Alternative stable states in the intestinal ecosystem: proof of concept in a rat model and a perspective of therapeutic implications

Abstract

Background: Chronic immune-mediated diseases are rapidly expanding and notoriously difficult to cure. Altered relatively stable intestinal microbiota configurations are associated with several of these diseases, and with a possible pre-disease condition (more susceptible to disease development) of the host-microbiota ecosystem. These observations are reminiscent of the behavior of an ecosystem with alternative stable states (different stable configurations that can exist under identical external conditions), and we recently postulated that health, pre-disease and disease represent such alternative states. Here, our aim was to examine if alternative stable states indeed exist in the intestinal ecosystem.

Results: Rats were exposed to varying concentrations of DSS in order to create a wide range of mildly inflammatory conditions, in a context of diet-induced low microbiota diversity. The consequences for the intestinal microbiota were traced by 16S rRNA gene profiling over time, and inflammation of the distal colon was evaluated at sacrifice, 45 days after the last DSS treatment. The results provide the first formal experimental proof for the existence of alternative stable states in the rat intestinal ecosystem, taking both microbiota and host inflammatory status into consideration. The alternative states are host-microbiota ecosystem states rather than independent and dissociated microbiota and host states, and inflammation can prompt stable state-transition. Based on these results, we propose a conceptual model providing new insights in the interplay between host inflammatory status and microbiota status. These new insights call for innovative therapeutic strategies to cure (pre-)disease.

Conclusions: We provide proof of concept showing the existence of alternative stable states in the rat intestinal ecosystem. We further propose a model which, if validated in humans, will support innovative diagnosis, therapeutic strategy, and monitoring in the treatment of chronic inflammatory conditions. This model provides a strong rationale for the application of combinatorial therapeutic strategies, targeting host and microbiota rather than only one of the two in chronic immune-mediated diseases. Video Abstract.

Keywords: Alternative stable states; Cure; GI tract; Host; Inflammation; Interaction; Intestinal ecosystem; Microbiota; Symbiosis; Therapy.

Fig. 1.

Experimental design. a Alternative stable states. Subpanel I, alternative stable states of an ecosystem as beads in a stability landscape. Dashed line, frontier between two basins of attraction. Subpanel II, alternative states (solid lines) can both exist under a range of identical conditions (bi-stable range). Dashed line, see subpanel I. Width and shape of the basins of attraction, and thereby the stability of the alternative states, change with changing conditions, as illustrated by the changing distances between solid lines and the dashed line [14]. When changing conditions push the ecosystem beyond a tipping point (sharp bend in the Z-shape curve), the limit of resilience where the basin of attraction of its present state disappears, it rapidly transits to an alternative state. Subpanel III, steep gradient where for any given condition only one stable state exists. Assuming that the original ecosystem state is represented by the red dot, models II and III both predict a change of ecosystem state when the external conditions change from c1 to c2. When the conditions change back to c1, model II predicts that the system remains in the alternative state (red triangle), while in model III the system returns to its original state. b Timeline of the experiment. Black and blue solid lines, different diets as indicated in the text (diet shift at T-31). Small arrows, fecal sampling time-points. DSS, DSS treatment periods (3 days each). T, time in days relative to start of first DSS treatment (T0). Reception of animals at T-58, sacrifice and distal colon histology at T75. c Reduction of microbiota diversity after diet shift (T-31). Time-points are indicated at the bottom of the figure. Each dot represents one rat. d Induction of low-grade inflammation through DSS treatments. Distal colon histology, 45 days after last DSS treatment. Size bar, 250 μm

Fig. 2.

Two microbiota states. a Principal Coordinates analysis (PCoA) of ecological divergence between microbiota samples (Jensen-Shannon Divergence), based on OTU data aggregated at genus level. Each dot represents one intestinal microbiota sample. The analysis includes the data from all experimental groups, for T-7 up to T68 (n = 562 samples; cf Table 1). The dashed red line corresponds to the separation between the two normal distributions in the inset (frequency distribution of PCoA1 using the same samples) and in c. b Subsets of the data in a, for three experimental groups at T-1 and T68, respectively. Colors represent treatment groups as indicated. c Bimodal distribution of microbiota status. PCoA1 coordinates from the ordination plot in a as a measure of microbiota status are divided in categories with a range of 0.01, and the frequency of occurrence of each category is plotted. Top left: combined data from all groups (0% up to 3% DSS) for T-7, T-1, T63, and T68 (n = 232 samples). Bimodal graph overlay (density) and coloration according to the results of finite Gaussian mixture modeling using Mclust 5.4 [18]. Blue and red represent basal state and alternative state microbiota, respectively. Other plots: data for groups (% DSS) and time-points as indicated at the left and at the top of the figure, respectively. Coloration according to the bimodal distribution in the top left plot. Scale of the vertical axis is adjusted for each plot individually. *, bimodal (mixed normal) distributions according to Mclust 5.4 [18] when analyzing per-plot data (maximal value of the Bayesian Information Criterion (BIC), corroborated by bootstrap sequential Likelihood Ratio Testing (LRT): a p = 0.001 (cf Additional Table 3); b p = 0.004; c p = 0.032; d p = 0.027; e p = 0.01; f p = 0.04; g p = 0.03)

Fig. 3.

Evolution of microbiota states and enterotyping-based clusters with time. a Percentage of rats with PCoA1-based basal state or alternative state microbiota (cf Fig. 2). b Percentage of rats with enterotyping-based cluster A, B, or C microbiota. For both panels, percentage in groups treated with 2% or 3% DSS as indicated in Fig. 1. Sampling points as indicated in Fig. 1. Total number of rats sampled is 20 for each time point, except for T27 (19 rats). T31 was omitted because of missing data (only 15 rats sampled)

Fig. 4.

Microbiota status versus host inflammatory status. a Microbiota status at T68 (PCoA1 coordinate from Fig. 2a), plotted against host inflammatory status at T75 (histological score of the distal colon (Additional Tables 4 and 5); higher scores indicate higher levels of inflammation). Histological scores are integers; jitter has been added to improve visibility of individual data-points. Each dot represents one rat (n = 58). Colors represent treatment groups as indicated. The side-panel shows the frequency distribution and density curve of alternative microbiota states from Fig. 2c (top left: all groups). The top-panel shows the frequency distribution of alternative host states from Additional Fig. 8a, to which a density curve was added according to the results of finite Gaussian mixture modeling using Mclust 5.4 [18]. b Relative frequencies of alternative microbiota states as a function of host inflammatory status (distal colon histology score). Relative frequencies are calculated from the data in panel a. When separately analyzing the limited numbers of observations in subgroups (same histology score H), bimodal (mixed normal) distributions could be confirmed for subsets H = 2 (p =0.04) and H = 6 (p = 0.05) (finite Gaussian mixture modeling using Mclust 5.4 [18] with bootstrap sequential Likelihood Ratio Testing (LRT)). *, one observation only; each of the other categories represents from 3 to 13 observations. For host status 0 to 8, microbiota state distribution strongly correlates with host status (linear regression, r² = 0.63, p = 0.01 (F-test)). c Relative frequencies of alternative host states as a function of microbiota status. Relative frequencies are calculated from the data in a. Microbiota status is expressed as PCoA1 coordinate (Fig. 2a) with binning in intervals of 0.02; the central value of each bin is indicated. *, one observation only; each of the other categories represents from 3 to 13 observations. For microbiota status categories − 0.09 to 0.03, host state distribution strongly correlates with microbiota status (linear regression, r² = 0.91, p = 0.0009 (F-test))

Fig. 5.

Model of the host-microbiota ecosystem. Microbiota status at T68 plotted against host inflammatory status at T75 (n = 58). The blue tentative Z-shape overlay characterizes alternative microbiota states (cf Fig. 1a). Continuous parts of the curve represent the cores of the alternative states, and are drawn in correspondence with the means of the frequency distributions in the side panel of Fig. 4a. The dashed diagonal represents the changing frontier between the two basins of attraction (a transition fold, see inset and Fig. 1a). A larger distance between a solid line and the dashed diagonal indicates a wider basin of attraction, resulting in a higher state stability as revealed by a higher relative frequency of observation (Fig. 4b). Approximate positions of the inflexion points (tipping points) are based on the frequency distributions of alternative microbiota states (Fig. 4b: change from two alternative states to one state). The red tentative S-shape overlay characterizes alternative host states. Continuous parts of the curve represent the cores of the alternative states, and are drawn in correspondence with the means of the frequency distributions in the top panel of Fig. 4a. Changing distances between solid lines and the dashed diagonal represent the observed changes in the relative frequencies of the two host states with microbiota status (Fig. 4c). Approximate positions of the inflexion points (tipping points) are based on the frequency distributions of alternative host states (Fig. 4c). Violet circles represent alternative stable states of the host-microbiota ecosystem (attraction points combining stable host and microbiota states); green circles represent fragile attraction points (less stable host and microbiota states, close to tipping points and frontiers between basins of attraction). Violet and green arrows illustrate predicted evolution of the ecosystem from different positions in the plot to the different attraction points (away from the dashed diagonals, towards the solid lines)

Fig 6.

Characterization of attraction points in the host-microbiota ecosystem model. a Stability of attraction points. For each of the quadrants 1A, 1B, 2A, and 2B from Fig. 5 the percentage of animals presenting a stable microbiota is indicated. Stable microbiota is defined as showing the same state (either basal state 1 or alternative state 2) at four consecutive sampling time-points (T31, T40, T63, T68; i.e., from just after the last DSS treatment onward). >, >>, and < signs are read from left to right and from top to bottom. *p = 0.003 (Fisher exact test). b Relative abundance of Akkermansia in different microbiota, host, and ecosystem states. Abundance is expressed as percentage of total number of sequence reads; median values for the rats in each state at T68 are represented. Statistically significant differences are indicated in red. Differences between ecosystem states (1A, 1B, 2A, 2B) were analyzed using a Kruskal-Wallis test with post hoc Dunn’s test and Holm correction for multiple comparisons; differences between microbiota states (1, 2) or host states (A, B) were analyzed using a Wilcoxon test with FDR adjustment. c Akkermansia, Bacteroides and Butyricimonas distributions in microbiota states 1 (m1) and 2 (m2). Combined data from T-7 to T68 (n = 463 samples for microbiota state 1, n = 99 samples for microbiota state 2 (cf Table 1)); each dot represents one intestinal microbiota sample. Abundance is expressed as number of sequence reads on a total of 38,000. Only genera for which the median abundances in the two microbiota states differ at least 1.2-fold with q < 0.05 (Wilcoxon test with FDR adjustment) are presented. d Spearman correlation (r) between Akkermansia, Bacteroides, and Butyricimonas abundances. q values, after Holm correction for multiple comparisons

Fig. 7

(Pre-)disease remediation strategies. Host-microbiota ecosystem model from Fig. 5. Yellow and green arrows show predicted requirements (solid lines) and outcomes (dashed lines) for disease remediation strategies based on host inflammatory status management (yellow line, bottom left), microbiota management (yellow line, top right), or combined host and microbiota management (green line). See main text for details