Greater expectations: meeting clinical needs through broad and rapid genomic testing

Background

Cancer arises from the neoplastic transformation of healthy cells driven by the accumulation of genetic and genomic changes over time. These changes, or variants, can be classified as inherited germline polymorphisms that are present at the time of conception or acquired somatic mutations that arise during the process of transformation [1]. Cancer cells acquire thousands of different mutations over their lifetime, but only a few of these changes are responsible for driving the neoplastic process. These “driver mutations” allow cells to bypass the normal regulatory mechanisms that inhibit unchecked cell growth and replication through a variety of mechanisms including the activation of sustained proliferative signaling and the avoidance of destruction by the immune system [2, 3]. The remaining somatic mutations that do not actively cause cancer are termed “passenger mutations” and thought to be acquired due to decreased DNA repair and increased mutagenesis.

Despite the more than 20,000 protein-coding genes encoded by the human genome, only an estimated 300–600 have been shown to be associated with cancer [1]. Simplified, these genes can be classified as either oncogenes or tumor suppressors depending on their function and how they contribute to cancer progression. Proto-oncogenes normally function to promote cell growth, division, and survival and are controlled under tight regulation [2]. When mutated, most variants found in oncogenes, such as BRAF, are gain-of-function, often leading to protein products that are constitutively active and unresponsive to normal regulatory signals [4]. On the other hand, tumor suppressor genes typically function to slow cell division, repair DNA mistakes, and activate cell death [2]. Variants in tumor suppressors, such as TP53, are often loss-of-function, leading to protein products that are inactive and can no longer work properly [2]. Lastly, some variants can be change-of-function, such as those seen in IDH1, which can cause the enzyme to produce the oncometabolite D-2-hydroxyglutarate instead of α-ketoglutarate [5].

Driver mutations encompass many variant types including relatively simple single nucleotide substitutions, also called single nucleotide variants (SNVs), insertions or deletions (indels) of one or more nucleotides, and more complex gains or losses up to and including partial or whole chromosomes that result in copy number variants (CNVs). Gene fusions represent another common variant type in which portions of two independent genes are rearranged so that they produce a single, hybrid mRNA transcript and protein. Characterization of the functional consequence of mutations detected in somatic sequencing is complex and relies heavily on prior knowledge of gene function, literature review of the specific mutation, landscape of other mutations in the tumor type, and predicted outcome of the mutation.

Identifying variants in a patient’s tumor can have important implications in the treatment, diagnosis, prognosis, risk assessment, and/or management of their cancer. The detection of variants with a molecularly targeted therapy that is approved or under investigation in a clinical trial is particularly important for clinical decision making as patients treated with matched targeted therapies often have improved overall and progression-free survival as well as decreased adverse side effects compared to those treated without [6]. Furthermore, the prevalence of actionable variants in solid tumors can be high, ranging from 25% for ovarian cancers up to 92% for endometrial cancers [7]. In hematological malignancies, their prevalence is estimated to be between 51% and 75% for all diagnoses [8, 9]. Therefore, identification of actionable variants can lead to improved clinical outcomes for a significant number of patients by guiding targeted treatment decisions.

In addition, some of these variants are hallmarks of certain types of cancer, such as PML::RARA fusions for acute promyelocytic leukemia (APL), which aids in the confirmation of a diagnosis [10]. These biomarkers also can be used to understand the prognosis of a patient’s cancer and better tailor treatment decisions based on its aggressiveness. For example, patients with mutations in TP53 typically have shorter survival or poorer responses to treatment predominantly in breast cancer, bone and soft tissue sarcomas, brain tumors, and adrenocortical carcinomas [11]. Finally, knowledge of a patient’s mutational status can be used in disease management to monitor their response to therapy, identify minimal residual disease (MRD), and detect the emergence of resistance mutations.

The importance of NGS for genomic profiling

Prior to the advent of next-generation sequencing (NGS) technology, the mainstay of genetic testing was Sanger sequencing. While this technique is low-cost and yields results with fast turnaround time (TAT), it can only sequence a single DNA amplicon of up to ∼1,000 base pairs (bp) per reaction, thus leading to lower sample throughput and limited sequencing information per sample (Figure 1). In addition, Sanger sequencing only affords a limit of detection (LOD) of ∼15–20% variant allele frequency (VAF) and low variant resolution as it only detects small genetic changes such as SNVs and short indels [12]. Given that oncogenic variants can be found in hundreds of different genes and include large changes like chromosomal gains and losses, Sanger sequencing is not practical for investigating tumor mutational landscapes.

Figure 1:

Comparison of Sanger and next-generation sequencing.

NGS has emerged as an alternative method for sequencing nucleic acids to overcome some of the drawbacks of Sanger sequencing. Also known as massively parallel sequencing (MPS), NGS is capable of sequencing millions of different DNA fragments in a single reaction, which allows for the coverage of entire genomes or more focused targets such as whole exomes or any number of specific genes (Figure 1). In addition, NGS reactions can be multiplexed with the use of unique “barcode” sequences that are added to individual samples during library preparation. Because of these features, NGS permits higher throughput and generates more information per sample than traditional Sanger sequencing, which is particularly important in oncology as hundreds of different genes are known to drive cancer and biopsy samples can be too limited to run on multiple tests.

The importance of NGS testing is underscored by professional guidelines from the European Society of Medical Oncology (ESMO) which recommend that it be performed as part of biomarker testing for certain cancers such as non-small cell lung cancer (NSCLC), prostate cancers, ovarian cancers, and cholangiocarcinoma [13]. With NSCLC, for example, NGS can be useful over the entirety of a patient’s clinical management. Identification of common actionable variants driving NSCLC tumors such as EGFR L858R and exon 19 deletions (15% prevalence), ALK fusions (5% prevalence), and MET exon 14 skipping (3% prevalence) can help to aid in the selection of the best matched induction therapy, which has been shown to improve overall survival [13]. During treatment, NGS can be used to identify the emergence of resistance mutations, like EGFR T790M and MET amplification, that can develop or be selected for in tumor cells in response to EGFR tyrosine kinase inhibitors (TKIs) and render them ineffective [13]. Finally, NGS profiling can be used for MRD assays, which quantify tumor-specific mutations over time and can detect relapsing clones often months before clinical detection.

NGS techniques

After extraction of nucleic acid from specimens, NGS assays involve three universal steps: library preparation, sequencing, and data analysis. For targeted NGS assays, library preparation also involves target enrichment, in which specific genomic regions of interest are enriched thousands of fold above the entirety of the remaining genomic background to increase the accuracy, specificity, and efficiency of sequencing [14]. Target enrichment is generally achieved through one of two main techniques: polymerase chain reaction (PCR)-based amplicon enrichment and hybrid capture enrichment. PCR-based amplicon enrichment utilizes multiple primer pairs, sometimes up to tens of thousands, to simultaneously amplify many regions of interest [14]. On the other hand, hybrid capture-based approaches utilize biotinylated oligonucleotide “probes” or “baits” to pull specific regions of interest out of solution while the remaining genomic background is washed away.

While both techniques generally achieve the same end goal, there are several similarities and differences. Hybrid capture target enrichment, for example, requires a relatively high amount of nucleic acid starting material and tends to have a slightly longer workflow due to the time required for hybridization (Figure 2). In addition, nucleic acids must be fragmented at the start of the assay, which may require purchasing additional instrumentation such as an ultrasonicator, although enzymatic-based methods may be used instead. While hybrid capture has some drawbacks, there is virtually no limitation on the number of gene targets that can be included in the panel, and there is a higher level of uniformity across targets, leading to better sequencing performance. Conversely, PCR-based amplicon enrichment requires a relatively low amount of input, which benefits small biopsy specimens in which little nucleic acid can be extracted, and it has a lower rate of off-target sequencing (Figure 2). However, the number of targets covered by these assays are typically smaller as they are limited by the number of primer pairs that can be multiplexed. They also have less coverage uniformity and are more susceptible to the introduction of PCR-based artifacts [15].

Figure 2:

Comparison of different target enrichment methods.

NGS can be used to detect a variety of different variant types including SNVs, indels, CNVs, and gene fusions. Oncogenic gene fusions frequently involve tyrosine kinases such as ALK and NTRK and can be found in up to 17% of all solid tumors [16, 17]. While DNA is often used as the starting material for the detection of many different variant types, there are several drawbacks when it comes to fusion detection, especially when the fusion breakpoint occurs across large or repetitive intronic regions and when the tumor percentage in the specimen is low. To overcome these shortcomings, RNA-based NGS assays can be used with greater chances of success as RNA transcripts do not contain intronic sequences that can require large chromosomal coverage or repetitive sequences that poorly align to the reference genome. RNA can also potentially increase the sensitivity for specimens with low tumor purity as highly expressed fusion transcripts can compensate for low VAF and make these fusions more likely to be detected [18]. Thus, clinical NGS assays need to be carefully designed to meet their intended use and maximize the overall sensitivity and specificity.

Broad genomic profiling

When designing an NGS panel, content, depth, and cost are related elements in sequencing that must be taken into account. NGS assays can range in content from small panels that cover specific “hotspot” positions across several genes or large panels that cover the entire coding regions of several hundred genes to the whole genome or exome. As more genomic regions are sequenced (content), either the depth of sequencing coverage decreases (sensitivity) or more sequencing is required (cost). Because of the heterogeneous nature of tumor specimens resulting from the presence of non-tumor cells and tumor subclones, somatic variants may be present at low VAFs. To confidently detect these lower frequency variants, higher depths of coverage are needed, which can be estimated based on the desired lower limit of detection (LLOD), sequencing error rate, and tolerance for false negative and false positive results [19]. Typically, depths of coverage between 250X–500X are sufficient to detect somatic variants in tumors with LLODs near 5% VAF, but samples with low tumor purity may require up to 1,000X coverage [19].

There are several advantages to broad genomic profiling that covers larger genomic content. For example, it can help clinical laboratories to prepare for the inevitable identification of future biomarkers that are relevant for oncology. While Memorial Sloan Kettering Cancer Center (MSKCC)’s OncoKB database currently reports 147 clinical implications across 43 genes and 2 mutational signatures for which there are Food and Drug Administration (FDA)-approved targeted therapies, numerous clinical trials are underway to expand that number [20]. In fact, between 2009 and 2020, the FDA approved 46 new anticancer drugs based on novel biological targets, leading to an average introduction of 3–4 new clinically relevant therapeutic biomarkers per year [21]. This figure is higher when the rate of new diagnostic and prognostic biomarkers is factored in. Validating a broad range of molecular targets for NGS assays both enables laboratories to stay on top of the continually evolving genomic environment in oncology and identify patients eligible for new clinical trials when other treatment options are exhausted. Laboratories can also choose to initially report a subset of the total genes being sequenced and then unmask additional targets as they become clinically meaningful. This strategy allows for the rapid introduction of new sequencing targets without having to validate a new assay from scratch.

Incorporating a broad range of sequencing targets in an NGS panel has additional benefits including the ability to detect complex biomarkers such as mutational signatures, Microsatellite Instability (MSI), and Tumor Mutational Burden (TMB). Mutational signatures are collections of somatic mutations in cancer genomes that are caused by multiple exogenous and endogenous mutational processes. They can be used to determine the likely etiology of the disease such as exposure to ultraviolet (UV) light or tobacco smoke or defects in DNA mismatch repair (dMMR) [22]. In some cases, they can also help to identify the origin of some carcinomas of unknown primary origin (CUPs). Because these signatures can include multiple variant types across many genes, broad genomic profiling is required for their identification. In addition, MSI and TMB are emerging as predictive and prognostic biomarkers for immunotherapy response in cancer patients. While MSI assays detect expansions and/or contractions found in microsatellite regions caused by somatic mutations in dMMR proteins and TMB assays calculate the number of somatic coding mutations per megabase (muts/Mb) of the cancer genome, both are used to predict the neo-antigen load of tumors. MSI has traditionally been tested using PCR- or immunohistochemistry (IHC)-based assays; however, NGS-based testing has multiple advantages over these techniques including the detection of inactivating mutations in dMMR genes and the interrogation of thousands of microsatellite loci, both of which increase the confidence in MSI detection [23, 24]. Although the gold standard for TMB calculations is using whole exome sequencing (WES), others have shown it can be accurately calculated using a 315-gene panel that covers ∼1.1 Mb of the genome [25]. Smaller gene panels covering fewer coding regions do not provide enough data to accurately estimate TMB, and thus broader sequencing is required. Despite the many benefits of using broad genomic profiling in oncology, it is not without its limitations.

Rapid genomic profiling

The average TAT for broad genomic profiling using hybrid capture-based NGS testing can range from 2 to 3 weeks [26, 27]; however, some clinical scenarios necessitate faster results to quickly implement treatment decisions that will most benefit the patient. One such situation stems from a recent FDA approval for the use of the immunotherapeutic nivolumab in combination with platinum-doublet chemotherapy for adult patients with resectable NSCLC in the neoadjuvant setting [28]. The caveat with this treatment is that patients with actionable EGFR or ALK variants are poor responders to PD-L1 inhibitors and subsequent treatment with targeted kinase inhibitors increases their risk of developing serious adverse events including interstitial pneumonitis, immune related adverse events (irAEs), prolonged QT intervals, hepatotoxicity, and/or hematotoxicity [28], [29], [30], [31]. Because it is important to start this treatment regimen as quickly as possible, a rapid TAT test that can detect multiple variants across multiple genes is required.

Other situations requiring rapid genomic profiling are associated with acute myeloid leukemia (AML) and metastatic colorectal cancer (mCRC). AML is an aggressive and genetically heterogeneous disease that can be driven by a number of different and potentially targetable variants in genes such as FLT3, IDH1, and IDH2, yet despite this diversity, patients with AML have traditionally been treated with a “one size fits all” chemotherapy approach [32]. Recently there have been a number of FDA-approved targeted therapies for patients with AML that have shown to improve overall survival compared to traditional chemotherapy regimens alone or placebo [33], [34], [35]. Because of the aggressiveness of the disease and the benefits of targeted therapy, it is therefore important to quickly determine if patients with AML have targetable variants that can be treated with precision medicine. Likewise, starting patients with mCRC on appropriate anti-cancer therapy is also time sensitive. The current standard of care for patients requiring rapid cytoreduction is treatment with anti-EGFR monoclonal antibodies (mAb) like cetuximab; however, numerous studies have shown that patients with activating driver mutations in KRAS and/or NRAS do not respond to this treatment and can even be harmed when combined with an oxaliplatin-based cytotoxic backbone [36]. According to professional guidelines from ESMO, at a minimum, tumor specimens from patients with mCRC should be sequenced to examine the presence of variants in exons 2, 3, and 4 in both KRAS and NRAS, although the full coding sequencing should be analyzed when possible [36]. In addition, these guidelines suggest that the TAT for RAS testing should be ≤7 working days from the time of specimen receipt to the time of issuing the final report, which requires rapid genomic profiling [36].

To decrease the TAT of genomic profiling, the breadth of genomic coverage included in the panel must be more limited to reduce the time associated with bioinformatic processing and the number of regions to review before case sign-out. This coverage can be limited by both reducing the number of genes sequenced and/or by reducing coverage to specific “hotspot” variants that are most frequently found in cancers. PCR-based amplicon methods for target enrichment are often also used as they do not require an extra day for the hybridization of the oligonucleotide probes that is necessary for hybrid capture-based methods. Because the breadth of coverage is more limited and PCR-based amplicon enrichment is often used, these rapid TAT tests usually require significantly less sample input, which enables sequencing of limited samples that would otherwise be put on hold until more material could be obtained. They can also tolerate poorer quality samples, such as formalin-fixed paraffin-embedded (FFPE) tissue that often have higher levels of DNA degradation. Lastly, additional steps may be taken to reduce the TAT of these assays such as automation of nucleic acid extraction and PCR plate setup with the use of liquid handlers or automation of the entire wet-lab process, including extraction, library preparation, and sequencing, using fully automated, integrated sequencing instruments.

Summary

Cancers are driven by genetic and genomic changes that dysregulate normal cellular functions. These variants are clinically important as they can be of therapeutic, prognostic (associated with risk assessment), and/or diagnostic value when a cancer is first detected as well as over the course of a patient’s treatment. As neoplasms can be driven by a large and diverse range of genetic abnormalities, NGS has become a mainstay of molecular testing for oncology as it can simultaneously interrogate numerous loci across many genes from a single patient sample. Depending on the intended use, clinical NGS assays can be designed for broad or rapid genomic profiling. While broad genomic profiling can be more informative and lend itself to the rapid adaptation of reporting new targeted regions as they become clinically important, its overall TAT can be too long for certain clinical scenarios. On the other hand, assays designed for rapid genomic profiling can produce results in less than one week but may be more limited in the number of genomic regions they cover, may not detect complex genomic events, and are typically more difficult to expand panel content as new targets emerge. Ultimately, in order to meet the clinical needs of all oncology patients, the ideal clinical diagnostic laboratory should have a variety of tests tailored to the clinical indication, that include RNA-based NGS to accurately detect fusion variants, DNA-based broad NGS to thoroughly characterize the tumor mutational landscape, and DNA-based rapid NGS to quickly identify actionable variants for time sensitive cases (Figure 3).

Figure 3:

Algorithm for clinically appropriate variant detection strategy.

Outlook

Although the clinical utility of NGS testing continues to expand, several hurdles prevent it from becoming routine across all clinics. For example, the cost and complexity of NGS testing is difficult for most laboratories to internally implement due to appropriately trained personnel and cost. NGS requires expensive equipment including electrophoretic and quantitative instruments to analyze sample quality, ultrasonicators to shear DNA, thermocyclers to amplify DNA, and of course the sequencing instrumentation itself. In addition, novel protocols and workflows are required to execute and coordinate nucleic acid extraction, library preparation, hybrid capture (if using for target enrichment), and sequencing. Furthermore, specialized knowledge of bioinformatics is necessary to setup pipelines that can analyze the enormous amount of data generated from sequencing runs.

Continued technological advancements such as the introduction of all-in-one, automated sequencers are lowering these barriers, making it easier for even those with no NGS experience to successfully run these assays and report results. The expense of sequencing has also been steadily declining over the past decade with it now costing on average ∼$500 to sequence an entire genome compared to ∼$6,500 just 10 years ago [37]. It should be noted, however, that the exact cost is dependent on a number of variables including testing volume, application (i.e. research vs. clinical), and the breadth of genome coverage [38]. With the continued advancements in precision medicine, technology, and sequencing costs, NGS testing is poised to be a routine order for all clinicians in the near future.