The insulating activity of the Drosophila BX-C chromatin boundary 1 Fub-1 is parasegmentally regulated by lncRNA read-through 2

Though long non-coding RNAs (lncRNAs) represent a substantial fraction of the Pol II transcripts 26 in multicellular animals, only a few have known functions. Here we report that the blocking 27 activity of the Bithorax complex (BX-C) Fub-1 boundary is segmentally regulated by its own 28 lncRNA. The Fub-1 boundary is located between the Ultrabithorax ( Ubx ) gene and the bxd/pbx 29 regulatory domain, which is responsible for regulating Ubx expression in parasegment 30 PS6/segment A1. Fub-1 consists of two hypersensitive sites, HS1 and HS2 . HS1 is an insulator 31 while HS2 functions primarily as a lncRNA promoter. To activate Ubx expression in PS6/A1 32 enhancers in the bxd/pbx domain must be able to bypass Fub-1 blocking activity. We show that 33 expression of the Fub-1 lncRNAs in PS6/A1 from the HS2 promoter inactivates Fub-1 insulating 34 activity. Inactivation is due to readthrough as the HS2 promoter must be directed towards HS1 to 35 disrupt blocking. 36


INTRODUCTION
The vast majority of Pol II transcripts encoded by the genomes of multicellular animals  In order to properly specify parasegment identity, the nine BX-C regulatory domains must 88 be functionally autonomous. Autonomy is conferred by boundary elements (insulators) that flank 89 each domain (Maeda and Karch, 2015). The role of boundaries is BX-C regulation is best 90 understood in the Abd-B region of the complex. Mutations that inactivate Abd-B boundaries cause 91 a gain-of-function (GOF) transformation in segment identity. For example, the Fab-7 boundary 92 separates the iab-6 and iab-7 regulatory domains which are responsible for specifying PS11(A6) 93 and PS12 (A7) identity respectively (Gyurkovics et al., 1990). When Fab-7 is deleted, the iab-6 94 initiator activates the iab-7 domain in PS11(A6) and iab-7 instead of iab-6 drives Abd-B . For example, Fab-7 has bypass activity and it mediate regulatory interactions between iab-102 6 and Abd-B. However, when Fab-7 is replaced by generic fly boundaries (e.g., scs or su(Hw)) or 103 by BX-C boundaries that lack bypass activity (Mcp) these boundaries block crosstalk but do not 104 support bypass (Hogga et al., 2001;Kyrchanova et al., 2017). As a consequence, Abd-B expression 105 in PS11/A6 is driven by iab-5 not iab-6 and this results in a loss-of-function phenotype.

106
In the Ubx region of the complex, the bxd/pbx regulatory domain is separated from the Ubx 107 gene by a boundary, Fub-1 (Fig. 1)

145
The bxd/pbx regulatory domain: boundaries and lncRNAs 146 The bxd/pbx regulatory domain controls Ubx expression in A1/PS6. This domain is off in      To test these predictions, we used  Fub-1 is not required to block crosstalk between abx/bx and bxd/pbx. 233 A BX-C boundary in the location of Fub-1 would be expected to have two regulatory 234 functions. One would be blocking crosstalk between abx/bx and bxd/pbx while the other would be 235 bypass so that enhancers in the bxd/pbx domain would be able to drive Ubx expression in PS6/A1.

236
To test for blocking activity, we generated a 1168-bp deletion (F1 attP ) that retains an attP site for  The Fub-1 element does not block enhancers in bxd/pbx from activating reporters. 257 Although Fub-1 defines one end-point of a TAD and marks the border for the PcG 258 H3K27me3 histone mark in PS5 nuclei, the finding that Fub-1 deletions have no phenotype raises 259 the possibility that it is actually "within" the bxd/pbx regulatory domain. If this is the case then 260 reporters inserted between Fub-1 and the element just upstream of the Ubx promoter should 261 respond to bxd/pbx enhancers, but be insulated from enhancers in the abx/bx regulatory domain.

262
To test this possibility, we analyzed the pattern of expression of reporters inserted into the F1 attP 263 site with and without the Fub-1 boundary. 264 We first examined the DsRed reporter used to mark the Fub-1 deletion in DsRed-F1attP.   embryos is similar to that of F7 attP50 ( Figure 4C).

367
Read-through transcription from HS2 abrogates HS1 insulator function 368 The finding that fragments containing only HS1 have insulating activity, while the full-369 length Fub-1 fragment does not, suggests that when HS2 is present it inactivates the HS1 boundary.   To test this hypothesis, we inserted a 229-bp SV40 transcription terminator in between HS2 376 and HS1 to give the HS2 505R -SV40-HS1 248R replacement ( Figure 5A). Unlike the starting HS2 505R -377 HS1 248R line, an A6-like segment is present in HS2 505R -SV40-HS1 248R males ( Figure 5B, S4). 378 However, blocking is not fully restored to the level of HS1 248R alone. In HS2 505R -SV40-HS1 248R 379 males, the A6 tergite is slightly reduced in size while the sternite is absent. In the embryonic CNS, 380 Abd-B expression in PS11 is clearly elevated, but less so than either that observed in either 381 HS2 505R -HS1 248R or F7 attP50 ( Figure 5C). To confirm that the partial reactivation of blocking activity 382 is due to a reduction in transcriptional readthrough, we generated a similar construct, HS2 505R -383 mSV40-HS1 248R , in which the polyadenylation sequences in the SV40 fragment were mutated. In 384 contrast to HS2 505R -SV40-HS1 248R , the insulating activity of HS2 505R -mSV40-HS1 248R is 385 substantially compromised and only a residual A6 tergite is observed ( Figure 5B, S4).

386
In these experiments the SV40 termination element will not only suppress transcription 387 through HS1, but also into the neighboring iab-7 regulatory domain. To rule out possibility that 388 the SV40 element "rescues" the blocking activity of HS1 by reducing transcription into the iab-7 389 regulatory domain, we placed it downstream of HS1 instead between HS2 and HS1. Fig. 5B and 390 S4 shows that just like HS2 505R -HS1 248R , HS2 505R -HS1 248R -SV40 has no boundary function: A6 is 391 absent in males carrying the replacement, indicating that cells in PS11/A6 have assumed a 392 PS12/A7 identity. The Abd-B expression pattern in the embryonic CNS also closely matches that 393 of HS2 505R -HS1 248R line (Fig. 5C). Similar results were obtained when the same construct (SV40-394 HS1 249 -HS2 505 ) was inserted in the forward orientation so that the HS2 element is "pointing" 395 towards iab-6 (Fig 5, S4).

396
To provide further evidence that transcriptional read-through from HS2 is responsible for 397 disrupting the blocking activity of HS1, we generated a replacement, HS2 505 -HS1 248R , in which the 398 5'→3' orientation of HS2 was inverted with respect to HS1. If transcription from HS2 is 399 unidirectional, then the blocking activity of HS1 should be unaffected when HS2 is "pointing" 400 away from HS1. Consistent with this prediction, HS1 retains blocking activity in HS2 505 -HS1 248R .   Table S2 for sequences). dCTCF sites in an HS1 Sph I restriction site located close to the endogenous dCTCF site. Fig 5B   423 shows that the addition of 4 dCTCF sites (HS2 505R -HS1 248R -CTCF x4 ) is sufficient to rescue the 424 blocking activity of HS1. As observed for HS1 alone, HS2 505R -HS1 248R -CTCF x4 males have a 425 tergite and sternite whose morphology resembles that seen in A5. Likewise, the pattern of Abd-B 426 expression in the CNS in HS2 505R -HS1 248R -CTCF x4 is similar to that observed for HS 248 (Fig. 5C).  On the other hand, antisense (proximal to distal) transcripts from this region are not 443 detected in WT embryos. Thus, it would be possible to assay the promoter activity of HS2 by 444 reversing its orientation in the genome. For this purpose, we generated two different Fub-1 445 replacements in which the 5'→3' orientation of the HS2 element was reversed so that it would 446 generate "antisense" transcripts. In the first, we inserted a fragment, HS1 248 -HS2 505R , into F1 attP 447 that contains both Fub-1 hypersensitive sites (Fig.5D). However, the 5'→3' orientation of HS2 in 448 this fragment was reversed so that it is pointing away from HS1. In this case transcripts expressed 449 from the HS2 promoter would extend into the bxd/pbx domain, instead of transcribing through 450 HS1. In the second, we inserted a Fub-1 fragment in F1 attP containing both hypersensitive sites, 451 HS2 505R -HS1 248R , but in the reverse orientation so that transcription from HS2 would read-through 452 HS1 towards the bxd/pbx regulatory domain (Fig.5D).

453
Antisense transcripts in WT, HS1 248 -HS2 505R and HS2 505R -HS1 248R embryos were then 454 visualized using smFISH probes spanning a 1463 bp bxd sequence located just distal to the attP 455 site (Fig.5D). Fig. 6 shows that antisense transcripts complementary to the smFISH probes are not 456 detected in WT embryos. In contrast, segmentally restricted antisense transcripts are expressed in 457 both of the Fub-1 replacements. Like the DsRed and GFP reporters described above (Fig. 3) interactions, and at least the initial activation in PS6 may depend upon the enhancers thought to be 467 located between the Ubx promoter and Fub-1. 468 Transcriptional read through is required to abrogate HS1 blocking activity in PS5/A1 469 The second prediction is that transcriptional read-through from HS2 into HS1 is required 470 to relieve the blocking activity of HS1, enabling regulatory interactions between the bxd/pbx 471 domain and Ubx in PS5/A1. This prediction also holds. As shown in Fig. 2, (Fig.5E). This replacement transforms the first abdominal segment into the third thoracic: the first 477 abdominal tergite is reduced or absent, in addition, the posterior third thoracic segment is partly 478 transformed to a posterior second thoracic segmentthe halters enlarged and pointing downward 479 (Fig. 2B). These phenotypes are known as bithoraxoid (bxd) and postbithorax (pbx) respectively 480 (Bender et al., 1983). HS1 248 -HS2 505R homozygotes also display very low viability and sterility.  However, in order to direct the proper expression of their target homeotic gene, all but three of the 512 regulatory domains must be able to bypass one or more intervening boundary elements. In the case 513 of the Ubx gene, the bxd/pbx regulatory domain is separated from its target promoter by the Fub-514 1 boundary, and by a second boundary element located close to the Ubx promoter.

515
The Fub-1 boundary is subdivided into two elements, HS1 and HS2. HS1 like other 516 boundaries in BX-C has a dCTCF site, is bound by dCTCF in vivo and is able to function as a 517 generic insulator. While HS2 may contribute to the boundary function of Fub-1, it does not have 518 insulating activity on its own. Instead, it has promoter activity, and can function as an enhancer 519 trap. In its normal context, the HS2 promoter is controlled by the bxd/pbx regulatory domain. The transcription from this promoter through HS1 disrupts Fub-1 boundary function (Fig. 7). This transcription termination polyadenylation element in between HS2 and HS1 partially rescues the 539 blocking defect. Rescuing activity is not due to the increased distance between HS2 and HS1 as a 540 SV40 element with mutations in the polyadenylation signal significantly compromising its 541 rescuing activity. In addition, the SV40 termination element must be located between HS1 and 542 HS2 as rescue is not observed when it is placed downstream of HS1. Second, when HS2 is inverted 543 so that transcription proceeds away from HS1, boundary function is also restored.

544
A similar mechanism appears to regulate Fub-1 boundary activity in its endogenous 545 context; however, in this case boundary activity is segmentally regulated. In PS5 and more anterior 546 parasegments, the bxd/pbx regulatory domain is maintained in an "off" state by a PcG based 547 mechanism. As a consequence, the Fub-1 HS2 promoter is inactive (Fig.7). In PS6 (A1) (and more 548 posterior parasgements) the bxd/pbx domain is in the "on" state and it activates transcription from 549 the Fub-1 HS2 promoter. When HS2 is oriented so that transcription proceeds through HS1, 550 boundary activity is abrogated in PS6/A1 and more anterior parasegments/segments (Fig. 7). This domain are unable to activate Ubx expression (Fig. 7). This leads to a LOF phenotype much like 557 that observed in PS11/A6 in Fab-7 replacement experiments when the HS1 boundary blocks 558 crosstalk but does not support bypass.  forth from an inactive and active state (Fig.7).

651
A related issue would be the mechanism activating HS2 transcription when the Fub-1 652 boundary is reversed (the HS1 248 -HS2 505R replacement) so that HS1 is interposed between HS2 and PS12/A7 is also observed when the Fub-1 replacement is inserted so that HS2 is next to iab-7.

668
Since iab-7 is "off" in PS11, this would imply that HS2 must be activated by the iab-6 regulatory  The strategy of the Fab-7 replacement lines creation is described in detail in .  was quenched by addition of 3.7mL of 2M Tris-HCl pH7.5 for 5 mins, washed twice with PBST.

778
Embryos were snap-frozen and stored at -80℃ until library construction.

779
Micro-C libraries were prepared as previously described  with the following 780 modification: we used 50uL of 12-16hr embryos, non-sorted for each biological replicate. 60U of 781 MNase was used for each reaction to digest chromatin to 80% mononucleosome vs 20% 782 dinucleosome ratio. Libraries were barcoded, pooled and subjected to paired-end sequencing on 783 an Illumina Novaseq S1 100 nt Flowcell (read length 50 bases per mate, 6-base index read).