Engineered bacteria can function in the mammalian gut long-term as live diagnostics of inflammation

Abstract

Bacteria can be engineered to function as diagnostics or therapeutics in the mammalian gut but commercial translation of technologies to accomplish this has been hindered by the susceptibility of synthetic genetic circuits to mutation and unpredictable function during extended gut colonization. Here, we report stable, engineered bacterial strains that maintain their function for 6 months in the mouse gut. We engineered a commensal murine Escherichia coli strain to detect tetrathionate, which is produced during inflammation. Using our engineered diagnostic strain, which retains memory of exposure in the gut for analysis by fecal testing, we detected tetrathionate in both infection-induced and genetic mouse models of inflammation over 6 months. The synthetic genetic circuits in the engineered strain were genetically stable and functioned as intended over time. The durable performance of these strains confirms the potential of engineered bacteria as living diagnostics.

Figure 1. Engineering a tetrathionate responsive memory device in E.coli NGF-1 a)

A bacterial memory device, PAS638, was constructed in mouse commensal E. coli NGF-1. S. typhimurium ttrR/S and P_ttrBCA drive Cro ‘trigger’ expression to switch a phage lambda-based memory circuit. In the presence of tetrathionate, TtrS becomes phosphorylated, in turn phosphorylating TtrR, which activates expression through P_ttrBCA in anaerobic conditions. Cro protein expression switches memory ON, accompanied by lacZ reporter expression. b) Testing of in vitro memory showed. c) a dose-response curve of PAS638 (EC50: 0.38–0.85μM 95%CI) but no response of the triggerless control PAS637 strain. Graph shows individual values from 6 replicate colonies from 2 separate experiments, non-linear fit ± SEM. d) PAS638 and S. typhimurium 14028s (S. tm) bacteria were administered by oral gavage to mice one day after streptomycin treatment, and fecal samples were analyzed d2–5 post administration (green stars) e) showing specific response of PAS638 on day 4 and/or 5 (see also Supplementary Fig. 2a) when co-infected with S. typhimurium ΔttrR bacteria (n=7) but not control (n=7) or S. typhimurium wt bacteria (n=6). f) Cumulative LCN-2 levels in mice administered PAS638+ S. typhimurium ΔttrR were higher in mice with PAS638 response (green lines) than in those without (black lines). Graph shows plots for individual mice (dotted) and ON or OFF averages (solid lines). ** t-ratio(5) = 4.3, p=0.03 using multiple t-tests with Holm-Sidak multiple comparisons test and each timepoint analyzed individually without assuming a consistent SD. d2 (t-ratio(5) = 1.2, p = 0.3), d3 (t-ratio(5) = 2.5, p = 0.1) and d5 (t-ratio(5) = 2.3, p = 0.1) were not significantly different.

Figure 2. Tetrathionate-sensing of inflammation in vivo

a) PAS638 and S. typhimurium ΔttrR bacteria were administered to cybb−/− and C57Bl/6 control mice (n= 5 per group) by oral gavage a day following streptomycin pre-treatment. b) Elevated PAS638 switching in the presence of S. typhimurium ΔttrR indicated the presence of tetrathionate in both strains. Graph shows max. percent over 5 day experiment, means are marked. *p=0.01 and ** p = 0.03 using Kruskal-Wallis test with Dunn’s multiple comparison correction. A comparison between the two S.tm ΔttrR samples was also done, p >0.99. c) Mass spectrometry from cecal extracts confirmed that tetrathionate was raised in cybb−/− and C57Bl/6 mice co-infected with S. typhimurium ΔttrR, along with intermediate levels in cybb−/− uninfected controls. Graph shows total tetrathionate detected per whole cecum; means are marked. *p<0.0001, **p0.0008, ***p=0.002 and **** p=0.0009; F(3,15) = 42.92 using one-way ANOVA with Tukey’s multiple comparison correction. Comparison of both infected groups (p=0.996) and C57Bl/6 control with cybb−/− infected mice (p<0.0001) were performed but not shown. d) PAS638 was administered to retired breeder IL10^−/− mice raised in gnotobiotic and barrier SPF conditions and retired breeder C57Bl/6 control mice raised in SPF conditions. e) Indicator plating of fecal samples from C57Bl/6 and IL10^−/− mice (> 1 week post administration) showed elevated tetrathionate response in IL10^−/− mice. Blue values: average of 3–4 measurements from individual mice. Grey values: single measurement from individual mice. Means are marked ** p=0.009 between groups (including only blue values) using a two-tailed Mann-Whitney test. f) PAS638 stably colonized IL10−/− mice (n=10) even without streptomycin pre-treatment. Graph shows CFU counts for individual mice (dotted lines) and average of all mice (solid) g) Balb/c, C57Bl/6 and 129X1/SvJ mice were colonized with PAS638 and the bacterial memory state was analyzed (green star = fecal sample collected). h) Balb/c and C57Bl/6 mice showed no switching, while a subset of 129X1/SvJ mice showed elevated response. Graph shows averages from at >3 days for individual mice and means. i) PAS638 levels broadly correlated with increased LCN-2 levels in five 129X1/SvJ mice. Timepoints in which bacteria were more highly switched into the memory-on state (memory values >1% highlighted green) tended to correspond with higher LCN-2 levels (values >20ng/g highlighted red).

Figure 3. PAS638 does not accrue mutations in synthetic elements over long periods

a) 129X1/SvJ mice were colonized with PAS638 for a 200-day period, >1600 bacterial generations. Mice were housed together for days 1–105, and then separated into two cages. b) PAS638 titers remained detectable following establishment in the streptomycin pre treated gut, without further antibiotic selection. c) Function was evident by switching at timepoints throughout the 200-day experiment. d) Ex vivo functional tests and next-generation sequencing analysis of PAS638 colonies following 159 and 200 days colonization confirmed that synthetic circuits did not mutate in any way over this period. Only one mutational insertion in the Colicin-like plasmid was identified in two cohoused mice, m1 and m2. e) In vivo function was further confirmed through administration of PAS638 and S. typhimurium ΔttrR to C57Bl/6 mice (n=5 per group), using PAS638 colonies isolated after colonization of three separate 129X1/SvJ mice for 200 days. Pooled colonies from a single mouse (94 colonies), or all three mice (280 colonies) were tested. f) Infection specific PAS638 switching response confirmed retained functionality in vivo. Graph shows max. switching percent measured from d3–5 post–infection; means are marked. ** p=0.008 using a two-tailed Mann-Whitney test of each experiment separately.