposure and to keep living conditions as identical as possible. Mice were fed with autoclaved R36 diet (Lactamin AB, Södertälje, Sweden) and had free access to water. Autoclaved aspen beddings were changed twice a week. After a 1-week adaptation to the test animal laboratory conditions and diet, ciprofloxacin (concentration 0.08 mg ml−1) and clindamycin (0.08 mg ml−1) were added to drinking water for 1 week. The combination of ciprofloxacin and clindamycin covers a wide spectrum of known gut-related microbes. Ciprofloxacin is effective against both Gram-negative and Gram-positive aerobic bacteria while clindamycin is active against anaerobic species.

1

Genetic relationships of the mouse strains used

graphic 
graphic 

The four congenic strains with C57BL/10-background differ from each other only in MHC-genotype (H-2-type). Five to seven animals of each strain were included. The different H-2-genotypes have different alleles mainly in MHC I and MHC II-regions [28].

Stool samples

Stool samples were collected from mice at the ages of 5, 6, 7, 8, 9, 11, 13 and 17 weeks. Only two persons took part in sample collection and animal treatment. The stool samples were collected in the morning. In one sample collection, from three to four fecal pellets per mouse were collected as rectal samples directly from the anus into plastic tubes and stored at −70°C within 1 h. Before GLC analysis, the stool samples were processed to separate bacteria from other fecal material as described before [11,15]. First, fibrous material and eukaryotic cells from the gut wall were allowed to sediment by suspending the fecal pellets (total weight ca. 0.1 g) in 5 ml physiological saline, gently mixing and allowing to suspend for 2 h at 4°C. The sample was then remixed and allowed to sediment for a further 15 min. The supernatant, which in addition to bacterial cells contains free fatty acids and fragments of eukaryotic and prokaryotic cell membranes, was removed and centrifuged (1000×g, 15 min at room temperature) to produce a pellet containing bacteria but not other fatty acids present in the feces.

For GLC the separated bacterial mass was saponified, methylated and analyzed as described previously [4]. In brief, the harvested bacterial mass was incubated for 30 min at 100°C with 15% (w/v) NaOH in 50% aqueous methanol and then acidified to pH 2.0 with 6 N aqueous HCl in CH3OH. The methylated fatty acids were then extracted with ethyl ether and hexane.

GLC and data analysis

In earlier studies, the GLC of stool samples has proved suitable for reliable analysis of a large number of samples in a short time [15,16]. In this study, GLC was performed using an HP 6890 gas chromatograph with an Ultra 2 (Crosslinked 5% PH ME Siloxane) 25 m×0.2 mm column (HP 19091B) combined with HP ChemStation analysis software. Ultra-high-purity helium was used as a carrier gas.

The analysis of GLC data is based on the computerized comparisons of bacterial CFA profiles. Fatty acid profiles of individual samples are compared to each other and cluster analysis reveals the relative similarity of the samples. For this purpose, a bacterial identification program, developed previously, [4] was used to analyze GLC data. The three chromatograms and corresponding comparisons between them are presented in Fig. 1.

1

Examples of CFA profiles produced by GLC. The length of the each peak represents the relative amount of one individual fatty acid in the sample. A, B and C are examples of fatty acid profiles of individual stool samples (B and C appear inverted). The peaks presenting the same fatty acids are located in the same vertical line and can be compared. Fatty acid profiles of individual samples are compared to each other, and the similarity indices between the samples are calculated by computer [4].

1

Examples of CFA profiles produced by GLC. The length of the each peak represents the relative amount of one individual fatty acid in the sample. A, B and C are examples of fatty acid profiles of individual stool samples (B and C appear inverted). The peaks presenting the same fatty acids are located in the same vertical line and can be compared. Fatty acid profiles of individual samples are compared to each other, and the similarity indices between the samples are calculated by computer [4].

To calculate the statistical significance of a difference between two groups, the variation of CFA profiles within each group was compared to that between groups. The variation within a group was determined by calculating the mean±standard deviation (SD) for all paired comparisons within the group. The variation between two groups was calculated by comparing each CFA profile in one group to all profiles in the other group. The mean±SD was calculated for all of these comparisons. Finally, the variation between the groups was compared to that within the groups by calculating a z-value in order to determine the P-value from the z-table [4,15,17].

Results

CFA profiles of samples from mice of the same strain appeared to be highly homogeneous at each time point. Similarity indices between mice of a particular strain varied from 82.8 to 99.3, over two thirds being higher than 90 (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity). The fecal community was strongly modified by antibiotic treatment. In Fig. 2, stool samples from each time point are compared to the first samples of the same strain. Antibiotic treatment was started after collection of the first stool sample and stopped after collection of the second sample. The average similarity index decreased to 7.3 between the first and the second samples, indicating a remarkable and predictable change in the fecal community when compared to the pre-antibiotic samples. Recovery of the community was rapid and its tendency to restore its original state appeared to be strong. When the stool samples collected 1 week after the antibiotic course are compared to the pre-antibiotic samples, the average similarity index rises to 72.1. Later, when samples from each time point are compared to the pre-antibiotic samples, the average similarity indices rose further reaching almost 80 at the end of the study. The dynamics of the community during and after antibiotic treatment were similar in all strains of mice.

2

Comparison of stool samples from each time point to the first samples of the same strain. For each group, a similarity index of 100 is given to the first samples collected at the age of 5 weeks. For other indices, the CFA profiles of the stool samples collected at a marked time point are compared to the CFA profiles of the first samples within the same strain (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity). Antibiotic treatment was started after collection of the first samples. The second samples were collected after 1 week of antibiotic treatment, after which normal drinking water was supplied. The third samples were taken 1 week after the second samples and so onwards. Intestinal communities recovered immediately after completion of antibiotic treatment and 10 weeks later the average similarity index was 78.2 when compared to the pre-antibiotic samples.

2

Comparison of stool samples from each time point to the first samples of the same strain. For each group, a similarity index of 100 is given to the first samples collected at the age of 5 weeks. For other indices, the CFA profiles of the stool samples collected at a marked time point are compared to the CFA profiles of the first samples within the same strain (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity). Antibiotic treatment was started after collection of the first samples. The second samples were collected after 1 week of antibiotic treatment, after which normal drinking water was supplied. The third samples were taken 1 week after the second samples and so onwards. Intestinal communities recovered immediately after completion of antibiotic treatment and 10 weeks later the average similarity index was 78.2 when compared to the pre-antibiotic samples.

In addition, comparisons between mouse strains at each time point were made. The GLC data for stool samples from different time points were pooled within C57BL/6J-, C57BL/10J- and A-backgrounds to assess the differences between the strains with different genetic backgrounds. The effect of host genetics on the fecal community is presented in Fig. 3, in which the similarity indices (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity) and statistical difference between the congenic strains of the same genetic background and the similarity indices and statistical difference between the strains of the different backgrounds are presented. At the beginning of the study period similarity indices were higher between congenic strains than between strains of different genetic backgrounds. Also P-values are higher for comparisons between congenic strains than between strains with different backgrounds. Antibiotic treatment diminished differences between the strains but differences appeared again when treatment was terminated. According to similarity indices and P-values the genetic background, i.e. genes other than MHC, seems to have a stronger influence on the CFA profiles, although there were also some significant differences between congenic strains.

3

The effect of genetic background on the fecal community. Similarity indices (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity) between congenic strains in the same genetic background and between strains in the different backgrounds are shown. Similarity indices are regularly higher between the congenic strains (triangles) than between the strains in the different genetic backgrounds (squares). Statistical differences between strains at each time point are shown as P-values of paired comparisons. P-values given indicate significance of the similarity of the samples within each group. P-values are generally less significant or insignificant in comparisons between congenic strains while in comparisons between the strains in the different backgrounds P-values were significant at each time point.

3

The effect of genetic background on the fecal community. Similarity indices (scale from 0 to 100; 100 indicating complete similarity and 0 complete dissimilarity) between congenic strains in the same genetic background and between strains in the different backgrounds are shown. Similarity indices are regularly higher between the congenic strains (triangles) than between the strains in the different genetic backgrounds (squares). Statistical differences between strains at each time point are shown as P-values of paired comparisons. P-values given indicate significance of the similarity of the samples within each group. P-values are generally less significant or insignificant in comparisons between congenic strains while in comparisons between the strains in the different backgrounds P-values were significant at each time point.

Discussion

The aim of this study was to determine the ability of a host to remodel its intestinal community after a temporary modification by GLC analysis of bacterial CFAs. Mice from eight strains were acclimatized to identical living conditions and ciprofloxacin and clindamycin were given for 1 week via drinking water to change the composition of the community. The antibiotic course changed the intestinal flora significantly as expected. Following a return to the normal diet, the community recovered remarkably quickly and 1 week after finishing antibiotic treatment the fecal CFA profiles were similar to those before treatment, demonstrating the community's ability to restore its original state. Recovery was also observed to be linked to the host genotype. Differences in the genetic background were the most important because the similarity indices between the strains with the same genetic background were regularly higher than those between the strains from different genetic backgrounds and statistical difference was correspondingly lower between congenic strains. Some differences between congenic strains were observed, but the influence of MHC was obviously weaker than that of background.

The composition of the fecal community and the factors influencing it are poorly understood. Our previous studies of intestinal ecology by GLC of bacterial CFAs indicate that intestinal communities have a tendency to retain stability in stable living conditions. Changes in diet and medication alter the flora but these alterations disappear rapidly after the modifying factor is removed [5]. Certain diseases appear to be linked to changes in fecal community, for example the pathogenesis of rheumatoid arthritis [18]. The response to effective diet treatment in rheumatoid arthritis has been connected with simultaneous alterations in the intestinal community [19,20]. There are reports that the gut community is a major immunogenic factor in Crohn's disease and a diet treatment altering the community relieves the symptoms [21–23]. There is evidence that the bacterial composition of the intestinal community is not determined coincidentally by microbes colonizing the gastrointestinal tract during ontogeny [24,25]. In our earlier studies, the inbred mouse strains have been observed to have different gut communities [11,14]. Whether this is caused by variable bacterial exposure during ontogeny or host-derived factors remains unknown. It is possible that the differences observed in this study were due to coprophagy by the mice, i.e. the single-caged mice would re-establish their gut community from their own feces. However, the regular change of the litter and the cleaning of the cages reduce this possibility. Baby mice are first colonized from their mothers’ birth canals and it can be speculated that the observed differences between strains are due to differences in the bacterial communities of the birth canals. However, in all strains the baby mice were from at least two mothers that had given birth on the same date. Thus, it is not probable that the exposure to the mothers’ birth canal community was responsible for the observed differences between mouse strains.

Possible evidence for the interaction between host genetics and gut bacteria can be found from recently published studies concerning the regulation of immune responses in gut. Certain bacteria were observed to upregulate genotype-linked factors activating lymphocytes in gut epithelium, thereby inducing induce gut inflammation [26]. Also, bacteria present in normal gut flora have been observed to regulate genes involved in intestinal functions like nutrient absorption [27]. According to these observations, characteristics of intestinal flora can modulate the gene expression of gut epithelial cells. The intestine is a complex ecosystem in which the host and bacteria influence each other. The present results support the hypothesis that the host-derived factors are able to modulate the composition of the intestinal flora. Exact molecular mechanisms behind this phenomenon are largely unknown but our results give further evidence for genetic determination of the composition of the gut community.

Acknowledgements

This study was supported by EVO of Turku University Central Hospital.

References

[1]
Moore
W.E.C.
Holdeman
L.V.
(
1974
)
Human fecal flora: the normal flora of 20 Japanese-Hawaiians
.
Appl. Microbiol.
 
27
,
961
979
.
[2]
Tannock
G.W.
(
1997
)
Normal microbiota of the gastrointestinal tract of rodents
In:
Gastrointestinal Microbiology
  (
Mackie
R.I.
White
B.A.
Isaacson
R.E.
Eds.), Volume 2, pp.
187
215
.
Chapman and Hall Microbiology Series
,
New York
.
[3]
Byrd
J.B.
Xu
H.
Colwell
R.R.
(
1991
)
Viable but nonculturable bacteria in drinking water
.
Appl. Environ. Microbiol.
 
57
,
875
878
.
[4]
Eerola
E.
Lehtonen
O.P.
(
1988
)
Optimal data processing procedure for automatic bacterial identification by gas-liquid chromatography of cellular fatty acids
.
J. Clin. Microbiol.
 
26
,
1745
1753
.
[5]
Peltonen
R.
Eerola
E.
(
1992
)
Direct automatic bacterial analysis of rat stools samples; the effects of diet and medical treatment studied by computerised gas-liquid chromatography of bacterial fatty acid
.
Microb. Ecol. Health Dis.
 
5
,
93
103
.
[6]
Satokari
R.M.
Vaughan
E.E.
Akkermans
A.D.L.
Saarela
M.
DeVos
W.M.
(
2001
)
Bifidobacterial diversity in human feces detected by genus-specific PCR and denaturing gradient gel electrophoresis
.
Appl. Environ. Microbiol.
 
67
,
504
513
.
[7]
Satokari
R.M.
Vaughan
E.E.
Akkermans
A.D.L.
Saarela
M.
DeVos
W.M.
(
2001
)
Polymerase chain reaction and denaturing gradient gel electrophoresis monitoring of fecal Bifidobacterium populations in a prebiotic and probiotic feeding trial
.
Syst. Appl. Microbiol.
 
24
,
227
231
.
[8]
Zoetendal
E.G.
Akkermans
A.D.L.
Akkermans-vanVliet
W.M.
DeVisser
J.A.G.M.
DeVos
W.M.
(
2001
)
The host genotype affects the bacterial community in the human gastrointestinal tract
.
Microb. Ecol. Health Dis.
 
13
,
129
134
.
[9]
Zoetendal
E.G.
Akkermans
A.D.L.
DeVos
W.M.
(
1998
)
Temperature gradient gel electrophoresis analysis of 16S rRNA from human fecal samples reveals stable and host-specific communities of active bacteria
.
Appl. Environ. Microbiol.
 
64
,
3854
3859
.
[10]
Conway
P.L.
(
1997
)
Development of intestinal microbiota
, p.
3
38
. In:
Gastrointestinal Microbiology
 , (
Mackie
R.I.
White
B.A.
Isaacson
R.E.
Eds.), Volume 2.
Chapman and Hall Microbiology Series
,
New York
.
[11]
Vaahtovuo
J.
Toivanen
P.
Eerola
E.
(
2001
)
Study of murine fecal microflora by cellular fatty acid analysis; effect of age and mouse strain
.
Antonie van Leeuwenhoek
 
80
,
35
42
.
[12]
Eerola
E.
Möttönen
T.
Hannonen
P.
Luukkainen
R.
Kantola
I.
Vuori
K.
Tuominen
J.
Toivanen
P.
(
1994
)
Intestinal flora in early rheumatoid arthritis
.
Br. J. Rheumatol.
 
33
,
1030
1038
.
[13]
Peltonen
R.
Ling
W.-H.
Hänninen
O.
Eerola
E.
(
1992
)
An uncooked vegan diet shifts the profile of human fecal microflora: computerized analysis of direct stool sample gas-liquid chromatography profiles of bacterial cellular fatty acids
.
Appl. Environ. Microbiol.
 
58
,
3660
3666
.
[14]
Toivanen
P.
Vaahtovuo
J.
Eerola
E.
(
2001
)
Influence of major histocompatibility complex on bacterial composition of fecal flora
.
Infect. Immun.
 
69
,
2372
2377
.
[15]
Peltonen
R.
(
1994
)
Studies on faecal microecology with reference to diet, medication and rheumatoid arthritis
.
Ann. Univ. Turku
 
143
,
1
96
.
[16]
Moss
C.W.
Nunez-Montiel
O.L.
(
1982
)
Analysis of short-chain acids from bacteria by gas-liquid chromatography with a fused-silica capillary column
.
J. Clin. Microbiol.
 
15
,
308
311
.
[17]
Altman
D.G.
(
1996
)
Practical Statistics for Medical Research
 .
Chapman and Hall
,
London
.
[18]
Midvedt
T.
(
1987
)
Intestinal bacteria and rheumatic disease
.
Scand. J. Rheumatol.
 
64
,
49
54
.
[19]
Peltonen
R.
Kjeldsen-Dragh
J.
Haugen
M.
Tuominen
J.
Toivanen
P.
Förre
Ö.
Eerola
E.
(
1994
)
Changes of faecal flora in rheumatoid arthritis during fasting and one-year vegetarian diet
.
Br. J. Rheumatol.
 
33
,
638
643
.
[20]
Peltonen
R.
Nenonen
M.
Helve
T.
Hänninen
O.
Toivanen
P.
Eerola
E.
(
1997
)
Faecal microbial flora and disease activity in rheumatoid arthritis during a vegan diet
.
Br. J. Rheumatol.
 
36
,
64
68
.
[21]
MacDonald
T.T.
Monteleone
G.
Pender
S.L.F.
(
2000
)
Recent developments in the immunology of inflammatory bowel disease
.
Scand. J. Immunol.
 
51
,
2
9
.
[22]
Malchow
H.A.
(
1997
)
Crohn's disease and Escherichia coli. A new approach in therapy to maintain remission of colonic Crohn's disease
J. Clin. Gastroenterol.
 
24
,
653
658
.
[23]
O'Sullivan
M.A.
O'Morain
C.A.
(
1998
)
Nutritional therapy in Crohn's disease
.
Inflamm. Bowel Dis.
 
4
,
45
53
.
[24]
Artwohl
J.E.
Savage
D.C.
(
1979
)
Determinants in microbial colonization of the murine gastrointestinal tract: pH, temperature, and energy-yielding metabolism of Torulopsis pintolopesii
.
Appl. Environ. Microbiol.
 
37
,
697
703
.
[25]
McCarthy
D.M.
Jenq
W.
Savage
D.C.
(
1987
)
Mitochondrial DNA in Candida pintolopesii, a yeast indigenous on the surface of the secreting epithelium of the murine stomach
.
Appl. Environ. Microbiol.
 
53
,
345
351
.
[26]
Tieng
V.
Le Bouguenec
C.
Du Merle
L.
Bertheau
P.
Desreumaux
P.
Janin
A.
Charron
D.
Toubert
A.
(
2002
)
Binding of Escherichia coli adhesin AfaE to CD55 triggers cell-surface expression of the MHC class I-related molecule MICA
.
Proc. Natl. Acad. Sci. USA
 
99
,
2977
2982
.
[27]
Hooper
L.V.
Wong
M.H.
Thelin
A.
Hansson
L.
Falk
P.G.
Gordon
J.L.
(
2001
)
Molecular analysis of commensal host-microbial relationships in the intestine
.
Science
 
291
,
881
884
.
[28]
Klein
J.
(
1975
)
Histocompatibility system of the mouse
, p.
151
177
. In:
Biology of the Mouse Histocompatibility-2 Complex
  (
Klein
J.
Ed.), Volume 1.
Springer and Verlag
,
New York
.