Comparative transcriptional profiling of the early host response to infection by typhoidal and non-typhoidal Salmonella serovars in human intestinal organoids

Salmonella enterica represents over 2500 serovars associated with a wide-ranging spectrum of disease; from self-limiting gastroenteritis to invasive infections caused by non-typhoidal serovars (NTS) and typhoidal serovars, respectively. Host factors strongly influence infection outcome as malnourished or immunocompromised individuals can develop invasive infections from NTS, however, comparative analyses of serovar-specific host responses have been constrained by reliance on limited model systems. Here we used human intestinal organoids (HIOs), a three-dimensional “gut-like” in vitro system derived from human embryonic stem cells, to elucidate similarities and differences in host responses to NTS and typhoidal serovars. HIOs discriminated between the two most prevalent NTS, Salmonella enterica serovar Typhimurium (STM) and Salmonella enterica serovar Enteritidis (SE), and typhoidal serovar Salmonella enterica serovar Typhi (ST) in epithelial cell invasion, replication and transcriptional responses. Pro-inflammatory signaling and cytokine output was reduced in ST-infected HIOs compared to NTS infections, consistent with early stages of NTS and typhoidal diseases. While we predicted that ST would induce a distinct transcriptional profile from the NTS strains, more nuanced expression profiles emerged. Notably, pathways involved in cell cycle, metabolism and mitochondrial functions were downregulated in STM-infected HIOs and upregulated in SE-infected HIOs. These results correlated with suppression of cellular proliferation and induction of host cell death in STM-infected HIOs and in contrast, elevated levels of reactive oxygen species production in SE-infected HIOs. Collectively, these results suggest that the HIO model is well suited to reveal host transcriptional programming specific to infection by individual Salmonella serovars, and that individual NTS may provoke unique host epithelial responses during intestinal stages of infection.


Introduction
Salmonella enterica greatly impacts human health causing an estimated 115 million infections worldwide every year and are one of the four leading causes of diarrheal diseases [1,2]. Salmonella enterica is a copious species consisting of over 2500 serovars and infects the intestinal epithelial layer causing diseases ranging from asymptomatic carriers to more severe systemic disease. Salmonella serovars are classified based on host specificity and disease outcomes. Host generalist serovars including Salmonella enterica serovar Typhimurium (STM) and Enteritidis (SE) infect a broad range of hosts and cause localized inflammation and self-limiting diarrhea in healthy individuals or more severe gastroenteritis in children and elderly. In contrast, host-restricted serovars including Typhi and Abortusovis infect only one host and cause more serious clinical manifestations including Typhoid fever in humans and abortions in mares respectively.
Although Salmonella enterica serovars share a conserved core genome, determinants of host specificity and varying clinical manifestations are poorly understood. The molecular basis for distinct host adaptation and disease outcome is likely to be multifactorial, mediated by bacteria and host-dependent mechanisms. Initial comparative genomic analyses identified specific signatures that may be indicative of some of these differences [3]. However, comparative host signatures across different serovars are still limited by host specificity and poorly representative model systems. Using human epithelial cell lines addresses host-specificity, but immortalized cell lines do not represent the multiple subsets of intestinal epithelial cells found in the gut and harbor mutations that likely alter cellular responses to bacterial infection.
Human intestinal organoids (HIOs) have emerged as an alternative in vitro model to study intestinal epithelial host responses to commensal microbiota and enteric pathogens. HIOs are differentiated from pluripotent stem cells into three-dimensional spheroids composed of a defined luminal space bound by a polarized epithelial barrier surrounded by mesenchyme. This is an improvement over existing models because the untransformed HIO epithelium is polarized, and contains multiple epithelial cell lineages found in the intestine [4]. Hill et al. showed that HIOs supported luminal growth of Escherichia coli following microinjection, and that physiological changes in the HIO occurred during colonization, such as an increase in mucus production, mirroring what happens in vivo during initial colonization [5]. Forbester

Salmonella serovars invade HIO epithelial cells and induce distinct patterns of mucus production.
To study initial host responses to Salmonella, we micro-injected bacteria into the luminal space of the HIO to allow luminal replication throughout the course of infection. This model better resembles the continuous interaction between bacteria and intestinal cells during the natural course of infection. We first determined whether different Salmonella serovars could colonize and invade HIO epithelial cells. We chose the most prevalent serovars that cause gastroenteritis, STM and SE, and a typhoidal serovar, ST. HIOs were microinjected with STM, SE or ST, and total bacterial burden per HIO was enumerated at 2.5 and 24 hours post infection (h pi) (Fig 1A). All serovars showed at least a 1.5 log increase in bacterial burden at 24h pi, relative to 2.5h pi. Intracellular bacterial burden was quantified by gentamicin protection assay, where HIOs were sliced open to expose luminal bacteria to gentamicin prior to quantifying colony forming units (Fig 1B). Intracellular bacterial numbers increased over time with all serovars, suggesting that intracellular replication or continued invasion contributes to the increase in bacterial load at 24h pi. Noticeably, STM infection resulted in a slightly higher bacterial burden in both total and intracellular CFU at 24h pi, compared to the other serovars. This observation suggests that STM invades HIO epithelial cells more efficiently than SE and ST. We also evaluated epithelial morphology by performing hematoxylin and eosin (H&E) staining ( Fig   1C). HIOs remained intact during infection with all serovars for the duration of the experiment. However, damaged regions of the HIO epithelial lining could sometimes be observed, especially during STM infection. In addition, periodic acid-schiff reagent (PAS) and alcian blue (AB) staining was also performed to detect mucus, as a recent study showed that HIOs increase mucin production during bacterial colonization [5]. In agreement with these findings, PAS and AB staining revealed an increase in mucus production in response to infection with all three serovars (Fig 1D). Interestingly, we observed unique staining patterns during infection with the different serovars. In STinfected HIOs, mucus appeared to accumulate within cells, while STM and SE infection resulted in evacuation of mucus into the HIO lumen. Thus, within a 24h period, all three Salmonella serovars can colonize and invade HIOs, inducing distinct patterns of mucus production without causing major destruction to the HIO epithelial layer.

Host transcriptional dynamics differ between Salmonella serovars.
To define the global host transcriptional response to the 3 Salmonella serovars, we performed RNA sequencing (RNA-seq) at 2.5h and 8h pi with HIOs that were infected with STM, SE or ST and compared transcript levels to control PBS-injected HIOs.
Principal component analysis (PCA) was performed on normalized gene counts to identify clustering patterns between conditions (Fig 2A). infected with STM, SE or ST (S1 and S2 Tables). We found that there were comparable numbers of DEGs during infection of all serovars at 2.5h pi (Fig 2B). Some of the DEGs were shared between infections with each serovar, which likely represents a core host response to Salmonella infection. However, infection with each serovar also resulted in induction and suppression of a unique set of DEGs. Notably, transcriptional dynamics showed that there was an increase in the number of DEGs at 8h pi in response to infection with the non-typhoidal serovars (NTS), STM and SE, while the number of DEGs during infection with ST decreased (Fig 2B). Collectively, the HIO responses represent two patterns; core transcriptional responses that are changed during Salmonella infection and serovar-specific responses.

Salmonella serovars differentially alter inflammatory, stress response, metabolism of lipid and amino acids and cell cycle pathways.
To identify biological pathways associated with DEGs from each infection condition, gene sets were separated into upregulated (increased) and downregulated (decreased) categories based on fold change relative to PBS-injected HIOs and imported separately into the Reactome pathway analysis tool (S3-S6 Tables). In the upregulated datasets, the majority of the significant pathways in all three infection conditions at both 2.5h and 8h belonged to the immune system and signal transduction category (Fig 3A). We found that infection induced a complex response in both innate immune and cytokine signaling pathways including, but not limited to, Toll-like receptors, interleukin mediators and Type I interferons (Fig 3B). Notably, only in ST-infected HIOs were some immune system pathways associated with downregulated DEGs, such as non-canonical NF-κB and Interleukin-1 signaling. These results revealed that inflammatory pathways were the primary responses during Salmonella infection and are consistent with the hypothesis that typhoidal serovar infection is relatively "silent", producing less inflammatory mediators compared to NTS infection.
Apart from the predominant inflammatory pathways, we also identified several differentially upregulated pathways in response to Salmonella serovars that have been linked to intestinal infection. These pathways included antigen presentation, extracellular matrix organization (ECM), lipid and amino acid metabolism and cellular stress responses including IRE1α-mediated unfolded protein response (UPR), mitophagy and the inflammasome (S1 Fig). Although there were genes in these pathways that were significantly upregulated in response to all three serovars, some were enriched only in response to a specific serovar. For example, we found that pathways belonging to ECM, UPR and tryptophan catabolism were significantly upregulated at 8h pi during STM infection but not during SE and ST infection. In contrast, we found that cholesterol metabolism pathways were highly enriched in ST infection while amino acid metabolism, cellular responses to hypoxia, the inflammasome and antigen presentation pathways were significantly induced only in SE-infected HIOs.
Most of the significantly down-regulated pathways during STM and ST infections belonged to cell cycle, DNA replication and repair, metabolism of protein and metabolism of RNA (Fig 3A), which point to a potential reduction in cell proliferation. Interestingly, in SE-infected HIOs, some of these categories including cell cycle and metabolism of protein were instead associated with upregulated DEGs at 8h pi (Fig 3A and B). To further investigate how cell cycle genes changed in response to each serovar, we generated volcano plots to identify the distribution of significant cell cycle genes in response to infection (Fig 3C-E

Salmonella serovars induce distinct HIO proinflammatory response profiles.
Intestinal epithelial cells initiate inflammatory responses via production of proinflammatory mediators. Because the most dramatic transcriptional responses we observed were related to immune signaling, we sought to identify the HIO signature of inflammatory mediators including chemokines, cytokines and antimicrobial peptides (AMP) in response to each Salmonella serovar (Fig 4A and S2 Fig). We found that all of these mediators were induced early during infection although with different magnitudes.  [6]. HIO production of IL17C and its known downstream proinflammatory mediators, including CSF3 and DEF4A, also suggest that IL17C signaling modulates human intestinal host defense against Salmonella infection. To validate these transcriptional results, we measured production of specific inflammatory mediators (cytokine, chemokines and AMP) in the HIO culture medium by ELISA. All three serovars induced production of these mediators (Fig 4B and S3 Fig). Collectively, the data indicate that each serovar, even the two non-typhoidal serovars, interacts distinctly with the host to tune production of inflammatory mediators during infection. Our data also reflect previous reports that ST infection is less inflammatory than infection by other Salmonella serovars and suggest that HIOs are a useful platform for studying ST interactions with human epithelium.

Mitochondrial processes are differentially regulated during NTS infections.
Although NTS cause similar disease manifestations in humans, they may interact with the intestinal epithelium by varied mechanisms as their genomes contain some different accessory genes [3]. Our data indicated that one of the most differentially regulated cellular processes between NTS was related to metabolism of proteins (Fig 3A). To further identify major pathways within this category that were differentially regulated during infection with NTS, we sorted the significant pathways that belonged to the metabolism of proteins category in the Reactome database to identify these pathways.
We found that pathways belonged to three major categories; translation, protein folding, and post-transcriptional regulation that were increased in SE-infected HIOs but decreased during STM infection (Fig 5A). Within the translation umbrella category, we found many mitochondrial-related processes, including mitochondrial translation, mitochondrial protein import and oxidative phosphorylation, were increased during SE infection but decreased during STM infection (Fig 5B), suggesting that mitochondrial functions may differentiate between the host response to NTS during early stages of infection. Because mitochondria produce reactive oxygen species (ROS) during metabolism, we monitored generation of ROS in HIOs during infection (Fig 5C). Our finding that the three Salmonella serovars showed ifferential regulation of cell cycle pathways was intriguing. Intestinal epithelial cells undergo self-renewal to maintain barrier integrity, and infection with enteric pathogens can accelerate or inhibit cell proliferation to gain a survival advantage in the gut [12]. For example, Citrobacter rodentium stimulates the proliferation of undifferentiated epithelial cells, which increases oxygenation of the mucosal surface in the colon to create a replicative niche [13]. By contrast, some enteric pathogens including STM, H. pylori and Shigella are equipped with virulence factors to counteract intestinal cell proliferation and rapid epithelial turnover to enhance virulence [12]. In our experiments, both STM and ST infections resulted in downregulation of many genes in the cell cycle pathway while SE infection resulted in upregulation of several of these genes. Of note, it was previously reported that STM blocks epithelial cell proliferation via Type 3 Secretion System-2 effectors SseF and SseG [14]. These effectors are also encoded in the ST and SE genomes, but it is unclear whether expression levels or kinetics of SseF and SseG might differ to allow fine-tuned control of cell proliferation and pathogenesis.
Although STM and SE cause similar diseases in humans, we were surprised to observe that these two serovars exhibited the most variation in HIO responses relative to each other, including regulation of mitochondrial function-related genes. Our prior research has been focused on how cellular stress pathways contribute to innate immunity and we have recently shown that mitochondrial ROS contributes to bacterial killing by macrophages [15]. Interestingly, we observed that many pathways involved in mitochondrial metabolism are upregulated during SE infection and downregulated during STM infection.

Accordingly, we found that SE infection increased generation of antimicrobial ROS in the
HIOs, suggesting that an increase in mitochondrial metabolism may be important in intestinal host defense. Indeed, mitochondrial integrity and function is required for the maintenance of healthy intestinal barriers to prevent bacterial translocation across the epithelial lining [16,17]. In addition, recent studies demonstrated that metabolites produced by microbes in the gut can influence mitochondrial biogenesis and inflammation [18]. Given that both STM and SE are present in the HIO lumen through the course of infection, it remains unclear whether SE uniquely increases expression of mitochondrial genes, or luminal bacteria generally increase expression of mitochondrial genes but STM uniquely decreases their expression, or both. SE encodes more than 200 genes that are absent in either the STM or ST genome, which are clustered in unique islands designated as "regions of difference" (ROD) [19]. Some of these additional genes have been linked to SE pathogenesis using a mouse model of Salmonella infection [20,21]. Therefore, we speculate that genes expressed only by SE might account for SE-specific HIO responses and further work is required to elucidate mechanisms by which SE induces these specific responses.
Altogether, our findings show that the HIOs are a productive model to study early interactions of Salmonella serovars with the intestinal epithelium. HIOs have been previously used to probe for transcriptional responses during STM infection [22], but to our knowledge this is the first study to directly compare non-transformed human intestinal epithelial responses between non-typhoidal and typhoidal serovars.

HIO Differentiation and Culture
HIO were generated by the In Vivo Animal and Human Studies Core at the University of Michigan Center for Gastrointestinal Research as previously described [23].

Immunohistochemistry and Immunofluorescence Staining
HIOs were fixed with either 10% neutral formalin or Carnoy's solution for 2 days and embedded in paraffin. 5 μm sections were collected by the University of Michigan Cancer Center Histology Core and stained with hematoxylin and eosin (H&E) staining. Carnoy'sfixed HIO sections were stained with periodic acid-Schiff (PAS) staining kit according to the manufacturer's instructions (Newcomersupply). H&E and PAS stained slides were imaged on Olympus BX60 upright microscope. All images were further processed using ImageJ.

RNA Sequencing and Analysis
Total RNA was isolated from groups of 5 HIOs per replicate with a total of 4 replicates per infection condition using the mirVana miRNA Isolation Kit (ThermoFisher). Cytosolic and mitochondrial ribosomal RNAs were removed from samples using the Ribo-Zero Gold Kit according to manufacturer's protocol (Illumina). RNA samples were used to prepare cDNA libraries by the University of Michigan DNA Sequencing Core and the quality of RNA was confirmed, RIN > 8.5, using the Agilent TapeStation system. Libraries were sequenced on Illumina HiSeq 2500 platforms (single-end, 50 bp read length).

Reactive Oxygen Species (ROS) Measurement
HIOs were re-plated onto glass-bottom petri dishes (MatTek) and microinjected with 1 μl of PBS/bacteria containing 50 ng/HIO of CM-H2DCFDA (ThermoFisher). HIOs were imaged using inverted widefield live fluorescent microscopy at indicated time points.
Images were analyzed by ImageJ.

Quantification and Statistical Methods
Data were analyzed using Graphpad Prism 7 and R software. Statistical differences were determined using one-way ANOVA or two-way ANOVA (for grouped analyses) and followed-up by Tukey's multiple comparisons test. The mean of at least 3 independent experiments were presented with error bars showing standard deviation (SD). P values of less than 0.05 were considered significant and designated by: *P < 0.05, **P < 0.01, ***P < 0.001 and **** P < 0.0001.

Data and Software Availability
Raw data are available upon request, which should be directed to the corresponding authors. Code for analyses can be found at: https://github.com/rberger997/HIO_dualseq2 and https://github.com/aelawren/Salmonella-serovars-RNA-seq.

Sequence alignment
Sequencing generated FASTQ files of transcript reads were pseudoaligned to the human genome (GRCh38.p12) using kallisto software [24]. Transcripts were converted to estimated gene counts using the tximport package [25] with gene annotation from Ensembl [26].

Differential gene expression
Differential expression analysis was performed using the DESeq2 package [27] with P values calculated by the Wald test and adjusted P values calculated using the Benjamani & Hochberg method [28].

Pathway enrichment analysis
Pathway analysis was performed using the Reactome pathway database and pathway enrichment analysis in R using the ReactomePA software package [29].

Statistical analysis
Analysis was done using RStudio version 1.1.453. Plots were generated using ggplot2 [30] with data manipulation done using dplyr [31]. Euler diagrams of gene changes were generated using the Eulerr package [32].