Salt inducible kinases as novel Notch interactors in the developing Drosophila retina

Developmental processes require strict regulation of proliferation, differentiation and patterning for the generation of final organ size. Aberrations in these fundamental events are critically important in tumorigenesis and cancer progression. Salt inducible kinases (Siks) are evolutionarily conserved genes involved in diverse biological processes, including salt sensing, metabolism, muscle, cartilage and bone formation, but their role in development remains largely unknown. Recent findings implicate Siks in mitotic control, and in both tumor suppression and progression. Using a tumor model in the Drosophila eye, we show that perturbation of Sik function exacerbates tumor-like tissue overgrowth and metastasis. Furthermore, we show that both Drosophila Sik genes, Sik2 and Sik3, function in eye development processes. We propose that an important target of Siks may be the Notch signaling pathway, as we demonstrate genetic interaction between Siks and Notch pathway members. Finally, we investigate Sik expression in the developing retina and show that Sik2 is expressed in all photoreceptors, basal to cell junctions, while Sik3 appears to be expressed specifically in R3/R4 cells in the developing eye. Combined, our data suggest that Sik genes are important for eye tissue specification and growth, and that their dysregulation may contribute to tumor formation.


Introduction
The generation of tissues and the establishment of cellular diversity in multicellular organisms relies upon cell-cell communication during development. One of the prominent and evolutionarily-conserved signaling pathways that regulates specification and patterning of tissues is the Notch signaling pathway. Notch has been established as a key regulator of development in both vertebrates and invertebrates (reviewed in [1]). Interestingly, many signaling pathways are involved in cell fate determination, including the Notch pathway, have been implicated in promoting or suppressing cancer when mutated or misregulated [2]. Furthermore, we show a genetic interaction between Siks and the Notch signaling pathway, and that increased or decreased activity of Sik promotes Notch-associated tumor-like growth of eye tissue. This suggests that Notch pathway members could represent direct or indirect downstream targets of Siks.

Generation of fly lines
The UAS-Sik3::T2A::mCherry construct was generated with the full-length cDNA of Sik3 isoform A. The pUAST-attB vector was modified to include the T2A::mCherry fusion protein. The T2A::mCherry plasmid was a kind gift from Dr. Stefan Fuss (Istanbul, Turkey). It was cloned into the pUAST-attB cloning site using the restriction enzymes XhoI and KpnI. The cDNA template FBcl0162999, corresponding to Sik3, was purchased from Drosophila Genomics Resource Center (DGRC). Sik3 cDNA was amplified by high fidelity DNA polymerase Advantage 2 (Clontech) and cloned into pUAST-attB-T2A::mCherry using the NotI and XhoI restriction sites, T2A and mCherry being N-terminal. Transgenic fly line was generated at Genetivision, Inc. (Houston, USA).

Immunohistochemistry
Whole mount of 0.5 μm-thick sectioned eye imaginal discs were dissected in cold PBS and fixed in 4% formaldehyde, at room temperature for 20 minutes. After 3 washes in PBX (0.3% TritonX-100 in 1X PBS) the discs were blocked for 1 hour at room temperature in BNT Buffer (10% donkey serum, 0.01% BSA, 250 mM NaCl, 0.1% Tween-20, PBS 1X). Eye imaginal discs were incubated with primary antibodies overnight at 4˚C. After 3 washes in PBX, the discs were incubated with secondary antibodies for 2 hours at room temperature, and after several washing steps, they were mounted in Vectashield © mounting medium.

Protein analyses
Crude protein was extracted from dissected fly tissues in following buffer: 50 mM NaCl, 50 mM Tris-Cl pH 7.5, 10% Glycerol, 320 mM Sucrose, 1% Triton-X100, Complete Protease Inhibitor Cocktail © Inhibitor (Roche). The tissues were homogenized and centrifuged for 10 minutes at 13,200 rpm, 4˚C. The crude protein was loaded on an 8% SDS-PAGE, after denaturation by DTT. Separated proteins were transferred to a PVDF membrane with a pore size of 0.45 μm. The membrane was blocked in 5% non-fat milk in TBS-T (20 mM Tris-Cl pH 7.6, 150 mM NaCl, 0.1% Tween-20) and blotted with primary antibody in 5% milk solution, overnight at 4˚C. Signals were detected by LumiGLO 1 Reagent (Cell Signaling).

Bioinformatic analyses
The protein sequences were downloaded from UniProtKB (http://www.uniprot.org/) and Protein similarity was calculated by EBI global protein alignment tool, Needle (EMBOSS) (http:// www.ebi.ac.uk/Tools/psa/emboss_needle/), by pairwise comparison of the protein sequences of interest.

Sik proteins regulate growth in a Drosophila tumor model
Our previous experiments in mammalian cell culture indicated that Sik proteins are involved in cancer progression [36]. Thus, we addressed whether they can have similar effects in vivo and employed "eyeful", a Drosophila Notch pathway-dependent eye tumor model. Overexpression of the Notch ligand Delta using the ey-GAL4 driver, which expresses a UAS target specifically in embryonic eye primordia and in the third instar larval stage in cells after the morphogenetic furrow (MF) (S1C Fig) [3], induces eye overgrowth and sensitizes the organ to tumorigenesis (Fig 1A) [4,54]. "Eyeful" flies, which overexpress Delta and concurrently carry the GS88A8 mutation (deregulation of the epigenetic silencers lola (CG12052) and psq (CG2368)), display further eye enlargement and tumorigenesis [2,54]. We tested Sik function both in the "sensitized" (ey>Delta) and "eyeful" (ey>Delta, GS88A8) backgrounds by crossing in different Sik transgenes.
All tested genetic combinations affected eye development to some extent. Affected eyes showed different phenotypes including tumor-like overgrowth, incidences of small ectopic eyes or full-size but split eyes, total loss of eye tissue, or metastasis that originated from the dissemination of transformed cells from the developing eye to the body (Fig 1A-1G). For each genetic combination, the ratio of affected eyes was quantified by analyzing and counting each (H) Sensitized (ey > Dl) and eyeful (ey > Dl, GS88A8) flies were crossed with the Sik transgenes: background, crossed with wild type w 1118 (-), Sik2 overexpression (>Sik2 OE), Sik2 knock-down (>Sik2 RNAi ), Sik2 constitutively active (>Sik2 S1032A ), Sik2 kinase-dead (>Sik2 K170M ), Sik3 overexpression (>Sik3 OE), Sik3 knockdown (>Sik3 RNAi ), and Sik3 constitutively active (>Sik3 S563A ). For each cross, normal eyes (similar to the sensitized background, like in (A)), and affected eyes (eye folding or overgrowth, split eyes or ectopic eye tissue, eye loss or antenna enlargement, and metastasis, like in (B-G)), in the progeny were quantified. The ratio of cumulative number of affected eyes over the total eye number is presented in the table H. The detailed analysis is shown in S1 eye independently. Eyes that resembled the parents were not considered as affected; any deviation from the most common baseline phenotype was considered to be affected. In the sensitized background, Sik2 overexpression (Sik2 OE) affected 23% of the eyes, Sik2 knock-down (Sik2 RNAi ) 12% of the eyes, Sik2 constitutively active (Sik2 S1032A ) affected 8% of the eyes, and Sik2 kinase-dead (Sik2 K170M ) affected 22% of the eyes. While Sik3 overexpression (Sik3 OE) did not cause any obvious eye phenotype when compared to the baseline, Sik3 knock-down (Sik3 RNAi ) affected 8% of eyes (Fig 1A-1C and 1H, S1 Table). The true eye tumor model (eyeful) results were difficult to interpret, since the genetic background alone already caused a high ratio of affected eyes; yet they were not contradictory to the ones obtained with the "sensitized" background ( Fig 1D-1F and 1H, S1 Table). Taken together, our results show that any modulation of Sik activity (both increases and decreases) enhanced both tumor-like phenotypes and eye loss phenotypes, possibly indicating a disturbance in eye development. Interestingly, the wild type overexpression of Sik2 (Sik2 OE) and misexpression of the kinase inactive form of Sik2 (Sik2 K170M ) led to similar phenotypes rather than opposite ones. This prevented us from classifying Sik2 as a tumor suppressor or oncogene. The observation that overexpression of wild type Sik3 (Sik3 OE) and knock-down of Sik3 (Sik3 RNAi ) had a milder effect than manipulating Sik2 would seem to suggest that Sik3 might not be involved in this process as much as Sik2; however constitutively active Sik3 (Sik3 S563A , which cannot be suppressed by cAMP signaling), caused the most pronounced phenotypes of all. Sik3 S563A was mostly lethal both in the sensitized and eyeful backgrounds, which might be due to ectopic expression of ey-GAL4 [55]. The few escapers that we were able to analyze displayed a high percentage of affected eyes (40% in the sensitized and 45% in eyeful backgrounds); also, severe tissue defects, including differentiation problems of eye and antenna growth ( Fig 1E) and rare instances of distal metastasis of eye tissue to the body ( Fig 1G) were detected. The observed eye loss phenotypes suggest a possible role for Sik in retinal fate determination.
Taken together, depletion or overexpression of Siks enhanced tumor-like growth in the eye in the sensitized and eyeful backgrounds. Constitutively active Sik3 may have additional effects, as can also be seen below (Fig 3).

Siks have a role in the developing eye
To understand if Siks are required for normal patterning of the eye disc, Sik activity was specifically altered via two strong drivers. In this experimental set-up, in addition to the ey-GAL4 driver described in the previous experiment, we introduced lGMR-GAL4 to drive expression in all PRs during larval stages and throughout adulthood [56], and UAS-Dicer2 to amplify the RNAi effect.
Modulating Sik protein levels with eye-specific drivers disrupted the structural integrity of the eye. Overexpression of wild type Sik2 (Sik2 OE) and Sik3 (Sik3 OE) or Sik2 point mutants (Sik2 K170M and Sik2 S1032A ), as well as Sik3 knock-down (Sik3 RNAi ), caused mild defects in the organization of ommatidia, cone cells, lens structure, and resulted in supernumerary bristles as shown by light microscopy and scanning electron microscopy (Fig 2A-2B 00 and 2E-2H 00 ).
Sik2 knock-down (Sik2 RNAi ) not only affected the cellular organization (intermediate knock-down, Fig 2D), but also drastically decreased eye size with full penetrance (strong knock-down, Fig 2C). Furthermore, it resulted in enlargement of the antenna in some instances, a phenotype also visible in eye-antennal imaginal discs (Fig 2C 00 ). This extreme phenotype could be attributed to off-target effects of the Sik2 RNAi line, which was shown to additionally downregulate Sik3 by a previous report [30]. The opposite phenotype of the Sik2 RNAi line, overgrowth, seen in the previous experiment (Fig 1) might be attributable to the fact that the eye tissue was already primed for growth by Delta overexpression. Overall, these results suggest that Sik2 and Sik3 activity may be necessary for proper eye development.
Ectopic expression of a kinase-dead form of Sik2 (Sik2 K170M ) or a constitutively active form of Sik2 (Sik2 S1032A ) not only resulted in cone and bristle cell defects, but also a ventral growth defect, a "notch" on the ventral side of the eye (Fig 2G-2H 00 , indicated with a green arrow) that is not present in controls (Fig 2A-2A 00 ). The phenotype obtained by intermediate knock-down of Sik2 resembles loss of the ventral half of the eye (Fig 2D), although this was not confirmed by loss of ventral eye markers at the larval level. These ventral eye phenotypes were seen with point mutation alleles of Sik2 but not with Sik3, suggesting that Sik2 can more potently affect ventral eye development than Sik3.
Ventral eye defects can be caused by misregulation of dorsal-ventral (D-V) patterning during development. It is long known that D-V polarity is established at early larval stages by the activation of the Notch receptor in the midline cells of the eye disc by Delta and Serrate from dorsal and ventral sides respectively [4] (S1B Fig). Our data indicate that disrupting the Sik balance results in phenotypes that resemble Notch pathway defects, suggesting a possible interaction of Siks with the Notch signaling pathway.

Siks interact with the Notch pathway
Our manipulations of Siks in the eyeful and sensitized backgrounds, and during normal eye development, indicate a possible interaction with the Notch pathway. To better understand this interaction, we used the driver ey-GAL4, lGMR-GAL4 with UAS-Dicer2 to overexpress or deplete Notch pathway components in combination with Sik transgenes.
We first tested the Notch ligand Delta, which is dorsally expressed in the late L2 stage to establish D-V polarity, and also expressed at late L3, immediately posterior to the MF, to determine the number of neurons specified from the epithelium (S1B and S1C Fig). As in the "sensitized" background ( Fig 1A), overexpression of Delta resulted in larger eyes ( Fig 3A). This phenotype was enhanced by adding the constitutively active (Sik2 S1032A ) or kinase-dead (Sik2 K170M ) forms of Sik2, which caused tissue folding and necrosis (Fig 3B and 3C) or in some cases, a total loss of the eye. Comparing these phenotypes to the Sik2 point mutation transgenes in the same background without coexpression of Delta, which produced a minor effect on eye size that is limited to the ventral eye ( Fig 2G and 2H), these results clearly indicate a genetic interaction between Sik2 and Delta. While Sik3 knock-down (Sik3 RNAi ) did not alter the phenotype (Fig 3D), constitutively active Sik3 (Sik3 S563A ), which caused excessive defects in the "sensitized" and "eyeful" backgrounds, resulted in phenotypes that are similar to those obtained with constitutively active and kinase-dead Sik2 lines, including both larger eyes with necrotic lenses and total eye loss ( Fig 3E). Sik2 knock-down (Sik2 RNAi ) in combination with Delta overexpression led to a phenotype very similar to those obtained with Sik2 point mutations; larger eyes and necrotic lenses (Fig 3F). Keeping in mind that Sik2 knock-down itself (Sik2 RNAi ) in the same background prominently blocked eye growth with full penetrance (Fig  2C), we can conclude that coexpression of Delta rescued the Sik2 knock-down phenotype. This experiment suggests that Delta acts downstream of Siks and thus, Siks might be regulating Delta or its main effectors.
Serrate is another Notch ligand, and is expressed in the ventral half of the eye imaginal disc. Depletion of Serrate using Serrate RNAi resulted in small eyes (Fig 3G). Sik2 overexpression (Sik2 OE) in a Serrate-depleted background enhanced the small eye phenotype and caused total loss of the eye in many instances (Fig 3H). Similar phenotypes were observed when Sik2 kinase-dead (Sik2 K170M ) or constitutively active Sik2 (Sik2 S1032A ) were used (Fig 3I and 3J), suggesting a genetic interaction between Sik2 and Serrate, which we confirmed using another  (Fig 3K) or Sik3 knock-down did not have any noticeable effect when combined with Serrate RNAi , just like in the Delta overexpression background ( Fig 3D). On the other hand, the Sik3 constitutively active form (Sik3 S563A ) enhanced the small eye phenotype and resulted in total loss of the eye (Fig 3L). The constitutively active Sik3 phenotype was exaggerated as in Figs 1 and 3E, which more closely resembled the phenotypes observed with Sik2 rather than wild type Sik3.
The glycosyltransferase Fringe is expressed in the ventral eye disc. It modifies Notch to promote Notch-Delta interaction, and suppresses interaction of Notch with Serrate except for at the dorsal-ventral border (S1B Fig). As previously reported, Fringe overexpression results in smaller eyes, presumably by disrupting the balance of Notch activation during development ( Fig 3M) [58]. The small eye phenotype of Fringe was not altered by Sik3 overexpression, but it was greatly enhanced when combined with the kinase-dead form of Sik2 (Sik2 K170M ) or the constitutively active version of Sik3 (Sik3 S563A ) (Fig 3N and 3O), indicating a genetic interaction between Siks and Fringe.
Taken together, these observations reveal a genetic interaction between Notch pathway members and Siks, in particular Sik2. Since Sik2 is a protein kinase, we hypothesized that it targets and phosphorylates members of the Notch pathway, or affects Notch signaling indirectly through the effectors of the Notch signaling.
In order to investigate if Notch pathway members could be directly targeted by Siks, we set out to find Sik phosphorylation motifs on Notch pathway members. First, we aimed to determine the amino acid sequence targeted by Siks. In addition to previously published Sik target motifs, recently identified targets were included in this study, to generate an updated target sequence. For this purpose, first, we aligned and compared experimentally-deduced Sik target sequences to generate an updated Sik target sequence (S4A Fig). Since only a few Sik targets are experimentally confirmed in Drosophila, we made use of proven targets from different organisms. This gave us a more generic, but a more conserved sequence of amino acids. SCARB1 and SREBP-1 were excluded from this analysis, since they do not contain any of the conserved residues. It is therefore possible that Siks have a second selection mechanism for phosphorylation. In particular, a 15 amino acid long region flanking the serine / threonine residue that is phosphorylated by Sik in different species was aligned to search for a target motif After determining the target motif, we scanned the Drosophila melanogaster proteome for proteins carrying this motif. In the whole fly proteome, this motif was identified in 307 annotated proteins, which include Notch (on serine 1066), Delta (on serine 740), and Fringe (on threonine 303), along with previously identified and confirmed targets (S4C Fig, S3 Table). However, the predicted residues are located in the non-cytosolic portions of these transmembrane proteins. Since Siks are known to function at nuclear or cytoplasmic compartments, it seems unlikely that these residues are phosphorylated by Sik proteins directly.

Siks are expressed in the developing retina
Our observations suggest a role for Siks in eye development, implying that Siks must be expressed in the developing retina. Thus, we set out to determine the localization of these proteins. Due to a high conservation of AMPK family members, most antibodies have the potential problem of cross-reactivity. To confirm their specificity, several commercially available antibodies were first tested in western blot (WB) analyses on crude fly extracts and immunofluorescence (IF) stainings on dissected larval and sectioned adult tissues. None of the commercially available α-SIK2 antibodies were able to specifically detect fly Sik2 in WB or IF. Sik3 has two known isoforms. The long isoform has previously been shown to interact with the Salvador (Sav) protein [30]. However, using the Abcam ab88495 α-human SIK3 antibody in WB, we observed predominantly the Sik3 short isoform. This antibody recognizes fly Sik3 at >75 kD (S2B Fig), slightly higher than the calculated size (77 kD), which was also confirmed by overexpression of Sik3. While we were able to show that Drosophila Sik3 is highly expressed in the developing larval nervous system in WB analyses (S2B Fig), no specific staining for Sik3 could be obtained in IF analyses using the same antibody.
Since none of the tested antibodies were specific enough to recognize and distinguish fly Sik2 and Sik3 in IF experiments, in order to observe endogenous protein localization, we generated BAC transgenic constructs in which Sik2 and Sik3 were tagged with fluorescent proteins. BAC clones spanning the entire coding sequence, UTRs and 14 and 6 kb upstream regulatory region of the Sik2 and Sik3 genes respectively were used; green fluorescent protein (Sik2) or mCherry (Sik3) sequences were inserted C-terminally, prior to the stop codon, using the P[acman] recombineering technique [41] (S2A Fig).
Eye imaginal discs of BAC transgenic flies were co-stained for cell-type specific markers and for GFP or mCherry to identify the localization of Sik2 and Sik3 proteins, respectively. Senseless (Sens) was used as a marker for PR R8 [59], Prospero (Pros) as a marker for R7 and cone cells [60], Spalt-major (Salm) as a marker for R3/R4 cells for the initial few rows posterior to the MF and R7/R8 for more posterior cells [61], Seven-up (Svp) was used as a marker for R3 and R4 (expressed stronger in R4 than in R3) and R1 and R6 [62], and Rough (ro) as a marker for PRs R2/R5 [59]. Additionally, N-Cadherin (N-Cad) and Armadillo (Arm) were used to mark the AJs. They are known to localize to the apical surface, between cone cells and PRs in pupal retinae [63,64].
In late 3 rd instar larval eye discs, a Sik2::eGFP signal was observed in the center of every ommatidium, starting 6-7 rows posterior to the MF (Fig 4A-4E). The localization does not appear to be nuclear, thus no co-localization was observed with Sens ( Fig 4A and 4A 0 ), Pros (Fig 4B), or Salm (Fig 4C), all central PR markers that are localized to the nucleus [59][60][61]. Sik2::eGFP was observed to be close to the N-Cad and Arm signals at the larval stage (Fig 4D  and 4E), though the Sik2 signal did not exactly co-localize with the junctional proteins. At 30% pupal development, the Sik2 signal in the PRs was localized basal to the AJ proteins Arm and N-Cad, and dispersed in the cytoplasm (Fig 4F and 4G).
In the L3 eye discs, mCherry tagged Sik3 appears to be nuclear. Co-staining of Sik3:: mCherry with Svp showed a co-localization with photoreceptors R3/R4 starting at row 3 ( Fig  4H), and not with Ro ( Fig 4I). Delta (Dl) was expressed both in Sik3-positive and Sik3-negative cells in the initial few rows of ommatidia (Fig 4J). In the anterior rows, Sik3 was clearly expressed in the anterior-most two PRs of every ommatidium, which represent the position of R3/R4 cells (Fig 4J 0 ). Posterior to this, Sik3 was expressed in most PR cells (Fig 4J).
In summary, Sik2 appears as a cytoplasmic, PR-specific protein, localizing basal to cell junctions; and Sik3 appears as an R3/R4-specific protein initially, that is later expressed panneurally during larval retina development.

Discussion
Precise regulation of signaling pathways and the interactions between them is required for proper tissue development. One of the fundamental aims of developmental biology is to identify the genetic interaction network that directs the development of a tissue. Although we are far from understanding the complete picture, dissecting the role of major interactors will get us closer to the answer. In this study, we focused on fly homologs of Sik2 and Sik3, their importance for development of the retina, and their involvement in tumor progression. Sik genes are members of the AMPK family that are conserved in animals and plants. They function as regulators of sodium sensing, cAMP signaling, metabolism, bone formation, immune cell maturation and neuroprotection (reviewed in [65]). Although some Sik functions like gluconeogenesis and lipogenesis have been thoroughly studied, knowledge about their involvement in neural development is quite limited.
Molecular analyses revealed that development and tumor progression actually share many aspects; there are fundamental molecular mechanisms common to both processes [2]. The Drosophila eye has been used as model to investigate developmental processes for decades. Lately, it has also been employed in tumorigenesis studies (reviewed in [66]). The eye undergoing metamorphosis is a great tool to study both the genetic interactions that control de novo tissue generation, and the molecular mechanisms that control cell division to establish proper tissue size. With this motivation, we tested Sik behavior in an insect tumor model.
Although well-known for its developmental role, Notch pathway activity was also shown to be disturbed in several cancer types, acting as tumor suppressor or oncogene in a contextdependent manner (reviewed in [67,68]). Our first experiment tested whether Siks could modulate the phenotype of a Notch-based tumor model in the Drosophila eye. We first tested in the context of gain of function of the Notch ligand Delta, which alone causes slight overgrowth and sensitization of the tissue to tumorigenesis, and then in the "eyeful" tumor model. While the contribution of Siks was less obvious in the stronger "eyeful" background, in the "sensitized" background changes in Sik2 levels clearly induce tumor-like structures (Fig 1, S1 Table). Paradoxically, our data show that both gain and loss of function of Siks result in similar phenotypes, preventing us from easily classifying Siks as inducers or inhibitors of growth.
The misregulation of Sik2 and Sik3 being detected in several cancer cases [39,48] at both genetic and epigenetic levels [69], and Sik3 being established as an ovary cancer marker [70] are consistent with our finding that Siks can promote tumorigenesis. In addition to the previous correlation studies, our experiment shows that Sik misregulation can be causative for tumor formation. Siks have been associated with certain cancer types, with both oncogenic and tumor suppressor roles, suggesting Sik involvement at a very fundamental level in cancerrelated mechanisms. This dual role is reminiscent of Notch-mediated neoplasias, where Notch acts either as an oncogene or tumor suppressor gene depending on the cellular context [71].
As demonstrated in Fig 1, Siks contribute to tumorigenesis and metastasis. It is known that reduction in E-Cadherin and destabilization of AJs is a key step for the epithelial-mesenchymal transition that is a prerequisite for metastasis [72,73], and any compromise in AJ stability is correlated with an invasive phenotype. The localization of Sik2 around the cell junctions ( Fig  4D-4G) appears to be in line with this finding. We have previously shown that Sik alters the invasiveness of tumor cells [36], and other reports propose Siks as anti-metastatic proteins [74,75] and show that loss of Sik function disables anoikis [76]. Combining the ability of Sik to contribute to invasiveness with their effect on survival, Siks are strong candidates for promoting cancer and anoikis.
The structural analysis by SEM of gain and loss of function alleles of Sik2 and Sik3 suggest that Sik proteins may play a role in the proper development of the eye. We observed that misregulation of Siks resulted in defects, however no obvious phenotypes were observed in mutants of Sik2 or Sik3 (S3 Fig). This discrepancy in our results might be attributed to functional compensation mechanisms that are employed in the absence of one of the Sik homologs, either with one another, or with other members of the AMPK family. The severe defects observed with the Sik2 RNAi line (Fig 2C and 2D), which is known to downregulate Sik3 as well as Sik2 [30], supports this idea, as loss of both homologs is expected to have a more severe phenotype than loss of either one alone.
It is intriguing that overexpression of a kinase-dead form of Sik2 (Sik2 K170M ) perturbed eye development and promoted tumorigenesis in a similar way to the wild type Sik2 protein (Figs 1, 2 and 3). This implies that the phenotypes are not wholly the result of aberrant phosphorylation of direct Sik targets. Sik2 K170M could presumably retain molecular interactions with its targets, without phosphorylating them, and possibly with other interactors such as scaffold proteins. This might result in an unforeseen dominant negative effect, effectively creating another Sik2 loss of function condition in addition to knock-out (full mutant) and knock-down (RNAi). The constitutively active transgenes, which are not subject to PKA suppression, represent gain of function alleles of Siks that would be expected to give stronger phenotypes than the wild type overexpression lines. Constitutively active Sik2 (Sik2 S1032A ) expressed in the eye gives similar results to wild type overexpression (Fig 2), lending further support to the hypothesis that these phenotypes are not wholly dependent on the kinase activity of Sik2.
Gain and loss of function of Sik3 via overexpression and knock-down resulted in defects in lens and bristle cells that were similar to each other, and to the Sik2 manipulations (Fig 2E-2F 00 ), perhaps indicating that Sik3 also has a role in eye development. Another surprise was the different effect of Sik3 knock-down by RNAi (disorganization of ommatidial packing and some supernumerary bristles) compared to Sik3 null mutant clones. This might be explained by the Sik3 RNAi line having off-target effects and affecting other homologs.
The effect of overexpressing constitutively active Sik3 (Sik3 S563A ) was stronger than the overexpression of wild type Sik3 in multiple backgrounds (Figs 1H and 3E, 3L and 3O). The difference was much more pronounced than the difference between wild type and constitutively active Sik2. One possible explanation is that PKA phosphorylation is more important for normal regulation of Sik3 activity than Sik2, which might be restrained by additional mechanisms. Alternatively, we may have disrupted the balance of AMPK family members by overexpressing the constitutively active Sik3 protein.
Overexpression of Sik2 K170M or Sik2 S1032A did not prevent the generation of eye tissue, but did induce a growth defect in the ventral eye (Fig 2G-2G 00 and 2H-2H 00 ). This highly resembles Notch pathway phenotypes [77] (reviewed in [78]). The observation of flies that appear to be missing the entire ventral eye following intermediate knock-down of Sik2 (Fig 2D) further suggests that Siks, particularly Sik2, are able to influence dorsoventral polarization, or growth or survival of ventral cells, in the developing eye tissue in a Notch-dependent manner. Our results imply that Sik can affect the activity of the Notch pathway, possibly by interacting with Notch pathway components or their effectors. However, the exact contribution of Siks in this process still needs to be established.
Having observed that Siks modify the Delta overexpressing background, and that manipulation of Siks can affect eye development in ways reminiscent of Notch pathway defects, we sought further evidence of genetic interactions by co-overexpressing Siks with Notch pathway components in the eye (Fig 3). We found that manipulation of Sik2 levels exaggerate the small eye phenotype of Serrate knock-down (Fig 3G-3J) and of Fringe overexpression (Fig 3M and  3N), although they do not affect eye size on their own (Fig 2B-2B 00 and 2G-2H 00 ). Unlike Fringe overexpression and Serrate knock-down, Delta overexpression increases eye size, since it is not only required for establishment of D-V polarity at the L2 stage, but also promotes neurogenesis behind the MF in the L3 stage [79]. Changes in Sik2 levels also exaggerate or modify this phenotype (Fig 3A-3C and 3F). Conversely, while Sik2 knock-down strongly suppresses eye growth (Fig 2C), Delta overexpression can rescue this phenotype (Fig 3F), suggesting that Delta acts downstream of Sik2. Therefore, although we have not verified a direct interaction between Siks and Notch pathway members, our data clearly show that they are epistatic to each other.
After confirming a genetic interaction between Siks and Notch, we asked whether Notch pathway members might be direct targets of Sik phosphorylation. By comparing the sequences of experimentally proven Sik targets, we identified a consensus phosphorylation motif (S4A and S4B Fig). Analysis of the fly proteome showed that >300 reviewed protein entries contain this deduced motif; the list includes Notch, Delta and Fringe from the Notch pathway [80]. Notch and its ligands are known to be regulated by phosphorylation of the intracellular parts of the proteins (reviewed in [81]). However, the motifs predicted to be phosphorylated by Siks are located in the extracellular regions of Notch and Delta. In the case of Fringe, the predicted Sik target motif would be located in the Golgi lumen. Since the Sik proteins function in the cytosol, these sequences cannot represent true Sik targets. We thus hypothesize that Siks might be modulating the Notch signaling pathway indirectly, maybe by phosphorylating a regulator of Notch or its ligands, or a major downstream effector.
Scanning the fly proteome revealed experimentally-confirmed Sik targets like Salvador, showing the validity of our screening. Novel predicted targets include several DNA and RNA binding proteins, transcription coactivators, proteins important for the RNAi machinery, epigenetic regulators, some proteins important in cell cycle and cancer, proteins related to cytoskeletal organization, proteins functioning in signal transduction, some proteins functioning in phototransduction, gustatory and odorant receptors, several ubiquitin-protein ligases, some proteins important in lipid synthesis, many mitochondrial proteins, and proteins with metallopeptidase activity that are important for migration (S3 Table). Our data hint that several proteins related to animal development and retina morphogenesis that could be Sik targets, including UVRAG from the Notch signaling pathway, Pannier, Slingshot, tiptop, hedgehog, hopscotch, Calpain-D, Neurobeachin, fork head, prospero, FGFR2, flamingo, elav, Dscam2, Patj, scribble, arrowhead and domeless. It will be interesting to explore some of these hits further in future experiments. Pannier is particularly interesting, since it is responsible for establishing the dorsal eye fate, and we show that Sik disruption leads to a ventral eye defect. Clearly, the predicted putative phosphorylation sites (S3 Table) need biochemical confirmation.
Recent findings point to a potential role of Sik kinases in animal development, mostly in the skeletal tissue [82]. We aimed to investigate their role in development, but were hampered by the lack of specific fly antibodies. To overcome this, we generated novel useful tools. Using the BAC recombineering technique, we created translational fusion proteins of Sik2 and Sik3 with fluorescent protein tags (S2A Fig). With the only working α-SIK3 antibody, our WB analyses showed that Sik3 was highly expressed in the larval nervous system (brain, ventral nerve cord and eye-antennal imaginal discs) (S2B Fig), unlike the ubiquitous expression shown previously [83]. In contrast to a previous report [30], we mainly detected the shorter isoform of Sik3 (S2B Fig) and thus used the Sik3-RA transcript to generate our fluorescently tagged genomic constructs (S2C Fig). IF analyses of developing retinae from the Sik fusion protein-expressing animals showed that the Siks are expressed in a specific pattern, for a limited period, during eye development. In our analyses, Sik3 expression was detected in the nucleus, only in late L3 stage eye discs, in photoreceptors R3/R4 for the first few rows after the MF, and in many other neural cells in the more posterior rows. This stepwise expression of Sik3 in PRs suggests a potential role of Sik3 in PR specification, especially in R3 and R4 (Fig 4H, 4J and 4J 0 ). We noted some overlap between Sik3-expressing cells and Delta-expressing cells at this stage, which sets the stage for a genetic interaction.
ptIn late L3 eye discs, Sik2::eGFP expression could be detected in center of each ommatidial cluster, starting 6-7 rows posterior to the MF, localized in the cytoplasm of photoreceptors, adjacent to the AJs in larval stage, and basal to the AJs at the pupal stage (Fig 4D-4G). This localization can be particularly meaningful. A previous report showed that AJs can be regulated by Sik3 and Lkb-1, the major regulator of Siks. It has been reported that, in mutants of Lkb1 and Sik3, AJs are defective, and Armadillo is mislocalized [84]. The detection of Sik2::eGFP not directly at the AJs but adjacent to AJs suggests the possibility that Sik2 may also have a developmental function involving regulation of cell-cell junctions, at early stages. The cytoplasmic rather than nuclear localization of Sik2 suggests that its nuclear targets such as HDACs [21] and CRTCs [9] may not be relevant in the eye at the times we examined.
Although Sik expression in earlier periods could not be assessed due to low levels of signal, we cannot exclude the possibility that they are expressed at an earlier stage like L2. So, the genetic interaction between Siks and Notch pathway members, or effectors, could be happening in L2, when D-V polarization and eye morphology are established. This could also explain how both Sik2 and Sik3 are able to modulate the Serrate and Fringe eye size phenotypes.
Our tagged Sik2 and Sik3 proteins were not expressed in the same subcellular compartments, or in the same cells at the same time. While this seems not to support the idea of redundancy between them, it is possible for tissues to adapt to the loss of one protein by upregulation of another. Compensatory mechanisms between Sik homologs and a molecular pathway involving feedback loops could explain the similarity in phenotypes observed for Sik2 and Sik3, despite their expression patterns being different during eye development (Fig 4). Also, we cannot eliminate the risk that the BAC clones used here do not contain all the enhancers necessary to recapitulate the endogenous expression patterns of the Sik proteins. Thus, the Sik expression patterns described here using tagged BAC transgenic lines would benefit from verification, if antibodies that work in immunofluorescent stainings can be developed.
Taken together, this study shows that fly Sik2 and Sik3 can promote tumor formation, that proper regulation of Sik2 and Sik3 is required for normal eye development, and that Siks functionally interact with the Notch pathway. Further studies will be necessary to uncover the molecular basis of this interaction, and to discover at which stages during development their interaction with the Notch signaling pathway takes place. The protein organization and genetic constructs of Drosophila salt inducible kinases. Sik2 and Sik3 protein kinase domains are shown in red, the UBA domains in green, and the key residues for suppression of Siks by PKA (S1032 and S563, respectively) are shown in blue. Sik2::eGFP and Sik3::mCherry are translational fusions, generated on BAC clones, where fluorescent proteins are added prior to the stop codon, after a flexible linker. Both clones comprise all the introns, exons, UTRs, and sufficient regulatory regions (14 kb and 6 kb for Sik2 and Sik3 respectively) to reflect the endogenous expression. UAS-Sik3::T2A::mCherry was generated using the EST clone encoding Sik3 CDS isoform A. mCherry was attached after a self-cleaving T2A linker. Sik3 Δ109 allele was generated by excision of P element P{EPgy2}CG42856 EY14354 . Mobilization created a deletion of 9.7 kb from exon 2-10. CG15071 gene in the intron 2 was excised, CG42855, which overlaps with the 5' of Sik3, remained intact. (B) Western blot on the wild type (w 1118 ) strain crude protein extract from adult head, adult body, larval CNS (brain, ventral nerve cord, and eye-antennal imaginal discs) and larval body. >75 kD band was revealed with the α-human SIK3 antibody (estimated Sik3 size~77 kD). Actin was used as loading control. (C) Western blot of fly head extract, from control (w 1118 ) and ectopic expression of Sik3 by eye-specific drivers (ey-GAL4, lGMR-GAL4 / +; UAS-Sik3::T2A::mCherry / +), revealed with α-DsRed antibody. The majority of the mCherry was already cleaved from the fusion protein, as seen at >25 kD, and a small portion was still attached to Sik3, as seen at >100 kD (estimated size~27 kD and 115 kD respectively).  Table. Eye phenotypes in sensitized and eyeful background. The eye phenotype quantification (A-A 0 ) in the "sensitized" (Delta overexpression) and (B-B 0 ) in the "eyeful" backgrounds (Delta overexpression in combination with GS88A8 epigenetic regulator mutation). Percentages for different backgrounds are (A,B) listed in the table and (A 0 ,B 0 ) shown in the histogram. The baseline eyes are similar to the sensitized parents', which is slightly bigger than the wild type flies. The affected eyes are subclassified as fold / overgrowth (bigger eyes with at least one fold, or overgrowth of the eye), ectopic-eyes (ectopic eye tissue on the head surface or spliteyes on one side of the head), eye loss (total loss of the eye tissue), and metastasis (ectopic eye tissue in the body or in the head, which is not exposed on the surface). The incidences of normal eyes and affected eyes were quantified, and the ratio over the total number of eyes is presented in the table. (A,B) The total ratio of affected eyes is shown in the last column for each background, together with the standard error of the mean obtained from 3 independent trials. The genotypes are as in Fig 1. The full genotypes are listed in S2 Table. (PDF) S2