Background

The Janus kinase-signal transducer and activator of transcription (JAK/STAT) signaling pathway was originally characterized in mammals as an intracellular signaling pathway regulating cytokine signaling [1,2,3,4,5] and was later found to be highly conserved in various organisms, including Drosophila [6,7,8,9]. Four JAKs and seven STAT gene products have been identified in mammals, whereas in Drosophila, the pathway components are comprised of simply one JAK and one STAT, making it an excellent model system to study their biological roles [8, 10]. In both mammals and Drosophila, the canonical signaling is initiated by the binding of an extracellular ligand to transmembrane receptors that induce dimerization and activate the JAKs associated with these receptors. Then, the activated JAKs phosphorylate tyrosine residues on the cytoplasmic tails of the receptors that serve as docking sites for cytoplasmic STAT transcription factors, which then are phosphorylated, dimerize and translocate to the nucleus to activate transcription of target genes [11,12,13]. It has been well-established across various species that the canonical JAK/STAT pathway performs essential roles in development, immune signaling and cancers by its direct transcriptional activation of target genes [8, 14,15,16,17,18].

The non-canonical JAK/STAT pathway was initially identified in Drosophila, in which both the STAT transcription factor and the protein inhibitor of activated STAT homologue (dPIAS) were found to be suppressors of variegation, a heterochromatin-mediated phenomenon [

Results

Analysis of various JAK /STAT mutant embryo transcriptome changes relative to wildtype show clustering based on genotype and embryo stage

We collected early Drosophila embryos (0–12 h eggs) from the following four different genotypes: Jak gain- (hop Tum-l/+), and loss-of-function (hop 3/+), Stat loss-of-function heterozygotes (Stat92E 06346/+), and the Stat loss-of-function maternal null (Stat92E mat-). We conducted microarray experiments to compare the transcriptome of these embryos (Additional file 1: Table S1). Gene expression levels were calculated as fold changes relative to wildtype (w 1118) embryos of the same stage. Hierarchical clustering analyses indicated that hop Tum-l/+ and Stat92E mat- embryos were most dissimilar, with many genes expressed in the opposite manner (Fig. 1A), consistent with gain- and loss-of-function of the JAK/STAT pathway. Analysis of differentially regulated probe sets found clustering of hop Tum-l/+ and hop 3/+ together, then Stat92E 06346/+, and Stat92E mat- (Fig. 1A; Additional file 2: Table S2). Next, we assessed the number of overlap** probe sets between Stat92E mat- and Stat92E 06346/+ upregulated and downregulated sets of genes (Additional file 3: Table S3). For the upregulated probe sets there were 49 that were shared between the 381 Stat92E mat- and 579 Stat92E 06346/+ and for the downregulated set, there were 47 that were shared between the 788 Stat92E mat- and 640 Stat92E 06346/+ significantly downregulated probe sets (Fig. 1B). Since Stat92E 06346/+ heterozygous embryos are expected to undergo normal development in spite of reduced Stat gene activity, the difference between the transcriptomes of Stat92E mat- and Stat92E 06346/+ might be due to a failure in activation of the zygotic genome in Stat92E mat- embryos, as we have previously shown [54]. The differentially regulated genes include many of the known canonical JAK/STAT pathway targets genes, such as the Turandot (Tot) family of humoral factors [55] (Additional file 2: Table S2). We validated the differential expression of several of these known JAK/STAT target genes using RT-PCR in the hop Tum-l/+ overexpression mutants compared to wildtype (Additional file 4: Fig. S1). These results suggest that the activation status and levels of JAK and STAT affect the transcriptome of the early Drosophila embryo. In our previously published publication, we have validated reduced expression in the early loss-of-function embryo of the dpp, Kr, tll, and eve target genes that we have also found in our current microarray analysis [54].

Fig. 1
figure 1

Hierarchical clustering of Jak gain-of-function and Stat loss-of-function mutants show separation by gene and stage of the embryo. (a) Differentially regulated genes were clustered by Pearson’s correlation complete distance separation. (b) Overlaps between Stat92E mat- and Stat92E 06346/+ embryos were determined and suggest that a large number of genes are regulated stage-specifically

Full size image

Canonical versus non-canonical JAK/STAT targets can be inferred from differential transcriptional analyses across the genome of various mutants

A hallmark of the noncanonical STAT pathway is that loss of STAT and JAK overactivation both result in reduced heterochromatin levels and previously demonstrated to be relevant in multiple heterochromatin-involved processes [26,27,28]. Thus, genes normally repressed by heterochromatin would be depressed as a result of reduced heterochromatin. These studies showed that in contrast, in the canonical pathway, loss of STAT and JAK overactivation have opposite effects on STAT target gene expression.

Since the hierarchical clustering analysis showed large differences in overall transcriptomes of Stat92E mat- and Stat92E 06346/+ embryos, we aimed to differentiate between canonical versus non-canonical JAK/STAT pathway by analyzing significantly regulated probe sets of Jak gain- (hop Tum-l/+) and loss-of-function (hop 3/+) and Stat loss-of-function (Stat92E 06346/+) separately from the maternal null (Stat92E mat-). For the putative canonical targets assessing overlaps among hop Tum-l/+ upregulated, hop 3/+ downregulated and Stat92E 06346/+ downregulated genes, we found 221 putative probe sets and for the putative non-canonical targets assessing overlaps among hop Tum−/+l upregulated, hop 3/+ downregulated and Stat92E 06346/+ upregulated genes, we found 371 putative probe sets (Fig. 2A). While comparing putative canonical targets by assessing overlaps among hop Tum-l/+ upregulated, hop 3/+ downregulated and Stat92E mat- downregulated genes, we found 66 putative probe sets and for the putative non-canonical targets assessing overlaps among hop Tum-l/+ upregulated, hop 3/+ downregulated and Stat92E mat- upregulated genes, we found another 66 putative probe sets (Fig. 2B). The total number of unique probe sets for further analyses was 258 putative canonical and 409 putative non-canonical probe sets (Additional file 5: Table S4).

Fig. 2
figure 2

Biologically relevant overlap** probe sets of putative canonical and non-canonical transcriptional targets inferred by microarray. Significant genes were assessed for overlaps among the Jak-Stat mutants previously determined to be relevant for the canonical versus non-canonical heterochromatin-mediated silencing mechanism [19, 26] for (a) embryo samples and (b) maternal null and embryo samples

Full size image

Non-canonical targets are significantly more enriched with heterochromatin markers

Since the distinct mode of the non-canonical JAK/STAT pathway involves heterochromatin stabilization by unphosphorylated STAT by its interaction with HP1a, we hypothesized that we would observe increased association of putative non-canonical target loci with key heterochromatin markers, HP1a, Su(var)3–9 and H3K9me3 [19, 26, 27, 29]. We therefore identified overlaps with the publicly available modENCODE ChIP-seq database annotating loci enriched with HP1a, Su(var)3–9 and H3K9me3 in various wildtype background samples at different developmental stages and Drosophila cell culture [56]. Among probe sets that were classified as putative canonical targets by our transcriptional analyses, 25.6% had HP1a, Su(var)3–9 and/or H3K9me3 enriched site overlap and 1.6% mapped to transposable elements, compared to 35.9% and 2.2% respectively in non-canonical targets. Our results comparing the putative canonical versus the putative non-canonical targets we identified from our transcriptome analyses thus indicate that as expected, there was significant difference between these two groups, in which the putative non-canonical transcriptional target group had a larger number of overlap sites with HP1a, Su(var)3–9 and/or H3K9me3, or are transposable element sites, compared to the group classified as canonical transcriptional targets (38.1% versus 27.1%, p = 0.004 Fisher’s exact test) (Fig. 3).

Fig. 3
figure 3

Non-canonical target sites have significantly more overlaps with heterochromatin sites. For both embryonic and maternal transcripts, overlaps between genomic loci corresponding to the probe set annotation and HP1a, Su(var)3–9 and/or H3K9me3 enriched loci listed in the relevant modENCODE database were tabulated. Probe sets annotated as transposable elements were also considered as heterochromatin sites. Non-canonical target probe sets had significantly higher proportion of such heterochromatin-related sites compared to canonical targets (p = 0.004, Fisher’s exact two-tailed test)

Full size image

Canonical and non-canonical targets have distinct biological roles

After having classified probe sets to canonical versus non-canonical JAK/STAT pathway transcriptional target genes, we sought the biological significance of these different targets. We used the DAVID Gene Ontology database to determine biological functions and pathways that are enriched for the canonical and non-canonical target gene panel [57, 58]. Canonical targets appeared to be mainly involved in development and innate immunity, as expected for their well-established role, as seen by the top 20 fold enrichment of GO Biological Functions terms (Fig. 4). Ten out of 20 top enriched terms were relevant to development, including “larval chitin-based cuticle development,” “establishment of epithelial cell apical/basal polarity,” “body morphogenesis,” “chitin-based cuticle development,” “metamorphosis,” “negative regulation of cell proliferation,” “chorion-containing eggshell formation,” “wing disc development,” “imaginal disc-derived wing morphogenesis” and “multicellular organism reproduction.” Terms relevant to innate immune response were also found four times, including “innate immune response,” “response to bacterium,” “Toll signaling pathway” and “defense response to Gram-positive bacterium.” Other enrichment terms suggested a role for the canonical pathway in olfactory and sensory response, including “detection of chemical stimulus involved in sensory perception of taste,” and “sensory perception of smell.” Additional GO terms suggested transport mechanisms, including, “microtubule-based process” and “transmembrane transport” and others included “cellular response to heat” and “proteolysis.”

Fig. 4
figure 4

Canonical transcriptional targets are generally enriched with biological processes relevant to development and innate immunity, whereas non-canonical transcriptional targets appear to be involved with metabolism. (a) Gene Ontology (GO) enrichment of Biological Processes terms was assed using DAVID and the 20 highest fold enrichment GO terms and (b) all hits of KEGG pathway enrichment terms are shown for the canonical versus non-canonical target probe sets, in order of fold enrichment

Full size image

On the other hand, non-canonical targets were enriched mainly with terms relevant to metabolism, including “glucose metabolic process,” “ecdysteroid metabolic process,” “pyruvate metabolic process,” and stress response including “detection of temperature stimulus involved in thermoception,” “cold acclimation,” “thermotaxis,” “mechanosensory behavior,” and “cellular response to oxidative stress.” Female sex and egg development also appeared to be prevalent among the GO enriched terms, including “imaginal disc-derived female genitalia development,” “eggshell chorion assembly,” “vitellogenesis” and “vitelline membrane formation involved in chorion-containing eggshell formation.” Regulation of visual perception also appeared multiple times including “compound eye retinal cell programmed cell death,” “visual perception” and “deactivation of rhodopsin mediated signaling.” “Regulation of G-protein coupled receptor protein signaling pathway” was also among the top fold enrichment. The observation that the highest GO term fold enrichment was seen for “negative regulation of RNA splicing” is noteworthy, and may be related to the role of the non-canonical pathway in epigenetic signaling. GO terms related to chorion formation were found in both canonical and non-canonical targets, suggesting that while a large number of target loci distinct to canonical or non-canonical pathways exist, there is also a possibility of shared processes and targets.

Discussion

Our current study aimed to differentiate between canonical versus non-canonical JAK/STAT target genes in Drosophila by using a genome-wide transcriptional analysis approach. The use of the Drosophila system with a simple JAK/STAT system that involves only one JAK and one STAT with mutant lines readily available facilitated this current study that aimed to distinguish between the two pathways by transcriptional analysis. The genetic analysis of the non-canonical JAK/STAT pathway and the establishment of the paradigm of unphosphorylated STAT as a key player in epigenetic regulation in the nucleus has been conducted in Drosophila [

Methods

Fly stocks/genetics and RNA sample preparation

All crosses were carried out at 25 °C on standard cornmeal/agar medium. Fly stocks of w 1118, hop3/FM7c, Stat92E 06346 /TM3, FRT 82B [ovo D1 , w + ]/TM3, and hop Tum-l /FM7c were obtained from the Bloomington Drosophila Stock Center (Bloomington, IN). For the preparation of heterozygous embryos, females from the mutant stocks were crossed with w 1118 males and the resulting hop 3/+, Stat92E 06346/+ or hop Tum-l/+ progeny were used to produce embryos, which were collected between 0 and 12 h after egg laying on apple agar plates with yeast paste. To generate Stat92E mat– embryos, hsp70-flp; FRT 82B Stat92E 06346 /TM3 females were crossed to hsp70-Flp; FRT 82B [ovo D1 , w + ]/TM3 males. Third instar larval progenies were heat-shocked at 37 °C for 2 h daily over 3 to 4 days, and resulting adult females of the genotype hsp70-flp; FRT 82B Stat92E 06346 /FRT 82B [ovo D1 , w + ] which were used to produce embryos lacking in maternal Stat92E gene products, as described previously in the dominant female-sterile “germline clone” technique [65]. Stat92E mat– and control w 1118 were collected between 1 and 2 h after egg laying on apple agar plates with yeast paste. The embryos were washed twice with deionized water and total RNA was prepared using the RNeasy Plus Mini kit (Qiagen) according to the manufacturer’s manual. RNA quality was assessed using the Agilent 2100 Bioanalyzer and the RNA 6000 Nano kit (Agilent Technologies Inc., Palo Alto, CA).

Microarray analyses

To prepare microarray samples from the RNA prepared, 200 ng of total RNA was used to prepare biotin-labeled RNA using Ambion MessageAmp Premier RNA Amplification Kit (Applied Biosystems, Forster City, CA). Briefly, the first strand of cDNA was synthesized using ArrayScript reverse transcriptase and an oligo(dT) primer bearing a T7 promoter. Then DNA polymerase I was used (in the presence of E. coli RNase H and DNA ligase) to convert single-stranded cDNA into double-stranded DNA (dsDNA), which was then used as a template for in vitro transcription in a reaction containing biotin-labeled UTP and T7 RNA Polymerase to generate biotin-labeled antisense RNA (aRNA). Twenty μg of labeled aRNA was fragmented and 15 μg of the fragmented aRNA was hybridized to Affymetrix Drosophila Genome 2.0 Array Chips according to the manufacterer’s Manual (Affymetrix, Santa Clara, CA). Array Chips were stained with streptavidin-phycoerythrin, followed by an antibody solution (anti-streptavidin) and a second streptavidin-phycoerythrin solution, performed by a GeneChip Fluidics Station 450. The Array Chips were scanned with the Affymetrix GeneChip Scanner 3000. For the numerical conversion to expression intensity and Present/Absent calls employing MAS5 [66] (Additional file 1: Table S1), the Genespring software (Agilent Technologies Inc., Palo Alto, CA) or the R package Affy was used [67, 68]. R version 3.1.3 was used for the analyses.

For each mutant genotype, control probe sets were filtered, as well as those where the wild-type and respective mutant intensities all had the “Absent” call. The top 10th percentile upregulated and downregulated log 2 fold change of all probes were found to be 1.027 and −1.047, respectively and therefore the 2-fold change cut-off was considered to be significantly differentially regulated genes for each mutant genotype (Additional file 2: Table S2). Pearson’s correlation with complete distance separation was used for the clustering and heatmap representation of the differentially regulated probe sets. For conducting RT-PCR, 0–12 h embryos were collected and total RNA was harvarested using the RNeasy kit (Qiagen). The SuperScript™ III Reverse Transcriptase kit (Invitrogen) was used to generate cDNA as a template for semi-quantitative PCR.

HP1a and Su(var)3–9 binding and the heterochromatic H3K9me3 enriched sites were obtained from the publicly available modENCODE ChIP-seq database for comparison with genomic sites associated with the relevant probe sets identified from the microarray analysis [56]. The following modENCODE data files were used: for HP1a binding sites, #3956 (OregonR 14–16 h embryo), #323 (S2 cells), #955 (OregonR 3rd instar larvae), #2074 (S2 cells), #2665 (OregonR 2–4 h embryo), #2666 (BG3-c2 cells), #2668 (S2 cells) and #3956 (OregonR 14–16 h embryo), for Su(var)3–9 binding sites, #952 (BG3-c2) and #2673 (S2), and for H3K9me3 enrichment, #971 (yellow cinnabar brown speck 0–4 h embryo) and #4939 (OregonR 14–16 h embryo). In our study, if one or more enrichment sites were found to overlap with the annotated genomic locus indicated by the microarray data, the gene was deemed to be relevant to heterochromatin. For the overlap between probe sets upregulated by hop Tum-l/+ and downregulated by hop 3/+, probe sets overlap** with downregulated Stat92E 06346/+ was categorized as putative canonical targets, whereas those upregulated were categorized as non-canonical targets. Similarly, Stat92E mat- was also analyzed by taking into consideration, overlaps with hop Tum-l/+ upregulated and hop 3/+ downregulated probe sets. Probe sets where the Stat92E 06346/+ and Stat92E mat- showed opposite trends were removed from the analyses, as well as hop Tum-l/+ and hop 3/+ overlaps without significant Stat92E 06346/+ or Stat92E mat- differential expression.

The two-tailed Fisher’s exact test was used to determine significance between the difference in the number of probe sets for which their relevant sites overlapped with HP1a, Su(var)3–9 and/or H3K9me3 enriched sites and transposable elements comparing putative canonical versus non-canonical targets. Database for Annotation, Visualization and Integrated Discovery (DAVID), version 6.8 Beta was used for functional annotation and assessing the top 20 Fold Enrichment of Gene Ontology terms of putative canonical and non-canonical sets [57].