A novel function for the IκB inhibitor Cactus in promoting Dorsal nuclear localization and activity in the Drosophila embryo

The evolutionarily conserved Toll receptor pathway is implicated in the control of development, proliferation and immunity. Toll signals are modulated at many levels, a characteristic that facilitates an assortment of possible outcomes (Mitchell et al., 2016). To understand Toll pathway architecture it is necessary to define quantitatively how each pathway element contributes to the final response. Inhibitor of NFκB (IκB) proteins comprise the Toll responsive complex together with NFκB effectors. Therefore, they are central elements of the Toll pathway that require careful analysis of their effects.

Drosophila is a unique system in which to study how IκB proteins tune Toll responses, as disturbances in IκB function can be analyzed concomitantly across a range of Toll activation levels. During embryogenesis, ventral-lateral activation of maternal Toll receptors leads to a ventral-to-dorsal nuclear gradient of the NFκB/c-Rel protein Dorsal (Dl). Inside the nucleus, the differential affinity of Dl-target genes for Dl enables the spatial control of gene expression along the dorsal-ventral (DV) axis (Rushlow and Shvartsman, 2012). High Toll signals lead to ventral mesodermal gene expression, intermediate signals induce lateral neuroectodermal genes, whereas low signals allow dorsal ectodermal gene expression. Toll signals are transduced through mobilization of adaptor proteins, protein kinase activation, and ultimately phosphorylation and proteasomal degradation of the sole Drosophila IκB protein Cactus (Cact). Cact degradation then exposes nuclear localization sequences in Dl and nuclear translocation ensues (Stein and Stevens, 2014).

The amount of nuclear Dorsal (nDl) is used as readout for the level of Toll pathway activation. Quantitative analysis of nDl in fixed and live embryos has shown that the Dl gradient is visible from early blastoderm cycle 9, increasing in amplitude with time until mitotic cycle 14 (Kanodia et al., 2009; Liberman et al., 2009; Reeves et al., 2012). Although the shape and dynamics of the Dl gradient have been described under wild-type conditions, how cactus (cact) disturbances alter these characteristics has not been investigated. Using quantitative analysis we show that, under specific circumstances, loss of cact flattens the nDl gradient, implying that cact is able to augment Dl nuclear accumulation in addition to its widely established role in inhibiting Toll signals.

In order to identify how Cact tunes NFκB responses, we investigated the effects of reducing cact on Dl nuclear localization and expression of Dl target genes using quantitative fluorescent immunolabeling and in situ hybridization. It has previously been reported that the nDl gradient expands and the ventral twist (twi) expression domain widens in embryos generated from mothers carrying cact loss-of-function germline clones (Roth et al., 1991). This effect results from depriving the embryo of Cact protein, which releases Dl inhibition in the cytoplasm. Smaller reductions in cact lead to milder effects along the DV axis, as revealed by the embryonic cuticle pattern, gastrulation movements and colorimetric in situ hybridization (Govind et al., 1993; Isoda and Nusslein-Volhard, 1994; Roth et al., 1991).

To quantify the effect of reducing cact, we progressively decreased maternal Cact activity using a series of loss-of-function allelic combinations (Fig. 1 and Figs S1 and S2). One dose of a strong cact loss-of-function allele induces 30% embryo lethality, despite only a mild decrease in the number of amnioserosa cells (cact[A2]/+ mothers) (Govind et al., 1993) and no detectable alteration in nDl (Fig. 1B,E,F,I,J and Table S1). Further decreasing cact function increases nDl in regions of the embryo that receive intermediate or low Toll signals (Fig. 1B-D): in cact[A2]/cact[011], intermediate (lateral) nDl levels expand to dorsal regions of the embryo (Fig. 1C). These results are in agreement with the embryonic phenotype reported for cact[H8] homozygotes (Roth et al., 1991), showing loss of dorsal zen gene expression and ventralized cuticles (Table S1 and Fig. S2G). Surprisingly, stronger loss-of-function allelic combinations, as in cact[A2]/Df(cact) (Fig. 1D), cact[A2]/cact[4] and cact[A2]/Df(2L)III18 (Fig. S2A-C,E) have the same effect on lateral and dorsal territories, but additionally reduce nDl in the ventral domain. As these three genotypes alter cact function, their antagonistic effects on ventral and lateral nDl are likely due to cact.

The changes in nDl shown above are reflected in the pattern of Dl target genes. The lateral domain of short gastrulation (sog) expression, a gene that requires intermediate levels of nDl for activation (Stathopoulos and Levine, 2002), expands ventrally and dorsally in several cact loss-of-function combinations (Fig. 1G,H,K,L). Dorsal sog expansion results from activation by increased nDl, while ventral sog expansion probably results from a decrease in nDl and the narrower domain of the Snail repressor. Consistent with this interpretation and with the ventral decrease in nDl, the ventral prospective mesodermal domains of snail (sna) and twi expression, genes that require high levels of nDl for activation (Ip et al., 1991; Papatsenko and Levine, 2005), reduce as Cact activity drops (Fig. 1G,H,L-N, Fig. S2B,C,E and S3B). Therefore, we find that Cact performs an additional function to enhance Toll signals, which is distinct from its established function in inhibiting Dl nuclear translocation.

Next, we investigated whether this positive effect depends on Dl, taking into consideration that the deficiency used in one of the allelic combinations displaying this effect deletes dl in addition to cact [Df(cact)]. In embryos laid by dl⁻ heterozygous mothers, nDl decreases ventrally and the sna domain narrows compared with wild type (Fig. 2A,C) (Fontenele et al., 2013; Kanodia et al., 2009). We reasoned that decreasing dl should sensitize the embryo to Cact reductions if the positive role of Cact depends on Dl. In agreement with this hypothesis, a mild reduction in Cact activity attained in dl[6]/cact[A2] significantly decreases nDl in the ventral and lateral domains of the embryo (Fig. 2B). This effect is not observed in cact[A2]/+ (Fig. 1B). In addition, in dl[6]/cact[A2] the lateral sog domain expands into the ventral, high Toll activity domain (Fig. 2C-H). Under this condition the ventral sna domain is stochastic, enabling ventral sog expression in nuclei devoid of the Sna repressor, in agreement with the nDl gradient format (Fig. 2B). However, the total sog plus sna domain is similar to wild type. In dl[6]/cact[4] (Fig. S6F), we see a reduction in the sna domain compared with dl[6]/+, strengthening the hypothesis that the positive role of Cact depends on Dl.

Next, as loss of Cact can result in a decrease in nDl, we tested whether altering Cact levels by overexpression results in a corresponding increase in nDl and Dl target genes. Two pathways control Cact levels and activity: Cact is phosphorylated at N-terminal serine residues, ubiquitylated and degraded through the proteasome in response to Toll pathway activation (Bergmann et al., 1996; Fernandez et al., 2001; Hecht and Anderson, 1993; Reach et al., 1996; Shelton and Wasserman, 1993). Cact is also subject to C-terminal CKII kinase phosphorylation through a Toll-independent pathway (Belvin et al., 1995; Liu et al., 1997; Packman et al., 1997), priming Cact for cleavage by the Calpain A (CalpA) protease (Fontenele et al., 2009). N-terminal truncated Cact, hereafter referred to as Cact[E10], is not degraded in response to the Toll pathway. Interestingly, CalpA activity generates Cact[E10], and possibly releases a second full-length free Cact molecule (Fontenele et al., 2013).

To test the effect of increasing Cact activity on the nDl gradient we used gain-of-function alleles and transgenic overexpression lines. The cact[E10] and cact[BQ] mutants were characterized as gain-of-function alleles that produce N-terminal-deleted proteins that were unresponsive to Toll (Bergmann et al., 1996; Roth, 2001), but that retain their ability to bind Dl (Fontenele et al., 2013). In trans to cact loss-of-function alleles, these mutant proteins inhibit Toll pathway activity along the DV axis: the severity of the gain-of-function phenotype increases with severity of the loss-of-function allele (Bergmann et al., 1996; Govind et al., 1993; Roth et al., 1991). In cact[E10]/cact[A2], the ventral sna domain decreases, consistent with Toll pathway inhibition due to cact[E10]. Surprisingly, no inhibition is seen dorsally as the lateral sog domain remains dorsally expanded, compared with cact[A2]/Df(cact) (Fig. 3A,B and Fig. 1H). These Dl-target effects are in agreement with nDl gradient alterations (Fig. 3C). Nonetheless, cact[E10] and stronger cact loss-of-function allelic combinations are lethal, impairing further analysis of a positive effect exerted in the presence of cact[E10].

To investigate the effects of altering Cact and Cact[E10] activity, we expressed GFP-tagged Cact chimeric proteins in oocyte and early embryos using a maternal promoter [αtub67C, referred to as CaM (Fernandez et al., 2001)]. In all genetic backgrounds tested, cact-eGFP expression either had no effect or decreased nDl levels and the size of Dl-target expression territories, conforming to Cact inhibitory function (Fig. 3D-K and Fig. S4). Importantly, when expressed in a cact[A2]/cact[011] loss-of-function background, cact-eGFP recovers the nDl gradient and reduces lateral sog to a wild-type pattern (Fig. S4A,F,I, compare with Fig. 1C), confirming that CaM>cact-eGFP produces functional Cact that binds to and inhibits Dl and responds to Toll signals (Fontenele et al., 2013).

Conversely, cact[E10]-eGFP expression increases the ventral sna and decreases the lateral sog domains that were modified in cact[A2]/Df(cact) (Fig. 3D,E,G,H,J,K). Therefore, in this genetic background Toll signals enhance ventrally but are inhibited laterally compared with cact[A2]/Df(cact). This result is consistent with the nDl gradient that reverts to an almost wild-type format (Fig. S5) and akin to the opposing effects shown in cact loss-of-function assays (Fig. 1 and Fig. S2 and S3). Importantly, cact[E10]-eGFP has no effect on Dl target genes when expressed in the less severe cact[A2]/cact[011] background (Fig. S4). Thus, the effects of Cact[E10]-eGFP depend on the severity of the cact loss-of-function background, similar to the behavior displayed by endogenous cact[E10]. This characteristic may stem from differences in functionality of mutant versus wild-type Cact protein, such as different Dl-binding affinities or interactions with other elements of the Toll-dependent and -independent pathways.

The ability of Cact[E10] to alter Toll signals is strongest in the presence of limited Dl, a condition that also reduces endogenous Cact levels (Whalen and Steward 1993; Bergmann et al., 1996). The lateral sog domain expands ventrally in cact[E10]/dl[6] (Fig. S6G-I) due to loss of inhibition by Sna, although no effect is observed in cact[E10]/+. Likewise, cact[E10]-eGFP has no effect on the nDl gradient in the presence of wild-type Dl levels (Fig. S4C) but produces an inhibitory effect comparable with Cact-eGFP in a dl⁻ heterozygous background (Fig. S6A,B compare with Fig. 2A). Collectively, the results described above indicate the existence of a delicate balance between levels of full-length Cact, truncated Cact (Cact[E10]) and Dl for proper Toll signaling events.

The phenotypes presented here resulting from cact loss-of-function and overexpression alleles show that cact is able to enhance Toll signals. Loss of cact function results in loss of this positive activity. cact[E10]eGFP expression recovers the loss of this positive function, suggesting it acts on the same mechanism that is impaired in specific cact allelic combinations. This interpretation is strengthened by the observation that under the loss-of-function and in the cact[E10] gain-of-function conditions, the effects are strongest by reducing cact and/or dl. However, these results raise several questions: by what mechanism does Cact augment Toll signals and what process does cact[E10] modify to recover the loss of this positive effect?

First and foremost, endogenous Cact[E10] is a product of the Toll-independent pathway. We have shown that the CalpA protease targets exclusively Dl-free Cact. Our data suggest that CalpA generates Cact[E10] and releases a full-length Cact molecule. Conversely, Cact[E10] inhibits CalpA activity, forming a regulatory loop (Fontenele et al., 2013). Mathematical modeling pointed out the importance of IκB proteins in modulating Toll signals in vertebrates and in Drosophila sp. (Ambrosi et al., 2014; Kearns and Hoffmann, 2009; O'Dea et al., 2007). In particular, it was shown that free IκB is an important regulatory target for the control of Toll signals (Konrath et al., 2014). In Drosophila, Cact partitions between free and NFκB-bound complexes. Dl-bound Cact corresponds to the Toll-responsive complex (1Cact:2Dl), whereas Dl-free Cact (2Cact) is a target of the Toll-independent pathway (Bergmann et al., 1996; Liu et al., 1997). Taking into account that CalpA enhances Toll signals and produces Cact[E10] from Dl-free Cact, the simplest interpretation of our results is that the positive function attributed to Cact stems from the Toll-independent pathway. Conforming to this hypothesis, CalpA activity is extremely sensitive to Dl and Cact levels, as it is reduced in dl⁻ and cact⁻ (Fontenele et al., 2009, 2013), reminiscent of the positive Cact effects here described.

Notably, vertebrate calpains target IκB proteins, and the pathways that control free and NFκB-bound Cact and IκBα are well conserved (Han et al., 1999; Li et al., 2010; Pando and Verma, 2000; Schaecher et al., 2004; Shen et al., 2001; Shumway et al., 1999). Therefore, our findings of a positive function for Cact may have important implications for the control of vertebrate Toll signals.

Based on the arguments above, we propose a model in which free Cact (unbound to Dl) is modified by the action of CalpA to enhance Toll signals in the embryo by replenishing Cact:2Dl complexes (Fig. 4). However, other mechanisms involving Cact and Cact[E10] could potentially enhance Toll. Appropriate Toll responses depend on pre-signaling complex mobilization (Marek and Kagan, 2012; Sun et al., 2004), endocytosis (Huang et al., 2010; Lund et al., 2010) and DNA-bound NFκB nuclear resident time (Mitchell et al., 2016; O'Connell and Reeves, 2015). Thus, mechanisms involving Cact and Cact[E10] that modify these functions may positively impact Toll signals and explain our results (Fig. 4B). Furthermore, it was recently proposed that Cact functions by a shuttling mechanism to concentrate Dl ventrally (Carrell et al., 2016 preprint), akin to the shuttling mechanism exerted by the BMP inhibitor Sog to concentrate BMPs dorsally (Mizutani et al., 2005; Shimmi et al., 2005; Umulis et al., 2006). Although further research is required to understand how Cact enhances nDl levels in the embryo and consequently Dl-target gene expression, we have clearly shown that Cact exerts a positive effect on Dl nuclear uptake, that this effect is strongest when Dl levels are limiting, and that Cact[E10] modifies an essential process responsible for generating this positive effect. Therefore, we have uncovered a novel function for the Cact inhibitor to enhance Toll responses in the Drosophila embryo.

Lines used in this study were: loss-of-function cact[A2] and cact[011], generously provided by Steve Wasserman; and dl[6] and cact[4], obtained from the Bloomington Indiana Stock Center. Df(2L)cact[255] and Df(2L)III18 were used as cact deficiencies. Df(2L)cact[255] deletes both dl and cact, whereas Df(2L)III18 does not delete dl. As Df(2L)III18 is not viable in several allelic combinations Df(2L)cact[255] was used in most panels unless stated otherwise and is referred to as Df(cact). Maternal overexpression lines were CaM>cact-eGFP and CaM>cact[E10]-eGFP, which have been described previously (Fontenele et al., 2013). All embryos were collected from mothers of the respective genotypes crossed with wild-type Canton S males.

Bleach dechorionated 30 min- to 1 h 30 min-old embryos of the appropriate genotypes were homogenized in lysis buffer (1 embryo/µl) and prepared for SDS-PAGE as described previously (Fontenele et al., 2009). Endogenous Cact was detected with monoclonal antiserum from Developmental Studies Hybridoma Bank (DSHB, anti-Cact, 1:500). Anti-α-Tubulin was used as loading control (DM1α, 1:3000, Sigma). Western blot quantitative data were generated by measuring band intensity relative to tubulin, from direct luminescence using a ImageQuant LAS 4000 (GE Healthcare).

For nDl gradient visualization, mutant and control Histone-GFP embryos were mixed, fixed and processed concomitantly as described previously (Fontenele et al., 2013). Primary antisera used were monoclonal anti-Dl (7A4, 1:100, DSHB) and anti-GFP (NB600308, 1:1000, Novus Biologicals, to detect control gradients). Dl target genes were visualized by in situ hybridization as described previously (Fontenele et al., 2009).

Images of the nDl gradient and Dl target genes were collected from mid-stage 14 embryos, as defined by the amount of membrane invagination around nuclei. Dl gradient quantification was as described previously (Kanodia et al., 2009) using Matlab, collected at 85% egg length using a microfluidic device to orient embryos for end-on imaging (Chung et al., 2011). For upright imaging, a Nikon 60× Plan-Apo oil objective was used, and images were collected at the focal plane ∼90 µm from the embryo anterior pole. For the overall effect on Dl target genes, embryos were imaged laterally. All genotypes were processed and analyzed in parallel; thus, the same wild-type control is shown in graphs. Images were acquired with a Nikon A1 or a Leica LSM confocal microscope.

Student's t-test was performed for all experiments. Results are displayed as mean±s.e.m. The level of significance is shown in each figure (***P≤0.001, **P≤0.01, *P≤0.05).

We thank Trudi Schupbach and Attilio Pane for critical reading of the manuscript, and are greatly indebted to Siegfried Roth for making available essential information on cact mutant data. We are grateful to the anonymous reviewers for helpful suggestions to improve the manuscript. Monoclonal antibodies anti-Dl and anti-Cact, originally developed by Ruth Steward, were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA. Stocks obtained from the Bloomington Drosophila Stock Center (supported by NIH P40OD018537) were used in this study.