Author response: Integration host factor regulates colonization factors in the bee gut symbiont Frischella perrara

Abstract

Full text Figures and data Side by side Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Bacteria colonize specific niches in the animal gut. However, the genetic basis of these associations is often unclear. The proteobacterium Frischella perrara is a widely distributed gut symbiont of honey bees. It colonizes a specific niche in the hindgut and causes a characteristic melanization response. Genetic determinants required for the establishment of this association, or its relevance for the host, are unknown. Here, we independently isolated three point mutations in genes encoding the DNA-binding protein integration host factor (IHF) in F. perrara. These mutants abolished the production of an aryl polyene metabolite causing the yellow colony morphotype of F. perrara. Inoculation of microbiota-free bees with one of the mutants drastically decreased gut colonization of F. perrara. Using RNAseq, we found that IHF affects the expression of potential colonization factors, including genes for adhesion (type 4 pili), interbacterial competition (type 6 secretion systems), and secondary metabolite production (colibactin and aryl polyene biosynthesis). Gene deletions of these components revealed different colonization defects depending on the presence of other bee gut bacteria. Interestingly, one of the T6SS mutants did not induce the scab phenotype anymore despite colonizing at high levels, suggesting an unexpected role in bacteria-host interaction. IHF is conserved across many bacteria and may also regulate host colonization in other animal symbionts. Editor's evaluation This fundamental work substantially advances our understanding of the genetic basis of how a very prevalent bee symbiont, Frischella perrara, colonizes the gut of these insects, by identifying novel players in this process and raising new questions related to their mode of action. The authors characterized spontaneous mutants in an important regulator, and showed that this regulator controls the expression of several genes required for gut colonization, they constructed deletion mutants on these genes and characterized these mutants both in vitro and in colonization assays in the presence and absence of other gut symbionts, and provide insights on the mode of action of the novel players identified during host colonization. The combination of approaches used is exceptional and established new standards in the field of host-microbe interactions aiming to understand the molecular players involved in the colonization of gut symbionts. https://doi.org/10.7554/eLife.76182.sa0 Decision letter Reviews on Sciety eLife's review process Introduction The digestive tract of many animals is colonized by specialized gut symbionts that occupy distinct physical niches and utilize diverse nutrients. Owing to the availability of genetic tools, gnotobiotic animal models, and multi-omics approaches, we can now study the genetic features that allow gut symbionts to colonize various animal hosts, including mammals, fishes, and insects (Wu et al., 2015; Lee et al., 2013; Townsend et al., 2019; Townsend et al., 2020; David et al., 2014; Nakajima et al., 2018; Powell et al., 2016a; Ohbayashi et al., 2015; Salem et al., 2017; Kikuchi et al., 2020). The Western honey bee, Apis mellifera, is a particularly interesting model to characterize colonization factors of bacterial gut symbionts due to its agricultural importance and the tractability of its gut microbiota. Honey bees harbor relatively simple yet highly specialized gut microbiota composed of 8–10 bacterial genera (Kwong and Moran, 2016; Bonilla-Rosso and Engel, 2018). The wide distribution of these communities across social bees suggests long evolutionary associations with the host (Kwong et al., 2017). Moreover, different members of the bee microbiota colonize distinct physical niches along the gut. Lactobacilli and Bifidobacteria predominate in the posterior hindgut (rectum), while Gammaproteobacteria (Frischella perrara and Gilliamella species) and a Betaproteobacterium (Snodgrassella alvi) preferentially colonize the anterior hindgut (ileum and adjacent pylorus, i.e., the transition zone between midgut and ileum) (Kwong and Moran, 2016; Engel et al., 2015a; Martinson et al., 2011; Powell et al., 2014). The partitioning of these bacteria into distinct gut compartments suggests the existence of specific bacterial and/or host mechanisms that facilitate colonization. Most bacteria of the bee gut microbiota can be cultured and experiments with gnotobiotic bees have been established (Emery et al., 2017; Raymann et al., 2018; Zheng et al., 2017; Brochet et al., 2021; Kešnerová et al., 2017). Moreover, genomic analyses have provided important insights about the functional potential of bee gut symbionts and their adaptation to the gut environment (Brochet et al., 2021; Ellegaard et al., 2019; Ellegaard and Engel, 2019; Engel et al., 2012; Kwong et al., 2014; Ludvigsen et al., 2018; Steele et al., 2017; Zheng et al., 2019). Yet, little is known about which genes are directly involved in establishing colonization in the bee gut and how these genes are regulated. The only symbiont that has been extensively studied in this respect is S. alvi. Using transposon sequencing and transcriptome analysis, Powell et al. determined genome-wide host colonization factors in S. alvi (Powell et al., 2016a). Most genes with strong fitness effects were found to belong to three major categories: extracellular interactions, metabolism, and stress response. In particular, genes for attachment and biofilm formation were highly beneficial for colonization, which is in agreement with the observation that this gut symbiont adheres to the host epithelium of the ileum and forms a multispecies biofilm with Gilliamella. The honey bee gut symbiont F. perrara belongs to the recently described family Orbaceae within the Gammaproteobacteria (Volkmann et al., 2010). It is taxonomically close to the bee gut symbiont Gilliamella and has a similar metabolism (Engel et al., 2013). However, compared to the other members of the bee gut microbiota, F. perrara shows a rather distinctive colonization phenotype. While S. alvi and Gilliamella have been reported to colonize the host epithelium of the entire ileum (Martinson et al., 2012; Leonard et al., 2018), F. perrara preferentially colonizes the transition zone between the midgut and the ileum, that is, the pylorus. Moreover, colonization with F. perrara leads to the appearance of a brown to black material on the luminal side of the epithelial surface between the cuticle layer of the host tissue and the adherent F. perrara cells (Engel et al., 2015a). This so-called scab phenotype forms after 5–7 d post-colonization and has so far not been reported to be triggered by any other gut symbiont than F. perrara. Transcriptome analysis of the host showed that F. perrara elicits a specific immune response that includes the upregulation of the host melanization pathway likely responsible for the formation of the scab phenotype (Emery et al., 2017). F. perrara is highly prevalent across worker bees and colonies of A. mellifera (Engel et al., 2015a; Kešnerová et al., 2020), and related bacteria have also been found in Apis cerana (Wolter et al., 2019). Moreover, between 25 and 80% of all worker bees of a colony harbor a visible scab phenotype in the pylorus region of the gut, which has been shown to strongly correlate with a high abundance of F. perrara (Engel et al., 2015a). However, the impact of these phenotype on the host has remained elusive. Genome sequencing of the type strain of F. perrara and comparison with other genomes of the Orbaceae family revealed the presence of several genomic islands that may be involved in the specific interaction of F. perrara with the host (Engel et al., 2015b). These include a biosynthetic gene cluster for the production of the genotoxic metabolite colibactin (Clb), two distinct type VI secretion systems (T6SSs) and associated effector proteins, type I secretion systems, and fimbrial low-molecular-weight protein (Flp) pili genes. However, currently no genetic tools are available for F. perrara precluding studies about the role of these genetic factors in gut colonization or the induction of the scab phenotype. Here, we report the isolation of a spontaneous mutant of F. perrara that possesses a strong colonization defect in vivo. Resequencing of the mutant revealed a single nonsynonymous point mutation in the gene encoding the alpha subunit of the DNA-binding protein integration host factor (IHF). Using a combination of gnotobiotic bee experiments, transcriptomics, and metabolite analyses, we characterized the genes regulated by IHF. We then established a gene-deletion strategy for F. perrara, which allowed us to knock out some of the IHF-regulated genes and show that they impact gut colonization and scab development to different extent in the presence and absence of a complex community. Results Isolation of spontaneous IHF mutants affecting growth and colony morphology of F. perrara Culturing F. perrara type strain PEB0191 (Engel et al., 2013) on modified tryptone yeast glucose (mTYG) agar resulted in the formation of yellow colonies. However, we occasionally observed the appearance of larger white colonies among the yellow ones (Figure 1A). Restreaking white colonies on fresh mTYG agar usually resulted in yellow colonies again. However, three white colonies that we identified in independent experiments did not change their appearance anymore, suggesting that we had isolated stable ‘white’ variants of F. perrara PEB0191 (Figure 1B, Figure 1—figure supplement 1). Genome sequencing of the white variants revealed the presence of three different non-synonymous point mutations in the genes encoding the IHF. IHF is a widely distributed DNA-binding protein consisting of the IhfA/B heterocomplex (Friedman, 1988; Freundlich et al., 1992; Goosen and van de Putte, 1995). Strikingly, two point mutations were identical to each other, but occurred in the different subunits of IHF (ihfA and ihfB), resulting in a proline to lysine change at amino acid positions 82 and 83, respectively (Figure 1C). The third point mutation resulted in a lysine to serine change at position 38 of IhfA (Figure 1C). Homology modeling showed that these amino acids are located in the region interacting with DNA, suggesting that the three mutations impact the DNA-binding properties of IHF (Figure 1D; Rice et al., 1996). As two of the isolated mutants occurred when generating gene deletions of F. perrara, they harbored additional genetic modifications (see ‘Materials and methods’). Therefore, we focused further characterization of IHF on the mutation Pro83Lys (hereafter ihfA*) that occurred in the wild type (wt) background of F. perrara PEB0191. While the ihfA* strain consistently formed larger colonies than the wt strain on mTYG agar (Figure 1A and B), there was no significant difference in growth in liquid culture (permutation test, p=0.097; Figure 1—figure supplement 2A). However, light microscopy showed that cells of the mutant strain were on average slightly longer than cells of the wt (Kolgomorov–Smirnov test p<0.0001, Figure 1—figure supplement 2B and C). Figure 1 with 2 supplements see all Download asset Open asset Isolation of a spontaneous ihfA* mutant of F. perrara displaying an altered colony morphotype. (A) Colonies of F. perrara PEB0191 (wt) after 48 hr of growth on modified tryptone yeast glucose (mTYG) agar. Arrowhead points at a larger white colony in between many yellow colonies. The area in the white square is magnified. (B) Colony morphology of F. perrara wt and the isolated white ihfA* mutant after growth on mTYG for 48 hr. (C) Protein sequence comparison of IhfA and IhfB of F. perrara wt and E. coli wt. The outlined positions refer to the residues mutated in the three spontaneous ihfA mutants: (i) lysine (Lys) to serine (Ser) at position 38 of F. perrara IhfA, (ii) proline (Pro) to lysine (Lys) at position 83 of F. perrara IhfA, and (iii) proline (Pro) to lysine (Lys) at position 82 of F. perrara IhfB. Note that the numbers given on top of the alignment refer to alignment positions and not to positions in the individual sequences. Secondary structures are depicted above as ribbons (α-helix) and arrows (β-sheet) and are numbered according to their appearance in the protein and the structure shown in (D). (D) Three-dimensional structure of E. coli IhfA/B heterocomplex with DNA (source protein databank NDB: PDT040). DNA is depicted in blue and IhfB in dark gray. IhfA is colored according to secondary structure: α-helix orange, β-sheet light blue, and the rest in light gray. α-helices and β-sheets are numbered. The mutated Pro83 and Lys38 residues of F. perrara IhfA and the Pro82 residue of IhfB are marked with an orange and green circle, respectively. F. perrara produces an aryl polyene secondary metabolite that is responsible for the yellow colony morphotype F. perrara PEB0191 encodes a genomic island that is homologous to aryl polyene (APE) biosynthetic gene clusters present in other Gammaproteobacteria (Figure 2A; Cimermancic et al., 2014). APEs are polyunsaturated carboxylic acids conferring a yellow pigmentation to bacterial cells (Goel et al., 2002; Poplawsky et al., 2000). To assess whether F. perrara wt, but not the ihfA* mutant, produces an APE, we analyzed cell extracts of both strains by liquid chromatography coupled to heated electrospray ionization high-resolution mass spectrometry (HPLC-HES-HRMS). The data revealed a strongly UV-Vis-absorbent ion peak at m/z 323.1647 [M+H]+, which had a suggested molecular formula of C21H23O3 (Figure 2B and C). In the ihfA* mutant, this ion was only present at trace amounts (Figure 2C). To characterize the metabolite in greater detail, a larger pellet of F. perrara wt cultures was extracted and purified by several HPLC runs. Mass spectrometry- (MS) and UV-Vis-guided fractionation yielded an enriched extract that was analyzed by nuclear magnetic resonance (NMR) spectroscopy. The characteristic ions detected in MS-MS fragmentation experiments (Figure 2—figure supplement 1), the UV-Vis spectrum with an absorption maximum at 415 nm (Figure 2—figure supplement 2) in conjunction with NMR data (Figure 2—figure supplements 3–10), suggest an aryl polyene structure identical to that reported in Cimermancic et al., 2014; Schöner et al., 2016; Figure 2D. Unfortunately, it was not possible to connect the NMR substructures because the central methines could not be assigned to chemical shifts (Figure 2D, Figure 2—figure supplement 10). Comparison of the organic extracts of F. perrara wt and Escherichia coli CFT073 provided further evidence that both produce the same compound (Figure 2—figure supplement 11). Combined, these results suggest that the APE pathway is responsible for the yellow color of the wt colonies of F. perrara and is suppressed in the ihfA* mutant. Figure 2 with 11 supplements see all Download asset Open asset Metabolite analysis of F. perrara wt and the ihfA* mutant. (A) Comparison of gene synteny and sequence similarity of the genomic islands of F. perrara PEB0191 (top) and E. coli CFT073 (bottom) encoding the aryl polyene (APE) biosynthesis genes. Gray lines indicate homologous regions based on tblastx analysis. Plots were generated with genoplotR (Guy et al., 2010). Transcripts per million (TPMs) are shown on top of the genomic island for one RNAseq replicate of each F. perrara wt (blue) and the ihfA* mutant grown in vitro. Coverage plots were generated with the Integrated Genome Browser v9 (Freese et al., 2016). (B) Total ion chromatogram (TIC) and UV trace (λ = 420 nm) of wt and ihfA*. A peak highly abundant in the wt was discovered at 18.85 min. Its high UV absorbance at λ = 420 nm indicated a conjugated carbon double bond system. (C) The normalized mass spectrum at 18.85 min reveals the ion m/z = 323.1647 Da to be approximately 50-fold more abundant in the wt compared to ihfA*. (D) Enrichment of the ion containing fraction by HPLC followed by nuclear magnetic resonance (NMR) experiments suggest a structure identical to that reported by Cimermancic et al., 2014. Reported (red) and observed (blue) 1H and 13C chemical shifts are shown. Central methines could not be assigned. The ihfA* mutant of F. perrara has a colonization defect and does not cause the scab phenotype As APEs have been shown to increase protection from oxidative stress and contribute to biofilm formation (Cimermancic et al., 2014; Schöner et al., 2016), we sought to test whether the ihfA* mutation impacts bee gut colonization. We mono-associated microbiota-free bees with either F. perrara wt or ihfA*. Colonization with the wt strain resulted in a visible scab in 50 and 80% of all bees after 5 and 10 d of colonization, respectively (n = 18 and n = 36 for both treatments for day 5 and day 10, respectively, Figure 3A and B). In contrast, none of the bees colonized with the ihfA* mutant developed a visible scab phenotype. To determine whether this difference was due to a general colonization defect of ihfA*, we quantified the colonization levels of F. perrara at day 5 and day 10 post colonization using colony-forming units (CFUs). While there was a trend towards lower colonization levels (fewer CFUs and more bees without detectable colonization) for the ihfA* mutant at day 5 post colonization, the difference was not statistically significant (Figure 3C, Wilcoxon rank-sum test p-value = 0.076). However, at day 10 post-colonization, bees colonized with the wt strain showed significantly higher CFUs than the ihfA* mutant (Figure 3C, Wilcoxon rank-sum test p-value<0.0001). In fact, in 50% of all bees (n = 36) the colonization levels of ihfA* were below the detection limit of 500 CFUs (Figure 3C). Figure 3 with 1 supplement see all Download asset Open asset F. perrara ihfA* mutant displays a colonization defect. (A) Light microscopy pictures of pylorus region of bees colonized with F. perrara PEB0191wt or ihfA* 10 d post colonization. (B) Quantification of scab phenotype of bees 5 and 10 d post colonization with n = 18 and n = 36 per treatment, respectively. (C) Quantification of colonization levels is measured by colony-forming units (CFUs) at day 5 (n = 18) and day 10 (n = 36) post colonization. Wilcoxon rank-sum test was used to assess significant differences. (D) Time-course experiment of bees colonized with F. perrara wt or ihfA*. Colonization levels were measured by CFUs every second day until day 10 and then at day 14 and day 21. n = 12 bees per time point per treatment. Wilcoxon rank-sum test was used to assess significant differences per time point. Error bars represent median and interquartile range. Data from three independent experiments. *p<0.05, **p<0.01, ***p<0.001. Figure 3—source data 1 contains the numeric values for the figures shown here. Figure 3—source data 1 Numeric data underlying the results shown in Figure 3 and Figure 3—figure supplement 1. https://cdn.elifesciences.org/articles/76182/elife-76182-fig3-data1-v1.xlsx Download elife-76182-fig3-data1-v1.xlsx As the quantification of F. perrara was based on CFUs obtained for whole-gut tissue, we carried out a second experiment, in which we specifically assessed the colonization levels in the pylorus and the ileum region of the honey bee gut, using both CFUs and quantitative PCR (Figure 3—figure supplement 1). The results were comparable to those obtained for the whole gut: at day 10 post colonization, there was a significant difference in the colonization levels of the wt and ihfA* in both the pylorus and the ileum (Wilcoxon rank-sum test p-value<0.05). To obtain a better understanding of the colonization dynamics of F. perrara wt and the ihfA* mutant, we conducted a third gnotobiotic bee experiment in which we inoculated microbiota-free bees with one of the two strains and followed the colonization levels over 12 time points from day 0 (i.e., 4 hr post inoculation) until day 22 post inoculation (Figure 3D). From the first time point at 4 hr post inoculation until day 4 post inoculation, the bacterial levels were below the detection limit (i.e., below 500 CFUs) in both conditions. Between day 4 and day 8 post inoculation, the abundance of the wt increased rapidly to about 106 CFUs per gut and then steadily further to 107 CFUs per gut until the last time point. In contrast, the levels of the ihfA* mutant remained low until day 10 post colonization and reached on average no more than 105 CFUs per gut until the last time point at day 22 post colonization. Notably, while we had used the same optical density of the two strains for colonizing microbiota-free bees, dilution plating revealed that there were fewer CFUs in the inocula for the wt compared to the ihf* mutant. Despite these differences, the wt colonized much better than ihfA*. In summary, these results show that ihfA* has a strong colonization defect. It has a delayed colonization dynamics compared to the wt, does not reach the same bacterial loads, and does not cause the scab phenotype, even though the bees were inoculated with more viable cells of ihfA* than the wt. Genes involved in symbiotic interactions are upregulated in F. perrara wt relative to the ihfA* mutant IHF may not have a direct effect on gut colonization, but rather regulate the gene expression of host colonization factors. To test this, we assessed the transcriptional differences by RNA sequencing (RNAseq) between the wt and ihfA* mutant when grown in vitro. We found that 358 out of 2337 genes encoded in the genome of F. perrara were differentially expressed with a log2-fold change >|2| between the two strains (Fisher’s exact test with p<0.05 and false discovery rate (FDR) <5%). Of those, 237 and 121 genes were up- and downregulated in F. perrara wt versus ihfA*, respectively (Figure 4A and B, Figure 4—source data 1). Among the genes upregulated in the wt, ‘Intracellular trafficking, secretion, and vesicular transport’ (COG U), ‘Extracellular structure’ (COG W), ‘Lipid transport and metabolism’ (COG I), ‘Mobilome: prophages and transposases’ (COG X), and ‘Secondary metabolites biosynthesis, transport and catabolism’ (COG Q) were significantly enriched (Fisher’s exact test, BH-adjusted p-value<0.01; see Figure 4—source data 2). Genes belonging to these three categories include different subunits and effectors of the two T6SSs of F. perrara, the Clb biosynthesis gene cluster, various components of the Flp pili, and an RTX (repeats in toxin) toxin belonging to the type I secretion system family (Figure 4—source data 1). Also, the genes of the APE biosynthesis gene cluster were among the upregulated genes, which is in line with the production of the corresponding metabolites in the wt but not in the mutant strain (Figure 2). Interestingly, a relatively large proportion of the upregulated genes encoded hypothetical or poorly characterized proteins. In fact, genes without COG annotation were also enriched relative to the entire genome of F. perrara (Fisher’s exact test, BH-adjusted p-value<0.01; see Figure 4—source data 2). Many of the upregulated genes were organized in genomic islands, with the largest one including the biosynthesis gene cluster of Clb and many hypothetical protein-encoding genes (Figure 4C). T6SS and Clb biosynthesis genes were among the genes with the highest fold changes relative to ihfA* mutant (Figure 4B, 28 of 32 genes with log2-fold change > 6). Moreover, 64% of the upregulated genes (152/237) belonged to the F. perrara-specific gene content as based on our previously published genome comparison of F. perrara PEB0191 with four other strains of the family Orbaceae (three of the genus Gilliamella and one of the genus Orbus, Figure 4—source data 1; Engel et al., 2015b). Figure 4 Download asset Open asset Differential gene expression between F. perrara wt and ihfA* mutant during in vitro growth. (A) Chromosomal localization of all genes significantly differentially expressed (2-log fold change = |2|, Fisher’s exact test p-value <0.05, false discovery rate [FDR] < 0.05) between F. perrara wt and the ihfA* mutant. Starting from outside, the first circle shows the scale of the genome representation of F. perrara in gray and white steps of 100 kb. The second and third circles (gray) depict the genes on the plus and minus strands of F. perrara. The fourth (beige) and fifth (light blue) circle depicts genes upregulated and downregulated in wt compared to ihfA*. Genomic islands are highlighted by coloration. (B) Bar plot of the genes differentially expressed between F. perrara wt and ihfA* with a log2-fold change > 2 (Fisher’s exact test p-value<0.05, FDR <0.05). (C) Comparison of the transcriptional profile of the genomic location encoding the colibactin biosynthetic gene cluster between F. perrara wt and the ihfA* mutant. Transcripts per million were visualized using the Integrative Genome Browser (Freese et al., 2016). The colibactin operon is schematically depicted below (green arrows). Figure 4—source data 1 and Figure 4—source data 2 contain the data used to produce the figure shown here. Figure 4—source data 1 List of F. perrara genes differentially expressed in vitro. https://cdn.elifesciences.org/articles/76182/elife-76182-fig4-data1-v1.xlsx Download elife-76182-fig4-data1-v1.xlsx Figure 4—source data 2 Clusters of orthologous groups for genes differentially expressed in vitro. https://cdn.elifesciences.org/articles/76182/elife-76182-fig4-data2-v1.xlsx Download elife-76182-fig4-data2-v1.xlsx Among the 121 downregulated genes, only COG category O (‘Posttranslational modification, protein turnover, chaperones’) was statistically enriched (Fisher’s exact test, BH-adjusted p-value<0.01, see Figure 4—source data 2). Moreover, only a small fraction (12%) belonged to the ‘F. perrara-specific genes,’ and fewer genes were organized into genomic islands. A more detailed inspection of the annotation revealed that a large number of the downregulated genes were involved in transport and metabolism (40 genes), transcriptional regulation (10 genes), and protein folding (8 genes), highlighting clear differences in the functional roles of the up- and downregulated genes. The two genes with the highest fold change (log2-fold change < -5) both encoded transcriptional regulators. One of them, dksA (Fpe_01158), is located upstream of the mrsA/mrsB antioxidant system (Fpe_01159 to Fpe_01162), which was also among the downregulated genes. The other one is part of the two-component regulator system basS/basR (Fpe_02097 and Fpe_02098), which has been reported to act as an iron- and zinc-sensing transcriptional repressor and activator in E. coli (Lee et al., 2005; Hantke, 2001). Taken together, these results show that many accessory genes known to be involved in symbiotic interactions (colibactin, Flp pili, T6SS) are upregulated in F. perrara wt as opposed to ihfA*, providing a list of candidate genes responsible for the colonization defect of the ihfA* mutant. T6SS, pili, APE biosynthesis, and Clb biosynthesis genes are expressed during bee gut colonization To test whether the genes upregulated in vitro in the wt relative to ihfA* were expressed in vivo, we determined the transcriptome of F. perrara wt at day 5 and day 10 post colonization. A total of 260 (149 up and 111 down) and 298 (162 up and 136 down) genes were differentially expressed at day 5 and day 10 post colonization relative to growth in vitro (log2-fold change >|2|, quasi-likelihood F-test with p<0.05 and FDR < 5%, Figure 5—source data 1). There was a considerable overlap of the differentially regulated genes between the two time points (115 and 80 shared up- and downregulated genes, respectively). At both time points, the COG category ‘Carbohydrate transport and metabolism’ (COG G) was significantly enriched among the genes upregulated in vivo relative to the entire genome (Figure 5—source data 2). In addition, at time point day 10, also the COG category (P) ‘Inorganic ion transport and metabolism’ was enriched (P adj<0.01, Fisher’s exact test, Figure 5—source data 2). Genes belonging to these two categories encoded transporters for different sugars (Phosphotransferase systems), iron, and transferrin (Figure 5—figure supplement 1; Figure 5—source data 1). In addition, a catalase gene and several genes for the biosynthesis of the amino acid tryptophan were upregulated at both time points. However, only 14 and 19 genes of those upregulated in vitro in the wt relative to the ihfA* mutant (see Figure 5) were also upregulated in vivo at day 5 and day 10 post-colonization, respectively (Figure 5—source data 1). This was expected because the in vitro RNAseq analysis had shown that these genes are already expressed in the wt when grown on mTYG agar, which we used as a reference condition for the in vivo analysis. Indeed, when comparing count-normalized gene expression (as measured by transcripts per million [TPM]) across the different conditions, we found that most of the T6SS machinery, APE biosynthesis, pilus, and iron uptake genes were expressed at both time points in vivo, and to similar levels as in vitro (Figure 5, Figure 5—figure supplement 2). Only the Clb genomic island and some of the VgrG-like T6SS effector genes had clearly lower TPM values in vivo than in vitro, yet higher than in ihfA* in vitro (Figure 5C and F). These results suggest that most of the genes upregulated in vitro in the wt relative to ihfA* are also expressed at high level by the wt in vivo. Figure 5 with 2 supplements see all Do