This is a particularly important need since many cancers produce multi-drug resistance by upregulating the expression of membrane efflux pumps [25], which are capable of speedily taking away cytotoxic prodrug metabolites from the cytoplasm. Ultimately, enzyme-prodrug methods that can target each dividing and non-diving cells could be essential to accomplish maximal tumor eradication in specified malignancies. JSI-124The inadequate catalytic performance of the HSV-tk/Figure 4. Bystander mobile killing mediated by TMPK-F105Y/AZT remedy drives a significant tumor mass reduction in a prostate cancer xenograft design. (A) Magnitude of bystander mobile killing was evaluated subsequent the shipping of the TMPK-F105Y suicide gene by immediate intratumoral injection of LV/TMPK into proven tumors in NOD/SCID mice (n=six). Reduction in tumor mass was assessed at the end of 6-working day AZT remedy program (at 50mg/kg/working day) by extraction of tumors and measurement of moist tumor bodyweight. Statistical significance is indicated by an asterisk ( p<0.05). (B) Weight of individual extracted tumors is shown for each animal in the AZT-treated and vehicle-untreated groups.ganciclovir-based suicide gene therapy system may have yielded mediocre bystander effects in previous work [26], examining bystander killing of the aggressive PC-3 prostate cancer cells. We hypothesize that the HSV-tk/ganciclovir system may not meet certain thresholds required for efficacious killing of bystander cells in PC-3 cells, and that our novel TMPK/AZT system [12] would result in better bystander effects in that same cancer model. We have therefore evaluated the magnitude of bystander effects mediated by our TMPK/AZTbased suicide gene therapy in PC-3 cells in vitro and in vivo. This unique system is based on an active-site engineered human TMPK, TMPK-F105Y, which is capable of selectively activating AZT. It is characterized by catalytic robustness, rapid antimetabolite accumulation as the enzyme overcomes the rate-limiting step in the AZT activation pathway, and cell killing driven by multiple mechanistic pathways, with toxicity against both DNA-replicating and non-dividing cells. Despite other reports that PC-3 cells did not exhibit strong bystander effects with the HSV-tk/GCV system [26,27], we were able to observe fairly significant bystander cell killing both in vitro and in vivo with our TMPK-F105Y/AZT suicide gene therapy. Our findings indeed support the lack of proper intracellular localization of connexin43 expression in PC-3 cells, but clearly indicate the existence of functional GJICs in these cells by dye-transfer experiments, which we believe are comprised at least in part of the pannexin family of proteins [28,29]. Bystander cell killing effect was completely abolished with the physical separation of the effector and bystander cells in transwell experiments as well as upon GJIC inhibitor treatment, highlighting the requirement for functional GJICs and indicating that the bystander effects observed in the TMPK/AZT system are not mediated by any soluble factor or free diffusion of antimetabolites to bystander cells. Finally, we underscore the therapeutic utility of the TMPK/AZT system for SGTC in a robust in vivo model of human prostate cancer xenografts. This study, in which efficient tumor regression was observed with significant contribution from bystander effects, provides the preclinical proof-of-principle for the application of the TMPK/AZT suicide gene system for lentiviral-based GDEPT of solid malignancies.Lung cancer kills more people than colorectal, prostate and breast cancer combined [1]. Squamous cell carcinoma (SCC) constitutes 26% of all lung cancer [2], making it one of the main histological subtypes besides small-cell and adenocarcinoma. Karyotypes of lung SCCs have revealed some commonality in the genomic landscape of these tumours, including distal amplification of 3q [3] and a more focal amplification at 8p12 [4], but as yet these findings have not translated into the clinic. SCC remains the most common lung cancer histotype for which no genomically targeted therapy currently exists [5]. The lack of such therapy prompted inclusion of the lung SCC subtype in The Cancer Genome Atlas (TCGA) project, an international collaboration aimed at cataloguing cancer-driving genetic variation within tumours using multiple high-throughput approaches. One such approach was Next-Generation Sequencing (NGS), which has been used to gain insights into disease development and progression in several types of cancer, including both lung adenocarcinoma and small-cell lung cancer (SCLC) [6,7]. The results of TCGA study of SCC revealed marked genomic complexity within lung SCC patient samples. However, pathPLOS ONE | www.plosone.org 1 way-specific alterations, hoped to yield therapeutic targets, did cluster by expression subtype, indicating the importance of integrating transcriptomic information in order to understand the phenotypic consequences of the plethora of genomic changes [8]. To understand how a more detailed, integrated analysis may aid inspection of lung SCC genomes, we deeply sequenced both the genome and transcriptome of LUDLU-1: a lung SCC cell line derived from a male patient whose smoking status is unknown. We also sequenced appropriate controls: the genome of an EBVtransformed lymphocyte cell from the same patient (cell line AGLCL) and the transcriptome of a normal bronchial epithelial cell line (LIMM-NBE1). To maximize our findings, we adopted an RNA sequencing method that captured both coding and noncoding RNA in a manner that retained information regarding the strand of origin. We have previously catalogued the transcriptional consequences of somatic structural variants in this cell line but here we focus on point mutations, aiming to see whether the mutational signature would give insight into disease etiology or carcinogenic mechanism, as it has for other cancer subtypes [6,9,10]. This type of in-depth characterisation of a given tumour LUDLU-1 and AGLCL were cultured as we have indicated previously [10]. LUDLU-1 was shown to be p63 positive and TTF-1 negative (data not shown) confirming a squamous carcinoma subtype. A549 was obtained from American Type Culture Collection (ATCC Manassas, USA) and cultured in Advanced DMEM-F12 medium (Life Technologies, 1263-4010) supplemented with 5% foetal bovine serum (Sigma, F7524), 2 mM GlutaMAXTM (Life Technologies, 3505-0087) and 50 U/ml penicillin and 50 mg/ml streptomycin (Life Technologies, 15070) at 37uC with 7.5% CO2.mapped reads, resulting in a per gene expression metric: Reads Per Kilobase per million Mapped (RPKM). Reads that had between two and five valid alignments were assigned a single location using SEQEM [14].A subset of 5% (108,362) of the germline polymorphisms in LUDLU-1 were selected, at random, for comparison with the somatic variants. All single nucleotide substitutions were annotated using a bespoke script that accessed Ensembl 60 via its Perl application programming interface. Additional information on the location of CpG islands was downloaded from the UCSC genome browser and incorporated into our in-house code. Pathogenicity scores were attained from MutPred [15]. Variable distributions or count data were tested between samples, or variant type, using an appropriate significance test (Wilcoxon and Chi-squared or Fisher's Exact respectively) at the 5% level but corrected for multiple testing using the Bonferonni correction. Allele counts for heterozygous variants were attained for both the genome and the transcriptome using SNVmix2 [16] (once Complete Genomics data had been converted using the cg2bam tools provided in the cgatools package). Allelic imbalance was tested only in those genes deemed as expressed in the tumour or normal and we required that a 10% change in absolute allele frequency been seen for either the DNA or RNA data, as per Tuch et al., 2010 [17]. Ratios were tested using Fisher's exact test at a false discovery threshold of 5%. To assess transcription-coupled repair, we used Cufflinks [18] to delineate transcripts within our LUDLU-1 RNAseq data, guided by Ensembl 60 gene annotations. The output included the strand each gene was located on. We recorded the number of mutations in each annotated, expressed gene, and whether each was on the transcribed or non-transcribed strand. We then calculated the rate of mutations per at risk base in the gene footprint. Significance (5% level with Bonferroni correction) and curve-fitting was performed using a zero-inflated, negative binomial regression model in R.This was performed as previously described in Stead et al. [10]. Briefly, Complete Genomics used their proprietary method to sequence DNA that we extracted from the LUDLU-1 and AGLCL cell lines. RNA, extracted from LUDLU-1 and LIMMNBE1, was sequenced by LGC Genomics on an Illumina HiSeq 2000 using 50 bp single end reads. Sequenced reads were aligned to the human reference genome, build 37, except in the case of miRNAs which were aligned to known miRNAs from build 36 using miRanalyzer [11]. All sequencing data have been submitted to the NCBI sequencing read archive (SRA: http://www.ncbi. nlm.nih.gov/sra) under accession numbers ERP001465 (LUDLU1 and LIMM-NBE1 RNA sequencing) and ERP001771 (LUDLU1 and AGLCL DNA sequencing).Complete Genomics call and score genomic variants during their local de novo assembly approach and output results into variant files, two of which (i.e. the tumour sample and matched normal) are compared, to identify somatic variants, using a tool called calldiff that is included in their proprietary cgatools 1.5.0, build 31, software suite. Somatic variants are each annotated with a somatic score [12]. Structural variants were extracted from the Junctions files, provided as part of the Complete Genomics sequencing report. Somatic mutations have been submitted to dbSNP (http://www.ncbi.nlm.nih.gov/SNP/) under the handle LIMM_YCR_PGG.Cisplatin was purchased from Sigma (P4394), dissolved in dimethyl sulfoxide (DMSO) at 3.3 mM and stored at 220uC.A flask of subconfluent cells was trypsinised recovered cells were washed once with PBS and then seeded in culture medium at a density of 1,000 to 4,000 cells per mL in 96 well plates. The plates were incubated at 37uC with 7.5% CO2 overnight to allow cells to adhere. Treated culture medium was prepared at a 2X concentration and 100 ml was added to each well. The cells were then treated with one of six concentrations of cisplatin (0. 3125, 0.625, 1.25, 2.5, 5, 10 mM) or the control (DMSO). The plates were incubated for a further 5 days. At this point the cells were stained with crystal violet, allowed to dry and 100 mL of 33% glacial acetic acid was added. The absorbance was read at 590 nm using the POLARstar OMEGA plate reader. IC50 values were calculated using Calcusyn V2 software (Biosoft, Cambridge, UK). Cell line sensitivity to cisplatin was compared by ANOVA using SPSS Statistics v19 (IBM). Cisplatin was purchased from Sigma (P4394), prepared at 3.3 mM in DMSO and stored at 220uC.50 coding and 50 non-coding single somatic variants were selected at random, and checked to ensure a normal distribution of somatic scores had been captured. Primer design failed for four but 96 underwent PCR and capillary sequencing, with seven giving ambiguous results. A receiver operating characteristic (ROC) curve was created using the validation results from the remaining 89 variants (data not shown), enabling us to set the somatic score threshold for single substitutions at 0.084, resulting in an estimated 100% specificity and 84% sensitivity. Indel validation was unsuccessful owing to ambiguous PCR/capilliary sequencing read-outs so all variants of this type were included if they had a somatic score .0. Somatic structural variants were validated as described in Stead et al. [10].Performed as described in Stead et al. [10]. Briefly, the number of reads aligning to exons, annotated as per Ensembl 60 [13], were counted and normalised by gene length and total number of The method was as described by Grenman et al [19]. In brief, a flask of subconfluent cells were trypsinised, washed once with PBS and seeded in culture media at a starting density of 1,200 cells per well in 96 well plates. A 1 in 10 serial dilution was made for A549 cells and for LUDLU-1, a 1 in 5 serial dilution was carried out. The plates were incubated at 37uC with 7.5% CO2 for 2 hours and then cells were exposed to 2, 4, 6 and 6 Grays of 137Cs (10 Gy, 1.958 Gy/min) in a GSR-D gamma irradiator. Plates were incubated for a further 7 to 14 days until colonies reached a size of 32 cells or more. Plating efficiency (PE) was calculated as described by Thilly et al [20] using Poisson statistics according to the formula PE = -In (neg wells/total wells)/number of cells plated per well. The fraction survival was expressed relative to the PE of the un-irradiated control. Radiation survival curves were compared by linear regression using SPSS Statistics v19 (IBM) as previously described [21].We sequenced the LUDLU-1 genome to an average coverage of 61x (i.e. each base was sequenced, an average, 61 times), and the matched lymphocyte to an average 55x. Copy number analysis shows that the SCC cell line is largely tetraploid and shares several features previously seen in lung SCC [10]. We identified 31,141 somatic single base substitutions (Supp. Table A in File S2), on average 10 per megabase (Mb). This exceeds the average mutation rate of 8.1 per Mb, ascertained from the sequencing of 178 lung SCC exomes as part of TCGA project in clinical samples [8]. 2832770A total of 181 somatic substitutions were located within coding regions of LUDLU-1 compared to an average of 360 in the series included in TCGA (Table 1). This included the only somatic mutation in the tumour cell line that is present in the Catalogue Of Somatic Mutations In Cancer (COSMIC) of clinical tumours [22]: a non-synonymous variant in TP53 (ID:10656). This somatic mutation causes Arg248Trp this is highly likely to inactivate the tumour-supressor function of p53, a protein involved in DNA repair, by removing the ability of Arg248 to directly contact the DNA response element via the minor groove [23]. No wild-type allele is present, as confirmed by the transcriptome sequencing. This specific p53 variant has been shown to be a gain-of-function mutation that promotes tumorigenesis [24]. We discovered 9431 somatic structural variants ranging from .1 bp substitutions to chromosomal translocations (Table 2 and Supp. Tables B and C in File S2). Included in the list of genes Table 1.
