Gene Expression Profiling in Multiple Myeloma: Redefining the Paradigm of Risk-Adapted Treatment

Multiple myeloma is a blood cancer characterized by clonal proliferation of plasma cells in the bone marrow. In recent years, several new drugs have been added to the therapeutic landscape of multiple myeloma, which have contributed to increased survival rates. However, while the use of therapeutics has evolved, there is still a group of high-risk patients who do not benefit from current treatment strategies. Risk stratification and risk-adapted treatment are crucial to identify the group of patients with urgent need for novel therapies. Gene expression profiling has been introduced as a tool for risk stratification in multiple myeloma based on the genetic make-up of myeloma cells. In this review we discuss the challenge of defining the high-risk multiple myeloma patient. We focus on the standardized analysis of myeloma cancer cells by gene expression profiling and describe how gene expression profiling provides additional insights for optimal risk-adapted treatment of patients suffering from multiple myeloma.

the number of myeloma patients is expected to rise as it follows the ageing population worldwide (2).
In the last 10 years, new therapies and novel mechanisms of action have been introduced in the clinical landscape of MM. The inclusion of immunomodulatory drugs (e.g. lenalidomide, thalidomide and pomalidomide), proteasome inhibitors, (e.g. bortezomib, carfilzomib) and therapeutic monoclonal antibodies (e.g. daratumumab, isatuximab) led to a significant improvement in survival (3). The median overall survival for newly diagnosed patients treated with high-dose therapy is between 4 and 10 years (4). Although some patients may have long term remission or "functional cure", MM is a chronic relapsing disease for the majority of patients. By combining up to 4 drugs (quadruplet regimens) with different mechanisms of action, the list of possible treatment regimens is endless, creating a complex decision tree in the clinical path for a newly diagnosed patient (5).
Despite the enormous progress that has been made in prolonging survival in MM, there is a fraction of patients who do not respond to any of the available treatments or relapse rapidly after an initial response and have reduced survival. In literature, these patients are referred to as high-risk patients. The definition of high risk has evolved over time and there are still many variations on how to describe high-risk disease characteristics. Risk stratification is crucial for better understanding of the disease prognosis and rational use of therapies to achieve risk-adapted treatment. Additionally, risk stratification is essential for understanding the risk-based diversity of patients in clinical trialswhy do certain patients respond better than others.
Studying the genetics of MM offers more insight into the cancer cells and molecular risk stratification. About 20% of newly diagnosed multiple myeloma (NDMM) patients have molecular abnormalities that account for high risk. Fluorescence in situ hybridization (FISH) can detect chromosomal aberrations, such as deletion, translocation and gain. Presence of one these aberrations (single hit) have been associated with worse outcomes in MM patients. Furthermore, presence of more (double or triple hit) of such aberrations indicate serious genomic instability and a very aggressive disease (6,7). Gene expression profiling (GEP) has also been introduced as a tool for risk stratification in MM based on the genetic make-up of myeloma cells and offers additional prognostic subgroups (4).
The recent introduction of novel agents with multiple modes of action has primarily benefited patients with standard-risk disease defined by current criteria. Although the treatment outcome of patients with high-risk disease has improved, the unfavorable impact of high-risk FISH abnormalities has not been abrogated. Therefore, this review will focus on the standardized analysis of myeloma cancer cells by GEP and describe how GEP can provide additional insights for optimal risk-adapted treatment of MM patients.

THE CHALLENGE OF DEFINING THE HIGH-RISK PATIENT
One of the controversies that limit risk-adapted treatment in MM is the challenge of defining the high-risk patient. Stratifying patients into different risk groups depends on several aspects. Molecular abnormalities are one way to determine high risk, but the clinical behavior of the patient is another one. In order to identify the group of patients that are not receiving the right treatment, both patient clinical behavior and molecular abnormalities need to be integrated into the definition of high risk.

Clinical Risk
Patient frailty, renal failure, extramedullary disease, tumor burden, early relapse and minimal residual disease can all predict high-risk disease (8)(9)(10)(11). The Durie-Salmon staging system, introduced in 1975, reflects the tumor burden by using immunoglobulin levels, hemoglobin and calcium concentration and the number of bone lesions as the classification criteria (9). Although widely accepted from its time of publication, the Durie-Salmon staging system lacks reproducibility and has problematic performance in patients under treatment (9,12).
In the following years, the relevance of two important and highly prognostic factors appeared in the clinical field. The first is proliferation rate and disease severity indicated by albumin and its inverse relation with interleukin-6a known growth and survival factor of myeloma cells (13,14). The second is tumor burden and renal function reflected by b-2 microglobulin (15)(16)(17)(18)(19)(20)(21)(22). Serum albumin and b-2 microglobulin have shown to be better indicators of prognosis and have outperformed the Durie-Salmon system (23). In 2005, a large international consortium of myeloma key opinion leaders defined a staging system on the basis of 10,750 patients from three continents that was about to be the new standard: the International Staging System (ISS), that based its three-group stratification on a combination of the two most powerful and reproducible markersalbumin and b-2 microglobulin (24). The role of tumor burden is however affected by age. Data from 3894 patients uniformly treated in the Myeloma XI trial shows that ISS plays the major role in older patients when defining the survival risk and is of less importance in younger patients (25).
Early relapse, that is a relapse occurring within 12 months from autologous stem cell transplantation (ASCT), is a marker of highrisk disease. Early relapse is associated with reduced survival even after an intensive first line of treatment (8,26). The first line of treatment in MM is considered crucial in order to prolong the duration of response and survival (8). In similar fashion, patients who do not achieve long lasting minimal residual disease (MRD) negativity are also considered high-risk (11). MRD negativity has become the new end point in treatment, especially for NDMM. Additionally, highly sensitive MRD monitoring may allow for better prediction of early relapse.
With chromosomal abnormalities obtained through FISH adding information to the prognosis of MM patients, the International Myeloma Working Group proposed the revised ISS (R-ISS) in order to add the presence or absence of cytogenetic markers t(4;14), t(14;16), del(17p) and serum lactate dehydrogenase (LDH) (44). Higher levels of LDH are a proxy for high proliferation rates or the presence of tumor mass leading to extramedullary and extraosseous disease and have shown prognostic value in various treatment settings (45)(46)(47)(48)(49).
To further add to the prognostic arsenal of clinical and biological markers, various molecular gene classifiers were developed to stratify patients on the basis of up-and downregulated genes ( Table 1

The Challenge of Defining High Risk in MM
Globally, there is no consensus on the definition of high risk, nevertheless clinical experts agree that it is never a single marker. Furthermore, clinical experts seem to also agree on an escalated treatment paradigm for high-risk patients with prolonged planned maintenance (55)(56)(57)(58). The National Comprehensive Cancer Network (NCCN) states "patients with cytogenetically and molecularly defined high-risk disease do not receive the same benefit from certain approaches as the low-risk patients and need alternative therapies". 2 The International Myeloma Working Group (IMWG) concludes that "risk stratification in MM is important to predict survival and to define a treatment strategy" (59). In none of the guidelines preferred risk-stratified treatment pathways have been described.
The challenge of defining high risk is also reflected in clinical trials studying the efficacy of treatment (combinations) ( Figure 1). We have performed a search on August 11, 2021 on ClincialTrials.gov on condition "multiple myeloma" in combination with the terms "newly diagnosed" and "high risk". The search resulted in 78 studies that were "not yet recruiting", "recruiting", "enrolling by invitation", "active, not recruiting" or "completed". We further analyzed 17 trials mentioning high-risk as eligibility criteria for enrollment and found 29 different high-risk markers. Figure 1 lists all 29 markers and shows the diversity in selecting markers to define high risk across clinical trials in MM.

GENE EXPRESSION PROFILING IDENTIFIES A UNIQUE GROUP OF HIGH-RISK PATIENTS
The developed GEP signatures provide additional insights into risk stratification in a robust manner. The clinical application of the GEP signatures in myeloma has been stagnant because of the lack of standardization and user-friendly platforms (60). Clinical development of SKY92 has overcome these issues, by providing a standardized, analytically validated and user-friendly tool (61). At this moment, the SKY92 signature is the only fully accredited GEP signature and has consistent performance in detecting high risk in MM (61). Therefore, from this section onwards we will focus on SKY92.

The GEP-Based Marker SKY92
The prognostic GEP-based marker SKY92 was developed as EMC-92 based on the HOVON65/GMMG-HD4 trial using a cohort of 290 NDMM patients. The prognostic power in the combination of 92 genes (including several known cancer genes) was generated by supervised principal component analysis in combination with simulated annealing (4). SKY92 provides a binary outcome and classifies a patient at either high risk with poor survival or at standard risk (61).
SKY92 is clinically validated for accurately predicting the prognosis of NDMM and relapsed/refractory multiple myeloma (RRMM) patients for both overall survival (OS) as well as progression free survival (PFS). After discovery in 2012, SKY92 has been independently validated in 16 patient cohorts totaling 3,339 patient cases ( Table 2) (73). The validation cohorts cover a wide variety of geographies and treatments, clinical trial and non-trial real world settings, and both transplant eligible as well as non-transplant eligible patients. The proportion of high-risk patients identified by SKY92 remains stable around 20% in the NDMM validation sets and is slightly higher for the poorer performing RRMM patients.
Longitudinal analysis of two publicly available MM patient data sets aimed to investigate the evolution of SKY92 risk classification with disease progression (74). SKY92 risk classification was compared between same-patient samples at diagnosis and relapse in the Multiple Myeloma Research Foundation (MMRF) CoMMpass data set and the University of Arkansas for Medical Sciences (UAMS) Total Therapy cohort (TTx). In the analysis of the CoMMpass data set, 31% of patients were classified high-risk at diagnosis. The number of high-risk patients increased to 46% at the second timepoint and 58% for patients with the latest timepoint more than 12 months after baseline. Furthermore, almost all patients in the CoMMpass data set died within 12 months after being classified as high-risk. In the TTx cohort the percentage of high-risk patients increased significantly from 12% at diagnosis to 28% at relapse. Analysis of these data sets show that repeated testing of risk signature provides additional prognostic information for standardrisk patients.
The PROMMIS trial (NCT02911571) investigated the impact of SKY92 on risk stratification and treatment plan (73). This

Combining GEP-Based Biomarkers With (R-)ISS
Besides the use of univariate prognostic markers such as (R-)ISS, single cytogenetic abnormalities or single gene expression classifiers, the combination of prognostic markers is being  increasingly investigated (34,75). This allows for more specific risk classifications, shifting from the ability to determine if a patient is high risk (and otherwise standard risk) towards a multi-categorical classification.
In a multivariate analysis, 20 clinical and biological risk markers were used independently to find the strongest predictor for prognosis as either a stand-alone marker or in combination (62). A total number of 4,750 patients were included from the APEX, HOVON65/GMMG-HD4, IFM-G, MRC-IX, TT2 and TT3 cohorts. The research showed that ISS is a valuable partner to both GEP classifiers and FISH markers. Ranking all existing and combined risk markers showed that GEP+ISS is the strongest predictor for OS, resulting in a 4-group risk classification. In this setting GEP, by means of SKY92, stratified patients into the highrisk and the standard-risk group. ISS sub-stratified the standard-risk patients into two intermediate-risk groups (ISS II + ISS III) and a low-risk group (ISS I group). The median survival was 24 months for the high-risk group, 47 and 61 months for the intermediate-risk groups, and the median survival was not reached after 96 months for the low-risk group.
The GEP classifier SKY92 has also been combined with R-ISS in the HOVON-87/NMSG-18 trial in an analysis of 168 older myeloma patients (76). Combining the R-ISS with SKY92 resulted in 3 risk groups: SKY-RISS I (SKY92 standard risk + R-ISS I, 15%), SKY-RISS III (SKY92 high risk + R-ISS II/III, 11%) and SKY-RISS II (all other patients, 74%). The 3-year OS rates for SKY-RISS I to III were 88%, 66% and 26% (p=6x10 -7 ) and were validated in the elderly patient subset from the CoMMpass dataset. Combining SKY92 with R-ISS resulted in a superior prognostic marker compared to either marker separately.

Unique Group of High-Risk Patients Is Identified With GEP
In order to make sure that a high-risk patient is correctly identified, newer, more robust, standardized and reliable technologies should be incorporated for risk stratification. There is substantial evidence that GEP is indispensable for correct assessment of high-risk MM patient population after which the well-established clinical and cytogenetic markers can further distinguish the low-risk from the intermediaterisk patients.
In 2019, Kuiper et al. analyzed PFS and OS related to highrisk outcomes based on ISS and SKY92 in non-transplant eligible patients from the HOVON87/NMSG18 study (66). In this cohort, 26% of patients were classified as high risk by ISS (ISS III). SKY92 classified 14% of patients as high-risk, with an overlap between the two groups of 5%. Thus, 9% of high-risk patients are misclassified as lower risk (ISS I or ISS II). As the R-ISS is becoming the preferred staging system, in 2020 Kuiper et al. compared prognostication between R-ISS and SKY92 in older myeloma patients (76). In the HOVON87/NMSG18 cohort, R-ISS III identified 7% of patients as high risk, where SKY92 identified 13%. Furthermore, in the CoMMpass cohort these percentages were 13% and 26%, respectively.
Several multivariate analyses have established that SKY92 is an independent prognostic marker and that combining SKY92 with ISS or R-ISS results in a marker with improved performance compared to either marker separately (34,66,76). But how does the prognostic power of SKY92 relate to several high-risk cytogenetic markers such as del(17p) and t(4;14)? Combining six datasets, for a total of 805 patients, SKY92 combined with ISS identified three risk groups: low, intermediate and high (77). The OS of the high-risk group was significantly shorter than the lowrisk group (hazard ratio 6.0, p<0.001). For all three risk groups, the comparison of FISH-positive and FISH-negative patients resulted in a non-significant OS difference. In the high-risk group (n=169), 53% of patients (90/169) were FISH-negative. The high-risk status of these 90 patients was overlooked by using only FISH for risk stratification.
UK OPTIMUM trial (MUKnine, NCT03188172) is a prospective study from 39 UK hospitals designed to identify ultra high-risk patients and provide risk-adapted treatment. Patients were centrally profiled for GEP high-risk signature and/or double hit disease. These patients were considered ultra high-risk and were treated with daratumumab, cyclophosphamide, bortezomib, lenalidomide, dexamethasone (Dara-CVRd) induction, augmented high-dose melphalan and ASCT (78). MRD status was 64% MRD-negative, 14% MRDpositive and 22% not evaluable at day 100 post ASCT (assessed by flow cytometry, sensitivity 10 -5 ). Despite overall high MRDnegativity, some early progressions indicate a group of patients with unmet clinical need. Furthermore, OPTIMUM trial is a digital comparator arm trial. PFS at 18 months was compared between patients in the OPTIMUM trial and matched ultra high-risk myeloma patients from the Myeloma XI trial treated with carfilzomib, lenalidomide, cyclophosphamide and dexamethasone, ASCT and lenalidomide maintenance or observation. Patients from the OPTIMUM trial, treated with the five-drug combination, were found to have significantly improved PFS; with an 18 months estimate of 81.7% (95% CI: 74.2-89.1) and 65.9% (95% CI: 57.3-74.4) for the OPTIMUM and Myeloma XI trial, respectively (79). The analyses of the OPTIMUM trial have shown that identifying the ultra highrisk patients and adapting the treatment can lead to high MRD negativity rates and improved survival in comparison to the standard of care. Lastly, OPTIMUM trial demonstrated the feasibility of incorporating GEP into clinical patient pathway across institutions and showed the value of combining cytogenetic information with GEP.

DISCUSSION
Several new drugs have been added to the therapeutic landscape of MM in recent years, which have contributed to increased PFS and OS. However, while the use of therapeutics in the clinical scenario has evolved, there is still a group of high-risk patients who do not benefit from current treatment strategies. The international guidelines on MM all recognize the relevance of having detailed information on the patient's disease and associated risk for progression in order to better evaluate the individual clinical care pathway (59). It has become clear that there is not a single marker capable of independently and accurately defining the high-risk MM patient. For this reason, multiple combinations are postulated by different study groups, but no real consensus has been formulated and used in clinical trials.
In this review, we aimed to describe the impact of the genetic make-up of cancer cells as a molecular risk factor in MM. In this section, we will discuss the integration of both clinical variables and molecular markers into the definition of high risk and the outlook of risk-adapted treatment in MM.
GEP-based marker SKY92 is a standardized tool for risk stratification that provides additional information to the anamneses of patients with MM. SKY92 allows for risk stratification in relatively homogenous subgroups of patients and provides added value in combination with clinical markers and FISH, which by itself does not capture the genomic complexity of MM (27,34). SKY92 stratifies MM patients into high-risk or standard-risk group irrespective of treatment regime and relapse setting. Risk stratification using SKY92 is important at diagnosis in order to choose the optimal first line of treatment for maximum effect and prevention of relapse.
Longitudinal monitoring of risk assessment using SKY92 allows for dynamic risk stratification (74). GEP in combination with FISH is a great way to track changes in molecular risk over time and should be performed at diagnosis and every relapse to correctly identify high-risk patients and guide treatment (80,81).
Risk can also change during the course of disease depending on response to treatment. Early relapse has strong association with reduced survival in both high-risk and standard-risk cytogenetic groups (8). MRD also plays a big role in patient prognosis. MRD-negative patients with high-risk cytogenetics have similar outcome as standard-risk patients (82). However, further research is needed to investigate the proper actions that should be taken regarding de-escalation of therapy of patients with sustained MRD-negativity.
There are still some limitations of using GEP in practice. Bone marrow sample availability is one of the challenges, as well as the lack of guidelines for optimal risk classification and corresponding treatment strategy. Additionally, SKY92 was developed and validated for active and symptomatic MM, therefore the prognostic value of SKY92 has yet to be assessed in (asymptomatic) smoldering myeloma. Lastly, a cost-benefit analysis of SKY92 still needs to be performed because current evidence is lacking. Such analysis should be done to investigate the cost impact of SKY92 on the use of high-cost anti-myeloma drugs.
Despite the current limitations, there is substantial evidence that GEP is an important tool in providing the most accurate risk assessment identifying the true high-risk population and can be used in combination with the well-established clinical and cytogenetic markers. After GEP risk assessment, ISS and FISH can further distinguish the intermediate-risk from the low-risk patients. Without GEP, many patients are misclassified using existing tools. Therefore, a new era beckons in which patients are routinely and accurately assigned risk and its relevance, considering the available treatment opportunities. In this new era: • GEP is used to identify high-risk MM patients; • GEP is part of routine diagnostic work-up to allow for riskadapted strategies; • Risk-adapted treatment is investigated in both clinical trials and the real-world setting.

CONCLUDING REMARKS AND FUTURE PERSPECTIVES
With the availability of new techniques and increasing knowledge of MM biology, the definition of high risk is evolving and should include personalized assessment of both clinical and molecular markers. Treatment should focus on biology of high risk in younger patients and on clinical behavior in older patients. Treatment intensification in molecular high-risk patients and dose reduction and deliverability in clinical high-risk and frail patients need to be researched. Treatment combining several modes of action and incorporation of novel immunotherapies, for example CAR Tcell therapy or bispecific antibodies (83), could be the next area to explore for the high-risk patients that represent the group with unmet clinical need. Risk-adapted therapy is crucial to achieve deep and sustained response in high-risk MM patients. In order to develop therapeutic strategies for specific risk groups, it is of utmost importance to use GEP as an eligibility marker. In many trials the risk groups are stratified along different randomization arms or inclusion criteria. For both misclassified high-risk patients in a low-risk trial and low-risk patients in high-risk trials, the final conclusions on the effectiveness of the investigated therapeutic regimen will be influenced. Clinical trials focusing on high-risk MM patients are crucial for identification of optimal therapy for improved survival.