NHR-49/PPAR-α and HLH-30/TFEB cooperate for C. elegans host defense via a flavin-containing monooxygenase

The model organism Caenorhabditis elegans mounts transcriptional defense responses against intestinal bacterial infections that elicit overlapping starvation and infection responses, the regulation of which is not well understood. Direct comparison of C. elegans that were starved or infected with Staphylococcus aureus revealed a large infection-specific transcriptional signature, which was almost completely abrogated by deletion of transcription factor hlh-30/TFEB, except for six genes including a flavin-containing monooxygenase (FMO) gene, fmo-2/FMO5. Deletion of fmo-2/FMO5 severely compromised infection survival, thus identifying the first FMO with innate immunity functions in animals. Moreover, fmo-2/FMO5 induction required the nuclear hormone receptor, NHR-49/PPAR-α, which controlled host defense cell non-autonomously. These findings reveal an infection-specific host response to S. aureus, identify HLH-30/TFEB as its main regulator, reveal FMOs as important innate immunity effectors in animals, and identify the mechanism of FMO regulation through NHR-49/PPAR-α during S. aureus infection, with implications for host defense and inflammation in higher organisms.


Introduction
In their natural habitat, C. elegans feed on microbes that grow on rotting vegetable matter, and thus face a high likelihood of ingesting pathogens (Schulenburg and Félix, 2017). To defend against infection, C. elegans possess innate host defense mechanisms that promote their survival (Ermolaeva and Schumacher, 2014;Kim and Ewbank, 2018). In the laboratory, model human pathogenic bacteria cause intestinal pathology and death through poorly understood mechanisms (Irazoqui et al., 2010a). Infected animals experience both chemical signals that reveal the pathogen's presence and organismal stress caused by the infection. Over the last 15 years, several studies have identified and characterized C. elegans gene expression changes in response to pathogenic bacteria, fungi, and viruses, mounted through evolutionarily conserved mechanisms (Irazoqui et al., 2010b;Kim and Ewbank, 2018). However, the relative contributions of pathogen sensing and organismal stress mechanisms to the total pathogen-induced response remain unclear.
We previously showed that ingested Gram-positive bacterium Staphylococcus aureus causes drastic cytopathology in C. elegans (Irazoqui et al., 2010a). Infection with S. aureus results in progressive effacement and lysis of intestinal epithelial cells, whole-body cellular breakdown, and death (Irazoqui et al., 2010a). Therefore, S. aureus-infected C. elegans experience dietary changes from its laboratory food of nonpathogenic E. coli, as well as intestinal destruction, cellular stress, and putative molecular signals produced by the pathogen.
In previous work, we showed that C. elegans mount a pathogen-specific transcriptional host response against S. aureus, which includes genes that encode antimicrobial proteins (e.g. lysozymes, antimicrobial peptides, and secreted C-type lectins) and cytoprotective factors (e.g. autophagy genes, lysosomal factors, and chaperones) that are necessary and sufficient for survival (Irazoqui et al., 2010a). Moreover, we discovered that C. elegans still induced select host defense genes even when exposed to heat-killed S. aureus, which did not cause intestinal destruction (Irazoqui et al., 2010a). However, the relative contributions of organismal stress and pathogen detection to the induction of the overall host defense response remain unknown.
We recently discovered that the induction of a large majority of the transcriptional host response to S. aureus requires HLH-30, the C. elegans homolog of mammalian transcription factor EB (TFEB) (Visvikis et al., 2014). TFEB belongs to the MiT family of transcription factors, which in mammals and C. elegans controls the transcription of autophagy and lysosomal genes in response to nutritional stress in addition to infection (Lapierre et al., 2013;Raben and Puertollano, 2016). HLH-30 and TFEB also regulate lipid store mobilization during nutritional deprivation (O'Rourke and Ruvkun, 2013;Settembre et al., 2013). Thus, HLH-30/TFEB could potentially integrate organismal stress, metabolism, and pathogen recognition to elicit coordinated host responses to infection. How HLH-30/TFEB integrates this information to produce stress-specific responses and what other factors are involved in such specificity are poorly understood. Specifically, the genes that are induced during infection independently of nutritional stress are not known.
Here, we report that S. aureus infection in C. elegans elicits a transcriptional response that is distinct from that induced by nutritional deprivation, thus defining an infection-specific transcriptional signature. Both the starvation response and the infection-specific signature were largely dependent on HLH-30/TFEB, highlighting its key role as a transcriptional integrator of organismal stress during infection. Moreover, we identified six genes that were specifically induced during infection even in the absence of HLH-30/TFEB, potentially revealing an alternative transcriptional host response signaling pathway. The induction of two of the six genes, fmo-2/FMO5 and K08C7.4, was entirely dependent on transcription factor NHR-49/PPAR-a (Van Gilst et al., 2005), suggesting that NHR-49/PPAR-a defines an additional host defense pathway during S. aureus infection. NHR-49/PPAR-a was required non cell-autonomously for fmo-2/FMO5 induction and host defense against S. aureus. Moreover, functional characterization of fmo-2/FMO5 suggested that its enzymatic activity is specifically required for host defense against S. aureus, revealing that FMO-2/FMO5 is a key host defense effector. Thus, our work demonstrates for the first time that flavin-containing monooxygenases are important for host defense against infection in animals.

Starvation and infection trigger distinct transcriptional responses
Our prior studies showed that S. aureus infection of C. elegans causes a robust host transcriptional response that results in the upregulation of 825 genes (Irazoqui et al., 2010a). Moreover, the 'early' phase of this response was already upregulated by 4 hr infection (Irazoqui et al., 2010a). It is likely that this transcriptional response to infection is compounded with nutritional stress, due to nutritional differences between laboratory food (nonpathogenic E. coli) and S. aureus, and due to intestinal destruction caused by the pathogen (Irazoqui et al., 2010a). To identify genes that are induced during infection independently of nutritional stress, we used whole-animal RNA-seq to directly compare infected and starved animals ( Figure 1A). We identified 388 genes that were differentially expressed between these two conditions ( Figure 1B,C, Supplementary file 1). About 70% (283) of differentially expressed genes were upregulated by starvation, while about 30% (105) were upregulated by infection (Supplementary file 1). Gene ontology (GO) analysis showed the starvationinduced genes to belong mostly to metabolic processes, whereas the infection-specific signature was highly enriched for innate immune response genes (Supplementary file 2). RT-qPCR of the 13 most highly infection-induced genes relative to animals that were starved or fed nonpathogenic E. coli laboratory food confirmed their S. aureus -specific induction ( Figure 1D, Figure 1-figure supplement 1A). Thus, we identified an infection-specific signature of genes that excludes expression changes that are caused by starvation, indicating that the host responses to nutritional deprivation and S. aureus infection have distinct and specific features.

HLH-30/TFEB is critical for host responses to starvation and infection
HLH-30/TFEB was shown to be important for gene induction during dietary challenge and during infection (O'Rourke and Ruvkun, 2013;Settembre et al., 2013;Visvikis et al., 2014). However, whether HLH-30/TFEB regulates the infection-specific response was not known. To assess the relevance of HLH-30/TFEB to the infection-specific signature, we compared starved and infected hlh-30/ TFEB loss-of-function mutants by RNA-seq. To our surprise, in hlh-30/TFEB mutants differential gene expression between starvation and infection was almost completely abrogated ( Figure   conditions. Synchronized young adults were subjected to either starvation or infection for 4 hr before RNA extraction. (B) Volcano plot of differentially expressed genes (P adj 0.01). Genes that were induced in each condition relative to the other are indicated in red (for starvation) and green (for infection). FC, fold change. P adj , adjusted p value. (C) Heat map of differentially expressed genes [Log 2 (FC)] comparing infection with S. aureus SH1000 to starvation by RNA-seq. The boxed area represents the designated infection-specific expression signature. (D) Heat map of a set of 13 genes most highly induced by S. aureus SH1000 compared to starvation, whose relative transcript levels were measured by RT-qPCR and plotted as row-normalized log 2 (relative expression), or -DCt. Conditions include nonpathogenic E. coli, S. aureus (4 hr), and starvation (4 hr). Columns represent independent biological replicates. The online version of this article includes the following source data and figure supplement(s) for figure 1: Figure supplement 1. Expression analysis of 13 most highly induced genes. Figure supplement 1-source data 1. mRNA levels of 13 highly induced genes in wild type animals fed nonpathogenic E. coli, infected with S. aureus, or starved. Figure supplement 1-source data 2. mRNA levels of 13 highly induced genes in hlh-30(-) mutant animals fed nonpathogenic E. coli, infected with S. aureus, or starved. C33A12.19, C54F6.12, K08C7.4, and Y47H9C.1 (Supplementary file 3). RT-qPCR confirmed the predicted results for the selected 13 top induced genes ( Figure 2C, Figure 1-figure supplement 1B). Particularly, we verified that fmo-2/FMO5 was partially induced in hlh-30/TFEB mutants compared to wild type ( Figure 2D). Partial induction of fmo-2/FMO5 in hlh-30/TFEB mutants was rescued by transgenic re-expression of hlh-30/TFEB driven by its endogenous promoter ( Figure 2D). Altogether, these results showed that HLH-30/TFEB is crucial for both the starvation and the infection- Figure 2. HLH-30/TFEB is critical for host responses to starvation and infection. (A) Volcano plot of differentially expressed genes in hlh-30/TFEB lossof-function mutants (P Adj . 0.01). Genes that were induced in each condition relative to the other are indicated in red (for starvation) and green (for infection). (B) Venn diagram representing genes that were upregulated during infection compared to starvation in wild type and hlh-30/TFEB mutants. A few selected genes are indicated for reference. (C) Heat map of RT-qPCR (-DCt) relative expression values of a set of 13 genes most highly induced by S. aureus v starvation , measured in wild type and hlh-30/TFEB mutants. Conditions include nonpathogenic E. coli, S. aureus, and starvation. Columns represent independent biological replicates. * indicates genes that were highly induced in wild type compared to hlh-30/TFEB mutants during infection, and thus were partially or completely HLH-30/TFEB-dependent. 'Starv.', starvation. (D) RT-qPCR of fmo-2/FMO5 transcript in wild type, hlh-30/TFEB loss-of-function mutants, and hlh-30(-); Phlh-30::hlh-30::gfp (complemented) animals fed nonpathogenic E. coli or infected with S. aureus (4 hr). Data are normalized to wild-type fed nonpathogenic E. coli, means ± SEM (3-4 independent biological replicates). ***p 0.001, ns = not significant, one-way ANOVA followed by Š ídá k's test for multiple comparisons. The online version of this article includes the following source data for figure 2: Source data 1. fmo-2 mRNA levels in wild type, hlh-30(-), and hlh-30(-); Phlh-30::hlh-30::gfp (complemented) animals fed nonpathogenic E. coli or infected with S. aureus. specific responses, and hinted at an HLH-30/TFEB-independent pathway for the induction of six infection-specific genes.
Previous studies identified NHR-49, a nuclear receptor homologous to human PPAR-a and HNF4a, as essential for fmo-2/FMO5 induction during exogenous oxidative stress (Goh et al., 2018;Hu et al., 2018). To examine the role of NHR-49/PPAR-a during S. aureus infection, we measured fmo-2/FMO5 expression in nhr-49/PPARA null mutants (Liu et al., 1999;Van Gilst et al., 2005). We found that in these mutants, expression of the fmo-2/FMO5 fluorescent transcriptional reporter was barely induced ( Figure 3F,G) and, importantly, was undetectable in the intestinal epithelium ( Figure 3H-K). In contrast, fmo-2/FMO5 induction by S. aureus was partially dependent on hlh-30/ TFEB, as predicted by RNA-seq ( Figure 3A-D and G, Figure 2); in these mutants, expression was preserved in the pharyngeal isthmus, pharyngeal-intestinal valve, and in the intestinal epithelium, albeit to lower levels compared to wild type ( Figure 3H,I and K). Thus, nhr-49/PPARA appeared to be essential for the induction of fmo-2/FMO5 in the entire body. Consistently, noninfected nhr-49/ PPARA mutants exhibited about 10-fold lower fmo-2/FMO5 expression than wild type by RT-qPCR ( Figure 3L). After infection, nhr-49/PPARA mutants completely failed to induce fmo-2/FMO5 ( Figure 3L). Transgenic rescue of nhr-49/PPARA driven by its endogenous promoter partially restored fmo-2/FMO5 induction ( Figure 3L). Therefore, nhr-49/PPARA was absolutely required for fmo-2/FMO5 expression in infected and noninfected animals. RT-qPCR of the other five HLH-30-independent genes showed that only K08C7.4 induction by S. aureus was also dependent on NHR-49/PPAR-a ( Figure 3-figure supplement 2). We generated a K08C7.4 GFP transcriptional reporter strain and examined fluorescence in infected and noninfected wild type and nhr-49/PPARA mutants. In this strain, GFP was visible in noninfected animals in the anterior and posterior intestinal epithelium but was strongest in two unidentified head sensory neurons (as previously reported, potentially AFD neurons [Lockhead et al., 2016;Mounsey et al., 2002]), additional unidentified head and nerve ring neurons, ventral nerve cord, and an unidentified pair of tail neurons (Figure 3-figure supplement 3A). In infected animals, this expression pattern remained unchanged but increased slightly in intensity (Figure 3-figure supplement 3B,E). In contrast, nhr-49/PPARA mutants exhibited similar GFP expression in infected and noninfected animals ( Figure 3-figure supplement 3C-E). Together, these data showed that nhr-49/PPARA was absolutely required for fmo-2/PPARA expression and induction, while it was only partially required for induction of K08C7.4. For this reason, we designated K08C7.4 as nfds-1 for 'NHR-forty-nine-dependent induction by S. aureus, member 1'; nfds-1 appears to have homologs only in nematodes. These data also suggested that NHR-49/PPAR-a contributes to the induction of some of the HLH-30-

NHR-49/PPAR-a functions in multiple tissues for host defense
nhr-49/PPARA is expressed in multiple tissues (Ratnappan et al., 2014). To identify specific tissues where nhr-49/PPARA is sufficient for host defense against S. aureus, we reintroduced wild type nhr-49/PPARA into nhr-49/PPARA mutants driven by tissue-specific promoters, including intestine, neurons, muscle, and epidermis (i.e. hypodermis). We examined these rescue lines for fmo-2/FMO5 induction and survival of infection. Intestinal rescue of nhr-49/PPARA fully restored both basal and induced fmo-2/FMO5 expression ( Figure 5A). Re-expression of nhr-49/PPARA partially restored fmo-2/FMO5 induction in other lines ( Figure 5D,G,J). Consistently, expression in each of these tissues also rescued the infection survival defect of nhr-49/PPARA mutants. In fact, intestinal, neuronal, and muscular expression enhanced infection survival compared to wild type ( Figure 5B,E,H), while epidermal expression rescued infection survival to a level similar to wild type ( Figure 5K). These results suggested that nhr-49/PPARA can function from any of these tissues to promote host defense against infection.
In contrast, tissue-specific complementation of nhr-49/PPARA had more complex effects on normal lifespan on nonpathogenic E. coli. Intestinal and epidermal re-expression prolonged lifespan compared to wild type ( Figure 5C,L). Neuronal expression rescued lifespan to the wild-type level ( Figure 5F), and muscle expression yielded partial rescue ( Figure 5I). Together, these data showed distinct and tissue-specific roles for nhr-49/PPARA in host defense and lifespan.

FMO-2/FMO5 is required for host survival of infection
So far, we focused on fmo-2/FMO5 as a useful reporter of the broader infection-specific signature, but its biological relevance to infection survival was unknown. FMO-2 and FMO5 belong to the evolutionarily conserved flavin-containing monooxygenase (FMO) protein family (Huijbers et al., 2014). In mammals, FMO proteins are primarily known to function in the detoxification of foreign substances (xenobiotics) with prominent roles in drug metabolism (Krueger and Williams, 2005). C. elegans FMO-2 exhibits homology to human proteins FMO1-5, with closest similarity to FMO5 (42% identity). Previously, FMO-2/FMO5 had been implicated in dietary-restriction-mediated lifespan extension, and its forced expression resulted in stress resistance (Leiser et al., 2015). In plants, FMOs participate in host defense against bacterial and fungal infections (Bartsch et al., 2006;Koch et al., 2006). Whether animal FMOs also function in innate host defense was not known.
To determine the physiological relevance of fmo-2/FMO5 during infection, we examined mutants homozygous for an fmo-2/FMO5 deletion predicted to result in a null allele (C. elegans Deletion Mutant Consortium, 2012). Compared with wild type, fmo-2/FMO5 mutants exhibited greatly compromised survival of S. aureus infection ( Figure 8A) but did not exhibit differences in survival of P. aeruginosa infection or in aging when fed nonpathogenic E. coli ( Figure 8B-C). Deletion of fmo-2/ FMO5 did not affect the induction of the nine most highly induced infection-specific signature genes

hlh-30
Wild type nhr-49 ( . These data suggested that FMO-2/FMO5 may play an important role for host defense specifically during S. aureus infection, which is independent of the induction of many other host defense genes. To determine whether such a role of FMO-2/FMO5 requires its catalytic activity, we used CRISPRmediated genome editing to modify key conserved residues in the FMO-2/FMO5 FAD-binding domain, the NADPH-binding domain, or both ( Figure 8-figure supplement 2A-B). Due to their conservation in FMOs from yeast, plants, and animals ( Figure 8-figure supplement 2B), these residues are predicted to be required for electron transfer from organic substrates to cofactors FAD and NADPH (Kubo et al., 1997;Rescigno and Perham, 1994). Remarkably, mutations of either cofactor binding motif caused infection survival defects that were barely weaker than that of null fmo-2/FMO5 mutants. Simultaneous mutation of both motifs produced susceptibility that was indistinguishable from that of fmo-2/FMO5 null mutants ( Figure 8D). These results suggested that both cofactor-binding sites were required for FMO-2/FMO5 function in host defense. In contrast, none of these mutations, alone or in combination, altered total lifespan on nonpathogenic E. coli ( Figure 8E). Together, these data indicated that FMO-2/FMO5 catalytic activity may be specifically required for host defense against infection.
Simultaneous deletion of nhr-49/PPARA and fmo-2/FMO5 resulted in similarly impaired infection survival as the single mutants ( Figure 9A), suggesting that nhr-49/PPARA and fmo-2/FMO5 function in the same pathway. On nonpathogenic E. coli, the lifespan defect of nhr-49(-) mutants was slightly improved by introduction of fmo-2(-) ( Figure 9B). In contrast, introduction of fmo-2(-) in the two nhr-49(gof) mutant backgrounds almost completely suppressed the enhanced infection survival ( Figure 9C), while their lifespan on nonpathogenic E. coli was also suppressed (for gf1) or slightly impaired (for gf2) ( Figure 9D). These results show that that fmo-2/FMO5 is absolutely required for the enhancement of host infection survival by nhr-49/PPARA gain-of-function, consistent with fmo-2/ FMO5 acting downstream of nhr-49/PPARA, while its effect on their lifespan is allele-specific.

Discussion
Because bacteria serve as nutritional source for C. elegans, and because intestinal infections cause destruction of the epithelium resulting in loss of nutrient absorption, transcriptional responses to nutritional challenges are likely intertwined with the transcriptional host defense response to infection itself. This raises the question of whether C. elegans senses infection as a stress per se, through its physiological consequences in the organism, or a combination of both. Here, by directly comparing transcriptomes of animals that were infected with S. aureus or were starved, we discovered that starvation and infection elicit large and distinct transcriptional signatures. This indicates that the C. elegans host response to S. aureus infection is not entirely the result of starvation, and enables the dissection of infection-specific and starvation-specific host response regulatory modules as shown here.
In the present study, we found that loss of HLH-30/TFEB almost completely abrogated differential gene expression between starvation and infection -implicating HLH-30/TFEB not just in a hypothetical overlapping response but also in each of these two distinct signatures. This strongly suggests that HLH-30/TFEB integrates metabolic and other stresses to contribute to stress-specific transcriptional responses. The molecular mechanisms that enable a single transcription factor to mediate specific transcriptional responses to distinct stresses may involve stress-specific signals or transcriptional co-factors.
By focusing on fmo-2/FMO5, which we first showed was highly and specifically induced by S. aureus infection (Irazoqui et al., 2008;Irazoqui et al., 2010a) and is only partially dependent on HLH-30/TFEB (Visvikis et al., 2014), we discovered a novel role for the nuclear receptor NHR-49/ PPAR-a in host defense against S. aureus infection. This discovery is related to prior studies that showed that Enterococcus faecalis, an enteric Gram-positive human pathogen unrelated to S. aureus, induces fmo-2/FMO5 during infection of C. elegans (Dasgupta et al., 2020). In the same work, RNAi of nhr-49/PPARA resulted in enhanced susceptibility to P. aeruginosa, S. enterica serovar Typhimurium, and Candida albicans, but not S. aureus (Dasgupta et al., 2020). These results suggest that nhr-49/PPARA may play other roles in host defense against infection, aside from the induction of fmo-2/FMO5 (which is not induced by P. aeruginosa [ Figure 3-figure supplement 1 and Irazoqui et al., 2008, Irazoqui et al., 2010a). Moreover, nhr-49/PPARA RNAi did not abrogate Pfmo-2::gfp expression in the pharynx (Dasgupta et al., 2020). The reasons of these discrepancies with our observations remain unclear, but could be related to the use of RNAi instead of null alleles or to pathogen-intrinsic differences between S. aureus and E. faecalis. Nonetheless, fmo-2/FMO5 was first described as the most highly induced gene in E. faecalis -infected animals compared with E. coli controls several years ago (Wong et al., 2007;Yuen and Ausubel, 2018). Subsequent work showed that nhr-49/PPARA silencing impaired infection survival of E. faecalis (Sim and Hibberd, 2016), but the connection between nhr-49/PPARA and fmo-2/FMO5 during infection was not established until our present work and that of others (Dasgupta et al., 2020). Thus, NHR-49/PPAR-a appears to play an important role in defense against a broad range of bacteria, a conclusion that is reinforced by the recent identification of small molecules that protect germline-defective C. elegans from P. aeruginosa -mediated killing via NHR-49/PPAR-a (Hummell et al., 2021).
Recently, NHR-49/PPAR-a was shown to mediate the defense response to exogenous oxidative stress (Goh et al., 2018;Hu et al., 2018). Thus, NHR-49/PPAR-a participates in host defense against biotic and abiotic stressors, and should be considered a key player in the organismal stress response alongside SKN-1/NRF2, DAF-16/FOXO3, and HLH-30/TFEB (Blackwell et al., 2015;Lin et al., 2018;Tissenbaum, 2018). Our analysis showed that NHR-49/PPAR-a is not required for as large a portion of the host response to infection as HLH-30/TFEB, even though NHR-49/PPAR-a is partially required for HLH-30/TFEB induction. The larger HLH-30/TFEB regulon implies that during infection signals in addition to NHR-49/PPAR-a activation contribute to HLH-30/TFEB regulation. Similar to HLH-30/ TFEB, how NHR-49/PPAR-a induces specific responses to distinct stresses is also unknown. These findings are relevant beyond nematodes, as PPAR-a regulates TFEB in mammalian cells (Kim et al., 2017). Moreover, the microbiota represses HNF4-a, a second NHR-49 homolog, in zebrafish and mice, to maintain intestinal homeostasis (Davison et al., 2017). Therefore, unraveling the control of NHR-49/PPAR-a in relation to intestinal microbiota and infection may provide useful information to understand vertebrate intestinal homeostasis and host defense.
NHR-49/PPAR-a is a known regulator of fat metabolism during starvation (Van Gilst et al., 2005). Lipid metabolism plays an important role in immune activation and host defense in C. elegans Nandakumar and Tan, 2008). Recent work showed that infection by E. faecalis results in NHR-49/PPAR-a-dependent upregulation of lipid catabolism and downregulation of lipid synthesis genes (Dasgupta et al., 2020). Whether this is also true of S. aureus infection is unknown, but our RNA-seq results suggest an enrichment of genes related to 'Lipid Metabolic Process' -including catabolism and anabolism, in infected wild type but not nhr-49/PPARA mutants. Thus, a mechanism of NHR-49/PPAR-a-mediated host defense could be by the regulation of fat metabolism, in parallel to induction of fmo-2/FMO5.
In addition, we found that loss of FMO-2/FMO5 causes a severe defect in infection survival without affecting longevity. Thus, FMO-2/FMO5 represents a novel host defense effector. We previously examined the requirement for fmo-2/FMO5 using RNAi-mediated silencing, but such manipulation failed to produce a phenotype for reasons unknown (Irazoqui et al., 2010a). Moreover, the failure of tissue-specific RNAi to elicit a phenotype and the toxicity of fmo-2/FMO5 extrachromosomal transgenic constructs precluded our investigation of the tissues of fmo-2/FMO5 action for host defense. However, single copy intestinal expression of FMO-2/FMO5 boosted host defense, suggesting that FMO-2/FMO5 could play a major role in the intestine, a hub for host defense in C. elegans (McGhee, 2007). Nonetheless, FMO-2/FMO5 induction appears to be a major mechanism of host defense in C. elegans. Exactly how FMO-2/FMO5 promotes host infection survival is poorly understood, but site-directed mutagenesis of the NADPH and FAD-binding sites revealed that the mechanism of action requires its catalytic activity. In addition, human FMO5 can generate large amounts of H 2 O 2 from O 2 (Fiorentini et al., 2016). Thus, it is possible that FMO-2/FMO5 is an infection-specific NADPH oxidase that generates H 2 O 2 with antimicrobial and signaling functions (McCallum and Garsin, 2016;Sies and Jones, 2020). The observed roles of fmo-2/FMO5 in survival of heat, di-thiothreitol, and tunicamycin stresses are consistent with a H 2 O 2 -mediated signaling role (Leiser et al., 2015). Alternatively, ER localization of FMOs may be important for regulation of ER stress or of the UPR ER . Several infections induce ER stress, and the UPR ER promotes host defense in many organisms (Choi and Song, 2019). In support of this, yeast FMO (yFMO), which localizes to the ER membrane, is activated by the UPR ER and is required for protein folding in the ER (Suh and Robertus, 2000;Suh et al., 1999). Moreover, the five human FMOs localize to the ER membranes of cells of the liver, lung, and kidney (Dolphin et al., 1996;Phillips et al., 1995). Human FMO5 is also expressed in the intestine (Hernandez et al., 2004;Zhang and Cashman, 2006) and murine FMO5 is expressed in the epithelial cells of small and large intestine (Scott et al., 2017). While its subcellular localization in intestinal epithelial cells was not characterized, it is possible that FMO5 localizes to their ER membrane (Phillips et al., 1995). Further research is necessary to understand the precise connections linking infection, ER stress, and FMO function in nematodes and higher organisms.
Despite the paucity of information, FMOs are emerging as important host defense factors across phylogeny. In plants, FMO1 is required for the conversion of pipecolic acid to N-hydroxypipecolic acid, which provides systemic acquired resistance to bacterial and oomycete infections (Hartmann et al., 2018). In mammals, FMO3 is an evolutionarily ancient FMO that exhibits unique substrate specificity and catalyzes multiple drugs that is important for their detoxification (Krueger and Williams, 2005). Moreover, mammalian FMO proteins were recently shown to promote cellular stress resistance and alter cellular metabolism. Overexpression of each of the five mouse FMOs (FMO1-FMO5) in human hepatocytes and kidney epithelial cells conferred resistance to Cd 2+ , arsenite, paraquat, UV radiation, and rotenone (Huang et al., 2021). Additionally, FMO overexpression increased mitochondrial respiration with concomitant decrease in glycolytic activity (Huang et al., 2021). However, to date no reports have indicated an important role for FMO5, or other FMOs, in mammalian (or any animal) innate immunity. In mice, FMO5 is expressed in many tissues and organs, including the liver and the epithelium of the gastrointestinal tract (Scott et al., 2017). Mouse FMO5 is required for sensing the microbiota, and Fmo5 -/mutants exhibit altered metabolic profiles and microbiomes compared with wild-type mice (Scott et al., 2017). Furthermore, Fmo5 -/mutants exhibit a 70% reduction in plasma TNF-a compared with wild type (Scott et al., 2017). Together, these observations suggest that FMO5 is an important microbiota sensor and effector that modulates the intestinal microbiota, but the mechanism of action is unknown. Therefore, elucidation of mechanisms of host defense mediated by FMO-2 in nematodes and FMO5 in mammals will provide fundamental insight into evolutionarily conserved mechanisms of host defense against infection and identify therapeutic opportunities for infections and inflammatory diseases.

Experimental model
The nematode C. elegans was used as the experimental model for this study. Strains were maintained at 15-20˚C on Nematode Growth Media (NGM) plates seeded with Str R E. coli OP50-1 strain using standard methods (Stiernagle, 2006).

Method details Infection assays
S. aureus SH1000 strain was grown overnight in tryptic soy broth (TSB) containing 50 mg/ml kanamycin (KAN). Overnight cultures were diluted 1:1 with TSB and 10 ml of the diluted culture was uniformly spread on the entire surface of 35 mm tryptic soy agar (TSA) plates containing 10 mg/ml KAN. Plates were incubated for 5-6 hr at 37˚C, then stored overnight at 4˚C. P. aeruginosa isolate PA14 was grown overnight in Luria broth. 10 ml of the overnight culture was uniformly spread on the entire surface of 35 mm NGM plates. Plates were incubated at 37˚C for 24 hr followed by 25˚C for 48 hr (Powell and Ausubel, 2008). Animals were treated with 100 mg/ml 5-fluoro-2 0 -deoxyuridine (FUDR) at L4 larval stage for~24 hr at 15-20˚C before transfer to S. aureus or P. aeruginosa plates. Three plates were assayed for each strain in each replicate, with 20-40 animals per plate. Infection assays were carried out at 25˚C except for the RNAi experiments. We found that feeding wild type animals with E. coli strain HT115(DE3) prior to putting them on SH1000 made them die much faster than when fed on regular E. coli OP50 food. Therefore, to determine any differences in survival to SH1000 upon RNAi of the target genes, infection assays were carried out at 20˚C to slow down the growth of S. aureus. Survival was quantified using standard methods (Powell and Ausubel, 2008). Animals that crawled off the plate or died of bursting vulva were censored. Infection assays were carried out at least twice.

S. aureus infection for RNA analysis
To prepare infection plates, S. aureus SH1000 was grown overnight in TSB containing 50 mg/ml KAN. 500-1000 ml of overnight culture was uniformly spread on the entire surface of freshly prepared 100 mm TSA plates supplemented with 10 mg/ml KAN. The plates were incubated for 6 hr at 37˚C, then stored overnight at 4˚C. To prepare P. aeruginosa plates, P. aeruginosa isolate PA14 was grown overnight in Luria broth. 1 ml of overnight culture was uniformly spread on the entire surface of freshly prepared 100 mm NGM plates. The plates were first incubated at 37˚C for 24 hr and then at 25˚C for 48 hr. To prepare control plates with nonpathogenic E. coli, 1 ml of 10-20X concentrated overnight culture of OP50-1 bacteria was spread on 100 mm NGM plates, incubated for 5-6 hr at 37˚C, and then stored at 4˚C, similar to S. aureus plates. To prepare plates for starvation, TSA plates were treated similarly to infection plates, except that nothing was added to them. Synchronized young adults of wild type and mutants were seeded the next day on S. aureus, P. aeruginosa, OP50-1, and starvation plates that were previously warmed to room temperature. After 4 hr incubation at 25˚C, animals for all conditions were washed 3-4 times in water, and then lysed in 1 ml of TRIzol reagent (Invitrogen). The samples were snap frozen in liquid nitrogen, then stored at À80˚C. RNA was extracted using 1-bromo-3-chloropropane (MRC) followed by purification with isopropanol-ethanol precipitation. RNA was analyzed by qPCR or sequencing. In some experiments, the magnitude of fmo-2 expression measured by RT-qPCR showed some inter-trial variation. The reasons for this are not fully understood, but could be related to the very low expression in infected animals or to variation in environmental variables, such as plate batch. Regardless, the results strongly support our conclusions regarding hlh-30 and nhr-49 dependency. For sequencing, RNA was additionally purified using PureLink RNA Mini Kit (Invitrogen). Four independent biological replicates were submitted to BGI for library preparation and sequencing using BGI-seq 500.

Quantitative RT-PCR
After each treatment, C. elegans were collected in sterile water and lysed using TRIzol Reagent (Invitrogen). Total RNA was extracted and purified as described before and then digested with DNAse (Bio-Rad). 100-1000 ng of total RNA was used for cDNA synthesis using iScript cDNA synthesis kit (Bio-Rad). RT-qPCR was performed using SYBR Green Supermix (Bio-Rad) using a ViiA7 Real-Time qPCR system (Applied Biosystems). Primer sequences are provided in Supplementary file 5. At least two independent biological replicates were used for each treatment and C. elegans strain. qPCR Ct values were normalized to the snb-1 control gene, which did not change with the conditions tested, to calculate RT-qPCR DCt values. Data analysis was carried out using the Pfaffl, 2001 method. Heat maps were generated using open access online tool Morpheus (https://software.broadinstitute.org/ morpheus).

RNA sequencing analysis
BGI provided clean reads in FASTQ format. Clean FASTQ files were verified using FastQC (https:// www.bioinformatics.babraham.ac.uk/projects/fastqc) using Bioconductor in RStudio (Loraine et al., 2015) and used as input for read mapping in Salmon v.0.9.1 (Patro et al., 2017) using WBCel.235. cdna from Ensembl (http://www.ensembl.org) as reference transcriptome. Salmon outputs in quant format were used for input in DESeq2 (Love et al., 2014) in Bioconductor in RStudio for count per gene estimation using batch correction. Total counts per gene tables from DESeq2 were used as input for DEBrowser (Kucukural et al., 2019) for verification of transcriptome replicate similarity, data analysis using the built-in DESeq2 algorithm for differential gene expression analysis (adjusted p value 0.01 was considered significant), visualization, and interactive data mining. Overlap between gene sets was determined using the Venn tool in BioInfoRx (https://bioinforx.com). GO representation analysis was performed using online tool g:Profiler (Raudvere et al., 2019) (https:// biit.cs.ut.ee/gprofiler/gost).

Longevity (aging) assays
Animals were transferred to 60 mm NGM plates seeded with 10-20X concentrated E. coli OP50-1 bacteria supplemented with 100 mg/ml FUDR. For consistency with infection assays, longevity assays were also performed at 25˚C. Three plates were assayed for each strain in each replicate, with 20-40 animals per plate. Experiments were repeated at least twice. Animals that did not respond to prodding were scored as dead, and the animals that died from bursting vulva or crawled off the plate were censored.

RNA interference (RNAi)
Knockdown of genes was carried out by feeding C. elegans the E. coli strain HT115(DE3) expressing double-stranded RNA against the target gene using standard methods (Kamath et al., 2001). Briefly, glycerol stocks of RNAi clones containing the appropriate vectors were streaked out on LB media plates containing ampicillin (50 mg/ml) and tetracycline (10 mg/ml) and the bacteria were allowed to grow at 37˚C for 16 hr. Single colonies of the clones were then grown in LB liquid media containing ampicillin (50 mg/ml) for 16 hr at 37˚C with constant shaking. The cultures were concentrated 10X and 300-500 ml were plated on NGM plates containing ampicillin (50 mg/ml) and IPTG (1 mM). Plates were dried in a fume hood and left overnight at room temperature for dsRNA induction. Next day, five L4 animals were placed on the plates and then allowed to grow at 20˚C for 4 days to get enough L4 progeny of the next generation. FUDR (100 mg/ml) was placed on the plates and next day young adult animals were picked and placed on TSA plates containing uniformly spread SH1000. Survival on SH1000 was performed as described above in Infection Assays. For assessing lifespan after RNAi of the target genes, L4 animals were picked and placed on OP50 containing NGM plates supplemented with 100 mg/ml FUDR and then placed at 25˚C. Lifespan was carried out as described above in Longevity (aging) assays.

Generation of transgenic strains
To construct Pnhr-49::nhr-49::gfp containing plasmid, a 6.6 kb genomic fragment of nhr-49 gene (comprising of 4.4 kb coding region covering all nhr-49 transcripts plus 2.2 kb sequence upstream of ATG) was cloned into the GFP expression vector pPD95.77 (Addgene #1495), as reported previously (Ratnappan et al., 2014). For generating tissue-specific constructs, the nhr-49 promoter was replaced with tissue-specific promoters using SbfI and SalI restriction enzymes. The primers that were used to amplify tissue-specific promoters are listed in Supplementary file 5. For the generation of rescue strains, each rescue plasmid (100 ng/ml) was injected along with pharyngeal musclespecific Pmyo-2::mCherry co-injection marker (25 ng/ml) into nhr-49(nr2041) mutant strain, using standard methods (Mello and Fire, 1995). Strains were maintained by picking animals that were positive for both GFP and mCherry.
To construct PK08C7.4::gfp containing plasmid, a 2 kb 5' upstream sequence (upstream of ATG) was amplified from C. elegans genomic DNA, and then cloned into the GFP expression vector pPD95.75 (Addgene #1494) using Gibson Assembly (Gibson et al., 2009). Promoter and vector sequences were assembled using Gibson Assembly Kit (NEB #E2611). For the generation of transgenic strains, PK08C7.4::gfp containing plasmid was injected into wild-type animals (100 ng/ml) along with rol-6(su1006) con-injection marker (for a total of 700 ng), using standard methods (Mello and Fire, 1995). Strains were maintained by picking animals that were rollers as well as positive for GFP. The primers that were used to amplify K08C7.4 promoter and pPD95.75 vector sequences used in the Gibson Assembly are listed in Supplementary file 5.
fmo-2(FAD), fmo-2(NADPH), and fmo-2(FAD+NADPH) strains were generated using CRISPR-Cas9 genome editing as described (Dokshin et al., 2018). Residues for mutation were selected based on protein sequence alignment and as previously reported (Bartsch et al., 2006). To isolate worms with mutated residue(s) in FAD or NADPH motifs, silent mutations that resulted in restriction enzyme sites (PvuII for FAD, and AvaII for NADPH) without any change in the amino acid(s) were created in the repair templates. A PCR fragment spanning the mutated nucleotides was amplified from the progeny of the injected worms, followed by digestion with the above-mentioned restriction enzymes. Mutations in FAD and NADPH motifs were confirmed by sequencing PCR fragments amplified from the corresponding regions in the mutant animals. To generate fmo-2(FAD+NADPH) double mutant, the fmo-2(NADPH) mutant strain was used as a background for a second round of CRISPR microinjections. Sequences for crRNAs, repair templates, and the genotyping primers used for the construction of these strains are listed in Supplementary file 5.

Image analysis
Images were captured using a Lionheart FX Automatic Microscope (BioTek Instruments) under a 4X or 20X objective. Twenty to 50 animals were anesthetized using 20-100 mM NaN 3 on a 2% agarose pad immediately prior to imaging. Comparable images were captured with the same exposure time and magnification. Fluorescence microscopy analysis was independently replicated at least three times. Fluorescence intensity was quantified using ImageJ. Intestinal fluorescent intensity was quantified by outlining the intestine in the DIC images and then quantifying the corresponding fluorescent signals in the GFP images using ImageJ. There are several potential factors that may cause differences between RT-qPCR of fmo-2 transcript and quantification of images of the Pfmo-2::gfp transcriptional reporter: (1) fluorescence microscopy does not possess the same sensitivity and linear range as RT-qPCR, which is the preferred method for expression analysis; (2) the Pfmo-2::gfp transcriptional reporter is a transgene that only contains 1536 bp of upstream (promoter) sequence plus the first 16 codons of fmo-2 (Goh et al., 2018), and thus could be missing regulatory elements that are present in the endogenous (chromosomal) locus; (3) steady-state mRNA levels (measured by RT-qPCR) are the net result of the rates of transcription and degradation, and so mRNA stabilization could play a role in increased fmo-2 expression, which would not be reproduced by the GFP reporter.

Quantification and statistical analysis
Prism 8 (GraphPad) was used for statistical analyses. Survival data were compared using the Log-Rank (Mantel-Cox) test. A p value 0.05 was considered significantly different from control. For comparisons to a single reference, two-sample, two-tailed t tests were performed to evaluate differences between DCt values (Schmittgen and Livak, 2008). For multiple comparisons, statistical significance was examined by one-way ANOVA followed by Š ídá k's post-hoc test. A p value 0.05 was considered significant.