NGS-based identification of druggable alterations and signaling pathways – hepatocellular carcinoma case report

E. A. Kotelnikova, M. D. Logacheva, Е. R. Nabieva © 2015 E. A. Kotelnikova et al.; Published by the Institute of Molecular Biology and Genetics, NAS of Ukraine on behalf of Biopolymers and Cell. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited UDC 57.032 NGS-based identification of druggable alterations and signaling pathways – hepatocellular carcinoma case report


Introduction
Carcinogenesis is considered to be caused by alterations in specific genes associated with dysfunction of regulatory networks [1].Therefore, reconstruction of regulatory interactions is necessary for understanding the processes of carcinogenesis in addition to the identification of molecular targets for the ISSN 1993-6842 (on-line); ISSN 0233-7657 (print) Biopolymers and Cell. 2015. Vol. 31.N 6. P 436-446 doi: http://dx.doi.org/10.7124/bc.000901antineoplastic drugs.The systems biology analysis of transcriptomic data makes it possible to identify and interpret the effects of mutations and gene expression deregulation.In cancer research, the goal of systems biology is to decipher the impact of genetic and epigenetic aberrations in cancer cells on their homeostasis, intercommunication and response to possible treatments [2].This approach is particularly important for precision oncology, since each tumor is unique in terms of genetics and pathological regulation of signaling pathways.The reconstruction of the patient-specific signaling pathways could help clinicians to identify the most effective treatment.One of the interdisciplinary tools of system biology is known as the next-generation sequencing (NGS) technology.NGS platforms perform massively parallel sequencing, so millions of DNA fragments are sequenced at a time.Such large-scale sequence analysis of the genome and transcriptome is vital for developing effective strategies in personalized cancer therapy.Specifically, this NGS-oriented approach is important for choosing between the treatment schemes, when selecting patients are likely to benefit from targeted therapies [3].The personalized NGS-based analysis promotes clinical decisions when standard therapy does not give the expected results or leads to tumor resistance.
Hepatocellular carcinoma (HCC) is one of the most often diagnosed types of liver cancer and occupies the 6th place in frequency of all cancer types [4].In this work we aimed to identify potential cancer driving or clinically relevant molecular events for a patient with HCC using NGS technology.

Samples collection and extraction of RNA/DNA
Genomic DNA and total RNA were isolated from fresh-frozen samples of hepatitis-negative HCC and adjacent non-cancerous tissue liver using Wizard SV Genomic DNA Purification System, Promega and PureLink RNA Mini Kit, Life Technologies with DNase treatment, respectively.Samples were collected from 66 years old male patient with histologically verified moderately differentiated HCC after tumor resection with informed consent, conforming to the ethical guidelines of the 1975 Declaration of Helsinki.
RNA quality was checked using Agilent 2100 Bioanalyzer; only samples with RIN (RNA integrity number) > 7 were taken for analysis.Before library preparation, ribosomal RNA was removed using Ribo-Zero Gold rRNA Removal Kit (Epicentre).rRNA-depleted RNA was then processed using TruSeq Stranded mRNA Library Prep Kit (Illumina).Libraries were sequenced on HiSeq2000 instrument with TruSeq v. 3 chemistry.Read length was 101 from each end of the fragment.

SNV and indel calling
In order to identify somatic and germline single nucleotide variants insertions/deletions, we ran VarScan2 [8] in the somatic mode on tumor and control samples.The discovered variants were annotated using the Annovar [9].The following parameters were used: • VarScan p-value < 0.05 (somatic p-value for somatic variants, variant p-value for germline variants) • Fraction of reads with alternative allele found in tumor sample > 20 % • Variant belonging to exonic or splicing region • >10 reads for alternative allele in tumor sample

Identification of damaging mutations
In order to assess mutation impact upon a protein function we utilized MutationAssessor [10] and PolyPhen2 [11].Additionally CHASM [12] software was used to differentiate between potential driver and passenger mutations.The following filters were applied: MutationAssessor score classification is high, low or medium OR Polyphen2 class is "delete-rious", OR CHASM score is less than 0.5, OR mutation is "nonsense".

Differential expression
For the differential expression analysis we followed the popular protocol [13], using Tophat2 for reads mapping and DESeq [14] for discovering genes with significantly different expression levels.We used 0.05 as a threshold for p-value, and left only genes for which expression levels ratio between normal and cancer tissues exceed 2. We also calculated logratio for each gene as log2 (expr. in tumor)/(expr.in normal).

RNA differential expression data analysis
As a result of RNA-seq data analysis we have identified 497 upregulated and 359 downregulated differentially expressed genes with FDR<0.05.No clear markers of pharmacological response (either FDA or preclinical) were found among them.In order to get indirect evidences about favorable pharmacological interventions we have classified obtained genes using different cancer hallmark processes (see Fig. 1) and checked the expression of genes, related to the pathways implicated in HCC treatment responses.
Sorafenib is a multikinase inhibitor and the first target drug approved by the FDA for the HCC treatment [15].In the studied tumor sample, PDGFA gene is upregulated relative to the control values, supporting the potential activation of the PDGFsignaling.We checked the CTD database [16] in order to define other cancer-driving differentially expressed genes, potentially affected by sorafenib action.Among the overexpressed genes is BIRC5 which is a negative regulator of apoptosis that prevents apoptotic cell death and that can be downregulated by sorafenib [17].Sorafenib can also inhibit HCC cell proliferation by blocking RAS/ RAF/MAPK and PI3K/AKT/mTOR pathways activated by overexpressed growth factor EGF [18].However, the genes described above could not be used for evaluation of sorafenib effectiveness in this case.
Alternatively, overexpressed EGF gene is a marker of EGFR/ERBB cascade activation with downstream PI3K/AKT1/mTOR and JAK/STAT signaling.In general, these cascades could be targeted by EGFR and ERBB2-inhibiting drugs Erlotinib and Lapatinib [19].A drug specific for PI3K/AKT1/ mTOR inhibition, Temsirolimus, could be specifically important because of the sorafenib ineffectiveness for this cascade.We further discuss the EGFR cascade and the corresponding drugs below in the context of the found genetic alterations.

Somatic SNVs and InDels
Exome sequencing revealed 9250 SNVs in the exonic or splicing regions, 77 somatic and 9173 germline variants.In order to identify somatic SNVs, potentially driving the cancer progression, we first filtered out dbSNP and silent mutations, leaving 23 missense or nonsense SNVs.Among these variants in the exonic or splicing regions, we identified 18 (see Table 1), predicted to be damaging by at least one of these tools: PolyPhen2 (PP2), MutationAccessor (MA) or CHASM (see Materials and Methods).
Using filtering, described in Materials and Methods, we have also identified 3 deletions in exonic regions, described in Table 2.

Somatically disturbed molecular pathways
All somatic SNVs and indels were manually curated in order to identify possible cancer driving pathways and potential pharmacological interventions.Some of the examples are presented below.

ERBB3 and EGFR pathway
EGFR/ERBB1, ERBB2 and ERBB3 comprise an EGFR family of tyrosine kinases.Interacting with corresponding ligands and forming the functional homo and hetero-dimers, EGFR/ERBB-receptors could transfer the signal inside the cell, regulating proliferation, migration and apoptosis.ERRB3, mutated in the studied tumor sample, can bind to the ligands but does not have its own kinase activity.Thus, ERBB3 could activate the downstream signaling only in complex with other ERBB receptors [20].
Mutation in ERBB3 is found as potentially driving by CHASM and statistically significant overexpression of EGF as well as less significant but coordinated overexpression of other members of this cascade, could characterize the aberrant activation of this mechanism in studied tumor.
The main signaling cascades activated downstream of EGFRs are PI3K/AKT1, MAP-kinase, and JAK/STAT (see Fig. 2).The activation of these cascades leads to the inhibition of apoptosis, uncontrolled cells proliferation and other pro-oncogenic processes.This activity can be suppressed by EGFR and ERBB2 inhibitors -Erlotinib and Lapatinib [19].There are several ongoing clinical trials, where these drugs are used as a second line therapy of HCC or in combination with sorafenib.
Alternatively, taking into account the PDGFA overexpression, the switch to the MTOR signaling is one of the probable scenarios.This cascade and its downstream targets could be suppressed by Temsirolimus.It could be specifically important because the mTOR activity is not targeted by standard sorafenib treatment.There are several clinical trials, where temsirolimus is used in combination with sorafenib for HCC treatment (NCT01008917).

BTG1 -potential driver
The gene BTG1 interacts with several nuclear receptors that could regulate differentiation of the cells [21], see Fig. 3.The somatic nonsense mutation K150*(chr12: 92537924) in BTG1 is probably damaging.It leads to the partial deletion of C-terminal region that is necessary for the BTG1 accumulation in nucleus and interaction with other proteins [22].Among the negative targets of BTG1 are antiapoptotic genes MMP9, BCL2 and CCND1, that could switch the tumor cells behavior towards the proliferative mode in response to the damaging BTG1 mutation.Additionally, BTG1 is shown to be downregulated in HCC [23].Summarizing, these evidences support the hypothesis about BTG1 as a driver gene in the case studied.FGFR2 expression in HCC is associated with unfavorable prognosis [24].The detected SNV in FGFR2 -V144F is considered as damaging.Possible activation of FGFR2 cascade provided by its ligands expression -FGF2 and FGF7 may suggest tyrosinekinase inhibitors therapy.

CIRBP
Somatic deletion in the CIRBP gene alters the polypeptide chain starting with the 101st residue, damaging the RGG domain that operates mRNA stability

Germline SNVs
Among the 9173 found germline SNVs in exonic regions we identified those 13 variants (Table 3) which were relevant to the drug toxicity and resistance according to PharmGKB database [25].In the studied case a possible effect of TP53 and DPYD germline mutations on tumor sensitivity to 5-fluorouracil was analyzed using information from scientific literature.Somatic SNV in the gene DPYD (C29R) activates the DPYD enzyme, which rapidly converts 5-FU to its inactive metabolite 5-dihydrofluo-rouracil [26].The identified TP53 polymorphism (R72P) also reduces the efficacy of the 5-FU therapy [27].Accordingly, the use of 5-FU therapy is likely to be ineffective in this case (see Fig. 5).SNV in the gene XRCC1 (R399Q) could be related to sensitivity to platinum therapies [28].Other germline SNVs also might be associated with therapy toxicity and adverse drug reactions.SNV in the gene MTHFR (E429A) might be associated with an increased risk of myelosuppression in the patients treated with methotrexate [29].SNV in CDA (K27Q) was shown to be associated with an increased severity of hematological toxicity, including neutropenia, in patients with pancreatic neoplasms treated with gemcitabine or cytarabine [30].SNV in XPC (Q902K), SLC22A2 (S270A), XRCC1 (R194W), LRP2 (K4094E) might be associated with an increased risk of drug toxicity when treated with cisplatin [31][32][33].SNV in UMPS (G213A) could be related with the increased likelihood of drug toxicity when treated with fluorouracil and leucovorin.ERBB2 polymorphism (I625V) may be associated with cardiotoxicity under trastuzumab treatment.SLC19A1 polymorphism (H27R) might be related with drug toxicity under methotrexate and mercaptopurine treatment.

Fig. 1 .
Fig. 1.The distribution of genes with altered expression across different cancer hallmark processes.