In total, 191 fecal samples of the three genotypes (Pdx1-Cre, n=17; KC, n=19; KPC, n=28) collected at the three indicated time points were analyzed. Following sequencing and rarefying of library sizes, 1560 phylotypes were observed. The phylotypes belonged to 99 genera and 10 phyla, with sequences of Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria and Deferribacteres comprising approximately 99% of the total bacterial community (Supplementary data 1 and 2).
Similarities in the bacterial community assemblages
Formal comparisons between the global bacterial assemblages were performed using a three-way design to analyze the influence of the factors: mouse genotype (Pdx1-Cre vs. KC vs. KPC), age and gender. PERMANOVA revealed that all three factors had a statistically significant influence on the community structure on the phylotype level (genotype, pseudo-F=6.52, p=0.001; gender, pseudo-F=6.26, p=0.001; age, pseudo-F=4.53, p=0.001). This difference retained significance in all three parameters at low taxonomic ranks up to the family level. At higher taxonomic ranks, only age had a significant impact (Additional File 1, Supplementary Table 2). Moreover, a statistically significant interaction in the effects of genotype and gender on the variability of the assemblages was observed. This effect was visible up to the genus level. Otherwise, a clear interaction was seen in the effects of genotype and age up to the class level. These significant effects are also indicated in the non-metric multidimensional scaling (nMDS) plot (Figure 1).
Figure 1: Differences in global bacterial community structures. In Pdx1-Cre, KC and KPC mice of 5, 11 and 17 weeks of age (or date of death see Supplementary Table 1) the global bacterial community structure was assessed by non-metric multidimensional scaling (nMDS). The global community structure is based on standardized genus abundance data. The community structures influenced by genotype, age (1=5 weeks, 2=11 weeks and 3=17 weeks) and gender are indicated. Gender: M=male, F=female.
Pairwise tests revealed that the differences between Pdx1-Cre versus KC or KPC mice were more pronounced compared to differences between KC versus KPC mice (Table 1). During ageing the most striking changes were observed between mice of 5 and 11 weeks of age whereas further changes in older mice remained significant but were less pronounced.
Pairwise analysis of the interactions between the different factors revealed differences between female and male mice: For example, the communities were significantly different on the genus and family level in KC and KPC mice only (Additional File 1, Supplementary Table 3). Moreover, assemblages in Pdx1-Cre versus KC or KPC mice were significantly different at all analyzed ages, including 5 weeks of age when morphological differences within the pancreas were minimal. Assemblages between KC versus KPC mice were significantly different only after 17 weeks (Additional File 1, Supplementary Table 3).
Table 1. Factors influencing global community structures as indicated by PERMANOVA
|
|
Phylotype
|
Genus
|
Family
|
Order
|
Class
|
Phylum
|
Factor
|
Groups compared
|
t
|
p(MC)
|
t
|
p(MC)
|
t
|
p(MC)
|
t
|
P(MC)
|
t
|
p(MC)
|
t
|
p(MC)
|
Geno-type
|
KC, KPC
|
1.70
|
0.003
|
1.66
|
0.018
|
1.54
|
0.045
|
1.04
|
0.325
|
1.03
|
0.335
|
0.43
|
0.844
|
KC, Cre
|
2.73
|
0.001
|
2.38
|
0.001
|
2.28
|
0.006
|
1.68
|
0.054
|
1.69
|
0.040
|
0.65
|
0.666
|
KPC, Cre
|
3.12
|
0.001
|
2.71
|
0.001
|
2.60
|
0.002
|
1.33
|
0.146
|
1.34
|
0.156
|
0.79
|
0.502
|
Age
|
1, 2
|
2.23
|
0.001
|
2.67
|
0.001
|
2.91
|
0.001
|
3.27
|
0.001
|
3.28
|
0.001
|
4.86
|
0.001
|
1, 3
|
2.52
|
0.001
|
2.50
|
0.001
|
2.60
|
0.001
|
2.91
|
0.002
|
2.91
|
0.002
|
4.14
|
0.001
|
2, 3
|
1.48
|
0.011
|
1.80
|
0.009
|
1.86
|
0.011
|
1.30
|
0.158
|
1.31
|
0.146
|
0.84
|
0.504
|
Gender
|
M, F
|
2.50
|
0.001
|
2.62
|
0.001
|
2.55
|
0.001
|
1.64
|
0.043
|
1.60
|
0.053
|
1.24
|
0.202
|
The influence of genotype, age and gender on microbial structures was calculated by PERMANOVA (pairwise test). The t statistics and the Monte Carlo p-values are given for each pair of groups of the factors performed at different taxonomic levels (from phylotype to phylum). Age: 1=5 weeks, 2=11 weeks and 3=17 weeks (or date of death); gender: M=male, F=female.
The diversity of fecal communities depends on genotype, age and gender
Further differences in the microbial communities were indicated by one-way analysis of α-diversity measures (Supplementary data 2; Additional File 1, Supplementary Table 4). Here, phylotype richness was significantly higher in Pdx1-Cre compared to KPC mice (562 ± 49 versus 533 ± 83 phylotypes, p=0.0391), with KC mice communities adopting mean values (544 ± 59). Also, older mice showed a higher phylotype richness compared to younger animals (554 ± 82 versus 529 ± 60 phylotypes, p=0.036) and females had a higher phylotype richness than male mice (557 ± 52 versus 536 ± 77 phylotypes, p=0.0224). Significant differences in diversity were observed between female and male mice (Shannon index H= 4.301 ± 0.28 versus 4.192 ± 0.366, p=0.0207; Simpsons index 1-l = 0.9711 ± 0.014 versus 0.9580 ± 0.021, p=0.0363 (Additional File 1, Supplementary Table 4 and Supplementary Figure 2).
More detailed differentiations could be obtained through 2-way ANOVA analyses, which showed that diversity increased with age in Pdx1-Cre but not in KC or KPC mice (Figure 2 and Additional File 1, Supplementary Table 5) such that 17 weeks old Pdx1-Cre mice exhibited a significantly higher diversity compared to KC or KPC mice at the same age.
Figure 2: Fecal bacterial community diversity is dependent on genotype and age (left), on genotype and gender (middle) and age and gender (left). Diversity is indicated (from top to bottom) by total phylotype number, Shannon diversity (H’), Pielou’s evenness (J’) and Simpsons diversity (1-l), respectively. Statistically significant differences between animals with the same genotype (Pdx1-Cre, KC and KPC) but different age (1=5 weeks, 2=11 weeks and 3=17 weeks), or of animals with the same age, but distinct genotypes is indicated on the left. Statistically significant differences between distinct gender (M, F) of the genotypes (Pdx1-Cre, KC or KPC) are indicated in the middle. Statistically significant differences between animals of different age (1=5 weeks, 2=11 weeks or 3=17 weeks) and gender (M, F) ofPdx1-Cre miceare indicated on the right. Significance is shown as *p<0.05 or **p<0.01 or **p<0.05. The mean is indicated by + and the median by a black line. The box represents the interquartile range. The whiskers extend to the upper adjacent value (largest value = 75th percentile + 1.5 x IQR) and the lower adjacent value (lowest value = 25th percentile - 1.5 x IQR) and dots represent outliers. Gender: M=male, F=female.
In addition, 2-way analysis revealed that whereas in female mice, the fecal communities of different genotypes did not differ significantly in diversity, specifically male KPC mice showed a rather low diversity, which was significantly lower than that of female mice (Figure 2 and Additional File 1, Supplementary Table 6). As diversity increased with age in Pdx1-Cre mice, it was also analyzed if this increase was influenced by gender. As shown in Figure 2 and Additional File 1, Supplementary Table 7 female Pdx1-Cre mice showed a significant increase in fecal microbial community diversity (Shannon and Simpsons index) and evenness and a trend in phylotype richness (p=0.066) with increasing age whereas a significant increase in phylotype richness only was visible in male Pdx1-Cre mice.
The microbial community composition depends on genotype, age and gender
Differences in microbial community composition in the context of the different genotypes, age and gender were initially evaluated independently by one-way analyses using the Kruskal-Wallis test (factor genotype), the Friedmann test (factor age) or the Mann-Whitney test (factor gender) considering genus level taxa. Out of the 99 identified genera or genus level taxa (Supplementary data 2), 74 were observed with a prevalence of >20% in at least one subgroup and thus subject to a differential abundance analysis. Overall, 59 out of 74 genera (79.7%) were influenced by at least one of these factors as summarized in Additional File 1, Supplementary Figure 3.
Out of the 15 genera and genus level taxa mainly influenced by age, the strongest effect was observed for Turicibacter, where the mean abundance increased by one order of magnitude (Figure 3 and Supplementary data 4). Ten genera and genus level taxa were mainly influenced by genotype (Figure 3 and Supplementary data 5). The most pronounced effects were observed for Monoglobus, which was practically absent from KPC mice and showed roughly one order of magnitude lower abundance in KC compared to Pdx1-Cre mice. Likewise, Enterocloster, Vampirivibrio and Saccharibacteria were all found in Pdx1-Cre mice in a mean abundance at least twice or thrice that of KC or KPC mice. Genera influenced by gender were scarce: Only Ligilactobacillus and Acetatifactor seemed to be influenced by gender only, being slightly more abundant in female mice. However, for some genera that were also influenced by genotype such as Faecalibaculum (6-fold higher mean abundance in male mice) or by age such as Bifidobacterium (more than 10-fold higher mean abundance in male mice), marked gender influences were visible. Overall, 18 genera and genus level taxa were significantly influenced by gender (Figure 3 and Supplementary data 5). Fifteen genera and genus level taxa, respectively, were influenced by age and genotype and 10 were affected by all three factors. For example, three Bacteroidales genera, i.e. Bacteroides, Prevotella and Parabacteroides, showed similar effects with increasing abundances both with age and from Pdx1-Cre over KC to KPC mice, with higher levels in females compared to male mice.
Figure 3: Relative mean abundance of genera and genus level taxa. The mean abundance of genera and genus level taxa influenced by genotype (light red, Pdx1-Cre mice; medium red KC mice; dark red KPC mice), age (light green, 5 weeks; medium green, 11 weeks; dark green, 17 weeks) and gender (light blue, female; dark blue, male mice) is displayed as well as the standard error of mean. The taxa are sorted according to the factors influencing their abundance.
A phylogenetic analysis indicated that Bacteroidales were typically affected by all three factors and out of other phylogenetic groups, Erysipelotrichiaceae were prone to be influenced by gender (Additional File 1, Supplementary Figure 4). In contrast, Proteobacteria were mainly influenced by genotype (Figure 3 and Additional File 1, Supplementary Figure 4). Various Clostridiales genera were influenced by both genotype and age. However, whereas for example Anaerotruncus and Anaerotignum of the Ruminococcaceae and Lachnospiraceae were of lower abundance in Pdx1-Cre mice, Monoglobus and Enterocloster from the same families exhibited a higher abundance in Pdx1-Cre mice. Therefore, the influence of the genotype was clearly observed at the genus level, but to a lesser extent at higher phylogenetic levels (Figure 4). Similarly, genotype effects were observed for various genera and families inside the Bacteroidales, but not at class level for the Bacteroidales itself (Figure 4).
Figure 4: Phylogenetic taxa (genera, families, order, classes and phyla) influenced by genotype, age and gender. Factors significantly influencing the relative abundance are indicated in yellow (age), dark green (genotype) or magenta (gender) if p<0.01 and by orange (age), light green (genotype) or light magenta (gender) if p=0.01-0.05. A significant increase (p<0.01) with age or from Pdx1-Cre over KC to KPC mice is indicated by ▲, a decrease is indicated by ▼. A significant increase/decrease with p=0.01-0.05 is indicated by ▲ or ▼. The comparisons given are 5 weeks/11 weeks, 5 weeks/17 weeks and 11 weeks/17 weeks and Pdx1-Cre/KC, Pdx1-Cre/KPC and KC/KPC, respectively. In case of gender, the gender with the higher abundance is indicated as F (female) or M (male).
As possible interactions between the different factors could be expected, we performed three-way ANOVA analysis on the genus level. Here, the results of the one-way analysis could be largely confirmed. In 44 of the 46 cases (96%) where a comparison by one of the methods had indicated a significant difference in abundance to be caused by one factor with p<0.001, the second method also indicated a significant influence. In addition, in 73 of the 90 cases (81%) where a significant influence of one of the factors with p<0.01 was observed by one method, the second method gave concordant results (Figure 5).
Figure 5: Genera influenced by genotype, age and gender as indicated by one-way and three-way analyses. Factors influencing the relative abundance significantly are indicated in yellow (age), green (genotype) or magenta (gender). *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001; t (trend), p<0.1. Possible significant interactions between factors were evaluated by three-way analysis.
In a few cases, a significant interaction between factors was observed. For Bacteroides all three factors significantly influenced its relative abundance (Figure 5). The abundance increased from Pdx1-Cre over KC to KPC mice at nearly all time points and in both males and females (Figure 6). Furthermore, interactions were observed for Parabacteroidesand specifically Prevotella between genotype and gender (Figure 6).
Figure 6: Relative abundance of Bacteroides, Parabacteroides, Prevotella, Catabacter and Oscillibacter as influenced by genotype, age and gender. The left graph (except for Oscillibacter) shows abundances separately for each genotype, age and gender. In case of Oscillibacter, the left graph shows the relative abundances for the different genotypes and gender over all ages. The right graph always shows the relative abundances for the different genotypes and gender over all ages. The mean is indicated by + and the median by a black line. The box represents the interquartile range. The whiskers extend to the upper adjacent value (largest value = 75th percentile + 1.5 x IQR) and the lower adjacent value (lowest value = 25th percentile - 1.5 x IQR) and dots represent outliers. F, female; M, male.
There was a pronounced effect of genotype on the abundance of Parabacteroides in female, however, only a poor effect in male mice. In case of Prevotella, a higher abundance was observed in female KC and KPC compared to Pdx1-Cre mice, whereas no such effect was visible in male mice.
A strong interaction was also observed between gender and genotype for the abundance of Catabacter: Whereas female mice of the different genotypes harbored similar abundances, the abundance of this genus was dramatically higher in male Pdx1-Cre compared to male KC or KPC mice. In the Oscillibacter genus strong interactions were visible between genotype and gender but even more drastically between genotype and age. As shown in Figure 6 only male but not female Pdx1-Cre mice showed higher abundances compared to KC or KPC animals and only in aged mice an effect of the genotype was visible.
Differentially distributed species level taxa
Three-way analyses have also been performed to evaluate significant differences in distribution of species level taxa. This has been done specifically to distinguish between different Bacteroides, Duncaniella, Parabacteroides and Prevotella species, as well as between different groups of unclassified Eggerthellaceae, unclassified Lachnospiraceae, unclassified Muribaculaceae and unclassified Ruminococcaceae (see Additional File 1, Supplementary Figure 5). In case of the Bacteroides genus nine of the clusters of similar sequence were observed in >20% of the samples and most of them were influenced by age, genotype and gender, with increasing abundance with age, increasing abundance from Pdx1-Cre over KC to KPC mice and a higher abundance in female mice indicating that most Bacteroides spp. behaved similarly. In addition, Parabacteroides diastonis showed a very similar behavior (Additional File 1, Supplementary Figure 6 and Supplementary data 7).
Generally, only in a few cases the 16S rRNA gene sequence gave an indication of the harboring species and frequently, closely related isolates of a defined species were not available. For example, 15 prevalent sequence types indicating the presence of different Duncaniella species were observed. These bacteria, with only three species being available so far (D. muris, D. dubosii, D. freteri), exhibited clearly different behavior regarding abundance depending on gender, genotype and age. Different behavior was also observed for Anaerotignum where the abundance of the major Anaerotignum cluster (Anaerotignum 1) decreased with age whereas the abundance of two other prevalent Anaerotignum clusters (cluster 4 and 5) increased with age. Obviously, different species of these genera have very distinct characteristics enabling them to react differently to environmental changes.
Similar observations were made for bacterial families, which could not be further classified to the genus or species level. A distinct behavior was evident for the 36 clusters of the unclassified Lachnospiraceae regarding their response to age and genotype. However, only a minority (three clusters) was influenced by gender. This contrasted with Bacteroidales for which various clusters were affected by gender. Analysis on the species level gave further insights into interactions between the analyzed factors, specifically on differences in gender-dependent genotype effects. Such gender-specific effects were observed for several of the analyzed taxa (Figure 7).
Figure 7: Relative abundance of Duncaniella duboisii, Alistipes 2, Lachnospiraceae 29, Bacteroides 3, Lachnospiraceae 27 and Duncaniella 9 as influenced by genotype, age and gender. For Duncaniella duboisii, Alistipes 2, Lachnospiraceae 29, Bacteroides 3,the left graph shows abundances separately for each genotype, age and gender, whereas the right graph shows the relative abundances for the different genotypes and gender over all ages. For Lachnospiraceae 27 and Duncaniella 9 the graphs show the relative abundances for the different genotypes and gender over all ages. The mean is indicated by + and the median by a black line. The box represents the interquartile range. The whiskers extend to the upper adjacent value (largest value = 75th percentile + 1.5 x IQR) and the lower adjacent value (lowest value = 25th percentile - 1.5 x IQR) and dots represent outliers. F, female; M, male.
Duncaniella duboisii belongs to the most abundant species observed here, with a mean abundance of 3%. In female mice its abundance decreased significantly from Pdx1-Cre over KC to KPC mice. However, this was not the case in male KC and KPC mice that showed an increased abundance. Similarly, a pronounced genotype effect was seen in female mice in which Alistipes 2 increased significantly, in contrast to male mice. Even more strikingly the abundance of Lachnospiraceae 29 increased from Pdx1-Cre over KC to KPC in female, but decreased in male mice. Of note, Bacteroides 3 was absent from all Pdx1-Cre animals with a reasonable abundance in KPC animals and in female KC animals only. Various species level taxa showed an interaction between the effects of genotype and age or between gender and age (Figure 7). As an example, Lachnospiraceae 27 bacteria were observed with a high abundance in Pdx1-Cre mice only with a age-dependent increase. Duncaniella 9 showed a constant abundance with age in Pdx1-Cre mice but decreased in abundance with age in KC and KPC mice.