Targeted nanopore sequencing for the identification of ABCB1 promoter translocations in cancer

Background Resistance to chemotherapy is the most common cause of treatment failure in acute myeloid leukemia (AML) and the drug efflux pump ABCB1 is a critical mediator. Recent studies have identified promoter translocations as common drivers of high ABCB1 expression in recurrent, chemotherapy-treated high-grade serous ovarian cancer (HGSC) and breast cancer. These fusions place ABCB1 under the control of a strong promoter while leaving its open reading frame intact. The mechanisms controlling high ABCB1 expression in AML are largely unknown. We therefore established an experimental system and analysis pipeline to determine whether promoter translocations account for high ABCB1 expression in cases of relapsed human AML. Methods The human AML cell line THP-1 was used to create a model of chemotherapy resistance in which ABCB1 expression was driven by a promoter fusion. The THP-1 model was used to establish a targeted nanopore long-read sequencing approach that was then applied to cases of ABCB1high HGSC and AML. H3K27Ac ChIP sequencing was used to assess the activity of native promoters in cases of ABCB1high AML. Results Prolonged in vitro daunorubicin exposure induced activating ABCB1 promoter translocations in human THP-1 AML cells, similar to those recently described in recurrent high-grade serous ovarian and breast cancers. Targeted nanopore sequencing proved an efficient method for identifying ABCB1 structural variants in THP-1 AML cells and HGSC; the promoter translocations identified in HGSC were both previously described and novel. In contrast, activating ABCB1 promoter translocations were not identified in ABCB1high AML; instead H3K27Ac ChIP sequencing demonstrated active native promoters in all cases studied. Conclusions Despite frequent high level expression of ABCB1 in relapsed primary AML we found no evidence of ABCB1 translocations and instead confirmed high-level activity of native ABCB1 promoters, consistent with endogenous regulation. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-020-07571-0.


Background
High level expression of the drug efflux pump ABCB1 leading to active efflux from blast cells of standard-of-care chemotherapy drugs such as daunorubicin and cytarabine is one of the strongest predictors for induction failure in acute myeloid leukemia (AML) [1]. Interestingly, recent studies have shown promoter translocations to be common drivers of ABCB1 expression in recurrent, chemotherapy-treated high-grade serous ovarian (HGSC) and breast cancer [2,3]. As a consequence ABCB1 is placed under the control of a strong promoter while leaving its open reading frame intact. The treatment-emergent selection of promoter-translocated ABCB1 variants provides compelling evidence for a central role of drug efflux in chemotherapy resistance, at least in cancers of ovary and breast. By contrast, the cellular mechanisms controlling ABCB1 expression in AML are largely unknown. Most publicly available genome sequencing data in AML were generated using whole-exome sequencing which does not detect structural variants (SVs) in non-coding sequences [4]. Furthermore, few studies have included samples from relapsed patients previously treated with intensive chemotherapy, raising a question as to whether ABCB1 promoter SVs might be present in ABCB1 high AML relapse cases. We therefore established an experimental system and analysis pipeline to determine whether promoter translocations account for high ABCB1 expression in cases of relapsed human AML.

Methods
Cell culture THP-1 cells were from DSMZ (Braunschweig, Germany) and cultured in RPMI 1640 medium (Sigma Aldrich) supplemented with 2 mM L-Glutamine (Life Technologies, Carlsbad, CA) and 10% fetal bovine serum (Sigma Aldrich). Whilst under drug selection cells were counted and replated every third day. Cell lines were confirmed mycoplasma-free and authenticated by short tandem repeat DNA profiling.

Primary human samples
Primary human AML and HGSC samples were from the Manchester Cancer Research Centre Tissue Biobank (approved by the South Manchester Research Ethics Committee). Their use was authorized by the Tissue Biobank's scientific sub-committee, with the informed consent of donors. For relapsed AML, the Biobank archive was searched for all samples. Where there were too few cryopreserved cells available for potential downstream analyses, or where the percentage of blasts in the cryopreserved sample was less than 80%, cases were excluded. This left 21 separate samples for analysis (Table  S1). Three presentation cases, selected at random, were also included for comparison. For ChIP sequencing, selected samples were thawed and immediately crosslinked.

Processing of primary HGSC samples
Samples for ABCB1 quantitative PCR and nanopore sequencing were taken from ex vivo models established using primary ascitic fluid samples, as previously described [5]. Briefly, ascites was centrifuged (500×g for 10 min at 4°C) and cell pellets pooled in HBSS (Life Technologies). Red blood cells were removed using a red blood cell lysis buffer (Miltenyi Biotec) as per the manufacturer's instructions. Tumour cells were plated into Primaria flasks containing OCMI. All cultures were incubated for 2-4 days at 37°C in a humidified 5% CO 2 and 5% O 2 atmosphere. Media was replaced every 3-4 days. Upon cell attachment, stromal cells were separated from the mixed sample using 0.05% trypsin-EDTA. Once tumour cells reached 95% confluency, cells were passaged using 0.25% Trypsin-EDTA, centrifuged in DMEM containing 20% FBS and re-plated at a 1:2 ratio. For long-term storage, cells were frozen in Bambanker (Wako pure chemical).
Cell viability assays 5 × 10 3 cells were plated in each well of a 96-well plate with media containing a serial dilution of daunorubicin. Plates were incubated for 72 h at 37°C. 20 μl of 140 μg/ mL resazurin (Sigma Aldrich) was added to each well. Plates were then incubated for a further 4 h and read using a POLARstar Omega plate reader (BMG Labtech, Aylesbury, UK).

RNA sequencing and data analysis
Total RNA was extracted from 5 × 10 5 cells using QIAshredder spin columns and an RNeasy® Plus Micro kit (Qiagen, Manchester, UK). Prior to sequencing RNA integrity was checked using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). PolyA libraries were prepared using a Sure-Select Strand Specific RNA Library Prep kit (Agilent Technologies) and samples were then barcoded and pooled. Sequencing was performed using a NextSeq system (Illumina, San Diego, CA). Two replicates were sequenced for each cell line (THP-1_S and THP-1_R). A single run of 150 bp paired-end sequencing produced a mean of 55.1 M reads per sample. Reads were aligned to the human genome (hg38) using STAR v2.4.2a [6]. DEseq2 was used to calculate FPKM (fragments per kilobase of transcript per million mapped reads) values for each transcript [7].

Quantitative PCR
cDNA was generated using a High Capacity Reverse Transcription kit (Applied Biosystems, Foster City, CA). qPCR reactions were performed in MicroAmp® optical 384-well reaction plates and analysed using a QuantStudio® 5 PCR system (Applied Biosystems). Reactions were performed in quadruplicate and included primers for β-Actin (ACTB) as a housekeeping gene. Primers were designed using the Universal Probe Library (UPL) Assay Design Center (Roche, Basel, Switzerland) and purchased from Integrated DNA Technologies (Coralville, IA). Raw fluorescence data was converted to CT values using the Thermo Fisher Cloud facility (Waltham, MA) and normalised to ACTB. Primers were (i) ACTB (F) ATTG GCAATGAGCGGTTC, (R) GGATGCCACAGGACTC CAT, UPL probe #11 and (ii) ABCB1 (F) GGAAATTTAG AAGATCTGATGTCAAAC, (R) ACTGTAATAATAGG CATACCTGGTCA, UPL probe #65.

Chromatin immunoprecipitation (ChIP) and next generation sequencing
ChIP was performed using anti-H3K27Ac (ab4729, Abcam, Cambridge, UK). 10 8 cells (THP-1) or 10 7 cells (primary AML samples) were used for each precipitation using the method described by Lee et al. [8]. Briefly, cells were cross-linked with 1% formaldehyde for 10 min at room temperature before the reaction was quenched with 0.125 M glycine. Cell pellets were washed twice with PBS and nuclear lysates sonicated for 6 cycles using a Bioruptor® Pico (Diagenode, Liege, Belgium). 10 μg of antibody bound to 100 μl of magnetic beads (Dynabeads Protein G, Invitrogen, Carlsbad, CA) was added to each sample and immunoprecipitation performed overnight on a rotator at 4°C and 20 rpm. After five washes with RIPA buffer (50 mM HEPES pH 7.6, 1 mM EDTA, 0.7% Na deoxycholate, 1% NP-40, 0.5 M LiCl), chromatin IPbound fractions were extracted by incubating for 15 min at 65°C with elution buffer (50 mM TrisHCl pH 8, 10 mM EDTA, 1% SDS). Crosslinking was then reversed by incubation at 65°C for 6 h. RNaseA (1 mg/ml) and proteinase K (20 mg/ml) were added to eliminate RNA and protein from the samples. DNA was extracted using phenol:chloroform:isoamyl alcohol and precipitated by adding 2 volumes of ice-cold 100% ethanol, glycogen (20 μg/μl), 200 mM NaCl and freezing at − 80°C for at least 1 h. Pellets were washed with 70% ethanol and eluted in 50 μl 10 mM TrisHCl pH 8.0.
Libraries were prepared for sequencing using a Microplex Library Preparation Kit (Diagenode). 200-800 bp fragments were selected using AMPure XP beads (Beckman Coulter, Brea, CA). Sequencing was performed using a NextSeq desktop sequencing system (Illumina) with 75 bp, paired-end high output generating 65-80 M (THP-1_S and THP-1_R) and 34-45 M reads per sample (primary AML samples). Reads were aligned to the human genome (hg38) using BWA-MEM v0.7.15 [9]. Read duplicates were removed using Picard v2.1.0. Reads were further filtered using Bedtools v2.25.0 to keep only paired reads that mapped to standard chromosomes and to remove reads with a mapping quality of less than 10. Reads mapped to blacklisted regions defined by EN-CODE were then removed using Bedtools (http://mitra. stanford.edu/kundaje).
PCR was performed with Expand Long Template Polymerase (Roche, #11759060001) using 3.2 μg of 4C template product and then purified using a High Pure PCR Product Purification Kit (Roche. #11732676001). Library quality was assessed using a 2100 Bioanalyzer (Agilent Technologies). Samples were sequenced with 10% phiX using a MiSeq desktop sequencing system (Illumina) with 75 bp single-end settings generating a mean of 1.4 M reads per sample. Sequencing data was deconvoluted using cutadapt v1.18. Reads were mapped and analysis performed using 4Cseqpipe [11].
Target enrichment and nanopore long-read sequencing Genomic DNA was extracted from 5 × 10 6 cells using a QIAamp DNA mini kit (Qiagen, Manchester, UK). 1 μg of genomic DNA was sheared with a g-TUBE™ (Covaris, Inc., Woburn, MA) using 50 μl sample volume and a 30s spin at 11,000 rpm followed by inversion and a further 30s at 11,000 rpm. Barcoded libraries were prepared from 45 μl of fragmented DNA using a Ligation Sequencing Kit SQK-LSK108 / SQK-LSK109 (Oxford Nanopore Technologies, Oxford, UK) according to the 1D Sequence Capture protocol with the following modifications: the pre hybridisation PCR was replaced by a custom PCR with 6 cycles using Herculase II fusion DNA polymerase (Agilent Technologies). The subsequent AMPure XP purification was performed using a 0.6x bead to sample volume ratio. All purified product was used for the hybridization using an oligonucleotide designed using the online Agilent Sure-Design service (https://earray.chem.agilent.com/suredesign/home.htm). The library targeted a 500 kb region that included ABCB1 and upstream neighbouring genes (RUNDC3B, SLC25A40, DBF4 and ADAM22; chr7:87,132, 949-87,632,948 hg19). Following pull-down, the postcapture amplification was replaced by a custom PCR with 14 cycles as described above. Amplified libraries were quantified by Qubit (Invitrogen) and TapeStation 2200 (Agilent Technologies). An additional amplification was performed using 10 cycles of PCR and the product was quantified. Amplified libraries were again quantified by Qubit and 1 μg of each library used for the End-prep. AMPure XP purification was performed using a 0.47x bead to sample ratio. Four end-prepped libraries were pooled into one adapter ligation reaction using 125 ng of each, as measured by Tapestation. A 0.47x bead to sample ratio was used for the AMPure XP bead binding. Sequencing was performed on the MinION using the R9.4 / 9.4.1 Flow Cell FLO-MIN106 / FLO-MIN106D (Oxford Nanopore Technologies) according to the manufacturer's instructions. Target enriched DNA from THP-1_R was also subjected to conventional sequencing using a MiSeq desktop sequencing system (Illumina) with 1% phiX and 251 bp, paired-end settings generating 1.92 M reads. Nanopore sequencing of THP-1_R generated 0.92 M reads with a mean read length of 1901 bp and a maximum read length of 224.8 kb. The coverage of the targeted region was 58.1x. Nanopore sequencing of AML/HGSC generated an average of 1.6 million reads per sample (range 0.8-3.0 million) with a mean read length of 1581 bp. FASTQ files were aligned against the GRCh38/hg38 reference human genome using LAST after training the aligner with a subsample of 10,000 reads [12]. Structural variant calling was performed using NanoSV [13]. Structural variants were confirmed through manual inspection and identification of chimeric reads using IGV (https://software.broadinstitute.org/software/igv/).

Flow cytometry & assessment of daunorubicin retention
Flow cytometry was performed using an LSR II flow cytometer (BD Biosciences, Franklin Lakes, NJ). FlowJo v10.1 (BD Biosciences) was used to analyze data. To assess daunorubicin retention 5 × 10 5 cells were resuspended in PBS containing 1 μM daunorubicin with 40 μM verapamil or 5 nM tariquidar or vehicle. Samples were incubated for 2 h at 37°C then placed on ice to prevent further efflux. Daunorubicin accumulation was assessed by flow cytometry.

Results
In the first instance we sought to establish whether activating promoter translocations could be experimentally induced in AML. ABCB1 promoter translocations have been described in several solid malignancy cancer cell lines but not in human myeloid leukemia cell lines [14]. We exposed the human AML cell line THP-1 to escalating doses of daunorubicin over 142 days and generated a resistant line (THP-1_R) which exhibited a 28.3-fold increase in daunorubicin IC 50 (Figs. S1A and S1B). RNA sequencing and confirmatory quantitative PCR indicated that ABCB1, which was not expressed in THP-1 cells, was one of the most highly expressed genes in THP-1_R cells (among the top 2.5% of genes by read count; Table  S2, Fig. S1C). Parental drug-sensitive THP-1 (THP-1_S) cells did not efflux daunorubicin whereas THP-1_R cells exhibited robust drug efflux that was completely reversed by either verapamil (a non-specific ABC transporter substrate) or tariquidar (a highly specific inhibitor of ABCB1) [15], confirming that efflux was due to ABCB1 (Fig. S1D). Importantly, the resistance of THP-1_R cells to daunorubicin was a stable cellular phenotype and persisted over 3 months of continuous culture without daunorubicin selection, suggesting constitutive expression (Fig. S1E).
ABCB1 codes for four protein coding transcript variants and contains two promoters separated by a~110 kb intron that is transcribed in variants 1-3. This large first intron, which overlaps part of the coding sequence for RUNDC3B (Fig. 1a), appears to be a common site for promoter translocations. In fact, all of the high-  [3] had at least one SV involving intron 1. We observed that THP-1_R cells expressed ABCB1 transcript variant 4 but inspection of RNA sequencing tracks revealed additional transcription of intron 1, starting~48 kb upstream of the variant 4 promoter (Fig. 1a). ChIP sequencing for H3K27 acetylation, a histone modification that marks active promoters and enhancers, found no acetylation of either ABCB1 promoter confirming that neither was active despite high levels of transcription (Fig. 1a). To identify the co-opted promoter we performed 4Csequencing of THP-1_S and THP-1_R cells with a viewpoint centered on the ABCB1 variant 4 promoter and observed in resistant cells both an abrupt loss of contacts within ABCB1 intron 1 and gain of a distant interaction with GTF2I ( Fig. 1a and b). The GTF2I promoter is located~12.94 million base pairs centromeric to the ABCB1 variant 4 promoter on chromosome 7q; it was strongly acetylated and GTF2I expression was among the top 25% of genes in both lines ( Fig. 1a and Table S2). These data indicate that a daunorubicininduced translocation fusing the GTF2I promoter with the coding sequence for ABCB1 was responsible for the acquisition of constitutive high level ABCB1 expression in THP-1_R cells.
Having demonstrated the presence of an experimentally initiated ABCB1-activating promoter translocation in THP-1_R cells, we developed a long-read sequencing technique to confirm this finding and to facilitate screening of primary patient AML samples for similar SVs. We performed targeted sequencing of a 500 kb region of chromosome 7 that contained both ABCB1 and SLC25A40, which is the most common translocation partner for ABCB1 in chemotherapy-treated breast and ovarian cancer patients [3]. The region was enriched using a library of biotinylated oligonucleotides that provided 95.5% coverage of the targeted sequence (Fig. 1c). Uniquely mapping oligonucleotides cannot be produced for regions containing repetitive sequences. Given that breakpoints frequently contain repeats we first compared conventional sequencing to long-read nanopore sequencing, which generates reads of up to 15 kb that can bridge translocations involving repetitive sequences (Fig. 1c) [16]. Both approaches identified the GTF2I promoter: ABCB1 translocation (Figs. 1a and 2a). However, despite lower coverage of the targeted region (58.1x versus 368.6x for MiSeq) nanopore long-read sequencing generated 52 chimeric reads that crossed the breakpoint and mapped unambiguously to GTF2I, compared with only six by conventional sequencing (Fig. 2b). These data demonstrate that long-read nanopore targeted sequencing efficiently identified the GTF2I promoter: ABCB1 translocation.
In chemotherapy-treated HGSC Christie et al. [3] identified ABCB1 fusions in 20/108 (18.5%) of cases, increasing to 31.3% if only ABCB1 high cases were considered. To assess the sensitivity of our nanopore long-read sequencing approach we quantified ABCB1 expression in 33 samples of metastatic HGSC recovered from ascitic fluid (Fig. 2c), and sequenced the 15 cases with the highest expression (Table S3). In 5/15 cases we identified translocations involving splicing of non-coding 5′ regions of partner genes to exon 2 of ABCB1 as a result of an intron 1 breakpoint (Table S4 and Fig. 2c-d); in one sample three separate fusion genes were identified. In addition to expanding the repertoire of reported ABCB1 fusions in HGSC, our data also demonstrate that targeted nanopore sequencing is at least as sensitive as the multimodal approach taken by Christie et al. [3].
We next identified in our Biobank 21 cases of relapsed AML with sufficient cryopreserved cells for downstream analyses, and once more determined ABCB1 expression levels by quantitative PCR (Fig. 2e). All but two patients had previously been treated with anthracycline-containing chemotherapy and all had initially achieved a complete morphological remission. Patient characteristics and molecular genetic features are shown in Table S1. At original presentation 16/21 cases exhibited a normal karyotype, 2/ 21 an MLL gene rearrangement, 1/21 a t (8;21), 1/21 inv. [16] and 1/21 trisomy 11 (Table S1). Blood samples exhibited high blast cell percentages (median 93.5%, range 82-100%). Targeted nanopore sequencing of 12 cases with high level ABCB1 expression (Fig. 2e) achieved similar coverage to that achieved in THP-1_R (mean 45.3x, range 16.5-82.3), but failed to identify any functional translocations involving ABCB1 intron 1 (Figs. S2A and S2B). H3K27Ac ChIP sequencing of nine of 12 of the nanopore long read sequenced samples performed in parallel revealed acetylation of one or both ABCB1 promoters in every sample, demonstrating that high level ABCB1 expression in relapsed AML is driven by native promoters (Fig. 2f), at least in the samples evaluated. These results are consistent with our recent finding that ABCB1 expression in primary AML is highly dynamic and regulated by several active enhancers [17].

Discussion
While we were able to experimentally induce in a human myeloid leukemia cell line the kind of treatmentemergent promoter-translocated ABCB1 variant found in chemotherapy-treated breast and ovarian cancers, we were unable to identify similar translocations in primary human relapsed AML samples from patients previously exposed to anthracycline-based chemotherapy. This is despite the strong association of ABCB1 with treatment failure and the high frequency of balanced translocations targeting a range of genes in AML [1,18]. Instead, as we have recently demonstrated, ABCB1 expression appears to be regulated by its native promoters and a range of cooperating enhancer elements [17]. One possible explanation for this difference is the relative genetic stability of AML which has one of the lowest mutation frequencies of any cancer [19]. By contrast ovarian and breast cancers have much higher rates of mutation and more frequent microsatellite and chromosomal instability that predicts for poor outcome [20][21][22]. The mechanism of activation may also depend on the state of the ABCB1 promoter in different cancers and their cells of origin. Hematopoietic stem cells normally express ABCB1 and its regulatory mechanisms appear to be preserved in AML, allowing for dynamic expression and adaptation following treatment [17]. In contrast, in tumours and tissues where the ABCB1 promoter is silenced, translocation may be the only mechanism by which expression can be induced, favouring the selection of structural variants that replace the silenced promoter. Despite the poor performance in clinical trials of inhibitors of ABCB1, recent findings continue to highlight the central clinical relevance of this drug efflux pump.

Conclusions
Our findings demonstrate the utility of long-read sequencing as an efficient technique for identifying SVs in human cancers and a promising approach for cost effective medical genetics [23]. Whilst our approach expanded the repertoire of reported ABCB1 fusions in HGSC, we did not identify similar structural variants in ABCB1 high AML, where expression appears to driven by native promoters. Further characterisation of diseasespecific mechanisms of ABCB1 regulation are needed to inform novel approaches for overcoming chemotherapy resistance. performed the long-read nanopore protocol. LN  (See figure on previous page.) Fig. 2 Nanopore sequencing identifies ABCB1 fusions in ABCB1 high relapsed primary human HGSC but not primary AML. a Image illustrates the GTF2I-ABCB1 translocation. Red Xs indicate absent transcription from normal ABCB1 promoters in THP-1_R AML cells. b Tracks show the GTF2I portion of chimeric reads spanning the GTF2I-ABCB1 translocation generated by nanopore long-read sequencing (MinION; upper panel) and conventional sequencing (MiSeq; lower panel) of THP-1_R. Pink or blue coloring of reads indicates orientation. c ABCB1 expression levels in primary HGSC samples relative to normal human CD34 + haematopoietic stem and progenitor cells (HSPCs), as determined by quantitative PCR. Targeted nanopore sequencing was performed on the samples highlighted in pink; samples with a functional ABCB1 intron 1 fusion are marked with asterisks. d ABCB1 intron 1 breakpoints identified in HGSC samples & THP-1_R cells are indicated, together with the partner gene. e ABCB1 expression levels relative to normal human CD34 + HSPCs as determined by quantitative PCR in primary relapsed AML samples. Targeted nanopore sequencing was performed on the samples highlighted in pink. Three presentation AML samples are also shown in green. f H3K27Ac ChIPseq tracks showing the acetylation of ABCB1 promoters for nine of the relapsed AML samples shown in (E). For (C), (E) and (F) number indicates Biobank identifier; T = timepoint