Cytokine signaling through the JAK/STAT pathway is required for long-term memory in Drosophila

Abstract

Cytokine signaling through the JAK/STAT pathway regulates multiple cellular responses, including cell survival, differentiation, and motility. Although significant attention has been focused on the role of cytokines during inflammation and immunity, it has become clear that they are also implicated in normal brain function. However, because of the large number of different genes encoding cytokines and their receptors in mammals, the precise role of cytokines in brain physiology has been difficult to decipher. Here, we took advantage of Drosophila’s being a genetically simpler model system to address the function of cytokines in memory formation. Expression analysis showed that the cytokine Upd is enriched in the Drosophila memory center, the mushroom bodies. Using tissue- and adult-specific expression of RNAi and dominant-negative proteins, we show that not only is Upd specifically required in the mushroom bodies for olfactory aversive long-term memory but the Upd receptor Dome, as well as the Drosophila JAK and STAT homologs Hop and Stat92E, are also required, while being dispensable for less stable memory forms.

Although it has long been apparent that the immune system and the brain are closely connected, and rapid progress in our understanding of the molecules involved in regulating both immune and neural functions has been made, the extent and the exact mechanisms of this connection are still matters of considerable debate (1). One of the most likely mediators for communication between these two systems is the cytokine network because both neural and immune cells express cytokine receptors that regulate numerous aspects of cell function such as survival, differentiation, and motility. Although a great deal of work is currently focused on understanding the role of cytokine signaling in neuropathologies ranging from Alzheimer's disease to depression, the role of cytokines in regulating normal physiological function within the brain has received much less attention (2). Motivated by a need to understand the effects of inflammation on brain function, several studies have addressed the relationship between cytokine signaling and learning and memory in mammals. Although a consensus has been established that experimentally elevated levels of proinflammatory cytokines are detrimental for memory, these studies are often difficult to interpret because it is not simple to distinguish between specific effects on memory and secondary effects on confounding factors such as motor skills, motivation, and general malaise induced by the inflammatory response (3). Furthermore, experiments in which genes encoding cytokines or their receptors have been ablated reveal widely different results, depending on the specific cytokine studied and the particular behavioral assay used. It is therefore not clear to what extent cytokine-dependent signaling might play a physiological role in learning and memory.

Drosophila is an especially suitable model organism with which to address this question not only because its repertoire of cytokines is significantly reduced but also because the signaling pathways regulated by these cytokines are similarly simplified (4). A large number of mammalian cytokines, including interleukins and interferons, modulate transcription by recruiting the JAK/STAT signaling pathway. In the canonical JAK/STAT signaling cascade, a cytokine ligand binding to its receptor triggers the recruitment of a tyrosine kinase (JAK) that phosphorylates and activates the transcription factor (STAT). Phosphorylated STAT is subsequently transferred to the nucleus where it regulates its target genes. Whereas mammals express at least four different JAK genes and seven different STAT genes, flies have only a single copy of each: the JAK homolog Hopscotch (Hop) and the STAT homolog Stat92E (Fig. 1A) (5, 6). This simplicity makes the Drosophila JAK/STAT pathway an ideal target for loss-of-function studies. Such studies have shown that Drosophila JAK/STAT signaling is involved in diverse biological processes, including early and late development, innate immunity, germ-cell adhesion, and inhibition of apoptosis (6–10).

Fig. 1.

JAK/STAT signaling components are expressed in MBs. (A) Schematic diagram showing the parallels between the Drosophila and vertebrate JAK/STAT cascades. TF, transcription factor. (B) Schematic representation of the adult Drosophila MBs. The MB cell bodies [Kenyon cells (KC)] are located at the dorsal cortex, extending their dendrites into the calyx (Ca), which receives olfactory information from the antennal lobes. More distally, MB axons project to the anterior portion of the brain via a dense structure known as the peduncle (P), where they give rise to five major lobes (α, α′, β, β′, and γ). PB, protocerebral bridge. (C) Immunostaining of the brain of a upd-GAL4 > UAS-GFP fly (updG4>gfp, green), labeled with α-Fasciclin II antibody, which labels the MB α/β and γ lobes (blue). Upd-expressing neurons represent a subset of α/β and γ neurons, as shown by their pattern of axonal projection. Shown is the anterior view of a 3D stack projection of one side of the brain. (D) Immunostaining of the brain of a upd-GAL4 > UAS-GFP fly (green), labeled with α-Upd antibody (red) showing strong Upd protein accumulation in the calyx (white arrow). (E) Immunostaining of the brain of a MB247 > ncCHY fly (red), labeled with α-Dome antibody (green). Dome is present in the MB cell bodies as indicated by the proximity of the α-Dome signal to the MB247-positive nuclei. (F) Immunostaining of the brain of a MB247 > ncCHY fly (red) labeled with α-Stat92E antibody (green). Stat92E is present in the MB cell bodies (green arrow) as well an in the calyx (white arrow). (G) Immunostaining of the brain of a upd-GAL4 > UAS-GFP fly (green), labeled with α-Dome antibody (red) showing colocalization of the two signals in the MB cell bodies. (H) Immunostaining of the brain of a upd-GAL4 > UAS-GFP fly (green), labeled with α-Stat92E antibody (red) showing colocalization of the two signals. (I) Analysis of the brain of a upd-GAL4 > UAS-GFP fly (green). Before branching into the lobes, Upd-expressing neurons form four axonal bundles within the peduncle, which more anteriorly merge into two bundles. Three single sections obtained in the posterior-to-anterior sequence are presented. The peduncle region is outlined by a dotted line. Except in C, all images depict single confocal planes.

Although genetically simple, Drosophila has been widely used as a model for complex behavioral traits such as the formation of multiple memory phases. Forward genetic screens, an array of tools enabling precise spatial and temporal control of both gene expression and neural activity, and the development of new technologies to image neural activity in vivo have facilitated the discovery of a number of genes and circuits required for learning, storage, and recall of different forms of memory (11, 12). At the molecular level, the signaling proteins underlying learning and memory in flies appear to be well conserved in mammals, and it is therefore likely that the knowledge acquired from behavioral studies in Drosophila will be applicable to more complex systems. One of the most frequently used behavioral paradigms for associative memory in flies is olfactory aversive learning, which is a form of classical Pavlovian conditioning. In this paradigm, flies are exposed to two distinct odors, one alone and the other accompanied by electric shock. The flies learn to avoid the previously punished odor, as measured in a T maze in which both odors are presented simultaneously (13). Memory can be tested 1–2 h after training [referred to as short-term memory (STM)] or at later time points. Two distinct forms of long-lasting memory have been characterized by using this protocol (14). The first, known as anesthesia-resistant memory (ARM), forms after a single presentation of odor and shock but is enhanced by massed repetition of training cycles. ARM is resistant to both cold-shock anesthesia and inhibition of protein synthesis by cycloheximide. In contrast, a protein synthesis–dependent form of memory, known as long-term memory (LTM), is formed if training cycles are not only repeated but also spaced by 15-min intervals (14). The mushroom bodies (MBs), a bilaterally symmetrical structure in the central brain, are critical components of the circuitry required for many forms of olfactory learning and memory (Fig. 1B) (15, 16).

To test whether cytokine-mediated JAK/STAT signaling plays a physiological role in learning and memory, we used RNAi-mediated knockdown and dominant-negative strategies to specifically impair JAK/STAT activation in the adult Drosophila memory center. Our data show that cytokine signaling within this group of cells is specifically required for LTM, while dispensable for less stable forms of memory. We therefore show that a pathway well known for its involvement in the stress response and immunity plays an important role in behavioral plasticity.

Results

Cytokine/JAK/STAT Signaling Genes Are Expressed in the Adult Drosophila MBs.

To determine whether components of the Drosophila cytokine cascade are expressed in the adult brain, we first examined the localization of transcripts of the cytokine Upd, the cytokine receptor Dome, the kinase Hop, and the transcription factor Stat92E by mRNA in situ hybridization on adult brain sections. We found that all of these mRNAs are visibly present in the MB region of the adult brain (Fig. S1). To achieve better resolution, we next examined the expression pattern of Upd, Stat92E, and Dome by using whole-mount immunofluorescence labeling with previously characterized antibodies (17–20) and a Upd enhancer trap line (upd-GAL4) (21). Although α-Stat antibodies have been well described (18), we confirmed the specificity of α-Upd and α-Dome antibodies by using the Gal4-Switch/RNAi system, which allows both spatial and temporal control of gene expression (22) (Fig. S2 C and D).

Analysis of upd-GAL4 brains, labeled with an antibody directed against Fasciclin II to visualize the MB lobes, revealed that upd-GAL4–expressing neurons comprise a subset of MB α/β and γ neurons (Fig. 1C). These neurons are frequently found to be organized into four distinct bundles, visible especially well at the axonal level within the MB peduncle. The four Upd axonal tracts initially enter the MB peduncle separately, but on a more anterior level they fuse into two final bundles (Fig. 1I). We next performed immunostaining experiments with α-Upd, α-Dome, and α-Stat antibodies. Staining of the Upd enhancer trap line with α-Upd antibody showed that the two signals largely colocalize, confirming the validity of the use of the enhancer trap line (Fig. 1D and Fig. S3). Furthermore, labeling with the α-Upd antibody showed that Upd protein is present in the MBs and accumulates in the calyx, the dendritic region of MBs (Fig. 1D), although it covers a broader brain region (Fig. S4A). Dome and Stat92E are also expressed widely in the adult brain (Fig. S4 B–D), and their signal can be observed in the MBs as defined by colabeling with the MB247-specific Gal4 driver–driven expression of a nuclear-targeted marker, Cherry (ncCHY) (Fig. 1 E and F). Stat92E and Dome can both be seen localized to the Kenyon cell bodies, and the former is also present in subdomains of the calyx (Fig. 1 E–H). Dome adopts punctate localization in the Kenyon cell bodies as well as in other brain regions (Fig. 1 E and G and Fig. S4B). We postulate that this punctate appearance may reflect the active state of the receptor because many membrane receptors, after being activated by their ligands, are internalized and later recycled to the cytoplasmic membrane in small vesicles that can often be visualized as puncta. In particular, the Dome receptor has been shown to undergo endocytosis after its activation by Upd in the Drosophila egg chamber as well as in cell-culture assays (23). Colabeling experiments show that both Dome and Stat92E proteins colocalize with the signal generated by the Upd enhancer trap line (Fig. 1 G–H and Fig. S4 C and D).

Negative Regulation of JAK/STAT Signaling in Adult MBs Impairs LTM.

Having established that the components of the JAK/STAT pathway are expressed in the MBs, we tested whether signaling through the JAK/STAT pathway is involved in olfactory aversive learning and memory. Classical null mutations in JAK/STAT pathway genes are embryonic lethal in Drosophila (5–7, 20). We therefore decided to use a panel of dominant-negative transgenes driven by the Gene-Switch system (22). Thus, both potential developmental defects and effects in secondary adult tissues that might be induced through loss of function of this essential signaling cascade can be avoided. In MB247-Switch (MB-Sw) flies, Gene-Switch-Gal4 is expressed in the α, β, and γ neurons of the MBs when the flies are fed with RU486 (RU) (24).

Adult-specific overexpression of wild-type SOCS36E, an endogenous inhibitor of JAK/STAT signaling (25), causes a striking defect in LTM (Fig. 2A) without affecting ARM (Fig. 2B) or STM (Fig. 2C). In contrast, overexpression of an inactive form of SOCS36E (SOCS36E-SHmut), used here as a control, shows no impact on LTM scores (Fig. 2A). Similar LTM-specific defects were induced by expression of dominant-negative alleles of Stat92E, StatΔN, and Stat(M647H) (26, 27) as well as the Dome dominant-negative allele DomeΔCyt (17) (Fig. 2 A–C). These results suggest that appropriate levels of JAK/STAT signaling within the Drosophila memory center are specifically critical for the formation or stability of LTM.

Fig. 2.

Negative regulation of JAK/STAT signaling in the adult MBs impairs LTM. UAS/MB-Sw flies (white bars) were compared with UAS/+ flies (gray bars). All flies were fed with RU. (A) Expression in the adult MBs of the dominant-negative proteins SOCS36E, StatΔN, Stat(M647H), or DomeΔCyt, but not the inactive control SOCS36E-SHmut, impairs LTM. Significant differences in LTM scores were observed (t test, ***P < 0.001, **P = 0.0044, *P = 0.0105; n ≥ 15). (B and C) Expression of none of the above dominant-negative proteins affects ARM (B; n ≥ 10) or STM (C; n ≥ 6).

Upd, Dome, Hop, and Stat92E Are Required in the MBs for LTM.

To test whether MB-specific expression of the cytokine Upd is required for LTM, and to confirm that signaling through the canonical JAK/STAT pathway plays a physiological role in the process, we used Drosophila lines expressing RNAi constructs designed to specifically target Upd, Dome, Hop, and Stat92E expression (Materials and Methods). The knockdown efficiency of these RNAi constructs was assessed by performing quantitative real-time PCR using the ubiquitous inducible da-Switch driver (28) (Fig. S2A) and the constitutive pan-neuronal elav-Gal4 driver (Fig. S2B). The MB-Sw driver was then used to target knockdown specifically to the adult MBs. We found that knockdown of each of the four tested genes for 2 d before conditioning specifically impairs LTM without affecting ARM or STM (Fig. 3 A–C). Furthermore, impairment of LTM was not caused by leaky expression of RNAi during development because the observed phenotypes depended on the presence of RU (Fig. S5). Because RNAi-mediated knockdown is prone to off-target effects, we constructed and tested additional RNAi lines with target sequences that did not overlap those of the previous constructs. We observed the same pattern of LTM-specific impairment in these independent experiments (Fig. S6). Importantly, we confirmed that the observed LTM defects are not caused by an impairment of olfactory sensitivity or response to electric shock (Fig. S7). The MB-Sw system is based on the MB247 driver, which might label other structures than the MBs. To ensure that the LTM defects were caused by a deficiency of the JAK/STAT pathway in the MBs, we tested stat-RNAi and hop-RNAi expression by using the Gal80ts;OK107 driver, which provides temperature-controlled spatiotemporal knockdown. We observed a significant LTM impairment when the driver was induced (at 29 °C) compared with the scores observed for flies incubated at 18 °C (Fig. S8A). In contrast, STM remained unaffected (Fig. S8B). We then tested two additional MB constitutive drivers: 238Y, which drives expression in all of the MB neurons, and c739, which drives a specific expression in MB α/β neurons (29). We again observed an LTM impairment, whereas STM was not affected (Fig. S8 C and D). These data corroborate our initial results and confirm that the LTM defect observed in JAK/STAT-compromised flies is specifically caused by an impairment of this pathway in the MB neurons, and in particular in the MB α/β neurons, that are known to be involved in aversive olfactory LTM (30, 31).

Fig. 3.

Upd, Dome, Hop, and Stat92E expression is required in the adult MBs for LTM. Flies were put on RU-containing medium for 48 h before conditioning, and until testing (A and B). UAS-RNAi/MB-Sw flies (white bars) were compared with UAS-RNAi/+ flies (gray bars). (A) Specific knockdown of endogenous Upd, Dome, Hop, and Stat92E in the adult MBs impairs LTM (t test, **P = 0.0073, *P < 0.05; n ≥ 12). (B and C) Knockdown of none of the above genes affects ARM (B; n ≥ 10) or STM (C; n ≥ 9).

Discussion

Using the Drosophila olfactory aversive learning paradigm in combination with a conditional tissue-specific expression system, we have shown that cytokine signaling through the JAK/STAT pathway is necessary for protein synthesis–dependent LTM but is dispensable for less stable forms of memory. All four major components of this pathway—the extracellular cytokine Upd, the cytokine receptor Dome, the tyrosine kinase Hop, and the transcription factor Stat92E—are required within the MBs, the major olfactory learning and memory center for LTM processing.

Although cytokine signaling may be required for normal health and physiology of the MBs, we do not favor this hypothesis because neither learning nor ARM formation are affected when this signaling pathway is compromised. Rather, we suggest that the JAK/STAT pathway is specifically recruited for LTM processing. The requirement for de novo gene expression during LTM formation has been widely observed in a number of different model systems (32). Much attention has been focused on the role of transcription factor cAMP response element-binding protein (CREB) as an LTM-specific regulator of gene expression in Drosophila (33, 34) and other species (reviewed in refs. 35 and 36). A number of other transcription factors have also been found to play an important role in LTM, including Adf-1 in Drosophila (37) and CCAAT/enhancer-binding protein (C/EBP), Zif-268, AP-1, and NF-κB in mammals (reviewed in ref. 35). Although the JAK/STAT pathway has been shown to be involved in diverse biological processes in flies (38, 39), our study identifies a role in Drosophila adult brain physiology and behavioral plasticity. In addition, despite the plethora of studies examining the impact of cytokines in memory formation, the experiments presented here demonstrate that JAK/STAT signaling contributes to the transcriptional regulation thought to underlie synaptic plasticity and long-lasting memory.

To understand how Stat92E modulates memory, it will be necessary to identify its transcriptional targets in the adult MBs. Identification of such target genes could be approached by using bioinformatics and/or transcription profiling. Recent profiling studies have identified a number of putative Stat92E target genes in the Drosophila eye disk and hematopoietic system (40, 41), some of which include Notch signaling pathway genes that have already been implicated in LTM (42, 43). Another mode of action of JAK/STAT signaling in LTM could be through chromatin remodeling. Recent findings have identified a noncanonical mode of JAK/STAT signaling that directly regulates heterochromatin stability and cellular epigenetic status, affecting expression of genes beyond those under direct Stat92E transcriptional control (44, 45). Finally, given that regulation of the actin cytoskeleton is central to both cell motility and neuronal structural plasticity, it will be interesting to determine whether some of the mechanisms by which JAK/STAT signaling drives border cell migration in the Drosophila germ line are also relevant to the formation of stable memories in the MBs (46, 47).

Our experiments demonstrate a clear positive role for signaling by the cytokine Upd in olfactory aversive memory, and, in doing so, they contribute to a lively debate as to the role played by cytokines in memory. Mammalian studies in which levels of proinflammatory cytokines are increased to pathogenic levels, either through direct injection or indirectly via induction of inflammation through injection of lipopolysaccharide or bacteria, tend to suggest that augmented cytokine signaling is detrimental for performance in a variety of learning and memory assays (48, 49). This negative impact of cytokine signaling on memory is supported by studies that take a loss-of-function approach to address the physiological function of interleukins and their receptors in different cognitive tasks under nonpathological conditions (50, 51). On the other hand, several studies describe learning and memory defects attributed to loss of function of other cytokines or their receptors, using a variety of behavioral assays (52, 53). Thus, despite significant efforts, our understanding of the molecular and cellular basis for interactions between the cytokine network and learning and memory remains limited. The complexity of mammalian cytokine signaling, with its vast array of genes encoding ligands, receptors and downstream regulators, and the substantial degree of crosstalk between pathways, ensures that this task remains an enormous challenge. By using Drosophila, a simplified model system encoding single JAK and STAT genes, we now show that signaling through a cytokine-regulated JAK/STAT pathway is critical for LTM. In contrast to the mammalian gene-disruption studies described above, we are able to rule out the possibility that the observed memory impairments are attributable to defects in development because targeting of gene expression in our study was limited to adult flies. The crucial role of JAK/STAT signaling in memory, if conserved in vertebrates, may explain why inappropriate up-regulation of the pathway appears to disrupt memory, thus shedding light on the large number of diseases in which neuroinflammation is thought to drive pathogenesis.

Materials and Methods

Generation of Upstream Activating Sequence (UAS)-RNAi Transgenes.

We amplified ∼500-bp-long fragments for each JAK/STAT signaling component by using PCR and Drosophila Genomics Resource Center cDNA clones as templates. Primer sequences are shown in Table S1. The first seven bases in each forward primer contain an AvrII restriction site, and the first seven bases of each reverse primer contain an NheI site. NheI and AvrII have compatible cohesive ends and were used for subsequent subcloning into the pWIZ vector (54). All PCR products were initially cloned into the pGEM-Teasy vector (Promega). To introduce the first inverted repeat into pWIZ, each fragment was excised from the pGEM-Teasy vector by using the NheI and AvrII restriction sites and then introduced into the AvrII site of the pWIZ vector. To introduce the second inverted repeat, each NheI/AvrII fragment was introduced into the NheII site of the first pWIZ cloning product. The pWIZ final cloning products were injected into w¹¹¹⁸ embryos (performed by BestGene).

Fly Stocks.

All lines were outcrossed to a Canton Special genetic background. Two UAS-RNAi lines (UAS-updRNAi1 and UAS-statRNAi1) were acquired from the National Institute of Genetics collection. The following lines were acquired from individual laboratories: MB-Sw (R. L. Davis, Scripps Florida, Jupiter, FL), upd-GAL4 > UAS-GFP (E. A. Bach, New York University, New York), UAS-domeΔCYT (J. C. Hombría, Centro Andaluz de Biología del Desarrollo, Sevilla, Spain), UAS-ΔNSTAT92 (J. E. Darnell, Jr., The Rockefeller University, New York), UAS-Stat-M647H-GFP (M. P. Zeidler, University of Sheffield, Sheffield, U.K.), UAS-Socs36E and UAS-Socs36E-SHmut (B. Mathey-Prevot, Dana-Farber Cancer Institute/Harvard Research Center, Boston), and Gal80ts;OK107 (Y. Zhong, Cold Spring Harbor Laboratory, New York).

Gene-Switch Experiments and Olfactory Classical Conditioning.

Behavior experiments were performed as described previously (14, 30) with flies raised at 18 °C unless otherwise stated. 3-Octanol (>95% purity, Fluka 74878; Sigma-Aldrich) and 4-methylcyclohexanol (99% purity, Fluka 66360; Sigma-Aldrich) odors were used at 0.360 mM and 0.325 mM, respectively. For LTM, a spaced training protocol comprising five conditioning cycles separated each by a 15-min rest was used. To induce transgene expression, the Gene-Switch system was used as described (24). Progeny flies were transferred 1–2 d after eclosion to fresh food vials supplemented with RU (mifepristone; Sigma) to a final concentration of 200 μM, for 2 d. When memory performance was tested 24 h after training, flies were transferred to fresh RU-containing food vials after conditioning and incubated at 18 °C until testing. All behavioral measurements were analyzed by using either two-tailed Student's t tests, unless otherwise stated, or one-way ANOVA followed, if significant at P < 0.05, by the Newman–Keuls multiple-comparisons test. Error bars represent the SEM.

Immunolabeling.

Fly heads were detached from the bodies in PBS and fixed in 4% paraformaldehyde for 3 h on ice. Brains were dissected and processed for whole-mount staining according to standard protocols. The antibodies used were: rabbit α-Dome (J. C. Hombría, Centro Andaluz de Biología del Desarrollo) preabsorbed on fly heads and used at a dilution of 1:1,000, rabbit α-Upd (D. Harrison, University of Kentucky, Lexington, KY) preabsorbed on fly heads and used at a dilution of 1:1,000, rabbit α-Stat92E (X. S. Hou, National Institutes of Health, Frederick, MD) used at 1:1,000, and monoclonal α-FascII (1D4; Developmental Studies Hybridoma Bank) used at 1:3. Preliminary experiments performed with an anti-Hop antibody previously characterized on Western blot (55) did not yield a clearly reproducible immunohistological signal.