A trade-off switch of two immunological memories in Caenorhabditis elegans reinfected by bacterial pathogens

Recent studies have suggested that innate immune responses exhibit characteristics associated with memory linked to modu-lations in both vertebrates and invertebrates. However, the diverse evolutionary paths taken, particularly within the invertebrate taxa, should lead to similarly diverse innate immunity memory processes. Our understanding of innate immune memory in invertebrates primarily comes from studies of the fruit fly Drosophila melanogaster , the generality of which is unclear. study profiles the immune memories during C. reciprocally, suggesting that a subtle regulation should occur within the two immunological memories.

Immune system of vertebrates produces a faster and increasingly more efficient immune response against the repeated infections because of immunological memory. The immune memory in vertebrates was thought to depend mainly on the memory T or B lymphocytes, i.e. the adaptive immune system. However, there is an increasing body of evidence supporting the hypothesis that a memory of past immune insults similarly exists in the prototypical cells of innate immunity of vertebrates, namely, the NK cells, monocytes, macrophages, and microglia (1)(2)(3)(4). The memory of innate immunity lasts for a few days to several months. It has also been described that in addi-tion to enhancing the protection against reinfections, innate immunity memory leads to immunity tolerance by dampening the inflammatory responses accordingly (3,5). Therefore, a more intriguing example about the modulation of innate immunity memory comes from the recent report on the microglia memory in Alzheimer's disease (1). It has been shown that two injections of lipopolysaccharide induced an enhanced innate immunity of microglia, characterized by elevated levels of proinflammatory molecules; in contrast, four lipopolysaccharide exposures caused immune tolerance, which affected disease progress (1).
Because invertebrates lack an adaptive immune response, they rely solely on innate immunity, which makes them ideal experimental models to investigate the hypothesis of immunological memory within the innate immune responses. Evidence for an innate immune memory in invertebrates was initially recognized in sponges that, compared with the first exposure, displayed a much stronger graft rejection when re-exposed to a graft of the same donor (6). Since the late 1970s, more invertebrates, including Porifera, Cnidaria, Annelida, and Echinodermata, have revealed a similar allograft rejection after tissue transplantation (7). Also, the repeated challenge of infective agents or noninfectious microorganisms (e.g. heat-killed or even probiotic bacteria) can prime the innate immunity in invertebrates. The memory of innate immunity in invertebrates is generally nonspecific or noncompletely specific as it triggers protective functions against a broad spectrum of microorganisms, which seems to be distinct from vertebrates. However, the specificity of innate immunity memory has been identified in a few Arthropods. For example, it was found that during the immune response of cockroaches against the inactivated bacteria, an acute nonspecific phase of immunological memory could be replaced by a more specific immunological protection with passage of immunization time at the fourth injection (8). Furthermore, Drosophila infected with Streptococcus pneumonia or Beauveria bassiana is better at antagonizing these pathogens than other microbes upon reinfection (9). The mosquito Anopheles gambiae also displays an increased resistance for Plasmodia after a second encounter of Plasmodium falciparum (10). Collectively, because of the many evolutionary paths, the memory of innate immunity appears to be diverse within the invertebrate taxa (2,11).
Caenorhabditis elegans lives in the soil, a complex environment where a variety of stressful factors may exist, including pathogenic microbes. C. elegans uses two major defense strategies when infected with pathogenic microbes: innate immune and aversive behavior. Innate immunity in C. elegans can eliminate the pathogenic microbes that invade the animal's body, and it has been suggested to involve transforming growth factor b-like, insulin-like (DAF-2/DAF-16), and p38 mitogen-activated protein kinase (MAPK) classical signaling pathways (12). Those inducible immunity mechanisms are efficacious but expensive in terms of energy and self-damage (13). Avoidance is the other strategy, which is mainly divided into instinctive and learned avoidance. As a more rapid response on exposure of pathogens, it is generally committed by neurons and neuropeptides and seems more economical and less harmful (14). Because learned avoidance has been shown to share molecules with the immune system, including TOL-1 (15,16), it is regarded as a form of immune memory in the broad sense (11,17). Although innate immunity and avoidance have been intensively investigated in the context of antagonizing pathogens, little knowledge has been obtained about their characteristics of immunological memory and any of its possible interactions between them. Here, we profiled avoidance and innate immunity as immunological memories elicited by bacterial pathogens upon reinfection and revealed a specificity of immunity memories in the invertebrate of C. elegans. Furthermore, we reveal a trade-off between these two immunity memories and detail the underlying molecular mechanism.

Results
An enhanced immunological memory of avoidance specific to the bacterial pathogens in C. elegans Although it has been described that the aversive behavior of C. elegans can be promoted after reinfection by several bacterial pathogens, such as Pseudomonas aeruginosa PA14, Serratia marcescens, or Staphylococcus aureus, whether the immune memory is specific to the pre-exposed pathogen or responds to a broad spectrum of microbes remains to be elucidated.
For that purpose, the avoidance index against different bacterial pathogens was determined in either naïve or trained worms (Fig. 1A). The assay of avoidance index demonstrated that, compared with the naïve animals, the trained worms indeed displayed increased avoidance upon the second encounter of the same pathogenic bacterium P. aeruginosa PA14 (Fig. 1B, left panel), which was similar to a result previously reported (15). In contrast, the worms that had been trained by pre-exposure to P. aeruginosa PA14 but then were re-exposed to either a Gram-positive pathogen, namely, S. aureus (Fig. 1B, middle panel), or another Gram-negative pathogen, namely, Salmonella typhimurium, did not show avoidance behavior (Fig. 1B, right panel); the avoidance index was fairly low in either naïve or trained worms, and no enhancement of avoidance was observed in each conditioning. Thus, our data suggested that the invertebrate C. elegans should possess a specific memory that promoted avoidance upon re-exposure to the same pathogen.
To further address whether the immunological memory of enhanced aversion existed when C. elegans were re-exposed to other bacterial pathogens, we assayed the Gram-positive pathogen S. aureus, which had also been described to induce the avoidance for C. elegans (18). Comparing pre-and re-exposure to the same bacterial strain S. aureus ATCC 25923, it was also found that the avoidance index was significantly increased in the trained worms (Fig. 1C, left panel). However, when the worms pre-exposed to S. aureus were re-exposed to either P. aeruginosa PA14 or S. typhimurium, the avoidance index remained comparable with that of the naïve ones (Fig. 1C, middle and right panels).
A trade-off of two immunological memories with promoted avoidance and suppressed innate immunity We next determined the innate immunity memory in the trained worms by assaying survival rates on a full bacterial lawn ( Fig. 2A). Contrary to the enhancement of avoidance index, the worms seemed more susceptible to the pathogenic bacterium upon pre-and reinfection with P. aeruginosa PA14, and therefore, the survival rates of the trained worms decreased significantly compared with the naïve ones (Fig. 2B, left panel). However, C. elegans that had been trained by P. aeruginosa PA14 but were reinfected by S. aureus ATCC 25923 retained the unchangeable survival rates (Fig. 2B, right panel). To further confirm the result above, a reverse assay was performed again in which the worms preconditioned with S. aureus were reinfected by S. aureus ATCC 25923 or P. aeruginosa PA14. The decreased innate immunity memory was similarly obtained upon pre-and reinfection of S. aureus (Fig. 2C, left panel), but no obvious change happened upon reinfection of P. aeruginosa PA14 (Fig. 2C, right panel).
To further answer the question of whether this negative regulation in innate immunity was triggered by the enhancement of avoidance, we determined the survival rates of S. typhimurium infection because no avoidance was observed by this pathogenic bacterium after the same treatment of either pre-or reexposure. Our data from S. typhimurium illustrated that, compared with the naïve worms, the survival rates indeed remained unchanged in the trained ones (Fig. 2D).
The pathogen's accumulation in the worm intestine has been reported to influence the susceptibility of infection, with the pumping rates (a type of pharynx movement when C. elegans ingests food) directly affecting the pathogen's colonization in intestine (19,20). Therefore, the pumping rates with or without P. aeruginosa PA14 preinfection were determined by recording the pharynx contraction during a 30-s period. It was found that both the naïve and trained animals exhibited a similar pumping rate at 1 and 3 h (Fig. S1A), suggesting that the changed survival rates were unlikely to be due to accumulation of pathogens in the worm's intestine.
Collectively, our results revealed specificity in the memories of aversive behavior and innate immunity when C. elegans faced with reinfection of the same bacterial pathogens and a trade-off with enhanced avoidance and reduced innate immunity reciprocally, suggesting that a subtle regulation should occur within the two immunological memories.

Dysfunction of either AWB or ADF affects the trade-off switch of immunological memories
It has been widely reported that the chemosensory neurons AWB and ADF play important roles in either the behavioral response to odors or damages caused by bacterial pathogens (21,22). Because the lim-4 is the gene required for the development and function of those two neurons (23,24), we assayed the avoidance and survival in lim-4 mutant animals. Compared with the WT N2, the mutants of lim-4(ky403) and lim-4(yz12) displayed a decreased avoidance index in either naive or trained animals (Fig. 3A). During re-exposure, the promoted lawn-leaving behavior, which was typical of N2 worms, was particularly and significantly inhibited in both lim-4 mutants (Fig. 3A). The innate immunity memory against P. aeruginosa PA14 also contrasted with that of N2, in which the survival rates after reinfection did not decrease in the lim-4(yz12) mutant or even showed a slight increase in the lim-4(ky403) mutant (Fig. 3B). To further identify the involvement of lim-4 in regulating the trade-off switch, we examined the phenotypes when rescued expressing lim-4 gene in lim-4(yz12) mutant. Compared with lim-4(yz12) mutants, lim-4(yz12);Ex[lim-4(1)] almost restored the avoidance enhancement and innate immunity suppression as N2 worms (Fig. 3, C and D). Similarly, to exclude the possibility of pathogen's accumulation in the lim-4 mutant worms, we detected their pumping rates again. Our experimental data indicated that the pumping rates in the lim-4 mutant were not different from N2, with or without pre-exposure (Fig. S1B).
To distinguish which pair of chemosensory neurons (AWB or ADF) mediated the trade-off of the two immunological memories, the phenotypes of worms without either neuron were further investigated. Although the avoidance index of str-1::mec-4(d) worms, which lack AWB neurons, was slightly decreased under naïve conditions (21, 25), the Profiling the immunological memory of avoidance upon re-exposure of bacterial pathogens. A, graphic representation of the avoidance assay for naïve and worms trained by exposure to bacterial pathogens. B, once pre-exposed to the bacterial pathogen P. aeruginosa PA14, the avoidance index of trained worms increased significantly upon the second encounter with the same bacterium (P. aeruginosa PA14), but not when the bacterial pathogen was S. aureus or S. typhimurium. C, similarly, once pre-exposed to the bacterial pathogen S. aureus, the avoidance index of trained worms increased significantly on the second encounter of S. aureus. These results are presented as means 6 S.D. of at least three independent experiments. Avoidance index were statistically analyzed using a t test. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001.
Two immunological memories of C. elegans enhanced avoidance was completed disappeared in the trained animals (Fig. 3E). Accordingly, the worms lacking AWB neurons exhibited unaltered survival rates after re-exposure (Fig. 3F). Meanwhile, we addressed the involvement of ADF neurons in this process. We observed that the avoidance index and survival rates in the N2;Ex [ADF::egl-1] mutants that lacked the functional ADF neurons were similar between naïve and trained animals, which meant that the worms without ADF function had also lost the two immunological memories (Fig. 3, E and F).

Coordination of dopamine and serotonin is responsible for the trade-off between the two immunological memories
Abundant experimental evidence suggests that C. elegans employs the neuronal signaling to regulate both the avoidance behavior and the innate immune response. Notably, serotonin Figure 2. Profiling the immunological memory of innate immunity upon reinfection of bacterial pathogens. A, graphic representation of the survival assay for naïve and trained worms infected by bacterial pathogens. B, the survival rates in the trained worms (preinfected by the bacterial pathogen P. aeruginosa PA1) decreased during the second encounter with the same bacterium P. aeruginosa PA14, but not when the pathogen was S. aureus. C, once preinfected by the bacterial pathogen S. aureus, the survival rates of trained worms also decreased upon the second encounter with S. aureus. D, reinfection of S. typhimurium that cannot enhance the immunological memory of avoidance has no negative impact on the innate immune memory. These results are presented as means 6 S.D. of at least three independent experiments. The statistical analysis of survival curve was performed by log-rank (Mantel-Cox) test. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001. and dopamine have been reported to be involved in the integration of behavior and immunity (26,27). Therefore, we tried to identify whether the signaling of serotonin or dopamine participated in modulating the two immunological memories during reinfection by bacterial pathogens.
We first assessed the effects of the serotonergic circuit by using tph-1 mutant strains that lacked tryptophan hydroxylase for serotonin biosynthesis and mod-1 mutant with a nonfunctional serotonin-gated chloride channel. After re-exposure to the bacterial pathogen P. aeruginosa PA14, although the aversive behavior in those mutants, including tph-1(mg280), tph-1 (n4622), and mod-1(ok103), did not increase as did that of the WT N2 (Fig. 4A), their survival rates were still reduced in the reinfected worms, similar to the innate immunity memory of N2 (Fig. 4B). With respect to the synthesis of dopamine, in the worms of cat-2(e1112) and cat-2(tm2261), whose gene mutation caused lower efficiency in the rate-limiting enzyme tyrosine hydroxylase, we found phenotypes similar to the animals with serotonin deficiency: they could not increase their lawnleaving behavior but retained the reduced innate immunity, as did the WT N2 (Fig. 4, C and D). To further validate the above results of dopamine, we conducted the assays of avoidance index and survival rates using a treatment of 6-hydroxydopamine (6-OHDA), which functions specifically to degenerate the dopamine neurons (28). The addition of 6-OHDA showed the same phenotype as the cat-2 mutants (Fig. 4, C and D). Thus, our data suggested that the serotonergic or dopaminergic circuit cannot sustain the trade-off switch of two immunological memories independently.
Surprisingly, because either gene of cat-1 and cat-4 functions in the common path for biosynthesis or transport of serotonin and dopamine, the increased avoidance index and reduced survival rates disappeared synchronously in the trained worms with cat-1 or cat-4 mutation (Fig. 4, E and F), implying that both neurotransmitters, instead of either serotonin or dopamine, were necessary in modulating the two immunological memories. To further address their combined roles of serotonin and dopamine, we assayed serotonin-deficient mutants by adding 6-OHDA. Compared with the tph-1 or cat-2 mutant that showed survival rates comparable with that of N2, Figure 3. Either the AWB neuron or the ADF neuron is required for the trade-off switch of two immunological memories with enhanced avoidance and reduced innate immunity. A and B, upon reinfection, the avoidance index and survival rates of lim-4 mutants did not change reciprocally as the trained N2 did but remained similar to that of the naïve worms. C and D, avoidance index and survival rates of the WT N2, lim-4(yz12), and lim-4(yz12);Ex[lim-4(1)] for naïve and trained conditions. The complement expression of gene lim-4 partially rescued the trade-off switch of two immunological memories in the mutant lim-4(yz12). E and F, effect of either AWB or ADF on the immunological memories of avoidance and innate immune response. The dysfunction of each chemosensory neuron disturbed the trade-off switch of two immunological memories. These results are presented as means 6 S.D. of at least three independent experiments. The statistical differences were analyzed using two-way ANOVA. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001. deficiencies in those two neurotransmitters completely disrupted the trade-off between avoidance and immunity (Figs. 4,  G and H).
Based on the above hypothesis that the two neurotransmitters redundantly regulate the trade-off between the two immunological memories during reinfection of pathogenic bacteria, the contents of serotonin and dopamine in the naïve and trained worms were determined in vivo using ultra performance liquid chromatography-tandem MS (UPLC-MS/ MS). A dramatic 7-10-fold increase of serotonin and dopamine was indeed observed in the trained worms (Fig. 4, I and J).
Either AWB or ADF neurons control the production of dopamine and serotonin To validate the hypothesis that the trade-off switch of two immunological memories was regulated by the chemosensory neurons via the two neurotransmitters of dopamine and serotonin, we observed under a fluorescence microscope the tph-1:: gfp and cat-2::gfp in wildtype N2 and lim-4 mutants that have effects on the development and role of both AWB and ADF. In the lim-4(yz12) mutant, the fluorescence of TPH-1::GFP was significantly weakened in the ADF neurons (Fig. 5A), and no CAT-2::GFP was observed in any dopaminergic neurons, such as the dorsal CEP, ventral CEP, ADE, and PDE neurons (Fig.  5B). We also assayed the mRNA levels of the tph-1 and cat-2 genes. Our results indicated that the mRNA of tph-1 and cat-2 was significantly decreased in lim-4(yz12) (Fig. S2, A and B), consistent with the results from GFP data. Because the ADF neurons are known to be in charge of the production of both serotonin and dopamine, we further determined the function of AWB control. After assaying the fluorescent changes of tph-1:: gfp and cat-2::gfp in the worm of str-1::mec-4(d) (AWB-killed animals), it was demonstrated that tph-1::gfp was significantly reduced in ADF neurons but partially retained in the NSM neurons (Fig. 5C). Moreover, the dopaminergic neurons PDE had no fluorescence of CAT-2::GFP (Fig. 5D), suggesting that AWB neurons control the production of both dopamine and serotonin.
Finally, we directly measured the contents of serotonin and dopamine using UPLC-MS/MS assay in a series of worm strains with dysfunction of AWB and/or ADF (Fig. 6, A-D).
The quantitative results showed that, compared with the WT N2, the concentrations of serotonin and dopamine were obviously reduced in lim-4(yz12), str-1::mec-4(d), and N2;Ex[ADF:: egl-1] (Fig. 6E). However, the tph-1(mg280) and tph-1(n4622) mutants had only decreased levels of serotonin but with increased levels of dopamine (Fig. S3, A and B), which also explained why the deficiency in gene tph-1 could not com-pletely disturb the trade-off between the two immunological memories.

Involvement of insulin-like signaling pathway in the trade-off switch of the two immunological memories
Tissue-specific activation of the genetic pathways, such as the DAF-2/DAF16 pathway, p38 MAPK pathway, and EGL-30 (Gaq)-UNC-73 (Trio RhoGEF)-RHO-1 (RhoA) pathway, control avoidance and innate immunity (29)(30)(31). After screening the signaling pathways above, the insulin-like pathway seemed to be involved in the trade-off switch of two immunological memories upon the second encounter of P. aeruginosa PA14: the mutant of daf-2(e1370) in the insulin-like signaling pathway obviously disrupted the enhanced avoidance and the reduced innate immunity after reinfection (Fig. 7, A and B), but the double mutation of daf-2 and daf-16 genes rescued the two types of immunological memories to a similar extent as N2 (Fig. 7, A  and B). However, few effects on the immunological memories were observed in the other two pathways.
Within insulin-like signaling pathway, activation of the transmembrane tyrosine kinase insulin receptor DAF-2 phosphorylates the transcriptional factor DAF-16 and retains it cytoplasmic accumulation, but the nuclear translocation of the latter is indispensable for the expression of antibacterial genes in C. elegans (12). Therefore, here, to further confirm the involvement of the DAF-2/DAF-16, we selected the downstream target genes of DAF-16 and employed qPCR to determine their expressions after reinfection with P. aeruginosa PA14. Among the 10 known target genes related to antimicrobial peptides, lysozyme genes of reactive oxygen species and immune defense response, six of them, namely F08G5.6, lys-2, pqm-1, aqp-1, spp-1, and acdh-1, had significantly lower mRNA levels compared with the naïve N2, with the slight reduction in sod-3 having no statistical significance, suggesting that reinfection by the bacterial pathogen P. aeruginosa PA14 should suppress the expression levels of DAF-16 target genes (Fig. 7C). When we remeasured the expression of those seven target genes in the daf-2(e1370) mutant, most of the genes above increased their expression levels upon reinfection (Fig.  7D). Consistent with the result from P. aeruginosa PA14, the worms pre-and reinfected by the bacterium S. aureus also showed the same expressional pattern with reduced target genes of DAF-16 upon the second encounter of the same pathogen (Fig. S4). Despite the similar general trend, in fact, there is still a difference in individual genes between P. aeruginosa PA14 and S. aureus (particularly e.g. pqm-1 and sod-3), implying the personalized genes of innate immunity should be induced by the different pathogens ( Fig. 7C and Fig. S4).  A and B, the dysfunction of the serotonergic circuit decreased the avoidance memory but retained the unaltered immunity after reinfection. C and D, a similar phenotype with decreased avoidance but unaltered innate immunity was shown in the absence of dopamine signaling. E and F, mutations in the catecholamine biosynthetic genes, such as cat-1(e1111) and cat-4(e1114), destroyed the trade-off switch of the two immunological memories after reinfection by P. aeruginosa PA14. G and H, serotonin biosynthetic mutants synchronously treated with 6-OH-DA confirmed the coordination of serotonin and dopamine in regulating the two immunological memories. I and J, assay for serotonin content (I) and dopamine content (J) using UPLC-MS/MS in the naïve and trained N2 in vivo. These results are presented as means 6 S.D. of at least three independent experiments. The statistical differences were analyzed using two-way ANOVA, and the quantitative analysis of serotonin and dopamine concentrations was done by t test. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001.

Two immunological memories of C. elegans
Next, we tried to verify the relationship between the chemosensory neurons (AWB and ADF) and the DAF-2/DAF-16 signaling pathway. We quantified the above target genes of DAF-16 in the mutant worms of lim-4(yz12), str-1::mec-4(d), and N2; Ex[ADF::egl-1]. Our results showed that in all of those mutants, the mRNA levels of the tested genes were significantly higher than those in N2 when reinfected with P. aeruginosa PA14 (Fig.  S5). To further confirm whether the DAF-2/DAF-16 pathway acts downstream of the neurotransmitters, we crossed tph-1 (mg280) or cat-2(e1112) with daf-16(mu86) to obtain double mutants. It was found that disruption of the downstream gene daf-16 reverted the phenotype of tph-1(mg280) or cat-2(e1112), with an avoidance index and survival rates similar to those of the WT N2 (Fig. 7, E and F).

The serotonin and dopamine neurotransmitters act on DAF-2/DAF-16
Because the DAF-2/DAF-16 pathway functions downstream of serotonin and dopamine, we further determined whether these two neurotransmitters directly regulated DAF-2/DAF-16. Using the reporter gene daf-16::gfp, we observed the effects on the nuclear translocation of DAF-16 when the exogenous serotonin and dopamine were added. Under the stressful situation of heat shock, DAF-16::GFP could be quickly imported into the nucleus within 10 min. However, the addition of 2 mM serotonin caused most of DAF-16::GFP to remain in the cytoplasm under the same conditions. The addition of 2 mM dopamine seemed to have had a weaker effect on nuclear accumulation of DAF-16::GFP. Meanwhile, the translocation of DAF-16:: GFP into the nucleus was similarly inhibited when serotonin and dopamine were added together (Fig. 7G).

Discussion
Until now, our knowledge on innate immune memory in invertebrate mostly came from the studies in the fruit fly Drosophila melanogaster (32). However, the invertebrates are composed of a variety of species that have the quite different immune and neural systems, so the results obtained from the model D. melanogaster cannot be easily generalized across invertebrates.
Innate immunity memory has been sporadically described in C. elegans. For example, the learned avoidance, which is believed to be a form of memory in the broadest sense, occurs during the re-exposure of the pathogen P. aeruginosa PA14 (15). Conditioning by either the probiotic bacterium Lactobacillus acidophilus or enteropathogenic Escherichia coli is . E, quantitative analysis to serotonin and dopamine contents using one-way ANOVA by comparing each mutant strain to N2. These results are presented as means 6 S.D. of at least three independent experiments. The statistical differences were analyzed using a t test by comparing each mutant strain to N2. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001. beneficial for the worms to improve their survival rates (27,33). Additionally, transgenerational protection against viruses can be provided by the RNAi machinery (34). However, the investigations dealing with the characteristics of those two immunological memories and their interrelationship are still incipient. Our current study suggested that, during reinfection by the pathogenic bacteria, C. elegans will prime the immune memory by promoting avoidance that is specific to the pre-exposure of bacterial pathogens. At the same time, a reduced innate immunity in the trained worms was modulated reciprocally with the enhanced avoidance. This hypothesis was validated by the experiment in which the worms were conditioned with the pathogenic bacteria S. typhimurium for 4 h that could not induce the much stronger aversive behavior, because the pathogens had little effect on innate immunity or even managed to activate the genes of innate immunity (Fig. 2D). This situation is more similar to a previous report from the enteropathogenic E. coli serotype 0127:H6 (27). However, the changed levels of the innate immunity after reinfection could be affected by the types of pathogenic bacteria, the conditioning time, and even the status of avoidance.
Although the immune memory of vertebrates is known for decades to result from the epigenetic reprogramming of trained immune cells carrying specific antigen receptors that underwent a process of somatic diversification at the level of the DNA, the mechanism of the immune memory remains unclear for invertebrates. A few recent discoveries from invertebrates led to the following proposed candidate mechanisms: long-lasting up-regulation of Toll or some other immune molecules (35), somatic diversification of the Down syndrome cell adhesion molecule (Dscam) in arthropods (36,37), fibrinogenrelated proteins in mollusks (38), and the scavenger receptor cysteine-rich proteins in echinoderms (39). Here we addressed the neurons, neuropeptides, and the related signaling pathways that are responsible for modulating the trade-off switch of the two immunological memories. During such a shift, C. elegans nonredundantly employs the chemosensory neurons AWB and ADF, which secrete the serotonin and dopamine neurotransmitters, and further activates the downstream DAF-2/DAF-16 pathway (Fig. 8). Chemosensory neurons detect the volatile and water-soluble environmental clues associated with food, pathogenic microbes, and other stimuli and subsequently modulate a variety of behaviors in C. elegans, including the life span, development, immune response, and chemotaxis or avoidance behavior. It has been demonstrated that the pathogen-avoidance behavior requires the AWB neurons (22). ADF serotonergic neurons affect olfactory learning and the immune response (15,26). These results imply that the bilaterally symmetric pairs of ciliated chemosensory neurons AWB and ADF mediate the defense, including behaviors and immunity, against pathogens. Furthermore, the serotonin and dopamine neurotransmitters are required for longevity, egg-laying behavior, feeding behavior, and the immune response, as well as naïve aversion behavior and learned choice in C. elegans (27,40,41). In humans, serotonin enhances the innate immunity by producing NK cells and interferon-g (42,43). Dopamine also mediates NK cells, T lymphocytes, T regulatory lymphocytes, B lymphocytes, and inflammatory factors (44,45). There is also evidence to support that serotonin and dopamine interact in mammals, D. melanogaster, and C. elegans (46,47). Thus, in our current data, although individually serotonin and dopamine cannot regulate the avoidance behavior and innate immunity, their coordination should be indispensable for the two types of immunological memories. The LIM-homeobox gene lim-4 has been described as necessary for the synthesis of serotonin (22). In both vertebrates and invertebrates, LIM-homeobox genes mediate the GABAergic phenotype, cholinergic neurons, dopaminergic neurons, serotonergic neurons, and some motor neurons and interneurons (48,49). Our results confirmed these findings, because the lim-4 mutation affected not only serotonin but also dopamine in C. elegans. In addition, our data showed that both ADF neurons, known as serotonergic neurons, and AWB olfactory neurons have effects on the serotonergic and dopaminergic phenotype. The daf-2 mutation led to a disturbance of the trade-off between the two immunological memories upon reinfection of P. aeruginosa PA14, but the daf-2;daf-16 double mutant rescued the phenotype to a similar extent as N2. C, the transcriptional levels of DAF-16 target genes decreased when the WT N2 was reinfected by P. aeruginosa PA14. D, the expressional levels of DAF-16 target genes in daf-2(e1370) worms were comparable or even higher upon reinfection than the naïve ones. E and F, the avoidance index and survival rates in the tph-1(mg280);daf-16(mu86) and cat-2(e1112);daf-16(mu86) double mutants. G, effects of the exogenous neurotransmitters serotonin or/and dopamine on DAF-16::GFP nuclear accumulation. These results are presented as means 6 S.D. of at least three independent experiments. The statistical differences were analyzed using one-way or two-way ANOVA, and the relative mRNA levels were analyzed by t test. n.s., P 0.05; *, P , 0.05; **, P , 0.01; ***, P , 0.001. Figure 8. A proposed model for the trade-off switch of two immunological memories with enhanced avoidance and reduced innate immunity in C. elegans upon reinfection by bacterial pathogens. In naïve worms without prior exposure to pathogens, the avoidance behavior is stimulated by chemosensory neurons, and the immune response is regulated by insulinlike signaling, p38 MAPK, or the noncanonical UPR pathways (19,20,25,54). After reinfection of bacterial pathogens, a trade-off switch of two immunological memories with enhanced avoidance and reduced innate immunity takes place that is mediated by the chemosensory neurons AWB and ADF, involving the coordination of serotonin and dopamine, and the downstream DAF-2/DAF16 signaling pathway.
Two immunological memories of C. elegans As one of the most important signaling pathways involved in the glucose and lipid metabolism of C. elegans, the insulin-like pathway has been reported to negatively regulate the pathogen avoidance behavior (31,50). Meanwhile, activation or suppression of DAF-16 that occurs during C. elegans antagonizes bacterial infection (12,51). Therefore, our current data undoubtedly provide additional experimental evidence to support a hypothesis that the innate immune memory from lower animals to mammals should be controlled by cellular metabolism (52).
In natural environments, it is common that organisms may develop a series of independent and/or redundant strategies when confronted with stresses, which increases their chances of survival. For example, under harmful stimuli, C. elegans either commits suicide of the damaged cells (apoptosis) or induces an animal-wide stressful resistance to antagonize stress. Once the apoptosis machinery is disturbed, the worms can switch to a more global animal-wide system of stress resistance (53). As the two major defense strategies for C. elegans to antagonize the infection of pathogenic microbes, although many investigations have focused on innate immune or avoidance behavior, respectively, their interrelationship has remained unclear. Our study suggests that there is an extensive interaction between the two responses. Our data imply that, of the two defensive responses, C. elegans may preferentially employ the aversive behavior, which is thought to be more economical than innate immunity (14). Depending on the activated avoidance behavior, C. elegans escapes from the invasive dangers more effectively and hence increases its survival upon the second encounter of the same pathogens.
Collectively, our investigation, for the first time, profiles two types of immune memories in the invertebrate model C. elegans and further reveals the chemosensory neurons, neurotransmitter(s), and their associated molecular signaling pathways are involved in a trade-off between the two immunological memories against the reinfection of bacterial pathogens.

Avoidance and survival assays
In our reinfection training, the synchronized animals were preinfected on a full lawn of the pathogen P. aeruginosa PA14/ NGM for 4 h at 20°C, followed by 1 h of recovery on E. coli OP50. The survival rate was also performed in PA14/NGM. The bacterium P. aeruginosa PA14 was cultured in LB medium overnight, and the culture was then spread on NGM plates. The bacterial culture was incubated at 37°C for 24 h and then 25°C for 12 h. Finally, ;80-100 naïve or trained worms were placed on the PA14/NGM plates with full bacterial lawn at 25°C, and the live nematodes were scored after 72 h. The survival rates were calculated as live animals/total animals.
Simultaneously, the avoidance behavior assay was conducted on NGM plates with a small lawn of P. aeruginosa PA14, in which 80-100 naïve or trained worms were added to the center of the bacterial lawn (22). After 2 h, the worms on the lawn or off the lawn were scored. The avoidance index was calculated by worms off the lawn/total numbers of worms.

Quantitative real-time PCR
Synchronized young adults, either naïve or post-pathogen exposure, were washed with M9 buffer. Total RNA was extracted using TRIzol (Tiangen Co., Tianjin, China). After random-primed cDNAs were generated, qPCR analysis was performed with SYBR Green JumpStart Taq ready mix for the qPCR kit (Sigma-Aldrich) following the manufacturer's instructions. The partial act-1 gene was used as an internal control. The PCR amplification used 40 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 40 s on ABI PRISM 7000 real-time PCR (Applied Biosystems, Foster City, California, USA). The real-time PCR experiments were repeated three times for each reaction using independent RNA samples.

6-Hydroxydopamine assays
L4 larvae cultured on NGM were transferred to the mixture containing 10 mM ascorbic acid and 50 mM 6-hydroxydopamine (28). The worms with the drug were treated for 1 h at 24°C and gently mixed every 15 min during treatment. Then the animals were washed in M9 buffer and spread on E. coli OP50-seeded NGM for 2 h. After treatment, the worms were assessed for survival and avoidance.

UPLC-MS/MS analysis
Quantification of serotonin and dopamine was performed using UPLC-MS/MS. Approximately 500 ml of synchronized worms, either naïve or post-pathogen exposure, were harvested from NGM, and 500 ml of 0.2 M HClO 4 was then added into each sample. The samples were then sonicated and centrifuged at 12,000 rpm for 15 min. The supernatants were prepared for the identification and quantification of dopamine and serotonin. The concentrations of serotonin or dopamine were calculated as the total amount in the sample/wet weight of sample worms.

Statistical analysis
All of the data are expressed as the means 6 S.D., and statistical comparisons were performed by GraphPad Prism5. The specific analysis methods were described in the corresponding figure legends. Every experiment of behavior and survival assay was performed at least three independent replicates.

Data availability
All data are contained within the article.