Transcription factor NF-κB is modulated by symbiotic status in a sea anemone model of cnidarian bleaching

Transcription factor NF-κB plays a central role in immunity from fruit flies to humans, and NF-κB activity is altered in many human diseases. To investigate a role for NF-κB in immunity and disease on a broader evolutionary scale we have characterized NF-κB in a sea anemone (Exaiptasia pallida; called Aiptasia herein) model for cnidarian symbiosis and dysbiosis (i.e., “bleaching”). We show that the DNA-binding site specificity of Aiptasia NF-κB is similar to NF-κB proteins from a broad expanse of organisms. Analyses of NF-κB and IκB kinase proteins from Aiptasia suggest that non-canonical NF-κB processing is an evolutionarily ancient pathway, which can be reconstituted in human cells. In Aiptasia, NF-κB protein levels, DNA-binding activity, and tissue expression increase when loss of the algal symbiont Symbiodinium is induced by heat or chemical treatment. Kinetic analysis of NF-κB levels following loss of symbiosis show that NF-κB levels increase only after Symbiodinium is cleared. Moreover, introduction of Symbiodinium into naïve Aiptasia larvae results in a decrease in NF-κB expression. Our results suggest that Symbiodinium suppresses NF-κB in order to enable establishment of symbiosis in Aiptasia. These results are the first to demonstrate a link between changes in the conserved immune regulatory protein NF-κB and cnidarian symbiotic status.


Results
The Aiptasia NF-κB protein has similarities in structure, activity, and regulation to mammalian non-canonical NF-κB. To investigate whether NF-κB plays a role in the regulation of cnidarian symbiosis, we chose to analyze NF-κB signaling using the model sea anemone Aiptasia. The analysis of transcriptomic and genomic databases of Aiptasia revealed a single gene encoding an NF-κB-related protein. The Aiptasia NF-κB protein (Ap-NF-κB) has a structural organization that is analogous to mammalian NF-κB p100: an N-terminal RHD, followed by a glycine-rich region (GRR), a series of ANK repeats, and putative regulatory sites of serine phosphorylation by IKK ( Fig. 1a and Supplementary Fig. 1).
To analyze the biological regulation and molecular properties of Ap-NF-κB, we first created pcDNA-FLAG mammalian expression vectors for full-length Ap-NF-κB and a C-terminally truncated mutant (Δ427) that approximates the structure of a processed form of Ap-NF-κB. These two vectors expressed appropriately sized proteins in human 293 cells (Fig. 1b). The Ap-NF-κBΔ427 protein migrated similarly to the Nematostella NF-κB protein (Nv-NF-κB), which lacks a C-terminal IκB-like domain [22][23][24] . In an electrophoretic mobility shift assay (EMSA) using an NF-κB-site probe, extracts from 293 cells transfected with the Δ427 mutant, but not full-length Ap-NF-κB, showed a high level of DNA-binding activity, which was comparable to that seen with Nv-NF-κB (Fig. 1c) 22 .
To determine the overall DNA binding site specificity of Ap-NF-κB, we assessed the DNA-binding profile of bacterially expressed, affinity purified Ap-NF-κB in a protein-binding microarray (PBM) consisting of 2592 κB-type sites and 1195 random background sequences (Supplementary Data 1). We have previously used similar PBM analyses to compare the DNA binding site preferences of mammalian NF-κB family proteins 15 . By comparing z-scores for binding to specific sites on the PBM, we found that the DNA-binding profile of Ap-NF-κB is similar to human NF-κB p50 and Nv-NF-κB, but is distinct from human c-Rel and RelA (Fig. 1d). Furthermore, a comparison of the DNA-binding profile of a variety of other Rel/NF-κB proteins across a broad evolutionary expanse clearly places Ap-NF-κB among the subfamily of NF-κB proteins ( Supplementary Fig. 2). In addition, a comparison of the general structure of the Ap-NF-κB protein shows that it is most similar to NF-κB subfamily proteins ( Supplementary Fig. 3).
NF-κB p100 protein activation requires phosphorylation by an upstream IKK, which is followed by C-terminal proteolytic processing and nuclear localization of the mature NF-κB p52 protein. Therefore, we analyzed Ap-NF-κB for these central properties of NF-κB signaling and regulation. First, by immunofluorescent staining of transfected chicken fibroblasts, we found that the full-length Ap-NF-κB localized exclusively to the cytoplasm, whereas the C-terminally truncated Ap-NF-κBΔ427 mutant was exclusively in the nucleus (Fig. 2a). As a control, we show that Nv-NF-κB, which has a structure similar to the Δ427 mutant, was also exclusively in the nucleus of transfected cells (Fig. 2a), consistent with our previous results 22 . Secondly, we found that Ap-NF-κBΔ427 transactivated a multimeric NF-κB-site reporter in 293 cells to approximately the same extent as Nv-NF-κB (Fig. 2b) 22 . In contrast, full-length Ap-NF-κB showed little ability to activate the reporter above vector control levels (Fig. 2b). Thus, our results show that the C-terminally truncated Ap-NF-κBΔ427 protein  Processing of Ap-NF-κB. (a) Indirect immunofluorescence (green) using anti-FLAG primary antiserum of DF-1 chicken fibroblasts transfected with the indicated FLAG-tagged proteins. Nuclear staining with DAPI is shown in blue. (b) An NF-κB-site luciferase reporter gene assay was performed in 293 cells with the indicated proteins. Luciferase activity is relative to that seen with the vector control (1.0), and is the average of three experiments done in triplicate. Standard error is indicated. (c) An in vitro kinase assay using FLAG-Ap-IKK and GST-fusion proteins containing C-terminal sequences (amino acids 783-812) of wild-type and 3X Ser-Ala mutant Ap-NF-κB. The top image is a phosphorimage of 32 P-labeled proteins, showing the relevant portion of the image; the free ATP images at the bottom on the gel are not shown. The bottom image shows a Coomassie blue-stained gel of GST proteins used in the kinase assay. An anti-FLAG blot of the 293 cell extract was also performed to confirm Ap-IKK expression (not shown). (d) 293 cells were transfected with full-length wild-type or 3X Ser-Ala mutant Ap-NF-κB in the absence (−) or presence (+) of an Ap-IKK expression vector. Extracts were analyzed by anti-Ap-NF-κB Western blotting, and the full-length and processed forms of Ap-NF-κB are indicated. At the bottom is an anti-FLAG Western blot showing expression of Ap-IKK, as determined by its migration against molecular weight standards. (e) 293 cells were transfected with full-length Ap-NF-κB and Ap-IKK in the absence (−) or presence (+) of proteasome inhibitor MG132. Extracts were analyzed by anti-Ap-NF-κB Western blotting. The filter was then stripped and probed for β-tubulin as a loading control. (f) Anti-Ap-NF-κB Western blotting of 293 cells expressing full-length and processed forms of Ap-NF-κB and lysates from symbiotic Aiptasia anemones. The only bands seen on the film are shown. ns, non-specific band. has increased DNA-binding activity, nuclear localization, and transactivation ability as compared to full-length Ap-NF-κB. Ap-NF-κBΔ427 shares these three properties with the processed p52 form of the mammalian NF-κB p100 protein, suggesting that the removal of Ap-NF-κB's C-terminal ANK repeat is necessary for activation of the protein.
In the mammalian non-canonical pathway, activated IKKα initiates processing of NF-κB p100 by phosphorylating three serine residues downstream of the p100 C-terminal ANK repeats, which then promotes processing of p100 to p52 by the proteasome 19 . Ap-NF-κB contains three serine residues in a similar arrangement and location as the serine residues in human p100 that are phosphorylated by IKKα ( Supplementary Fig. 1). A BLAST analysis identified a single IKK-like protein in Aiptasia, which, by MEME and phylogenetic analyses, was indicated to be an IKKα/β-like protein homolog (Supplementary Fig. 4 and Supplementary . In an in vitro kinase assay, Aiptasia IKK (Supplementary Data 5) and human IKKα and β can all phosphorylate a peptide containing C-terminal sequences (aa 783-812) of Ap-NF-κB, but cannot phosphorylate the same peptide in which the three Ser residues are converted to Ala ( Fig. 2c and Supplementary Fig. 5). Moreover, co-transfection of 293 cells with an expression vector for Ap-IKK induces processing of wild-type Ap-NF-κB to a protein of approximately 52 kDa, but Ap-IKK did not induce processing of the SSS/AAA mutant of Ap-NF-κB under the same conditions (Fig. 2d). Ap-IKK-induced processing of Ap-NF-κB in 293 cells is reduced in the presence of the proteasome inhibitor MG132 (Fig. 2e). Co-transfection of Ap-NF-κB with human NIK (to activate endogenous IKKα) or a constitutively active human IKKβ also induced processing of Ap-NF-κB in 293 cells ( Supplementary Fig. 6). Finally, we show that lysates from Aiptasia contain forms of Ap-NF-κB that co-migrate with both full-length and processed forms of Ap-NF-κB that were detected upon expression in 293 cells (Fig. 2f). These results identify phosphorylation sites and conditions that are necessary for Ap-IKK induced processing of Ap-NF-κB, and suggest that similar processing events occur in Aiptasia. In contrast, NIK, IKKβ, and Ap-IKK did not induce cleavage of the sponge Aq-NF-κB protein, which lacks clear IKK phosphorylation sites ( Supplementary Fig. 3), upon co-transfection in 293 cells ( Supplementary Fig. 7).
Taken together, these results show that Ap-NF-κB has structural, functional and regulatory similarities to the non-canonical mammalian NF-κB protein p100, and that these properties are conserved even upon expression of Ap-NF-κB in human cells. Furthermore, the co-migration of full-length and processed forms of Ap-NF-κB expressed in 293 cells with two forms of Ap-NF-κB detected in extracts from anemones suggests that a similar processing event occurs in Aiptasia.

NF-κB mRNA, protein, and activity increase with loss of symbiosis in Aiptasia.
To determine whether Ap-NF-κB is affected by the symbiotic status of Aiptasia, we induced loss of symbiosis in anemones by two commonly used methods, namely elevated water temperature and menthol treatment. Elevated temperature has been used to mimic climate change-induced bleaching, and in the laboratory, symbiotic anemones can be bleached by increasing water temperature from 25 °C to 32 °C for 6 days. Menthol treatment produces bleached anemones after just three days of treatment 26 . With either treatment, the anemones take on a strikingly bleached appearance (Fig. 3a, top right), and they show greatly reduced numbers of symbionts by fluorescence (Fig. 3a, bottom right) and by qPCR of symbiont 28S RNA (Fig. 3b). In this paper, we refer to our heat-and menthol-treated anemones as aposymbiotic based on the greatly reduced numbers of symbionts present after either treatment.
To determine whether NF-κB expression is altered by symbiotic status in Aiptasia, we compared the protein and mRNA levels of Ap-NF-κB between symbiotic and aposymbiotic animals. The levels of processed Ap-NF-κB protein were increased approximately three-fold in heat-and menthol-induced aposymbiotic anemones as compared to symbiotic anemones (Fig. 3c) and similarly, the level of Ap-NF-κB mRNA was increased by approximately three-fold in menthol-induced aposymbiotic anemones as compared to symbiotic anemones (Fig. 3d). There was also an approximately six-fold increase in NF-κB site DNA-binding activity in extracts from mentholand heat-induced aposymbiotic Aiptasia ( Fig. 3e; Supplementary Fig. 8). That the increased DNA-binding activity in aposymbiotic anemone extracts is due to Ap-NF-κB is suggested by the similar migration of the DNA-protein complex with the complex generated by the Δ427 protein expressed in 293 cells and by the ability of Ap-NF-κB antiserum (but not preimmune serum) to supershift the complex from aposymbiotic anemones ( Fig. 3e; Supplementary Fig. 8).
To compare the whole organism pattern of Ap-NF-κB expression, we performed anti-Ap-NF-κB immunofluorescence on cryosections of symbiotic and heat-and menthol-induced aposymbiotic Aiptasia (Fig. 3f). At low magnification imaging of tentacle cross-sections (Fig. 3f, top), Ap-NF-κB staining was noticeably brighter in aposymbiotic anemones. At higher magnification (Fig. 3f, bottom), NF-κB was detected in symbiotic animals in a few cells primarily in the epidermis and away from cells containing intracellular symbionts located in the gastrodermis. After heat-or menthol-induced bleaching, there was an approximately three-to four-fold increase in NF-κB-positive cells, which are mostly in the gastrodermis where symbionts were greatly reduced in number. The three-to four-fold increase in Ap-NF-κB-positive cells seen in aposymbiotic Aiptasia tissue sections is consistent with the increased levels of Ap-NF-κB protein seen by Western blotting of whole animal extracts (Fig. 3c). The staining of Ap-NF-κB in tissue sections from both symbiotic and bleached anemones coincides with nuclear DAPI staining (Fig. 3f, right). Overall, these results show that loss of Symbiodinium from Aiptasia is correlated with increased NF-κB protein levels and DNA-binding activity as well as expanded tissue expression of nuclear NF-κB.

Increased NF-κB protein expression in Aiptasia follows loss of symbiosis.
To investigate the kinetics of increased NF-κB expression during loss of symbiosis, Ap-NF-κB expression and symbiont levels were quantified at multiple time-points during the process of menthol-induced loss of symbiosis. In these experiments, symbiotic anemones were treated for three days with menthol and were then given regular seawater for the remainder of the experiment. Aiptasia were collected during the first two menthol treatments (Days 1 and 2, i.e., D1 and D2) and at three time-points after menthol treatment when the anemones were in seawater (D4, D7 and D14). As shown by qPCR, menthol treatment induces a rapid and progressive loss of symbionts, with an ~80% loss of symbionts at D1 as compared to control anemones, and an approximately 99% decrease in symbionts by D4 (Fig. 4a). After D4, the symbiont densities remained at this low level through D14 (Fig. 4a), indicating that the symbionts were not repopulating the animal after removal of the menthol. By Western blotting, Ap-NF-κB protein levels in anemones at D1 and D2 of menthol treatment were similar to those seen in control, untreated In the bottom image, Symbiodinium fluoresce red due to chlorophyll autofluorescence. (b) qPCR using primers against Symbiodinium Clade B 28 S rRNA was performed on Aiptasia treated as indicated. Symbiodinium 28 S rRNA levels are relative to the amount in symbiotic control anemones (100). n = 3. Statistical significance compared to sym determined using an unpaired T-test: ***p ≤ 0.001 (c) Aiptasia extracts (symbiotic, heat-bleached or menthol-bleached) were analyzed by anti-Ap-NF-κB Western blotting. The relative amount of Ap-NF-κB protein was determined by band quantitation using ImageJ on Western blots of three biological replicates per group, and values were normalized to the amount of β-tubulin in each sample. Statistical significance of the increases in Ap-NF-κB in aposymbiotic animals as compared to those in symbiotic anemones was determined using an unpaired T-test: *p < 0.05. The Ap-NF-κB bands were detected as in Fig. 2. β-tubulin was detected following stripping of the primary anti-Ap-NF-κB Western blot. (d) Relative amounts of Ap-NF-κB mRNA were determined by qPCR of RNA from symbiotic and mentholbleached Aiptasia. Values are relative to the amount of NF-κB mRNA in control symbiotic anemones (1.0). Sym n = 6, apo n = 3. Statistical significance compared to sym was determined using an unpaired T-test: *p < 0.05. (e) An NF-κB-site EMSA was performed using extracts from symbiotic (SYM) and menthol-bleached (APO) Aiptasia. Δ427 indicates that extracts from 293 cells transfected with pcDNA-FLAG-Ap-NF-κBΔ427 were used (as in Fig. 1c). Supershifts were performed with Ap-NF-κB antiserum (Ab), preimmune serum, or no antiserum (−). The relevant portion of the image containing the NF-κB-DNA complex (NF-κB) or the supershifted (SS) complex is shown; not shown is the free probe that migrates at the bottom on the gel (see Fig. 1c). (f) On the left is shown indirect immunofluorescence analysis of whole-body sections from symbiotic and aposymbiotic (heat-and menthol-threated) anemones. 20 μm frozen sections were stained with Ap-NF-κB primary antiserum and Alex Fluor 488-conjugated secondary antiserum. Ap-NF-κB staining (green); DAPI staining (blue); Symbiodinium cells (red, autofluorescence). Top images were taken at 4X, bottom images at 40X (scale bar: 20 μm). e, epidermis; g, gastrodermis. The graph shows quantitation using ImageJ of relative NF-κB puncta in three different sections of tentacles for each type of anemone. Statistical significance compared to Sym was determined using an unpaired T-test: *p < 0.05. The right panels are high magnification images showing that NF-κB staining (A488) coincides with nuclear DAPI staining in anemone sections. Scale bar: 2 μm (right). Anemones were incubated with menthol on D1-D3 and were placed in fresh ASW from D4-D14. Symbiodinium 28 S rRNA levels are relative to the amount of 28 S rRNA in symbiotic animals (100). Sym n = 6, all other groups n = 3. Statistical significance for each group compared to symbiotic was determined using an unpaired T-test: **p < 0.01, ****p < 0.0001. (b) Representative anti-Ap-NF-κB Western blot of extracts from symbiotic and menthol-treated Aiptasia extracts prepared from animals treated as in (a). Shown is the portion of the gel containing the processed form of Ap-NF-κB. At this exposure, the much weaker full-length form of Ap-NF-κB was not seen (e.g., see Fig. 2f). The filter was then stripped and reprobed for β-tubulin. Quantitation (bottom) is based on the relative amount of Ap-NF-κB protein from band quantitation using ImageJ of all biological replicates for each group. In each case, Ap-NF-κB values normalized to the amount of β-tubulin in each sample, and values are relative to the amount of NF-κB in the Sym animals (1.0). Statistical significance was determined using an unpaired T-test for each group as compared to the symbiotic control group. *p < 0.05. (c) Quantitation of relative Ap-NF-κB protein from symbiotic anemones (1.0) compared to anemones that were maintained in the aposymbiotic for 73 days. NF-κB was quantified from Western blots as described for (b). Sym n = 5, Apo n = 3. Statistical significance was determined using an unpaired T-test as above: *p < 0.05.
Scientific REPORTS | 7: 16025 | DOI:10.1038/s41598-017-16168-w animals (Fig. 4b). However, by D4 Ap-NF-κB protein levels were statistically increased (i.e., 3.8-fold) as compared to control anemones, and the levels of Ap-NF-κB protein remained statistically higher than in control animals even at D14 (Fig. 4b). In addition, anemones that have been kept aposymbiotic for a longer period of time, i.e. 73 days, have significantly higher NF-κB expression as compared to symbiotic anemones (Fig. 4c). These results show that increased levels of NF-κB occur only after there has been a substantial loss of symbionts and that the increase in NF-κB protein levels is maintained in aposymbiotic animals for over two months after bleaching is initially induced.

Introduction of Symbiodinium into naïve Aiptasia larvae leads to reduced NF-κB expression.
Given that loss of Symbiodinium was correlated with increased Ap-NF-κB expression, we hypothesized that introduction of Symbiodinium into aposymbiotic Aiptasia would decrease NF-κB expression. For these experiments, naturally and completely aposymbiotic Aiptasia larvae were inoculated with Symbiodinium, and Ap-NF-κB expression was then assessed using indirect immunofluorescence of whole-mount specimens. This approach was used for two reasons: 1) we could ensure that the larvae were aposymbiotic and had never seen Symbiodinium; and 2) the population of anemones with Symbiodinium could be done within a much shorter period (i.e., about 11 days) than the lengthy period (months) required for repopulating aposymbiotic adults. In these experiments, larvae at 4-5 days post-fertilization (dpf) were inoculated with Symbiodinium, and then 5-6 days after infection the larvae were fixed and stained with Ap-NF-κB antiserum. In larvae that were successfully infected, the numbers of symbionts present ranged from 2-15 cells (mean of 7.8 Symbiodinium/larva) (Fig. 5a). Larvae infected with Symbiodinium showed a 92% decrease in Corrected Total Cell Fluorescence (CTCF) for Ap-NF-κB as compared to control uninfected larvae (Fig. 5b,c). This result demonstrates that Ap-NF-κB protein expression is down-regulated in Aiptasia upon infection with Symbiodinium.

Discussion
Overall, the results presented herein correlate NF-κB transcription factor expression and activity with symbiotic status in Aiptasia, demonstrating that NF-κB expression is suppressed by the establishment of symbiosis and is increased by loss of symbiosis. The association between increased NF-κB expression/activity and loss of symbiosis in Aiptasia is especially noteworthy given that Aiptasia is a laboratory model for the pervasive climate change-induced coral bleaching that is occurring in the environment. Our results also suggest that a form of non-canonical NF-κB processing is present in the sea anemone Aiptasia. We show that basal NF-κB proteins have DNA-binding specificities and overall structures that are more similar to human NF-κB proteins than to human Rel proteins. Moreover, the Aiptasia NF-κB protein undergoes processing in human cells when human or Aiptasia IKK proteins are overexpressed.
Mammals and some arthropods have multiple Rel/NF-κB proteins, whereas all basal metazoan or pre-metazoans have single NF-κB-like proteins, based on genomic and phylogenetic studies 11,14 . Our PBM-based DNA-binding analyses also indicate that invertebrate NF-κB-like proteins (from Capsaspora, sponge, anemones, and Drosophila) are more similar to human NF-κB p50 and p52 proteins than to c-Rel and p65 proteins ( Supplementary Fig. 2). Thus, it is almost certain that NF-κB proteins arose first and later expanded and diversified into the Rel family, consistent with earlier predictions 27 .
Our PBM analyses demonstrate that NF-κB DNA-binding specificity is highly conserved across a great evolutionary distance. That is, there is little difference between the DNA-binding site profiles of human NF-κB p50 and either sponge or anemone NF-κB proteins. These results suggest that there has not been substantial diversification in the DNA-binding site preference of NF-κB proteins along the main evolutionary branch from sponges to humans. Therefore, we propose that differences in gene networks required for new biological processes controlled by NF-κB proteins arose primarily due to mutations in the transcriptional regulatory DNA elements of new target genes. For example, the specialized developmental processes (cnidocyte formation and dorsal-ventral polarity) controlled by Nematostella and Drosophila NF-κB proteins 28,29 likely came about due to the acquisition of NF-κB binding sites in DNA regulatory elements of genes required for cnidocyte development and embryonic polarity, respectively.
In the vertebrate non-canonical pathway, the inactive, cytoplasmic NF-κB complex is a p100-RelB heterodimer. The active p52/RelB heterodimer is generated by activation of the upstream kinase NIK that phosphorylates IKKα which then phosphorylates C-terminal residues of p100 to induce proteasomal processing of p100 to p52. As we show here, Ap-NF-κB has three C-terminal Ser residues that are similar to those in human p100 ( Supplementary Fig. 1) and can be phosphorylated by human and Aiptasia IKKs (Fig. 2c and Supplementary  Fig. 5). Moreover, overexpression of human NIK (and consequent activation of IKKα), human IKKβ, or Ap-IKK in human 293 cells can induce processing of Ap-NF-κB ( Fig. 2d and Supplementary Fig. 6) to essentially the same extent as IKK phosphorylation has been shown to induce processing of human p100 30,31 . Of note, human IKKβ induced phosphorylation and processing of Ap-NF-κB more efficiently than human IKKα ( Supplementary  Fig. 6). Furthermore, there is a single IKKα/β homolog in Aiptasia (and other basal animals) ( Supplementary  Fig. 4), suggesting that basal IKKs and their NF-κB substrates have properties that they share with both canonical and non-canonical signaling proteins of mammals. Furthermore, Ap-NF-κB has a glycine-rich region (GRR) between the RHD and the ANK repeat domain, and a similar GRR acts as a stop signal for the proteasome-mediated C-terminal processing of human p100 32 . However, it is important to note that there are no known RelB-like proteins (or any Rel-like protein) in Aiptasia 33 nor any other basal organism, and we have been unable to identify a NIK-like protein in any basal organism.
Our results and others suggest that there has been co-evolution of the IKKs and their NF-κB/IκB substrates. For example, the nematode C. elegans has neither IKKs nor NF-κB/IκB proteins. The sponge Amphimedon queenslandica has an NF-κB protein without apparent C-terminal IKK phosphorylation sites ( Supplementary  Fig. 3), and Amphimedon lacks an IKKα/β ortholog (Supplementary Fig. 4). Moreover, the Relish protein of Scientific REPORTS | 7: 16025 | DOI:10.1038/s41598-017-16168-w Drosophila lacks a GRR and C-terminal IKK phosphorylation sites, and the single Drosophila IKKα/β homolog does not appear to be involved in NF-κB processing or signaling 29 .
The truncated Ap-NF-κB protein (Δ427) enters the nucleus, binds DNA efficiently, and has intrinsic transactivation activity (Figs 1c, 2a,b). Thus, the Aiptasia NF-κB pathway appears to consist of a p100-like homodimer and an upstream IKK-like kinase, and activation of the Ap-IKK-like kinase presumably leads to phosphorylation of the Ap-NF-κB C terminus and proteolytic processing. Nevertheless, even though we can promote processing of Ap-NF-κB by overexpression of Ap-IKK in human cells in culture, the vast majority of Ap-NF-κB in anemones is in the processed form (Fig. 2f) and in the nucleus of cells (Fig. 3f) regardless of whether the anemones harbor Symbiodinium or have been rendered aposymbiotic by heat or menthol treatment. Whether Ap-NF-κB is always constitutively processed in vivo in anemones or there are factors or conditions that regulate its processing is not clear at this time. It is interesting to note that the majority of mouse NF-κB p100 is in its processed form Symbiodinium. The number of symbiont cells in individual larvae were counted by manually scanning through the z-plane in confocal images. Apo, n = 8; Sym, n = 6. No symbionts were seen in Apo larvae. (b) Ap-NF-κB fluorescence in Apo and Sym Aiptasia larvae was quantified using Corrected Total Cell Fluorescence (CTCF) of confocal images measured using ImageJ. Values are relative to the CTCF value of Apo larvae (100). Five larvae were quantified for each group. ***p < 0.0001. (c) Representative immunofluorescence images of naïve aposymbiotic larvae and larvae infected with Symbiodinium. NF-κB panels, NF-κB staining alone. Merged panels: Green, NF-κB; Blue, nuclei; Red, Symbiodinium (e.g., white arrows); Magenta, α-tubulin.
when analyzed in extracts taken directly from liver, lymph nodes, and bone marrow 30 , but is primarily in its non-processed form when overexpressed in cells in culture 30 .
Our findings that laboratory-induced bleaching of Aiptasia is associated with increased expression and activity of NF-κB and that infection of naïve aposymbiotic Aiptasia larvae with Symbiodinium reduces NF-κB expression are consistent with recent findings showing that introduction of Symbiodinium into Aiptasia larvae resulted in decreased NF-κB mRNA five days later 34 . A simple interpretation of these collective results is that Symbiodinium down-regulates NF-κB expression in Aiptasia. We propose that NF-κB controls an immune pathway that must be suppressed for the establishment of symbiosis in Aiptasia. Consistent with this proposal, symbiotic Aiptasia (i.e., with reduced NF-κB) have a reduced capacity to respond to immune elicitors as compared to aposymbiotic (i.e., high NF-κB) anemones 35 . In addition, RNA-seq data comparing aposymbiotic and symbiotic adult Aiptasia show that the establishment of symbiosis is associated with down-regulation of an inflammation gene set 36 . In at least two symbiotic corals, Orbicella faveolata and Acropora palmata, increased levels of NF-κB transcripts have been found in bleached animals 10,25 .
It is important to note that the increase in Ap-NF-κB protein and activity that is seen with loss of symbiosis is not analogous to the rapid activation of NF-κB that occurs in many vertebrate and insect systems, i.e., wherein cytoplasmic NF-κB is rapidly freed from IκB inhibition to enter the nucleus. That is, the increases in Ap-NF-κB expression and activity following loss of symbiosis appear to be due to a corresponding increase in the total number of cells with nuclear Ap-NF-κB that arise following menthol or heat treatment. Thus, loss of symbiosis appears to result in increased transcription of Ap-NF-κB in a subset of gastrodermal cells only after a significant loss of symbiosis has occurred, and the resultant increase in expression of Ap-NF-κB protein is maintained for over two months in the absence of Symbiodinium. How Symbiodinium may suppress Ap-NF-κB mRNA and/or protein levels is not clear at this time.
NF-κB is involved in the control of innate immunity across a broad range of species 11 , and many pathogens and symbionts have mechanisms to modulate innate immunity, in some cases likely through effects on NF-κB, to ensure their survival within in the host. Indeed, NF-κB activity has been shown to be reduced by viral infection in humans, by an algal symbiont in salamanders, by bacterial symbionts in squid, and by commensal bacteria in the Drosophila gut microbiome [37][38][39][40][41][42] . Future studies will be aimed at determining the role of Ap-NF-κB in immunity and determining if NF-κB activity is increased by bleaching in other cnidarian dysbiosis settings, e.g., in corals, which are more deleteriously affected by loss of symbiosis.

Methods
Plasmid constructions and cell culture. Details about plasmids and plasmid constructions, as well as primers used for PCR amplification, EMSAs and qPCR, are included in Supplementary Tables 1 and 2. Plasmids were verified by restriction enzyme mapping and/or DNA sequence analysis.
DF-1 chicken fibroblasts and human 293 cells were grown in Dulbecco's modified Eagle's Medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (Biologos), 50 U/ml penicillin, and 50 μg/ml streptomycin. Transfection of cells with expression plasmids was performed using polyethylenimine (Polysciences, Inc.) as described previously 23,24 . Cells were seeded to be ~60% confluent on the day of transfection at which time cells were incubated with DNA/PEI at a ratio of 1:6 in serum-free media (300 µl for a 60 mm plate) for 15 min at room temperature. Following incubation, this mixture was added to cells with 3 ml DMEM/10% FBS. Twenty-four hours later, the medium was replaced with 3 ml of fresh DMEM/10% FBS. The following day, cells were harvested. If cells were used for immunofluorescence, they were passaged onto glass coverslips on the day prior to fixation. If cells were used for Western Blotting they were lysed by boiling for 10 min in 2x SDS sample buffer followed by a 10 min centrifugation to clarify the samples. If cells were used for EMSA they were lysed in AT lysis buffer supplemented with protease inhibitors (20 mM HEPES, pH 7.9, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 20% wt/ vol glycerol, 1% [wt/vol] Triton X-100, 20 mM NaF, 1 mM Na 4 P 2 O 7 ·10H 2 O, 1 mM dithiothreitol, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 1 μg/ml leupeptin, 1 μg/ml pepstatin A, 10 μg/ml aprotinin) followed by passing the sample five times through a 27.5 gauge syringe (as described previously 23,24 . Samples were centrifuged at 14,000 rpm and the supernatants were used for EMSAs, as described below. Anemone maintenance and preparation of anemone lysates. Exaiptasia pallida strain H2 (originally from Kaneohe Bay, Oahu, HI) animals, symbiotic with Symbiodinium minutum (clade B1), were maintained in 150 mls of artificial sea water (ASW) in glass dishes at 25 °C with light provided by Sylvania Gro-Lux (GRO/Aq/RP) fluorescent bulbs at approximately 20 µmol photons/m 2 /sec with a 12 h:12 h light:dark cycle. Menthol-induced loss of symbiosis was achieved by treatment of symbiotic anemones in a glass dish with 150 ml of 0.58 mM menthol in ASW (that had been aerated for 4 h prior to use). Anemones were incubated in the dark with the menthol/ASW solution with fresh menthol changes daily for three 24 h treatment periods 26 . After the last treatment, anemones were placed into fresh, aerated ASW and collected at either day 4, day 7, or day 14 (days numbered from beginning of treatment). Long-term aposymbiotic anemones were isolated by bleaching anemones in menthol and then storing the animals in the dark with regular three times per week artemia feedings for 73 days before assaying by the levels of NF-κB protein by Western blotting. Heat treatments were performed by placing symbiotic anemones into 150 ml of ASW in a glass bowl and the temperature was raised from 25 °C to 32 °C by increasing the temperature 1 °C every hour for 7 h. Then anemones were kept at 32 °C for 6 days (with daily water changes to maintain salinity levels at 32 ppt) to induce bleaching, after which they were collected for assaying. Anemones (average oral disc: 3 mm) were then lysed for Western blotting or EMSA, or were flash frozen for later use in qPCR. For the preparation of lysates for Western blotting, single anemones were homogenized in 20 µl of 2x SDS sample followed by boiling for 10 min and a 10-min centrifugation to remove cell debris. For extracts for EMSAs, Aiptasia were first homogenized in AT lysis buffer (as above), and samples were then centrifuged at ~19,000 × g, and the supernatants were used for EMSAs.
Scientific REPORTS | 7: 16025 | DOI:10.1038/s41598-017-16168-w Generation of anti-Ap-NF-κB antibody. A GST-Ap-NF-κB RHD fusion protein (containing amino 2-422 of Ap-NF-κB), encoded by plasmid pDEST15-Ap-NF-κB-RHD, was expressed in BL21 bacterial cells. The GST-Ap-NF-κB protein was then purified using glutathione affinity column chromatography. Seven mg of purified GST-Ap-NF-κB protein was sent to Thermo Fisher Pierce for custom antibody production in rabbits, and the resulting antiserum was further purified by passage through an anti-Ap-NF-κB column and an anti-GST column to obtain antibodies to only Ap-NF-κB.
Western blotting. Western blotting was performed essentially as described previously 23 . Briefly, cell extracts were separated on a 7.5% SDS-polyacrylamide gel. Proteins were then transferred to nitrocellulose in transfer buffer (20 mM Tris, 150 mM glycine, 10% methanol) at 4 °C at 250 mA for ~1 h followed by 160 mA overnight. The membrane was blocked in TBST (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.1% v/v Tween 20) containing 5% powered milk (Carnation) for 1 h at room temperature. Filters were incubated at 4 °C with primary antiserum diluted in 5% milk TBST as follows: FLAG antiserum (1:1000, Cell Signaling Technology), Ap-NF-κB antiserum (1:10,000; see above) or human β-tubulin antiserum (1:1000, Cell Signaling Technology) that cross-reacts with Aiptasia β-tubulin. Membranes were washed three times for 10 min each in TBST and then incubated with anti-rabbit HRP secondary antiserum (Cell Signaling) at 1:2500 in TBST/5% milk for 1 h at room temperature with gentle shaking. Membranes were washed as above and reactive bands were detected using SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Pierce) on Blue Basic Autoradiography Film (BioExpress). The identities of protein bands were assessed in relation to either known protein molecular weight standards or internal control proteins on the Western blots. Quantification of bands on Western blots was performed using ImageJ on scans of the original film images. For each sample, Ap-NF-κB protein expression was normalized to β-tubulin, and the average and standard error of each treatment were calculated across three biological replicates per treatment. Each treatment was compared to control, symbiotic anemones and a statistical analysis was performed with an unpaired, two-tailed t-test using the GraphPad QuickCalcs online software. Differences in protein expression were considered statistically significant if p < 0.05 and are denoted by (*) in figures. For representation of Western blots in figures, films were first scanned using uniform scanning settings with an Epson Perfection V500 scanner, and images were mocked up using Adobe Illustrator C56 software.
Electrophoretic mobility shift assays. EMSAs were performed essentially as described 22 . A double-stranded κB-site oligonucleotide (GGGAATTCCC) was end labeled with T4 polynucleotide kinase (New England Biolabs) in 15 μl reaction with 1x T4 polynucleotide kinase buffer, 1 μg double-stranded probe, and 20 μCi [γ-32P]-dATP (Perkin Elmer) at 37° for 2 h. The reaction volume was then brought up to 100 μl with TE and excess label was removed by filtration through a Bio-Spin 6 Chromatography Column (Bio-Rad). For binding reactions using lysates from 293 cells, approximately 5-10 μg total protein and 100,000 cpm of the NF-κB-site probe were used. For Aiptasia lysates made in AT lysis buffer, approximately 50-60 μg of total protein and 200,000 cpm of the radiolabeled NF-κB-site probe were used. Reactions were carried out in either animal lysate binding buffer (10 mM Hepes pH 7.8, 50 mM KCl, 1 mM DTT, 1 mM EDTA, 4% w/v glycerol) or cell lysate binding buffer (10 mM Tris pH 7.4, 50 mM NaCl, 1 mM DTT, 1 mM EDTA, 4% w/v glycerol) for 30 min at 30 °C. Supershifts were performed by incubating the samples with 2 µl of antiserum for 2 h on ice after completion of the binding reaction. Samples were analyzed on 5% polyacrylamide gels, and bands were visualized on dried gels by autoradiography or phosphorimaging. Final images were prepared by uniform scanning and mocking up of images using Adobe Illustrator C56 software.

Protein-binding microarray experiments and analysis.
PBM experiments were carried out using custom NF-κB oligonucleotide arrays designed as part of this study (Agilent Technologies, AMADID 045485), based on a previously published "10-mer" κB microarray used to assay human and mouse NF-κB proteins 15 . DNA probe sequences synthesized on the custom-designed arrays are listed in Supplementary Data 1. Proteins used for PBM analysis were expressed in BL21 cells from Gateway pDEST15 GST-tagged vectors. BL21 cells were grown at 37 °C in LB containing chloramphenicol (50 µg/ml) and ampicillin (100 µg/ml), and GST protein expression was induced with 1 mM IPTG. Cells were mechanically lysed using a French press in the presence of protease inhibitors. GST-tagged proteins were purified with glutathione agarose (Thermo Fisher Scientific) and stored at −80 °C in 10% glycerol, before use. Probes on PBMs were made double-stranded by incubating the array with dNTP annealing mix at 85 °C for 10 min, 75 °C 10 min, 65 °C 10 min, then 60 °C for 90 min. The array was blocked at room temperature in filtered PBS with 2% milk for 1 h followed by washing in 0.1% PBS, Tween-20 (5 min) and 0.01% PBS, Triton-X (2 min). Arrays were incubated in the dark for 1 h at room temperature with ~300 nM protein and protein binding buffer (6 mM HEPES pH 7.8, 80 mM KCl, 0.5 mM EDTA, 0.5 mM EGTA, 6% glycerol, 35 ng/µl poly-dIdC, 1% milk,). Arrays were then washed with 0.05% PBS Tween-20 (3 min) and 0.01% PBS TX-100 (2 min). An A488-conjugated anti-GST (Invitrogen) antibody was used at 1:40 dilution in PBS containing 2% milk and incubated in the dark for 20 min at room temperature. Arrays were then washed as above followed by an additional wash in PBS (2 min) and then scanned on a GenePix 4400 A Scanner (Molecular Devices) and fluorescence was quantified with GenePix Pro 7.2 (Molecular Devices). PBM probe fluorescence values were spatially averaged and normalized using MicroArray LINEar Regression 43 as described 44 . Replicate experiments were combined using quantile normalization of probe fluorescence values using the normalize.quantiles method in the 'R' statistical package (www.r-project.org). For each unique DNA binding sequence, median fluorescence values were determined over eight replicate probe measurements. Log median fluorescence values (i.e., log(F)) were transformed into a 'z-score' using the mean (μ) and variance (σ) of the log median fluorescence values for the 1195 random background DNA sequences: z = (log(F) − μ)/σ). split into two groups, one was infected with Symbiodinium and the other served as an uninfected control. Prior to infection, axenic Symbiodinium (strain SSB01 47 ) cells were rinsed three times by centrifugation at 3000 × g in ASW and were counted using a Guava Flow Cytometer (Millipore) 48 . Infections were performed at 50,000 cells/ml for 5-6 days in 50 ml of ASW. Control and infected larvae were fixed in 4% formaldehyde for 4 h at room temperature. Fixed larvae were permeabilized and blocked as described above for staining of anemone tissue sections. Larvae were then incubated overnight at room temperature with rabbit Ap-NF-κB (1:10,000) and goat anti-tubulin (Sigma; 1:200) primary antisera. Larvae were then washed three times in PBS, and were incubated with Alexa fluor-488-conjugated goat anti-rabbit secondary antiserum (Invitrogen; 1:500), Alexa fluor-649-conjugated mouse anti-goat secondary antiserum (Invitrogen; 1:200), and Hoechst (Sigma; 5 µM) followed by three washes in PBS. Larvae were then pipetted onto 10 µl of Prolong Gold on SuperFrost Plus slides and mounted with a coverslip. Larvae were imaged as above using A488, A649, DAPI, and mCherry channels. Corrected Total Cell Fluorescence (CTCF) of NF-κB was quantified using ImageJ. For each image, larvae were outlined using the circle tool and area, integrated density, and mean gray value were measured as well as the mean gray value for background fluorescence. CTCF was calculated by the formula (CTCF = Integrated Density − (Area × Mean gray value of background)). Five aposymbiotic and five symbiotic larvae from two trials were used. Statistical significance was determined using an unpaired, two-tailed T-test. ***p < 0.0001.