Kalyanasundaram Subramanian

A systems biology based integrative framework to enhance the predictivity of in vitro methods for drug-induced liver injury

Background: Liver injury is the most common cause of postmarketing withdrawal of drugs. Traditional animal toxicity testing methods have proved to be imperfect tools for predicting toxicity observed in the clinic. Objective: Predictive methods that integrate data and insights from several in vitro methods to provide a deeper understanding of the impact of a drug on the liver are the need of the hour. Method: A systems approach based on mathematical modelling using the kinetics of biochemical pathways involved in liver homeostasis coupled with in vitro measurements to quantify drug-induced perturbations is described here. Conclusions: Integrating in silico and in vitro methods provides a powerful platform that allows reasonably accurate and mechanistic-level prediction of drug-induced liver injury. The method demonstrates that several physiological situations can be accurately modelled as can the effect of perturbations induced by drugs. It can also be used along with high-throughput ‘omic’ data to generate testable hypotheses leading to informed decision-making.

Detection of high frequency of mutations in a breast and/or ovarian cancer cohort: implications of embracing a multi-gene panel in molecular diagnosis in India

Breast and/or ovarian cancer (BOC) are among the most frequently diagnosed forms of hereditary cancers and leading cause of death in India. This emphasizes on the need for a cost-effective method for early detection of these cancers. We sequenced 141 unrelated patients and families with BOC using the TruSight Cancer panel, which includes 13 genes strongly associated with risk of inherited BOC. Multi-gene sequencing was done on the Illumina MiSeq platform. Genetic variations were identified using the Strand NGS software and interpreted using the StrandOmics platform. We were able to detect pathogenic mutations in 51 (36.2%) cases, out of which 19 were novel mutations. When we considered familial breast cancer cases only, the detection rate increased to 52%. When cases were stratified based on age of diagnosis into three categories, ⩽40 years, 40–50 years and >50 years, the detection rates were higher in the first two categories (44.4% and 53.4%, respectively) as compared with the third category, in which it was 26.9%. Our study suggests that next-generation sequencing-based multi-gene panels increase the sensitivity of mutation detection and help in identifying patients with a high risk of developing cancer as compared with sequential tests of individual genes.

Identifying a Clinically Applicable Mutational Burden Threshold as a Potential Biomarker of Response to Immune Checkpoint Therapy in Solid Tumors

Purpose An association between mutational burden and response to immune checkpoint therapy has been documented in several cancer types. The potential for such a mutational burden threshold to predict response to immune checkpoint therapy was evaluated in several clinical datasets, where mutational burden was measured either by whole-exome sequencing (WXS) or using commercially available sequencing panels. Methods WXS and RNA-seq data of 33 solid cancer types from TCGA were analyzed to determine whether a robust immune checkpoint activating mutation (iCAM) burden threshold associated with evidence of immune checkpoint activation exists in these cancers that may serve as a biomarker for response to immune checkpoint blockade therapy. Results We find that a robust iCAM threshold, associated with signatures of immune checkpoint activation, exists in 8 of 33 solid cancers: melanoma, lung adenocarcinoma, colon adenocarcinoma, endometrial cancer, stomach adenocarcinoma, cervical cancer, ER+HER2- breast cancer, and bladder-urothelial cancer. Tumors with mutational burden higher than the threshold (iCAM+) also had clear histologic evidence of lymphocytic infiltration. In published datasets of melanoma, lung adenocarcinoma and colon cancer, patients with iCAM+ tumors had significantly better response to immune checkpoint therapy compared to those with iCAM- tumors. ROC analysis using TCGA predictions as gold standard showed that iCAM+ tumors are accurately identifiable using clinical sequencing assays, such as FoundationOne or StrandAdvantage. Using the FoundationOne derived threshold, analysis of 113 melanoma tumors, showed that iCAM+ patients have significantly better response to immune checkpoint therapy. iCAM+ and iCAM- tumors have distinct mutation patterns and different immune microenvironments. Conclusion In 8 solid cancers, a mutational burden threshold exists that may predict response to immune checkpoint blockade. This threshold is identifiable using available clinical sequencing assays.

StrandAdvantage test for early-line and advanced-stage treatment decisions in solid tumors

Comprehensive genetic profiling of tumors using next‐generation sequencing (NGS) is gaining acceptance for guiding treatment decisions in cancer care. We designed a cancer profiling test combining both deep sequencing and immunohistochemistry (IHC) of relevant cancer targets to aid therapy choices in both standard‐of‐care (SOC) and advanced‐stage treatments for solid tumors. The SOC report is provided in a short turnaround time for four tumors, namely lung, breast, colon, and melanoma, followed by an investigational report. For other tumor types, an investigational report is provided. The NGS assay reports single‐nucleotide variants (SNVs), copy number variations (CNVs), and translocations in 152 cancer‐related genes. The tissue‐specific IHC tests include routine and less common markers associated with drugs used in SOC settings. We describe the standardization, validation, and clinical utility of the StrandAdvantage test (SA test) using more than 250 solid tumor formalin‐fixed paraffin‐embedded (FFPE) samples and control cell line samples. The NGS test showed high reproducibility and accuracy of >99%. The test provided relevant clinical information for SOC treatment as well as more information related to investigational options and clinical trials for >95% of advanced‐stage patients. In conclusion, the SA test comprising a robust and accurate NGS assay combined with clinically relevant IHC tests can detect somatic changes of clinical significance for strategic cancer management in all the stages.

Expression of tripartite motif-containing protein 28 in primary breast carcinoma predicts metastasis and is involved in the stemness, chemoresistance, and tumor growth:

The prediction of who develops metastasis has been the most difficult aspect in the management of breast cancer patients. The lymph node metastasis has been the most useful predictor of prognosis and patient management. However, a good proportion of patients with lymph node positivity remain disease free for 5 years or more, while about a third of those who were lymph node negative develop distant metastasis within the same period. This warrants a robust biomarker(s), preferably gene expression based. In order to elucidate gene-based biomarkers for prognosis of breast cancers, gene expression profiling of primary tumors and follow-up for over 5 years has been performed. The analysis revealed a network of genes centered around the tripartite motif-containing protein 28 as an important indicator of disease progression. Short hairpin RNA–mediated knockdown of tripartite motif-containing protein 28 in breast cancer cells revealed a decreased expression of epithelial-to-mesenchymal transition markers and increased expression of epithelial markers, decreased migration and invasion, and increased chemosensitivity to doxorubicin, 5-fluorouracil, and methotrexate. Furthermore, knockdown of tripartite motif-containing protein 28 resulted in the decrease of stemness as revealed by sphere formation assay as well as decreased expression of CD44 and Bmi1. Moreover, tripartite motif-containing protein 28 knockdown significantly reduced the tumor size and lung metastasis in orthotopic tumor xenograft assay in immunocompromised mice. The tumor size was further reduced when these mice were treated with doxorubicin. These data provide evidence for tripartite motif-containing protein 28 as a biomarker and a potential therapeutic target for breast cancer.

Identification of a novel glucokinase mutation in an Indian woman with GCK-MODY

Glucokinase-maturity-onset diabetes of the young (GCK-MODY; also known as MODY 2) is believed to cause 1–2% of cases diagnosed as gestational diabetes. Pregnant women with GCK-MODY should be diff erentiated from those with gestational diabetes, because diff er ent management is needed. The prevalence of GCK-MODY in Asians is unclear because of a paucity of epidemiological data, although Rudland and colleagues estimated a prevalence of about 1–1·9 per 100 for Indian women diagnosed with gestational diabetes. We identifi ed a 29-year-old Indian woman with mild fasting hyperglycemia during pregnancy. 1 year before gestation, routine biochemical testing had shown fasting plasma glucose (FPG) of 6·1 mmol/L. At this time, a 2-h oral glucose tolerance test had shown an FPG concentration of 6·9 mmol/L and 2-h FPG concentration of 7·8 mmol/L. Further testing before pregnancy showed a fasting C-peptide concentration of 0·9 ng/mL, an HbA1c of 48 mmol/mol (6·5%), a fasting insulin concentration of 6 μIU/mL, a GAD-65 autoantibody titre of less than 5 IU/mL, an islet cell antibody titre of less than 1:4, and a BMI of 19 kg/m2. A diagnosis of diabetes was then made on the basis of the HbA1c measurement before pregnancy. However, negative antibody titres and an absence of clinical or biochemical features consistent with insulin resistance precluded type 1 or type 2 diabetes. Genetic testing before pregnancy identifi ed a variant in the GCK gene: c.1030G>T; p.Asp344Tyr, but it was not known whether this variant was clinically relevant. After conception, FPG was about 6·1 mmol/L. We expected glucose levels to come down after conception, which usually happens in a normal pregnancy. However, when FPG remained high during the fi rst trimester, and BMI being low, we made the diagnosis of GCK-MODY and began monitoring fetal size. Fetal growth was monitored with serial ultrasounds, and the patient was managed with lifestyle modifi cation alone. 2-h postprandial plasma glucose remained in the 6·1–6·7 mmol/L range. Since the fetus was growing appropriately, we assumed that it had inherited the same mutation, because had it not inherited the mutation, it would be at high risk of macrosomia. At 38 weeks gestation, the patient gave birth to a healthy boy (3·3 kg). The baby did not develop macrosomia, possibly in part because the mother’s glycaemic control was overall at target. Direct maternal gene sequencing (saliva) was repeated, confi rming a heterozygous variant of unknown clinical signifi cance in exon 9 of the GCK gene (chr7:44185319C>A, c.1030G>T, p.Asp344Tyr). The variant was in the vicinity of other missense variants that are probably pathogenic (p.Ser340Gly and p.Ile348Asn), and was predicted to be damaging because of its conserved nature and proximity to other previously reported pathogenic variants. This variant is not among the known 620 GCK mutations that have been identifi ed in 1441 families. We did a family segregation analysis in the patient’s immediate family members, as well as her newborn baby (her spouse was not tested because fasting hyperglycemia had not been documented). The patient’s father and brother had the same mutation. Her 54-year-old father (BMI 29·5 kg/m2) had an FPG of 6·6 mmol/L and an HbA1c of 6·5%. Her 20-year-old brother (BMI 28·65 kg/m2) had an FPG of 6·3 mmol/L and an HbA1c of 6·0%. The baby was found not to have inherited the mutation and was not tested for diabetes. To our knowledge, this GCK variant has never been described in any ethnic group. We believe that the variant is pathogenic, since all aff ected family members have FBG and HbA1c measurements consistent with a GCK-MODY phenotype.

Big Data Analytics and Molecular Medicine

Diagnostics have a major role to play in improving patient care, protecting consumer health and reducing health care costs. The quality of patient care can be significantly improved by detecting and diagnosing disease earlier and more rapidly. This is especially true in the case of cancer where accurate and early diagnosis can provide more targeted and effective treatment options leading to better outcomes. Diagnostic tests can also provide companies with accurate quality checks of their products thus ensuring product safety and consequently protecting consumer health.

Abstract 1424: Predicting response to immune checkpoint therapy using an mutation burden threshold

Treatment with antibodies to PD-1 or CTLA-4 leads to prolonged response in some cancer patients. Patients whose tumors have a high mutational burden have a better response to anti-CTLA-4 treatment with ipilimumab in melanoma5, and to anti-PD-1 treatment with pembrolizumab in non-small cell lung6 and colorectal cancer7. These results suggest that a sufficiently high non-synonymous somatic mutation burden in the tumor may lead to protein alterations that serve as neo-antigens to stimulate CD8 T-cell immune response. This immune response may be blocked by the tumor by engaging the immune checkpoint pathway, making such tumors especially vulnerable to immune checkpoint disruption. However, the specific criteria to identify patients likely to respond to immune checkpoint therapy have remained elusive so far. Using somatic mutation and RNA-Seq data from TCGA, we propose that there exists a mutational threshold, which we call the “Activated Immune Mutational” (AIM) Threshold, which can identify patients likely to respond to immune checkpoint therapy. Compared to AIM- patients (potential non-responders), AIM+ patients (potential responders) have tumors that meet four criteria: (1) a high non-synonymous mutation burden; (2) a high level of immune infiltration in the tumor; (3) a high CD8 T cell fraction in the leukocyte component of the immune infiltrate plus high expression of the T cell marker CD8A; and (4) a high expression of immune checkpoint genes PD-1, PD-L1, PD-L2, CTLA-4. We find that in melanoma, endometrial, colon, cervical and breast cancer, a clear AIM threshold satisfying all four criteria exists. The distribution of mutations in AIM- versus AIM+ tumors for these cancers is very different. In the former, there are high frequency somatic mutations in only a few genes, while in the latter, high frequency somatic mutations are found in many genes distributed over the whole genome. Pathological analysis of high-resolution images from 150 TCGA tumors (30 from each of the five tumor types) validated the presence of significantly higher lymphocytic infiltration in the AIM+ tumors compared to AIM- tumors. We show that AIM+ tumors can be identified using sequencing assays which are currently in clinical use. Finally, survival analysis in melanoma patients from TCGA treated with immunotherapy will be presented. Citation Format: Anshuman Panda, Anil Betigeri, Kalyanasundaram Subramanian, Kim Hirshfield, Lorna Rodriguez, Shridar Ganesan, Gyan Bhanot. Predicting response to immune checkpoint therapy using an mutation burden threshold. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1424.

Abstract 4878: Analytical and technical validation of a cost-effective diagnostic test for BRCA1, BRCA2 and TP53

Proceedings: AACR 106th Annual Meeting 2015; April 18-22, 2015; Philadelphia, PA BRCA1, BRCA2 and TP53 encode tumor suppressor proteins in humans that help repair damaged DNA and play critical roles in ensuring genome stability. Several inherited mutations in any of the above 3 genes substantially increase the risk of cancer. Together, BRCA1 and BRCA2 mutations account for about 20 to 25 percent of hereditary breast and ovarian cancers. Germline mutations in TP53 are the most common cause of Li-Fraumeni syndrome, a rare disorder that increases the risk of developing multiple tumors such as breast, soft-tissue and leukemias, in children and young adults. In India, where the incidence of cancer has seen a steep rise in the last decade, there is a pressing need to develop cost-effective screening tests that can identify known and novel mutations in commonly associated genes. We have developed and offer a 3-Gene panel (Strand® - 3 gene) covering all known HGMD/ClinVar mutations and all coding exons of BRCA1, BRCA2 and TP53 genes. Current Sanger based methods query for restricted loci across these genes. Our test is based on an NGS enrichment protocol using xGen lockdown probes that allows parallel sequencing of upto 32 - 96 samples. The test would be offered at a tenth of the cost of current Sanger based tests anywhere in the world. In this study we present the technical and clinical validation data obtained from this assay. For technical validation, we included “gold standard” HAPMAP characterized as part of 1000 Genome Project, seven cell lines with known BRCA and TP53 mutations. For clinical validation, we enrolled thirty seven (37) patients who were consented on an IRB-approved study at HCG hospital for collecting saliva / blood. These patients were stratified / selected based on their family history, known risk of hereditary cancers and availability of previously characterized clinical samples. The overall sensitivity and specificity of this panel is 99.78% and 99.74% respectively with a reproducibility of 100%. On an average, 99.75% and 97% of the bases are covered at 0.2x and 0.5x mean coverage and the average gap (<20 reads per base) is 0.0056% in the validation study. We have identified 3 separate cases of Li-Fraumeni from the cohort of thirty seven patients. We further present clinical validation data from these 3 case studies in which we have identified both known and novel mutations. Further clinical validation of panel is ongoing. In summary this panel will provide a cost effective screening method for early detection of pathogenic variants in pre-symptomatic individuals and in families with known risk of hereditary cancer. Citation Format: Manimala Sen, Pooja Agrawal, Vikram Vittal P, Mithua Ghosh, M.L Sheela, Divya Vishwanath, Kiran Kumari, Swetha N.S.N, Vaibhavi Pathak, Gouri Deshpande, Ashraf Mannan, Rupali Gadkari, Suman Kapoor, Jamuna Yadhav, Mohammed Yousuff, Satish Sankaran, Ramesh Hariharan, Preveen Ramamoorthy, Kalyanasundaram Subramanian, Vaijayanti Gupta. Analytical and technical validation of a cost-effective diagnostic test for BRCA1, BRCA2 and TP53. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4878. doi:10.1158/1538-7445.AM2015-4878

Integrated Systems Biology Approaches to Predicting Drug-Induced Liver Toxicity: A Dynamic Systems Model of Rat Liver Homeostasis Combined with In Vitro Measurements to Predict In Vivo Toxicity

Liver injury is the most common cause of post-marketing drug withdrawal. Predicting toxicity observed in the clinic, especially idiosyncratic toxicity, is extremely challenging. In this chapter we developed a predictive system that integrates different data types and provides insight into the mechanisms of drug-induced liver injury. This is a dynamic systems approach based on the mathematical modeling of the kinetics of metabolic pathways involved in liver homeostasis. Drug-induced perturbations to this homeostasis that lead to toxicity can be measured by targeted in vitro assays. Several physiological and pathological situations can be accurately modeled by integrating in silico and in vitro methods. What we also demonstrate is that the method is flexible enough to allow an understanding of the mechanistic basis for idiosyncratic toxicity and individual variations in toxic responses. It can also be used along with functional genomic data to generate mechanistic hypotheses of drug action.