In vivo dissection of a clustered-CTCF domain boundary reveals developmental principles of regulatory insulation – Nature.com

A genetic setup to investigate boundary function in vivo

We previously demonstrated that a 150-kilobase (kb) region, the EP boundary, is sufficient to segregate the regulatory activities of the Epha4 and Pax3 TADs10 (Extended Data Figs. 1 and 2). The DelB background carries a large deletion that removes this boundary region, and the Epha4 gene, resulting in the ectopic interaction between the Epha4 limb enhancers and the Pax3 gene. This causes Pax3 misexpression and the shortening of fingers (brachydactyly) in mice and in human patients. In contrast, the DelBs background carries a similar deletion but not affecting the EP boundary, which maintains the Epha4 and Pax3 TADs and confines the Epha4 enhancers within their own regulatory domain (Fig. 1a and Extended Data Fig. 1).

a, cHi-C maps from E11.5 distal limbs from DelBs mutants at 10-kb resolution. Data were mapped on a custom genome containing the DelBs deletion (n=1 with an internal control comparing 6 different experiments; Methods). The red rectangle marks the EP boundary region. Insets represent a magnification (5-kb resolution) of the centromeric (left) and telomeric (right) loops highlighted by brackets on the map. Cen, Centromeric; Tel, Telomeric. Arrowheads represent reverse- (light blue) and forward- (orange) oriented CBSs. Below, Lac-Z staining (left) and WISH (right) of E11.5 mouse forelimbs show activation pattern of Epha4 enhancers and Pax3 expression, respectively. b, CTCF ChIPseq track from E11.5 mouse distal limbs. Schematic shows CBS orientation. c, Insulation score values. The gray dot represents the local minima of the insulation score at the EP boundary. BS, boundary score. d, Relationship between BS and the number of CBSs (data from ref. 26). The boxes in the boxplots indicate the median and the first and third quartiles (Q1 and Q3). Whiskers extend to the last observation within 1.5 times the interquartile range below and above Q1 and Q3, respectively. The rest of the observations, including maxima and minima, are shown as outliers. N=8,127 insulation minima found in mESC Hi-C matrices. e, WISH shows Pax3 expression in E11.5 forelimbs from CBS mutants. Note Pax3 misexpression on the distal anterior region in R1, F1 and F2 mutants (white arrowheads). Scale bar, 250m. f, Pax3 qPCR analysis in E11.5 limb buds from CBS mutants. Bars represent the mean and white dots represent individual replicates. Values were normalized against DelBs mutant (Ct) (two-sided t-test *P0.05; NS, nonsignificant; P values from left to right: DelBs versus R1: 0.02; DelBs versus R2: 0.11; DelBs versus F1: 0.02; DelBs versus R3: 0.23; DelBs versus F2: 0.02; DelBs versus R4: 0.73). Cen, Centromeric; Tel, Telomeric.

To characterize the EP boundary in vivo, we performed CTCF ChIPseq on developing limbs. This analysis revealed the presence of six clustered CBSs at the EP boundary region (Fig. 1a,b and Extended Data Fig. 2), a profile that is conserved across tissues25,26. CTCF motif analyses confirmed the divergent orientation of these sites, a signature of TAD boundaries, with four CBSs in reverse (R) and two in forward orientation (F). Other features associated with boundaries, such as active transcription or housekeeping genes, were not found in the region27 (Extended Data Fig. 3). cHi-C data from DelBs stage E11.5 distal limbs28 revealed chromatin loops connecting the two forward-oriented CBSs (F1 and F2) with the telomeric boundary of the Pax3 TAD, and the centromeric boundary of the Epha4 TAD with the reverse-oriented CBSs R1, R2 and R3 (Fig. 1a,b). However, the close genomic distances between R2 and F1 and between R3 and F2 preclude the unambiguous assignment of loops to specific sites. RAD21 (cohesin subunit) ChIPseq experiments in E11.5 distal limbs revealed that R1, F1 and F2, as well as R2 and R3 to a lesser degree, are bound by cohesin (Extended Data Fig. 3), an essential component for the formation of chromatin loops21,29,30. These results delineate the EP element as a prototypical boundary region with insulating properties likely encoded and controlled by CBSs.

Boundary regions are predominantly composed of CBS clusters31, suggesting that the number of sites might be relevant for their function. We explored this by calculating boundary scores32 on available Hi-C maps26, and categorizing boundaries according to CBS number. We observe that boundary scores increase monotonically with CBS number, reaching a stabilization at ten CBSs (Fig. 1d). According to this distribution, the EP boundary falls within a range where its function might be sensitive to alterations on CBS number. To test this, we employed a mouse homozygous embryonic stem cell (mESC) line for the DelBs background28, which we edited to generate individual homozygous deletions for each of the six CBSs of the EP boundary region (Supplementary Fig. 1). ChIPseq experiments revealed that the disruption of the binding motif was sufficient to abolish CTCF recruitment (Supplementary Fig. 2). Subsequently, we employed tetraploid complementation assays to generate mutant embryos and measure the functional consequences of these deletions in vivo33,34.

Whole-mount in situ hybridization (WISH) on E11.5 mutant embryos revealed that the insulation function of the EP boundary can be sensitive to individual CBS perturbations (Fig. 1e). However, this effect was restricted to CBSs displaying prominent RAD21 binding (R1, F1 and F2) (Extended Data Fig. 3). The altered boundary function was evidenced by Pax3 misexpression on a reduced area of the anterior limb, while the expression domains in other tissues remained unaltered (Supplementary Fig. 3). The disruption of the other CBSs (R2, R3 and R4) did not alter Pax3 expression, demonstrating that the EP boundary can also preserve its function despite a reduction in CBS number.

To quantify Pax3 misexpression, we performed quantitative PCR (qPCR) in E11.5 forelimbs. Similarly, we observed a modest, but significant, upregulation in R1, F1 and F2 mutants (Fig. 1f). Importantly, the functionality of individual CBSs is not strictly correlated with CTCF occupancy as the deletion of R3, displaying the highest levels of CTCF binding among the cluster (Fig. 1b and Extended Data Fig. 3), does not result in measurable transcriptional changes (Fig. 1f). Thus, while CBS number influences insulation, the characteristics of individual sites are major determinants of boundary function.

To explore CBS cooperation, we retargeted our R1 mESC line to generate double knockout mutants with different (R1+F2) or identical CBS orientations (F1 and F2 in F-all) (Fig. 2a). WISH revealed an expanded Pax3 misexpression towards the posterior region of the limb, demonstrating that the EP boundary is compromised in both mutants. Next, we determined the nature of CBS cooperation by qPCR. These experiments revealed that, in both mutants, Pax3 misexpression exceeded the summed expression levels from the corresponding individual deletions (Fig. 2b). These negative epistatic effects indicate that CBSs are partially redundant, compensating for the absence of each other.

a, WISH shows Pax3 expression in E11.5 forelimbs from CBS mutants. Arrowheads represent reverse- (light blue) and forward- (orange) oriented CBSs. Crosses indicate deleted CBSs. Note increased Pax3 misexpression towards the posterior regions of the limb. Scale bar, 250m. b, Pax3 qPCR analysis in E11.5 limb buds from CBS mutants. Bars represent the mean and white dots represent individual replicates. Values were normalized against DelBs mutant (Ct) (**t-test **P0.01; R1+F2 versus F-all: 0.008). c, cHi-C maps from E11.5 mutant distal limbs at 10-kb resolution (top). Data were mapped on a custom genome containing the DelBs deletion (n=1 with an internal control comparing 6 different experiments; Methods). Insets represent a magnification (5-kb resolution) of the centromeric (left) and telomeric (right) loops highlighted by brackets on the map. Gained or lost chromatin loops are represented by full or empty dots, respectively. Subtraction maps (bottom) showing gain (red) or loss (blue) of interactions in mutants compared with DelBs. d, Insulation score values. Lines represent indicated mutants. Dots represent the local minima of the insulation score at the EP boundary for each mutant. e, Virtual 4C profiles for the genomic region displayed in c (viewpoint in Pax3). The light-gray rectangle highlights the Epha4 enhancer region. Note increased interactions between the Pax3 promoter and the Epha4 enhancer in R1+F2 and F-all (purple and orange) compared with DelBs mutants (gray).

To gain insights on the mechanisms of CBS cooperation, we generated cHi-C maps of the EP locus from E11.5 distal limbs (Fig. 2c and Supplementary Fig. 4). Maps from R1+F2 embryos denoted a clear partition between the EphaA4 and Pax3 TADs, analogous to DelBs control mutants (Fig. 2c). However, subtraction maps revealed decreased intra-TAD interactions for the Epha4 and Pax3 TADs, and a concomitant increase in inter-TAD interactions. In addition, we observed the appearance of a loop connecting the outer boundaries of the Epha4 and Pax3 TADs (meta-TAD loop; Extended Data Fig. 4)35. Accordingly, the boundary score of the EP boundary in R1+F2 mutants was decreased, reflecting a weakened structural insulation (Fig. 2d). Virtual Circular Chromosome Conformation Capture (4C) profiles revealed increased chromatin interactions between the Pax3 promoter and the Epha4 limb enhancers (Fig. 2e), consistent with the upregulation of Pax3. In addition, two of the chromatin loops that connect the EP boundary and the telomeric boundary were abolished, due to the deletion of the F2 anchor and the associated loss of RAD21 (Fig. 2c and Extended Data Figs. 4 and 5). Consequently, the adjacent chromatin loop exhibited a compensatory effect, with increased interactions mediated by the F1 anchor, consistent with higher RAD21 occupancy (Extended Data Figs. 4 and 5). At the centromeric site, the deletion of R1 causes the relocation of the loop anchor towards an adjacent region containing a reverse-oriented (R2) and the only remaining forward CBS (F1). While the loop extrusion model would predict a stabilization at a reverse CBS15,16, the short genomic distance between R2 and F1 precludes an unambiguous assignment of the loop anchor. We also observed increased contacts at R3 and R4, suggesting that these sites are functionally redundant.

Then, we examined cHi-C maps from F-all mutants, which display a more pronounced Pax3 misexpression (Fig. 2b). Interaction maps revealed a partial fusion of the Epha4 and Pax3 domains (Fig. 2c), accompanied by a notable decrease of the boundary score (Fig. 2d). Virtual 4C profiles confirmed increased interactions between Pax3 and the Epha4 enhancers in F-all compared with R1+F2 mutants, in agreement with the more pronounced Pax3 upregulation (Fig. 2e). The deletion of all CBSs with forward orientation abolishes the chromatin loops connecting with the telomeric Pax3 boundary (Fig. 2c and Extended Data Fig. 4). Towards the centromeric side, R1 maintains RAD21 binding and its chromatin loop with the centromeric Epha4 boundary (Extended Data Figs. 4 and 5). However, other chromatin loops are still discernible and anchored by the R3 and R4 sites, confirming that these sites perform distinct yet partially overlapping functions. These results demonstrate that CBSs can cooperate but also partially compensate for the absence of each other, conferring functional robustness to boundaries.

Chromatin loops are predominantly anchored by CBS pairs with convergent motif orientation14,36. Intriguingly, we observed that the combined F1 and F2 deletion (F-all) not only disrupts the loops in the expected orientation (telomeric), but also impacts the centromeric one, as observed in the subtraction maps (Fig. 2c). This effect is noticeable at the R2/F1 site, which was associated with a centromeric chromatin loop in the DelBs background (Fig. 1a). This demonstrates that the main loop anchor point was not the R2 but the F1 site (Extended Data Fig. 4), suggesting that this CBS can form loops in a nonconvergent orientation. Such mechanism is described by the loop extrusion model, which predicts that loops could create steric impediments that might prevent additional cohesin complexes from sliding through anchor sites15,16. This effect would stabilize these additional cohesin complexes, resulting in the establishment of simultaneous and paired nonconvergent and convergent loops (Fig. 3a).

a, Schematic of a convergent loop that indirectly generates a nonconvergent loop in the opposite direction. b, Percentage of loop anchors establishing bidirectional loops (n=12,635 loops from mESCs from ref. 26). Anchor categories: convergent-only (only CBSs oriented in the same direction as their anchored loops, n=7,769), nonconvergent (anchor loops in a direction for which they lack a directional CBS, n=960) and no-CTCF (no CBS, n=3,906). c, Loop strengths in pairs of convergent/nonconvergent loops classified into Non-conv.-associated (nonconvergent loop sharing the nonconvergent anchor with a convergent loop in the opposite direction, n=322) and Conv.-associated (convergent loop sharing one anchor with a nonconvergent loop in the opposite direction, n=496). Boxplots defined as in Fig. 1c. Two-sided BenjaminiHochberg-corrected MannWhitney U-test P=6.2106. d, Aggregated loop signal for categories in c. Arrows represent CBS orientation. e, Pax3 WISH in E11.5 forelimbs from CBS mutants. Arrowheads represent reverse- (blue) and forward- (orange) oriented CBSs. Crosses indicate deleted CBSs. Note the positive correlation between expanded Pax3 misexpression and increased number of deleted CBSs. Scale bar, 250m. f, Pax3 qPCR analysis in E11.5 limbs from CBS mutants. Bars represent mean and dots individual replicates. Values were normalized against DelBs mutant (Ct). Note the positive correlation of Pax3 misexpression with the increase in deleted CBSs (Pearson correlation significantly>0; ***P0,001). g, cHi-C maps from E11.5 mutant distal limbs at 10-kb resolution (top). Data were mapped on a custom genome containing the DelBs deletion (n=1 with an internal control comparing 6 different experiments; Methods). Insets represent a magnification (5-kb resolution) of the centromeric (left) and telomeric (right) loops highlighted by brackets on the map. Gained or lost chromatin loops are represented by full or empty dots. Subtraction maps (bottom) showing gain (red) or loss (blue) of interactions in mutants compared with DelBs. h, Insulation score values. Dots represent the local minima of the insulation score at the EP boundary for each mutant. i, Virtual 4C profiles for the region in g (viewpoint in Pax3). The gray rectangle highlights Epha4 enhancers. Note increased interactions between the Pax3 promoter and the Epha4 enhancers in R-all compared with DelBs.

We searched for further biological indications of this mechanism by analyzing ultra-high-resolution Hi-C datasets26. First, we identified loop anchors and classified them according to the orientation of their CBS motif and associated loops. Loop anchors were split into convergent-only (only CBSs oriented in the same direction as their anchored loops), nonconvergent (anchor loops in a direction for which they lack a directional CBS) and no-CTCF (no CBS). While most loop anchors belong to the convergent-only category14,36, 7.6% of them were classified as nonconvergent. Then, we explored whether these nonconvergent loops could be explained by the nonconvergent anchor simultaneously establishing a convergent loop in the opposite direction (Fig. 3a). We calculated the frequency of anchors involved in bidirectional loops for each category and discovered that, while only 5% of convergent-only or no-CTCF anchors participate in bidirectional loops, this percentage increases significantly up to 45% for nonconvergent anchors (Fig. 3b; chi-squared test, P<10225). To gain further insights into the mechanisms that establish convergent/nonconvergent loop pairs, we calculated the strength of each corresponding paired loop22. We observed that the convergent loops linked to a nonconvergent loop are significantly stronger than their nonconvergent counterparts (Fig. 3c,d; MannWhitney U-test, P=6106). Next, we explore if convergent loops paired to nonconvergent loops are particularly strong in comparison with other types of convergent loops. This analysis revealed that the strength of these convergent loops is similar to other unpaired convergent loops across the genome (Extended Data Fig. 6; single-sided convergent category). However, paired convergent/nonconvergent loops appear to be mechanistically different from unpaired loops, as they are more often associated with TAD corners (Extended Data Fig. 6c; chi-squared test, P<3.5106) and therefore connect anchor points that are located farther away in the linear genome (Extended Data Fig. 6d; MannWhitney U-test, P<4.8108). A comparison against pairs of convergent/convergent loops, which are similarly associated with TAD corners (Extended Data Fig. 6b; category double-sided convergent), revealed that the convergent loops in convergent/nonconvergent pairs are on average stronger (MannWhitney U-test, P=7105). This type of convergent/nonconvergent loops can be observed at relevant developmental loci, such as the Osr1, Ebf1 and Has2 loci (Extended Data Fig. 7). Overall, our analyses suggest that a considerable number of nonconvergent loops could be mechanistically explained by the presence of a stronger and convergent chromatin loop in the opposite orientation and anchored by the same CBS.

To validate these findings in vivo, we sequentially retargeted our R1 mESCs to create a mutant that only retains the forward F1 and F2 sites, which have strong functionality (Fig. 2a,b). During the process, we obtained intermediate mutants with double (R1+R3) and triple CBS deletion (R1+R3+R4), as well as the intended quadruple knockout lacking all reverse CBSs (R-all). WISH revealed an expanded Pax3 expression pattern towards the posterior limb region, an effect that increases with the number of deleted CBSs (Fig. 3e). Expression analyses by qPCR confirmed a significant increasing trend in Pax3 misexpression levels across mutants (Fig. 3f; Pearson correlation>0, P2107). These results demonstrate again that R2, R3 and R4 are functionally redundant sites, despite the absence of measurable effects upon individual deletions (Fig. 1b). However, we noted that Pax3 levels were only moderately increased (threefold) compared with the expression in mutants retaining only-reverse CBSs (ninefold, F-all). Importantly, R-all mutants retain two intact CBSs in the forward orientation, while up to four CBSs are still present in F-all mutants, suggesting that these two forward CBSs (F1 and F2) grant most of the insulator activity of the EP boundary. These experiments indicate that the functional characteristics of specific CBSs can outweigh other predictive parameters of boundary function such as the total number of sites.

As expected, cHi-C maps from R-all mutant limbs revealed a clear partition between the Epha4 and Pax3 TADs (Fig. 3g), consistent with the reduced Pax3 misexpression. Boundary scores at the EP boundary were also only moderately reduced (Fig. 3h), in comparison with the broader effects of the F-all mutant (Fig. 2d). Accordingly, intra-TAD interactions modestly decreased while inter-TAD interactions increased, as also observed in virtual 4C profiles (Fig. 3i). Despite the multiple deletions, the telomeric chromatin loops remained unaffected and anchored by the F1 and F2 sites, both occupied by RAD21 (Fig. 3g and Extended Data Figs. 4 and 5). However, we noticed the persistence of centromeric chromatin loops anchored by the F1 and F2 sites, despite their nonconvergent forward orientation. A higher contact intensity is observed at F1, which would be the first CBS encountered by cohesin complexes sliding from the centromeric side (Extended Data Figs. 4 and 5).

Finally, we investigated if the formation of nonconvergent loops might be associated with the accumulation of cohesin complexes over a limited number of CBSs. We generated a mutant that only retains the R3 CBS (R3-only), which is prominently bound by CTCF (Fig. 1b). We hypothesized that, in the absence of others, this CBS may accumulate the cohesin and form a nonconvergent loop. However, although R3 was the only site able to stall cohesin in this background (Extended Data Fig. 4), cHi-C maps revealed a single convergent loop towards the centromeric side (Extended Data Fig. 8). This loop displays a weak insulator function, denoted by a decreased boundary score, an Epha4 and Pax3 TAD fusion and prominent Pax3 misexpression. Therefore, our results in transgenic mice support our findings at the genome-wide level (Fig. 3ac), demonstrating that specific CBSs can create chromatin loops independently of their motif orientation, seemingly through loop interference.

Previous studies identified divergent CBS clusters as a signature of TAD boundaries, suggesting a role on insulation13,31. While our analysis on mutants with reverse-only CBS orientation (F-all) showed a severe impairment of boundary function (Fig. 2c), this was not the case for R-all mutants, which retain CBSs only in the forward orientation (Fig. 3f). Indeed, the levels of Pax3 misexpression evidenced that insulation is more preserved in R-all than in R1+F2 mutants, which still conserve a divergent CBS signature (Fig. 2c).

This prompted us to explore the relation between CBS composition at boundaries and insulation strength. We examined available Hi-C datasets, classifying boundary regions according to different parameters of CBS composition (that is, number and orientation) and calculating boundary scores (Fig. 4a). Our analysis revealed that, for the same CBS number, boundaries with divergent signatures generally display more insulation than their nondivergent counterparts. However, up to 6% of nondivergent boundaries display scores above 1.0, a value associated with robust functional insulation (Fig. 1c). Manual inspection at specific loci showed that nondivergent boundaries with strong boundary scores present clear TAD partition and no evidence of coregulation for genes located at either side (Extended Data Fig. 9). These results suggest that a divergent signature is not strictly required to form strong functional boundaries.

a, Relation between BSs and the number of CBSs for divergent and nondivergent boundaries in mESC Hi-C data26. Boxplots defined as in Fig. 1c. b, WISH shows Pax3 expression in E11.5 forelimbs from CBS mutants. Arrowheads represent reverse- (light blue) and forward- (orange) oriented CBSs. Crosses indicate deleted CBSs. Light-gray rectangle marks inverted region. Note similar Pax3 misexpression pattern between F-all-Inv and F-all mutants. Scale bar, 500m. c, Pax3 qPCR analysis in E11.5 limb buds from CBS mutants. Bars represent the mean and white dots represent individual replicates. Values were normalized against DelBs mutant (Ct) (two-sided t-test P value). d, cHi-C maps from E11.5 mutant distal limbs at 10-kb resolution (top). Data mapped on custom genome containing the DelBs deletion and the inverted EP boundary (n=1 with an internal control comparing 6 different experiments; Methods). Insets represent a magnification (5-kb resolution) of the centromeric (left) and telomeric (right) loops highlighted by brackets on the map. Gained or lost chromatin loops are represented by full or empty dots, respectively. Subtraction maps (bottom) showing gain (red) or loss (blue) of interactions in mutants compared with DelBs. e, Insulation score values. Lines represent mutants. Dots represent the local minima of the insulation score at the EP boundary for each mutant. f, Virtual 4C profiles for the genomic region displayed in d (viewpoint in Pax3). Light-gray rectangle highlights Epha4 enhancer region. Note similar interaction profile between F-all-Inv (yellow) and F-all mutants (orange).

Next, we explored if the genomic contexts might explain the prominent insulation differences between only-reverse (F-all) or only-forward (R-all) mutants. To evaluate this, we generated a mutant with a homozygous inversion of the boundary region, on the F-all background (F-all-Inv) (Fig. 4b and Supplementary Fig. 5).

WISH and qPCR experiments showed that Pax3 expression is almost indistinguishable from the F-all mutants, both spatially and at the quantitative level (Fig. 4b,c). Moreover, cHi-C maps from F-all-Inv mutants revealed a similar fusion of the Epha4 and Pax3 TADs (Fig. 4d). However, subtraction maps showed a redirection of chromatin loops, which now interact mainly with the telomeric Pax3 boundary instead of the centromeric Epha4 boundary. These ectopic loops are mainly anchored by the R1 site, which preserves its marked functionality. Despite these local differences, boundary scores and virtual 4C profiles remained comparable between F-all-Inv and F-all mutants (Fig. 4e,f). These results suggest that the orientation of entire boundary regions, as well as the differences in the surrounding genomic context, play a minor role in insulator function.

To determine to what extent CTCF binding contributes to the EP boundary function, we generated a sextuple knockout with all CBSs deleted (ALL). WISH revealed a further expansion of Pax3 misexpression, covering the distal limb entirely. This expanded expression mirrors that of DelB mutants, in which the entire boundary region is deleted (Fig. 5a). Expression analyses revealed that Pax3 misexpression in ALL mutants exceeds the combined sum of expression from R-all and F-all mutants (Fig. 5b), again indicating the cooperative and redundant CBS action. Intriguingly, Pax3 misexpression in the R3-only background was comparable to ALL, suggesting that a functionally weak CBS is not sufficient to hinder enhancerpromoter communication (Extended Data Fig. 8). Nevertheless, ALL mutants only reach 65% of the Pax3 misexpression observed in DelB mutants (Fig. 5b), which may be attributed to the 150-kb inter-CBS region that differentiates both mutants.

a, WISH shows Pax3 expression in E11.5 forelimbs from CBS mutants. Arrowheads represent reverse- (light blue) and forward- (orange) oriented CBSs. Crosses indicate deleted CBSs and the gray rectangle represents the deleted region. Note the similarities in expression pattern between mutants. Scale bar, 250m. b, Pax3 qPCR analysis in E11.5 limb buds from CBS mutants. Bars represent the mean and white dots represent individual replicates. Values were normalized against DelBs mutants (Ct) (*two-sided t-test P0.05, ALL versus DelB: 0.03). c, cHi-C maps from E11.5 mutant distal limbs at 10-kb resolution (top). Data mapped on custom genome containing the DelBs deletion (n=1 with an internal control comparing 6 different experiments; Methods). Insets represent a magnification (5-kb resolution) of the centromeric (left) and telomeric (right) loops highlighted by brackets on the map. Gained or lost chromatin loops represented by full or empty dots, respectively. Subtraction maps (bottom) showing gain (red) or loss (blue) of interactions in mutants compared with DelBs (left) and DelB (right). d, Insulation score values. Lines represent mutants. Dots represent the local minima of the insulation score at the EP boundary for each mutant. e, Virtual 4C profiles for the genomic region displayed in c (viewpoint in Pax3). Light-gray rectangle highlights Epha4 enhancer region.

To investigate the reduced Pax3 misexpression in ALL, compared with DelB mutants, we performed cHi-C experiments (Fig. 5c). These experiments revealed a prominent Epha4 and Pax3 TAD fusion, with increased intensity of the meta-TAD loop (Extended Data Fig. 4). This results from the severe disruption of the EP boundary, denoted by a reduced boundary score (Fig. 5d) and the complete absence of RAD21 binding or anchored loops (Extended Data Figs. 4 and 5). In fact, the interaction profile at the EP boundary is not different from other internal locations of the Epha4 TAD (Fig. 5c). Of note, higher insulation is observed in R3-only compared with ALL, despite the comparable Pax3 misexpression between both genetic backgrounds (Extended Data Fig. 8). However, virtual 4C profiles from ALL and R3-only mutants confirmed a similar interaction between Epha4 enhancers and Pax3 (Fig. 5e and Extended Data Fig. 8). These enhancergene interactions were reduced in comparison with DelB, in which Pax3 misexpression is more prominent (Fig. 5e and Extended Data Fig. 8). ChIPseq datasets for epigenetic marks did not reveal additional regions with regulatory potential within the 150-kb region (Extended Data Fig. 3), indicating that the enhanced Pax3 misexpression in DelB mutants is unlikely caused by the deletion of regulatory elements. Taken together, these results suggest that enhancerpromoter distances might influence gene expression levels.

PAX3 misexpression during limb development can cause shortening of thumb and index finger (brachydactyly), in human patients and mouse models10. Therefore, our mutant collection provides an opportunity to study how boundary insulation strengths translate into developmental phenotypes.

We obtained mutant E17.5 fetuses and performed skeletal stainings, measuring relative digit length as a proxy for the phenotype (Fig. 6a,b). First, we analyzed R1 mutants, which displayed moderate Pax3 misexpression in the anterior distal limb (Fig. 1f). Finger length ratios revealed that R1 limbs develop normally, demonstrating that the detrimental effects of Pax3 misexpression can be partially buffered.

a, Skeletal staining of forelimbs from E17.5 mutant and control fetuses. White arrowheads indicate reduced index finger lengths. Black bracket shows the region of the finger measured for the quantification. Finger length correlates negatively with increased Pax3 misexpression. Scale bar, 500m. b, Index lengths relative to ring finger lengths in E17.5 mouse forelimbs. Bars represent the mean and white dots represent individual replicates. Values were normalized on control (CTRL) animals (two-sided t-test **P0.01; two-sided t-test ***P0.001; R1+F2 versus CTRL: 0.007; F-all versus CTRL: 0.0002). c, Correlation between the number of remaining CBSs at the EP boundary and the levels of Pax3 expression in the different mutants described in this study. Pearson regression lines are shown together with R2 values, both for the whole collection of mutants (black) and discarding combined CBS deletions involving CBSs with forward orientation (turquoise). d, Correlation and R2 between BSs and the brachydactyly phenotype penetrance measured as the index to ring finger length ratio for controls, R1+F2 and F-all mutants. The color of the dots represents the level of Pax3 limb misexpression as measured by qPCR. e, Model for boundary insulation as a quantitative modulator of gene expression and developmental phenotypes. Left, a strong boundary (B) efficiently insulates gene A from the enhancers located in the adjacent TAD (E). The boundary shows a cluster of CBSs with different orientations represented with arrowheads. The colored arrow represents a CBS with prominent contribution to boundary function. Middle, the absence of specific CBSs results in a weakened boundary, moderate gene misexpression (limb, indicated in yellow) and mild phenotypes (reduced digits, indicated in red and pointed out by white arrowhead). Right, the absence of the boundary causes a fusion of TADs, strong gene misexpression and strong phenotypes.

In contrast, R1+F2 mutants displayed a moderate reduction of index digit length (Fig. 6a,b), consistent with their increased Pax3 misexpression (Fig. 2b). This demonstrates that weakened boundaries can be permissive to functional interactions between TADs, resulting in altered transcriptional patterns and phenotypes. Importantly, the phenotypes of R1+F2 mutants occur despite an observable partition between Epha4 and Pax3 TADs and across a boundary region displaying high boundary scores (Fig. 2c,d; boundary score=0.8). Analyses on ultra-high-resolution Hi-C datasets26 revealed that many boundary scores fall within the ranges described in our mutant collection (Extended Data Fig. 10). Of note, 40% of boundaries display scores lower than 0.8. According to our observations, those boundaries could be permeable for functional interactions across domains.

Finally, we analyzed the F-all mutants, in which the Epha4 and Pax3 TADs appear largely fused (Fig. 2c). This disruption of TAD organization led to a prominent reduction of digit length (Fig. 6a,b), consistent with the higher Pax3 misexpression (Fig. 2b). Overall, these results illustrate how boundary insulation strength can modulate gene expression and developmental phenotypes, by allowing permissive functional interactions between TADs.

Continued here:
In vivo dissection of a clustered-CTCF domain boundary reveals developmental principles of regulatory insulation - Nature.com