Escherichia coli Metabolite Profiling Leads to the Development of an RNA Interference Strain for Caenorhabditis elegans

The relationship of genotypes to phenotypes can be modified by environmental inputs. Such crucial environmental inputs include metabolic cues derived from microbes living together with animals. Thus, the analysis of genetic effects on animals’ physiology can be confounded by variations in the metabolic profile of microbes. Caenorhabditis elegans exposed to distinct bacterial strains and species exhibit phenotypes different at cellular, developmental, and behavioral levels. Here we reported metabolomic profiles of three Escherichia coli strains, B strain OP50, K-12 strain MG1655, and B-K-12 hybrid strain HB101, as well as different mitochondrial and fat storage phenotypes of C. elegans exposed to MG1655 and HB101 vs. OP50. We found that these metabolic phenotypes of C. elegans are not correlated with overall metabolic patterning of bacterial strains, but their specific metabolites. In particular, the fat storage phenotype is traced to the betaine level in different bacterial strains. HT115 is another K-12 E. coli strain that is commonly utilized to elicit an RNA interference response, and we showed that C. elegans exposed to OP50 and HT115 exhibit differences in mitochondrial morphology and fat storage levels. We thus generated an RNA interference competent OP50 (iOP50) strain that can robustly and consistently knockdown endogenous C. elegans genes in different tissues. Together, these studies suggest the importance of specific bacterial metabolites in regulating the host’s physiology and provide a tool to prevent confounding effects when analyzing genotype-phenotype interactions under different bacterial backgrounds.

Historically, a strain of Escherichia coli, known as OP50 has been chosen for growing C. elegans in laboratory conditions (Brenner 1974). OP50 is a uracil-requiring B-type E. coli and carries a number of positive traits for C. elegans husbandry: it forms a thin bacterial lawn due to its uracil requirements, it is relatively sticky which facilitates worm transfer, and it is mostly considered non-pathogenic (Couillault and Ewbank 2002). Alternatively, an E. coli B-K-12 hybrid strain HB101 and a 'wild type' E. coli K-12 strain MG1655 are both healthier than OP50 and can grow into a thicker lawn to support a larger number of C. elegans (MacNeil and Walhout 2013; Lin and Wang 2017). Furthermore, another E. coli K-12 strain, HT115, has been utilized for C. elegans gene knockdown using RNA interference (RNAi). HT115 has been rendered capable of housing RNAi inducing double stranded RNA through two major mutations, the loss of the rnc allele encoding RNase III, and the introduction of an IPTG inducible T7 polymerase. The phenotypic differences between C. elegans grown on OP50 and alternative -often K-12 derived strains -have been a recent focus of the scientific community. C. elegans have been shown to exhibit bacteriadependent molecular and physiological changes, including alterations to development, innate immunity, lifespan, reproductive lifespan, and metabolism (Coolon et al. 2009;Soukas et al. 2009;Reinke et al. 2010;Pang and Curran 2014;Watson et al. 2014;Sowa et al. 2015;Han et al. 2017;Revtovich et al. 2019). These phenotypic differences in C. elegans are likely associated with metabolic cues derived from different bacterial strains.
In our studies, we have performed high-throughput metabolomic profiling to systemically analyze metabolic differences among three different E. coli strains, OP50, HB101, and MG1655. We found that specific bacterial metabolites, rather than gross metabolic patterns, actively influence phenotypic characteristics of C. elegans grown on these bacterial strains. Given that C. elegans exhibit drastic differences in their mitochondrial morphology, fat storage, and reproductive span when grown on OP50 and HT115, we have generated an additional OP50 strain that can effectively induce RNAi knockdown and confirmed its efficacy for different tissues and for a variety of endogenous genes. This strain shares an ancestry, but differs in its method of creation when compared to the other OP50 RNAi strain generated by the Xu lab (Xiao et al. 2015). Together the metabolite profiles of different bacterial strains and the new OP50 RNAi strain provide useful resources for understanding microbe-host interactions in regulating different physiological activities and for large-scale genetic analysis using RNAi under different bacterial backgrounds.

Bacterial strains used
The following bacterial strains were used in this study. E. coli wild type (MG1655) was a gift from J.J. Wang, and CH1681 was a gift from C. Herman. The E. coli strains OP50 and HB101 were obtained from the CGC. The E. coli strain HT115(DE3) was obtained from the Ahringer (Kamath et al. 2003) RNAi library.
Producing chemically competent iOP50 bacteria All steps are performed using sterile technique. Bacterial iOP50 stocks from -80°were streaked onto Tetracycline (Tet at 50ug/ml) and Chloramphenicol (CAM at 17mg/ml) containing LB plates and grown overnight at 37°. Two tubes of 2ml LB with Tet and CAM (50mg/ml and 17mg/ml) were inoculated with single colonies of iOP50 and grown overnight in a 37°shaker. The next day, 1ml of overnight culture was used to inoculate 100ml of LB without antibiotics. The culture was shaken at 37°for 1.5 to 3 hr, until the OD600 was between 0.2 and 0.5. During the shaking, sterile microcentrifuge tubes (.50), 50ml centrifuge tubes, 100mM CaCl 2 , and 100mM CaCl 2 /15% Glycerol were cooled to 4°. The bacteria were then transferred to two 50ml cold centrifuge tubes. Cells were chilled on ice for 10 min, and all future steps were conducted in the 4°cold room, in a pre-chilled 4°centrifuge, or with the bacteria kept on ice. Cells were harvested by centrifugation at 4000rpm for 3 min, and supernatant was then discarded. Cells were then gently re-suspended in 5ml of cold 100mM CaCl 2 . Cells were incubated on ice for 20 min and then harvested by centrifugation at 4000rpm for 3 min. Cells were returned to ice and the supernatant was discarded. Cells were gently re-suspended in 2ml of cold 100mM CaCl 2 /15% Glycerol and then dispensed into microcentrifuge tubes in 50ml aliquots. Cells are stored at -80°. To use, cells were thawed on ice and transformed with custom plasmids, or those obtained from the Ahringer or Vidal library (Kamath et al. 2003;Rual et al. 2004a).

Transformation of iOP50
Competent cells were transformed using standard methodologies described in the New England Biotechnology protocol (C2987H/C2987I).
Metabolomics analysis E. coli metabolite profiling was performed at Metabolon, Inc (Durham, NC) with 5 biological replicates of each test bacterial strain (MG1655, OP50, HB101). E. coli colonies were inoculated in LB, grown overnight, and then plated on Nematode Growth Medium (NGM) Agar plates to form lawns for three days at 20°. Lawns were then scraped off of the surface of the agar and collected in distilled deionized water. Samples were then washed 3 times. Each sample contained approximately 0.1mg of pooled bacterial pellet which was then flash frozen using liquid nitrogen. Metabolomics profiling was performed in the method described previously (Evans et al. 2009). Statistical analysis was performed using Welch's two-sample t-test to identify compounds that differed significantly between bacterial strains.
Measurement of lipid content levels in C. elegans intestine by SRS microscopy C. elegans (N2) were raised from L1 on the specified food source/ supplement (Betaine at 200mM or Glucose at 0.4% final volume compared to agar). Day-2-old adults ($10 for each of three biological replicates) were anesthetized in 1% sodium azide in M9 buffer (3g KH2PO4, 6g Na2HPO4, 5g NaCl, 1ml 1M MgSO4, H2O to 1 liter, sterilized by autoclaving), and mounted on 2% agarose pads sandwiched between glass microscopic slides and coverslips. Images were taken using the SRS system as described in previous studies (Ramachandran et al., 2015). In brief, when using Stokes beam at 1064nm and pump beam at 817nm, the energy difference between the Stokes and pump photons resonates with the vibrational frequency of CH2 bonds (2845cm-1). Due to the fact that the CH2 chemical bonds are highly enriched in lipid carbon chains, we quantitatively image total fat content levels by detecting the emitted Raman signals of CH2 bonds.

Mitochondrial morphology observation and dynamics measurement
The intestinal mitochondrial morphology was examined using day-3-old adult transgenic C. elegans strain raxIs51[Pges-1::mitogfp] raised from L1 on the specified food source, with the specified supplement (Betaine at 200mM or Glucose at 0.4% final volume compared to agar). Images are presented as single-layer confocal images in anterior regions for consistency. For sample preparation, worms were anesthetized in 1% sodium azide (NaAz) in M9 buffer and mounted on 2% agarose pads sandwiched between glass microscopic slides and coverslips. Confocal images were taken using an IX81 microscope (Olympus) connected to an AxioCam ICc3 camera (Zeiss). For morphological categorization, if the lengths of a majority of mitochondrial filaments in a cell were longer than $4mm, it would be considered "filamented"; if the lengths of a majority of mitochondrial filaments in a cell were shorter than $2mm, it would be considered "fragmented"; the rest belonged to the category "intermediate". Categorical rankings were determined by three blinded scorers. Rankings were only determined for cells on which at least 2 out of 3 of the scorers agreed. An individual cell was utilized as the unit. At least three trials were performed and the accumulated counts from each genotype and environmental condition were summed up across biological replicates. Percentages were calculated and reported as bar representations. The chi-squared test for trend was conducted for statistical analysis.
Inducing RNAi using HT115 or iOP50 HT115 or iOP50 bacterial cultures were grown in LB with Carbenicillin (2.64mM) for 14h, and 18h respectively. 150ml of bacterial culture was then plated on RNAi plates (standard nematode growth medium (NGM) agar plates with IPTG (1mM) and Carbenicillin (2.64mM)) and allowed to dry. Production of dsRNA was induced overnight (18h) at room temperature. C. elegans were then added to the plates.
All noted phenotypes or images were taken in P0 animals grown 62-72h at 20°.

Florescent imaging for determining RNAi efficacy
Well-fed N2 animals were raised for two generations on standard NGM plates seeded with OP50. Animals were synchronized using a variation of the "egg prep" methodology described in (Porta-de-la-Riva et al. 2012). Approximately 100 animals were grown from the L1 stage on RNAi plates with bacteria containing L4440 based plasmids targeting GFP. At 72h post plating, $10 day-1-adult animals were picked into a pool of 8ml 2% sodium azide in M9 buffer located in the center of a freshly made empty NGM plate. Animals were paralyzed, then arranged by gentle prodding and by increasing the surface area of the NaAz pool while allowing it to evaporate. Plates were then immediately imaged using a Hamamatsu digital camera (C11440), utilizing the x-cite xylis light source, and a Nikon SM218 microscope.

Determining RNAi induced phenotypic alterations
Well-fed N2 animals were raised for two generations on standard NGM plates seeded with OP50. Animals were synchronized using a variation of the "egg prep" methodology described in (Porta-de-la-Riva et al. 2012). 6cm plates of bacteria containing L4440 based plasmids targeting the loci of interest from the Ahringer or Vidal library were prepared. Approximately 100 animals were grown from the L1 stage on each of the 6cm RNAi plates. Five independent trials were conducted, and the results were collated. Animals were grown for $72h at 20°. Phenotypes were then noted for all animals on the plates and penetrance was determined.

Methodology for performing Reproductive Lifespans
Reproductive lifespans (RLS) were performed at 20°on animals that were developmentally synchronized using a variation of the "egg prep" methodology described in (Porta-de-la-Riva et al. 2012). Well-fed animals were raised for two generations on standard NGM plates seeded with OP50. The third generation of animals were plated on antibiotic free NGM plates seeded with the test bacteria, and if necessary, an inducible agent (IPTG) or supplement (Betaine at 20mM during development, and 100mM after L4, based upon plate volume). Upon development to the L4 stage, worms were singled to individual plates containing their specified food source. The animals were then transferred to fresh plates every day at the same time until reproduction cessation, determined as the point when no progeny were produced for three consecutive days. Two days after the individual worms were transferred, plates were checked, and double checked for progeny. Animals that bag or die were noted as censors. Reproductive span curves were calculated using Kaplan-Meier survival analysis and compared using the log-rank test. At least two independent trials were conducted.

RESULTS AND DISCUSSION
Revealing metabolomic diversity among different E. coli strains A microbial community carries diverse species of bacteria, and within each species, there are different strains. Interestingly, C. elegans exposed to different strains of E. coli exhibit different physiological phenotypes (Brooks et al. 2009;MacNeil and Walhout 2013;Sowa et al. 2015;Han et al. 2017).
For example, the reproductive lifespan and intestinal fat levels are increased when C. elegans are grown on the OP50 E. coli strain compared to those animals grown on the HB101 or MG1655 E. coli strains (Brooks et al. 2009;Sowa et al. 2015;Lin and Wang 2017). Using stimulated Raman scattering (SRS) microscopy (Yu et al. 2014;Ramachandran et al. 2015), we quantitatively examined fat content levels in C. elegans exposed to those different bacterial strains. In line with previous studies (Brooks et al. 2009), we found that C. elegans on OP50 show more fat storage than those on HB101 or MG1655 in the intestine, the major fat storage tissue ( Figure 1A). In parallel, we examined mitochondrial fission-fusion dynamics which have previously been reported to alter in a bacteria-dependent manner (Reinke et al. 2010;Han et al. 2017;Revtovich et al. 2019), and have been linked to host's lipid metabolism in response to bacterial inputs (Lin and Wang 2017). Using a C. elegans transgenic reporter strain expressing mitochondrial localized GFP (mito-GFP) under the control of an intestine-specific promoter, we imaged mitochondrial morphology and scored the images as one of three distinct categories: filamented, intermediate, and fragmented ( Figure 1B). Similar to the difference in fat storage, worms grown on OP50 show distinct mitochondrial morphology when compared to those grown on HB101 or MG1655, as evidenced by increased mitochondrial fragmentation in the intestine ( Figure 1B).
To understand whether the difference in these physiological features of C. elegans are associated with the difference in the metabolic features of E. coli strains, we systemically analyzed the metabolomic profiles of OP50, HB101, and MG1655 E. coli (Supplemental table 1). To our surprise, PCA analysis of all detected metabolites shows that the metabolic profiles of HB101 and OP50 cluster together more readily than OP50 or HB101 do with MG1655 ( Figure 1C & 1D). Thus the phenotypic difference of C. elegans on different E. coli strains is not simply a response to gross metabolic input alterations in those E. coli strains, but might be associated with specific metabolites derived from those E. coli strains.
We then searched for metabolites that show differences in both HB101 and MG1655 when compared to OP50, which might contribute to the observed C. elegans phenotypic differences. Our analysis identified a total of 42 carbohydrate-related metabolites in our bacterial samples (Supplemental table 1). Twelve of these metabolites show an increase in both HB101 and MG1655 when compared to OP50 and one shows a decrease ( Figure 1E). An inverse relationship is observed in regards to lipid-related metabolites when comparing HB101 and MG1655 to OP50: seven metabolites are differentially observed and six out of the seven exhibit decreased levels in HB101 and MG1655 ( Figure 1E, Supplemental table 1). These results suggest that HB101 and MG1655 may have a higher activity of glycolysis and less fermentation than OP50. Among the groups of amino acids and nucleotides, there are also other specific metabolites that show similar changes in HB101 and MG1655, although there is no clear trend as a group.

Bacterial betaine regulates lipid metabolism in the host
Our previous studies have linked the bacterial one-carbon methyl cycle with lipid metabolism in the host (Lin and Wang 2017), while others have linked this cycle with development and mitochondrial dynamics (Watson et al. 2014;Revtovich et al. 2019). Among the four metabolites directly involved in the methyl cycle (Figure 2A), methionine is decreased by 50% in HB101 compared to OP50, but not in MG1655, neither dimethyglycine nor homocysteine is significantly changed in HB101 and MG1655, but betaine is increased by twofold in HB101 and by threefold in MG1655 compared to OP50 ( Figure 2B). In line with this increase in bacteria, C. elegans grown on the K-12 strain HT115 show increased betaine levels when compared to worms grown on OP50 (Reinke et al. 2010).
Next, we examined whether changes in bacterial betaine levels are associated with fat storage differences in the host C. elegans. Interestingly, the supplementation of exogenous betaine is sufficient to reduce the high fat storage of C. elegans on OP50 to a level comparable to those grown on HB101 or MG1655 ( Figure 2C), while betaine supplementation does not further decrease fat storage in worms on HB101 or MG1655. Given the significant induction of carbohydrate-related metabolites in HB101 and MG1655 compared to OP50 ( Figure 1E), we also examined whether increased sugar levels contribute to the low fat content levels in C. elegans grown on HB101 and MG1655. We supplemented glucose to C. elegans grown on different E. coli strains, and found that the glucose supplementation is not sufficient to suppress the fat storage difference ( Figure 2C). Together, these results suggest betaine as a key metabolite in regulating lipid metabolism in the host C. elegans.
We also examined whether betaine regulates mitochondrial fissionfusion dynamics, a trait that has been linked to host's lipid metabolism in response to bacterial inputs (Lin and Wang 2017). We supplemented betaine to worms grown on different bacterial strains and examined the effect of this supplementation on mitochondrial dynamics. We found that, unlike the fat phenotype, betaine supplementation is not sufficient to suppress the mitochondrial fragmentation phenotype observed in worms grown on OP50 ( Figure 2D). In addition, the supplementation of glucose to worms grown on MG1655, HB101, or OP50 also fails to significantly alter mitochondrial dynamics trends when compared the trends observed in non-supplemented animals ( Figure 1B). Therefore, betaine can regulate lipid metabolism in the host via a mitochondrial dynamics' independent mechanism.
Together the high-throughput metabolite profiles demonstrate that E. coli B and K-12 strains exhibit drastic difference in their metabolism. However these global metabolomics patterns of bacteria are not directly associated with their distinct impacts on the physiology of the host. Instead, specific bacterial metabolites contribute to those differences. In the example put forward, bacterial betaine specifically regulates fat content levels in the host C. elegans but has no effect on mitochondrial dynamics. On the other hand, supplementation of vitamin B12 has been reported to rescue the OP50-induced mitochondrial fission phenotype (Revtovich et al. 2019). Thus, the association between bacterial metabolism and C. elegans physiology is complex and multifaceted, and specific metabolite signals derived from different bacterial strains can be key confounders interfering with the genetic analyses of C. elegans phenotypes. The E. coli strain HT115 is a K-12 strain that has been utilized for over a decade to perform RNAi-based screens, verify mutant phenotypes, and generate hypomorphic conditions for otherwise lethal loss of function mutations (Kamath et al. 2001). However, it has been noted that phenotypes reported in animals raised on OP50 have the potential to differ significantly from the phenotypes of animals grown on HT115 (Reinke et al. 2010;Pang and Curran 2014;Sowa et al. 2015;Xiao et al. 2015;Revtovich et al. 2019).

Development of an RNA interference strain using OP50
Almost two decades ago, two large RNAi libraries were generated by Ahringer and Vidal laboratories (Kamath et al. 2003;Rual et al. 2004b). Both of these libraries house their RNAi vectors in the HT115 (DE3) strain of E. coli, which contains a deletion RNase III (rnc) allele and an IPTG-inducible T7 RNA polymerase, changes which render the individual bacterium capable of producing and maintaining double stranded RNA from the L4440 double-T7 vector (derived from pPD129.36 (Timmons and Fire 1998)). Similar to MG1655 in being Figure 1 Metabolomic profiles of three E. coli strains and their distinct effects on C. elegans fat storage and mitochondrial dynamics. A) Fat content levels of day-2-old C. elegans adults are quantified using SRS microscopy imaging. When compared to those grown on OP50, worms grown on HB101 and MG1655 show decreased fat storage in the intestine. Example SRS images are presented from three independent trials and intestinal areas for quantification are outlined. B) Mitochondrial morphology is examined in day-3-old C. elegans adults carrying a Volume 10 January 2020 | E. coli Metabolomics & New RNA Strain | 193 a K-12 derived strain, HT115 causes reduced fat storage in C. elegans when compared to OP50 ( Figure 3A) and the reduction level is similar to that caused by either HB101 or MG1655 ( Figure 1A). We also examined intestinal mitochondrial morphology and found that worms on HT115 show a more filamented mitochondrial network than that seen in worms on OP50 ( Figure 3B). Similar OP50-induced mitochondrial fragmentation is also observed in C. elegans body-wall muscle cells (Revtovich et al. 2019). In addition, other physiological characteristics are also different between worms grown on OP50 and on HT115, including reproductive lifespan, innate immunity, developmental timing, and lifespan (MacNeil and Walhout 2013; Pang and Curran 2014;Sowa et al. 2015;Revtovich et al. 2019). Therefore, when comparing phenotypes of C. elegans caused by HT115 induced RNAi knockdown, to those exhibited in genetic mutants grown on OP50, the bacterial strain backgrounds introduce an additional confounder that might mislead the interpretation.
To override this problem, we have developed an RNAi competent OP50 bacterial strain using phage transduction of two loci required for Figure 2 Bacterial metabolite Betaine regulates fat storage in the host. A) Simplified one-carbon methyl cycle in E. coli, highlighting the methyl donation interactions between four key metabolites, methionine, homocysteine, dimethylglycine, and betaine. Arrows indicate alterations in MG1655 (green) and HB101 (purple) when compared to OP50. B) Quantitative comparison of one-carbon methyl cycle metabolite levels show that betaine levels are increased over threefold and twofold in MG1655 and HB101, relative to OP50, respectively, and the methionine level in HB101 is decreased compared to that in OP50. C) Supplementation of betaine but not glucose suppresses the fat storage increase in C. elegans grown on OP50. Graphical representation of relative whole intestine fat content levels as measured by SRS microscopy from three independent trials. D) Neither betaine nor glucose supplementation affects morphological difference of mitochondria in C. elegans grown on HB101, MG1655, and OP50. Graphical representation of mitochondrial dynamics in the intestinal cells from double-blind analyses of three independent trials. Statistical significance indicated by asterisk, Ã P = ,0.05, ÃÃ P = ,0.005. Error bars represent SEM. Statistical analysis performed using Student's t-test (B & C), or Chi-Squared test (D). mitochondria-localized-GFP reporter in the intestine, and classified into three categories for quantification: filamented, fragmented, or intermediate. When compared to those grown on OP50, worms grown on HB101 and MG1655 show decreased mitochondrial fragmentation in the intestine. Phenotypes were noted in a double-blind fashion from three independent trials. C) PCA analysis of E. coli metabolomics profiles shows clustering between OP50 and HB101 that is separated from MG1655. D) Clustering of different E. coli samples based upon metabolite levels shows that HB101 and OP50 have a similar pattern distinct to MG1655. E) Heat maps show relative fold changes in a log scale of significantly altered metabolites in MG1655 and HB101 when compared to OP50. Five independent metabolomics profiles for each strain are presented. Statistical significance indicated by asterisk, Ã P = ,0.05, ÃÃ P = ,0.005. Error bars represent SEM. Statistical analysis performed using Welch's two-sample two-sided t-test (E), Student's t-test (A), or Chi-Squared test (B). double stranded RNA production and retention ( Figure 3C). The RNase III (rnc) allele from HT115 (rnc:14::DTn10) was transduced into the CGC supplied "wild type" OP50 bacteria. This allele provides tetracycline resistance, a trait used for selection of bacterial colonies. Allele introduction was verified using PCR of the appropriate loci (Supplemental figure 1). Following introduction of tetracycline resistance, the laczgΑ::T7pol camFRT allele was introduced by phage transduction to facilitate production of double stranded RNA from the L4440 double T7 vector found in both the Ahringer and Vidal RNAi libraries. This allele is selected for using Chloramphenicol, and its presence was further confirmed using PCR (Supplemental figure 1). Following these two phage transduction events, continued rounds of selection on Tetracycline and Chloramphenicol were performed to purify the background of the RNA interference competent OP50 strain (rnc:14::DTn10; laczgΑ::T7pol camFRT) herein termed as RNAi competent OP50 or iOP50.
iOP50 E. coli carrying the L4440 plasmid show a distinct growth pattern when compared to HT115. We grew iOP50 E. coli at 37°overnight in LB with Carbenicillin (50mg/ml) and measured OD hourly ( Figure 3D). We found that a 10 to 14-hour growth period is required for HT115 to enter a stationary phase, while iOP50 requires approximately 18 hr to reach a stationary phase, a growth rate similar to that of non-transformed OP50. Thus, an increased incubation time is necessary for iOP50 strain to provide sufficient, robust, and repeatable RNAi knockdown.

Efficacy of OP50 RNAi strains in gene inactivation
To confirm the knockdown efficacy of iOP50 in different tissues, we first transformed iOP50 with the GFP RNAi plasmid, and supplied those GFP RNAi strains to transgenic C. elegans strains expressing GFP in the intestine, muscle, and hypodermis. We found iOP50-induced knockdown of GFP in a qualitatively comparable level to that observed when inducing knockdown using HT115 ( Figure 4A). Thus, iOP50-mediated RNAi is sufficient to knockdown genes in different tissues. For these experiments, various growth times have been assessed for iOP50; the most robust knockdown is obtained when overnight liquid culture periods range from 18-to 22-hours followed by overnight growth on standard NGM plates with 1mM IPTG (data not shown).
Next, we examined the knockdown efficacy of iOP50 for numerous endogenous genes, where RNAi can induce diverse phenotypes. Figure 3 Development of an RNA interference competent OP50 strain. A) Quantification based on SRS microscopic imaging shows that worms grown on HT115 have decreased fat storage compared to those on OP50. Example SRS images are presented with intestinal areas outlined for quantification. Graphical quantification is from three independent trials. B) Mitochondrial fragmentation, visualized by the intestinal mito-GFP reporter, is reduced in worms grown on HT115 compared to those on OP50. Representative data from two independent trials. C) Schematic for the generation of an RNA-interference-capable OP50 bacterial cell line. D) The growth rate of OP50 is lower than that of HT115. Bacterial growth curves for vector carrying OP50 or HT115 show OD 600 measured every two hours for 20 hr, representative of three independent trials. Statistical significance indicated by asterisk, Ã P = ,0.05, ÃÃ P = ,0.005. Error bars represent SEM. Statistical analysis performed using Student's t-test (A), or Chi-Squared test (B).
When comparing the phenotypes of C. elegans caused by RNAi knockdown using either iOP50 or HT115, we observed two major categories of results, those where knockdown efficacy is comparable between iOP50 and HT115 ( Figure 4B), and those where knockdown using iOP50 gives alternative phenotypes ( Figure 4C). The first class includes K04C2.2, prp-8, nud-1, dpy-13, act-5 and par-1. Among them, RNAi knockdown of prp-8 using HT115 causes developmental arrest by the second larval stage (L2), similarly in iOP50, the majority of animals arrest by L2, although there are few escapers that arrest at the third larval stage (L3). RNAi-induced knockdown of dpy-13 leads to decreased animal body size, but worms are slightly larger when grown on iOP50 than on HT115. The second class includes let-711, gpb-1, npp-9, cars-1 and qars-1. RNAi knockdown of let-711 using HT115 causes over 90% of the worms to arrest at L3, however when using iOP50, over 90% of worms manage to reach adulthood, but become sterile. For RNAi knockdown of gpb-1, the HT115 background generates over 60% dead adults due to explosion through the vulva, but the iOP50 background generates over 60% sterile adults with increased body length. RNAi knockdown of npp-9 using HT115 gives sterile sick adults, but its knockdown using iOP50 gives grossly healthy adults whose progeny is embryonic lethal. For cars-1 or qars-1, its RNAi inactivation causes developmental arrest when using HT115, but completely penetrant adulthood sickliness and partial adulthood sterility when using iOP50. The differences between HT115 and iOP50 RNAi Figure 4 OP50 induced RNA interference is efficient to induce gene knockdown in different tissues. A) Example images of GFP knockdown using RNAi in either HT115 or iOP50 background show efficacy in different tissues including the intestine, hypodermis, or muscle. B) Gene knockdown using RNAi in either HT115 or iOP50 background gives almost indistinguishable phenotypes. C) A number of RNAi lines cause variable phenotype when using HT115 vs. iOP50, with less severity using iOP50 than using HT115. might be related to the strength of gene knockdown, which may be weaker with iOP50, and might also be related to the specific role of certain C. elegans genes in response to different bacterial strains. Together, these studies demonstrate the efficacy of iOP50 in executing RNAi knockdown in different tissues and for various genes, and also highlight the importance of examining the effect of C. elegans genes under different bacterial strain backgrounds.

CONCLUSION
In summary, C. elegans exhibit drastically different phenotypes when exposed to different bacterial strains, which are unrelated to genetic alterations in C. elegans. These differences could introduce confounders that complicate analysis of genetic regulation, but also provide researchers the opportunity to investigate environment-microbe-host interactions. Our metabolomics studies systematically reveal metabolite difference among different E. coli strains and found that C. elegans phenotypic changes are directly associated with specific bacterial metabolites. In particular, we have linked betaine, a metabolite derived from the methyl cycle, with lipid metabolism in worms, which does not contribute to mitochondrial dynamics or reproductive span (Supplemental figure 2). This one-to-one relationship between a microbial metabolite and a host's phenotype highlights the importance of investigating the mechanistic impact of microbial metabolism on host's physiology beyond profiling microbial phylogenetic composition. Moreover, to expand the analysis toolkit of genetic regulation and microbe-host interaction, we have rendered the common E. coli strain OP50 competent for the induction of the RNAi response in C. elegans. This bacterial strain contains the same two alleles that render HT115 capable of inducing the RNAi response. These changes allow iOP50 to induce RNAi in multiple tissues to a level similar to that found in HT115fed animals. Both OP50 (Xu363) (Xiao et al. 2015) and iOP50 have been used as the basis for an RNAi screen (96 wells) and have proven amenable to en-mass transformation. Phenotypes observed and described in the iOP50 screen hold true for the reciprocal HT115 screen (data not shown). The strain is available by request and has been deposited with the CGC for academic use. If a mutant C. elegans strain fails to demonstrate the same phenotype as an RNAi-fed animal or vice-versa, the bacterial strain and specific metabolites may be responsible for the discrepancy, and iOP50 could facilitate phenotypic validation.