acal is a Long Non-coding RNA in JNK Signaling in Epithelial Shape Changes during Drosophila Dorsal Closure

Dorsal closure is an epithelial remodeling process taking place during Drosophila embryogenesis. JNK signaling coordinates dorsal closure. We identify and characterize acal as a novel negative dorsal closure regulator. acal represents a new level of JNK regulation. The acal locus codes for a conserved, long, non-coding, nuclear RNA. Long non-coding RNAs are an abundant and diverse class of gene regulators. Mutations in acal are lethal. acal mRNA expression is dynamic and is processed into a collection of 50 to 120 bp fragments. We show that acal lies downstream of raw, a pioneer protein, helping explain part of raw functions, and interacts genetically with Polycomb. acal functions in trans regulating mRNA expression of two genes involved in JNK signaling and dorsal closure: Connector of kinase to AP1 (Cka) and anterior open (aop). Cka is a conserved scaffold protein that brings together JNK and Jun, and aop is a transcription factor. Misregulation of Cka and aop can account for dorsal closure phenotypes in acal mutants.


Introduction
A large fraction of the eukaryotic genome codes for non-coding RNAs (ncRNAs), which are very abundant and diverse, yet mostly uncharacterized and of unknown functions [1,2]. Nevertheless, some are critical players in gene expression regulation. Non-coding RNAs encompass different classes of molecules. The best-characterized ncRNAs are small RNAs. Broadly, small RNAs recruit silencing machinery to mRNAs to inhibit translation and/or target them for degradation [3]. Small ncRNAs characterization was facilitated because they base-pair with cognate mRNA targets, and interact with common protein complexes to regulate gene expression [4].
In contrast, long non coding RNAs (lncRNAs) encode a diverse group of ncRNAs, which may be part of different molecular complexes, and can stimulate or inhibit gene expression [5]. Recognized lncRNAs range from 200 nucleotides to several kilobases [6]. Since they do not have evident sequence motifs for annotation, their identification and relevance has remained elusive. In addition, many have low expression levels, hindering isolation and characterization. Next-generation RNA sequencing uncovered many lncRNAs (over 50% of transcribed species [2,7]), increasing dramatically their number and repertoire [8]. Despite this, in most cases contribution of lncRNAs to gene expression regulation or other still awaits genetic and functional validation.
In Drosophila melanogaster some lncRNAs have been annotated [9], and play important roles in gene expression, as in vertebrates [10]. In spite of extensive genetic screens, forward genetics identification of lncRNAs has been very limited, as they are found to fine-tune gene expression with mild phenotypical contributions.
Here we characterize a Drosophila lncRNA with strong embryonic phenotypes and lethality. Mutations in this locus, acal, result in partially penetrant dorsal closure (DC) defects due to Jun N-terminal kinase (JNK) signaling over-activation. This results in failure of DC, leading to a lethal dorsal open phenotype.
During DC, lateral epidermal sheets stretch over a dorsal extra-embryonic cell layer, the amnioserosa, and fuse at the dorsal midline [11]. The JNK signaling pathway regulates DC, and is activated at the dorsalmost row of epidermal cells, the leading edge cells (LE), which act as a signaling center for DC ( [12], reviewed in [13]).
In these epidermal cells, JNK activation induces expression of cytoskeleton and adhesion regulators for cell stretching [14]. JNK activation at the LE spreads the morphogenetic rearrangement by inducing the signaling ligand decapentaplegic (dpp), a fly homolog of vertebrate BMPs. Dpp signals to lateral epidermal cells and amnioserosa to promote cell shape changes [15,16]. This response induces stretching of lateral epithelia in absence of cell proliferation, and final zippering.
It is not known how JNK activation is triggered at the LE, nor completely known how LE restriction occurs. The JNK signaling pathway is an evolutionarily conserved, MAPK-type signaling pathway. It consists, at its core, of a cascade of kinases, from JN4K (the gene misshapen in flies), to JN3K (slipper), JN2K (hemipterous), and finally JNK proper (basket). Activated JNKbound to a scaffold protein called Connector of kinase to AP1 (Cka)-, phosphorylates the transcription factor Jra (Drosophila Jun). Jra, together with Kayak (Drosophila Fos), constitute the AP-1 transcription complex activating JNK target genes. Besides Jra, JNK also phosphorylates Anterior open (Aop), leading to Aop nuclear export, de-repressing JNK target genes. The JNK pathway is required for orchestration of embryonic dorsal closure, but also for wound healing and in response to certain stressful conditions [13].
During DC JNK activity is restrained from the lateral epidermis partly by aop, although this transcription factor also functions earlier in the tissue in a positive way by promoting epidermis differentiation, preventing ectopic mitoses [15]. The 'raw-group genes': raw, a pioneer protein, and ribbon, a transcription factor (rib, [17,18]) also restrict JNK activation in the lateral epithelia. Another member, puckered (puc), coding for a JNK phosphatase, acts in a feedback loop in the LE, dephosphorylating active JNK, and stopping signaling. JNK activation is also antagonized in the amnioserosa by pebbled (peb), which codes for a transcription factor [19].
We show here that acal lncRNA partly mediates Raw JNK signaling antagonism in the lateral epidermis. acal partakes in regulating expression of the scaffold protein Cka [20]. This explains partly raw function (raw as an acal regulator), and provides a framework for JNK activity down-regulation at the lateral epidermis during DC. We find that mutations in acal also alter the expression of aop, balancing JNK activation. acal also shows a genetic interaction with Polycomb, suggesting a relationship between acal and the Polycomb repressive complex.
Overall, this provides a rationale for acal DC phenotypes.

acal is a novel 'dorsal open' gene
Mutations in fly JNK signaling genes-like bsk, the fly JNK gene [12], Cka [20], and aop [15]lead to an embryonic lethal condition with a dorsal hole in the cuticle (Fig. 1B-D). This 'dorsal open' phenotype has been used to identify JNK and other DC genes [13].
We identified a lethal locus where a fraction of mutant embryos die with dorsal closure defects. We called it acal, or 'boat' in Náhuatl, due to the cuticular appearance of mutants. Two alleles, acal 1 and acal 2 (Fig. 1E-F), are lethal excisions of P{KG09113}, a viable and fertile Pelement transposon insertion at 47A13 (Fig. 1M''). Three more EMS alleles were isolated over acal 1 and acal 2 by lack of complementation (acal 3 , acal 4 , and acal 5 , Fig. 1G-J). A sixth lethal mutant allele is a mapped P element insertion in the locus, P{GS325ND3} (Fig. 1K-L and M''). All mutations are lethal, with embryonic, larval, and pupal phenocritical periods. No adults are ever observed. A fraction of mutant embryos have DC defects, with holes in either the dorsal or anterior ends of embryos (Figs. 1E-L and S1A). Dorsal or anterior holes are the only cuticular defects seen in these mutants.
acal mutants have an extended phenocritical period, since a fraction of mutant embryos die without cuticular defects, and finally another fraction of mutant embryos survive embryogenesis and die later during larval stages up to the beginning of pupariation. This means that acal is required at various times during development. Larval lethality varies from 10 to 50% depending on the allele studied, and pupal death from 3 to 40%. As stated, no adult homozygous mutant flies are ever recovered, except 1-2% adult acal 6 escapers. In this work, we focus on the embryonic mutant phenotypes. Using these embryonic phenotypes, mutant alleles conform to an allelic series with acal 5 as the strongest, and acal 1 as the weakest (S1A Fig.).
acal maps to an intergenic unannotated region and to SD08925 We used deficiencies uncovering P{KG09113} to map acal in complementation tests, and established a 100 kb interval where acal maps. This interval includes three annotated genes: longitudinals lacking (lola), pipsqueak (psq), and CR45135 (Fig. 1M). acal mutations complement lola and psq alleles (S1 Table). Besides, lola and psq have different embryonic mutant phenotypes from acal, with no cuticular phenotypes, like dorsal or anterior holes ( [21,22] and S7A Fig.). CR45135 is not characterized, but partially overlaps a lola exon, so CR45135 mutations should affect lola. Moreover, many lola lethal insertions overlap this gene [23].
A previous report showed that there is at least another locus whose mutation complements lola and psq, despite being unannotated [24]. This mutation, l(2)00297, suppresses the peb 1 heat-shock sensitive rough eye phenotype. acal mutant alleles also suppress, in our experimental conditions, this subtle but clear phenotype (S1B Fig.). This suggests that acal and l(2)00297 are allelic, and is independent evidence of a lethal locus between lola and psq.
Another paper independently reports the existence of a long non-coding transcript approximately 4 kb from the start of lola [21]. The size of this transcript is stated as 4 kb, but this was an approximation, as the evidence supports a maximum size of 3.4 Kb, since this cDNA hybridized to a 2.8 and a 0.6 Kb neighboring restriction fragments in Southern blots ( [21], and Edward Giniger, personal communication).
An EST in Flybase, SD08925, sequenced partially at the 5' end, maps 3.8 kb away from lola, precisely where Giniger et al. report the location of their 'alpha' long non-coding transcript [21]. We believe this transcript is the same as SD08925. We sequenced fully this EST (Genbank accession number # KJ598082). SD08925 is 2.3 kb long, poly-adenylated, and a single exon. Its genomic 3' end bears an A-rich stretch ( Fig. 1M''), characteristic of poly-adenylation sites [25]. We also obtained a 2.3 kb weak Northern blot signal (compared to control Rp49), demonstrating SD08925 expression, and confirming that SD08925 is a full-length cDNA. This band was detected throughout the life cycle and was nearly absent in the Δ18 deficiency strain uncovering the locus by semi-quantitative RT-PCR (S1C-D Fig.). Consistent with the weak Northern signal, RNA-seq and microarray data show low SD08925 expression levels [26,27].
We sequenced 4.7 kb of the SD08925 genomic region in acal mutant and control lines. In these and following experiments, we selected and separated mutant embryos from heterozygous controls by lack of embryonic GFP-expression due to a balancer chromosome (see Materials and Methods). Within this region, we found a 20 base pair insertion in acal 1 . The acal 1 lesion is 1031 bp away from P{KG09113}, at +931 of the cDNA, a position conserved in closely related species. Similarly, we found a 2 bp deletion 1630 bp away from P{KG09113} within the 5' end of SD08925 in acal 2 , in a conserved region. In acal 5 SD08925 expression is significantly reduced (Fig. 1M'', Fig. 2A and D and 3A).
In acal 1 and acal 2 , no trace of the original P element was found. We reasoned that both lesions, found 1-2 Kb away from the original insertion site (which is in itself viable and fertile), are likely consequences of P-element imprecise excisions and repair, due to nicking and repair of the DNA break, leading to lethality, and responsible for the acal 1 and acal 2 mutant phenotypes, respectively. In both cases, the changes are not present in any of the other strains (genome reference assembly, yw wild type control, parental strain, or other acal mutants). acal 6 is a P-element insertion mapping to the locus, 3 Kb 5' from the transcription start site. We found no differences in the sequence in the 4.7 kb genomic region sequenced centered around  Compared with acal 5/5 mutants, the cuticular phenotypes are the same, but they differ significantly in the abundance (expression of acal significantly reduces the number of mutant embryos that die and that have cuticular phenotypes). In these and following experiments, cuticular phenotypes do not change, unless otherwise stated. No new cuticular phenotypes are found; thus, examples are not depicted in all figures. Numbers analyzed: acal 5/5 ,UAS-acal/+ = 240, acal 5/5 ;69B-gal4/+ = 441, acal 5/5 ;69B>acal = 237, acal 5/5 ;pnr-gal4/+ = 435, acal 5/5 ;pnr>acal = 130. Statistical significance was calculated using chi square the SD08925 transcribed region for acal 3 , acal 4 , and acal 5 . The sequenced genomic region includes the whole transcript, plus some neighboring 5' and 3' sequences ( Fig. 1M''), suggesting that mutations in acal 3 , acal 4 , and acal 5 lie outside the transcript. This is consistent with a regulatory nature for these alleles, rather than molecular lesions within the transcribed region, like acal 6 .
We performed a genomic rescue experiment (Fig. 1M'-N), rescuing fully the embryonic lethality of acal 1 and acal 2 mutants with a *20 kb genomic construct called 178D09. 178D09 spans a small portion of the 5' region of lola, wholly the SD08925 transcript region, plus *7 kb upstream of SD08925 in the intergenic region between SD08925 and psq (Fig. 1M'). Besides the 5'-most exon of lola, which is part of the untranslated 5' leader sequence of some lola splice variants and CR45135, no other transcript is included, except SD08925. SD08925 expression levels are not changed in rescued embryos versus controls, presumably due to endogenous regulatory sequences maintaining overall low expression levels (S1E Fig.). SD08925 is conserved in eleven Drosophila species surveyed, and recognizable in A. gambiae ( Fig. 1O; 3A). Together, these data are consistent with a low-expression locus involved in DC.
We quantified SD08925 expression during DC in acal 2 and acal 5 mutant embryos by qRT-PCR. acal 2 expression is similar to wild type, but, as stated, acal 5 expression is significantly reduced ( Fig. 2A). This is consistent with acal 5 being a regulatory mutant, although other interpretations are possible. To ensure consistency, whenever possible we performed experiments with multiple acal mutant alleles, with similar results. These data are also consistent with SD08925 coding for acal. In contrast, using primers that detect all lola isoforms, or all psq isoforms (both loci have alternative splicing) no significant differences in expression levels are seen in acal 1 , acal 2 , acal 5 , and acal 6 (S7C Fig.).
In order to more rigorously test whether SD08925 is acal, we generated a UAS transgene containing only the complete SD08925 cDNA (2.3 kb). We expressed this construct in acal 5 mutants, using the 69B ectodermal driver. This resulted in significant rescue of acal 5 DC defects and reduction of embryonic lethality ( Fig. 2B-B'''). 69B over-expression of UAS-acal in acal 5 mutants significantly increased SD08925 expression (40%, S2A Fig.). We also show that pnr MD237 gal4 (pnr-gal4) driven expression of SD08925 rescues DC defects in acal 5 (Fig. 2B). Regardless of the nature of acal 5 , augmenting SD08925 expression in the mutant significantly rescues the embryonic phenotypes, arguing that reduction of acal expression is responsible for the acal 5 DC and embryonic lethality phenotypes. Taken together, (1) the complete lack of complementation to each other and similar mutant phenotypes of all six acal alleles (but complementation to and dissimilar to mutant phenotypes of the neighboring lola and psq loci), (2) the molecular mapping of acal 1 , acal 2 , and acal 6 to the SD08925 locus, (3) the significant genomic and UAS-acal rescues for acal 1 , acal 2 , and acal 5 , and (4) the significantly reduced SD08925 expression in acal 5 both by qRT-PCR and in situ hybridization (see below), compared to no differences in lola and psq expression levels in acal 1 , acal 2 , acal 5 , and acal 6 establish the SD08925 locus as the acal locus.

acal embryonic expression
In situ hybridization with SD08925 in wild type embryos revealed a dynamic expression pattern, mainly in two tissues: the lateral epidermis and the central nervous system. At early stages, the transcript is detectable throughout the embryo but falls at gastrulation (Fig. 2D). From germband retraction to DC, expression is seen in germband derivatives, mainly the lateral epidermis and nervous system (Fig. 2D). This is consistent with a role in DC and the UAS-acal rescue in the ectoderm and/or lateral epidermis (Fig. 2B). Later, expression appears in the mesoderm, but the bulk of expression is in the condensing nervous system and closing lateral epithelia. At the end of embryogenesis, expression occurs in the nervous system (Fig. 2D).
In acal 5 mutant embryos, expression is significantly reduced in the lateral epithelium during DC, consistent with acal 5 being a regulatory mutant and with the UAS rescue ( Fig. 2B-D). Nervous system expression is not altered in acal 5 during DC stages, but decreases significantly later (S2B-C Fig.).
acal is a processed, long non-coding RNA Is SD08925, from now on referred to as the acal transcript, translated? The acal transcript is 2.3 kb long and could potentially code for protein(s). The locus is also evolutionarily conserved in other dipteran species (Fig. 3A), but not outside Diptera.
We find that in the dipteran species examined, the locus is devoid of conserved open reading frames (ORFs), sporting only variable, non-conserved ORFs ranging from 33 up to 243 nucleotides in size (S3A Fig.), even though other regions of the transcript are conserved ( Fig. 3A; S3B Fig.). Unlike short ORF-bearing transcripts like polished rice (pri, [28]), the putative short ORFs in SD08925 and homologues have no common motifs. The deduced protein sequences of these ORFs in D. melanogaster show no homology with annotated proteins, and are not present in Drosophila proteomics databases (see Materials and Methods). Using an algorithm that measures coding potential [29], based on six different criteria [(1) feasibility of ORFs, (2) coverage of ORFs within transcript, (3) presence of in-frame initiation and stop codons, (4) number of hits in BLASTX, (5) high quality in any given BLASTX hit, and (6) whether any given hit is in frame with the predicted ORFs], we determined that acal has an extremely low coding potential, similar to other characterized long non-coding RNAs (Fig. 3B).
Genes harboring short ORFs have been erroneously classified as long non-coding RNAs, and later found to code for small peptides [28,30,31]. Does acal code for short, translated ORFs? Genes coding for short ORFs still have a protein coding score, unlike acal (Fig. 3B), and are exported to the cytoplasm for translation. In contrast, many non-coding RNAs reside in the nucleus. We detected acal transcripts enrichment in a nuclear fraction, different from Rp49, a protein-coding mRNA ( Fig. 3C-D). As control, we also detected a small enrichment in the nuclear fraction of pre-bantam (ban), a well-known miRNA precursor. The enrichment is . acal is a processed long non-coding RNA. (A) acal full-length transcript sequence conservation plot compared to other dipterans, and pairwise alignments between acal in D. melanogaster and its homologs (adapted from the UCSC genome browser). Location of molecular lesions in acal 1 (1) and acal 2 (2) are marked within brackets after an 'X' in the transcript (gray arrow). Above the plot is the location of the probes used for small RNA Northern blots in (E) and in S3 Fig small, likely because it is quickly processed to mature ban. acal nuclear enrichment also suggests a non-coding nature.
High-throughput RNA sequencing projects have deeply annotated small RNAs in the Drosophila genome [32]. Within acal, we found two groups of sequence reads reported to appear under various experimental conditions that did not qualify as functional small RNAs. We designated these groups of reads as acal-A and acal-C (Fig. 3A), and performed small RNA Northern blotting, reasoning these could be the mature products of the acal transcript. With this Northern protocol, larger RNA fragments are not detected, but small RNA species are well separated. Using the acal-A probe, we found evidence of acal fragmentation throughout the life cycle of the fly, particularly during pupal stages ( Fig. 3E; S3C). Instead of a roughly 22-nucleotide fragment, we found a group of small bands from about 50 to 118 nucleotides (Fig. 3E). These bands are not detectable in regular agarose gels due to their size. In contrast, acal-C did not reveal evidence of processing (S3C Fig.). We used two additional probes, acal-B and acal-D, to look for evidence of further processing. acal-D revealed no processing (S3C Fig.), but acal-B revealed at least a 59 nucleotides-long fragment (Fig. 3E). Both acal-A and acal-B fall within conserved positions of the gene at the 3' end (Fig. 3A); acal-B lies within a 60 nucleotide-long conserved region, a size coincidental with the band we found in small RNA Northern blots (for detail, see S3B Fig.). In conclusion, we have evidence that the 3' half of the acal transcript undergoes processing, whereas the 5' half does not. Unfortunately, as the signal we detected was mostly from pupae, we could not test mutants, because mutants die as embryos and larvae, and never reach pupation proper. Taken together, we propose acal is an unconventional long non-coding RNA partially processed into small fragments.
acal down-regulates JNK signaling during DC We next studied JNK signaling, the DC trigger, to characterize DC defects in acal. We tracked and quantitated expression of a DsRed reporter under control of an AP-1 response element, TRE (Tetradecanoylphorbol acetate Response Element, see Fig. 4A and [33]). In parallel, we tagged cellular membranes using sGMCA, a fusion protein of the actin-binding domain of Moesin and GFP [34]. These constructs were expressed in an acal mutant and wild type backgrounds. In stage 13 wild type embryos, TRE-DsRed marks only the LE [a very faint amnioserosa DsRed positive staining is also present (Fig. 4B, 4B'')]. At this stage, lateral epidermal cells are stretching towards the dorsal side (Fig. 4B, B'). In acal mutants, the TRE-DsRed is ectopically expressed in the lateral epidermis and in some amnioserosa cells ( Fig. 4C-E). Quantitation of the signal uncovers significantly higher JNK activation levels in the lateral epidermis of acal mutants, and a significantly higher sGMCA signal, consistent with the observed mis-stretching and local accumulation of lateral epithelial cells at the border of the epithelium during DC (Fig. 4E). Ectopic JNK activation coincides with mis-stretching of the lateral epidermis, seen as a contraction of the epithelium where JNK ectopic activation occurs ( Fig. 4C-D). Similar results are seen at later DC stages (S4A-D Fig.). JNK ectopic activation in the amnioserosa of acal mutants is likely a consequence of lateral epithelium defects, as acal rescue solely in the lateral epithelium rescues DC defects. Significant ectopic JNK reporter activity in the lateral epidermis underneath the leading edge cells and in the amnioserosa was also observed using another weaker acal allele (acal 2 ) (S4E-G Fig.), with a GFP-based TRE JNK activity reporter [33]. Results for acal 2 are normalized for nuclear density. This strongly suggests that acal DC defects are due to JNK gain-of-function signaling in more lateral epidermal cells, resulting in disorganized stretching.
As a second and independent means to study the consequences of acal mutations in JNK activity, we over-activated JNK signaling in acal mutant and control embryos by expressing both a constitutively active form of the JNK kinase, Hep (Hep ACT ), and a wild type version of the same gene (Hep) in the ectoderm (using 69B-gal4). We found that expressing hep ACT in acal heterozygotes results in significantly stronger and more penetrant closure defects compared to wild type embryos expressing hep ACT (S4I Fig.), exacerbating the hep ACT over-expression phenotypes and supporting the notion that acal mutants promote increased JNK activity. We also see the appearance of a new phenotype, 'early' embryonic death, where embryos die without forming a cuticle. Over-expression of hep in acal heterozygotes has milder effects, but significantly leads to an increase in death of fully formed embryos in acal heterozygotes, compared to wild type embryos over-expressing Hep (S4H Fig.). Homozygosity for acal 5 together with hep ACT over-expression, or for acal 2 together with hep over-expression, lead to significant further increases of the respective mutant phenotypes (S4H-I Fig.). These hep gain-of-function experiments together with acal loss-of-functions are all consistent with a role for acal in negative regulation of JNK activity during dorsal closure in the lateral epithelium.
We then corroborated this in two other independent ways: using a second JNK signaling reporter, and performing genetic interactions with JNK mutants. If acal acts to negatively regulate JNK activity, then in acal mutants we should see enhanced JNK reporter line expression (similar to the TRE results, above), or suppression of acal phenotypes in partial JNK loss-offunction.
puc is a transcriptional target of JNK signaling coding for a JNK phosphatase, a negative feedback loop that halts JNK. puc E69 is a lacZ enhancer trap allele (from now on referred to as puc lacZ ). Heterozygosity for puc lacZ provides a JNK sensitive signaling reporter, and a sensitized background disrupting the feedback loop [18]. Despite decreased puc dosage in puc lacZ heterozygotes, in approximately 90% of embryos, JNK signaling is restricted to LE (Fig. 4F-H). In acal homozygotes (acal 1 , acal 2 , or acal 5 ), heterozygous for puc lacZ , the proportion of embryos with ectopic puc lacZ expression in amnioserosa and lateral epidermis significantly grows to around 50% or more, similar to JNK ectopic activation with TRE-constructs (Figs. 4G-H and S4A-G).
acal mutants heterozygous for puc lacZ also show more extreme cuticle phenotypes. This is consistent with a potentiation of the puc effect on acal phenotypes, both genes acting to counteract JNK signaling (S5A Fig.).
Finally, we reduced JNK dosage to see whether it would ameliorate acal mutant phenotypes. Heterozygosity for bsk 1 , a JNK loss-of-function mutation, significantly suppresses acal 5 phenotypes (Fig. 4I), meaning that JNK signaling over-activation of acal mutants depends on JNK itself. Conversely, acal 5 heterozygosity does not alter the bsk 1 mutant phenotype (Fig. 4I), consistent with acal mutant phenotypes dependent on JNK. We confirmed this genetic interaction in a different genetic background and found similar results (S5B Fig.). Taken all data together, the (1) two types of JNK reporters (TRE-based and puc lacZ ) tested with several acal alleles, and (2) genetic interactions for several acal alleles with bsk and puc (loss-of-function), and hep (ectopic expression), all establish acal as a negative regulator of JNK signaling.
We next tested if acal mutations interact with dorsal open group mutations known to act in the amnioserosa, in order to determine if acal could have a function in this tissue. We used peb 308 , a hypomorphic mutation that affects peb expression in the amnioserosa, and, consequently, DC [19]. The peb 308 DC defects were not modified in an acal sensitized background, consistent with acal required only in the lateral epidermis (S5C Fig.).
The pioneer protein Raw and acal act together to counteract JNK signaling acal might interact with other negative regulators of JNK signaling. raw, a conserved gene of unknown molecular function, counters JNK signaling in the lateral epithelium, like acal. raw and acal have very similar expression patterns during DC (Fig. 2D for acal; for raw see [17]). Raw function is genetically positioned at the level of JNK [35], like acal, and Jra [17]. raw mutant cuticular phenotypes are stronger than acal, with dorsalized embryos consequence of dpp ectopic expression at the lateral epidermis (see Fig. 5A-B and Fig. 6E-F, [17]).
raw mutants have dpp ectopic expression during DC in rows of cells ventral to the leading edge ( [17] and S6C Fig.). acal ectopic expression in the posterior part of each segment of these embryos (en-gal4 driver) reduces cell-autonomously dpp over expression (S6D Fig.). This is consistent with acal acting to inhibit JNK pathway on the lateral epithelium during DC. On their own, acal mutant embryos are not dorsalized (Fig. 1E-L), and show no clear ectopic dpp over-expression (S6A-B Fig.), yet at least under conditions of dpp over-expression, acal can repress dpp. Altogether, this argues that acal is a negative regulator of JNK activity, and that it has a subtler effect to that of raw.
We analysed genetic interactions between acal and raw. Heterozygosity for raw 2 increases the embryonic lethality of acal 5 embryos, with a quarter phenocopying the raw 2 dorsalized phenotype ('raw-like phenotype', Fig. 5C). acal 5 heterozygosity does not enhance the already strong raw 2 homozygous condition, and double homozygotes do not show a stronger   (S6E Fig.). These genetic interactions are consistent with acal acting downstream of raw.
We looked for acal expression in raw mutants, and found that acal lateral epidermis expression is significantly decreased in these embryos (Fig. 5D-E). This suggests that raw regulates acal lateral epidermis expression. Nervous system acal expression is unchanged in these raw mutant embryos, acting as an internal control. This also ties in with acal 5 being a regulatory mutant with reduced lateral epithelium expression (Fig. 2D), and suggests a raw dependent acal control module for the lateral epidermis. Consistent with acal acting downstream of raw, as a raw effector, 69B-gal4 UAS-acal expression partially rescues the raw dorsalized cuticle phenotypes (S6F-H Fig.), as expected from the dpp in situ hybridization in raw mutants over-expressing acal (S6D Fig.). Altogether, these results indicate that acal mediates part of Raw function during DC.
raw over-expression in the lateral epidermis of wild type embryos has no effect in DC, in spite of the strong raw loss-of-function phenotype [17]. Likewise, acal over-expression in wild type embryos has no DC effects (embryonic lethality of 69B>acal embryos is around 1%, n = 791; also pnr MD237 >acal shows no lethality or overt phenotypes, with equal numbers of control and over-expression siblings, n = 104). These lack of gain-of-function phenotypes for both genes indicate that raw/acal JNK negative regulation has a limit beyond which it cannot be exercised further during dorsal closure, suggesting other regulatory mechanisms might act in concert.
Thoracic closure during metamorphosis also depends on JNK signaling, much like DC does [36,37] Altering JNK signaling at the notum anlagen results in an antero-posterior cleft spanning the adult thorax ( Fig. 5G-H). Some JNK signaling loss-of-function conditions have cleft thoraces [38,39]. This process has been used also in miss-expression screens for JNK signaling, where gain-of-function phenotypes are repeatedly encountered [40]. We thus used thorax closure as a model where we might be able to study acal and raw gain-of-function conditions. For this, we hypothesized that acal and/or raw over-expression at the notum anlagen could result in a cleft phenotype, using pnr-gal4 (expressed at the dorsal heminota, equivalent to the lateral epidermis, Fig. 5F-H). pnr-gal4 also provides a sensitized genetic background, as heterozygosity for this allele generates a mild thoracic phenotype, increasing the sensitivity of the assay [41].
Thoracic cleft was not enhanced by raw notum over-expression, as quantified by a 'Thoracic cleft' index ( Fig. 5I-J; Materials and Methods). Similar results were observed for RawRA or RawRB isoforms, generated by alternative splicing (Fig. 5J). In contrast, two copies of the UASacal transgene had a significant effect (one copy was ineffective; Fig. 5H-J). If raw over-expression induces acal, adding more acal would then lead to a more extreme phenotype, as adding two transgene copies of acal do in the dorsal thorax. Simultaneous over-expression of acal and raw single copy transgenes had a significant synergistic effect, inducing stronger cleft formation (Fig. 5J). Both raw isoforms show synergism with acal. This is consistent with raw positively regulating acal, and both counteracting JNK signaling. Taken together, these results show that acal and Raw act together to counteract JNK signaling, acal downstream of raw. This is consistent with Raw acting at the level of JNK and/or Jra, as our experiments with acal point.

Expression of Cka, a JNK scaffold, is regulated by Raw and acal
Both raw and acal act genetically at the level of JNK ( Fig. 4I; [35]). Yet raw has been shown to act also at the level of Jra [17]. Could raw do both? Could raw via acal modulate a mediator of both JNK and Jra, for example? Cka is a scaffold protein that interacts with JNK and Jra [20]. We tested whether acal and Cka interact genetically. acal DC defects are dependent on Cka, since heterozygosity for Cka 1 rescues the acal mutant phenotype (Fig. 6A). This is in fact a stronger genetic interaction than that seen with bsk (JNK) mutants. Moreover, Cka expression is similarly significantly increased (to 50% above control values) in both acal and raw mutants (Fig. 6B). Conversely, acal embryonic over-expression (gain-of-function) leads to a significant reduction in Cka expression in otherwise wild type embryos, although this condition bears no observable phenotypic consequences (Fig. 7D).
To test the raw-acal-Cka genetic pathway, we silenced Cka to see if it suffices to rescue the raw mutant phenotype, using again the strong raw mutation-dependent dpp over-expression phenotype as readout (Fig. 6E-F). If this pathway is true, expressing Cka RNAi in the en-gal4 expression pattern in raw mutants would silence dpp expression only at posterior compartments of each segment, the anterior compartments serving as internal controls. We found that Cka knockdown suppresses cell-autonomously raw ectopic dpp expression (Fig. 6F versus 6H), similar to rescuing raw function with UAS-rawRA (Fig. 6F-G). Taken together, both the lossof-function as well as the ectopic expression results argue in favor of raw acting through acal down-regulating JNK signaling by regulating Cka expression, and accommodates the apparently conflicting results that raw acts at the JNK and/or Jun levels. Genetic interactions do not imply direct interactions, and so, acal, as a raw mediator, could act indirectly on Cka and regulate in this way JNK and Jra.
Cka sole over-expression should alter DC. Using a heat-shock inducible Cka, hs-Cka [20], we found that 29°C treatment of embryos throughout development significantly induces DC defects (Fig. 6C). Heat-shock induction indeed increases Cka expression (Fig. 6D). Taken together, these results define a novel regulatory path for JNK signaling, raw counteracting Cka expression through acal.
acal fine-tunes DC by regulating aop expression Does acal interact with other DC regulators? Aop serves at least two functions during DC. In the lateral epidermal cells, Aop promotes tissue differentiation by preventing ectopic mitoses. As an extension, in the LE, it has a second function preventing ectopic and precocious expression of JNK activity and of AP-1 target genes, and thus, acting against dorsal closure, an activity countered by JNK via phosphorylation. Hence, over-expression of a constitutively active Aop (aop ACT ) that cannot be inactivated by JNK in the LE, results in DC defects due to lack of JNK targets expression. On the other hand, aop loss-of-function leads to anterior holes in the cuticle and faulty DC resulting from both ectopic mitoses in the tissue, and JNK over-activation. The wild type stretching lateral epithelium is a post-mitotic, competent tissue for DC in part due to aop dual roles (i.e., mitosis and precocious DC repression [15]). So, aop regulation can be a nodal, sensitive point for DC regulation. Genetic analyses show that acal and aop interact. acal mutations powerfully rescue aop loss-of-function, but aop loss-of-function does not alter acal mutants, showing that acal acts above or at the level of aop (Fig. 7A). Indeed, we found that aop expression is significantly increased in acal and raw mutants (Fig. 7B). Conversely, acal over-expression in wild type embryos leads to a significant reduction of aop expression (Fig. 7D), without phenotypic consequences (similar to Cka), consistent with a modulatory role for acal.
In accordance with an acal role as JNK negative regulator, aop ACT that presumably locks in the post-mitotic nature of the lateral epithelia, while preventing JNK target gene expression, is countered by acal mutants, more so if aop ACT expression is less widespread (Fig. 7C). A possible explanation for the puzzling interactions between acal and aop, resides in the dual nature of aop function in the LE and the lateral epidermis. In wild type embryos, aop is only phosphorylated and inactivated at the LE by JNK. Expressing aop ACT conceivably affects only the JNKregulated function of aop (i.e., its role in the LE). Thus, the JNK over-activation seen in acal mutants might compensate partially for LE aop ACT . In contrast, in aop mutants, low aop levels throughout the epidermis might be partially restored to more wild type levels by acal loss-offunction conditions, as acal mutants over-express aop. Taken together, by regulating aop expression, acal can powerfully influence DC. In summary, we show that acal critically regulates both positive (Cka) and positive/negative (aop) components of JNK signaling.
acal acts in trans during DC How does acal regulate expression of other genes? One possibility is that acal acts in cis, exerting an effect on neighboring genes, something reported for other lncRNAs [10,41]. To study this possibility, we made use of a Gene Search insertion line, a transposon that can lead to ectopic expression outwards in both directions from the insertion side. GS88A8 maps in the same interval as acal. Using gal4 lines, GS88A8 generates ectopic expression of neighboring genes, namely lola and psq [42]. We used the 69B-gal4 line to drive ectopic expression from GS88A8 in the embryonic ectoderm. If acal negatively regulates neighboring genes (lola and psq) and they are relevant for DC, then ectopic expression from GS88A8 should phenocopy acal mutant phenotypes. Instead, this condition mainly leads to larval and pupal lethality plus some adult escapers (S7B Fig.), with hardly any embryonic phenotypes, consistent with unaltered lola and psq expression profiles in acal mutants (S7C Fig.). Moreover, acal mutations complement lola and psq mutations (S1 Table), and mutations in these genes lead to different embryonic phenotypes (S7A Fig.). Altogether, we conclude that it is not very likely that acal acts in cis during DC.
Some long non-coding RNAs work in trans. One possibility is as part of chromatin remodeling complexes, for instance, although other mechanisms are possible [43,44]. In order to explore whether acal might function through remodeling complexes, we performed genetic interactions with Polycomb 3 (Pc 3 ) mutants. Pc is a critical member of the Pc chromatin repressive complex 1 (PRC1, [45]). acal and Pc 3 double heterozygotes show novel strong wing defects [significantly twisted wings, with reduced posterior territory (S8A and F Fig.)]. Consistently, this interaction is seen with raw mutant alleles, albeit to a lesser degree (S8B Fig.). Cka and aop do not show this interaction (S8C-D Fig.).
acal 5 also significantly enhances the Pc 3 extra sex combs phenotype, as does raw 1 (aop 1 reduces extra sex combs, presumably due to a repressive function in sex combs formation, S8G-I Fig.). These non-allelic non-complementation results are reminiscent of the 'extended-wing' phenotype used to detect direct interactors / components of other chromatin remodeling complexes, like the brahma or kismet complexes [46], suggesting a common regulatory pathway for acal, downstream of raw, with PRC1. The fact that raw and acal show the same genetic interactions with Pc, but not Cka and aop, underscores the fact that Cka and aop are downstream from the former two (raw and acal), as regulatory targets in the same signaling cascade.

Discussion
Several lines of evidence show that acal is a putative nuclear, processed, lncRNA, whose mutations are homozygous lethal, with requirements in DC. We show that acal mutants (three of which map molecularly to the SD08925, or acal, locus) fail to complement each other, but complement neighboring genes (lola and psq), regulate the same downstream genes (Cka and aop), have the same JNK negative regulation function [evidenced by two different reporter gene systems and genetic interactions with several JNK pathway component (bsk, hep, and Cka) and regulator genes (puc, raw, and aop)], are rescued by the SD08925 locus, with one allele having significantly reduced SD08925 expression, and all having similar mutant phenotypes and phenocritical periods. The locus produces a nuclear 2.3 Kb transcript with low coding potential that is processed into smaller pieces. Altogether, these data establish acal as a lncRNA locus in DC.
lncRNAs are ubiquitous species, probably tied to most biological processes. Yet, few have been characterized [47]. They fine-tune target abundance in multiple ways [1]. lncRNA species probably form diverse complexes with multiple mechanisms of action, and are not a rigorously defined class of molecules. Identifying lncRNA mutants with clear phenotypes in genetic models is a clear way forward. acal has an extremely low coding potential and no evolutionarily conserved ORFs. The putative acal ORFs show no homology to known peptides / proteins. Unlike short-ORF bearing transcripts [48], these putative ORFs have no sequence redundancy among them, reliable Kozak sequences, or high coding potential. This places acal as a nontranslated transcript.
acal is a Dipteran-conserved locus. As many lncRNAs, it is not broadly conserved across species. acal RNA is processed into smaller fragments. These fragments have probably been missed in high-throughput sequencing projects due to RNA cut-off sizes [32]. acal functions in trans to regulate DC, as DC defects can be rescued in trans with a wild type full-length cDNA targeted to the epidermis.
acal foci for DC is the lateral epidermis. JNK loss-of-function result in lack or reduction of lateral epidermal cells stretching. In contrast, negative regulators, such as peb, puc, and raw also cause DC defects, due to disorganized and / or precocious cell stretching [13]. In acal mutants, cells also stretch in a disorganized manner. Predictably, reducing bsk gene dosage significantly reduces acal defects in DC, and JNK activity monitors show enhanced JNK function in acal mutants. This prompted study of acal and other negative regulators.
raw male gonad morphogenesis defects are ameliorated by bsk gene dosage reduction [35]. acal and raw collaborate to down-regulate JNK signaling at the lateral epidermis. In situ hybridization experiments, embryonic genetic interactions, and raw and acal over-expression studies in the thorax, put raw upstream of acal, yet they do not strictly phenocopy. This means that raw has other effectors. raw mutant embryos are dorsalized due to strong ectopic dpp expression in the lateral epidermis [17]. acal mutants do not dorsalize or show overt dpp misregulation. Furthermore, since Raw is localized in the cytoplasm, and acal is present in the nucleus, it is reasonable to assume that Raw regulates acal expression indirectly. Over-expression of either raw or acal in wild type embryos has no discernible effect on DC ( [17] and this work), yet simultaneous over-expression in the thorax disrupts thorax closure, and acal over-expression reduces both Cka and aop transcript levels.
Cka expression is elevated in raw and acal mutants. Genetically decreasing Cka levels rescues DC defects in raw and acal mutants. Consistently, Cka over-expression itself causes DC defects. Work in the mammalian Cka functional homologue JIP-3 has shown this scaffold protein up-regulated under stress conditions. JIP-3 depletion and over-expression directly influence JNK activity [49]. Together with our results, these findings show that JNK scaffolds are an important point for JNK regulation, since augmenting scaffold availability augments JNK activity. In fact, this has been demonstrated biochemically for Cka [20].
Another critical JNK regulatory point by acal is aop. aop is also over-expressed in acal mutants, and down-regulated in acal over-expression. aop plays a complex role in DC making the tissue ready early, before DC, by preventing mitosis and early activation of JNK targets. For this, Aop function is required in the LE and lateral epidermal cells ventral to the LE [15]. Later, aop is inhibited by the JNK pathway only in the LE, allowing target gene expression there. In cells ventral to the LE, acal (together with raw) prevents JNK activation. In these cells there is no JNK activity to inhibit aop, yet aop expression regulation must be effected. This may be accomplished by acal (and raw). Thus, controlling aop levels, like Cka levels, may be a critical point for DC. For both genes, acal acts genetically as a repressor, with the net result of allowing JNK activity only in the LE in wild type.
In acal mutants, Cka over-expression leads to JNK ectopic activation in more lateral epithelium cells. However, the fact that aop is also over-expressed in acal mutants may mask a stronger JNK over-activation, as JNK signaling may inhibit at least in part Aop function, Aop being a JNK direct target. Future work will focus on how acal fine-tunes these signaling genes (Fig. 8).
We show here that acal function modifies gene expression. Whether this is direct, as has been proposed for other lncRNAs [45,50], or is done in some other fashion, is not yet clear. acal interacts genetically with Pc (raw, that regulates acal, also interacts with Pc). This opens the interesting possibility that acal acts directly with chromatin remodeling complexes, as has been demonstrated for other Pc interactors [42,[51][52][53]. Yet other possibilities and/or indirect modes are clearly plausible.
In conclusion, we show that a lncRNA locus is required for DC regulation and JNK signaling. lncRNAs have not been implicated before in JNK signaling, and hence, this adds a new layer of control in JNK regulation.

Materials and Methods
For detailed protocols, please refer to S1 Protocol.

Fly husbandry
Detailed description of stocks is available in S1 Protocol. Crosses were done at 25°C under standard conditions. For heat-shock treatments (hs-Cka and peb 1 experiments), animals were maintained at 29°C throughout development.

Cuticle preparations
Mutant embryos were processed as in [54]. Cuticles were examined under a dark field microscope.

Molecular mapping and rescue
4.7 kb of genomic DNA encompassing the acal locus of homozygous embryos was PCR amplified, cloned, and sequenced. The SD08925 clone was sequenced fully. A genomic (178D09) and a UAS-cDNA (UAS-acal) rescue constructs were synthesized and employed.

In situ hybridization and LacZ staining
For in situ hybridization, embryos were fixed, hybridized with digoxigenin-labeled RNA probes, and developed with alkaline phosphatase-NBT/BCIP. Pixel intensity was quantified using ImageJ. For LacZ staining, embryos were fixed and developed with X-Gal.

Northern blots
Northern blots were done according to standard procedures [55]. Small RNA Northern blots were resolved by PAGE, and hybridized with labeled DNA oligonucleotides.

Nuclear fraction preparation
Embryos were dechorionated and processed for nuclei low-salt extraction by centrifugation. Supernatant was recovered as the cytoplasmic fraction. Both fractions were homogenized in TRIzol.
cDNA synthesis, semi-quantitative PCR, and real time PCR RNA was purified and retro-transcribed with M-MLV reverse transcriptase. Semi-quantitative PCR and real time PCR were done by standard methods. Densitometry analysis was done using ImageJ. Primers used are listed in S2 Table. Fluorescence microscopy Z-stacks of embryos undergoing DC were obtained using a Zeiss LSM 780 microscope and processed using ImageJ. Fluorescence intensity was calculated using ImageJ.

Scanning Electron Microscopy
For imaging eyes and thoraces, animals were cold anesthetized, attached to SEM stubs with carbon paint and glue, and imaged under high vacuum in a JEOL JSM-6060.

Thoracic cleft index and wing angle quantification
Measurements were done with iVision software.

Adult leg preparation
Adult male thoraces were digested with potassium hydroxide and mounted for viewing and quantification of sex combs.
Bioinformatics acal homologous sequences were extracted from the Flybase Genome Browser [56]. Multiple sequence alignments were done with the CLC sequence viewer software (http://www.clcbio. com). Putative translated ORFs were analyzed with BLAST and compared in the Drosophila PeptideAtlas server [57]. Protein coding capacity was calculated according to [29].