Precision medicine is a clinical approach that tailors treatment to a patient’s specific form of disease. Molecular testing, such as genomic or proteomic analysis, often informs the treatment approach by enabling physicians to better understand a patient’s condition and best select a treatment to which the patient’s disease is most likely to respond.
Historically, patients presenting with the same condition have been given the same treatment, a one-size-fits-all type of clinical approach. While the treatment may work for some patients and their condition improves, other patients may show limited, if any, improvement or their condition may even worsen. Precision medicine, on the other hand, leverages data specific to a given patient, leading to treatment selection likely to be most effective for that patient.
The promise of precision medicine is to improve patient care by tailoring treatments to address a patient’s specific form of disease. In this way, the correct treatment can be more efficiently identified while avoiding the side effects resulting from treatments that may be of limited or no benefit to the patient. If successful, precision medicine will produce a higher success rate in treating diseases like cancer than traditional standard of care.
While the terms “precision medicine” and “personalized medicine” are often used interchangeably, they have different connotations. Personalized medicine implies that a treatment regimen is devised specifically for one individual, whereas precision medicine assigns, or classifies, an individual to a group with which the individual’s disease shares common traits. By extension, if an individual and group share common traits, the individual will likely benefit from a similar therapeutic approach to that of the group.
Precision medicine has become increasingly important in oncology. A variety of treatment options exist for different cancers and knowing more about the biochemical makeup of an individual’s tumor can help determine which treatment is likely to be most effective. This is particularly important, for example, in breast cancer where hormone receptor status is a key factor in treatment responsiveness. For instance, triple negative cancers are more likely to show a pathologic complete response to chemoradiation therapy than ER positive/HER2 negative tumors. Collectively, the interrogation of biomolecules present in a patient's tumor is known as molecular profiling.
Molecular profiling informs treatment decisions based on the biomolecules prevalent in a patient’s cancer. Traditionally, molecular profiling has focused on genes, but more recently, a higher emphasis is being placed on proteins and metabolites, such as lipids. The reason for this shift in focus is that proteins and metabolites represent the active state of cells of interest, and expression of these biomolecules can rapidly change. Alternatively, genes indicate the propensity for a condition but do not directly provide phenotypic information.
By understanding the active state of the cancer, physicians can better select a treatment that targets activated pathways. This ability to efficiently identify the treatment that will most effectively benefit the patient can significantly improve patient outcomes while, at the same time, reduce the unnecessary time, cost, and physical and emotional stress incurred due to an iterative clinical approach.
Molecular profiling tests are typically used as ancillary tests to provide a physician with additional information after an initial, or suspected, diagnosis of disease. Molecular profiling can aid in disease risk assessment, prevention, and screening; difficult case disease diagnosis; determination of disease subtype, stage, and grade; and prediction of disease treatment responsiveness, progression, and recurrence.
Molecular profiling tests are frequently used in oncology for screening as well as for improving diagnosis and determining the best treatment option. Two of the best known oncology screening tests are BRCA1 and BRCA2 mutations for breast and ovarian cancer and the PSA test for prostate cancer, genetic and proteomic tests, respectively. These tests are used to help determine those patients who may want to consider prophylactic measures or need to be monitored more closely for the potential development of these cancers. Molecular profiling tests are also used to subtype breast cancers to determine hormonal status, which helps to inform the physician as to what type of treatment is likely to be most effective for the patient. Molecular profiling tests are likewise used to aid in cancer diagnosis for histologically challenging cases, such as indeterminant cutaneous melanocytic lesions. Additionally, molecular profiling tests are increasingly used for prognostic determinations, such as predicting likely response to chemotherapy, radiation, or immunotherapy.
The development of companion diagnostics, which link a specific biomarker to a specific drug to predict its efficacy, has increased significantly within the oncology space. These tests look for the presence or absence of a particular gene or protein that is used to determine if a certain drug is likely to be effective against a patient’s cancer. Unlike molecular profiling tests that interrogate many biomolecules for diagnostic or prognostic analytical purposes, companion diagnostics typically look at only one or a handful of biomarkers that are linked directly to a specific drug treatment. While companion diagnostics can be highly useful for their specialized purpose, they do not provide wide applicability to disease management.
Diagnostic tests are used by physicians to help determine if a patient has a particular disease or a specific disease subtype. Such tests may supplement clinical findings by aiding in the determination, for example, of HER2 status in breast cancer or whether a prostate cancer is hormone refractory. Prognostic tests, on the other hand, serve to help determine whether a particular patient outcome is likely. These tests may be used to aid physicians in determining, for example, if a patient’s cancer is likely to recur after surgery, thereby necessitating adjuvant therapy, or if a patient is likely to show a pathologic complete response to neoadjuvant chemoradiotherapy, thereby indicating that the patient can be placed on an active surveillance plan as opposed to undergoing an unnecessary surgical procedure.
Following a similar trend to the evolution from genomics to proteomics in basic research, the majority of early molecular profiling tests focused on gene analysis. The early molecular profiling tests that analyzed proteins considered only one or a few specific targeted proteins as opposed to developing a broader molecular fingerprint.
Immunohistochemistry was first introduced in the 1940s for the detection and visualization of antigens in tissue sections. This technique can be used to determine the presence, localization, and relative abundance of specific markers in processes such as cell proliferation and apoptosis in tumors.
IHC takes advantage of highly specific interactions between antibodies and antigens (proteins) in tissue sections. A solution containing an antibody conjugated to a reporter molecule is incubated with the tissue. Once the antibody is bound to the targeted antigen, and excess unbound antibody is washed away, the reporter molecule (typically a dye or a fluorophore) is released in situ and the staining is evaluated by a pathologist. Commonly used markers in oncology include ki67, which indicates cell proliferation, and cytokeratins for the identification of cancerous cells. More specialized applications include the determination of estrogen receptor and progesterone receptor status in breast and gynecologic cancers as well as differentiation of B-cell and T-cell lymphomas.
IHC can be quite useful as an aid in the diagnosis of cancer, particularly in cases where histological evaluation can be ambiguous. Antibodies can be highly specific for a target protein, allowing for confidence in the determination of its presence or absence in a specimen. The use of fluorescent probes for detection of antibody binding can allow for high sensitivity in analyte detection.
While IHC has proven highly effective over the past 70+ years, it is not without its limitations. Because IHC relies on the use of a target-specific antibody, the target protein must be known in order for IHC to be applicable. Additionally, because IHC is reliant on the use of antibodies, which are large biomolecules, as well as reporter dyes, the number of analytes that can be detected in a single tissue section is limited to only a handful as a result of steric hinderance and the number of distinct stains that can be visualized. Finally, the interpretation of IHC staining is subjective, resulting in different observers reporting different results.
Fluorescence in situ hybridization is a technique that uses fluorescently labeled probes to bind to specific sequences of DNA or RNA. It can be used to look for mutations or abnormalities to genes or known gene sequences that can be indicative of cancer or other disease. The probes must be long enough to precisely bind to the target of interest but short enough that they can reach their target within the tissue.
FISH involves the denaturization of DNA within a tissue section before the binding of an oligomer to the target region of interest. Visualization is achieved either through a fluorophore that is bound directly to the probe or through antibody or biotin tagging to the oligomer with a fluorescent reporter molecule. Localization is then determined through excitation of the dye, and visualization and imaging are performed using a fluorescence microscope.
Because FISH can be carried out directly on a tissue section, it has the benefit of providing localization information as the analysis is carried out in situ. Analysis directly from tissue sections minimizes the amount of material needed for interrogation. FISH is also faster than other gene-based techniques as no cell culturing or DNA amplification is carried out prior to analysis.
Like other tissue staining techniques, FISH requires a pathologist to interpret the results. These evaluations can be subjective with different pathologists providing different diagnoses. Similar to IHC, only a limited number of targets can be analyzed in a single tissue section due to limitations on the visualization of the fluorescent probes. Non-specific binding of FISH probes can also lead to a decreased sensitivity and specificity when using this technique. FISH requires that the target segment of the gene be known so that an adequate probe to target the gene segment can be created.
Comparative genomic hybridization works on a similar principle to FISH but compares a suspicious specimen (e.g., tumor) to a normal, or reference, specimen to determine differences in the copy numbers of particular genes. These copy number differences, either increases or decreases, can be indicative of diseases such as cancer.
In the CGH technique, DNA is extracted from both normal and diseased tissue or biofluids. Each specimen is hybridized with probes labeled with distinct fluorophores. The normal and diseased specimens are mixed in equal amounts, and the relative levels of the fluorophores conjugated to the diseased specimen serve to provide information as to increases or decreases in copy numbers of specific genes as compared to the normal specimen. CGH can be carried out on whole specimens or in an array format (aCGH) where specific genes are bound to specific locations on a chip.
CGH can prove highly sensitive for detecting changes in chromosomal copy numbers due to its use of computer software in determining intensity differences between control and diseased specimens. In the aCGH approach, genome-wide analysis can be carried out on a chip, targeting copy number changes in 5-10 kilobase DNA segments.
Because DNA must be extracted from a specimen to carry out CGH analysis, information as to the cells of origin exhibiting the copy number changes is lost. This issue can be somewhat alleviated through the use of microdissection to target regions within a specimen, but this technique does not provide the spatial resolution achieved by tissue section-based approaches. CHG also suffers from the limitation that it can only detect changes in the copy number of genes, not their translocation, inversions, or mutations that can also be indicators of disease. Finally, CHG requires 0.5-1.0 µg of DNA to carry out analysis, which may be more material than is available from small diagnostic biopsies.
Gene sequencing can be carried out to determine if mutations are present that are indicative of cancer or a particular subtype of cancer. Traditional methods such as Sanger sequencing were slow, tedious, and expensive. Technological advancements in gene sequencing from the late 1990s through current day have allowed gene sequencing to be carried out much more rapidly, facilitating their use on timescales conducive to clinical analyses. Collectively, these advanced sequencing technologies are referred to as next generation sequencing and high throughput sequencing.
Most current DNA sequencing methods rely on the synthesis of DNA molecules to determine the sequence of DNA bases using a reporter. This may occur as an observable flash of light or a fluorescent reporter or through the release of detectable hydrogen through base incorporation. Monitoring these reactions and the reporters released can be translated into DNA sequences of varying lengths based on the technology being used. Other sequencing technologies include nanopore sequencing, tunneling current sequencing, sequencing by hybridization, and MALDI mass spectrometry. Current sequencing technology in the medical diagnostic field is often used in tandem with polymerase chain reaction (PCR) to amplify target genes prior to sequencing in order to characterize mutations. Through parallel processing, these newer technologies can allow for the determination of millions of nucleotides per second.
NGS and HTS allow for genetic sequence information to be generated at timescales and costs that are conducive to clinical applications. These techniques provide a highly detailed view of an individual’s genome including gene mutations, allowing for the rapid determination of the presence, and exact type, of disease and insight into the most appropriate treatment options.
While considerable advances have been made in the speed and accuracy of gene sequencing, there are still limitations to the information it can provide. All sequencing technologies require that the specimen be homogenized and the DNA or RNA be extracted, thus eliminating all spatial information from which the signal originated. Additionally, there is not necessarily a direct correlation between genotypic and phenotypic information in the cells. In other words, the presence of a gene does not necessarily mean that it is expressed.
Current molecular profiling tests employ more objective techniques to differentiate between disease states. These tests have moved away from the subjective evaluation of staining to instead rely on machine learning algorithms to better recognize the molecular signatures that differentiate cancer from normal specimens and determine which treatment is likely to be most effective for a particular cancer.
Gene expression profiling examines genes that are actively being expressed in cells, typically tens to thousands of genes, to determine a molecular profile that is consistent with a disease state or outcome.
GEP targets messenger RNA (mRNA), nucleic acids that are transcribed from actively expressed genes in the tissue of interest. The mRNA is transcribed to complementary DNA (cDNA) and amplified through reverse transcriptase polymerase chain reaction (RT-PCR). A fluorescent tag is typically incorporated during the process. Single strand DNA is immobilized on a chip and the amplified cDNA is incubated with it to allow binding to occur. After washing of unbound material, the fluorescence at each spot of the microarray is evaluated to determine the relative expression levels of the target genes of interest.
GEP has shown great improvement over traditional gene sequencing approaches. By targeting mRNA over DNA, this technique interrogates genes that are activated and relevant to the target tissue of interest. By further limiting the analyzed genes to those that have been shown to be relevant in differentiating disease states or outcomes, GEP is better able to correlate gene expression with clinical findings, helping to provide increased benefit to patients.
While targeting of mRNA provides a better picture of the cellular processes than does DNA, there is still not a direct correlation between mRNA and protein expression. Collectively, it is the proteins and their posttranslational modifications that carry out the functions of the cell. GEP, like other gene-based techniques, requires extraction of the genetic material from the specimen prior to analysis. While microdissection or macrodissection can be incorporated into the workflow to target areas enriched in tumor, the exact spatial localization is lost. Also, GEP can be vulnerable to contamination as inadvertently introduced DNA will be amplified along with the endogenous DNA in the specimen.
Emerging molecular profiling tests employ an approach called multi-analyte analysis with algorithm (MAAA) in which hundreds to thousands of analytes are measured and machine learning algorithms are used to differentiate between disease states. Additionally, many of these tests focus on the analysis of proteins or metabolites as opposed to genes. The basis for this shift in focus is that proteins and metabolites provide a snapshot of the active state of targeted cells, not just the genetic propensity for disease.
Histology guided mass spectrometry profiling represents a significant advance in molecular testing, capable of aiding physicians in both diagnostic and prognostic evaluations. HGMS profiling can be thought of as a specialized application of mass spectrometry imaging (MSI). This technology allows for the in situ analysis of biomolecules directly from tissue sections using histopathological evaluation to target precise tissue locations for data collection.
HGMS profiling uses thin tissue sections, similar to those used in traditional pathology applications. One tissue section is collected on an indium tin oxide (ITO) coated conductive glass slide that is compatible with a matrix assisted laser desorption/ionization (MALDI) time of flight (TOF) mass spectrometer, and a serial tissue section is collected on a standard microscope slide for histological staining. A digital microscopy image of the stained tissue section is generated and reviewed by a pathologist who places color coded annotations (typically 50-100 µm in diameter) on the digital image. The annotated microscopy image is merged with a digital image of the serial unstained tissue section on the ITO slide. Appropriate sample preparation for the biomolecular class of interest is carried out on the tissue section. This process may include deparaffinization, antigen retrieval, enzymatic digestion, and/or matrix application. The specimen slide is then loaded into the mass spectrometer, and the merged digital image is aligned to the specimen slide using fiducial markings. Data are collected only from the locations corresponding to the annotations designated by the pathologist. This approach allows for high throughput analysis of tissue specimens and more efficient statistical analysis as compared to traditional MSI due to the decreased data volume.
Data collected from HGMS profiling is subjected to machine learning to develop classification algorithms that can be used to distinguish between disease states, such as malignant versus benign or treatment responder versus non-responder. The use of artificial intelligence provides an objective classification as to which group a patient belongs. In order to develop this classification capability, data from specimens originating from patients with known clinical outcomes are used to train the computer to recognize the molecular fingerprint associated with each relevant outcome.
HGMS profiling offers a number of benefits as compared to other molecular profiling tests currently in use. Unlike CGH, GEP, and LC-MS (discussed below), the spatial localization of biomolecules is maintained throughout the process, allowing for the determination of the cells from which the observed signals originate. Further, HGMS profiling provides objective classification results based on machine learning algorithms, whereas IHC and FISH are dependent upon a pathologist’s subjective evaluation of staining. In addition, while FISH, CGH, and GEP evaluate genomic content, HGMS profiling examines proteomic content, providing a molecular snapshot of the active state of targeted cells. This technique comports with a pathologist’s typical section collection workflows and requires less tissue material (two, 5 µm thick tissue sections) than alternative techniques, particularly as compared to CGH. HGMS profiling represents an efficient and high throughput analytical process that offers physicians quicker test result turnaround (within a few business days) as compared to existing molecular profiling tests.
While HGMS profiling presents significant advantages over other technologies, it is also not without limitations. HGMS profiling can only be applied to tissue-based diseases, as it cannot be applied to non-invasively collected specimens, such as serum, plasma, or urine. In addition, current instrumentation and procedures limit the minimum size of areas targeted for HGMS profiling to approximately 50 µm. When evaluating some diseases, this target area may not be small enough to evaluate pure cell populations of interest, and other cell types and molecules expressed within the target area may contaminate the signal obtained.
Clinical mass spectrometry profiling can also be applied to biofluids and cytology specimens. For biofluids, this technique is most effective when the fluid being analyzed is directly related to the disease of interest. While fluids such as serum and plasma have been successfully used in some applications, they can pose challenges for detecting relevant low abundance proteins amidst a large dynamic range of “housekeeping” proteins. Biofluids, such as urine, cerebrospinal fluid, and saliva, among others, show greater applicability for clinical MS profiling as the relevant biomarkers are less likely to be subjected to dilution effects. Cytology specimens, such as cellular smears and fine needle aspirates, have also been successfully analyzed via clinical MS profiling. Biopsies using these techniques are sometimes preferred due to their less invasive nature.
Specimens analyzed by clinical MS profiling are generally subjected to a cleanup procedure to remove biological salts that can suppress protein signal. For biofluids, this preparation is typically performed using a ZipTip® or magnetic beads that have an appropriate stationary phase immobilized on them to allow binding of proteins. The stationary phase material can be washed to remove salts and then the proteins eluted using a suitable solvent. In addition to desalting the specimen, this procedure also serves to concentrate and normalize the protein levels across specimens. Next, the specimen is spotted to a MALDI target and mass spectra are collected.
Cytology specimens are generally collected into a liquid buffer or fixative before being cytospun into a thin layer of cells on a mass spectrometry compatible slide. Once adhered, the cells are washed with an alcohol solution to remove any residual salts from the buffer. Matrix is then applied in the same manner as for HGMS profiling and mass spectra are collected from the cells.
Machine learning is then applied to the collected biofluid or cytology data, similar to HGMS profiling, to determine a molecular fingerprint indicative of diagnosis or outcome.
Collection of biofluids and cytology specimens is less invasive than tissue biopsies, making them preferred specimen collection methods when possible. The specimen preparation process for biofluids is relatively simple and can be automated through the use of robotics. In addition, clinical MS profiling of biofluids tends to offer very high specimen throughput, with the ability to analyze hundreds of specimens on a single target, and to attain data collection speeds of less than a second per specimen. Clinical MS profiling of cytology specimens can also realize relatively high throughput as compared to tissue sections, largely because histological review and annotation prior to data collection is generally unnecessary for these specimens.
Clinical MS profiling is performed without extensive sample preparation, such as enzymatic digestion, allowing for the analysis of intact proteins from the specimen. However, most MALDI mass spectrometers display the greatest sensitivity for lower molecular weight proteins (<30 kDa), which limits the proteins that can be detected using this technique. In order to improve this technique’s chances of analytical success, clinical MS profiling should be applied to fluids or cells directly related to the disease (e.g., urine for bladder cancer and Pap smear cells for cervical cancer) where there is less of a dilution effect on the relevant molecules or cells in question. Given this, clinical MS profiling may not be relevant for all types of cancer.
Surgical mass spectrometry imaging technologies are gaining acceptance in the operating room. Both the iKnife and the MasSpec Pen are in trials for use during surgery to evaluate tumor margins in real time.
The iKnife is a modification to a traditional cauterizing knife that is routinely used in surgical procedures. When tissue is ablated, smoke is generated that can be transferred via tubing to a nearby mass spectrometer. The molecules present in the smoke are then analyzed, and a machine learning algorithm is applied that can differentiate between cancerous and normal tissue. The iKnife is being used in clinical trials in the UK to better determine breast tumor margins during surgery.
The MasSpec Pen uses a 3D printed device the size of a pen to deposit a droplet of sterile water onto the surface of tissue. Biomolecules are dissolved into the water droplet, which is then extracted from the tissue surface and transferred via tubing to a nearby mass spectrometer. The MasSpec Pen also uses machine learning to differentiate between cancerous and normal tissue. The MasSpec Pen is being evaluated for use in surgery to non-destructively determine tumor margins in various cancers. It has even been featured on the hit TV show Grey’s Anatomy.
These technologies allow for rapid determination of lipid and metabolite fingerprints with little or no sample preparation. Through the application of machine learning algorithms, these techniques can be used to determine if a tumor has been removed in its entirety while the patient is still in the operating room. Being able to determine if a patient has clean surgical margins in real time will help to prevent repeat surgeries (20-30% rate in breast cancer) due to insufficient margins being found upon histopathological review. It is likely that one or both of these technologies will become routinely used in surgical procedures within the coming years.
While surgical mass spectrometry imaging technologies show great promise for improved patient care in the operating room, they have limitations in what they can deliver. Each technique described is limited to the detection of small molecules, such as lipids and metabolites. These techniques are unable to detect and analyze larger molecules, such as proteins and genes. The iKnife requires that the molecules survive being vaporized and not be destroyed by the burning process that creates the smoke. The iKnife also causes damage to the tissue that it cuts, compromising the tissue quality for subsequent histopathological review. The MasSpec Pen can only detect molecules that are water soluble, limiting the number and types of molecules that can be detected.
These technologies, based on their current applications, require that the patient already be undergoing surgery for the treatment of his or her disease. While these technologies show great promise for helping to drive surgical decisions, they are not able to prevent unnecessary surgeries, nor can they predict patient response to less invasive treatments, such as chemotherapy, radiation, and immunotherapy. Additionally, in order to achieve widespread utility, these devices will be subject to regulatory approval processes involving clinical trials that can be lengthy and expensive.
Liquid chromatography-mass spectrometry has long been used in the analysis and quantitation of drugs and other small molecules (e.g., vitamins and metabolites) from tissue or blood specimens. This technology is also frequently used in the proteomic analysis (e.g., amyloidosis typing) of biological specimens, both in a qualitative and quantitative manner. Due to its wide-ranging capabilities, LC-MS is used in numerous clinical testing applications.
LC-MS couples two very powerful analytical techniques, liquid chromatography and mass spectrometry. The specimen to be analyzed must be in a liquid form, either naturally (serum, plasma, urine) or through homogenization (tissue). The specimen is then passed through a column packed with a stationary phase that allows molecules to be separated based on their chemical properties, such as hydrophobicity and size. Additionally, the specimen is pre-concentrated. In other words, the concentration of a biomolecule analyte is higher in the eluant volume compared to within the original specimen volume. The eluant is then directly transferred to a mass spectrometer where the separated molecules are detected and identified via fragmentation.
Because of the separation and enrichment that occurs through the chromatographic process and the ability to target specific molecular weights in the mass spectrometer, LC-MS can be highly sensitive, detecting very low levels of analytes in biological specimens. The ability to use isotopically labeled internal standards also allows for highly accurate quantitation of target analytes.
The biggest limitation of LC-MS technology is that the specimen must be in a liquid phase for analysis. In the case of tissue specimens, the specimen must be homogenized, resulting in all spatial information related to expression of specific biomolecules being lost. Simply put, this means that the cells of signal origin cannot be determined. Additionally, biomolecules that are highly localized to specific cells or structures may not be detected due to dilution effects during homogenization.
Precision medicine has altered the historical standard of care, advancing clinical decision making from a one-size-fits-all approach to one where the molecular profile of a patient’s tumor is frequently taken into consideration when determining the best course of treatment. In much the same way, molecular profiling has evolved from earlier subjective, gene focused tests that provide more limited insight to emerging objective, protein and metabolite focused tests that provide a broader molecular snapshot of a patient’s tumor along with the phenotypic information that is critical to delivering on the promise of precision medicine.
Delivering on the promise of precision medicine will involve the aggregation of significant amounts of patient specific data as well as the ability to harness this information in ways that improve clinical decision making and thereby improve patient care. The benefits of executing on this promise are substantial, including improvements in patient screening, risk assessment, and diagnosis as well as disease treatment responsiveness, progression, and recurrence. At the same time, executing on this promise also reduces the precious time wasted by the patient when trying ineffective treatment options and the corresponding financial cost and physical and emotional stress incurred. The impact of precision medicine will continue to evolve and will become increasingly relevant in clinical applications beyond oncology.
While molecular profiling will not single handedly deliver the patient specific data necessary to deliver on the promise of precision medicine, molecular tests will play a significant role in this effort. The evolution of molecular profiling will serve to improve the type and quality of information provided to physicians, positively impacting their treatment decisions. The trend in molecular profiling is clear – moving towards objective, targeted classification algorithms that provide useful phenotypic information by focusing on biomolecules beyond the genome. In this way, these tests will increasingly provide a holistic molecular fingerprint of a patient’s disease state. As molecular profiling further differentiates and segments a disease population, physicians will increasingly be able to assign a patient to a group with whom the patient shares an ever-greater number of common traits, thereby further refining the selection of treatment options. Given this, molecular profiling is likely to become standard of care as an aid in the diagnosis and treatment of cancer.