Co-expression of Dorsal and Rel2 Negatively Regulates Antimicrobial Peptide Expression in the Tobacco Hornworm Manduca sexta

Nuclear factor κB (NF-κB) plays an essential role in regulation of innate immunity. In mammals, NF-κB factors can form homodimers and heterodimers to activate gene expression. In insects, three NF-κB factors, Dorsal, Dif and Relish, have been identified to activate antimicrobial peptide (AMP) gene expression. However, it is not clear whether Dorsal (or Dif) and Relish can form heterodimers. Here we report the identification and functional analysis of a Dorsal homologue (MsDorsal) and two Relish short isoforms (MsRel2A and MsRel2B) from the tobacco hornworm, Manduca sexta. Both MsRel2A and MsRel2B contain only a Rel homology domain (RHD) and lack the ankyrin-repeat inhibitory domain. Overexpression of the RHD domains of MsDorsal and MsRel2 in Drosophila melanogaster S2 and Spodoptera frugiperda Sf9 cells can activate AMP gene promoters from M. sexta and D. melanogaster. We for the first time confirmed the interaction between MsDorsal-RHD and MsRel2-RHD, and suggesting that Dorsal and Rel2 may form heterodimers. More importantly, co-expression of MsDorsal-RHD with MsRel2-RHD suppressed activation of several M. sexta AMP gene promoters. Our results suggest that the short MsRel2 isoforms may form heterodimers with MsDorsal as a novel mechanism to prevent over-activation of antimicrobial peptides.

the M. sexta EST database, we performed PCR amplification and RACE to obtain the full-length cDNAs of two M. sexta Relish isoforms, MsRel2A (GenBank accession number: HM363513) and MsRel2B (GenBank accession number: HM363514), and a Dorsal homologue (GenBank accession number: HM363515). MsRel2A cDNA is 1677 bp long with an opening reading frame (ORF) of 1191 bp, which encodes a putative protein of 397 amino acids. MsRel2B cDNA is 2057 bp with an ORF of 1326 bp encoding a putative protein of 442 amino acids. MsRel2A and MsRel2B have an identical Rel homology domain (RHD) and only differ at the C-terminal regions. MsRel2A and MsRel2B share 91.7% identity, but MsRel2B is 45 amino acids longer at the C-terminus. MsDorsal RHD is 263 amino acids long. Sequence analysis showed that MsDorsal-RHD is most similar to RHDs of the class II NF-κ B, while MsRel2-RHD is most similar to RHDs of class I NF-κ B ( Fig. S1A and S1B). Both MsRel2A and MsRel2B lack the ankyrin-repeat inhibitory domain, which is presence in the full-length Relish.

Expression profile of M. sexta Dorsal and Rel2. Tissue distribution profile of MsDorsal, MsRel2A
and MsRel2B in M. sexta naïve larvae was determined by real-time PCR. Since MsRel2A and MsRel2B cDNA sequences are highly identical, we cannot design primers specific for MsRel2A. But MsRel2B cDNA is longer than MsRel2A at the 3′ end. Thus, we designed primers for MsRel2 (MsRel2A + MsRel2B) and specific primers for MsRel2B in real-time PCR reactions. The results showed that MsDorsal and MsRel2 mRNAs were highly expressed in epidermis compared to other tissues (hemocytes, fat body, midgut and testis), and only MsRel2B was also expressed at a high level in the midgut ( Fig. 1A-C). To determine induced expression of these NF-κ B factors by microbial infection, M. sexta larvae were injected with Staphylococcus aureus, Escherichia coli and Saccharomyces cerevisiae, and MsDorsal and MsRel2 transcripts were measured by real-time PCR. Compared to the naïve larvae, expression of MsDorsal, MsRel2 and MsRel2B mRNAs in fat body, midgut and hemocytes was significantly induced by injection of microorganisms, but the overall induced expression level was not high, and the induction of individual NF-κ B factor genes depends on the tissues and microorganisms injected (Figs 1D-I and 2A-C). To test whether MsDorsal protein is also induced by microbial injection, Western blot was performed for cytoplasmic and nuclear protein extracts from hemocytes of naïve and E. coli-injected M. sexta larvae. The results showed that MsDorsal protein was detected at high levels in the cytoplasmic proteins of both naïve and E. coli-injected larvae (Fig. 2D, lanes Cp, arrow head); however, MsDorsal protein was detected only in the nuclear proteins of E. coli-injected larvae but not in the nuclear proteins of naïve larvae (Fig. 2D, lanes Nu), suggesting that bacterial infection can induce translocation of MsDorsal from the cytoplasm to the nucleus to activate gene expression. sexta AMP genes can be significantly upregulated by various microbial components 50 . Since NF-κ B transcription factors play central roles in activation of AMP genes, we carried out dual-luciferase reporter assays in both D. melanogaster S2 cells and Spodoptera frugiperda Sf9 cells to determine whether MsDorsal and MsRel2 can activate different AMP gene promoters, as M. sexta and D. melanogaster AMP gene promoters are activated differently in S2 and Sf9 cells 48 . Recombinant MsDorsal-RHD (Dl-RHD) and MsRel2-RHD (Rel2-RHD) (only the RHD domains), as well as MsRel2A and MsRel2B (full length proteins) were successfully expressed in both S2 cells and Sf9 cells (Fig. 3), which can be detected in both the cytoplasm (Fig. 3, lanes Cp) and the nucleus (Fig. 3, lanes Nu). Overexpression of recombinant Dl-RHD and Rel2-RHD in S2 cells can significantly activate promoter activity of several M. sexta AMP gene promoters and B. mori lebocin-4 promoter (Fig. 4A). Most AMP gene promoters, including M. sexta moricin, defensin-1 and two attacins, and B. mori lebocin-4, were activated to significantly higher levels by Rel2-RHD than by Dl-RHD, while M. sexta cecropin promoter was activated to an equally high level by Dl-RHD and Rel2-RHD, but M. sexta lysozyme promoter was activated to a significantly higher level by Rel2-RHD than by Dl-RHD (Fig. 4A). Similarly, overexpression of the full length Rel2A and Rel2B can also significantly stimulate the activity of AMP gene promoters, and Rel2A and Rel2B had similar activity in activation of most AMP gene promoters (except M. sexta attacin-2 and lysozyme) (Fig. 4B). Activation of moricin and lysozyme promoters by MsDorsal and MsRel2. We have previously characterized a moricin promoter (1400 bp) and also cloned a lysozyme promoter (1203 bp) in M. sexta. Five NF-κ B sites were predicted in the MsMoricin promoter, but only the proximal NF-κ B5 (Mor-NF-κ B5) is activated in Sf9 cells by peptidoglycan from E. coli 48 . To test whether MsMoricin can be activated by MsDorsal and/ or MsRel2 and whether the five predicted NF-κ B sites are all functionally active, reporter luciferase assays were performed with MsMoricin promoter and its deletion and mutation promoters in Sf9 cells since MsMoricin promoter showed low activity in S2 cells but high activity in Sf9 cells 48 . All the MsMoricin promoters showed almost no activity in Sf9 cells after overexpression of MsDorsal-RHD (Fig. 5A,B), indicating that MsMoricin is not activated by Dorsal. MsMoricin (1400 bp), MsMoricin-725 (725 bp) and MsMoricin-242 (242 bp) promoters were activated to similar high levels by MsRel2-RHD, but MsMoricin-40 (40 bp) promoter did not have any activity (Fig. 5A). Deletion of the predicted NF-κ B1, NF-κ B2, NF-κ B3 or NF-κ B4 site did not have an effect on the activity of MsMoricin-242 promoter activated by MsRel2-RHD, but deletion or mutation of NF-κ B5 site significantly decreased the activity of MsMoricin-242 promoter (Fig. 5B), indicating that MsMoricin is activated by Rel2 and only the NF-κ B5 site is functionally active.
We showed that MsLysozyme promoter was activated by MsDorsal-RHD (Fig. 4A), and only one NF-κ B site was predicted in the lysozyme promoter. Moricin NF-κ B site differs from lysozyme NF-κ B site only at two 3′ nucleotides but the two NF-κ B sites have opposite direction (Fig. 5C). To test whether the consensus sequence and direction of NF-κ B sites as well as other transcription factor binding sites are required for activation of AMP gene promoters by Dorsal and Rel2, we made several mutations in the Mor-242 promoter (242 bp) by replacing NF-κ B5 site with lysozyme NF-κ B site (Lyz-κ B, with an opposite direction to NF-κ B5) or reversed lysozyme NF-κ B site (Lyz-κ B-Rev, with the same direction to NF-κ B5) and with or without GATA-1 site (Fig. 5E), since GATA-1 is required for NF-κ B5 to activate moricin promoter 48 . It has also been reported that Drosophila Dif and Relish may form heterodimers to synergistically activate AMPs 17 . Thus, we also test whether co-expression of MsDorsal-RHD and MsRel2-RHD has an effect on the activity of the moricin promoters. Among the five moricin promoters, only Mor-242 promoter (containing both NF-κ B5 and GATA-1 sites) was activated by MsRel2-RHD, and only Mor Lyz-κB-Rev promoter (containing both the reversed lysozyme NF-κ B site that has the same direction to NF-κ B5 and GATA-1) was activated by MsDorsal-RHD, but none of the five promoters was activated by co-expression of MsRel2-RHD and MsDorsal-RHD (Fig. 5D).
To determine activation of MsLysozyme promoter by Dorsal and Rel2, we constructed four deletion promoters and four mutation promoters by replacing lysozyme NF-κ B with moricin NF-κ B5 (opposite direction to lysozyme NF-κ B) or reversed NF-κ B5 (same direction to lysozyme NF-κ B) with or without GATA-1 site (Fig. 6C). Since lysozyme promoter showed similar high activities in both Drosophila S2 and S. frugiperda Sf9 cells 48 , activation of lysozyme promoters was performed in S2 cells. The results showed that only the lysozyme promoter (1203 bp) but not the four deletion promoters was activated by MsDorsal-RHD, and all five lysozyme promoters showed low basal activities when MsRel2-RHD was overexpressed (Fig. 6A), indicating that the distal lysozyme NF-κ B is functional active and it binds to Dorsal but not Rel2. Among the lysozyme promoter and the four mutated promoters, only lysozyme promoter was activated by MsDorsal-RHD, and only Lyz Mor-κB5-GATA promoter (containing moricin NF-κ B5 and GATA-1 sites) was activated by MsRel2-RHD, and none of the five lysozyme promoters was activated by co-expression of MsDorsal-RHD and MsRel2-RHD (Fig. 6B).  in S2 cells and performed co-immunoprecipitation (Co-IP) experiments. Expression of MsRel2-RHD-Flag and MsDorsal-RHD-V5 in S2 cells was confirmed by Western blot analysis with monoclonal anti-Flag or anti-V5 antibody (Fig. 7A,B,D,E, lanes 2 and 3 for input), and expression of MsDorsal-RHD-V5 was also confirmed by polyclonal rabbit anti-Dorsal antibody (Fig. 7C, lane 3). Since only the RHD domains of MsRel2 and MsDorsal were expressed and the two RHD domains have similar size but different tags, they were recognized by anti-V5 and anti-Flag antibodies, respectively, but appeared at the same location on the Western blot. Co-IP experiments showed that anti-Flag antibody can pull down MsRel2-RHD-flag (Fig. 7A, lane 4) and co-precipitated MsDorsal-RHD-V5, which was recognized by anti-V5 antibody (Fig. 7B, lane 4) and anti-Dorsal antibody (Fig. 7C, lane 4). Likewise, anti-V5 antibody can pull down MsDorsal-RHD-V5 (Fig. 7D, lane 4) and co-precipitated MsRel2-RHD-Flag, which was recognized by anti-Flag antibody (Fig. 7E, lane 4). These results suggest that M. sexta Dorsal and Rel2 may form heterodimers in vivo.

Discussion
In this study, we identified a Dorsal homolog (MsDorsal) and two short isoforms of Relish (MsRel2A and MsRel2B) in M. sexta, and investigated their roles in activation of AMP gene promoters. All three NF-κ B factors were highly expressed in epidermis, but only MsRel2B was also expressed in the midgut. Expression of the three NF-κ B factors in M. sexta larvae was induced in response to microbial infections depending on the tissues and microorganisms, but the induction levels were not high. Interestingly, E. coli injection induced translocation of MsDorsal from the cytoplasm to the nucleus, suggesting that induced translocation of NF-κ B factors in the nucleus is the key to activate gene expression. MsDorsal, MsRel2A and MsRel2B were functionally active NF-κ B factors that can activate M. sexta AMP gene promoters differently. Importantly, MsDorsal can interact with MsRel2 to form Dorsal-Rel2 heterodimers, which may serve as negative regulators in activation of M. sexta AMP genes, a novel mechanism to prevent over-activation of AMPs.
In mammals, NF-κ B factors (p65, RelB, c-Rel, p50 and p52) can form homodimers and heterodimers to activate gene expression 51 . In D. melanogaster, it has been suggested that Dif and Relish may form heterodimers, and peptide linked Dif-Relish-N (N-terminal fragment of Relish) heterodimers can activate AMPs regulated by both the Toll and IMD pathways 17 . However, there has been no direct evidence for interaction of Dif (or Dorsal) with Relish, and peptide linked Dif-Relish-N dimers may not function as heterodimers. This is because the covalent linked Dif-Relish-N dimers may form non-covalent dimers of Dif-Relish-N dimers, in which Dif-Dif can be on one end, while Relish-N-Relish-N can be on the other end. Thus such non-covalent dimers of Dif-Relish-N heterodimers may still function as Dif-Dif and Relish-Relish homodimers. We for the first time demonstrated that MsDorsal-RHD and MsRel2-RHD can interact with each other and may form heterodimers, which suppressed the promoter activity of M. sexta AMP genes. By co-expressing MsDorsal-RHD and MsRel2-RHD in S2 and Sf9 cells, both homodimers of MsDorsal and MsRel2, as well as heterodimers of MsDorsal-MsRel2 can form. We already showed that homodimers of MsDorsal and MsRel2 can activate AMP gene promoters, therefore, suppression of the promoter activity of M. sexta AMP genes by co-expression of MsDorsal and MsRel2 in Sf9 cells must be due to formation of MsDorsal-MsRel2 heterodimers. Dorsal-Rel2 heterodimers as negative regulators in activation of AMP genes may be a new mechanism to prevent over-activation of the Toll and IMD pathways, since over-activation of AMPs and other immune-related genes can be detrimental to hosts 1,52 . We also noticed that co-expression of MsDorsal and MsRel2 in S2 cells only inhibited activation of Drosophila attacin promoter, but stimulated activation of cecropin, diptericin and metchnikowin promoters to significantly higher levels than by MsDorsal-RHD or MsRel2-RHD alone. This may be because activation of dipteran and lepidopteran AMP gene promoters in S2 cells (a dipteran cell line) and in Sf9 cells (a lepidopteran cell line) differs as demonstrated previously in our lab 40 , or M. sexta Dorsal and Rel2 proteins may not work the same way as D. melanogaster Dif (or Dorsal) and Relish. In addition, there has been no direct evidence for interaction of Drosophila Relish with Dif or Dorsal, and short isoforms of Relish have not been identified so far in D. melanogaster. Thus, it is not clear how Dif-Relish or Dorsal-Relish heterodimers are formed in vivo in Drosophila, since both the Toll and IMD pathways must be activated at the same time to release Dif (or Dorsal) and generate Relish-N for formation of heterodimers. Future work is to identify short isoforms of Relish, verify interaction between Relish-N and Dif (or Dorsal), and to determine regulation of AMP genes by NF-κ B heterodimers in Drosophila.
In D. melanogaster, AMP genes are regulated by the Toll pathway via Dorsal/Dif and by the IMD pathway via Relish [53][54][55] . Based on the NF-κ B sites of Drosophila AMP gene promoters, consensus sequences of NF-κ B sites for Dorsal/Dif and Relish have been proposed 56 . In lepidopteran insects, components of the Toll and IMD pathways as well as NF-κ B factors have been identified 57,58 , suggesting that the two signaling pathways in regulation of AMPs are conserved in lepidopteran insects. But it is not clear whether lepidopteran Dorsal and Relish proteins bind to the same NF-κ B consensus sequences from Drosophila, and whether the direction of NF-κ B sites as well as the non-consensus nucleotides also plays a role in selection/binding of Dorsal and Relish. Alignment of the active NF-κ B sites from Drosophila and M. sexta AMP gene promoters showed that only the first three nucleotides (GGG) are highly conserved (Fig. S1C). We have identified two NF-κ B sites with high selectivity for Relish (M. sexta moricin) and Dorsal (M. sexta lysozyme), respectively. Interestingly, the two NF-κ B sites have opposite direction and only differ in two nucleotides at the 3′ -end. By replacing the NF-κ B site in the moricin promoter with the NF-κ B site of lysozyme, and by replacing the NF-κ B site in the lysozyme promoter with the NF-κ B site of moricin in the presence or absence of GATA-1 site, we showed that both the direction and sequence (including the non-consensus sequence) of the NF-κ B site are important for selection of NF-κ B factors (Dorsal/Dif or Relish), and other transcription factors (for example, GATA-1 factor in moricin promoter) are also important for activation of AMP genes 48,56,59 . Thus, regulation of AMP gene expression by NF-κ B factors in lepidopteran insects may differ from that in Drosophila. Future work is to compare gene regulation by NF-κ B factors between dipteran and lepidopteran insects. were purchased from American Type Culture Collection (ATCC), and Spodoptera frugiperda Sf9 cells were purchased from Invitrogen (12552-014, Invitrogen).

Cloning and sequence analysis of M. sexta Dorsal and Rel2 cDNAs.
In the M. sexta EST library (http://entoplp.okstate.edu/profiles/jiang.htm), two EST fragments were predicted to encode Rel-homology domain (RHD)-containing proteins (manduca.Contig2427 and manduca.Contig7025). Gene specific primers were designed based on the EST sequences to clone the full-length cDNAs. Total RNA was prepared from the fat body of day 3 M. sexta naïve larvae using TRIzol ® Reagent (T9424, Sigma-Aldrich), and contaminated genomic DNA was removed by RQ1 RNase-free DNase I (Promega). Reverse transcription was performed using oligo(dT) primer (Promega) and ImProm-II reverse transcriptase (Promega) following the manufacturer's instructions. The 5′ and 3′ RACE reactions were performed using smarter race kit (Clontech). The opening reading frame (ORF) was predicted from the nucleotide sequence using DNAMAN (Lynnon Corporation, Quebec, Canada). BLASTP (http://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to search homologous RHD sequences. RHD sequences from various NF-κ B factors were aligned with the MUSCLE module of MEGA 6.0. The aligned sequences were used to construct a neighbor-joining tree with 1000 Bootstrap Replications 61 . RHD sequences were aligned with Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/), and the alignment result was decorated with ESPript 3.0 62 . Consensus sites of κ B sites were displayed with WebLogo3 63,64 .

Construction of luciferase reporter plasmids.
To construct different mutated promoters, site-directed mutagenesis was performed using the wild-type M. sexta lysozyme promoter (1203-bp) and moricin deletion promoter (242-bp) as templates 48  Expression and purification of M. sexta Dorsal in bacteria and preparation of polyclonal rabbit antiserum. RT-PCR was performed to obtain cDNA sequence encoding MsDorsal-RHD domain (residues 92-263). The PCR fragment was ligated into the Nco I/Xho I digested expression vector pGEX-5X, and then transformed into competent E. coli BL21 (DE3) cells. Recombinant plasmids were prepared and confirmed by restriction enzyme digestion and DNA sequencing. To express and purify recombinant GST-MsDorsal-RHD fusion protein, overnight culture of a single bacterial colony in LB medium containing ampicillin (100 μ g/ml) was diluted 1:100 in LB medium and incubated at 37 °C to OD 600 = 0.8 and then isopropyl-D-thiogalactoside (IPTG) was added (at 0.5 mM final concentration) to induce protein expression. Recombinant protein was purified using Ni-NTA agarose beads (Qiagen) under native conditions following the manufacturer's instructions. The purified recombinant GST-MsDorsal-RHD fusion protein was cleaved by thrombin, and the cleavage products were separated on 12% SDS-PAGE. The gel slice containing recombinant MsDorsal-RHD was used as an antigen to produce rabbit polyclonal antiserum at Cocalico Biologicals, Inc (Pennsylvania, USA).

Construction of recombinant pAC5.1⁄V5-His A and pIZ/V5-His expression vectors.
cDNA fragments encoding MsRel2A (residues 1-397), MsRel2B (residues 1-422), MsRel2-RHD (residues 58-227, identical for MsRel2A and MsRel2B), and MsDorsal-RHD (residues 92-263) were amplified by PCR. For pAC5.1⁄V5-His A vector (in S2 cells), forward primers for MsRel2A, MsRel2B and MsDorsal-RHD contain 5′ non-coding region recognized by Drosophila ribosome, followed by a start codon and a Kpn I site, while reverse primers contain an Apa I site. Forward primer for MsRel2-RHD contains a start codon and an EcoR I site and the reverse primer contains a Not I site followed by an in-frame Flag sequence and a stop codon. For pIZ/V5-His vector (in Sf9 cells), forward primers contain 5′ non-coding region suitable for Sf9 cell expression, followed by a start codon and a Kpn I site, while reverse primer for MsDorsal-RHD contains an Xba I site and the reverse primer for MsRel2-RHD contains an Xba I site followed by an in-frame Flag sequence and a stop codon. All primers are listed in Table S1. PCR reactions were performed with the following conditions: 94 °C for 3 min, 35  Insect cell culture and Transient transfection. D. melanogaster Schneider S2 cells or S. frugiperda Sf9 cells were maintained at 27 °C in Insect Cell Culture Media (SH30610.02, Hyclone), supplemented with 10% heat-inactivated fetal bovine serum (#10082063, Invitrogen) containing 1% penicillin-streptomycin solution (G6784, Sigma-Aldrich). For DNA transfection, cells were placed overnight to 70% confluence prior to transfection in serum-free medium (SH30278.01, Hyclone). GenCarrier-1 TM transfection reagent (#31-00110, Epoch Biolabs) was used for transient transfection based on the manufacturer's instructions. After 7 h transfection, S2 or Sf9 cells were centrifuged and resuspended in complete growth medium to induce protein expression for 48 h. The cell culture media and cell lysates were analyzed by Western blot.
Western blot analysis and Co-immunoprecipitation (Co-IP) assay. For Western blot analysis of endogenous MsDorsal, hemocytes were collected from M. sexta fifth instar naïve larvae or larvae injected with heat-killed E. coli strain XL1-blue (5 × 10 7 cells/larva) at 24 h post-injection. Nuclear and cytoplasmic proteins were extracted from hemocytes using nuclear extraction kit (Millipore, Cat. No. 2900). The concentrations of nuclear and cytoplasmic proteins were measured by Nano-Drop using BSA as the standard.
Co-immunoprecipitation (Co-IP) assays were performed the same as described previously using cell extracts containing recombinant proteins with anti-Flag M2 or anti-V5 antibody, and immunoprecipitated proteins were analyzed by immunoblotting 47 .
Statistical analysis of the data. At least three replicates of each sample were analyzed for each experiment, and experiments were repeated with three independent biological samples (or three independent cell cultures), and a typical set of data was used to make figures. Figures were made from means of three independent biological replicates with the GraphPad Prism. Significance of difference was determined by one way ANOVA followed by a Tukey's multiple comparison tests using GraphPad Prism (GraphPad, CA). Identical letters are not significant difference (p > 0.05) while different letters indicate significant difference (p < 0.05) determined by one-way ANOVA followed by a Tukey's multiple comparison test. Asterisks indicate significant difference (*p < 0.05; **p < 0.01) determined by one-way ANOVA.