Repression of the Hox gene abd-A by ELAV-mediated Transcriptional Interference

Intergenic transcription is a common feature of eukaryotic genomes and performs important and diverse cellular functions. Here, we investigate the iab-8 ncRNA from the Drosophila Bithorax Complex and show that this RNA is able to repress the transcription of genes located at its 3’ end by a sequence-independent, transcriptional interference mechanism. Although this RNA is expressed in the early epidermis and CNS, we find that its repressive activity is limited to the CNS, where, in wild-type embryos, it acts on the Hox gene, abd-A, located immediately downstream of it. The CNS specificity is achieved through a 3’ extension of the transcript, mediated by the neuronal-specific, RNA-binding protein, ELAV. Loss of ELAV activity eliminates the 3’ extension and results in the ectopic activation of abd-A. Thus, a tissue-specific change in the length of a ncRNA is used to generate a precise pattern of gene expression in a higher eukaryote.

Introduction Several noncoding RNAs (ncRNAs) have been identified from the Hox clusters of different species; a few of these have been shown to play key roles in gene regulation [1][2][3][4][5][6][7][8][9][10]. One of these ncRNAs is the 92 Kb, spliced and polyadenylated transcript called the iab-8 ncRNA. Located within the Drosophila Bithorax Complex (BX-C), the iab-8 ncRNA originates from a promoter located about 4.5Kb downstream of the Abd-B transcription unit and continues until within about 1 Kb of the abd-A promoter. In situ hybridization experiments show that it is transcribed specifically in the very posterior epidermis of the embryo from the cellular blastoderm stage. From later embryonic stages, its expression becomes limited to parasegments (PS) 13 and 14 of the CNS [1,2,[11][12][13][14]. Loss of the iab-8 ncRNA has been shown to result in both male and female sterility, likely due to problems in the innervation of muscles important for reproduction [14,15]. Much of its function has been attributed to a microRNA located between its sixth and seventh exons, called miR-iab-8 (miRNA). miR-iab-8 targets multiple transcripts including the Ubx and abd-A homeotic genes and their cofactors hth and exd [1,[15][16][17]. Indeed female sterility has been directly linked to ectopic hth, Ubx and abd-A in the CNS [1,[15][16][17].
In the embryonic CNS, abd-A expression is normally limited to PS7-12. Studies performed by our lab and others have shown that the restriction of abd-A expression from PS13 in the CNS is dependent upon expression of the iab-8 ncRNA [1,14,15,18]. Although, the miR-iab-8 miRNA plays a part in the repression of abd-A in PS13, a deletion of the miRNA template sequence only results in a mild derepression of abd-A in PS13 (see Fig 1D) [14]. On the contrary, mutations preventing the production of the iab-8 ncRNA cause a complete de-repression of abd-A, such that the abd-A expression pattern in PS13 mimics that of PS12 (see Fig  1C), suggesting the existence of a second repression mechanism. Here, we explore the mechanism by which the iab-8 ncRNA represses abd-A. Using deletions spanning different regions of the iab-8 transcript, we were unable to identify specific parts of the transcript that can account for the additional repression of abd-A by the iab-8 ncRNA. Furthermore, we find that the iab-8 transcript can repress an exogenous reporter gene placed downstream of its sequence. Based on these findings, we conclude that it is the act of transcription that is necessary for repression, rather than the sequence transcribed (a phenomenon called transcriptional interference). Examination of the iab-8 transcript in the CNS, shows that there is a 3' extension made specifically within the CNS. This elongated transcript seems to be essential for abd-A down-regulation and requires the neuronal-specific, RNA-binding protein ELAV (or its paralogue FNE) for its creation. Overall, our work suggests that ELAV mediates a 3' extension of the iab-8 ncRNA that, in turn, allows it to specifically repress abd-A expression in the posterior CNS through transcriptional interference.

CRISPR-mediated deletions within the transcriptional unit of the iab-8 ncRNA do not affect the microRNA-independent repression of abd-A in the CNS
In order to identify the second element in the iab-8 transcript that is required for abd-A repression, a series of CRISPR-mediated deletions were created within the iab-8 ncRNA template. Previously, we used classically isolated deletions and chromosomal breaks to eliminate much of the 5' sequences as containing this second activity. From these studies, we were able to eliminate all but the final two exons of the iab-8 ncRNA [14]. Here, we examined deletions that remove large portions of the iab-8 ncRNA starting from the 3' end, but leave the iab-8 promoter and first two exons intact for continual transcription. The largest of the deletions examined was del(ex3-8) [19]. This deletion leaves only the first two exons of the iab-8 ncRNA, leading to a juxtaposition of these exons with the abd-A promoter (Fig 1A). Although this deletion removes almost the entirety of the iab-8 transcription unit, including the iab-8 miRNA, del(ex3-8) shows only a mild de-repression of abd-A in PS13 of the CNS (Fig 1E). By contrast, a deletion that removes the iab-8 ncRNA promoter (Fab-8 64 ) shows a complete derepression of abd-A, such that abd-A expression in PS13 resembles that found in PS12 ( Fig  1C). In fact, the mild de-repression seen in del(ex3-8) is reminiscent of the phenotype caused by the clean deletion of mir-iab-8 alone (Fig 1D), which is also deleted in this mutant. Since the del(ex3-8) mutation does not show more extensive abd-A derepression than the deletion of mir-iab-8 alone, we conclude that the sequence deleted in del(ex3-8) does not contain the secondary repressive element encoded by the iab-8 ncRNA. Based on the analysis of del(ex3- 8), the second repressive element should be located upstream of exon 3 in the iab-8 transcript. This was at odds with our previous results, where we concluded that the 3' region of the transcript was important for abd-A repression. Analysis of a chromosomal inversion of the iab-8 exon 3-8 region, helped to shed light on this discrepancy. During the course of generating our deletions, we also obtained, at high frequency, inversions of the areas targeted for deletion (Fig 1A). One of these deletions, inverts the exact sequences deleted in del(ex3-8) that we call inv(ex3-8) [19]. In inv(ex3-8), the iab-8 promoter and first two exons are untouched and, by in situ hybridization using probes directed against these exons, seem to be transcribed normally (S1 Fig). We reasoned that if these exons produce something that is important for abd-A repression, inv(ex3-8) should repress abd-A about as well as del(ex3-8). However, examining abd-A expression in inv(ex3-8), we see a total derepression of abd-A in PS13 of the CNS (Fig 1F), thus suggesting that the sequences of exons one and two are not, by themselves, crutial for repressing abd-A.
Based on the data from the deletions and inversions, no particular sequence can account for the miRNA independent repression of abd-A. We cannot, however, rule out the importance of both the iab-8 and abd-A promoter regions. Previously, we published 3'RACE results that showed that about 5% of the iab-8 transcripts in embryos do not terminate at the end of exon 8 [14]. We wondered if this "read-through" transcription could be important for abd-A repression. To visualize this transcription, we performed in situ hybridization using a probe against the intergenic sequence located between the terminal exon of the iab-8 ncRNA and It is interesting to note that in wildtype, although the iab-8 ncRNA can be detected using an exon 8 probe in the epidermis, the intergenic transcription, can only be seen in the developing CNS, which probably accounts for the low proportion of this transcript detected in the 3'RACE studies. Thus, these results suggest that there is a tissue-specific extension of the iab-8 ncRNA that allows the iab-8 transcript to extend past exon 8 towards the abd-A promoter region. Moreover, this extended transcript is eliminated in the inv(ex3-8) embryos, where abd-A repression is lost.

CNS-specific transcriptional read-through of the iab-8 ncRNA
The probe directed against the intergenic region between the iab-8 ncRNA and abd-A, demonstrates that there is significant transcriptional read-through from the iab-8 ncRNA in PS13 and 14 of the embryonic CNS. Examining recently published RNA-seq data on timed embryo collections, the presence of the readthrough transcripts can be verified (see below) [20]. In order to explore the extent of the transcriptional read-through, we tested additional probes within the abd-A transcription unit. The most-3' probe was generated against the 3'UTR of abd-A. This probe detects abd-A transcription in the CNS, in a domain that spans from PS7 to PS14 (Fig 2). This domain of transcription extends beyond the normal PS7-PS12 domain of expression found by ABD-A protein immunostaining. Interestingly, the transcript signal detected with the abd-A 3'UTR probe shows two distinct patterns: in PS7 to PS12, the domain that corresponds to ABD-A protein expression, the staining is concentrated in certain clusters of cells, giving a characteristic, aspect to each parasegment that we call the "domino" pattern ( Fig 2B, black arrows). Meanwhile, in PS13 and PS14, a diffuse, uniform signal is observed (Fig 2B, grey arrows).
No previous evidence has shown that the abd-A promoter is highly active in PS13-PS14 of WT embryos. In light of our observation of high levels of read-through transcription in the CNS and the different staining pattern in the most-posterior parasegments, we asked if the staining observed in these parasegments could correspond to transcriptional activity emanating from the different promoters (abd-A vs iab-8). To test this, we performed in situ hybridizations using the same abd-A 3'UTR probe on Fab-8 64 embryos, which lack the iab-8 ncRNA promoter. In these mutants, we observe that the uniform signal normally present in PS13 and 14 (Fig 2C, white arrow) is gone. However, the "domino" pattern remains and actually invades PS13 (Fig 2C, black arrow). These results are consistent with the idea that the diffuse pattern represents transcriptional read-through by the iab-8 ncRNA into the abd-A transcription unit and that, in the absence of the iab-8 ncRNA, the abd-A enhancers are able to drive expression of abd-A in PS13 (but not PS14).
In order to show that the iab-8 ncRNA continues into the abd-A transcription unit within the CNS, we performed RT-PCR experiments on 0-24 hour-old embryos to identify iab-8/ abd-A fusion transcripts. The RT-PCR experiments were performed using primers in different exons of the iab-8 ncRNA and abd-A. Using this method, five differentially spliced isoforms were identified; all of the isoforms were composed of 5' exons from the iab-8 ncRNA and 3'exons of abd-A. For the most part, the alternative isoforms were generated by splicing from exons 6 or 7 of the iab-8 ncRNA, to exons 2, 4 or 5 of abd-A. All variants lacked exon 8 of the iab-8 ncRNA and all but one isoform lacked the initial ATG of abd-A (iab-8(7)-abd-A(2)) (see Fig 3, and also the discussion regarding the ATG-containing isoform).

Transcription of the iab-8 ncRNA represses the expression of a downstream reporter gene
Thus far, our results indicate that the iab-8 ncRNA extends well into the abd-A gene in PS13 and 14 of the CNS. Furthermore, we see read-through transcription in embryos where abd-A is repressed (like wild type and del(ex3-8) embryos), but not in cases where abd-A is ectopically expressed (like Fab-8 64 and inv(ex3-8)). If abd-A is repressed by the extension of the iab-8 transcript, then one likely mechanism by which this could occur is transcriptional interference. Transcriptional interference relies on the act of transcription across a target promoter/gene to repress its transcription, presumably by either physically removing a downstream RNA polymerase or changing the chromatin environment to make it less capable of initiating transcription. Thus, it is independent of the sequence of both the repressing and the targeted transcripts. To determine if transcriptional interference could explain the additional repression of abd-A in the embryonic CNS, we decided to recreate the transcriptional situation present at the abd-A promoter, using a reporter system that would be independent of trans-acting factors produced by the ncRNA.
In the reporter line, we fused the EGFP cDNA to the abd-A promoter and the first 140bp of the abd-A 5'UTR. We then integrated this construct (which also contains iab-8 exon 8) into the genomic locus of exon 8 of the iab-8 ncRNA (abd-A:GFP: Fig 4) to recreate the abd-A promoter region downstream from the endogenous iab-8 ncRNA in the BX-C, but with the EGFP sequence replacing abd-A. This modified gene should respond to the BX-C enhancers that drive localized expression of abd-A, but, since it does not contain the vast majority of normal abd-A sequence (including its 3'UTR), should not be repressed by any sequence-specific, trans-acting factors produced by the ncRNA (like mir-iab-8).
Examining EGFP expression in these embryos, we see that the EGFP reporter, like the wild type abd-A protein (Fig 4B), is strongly expressed in the embryonic CNS from PS7 to PS12 https://doi.org/10.1371/journal.pgen.1009843.g003 (Fig 4C). However, unlike the wild type ABD-A protein, there is a slight ectopic expression of EGFP in PS13 of the CNS (Fig 4C, bracket). This ectopic expression is similar to the pattern of ABD-A in miR-iab-8 mutants (Fig 1D). As the abd-A:GFP reporter lacks microRNA binding sites, this result was expected. Importantly, however, EGFP expression in PS13 is significantly lower than that seen in PS12. Thus, the second repressive mechanism is still present and can prevent the full expression of GFP in this parasegment. To prove that this repressive mechanism is dependent on iab-8 transcription, we then recombined an iab-8 promoter deletion (Fab-8 64 ) onto the abd-A:GFP chromosome. Strikingly, this leads to a full de-repression of EGFP in PS13 of the CNS (Fig 4D), proving that the expression of the iab-8 ncRNA is required for the repression of the abd-A:GFP gene in this parasegment. Finally, as transcriptional interference requires the interfering transcript to be on the same chromosome, we compared the expression of the GFP reporter when it was placed in cis or in trans to the iab-8 promoter deletion. As seen by comparing Fig 4D to Fig 4E, the complete derepression of the GFP reporter requires that the deletion be placed in cis to the reporter. Based on these experiments, we conclude that transcriptional interference limits the spatial domain of abd-A expression.
The neuronal specific gene ELAV enhances the production of the transcriptional read-through of the iab-8 ncRNA and is necessary for transcriptional interference of abd-A As mentioned above, the read-through transcription of the iab-8 ncRNA seems to be tissuespecific and is restricted to the CNS. Indeed, while a probe against exon 8 of the iab-8 ncRNA is able to detect expression of this gene in the ectoderm of early embryos (Fig 5B) and the CNS of late embryos (Fig 5C), none of the other probes targeting the genomic areas downstream from exon 8 of the iab-8 ncRNA is able to detect significant transcription in PS13-14, in any tissues outside of the CNS (Fig 5).
The CNS-specific nature of this transcriptional read-through made us wonder if it could be due to the action of a neuronal tissue-specific factor. The neuronal RNA binding protein ELAV has previously been described [21] as a protein responsible for the alternative polyadenylation and splicing of neuronal-specific isoforms of numerous genes. In fact, the homeotic genes, including abd-A, are known targets of ELAV, whose activity causes an elongation of the 3'UTR, often making them more susceptible to miRNA regulation [22]. Following this idea, we examined the expression of the iab-8 ncRNA in elav mutants by in situ hybridization using probes targeting two different areas of the main transcription unit (iab-8ex1-2 and iab-8ex8) and an additional probe targeting the intergenic region between iab-8 and abd-A. As illustrated in Fig 6, although elav mutants show similar levels of transcription emanating from the iab-8 ncRNA promoter (as visualized by the iab-8ex1-2 and iab-8ex8 probes, Fig 6A-6D), they show a strong reduction in the amount of transcript that continues past iab-8 exon 8 (as visualized by staining with the intergenic probe, Fig 6E and 6F). Thus, ELAV is required, at least in part, to extend the iab-8 transcript beyond exon 8.
Since abd-A repression in PS13 seems to depend on read-through transcription from the iab-8 transcript and the read-through transcript seems to be ELAV dependent, it follows that abd-A repression should be ELAV dependent. We tested this by examining ABD-A expression in elav embryos. Indeed, immunostaining of elav embryos against ABD-A shows a noticeable level of ABD-A protein derepression in PS13 of the CNS (Fig 6H).
It is noteworthy that the elav embryos used in the experiments above still express the miriab-8 miRNA. Also, recent studies have shown that in the absence of elav, its paralogue fne is differentially spliced to make a form of FNE capable of performing some of ELAV's functions [20,23]. Our in situ results show that a small amount of read through transcription continues in elav mutants (Fig 6F). However, the recent transcriptomic analysis of elav mutants shows that there is a reduced, but noticeable amount of the read-through transcription in elav mutants. This amount of readthrough transcription is further reduced by also removing the elav homologue fne. This can be seen in Fig 7A and 7B, where RNA isolated from embryos of different ages were examined for read through transcription. To verify an effect on ABD-A expression, we examined ABD-A protein levels in embryos lacking elav, fne and miR-iab-8. As seen in Fig 1G, removal of these three elements is able to very strongly derepress abd-A in PS13 (see Discussion). Overall, these observations led us to conclude that the spatial restriction of abd-A expression in the CNS is determined by the iab-8 miRNA along with transcriptional interference of the abd-A gene by the iab-8 ncRNA in an ELAV/FNE dependent fashion.

Discussion
Previously, we reported that the main transcript of the iab-8 ncRNA terminates~1 kb upstream of the abd-A transcription unit. This finding was based on 3'RACE experiments performed on RNA isolated from relatively early embryos (6-12 hours). Here we show that, in later embryos, where the iab-8 ncRNA is restricted to the CNS, the iab-8 transcript extends well into the abd-A transcription unit. In every condition tested, we see that abd-A is repressed when there is an  [20]. B. RT-qPCR quantification of iab8 ncRNA and the intergenic region RNA in wild-type, Δelav, and ΔelavΔfne embryos. RNA was extracted from whole embryos 14-16h AEL (after egg laying) and 18-22h AEL. RNA levels were normalized to iab8 ncRNA levels. Error bars represent mean ± SD of three biological replicates (10-15 embryos per replicate) for each genotype. P-values are indicated (one-way ANOVA testing the difference to wild type). As typical for ELAVdependent RNAs, the intergenic region is more significantly downregulated in ΔelavΔfne than in Δelav embryos at both time points. N.S.: non-significant.
https://doi.org/10.1371/journal.pgen.1009843.g007 extended transcript. However, outside of the sequences required for normal transcription and RNA processing, the iab-8 transcript, itself, does not seem to require any specific sequences to mediate this repression. Based on these findings, and the cis nature of this repression, we conclude that the act of transcribing the extended iab-8 ncRNA is what represses abd-A expression in PS13 of the CNS. This type of inhibition is called transcriptional interference.
In transcriptional interference the transcription of one gene spreading over the coding or regulatory sequences of another gene is able to downregulate the target gene's expression [24]. The mechanisms mediating transcriptional interference seem to depend on the relative position of both promoters. In the case of the iab-8 ncRNA and the repression of abd-A, we have a case of tandem promoters, where the genes are transcribed in the same direction and the upstream transcript transcribes over portions of the downstream gene (promoter, enhancers, transcription unit). Studies performed in single cell organisms (yeast and bacteria) suggest that there are two main mechanisms mediating transcriptional interference of tandem promoters. The first is called the "sitting duck" mechanism, where an initiating RNA polymerase or activating transcription factors are knocked off of the target gene by the passing polymerase. The potential second mechanism to mediate transcriptional interference is called the "occlusion" mechanism, where activating transcription factors (or RNA polymerase itself) for the downstream gene are prevented from binding to their binding sites by the passing RNA polymerase or by the modified chromatin structure following the passage of an elongating polymerase. Thus far, we cannot distinguish between these two mechanisms in our system. However, both mechanisms have been shown to be dependent upon the strength of the silencing transcript's promoter relative to the target transcript promoter. The stronger the promoter activity from the upstream gene, the stronger the repression of the downstream gene [25]. In the case of transcriptional interference by the iab-8 ncRNA, we believe that its level of transcription is approximately equivalent to that of abd-A. Indeed, using an abd-A intronic probe to compare levels (a probe not subject to possible stabilization of the exonic probes of the abd-A coding mRNA), we see a similar level of transcription from both the iab-8 (PS13 and 14) and abd-A promoters (PS7 through PS12) (S3 Fig). Given the slower nature of transcription initiation vs transcriptional elongation, this high level of transcription might favor downstream gene repression.
From work on mammalian cells, it has long been known that the final exon of coding genes often promotes termination by the recruitment of the termination machinery to the poly(A) site [26]. Although in recent years, ELAV has been studied as a protein whose function lies in extending the 3'UTRs of neuron-specific genes [21] by altering the selection of poly-A signals [27,28], RT-PCR results suggest that, here, ELAV may play a role in the alternative splicing of the final exons of the iab-8 transcript. This function in alternative splicing is consistent with the role described for ELAV as an RNA binding protein involved in the alternative splicing of the neuronal isoforms of the Nrg [29,30] and fne gene products [20,23]. In fact, ELAV family members have been shown to be particularly important for splicing into a terminal exon [31]. Thus, ELAV might extend the iab-8 ncRNA by suppressing the ability of the iab-8 transcript to splice into its normal terminal exon. This would then prevent the transcribing RNA polymerase from terminating, causing it to continue transcribing until it finds a new terminal exon. Published ChIP-seq experiments (where nascent transcripts were cross-linked to the genomic DNA along with proteins) on ELAV from early and later embryos support this interpretation. According to these results, there is additional ELAV binding at the junction between intron 7 and exon 8 of the iab-8 transcript in later embryos when iab-8 is expressed only in the CNS [32].
Interestingly, among the spliced fusions between iab-8 and abd-A, we found one isoform that contains the abd-A ATG sequence. This would seem counter-productive, if the function of transcriptional interference is to prevent abd-A expression. Although we cannot judge the amount of this specific transcript based on our experiments, previous results from our lab have suggested that exons one and two of the iab-8 ncRNA act as translational repressors. Indeed, the MSA RNA [19], which is identical to the iab-8 ncRNA except that the first two exons of iab-8 are replaced by an alternative first exon, actually codes for a peptide whose coding sequence lies in the shared last exon [33]. A GFP fusion to this peptide and other reporters placed in the iab-8 sequence have shown that these proteins are never expressed in the CNS, but can be expressed in the male accessory gland, where MSA is expressed [33]. Thus, even if this form is produced in a significant quantity, it seems that the embryos have further buffered themselves against ectopic abd-A, by repressing its translation.
Lastly, it is of note that even in the elav, fne, mir-iab-8 triple mutant, the derepression of abd-A, while strong, may not be complete. There are a few cells, that still seem to repress abd-A in the posterior CNS (compare Fig 1C or 1F with 1G). At the moment, we cannot explain this result. We believe that some of this change may be due to fact that elav, fne, miR-iab-8 mutant nerve cords are very much abnormal and may have certain cellular defects. We noticed, for example that these nerve cords were more difficult to dissect as they were to the extremely fragile relative to wild type. However, it is also possible that there are additional factors that allow transcriptional readthrough in these embryos or perform a repressive function on abd-A by another mechanism. Interestingly, our RT-qPCR results still seem to detect a low level of transcriptional readthrough even in elav, fne double mutants, hinting that some transcriptional interference might occur even in the absence of these factor. One possible candidate to mediate this transcriptional readthrough is the rbp9 gene, the third elav homologue in Drosophila. Like elav and fne, rbp9 that is expressed in neurons and has been shown to be capable of promoting 3'RNA extensions when ectopically expressed in cultured cells [23].

Transcriptional interference and the Hox clusters
As a mechanism of transcriptional repression, transcriptional interference has mostly been found in organisms with compact genomes like yeast and bacteria. Because most of the multicellular eukaryotes studied in the lab have much larger genomes, containing a large proportion of "non-essential" DNA, transcriptional interference has often been disregarded as a common mechanism for gene repression. However, due to co-regulation and/or gene duplication events, eukaryotic genes may be more compact at certain locations than generally assumed. This is very evident in the HOX gene clusters where there are numerous examples of tightly packed or overlapping transcription units. With all of these examples of overlapping transcription units and possible transcriptional interference, it is interesting to ask if this association could relate to an ancient gene regulatory mechanism. Within the Hox genes there is a known phenomenon called posterior dominance. According to the principle of posterior dominance, the more posterior Hox gene expressed in a segment generally plays the dominant role in patterning the segment. In Drosophila, this is often seen by down-regulation of the more-anterior gene. It is interesting to note that in the most studied Hox clusters, the Hox genes are organized on the chromosome in a way in which each Hox gene is located directly 3' to the next more-posterior segment specifying Hox gene. If we consider that the Hox clusters are thought to have arisen from successive gene duplication events and after such duplication events, the two genes should have equal regulatory potential, then how could the upstream gene consistently take on a more dominant role? Transcriptional interference provides a possible explanation for this. According to this model, the upstream gene might have a slight advantage over the downstream gene due to transcriptional interference. This advantage, although potentially weak in many cases, could then be intensified and fixed through evolving cross-regulatory interactions. In our case, the finding that ectopic abd-A in the posterior CNS leads to female sterility would help to drive such interactions.
Although we have studied this phenomenon in a HOX cluster, other situations might exist where genes are located in similar tight configurations that induce transcriptional interference. An interesting bioinformatic analysis of nested genes in Drosophila suggests that transcriptional interference might be a natural consequence of tight, tandem gene arrangement. In this study, the authors showed that there is a significantly lower number of nested genes transcribed from the same strand in the Drosophila genome [34]. Furthermore, nested gene in the same orientation contained fewer or no introns. Examining the expression data of the tandem, nested genes showed that these genes were often downregulated in tissues where the upstream gene was expressed, leading the author to suggest that the genetic arrangement of the genes might lead to transcriptional interference through mechanisms like unnatural splicing [34]. This is very similar situation to what we find in the Hox complex and may hint that transcriptional interference exists at other loci displaying a similar arrangement of genes. Examining the mechanism that mediates transcriptional interference at model loci like iab-8 may help to define the conditions necessary for transcriptional interference to occur and potential lead to the identification other loci regulated in similar fashion.

Materials and methods
Standard molecular biology techniques were used. Primers and G-blocks used for the construction of plasmids in this study are shown in S1 and S2 Tables.

Fly strains
The Canton S strain was considered as WT for all studies. All strains were raised in standard cornmeal-agar food supplemented with propionic acid and nipagin. The crosses and egg laying for embryo collection were made at 25˚C, and the fly stocks were kept at 18˚C whenever possible.

Construction of the abdA:GFP flies
The plasmid pYex8abdAGFP was first constructed by Gibson Assembly (Gibson, Young et al. 2009), using the "Gibson Assembly Master Mix" kit from NEB (Mass., USA). The backbone plasmid, pY25, a derivative of the pw25 in which the white reporter was replaced by yellow [38]. The Gibson assembly required five fragments: 1. pY25 linearized by Not I and ClaI digestion, 2. Gblock attBexon8flk3' containing an attB site and the region flanking the 3' end of the exon 8 of the iab-8 ncRNA (see S1 Text and S2 Table), ordered from Eurofins MWG Operon, 3. A Kozak sequence and EGFP was amplified from the plasmid pGSA-iab8 amplified with the primers PYexC1GFPFw and PYex8C1ex8Fw (see S1 Text and S1 Table). 4. Exon 8 of the iab-8 ncRNA and initial 140bp of abd-A, was amplified from Drosophila genomic DNA with the primers PYex8C1GFPRv and PYex8C1intFw (see S1 Text and S1 Table), and 5. The region flanking the 5' end of the exon 8 of the iab-8 ncRNA was amplified from Drosophila genomic DNA with the primers PYex8C1ex8Rv and PYex8C1intRv (see S1 Text and S1 Table).
This plasmid was then injected into embryos carrying the del(ex8) line (which contains an attB site in place of the iab-8 exon 8) [33] and the vas-PhiC31 integrase on the X [39].

Is situ hybridization and immunostaining
In situ hybridizations and immunostainings were performed using standard protocols. The creation of all probes was done with the primers listed in S3 Table and the antibodies used in the immunostainings and in situs are listed in S5 Table. Microscopy Fluorescence microscopy images were obtained on a Zeiss LSM710 or LSM800 confocal microscope and treated with the Fiji image analysis software [40] or a Zeiss Axioplan fluorescent microscope with an X-lite 120 lamp. All fluorescent colors presented are pseudo-colors generated by the Fiji software to the different channels collected. Confocal images are generally shown as maximum image projections from stacks of images, unless otherwise noted in the figure legends. Non-Fluorescent images were obtained using a Zeiss axiophot microscope using an Optronics camera and the MagnaFire v2.0 software. Images were cropped and mounted for figures in Microsoft Powerpoint and Adobe Photoshop.

Annotation of the ELAV-binding coordinates in the iab-8 ncRNA transcriptional unit
Chromosomal coordinates provided by [32] converted from the dm3 coordinates to dm6 coordinates (release of 2014) using the UCSC Batch Coordinate conversion tool (https:// genome.ucsc.edu/cgi-bin/hgLiftOver). In the dm6 release, the iab-8 ncRNA coordinates correspond to: chr3R:16831116-16923127. ChIP-seq data for ELAV binding in coordinates comprised on this interval were inspected manually (For coordinates, see S6 and S7 Tables).

Analysis of differentially spliced transcripts
Total RNA from an overnight collection of embryos was used with primers iab8 Ex6-7F (spanning the iab-8 exon 6 to exon 7 junction) and abdA Ex5R (or iab8 Ex8rev) according to the instructions of the Qiagen One-Step RT-PCR kit. Products from the reaction (a mixture of bands, due to the differentially spliced products) were run on agarose gels and cloned into pGemTeasy (Promega Corp, Wisconsin, USA). Different sized inserts were selected and sequenced.

mRNA-seq data processing and visualization
Published mRNA-seq data (GSE146986 [20]) were analyzed using the RNA-seq module from snakePipes [41] with default parameters. Reads were mapped to the dm6 genome annotation (Ensembl release 96) and visualized using IGV_2.8.2.

Embryo collection and RT-qPCR
Embryos were raised and collected as described in Carrasco et al., 2020 [20]. Genotypes of embryos selected for RT-qPCR are: wild type: w 1118 /Y. Δelav: elav CDS20 /Y. ΔelavΔfne: elav CDS20 , fne Δ /Y. For each condition, 10-15 embryos were homogenized in TRIzol (Invitrogen) in 3 biological replicates, and total RNA was extracted according to the manufacturer's instructions. 300ng of total RNA were used for each RT-qPCR reaction. Reverse transcription used iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad). RT-qPCR was performed in a LightCycler 480 II instrument using FastStart SYBR Green Master (Roche). RT-qPCR primer sequences are listed in S1 Text and S4 Table. Control reactions to measure the levels of elav and fne are shown in RT-qPCR quantification of elav and fne coding sequence (CDS) RNAs in wild-type, Δelav, and Δelav Δfne embryos. RNA was extracted from whole embryos 14-16h AEL (after egg laying) and 18-22h AEL. RNA levels were normalized to RpL32 (rp49) mRNA levels. Error bars represent mean ± SD of three biological replicates (10-15 embryos per replicate) for each genotype. (TIFF) S1  Table. Manual annotation of the genomic coordinates of the BX-C region between abd-A and Abd-B where ELAV binding has been detected by ChIP-seq in 6-8h embryos [32].