Biologic Frontiers in Multiple Myeloma: From Biomarker Identification to Clinical Practice

Multiple myeloma is a hematologic disease that is characterized by the proliferation of abnormal bone marrow plasma cells (BMPC) and immunoglobulin or light chain overproduction, with evidence of end-organ damage. Before 1997, the median survival of patients with newly diagnosed multiple myeloma was approximately 2.5 years (1). From 1997 to 2006, the use of high-dose anti–multiple myeloma therapy, stem cell transplantation, and novel agents increased the median overall survival (OS) to nearly 4 years, a 50% improvement (1). In some patients who are currently receiving such agents, improved relapse-free survival has increased to >10 years (2), raising the potential for long-term disease control and care.

Although the multiple myeloma treatment landscape has improved since the mid-1990s, the intraclonal heterogeneity (ICH) of malignant plasma cells (PC), the interaction of PCs with host factors, and the bone marrow microenvironment contribute to disease progression and the molecular evolution of the disease according to Darwinian principles (3, 4). Taken together, these factors contribute to the generation of treatment-resistant PCs, which ultimately leads to relapsed and refractory myeloma, the disease stage that is frequently used for drug development. However, as the nature of this disease stage continues to change as a consequence of patients receiving multiple lines of therapy, it is essential that we develop novel approaches to drug development.

If patients are to overcome treatment resistance, one approach that may improve clinical outcomes is to select treatments that are well matched to a patient's molecularly defined disease subtype (5). Another promising approach is to identify new biomarkers and tools that can identify onset or worsening of disease. According to the National Cancer Institute (NCI), a biomarker is a molecule that is found in blood and other body fluids or tissues that can serve as an indicator for a normal or an abnormal process, condition, or disease (6). Herein, we review biomarkers that have been used in multiple myeloma that meet the NCI's definition of a biomarker [e.g., monoclonal (M) protein]; we will also discuss molecular markers and tools that have the potential to dramatically alter how clinicians diagnose, stage, and treat patients with multiple myeloma or asymptomatic “early” multiple myeloma.

In the 1960s, multiple myeloma researchers began to identify biomarkers that were independent predictors of survival, including hemoglobin, serum calcium, serum creatinine, and bone lesion severity (7, 8). In 1975, Durie and Salmon (9) introduced a staging system that used M protein, hemoglobin, calcium, and a number of bone lesions to predict multiple myeloma cell tumor burden. In the 1980s, serum β₂-microglobulin (β₂M) was found to be a simple, yet reliable prognostic marker for disease staging; serum β₂M also enabled clinicians to predict a patient's likelihood of survival (10, 11). Subsequently, albumin (12), C-reactive protein (13), and BMPC proliferation indices (14, 15) were found to be reliable prognostic factors, but they did not come into general use. In 2005, an international consortium of researchers used serum β₂M and serum albumin to create the International Staging System (ISS), which enabled clinicians to stage patients and predict their long-term prognosis (16). This has been further refined by combining FISH data with ISS (17, 18).

Other types of biomarkers [e.g., serum free light chain (sFLC) ratio and cytogenetic markers] are also beginning to provide prognostic information on multiple myeloma and patients with asymptomatic disease (18–20). There have been significant developments in the diagnostic platforms that are available to assess the molecular features of multiple myeloma that provide extensive novel information as well as new diagnostic tests. The current emphasis is on using multiparallel genome-wide approaches, which yield huge amounts of data and offer the advantage of being able to multiplex genomes from many patients in a single run, which can reduce the cost per test. However, these technologies also raise significant issues with regard to the generation of complex datasets and how they can be reliably analyzed. These technologies will need to be robust and applicable if they are to be widely used in a routine setting. It is also appropriate to ask whether we should take a genome-wide approach to biomarker development or whether we should focus on a smaller number of “multiple myeloma relevant” markers that can be analyzed using simpler and potentially more robust techniques.

Changes at the DNA level determine how a cancer cell behaves. If we can understand these molecular events, we should be able to predict the behavior of cancer. Chromosomal translocations are primary genetic events that occur early in the disease and are seen in roughly 50% of patients with multiple myeloma. As a consequence of the translocation events, there is overexpression of partner genes, leading to abnormal clonal PC behavior and differing clinical outcomes (17, 21–27). For example, 4-year progression-free survival (PFS) and OS rates for patients with the t(4;14) have been estimated to be 11% and 35% versus PFS and OS rates of 32% and 60% for patients without the t(4;14), respectively (18).

Other chromosomal translocations have been investigated for their prognostic potential. The t(14;16) is present in <5% of patients with newly diagnosed multiple myeloma and defines a group of patients who often have poor survival (17, 18, 28, 29). Although the t(14;16) is relatively rare, the MAF gene is overexpressed in up to 30% of patients with multiple myeloma, defining a greater number of patients with a unifying biologic feature (4). The t(11;14) is observed in approximately 20% of patients and is associated with a favorable prognosis, higher rates of CD20 expression, lymphoplasmacytic morphology, hyposecretory disease, λ-light chain usage, and nuclear cyclin D1 (CCND1) expression and dysregulation (18, 21, 23, 30–32).

Hyperdiploidy of the odd-numbered chromosomes is the other major primary genetic event in multiple myeloma that defines the other 50% of multiple myeloma cases (Table 1; refs. 4, 33–36). Although the exact cause of hyperdiploidy is unknown, it seems to define a group of patients with a more favorable prognosis (37). In addition to chromosomal translocations and hyperdiploidy, inherited genetic variation is an important factor in the etiology of multiple myeloma. Molecular epidemiology techniques have shown that multiple myeloma is associated with single-nucleotide polymorphisms (SNP) at the chromosomal regions 2p, 3p, 6p, 7p, 17p, and 11q (4, 34, 35). However, at this point, no prognostic value has been seen in association with these variables. Inherited variation can also affect response to chemotherapy and the side-effect profile of a drug. To date, very few studies have been carried out to test this hypothesis, and they have not as yet defined relevant markers that can be used in this way in multiple myeloma (38).

Secondary genetic events, which are important in disease progression, also affect various molecular pathways and contribute to the biologic heterogeneity of PCs (Table 2). In this context, prognostically important copy number abnormalities (including gains or deletions of whole chromosomes or interstitial copy number gains) have been identified using SNP mapping arrays (4, 18, 33, 39). Clinically relevant regions that have been identified using this approach include 1q+, 1p32−, 1p11−, and 17p−. The International Myeloma Working Group (IMWG) has recommended assessing these specific cytogenetic lesions at disease presentation, as they may have a prognostic value that is clinically relevant. The study of copy number change has also improved our understanding of the biology of multiple myeloma, leading to the identification of 16q, a WW-domain–containing oxidoreductase gene (WWOX), CYLD (a negative regulator of the NF-κβ pathway), 11q− (BIRC2 and BIRC3), and 14q (TRAF3) as being relevant recurrent events (40, 41). These findings reinforce the importance of using therapeutic approaches to target the NF-κβ pathway.

The use of massively parallel sequencing has identified hundreds of mutations in multiple myeloma (42, 43); however, no consistently mutated gene has been identified that characterizes all occurrences of the disease. Nevertheless, these studies are in their infancy; the prognostic relevance of the multitude of exonic mutations that have been identified has yet to be determined. Although potentially useful, the challenge with interpreting sequencing data is defining which mutations are so-called “driver variants” that are important to disease pathogenesis rather than being “passenger variants” that are present by virtue of being carried forward by their association with other pathogenetically important genes. Driver mutations identified by this approach may be important therapeutic targets that could be used to design targeted treatment approaches.

If we are to use tumor-acquired genetic variants as biomarkers, it is important to consider their specificity and sensitivity for identifying the clinical outcome of interest (e.g., PFS, OS, or side effects). There are now considerable data on the use of “FISH-detected” cytogenetic markers to predict clinical outcomes. What is clear from these data is that no single marker offers either good sensitivity or specificity for the prediction of either PFS or OS. If we are to alter the treatment following the use of a FISH-based cytogenetic approach, we will need to use a comprehensive panel of markers to define risk based on both the number and nature of the adverse markers detected. Interphase FISH (iFISH) combined with the ISS is useful, but general applicability and issues with sensitivity and specificity remain. Some of these challenges can be overcome by counting the number of abnormal lesions that are present and using the number present as a way of defining risk. This approach, combined with the use of a comprehensive panel of markers, including the adverse translocations t(4;14) and MAF together with the adverse copy number variables 1q+, 1p−, and 17p−, can provide very complete clinical data.

Global gene expression profiling (GEP) and prognostic signatures provide an alternative approach to FISH (44). Although GEP and prognostic signatures offer greater specificity in patients who are identified as having a poor prognosis, they lack biologic relevance and remain difficult to apply in a routine clinical setting. However, GEP does provide a massive amount of data and can define the groups identified in the translocation cyclin D classification (TC) as well as detect prognostic signatures. At this time, GEP cannot usefully identify the adverse groups defined by 1q+ and 17p−; these two subgroups with a poor prognosis still need to be detected by iFISH. The eventual clinical application and regular use of such biomarkers depends upon their clinical applicability, together with the availability of effective treatments for use in specific patient subgroups.

The technology used to detect a cytogenetic biomarker is also of critical importance. Historically, metaphase cytogenetic analysis was used, but this only gave results in 18% of patients and was therefore not widely adopted. Although FISH on CD138-selected PCs is applicable and gives results in nearly 100% of patients if adequate numbers of tumor cells are obtained, this method is slow and expensive, and up to now has lacked sensitivity and specificity. For predicting clinical outcomes, new genome-wide technologies are useful, yet their design and application are still in their infancy. Alternative strategies that may be more applicable for detecting copy number changes include multiplex ligation–dependent probe amplification (MLPA), which can detect the clinically relevant multiple myeloma–associated copy number variants at a fraction of the cost of FISH. MLPA can also deliver results in a more timely fashion, whereas GEP can define risk status based on poor prognosis signatures. An alternative expression-based approach is to design quantitative real-time PCR (qRT-PCR) assays that can identify TC groups and prognostically relevant genes (45). If these assays are combined with MLPA copy number variant assays, this approach can provide all of the relevant prognostic variants in a set of tests that are readily applicable.

MicroRNAs (miRNA), which are more stable than conventional RNA, can be detected in serum as well as in malignant PCs (46). In this context, most work on multiple myeloma biomarkers has focused on BMPCs; little work has been done on circulating PCs or serum DNA or RNA. These areas offer considerable opportunities for future development (3, 47). In addition, important epigenetic events—including global DNA hypomethylation and gene-specific DNA hypermethylation—are beginning to provide insights into the etiology of disease progression and could lead to the identification of clinically useful biomarkers (4, 48, 49).

The development of clinically relevant biomarkers is now being shaped by the regulatory framework for such testing. The Clinical Laboratory Improvement Amendments (CLIA) approach suggests that the development of biomarkers needs to be done in defined laboratories that work according to relevant standard operating procedures. This implies that all biomarkers need to be developed in the context of clinical trials in which data are collected according to the principles of good clinical practice, and the biomarkers are determined in a laboratory working as per CLIA standards. In the absence of such an approach, biomarkers cannot be adopted into clinical practice.

The identification of patient-specific cytogenetic abnormalities and patients' gene mutation status may be particularly useful in helping to direct treatment in patients with multiple myeloma who possess a specific oncogenomic profile (5). It is now possible to define low-risk and high-risk disease subsets using GEP and FISH (50). GEP is useful in defining the molecular subtypes of multiple myeloma as well as being able to define signatures of high-risk disease (50–53).

On the basis of data analyses from platforms that are derived from completed studies, we are making progress toward improving the outcomes of patients with standard-risk disease. Nevertheless, we have made little progress toward treating or managing patients with high-risk disease. These observations argue in favor of moving toward the design of clinical trials that specifically develop treatments for high-risk disease. However, before we can design such studies, the biomarkers for robustly identifying these risk groups need to be validated in order for us to screen an adequate number of patients in the molecular subgroup of interest. This clinical trial approach will require us to fully embrace modern genomic technologies that are applicable in routine diagnostic facilities.

Changes in BMPCs, paraprotein, and light chains form the basis of assessing patient response to therapy and could be used as an alternative treatment endpoint. The standard approach to using these data is laid out in the IMWG response criteria (54). The achievement of complete response (CR) is an important endpoint when assessing new therapies (Table 3); it seems likely that the deeper the level of response, the better the clinical outcome will be. Indeed, the achievement of a CR with no detectable disease is the essential prerequisite for a cure. Thus, the definition of stringent CR (sCR) and the criteria defining this state are of some importance. The IMWG defines sCR similarly to CR, but patients with multiple myeloma must also have a normal FLC ratio with no evidence of clonal BMPCs (assessed using immunohistochemistry or two- to four-color flow cytometry; ref. 54). A consensus is beginning to emerge on the most appropriate panel of antibodies and the strategy for detecting malignant PCs, with considerable enthusiasm being directed toward their use as a definitive clinical endpoint. Flow cytometry has considerable merit for monitoring residual disease because it offers sensitivity and does not require one to design patient-/clonospecific approaches; it can also be performed in most hematooncology diagnostic laboratories.

A recent survey that included 30 major medical institutions in the United States found that the application of multicolor flow cytometry for minimal residual disease (MRD) in multiple myeloma varies greatly (55). Indeed, the definition of abnormal PCs differed substantially among institutions, with some relying on CD19 and CD45 negativity with or without CD56 positivity to determine the extent of MRD, despite previous studies showing that normal PC subpopulations can be negative for CD19 and CD45 or CD56 positive (56, 57). More specific antigens such as CD27, CD81, and CD117 were used by less than half of the institutions. Even more importantly, there is considerable variation in the number of BMPCs analyzed (i.e., events) and the number of abnormal PCs that are needed to determine the presence of MRD, which affects the maximum possible sensitivity of the assay. In this survey, it was found that the maximum detection sensitivity ranged from 0.0005% to 0.02%, a 100-fold difference in sensitivity. Such data copy the establishment of standardized approaches and external quality control programs. It is important to contrast this with multiparameter flow cytometry with the application of clonospecific PCR for the same purpose. In most settings, the clonospecific PCR offers greater sensitivity, but this technique has been hampered by the necessity to develop clonospecific probes, which has made its general application cumbersome and expensive. This situation is now changing with the advent of massively parallel sequencing approaches, which use probes that can be used to tile entire immunoglobulin heavy chain gene locus (IgH) regions (43). This approach is sensitive, high throughput, and universally applicable.

In recent years, it has become evident that patients with the t(4;14) have more favorable outcomes when they are treated with proteasome inhibitors. A group at the Mayo Clinic recently discussed the concept of “risk-adapted therapy” (5), whereby patients with high-risk disease [i.e., patients with the t(14;16), t(14;20), and/or del17p] may be better candidates for triplet combination therapy (e.g., bortezomib–lenalidomide–dexamethasone) compared with patients with intermediate-risk or standard-risk disease. Although triplet combinations may have a less favorable side-effect profile relative to single-agent therapies, triplet combination regimens are more appropriate for patients with high-risk disease because median OS in that cohort is only 3 years compared with 4 to 5 years and 8 to 10 years for patients with intermediate-risk and standard-risk disease, respectively. In the absence of using cytogenetic markers to risk-stratify transplant-eligible and transplant-ineligible patients, it would be difficult to determine which treatment regimen is most appropriate for patients with multiple myeloma.

Nevertheless, the t(4;14) cannot be considered to be a true prognostic biomarker because many patients who lack this translocation also respond and have good clinical outcomes following proteasome inhibitor therapy. Moving forward, information from genome-wide sequencing studies is leading to the identification of frequent pathogenetically important mutations in multiple myeloma, including those involved in the extracellular signal-regulated kinase (ERK) signal transduction pathway. These biomarkers include NRAS, KRAS, and BRAF (4). In recent years, BRAF-V600E has emerged as one of the most promising ERK mutations that can be targeted, validating the concept of targeted treatment. For example, a patient with multiple myeloma with extramedullary disease and the BRAF-V600E mutation responded well to low doses of vemurafenib, a mutation-specific BRAF inhibitor (58), making it a true predictive biomarker in the absence of which a response was not observed.

One potential issue with use of targeted therapy is the presence of ICH relating to subclonal variability. ICH is the essential substrate for clonal evolution according to the principles of Darwinian evolution (4). In recent years, three patterns of clonal evolution have been identified: relapse can be genetically stable; it can linearly evolve with several new genetic variants at relapse; or it can come from a clone that was not present at the time of diagnosis (59). Paired sample analysis of 28 patients with multiple myeloma revealed that patients with standard-risk cytogenetics typically have PC clones that are genetically stable (59). In contrast, patients with high-risk cytogenetics have PC clones that undergo many more genetic modifications over time.

The presence of ICH has important implications for the development of biomarkers. Although it was once acceptable to determine whether a molecular target was present, it is now essential to obtain an idea of the size of the subclone carrying the target. The size of the clone has important therapeutic implications for targeted treatment because completely eradicating a clone present only 5% of the time would have little importance clinically, whereas in a clone present 90% of the time, there would be a significant clinical response.

The presence of ICH also has an effect on the assessment and treatment of relapse. Traditionally, clinical data were used to define prognosis, response to therapy, duration of prior response, rate of relapse, the presence of extramedullary or blastic disease, performance status, and the toxicity of prior treatment. In the era of molecular medicine, cytogenetics, focal lesions, tidal clone status (i.e., same clone, new mutant, or prior clone), GEP signature, and methylation status are beginning to provide clinicians with more information about the clinical behavior of the disease at relapse.

In the 1970s, Kyle and Greipp began to use the phrases monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM) to describe premalignant PC disorders that are not associated with end-organ damage or treatment (60, 61). Currently, asymptomatic multiple myeloma is an active area of biomarker research in which the challenge is to predict those patients who are at high risk of disease progression for whom treatment intervention is essential to prevent the emergence of significant end-organ damage. The use of many of the molecular markers described above is difficult because the molecular features of multiple myeloma that require treatment are often present in patients with asymptomatic disease (62, 63), meaning that alternative approaches to predict risk status are required.

In 2003, the IMWG used laboratory and clinical markers (e.g., serum and urine M protein, clonal BMPCs, and sFLC) to define patients with asymptomatic multiple myeloma in which treatment was not indicated (64). The sFLC ratio is one of the most promising biomarkers in asymptomatic multiple myeloma; it has been used as a prognostic indicator in patients with MGUS (65), SMM (66–68), and newly diagnosed multiple myeloma (69), but it is not without controversy. The Spanish Programa Español de Tratamientos en Hematología (PETHEMA) and Mayo Clinic groups both used sFLC ratios, circulating PCs, and PC proliferation rates to create two models that classify asymptomatic patients as being at low, intermediate, or high risk of disease progression (66, 70). A group at the NCI used the Spanish PETHEMA model (70) and the Mayo Clinic model (66) to categorize 77 patients with SMM as being at low, intermediate, or high risk of developing active multiple myeloma (71). In the NCI's study, concordance between the PETHEMA and Mayo Clinic models was low (28.6%); there was also significant discordance between the ways in which each model classified patients' risk status. Thus, although characterization of PC immunophenotype and measuring sFLC ratios both have prognostic potential, they lack specificity and do not have enough positive predictive value to be useful in determining when treatment should be initiated in patients with asymptomatic disease.

Given that nearly all patients with symptomatic multiple myeloma develop osteolytic lesions, clinicians use bone marrow aspirates and trephine samples to monitor bone morphology throughout the course of disease progression. Although these tools are useful, there can be variability in the distribution of BMPCs throughout the body, increasing the likelihood of heterogeneous sample recovery and sampling error, particularly in patients with asymptomatic disease. Given the importance of understanding the underlying biology of bones in multiple myeloma, clinicians use X-rays to complete skeletal surveys, a technique that is currently the gold standard for evaluating severity of bone disease. Despite the utility of skeletal surveys, MRI, computed tomography (CT), and positron emission tomography (PET)/CT scanning are becoming increasingly important tools for assessment of disease severity. MRI is a useful noninvasive technique for imaging the spine, soft tissue, and infiltration of BMPCs, and is particularly useful for patients with SMM because it can detect diffuse lesions, the presence of which is an adverse prognostic factor for PFS (72). In addition, detection of seven or more focal lesions with MRI is considered to be an adverse prognostic factor in patients with multiple myeloma (73). If MRI is unavailable, CT may be helpful for assessing the extent of damage to the spine or soft tissues. Nevertheless, its use should be limited because CT can expose patients to nearly three times more radiation than conventional MRI. Finally, fusion of CT and fluorodeoxyglucose (FDG) PET images can be very useful for the evaluation of patients with symptomatic disease and has the potential to be useful in patients with MGUS or SMM (74). In the future, functional imaging techniques that are used in combination with biomarker data are likely to contribute to the management of patients with “early” multiple myeloma and the development of novel MRD assays.

Cytotoxic chemotherapy, immunomodulatory agents, and proteasome inhibitors have been highly effective and universally applicable in multiple myeloma. Going forward, the next generation of targeted treatments will depend not only on the presence of a target, but also on their ability to modulate their target. This will be especially important during the early phases of clinical development of targeted treatments, in which target modulation should be considered to be an important endpoint. In multiple myeloma, it is, perhaps, not optimal to use circulating lymphocytes as targets to assess target modulation because of differences in cellular biology and tissue penetration. Consequently, access to and characterization of BMPCs is important and can be achieved using immunohistochemistry on bone marrow trephines or flow cytometry on aspirate samples. Examples include the assessment of antiapoptotic proteins to predict a patient's response to agents that target these pathways, the demonstration of downregulation of pERK, or evidence of changes in histone methylation states following evaluation of MMSET inhibitors.

Over the past 40 years, biomarkers such as M protein and serum β₂M have contributed innumerable insights into our present understanding of multiple myeloma. Today, many cytogenetic markers are beginning to provide information about the severity of the disease; these markers are also beginning to inform clinicians about which anti-multiple myeloma treatment regimen is most appropriate for a particular patient. It is clear that rapid technological advances are changing the way biomarkers are perceived as well as their clinical relevance (Fig. 1). In the near future, diagnostic tests based on massively parallel sequencing approaches will enable detection of recurrent molecular abnormalities as well as actionable mutations that are rarely observed but which could use treatments that are used in other disease settings. Massively parallel sequencing approaches will also enable clinicians to monitor disease longitudinally, which may provide opportunities for early intervention before clonal PC expansion and disease complexity render existing treatments ineffective. Finally, combining functional imaging techniques with assays that use biomarkers has the potential to change clinical practice in patients with early myeloma and patients with symptomatic disease.

No potential conflicts of interest were disclosed.

Conception and design: O. Landgren, G.J. Morgan

Development of methodology: O. Landgren, G.J. Morgan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): O. Landgren, G.J. Morgan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): O. Landgren, G.J. Morgan

Writing, review, and/or revision of the manuscript: O. Landgren, G.J. Morgan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.J. Morgan

Study supervision: O. Landgren, G.J. Morgan

The authors thank A. Peter Morello and Thomas Renau (Onyx Pharmaceuticals, Inc., South San Francisco, CA) for providing medical writing and editorial assistance, which was funded by Onyx Pharmaceuticals, Inc.

This study was supported by Celgene, Janssen, Novartis, Onyx, and Millennium (honoraria; to G.J. Morgan).