Can pharmacogenomics help to improve therapy in patients with high-grade osteosarcoma?

Abstract

Current standard treatment of patients with high-grade osteosarcoma (HGOS) includes complete surgical resection of the tumor and chemotherapy, most often with high-dose methotrexate, doxorubicin and cisplatin. With this approach > 60% of patients can be cured. However, conventional anticancer drugs have a narrow therapeutic index, and efficacy and toxicity vary considerably among patients. Pharmacogenomics aim to identify key genomic factors for drug effects (either desired or adverse) in normal host cells (germ-line variants) and cancer cells (somatic variants), and if an association between a genotype and a drug phenotype has been identified, validated and demonstrated to have a large effect size, these genotypes may be used to tailor therapy. In addition, pharmacogenomic models can be used to identify novel therapeutic targets. For example, germ-line variants in genes which potentially influence the disposition of methotrexate and cardiotoxicity of doxorubicin have recently been identified. Moreover, next-generation sequencing combined with several analytical methods has identified the phosphatidylinositol 3-kinase/mammalian target of rapamycin (PI3K/mTOR) pathway as a potential therapeutic target in HGOS. Herein, we discuss whether and how these novel pharmacogenomic insights may help to improve future therapy in HGOS.

Keywords:

• efficacy
• osteosarcoma
• pharmacogenomics
• toxicity

High-grade osteosaroma (HGOS) is the most common primary malignant tumor of bone. HGOS is an ultra-orphan disease (i.e., 1 – 3 patients/million/year), and the highest incidence is reported during the puberty growth spurt. HGOSs are highly aggressive and ∼ 20% of patients have radiographically detectable metastases and ∼ 90% of patients have micrometastases [1].

Current multi-modal front-line therapy for patients with HGOS encompasses complete resection of all radiographically detectable lesions (primary tumors and metastases) and control of microscopic metastatic disease via systemic neo- and adjuvant polychemotherapy. While complete surgical resection is of paramount importance, surgery alone can cure < 20% of patients [1].

Based on the results of numerous clinical trials, there are currently three chemotherapeutic agents which might be considered standard in front-line HGOS therapy, namely high-dose methotrexate (HDMTX) with leucovorin rescue, doxorubicin (DOX) and cisplatin (CDDP) (so-called ‘MAP backbone’) [2]. Other agents with efficacy include ifosfamide, carboplatin and etoposide [1]. In the yet largest, recently completed European-American osteosarcoma treatment trial 1 (EURAMOS-1), MAP was given preoperatively (neoadjuvant) for 10 weeks. At definitive surgery, 50% of patients had < 10% tumour viability (i.e., good response) in the resected specimens [2]. Response to neoadjuvant chemotherapy is an important prognostic factor, and patients whose tumors responding well to preoperative therapy can achieve long-term overall survival rates of about 80% [3]. Based upon their exciting results in a small cohort of patients whose tumors had responded poorly to neoadjuvant chemotherapy, Rosen et al. had hypothesized back in 1982 that altering postoperative chemotherapy (i.e., intensification) should improve outcomes in such patients. However, subsequent trials addressing this issue with larger patient numbers, including EURAMOS-1 as the only trial employing a prospectively randomized design, did not validate this hypothesis; so that currently there is an urgent need for novel therapies especially for this sub-cohort of patients with a poor overall survival rate.

Current standard HGOS front-line chemotherapy can cause significant short- and long-term morbidity (e.g., nephrotoxicity of MTX, cardiotoxicity of anthracyclines, oto- and nephrotoxicity of CDDP, etc.), may predispose to secondary cancer or can even result in death due to lethal complications [4]. Obviously, there is a need to a priori identify patients at highest risk for severe complications, with the aim to modify therapy accordingly. However, a simple dose reduction of drugs may not do the job, as this can also place the patient at higher risk of poor tumor response to therapy and poor outcome. Therefore, profound insights into pharmacokinetics and pharmacodynamics of drugs are essential prerequisites before one should think about individualizing drug therapy.

Conventional anticancer drugs have a narrow therapeutic index, and efficacy and toxicity vary considerably among patients. Variation in drug effects result from the interplay of many variables, including for example, age, sex, ethnicity, nutritional status, renal and liver function, concomitant illnesses, other medications and genetics. In HGOS, for example, age and sex have been identified as independent prognostic variables, and children and females have obtained better outcomes than older patients and males, respectively [3].

Pharmacogenomics aim to establish clinically useful models by integrating information from functional genomics, high-throughput molecular analyses and pharmacodynamics. Approaches to establish pharmacogenomic models include candidate gene analyses and genome-wide analyses. The ultimate goal of pharmacogenomics is to use these models to maximize efficacy and reduce toxicity of existing medications, as well as to identify novel therapeutic targets.

This editorial accompanies a recent review article on the ‘Role of pharmacogenetics of drug metabolizing enzymes in treating osteosarcoma’ – which focused on the analysis of three drug metabolizing enzyme families as candidate genes. We herein intend to extend this view by focusing on selected novel insights into HDMTX and DOX pharmacogenomics, and on a novel insight on a possible new therapeutic target in HGOS, which has been identified using next-generation sequence analyses.

HDMTX (typically 12 g/m² infused over 4 h with alkalinisation, hydration, leucovorin rescue, and monitoring of MTX plasma levels) is an essential component of HGOS therapy. Major adverse drug reactions include neurotoxicity (0.4 – 5% of patients), and nephrotoxicity (up to 1.8% of patients) [4]. Despite the use of hemodialysis or administration of glucarpidase (which hydrolyses MTX into less toxic metabolites) to rescue patients with acute HDMTX-induced nephrotoxicity, severe and sometimes fatal intoxications can occur. Other, less severe side effects include mucositis and myelotoxicity; and most side effects result from delayed MTX excretion and altered pharmacokinetics. A decrease in MTX dose intensity is a risk factor for treatment failure in HGOS. The importance of MTX serum peak levels, however, is less well established.

After HDMTX infusion, there is substantial inter- but also intra-individual variability in MTX pharmacokinetics. The main route of elimination of MTX and its metabolite 7-OH-MTX is renal excretion, and biliary secretion contributes < 30%. Transport of MTX via membranes (e.g., in tumor, kidney and liver cells) involves several transporters, including the solute carrier family members (e.g., SLC19A2), the solute carrier organic anion transporter family members (e.g., SLC22A6, SLC22A7, SLC22A8 and SLCO1B1), and ATP-binding cassette transporters (e.g., ABCC2, ABCC4 and ABCG2); functional relevant germ-line variants in genes encoding these transporters can potentially influence MTX pharmacokinetics and pharmacodynamics. Among these transporters, only functional germ-line variants in SLCO1B1 have been validated to influence MTX disposition. Using a genome-wide approach, researchers from the St. Jude Children’s Research Hospital were able to find an association between germ-line variants in SLCO1B1 and altered MTX clearance in children with acute lymphoblastic leukemia (ALL); and these results were later robustly confirmed with five different MTX treatment regimens in > 1000 ALL patients [5]. This transporter, however, is associated with hepatobiliary excretion, which is only a minor path for MTX elimination. Therefore, the overall contribution of SLCOB1B variants to explaining interindividual variability in MTX pharmacokinetics is < 15%, and the major genetic contributors are unknown yet. Recently, Goricar et al. found that ‘low-clearance’ SLCOB1B variants were associated with increased MTX disposition (serum levels at 24 h after HDMTX and AUC) and better outcomes in patients with HGOS [6]. As this investigation encompassed only 44 patients, confirmation of the findings in large cohorts are necessary before any firm conclusions can be drawn.

The anthracycline DOX is given at relative high doses in patients with HGOS, and the scheduled cumulative dose in EURAMOS-1, for example, was 450 mg/m² [2]. Cardiac toxicity is a major concern in anthracycline therapy, and acute (occurring after a single dose or course), chronic and late onset (heart failure occurs > 1 year after completion of therapy) congestive heart failure (CHF) have been described. Clinically apparent anthracyline-induced cardiotoxicity (ACT) has been reported up to 4% in Phase III HGOS treatment trials; and age (higher risk < 4 and > 40 years), cumulative dose and female sex are well-defined risk factors [4]. A meta-analysis concluded that prolonged DOX infusion (> 6 h) decreased the risk of CHF, especially in adults [7]. However, a prospective randomized trial in young children with ALL did not find a difference in CHF between children who received bolus or long-term infusions [7].

Recent insights provide evidence that ACT is at least partially mediated through mitochondrial iron accumulation, and dexrazoxane (a drug that attenuates ACT) is able to reduce these toxic iron levels [8]. Dexrazoxane therapy, however, has been reported to be possibly associated with a higher risk of secondary cancers in children and adolescents, and the European Medicine Agency recommends that dexrazoxane should not be used in this age group (http://www.ema.europa.eu/ema/).

Marked heterogeneity in individual susceptibility for ACT has been observed, and recent pharmacogenomic investigations (including patients with HGOS) have used different approaches to identify genomic risk factors for ACT in children. The North American Children’s Oncology Group (COG), for example, used a candidate gene approach and a case-control design to analyze whether germ-line variants in carbonyl reductase-genes (CBR1 and CBR3; which catalyze reduction of anthracyclines) are associated with ACT; and identified CBR3 variants to be associated with a higher risk for ACT in patients who were treated with anthracycline doses < 250 mg/m², but not after higher doses [9]. Canadian and Dutch researchers used a candidate gene approach (2977 single-nucleotide polymorphisms [SNPs] in 220 key drug biotransformation genes) and identified a number of germ-line variants (mainly in genes encoding transporters like SLC28A3, SLC28A2, ABCB1, ABCB4, ABCC1, SLC10A2) to be associated with ACT, and the strongest association was found for SLC28A3 variants [10]. Although included, the CBR3 variants were not found to be associated with a higher risk of ACT in this study [10]. In a subsequent COG investigation, a case-control candidate gene approach (using the ITMAT/Broad CARe [IBC] cardiovascular SNP array, which profiles SNPs in 2100 genes considered to be relevant for cardiovascular disease) identified germ-line variants in the hyaluron synthase 3 gene (HAS3) to be associated with an 8.7-fold higher risk for ACT in patients exposed to high cumulative anthracycline doses (> 250 mg/m²) [11]. Of note, neither the CBR3 nor the SLC28A3 genes were included in this study. Unfortunately, none of the identified variants can currently be used in clinical routine to identify patients with a high risk for ACT, and more investigations are necessary before such findings can be used to tailor DOX therapy. On the other hand, different approaches to alter the pharmacological properties of anthracyclines, for example by conjugating DOX to squaline, may also help to enhance efficacy and reduce toxicity of anthracyclines [12].

Next-generation sequencing (NGS) technologies have shed light into the complexity of HGOS genomes, but – not surprisingly – were unable to identify driver oncogenes [13,14]. Whereas most pediatric cancers typically have a low somatic mutation rate with a median of ∼ 0.1 mutations/megabase, HGOS often carries extremely complex genotypes with a median of ∼ 1.2 mutations/megabase [14]. In order to identify therapeutic targets in genetically complex HGOS, Perry et al. used a sophisticated pharmacogenomic approach and explored NGS data with a number of analytic methods. Pathway analyses, for example, validated results from Chen et al. showing the p53 and the retinoblastoma pathways to be altered in most of the tumors [13,14]. Using other tools (e.g., geneset enrichment analysis, Molecular Signature Database, etc.), the phosphatidylinositol 3-kinase/mammalian target of rapamycin (PI3K/mTOR) pathway was identified as being altered in 24% of tumors. Based on this data, Perry et al. successfully treated HGOS cell lines with dual PI3K/mTOR inhibitors, even in the absence of pathway mutations [14]. Collectively, these results can be interpreted as that mTOR is a signaling ‘choke point’ in HGOS where many upstream signals converge, and mTOR inhibition may be a novel potential therapeutic target in this disease.

Expert opinion

There is an urgent need to improve outcome in HGOS, and strategies to achieve this can focus either on improving the use of current (standard) medications or on identifying novel therapeutic targets in those patients whose tumors do not sufficiently respond to standard therapy.

In about 50% of patients, tumors respond ‘well’ to standard MAP therapy and 80% of these patients can be cured. In this subcohort, pharmacogenomic investigations, which focus on the optimization of MAP drugs, may help to further ‘fine-tune’ MAP therapy based upon genotypes, potentially resulting in better outcomes. Currently identified major germ-line variants in genes that potentially influence MAP drugs pharmacology (e.g., SLCOB1B, SLC28A3, CBR3, HAS3, etc.), however, have not yet been validated in large cohorts of patients with HGOS. Future HGOS trials might, therefore, aim to include pharmacogenomic studies with well-defined pharmacodynamic end points. The investigation of somatic variants to study resistance mechanisms to MAP drugs (e.g., reduced levels of active polyglutamated MTX metabolites ‘MTXPGs’ in tumor cells due to reduced formation, enhanced degradation or export) is another important field of research, but was not considered in this editorial due to the lack of space.

In the 50% of patients whose tumors respond ‘poorly’ to neoadjuvant therapy and who have a poor outcome with survival rates of about 50% [3], pharmacogenomic investigations might focus on the identification of novel therapeutic targets. Indeed, recent insights from NGS analyses identified the PI3K/mTOR pathway as a promising therapeutic target [14]. In-line with this, a non-randomized Phase II trial from the Italian Sarcoma Group used an oral combination therapy with sorafenib and everolimus in heavily pre-treated patients with relapsed HGOS and found better outcomes with this combination compared to sorafenib alone [15]. Everolimus is a rapalog (synthetic analog of rapamycin) with only incomplete inhibition of mTOR, and dual inhibitors of PI3K/mTOR (e.g., BEZ235 or GSK2126458) with stronger effects on the pathway may be promising novel drugs on the horizon for patients with HGOS.

In conclusion, results from pharmacogenomic studies have so far not changed clinical practice in HGOS treatment. Pharmacogenomic investigations, however, should be included in future clinical HGOS treatment trials, because there is considerable potential to improve HGOS therapy. Important prerequisites for successful implementation include the collection of blood (germ line) and tumor material (somatic) at defined time points (e.g., tumor material from diagnosis and definitive surgery; blood from diagnosis, etc.), precisely up-front defined pharmacological end points, and long-term follow up. Whereas in the past pharmacogenomic research focused mainly on non-synonymous SNPs (i.e., variants in coding regions), future research should also include synonymous SNPs (i.e., variants in non-coding regions), variants in micro RNAs (so called ‘miRSNPs’) and epigenetic variation (so called ‘pharmacoepigenetics’), all of which have already been proven to influence drug effects.

Declaration of interest

L Kager is on the advisory board for Takeda. S Bielack is a consultant and/or is on the advisory board for Merck, Roche, IDM/Takeda, Celgene, Chugai and Bayer Healthcare. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.