Enhancer hijacking activates oncogenic transcription factor NR4A3 in acinic cell carcinomas of the salivary glands

The molecular pathogenesis of salivary gland acinic cell carcinoma (AciCC) is poorly understood. The secretory Ca-binding phosphoprotein (SCPP) gene cluster at 4q13 encodes structurally related phosphoproteins of which some are specifically expressed at high levels in the salivary glands and constitute major components of saliva. Here we report on recurrent rearrangements [t(4;9)(q13;q31)] in AciCC that translocate active enhancer regions from the SCPP gene cluster to the region upstream of Nuclear Receptor Subfamily 4 Group A Member 3 (NR4A3) at 9q31. We show that NR4A3 is specifically upregulated in AciCCs, and that active chromatin regions and gene expression signatures in AciCCs are highly correlated with the NR4A3 transcription factor binding motif. Overexpression of NR4A3 in mouse salivary gland cells increases expression of known NR4A3 target genes and has a stimulatory functional effect on cell proliferation. We conclude that NR4A3 is upregulated through enhancer hijacking and has important oncogenic functions in AciCC.

index compared to control cells, suggesting that NR4A3 has transforming potential in mouse cells. The authors conclude that the t(4;9) link active enhancer elements from highly expressed salivary genes in the SCPP cluster to the NR4A3 upstream region, resulting in activation of NR4A3 expression. It is also suggest that NR4A3 is an oncogenic driver in AciCC. This is an ambitious study employing state-of-the-art methodology. The epigenetic analyses are comprehensive and solid. The study is, however, mainly descriptive but contains some new information. NR4A3 is a known oncogene previously shown to be activated by chromosome translocations in sarcomas. A major concern is the lack of convincing functional evidence proving that NR4A3 is an oncogenic driver in AciCC. The authors only present physical evidence of a t(4;9) by WGS in 6 cases and show that the translocation leads to overexpression of NR4A3. No breakpoints were identified in any genes in 4q13 or 9q31 and no functional fusion transcript were identified.
To substantiate the conclusion that NR4A3 is an oncogenic driver in AciCC the authors should knock-down the expression of the gene in human AciCC cells and study the effects of knock-down on proliferation, apoptosis etc. They should also study the transforming activities of NR4A3 in human salivary or mammary cells by analyzing the effects on proliferation, viability, spherogenesis, growth in soft agar etc.
Another weakness of the paper is that only one target gene of NR4A3 is studied (CCND1). Additional target genes should be analyzed in order to further characterize the molecular consequences of activation of NR4A3 in AciCC.
The SCPP gene locus (se Introduction lines 50-51 and elsewhere) is not a specific locus (gene) but a cluster of several different genes. It should therefore not be designated the " SCPP gene locus". In addition, information about the genes included in this cluster, their functions and role in salivary glands should be added.
The FISH image shown in Fig. 1D is difficult to interprete since the boundries of individual nuclei are impossible to discern and therefore it is not possible to count the signals in these nuclei. The image should also include several nuclei showing the same FISH pattern.
The quality of the H&E sections of normal parotid gland and AciCC (Fig. 4A) could be substantially improved.
The authors should include more detailed information about the histology of their AciCC cases, that is whether they have a conventional histology or are cases with high-grade transformation. AciCC5 in Fig 1B looks heavily rearranged and could represent a case with high-grade transformation? Fig. 2, showing the translocation breakpoints (TXs), is somewhat difficult to understand and needs to be explained or better illustrated. Do AciCC3-6 have inversions at the chromosome 4 bp (cf. Fig  3D)? Please provide a more detailed explanation of secondary events (e.g. inversion of the SCPP gene cluster) occurring at the breakpoint sites on 4q and 9q. It would be interesting to see the expression level of NR4A3 for each AciCC and to compare it to the type of translocation (where the breakpoints are located) and to the amount of activating chromatin marks in the vicinity of NR4A3.
I suggest that the authors use at least two software programs for detection of gene fusions. I also suggest that the authors show cell counts over time instead of cell index and viability (Fig 4 I and G). There is a big difference between the effects presented in Fig 4G and 4I. The authors should also include a western blot of the transfected mouse cells and not only show the mRNA expression (Fig. 4H).
Reviewer #3 (Remarks to the Author): The authors describe recurring rearrangements in acinic cell carcinoma (AciCC), and perform genome, transcriptome, and epigenomic profiling, implicating highly recurrent inter-chromosomal translocations in enhancer hijacking activating the oncogenic transcription factor NR4A3. Further gene expression and in-vitro analyses are presented corroborating a likely oncogenic role of NR4A3 in AciCC via Cyclin D1, including 2.5-fold upregulation of mouse Ccnd1 mRNA, increased Cyclin D1 protein levels and increased cell index. This is an interesting contribution describing interesting and novel finding in a rare tumour entity. It is indeed likely that the active enhancers juxtaposed to NR4A3 can drive NR4A3 gene expression. This is a timely result that should be published soon.
My only major criticism pertains to the wording used by the authors. The authors state that 'the rearrangements translocate active enhancer regions from highly expressed salivary gland genes to the NR4A3 upstream region, resulting in upregulated 2 expression and nuclear accumulation of NR4A3. ' The word "resulting" is an overstatement, and should be toned down, since the authors did not present direct experimental evidence for interaction between the enhancer elements and the NR4A3 promoter (as e.g. pursued in PMID:27869826 using 4-C sequencing). At least the remaining limitation that this test for direct interaction has not been pursued should be made clear in the Discussion.
Additional points: - Figure 1: I agree NR4A3 seems to be a likely target here, but this figure needs to clarify a few aspects. Please indicate the p-value that '***' refers to. Please compute p-values for other genes, such as SEC61B, which might be significant too? What is the difference in fold change between NR4A3 and other genes. Please clarify whether there is sufficient evidence to rule out that the expression of other genes may have an additional role in AciCC.
-An analysis of common (cell type invariant) TADs (topologically associating domain structures) at the affected loci would be potentially useful for the readers. I assume the breakpoints will likely affect TAD boundaries with potentially insulating effect. It should be reassured though that there are no TAD boundaries between the active enhancer elements shown in Figure 3 and the target gene NR4A3. Datasets calling TAD boundaries present across distinct cell types could be used for such analysis. The manuscript by Haller et al. describes the novel finding that recurrent translocation of SCPP locus and NR4A3 might lead to upregulation of NR4A3, which might drive the oncogenic pathways in AciCC. The authors show differentially over expressed genes with active histone marks regions are enriched for NR4A target motif sites, adding more support to the hypothesis that oncogenic expression of many genes is driven by NR4A3 activity.
Overall the genetic data is very convincing, the findings are novel, the manuscript is nicely written, which warrant publication.
> We thank the reviewer for his/her very positive comment, and appreciate his/her careful evaluation of our work.
The authors show that the translocation breakpoints in the SCPP locus are in active (H3K27Ac) regions, which juxtaposes this with NR4A3, and lead to upregulation of NR4A3. It is suggested that this is because of "super-enhancer" (SE) juxtaposition. However, in not every case there seems to be detectable SE activity ( Fig. 2 and 3  > We provide hockey-stick plots for all three AciCC samples as a novel Supplemental Figure   S5, demonstrating that enhancer regions from the SCPP gene cluster in close proximity to the breakpoints rank among the strongest enhancer regions in two samples, and fulfill the criteria for super-enhancers in the third sample. This is now specifically described in the revised manuscript (page 8).
The genetic data and some of the ChIP-seq suggest an activation mechanism, even if there is no "SE" activity. There is a reduction in H3K27me3 (bivalency states). May be juxtaposed regular enhancer, not necessarily SE, is sufficient for activation of NR4A3 and drive the phenotypes.
> We agree with the reviewer that juxtaposition of a regular enhancer might be sufficient for activation of NR4A3, e.g. similar mechanisms were shown to drive TAL1, TLX1/TLX3, HOXA10 activation in T-ALL (reviewed in Belver & Ferrando, Nat Rev Cancer, 2016). We have now included the observation of reduced H3K27me3 marks in the revised manuscript (page 9).
Are the tissue samples amenable to conformational capture assay, to directly show that the enhancer activity from the SCPP locus juxtaposes to the NR4A3 promoter. Experimental evidence for such interaction will be a key strengthening point. In the absence of data to directly show the enhancer activity by interaction, these caveats should be discussed. > As also suggested by reviewer 3, we aimed at providing circularized chromatin conformation capture combined with next generation sequencing (4C-seq) from patient tissues AciCC 1-3 using the NR4A3 breakpoint region as view point to clarify the possible interactions. Unfortunately, and as the reviewer was already concerned about, the amount of frozen tumor material that was left after the extensive ChIP-seq analysis that had been performed in preparation of the original submission was too small to yield sufficient data in the new 4C-seq analysis. The 4C-protocol we used is included at the end of the cover letter, the amount of chromatin we had available for this experiment was in the range of <10% of the recommended minimum. To still further confirm the enhancer activity within the SCPP gene cluster, we now provide novel data from an experiment where we used a dual luciferase reporter assay that indeed demonstrated enhancer activity within the SCPP gene cluster (page 9, Figure 5, Supplemental Table S6), and we appropriately discuss this data in the revised discussion (page 15).
Is it possible to grow AciCC in nude mice to test if enhancer-inhibitor strategy (such as BETi) will reduce tumor growth. This will further support the enhancer/SE hijacking hypothesis. Or, will knocking out NR4A3 block tumor growth in the nude mouse models? > We agree that experiments in mouse xenotransplantation models would additionally strengthen the concept of enhancer hijacking and the oncogenic role of NR4A3. However, we have not identified any patient-derived AciCC cell line. The establishment of PDX mouse lines is in preparation but will not be ready for the resubmission. In addition, it was not possible to generate AciCC mouse models with stable NR4A3 knock-down/knockout as the Nr4a3 gene was not expressed endogenously in the cell lines that we isolated and immortalized from normal mice. We could not find any mouse models of AciCC which might have been used to establish primary AciCC cells. These would have been prerequisite to have to attempt stable knock-down/knockout of the Nr4a3 gene. Please see also our extensive reply to a similar suggestion made by reviewer #2.
Additional points: -In text relating to Fig. 4, expression of human NR4A3 is suggested to increase endogenous mouse Nr4a3. However the data pertaining to Hprt and Nr4a3 expression is not available.
Only human NR4A3 expression is shown, which is exogenously over-expressed.  Tables 13 and 14). Indeed, mouse Nr4a3 was not expressed in the immortalized mouse salivary gland cells, this has been corrected in the text.
-NR4A3 gene body methylation being significantly lower in two of the AciCC is correlated to sustained activated expression of NR4A3. This is not clear as most previous studies have suggested that gene-body methylation positively correlated with gene activity (increased expression associated with increased gene body methylation).
> DNA methylation at gene body has indeed been observed to exhibit positive correlation with the expression level of respective genes. The seeming paradox, given the silencing function of this mark at promoter regions, can be resolved if one considers the fact that most of the genes have alternative TSSs up-and down-stream of the canonical one. It is most likely that under normal circumstance the main purposes of gene-body methylation are to silence alternative TSS and prevent aberrant splicing. Since gene expression level of a gene measured in bulk tissue samples is impacted by the relative fraction of cells expressing this gene, its positive correlation with gene-body methylation can be readily explained through cellular heterogeneity. An alternative explanation for these observations is that aberrant transcription might be induced from several alternative gene-body TSSs. Verifying this hypothesis would require dedicated experiments, involving e.g. CAGE-Seq, which, however, we feel to be out of the scope of the present study.

Reviewer #2 (Remarks to the Author):
This manuscript describes a genomic, transcriptomic, and epigenomic profiling of salivary acinic cell carcinomas (AciCC). WGS of 6 AciCCs revealed t(4;9)(q13;q31) translocations in all cases. There were no recurrent copy number alterations or mutations. The 4q13 breakpoints clustered within the secretory Ca-binding phosphoprotein (SCPP) gene cluster and the 9q31 breakpoints were located upstream of the orphan nuclear receptor gene NR4A3. RNA seq did not reveal any functionally fusion transcripts but showed signficant upregulation of NR4A3 compared to normal parotid gland (NPG). No other genes up-or downstream of NR4A3 were overexpressed compared to NPG. The 4q13 breakpoints mapped to regions with enrichment of active chromatin marks associated with abundantly expressed salivary gland genes. Moreover, active chromatin marks from the SCPP cluster were located directly at the 9q31 breakpoint region upstream of NR4A3 in AciCCs. DNA methylation throughout the NR4A3 gene was significantly lower in two out of three AciCCs compared to normal salivary gland, consistent with activation of the gene. Mouse salivary gland cells stably transduced with a wildtype NR4A3 cDNA showed increased viability and cell index compared to control cells, suggesting that NR4A3 has transforming potential in mouse cells. The authors conclude that the t(4;9) link active enhancer elements from highly expressed salivary genes in the SCPP cluster to the NR4A3 upstream region, resulting in activation of NR4A3 expression. It is also suggested that NR4A3 is an oncogenic driver in AciCC. This is an ambitious study employing state-of-the-art methodology. The epigenetic analyses are comprehensive and solid.
> We thank the reviewer for his/her positive comment, and appreciate his/her favorable evaluation of our work.
The study is, however, mainly descriptive but contains some new information. NR4A3 is a known oncogene previously shown to be activated by chromosome translocations in sarcomas. A major concern is the lack of convincing functional evidence proving that NR4A3 is an oncogenic driver in AciCC. The authors only present physical evidence of a t(4;9) by WGS in 6 cases and show that the translocation leads to overexpression of NR4A3. No breakpoints were identified in any genes in 4q13 or 9q31 and no functional fusion transcript were identified. To substantiate the conclusion that NR4A3 is an oncogenic driver in AciCC the authors should knock-down the expression of the gene in human AciCC cells and study the effects of knock-down on proliferation, apoptosis etc. > We agree that the suggested NR4A3 knock-down experiments in AciCC cell line models would further establish the oncogenic role of NR4A3 in AciCC. However, no commercially available or published human patient-derived AciCC cell lines exist that could be employed in a NR4A3 knock-down experiment. A likely explanation for this is the fact that the majority of AciCCs are slowly proliferating low-grade tumors that also do not grow in nude mice or as spheroids. In contrast, high-grade transformed and fast proliferating AciCCs that could potentially be employed to generate AciCC mouse models for further stable NR4A3 knockdown are relatively rare, e.g. they represent only 5% among all AciCCs (Scherl et al. Another weakness of the paper is that only one target gene of NR4A3 is studied (CCND1).
Additional target genes should be analyzed in order to further characterize the molecular consequences of activation of NR4A3 in AciCC.
> We already showed a significant correlation between higher NR4A3 ChIP-seq peaks and upregulated genes in the AciCC tumors and a significant enrichment of the main NR4A3 binding motif NBRE among differential H3K27ac peaks in our manuscript, which is a strong indicator of the molecular consequence of NR4A3 activation in AciCC.
To identify further NR4A3 target genes that could be validated for the revision, we employed which is difficult to analyze in a cell monoculture. We thus chose CCND1 and ENO3 to be further validated as NR4A3 target genes.
Regarding our own data, ENO3 and VTN were within our RNA-seq list of significantly upregulated genes in the human AciCCs (Supplemental Table S7, Supplementary Figure S10) with CCND1 being higher expressed as well but not reaching significance level. We now also show a significant upregulation of CCND1 and ENO3 in human AciCC tissue samples by immunohistochemistry (Supplemental Figure S11, Supplemental Table S4). Employing our invitro systems, we generated RNA-seq and mass spec data from our NR4A3-transduced immortalized mouse salivary gland model and control cells having been stably transduced with red firefly luciferase. Ccnd1 and Eno3 were upregulated on mRNA and protein level, this is now included in the revised manuscript ( Figure 6, Supplementary Tables S13 and 14, Supplementary Figure S12).
In summary, combining intensive literature research, published gene expression data sets and newly data we have collected in the salivary gland cell context, we suggest that target gene specificity of NR4A3 is cell context dependent, with CCND1/Ccnd1 and ENO3/Eno3 being reproducibly identified as NR4A3 transcriptional target genes also by our own data.
The validation of ENO3 as an additional NR4A3 target gene is included in the revised manuscript while the impact of NR4A3 on the expression of CCND1 had been described already in our original submission.
The SCPP gene locus (see Introduction lines 50-51 and elsewhere) is not a specific locus (gene) but a cluster of several different genes. It should therefore not be designated the "SCPP gene locus". In addition, information about the genes included in this cluster, their functions and role in salivary glands should be added.
> We agree that the term SCPP gene cluster is a more appropriate designation for this genomic region, and this has been changed in the revised manuscript. The description of the function of these genes has been included in the introduction of the revised manuscript (first paragraph of the introduction, page 4).
The FISH image shown in Fig. 1D Figure S11).
The authors should include more detailed information about the histology of their AciCC cases, that is whether they have a conventional histology or are cases with high-grade transformation. AciCC5 in Fig 1B looks heavily rearranged and could represent a case with high-grade transformation? > We thank the reviewer for this very relevant suggestion. AciCC5 indeed represents a case with high-grade transformation. Additional data including histology and clinical follow-up from an extended cohort has been included in the revised manuscript (Supplemental Table   S4).  Fig 3D)? Please provide a more detailed explanation of secondary events (e.g. inversion of the SCPP gene cluster) occurring at the breakpoint sites on 4q and 9q. > In the revised figure legend for Figure 2, the graphical visualization of the translocation breakpoints has been explained more in detail, and structural rearrangements at the breakpoint sites are included in the revised Figure 2. Presenting 9 additional AciCCs with [t(4;9)(q13;q31)] aberration identified by hybrid capture NGS in the revised manuscript, we now provide evidence for three subgroups differing by the 4q13 breakpoint location and also by the orientation (e.g. inversion) of the 4q13 genomic region. This is explained in detail in the revised manuscript ( Figure 2, Figure 3, page 7-8). > For all samples, frozen tissue sections were evaluated for regions with tumor cell content >80% by an experienced head and neck pathologist, and only these regions were microdissected for DNA and RNA extraction. AciCC1 is a small low-grade conventional AciCC from an adolescent with no recurrence, and showed a highly stable genome with almost no gains or losses, and only 7 non-synonymous SNVs. Since we cannot provide another explanation, we suggest that this case represents a very early stage AciCC with no further secondary genetic events apart from the t(4;9)(q13;q31) translocation, thus closely resembling normal parotid gland tissue in the DNA methylation and RNA expression analysis.
It would be interesting to see the expression level of NR4A3 for each AciCC and to compare it to the type of translocation (where the breakpoints are located) and to the amount of activating chromatin marks in the vicinity of NR4A3.
We fully agree with the reviewer that the respective breakpoint region might affect expression levels of NR4A3, however, we think that the number of three samples is too low to draw further conclusions, and thus suggest not to include these values in the figure.
I suggest that the authors use at least two software programs for detection of gene fusions. I also suggest that the authors show cell counts over time instead of cell index and viability (Fig 4 I and G). There is a big difference between the effects presented in Fig 4G and 4I.
> Following the suggestion of the reviewer we have performed two different types of proliferation assay using the stably transfected mouse salivary gland as well as human mammary MCF10A cells, now included in the extensively revised former Figure 4 (now Figure 6). First, we did microscopic count of nuclei at time points 0 and 24 hours and then computed the level differences in the counts within control and NR4A3-transfected cells.
Secondly, we incubated control and NR4A3 transfected cells with BrdU and 7AAD and performed FACS analysis after cell fixation. While the former assay informed on differences in absolute numbers of cells, the latter provided information on potential alterations in ratios of cells that were 'caught' in the different cell cycle phases. Indeed, data from both, the nuclei count and cell cycle assays, strongly support our claim that NR4A3 contributes to the enhancement of oncogenic cell growth. The data is presented in the revised manuscript in the extended Figure 6 for both cell line models.
The authors should also include a western blot of the transfected mouse cells and not only show the mRNA expression (Fig. 4H).
> In response to the reviewer's comment on validation of protein expression we tested three commercial antibodies that are advertised to detect the NR4A3 protein (Santa Cruz order numbers SC393902 and SC393903, and Biorbyt orb256728). However, even though these antibodies are advertised to specifically detect a protein with an apparent molecular weight of about 68 kDa, they did not show specific bands in the Western blots we performed with the mouse and MCF-10A cell lines used in our study. Given that we could confirm expression of NR4A3 protein by mass spectrometry we conclude that none of these antibodies is of sufficient quality. Hence, we base our claims on ectopic expression of NR4A3 in the mouse salivary gland cell line model on the results of the mass spectrometry experiment and these are unequivocal: the cell line having been stably transfected with NR4A3 does express this at the RNA (RNA-seq and qRT-PCR data) and protein (mass spectrometry) levels, while the control cell line does not.
We thus suggest to present the novel mass spectrometry data which is now included in the revised Figure 6 as well as Supplemental Table S14 and Supplemental Figure S12. Making use of the same proteomic data sets, we could identify mouse Cyclin D1 and Eno3 proteins and their higher levels in NR4A3 vs. red firefly luciferase transduced mouse salivary gland cells, with higher protein level of Cyclin D1 also shown in a Western blot (Supplemental Figure   S12d).

Reviewer #3 (Remarks to the Author):
The authors describe recurring rearrangements in acinic cell carcinoma (AciCC), and perform genome, transcriptome, and epigenomic profiling, implicating highly recurrent interchromosomal translocations in enhancer hijacking activating the oncogenic transcription factor NR4A3. Further gene expression and in-vitro analyses are presented corroborating a likely oncogenic role of NR4A3 in AciCC via Cyclin D1, including 2.5-fold upregulation of mouse Ccnd1 mRNA, increased Cyclin D1 protein levels and increased cell index. This is an interesting contribution describing interesting and novel finding in a rare tumour entity. It is indeed likely that the active enhancers juxtaposed to NR4A3 can drive NR4A3 gene expression. This is a timely result that should be published soon.
> We thank the reviewer for his/her positive comment, and appreciate his/her favorable evaluation of our work.
My only major criticism pertains to the wording used by the authors. The authors state that 'the rearrangements translocate active enhancer regions from highly expressed salivary gland genes to the NR4A3 upstream region, resulting in upregulated expression and nuclear accumulation of NR4A3.' The word "resulting" is an overstatement, and should be toned down, since the authors did not present direct experimental evidence for interaction between the enhancer elements and the NR4A3 promoter (as e.g. pursued in PMID:27869826 using 4-C sequencing). At least the remaining limitation that this test for direct interaction has not been pursued should be made clear in the Discussion.
> We agree with this comment. This limitation has now been made clear in the revised discussion (page 14-15), and the cited statement from the abstract has been revised (abstract). To further address and clarify this issue we aimed at providing circularized chromatin conformation capture combined with next generation sequencing (4C-seq) from patient tissues AciCC 1-3 using the NR4A3 breakpoint region as view point to clarify the possible interactions. Unfortunately, the remaining little amount of frozen tumor material was consumed without yielding sufficient data. We provide novel data on enhancer activity within the SCPP gene cluster instead (page 9, Figure 5, Supplemental Table S6).
Additional points: - Figure 1: I agree NR4A3 seems to be a likely target here, but this figure needs to clarify a few aspects. Please indicate the p-value that '***' refers to. Please compute p-values for other genes, such as SEC61B, which might be significant too? What is the difference in fold change between NR4A3 and other genes. Please clarify whether there is sufficient evidence to rule out that the expression of other genes may have an additional role in AciCC.
> The explanation for the *** P-value has been included in the revised manuscript ( Figure   legend for Figure 1). An additional Supplemental Table S3 has been included in the revised version of the manuscript showing that only NR4A3 is significantly upregulated among these genes.
-An analysis of common (cell type invariant) TADs (topologically associating domain structures) at the affected loci would be potentially useful for the readers. I assume the breakpoints will likely affect TAD boundaries with potentially insulating effect. It should be reassured though that there are no TAD boundaries between the active enhancer elements shown in Figure 3 and the target gene NR4A3. Datasets calling TAD boundaries present across distinct cell types could be used for such analysis.
> We fully agree with this excellent comment. TAD boundaries and HiC contact maps from published data sets have been included in the revised figure 2, and show indeed a strong correlation with our novel ChIP-seq and RNA-Seq data from normal salivary gland tissue. This was also was very helpful for the further identification and characterization of three subgroups according to different 4q13 breakpoint regions. This important observation has been included in the revised manuscript (page 6 and 7, Figure 2, novel Figure 5).
Description of the protocol used for circularized chromatin conformation capture combined with next generation sequencing (4C-seq) (for review process only) All remaining frozen tumor biopsies from AciCC 1-3 (70 -120 mg) were treated with collagenase A (Roche; catalog no. 10103578001) to obtain tumor cell suspensions as described by Matelot et al. 2016. Circularized chromatin conformation capture combined with next generation sequencing (4C-seq) was adapted from the protocol described by van de Werken et al. 2012. In brief, chromatin of suspended tumor cells was crosslinked with formaldehyde and subjected to a first restriction digestion using restriction enzyme HindIII suited for cutting in crosslinked chromatin and known to cut within the NR4A3 promoter, the 4C viewpoint. After a first ligation which favorably joined the ends of the NR4A3 promoter-derived HindIII fragment with themselves but also with unknown HindIII fragments located in close vicinity due to chromatin crosslinking and, hence, likely derived from enhancer regions, the chromatin was decrosslinked. The resulting circularized DNA molecules were subjected to a second restriction digestion using the frequent cutter DpnII which also cuts in the viewpoint and with high likelihood in the unknown connected HindIII fragments. A second ligation with diluted DpnII-cut DNA favored self-ligation and, thus, circularization of DpnII fragments. Inverse PCR using primers pointing outwards of the known NR4A3 promoter-derived HindIII-DpnII fragment were employed to generate PCR products of the joined unknown DNA fragments, which were subsequently analyzed by highthroughput next generation sequencing on a HiSeq2000 sequencer (v4, single read 50 bp).

Results
According to the low amount of input material, DNA yield after the two restriction digestion/ligation rounds was low (35 -120 ng). For comparison, in the standard 4C protocol with 10 million cultivated cells, the DNA yield after the two restriction digestion/ligation rounds usually amounts to 20-30 µg. As a consequence of the low yield from the AciCC biopsies, the amount of library PCR products was also low with not more than 3-4 ng per PCR using each half of the available template as input. The PCR libraries were pooled, and the pool was sequenced in one lane. Bioinformatics analysis revealed that although 72,205,232 reads were generated, the vast majority mapped to PhiX (72,065,182; 99.7%) and after