Introduction

The precise regulation of gene expression is a fundamental pillar of developmental biology, guiding the development of complex organisms. Enhancers play a crucial role in orchestrating the finely tuned control of gene expression across diverse tissues and developmental stages. Enhancers function through the coordinated binding by transcription factors which in turn recruit components of regulatory complexes such as mediator; this complex of regulatory factors directly interacts with the RNA polymerase complex recruited to gene promoters to enact a cis-regulatory effect [1–4]. Although potentially separated by megabases of DNA, enhancers are brought into close spatial proximity to their gene targets and interact with RNA polymerase at the gene promoter [5,6]. These long range interactions can be mediated by the Cohesin complex [7] which has been proposed to extrude intervening chromatin through its ring-like structure, bringing distal regions into close proximity [8]. The CCCTC-Binding Factor (CTCF) has been shown to have a stabilizing role on Cohesin and aids in the formation of long-range chromatin interactions [9,10], however, the necessity of CTCF in long-range enhancer promoter communication globally has been questioned as the removal of these bound regions can have negligible impact on local chromatin topology and gene expression [10–12]. Global chromatin interactions partition the genome into Topologically Associated Domains (TADs), demarcated by boundaries wherein chromatin interactions occur much more frequently within a TAD than with neighbouring regions [13,14]. Regions within TADs show concerted epigenomic signatures and coordinated gene expression highlighting that at a higher level these can act as units of transcriptomic regulation [15–17].

During evolution, gene duplication events can give rise to paralogous genes encoding proteins that participate in shared cellular pathways. When both genes are maintained in the genome, the function of each usually diverges, either through modulating gene expression patterns or through changes to the coding sequence that alters protein function [18]. In tandem duplication events, paralogs remain in close proximity in the genome and are transcribed from the same DNA strand [19]. Although these duplicated genes often share expression patterns and appear co-regulated, it has been proposed that after gene duplication there is selection pressure to maintain gene dosage, by reducing paralog expression, so that stochiometric ratios of proteins are maintained [19–23]. In Drosophila, however, paralogs show a high degree of co-regulation, over 75% of paralogs with similar expression patterns display evidence of regulation by shared enhancers or the additive effects of multiple shared enhancers [24]. Furthermore, enhancer deletion in mouse embryonic stem cells (ESCs) revealed that gene paralogs can be co-regulated by shared enhancers [25]. These data fit with the gene neighborhood theory of gene expression regulation, whereby genes within the same TAD are largely co-regulated by shared enhancers [26–29], raising the question, how does gene expression control divergence for paralogs in a tandem duplication? Genomic separation of the paralogs appears to be a key factor in expression divergence. Studies have identified a strong correlation in the expression of tandem duplicated genes which are located within one megabase of each other, whereas paralogs which are separated in one species to different chromosomes show much lower correlation in their co-expression patterns [19]. This finding holds true when controlling for the age of the duplication as over longer timeframes, duplicated genes are more likely to reside on separate chromosomes [19,30,31].

Here, we focus on investigating the regulatory mechanisms controlling Lefty1 and Lefty2 transcription. These are paralogous genes central to both stem cell maintenance and the intricate processes of left-right patterning during vertebrate development. In Fugu and Flounder, only a single Lefty gene is present which covers the functions of both genes present in other species [32]. In zebrafish, both paralogs are present but arose from a whole genome duplication event in ray finned fishes [32]. In mammals, the tandem duplication event leading to the generation of the Lefty1 and Lefty2 paralogs was proposed to have occurred separately in mice and humans [33]. Ensembl gene trees predicts that these are separate events [34] and represent an evolutionarily young event [35]. Due to the nature of the duplication and gene function, Lefty2 has been proposed as the ancestral gene [32]. Given what is known of recently duplicated paralogs, these genes would be expected to have shared regulatory control in mammals and similar expression patterns, however, although both Lefty genes are expressed in ESCs and participate in pluripotency maintenance [36], they diverge in their tissue expression patterns in driving left-right axis patterning during embryonic development, with Lefty1 primarily being expressed in the prospective floor plate whereas Lefty2 is expressed in the lateral plate mesoderm [37]. Single-cell RNA-seq has identified that in addition to their divergence in tissue expression patterns, the temporal expression of each gene has diverged as well, with Lefty1 being expressed prior to E (embryonic day) 6.5 and losing expression by E7.4 while Lefty2 increases in expression at E6.7 and loses expression at E8 [38].

Despite their joint participation in pluripotency maintenance, and physical presence within the same TAD [39], which would suggest enhancer regulatory cross-talk occurs, we determined that transcription of the Lefty genes is governed by separate discrete enhancer elements in ESCs. These two enhancers are located upstream of each gene; the Lefty2 enhancer overlaps the previously identified asymmetric specific enhancer (ASE), whereas the Lefty1 enhancer is located upstream of the more proximal neural plate enhancer (NPE) and lateral plate mesoderm-specific enhancer (LPE). The ASE, NPE, and LPE were previously identified by injecting mouse pronuclei with enhancer constructs fused to LacZ to identify expression patterns in the developing embryo [37]. Through allele-specific CRISPR-Cas9 mediated deletion, we dissected the elements between the Lefty genes that functionally insulate their respective enhancers. By removing 44 kb of intervening chromatin, we show that the Lefty2 enhancer can regulate Lefty1 when both are placed in closer proximity. Incremental deletions, which modulate the linear chromatin distance between these two genes, reveal that the Lefty2 enhancer exerts a distance-dependent dosage effect on Lefty1 expression, despite the relatively small separation distance of these two genes. These data reveal that although the Lefty genes are co-expressed in ESCs and located within the same TAD, they are controlled by separate cis-regulatory elements.

Results

To identify which regions of the Lefty locus are responsible for the expression of Lefty1 and Lefty2, we performed CRISPR-Cas9 mediated deletions, using an F1 mouse ESC line (Mus musculus¹²⁹ × Mus castaneus) to allow for allele-specific deletion, and RNA quantification as described in [40]. This approach can identify cis-regulatory elements by creating heterozygous deletions, avoiding the potentially confounding trans-effects of a homozygous deletion that abolishes target gene expression [25,41]. Upstream of either Lefty gene are regions we identified as potential enhancers of Lefty gene transcription (“Lefty1enh”, “Lefty2enh”, Fig 1A), as these regions contained multiple binding events for core ESC transcription factors, as well as occupancy of the Med1 subunit of the mediator complex (Fig 1A). Furthermore, GRO-seq signal was observed on both the positive and negative strand, corresponding to bidirectional transcription (Fig 1B), a feature often found at enhancer regions [42]. The identified transcription factor bound region upstream of Lefty2 overlaps part of the previously identified left-side specific ASE enhancer [37]. Intriguingly, both mouse enhancers are also conserved in the human genome, at a sequence level and display enhancer features in human ESCs [43].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Lefty1 and Lefty2 are regulated by separate proximal enhancer regions in mouse embryonic stem cells.

A) The Lefty locus is displayed on the UCSC genome browser (mm10). The targeted enhancer elements upstream of either Lefty gene are highlighted. Transcription factor bound regions derived from mouse ESC ChIP-seq data sets compiled in the CODEX database are shown above ChIP-seq data for H3K27ac, the mediator subunit MED1, and CTCF. Arrows denote CTCF motif orientation. DNAse1-seq data shows accessible chromatin. RNA-seq above the genes shows gene expression in wild-type mouse ESCs. B) Zoomed in view of each Lefty enhancer located upstream and proximal to the Lefty genes. Strand specific GRO-seq data is shown, bidirectional transcription is a marker of active enhancers. The locations of previously identified enhancers active at embryonic day 8.5 are shown (ASE, LPE, NPE). C) Allele-specific primers detect 129 or Castaneus transcripts from Lefty1, Lefty2, or Pycr2. Expression from deleted alleles are expressed as a percentage of the total transcript amount normalized to the percent expression from the matched WT allele with individual clonal isolates shown as biological replicates. Deletion of the Lefty1 enhancer leads to a significant reduction of Lefty1 transcript abundance with a mild effect on Pycr2. Deletion of the Lefty2 enhancer abolishes Lefty2 expression while significantly decreasing both Lefty1 and Pycr2 transcripts. Error bars represent SD, one way ANOVA significant differences are indicated; (*) P < 0.05, (**) P < 0.01, (***) P < 0.001, (****) P < 0.0001, (ns) not significant.

https://doi.org/10.1371/journal.pgen.1011513.g001

Removal of the transcription factor bound region upstream of Lefty2 (ΔLefty2enh) led to a complete abolishment of Lefty2 transcription while also leading to a significant but minor decrease in transcript abundance for Lefty1 (~30%) and Pycr2 (~30%) (Fig 1C), highlighting that this enhancer predominantly regulates Lefty2 while also contributing in a more minor way to transcription of the other genes in this locus. As Lefty1 expression was still largely maintained after the loss of the Lefty2 enhancer, in a separate experiment, a ~1kb transcription factor bound region ~9.5kb upstream of Lefty1 was removed (ΔLefty1enh). Deletion of this region led to a decrease in Lefty1 expression (~70%) while having no effect on Lefty2 and only a minor, yet significant, effect on Pycr2 (~10% reduction) (Fig 1C). We note that the remaining Lefty1 expression in the Lefty1 enhancer deleted cell line is similar to the decrease seen when the Lefty2 enhancer was deleted. Although this may be coincidental, it suggests Lefty1 is regulated by both the Lefty2 and Lefty1 enhancers in a roughly additive manner. In addition, the complete loss of Lefty2 expression after deletion of the Lefty2 enhancer revealed that the intact Lefty1 enhancer is unable to compensate for the loss of the Lefty2 enhancer (Fig 1C).

We next asked whether the specificity in regulation by the Lefty enhancers was due to a strong architectural insulation of the two genes, segregating them into separate TADs. Between the Lefty1 enhancer and Pycr2 are CTCF bound regions which overlap CTCF motifs both convergent and divergent with Lefty1, suggesting this region may act as an architectural boundary (Fig 2). We evaluated chromatin capture data for this locus to identify if there is evidence of chromatin interactions between the Lefty genes and enhancers. Hi-TrAC chromatin capture utilizes two linked Tn5 transposases to capture nearby open chromatin and identify long-range chromatin interactions from nearby accessible regions with increased resolution compared to Hi-C [39]. Both Hi-TrAC and Hi-C data reveals the Lefty genes reside within a shared TAD (Figs 2 and S1). To evaluate the association between the Lefty enhancers and the genes within this TAD, we generated virtual 4C from the Hi-TrAC data with viewpoints centered on either identified enhancer element. The Lefty2 enhancer shows an association with both the Lefty1 enhancer and gene, whereas the Lefty1 enhancer shows an interaction with the Lefty2 enhancer but minimal signal at the Lefty2 gene. There is no apparent interaction of the Lefty2 enhancer with the prominent binding site for CTCF present just upstream of the Lefty1 enhancer. Noting that the Lefty enhancers both physically interact in the nucleus, the question remains as to why these enhancers function separately in the regulation of their cognate Lefty genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Both Lefty enhancers physically interact as identified by Hi-TrAC chromatin capture.

CLoops2 was used to evaluate Hi-TrAC chromatin capture data. ChIP-seq, and DNase1-seq is presented above the collapsed 1-dimensional chromatin capture signal generated from cLoops2, which highlights regions of frequent interactions. The virtual 4C plot depicts the interaction profile originating from a specific viewpoint, in this case, either the Lefty1 or Lefty2 enhancers. The y-axis shows the log2-transformed interaction frequencies, representing the number of paired-end tags (PETs) mapped between the selected viewpoint and other interacting regions. Dotted lines denote the promoter of both Lefty genes while the enhancers have coloured overlays. Below, a heatmap represents the interaction frequencies between all pairs within the locus.

https://doi.org/10.1371/journal.pgen.1011513.g002

Previous studies have suggested enhancer-promoter specificity may act as a layer of cis-regulation in itself, with a ‘lock and key’ type mechanism, preventing off-target enhancer activation, independent of chromatin interactions [44,45]. Next, we investigated to what extent the observed specificity of the Lefty2 enhancer for the Lefty2 gene promoter may be due to an enhancer-promoter specificity mechanism. To evaluate this possibility, we created a deletion of the intervening chromatin from just upstream of the Lefty1 promoter region to just downstream of the Lefty2 enhancer which places the Lefty1 promoter immediately downstream of the Lefty2 enhancer (ΔLefty1 to Lefty2enh) (Fig 3A). This 44kb deletion resulted in an increase in the expression of Lefty1 (Fig 3B), significantly above the level of expression observed in WT cells, indicating that there is no apparent enhancer-promoter incompatibility between the Lefty2 enhancer and the Lefty1 promoter. In fact, the observed increase in Lefty1 expression, when under control of the Lefty2 enhancer, mirrors the observed higher expression levels of Lefty2 compared to Lefty1 in ESCs (Fig 1A). To confirm that there were no other effects on Lefty1 expression from including a deletion of the sequences between the Lefty1 enhancer and the more upstream promoter region of Lefty1, we created a deletion (ΔLefty1 upstream) which removed this region and confirmed there was no significant difference in expression between this deletion and the previous enhancer deletion (S2 Fig). These findings indicate that the 44kb intervening chromatin region does insulate the Lefty1 gene from the stronger Lefty2 enhancer.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The Lefty2 enhancer can enhance the expression of Lefty1 when placed into close linear proximity.

A) The Lefty locus is displayed on the UCSC genome browser (mm10). The targeted deletions are highlighted along with the previously identified Lefty enhancers. Transcription factor bound regions derived from mouse ESC ChIP-seq data sets compiled in the CODEX database are shown above ChIP-seq data for the mediator subunit MED1 and CTCF. Arrows denote CTCF motif orientation. DNAse1-seq data shows accessible chromatin. B) Allele-specific primers detect 129 or Castaneus transcripts from Lefty1 or Lefty2. Expression from deleted alleles are expressed as a percentage of the total transcript amount normalized to the percent expression from the matched WT allele with individual clonal isolates shown as biological replicates. Deletion of the Lefty1 enhancer leads to a significant reduction of Lefty1 transcript. A deletion which places Lefty1 directly adjacent to the Lefty2 enhancer causes an increase in Lefty1 transcript abundance. Error bars represent the SD, one way ANOVA significant differences are indicated. (*) P < 0.05, (**) P < 0.01, (***) P < 0.001, (****) P < 0.0001, (ns) non-significant differences are not marked.

https://doi.org/10.1371/journal.pgen.1011513.g003

Next we asked, in the absence of boundary-like architectural insulation or promoter incompatibility, what could cause the transcriptional insulation between the two Lefty genes. Pycr2 is ubiquitously transcribed and needed for proline biosynthesis. We considered the possibility that the Pycr2 transcription unit is able to act as a sufficient barrier to crosstalk between the two Lefty regulatory units. We created deletions of both Pycr2 alone (ΔPycr2), and in conjunction with deletion of the Lefty1 upstream region (ΔLefty1 upstream + Pycr2), to see if the absence of the Pycr2 transcription unit would cause an increase in transcription of Lefty1 (Fig 4A). However, we observed no effect on the expression of Lefty1 or Lefty2 when Pycr2 is deleted alone, and when deleted in compound with the Lefty1 enhancer, there was no rescue of Lefty1 expression compared to deletion of the Lefty1 enhancer alone (Fig 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The universally transcribed Pycr2 gene is not responsible for the transcriptional insulation of the Lefty genes.

A) The Lefty locus is displayed on the UCSC genome browser (mm10). The targeted deletions are highlighted along with the previously identified Lefty enhancers. Transcription factor bound regions derived from mouse ESC ChIP-seq data sets compiled in the CODEX database are shown above ChIP-seq data for the mediator subunit MED1 and CTCF. Arrows denote CTCF motif orientation. DNAse1-seq data shows accessible chromatin. B) Allele-specific primers detect 129 or Castaneus transcripts from Lefty1, Lefty2. Expression from deleted alleles are expressed as a percentage of the total transcript amount normalized to the percent expression from the matched WT allele with individual clonal isolates shown as biological replicates. Deletion of Pycr2 alone or in conjunction with the region upstream of Lefty1 overlapping the enhancer do not significantly affect Lefty1 expression as compared to WT or parent enhancer deleted clone. Error bars represent the SD, one way ANOVA significant differences are indicated. (*) P < 0.05, (**) P < 0.01, (***) P < 0.001, (****) P < 0.0001, non-significant differences are not marked.

https://doi.org/10.1371/journal.pgen.1011513.g004

To identify what element or elements are responsible for the insulation of Lefty1 from the Lefty2 enhancer, we performed further deletion experiments to remove portions of the intervening chromatin between these two genes. We performed a stepwise deletion approach, removing larger and larger portions of the chromatin that separates the genes. We created three deletions, 1) removing ~31kb, placing Lefty1 downstream of Lefty2 (ΔL1-31kb), 2) removing ~24.5kb placing Lefty1 downstream of Pycr2 (ΔL1-24.5kb), and finally 3) removal of ~14.5kb and the most prominent CTCF bound region upstream of the Lefty1 enhancer (ΔL1-14.5kb); these are each compared with the previous Lefty1 upstream deletion which removed ~9kb and the Lefty1 enhancer (ΔLefty1 upstream) (Fig 5A). We observed an increase in the expression of Lefty1 as larger portions of the intervening chromatin were removed. In the ΔL1-14.5kb deletion, we observe a great deal of variation between the different independent clonal isolates such that, although the average decrease in expression was 55% for these clones, these levels were not significantly different from either the enhancer deleted ΔLefty1 upstream clones or the wild-type expression levels (Fig 5B). The clonal variation did not appear to correspond to any differences in the deletion boundaries inherent to CRISPR-Cas9 based deletions (S3 Fig and S1 Table). Furthermore, independent deletions removing only the most prominent CTCF peak with or without a deletion of the Lefty1 enhancer revealed that removing the CTCF bound region had no effect on Lefty1 expression (S4 Fig). As larger deletions were made, shortening the separation distance between Lefty1 and the Lefty2 enhancer, we observed further increased Lefty1 expression which reached wild-type levels after the 31kb deletion (ΔL1-31kb deletion). In all cases, there was no significant change in the expression of Lefty2 across the clones, highlighting that the Lefty2 enhancer crosstalk regulating Lefty1 does not affect its efficacy in enhancing the expression of Lefty2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Regulatory capacity of the Lefty2 enhancer on Lefty1 expression is modified by the intervening chromatin.

A) The Lefty locus is displayed on the UCSC genome browser (mm10). The targeted deletions are highlighted along with the previously identified Lefty enhancers. Transcription factor bound regions derived from mouse ESC ChIP-seq data sets compiled in the CODEX database are shown above ChIP-seq data for the mediator subunit MED1 and CTCF. Arrows denote CTCF motif orientation. DNAse1-seq data shows accessible chromatin. B) Allele-specific primers detect 129 or Castaneus transcripts from Lefty1, Lefty2. Expression from deleted alleles are expressed as a percentage of the total transcript amount normalized to the percent expression from the matched WT allele with individual clonal isolates shown as biological replicates. A dosage dependent effect of Lefty2 enhancer crosstalk on Lefty1 transcription is observed as more intervening chromatin is removed. Error bars represent the SD, one way ANOVA significant groups with P < 0.05 are indicated with letters.

https://doi.org/10.1371/journal.pgen.1011513.g005

Discussion

Using CRISPR-Cas9 mediated genome editing to remove sections of the Lefty locus, we identified the cis-regulatory elements responsible for Lefty1 and Lefty2 transcriptional control in ESCs, as well as identifying a curious case of transcriptional insulation in the absence of obvious boundary-like architectural insulation. We determined that the Lefty1 and Lefty2 genes are both primarily regulated by separate enhancer elements located just upstream of either gene. The entirety of Lefty2 transcriptional regulation in ESCs can be ascribed to a ~1.9kb region located ~3.7kb upstream, that overlaps the ASE. Conversely, although the majority of Lefty1 transcriptional regulation is due to a ~1kb transcription factor bound region located ~9.5kb upstream, there is an apparent additive effect on Lefty1 transcription from the Lefty2 enhancer which accounts for ~30% of Lefty1 expression. The region we identify as the required cis-regulatory element for Lefty1 differs from the previously identified more proximal ESC enhancer element (located 1.3kb upstream of Lefty1); identified by luciferase reporter assays rather than genetic perturbation [46]. The ESC Lefty2 enhancer overlaps the previously established ASE responsible for driving Lefty2 expression in the lateral plate mesoderm during embryonic development [37], indicating this element is active in more than one context. In addition, we show that these two genes and their cognate ESC enhancers interact in conformation capture assays that show no evidence of a strong architectural boundary between the two genes. Despite this apparent lack of architectural insulation, we find that the 44kb of chromatin separating these two genes does transcriptionally insulate the Lefty2 enhancer from the Lefty1 gene, however, we cannot rule out that intervening elements may be acting to attenuate stronger chromatin interaction between these genes. Furthermore, partial deletions of this region tuned Lefty1 transcript levels in relation to the separation distance from the Lefty2 enhancer, suggesting insulation at this locus is highly distributed, similar to what has been observed for some TAD boundaries [47]. Other studies have determined that linear chromatin distance does impact enhancer activity when moving the β-globin enhancer (LCR) further away from its promoter when inserted at ectopic loci [48]. Distance also plays a more pronounced role in regulating the activity of weaker enhancers, as a reporter assay testing the effect of distance on enhancer activity found that the activity of the Nanog enhancer rapidly dropped off as it was shifted from 25kb away from the target gene to 75kb [49]. In Drosophila, shifting the GMR enhancer from a few hundred base pairs away to 3kb away completely attenuated its ability to activate a target hsp70 promoter [50]. Our study reveals the distance dependent insulation of the Lefty1 gene from the stronger Lefty2 enhancer which may have been involved in the functional divergence of these gene paralogs by allowing for downregulated Lefty1 dosage after the duplication event.

Insulation at a number of other loci has been linked to CTCF binding events [51–53] and convergent CTCF motifs are enriched at adjacent TAD boundaries, and required for TAD boundary maintenance [13,54–56]. Given this role for CTCF at other regions of the genome, we were surprised to find that deletion of the most prominent CTCF bound region that separates the two Lefty genes did not abolish insulation in any clones, but instead when removed in conjunction with the 14.5kb region upstream of Lefty1, caused clonal variation in Lefty1 expression levels, ranging from completely insulated to a restoration of wild-type expression levels. This finding indicates that the intervening chromatin between Lefty1 and Pycr2, confers robustness to the insulation between the Lefty paralogs, but this insulation can be maintained in the absence of the CTCF bound region. Two additional regions between the Lefty genes display CTCF binding and removal of one of these regions with the 24.5kb deletion did relieve additional insulation of the Lefty1 gene, while still displaying clonal variation. Interestingly, the largest deletion of 31kb that leaves both genes intact did not remove any additional CTCF bound regions, compared to the 24.5kb deletion, but did abolish insulation, bringing the expression of Lefty1 to wild-type levels. Together these findings suggest a primary role for chromatin distance in insulating gene expression domains in the Lefty locus. Genome-wide insulator boundaries do not seem to exist as discrete elements; a striking ~20% of TAD boundaries remain stable upon the degradation of CTCF [54] and investigations into the effects of removing TAD boundaries have mapped insulator regions of up to 80kb in size which are responsible for boundary maintenance [57]. Other studies have shown that modulating the distance between enhancer-promoter pairs in endogenous and artificial loci does modulate enhancer activity, indicating chromatin distance can have an insulator effect [48,50].

Highlighting the complex interplay of chromatin architecture and gene regulation, studies which have temporally degraded architectural factors, such as CTCF or the RAD21 subunit of the Cohesin complex, identified surprisingly few changes in global transcription even upon a near complete loss of TAD structures [54,58,59]. However, it is difficult to draw concrete conclusions from these data about how CTCF degradation impacts Lefty expression, as we see conflicting results from different studies. For example, after 24–48 hours of CTCF ablation using the Auxin-inducible degron system, RNA-seq data shows a significant increase in expression of both Lefty1 and Lefty2 [54], which could be due to a loss of insulation. Another study degrading CTCF over 3, 12, or 24hrs found a modest but significant increase in Lefty1 at 24hrs by RNA-seq, however, the same study measured a steady and significant decrease in Lefty1 nascent transcript by mNET-seq at 3hr, which grew more severe after 24hr of CTCF degradation [58]. These confounding results highlight issues with global protein degradation as the trans-regulatory environment is profoundly affected, causing both direct and indirect gene expression changes.

Additional complex mechanisms can tune gene expression levels, for example promoters can compete for enhancer activity which may act to dilute the activity of an enhancer away from other nearby genes [50,60], although in the α-globin locus one enhancer acts to regulate multiple genes by forming a single complex hub, arguing competition may not occur in all cases [61]. We cannot rule out the possibility of simultaneous multiway interactions occurring between the Lefty genes and enhancers as the Hi-C and Hi-TrAC data we presented identifies only pairwise interactions. Furthermore, we saw evidence for promoter competition at the Lefty locus as the Lefty2 enhancer exerts the greatest effect on Lefty1 expression when the Lefty2 gene was also deleted in the 44kb deletion. In this deletion, Lefty1 expression levels were observed to increase above the WT levels for this gene due to control by the stronger Lefty2 enhancer. It is also possible, however, that this apparent promoter competition is simply the result of positioning the Lefty1 gene even closer to the Lefty2 enhancer as we are comparing a 31kb and 44kb deletion in this case.

We conclude that although the Lefty enhancers contact each other as well as both Lefty genes, as detected by chromatin conformation capture methods, there is a distance dependent insulation between the two genes that allows for reduced Lefty1 expression. Other studies have shown that tethering an enhancer to an off-target gene can lead to aberrant transcription, opposing our findings [62,63], however, there may be differences in the magnitudes of the chromatin interaction in these two cases. It has also been observed that by shifting an enhancer and it’s cognate gene closer together, one can increase and maintain gene expression over time in an artificial locus [48], similar to the increased Lefty1 expression driven by the Lefty2 enhancer when these elements are closer together. As there are several examples of separation distance tuning enhancer strength at engineered loci [48–50,64,65], our observation that this also occurs at the Lefty locus suggest it may be a relatively common phenomenon.

Materials and methods

Supporting information

Acknowledgments

We thank Barbara Panning for sharing the F1 ESC line and members of the Mitchell lab for their guidance and feedback on this project. This research was enabled in part by computing support provided by Compute Ontario (https://www.computeontario.ca/) and the Digital Research Alliance of Canada (https://www.alliancecan.ca).