Prediction and identification of transcriptional regulatory elements at the lung cancer-specific DKK1 locus.

The glycoprotein dickkopf 1 (DKK1) is highly expressed in lung cancer cell lines and tissues. Our previous study demonstrated that DKK1 promoter activity is low in lung cancer cell lines. This may be because it lacks the necessary transcriptional regulatory elements (TREs) required for higher activity levels. However, it is difficult to computationally predict functionally significant TREs, as TREs from different locations can affect large segments of distant DNA. The Encyclopedia of DNA Elements project features multiple integrated technologies and approaches for the discovery and definition of functional elements, including enhancer elements and enhancer-blocking insulators. In the present study, DNase I hypersensitive sites and histone modifications of DKK1 were investigated in the A549 lung cancer cell line using the UCSC Genome Browser. A set of cis-acting enhancer elements were identified by a dual-luciferase reporter gene assay system to increase activity of the DKK1 promoter with lung cancer specificity. To the best of our knowledge, these data provide the first insight into the role of the DKK1 locus in lung cancer, and confirm the contribution of intronic cis-acting elements to the regulation of DKK1 expression, providing a new insight into gene regulation in lung cancer, which could inform the development of targeted therapy.

Introduction

Non-small cell lung cancer (NSCLC) is the most common type of lung cancer, and a major public health burden (1). Therefore, the molecular analysis of NSCLC is necessary to identify novel molecular targets for determining the disease prognosis and informing the design of targeted therapies.

There are 4 dickkopf (DKK) proteins: DKK1, 2, 3 and 4. DKK1 antagonizes the canonical Wnt signaling associated with the pathogenesis of malignant tumors by forming a ternary complex with LDL receptor-related proteins (2,3). High levels of DKK1 expression have been observed in multiple lung cancer cell lines and lung cancer tissue samples (4). DKK1 is expressed at very low levels in normal adult human tissues, and higher levels in embryonic tissues (5). The use of DKK1 and other biomarkers may increase the sensitivity of lung cancer detection. Our previous study demonstrated that the activity of the DKK1 promoter was low in lung cancer cells (6). Thus, it is likely that enhancer elements outside the basal promoter region contribute to its wide range of expression patterns (7).

Promoters, located at the 5′ends of genes, surround the transcriptional start site and act to initiate transcription (8,9). However, it is not fully understood how one transcriptional regulatory element (TRE) can precisely define a human genetic locus. Furthermore, not all functional elements have yet been determined for each transcript. In general, both a promoter and its associated transcript are embedded in a collection of positively and negatively regulating elements, which can be positioned anywhere in relation to the transcriptional start site, including at 5′, 3′ and intronic locations, and may act across large segments of DNA up to megabases in length (10). Therefore, the computational prediction of functionally significant TREs is challenging. The Encyclopedia of DNA Elements (ENCODE) project provides systematically mapped regions of transcription, transcription factor association, chromatin structure and histone modification (11,12). It has integrated multiple technologies and approaches to identify and define functional elements, including enhancer elements and enhancer-blocking insulators. This collective effort may be used to screen for potential evolutionarily conserved TREs of the DKK1 promoter.

Regulatory elements have been characterized downstream of a range of human genes, so analyzing multiple intronic and extragenic DNase I hypersensitive sites (DHSs) at +27 kb around a gene in different epithelial cell types can identify regions that are often associated with gene regulatory elements (13,14). This strategy has previously been successfully used to identify enhancers of the c-Myc and cystic fibrosis transmembrane conductance regulator genes (15,16). Not all DHSs contain TREs; they may be associated with structural elements that function in chromatin organization instead, which may have a major effect on the combination of DNA with transcriptional regulation factors (17). Therefore, the extent to which gene-specific combinatorial patterns of histone modifications exist remains to be determined. Numerous covalent histone modifications, including methylation and acetylation, which occur mainly at N-terminal tails, affect gene transcription (18). However, the extent to which combinatorial patterns of histone modifications exist in the DHS remains unknown. Histone modification is one of the most important epigenetic factors in the regulation of gene expression (19,20). Thus, active promoters are often marked by the acetylation of various histone residues, including H3K4ac, H3K27ac and H3K9ac, or methylation, including H3K4me1/2, H3K27me3 and H3K9me3 (21). This modification is largely conserved across species.

With the development of methodology that can evaluate regulatory elements in the whole genome (22), key regulatory elements of the expression and function of one large proximal promoter fragment of the DKK1 locus were investigated. The present study aimed to comprehensively predict potential regulatory elements of the lung cancer-specific DKK1 locus to identify widespread enhancers.

Materials and methods

Cell lines and cell culture

A549, H446 and H460 human lung cancer cell lines, and HEK293(T) cells were obtained from the American Type Culture Collection (Manassas, VA, USA). The squamous cell carcinoma cell line, Eca-109, and human umbilical vein endothelial cells (HUVECs) were supplied by the Cell Bank of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The 293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Gibco; Thermo Fisher Scientific, Inc.). All other cells were cultured in RPMI 1640 medium (HyClone; GE Healthcare Life Sciences, Logan, UT, USA) containing 10% FBS, and all cells were maintained at 37°C in 5% CO2.

Computational prediction of TREs at the DKK1 locus

To screen for evolutionarily conserved TREs in the genetically identified DKK1 promoter, computational tools were used to download and process information from the UCSC Genome Browser (https://genome-cancer.ucsc.edu/), released by the ENCODE project (23). A human genome sequence of ~27 kb surrounding the lung cancer-specific DKK1 locus was downloaded. Enhancer predictions were performed by separately analyzing histone modification patterns and the DHS of DKK1 in A549 cells, HUVECs and normal human lung fibroblasts (NHLFs). CCCTC binding factor (CTCF) is a multi-zinc finger protein that binds to insulators (24). To investigate whether histone methylation-defined chromatin domains are demarcated by CTCF-bound insulators (25), the genome-wide distribution of CTCF was determined in these cells. The enhancer modules were extended to include the most adjacent blocks with high conservation scores by using PhastCons conservation track and regulatory potential track at the UCSC genome website. The top 12 enhancers located within 3,500 bp upstream and 4,700 bp downstream of the DKK1 transcription start site in each cell were selected. Sequences and detailed alignments are presented in Table I.

Table I.

DKK1 distal promoter fragment primers.

Promoter fragment	Length (bp)	Primer sequences	Restriction enzymes
DKK1-9 (−938-40)	978	F: TCTCCACATTAGCCCACCAC	HindIII/XhoI
		R: CTGCGGTCCCAGAGTCCT
DKK1-5 (533–40)	573	F: ACTGCGACTCTAAAGGGTTAATG	HindIII/XhoI
		R: CTGCGGTCCCAGAGTCCT
DKK1-3 (−350-40)	390	F: CCCCTCGGCTCTGTAAAGTAT
		R: TGCGGTCCCAGAGTCCT
DKK1-2 (−282-40)	320	F: CAAGTTCCCAGAGTTCCTGCT	HindIII/XhoI
		R: CTGCGGTCCCAGAGTCCT
SV40	420	F: CGCAGCACCATGGCCTGA	BgIII/Xho I
		R: TTGCAAAAGCCTAGGCCTCCA

DKK1, dickkopf 1; F, forward; R, reverse.

Strongest candidate TREs distally regulate the DKK1 promoter

To interpret these datasets, it was confirmed that the 12 candidate DNA sequences function as enhancers of DKK1. Broad classes of chromatin states were distinguished, referred to as promoters, enhancers, and insulators. To ensure that elements capable of overcoming the effects of strong enhancers were identified, the assay utilized an enhancer from the human β-globin locus control region. Known as DHS II (HS2), this enhancer functions in multiple cell lines, with multiple promoters (8,26). Furthermore, the presence of HS2 provides a large window of expression to reliably measure loss-of-function effects (27). The HS2 enhancer was incorporated into silencer and TRE plasmids to confirm the ability to overcome strong activating events, and as a measure of the interference between promoter and enhancer interactions (28). However, it is not known whether HS2 has an enhancing effect on the DKK1 promoter. The expression clones prepared for transfection are listed in Table II.

Table II.

Primer sequences of the 12 candidate TREs.

TRE	Locus	Length (bp)	Primer sequences
TRE1	chr10:53,992,381–53,992,930	550	F: TGGTTCATATTTTGTTTTTCTTGTG
			R: ACCTAAGTTATTAAGTTTTGTCTCA
TRE2	chr10:54,016,954–54,017,517	565	F: TTTATGCTAAGACAAGGAGGTGT
			R: TTTTGAAGAATAAACATAACATGAGAA
TRE3	chr10:54,073,726–54,074,219	498	F: GCAAGGGCACCCAAGTTC
			R: CCAGAGCCATCATCTCAGAA
TRE4	chr10:54,078,638–54,079,344	707	F: AATGTCTGTTGTTGTTGCTGTG
			R: CCAGGCTCATTCTTATCAGTAG
TRE5	chr10:54,086,158–54,086,541	384	F: TTTTCATCCCTTTCCCTCACT
			R: GGGCAAGGAGAATCAGCTC
TRE6	chr10:54,128,561–54,128,810	250	F: CATGCCAGGCTCTCAGTAAG
			R: CTGTTGAGTCAGGGGTTTGG
TRE7	chr10:54,203,100–54,204,074	975	F: TCGCCTAGTGTATCTTTTAG
			R: TTTGTTGATATTTCATAATCATTGGA
TRE8	chr10:54,211,579–54,212,928	1,350	F: TATTCATTTGCATAAAAGATAAAGCC
			R: ATTTTCTTATTCATTCATTTCCTACCG
TRE9	chr10:54,218,904–54,219,195	293	F: GATGCTAAACATAGGTACTTTTGAA
			R: AATTTGACTATGGGCTTTTAGG

TRE, transcriptional regulatory element; chr, chromosome.

Construction of reporter plasmids

A total of 4 different fragments of the DKK1 promoter were amplified from A549 cell line genomic DNA, then cloned into HindIII and XhoI sites upstream of the firefly luciferase reporter gene in the pGL3-basic vector (Promega Corporation, Madison, WI, USA) to produce pGL3-DKK1-2, pGL3-DKK1-3, pGL3-DKK1-5 and pGL3-DKK1-9 constructs. The TK promoter-driven Renilla luciferase reporter gene, derived from a commercial vector (pRL-TK; Promega Corporation), was used to normalize for firefly luciferase activity. As a control, the SV40 enhancer/promoter fragment was excised from pRL-TK and inserted in forward and reverse orientations into the pGL3-basic plasmid at the XhoI and BglII sites to produce pGL3-SV40. Genomic DNA extraction, polymerase chain reaction (PCR) amplification and concentration quantification were performed as previously described (6). All primer sequences and cloning sites are listed in Table I. Subcloned fragments were confirmed by sequencing. Predicted TREs were amplified by PCR according to the same protocol (6); Table II presents the primer sequences. The products were ligated into the multiple cloning site of the pGL3-DKK1-5 vector, upstream of a candidate DKK1 promoter-driven luciferase reporter gene.

Plasmid transfection

Plasmids were purified using a Plasmid Midi kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's protocol. Cells were plated into 96-well plates at 8×103 cells per well, 24 h prior to transfection, and were transfected at 80–90% confluence. The cells were co-transfected with the recombined vectors carrying the firefly (50 ng) and Renilla (1 ng) luciferase reporter genes. X-tremeGENE HP (Roche Diagnostics, Basel, Switzerland) was added at 0.25 µl/well, according to the manufacturer's protocol. The pGL3-basic vector (10 ng) and recombined pGL3-SV40 (10 ng) were used as negative controls.

Firefly/Renilla luciferase quantification

Luciferase activity was measured 48 h post-transfection using Dual-Luciferase® Reporter Assay System (Promega Corporation), according to the manufacturer's protocol. Firefly luciferase reporter gene activity was normalized to Renilla luciferase activity by calculating the Renilla/firefly luciferase activity ratio for each construct. The promoter control, DKK1-5, exhibited a baseline level of luciferase activity, which was normalized to 1. The activity of each test construct was calculated relative to the DKK1 promoter construct. Transfections were performed in triplicate, in 3 independent experiments.

Statistical analysis

The data are presented as the mean ± standard deviation and were analyzed using SPSS 17.0 software (SPSS, Inc., Chicago, IL, USA) by one-way ANOVA with a Student-Newman-Keuls post hoc test. P<0.05 was considered to indicate a statistically significant difference.

Results

Identification of DKK1 DHSs

To identify evolutionarily conserved regulatory elements, the genomic DNA sequence of A549 cells was downloaded from the UCSC Genome Browser. DHSs were identified within the 27-kb region of the human genome sequence surrounding DKK1 in the non-cancer human cell lines, HUVECs and NHLFs. A total of 12 potential TREs were identified, the majority of which were located within highly conserved sites of the DKK1 locus (Fig. 1A). The locations of these modules, termed TRE 1–12, are presented in Table II. The strongest signals, representing higher DNase I hypersensitive sites and larger areas of histone interaction, occurred around chr10: 54,073,726–54,074,219 (TRE3). Many of the cancer-associated TREs are located within, or in close proximity to, predicted DHSs, including TRE8, 10 and 12. TRE1, 2, 5 and 6 were associated with weak signaling compared with TRE3 and TRE4.

Figure 1.

Computationally predicted TREs in DKK1, a lung cancer-associated gene. (A) A major DHS was identified at the distal DKK1 promoter in 3 cell lines that express the DKK1 gene, including a 100% vert.cons region. (B) Histone modification patterns at the 2 DHSs of the DKK1 gene in A549; the highlighted sequence is the previously identified TRE3. TRE, transcriptional regulatory element; DKK1, dickkopf 1; DHS, DNase I hypersensitive sites; vert.cons, conserved by vertebrates; chr, chromosome; CTCF, CCCTC binding factor; HUVECs, human umbilical vein endothelial cells; NHLF, normal human lung fibroblasts.

Analysis of DKK1 histone modification patterns

Fig. 1B illustrates the histone modification patterns at the 2 hypersensitive sites of the DKK1 gene in A549 cells. TRE maps of 6 histone acetylation sites (H3K4me1, H3K4me2, H3K4me3, H3K4ac, H3K9ac and H3K27ac), histone variant H2A.Z and a tissue-specific CTCF recognition site in the data from 3 cell lines from ENCODE were also analyzed, with no explicit knowledge of any previous annotation (29). Data analysis revealed H2A.Z, H3K4me1, H3K4me2, H3K4me3, H3K4ac and H3K27ac signals in the intergenic region between TRE3 and 4, and TRE7, 8, 11 and 12 in both A549 and HUVEC cells, but not in NHLF cells. Enrichment patterns of these modifications were mapped, and the chromatin signatures for TRE1, 2, 5, 6, 9 and 10 were described. H3K4me1, H3K4me2 and H3K27ac signals were highly localized to intronic DHSs of TRE7 and 11 and 12 in HUVEC cells, while the other signals were distributed over a broader region. Furthermore, the significantly enriched CTCF signals at TRE6 and 10 suggest that enhancer-blocking insulators flank the DKK1 locus.

DKK1-5 has high functionality as a proximal promoter in different cells

Differences in relative luciferase activity were observed among candidate promoters (Fig. 2). As was hypothesized, compared with an SV40 promoter, the DKK1 promoter greatly increases luciferase activity in NSCLC cell lines (A549 and H460) but not in SCLC cell lines (H446). In contrast, DKK1 activity was lower than that of SV40 in 293(T) cells. There was no significant difference among the 4 segments (pGL3-DKK1-2, pGL3-DKK1-3, pGL3-DKK1-5 and pGL3-DKK1-9) among the different cell types. Once length was taken into consideration, the newly screened DKK1-5 fragment was determined to have the most potential for application as a proximal promoter in subsequent experiments.

Figure 2.

Recombination of pGL3-DKK1 fragments and pGL3-SV40, and measurement of their activity. (A) Verification of reconstructed plasmid by restriction enzyme digestion. (B) Luciferase activity of DKK1-luciferase constructs in A549, H460, H446 and 293 cell lines. *P<0.05 vs. pGL3-b DKK1, dickkopf 1; M, marker.

Construction of TRE-luciferase plasmids

To determine whether the computationally identified TREs possessed biological functionality, they were cloned upstream of either a DKK1-5 promoter-driven luciferase reporter construct or a native HS2 enhancer driven luciferase reporter construct (Fig. 3A). The HS2 enhancer is well characterized and has been used in similar reporter assays (30–32). The candidate enhancers were cloned into the enhancer site of the pGL3 DKK1-5 promoter vector. The observed high CTCF signal predicted 2 insulating fragments, TRE6 and 10. As a silencing test, the insulator was subcloned upstream of the DKK1-5 promoter and downstream of the HS2 enhancer to determine whether the elements had a cooperative effect on luciferase expression (Fig. 3A). The reconstructed TRE reporter plasmids were validated by restriction enzyme digestion (Fig. 3B).

Figure 3.

Recombination of pGL3-DKK1-5-TREs. (A) The location and order of TREs, HS2, DKK1-5 and luciferase in the constructs. The candidate TREs were cloned in front of a DKK1-5 promoter-driven luciferase reporter pGL3-basic vector. (B) Identification of reconstructed TRE reporter plasmids by restriction enzyme digestion. The position on the 1.0% agarose gel corresponds to the specific TRE, as indicated. DKK1, dickkopf 1; TRE, transcriptional regulatory element; HS2, DNase I hypersensitive site II; Poly-A, polyadenylated tail; LUC, luciferase.

Functional validation of transcriptional regulatory elements predicted at the DKK1 locus

The relative luciferase activity of TRE-DKK1-luciferase constructs was investigated in A549 and H460 NSCLC cells, the small cell lung cancer cell line H446, the esophageal cancer cell line Eca-109, and 293 cells (Fig. 4). The enhancer-control revealed the upper threshold of expression contributed by the HS2 enhancer, and was used as a normalization control. Luciferase reporter assays in A549 cells revealed that TRE7 and 12 act as strong transcriptional enhancers, leading to an 8-fold increase in DKK1 promoter activity. TRE3, 4, 6 and 8 also increased the activity of the DKK1 promoter in A549 cells, reflecting the DHS and histone modification results (Fig. 1). Importantly, HS2 together with TRE6 increased DKK1 promoter activity in A549 and H446 cells, whereas HS2 together with TRE10 and HS2 alone had a minor or no effect. This was also observed in H460 cells, albeit to a lesser extent. This suggests that TRE6 and TRE10 do not act as a silencer in A549, H446 and H460 cells. In H460 cells, it was demonstrated that TRE3, 4 and 12 significantly increased the activity of the DKK1 promoter, while TRE6, 7 and 8 had little effect. Furthermore, TRE3, 4, 7 and 12 had no effect on DKK1 activity in H446 cells. The majority of the putative enhancers did not significantly stimulate the DKK1 promoter in Eca-109 or 293, except for TRE6 which resulted in high DKK1 promoter activity in Eca-109 cell line.

Figure 4.

Enhancer dual luciferase reporter assay analysis of functionality. The relative luciferase activity of the candidate promoters in (A) lung cancer cells A549, H460, H446 and (B) the squamous cell carcinoma cell line Eca-109 and normal cell 293T. *P<0.05 vs DKK1. DKK1, dickkopf 1; TRE, transcriptional regulatory element; HS2, DNase I hypersensitive site II.

Collectively, these results demonstrate that the 6 identified DNA sequences act as efficient TREs. TRE3, 4, 6, 7, 8 and 12 significantly increased the activity of the DKK1 promoter in A549 cells, and TRE7 increased the activity to the highest degree. TRE3 and 4 increased promoter activity in H460 cells, while TRE12 showed a strong effect in A549, H460 and Eca-109 cells. Furthermore, TRE7 and 8 did not increase promoter activity in H460 cells, suggesting that TRE3 and 4 may have NSCLC-specificity.

Discussion

DKK1 has been reported to be a biomarker in lung cancer, with high levels of specificity and expression (4). However, our previous study demonstrated low DKK1 promoter activity in lung cancer cells (6). In the present study, 12 TREs were predicted using DHS and histone modification analysis of the lung cancer-specific DKK1 locus in A549 cells using the UCSC genome browser. Several of these protocols were applied to distinguish between active and poised enhancers. The functionality of the TREs was investigated using a dual-luciferase reporter assay in 5 different cell types. It was demonstrated that several of these DHS regions possess cooperative enhancer activity in vitro.

It has been identified that specific TREs may be associated with low promoter activity (7,33). However, to the best of our knowledge, no TREs were previously defined for the lung cancer-specific expression of DKK1. Methods to detect DNA sequence elements on a genome-wide scale and to examine the role of different one-dimensional regulatory signals remain a challenge to develop (34,35). However, this work uses a computational approach to predict TREs that contain cell line-specific long-range interactions between enhancers and promoters. The present study also presents methods to distinguish between different classes of functional TREs, which may contribute to the elucidation the mechanisms for cell type-specific gene expression.

The majority of predicted enhancers are distal to core promoters, and can be detected by locating tissue-specific DHSs within and around the gene (36). Previous studies have demonstrated that a number of histone modification patterns of promoters, including histone acetyltransferase p300, are enriched at the enhancer site in fetal mouse tissues (37,38). Thus, it is likely that TREs are highly evolutionary conserved and are marked by histone acetylation (39) or the binding of coactivator proteins such as p300 and MED1 (26). The effect of insulators at human enhancers may contribute to the understanding of how enhancers function in tissue-specific gene regulation (40). H3K4me1/2, H3K27ac, H3K9ac and H3K4ac modifications are associated with active enhancers (39,41). The histone variant H2A.Z associates with functional regulatory elements, and was previously demonstrated to be enriched in active promoters and strong enhancers, while the insulator binding protein, CTCF, marks the boundaries of histone methylation domains (42). However, in the present study, H3K9me1/3 and H3K27me3 signals were modestly elevated at silent promoters, and little change was observed in the intergenic regions of the cell lines. By integrating the predictions of DHSs and selected histone modifications in A549, HUVEC and NHLF cell ENCODE datasets, 12 candidate DNA sequences were proposed as potential TREs. A set of cis-acting elements were characterized within the DKK1 locus, and demonstrated that enhancer elements located within cancer-associated regions can regulate DKK1 promoter activity (Fig. 4). As the activity of the DKK1 promoter was only increased in NSCLC cells, it is likely that the most crucial sequences (TRE3, 4, 7 and 8) are NSCLC-specific.

An important aspect of this study was its initial predictive framework for the computational prediction of enhancer-promoter interactions from datasets. The observed changes in reporter activity indicate that these constructive TREs may have potential regulatory ability. However, it should be noted that only the specific TREs of DKK1 expression in cell lines were examined in the present study, and that additional in vivo studies are required to verify the results. Finally, the power of these techniques as a strategy to predict and identify TREs on the basis of their epigenetic characteristics, independent of motifs or other sequence features, was demonstrated in the present study, with particular respect to DHS mapping and histone acetylation patterns. A detailed enhancer map may be used to inform future therapeutic genome editing (43), as demonstrated by a recent report applying the CRISPR-Cas9 system to enhancers in situ, which confirmed that the BCL11A enhancer may be a novel target for intractable diseases (44).

In conclusion, the present study confirms the value of using computational tools to predict TREs, and provides the first insight into the intronic cis-acting elements that control the regulation of DKK1 expression. A systematic strategy to predict cell line-specific enhancer-promoter interactions using a minimal dataset is proposed. Furthermore, the present study suggests that the active global enhancer network offers important insights into cancer development and potentially, targeted therapy.