A boundary-defining protein facilitates megabase-scale regulatory chromosomal loop formation in Drosophila neurons

In this study, Mouginot et al. describe the formation and function of meta-domains, megabase-range chromatin contacts that regulate neuronal gene expression in Drosophila. They identify Cp190 and Lola-I as meta-loop anchor-bound proteins that mediate meta-loop formation and underpin the complexity and specificity of neuron-specific genome architecture.

Abstract

Regulatory elements, such as enhancers and silencers, control transcription by establishing physical proximity to target gene promoters. Neurons in flies and mammals exhibit long-range three-dimensional genome contacts, proposed to connect genes with distal regulatory elements. However, the relevance of these contacts for neuronal gene transcription and the mechanisms underlying their specificity necessitate further investigation. Here, we precisely disrupt several long-range contacts in fly neurons, demonstrating their importance for megabase-range gene regulation and uncovering a hierarchical process in their formation. We further reveal an essential role for the chromosomal boundary-forming protein Cp190 in anchoring many long-range contacts, highlighting a mechanistic interplay between boundary and loop formation. Finally, we develop an unbiased proteomics-based method to systematically identify factors required for specific long-range contacts. Our findings underscore the essential role of architectural proteins such as Cp190 in cell type-specific genome organization in enabling specialized neuronal transcriptional programs.

Gene transcription in multicellular organisms occurs in specific spatiotemporal patterns controlled by regulatory elements, such as enhancers and silencers, that activate or silence transcription. Functional communication between regulatory elements and target promoters depends on their physical proximity, though proximity does not always correlate with regulation. For this reason, gene regulation typically occurs locally within topologically associating domains (TADs), dynamic structures spanning tens of kilobases in flies and hundreds of kilobases in mammals, where regulatory elements and promoters interact more frequently with each other than with those in another TAD.

However, recent studies of genome folding—mainly using chromosome conformation capture techniques to generate genome-wide maps of chemical cross-linking frequencies between pairs of genomic loci at restriction fragment (in Hi-C) or nucleosome (in Micro-C) resolution—have revealed long-range physical interactions between loci in distant TADs. These interactions are mediated by various mechanisms, including cohesin-driven extrusion of chromosomal loops, compartmentalization of active and inactive chromatin, and direct tethering of distant interacting loci (Zunjarrao and Gambetta 2025). Importantly, some distant interactions are regulatory, demonstrating that gene regulation can also occur across several TADs.

In particular, fly and mammalian neurons exhibit striking long-range gene contacts, suggesting that genome folding may enable specialized transcriptional programs critical for neuronal function (Bonev et al. 2017; Monahan et al. 2019; Winick-Ng et al. 2021; Mohana et al. 2023; Pourmorady et al. 2024). For example, mammalian neurons organize olfactory receptor (OR) genes into a silent heterochromatic hub or a multichromosomal enhancer hub containing a single OR gene chosen for expression (Clowney et al. 2012; Monahan et al. 2019; Pourmorady et al. 2024). Although these hubs are essential to transcriptionally activate a single OR gene, their complexity likely underlies functional redundancy, complicating the identification of specific ultra-long-range regulatory interactions.

In fly neurons, we reported that specific pairs of TADs interact over megabases during neuronal differentiation to form higher-order structures called meta-domains (Mohana et al. 2023). Within meta-domains, promoters of genes involved in axon guidance and signaling loop to each other or to noncoding DNA elements in the distant TAD. We refer to these megabase-range punctual interactions as meta-loops, and to the interacting loci as meta-loop anchors. Meta-loop anchors are classified as promoter anchors when within ±200 bp of a transcription start site, or otherwise as intergenic anchors. Unlike other known long-range gene associations, Drosophila meta-loop interactions are simple (predominantly pairwise) and highly reproducible in single neurons. Many meta-loops are conserved over 40 million years of Drosophila evolution, suggesting that they are under selection and hence functional. These characteristics establish Drosophila meta-loops as a unique paradigm for understanding long-range regulatory interactions in a neuronal genome.

By precisely deleting an intergenic anchor that loops over 1.6 Mb to the promoter of the no long nerve cord (nolo) gene, we demonstrated that full transcriptional activation of nolo requires the intergenic anchor (Mohana et al. 2023). By mutating transcription factors such as CTCF, which binds to and is required to form a subset of meta-loops, we showed that CTCF is required for the timely transcriptional activation of beaten path IV (beat-IV) and Multiplexin (Mp), both of which are involved in axon guidance (Mohana et al. 2023). However, it remains unclear whether transcriptional defects observed in CTCF mutants result from the loss of meta-loops in the beat-IV and Mp meta-domains or from the loss of CTCF's functions beyond its architectural role in meta-loop formation. Thus, the functional relevance of meta-loops in long-range neuronal gene regulation requires further investigation.

Apart from their function, another basic question remains regarding how meta-loop anchors pair with such remarkable specificity. Elucidating the mechanism of meta-loop formation is important to understand how precise contacts are established between genes and distant regulatory elements. Through Micro-C analyses of larval central nervous systems (CNSs), we identified 58 meta-loops and 28 meta-domains (Mohana et al. 2023). The fact that most meta-domains contain two or more meta-loops hints at potential cooperativity in the emergence of meta-loops in the same meta-domain. However, the contribution of the TADs themselves for meta-loop emergence remains unclear.

Which proteins mediate meta-loop formation also remains largely unknown. It is unclear how DNA-binding proteins, each of which typically binds >1000 loci, confer specificity to the pairing of the meta-loop anchors they bind. To date, two DNA-binding proteins, CTCF and GAF, have been shown to bind and be required to form specific subsets of meta-loops, together accounting for only 14% (eight out of 58) of meta-loops (Mohana et al. 2023).

CTCF, known primarily for its role in forming TAD boundaries that limit regulatory cross-talk in flies and mammals (Nora et al. 2017; Wutz et al. 2017; Kaushal et al. 2021; Chakraborty et al. 2025; Tonelli et al. 2025), collaborates with a cofactor in Drosophila called centrosomal protein 190 kDa (Cp190), which CTCF requires to form robust physical and regulatory boundaries (Kaushal et al. 2021, 2022). Cp190 is recruited by CTCF and other DNA-binding proteins and is essential to form most TAD boundaries that form independently of transcription, making it the primary TAD boundary-forming protein in Drosophila (Kaushal et al. 2022). This raises the question of whether Cp190 is required for CTCF-dependent and additional meta-loops in fly neurons.

Finally, it is important to note that all meta-loop-forming proteins identified so far have emerged through candidate-based approaches, highlighting the need for an unbiased method to identify additional factors.

Here, we clarify the functional relevance of meta-domains by precisely perturbing meta-loops individually or in combination in three selected meta-domains. We demonstrated that meta-loops regulate neuronal genes over 5.1 Mb, representing the longest-range natural regulatory interactions identified in a genome. We found that two primary functions of meta-domains are either to enable efficient transcriptional activation of neuronal genes by regulatory elements in a distal TAD or to facilitate cotranscription of two neuronal promoters connected by a meta-loop.

In terms of how meta-domains form, we uncovered a hierarchical process of meta-loop formation within several meta-domains. We also discovered a critical interplay between Cp190-dependent TAD boundaries and meta-loop anchors, whereby Cp190 is required to form over half of all meta-loops. Finally, we developed a proteomics-based method for unbiased identification of meta-loop anchor-associated proteins, leading to the characterization of a specific isoform of a protein known for its role in axon guidance and signaling. This isoform binds to half of all meta-loop anchors in vivo and is required to form a small subset of meta-loops.

Our work establishes a framework for systematic investigations of proteins mediating meta-loop formation, essential for understanding the molecular basis of their remarkable specificity. These insights underscore the importance and complexity of neuronal-specific genome folding, driven by architectural proteins, for enabling robust neuronal transcriptional programs.

Results

Meta-loops can form hierarchically or independently

Most (19 out of 28; 68%) meta-domains in larval CNSs contain multiple (two to four) meta-loops detected by Micro-C (Mohana et al. 2023). In three meta-domains analyzed here, each contains an intergenic–intergenic (I-I) meta-loop and a second meta-loop connecting either the promoters of Mp or beat-IV to an intergenic anchor or the promoters of glutamate receptors IA (GluRIA) and IB (GluRIB), encoding glutamate-gated ion channels at excitatory synapses, to each other (Fig. 1A–D). Chromatin at the I-I loop anchors becomes accessible earlier in embryonic neuronal differentiation, as shown by published DNase sequencing on purified embryonic neurons (Supplemental Fig. S1A–C; Reddington et al. 2020). Published Hi-C data on neuronal nuclei from early (6–8 h), mid (10–12 h), and late (14–16 h) embryos (Mohana et al. 2023) further suggest that I-I meta-loops form first, preceding intergenic–promoter (I-P), and promoter–promoter (P-P) meta-loops in the same meta-domain (Supplemental Fig. S1A–C). This sequential formation raises the possibility that I-I meta-loops may nucleate the later formation of promoter-involved meta-loops.

Figure 1.

Meta-loop requirements for meta-domain formation and neuronal gene expression. (A) Cartoon Hi-C maps at the chromosome level (left) and a zoomed-in view of a meta-domain (beat-IV; right). (B–D) Micro-C maps of WT third-instar larval CNSs showing the Mp (B), beat-IV (C), or GluRIA–GluRIB (D) meta-domains. (Bottom) Zoom-in on left (X-axis) and right (Y-axis) anchors of meta-loops (black for I-I anchors, green for I-P or P-P anchors; numbered as in Mohana et al. 2023) with DNA-FISH probes (gray), gene tracks (longest isoforms colored by transcription direction), published pseudobulk ATAC-seq in larval CNSs (Mohana et al. 2023), and anchor deletions. (E–G) Violin plots (central horizontal lines mark medians; boxes mark interquartile ranges) of distances (in micrometers; Y-axis) between meta-loop anchors measured by DNA-FISH in n ELAV⁺ neuronal nuclei (blue) and neighboring nonneuronal nuclei (gray) of nerve cords of N independent embryos for the indicated genotypes (X-axis). Percentages of nuclei with colocalized anchors (<250 nm apart) are indicated. (H–J) RT-qPCR fold changes in Mp (H), beat-IV (I), or GluRIA or GluRIB (J) mRNA levels (normalized to the RpL15 housekeeping internal control gene) in adult heads of the same genotypes as in E–G. See also Supplemental Figures S1 and S2 and Supplemental Table S1.

To test this hypothesis, we generated single or double deletions (199–590 bp) of accessible chromatin peaks at meta-loop anchors in three studied meta-domains and assessed their impact on meta-domain formation using DNA fluorescence in situ hybridization (DNA-FISH) with oligo-FISH probes (Supplemental Table S1) targeting each anchor in late (stage 16) embryos. In wild-type (WT) embryos, DNA-FISH confirmed the presence of meta-domains containing Mp, beat-IV, or GluRIA–GluRIB in neuronal nuclei but not in neighboring, nonneuronal nuclei (Fig. 1E–G; Supplemental Fig. S2A,B). We verified that differences in relative distances between meta-loop anchors in neurons versus nonneurons were not explained by the small difference in nuclear size between these two populations (Supplemental Fig. S2C,D).

Deleting an intergenic anchor of the I-I L53 meta-loop (L53^ΔA63) strongly destabilized the beat-IV meta-domain and impaired subsequent I-P meta-loop formation involving the beat-IV promoter (Fig. 1F). Similarly, deleting an intergenic anchor of the I-I L36 meta-loop (L36^ΔA46) strongly destabilized the GluRIA–GluRIB meta-domain and impaired P-P meta-loop formation (Fig. 1G). In contrast, deleting an intergenic anchor of the I-I L34 meta-loop (L34^ΔA39) did not affect the Mp meta-domain (Fig. 1E). Partial destabilization of the Mp meta-domain occurred with deletion of the I-P L33 meta-loop (L33^ΔA38), and full disruption required combined deletions of both the I-I and I-P anchors (L33^ΔA38 L34^ΔA39) (Fig. 1E).

We conclude that meta-loop formation can be hierarchical, depending on the meta-domain. For beat-IV and GluRIA–GluRIB meta-domains, I-I meta-loops formed early in neuronal differentiation nucleate later I-P or P-P meta-loops. In contrast, the Mp meta-domain is redundantly formed by independent I-I and I-P meta-loops.

beat-IV and Mp meta-domains boost transcription of genes at meta-loop anchors

Some single or double meta-loop anchor deletions caused specific developmental defects, such as bristle abnormalities in L53^ΔA63 (Supplemental Fig. S1D) or female sterility in L54^ΔA65, indicating potential impacts on gene transcription. To investigate whether genes at meta-loop anchors are misexpressed when the meta-loops they engage in, or the meta-domains they are in, are disrupted, we quantified neuronal gene expression using reverse transcriptase qPCR (RT-qPCR) on total RNA from adult heads, bypassing the challenges of measuring rapidly changing embryonic gene expression.

In the Mp meta-domain, deleting the I-P meta-loop anchor (L33^ΔA38) partially destabilized the Mp meta-domain and reduced Mp expression by ∼40%, whereas deleting the I-I meta-loop anchor (L34^ΔA39) had no effect on the meta-domain and only slightly reduced Mp mRNA levels (Fig. 1E,H). The double deletion (L33^ΔA38 L34^ΔA39) reduced Mp mRNA levels to an extent similar to that of L33^ΔA38 alone (Fig. 1H). These findings suggest that the intergenic anchor to which the Mp promoter loops (L33^A38) is critical for efficient Mp transcriptional activation, potentially by overlapping a distant Mp enhancer or partially destabilizing the meta-domain. The lack of further transcriptional reduction after complete meta-domain disruption supports the idea that L33^A38 may be a key distant enhancer or that partial meta-domain destabilization is sufficient to impair transcription. Overall, these results indicate that Mp is transcriptionally regulated over 2 Mb.

In the beat-IV meta-domain, deletion of the I-I meta-loop anchor (L53^ΔA63) or the I-P meta-loop anchor (L54^ΔA65) each reduced beat-IV expression by ∼30%, whereas the double deletion (L53^ΔA63 L54^ΔA65) reduced expression by ∼60% (Fig. 1I). Reduced beat-IV expression in L54^ΔA65 with an otherwise intact beat-IV meta-domain suggests that L54^A65 may act as a beat-IV distant enhancer (Fig. 1F,I). Previous transgenic enhancer assays demonstrated that L54^A65 can drive neuronal transcription (Mohana et al. 2023), supporting its potential role as an enhancer activating beat-IV over 5.1 Mb. Similarly reduced beat-IV expression in L53^ΔA63, which strongly destabilized the meta-domain, suggests that strong meta-domain destabilization impairs transcriptional activation (Fig. 1F,I). However, we cannot rule out that L53^ΔA63 deletion impacts transcription of beat-IV, located 18 kb away, independently of the meta-domain. We note that our attempts to delete L53^A66 instead of L53^A63 were unsuccessful despite having tested eight independent guide RNAs.

In the GluRIA–GluRIB meta-domain, deletion of the I-I meta-loop anchor (L36^ΔA46) disrupted the meta-domain but did not impair GluRIA or GluRIB expression, which instead showed a slight upregulation (Fig. 1G,J).

We conclude that the beat-IV and Mp meta-domains facilitate transcriptional activation over megabases, with intergenic anchors directly looping to their promoters playing a role in this activation (Fig. 1H,I). In contrast, the GluRIA–GluRIB meta-domain appears dispensable for transcriptional activation of these genes.

The GluRIA–GluRIB meta-domain facilitates paralogous gene cotranscription

We hypothesized that rather than enabling high levels of gene expression, the GluRIA–GluRIB meta-domain may facilitate cotranscription of these paralogs. This hypothesis stems from recent findings that intra-TAD chromatin loops linking paralogous genes facilitate their cobursting in fly embryos (Levo et al. 2022) and that paralogous gene loops enable cross-regulation by shared regulatory elements (Mohana et al. 2023; Pollex et al. 2024). To test this, we performed two-color single-molecule RNA-FISH (smRNA-FISH) on late (stage 16) embryonic nerve cords and found that GluRIA and GluRIB were more frequently cotranscribed in WT than in L36^ΔA46 mutants (Fig. 2A,B). In WT embryos, GluRIA and GluRIB transcription foci often colocalized, a feature lost in L36^ΔA46 nuclei (Fig. 2C).

Figure 2.

The GluRIA–GluRIB meta-domain facilitates paralogous gene cotranscription. (A) smRNA-FISH images of GluRIA (green) and GluRIB (pink) mRNAs in WT (top) and L36^ΔA46 (bottom) embryonic nerve cords. Arrowheads mark colocalized GluRIA and GluRIB transcription foci. (B) The percentage of neurons exhibiting a nascent transcription site for GluRIA, GluRIB, or both, shown relative to either the total neuron count (left bar plot) or neurons with nascent foci for at least one of the two genes (right bar plot). Data are presented for WT and L36^ΔA46. (C) Violin plots of distances (in micrometers; Y-axis) between GluRIA and GluRIB transcription sites in embryonic nerve cord nuclei coexpressing both genes in the indicated genotypes (X-axis).

We conclude that the GluRIA–GluRIB meta-domain primarily supports cotranscription, potentially through direct sharing of transcription factors rather than high levels of expression.

Cp190 is essential for forming CTCF-dependent and additional meta-loops

The importance of I-I meta-loops for forming some meta-domains (Fig. 1F,G) underscores the role of structural proteins binding to intergenic anchors in meta-domain formation and function. Approximately 50% of meta-loop anchors overlap with TAD boundaries in larval CNS Hi-C maps (Fig. 3A), and we previously demonstrated that CTCF is required to form a subset of these TAD boundaries (Kaushal et al. 2021) and meta-loops (Mohana et al. 2023). Cp190—a Broad complex, Tramtrack, and Bric-a-brac (BTB; also known as POZ) zinc finger (ZnF) protein—is essential to form most TAD boundaries distal to transcribed promoters, unlike TAD boundaries overlapping highly transcribed promoters, which form independently of Cp190 (Kaushal et al. 2022). In larval CNSs, TAD boundaries overlapping meta-loop anchors are generally distal to transcribed promoters and likely form via architectural proteins independently of transcription (Fig. 3A).

Figure 3.

Cp190 is essential for forming CTCF-dependent and additional meta-loops. (A, right) Distribution of TAD boundaries called in WT larval CNSs (first lane), Cp190 or CTCF ChIP-seq peaks or differential Cp190 ChIP-seq occupancy in CTCF⁰ versus WT larval CNSs (second through fourth lanes, colored by average best.logFC), and expressed transcription start sites in WT larval CNSs (fifth lane), in 1 kb bins ±25 kb around all meta-loop anchors (rows) ordered by distance to the nearest TAD boundary (data are from Kaushal et al. 2021). Average signals across all anchors are shown above. (Left) Classification of anchors as intergenic (I; gray), promoter (P; blue), or Cp190-dependent (pink) or Cp190-independent (gray). (B) Differential analysis of strengths of loops involving meta-loop anchors (dots) measured by Hi-C in CTCF⁰ (top) or Cp190⁰ (bottom) versus WT 16–24 h embryo neurons. Labeled loops (red dots) were significantly weakened (Padj ≤ 0.05, |fold change| ≥ 2.5). (C) Overlap between all meta-loops and those dependent on CTCF and/or Cp190. (D,E) Same as Figure 1, B–D, but showing Hi-C maps of neurons from 16–24 h embryos of the indicated genotypes, with published CTCF and Cp190 ChIP-seq tracks and called peaks in larval CNSs (Kaushal et al. 2021) and single-cell ATAC-seq (sci-ATAC-seq3) in 16–18 h embryonic nerve cords (Calderon et al. 2022), showing a CTCF- and Cp190-dependent meta-domain (Mp; D) and a Cp190-dependent, CTCF-independent meta-domain (E). See also Supplemental Figure S3 and Supplemental Table S2.

To investigate whether TAD boundary-forming proteins contribute to meta-loop formation, we examined Cp190 binding. Published Cp190 chromatin immunoprecipitation sequencing (ChIP-seq) data in WT larval CNSs (Kaushal et al. 2021) showed that Cp190 binds within ±500 bp of 41% of meta-loop anchors, often overlapping TAD boundaries (Fig. 3A). CTCF recruits Cp190 to CTCF binding sites (Fig. 3A) to establish robust physical and regulatory boundaries (Kaushal et al. 2021, 2022), raising the question of whether Cp190 is essential for forming CTCF-dependent meta-loops. Cp190 is recruited to additional loci by DNA-binding proteins other than CTCF, raising the question of whether Cp190 contributes to additional meta-loops beyond those mediated by CTCF.

CTCF⁰ and Cp190⁰ mutants, which completely lack both maternal and zygotic CTCF or Cp190 protein, arrest development at pupal or early larval stages, respectively (Kaushal et al. 2021, 2022). To test whether Cp190 is required for meta-loop formation, we therefore performed Hi-C on neurons from old (16–24 h) WT, CTCF⁰, or Cp190⁰ embryos. Because CTCF is known to form three meta-domains in larval CNSs (Mohana et al. 2023), it served as a positive control for analyzing meta-loop formation in embryonic neurons.

Neurons were isolated by fluorescence-activated cell sorting (FACS) from fixed embryos using an ELAV neuronal marker antibody (Supplemental Fig. S3A). Both CTCF⁰ and Cp190⁰ embryos exhibited normal neuronal counts (Supplemental Fig. S3A), suggesting that neuronal differentiation was unaffected in these mutants. Hi-C was then performed on these neurons (Supplemental Fig. S3B; Supplemental Table S2).

Differential analysis of meta-loop strength in CTCF⁰ versus WT embryonic neurons revealed six disrupted meta-loops, consistent with those previously identified in CTCF⁰ larval CNSs (CTCF-dependent meta-loops) (Fig. 3B; Mohana et al. 2023). In contrast, analysis of Cp190⁰ neurons showed a broader weakening of meta-loops to different degrees (Fig. 3B), including all CTCF-dependent meta-loops and additional CTCF-independent meta-loops (Fig. 3B–E; Supplemental Fig. S3C–E). Guided by visual inspection of affected meta-loops, 30 out of 58 (52%) meta-loops were classified as Cp190-dependent, though additional meta-loops were also visibly affected (Fig. 3B).

Among Cp190-dependent meta-loops, most (25 out of 30; 83%) were either directly bound by Cp190 at both anchors (15 out of 30; 50%) or located within meta-domains containing other Cp190-bound and Cp190-dependent meta-loops (10 out of 30; 33%) (Fig. 3D,E; Supplemental Fig. S3D,E). This explains why some Cp190-dependent meta-loops lack direct Cp190 binding (Fig. 3A) and suggests that in several meta-domains, formation of a Cp190-dependent meta-loop nucleates additional meta-loops with non-Cp190-bound anchors, as we demonstrated for the beat-IV meta-domain (Fig. 1F). We also demonstrated this for the L23–L24 meta-domain, which relies on Cp190 even though Cp190 only binds to the anchors of L24 (Fig. 3E).

We conclude that Cp190 is essential for forming both CTCF-dependent meta-loops and additional Cp190-dependent meta-loops formed independently of CTCF. Overall, Cp190 is required for the efficient formation of approximately half of all meta-loops in Drosophila neurons, highlighting its key role in meta-loop formation.

Unbiased identification of meta-loop anchor-associated proteins identifies Lola-I as a candidate factor for meta-loop formation

Because Cp190 does not bind DNA sequence-specifically (Kaushal et al. 2022), we sought to identify other sequence-specific DNA-binding proteins, aside from CTCF, that might recruit Cp190 to Cp190-dependent meta-loop anchors or synergize with Cp190 in another way. To do this, we used mass spectrometry to identify proteins associated with accessible chromatin at meta-loop anchors. We analyzed intergenic anchors of a CTCF- and Cp190-dependent meta-loop (L34) (Fig. 3D) as control and a Cp190-dependent, CTCF-independent meta-loop (L24) (Fig. 3E) as a test. We adapted a previously published strategy (Baldi et al. 2018) in which ∼400 bp bait DNA was PCR-amplified, concatemerized, biotinylated, immobilized on streptavidin beads, chromatinized (i.e., wrapped into nucleosomes) in vitro with young Drosophila embryo extract, and then incubated with nuclear protein extract prepared from adult fly heads expected to contain factors required for meta-loop formation (Fig. 4A; Supplemental Fig. S4A). Proteins enriched at meta-loop anchor baits were identified by mass spectrometry and compared with a negative control DNA sequence known to bind Phaser (Phs), a nucleosome-positioning protein (Supplemental Table S3; Baldi et al. 2018). The top enrichments of Phs in the negative control pull-down and of CTCF in the L34 meta-loop anchor pull-downs validated the method (Fig. 4B).

Figure 4.

Lola-I copurifies with certain meta-loop anchors and is required for axon guidance. (A) Workflow for unbiased identification of meta-loop-associated proteins. (B,C) Volcano plots showing fold change (X-axis) and P-value (Y-axis) of proteins enriched at intergenic anchors of the CTCF- and Cp190-dependent meta-loop L34 (B) or the Cp190-dependent, CTCF-independent meta-loop L24 (C) compared with the Phs control bait in triplicate pull-down experiments. (D) lola gene structure with common BTB-encoding exons (orange) spliced to ZnF-encoding exons (green) in isoforms (rows; gray indicates noncoding exons). Blue lines mark peptides enriched (>10-fold) in any meta-loop anchor pull-down relative to the Phs control. Red shading indicates deleted regions in lola-I^KO and lola-G^KO. Published DNA-binding motifs for different isoforms (Enuameh et al. 2013) are shown at the right. (E) Percentages of the indicated genotypes (X-axis) that completed the indicated developmental transitions across biological replicates (dots). Horizontal lines indicate means. (F) α-Tropomyosin (green) and α-Fasciclin II (red) immunostaining of old (stage 16) embryos of the indicated genotypes (columns) and merged images with DAPI-stained DNA (blue). See also Supplemental Figure S4 and Supplemental Table S3.

Of note, the TAD boundary-associated insulator proteins Ibf1 and Ibf2 (Ramírez et al. 2018) were highly enriched in L24 meta-loop anchor pull-downs but less so in L34 meta-loop anchor pull-downs (Fig. 4B,C). Ibf1 and Ibf2 bind DNA in a sequence-specific manner at sites largely distinct from CTCF (Ramírez et al. 2018). They copurify with Cp190 and are thought to recruit Cp190 to their binding sites (Cuartero et al. 2014). This suggests that Ibf1/Ibf2 may function similarly to CTCF in recruiting Cp190 to meta-loop anchors in fly neurons. However, instead of acting at CTCF-dependent meta-loops, they would do so at a distinct set of Ibf1/Ibf2-bound meta-loop anchors.

Among other interesting candidate proteins, a specific isoform (called “I”) of longitudinal lines lacking (Lola) was highly enriched in pull-downs with anchor A39 of L34 and with anchors A28 and A32 of L24 (Fig. 4B,C; Supplemental Table S3). Lola is a BTB ZnF protein, similar to Cp190, that plays a crucial role in guiding CNS axons and certain peripheral motoneurons (Seeger et al. 1993; Giniger et al. 1994; Dinges et al. 2017). Lola likely regulates this process by controlling the expression of axonal growth and guidance proteins (Crowner et al. 2002; Dinges et al. 2017). We previously found that Lola copurifies with Cp190 from nuclear extracts of Drosophila embryos (Kaushal et al. 2021). This prompted us to investigate whether Lola is required for meta-loop formation.

Lola is alternatively spliced, with universal exons encoding a shared N-terminal BTB protein–protein interaction domain spliced to diverse C-terminal ZnF DNA-binding domain-encoding exons (Fig. 4D; Goeke et al. 2003). Proteomic analysis of Lola peptides copurifying with meta-loop anchors identified the Lola-I, Lola-G, and Lola-Y isoforms (Fig. 4D). Lola-I and Lola-G share conserved ZnF domains (Supplemental Fig. S4B; Goeke et al. 2003) and recognize a similar DNA motif, distinct from other Lola isoforms, as shown by published bacterial one-hybrid experiments (Fig. 4D; Enuameh et al. 2013). We therefore further investigated Lola-I and Lola-G as potential meta-looping factors, though we cannot rule out that Lola-Y (and potentially other Lola isoforms not identified in our pull-downs) may also be relevant for meta-loop formation.

Different lola isoforms are expressed in different patterns in the embryo and are thought to control different axon guidance decisions in the nerve cord or peripherally (Goeke et al. 2003). To assess the relevance of Lola-I and Lola-G for meta-loop formation, we first reexamined their expression in embryos using isoform-specific RNA-FISH. Consistent with the original report (Goeke et al. 2003), lola-I was strongly enriched in the nervous system, whereas lola-G seemed enriched in the mesoderm (Supplemental Fig. S4C). This suggests that Lola-I may be more likely to play a role in neuronal meta-loop formation despite enrichment of both isoforms in the in vitro anchor pull-down.

To compare the lethal stages and developmental phenotypes of lola-I and lola-G mutants, we generated isoform-specific knockouts by deleting their ZnF-encoding exons using CRISPR/Cas9-mediated genome editing (Fig. 4D). Animals transheterozygous for two independent knockouts (lola-I^KO and lola-G^KO) were analyzed. Only ∼25% of lola-I^KO and ∼50% of lola-G^KO embryos hatched into larvae, with none (lola-I^KO) or ∼10% (lola-G^KO) developing into pupae (Fig. 4E). Expressing a hemagglutinin (HA)-tagged lola-I transgene under the control of lola-I 5′ and 3′ genomic sequences rescued lola-I^KO animals into viable and fertile adults (Fig. 4E).

Immunostaining muscles (α-tropomyosin) and axons (α-Fasciclin 2) revealed neuronal branching and muscle organization defects in both lola-I^KO and lola-G^KO mutants compared with WT controls (Fig. 4F), consistent with previous reports that lola-I mutants have defective motoneuron pathfinding (Goeke et al. 2003; Dinges et al. 2017). These findings raise the possibility that lola-I and lola-G may play key roles in neurons and muscle precursors, respectively, to ensure proper neuromuscular junction formation.

We conclude that Lola-I copurifies with a meta-loop anchor DNA bait (Fig. 4A–D), is enriched in the nervous system (Supplemental Fig. S4C), and plays a critical role in axon guidance (Fig. 4E,F), warranting further investigation into its potential role in higher-order genome folding in neurons.

Lola-I is required to form a single Cp190-dependent meta-domain

To assess the prevalence of Lola-I binding at meta-loop anchors, we mapped Lola-I binding sites in larval CNSs by anti-HA ChIP-seq in lola-I^KO mutants rescued by HA-lola-I cDNA driven by lola regulatory sequences (Fig. 4E). We found that 3412 Lola-I ChIP peaks were defined as enriched in HA-Lola-I-rescued animals compared with WT animals lacking the HA-lola-I transgene. Lola-I bound within ±500 bp of 47% of meta-loop anchors (Fig. 5A), including the L24 anchors initially used to pull down Lola-I. De novo motif discovery under Lola-I ChIP-seq peaks confirmed the previously identified lola-I motif from bacterial one-hybrid experiments (Supplemental Fig. S5A; Enuameh et al. 2013). Genome-wide, 54% of Lola-I peaks colocalized with Cp190 peaks (Fig. 5B). We conclude that Lola-I is often bound at meta-loop anchors and more broadly near Cp190 peaks.

Figure 5.

Lola-I is required to form a Cp190-dependent meta-domain. (A) Similar to Figure 3A but comparing Lola-I ChIP peaks with published Cp190 and CTCF ChIP peaks (Kaushal et al. 2021) around meta-loop anchors in WT larval CNSs. (B) Overlap between Lola-I (white), Cp190 (blue), and CTCF (gray) peaks in WT larval CNSs, with some peaks split for three-way comparisons (see the Materials and methods). (C) Percentages (Y-axis) of mCherry⁺ neuronal nuclei FACS-purified from dissociated 16–24 h embryos of the indicated genotypes (X-axis). (D) Differential analysis of strengths of loops involving meta-loop anchors (dots) measured by Hi-C in lola-I^KO versus WT embryonic neurons. Labeled loops (red dots) were significantly weakened (Padj ≤ 0.01, |fold change| ≥ 2.5). (E) Same as Figure 3D but showing Hi-C maps of neurons from 16–24 h embryos of the indicated genotypes, with Lola-I and published Cp190 (Kaushal et al. 2021) ChIP-seq tracks in larval CNSs, showing a meta-domain dependent on Lola-I and Cp190. See also Supplemental Figure S5 and Supplemental Table S4.

To test whether Lola-I is required for meta-loop formation, we performed Hi-C on FACS-purified neurons from two independent lola-I^KO alleles and WT control embryos. To obtain sufficient lola-I^KO embryos, these were sorted from a GFP-balanced stock using a particle sorter. Neurons expressing mCherry under elav regulatory sequences were isolated by FACS from late (16–24 h) embryos of two independent lola-I^KO lines and WT controls. Although up to 15% of sorted nuclei from WT embryos were mCherry⁺ (neurons), only ∼5% of nuclei were mCherry⁺ in lola-I^KO embryos (Fig. 5C; Supplemental Fig. S5B). Our results suggest that lola-I may be required for proper neuronal differentiation or for the ability of neurons to remain differentiated. Although lola is required in neural stem cells for neuronal differentiation and in neurons for maintaining their differentiated state (Neumüller et al. 2011; Southall et al. 2014; Wissel et al. 2016), lola-I was not specifically known to be required for either of these functions.

Hi-C analysis (Supplemental Fig. S5C; Supplemental Table S4) revealed that only the L24 I-I meta-loop, which was used to purify Lola-I (Fig. 4C), and the second L23 I-I meta-loop present in the same meta-domain were disrupted in lola-I^KO embryonic neurons (Fig. 5D,E). Lola-I binds strongly to L24 anchors and weakly to L23 anchors (Fig. 5E). To clarify whether loss of L23 and/or L24 in lola-I^KO mutants would destabilize the meta-domain, we deleted an intergenic anchor of either L23 or L24 (Supplemental Fig. S5D). DNA-FISH showed that disrupting either L23 or L24 strongly destabilized the meta-domain. Therefore, destabilization of either L23 or L24 in lola-I^KO mutants could explain the destabilization of the meta-domain (Fig. 5E). Visualizing L23 and L24 by DNA-FISH in WT embryos also revealed that these meta-loops are significantly less stable than others that we visualized previously, with both present in only ∼6% of WT embryonic neurons (in contrast to ∼30% of neurons, as observed for the Mp, beat-IV, and GluRIA/GluRIB meta-domains), potentially due to their large size (13.4 and 13.3 kb, respectively) (cf. Fig. 1E–G and Supplemental Fig. S5D).

We conclude that the meta-domain containing L23 and L24—a meta-loop that is cobound by Cp190 and Lola-I at its anchors—requires both proteins for its formation (Figs. 3B,E, 5D,E). Despite binding to nearly half of all meta-loop anchors, Lola-I is required to form only two meta-loops (L23 and L24), which are already unstable in WT embryos. We conclude that our unbiased approach identified Lola-I as a meta-loop anchor-associated protein required, along with Cp190, for the formation of a specific meta-domain.

Discussion

In this study, we investigated the functional relevance and formation mechanism of meta-loops in fly neurons. We precisely perturbed meta-loops in four meta-domains and investigated the roles of Cp190, a key player in TAD boundary formation, and Lola-I, a protein isoform identified through an in vitro pull-down using meta-loop anchor DNA. We reached the following conclusions: (1) Meta-domains containing beat-IV and Mp facilitate long-range transcriptional activation of these genes over 5.1 Mb and 2 Mb, respectively, demonstrating that meta-loops represent ultra-long-range regulatory interactions (Fig. 1). (2) A meta-domain connecting GluRIA–GluRIB paralogous genes is dispensable for high expression levels but facilitates their cotranscription (Figs. 1, 2). (3) Cp190 is critical for forming half of all meta-loops and is required for CTCF-dependent meta-loop formation (Fig. 3), representing the largest role of any protein in meta-loop formation described thus far. (4) An unbiased proteomics approach to purify meta-loop-associated proteins identified Lola-I as a new protein bound to nearly half of all meta-loop anchors and necessary to form one specific Cp190-dependent meta-domain (Figs. 4, 5). Below, we discuss how these findings advance our understanding of how meta-loops function and form.

Insights into meta-loop function

A key challenge in understanding the relevance of genome folding for gene regulation is to distinguish structural features that actively influence gene expression from those that simply reflect the processes that form them (Solovei and Mirny 2024). By studying two meta-domains with meta-loops that connect the promoters of neuronal genes (beat-IV and Mp) to intergenic anchors in distant TADs, we demonstrated that these intergenic anchors are required for wild-type expression of these genes. Disrupting the meta-domains leads to a twofold to threefold reduction in beat-IV and Mp expression, consistent with previous findings from the nolo meta-domain (Mohana et al. 2023). Our results suggest that, depending on the intergenic anchor, this anchor may function as a distant enhancer or as a structural tether to bridge beat-IV and Mp promoters to a distant enhancer. This suggests that meta-domains play an active role in long-range gene activation rather than simply reflecting transcriptional activation.

Unlike promoters involved in I-P meta-loops, the GluRIA–GluRIB meta-domain is required for the cotranscription of these paralogs. Like intra-TAD loops that enable cobursting of paralogous gene pairs in early fly embryos (Levo et al. 2022), meta-loops may facilitate cotranscription of paired genes within shared transcriptional hubs, likely by pooling transcriptional activators and RNA polymerase II. Physically separating GluRIA and GluRIB does not strongly affect their transcription levels, possibly because these genes may have been duplicated along with their local enhancers. Thus, their physical tethering may be more important for cotranscription than for transcriptional activation per se.

The conserved physical pairing of GluRIA and GluRIB over 40 million years of Drosophila evolution (Mohana et al. 2023) suggests selective pressure for cotranscription. An independent study recently addressed the relevance of this pairing by deleting the GluRIB promoter, which partially destabilized the GluRIA–GluRIB meta-domain (Pollex et al. 2024), suggesting that the second I-I meta-loop may have been retained. This deletion strongly upregulated GluRIA expression by up to 10-fold, suggesting that GluRIA–GluRIB pairing is required to dampen their expression (Pollex et al. 2024). In contrast, our deletion of an anchor of the I-I meta-loop fully disrupted the GluRIA–GluRIB meta-domain while leaving both genes intact (Fig. 1G) and did not significantly upregulate either gene (Figs. 1J, 2B). Although GluRIA–GluRIB pairing may dampen each gene's expression through competition of both promoters for shared resources required for transcription, we propose that the primary function of pairing is to ensure transcription of both paralogs in shared neuronal subtypes. Physical gene pairing may mitigate the loss of critical regulatory elements near one gene during both genes’ independent evolution after gene duplication. Furthermore, physical GluRIA–GluRIB pairing may ensure proper stoichiometric production of GluRIA and GluRIB proteins, which may in turn be critical for their assembly into heterotetramers, given that their mammalian homologs are known to form heterotetramers (Shi et al. 2001; Greger et al. 2017).

To reconcile the roles of meta-domains in controlling the transcription of single neuronal genes looping to a distant intergenic anchor or of pairs of distant genes, we propose that meta-domains enable sharing of regulatory elements and transcription factors between two distant TADs.

We observed a notable discrepancy between the severity of meta-loop and neuronal specification defects in Cp190⁰ (with widespread meta-loop disruption but largely normal neuronal differentiation) and lola-I^KO embryonic neurons (with neuronal differentiation defects but few meta-loop disruptions). This discrepancy suggests that meta-loop defects in Cp190⁰ neurons are not secondary effects of developmental or transcriptional alterations, which may be more pronounced in lola-I^KO mutants. These findings highlight a dichotomy between factors like Lola-I, which may control neuronal fate largely independently of inter-TAD interactions, and structural proteins like Cp190, which may fine-tune the transcription of a subset of neuronal genes and enable ultra-long-range gene regulation.

Insights into meta-domain formation

One of the most striking yet poorly understood aspects of meta-domain formation is the remarkably specific pairing of their interacting TAD partners. Precise deletions of meta-loop anchors in four studied meta-domains confirm that meta-domains are tethered by meta-loops. Deleting individual or combined meta-loop anchors was sufficient to destabilize the corresponding meta-domain, depending on its structure (Fig. 1B–G).

We further demonstrated that meta-loop formation is often hierarchical, with structural meta-loops nucleating formation of others in the same meta-domain. In two studied meta-domains, the I-I meta-loop formed earlier during embryonic neurogenesis and facilitated later I-P or P-P meta-loop formation (Fig. 1F,G). In contrast, the I-I and I-P meta-loops formed independently in the Mp meta-domain (Fig. 1E). Currently, predicting the impact of deleting a specific intergenic anchor on meta-domain formation remains challenging. Cp190, which we showed plays key role in meta-loop formation, frequently only binds to I-I meta-loop anchors but also binds to the Mp I-P meta-loop anchors, potentially explaining why this I-P meta-loop can form independently of the I-I meta-loop (Fig. 3D). Cp190-dependent meta-loops not directly bound by Cp190 at their anchors often occurred within the same meta-domain as other Cp190-bound meta-loops, further suggesting that Cp190-bound structural meta-loops nucleate others in the same meta-domain (Fig. 3A,E; Supplemental Fig. S3D,E). In brief, these findings highlight the central role of structural meta-loops, such as those bound by Cp190, in guiding the formation of additional nearby meta-loops and maintaining meta-domain stability.

We also uncovered an interplay between TAD boundaries and meta-loop anchors. Approximately half of meta-loop anchors overlap with TAD boundaries, and we demonstrated that Cp190, a key structural protein required to form most TAD boundaries distal to transcribed promoters, is essential to form half of all meta-loops (Fig. 3). CTCF recruits Cp190 to form a specific subset of Cp190-dependent meta-loops (Fig. 3A–C), and we speculate that Ibf1/Ibf2 may similarly recruit Cp190 to form a different subset of Cp190-dependent meta-loops (Fig. 4B,C).

However, Cp190 and its recruiting cofactors (CTCF and potentially Ibf1/Ibf2) alone are not sufficient for meta-loop formation, as only a small subset of Cp190-dependent TAD boundaries function as meta-loop anchors. We propose that Cp190-dependent meta-loops require additional proteins bound to their anchors, which distinguish them from more widespread TAD boundaries. One meta-loop anchor candidate is Lola-I (Figs. 4B–D, 5). Other proteins essential for meta-looping may be identified by systematically testing proteins enriched in our in vitro meta-loop anchor DNA pull-downs.

How does Cp190 contribute to meta-loop formation? We considered three possible scenarios: (1) Cp190's role in forming TAD boundaries may contribute to meta-loop formation. Although it remains unclear whether Drosophila CTCF and Cp190 create barriers to chromosomal loop extrusion as mammalian CTCF does, both proteins copurify with cohesin subunits (Kaushal et al. 2021, 2022). However, cohesin-mediated loop extrusion seems unlikely to underlie meta-loop formation due to the vast size of meta-loops, surpassing cohesin's estimated processivity on DNA (Chan and Rubinstein 2023; Han et al. 2023) and overcoming several tens of intervening boundaries. Still, colliding cohesin complexes extruding independent chromosomal loops can bridge loci that are further apart than sizes of typical cohesin loops (Hung et al. 2024), and loop extrusion therefore cannot be completely ruled out for a potential role in meta-loop formation. (2) Cp190 may help recruit other proteins needed for meta-loop formation to meta-loop anchor DNA. (3) Cp190 may directly mediate physical interactions between distant binding sites; for example, through homotypic BTB domain interactions or heterotypic interactions with other BTB ZnF proteins at distant meta-loop anchors. Interestingly, three of the four proteins shown so far to be required for meta-loop formation (GAF, Cp190, and Lola-I) belong to the BTB ZnF family. CTCF is the exception but may function primarily to recruit Cp190 to facilitate meta-loop formation (Fig. 3). Our findings therefore suggest that BTB ZnF proteins may play a central role in forming long-range interactions with unequal contributions: GAF and Lola-I are each required for 3% of meta-loops, whereas Cp190 is essential for ∼50%, highlighting its dominant role in meta-loop formation. Future work should clarify potential interactions between BTB ZnF proteins bound to different meta-loop anchors.

It remains to be determined whether Lola-I synergizes with Cp190 to form meta-loop L24 by recruiting Cp190 to L24 anchors through its sequence-specific DNA-binding activity (Fig. 5) and its ability to interact with Cp190 (Kaushal et al. 2022). Lola-I was recently identified as a pioneer factor (Ramalingam et al. 2023), potentially supporting its role in facilitating Cp190 recruitment to L24 anchors.

To conclude, our results highlight an interplay between TAD boundaries and ultra-long-range gene regulatory loops. Future studies will unravel the molecular underpinnings of this interplay and its relevance for long-range gene regulation in other genomes. Althgouh long-range physical contacts are observed in the mammalian genome, many studies to date have focused on architectural proteins required to form multivalent interactions (Monahan et al. 2019; Friman et al. 2023; Hu et al. 2023; Aboreden et al. 2025), potentially due to the challenges of identifying more specific long-range interactions in large genomes. Pushing the boundaries of high-resolution chromatin contact mapping may enable specific meta-loop-like interactions to be uncovered in more complex genomes.

Materials and methods

Drosophila melanogaster strains

Drosophila melanogaster was cultured under standard laboratory conditions at 25°C. Embryos and larvae were not sexed, but adults were analyzed in 50:50 pools of females:males.

CRISPR/Cas9-mediated intergenic anchor knockouts

Meta-loop anchors were precisely deleted by CRISPR/Cas9-mediated genome editing using small guide RNAs flanking loci chosen for deletion: 222 bp (dm6 coordinates chromosome 3R: 23,538,954–23,539,175) for L53^ΔA63, 199 bp (dm6 coordinates chromosome 3R: 28,654,618–28,654,816) for L54^ΔA65 as published previously (Mohana et al. 2023), 560 bp (dm6 coordinates chromosome r3L: 9,330,073–9,330,632) for L36^ΔA46, 590 bp (dm6 coordinates chromosome 3L: 4,800,693–4,801,282) for L33^ΔA38, 345 bp (dm6 coordinates chromosome 3L: 4,805,728–4,806,072) for L34^ΔA39, 859 bp (chromosome 2R: 8,686,315–8,687,173) for L23^ΔA27, and 653 bp (chromosome 2R: 8,715,570–8,716,222) for L24^ΔA28. Four to six sgRNAs per targeted intergenic anchor were cloned (using primers listed in Supplemental Table S1) downstream from U6:1, U6:2, and U6:3 promoters (Port et al. 2014); assembled into a transgene; and integrated by site-specific recombination into the ZH86Fb landing site (Bischof et al. 2007). All Drosophila injections were performed by FlyORF. Males expressing both sgRNAs and Cas9 in their germline (nanos-Cas9) (Port et al. 2014) were crossed to balancer females, and crosses of single sons were set up with balancer females. After 3 days, the males were sacrificed and genotyped by PCR and confirmed by Sanger sequencing. Double KOs were generated by deleting an anchor in the genetic background of a pre-existing anchor deletion. Males and females harboring two independently isolated CRISPR KO alleles were crossed, and the resulting transheterozygous embryos were analyzed. Two exceptions were L54^ΔA65 and L53^ΔA63 L54^ΔA65 double-KO animals, which were female sterile and for which both independent CRISPR KO alleles were analyzed in parallel experiments and compared to verify that similar results were obtained in both.

lola isoform-specific knockouts and rescue

Exons specific to lola-I and lola-G were knocked out as described above by CRISPR/Cas9-mediated genome editing using small guide RNAs flanking the loci chosen for deletion: 4462 bp (dm6 coordinates chromosome 2R: 10,481,824–10,486,285) for lola-I^KO and 3493 bp (dm6 coordinates chromosome 2R: 10,506,737–10,510,229) for lola-G^KO. lola-I^KO animals were then rescued by a transgene containing lola-I cDNA flanked in 5′ by 2.7 kb of genomic sequence extending from 1526 bp upstream of the lola-I transcription start site and comprising its entire 5′ untranslated region and in 3′ by 1492 bp of genomic sequence downstream from the lola-I stop codon. The 5′ and 3′ lola-I genomic sequences specifically corresponded to chromosome 2R: 10,533,735–10,536,470 and chromosome 2R: 10,506,863–10,508,346, respectively. The rescue transgene was inserted into the ZH86Fb landing site (Bischof et al. 2013). All experiments were performed in animals transheterozygous for two independent CRISPR KO alleles.

Drosophila viability tests

Thirteen sets of 60 embryos on average of desired genotypes were transferred into separate vials, and the numbers of hatched embryos, larvae, pupae, and fully hatched adults were recorded.

Embryo immunostaining

Embryos were dechorionated in 50% bleach for 2 min, fixed in 4% paraformaldehyde for 30 min at room temperature, devitellinized in methanol:heptane (2:1) for 5 min at room temperature, washed, and then stored in 100% MeOH at −20°C. Samples were rehydrated in PBS with 0.1% Tween 20 and washed three times in PBT (PBS with 0.1% Tween 20) and six times in BBT (1× PBS, 1% BSA, 0.1% Triton X-100). Primary antibody against Fasciclin II (mouse monoclonal 1D4, Developmental Studies Hybridoma Bank) diluted 1:20 in BBT was added, and tropomyosin (rat monoclonal BB5/37.1, Developmental Studies Hybridoma Bank) diluted 1:6 in BBT was incubated with embryos overnight at 4°C.

The next day, immunostained embryos were washed six times in BBT and incubated with secondary antibody Alexa 594 antirat IgG (Thermo Fisher A21209) and Alexa 680 antimouse IgG (Thermo Fisher A21058) diluted 1:100 in BBT overnight at 4°C.

The next day, immunostained embryos were washed twice in BBT and four times with PBS with 0.1% Triton X-100. DAPI was added on the last wash. Images were acquired on a Stellaris 5 WLL microscope with a 20× objective and visualized with Fiji software v2.14.0/1.54f.

Oligo-DNA-FISH probe preparation

Oligonucleotides for DNA-FISH (Supplemental Table S1) were designed using the PaintShop (Hershberg et al. 2021) server and ordered from GenScript. Oligonucleotide primary probes contained three regions: (1) a region of genomic homology composed of 30–40 nt complementary to the target locus; (2) three flanking 20 nt regions, unique to each locus, complementary to a fluorescently labeled readout oligo (two repeats at 5′ and one at 3′); and (3) two flanking 20 nt regions containing primers allowing PCR amplification of either the whole DNA or RNA oligopaints library. With this design, three readout sequences could be hybridized per primary oligo. For each locus, we used either 130 primary oligos spanning an ∼10 kb locus (for L33_A43, L34_A39, L36_A41, and L36_A46) or 400 primary oligos spanning an ∼25 kb locus (for L54_A64 and L54_A65). The DNA oligo-FISH library was amplified by emulsion PCR (Gizzi et al. 2020) using a universal pair of primers (Supplemental Table S1). After cleaning the PCR product, limited cycle PCR was performed using the same pair of primers with the following cycling conditions: 2 min at 98°C; 30 sec at 98°C; eight cycles of 20 sec at 98°C, 20 sec at 62°C, and 10 sec at 72°C; and 7 min at 72°C. The reverse primer contained an additional T7 promoter sequence (5′-TAATACGACTCACTATAGGG-3′). The PCR was purified with a MinElute PCR purification kit (Qiagen 28004) and transcribed using a HiScribe T7 Quick high-yield RNA synthesis kit (NEB E2050S) overnight at 37°C. Reverse transcription was done using M-MLV (Thermo Fisher 28025013) and the forward primers following the manufacturer's instructions. RNA was degraded by alkaline hydrolysis by adding 100 mM EDTA and 200 mM NaOH in the reverse transcription tube. The final probe was purified by phenol/chloroform extraction followed by ethanol precipitation, dissolved in 100% formamide, and stored at −20°C.

Oligo-DNA-FISH

WT and mutant embryos (0–24 h old) were dechorionated, fixed in 4% formaldehyde for 20 min at room temperature, washed, and then stored in 100% MeOH at −20°C. For FISH, samples were rehydrated in 2× SSCT (30 mM sodium citrate dihydrate at pH 7.4, 300 mM NaCl, 0.1% Tween 20), treated with RNase A for 30 min at room temperature, washed in 2× SSCT, permeabilized with 0.2 mol/L 37% HCL for 10 min at room temperature, washed in 2× SSCT, equilibrated to 50% formamide in 2× SSCT, and incubated for 2 h at 37°C. DNA-FISH probes mixed with embryos in 25 µL total of 2× SSC, 10% (w/v) dextran sulfate, 50% formamide, and 0.1% Tween 20 were heated for 10 min to 80°C and then incubated overnight at 37°C in the dark. The next day, the embryos were washed in 2× SSCT, and fluorescent readout probe (Supplemental Table S1) was mixed with 2× hybridization buffer (5× SSC, 25% [w/v] dextran sulfate, 50% formamide, 0.25% Tween 20), added to the embryos, and incubated for 1 h. Embryos were washed in 50% formamide in 2× SSCT, then 20% formamide in 2× SSCT, and then 2× SSCT at 37°C. Embryos were then immunostained for 2 h at room temperature with anti-lamin (mouse monoclonal clone ADL67.10, Developmental Studies Hybridoma Bank) diluted 1:10 in BBT (1% BSA, 1× PBS, 0.1% Triton X-100) and anti-ELAV (rat monoclonal clone 7E8A10, Developmental Studies Hybridoma Bank). Next, they were washed twice in BBT and immunostained for 1 h at room temperature in the dark with Alexa 680 antimouse (Thermo Fisher A21058). Images were acquired on a Stellaris 5 WLL microscope with a 63× oil objective and visualized with Fiji software v2.14.0/1.54f.

Oligo-DNA-FISH image analysis

A custom script (available at https://github.com/ariannaravera/OligoDNA-FISH) was developed to segment nuclei using the Cellpose cyto3 model, with a stitching threshold of 0.5 applied in the z-dimension to ensure accurate 3D reconstruction. Masks were generated to distinguish neurons (immunostained with α-Elav) from nonneurons. DNA-FISH spots were then detected within these segmented nuclei. Nuclei with more than one DNA-FISH spot per meta-loop anchor were excluded from downstream analysis (note that homolog chromosomes are paired in Drosophila). The diameters of the valid nuclei and the distances between the meta-loop anchors were measured and plotted using R.

In the case of beat-IV DNA-FISH experiments, which exhibited elevated background noise, the segmentation process was complemented by analysis in Imaris software. After nucleus segmentation with the Cellpose cyto3 model, the resulting masks were imported into Imaris for further refinement and processing. Preprocessing of the images in Imaris was performed using a median filter to enhance signal quality of the spots. Subsequently, individual loci were identified using the Imaris spot detection tool, generating spot-specific masks. These spot masks were integrated into the nuclear masks to associate loci with their respective nuclei. The distances between pairs of spots were measured and exported for downstream analysis.

Oligo-RNA-FISH probe preparation

Unlabeled primary probes were designed using the PaintShop (Hershberg et al. 2021) server and were prepared similarly to the DNA-FISH probes (Supplemental Table S1). The secondary fluorescent probes were conjugated to either a 6FAM or a Cy3 moiety through 5′ amino modifications. Fluorescent oligos were resuspended in 10 mM Tris-EDTA (pH 8) at a final concentration of 100 µM.

Oligo-RNA-FISH

Embryos were dechorionated in 50% bleach for 2 min, fixed in 4% paraformaldehyde for 30 min at room temperature, devitellinized in methanol:heptane (2:1) for 5 min at room temperature, washed, and then stored in 100% MeOH at −20°C. Embryos were rehydrated in PBT (1× PBS, 0.1% Tween) for 10 min at room temperature, equilibrated to 30% formamide in 2× SSC, and incubated for 1 h at 37°C. Unlabeled primary probes against GluRIA and GluRIB mRNA (Supplemental Table S1) were mixed with the embryos in 100 µL total of 2× SSC, 10% (w/v) dextran sulfate, and 30% formamide and then incubated overnight at 37°C in the dark while shaking. The next day, the embryos were washed in 2× SSCT, and the fluorescent readout probes (Supplemental Table S1) were mixed with 2× hybridization buffer (5× SSC, 25% [w/v] dextran sulfate, 50% formamide, 0.25% Tween 20), added to the embryos, and incubated for 1 h. After that, embryos were washed in RNA-FISH wash buffer (30% formamide in 2× SSC) four times for 15 min each at 37°C. Embryos were then immunostained for 2 h at room temperature with antispectrin (mouse monoclonal clone 3A9, Developmental Studies Hybridoma Bank) diluted 1:10 in BBT (1% BSA, 1× PBS, 0.1% Triton X-100). Next, they were washed twice in BBT and immunostained for 1 h at room temperature in the dark with Alexa 680 antimouse (Thermo Fisher A21058). Images were acquired on a Stellaris 5 WLL microscope with a 63× oil objective and visualized with Fiji software v2.14.0/1.54f.

GluRIA–GluRIB cotranscription analysis

Cells were segmented using the Cellpose cyto3 model, with a stitching threshold of 0.5 applied in the z-dimension. The resulting segmentation masks were imported into Imaris for further analysis. Image preprocessing in Imaris was performed using a median filter to improve signal quality. Transcription sites were identified using the Imaris spot detection tool by applying thresholds for size, intensity, and quality. The resulting spot masks were integrated into the nuclear masks. We quantified the number of cells containing either a GluRIA or GluRIB transcription site or both. For nuclei with both transcription sites, the distances between GluRIA and GluRIB transcription sites were measured.

TSA-RNA-FISH

RNA-FISH DNA templates for lola-I or lola-G isoforms were PCR-amplified from genomic DNA using the primer sequences listed in Supplemental Table S1. These inserts were cloned upstream of a T7 RNA polymerase promoter oriented such that it transcribed an antisense transcript to the mRNA. Labeled RNA probes were then generated from template DNA by in vitro transcription with Dig-UTP labeling mix (Roche 11277073910) and T7 RNA polymerase (Roche 10881767001). After DNase I digestion for 20 min at 37°C, probes were fragmented by incubation for 20 min at 65°C in 60 mM Na₂CO₃ and 40 mM NaHCO₃ (pH 10.2); precipitated in 300 mM sodium acetate (pH 5.2), 1.25 M LiCl, 50 mg/mL tRNA, and 80% EtOH; resuspended in 50% formamide, 75 mM sodium citrate (pH 5), 750 mM NaCl, 100 mg/mL salmon sperm DNA, 50 mg/mL heparin, and 0.1% Tween 20; and stored at −20°C. Embryos were fixed in 4% formaldehyde for 30 min at room temperature, washed, and then stored in 100% MeOH at −20°C. Samples were rehydrated in PBS with 0.1% Tween 20, postfixed in 4% formaldehyde for 20 min at room temperature, progressively equilibrated to hybridization buffer (50% formamide, 75 mM sodium citrate at pH 5, 750 mM NaCl), and heated to 65°C. RNA probes were diluted 1:50 in hybridization buffer, denatured for 10 min at 80°C, placed on ice, and added to the samples overnight at 65°C with shaking. Samples were washed six times for 10 min in hybridization buffer at 65°C and then progressively equilibrated to PBS with 0.1% Triton X-100. Samples were incubated overnight at 4°C in anti-digperoxidase (Roche 11207733910) diluted 1:2000 in PBS, 0.1% Triton X-100, and 1× western blocking reagent (Sigma-Aldrich 1921673). Samples were washed six times for 10 min in PBS with 0.1% Tween 20, labeled with Cyanine 3 tyramide in the TSA Plus kit (PerkinElmer NEL753001KT) for 3 min at room temperature, washed six times for 10 min in PBS with 0.1% Tween 20, and finally mounted with DAPI to stain DNA. Images were acquired on a Stellaris 5 WLL microscope with a 20× objective and visualized with Fiji software v2.1.0/1.53c.

RT-qPCR on adult heads

For RT-qPCR on adult heads shown in Figure 1, total RNA was extracted from 10 heads (five female and five male) of WT or KO 1–3 day old flies (transheterozygous for two independently generated CRISPR KO alleles) in biological triplicates with TRIzol. Five-hundred nanograms of RNA was reverse-transcribed with GoScript (Promega A2801) with random hexamers following the manufacturer's instructions. qPCR was performed with GoTaq (Promega A600A) on a QuanStudio 6 Flex real-time PCR system (Thermo Fisher). Fold changes of gene expression in mutants relative to WT controls were calculated using the ΔΔC_T method, referring to the RpL15 signal and normalizing by the mean expression level of WT for each primer pair. A two-sided unpaired heteroscedastic t-test was performed to test the significance of differential expression in knockout versus WT backgrounds.

Embryo fixation for FACS

WT, CTCF⁰, and Cp190⁰ embryos (16–24 h old) were collected on yeasted apple juice agar plates, washed off the collection plates, dechorionated by stirring in 50% bleach for 2 min, extensively rinsed with tap water, dried with paper towels, transferred into 50 mL Falcon tubes containing 4 mL of cross-linking solution (50 mM HEPES at pH 8, 1 mM EDTA, 0.5 mM EGTA, 100 mM NaCl, 1,8% formaldehyde) and 12 mL of heptane, and fixed by vigorous shaking for 15 min at room temperature. Falcon tubes were centrifuged at 500g for 1 min, and cross-linking solution and heptane were decanted. Formaldehyde cross-linking was quenched by vigorously shaking embryos for 1 min in 10 mL of 125 mM glycine in PBS with 0.1% Triton X-100. Falcon tubes were centrifuged for 1 min again, and the buffer was decanted. Embryos were rinsed in embryo wash (PBS, 120 mM NaCl, 0.1% Triton X-100). Falcon tubes were centrifuged for1 min, and the buffer was decanted.

Isolation of WT, CTCF⁰, and Cp190⁰ embryonic neurons

Embryos were fixed (see “Embryo Fixation for FACS”) and transferred to 1 mL microtubes. Embryos were snap-frozen dry and stored at –80°C until a total of 0.5–1 g of embryos was collected. Frozen embryos were dounce-homogenized in 10 mL of cold homogenization buffer (15 mM Tris-HCl at pH 7.4, 15 mM NaCl 60 mM KCl, 340 mM sucrose, 0.2 mM EDTA at pH 8, 0.2 mM EGTA at pH 8, 1× complete protease inhibitor cocktail) with 20 loose pestle strokes and 15 tight pestle strokes. The lysate was filtered through two layers of miracloth rotated by 90° to each other's grain. Nuclei were pelleted by centrifuging at 3000g for 10 min at 4°C and blocked in 1 mL of PBTB-NP40 (PBS, 0.1% Triton X-100, complete, 5% BSA, 0.1% NP-40) in 2 mL Eppendorf tubes by rotating for 1 h at 4°C. Primary antibodies against ELAV (rat monoclonal 7E8A10, Developmental Studies Hybridoma Bank) were added at a 1:40 dilution and incubated with nuclei overnight at 4°C. The next day, immunostained nuclei were washed three times in PBTB-NP40 (incubated with PBTB-NP40 for 10 min at 4°C and centrifuged at 5000g for 3 min at 4°C) and incubated with secondary antibody Alexa 647 antirat IgG (Thermo Fisher A21247) diluted 1:100 in PBTB without NP-40 for 6 h at 4°C. Immunostained nuclei were washed three times in PBTB (incubated with PBTB for 10 min at 4°C and centrifuged at 5000g for 3 min at 4°C), resuspended in PBS and 0.1% Triton X-100 after the last wash, and incubated with DAPI overnight at 4°C. The next day, nuclei were passed up and down 10 times through a 22 gauge needle, filtered through a 35 mm nylon mesh, and FACS-sorted into PBT (PBS, 0.1% Triton X-100, 1× complete protease inhibitor cocktail) on a Beckman Coulter Astrios EQ instrument. One million nuclei were collected in biological duplicates. Tubes were centrifuged at 5000g for 5 min at 4°C in a swinging bucket centrifuge. PBT was removed, and tubes were snap-frozen and store at –80°C.

Isolation of WT and lola-I^KO embryonic neurons

Embryos from two independent CRISPR lola-I^KO lines balanced with a GFP-expressing balancer and with mCherry fused to a nuclear localization signal knocked into the C terminus of the elav open reading frame downstream from a self-cleaving T2A peptide were collected and fixed (see “Embryo Fixation for FACS”). Neurons from WT animals with the same elav^mCherry knock-in were collected in parallel as positive control. mCherry-expressing neurons were then FACS-purified from WT and each lola-I^KO line instead of immunostaining embryos with anti-ELAV as performed for CTCF⁰ and Cp190⁰ embryos, because lola-I^KO embryos were rate-limiting for neuron isolation. Embryos were resuspended in 100 mL of PBT (PBS, 0.1% Triton X-100). A BioSorter (Union Biometrica) particle sorter was used to select GFP-negative and mCherry-positive embryos. Sorted embryos were transferred to 1.5 mL micro-tubes and snap-frozen dry until 5640 and 2906 embryos of each independent lola-I^KO allele (considered to be biological duplicates) were collected. Frozen embryos were resuspended in lysis buffer (10 mM Tris-HCl at pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% Triton X-100, 10% BSA, 1 mM DTT, 1× complete, 5× DAPI), transferred to a glass dounce, and dounce-homogenized in 1 mL of lysis buffer with 20 loose pestle strokes and 40 tight pestle strokes. The lysates were filtered through a 35 mm nylon mesh, passed up and down 10 times through a 22 gauge needle, filtered through a 35 mm nylon mesh, and FACS-sorted into PBT (PBS, 0.1% Triton X-100, 1× complete protease inhibitor cocktail) on a Beckman Coulter Astrios EQ instrument and FACSARIA II SORP. A total of 350,000 or 800,000 nuclei was collected for each replicate in 1.5 mL tubes and centrifuged at 5000g for 5 min at 4°C in a swinging bucket centrifuge. PBT was removed, and tubes were snap-frozen and stored at –80°C.

Hi-C on FACS-sorted nuclei

One million FACS-sorted embryonic nuclei per biological replicate were restricted with MseI and Csp6I. Restricted ends were marked with biotin and then ligated. DNA was purified by proteinase K digestion and reverse cross-linking for 6 h at 65°C, sonicated in AFA microtubes in a Covaris S220 sonicator, and purified on SPRIselect beads (Beckman Coulter B23319). DNA was end-repaired, A-tailed, and ligated to barcoded adapters using the NEBNext Ultra II DNA library preparation kit for Illumina (NEB E764S) and then enriched for pairwise DNA junctions by biotin pull-down using Dynabeads MyOne Streptavidin T1 (Invitrogen 65601) beads following the manufacturer's instructions. Libraries were amplified using KAPA HiFi HotStart ready mix (Roche KK2501) and purified on SPRIselect beads. Equimolar pools of multiplexed Hi-C libraries were subjected to 150 bp paired-end sequencing on a NovaSeq6000 instrument.

Hi-C analysis

FASTQ files were mapped to the dm6 D. melanogaster reference genome using distiller v0.3.3 (https://github.com/open2c/distiller-nf), retaining unique pairs of reads with high mapping quality scores on both sides (MAPQ ≥ 30).

For Hi-C map visualization, Hi-C maps at 1 kb resolution for pooled replicates were downsampled to 30 million read pairs using the “random-sample” command in cooltools v07.1.1 (Open2C et al. 2024). Downsampled multiresolution cooler files (1, 2, 4, 8, 16, 32, 64, 128, 256, and 512 kb) were generated using cooler v0.10.2 and normalized using the iterative correction procedure (Imakaev et al. 2012) implemented in cooler. Downsampled Hi-C maps were visualized in HiGlass (Kerpedjiev et al. 2018) and R v4.4.2.

For differential meta-loop strength analysis, distiller-generated 1 kb resolution Hi-C maps for each replicate were used. For each loop involving a meta-loop anchor described in WT larval CNSs in Supplemental Table S2 of Mohana et al. (2023) and each replicate in the Hi-C data set, the total Hi-C count was obtained by summing the Hi-C contacts (raw counts) between 11 kb (for WT, CTCF⁰, and Cp190⁰ Hi-C maps in Fig. 3B) and 21 kb (for WT and lola-I^KO Hi-C maps in Fig. 5D) windows centered on the 1 kb bins containing the summits of both meta-loop anchors engaging in a meta-loop. The resulting count table was used as input for differential analysis using the DESeq2 (Love et al. 2014) v1.46.0 package in R. DESeq2 size factors (data normalization) were evaluated by adding 10,000 randomly shifted meta-loops (with the same sizes as the original meta-loops).

Preblastoderm embryo extract preparation

A procedure similar to that originally described by Baldi et al. (2018) was followed. Embryos (0–90 min old) were collected on yeasted apple juice agar plates from eight cages, each with 25 g of flies reared at 25°C, and successive collections were stored for <1 day at 4°C. Embryos were washed in embryo wash buffer (0.7% NaCl, 0.04% Triton X-100) and dechorionated in 200 mL of embryo wash buffer and 60 mL of 13% sodium hypochlorite for 3 min at room temperature with stirring. Embryos were rinsed for 5 min with cold tap water and transferred into a glass cylinder with embryo wash buffer. The buffer was decanted, and the embryos were washed first in 0.7% NaCl and then in extract buffer (10 mM HEPES at pH7.6, 10 mM KCI, 1.5 mM MgCl₂, 0.5 mM EGTA, 10% glycerol, 10 mM β-glycerophosphate; 1 mM DTT, 1× complete). Embryos were dounce-homogenized by 10 strokes with a motorized homogenizer (Glas-Col) at 4°C. Homogenate volume was recorded, and 1/200 volume of 1 M MgCl₂ was added. Homogenate was centrifuged at 16,000g for 15 min at 4°C. The white lipid layer was discarded, and the supernatant was ultracentrifuged in an Optima Max benchtop ultracentrifuge with a MLS-50 rotor at 268,000g for 1 h at 4°C. The white lipid layer was discarded, and proteins in the supernatant were quantified with Qubit protein assay (Q33211), snap-frozen, and stored at −80°C.

Soluble nuclear extracts from adult heads

Soluble nuclear protein extracts were prepared from heads extracted from 25 mL of WT (OregonR) adults and collected on a sieve. Heads were dounce-homogenized in 30 mL of NU1 buffer (15 mM HEPES at pH 7.6, 10 mM KCl, 5 mM MgCl₂, 0.1 mM EDTA at pH 8, 0.5 mM EGTA at pH 8, 350 mM sucrose, 0.15% NP-40, 1× complete, 1 mM PMSF, 0.5 mM DTT) by 20 manual strokes. The lysate was filtered through a double layer of miracloth rotated by 90° relative to each other's grain. The homogenate was centrifuged at 10,000g for 20 min at 4°C. The nucleus pellet was resuspended, lysed in 24 mL of high-salt buffer (15 mM HEPES at pH 7.9, 400 mM KCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 20% glycerol, 2× complete, 0.5 mM DTT) with rotation for 20 min at 4°C, and centrifuged at 14,000g for 1 min at 4°C. The supernatant was ultracentrifuged in tubes (Beckman Coulter 349623) in an Optima Max benchtop ultracentrifuge with a MLS-50 rotor at 50,000 rpm for 1 h at 4°C. The white lipid layer was discarded, and the soluble nuclear extract was dialyzed twice for 1 h at 4°C into dialysis buffer (50 mM Tris at pH 7.5, 20% glycerol, 2 mM MgCl₂, 0.2 mM EDTA at pH 8, 0.5 mM DTT, 150 mM potassium acetate) with a 6–8 kDa molecular weight cutoff membrane. The final extract was snap-frozen in liquid nitrogen and stored at −80°C.

Chromatinized DNA pull-downs

The following procedure was adapted from Baldi et al. (2018). Meta-loop anchor or negative control DNA (332–451 bp; chromosome 3L: 4,805,785–4,806,195 for L34_A39, chromosome 3L: 6,995,497–6,995,828 for L34_A42, chromosome 2R: 21,977,919–21,978,369 for L24_A32, chromosome 2R: 8,715,653–8,716,054 for L24_A28, and chromosome 2L: 9,629,067–9,629,471 for the Phs-bound region [Baldi et al. 2018] used in our study as negative control) was amplified from genomic DNA (using primers in Supplemental Table S1) and individually cloned into plasmids. Eight-hundred microliters of PCR reactions using these plasmids as templates with cloned genomic regions was purified with a PCR clean-up kit and eluted in 44 µL. Amplified fragments were phosphorylated for 2 h at 37°C in phosphorylation mix (1× T4 ligase buffer, 1.2% PEG4000, 25 U of T4 polynucleotide kinase). Phosphorylated fragments were concatemerized overnight at 16°C after adding 800 U of T4 DNA ligase. The next day, fragment concatemerization was checked on an agarose gel, and the ligated DNA was purified by phenol/chloroform extraction followed by ethanol precipitation and dissolved in 38 µL of H₂0. DNA ends were biotinylated in 1× NEBuffer 2, 0.04 mM Biotin-14-dATP, and 15 U of Klenow (exo⁻) overnight at 37°C. Biotinylated DNA was purified with a PCR clean-up kit.

Triplicates of each amplified anchor and control (Phs) DNA were used for chromatin assembly and subsequent mass spectrometry. For each replicate, 2 µg of biotinylated DNA was coupled to 15 µL of M-280 Streptavidin Dynabeads (Life Technologies 112.06D) for 1 h at 25°C. The beads had been washed before with 1 mL of 1× B&W buffer (5 mM Tris-HCl at pH 7.5, 0.5 mM EDTA, 1 M NaCl) and resuspended in 15 µL of 2× B&W buffer (10 mM Tris-HCl at pH 7.5, 1 mM EDTA, 2 M NaCl). Chromatin assembly was performed by incubating DNA coupled to beads with assembly mix (100 µL of preblastoderm embryo extract, 3 mM MgCl₂, 3 mM ATP at pH 8, 30 mM creatine phosphate, 25 ng/µL creatine kinase [type I; Sigma], 10 mM β-glycerophosphate, 1 mM DTT, 10 mM HEPES at pH 7.6, 50 mM KCl, 0.5 mM EGTA, 10% glycerol) for 1 h at 26°C. Two-hundred microliters of soluble nuclear protein extracts prepared from adult heads was then added to each in vitro chromatin assembly reaction, adjusted to 0.1% Igepal, and incubated for 1 h at 26°C. The beads were washed once with dialysis buffer (50 mM Tris at pH 7.5, 20% glycerol, 2 mM MgCl₂, 0.2 mM EDTA at pH 8, 0.5 mM DTT, 150 mM potassium acetate), and proteins were eluted in 2× SP3 buffer (100 mM Tris-HCl at pH 7.5, 4% SDS, 20 mM DTT) for 10 min at 75°C.

Mass spectrometry sample preparation and protein digestion

Samples eluted in 2× SP3 buffer were diluted 1:1 with H₂O and then digested on an automated pipetting system (Integra Assist Plus) following the SP3 method (Hughes et al. 2019) with 50 mg/mL magnetic Sera-Mag Speedbeads (Cytiva 45152105050250). Proteins were first alkylated with 32 mM (final) iodoacetamide for 45 min at room temperature in the dark. Beads were added at a ratio 10:1 (w:w) to samples, and proteins were precipitated on beads with ethanol (final concentration: 60%). After three washes with 80% ethanol, beads were digested in 50 µL of 100 mM ammonium bicarbonate with 0.6 µg of trypsin (Promega V5113). After 1 h of incubation at 37°C, the same amount of trypsin was added to the samples for an additional 1 h of incubation. To remove traces of SDS, two sample volumes of isopropanol containing 1% trifluoroacetic acid (TFA) were added to the digests, and the samples were desalted on a strong cation exchange (SCX) plate (Oasis MCX, Waters Corp.) by centrifugation. After washing with isopropanol/1% TFA and 2% acetonitrile/0.1% formic acid (FA), peptides were eluted in 150 μL of 40% MeCN, 59% water, and 1% (v/v) ammonia and dried by centrifugal evaporation.

Liquid chromatography-mass spectrometry analyses

LC-MS/MS analyses were carried out on a TIMS-TOF Pro (Bruker) mass spectrometer interfaced through a nanospray ion source (“captive spray”) to an EvoSep One liquid chromatography system (EvoSep). Peptides were separated on a reversed-phase 15 cm C18 column (150 μm ID, 1.5 µm; EvoSep Product EV1137) at a flow rate of 220 nL/min with a 15 sample per day method (run time: 88 min; solvents were water and acetonitrile with 0.1% FA; useful gradient 0%–35%).

Data-independent acquisition was carried out using a method similar to a standard DIA-PASEF method reported previously (Meier et al. 2020), with ion accumulation for 100 msec each for the survey MS1 scan and MS2 scans. Duty cycle was kept at 100%. Precursor ions were chosen within a reduced mobility range 1/k0 from 0.8 to 1.3 and m/z between 400 and 1200. Collision energy was ramped linearly based uniquely on the 1/k0 values from 20 eV (at 1/k0 = 0.6) to 59 eV (at 1/k0 = 1.6). Per cycle, the mass range of 400–1200 m/z was covered by a total of 32 windows, each 25 Th wide. Two windows were acquired per TIMS scan (100 msec) so that the total cycle time was 1.8 sec.

Proteomics data processing

Identification of peptides directly from DIA data was performed with Spectronaut 19.5 with the Pulsar engine using the “deep” setting and searching the D. melanogaster reference proteome (https://www.uniprot.org) database of February 14, 2024 (22,056 sequences), and a contaminant database containing the most usual environmental contaminants and enzymes used for digestion (Frankenfield et al. 2022). For identification, peptides of 7–52 amino acid length were considered, cleaved with trypsin/P specificity and a maximum of two missed cleavages. Carbamidomethylation of cysteine (fixed), methionine oxidation, and N-terminal protein acetylation (variable) were the modifications applied. Spectra, peptide, and protein identifications were all filtered at 1% FDR against a decoy database. Ion mobility for peptides was predicted using a deep neural network and used in scoring. The library created contained 100,364 precursors overall.

Peptide-centric analysis of DIA data was done with Spectronaut 19.5 using the library generated by Pulsar from DIA data. Single-hit proteins (defined as matched by one stripped sequence only) were kept in the Spectronaut analysis. Peptide quantitation was based on the XIC area, for which a minimum of one and a maximum of three (the three best) precursors were considered for each peptide, from which the median value was selected. Quantities for protein groups were derived from interrun peptide ratios based on the MaxLFQ algorithm (Cox et al. 2014). Global normalization of runs/samples was done based on the median of peptides.

Overall, 99,341 precursors were quantified in the data set and mapped to 5036 protein groups. There were 51,442 precursors (4155 protein groups) that had full profiles; i.e., were quantified in all samples. The average number of data points per peak was 6.9.

Proteomics data analysis

The output result table obtained after proteomics data processing (explained in the previous section) was filtered to remove contaminant proteins and retain proteins identified by a minimum of two peptides. Missing values were imputed with a constant value of 28.92, corresponding to the 1% bottom left values. The DEP R package (Zhang et al. 2018) was used to normalize and calculate P-values and log₂FC, resulting in Supplemental Table S3. The Gviz R package (Hahne and Ivanek 2016) was used to plot lola isoforms and Lola peptides enriched at any meta-loop anchor in Figure 4D. Peptide enrichment at these anchors was calculated using peptide pseudocounts, with missing values imputed with a constant value of 0.1. The mean of each triplicate was calculated, and means were divided by the mean of the Phs control. A threshold was set to FC > 10.

Chromatin preparation from larval CNSs

Thirty third-instar larval cuticles per biological replicate (three biological replicates per sample) were dissected in ice-cold PBS and then cross-linked for 15 min at room temperature in 1.8% (v/v) paraformaldehyde, 50 mM HEPES (pH 8), 100 mM NaCl, 1 mM EDTA, and 1 mM EGTA. Cross-linking was stopped by washing for 10 min in 1 mL of PBS, 0.01% Triton X-100, and 125 mM glycine, and cuticles were washed for 10 min in 10 mM HEPES (pH 7.6), 10 mM EDTA, 0.5 mM EGTA, and 0.25% Triton X-100. CNSs were dissected from the cuticles in 10 mM HEPES (pH 7.6), 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and 0.01% Triton X-100 and then sonicated in 120 μL of RIPA buffer (10 mM Tris-HCl at pH 8, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, protease inhibitor cocktail) in AFA microtubes in a Covaris S220 sonicator for 5 min with a peak incident power of 140 W, a duty cycle of 5%, and 200 cycles per burst. Sonicated chromatin was centrifuged to pellet insoluble material and snap-frozen.

ChIP-seq

Half of each biological replicate chromatin was incubated with 2 μL of rabbit polyclonal anti-HA antibody (Abcam Ab9110) overnight at 4°C. Twenty-five microliters of premixed Protein A and G Dynabeads (Thermo Fisher 100-01D and 100-03D) was added for 3 h at 4°C and then washed for 10 min each—once with RIPA, four times with RIPA with 500 mM NaCl, once in LiCl buffer (10 mM Tris-HCl at pH 8, 250 mM LiCl, 1 mM EDTA, 0.5% Igepal CA-630, 0.5% sodium deoxycholate), and twice in TE buffer (10 mM Tris-HCl at pH 8, 1 mM EDTA). DNA was purified by RNase digestion, proteinase K digestion, reversal of cross-links for 6 h at 65°C, and elution from a Qiagen Minelute PCR purification column. ChIP-seq libraries were prepared using the NEBNext Ultra II DNA library preparation kit for Illumina. An equimolar pool of multiplexed ChIP-seq libraries was sequenced 150 bp paired end on a NovaSeq6000 instrument.

ChIP-seq analysis

Paired-end ChIP-seq reads were demultiplexed and mapped to the dm6 genome using bowtie2 v2.5.1 (https://github.com/BenLangmead/bowtie2) (Langmead and Salzberg 2012). Only chromosomes 2, 3, 4, and X were used. Reads mapping to ENCODE-blacklisted regions for dm6 (Amemiya et al. 2019) were discarded. ChIP-seq peaks were called using the R package csaw v1.40.0 (http://bioconductor.org/packages/csaw) (Lun and Smyth 2015) using a window width of 20 bp and spacing of 10 bp, ignoring duplicate reads and reads with a mapping quality score <20. A background enrichment was evaluated as the median over all samples in the comparison of the average number of reads per 2 kb bins. Windows with less than fourfold enrichment over background were filtered out. Data were normalized using the trimmed mean of M-values method (Robinson and Oshlack 2010) implemented in csaw. Differential binding analysis in csaw was based on the quasilikelihood framework implemented in the edgeR package v4.4.0 (http://bioconductor.org/packages/edgeR) (Robinson et al. 2010). Combined P-values were evaluated for each region using csaw, and the Benjamini and Hochberg method was applied to control the false discovery rate (FDR). Regions with FDR < 0.01 were considered as differential binding regions. Genuine Lola-I peaks were identified by differential analysis of ChIP-seq signals in lola-I^KO animals rescued by an HA-lola-I transgene versus WT animals not expressing HA-lola-I as being lower in the WT samples relative to the HA-lola-I-rescued animals. We defined ChIP occupancy as the best.log₂FC obtained from csaw in the differential analysis. We defined peak positions as the best.pos obtained from csaw. To count overlaps between Lola-I, CTCF, and Cp190 peaks in three-way comparisons, some Lola-I, CTCF, and Cp190 peaks were split into up to three subregions. Specifically, 3412 Lola-I ChIP peaks were split into 3426 peaks, 740 published WT CTCF peaks (Kaushal et al. 2021) were split into 768 peaks, and 6473 published WT Cp190 peaks (Kaushal et al. 2021) were split into 6554 peaks.

Lola-I motif discovery

To find DNA-binding motifs enriched under Lola-I ChIP-seq peaks, Lola-I ChIP-seq peaks were ranked by best.log₂FC, and the 500 peaks with the highest best.log₂FC were extracted. We than analyzed ±200 bp of genomic sequence around each of these peaks’ centers using the Xstreme tool (Grant and Bailey 2021) from the MEME suite with default parameters.

Reuse of published data sets

To visualize DNase-hypersensitive sites mapped by DNase-seq in WT D. melanogaster embryos (Reddington et al. 2020), a published data set was lifted over from dm3 to dm6 coordinates using the LiftOver tool (http://genome.ucsc.edu/cgi-bin/hgLiftOver) (Hinrichs et al. 2006).

To visualize scATAC-seq data in larval CNSs (Mohana et al. 2023), a published data set was downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE214707).

To visualize Hi-C data in neurons FACS purified from a developmental time course (Mohana et al. 2023), a published data set was downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE214707).

To visualize Hi-C and CTCF and Cp190 ChIP-seq data in larval CNSs and RNA-seq data in larval CNSs (Kaushal et al. 2021), published data sets were downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE146752).

To visualize sciATAC-seq3 data from the ventral nerve cord (cluster 1) of 16–18 h old embryos (Calderon et al. 2022), a published data set was downloaded from https://shendure-web.gs.washington.edu/content/members/DEAP_website/public/ATAC/revision/bigwigs.

Statistics and reproducibility

For all box plots in the figures, the center line denotes the median, box limits are upper and lower quartiles, the upper whiskers extend to the largest value no further than 1.5× the interquartile range from the upper hinge, and the lower whiskers extend to the smallest value no further than 1.5× the interquartile range from the lower hinge.

Materials availability

All plasmids and fly stains generated in this study are available on request.

Data and code availability

All sequencing data that support the findings of this study were deposited at NCBI Gene Expression Omnibus (GEO) with accession codes GSE286359 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE286359) and GSE286361 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE286361).

Mass spectrometry proteomics data were deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD060311.

All original code is publicly available at GitHub (https://github.com/ariannaravera/OligoDNA-FISH).

Any additional information required to reanalyze the data reported here is available on request.

Competing interest statement

The authors declare no competing interests.