Gene Regulatory Interactions at Lamina-Associated Domains

The nuclear lamina provides a repressive chromatin environment at the nuclear periphery. However, whereas most genes in lamina-associated domains (LADs) are inactive, over ten percent reside in local euchromatic contexts and are expressed. How these genes are regulated and whether they are able to interact with regulatory elements remain unclear. Here, we integrate publicly available enhancer-capture Hi-C data with our own chromatin state and transcriptomic datasets to show that inferred enhancers of active genes in LADs are able to form connections with other enhancers within LADs and outside LADs. Fluorescence in situ hybridization analyses show proximity changes between differentially expressed genes in LADs and distant enhancers upon the induction of adipogenic differentiation. We also provide evidence of involvement of lamin A/C, but not lamin B1, in repressing genes at the border of an in-LAD active region within a topological domain. Our data favor a model where the spatial topology of chromatin at the nuclear lamina is compatible with gene expression in this dynamic nuclear compartment.


Introduction
Adequate control of gene expression at the periphery of the nucleus is critical for the proper regulation of differentiation [1]. At the nuclear periphery, chromatin interacts with the nuclear lamina (NL), a meshwork of A-and B-type lamins (lamins A/C, B1 and B2), via 10-kb to 10-mb lamina-associated domains (LADs) [2,3]. LADs are enriched in diand tri-methylated H3K9 and contain a low density of mostly repressed genes. Whereas most LADs are conserved between cell types (these are often called constitutive LADs or cLADs), some are more variable (facultative LADs or variable LADs; vLADs) and differ between cell types [4,5] through repositioning during differentiation [6][7][8][9][10]. This dynamics, together with LAD alterations reported in laminopathies [11][12][13][14], cancer and senescence [15], underlines the functional relevance of these genome organizers.
Not all LADs are uniform in their chromatin composition. About 10% of genes in LADs are expressed and escape the repressive environment of the NL by being in euchromatic regions within LADs [16][17][18]. We have reported, within cLADs, active regions containing differentially expressed (DE) genes during early stages of the differentiation of human adipose stem cells (ASCs) into adipocytes [10]. In line with other findings [16,17,19], these regions are depleted of H3K9me3 and show lower LMNB1 enrichment than the rest of the LAD. How DE genes in LADs are regulated is largely unknown.
Association of multiple enhancers with genes is a safeguard mechanism of gene expression control [20]. Recent enhancer-capture Hi-C (ECHi-C) data show that enhancers exist as intra-connected "communities" whose remodeling via gains and losses of contacts upon differentiation contributes to defining lineage specificity [21]. Moreover, enhancers, genes or entire gene units are released from the NL during neuronal [6], myogenic [22] or

Fluorescence In Situ Hybridization (FISH)
Fosmid FISH probes were purchased from BACPAC Genomics (Emeryville, CA, USA) (bacpacresources.org; accessed on 16 January 2023). Probes were designed based on RNA expression, ChIP-seq profiles of histone modifications and enhancer identification [10] ( Table S1). FISH was done as described [26]. In short, cells were swollen in hypotonic buffer, fixed in ice-cold methanol/acetic acid and dropped on slides. Probes covering genes and enhancers were respectively labeled with digoxigenin-11-dUTP and Biotin-16-dUTP. Per slide, 80 ng probe was mixed with 8 µg Cot-1 DNA and 10 µg salmon sperm DNA, precipitated and DNA dissolved in 12 µL hybridization mix. Slides were RNase-treated and washed in 2x SSC, dehydrated in ethanol, air-dried, denatured for 1 min 20 s at 70 • C in 70% deionized formamide/2x SSC, dehydrated and air-dried again. Probes were denatured at 70 • C, pre-annealed for 15 min at 37 • C and 15 µL was applied onto 22 × 22 mm coverslips, which were then mounted on slides with nuclei.

Knock-Down of LMNA/C and LMNB1
ASCs were plated at 50% confluency in proliferation medium. After 24 h, knock-down of LMNA/C, LMNB1 or both LMNA/C and LMNB1 was done using Lipofectamin RNAiMAX (Thermo Fisher Scientific, Oslo, Norway) according to manufacturer's recommendations using siLMNA/C:siRNA IDS8221, siLMNB1:siRNA IDS8224 or siScr:siRNA ID4390844 as the control (Ambion; Thermo Fisher Scientific) at final concentrations 40 nM. After 24 h, the media was replaced, and cells kept proliferating for five days. Cells were reseeded, plated at confluency and differentiated as described in Section 2.1.
ChromHMM [33] was used to identify chromatin states using aligned reads. Options were selected to learn a 15-state model using the Baum-Welch training algorithm [34] on the grounds that it represented non-redundant states concordant with genomic locations.
Fold-enrichment of ECHi-C targets was calculated as follows: ((No. bases in state and external annotation)/(No. bases in state))/((No. bases in the external annotation)/(No. bases in the genome)) [33].

ATAC-Seq Data Reprocessing
Assay for Transposase-Accessible Chromatin (ATAC)-seq data were downloaded from GEO GSE143450 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE143 450; accessed on 19 January 2023) [37]. Coverage tracks were lifted from hg19 to hg38 for samples from gluteofemoral fat. Both ATAC-seq replicates from each of 10 donors in that study were merged, and mean coverage was used to generate an ATAC-seq profile at the TSS.

Gene Expression Analysis
The RNA-seq data used here were published by us [10] and downloaded from GEO GSE 185066. An expressed gene was defined by a transcript normalized read count of ≥15 [10]. For reverse transcription (RT)-qPCR, RNA was isolated using the RNAeasy mini kit (Qiagen). In addition, cDNA was synthesized with high-capacity cDNA reverse transcription kit (Applied Biosystems), and qPCR was done using IQ SYBR green (Biorad) with SF3A1 as reference gene and using primers listed in Table S2. Gene Ontology was analyzed using Protein ANalysis THrough Evolutionary Relationships (PANTHER) (http: //www.pantherdb.org/; accessed on 16 January 2023) [41].

cLADs Contain Active Genes in Euchromatic Regions of Low Enrichment in A-and B-Type Lamins
We recently reported transcriptionally active regions in LADs common to undifferentiated ASCs and early adipocytes (cLAD) as determinants of LAD repositioning in the first 3 days of adipogenic differentiation of primary human ASCs [10]. These regions are delineated by H3K4me1 enrichment (H3K4me1 high ), low in LMNB1 (LMNB1 low ) and depleted of H3K9me3 [10]. Extending these findings in the same adipogenic differentiation system, we now report 244 expressed genes in adipogenic cLADs (Table S3). Gene ontology analysis reveals implications in cell adhesion, signaling and developmental processes, which are typical for genes in LADs [5,10] (Table S4). Using our previous histone ChIP-seq data [10], we find that expressed cLAD genes are enriched in H3K4me3, H3K4me1 and H3K27ac around the TSS (Figure 1a). Further, by integrating recent ATAC-seq data for human adipose tissue from several donors [37], we show that the TSS of expressed cLAD genes also reside at sites of open, accessible chromatin ( Figure 1b). chromatin through shared and distinct LADs [43]. We established, by ChIP-seq of LMNA/C ( Figure S1a), that LMNB1 cLADs we previously mapped during early adipogenic differentiation [10] were co-enriched in LMNA/C (Figure S1b), and that H3K4me1 high cLAD regions, which are LMNB1 low , were also low in LMNA/C compared to the rest of the LADs (p < 10 −4 , unpaired two-tailed t-test with Welch's correction; Figure  1c). Thus, euchromatic active cLAD regions showed low contact frequencies with both Aand B-type lamins, suggesting that they were locally detached from the NL.  [37]. (c) LMNB1 and LMNA/C enrichment in H3K4me1 and non-H3K4me1 cLAD regions, and in whole cLADs; bar, median; cross, mean; box, 25-75% percentile; whiskers, min-max **** p < 10 −4 , unpaired two-tailed t-test with Welch's correction. A H3K4me1 region here is a cLAD region containing ≥ 1 H3K4me1 peak; LMNB1 and H3K4me1 data are from [10].

ECHi-C Data Provide Evidence of Enhancer Connectivity at the Nuclear Periphery
Whereas transcription can elicit the local dissociation of chromatin from the NL [18,44], chromatin looping involving CTCF and cohesin has been shown to also mediate detachment from the NL [19]. Thus, we examined whether active cLAD regions were involved in chromatin interactions which could account for the expression of cLAD genes. To this end, we took advantage of recent ECHi-C data for human mesenchymal stem cells induced to differentiate into adipocytes [21]. After lifting interaction bed files from the hg19 to hg38 genome assembly, we pooled all ECHi-C baited enhancers ("baits") and corresponding significant targets (defining enhancer connections) for all differentiation timepoints confounded in that study. By intersecting the genomic coordinates of these 18,710 baits with the cLAD coverage of our study (through D0, D1 and D3 of differentiation), we identified 768 baits localized within our adipogenic cLADs ( Figure 2a).  [37]. (c) LMNB1 and LMNA/C enrichment in H3K4me1 and non-H3K4me1 cLAD regions, and in whole cLADs; bar, median; cross, mean; box, 25-75% percentile; whiskers, min-max **** p < 10 −4 , unpaired two-tailed t-test with Welch's correction. A H3K4me1 region here is a cLAD region containing ≥ 1 H3K4me1 peak; LMNB1 and H3K4me1 data are from [10].
The NL consists of networks of A-and B-type lamins [42], which can interact with chromatin through shared and distinct LADs [43]. We established, by ChIP-seq of LMNA/C ( Figure S1a), that LMNB1 cLADs we previously mapped during early adipogenic differentiation [10] were co-enriched in LMNA/C (Figure S1b), and that H3K4me1 high cLAD regions, which are LMNB1 low , were also low in LMNA/C compared to the rest of the LADs (p < 10 −4 , unpaired two-tailed t-test with Welch's correction; Figure 1c). Thus, euchromatic active cLAD regions showed low contact frequencies with both A-and B-type lamins, suggesting that they were locally detached from the NL.

ECHi-C Data Provide Evidence of Enhancer Connectivity at the Nuclear Periphery
Whereas transcription can elicit the local dissociation of chromatin from the NL [18,44], chromatin looping involving CTCF and cohesin has been shown to also mediate detachment from the NL [19]. Thus, we examined whether active cLAD regions were involved in chromatin interactions which could account for the expression of cLAD genes. To this end, we took advantage of recent ECHi-C data for human mesenchymal stem cells induced to differentiate into adipocytes [21]. After lifting interaction bed files from the hg19 to hg38 genome assembly, we pooled all ECHi-C baited enhancers ("baits") and corresponding significant targets (defining enhancer connections) for all differentiation time-points confounded in that study. By intersecting the genomic coordinates of these 18,710 baits with the cLAD coverage of our study (through D0, D1 and D3 of differentiation), we identified 768 baits localized within our adipogenic cLADs ( Figure 2a).
From the ECHi-C dataset, we then identified 6108 targets (connections) from the K4 baits and 7988 targets from the nonK4 baits (Figure 2c), highlighting a higher density of connections from K4 baits (mean: 27.5/bait) than from nonK4 baits (mean: 14.7/bait) within cLADs (Figure 2d). Although this was anticipated, this finding argues that active genomic regions at the NL are able to contact other parts of the genome. This is exemplified in a > 1 mb cLAD region on chromosome 2, which displayed several K4 baits connecting weak (H3K4me1) or active (H3K4me1/H3K27ac) enhancers within the PXDN locus or between PXDN and SNTG2 or TMEM18, ~1 mb upstream (Figure 2e). We also noted a nonK4 bait within an H3K9me3 segment of the same LAD, connecting two unannotated loci 140 kb apart (Figure 2e). Contact density was extremely low from this nonK4 bait, supporting our quantification (Figure 2d) and reflecting the overall compact and inaccessible state of H3K9me3 chromatin (Figure 2e). bait; bar, median; cross, mean; box, 25-75% percentile; whiskers, min-max; **** p < 10 −4 ; unpaired t-test with Welch's correction. (e) Genome browser view of ECHi-C interactions originating from K4 baits (red arrows) and from a nonK4 bait (black arrow) in a cLAD, with underlying histone modifications and ATAC sites. Gene expression is shown (RNA). Histone ChIP-seq and RNA-seq data are from [10]; ATAC-seq data are reprocessed from [37].
Next, we cataloged baits overlapping by ≥ 1 bp with H3K4me1 peaks in cLADs. We focused on this subset of baits because coincidence with H3K4me1 would also confirm an enhancer marking in our adipogenic system, and H3K4me1 high regions demarcate H3K9me3-depleted and LMNB1 low cLAD regions containing active genes [10] (see Figure 1d). We found 225 such H3K4me1-marked baits, henceforth called "K4 baits," and 543 baits not overlapping with any H3K4me1 peak in cLADs ("nonK4 baits;" Figure 2b).
From the ECHi-C dataset, we then identified 6108 targets (connections) from the K4 baits and 7988 targets from the nonK4 baits (Figure 2c), highlighting a higher density of connections from K4 baits (mean: 27.5/bait) than from nonK4 baits (mean: 14.7/bait) within cLADs (Figure 2d). Although this was anticipated, this finding argues that active genomic regions at the NL are able to contact other parts of the genome. This is exemplified in a > 1 mb cLAD region on chromosome 2, which displayed several K4 baits connecting weak (H3K4me1) or active (H3K4me1/H3K27ac) enhancers within the PXDN locus or between PXDN and SNTG2 or TMEM18,~1 mb upstream (Figure 2e). We also noted a nonK4 bait within an H3K9me3 segment of the same LAD, connecting two unannotated loci 140 kb apart (Figure 2e). Contact density was extremely low from this nonK4 bait, supporting our quantification (Figure 2d) and reflecting the overall compact and inaccessible state of H3K9me3 chromatin (Figure 2e).
To appreciate the active gene-enhancer proximity in cLADs at the cellular level, we analyzed gene-enhancer distances on D0, D1 and D3 of adipogenic differentiation by dualcolor FISH labeling an expressed LAD gene and an enhancer located >200 kb from the gene probe position, marked by H3K27ac in our dataset (Table S1). We found that the CMKLR1 locus displayed increasing gene-enhancer distances as CMKLR1 was downregulated in this time course (Figure 3a,b). This occurred at the nuclear periphery (≤2 µm from the DAPI border), suggesting chromatin repositioning in the vicinity of the NL (Figure 3c). In contrast, the OTUD1 locus, upregulated after adipogenic induction, displayed increased gene-enhancer proximity over time (Figure 3d-f). Not all genes, however, displayed significant variations in gene-enhancer distances at the nuclear periphery despite expression changes, as exemplified by the LRATD1 locus (Figure 3g-i). To appreciate the active gene-enhancer proximity in cLADs at the cellular level, we analyzed gene-enhancer distances on D0, D1 and D3 of adipogenic differentiation by dual-color FISH labeling an expressed LAD gene and an enhancer located >200 kb from the gene probe position, marked by H3K27ac in our dataset (Table S1). We found that the CMKLR1 locus displayed increasing gene-enhancer distances as CMKLR1 was downregulated in this time course (Figure 3a,b). This occurred at the nuclear periphery (≤2 μm from the DAPI border), suggesting chromatin repositioning in the vicinity of the NL (Figure 3c). In contrast, the OTUD1 locus, upregulated after adipogenic induction, displayed increased gene-enhancer proximity over time (Figure 3d-f). Not all genes, however, displayed significant variations in gene-enhancer distances at the nuclear periphery despite expression changes, as exemplified by the LRATD1 locus (Figure 3g-i).
These results provide evidence of enhancer connectivity at the nuclear periphery. They also suggest that the chromatin looping connecting genes and regulatory elements at the NL are plausible and may provide a means of regulation of gene expression in LADs.  (a-c) CMKLR1 locus: (a) FISH images of changes in proximity of the CMKLR1 gene (red probe) and a putative (GeneHancer-predicted and ECHi-C-predicted) enhancer 480 kb away (green probe) on D0, D1 and D3 of differentiation; blue, DAPI; scale bars, 5 µm in main, 0.5 µm in insets; a z-score RNA-seq heatmap is shown for D0, D1, D3 (left to right), adapted from [10]; (b) probe-probe distances; ** p = 0.04; unpaired two-tailed t-test; (c) distances from probe to nuclear periphery outlined by DAPI staining (green probes). (d-f) FISH analysis of the OTUD1 locus and a putative enhancer 348 kb away, as in (a-c); * p = 0.08, ** p = 0.04; unpaired two-tailed t-test. (g-i) FISH analysis of the LRATD1 locus and a putative enhancer 510 kb away, as in (a-c). For all loci, n = 50 nuclei each with two green and two red FISH signals; box plots: bar, median; cross, mean; box, 25-75% percentile; whiskers, min-max. These results provide evidence of enhancer connectivity at the nuclear periphery. They also suggest that the chromatin looping connecting genes and regulatory elements at the NL are plausible and may provide a means of regulation of gene expression in LADs.

Predominance of Enhancer Connections within cLADs
Concordant with earlier findings [16], we previously found that enhancers from the GeneHancer database (double elite set) [45] imputed to genes in LADs are mainly shortrange enhancers [10]. To examine this further, we determined the proportions of ECHi-C cLAD bait connections which are localized within or outside LADs.
Data are shown in Figure 4a. (i) Over 70% of K4 bait targets were within cLADs, with about one-third of these at an H3K4me1 site. (ii) More than 20% of K4 and nonK4 baits interacted with targets in vLADs or in inter-LADs, which is consistent with our predictions of gene-enhancer contacts [10]. (iii) The majority of nonK4 baits also targeted sites within cLADs, with only a minor fraction targeting an H3K4me1 site. (iv) Lastly, most K4 and nonK4 baits targeted sites outside H3K4me1, largely in domains of H3K9me3-marked heterochromatin (see also Figure 2e). We conclude that most baited enhancers in cLADs form intra-cLAD interactions and, among these, K4 baits contact weak or strong enhancers more frequently than non-K4 baits. This suggests that enhancer-enhancer connections can occur within LADs, either within or between active regions in these LADs.

Predominance of Enhancer Connections within cLADs
Concordant with earlier findings [16], we previously found that enhancers from the GeneHancer database (double elite set) [45] imputed to genes in LADs are mainly shortrange enhancers [10]. To examine this further, we determined the proportions of ECHi-C cLAD bait connections which are localized within or outside LADs.
Data are shown in Figure 4a. (i) Over 70% of K4 bait targets were within cLADs, with about one-third of these at an H3K4me1 site. (ii) More than 20% of K4 and nonK4 baits interacted with targets in vLADs or in inter-LADs, which is consistent with our predictions of gene-enhancer contacts [10]. (iii) The majority of nonK4 baits also targeted sites within cLADs, with only a minor fraction targeting an H3K4me1 site. (iv) Lastly, most K4 and nonK4 baits targeted sites outside H3K4me1, largely in domains of H3K9me3-marked heterochromatin (see also Figure 2e). We conclude that most baited enhancers in cLADs form intra-cLAD interactions and, among these, K4 baits contact weak or strong enhancers more frequently than non-K4 baits. This suggests that enhancer-enhancer connections can occur within LADs, either within or between active regions in these LADs.  ASCs. TssA, active TSS; EnhATss, active enhancer at TSS; EnhWkTss, weak enhancer at TSS; En-hBivTss, bivalent enhancer at TSS; EnhBiv, bivalent enhancer; ReprPc, repressed by Polycomb; Quies, quiescent; Tx, transcription; HetG, heterochromatin in gene; Het, heterochromatin; EnhWk, weak enhancer; EnhWkG, weak enhancer in gene; EnhAG1, active enhancer in gene, type 1; EnhA, active enhancer; EnhAG2, active enhancer in gene, type 2 (weaker than EnhAG1).

Chromatin Landscape of ECHi-C Connections
We next characterized the chromatin landscape of these baited enhancer targets. To this end, we initially applied the 25-state ENCODE ChromHMM [46] to ASCs, but found the ENCODE "Quiescent" state (devoid of any detectable mark) to be effectively partly enriched in H3K9me3 in our ASC dataset [28]. Thus, we profiled by ChIP-seq H3K4me1, H3K4me3, H3K27me3, H3K27ac, H3K9me3 and H3K36me3 in ASCs, and using ChromHMM, learned a 15-chromatin state model recapitulating the chromatin landscape of ASCs (Figure 4b). We then determined the distribution of cLAD baited enhancer targets across all 15 chromatin states, considering ≥ 1 bp overlap between a target with a given state to attribute a target to that state.

Lamin A/C Restricts Gene Expression in LADs
Elevated LMNB1 and LMNA/C levels flanking active cLAD regions could restrict conformational changes of chromatin [47] and hence would contribute to gene repression in these regions. To test this, we knocked down using siRNAs, LMNA/C, LMNB1 or both in undifferentiated ASCs (Figure 5a and Figure S2) and induced differentiation. We assessed by RT-qPCR expression of the repressed IL5RA gene, which lies immediately 5 of the TRNT1-CRBN active region in cLAD of chromosome 3. (Figure 5b). Interestingly, IL5RA is in the same topologically associated domain (TAD) as TRNT1 and CRBN in ASCs [28] ( Figure 5b) and shares enhancers with these genes as part of an enhancer hub (Figure 5c).
Whereas the expression of TRNT1 and CRBN was maintained in knocked down cells, IL5RA expression was induced by as early as D3 after the knock-down of LMNA/C, or of both LMNA/C and LMNB1, but not after the knock-down of LMNB1 alone (Figure 5d). IL5RA expression increased by D6 and reached a similar level whether only LMNA/C or LMNA/C and LMNB1 were knocked down. In contrast, the neighboring CNTN4 gene, located in a separate TAD part of a larger assembly of TADs forming a repressive TAD clique at the NL [28] (Figure 5b), was not activated after LMNA/C or LMNB1 knock-down (not shown). Thus, downregulation of LMNA/C is sufficient to elicit activation of the silent IL5RA locus within the TRNT1-CRBN TAD, but in the neighboring TAD. Genes 2023, 14, x FOR PEER REVIEW 10 of 14

Discussion
We report the localization of >200 active genes in cLADs defined across early adipose differentiation time-points within euchromatic regions of low-contact frequencies with LMNB1 and LMNA/C. Of note, recent work has shown that RNA-seq underestimates the total number of effectively active (transcribing) genes at the NL, as shown by mapping nascent transcripts using global run-on sequencing (GRO-seq) [16]. Thus, transcriptional activity in LADs is expectedly to be higher than what is reported here. Lower lamin enrichment in active LAD regions [17][18][19] implies chromatin dissociation from the NL. This may be mediated by transcription [18] or by CTCF and cohesin, which also occupy active sites in LADs [19].
NL detachment can extend beyond the activated gene into non-transcribed regions, non-activated neighboring genes or whole transcription units [18]. This could however depend on the chromatin context surrounding the active locus. In ASCs, the repressed IL5RA gene, flanking the active TRNT1-CRBN locus, is enriched in LMNB1 and H3K9me3. All three genes reside in the same TAD [28] but only the TRNT1-CRBN locus is depleted of H3K9me3 and LMNB1 low . Thus, proximal interactions with lamins may prevent activation of entire transcription units (as TADs) at the NL, as suggested by IL5RA activation after the downregulation of LMNA/C. Several possibilities may account for the effect of LMNA/C, and not LMNB1, on the regulation of IL5RA, and probably of other genes, in the flanks of active regions in LADs. Of note, in our system, knock-down of LMNB1 was incomplete, so we cannot formally exclude an involvement of LMNB1 in gene regulation near these regions. (i) LMNA/C is required for proper genomic localization of H3K27me3 [48]. By displacing H3K27me3, elimination of LMNA/C may favor spreading of H3K27ac into the LAD and enable IL5RA activation, similar to the activation of the NL-associated Tcrb locus after deletion of a LAD border [23]. (ii) LMNA/C, but not LMNB1, has been shown to anchor chromatin at the nuclear envelope [49]; downregulation of LMNA/C may elicit local detachment of chromatin domains containing repressed genes from the NL, creating a permissive environment compatible with induction of IL5RA. (iii) LMNA/C restricts chromatin mobility in the nucleus [47], so knock-down of LMNA/C may enhance chromatin motion; this could, in turn, facilitate chromatin looping events and enhancer connectivity, as well as accessibility to the transcription machinery. (iv) We have earlier shown, using a ChIP-on-chip approach, that LMNA/C can bind promoters, an event that correlates with transcriptional inactivity of the gene [50]; LMNA/C downregulation may alleviate this local repression on the IL5RA promoter by mechanisms involving one of more of the aforementioned possibilities. These studies altogether suggest that local detachment of chromatin from the NL may facilitate the formation of regulatory loops at or near the NL, which could bring multiple regulatory elements close to active genes.
Accordingly, we provide evidence of multiple connections between ECHi-C baited enhancers in LADs and other enhancers also in these LADs. In bone marrow mesenchymal stem cells, enhancers have been shown using ECHi-C to exist as distinct "communities" that can be reconfigured via changes in enhancer interactions [21]. Mapping these ECHi-C baited enhancers onto cLAD in our adipogenic system suggests that some of these networks pre-exist at the NL in the form of hubs restricted to active cLAD regions (short-range interactions) or as wider networks spanning larger genomic distances. Detachment of chromatin from the NL in these regions may facilitate the remodeling of these enhancer networks.
Enhancers of active genes in LADs have earlier been inferred based on GRO-seq data and histone modifications (H3K4me1, H3K27ac) combined with a nearest target gene model [16]. In this approach, close-range enhancers are associated with highest gene activity, irrespective of NL association; however, most enhancers of NL-associated genes are found to lack NL association [16], supporting a view of long-range interactions. This could be because NL-associated genes are expressed at a lower level than genes in the nuclear interior, and that longer gene-enhancer distances correlate with lower gene expression levels [16]. In contrast, using the double-elite enhancer set from GeneHancer enriched in H3K27ac in our adipose system, we previously found that the majority of enhancers imputed to differentially expressed genes in cLADs are in cLADs [10], corroborating our ECHi-C analysis. Our study further argues that cLAD gene enhancers are weak enhancers (with low H3K27ac), which could also account for the lower expression level of genes at the NL.
The distinct outcomes of the two studies [10,16] may be influenced by the inference method used, the nature of the expressed LAD genes (e.g., housekeeping or cell type-specific [10,17]) and the expression level of the gene [16]. Yet, both raise the possibility of chromatin loops bringing LAD genes and enhancers in proximity at the nuclear periphery. Chromatin at the NL emerges as a dynamic compartment undergoing topological remodeling at various scales [10,18,28,51] to ensure lineage specificity of gene expression.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki and approved by the Norwegian Regional Ethics Committee of Southern Norway (REK 2013/2102).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.