Studying Escherichia coli in the mouse intes nal environment using 16S rRNA surveys Jon Lund Steffensen Master’s thesis project University of Copenhagen In collabora on with Statens Serum Ins tut and University of Rhode Island
April 4, 2014
1 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 2 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 3 / 36
Mo va on
E. coli is part of the intes nal microbial community along with many other bacteria, mostly anaerobes. ▶
It is hypothesized that E. coli locally associates with specific bacteria forming mutualis c rela onships.
4 / 36
Mo va on
E. coli is part of the intes nal microbial community along with many other bacteria, mostly anaerobes. ▶
It is hypothesized that E. coli locally associates with specific bacteria forming mutualis c rela onships.
▶
In addi on E. coli may be able to affect the global composi on of the microbial community.
4 / 36
Mo va on
E. coli is part of the intes nal microbial community along with many other bacteria, mostly anaerobes. ▶
It is hypothesized that E. coli locally associates with specific bacteria forming mutualis c rela onships.
▶
In addi on E. coli may be able to affect the global composi on of the microbial community.
▶
Different strains of E. coli may have dis nct effects on the surrounding microbial community.
4 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 5 / 36
Mouse intes ne
▶
The mouse intes ne is used as a model
▶
Treatment with streptomycin, then E. coli strain
▶
Samples from conven onal mice also included
6 / 36
Strains
Three different types of strains are used as treatment. MG1655 Commensal lab strain EDL933 Pathogenic strain (EHEC) Nissle1917 Probio c strain
7 / 36
Sample types
Three different types of samples are obtained. ▶
Mucus samples taken from live mice intes nal mucus scrape
▶
Laser capture microdissec on (LCM) samples also taken from intes nal mucus but selec ng local environment of E. coli
▶
In vitro samples taken from a system that is set up to mimic the intes nal environment
8 / 36
Sample types
Three different types of samples are obtained. ▶
Mucus samples taken from live mice intes nal mucus scrape
▶
Laser capture microdissec on (LCM) samples also taken from intes nal mucus but selec ng local environment of E. coli
▶
In vitro samples taken from a system that is set up to mimic the intes nal environment
In total, 96 samples of different types and treatments, sequenced 16S rRNA gene.
8 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 9 / 36
Sequence Prepara on Mothur so ware
Using so ware package Mothur [Schloss et al., 2009].
10 / 36
Sequence Prepara on Mothur so ware
Using so ware package Mothur [Schloss et al., 2009]. ▶
Align sequences
▶
Remove sequences that cannot be aligned properly or that are otherwise determined to contain errors
10 / 36
Sequence Prepara on Mothur so ware
Using so ware package Mothur [Schloss et al., 2009]. ▶
Align sequences
▶
Remove sequences that cannot be aligned properly or that are otherwise determined to contain errors
▶
Classify aligned sequences (taxonomic assignment)
10 / 36
Sequence Prepara on OTU-based distances
▶
Form opera onal taxonomic units (OTUs) (97%)
▶
Corresponding roughly to species level
11 / 36
Sequence Prepara on OTU-based distances
▶
Form opera onal taxonomic units (OTUs) (97%)
▶
Corresponding roughly to species level
▶
Calculate sample dissimilari es (two different measures)
11 / 36
Sequence Prepara on OTU-based distances
▶
Form opera onal taxonomic units (OTUs) (97%)
▶
Corresponding roughly to species level
▶
Calculate sample dissimilari es (two different measures)
▶
Provides es mate of sample richness
11 / 36
Sequence Prepara on Tree-based distances
▶
Form a tree of the sequences (FastTree).
12 / 36
Sequence Prepara on Tree-based distances
▶
Form a tree of the sequences (FastTree).
▶
Calculate sample distances (Unifrac).
12 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 13 / 36
Alpha-diversity
▶
Measure of intra-sample diversity
▶
Expecta on: More intensive sequencing should result in diminishing returns of new taxa
14 / 36
Alpha-diversity
▶
Measure of intra-sample diversity
▶
Expecta on: More intensive sequencing should result in diminishing returns of new taxa
▶
Rarefac on curve: How many unique taxa are found at a certain sampling effort? (Schloss et al., 2009)
14 / 36
Alpha-diversity Rarefac on curve (genus level)
LCM
In vitro
MJ058
100
120
Mucus
MJ048
80
MJ065 MJ070 MJ056 MJ055
60
Taxa
MJ042
MJ004
MJ002 MJ095 MJ034 MJ087 MJ088 MJ035 MJ032 MJ044 MJ037 MJ059 MJ005MJ036 MJ007 MJ039 MJ091 MJ023 MJ024 MJ003MJ090 MJ096 MJ031 MJ011 MJ019 MJ029 MJ009 MJ038 MJ040 MJ018 MJ001 MJ033 MJ008 MJ010 MJ089 MJ028 MJ022 MJ030 MJ020 MJ014 MJ027 MJ094 MJ021 MJ016 MJ041 MJ012 MJ086 MJ025 MJ015 MJ026 MJ017 MJ013
MJ049 MJ054 MJ047 MJ045 MJ050 MJ052 MJ046 MJ051
0
20
40
MJ053
MJ068 MJ064 MJ080 MJ067 MJ074 MJ079 MJ060 MJ061 MJ073 MJ063 MJ084 MJ057 MJ082 MJ066 MJ076 MJ072 MJ069 MJ075 MJ077 MJ071 MJ093 MJ078 MJ092 MJ081 MJ062 MJ083 MJ085
MJ043 MJ006
0
40000
80000
120000 0
40000
80000
120000 0
40000
80000
120000
Number of tags sampled
More richness in LCM and in vitro samples. 15 / 36
Alpha-diversity Rarefac on curve (OTU-based, 97%)
3000
Mucus
LCM
2500
MJ092 MJ091
MJ076
2000
MJ029 MJ020 MJ042
MJ070 MJ060
1500
MJ018MJ059 MJ004 MJ043 MJ012 MJ036 MJ024 MJ021 MJ039 MJ022 MJ026MJ096 MJ019 MJ090 MJ028 MJ044 MJ023 MJ006 MJ015MJ037 MJ010 MJ002 MJ030 MJ016 MJ087 MJ095 MJ003 MJ032 MJ007 MJ014 MJ094 MJ005MJ034 MJ040 MJ086 MJ013 MJ027 MJ038 MJ011 MJ088 MJ035 MJ033 MJ009 MJ017 MJ001 MJ041
1000
OTUs
In vitro
500
MJ008MJ025
MJ061
MJ067 MJ068 MJ074 MJ058 MJ078 MJ065 MJ077 MJ066 MJ072 MJ063
MJ050 MJ052 MJ046 MJ048 MJ051 MJ045 MJ047 MJ054
MJ056 MJ062 MJ075MJ057 MJ069 MJ093 MJ055 MJ085 MJ082 MJ084 MJ080 MJ073 MJ083 MJ071 MJ079 MJ064
MJ049
MJ053
MJ081
0
MJ089
0
40000
80000
120000 0
40000
80000
120000 0
40000
80000
120000
Number of tags sampled
No plateau. 16 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 17 / 36
Beta-diversity
▶
Measure of inter-sample diversity
▶
Visualized by ordina on
▶
3 sample types, 3 dissimilarity measures
▶
Choice of distance measure is important for interpreta on
18 / 36
Beta-diversity
▶
Measure of inter-sample diversity
▶
Visualized by ordina on
▶
3 sample types, 3 dissimilarity measures
▶
Choice of distance measure is important for interpreta on
▶
Expecta on: Clustering of samples of similar type and/or treatment
18 / 36
Beta-diversity All samples, Yue & Clayton, PCoA ordina on
0.2
0.4
In vitro LCM Mucus 0 days 3 days 5 days 7 days 10 days
MJ034
X2
MJ003
MJ023
MJ090
MJ067 MJ037 MJ048
MJ063 MJ022
MJ019 MJ015 MJ051 MJ007 MJ047 MJ075 MJ008 MJ026 MJ050 MJ046 MJ040MJ052 MJ058 MJ017 MJ045 MJ093 MJ065 MJ057 MJ069
MJ032 MJ086 MJ094
MJ066 MJ071 MJ084 MJ068 MJ073
MJ059 MJ030 MJ080
MJ010
MJ006 MJ021
−0.2
MJ085 MJ078 MJ087
MJ038 MJ011 MJ070 MJ077
−0.4
MJ002 MJ039
−0.4
MJ091 MJ096 MJ092
MJ033 MJ014
MJ036 MJ054 MJ079 MJ049 MJ076 MJ005 MJ053 MJ004
MJ055
−0.6
MJ088
MJ024 MJ001 MJ025
MJ028 0.0
MJ043 MJ042 MJ035 MJ027 MJ044 MJ029 MJ041 MJ018 MJ020 MJ095
−0.2
0.0
MJ009
MJ016
MJ013 MJ012 MJ061 MJ074MJ082 MJ031 MJ064 MJ083 MJ062 MJ060MJ056 MJ081 MJ072 MJ089 0.2
0.4
X1
No clear clustering based on sample type
19 / 36
Beta-diversity All samples, Jaccard index, NMDS ordina on MJ028 0.2
MJ017 MJ036 MJ051 MJ020MJ044 MJ008 MJ022 MJ043 MJ015 MJ034 MJ029 MJ023 MJ041 MJ019 MJ012 MJ045 MJ007 MJ054 MJ026MJ018 MJ042 MJ027 MJ049 MJ091 MJ035 MJ047 MJ024 MJ067 MJ005 MJ004 MJ087 MJ038 MJ059 MJ095 MJ040 MJ014 MJ025 MJ032 MJ003 MJ021 MJ013 MJ088 MJ030 MJ090 MJ006 MJ033 MJ092 MJ001 MJ096 MJ011 MJ010 MJ066 MJ016 MJ053 MJ082MJ002 MJ037MJ039 MJ046 MJ078 MJ061 MJ089 MJ077 MJ009 MJ074 MJ093 MJ072 MJ094 MJ070 MJ063 MJ086 MJ056 MJ050 MJ055 MJ083 MJ069 MJ065 MJ064 MJ048 MJ057 MJ058 MJ085 MJ081 MJ068 MJ079 MJ073 MJ080 MJ084 MJ075 MJ060 MJ071 MJ062
−0.4
−0.3
−0.2
−0.1
X2
0.0
0.1
MJ031
In vitro LCM Mucus 0 days 3 days 5 days 7 days 10 days −0.6
MJ0
MJ076 −0.4
−0.2
0.0
0.2
0.4
X1
Separa on of mucus and LCM, but this is expected given rarefac on 20 / 36
Beta-diversity Mucus samples, Yue & Clayton, PCoA ordina on
MJ028 MJ019 MJ007 MJ040 MJ008 MJ015
MJ023 MJ022
0.2
MJ017
MJ034
MJ032 MJ026
MJ086
MJ041
MJ094
MJ037 MJ005
0.0 X2 −0.2 −0.4
MJ020 MJ095 MJ018 MJ088
MJ036 MJ004
Conventional EDL933 MG1655 MG1655 flhD MG1655 mot1 MG1655 Oxygen− Nissle 1917 Control 0 days 5 days 10 days −0.4
MJ043 MJ042 MJ027 MJ035 MJ044 MJ029
MJ003
MJ006 MJ014
MJ001 MJ024 MJ025
MJ090 MJ096 MJ091
MJ033 MJ010 MJ087
MJ030 MJ059
MJ016
MJ002 MJ039
MJ021 MJ011 MJ038 MJ009
MJ089 MJ013 MJ031 MJ012 −0.2
0.0
0.2
0.4
X1
No clear clustering based on treatment
21 / 36
Beta-diversity In vitro samples, UniFrac, PCoA ordina on
MJ048
Nissle 1917 Control 3 days 7 days
0.2
MJ049
0.0
X2
0.1
MJ054
MJ051
−0.2
−0.1
MJ047MJ045 MJ046 MJ050 MJ052
−0.3
MJ053
−0.2
0.0
0.2
0.4
0.6
X1
Separa on of control and Nissle samples
22 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 23 / 36
Taxon abundance
▶
Which taxa are actually present?
▶
Is plo ed as rela ve abundance of each sample
24 / 36
Taxon abundance
▶
Which taxa are actually present?
▶
Is plo ed as rela ve abundance of each sample
▶
phylum and genus level are considered
▶
Only taxa above a threshold are shown
24 / 36
Taxon abundance
▶
Which taxa are actually present?
▶
Is plo ed as rela ve abundance of each sample
▶
phylum and genus level are considered
▶
Only taxa above a threshold are shown
▶
Can any obvious differences be found?
24 / 36
7
4
J0 9
6
J0 8
1
J0 4
0
J0 4
4
J0 2
8
J0 0
M
M
4
3
J0 4
J0 4
2
J0 4
0.2
0.0
0.4
0.4
0.4
0.4
0.4
0.8
0.8
0.8
0.8
0.8
0.6
0.6
0.6
0.6
0.6
1.0
1.0
1.0
1.0
1.0
Control
M
0.2
0.2
0.2
0.0
0.0
0.0
0.2
EDL933
M J M 003 J M 018 J M 025 J M 033 J M 034 J M 035 J M 036 J M 090 J M 091 J0 96
M
M
M
M
M
M
9
J0 0
J0 5
2
1
0
J0 3
J0 3
J0 3
2
1
J0 0
0.0
MG1655
M
M
M
M
M
M
8
J0 0
J0 2
7
6
7
J0 2
J0 2
J0 1
6
5
J0 1
J0 1
Nissle1917
M
M
M
M
M
M
M
Taxon abundance
Mucus samples (stacked bar plot, phylum level) Conventional
Proteobacteria Bacteroidetes Firmicutes Deferribacteres Actinobacteria Other
25 / 36
Taxon abundance
0.4 0.1
0.2
0.3
Nissle1917 MG1655 EDL933 Control Conventional
0.0
Relative abundance (±std.err.)
0.5
Mucus samples (rela ve abundance, genus level)
26 / 36
Taxon abundance In vitro (stacked bar plot, phylum level)
0.8 0.6
1.0
Control
Proteobacteria Bacteroidetes Firmicutes Deferribacteres Actinobacteria Other
54 J0 M
53 J0 M
49 J0 M
48 J0 M
52
51
J0 M
50
J0 M
J0 M
47 J0 M
M
J0
46
0.0
0.2
0.4
0.8
M
J0
45
0.0
0.2
0.4
0.6
1.0
Nissle1917
27 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 28 / 36
Sta s cal model
Since the explora ve analysis does not show any obvious major pa erns, a sta s cal model was tried to detect more subtle pa erns in the data. ▶
Model is implemented in R package mvabund
▶
Specifically designed for abundance data [Warton et al., 2012]
29 / 36
Sta s cal model
Since the explora ve analysis does not show any obvious major pa erns, a sta s cal model was tried to detect more subtle pa erns in the data. ▶
Model is implemented in R package mvabund
▶
Specifically designed for abundance data [Warton et al., 2012]
▶
Model is fi ed to raw abundance data of mucus samples
▶
ANOVA test determines if samples are drawn from pools that are sta s cally significantly different
29 / 36
Sta s cal model Results
▶
Model is a good fit to this data
30 / 36
Sta s cal model Results
▶
Model is a good fit to this data
▶
Sta s cally significant differences between treatments of Nissle1917, MG1655 and control (P = 0.001)
30 / 36
Sta s cal model Results
▶
Model is a good fit to this data
▶
Sta s cally significant differences between treatments of Nissle1917, MG1655 and control (P = 0.001)
▶
but abundance of E. coli itself seems to be responsible for the difference
30 / 36
Outline Introduc on Mo va on Obtained Samples Materials and Methods Sequence prepara on Results Alpha-diversity Beta-diversity Taxon abundance Sta s cal model Discussion Conclusions 31 / 36
Conclusions
▶
A method of sequence prepara on was established
▶
Some minor differences were shown but more data is needed
32 / 36
Outlook
▶
Sequence prepara on takes a long me there seems to be at least some room for improvement in so ware packages
▶
The mouse gut is a noisy environment and more samples may be needed to show a difference in treatments
▶
Abundance of E. coli needs to be handled
▶
The sta s cal model could be applied to other samples than mucus samples, and to sample types
33 / 36
Thanks
▶
Dr. Paul Cohen, URI
▶
Dr. Ying Zhang, URI
▶
Prof. Karen Krogfelt, SSI
▶
Prof. Anders Krogh, KU
34 / 36
16S rRNA Survey
▶
rRNA gene useful down to genus level
▶
Contains variable regions among highly preserved regions
▶
Regions have different u lity in different scenarios
35 / 36
16S rRNA Survey
▶
rRNA gene useful down to genus level
▶
Contains variable regions among highly preserved regions
▶
Regions have different u lity in different scenarios
▶
Prac cal considera ons: Established protocols and sequencing technology
35 / 36
Sta s cal model
▶
Uses raw abundance data
▶
Fits a generalized linear model to each taxon seperately
▶
Assumes nega ve binomial distribu on of abundance counts
36 / 36