Colonization of the Caenorhabditis elegans gut with human enteric bacterial pathogens leads to proteostasis disruption that is rescued by butyrate

Protein conformational diseases are characterized by misfolding and toxic aggregation of metastable proteins, often culminating in neurodegeneration. Enteric bacteria influence the pathogenesis of neurodegenerative diseases; however, the complexity of the human microbiome hinders our understanding of how individual microbes influence these diseases. Disruption of host protein homeostasis, or proteostasis, affects the onset and progression of these diseases. To investigate the effect of bacteria on host proteostasis, we used Caenorhabditis elegans expressing tissue-specific polyglutamine reporters that detect changes in the protein folding environment. We found that colonization of the C. elegans gut with enteric bacterial pathogens disrupted proteostasis in the intestine, muscle, neurons, and the gonad, while the presence of bacteria that conditionally synthesize butyrate, a molecule previously shown to be beneficial in neurodegenerative disease models, suppressed aggregation and the associated proteotoxicity. Co-colonization with this butyrogenic strain suppressed bacteria-induced protein aggregation, emphasizing the importance of microbial interaction and its impact on host proteostasis. Further experiments demonstrated that the beneficial effect of butyrate depended on the bacteria that colonized the gut and that this protective effect required SKN-1/Nrf2 and DAF-16/FOXO transcription factors. We also found that bacteria-derived protein aggregates contribute to the observed disruption of host proteostasis. Together, these results reveal the significance of enteric infection and gut dysbiosis on the pathogenesis of protein conformational diseases and demonstrate the potential of using butyrate-producing microbes as a preventative and treatment strategy for neurodegenerative disease.


Introduction
Neurodegenerative protein conformational diseases (PCDs), including amyotrophic lateral sclerosis (ALS), Alzheimer's, Huntington's, and Parkinson's disease, are characterized by the misfolding and aggregation of metastable proteins that reside within the proteome, often resulting in loss of tissue function that manifests in disease progression [1]. Despite the high prevalence and enormous financial and social burdens imposed on afflicted individuals and their families [2], no effective treatment or cure has been found; moreover, the etiology of these diseases remains largely unknown [3]. Factors such as age, diet, stress, trauma, toxins, infections, or antibiotics, have been shown to increase the risk of PCDs [4][5][6][7][8][9][10]. Notably, these triggers are also associated with changes in the microbiome [11][12][13], suggesting that bacteria may contribute to the pathogenesis of PCDs, which may explain their sporadic onset. Nonetheless, the relationship between the microbiome and disease progression remains poorly defined.
The human gut microbiota (HGM) is a complex assembly of microorganisms capable of synthesizing molecules that can impact the host's physiology [14]. A healthy commensal relationship between a balanced microbiome and the host is fostered by their cross-talk [15]. Conversely, gut dysbiosis results from the loss of beneficial bacteria accompanied by the overgrowth of opportunistic pathogens, a condition often induced by antibiotics [16,17]. Recent evidence has established a link between the HGM and PCDs, whereby gut dysbiosis or direct enteric infection exacerbate the disease [18][19][20]. Commensal residents of the human microbiota, specifically those involved in the synthesis of short-chain fatty acids (SCFAs) such as butyrate, were shown to be beneficial to the host [21]. Unfortunately, environmental factors and the complexity of the human microbiome often hinder the consistency of results from change between E. coli OP50 and P. aeruginosa PAO1 -two bacterial strains that elicited the lowest and highest intestinal aggregation, respectively (S1B Fig). Moreover, western blot analysis of the insoluble polyQ44::YFP fraction confirmed bacteria-mediated increase in aggregation (S2A Fig), as it parallels the aggregation profile of worms fed E. coli OP50 and P. aeruginosa PAO1 (Fig 1B). These results are intriguing as they demonstrate that common human enteric pathogens significantly affect polyQ aggregation, and therefore host proteostasis. We next asked whether gram-negative bacteria that are associated with commensal microflora would also enhance aggregation. We tested three strains known to be part of the commensal microbiome: K. oxytoca HM-624, Prevotella disiens HM-1171, and Prevotella . Each data point is an average of a minimum of three independent experiments with a total of at least 100 animals. Black circles represent control bacteria and bacteria that did not have any significant effect on polyglutamine aggregation upon colonization of the C. elegans intestine. Red solid circles represent bacteria that significantly enhanced aggregation by <3-fold. Red open circles represent bacteria that significantly enhanced aggregation by >3-fold (arbitrary threshold). Arrows represent bacteria that were chosen for follow-up experiments. C) The effect of select commensal bacteria on polyQ aggregation in the C. elegans intestine. Data are represented as the average number of aggregates per C. elegans intestine normalized to E. coli OP50 control strain. Each data point is an average of a minimum of three independent experiments with a total of at least 100 animals. D) PCR confirmation of E. coli MG1655 curli mutant strains. WT, csgA::kan (left two bands) amplified with primers flanking the ΔcsgA locus. WT, csgD::kan (right two bans) amplified with primers flanking the ΔcsgD locus. E) Phenotypic confirmation of the curli-deficient ΔcsgA and ΔcsgD mutant strains using Congo Red plate assay. F) ProteoStat staining of total aggregates produced by E. coli MG1655 wild-type (WT), MG1655 ΔcsgA, and MG1655 ΔcsgD. Data are represented as the average fluorescent signal per bacterial strain stained with ProteoStat. Each data point is an average of two independent experiments with three replicates per run. G) The effect of E. coli MG1655 curli mutants on polyQ (polyQ44) aggregation in the C. elegans intestine. Data are represented as the average number of aggregates per C. elegans intestine. Each data point is an average of three independent experiments with a total of 90 animals. Error bars represent standard error of the mean (SEM). Statistical significance was calculated using one-way analysis of variance (ANOVA) followed by multiple comparison Dunnett's post-hoc test ( ���� p< 0.0001). https://doi.org/10.1371/journal.ppat.1009510.g001

PLOS PATHOGENS
corporis HM-1294. While K. oxytoca enhanced aggregation, as we have seen with the other Klebsiella spp., both Prevotella spp. suppressed aggregation relative to E. coli OP50 (Fig 1C). In fact, P. corporis almost completely suppressed aggregation.
To begin understanding how bacteria could contribute to the disruption of host proteostasis, we used E. coli defective in the production of curli, which are aggregation-prone amyloids that are specific to Enterobacteriaceae [24]. We generated two mutant E. coli strains: one deficient in csgA, which encodes an amyloid structural subunit of curli fimbriae, and one deficient in csgD, a transcription factor that positively regulates the curli operon (Fig 1D). While staining with congo red confirmed a deficiency in curli production (Fig 1E), staining for bacterial protein aggregates using ProteoStat, which provides superior sensitivity in aggregate detection [25], revealed that only csgD mutant cells have significantly fewer aggregates (Fig 1F). In agreement with these results, we found that colonization of C. elegans only with the ΔcsgD E. coli strain led to a significant decrease in the number of polyQ aggregates per intestine (Fig 1G). These results suggest that bacteria-derived aggregates significantly contribute to the disruption of host proteostasis.
Tissue non-autonomous effects between C. elegans neurons and other somatic cells (i.e., intestine and epithelium) have been recently described where perturbation of proteostasis in one tissue impacts another [26][27][28]. As such, we postulated that bacteria-induced disruption of proteostasis in the intestine could affect other tissues. To begin assessing the impact of bacteria on muscle, we examined motility. The nematodes that express intestinal polyQ have an integrated rol-6 marker, which results in a roller phenotype and renders animals incompatible with known motility assays, such as thrashing. Thus, we developed a novel motility readout that entails sliding an eyebrow hair under the mid-section of the worm and counting the number of seconds until the worm completely crawls off-we refer to this phenotype as time-offpick (TOP) (Fig 2A). We found that TOP increases with age and the decrease in motility is dependent on polyQ when compared to control roller worms that do not express polyQs, as well as those that express non-aggregating polyQs (Fig 2B). To ensure that the TOP method measures polyQ-specific defects and not an undesired effect related to the roller phenotype or bacterial pathogenicity, we used non-roller strains that express muscle-specific polyQ35 (unc-54p::Q35::YFP, AM140) and polyQ40 (unc-54p::Q40::YFP, AM141). Both of these strains are known to exhibit polyQ-length and age-dependent motility defects [29]. In support of the functionality of our assay, we found that worms expressing these constructs also exhibited a significant polyQ-length and age-dependent TOP motility defect that correlated with the enumeration of body bends, which is a well-established assessment of motility in C. elegans (S3A Fig). Additionally, colonization of these polyQ-expressing strains with P. aeruginosa PAO1 inhibited motility in both assays yet had little to no effect in wild-type N2 animals. These findings indicate that polyQ toxicity, and not bacterial pathogenicity, is the factor that is measured by TOP (S3B Fig). Therefore, we further used the TOP method to determine whether colonization of the C. elegans intestine with bacteria that induced polyQ aggregation aggravates this phenotype. We chose two bacterial strains: K. pneumoniae KP182 (medium inducer); P. aeruginosa PAO1 (high inducer) and found that the "time-off-pick" was significantly increased by both strains (Fig 2C), supporting the trend seen in their respective aggregation profiles (Fig 1B). We did not detect any effect on the TOP phenotype in roller controls or in roller nematodes expressing a non-aggregating and intestine-specific polyQ33, confirming that this motility defect was the result of bacteria-induced polyQ-dependent toxicity rather than general pathogenicity from a bacterial infection (Fig 2C). While it is known that worms expressing polyQ33 do not exhibit age-dependent aggregation, we wanted to ensure that aggregation does not occur when these animals are colonized by bacteria that induce polyQ44 aggregation. We did not detect polyQ33 aggregates in animals colonized with the strongest inducer, P. aeruginosa PAO1 (S4 Fig). Based on our results, we conclude that bacterial colonization of the C. elegans intestine affects protein folding in the immediate environment and that such localized intestinal protein aggregation influences motility, likely by acting on muscle or neuronal tissues.

Bacterial colonization of the C. elegans intestine disrupts protein folding in distal tissues
Muscle tissue. Because bacteria-induced polyQ aggregation in the intestine influenced C. elegans motility, an effect that likely involves distal tissues, we sought to determine whether the influence of bacteria on protein aggregation is confined to the intestine or if bacteria can impact protein folding across other tissues. We employed nematodes expressing muscle-specific polyQ35 (unc-54p::Q35::YFP) as a proxy to detect changes in proteostasis in that tissue. Different tissues have different thresholds for polyQ aggregation; therefore, while polyQ33 did not aggregate in the intestine, polyQ35 aggregates in the muscle. This reporter exhibits age-dependent and muscle-specific aggregation and the associated motility defect [29]. We colonized the intestines of these worms with two strains that enhanced intestinal polyQ aggregation: K. pneumoniae KP182 and P. aeruginosa PAO1. Animals expressing muscle polyQs exhibit a more robust aggregation profile compared to the intestinal polyQ model; thus, we had to quantify the aggregates on day three instead of day four. The results paralleled the trend seen in the aggregation profile of the intestinal polyQ model, where K. pneumoniae enhanced aggregation compared to control E. coli OP50 and P. aeruginosa PAO1 showed the most robust enhancement (Fig 3A). To determine whether the bacteria-induced proteostasis imbalance detected by the polyQ aggregation in the muscle also leads to tissue dysfunction, we employed a previously characterized motility readout to enumerate the body bends of polyQ-expressing worms fed these two bacterial strains. The animals expressing muscle polyQ exhibit a sharp and age-dependent decrease in body bends (Fig 3B). We found that colonization of the C. elegans intestine with enteric bacterial pathogens led to a significant defect in motility, particularly in worms fed the most potent enhancer of aggregation, P. aeruginosa, which decreased the number of body bends 2-fold when compared to the control E. coli OP50; a larger and more significant defect than what was observed in control wild-type animals (Fig 3C and 3D).
Neuronal tissue. To determine whether proteostasis in neuronal tissue is also affected by bacterial colonization of the intestine, we utilized worms carrying neuronal polyQ40 (F25B3.3p::Q40::YFP) [30]. The small size and high concentration of neurons preclude us from accurately quantifying the exact number of aggregates in this tissue; as such, we measured toxicity by assessing TOP motility. Animals carrying this reporter exhibit an age-dependent decrease in motility compared to control N2 worms (Fig 3E). We found that colonization of the intestine with K. pneumoniae and P. aeruginosa led to a significant defect in motility when compared to the control E. coli OP50 (Fig 3F and 3G).
Gonad. Thus far, we demonstrated that colonization of the C. elegans gut with enteric bacterial pathogens leads to proteostasis imbalance and proteotoxicity in proximal tissues, including muscle and neurons. To further examine the extent to which gut bacteria influence host Each bar is an average of three independent experiments with a total of 100 animals. B) Age-dependent decline in motility assessed by enumerating body bends per 30 seconds in muscle-specific polyQ35 and N2 control worms. The data are represented as the average number of body bends quantified on day 3-5 posthatching in a total of 15 animals per each day. The effect of bacteria on the motility of C) control animals and D) animals expressing muscle-specific polyQ35. Data are represented as the average number of body bends per worm normalized to worms fed E. coli OP50 control strain. Each bar is an average of three independent experiments with a total of 45 animals. E) Age-dependent decline in motility assessed by increased TOP. The effect of bacteria on the motility of F) control animals and G) animals expressing neuron-specific polyQ40. Data are represented as the average TOP per worm normalized to animals fed E. coli OP50 control strain. Each bar is an average of three independent experiments with a total of 60 animals. H) A cartoon depicting experiments which demonstrated that bacterial colonization of the C. elegans intestine affects the F1 generation. I) Quantification of intestinal aggregates in the F1 and F2 generations from parental animals that were fed select test and control bacteria. Each bar represents the average number of aggregates per intestine normalized to the control (E. coli OP50). Data are representative of three (F1) and one (F2) independent experiments with a total of 100 and 30 animals, respectively. Error bars represent SEM. Statistical significance was calculated using one-way ANOVA followed by multiple comparison Dunnett's post-hoc test ( � p<0.05, �� p<0.01, ���� p<0.0001).
https://doi.org/10.1371/journal.ppat.1009510.g003 proteostasis, we asked whether colonization of the C. elegans intestine had any effect on the gonad. If the colonization of the intestine affects the muscle and neurons, we reasoned that bacteria might also affect proteostasis in the germline-if this is the case, then we should be able to detect changes in polyQ aggregation in the progeny of parents that were colonized by pathogenic bacteria. To test the effect of bacteria on the next generation, we fed C. elegans either control E. coli OP50, K. pneumoniae KP182, or P. aeruginosa PAO1 for three days, followed by "bleaching" adults, and age-synchronizing the progeny. C. elegans synchronization involves dissolving gravid adults by hypochlorite treatment leaving intact and sterile embryos which hatch into L1 larvae. Synchronized F1 animals were placed on fresh plates containing a lawn of control E. coli OP50 and intestinal aggregates were enumerated after 92 hours (h). The F2 generation was isolated and assessed using the same method (Fig 3H). Strikingly, we found that colonization of the parental intestines with the select bacteria enhanced polyQ aggregation in the F1 progeny, despite the fact that they never encountered the bacteria that were fed to the parental generation (with the exception of E. coli OP50) (S5 Fig). The extent of the enhancement in the number of polyQ aggregates per worm correlated with previously assessed phenotypes where P. aeruginosa showed the highest induction of aggregation (Fig 3I). Since P. aeruginosa induced the strongest effect, we looked at the F2 progeny from a parental generation fed these bacteria. We saw no enhancement of aggregation compared to control E. coli, indicating that only F1 generation is affected (Fig 3I). Using western blotting, we ruled out the possibility that these results were due to changes in polyQ44::YFP expression level (S1E

Butyrate suppresses bacteria-induced polyglutamine aggregation and aggregate-dependent toxicity across tissues
The depletion of butyrogenic bacteria in the gut establishes an environment in which pathogenic bacteria, particularly Enterobacteriaceae, can flourish [31]. In addition, butyrate itself has been found to suppress the growth of pathogenic bacteria and was shown to protect against neurodegeneration [32][33][34]. To assess the effect of exogenous butyrate on the observed bacteria-induced aggregation, we first tested its effect on intestinal polyQ aggregation in C. elegans. We used all bacterial genera that induced protein aggregation by three-fold or more (Fig 1B). The physiological concentration of human colonic butyrate is estimated to be 10-20 mM, and a therapeutic dose of 150 mM butyrate was safely delivered via enema directly into the human large intestine [35,36]. Therefore, we tested the effect of butyrate on bacteria-induced protein aggregation at concentrations ranging from 10-100 mM. Unexpectedly, we found that lower concentrations of butyrate differentially affected host proteostasis where enhancement of aggregation was dependent on the type of bacteria colonizing the intestine: 10 mM: S. enterica 12023, P. mirabilis; 25 mM: P. mirabilis, P. aeruginosa PAO1; 50 mM: P. aeruginosa PAO1. However, at higher butyrate concentrations (>25 mM), bacteria-induced aggregation was suppressed in a dose-dependent manner in worms colonized with all bacteria except P. aeruginosa. Supplementation with 100 mM butyrate suppressed bacteria-induced aggregation across all strains. (Fig 4). These results are unexpected, as they suggest that at low physiological concentrations (<25 mM), butyrate may contribute to the bacteria-mediated disruption of host proteostasis, as demonstrated by differentially enhanced polyQ aggregation; however, with the exception of P. aeruginosa PAO1, higher concentrations (50-100 mM) of butyrate suppressed aggregation induced by all other bacterial strains, which suggests therapeutic potential. We employed western blot analysis of polyQ44::YFP levels to eliminate the possibility that these results are due to changes in protein level (S1C and S1D Fig), and we confirmed these results by assessing insoluble fractions (S2B and S2C Fig).
Furthermore, to rule out the possibility that butyrate decreased bacteria-mediated protein aggregation by merely killing bacteria, we tested the effect of butyrate on bacterial growth and viability in liquid medium and on solid NGM plates, respectively, at concentrations up to 100 mM. While butyrate did affect the growth and viability of some bacteria, all of the strains tested remained viable. We found that most overnight cultures supplemented with butyrate at

PLOS PATHOGENS
concentrations that inhibited aggregation reached optical densities comparable to unsupplemented controls (S6 Fig). We also tested the effect of butyrate on bacteria cultured on solid NGM media and found that, while some strains were affected, all were viable across concentrations (S7 Fig). Additionally, feeding C. elegans dead bacteria resulted in higher aggregation profiles that are distinct from those treated with 100 mM butyrate, further suggesting that butyrate does not suppress bacteria-mediated aggregation by solely affecting bacterial viability (S8 Fig). Next, we asked whether butyrate could affect colonization of the intestine. We found that at 25 mM, butyrate significantly enhanced intestinal colonization by E. coli OP50, but exhibited no effect on colonization at 100 mM (S9 Fig). These results indicate that it is unlikely that the butyrate's effect on bacterial colonization affects aggregation.
A trade-off between organismal resources available to maintain either proteostasis or fecundity has been proposed [37]. As such, we asked whether butyrate affects fecundity in our experiments, which may explain the suppression of aggregation. Indeed, we found that butyrate decreases the number of progeny per worm, but the effect seems to depend on the presence of polyQ (S10A and S10B Fig). Moreover, 25 mM butyrate significantly decreased the number of progeny in worms colonized by P. aeruginosa (S10C Fig). Yet, we found that under the same conditions, butyrate increased polyQ aggregation in animals colonized by P. aeruginosa (Figs 4 and S2C). These results indicate that a decrease in fecundity alone does not explain the observed suppression of polyQ aggregation.
To determine whether exogenous butyrate would also rescue the aggregate-dependent motility defect observed in the intestinal polyQ44 model, we measured the TOP phenotype in animals colonized by E. coli OP50, K. pneumoniae KP182, and P. aeruginosa PAO1 in the absence or presence of 100 mM butyrate. While little to no changes in TOP were observed in the roller controls, exogenous butyrate supplementation (100 mM) rescued motility defects in intestinal polyQ44 worms (Figs 2C and 5). We further explored the ability of butyrate to suppress aggregation in other tissues by repeating the above experiment using the muscle polyQ model and assessed the effect of butyrate on bacteria-induced protein aggregation at

PLOS PATHOGENS
concentrations ranging from 0-50 mM. Unlike the intestinal polyQ44 model, worms expressing muscle polyQ35 did not experience a significant enhancement of aggregation at lower butyrate concentrations and exhibited a dose-response or near dose-response suppression of aggregation across all bacterial strains except for P. mirabilis and P. aeruginosa PAO1, wherein the only suppressive concentrations of butyrate were 25 and 50 mM, respectively (Fig 6). It is worth noting that the trends of butyrate-suppressed aggregation in worms expressing muscle polyQ parallel those seen in worms expressing intestinal polyQs (Fig 4). For example, animals

PLOS PATHOGENS
cultured on plates containing 0-50 mM butyrate exhibited a significant dose-dependent decrease in intestinal and muscle aggregation when colonized by E. coli OP50 or Klebsiella spp. On the other hand, lower concentrations of butyrate did not significantly suppress aggregation in either tissue when animals were fed P. aeruginosa PAO1, S. enterica 12023, and P. mirabilis, suggesting that the beneficial effect of butyrate is, at least in part, facilitated by bacteria (Figs 4 and 6).
To gain a mechanistic understanding of how butyrate suppresses bacteria-mediated aggregation, we investigated the dependence of the observed response on three major evolutionaryconserved transcription factors, HSF-1, SKN-1/Nrf2, and DAF-16/FOXO. Each of these transcription factors was previously shown to provide protection against proteotoxicity [38][39][40]. Because butyrate suppressed aggregation upon colonization with multiple strains of bacteria, including E. coli OP50 (Fig 4), we investigated the dependence of butyrate-mediated suppression of aggregation on the abovementioned transcription factors by knocking-down each candidate using RNAi. C. elegans carrying intestinal polyQ44 that were fed E. coli HT115 (DE3) expressing empty vector control (L4440) RNAi, exhibited a significant suppression of aggregation in the presence of butyrate (10-100 mM) (Fig 7A). While butyrate-mediated suppression of aggregation was not affected when HSF-1 was knocked-down, the beneficial effect of butyrate was abolished upon downregulation of SKN-1/Nrf2 and DAF-16/FOXO transcription factors. As expected, HSF-1 knockdown enhanced aggregation in the absence of butyrate when compared to control animals (Fig 7A) [41]. To confirm that the knockdown is functional, we used fluorescent reporters to assess the expression of downstream target genes of HSF-1 (hsp70p::GFP also known as C12C8.1p::GFP) and SKN-1 (gcs-1p::GFP), and we used DAF-16

PLOS PATHOGENS
(DAF-16::GFP) fluorescent fusion (Fig 7B). The expression of all fluorescent reporters was attenuated upon knockdown of the respective target genes.
Together, these results indicate that butyrate can suppress bacteria-induced polyQ aggregation across tissues and rescues aggregate-dependent toxicity in the host. Such protective effect of butyrate seems to be mediated by SKN-1 and DAF-16 transcription factors.
Colonization of the C. elegans intestine with E. coli engineered to overproduce butyrate suppresses protein aggregation across C. elegans tissues SCFA-producing bacteria are common residents of the human gut microbiota. Many of these bacteria, such as those belonging to the Firmicutes phylum, are known to produce butyrate [42]. Butyrate has been demonstrated to provide numerous health benefits against a variety of ailments, and neurodegenerative diseases are no exception [32]. To determine the effect that butyrogenic bacteria have on organismal proteostasis, we used an E. coli strain, LW393 (E. coli Bt ), engineered to conditionally overproduce butyrate [43]. Colonizing C. elegans intestines with this strain allows us to model simplified physiological, yet controlled, conditions. The strain was derived from E. coli W (E. coli WT ), a non-pathogenic strain closely related to commensal bacteria by first creating a quadruple deletion mutant (ldhA, frdABCD, ackA, and adhE), E. coli BEM9 (E. coli Δ ) strain, followed by overexpression of enzymes that channel intermediates into the pathway and which are essential for butyrate synthesis [43]. Therefore, we used E. coli WT and, in some instances, E. coli Δ , as controls. E. coli Bt synthesizes butyrate in the presence of 5-and 6-carbon sugar substrates to levels up to three times higher than any other available commercial strain [43]. We first used this strain to colonize the intestine of C. elegans expressing the intestinal polyQ44 reporter over a period of four days. We demonstrated that, on days three and four, C. elegans intestines are colonized by viable bacteria (S11 Fig). While glucose is a suitable substrate for butyrate synthesis, we found that it affects polyQ aggregation and thus cannot be utilized in our experiments (S12 Fig); however, L-arabinose is another butyrate synthesis substrate and does not affect aggregation. To confirm that L-arabinose from the NGM plates is available to intestinal bacteria, we colonized C. elegans intestines with bacteria that express an arabinose-inducible fluorescent reporter and cultured them in the absence of the sugar. We then transferred these animals onto plates containing E. coli OP50 and 3% Larabinose and detected a strong induction of fluorescence, indicating that L-arabinose is accessible to the intestinal bacteria (S13 Fig). We supplemented NGM plates with 3% L-arabinose and quantified the number of aggregates after four days feeding polyQ-expressing animals the three control (E. coli OP50, E. coli WT , and E. coli Δ ) and the butyrate-producing (E. coli Bt ) strains. We did not detect any effect on polyQ aggregation in worms colonized with each of the control E. coli strains; however, supplementation of L-arabinose did result in significant suppression of polyQ aggregation in animals colonized by E. coli Bt (Fig 8A). To determine whether the aggregate-suppressing effect of E. coli Bt would also rescue polyQ-associated motility defects, we measured the TOP phenotype in worms expressing intestinal polyQ44. In agreement with the observed suppression of aggregation, supplementation of L-arabinose to animals colonized by E. coli Bt also improved motility, indicating that bacteria-derived butyrate can rescue aggregation and aggregate-associate toxicity (Fig 8B).
We further investigated whether the effect of endogenous butyrate produced by E. coli Bt strain was limited to the intestine or was ubiquitous across tissues, as seen when the animals were exposed to exogenous butyrate. To accomplish this, we colonized the intestines of animals expressing muscle and neuronal polyQ with E. coli Bt . While supplementation of 3% Larabinose into NGM plates did not affect intestinal polyQ aggregation in animals colonized by control E. coli strains (Fig 8A), it did enhance aggregation in animals expressing muscle-specific polyQs (Fig 9A). Intriguingly, this enhancement of aggregation was significantly reduced in the presence of L-arabinose in animals colonized by E. coli Bt , suggesting that endogenously produced butyrate suppresses protein aggregation in the muscle tissue as well (Fig 9A). We then assessed the effects of this strain on aggregate-dependent toxicity in this model utilizing the body bend assay. Our results show that worms expressing muscle-specific polyQ have 2.5x more body bends when E. coli Bt is endogenously producing butyrate, compared to worms colonized with non-butyrogenic bacteria or butyrogenic bacteria in the absence of the sugar substrate (Fig 9B). The observed beneficial effect of endogenous butyrate is dependent on the presence of polyQ since endogenous butyrate did not increase the number of body bends in control wild-type worms (Fig 9C). In fact, the presence of endogenous butyrate was slight, but significantly detrimental to wild-type C. elegans. These results support the beneficial effect of butyrate in the suppression of aggregation and the associated proteotoxicity in the muscle tissues.
We expanded our investigation to animals carrying a neuronal polyQ40 reporter by measuring toxicity using the TOP phenotypic readout. We employed this method to assess the effect of endogenous butyrate on aggregate-dependent toxicity. The availability of L-arabinose to control E. coli strains (E. coli OP50 and E. coli WT ) did not improve motility. Still, the butyrate produced by E. coli Bt significantly enhanced motility in animals carrying the neuronal polyQ reporter (Fig 9D), but not in N2 control worms (Fig 9E). While further investigation is needed to elucidate the basis of this response, the result could be key to understanding the mechanism(s) by which butyrogenic bacteria affect host proteostasis.

Co-colonization of the C. elegans intestine with butyrogenic bacteria suppresses bacteria-induced aggregation
The human gut microbiota is a polymicrobial community and the effect of the interactions between individual strains and the host is hindered by the complexity and wealth of its composition. We co-colonized the C. elegans gut with pathogenic and butyrogenic bacteria as a strategy to begin deconvoluting complex polymicrobial interactions within the human

PLOS PATHOGENS
microbiome and their effect on the host. We fed worms cultures of E. coli Bt with either E. coli OP50 as a control or P. aeruginosa PAO1, which showed the most robust enhancement of aggregation. In the presence of L-arabinose, we found a significant decrease in aggregation of intestinal polyQs in co-cultures with both strains (Fig 10A). E. coli Bt suppressed E. coli OP50and P. aeruginosa PAO1-mediated aggregation by approximately 4-and 2-fold, respectively. These results indicate that endogenous butyrate synthesized by butyrogenic bacteria can enhance host proteostasis and suppress protein misfolding. Furthermore, our experiments demonstrate the potential benefits of butyrate and butyrogenic bacteria in the suppression of bacteria-induced proteotoxicity associated with metastable proteins and also emphasize the importance of having a balance between commensal butyrogenic and enteropathogenic microbes (Fig 10B).

Discussion
We used a simple metazoan model, C. elegans, to study the effect of bacterial gut colonization on organismal proteostasis. We found that the colonization of the C. elegans intestine with enteric bacterial pathogens disrupts organismal proteostasis and enhances proteotoxicity associated with polyQ. Importantly, our results demonstrate that bacteria alone do not cause major toxicity in control animals that do not express polyQ, which would hinder the interpretation of our data (Figs 2B, 2C, 3C, and 3F). It is not surprising that the detrimental effect of bacteria on polyQ aggregation is observed in the most proximal tissue, the intestine, as similar results were previously observed [23]. However, to our knowledge, this is the first report of bacteria having detrimental, tissue non-autonomous effects on protein folding. Recently, Bacillus subtilis, a probiotic strain that is also part of the human commensal microbiota, was shown to inhibit α-synuclein aggregation in C. elegans muscle [44]. The fact that certain bacteria disrupt host proteostasis while others can enhance it emphasizes the importance of microbes in the pathogenesis of PCDs. Previous research suggests that increased presence of bacterial species such as K. pneumoniae and P. aeruginosa in the human gut is associated with an increased prevalence of PCDs [45,46]. Interestingly, among all of the strains that we tested, Klebsiella spp. and P. aeruginosa were the most potent inducers of polyQ aggregation-this was the case even for a commensal strain, K. oxytoca (Fig 1B and 1C). Both of these species are ubiquitous in the environment, and multidrug-resistant strains of these bacteria are often associated with nosocomial and opportunistic infections [47]. Due to their rapidly increasing resistance to antibiotics, they are classified by the Centers for Disease Control and Prevention as serious threats [48]. While these strains are often associated with infections, their enteric presence is also evident in healthy individuals and may possibly support their contribution to neurodegenerative diseases [49,50]. In fact, both of these microbes have recently been associated with neurodegenerative diseases in humans [45,46], a correlation that is in agreement with our results. The increasing prevalence of antibiotic resistance among Klebsiella spp. and P. aeruginosa strains may specifically enrich for their growth within the human microbiome upon an antibiotic treatment while decreasing the abundance of beneficial bacteria. In agreement with

PLOS PATHOGENS
this hypothesis, population-based studies of patients with Parkinson's and ALS revealed that a history of antibiotic treatment is associated with an elevated risk for these diseases [7,8].
Another notable strain that induced aggregation by more than three-fold is P. mirabilis. This gram-negative bacterium is part of human commensal microbiota but can also become an opportunistic pathogen, most often leading to urinary tract infections, and eventually, bacteremia [51]. Interestingly, in one study, blood cultures of PD septic patients had a five-fold higher prevalence of P. mirabilis compared to non-PD septic patients [52]. Recently, it was demonstrated in a mouse model that P. mirabilis contributes to PD pathogenesis, including the induction of α-synuclein in the mouse brain [53]. Furthermore, colonization of the gut with P. mirabilis and Enterobacteriaceae, in particular, K. pneumoniae, was shown to be accelerated by antibiotic treatment, leading to detrimental dysbiosis in mouse microbiota [54]. On the other side of the aggregation spectrum, the non-pathogenic E. coli HB101 did not significantly enhance aggregation in our C. elegans polyQ model. The E. coli HB101 strain is also a common laboratory food for C. elegans; hence, we did not expect any enhancement with these bacteria. However, both P. disiens HM-1171 and P. corporis HM-1294 suppressed aggregation relative to E. coli OP50 (Fig 1C). The mechanism by which these Prevotella spp. suppress aggregation does not depend on butyrate as both strains were deficient in its production [55,56]. Interestingly, Prevotella was the only bacterial genus whose abundance significantly decreased in patients suffering from constipation, a condition that is a hallmark in Parkinson's disease and is associated with up to 80% of cases [57][58][59]. Further evidence that these two species could be beneficial to host proteostasis comes from a recent study that demonstrated a negative correlation between the abundance of Prevotella in the gut of Parkinson's disease patients and the severity of symptoms [60]. A negative correlation between the abundance of Prevotella and pathogenesis of Alzheimer's disease mouse model has also been observed [61]. Collectively, these results suggest that the members of the Prevotella genus could play a more direct role in enhancing host proteostasis, and that the mechanisms by which bacteria affect host proteostasis in our C. elegans model may be conserved in higher organisms.
Although P. aeruginosa was previously found to induce neurodegeneration in C. elegans [62], the results of decreased polyQ-dependent motility of animals colonized with P. aeruginosa and K. pneumoniae were unexpected and suggest that either bacteria, or host responses to these bacteria, affect the function of distal tissues (i.e., muscle or neurons), but only in animals that express metastable aggregation-prone proteins such as polyQs (Figs 2B, 2C, and 3A-3G). Recent studies show that colonization of the C. elegans intestine with pathogenic bacteria affects bacterial avoidance behavior [63] and consequently increases lifespan [64]. While it is possible that feeding behavior might have been affected in our experiments, it is unlikely that avoidance would contribute to an increase in polyQ aggregation and the associated motility defects. In fact, if the animals in our experiments experienced food avoidance behavior, we would expect to see enhanced proteostasis, as dietary restriction was previously shown to suppress polyQ aggregation [65].
It is known that bacterial colonization of the C. elegans intestine increases the production of reactive oxygen species (ROS), which in turn enhances aggregation in that tissue [23]. Oxidative stress is known to affect overall animal physiology, including motility [66]. Since our results show the bacteria-mediated motility defects only in the presence of aggregation-prone polyQ44, but not in non-aggregating polyQ33 or non-polyQ controls (Fig 2C), it is possible that ROS generated by the worm in response to bacterial colonization enhanced polyQ aggregation, which further affected the function of other tissues involved in motility. Such a mechanism likely contributes to the observed bacteria-mediated disruption of host proteostasis which is supported by our results demonstrating that butyrate rescued polyQ aggregation in a DAF-16 and SKN-1-dependent manner (Fig 7A). However, ROS is likely not the sole contributor since our results also indicate that bacterial aggregates affect host proteostasis and consequently increase aggregation of host polyQ (Fig 1E-1G).
Tissue non-autonomous regulation has been noted in C. elegans under various conditions [27,67]. Our results demonstrating that bacterial colonization of the C. elegans intestine affects proteostasis across tissues are intriguing, as they suggest that bacteria impose tissue nonautonomous effects on the host. In human and mouse studies, the effect of bacterial gut colonization on host proteostasis has been primarily attributed to non-specific mechanisms, such as systemic inflammation, which broadly affects many tissues [68]; however, C. elegans lacks an inflammatory response, indicating that other and possibly more direct mechanisms may be involved. Bacterial amyloids have also been shown to specifically contribute to pathogenesis by affecting metastable aggregation-prone proteins [69]. The introduction of metastable proteins is known to affect the stability and folding of other proteins within the proteome [70,71]. As such, it is likely that bacteria may be secreting factors that translocate from the intestine across other tissues and affect proteostasis. A cross-seeding between disease-associated proteins (i.e., Aβ) with bacterial amyloids, including those from E. coli and P. aeruginosa, was previously observed [72,73]. Also, an enhancement of alpha-synuclein aggregation mediated by bacterial curli was shown in mouse and C. elegans Parkinson's disease models [74]. While our results indicate that E. coli curli fimbriae were not responsible for the enhancement of polyQ aggregation, E. coli aggregates did play a contributing factor (Fig 1G). It is possible that alpha-synuclein and polyQ are affected differently by bacterial amyloids. In addition to regulating the csg operon, CsgD also functions as a global transcriptional regulator and its target gene(s), other than csgA, could affect host proteostasis [75]. It is intriguing to speculate that the transmission of aggregation-prone proteins between tissues [67], may underlie how bacterial colonization of parental intestines affects protein aggregation in the progeny (Fig 3I).
A symbiosis between commensal and pathogenic bacteria can be disrupted by many factors, including infections, diet, and antibiotics [76]. While dysbiosis of the human gut microbiota has been associated with the pathogenesis of PCDs, the mechanisms that contribute to the disruption of host proteostasis and disease progression remain poorly understood. Our results demonstrate that butyrate, a common metabolite produced by commensal microbiota, can suppress aggregation and the associated toxicity when supplied exogenously (Figs 4 and 6) or produced by intestinal bacteria (Figs 8 and 9). We also showed that butyrate did not simply inhibit intestinal colonization (S9 Fig). In agreement with our results, the protective effects of butyrate have been previously demonstrated in a mouse model and in humans [21,32]. Our data indicate that the benefits of butyrate depend on SKN-1/Nrf2 and DAF-16/FOXO transcription factors, suggesting that butyrate may inhibit bacteria-induced aggregation by activating protective stress responses, in particular, oxidative stress responses. Indeed, bacteria are known to trigger oxidative stress, which can contribute to polyQ aggregation [23]. It is possible that various bacteria introduce different levels of oxidative stress and the strongest contributors affect host proteostasis, consequently leading to misfolding and aggregation of metastable proteins present within the proteome. In fact, oxidative stress is one of the major contributors to PCDs [77]. Recently, it was found that butyrate and β-hydroxybutyrate provide Nrf2-and FOXO-dependent protection against oxidative stress, respectively [78,79]. In addition, both SKN-1 and DAF-16 were shown to be involved in β-hydroxybutyrate-mediated lifespan extension [80]. These results, along with our data, collectively suggest that butyrate activates SKN-1/ Nrf2 and DAF-16/FOXO, which attenuate bacteria-induced oxidative stress. The effect of butyrate on polyQ aggregation also seems to depend on the type of pathogenic bacteria present, suggesting that butyrate may contribute to host proteostasis by acting on bacteria in addition to the host. These results are consistent throughout the intestinal and muscle-specific polyQ models. Additionally, the co-colonization experiment further supports that the extent of the beneficial effect of butyrate depends on the bacteria present (Fig 10A). The fact that bacteria are viable in the presence of butyrate at concentrations that suppress aggregation (S6 and S7 Figs) and that dead bacteria exhibit a distinct effect on aggregation (S8 Fig) suggest that microbial processes which affect the host may be targeted by butyrate. Whether butyrate affects specific bacterial signals that contribute to PCDs remains to be determined and further studies are underway to identify bacterial factors that enhance aggregation of host metastable proteins. Whatever the mechanism may be, our results suggest a therapeutic potential of butyrate in the prevention or treatment of PCDs.
High metabolic rates of bacteria can lead to depletion of available oxygen and introduce a hypoxic environment on the plates. Low oxygen levels are known to disrupt cellular processes leading to increased polyQ aggregation [81]. One of the possible explanations regarding how butyrate may inhibit bacteria-mediated polyQ aggregation is by suppressing bacterial metabolism, which restores the normoxic environment. However, if this were true, we would expect to have seen the restoration of a normoxic environment when we fed worms heat-killed bacteria, which would result in polyQ aggregation similar to 100 mM butyrate (S8 Fig). Therefore, it is likely that butyrate could have some additional effects on bacteria in addition to the host. In fact, it is known that butyrate can affect bacterial pathogenicity [34,82,83]. Further research is underway to determine what changes butyrate might exert on bacteria and how that influences the host.

Bacterial strains
Detailed information about bacterial strains used and generated in this study is provided in Table 1.

Caenorhabditis elegans maintenance and strains
Nematodes were maintained according to previously established protocols [89]. All experiments were performed at room temperature (RT,~23 o C) unless described otherwise. All C. elegans strains used in this study can be found in Table 1.

Bacterial culture conditions
Culture conditions for bacteria used in endogenous and exogenous butyrate treatment can be found in "Endogenous Butyrate Production" and "Exogenous Butyrate Treatment", respectively. DC228 was inoculated in the presence of 2 μg/mL gentamicin. Strains were plated onto Nematode Growth Media (NGM) +/-3% arabinose and were left to dry overnight at RT prior to having worms plated on them. Curli-mutant bacteria and wild-type control (MG1655) were plated on NGM and kept at RT for 2 days prior to having worms plated on them. K. oxytoca (HM-624), Prevotella disiens (HM-1171), Prevotella corporis (HM-1294), and their E. coli OP50 control were grown in Reinforced Clostridial Broth (RCB). Unless otherwise indicated, all other bacteria were grown overnight in a 37˚C incubator shaking at 220 revolutions per minute (RPM) in Lennox Luria Broth (LB), were seeded on NGM, left to dry overnight, and were kept at 4˚C for another day prior to having worms plated on them.

Transduction and verification of curli (-) mutants
The bacterial strains used in this study are listed in Table 1. All solutions were prepared using ultrapure water, 18.2 MO/cm (Barnstead). Bacterial strains were grown in LB (5 g yeast extract, 10 g Bacto tryptone, and 10 g NaCl per liter of water), pH 7.4, at 37˚C, with shaking (250 RPM), unless otherwise indicated. Transduction with P1vir was done to introduce gene deletions from E. coli donor strains from the Keio library [85]. Gene deletions were confirmed using PCR amplification of the target gene sequence using the primers listed in Table 1. The PCR products were visualized on an ethidium bromide-stained agarose gel.

Congo red assay
Curli production of each strain was determined by Congo Red plate assay. Congo Red plates were made using LB medium without salt (5 g yeast extract and 10 g Bacto tryptone per liter of ultrapure water), to which 50 μg/mL filter sterilized Congo Red and 1 μg/mL filter sterilized Brilliant Blue G250 were added after autoclaving. Overnight cultures were pelleted and washed once in LB without salt. Cells were then resuspended in LB without salt and cell density (measured by OD 600 ) was normalized across all samples. Twenty-five microliters of each strain was spotted in a separate quadrant of Congo Red agar plate, which was then incubated at 26˚C for 48 h. Plates were imaged after 48 h incubation at 4˚C.

ProteoStat assay
All bacteria were grown in LB in a 37˚C incubator shaking at 220 RPM overnight. Bacteria were diluted to OD 600 of 0.5 and seeded in the center of a 96-well plate. Outer wells were filled with water and the plate was sealed with parafilm. For consistency, each bacterial strain was tested in a separate plate. Plates were incubated for~44 h at 26˚C. Following the incubation, LB was removed, and wells were washed twice with ddH 2 0. Adhered biofilm was resuspended by pipetting with 100 μL 1x ProteoStat Assay Buffer. Two microliters of 1x ProteoStat Detection Reagent was dispensed into wells in a black, clear-bottom 96-well plate and 98 μL of resuspended biofilm was added and mixed. The plate was incubated for 15 minutes (min) at RT away from light. Following the incubation, fluorescence was read at excitation wavelength 550 nm and emission wavelength 600 nm, gain 100, in a Tecan Infinite 200 Pro plate reader. Assays were performed in triplicates.

Plasmid construction
pXDC18.mCherry was digested with EcoRI and AseI to remove lacI q and Ptac. A cassette harboring the araC gene and ParaBAD promoter from plasmid pKD46 was amplified with primers MJG368 and MJG369, digested with EcoRI & AseI and cloned into the pXDC18.mCherry backbone to yield plasmid pMJG125. The sfGFP cassette from plasmid pBSK-sfGFP was amplified with oligonucleotide primers MJG1004 and MJG1005. The resulting PCR product

PLOS PATHOGENS
was digested with NdeI and XbaI and subsequently cloned into pMJG125 to yield pMJG125. sfGFP.

Generation of bacterial strains
The arabinose-inducible E. coli reporter strain (DC228) was generated by the electroporation of arabinose-inducible fluorescent reporter encoded on plasmid pMJG125.sfGFP into E. coli LW393 using the BioRad Gene Pulser Xcell electroporator according to the manufacturer's instructions.

Exogenous butyrate treatment
All bacteria were grown aerobically in LB and seeded on NGM plates supplemented with indicated concentrations of sodium butyrate or controls lacking the compound. To ensure uniform coverage of the bacterial lawn, cultures were spread evenly across the plate. Plates were left to dry overnight at RT and were placed at 4˚C one day prior to plating worms. Control experiments compared the aggregation profiles of worms fed heat-killed bacteria versus bacteria cultured in the presence of 100 mM sodium butyrate. Bacterial cultures were collected in 15 mL conical tubes and were killed by heating for 60 min at 70˚C water bath. Bacterial viability was assessed by inoculating 10% of the culture in LB followed by overnight growth at 37˚C shaking at 220 RPM. No growth confirmed bacteria to be dead. To bypass developmental delay that could result from feeding worms dead bacteria, animals were age synchronized, and cultured on plain NGM seeded with E. coli OP50 at 20˚C for 48 h. After 48 h, animals were washed 6x with 5 mL M9 and transferred to NGM plates containing either heat-killed bacteria or 100 mM sodium butyrate with live bacteria. Worms were incubated at RT for three additional days and aggregates were quantified as described in "Aggregate Quantification".

Endogenous butyrate production
Overnight cultures of E. coli Bt and control strains, E. coli WT and E. coli OP50, were grown anaerobically. E. coli OP50 was grown anaerobically only in experiments where it was used as a control for E. coli Bt . E. coli Δ was grown aerobically, because this strain is not capable of anaerobic growth. To ensure uniform coverage of the bacterial lawn, bacterial cultures were spread evenly across the plate. The NGM plates +/-3% L-arabinose were seeded with the bacteria and dried overnight. E. coli Bt and control strains, E. coli WT and E. coli OP50, were placed in airtight containers with anaerobic gas packs, and all plates were stored at 4˚C for one to two days prior to the experiment. Synchronized animals were plated and fed on the bacteria for 72 h (muscle polyQ and control) or 92 h (intestinal polyQ, neuronal polyQ and controls) prior to assessing aggregation or motility. Aggregates were quantified as described in "Aggregate Quantification". The same method was followed when using 3% glucose rather than 3% L-arabinose.

Aggregate quantification
Worms were cultured and maintained as described in "Caenorhabditis elegans Maintenance and Strains". Following age synchronization, animals were plated on NGM plates seeded with specific bacteria. Unless otherwise indicated, after 92 h (intestinal polyQ44) or 72 h (muscle polyQ35) of feeding, animals were picked from the plates and placed in 96-well plates containing M9/levamisole, and frozen at -20˚C. Aggregates were quantified between 18-48 h postfreezing. PolyQ aggregates form rapidly, and their quantification is labor-intensive; therefore, the freezing step is essential to halt the formation of new aggregates during the quantification process. Worms expressing intestinal polyQ (Fig 1A and 1B) were imaged using a Nikon Eclipse 80i epifluorescence microscope equipped with a 4X Plan Apochromat objective (0.2 NA), metal halide lamp, and Ex450-490nm/Em500-550nm excitation filter. For all other aggregate quantification assays, worms were viewed using a Zeiss Axiovert S100 equipped with an Achrostigmat 10X Ph1 phase-contrast infinity microscope objective (0.25 NA), Chroma EGFP/FITC long-pass filter set (19002), and a mercury lamp.

RNAi knockdown
RNAi-mediated knockdown was carried out according to a previously published protocol [90]. Briefly, E. coli carrying DAF-16, SKN-1, and HSF-1 RNAi constructs and empty vector (L4440) were cultured overnight in a 37˚C incubator shaking at 220 RPM in LB medium supplemented with 5 μg/mL tetracycline and 50 μg/mL ampicillin. Overnight cultures were induced for 4 h with 1 mM Isopropyl β-d-1-thiogalactopyranoside (IPTG) prior to seeding NGM plates supplemented with 12.5 μg/mL tetracycline, 100 μg/mL ampicillin, and 0.4 mM IPTG. Where specified, plates were also supplemented with 10, 25, 50, and 100 mM sodium butyrate. The seeded plates were allowed to dry overnight and were kept at 4˚C for a maximum of two days.

Colonizing the C. elegans intestine with polyculture
All strains were grown as described in "Bacterial Culture Conditions", the cultures were normalized to OD 600 of 1.0, and were plated in equal parts on NGM plates +/-3% L-arabinose. Worms were cultured as described in "Caenorhabditis elegans Maintenance and Strains", and aggregates were quantified as described in "Aggregate Quantification".

Motility assays
All assays were performed at RT on age-synchronized animals. Unless otherwise indicated, worms were 72 h old at the time of body bend assays and 92 h old at the time of time-off-pick (TOP) assays. A body bend was defined as a change in the direction of the midline constituting approximately a 90˚bend and a significant decrease in nose-to-tail distance. The rate was determined by placing animals in a drop of M9, allowing them to recover for 1 min, and counting the number of body bends in 30 s. TOP motility was defined by the time it took a worm to crawl completely off an eyebrow hair that was placed under its mid-section (Fig 2A).
Worms were visualized using Zeiss Stemi 305 stereo microscopes.

PolyQ aggregation in the progeny
Worms were cultured and maintained as described in "Caenorhabditis elegans Maintenance and Strains". Following synchronization, parental generation worms (P) were plated on NGM plates seeded with bacteria. After 72-78 h, animals (P) were bleached and the F1 progeny were age-synchronized, followed by plating on NGM plates containing control E. coli OP50 lawn. The F2 progeny were isolated by repeating the same method. F1 and F2 progeny were grown for 92 h at RT prior to aggregate quantification.

Confirmation of bacterial killing after embryo isolation and agesynchronization of progeny
Worms were cultured and maintained as described in "PolyQ Aggregation in the Progeny". Following both the bleaching of the parental generations (P), and overnight synchronization of F1 L1s, 100 μL of M9 containing embryos and L1s, respectively, were spread across LB plates, put at 37˚C overnight, and examined the following day for the presence or absence of bacterial growth.

Bacterial colonization assays
NGM plates were supplemented with indicated concentrations of sodium butyrate or 3% Larabinose and seeded with bacteria. E. coli OP50 and P. aeruginosa PAO1 were grown under aerobic conditions and E. coli Bt was grown anaerobically. The plates were then dried. Worms expressing intestinal polyQ44 were age-synchronized and plated on NGM seeded with the above bacteria, followed by a 3-and 4-day incubation. To remove any extracellular bacteria, 12 or more worms from the plate of interest were transferred onto a plain NGM plate for approximately 30 min. Ten worms were then transferred from each plain NGM plate into 1 mL of M9 supplemented with 80 μM levamisole and the tube was agitated to move worms around in solution. Once worms settled to the bottom, 950 μL of the supernatant was gently removed by pipetting. To remove any remaining external bacteria, worms were washed twice, each wash lasting 10 min, with M9 containing gentamicin (100 μg/mL) and levamisole (80 μM). Worms were then washed three times with M9/levamisole and the last wash was saved. To account for any external bacteria, the last wash was plated onto LB plates. Two hundred microliters of 1% triton X-100 was added to the worms sitting in 50 μL M9/levamisole. A handheld homogenizer (Bel-Art) was used to grind worms until no large remains were visible. The samples were then serially diluted with M9 and spotted onto LB plates. Plates were incubated at 37˚C overnight and colonies were quantified the next day.

Butyrate dose response in axenic culture
Bacterial strains were cultured in LB. Strains were grown overnight in a 37˚C incubator shaking at 220 RPM. Two-fold serial dilutions of sodium butyrate were prepared in LB in a 96-well plate. The concentrations ranged from 0 to 200 mM. Three microliters of culture adjusted to OD 600 of 1.0 were added to 1.5 mL of LB media. An equal volume of this culture was added to the wells of the 96-well plate, which diluted the final concentrations of sodium butyrate by half (0 to 100 mM). The plate was incubated at 37˚C in a Tecan Infinite 200 Pro plate reader for 24 h. Every 30 min, the plate was shaken for 5 s and OD 600 readings were taken.

Bacterial viability on NGM supplemented with butyrate
All bacteria were grown aerobically in LB overnight in a 37˚C incubator shaking at 220 RPM. The next day, cultures were adjusted to OD 600 of 0.5 and 10 μL of a 10 −1 dilution were spotted onto NGM plates supplemented with 0, 10, 25, 50, or 100 mM sodium butyrate. The plates were then placed in a 37˚C incubator for 48 h. After incubation, the spots of bacteria were scraped with an inoculating loop, suspended in 1 mL LB, and 10 μL were spotted onto LB agar plates and placed in a 37˚C incubator overnight. Images were taken the next day and viability was assessed by the presence or absence of bacterial lawns or distinguishable colonies.

Live imaging
Nematodes expressing intestinal polyQ33 or polyQ44 fed E. coli OP50 and P. aeruginosa PAO1 for four days were mounted in a drop of 1 mM levamisole on a 3% agarose pad and covered with a coverslip. GFP fluorescence and Nomarski images were taken using a Zeiss Axio Observer 7 microscope equipped with an Axiocam 503 mono camera, Solid-State Light Source Colibri 7, and a 5x Plan-Neofluar objective (0.16 NA). Pharyngeal mutant worms (DA597) were imaged using a 20x Plan-Neofluar objective (0.5 NA) and were processed using ZEN Tiles & Positions Module in Zeiss ZenPro software.

Arabinose availability assay
Bacteria expressing the arabinose reporter (E. coli DC228) were inoculated with 2 μg/mL gentamicin and were cultured as described in "Bacterial Culture Conditions". Bacteria were seeded onto NGM plates +/-3% L-arabinose and kept at RT for two days before worms were plated on them. Pharyngeal mutant animals (DA597) were age-synchronized and cultured for four days on E. coli bacteria (DC228) carrying an L-arabinose-inducible reporter (Ara-sfGFP) on NGM +/-3% L-arabinose (positive and negative control, respectively). Animals used to test the availability of arabinose to the intestine were cultured for two days on E. coli bacteria carrying L-arabinose inducible reporter in the absence of L-arabinose followed by their transfer onto 3% L-arabinose plates containing E. coli OP50. Image acquisition is described in "Live Imaging".