Prognostic and Therapeutic Biomarker Developments in Multiple Myeloma

New approaches to stratify multiple myeloma patients for informing on prognosis and therapy selection are needed since patients are currently managed in a similar manner regardless of individual risk factors. However, despite new and improved biomarkers for determining the prognosis of patients, there is currently insucient information to utilise biomarkers to intensify, reduce or altogether change treatment, nor to target patient-specic biology. The ever-increasing number and complexity of drug classes to treat multiple myeloma have improved response rates and so clinically useful biomarkers will need to be relevant in the era of such novel therapies. Therefore, the eld of multiple myeloma biomarker development is rapidly progressing, spurred on by new technologies and therapeutic approaches, and underpinned by a deeper understanding of tumour biology with individualised patient management the goal. In this review we describe the main biomarker categories in multiple myeloma and relate these to prognostic and therapeutic applications.


Introduction
Multiple myeloma (MM) is an incurable haemopoietic malignancy caused by uncontrolled proliferation of neoplastic plasma cells. It is the second most common blood cancer with approximately 140,000 newlydiagnosed patients each year worldwide and 1,876 new cases diagnosed in Australia in 2018 1 . Despite recent advances in therapy, the 10-year survival rate remains at only 17% 2 . In MM, a complex array of genetic and epigenetic changes lead to neoplastic transformation of plasma cells, resulting in their uncontrolled growth within the bone marrow (BM) and secretion of large amounts of non-functional monoclonal antibody (known as paraprotein or M protein) into the circulation 3,4 . The main clinical manifestations of MM are the development of osteolytic bone lesions resulting in bone pain and fracture, hypercalcaemia, renal insu ciency, and those relating to bone marrow failure. MM encompasses a spectrum of clinical variants ranging from benign monoclonal gammopathy of uncertain signi cance (MGUS) and smouldering/indolent MM (SMM) to more aggressive, disseminated forms of MM and plasma cell leukaemia. While the emergence of proteasome inhibitors such as bortezomib and car lzomib, and immunomodulatory drugs (IMiDs) such as lenalidomide, has seen overall survival (OS) increase from 3 to 6 years in the past two decades, relapse usually occurs. New approaches to better risk stratify MM patients for informing on prognosis and therapy selection are needed since most patients are managed in a similar manner regardless of individual risk factors 1,4 .
A biomarker is de ned as a molecule that is found in blood or other bodily uids that serves as an indicator of normal and abnormal processes and can be used for diagnostic, prognostic or treatment response monitoring purposes 5 . Biomarkers can be in the form of genomics, transcriptomics, proteomics, and clinicopathologic variables including imaging. Despite new and improved biomarkers for determining the overall prognosis of MM patients, there is currently insu cient information to utilise biomarkers to intensify treatment for high-risk MM, reduce treatment for low-risk MM or for changing to an alternative treatment strategy altogether. In this review we describe the main biomarker categories in MM, relating these to prognostic and therapeutic applications.

Current Staging Systems
In 2014 the International Myeloma Working Group (IMWG) revised the de nition and diagnostic criteria for MM 6 . The revised de nition includes patients who are symptomatic due to the presence of one or more CRAB criteria (hypercalcemia, renal insu ciency, anaemia and bone lesions) together with > 10% clonal plasma cells in the bone marrow biopsy and those who lack CRAB criteria though have either > 60% clonal plasma cells in the bone marrow, a serum involved/uninvolved free light chain ratio of > 100 or more than one focal lesion > 5 mm on MRI imaging of the skeleton 6 . The rst staging system by Durie-Salmon included the CRAB criteria and paraprotein level for prognostication, however, there are di culties in its application due to inconsistencies in the interpretation of the number of bone lesions whilst elevated serum creatinine and lower haemoglobin levels have aetiologies that may not necessarily relate to MM 7 . In 2005, the International Staging System (ISS) was developed utilising only the biomarkers albumin and β 2 -microglobulin 8 . The ISS performed well at the time of its creation but could not consistently predict clinical outcomes in the era of novel agents with no differences in OS between staging groups 8 . Moreover, higher values of β 2 -microglobulin and lactate dehydrogenase (LDH) are correlated with inferior survival and associated with greater tumor burden, however are also affected by renal failure and other comorbidities which can confound the results and limit the utility of these biomarkers. To overcome these di culties, the Revised International Staging System (R-ISS) also utilised β 2 -microglobulin and LDH but with the addition of high risk cytogenetics 9 . The presence of high risk cytogenetics by uorescence in situ hybridization (FISH) automatically classi es the patient as Stage III regardless of LDH or β 2 -microglobulin elevation.
Response to treatment is generally followed through paraprotein measurements in serum or urine, or through serum free light chains provided the ratio of the involved to uninvolved free light chains is abnormal. The small subset of non-secretory MM that lacks these measurable serum biomarkers is followed by plasma cell percentage and more recently skeletal imaging using MRI or PET/CT 10 . Whilst designed for informing prognosis, unfortunately, none of these staging systems are useful for making therapeutic decisions.
The asymptomatic stages MGUS and SMM are generally characterised by a lower percentage of malignant plasma cells and lower levels of serum M protein, light chains and other traditional biomarkers. MGUS is present in approximately 3.5% of the general population over the age of 50 and progresses to MM at the rate of approximately 1% of patients per annum 1,2 . For SMM, the risk is higher at 10% per annum in the rst 5 years, 3% per annum for the subsequent 3 years and 1% thereafter 1,2 . Up until recently there were no reliable biomarkers to sub-classify SMM patients into those with a high versus low risk of progression to MM. The Mayo Clinic and Spanish models were developed for more robust prognostication of SMM patients (Table 1) and together proposed new criteria for identifying high risk SMM 11,12 . Moreover, the Spanish myeloma group was able to show that high risk SMM patients treated with lenalidomide and dexamethasone had delayed progression and improved OS compared with observation alone 13 . Table 1 Risk strati cation of patients with SMM according to Mayo Clinic and Spanish models 11

Heavy And Light Chains As Prognostic Biomarkers
Currently, serum free light chain (FLC) levels are measured to monitor therapeutic responses, de ne complete response (CR) and to herald relapse in oligosecretory MM 10 . Moreover, an abnormal involved to uninvolved FLC ratio appears to be an accurate predictor of progression for patients with SMM and of survival and therapeutic response in patients with MM [15][16][17] . An abnormal FLC ratio before autologous stem cell transplantation (ASCT) predicted early progression afterwards and a one-third reduction in FLC levels within 30 or 60 days predicted a favourable prognosis 18, 19 . The prognostic value of FLC was independent of high risk translocations such as t(4;14) and t(14;16), although they were positively correlated 20 . As MM is characterized by spatial and temporal clonal heterogeneity, determining serum M protein and FLC levels likely re ects the overall MM disease burden in the body 21 . Moreover, measurement of FLCs may even be better than serum M protein in informing on prognosis. Compared with MM relapse associated with an increase in M protein only, relapse de ned as an increase in FLC alone predicted a shorter time to second line therapy, increased risk of progression and mortality 22,23 .

Prognostic Utility Of Imaging Modalities
Bone marrow and other traditional modalities for staging MM do not take into account the spatial heterogeneity of disease and thus imaging techniques are required for obtaining more complete information on disease burden. Skeletal survey using X-rays has long been the method used to identify bone lytic lesions, however, this cannot detect extramedullary disease nor cord involvement with MM and is not sensitive in detecting small lytic lesions 24 . The skeletal survey has essentially been replaced by more sensitive imaging modalities including whole body low-dose CT, MRI and 18 F-uorodeoxyglucose PET/CT which have now been incorporated into diagnostic and response assessment criteria [24][25][26] . CT can detect early destructive bone lesions but is unable to detect active MM in areas of prior destruction nor extramedullary sites of MM 24,27 . Conversely, MRI is su ciently sensitive to detect early marrow in ltration, can differentiate between benign and malignant osteolytic lesions and is useful for detecting the pattern and extent of bone marrow involvement with MM 28 . When used in asymptomatic MM, identi cation of at least one lesion, a diffuse in ltrative marrow pattern and 20% or more marrow in ltration predicted progression to symptomatic MM 29 . FDG-PET has similar advantages to MRI however results are obtained in a more reasonable time frame whilst CT shows early bone destruction 25 .
Thus, PET/CT appears to be the ideal imaging modality for both disease staging and follow-up [30][31][32] . The metabolic response and number of focal lesions found on PET/CT in patients with newly-diagnosed MM after treatment had independent prognostic value 33,34 . Therefore, these new imaging modalities are capable of accurately differentiating early-stage MM from MGUS and SMM, for which X-rays usually are negative, and are better at predicting disease progression.

Bone Turnover Markers
These markers can be divided into two categories: collagen fragments released from bone matrix during degradation and enzymes released from either osteoblasts or osteoclasts and can be utilized as noninvasive tools for detecting skeletal morbidity risk and response to anti-resorptive treatment 35 .
Biomarkers re ecting osteoclast-mediated collagen degradation such as N-terminal crosslinking telopeptide of type-1 collagen (NTX), C-terminal crosslinking telopeptide of type-1 collagen (CTX), Cterminal crosslinking telopeptide of type-1 collagen generated by metalloproteinase (ICTP) and deoxypyridinoline, provide information on the remodelling process and re ect whole body bone turnover 36 . Urinary NTX, serum CTX and serum ICTP levels are elevated in MM patients and correlate with advanced osteolytic bone disease [37][38][39] . Moreover, urinary NTX and serum ICTP correlate with risk for skeletal complications, progression-free survival (PFS) and OS 38,40,41 . Finally, procollagen type-1 N-propeptide and procollagen type-1 C-propeptide correlate with new bone formation whilst receptor activator of nuclear kappa B ligand (RANKL) and osteoprotegerin are also important markers of bone turnover which were found to normalize after ASCT 41-43 . 6. Genomic Biomarkers  51,52 . Loss of 1p affecting FAF1 and CDKN2C genes have also been associated with shortened survival 53 .
The site of the immunoglobulin heavy chain locus on chromosome 14q32 is the most involved chromosomal translocation locus in MM. A notable favourable prognostic MM feature is translocation t (11;14), which is associated with higher CD20 expression, lymphoplasmacytic or small mature plasma cell morphology, hyposecretory disease and nuclear cyclin D1 expression and dysregulation 54,55 . This genetic defect can be targeted with the Bcl-2 inhibitor venetoclax although recent clinical trials in relapsed/refractory MM showed increased mortality due to infection with venetoclax resulting in the FDA placing a partial hold on these studies. The t(4;14) (FGFR/MMSET) translocation was once thought to be a poor risk feature but in the era of proteasome inhibitors, has a more favourable outcome 56 . High risk translocations such as t(14;20) and t(14;16), and del(17p) which affects TP53 continue to have poor prognosis despite advances in therapeutics 51,57 . However, patients with these features were found to do better with triplet therapy (bortezomib, lenalidomide and dexamethasone) compared with intermediate or standard risk disease 58 .
Current risk strati cation is based on individual cytogenetic abnormalities without consideration of more than one being concurrently present, thus potentially rendering the disease course more unpredictable (Table 2). Kumar et al reviewed 500 patient FISH analyses and found that only 3% of MM patients had no discernible cytogenetic abnormality 59 . One third of patients were found to have a translocation event with the most common being t(11:14) at 18%. A further 12% had an abnormality in the IgH locus. Trisomies predominated with 57% of patients having at least one chromosomal trisomy, and 48% having at least two chromosome trisomies 59 . Monosomy 13 was seen in 47% and only 13% had 17p deletion (as either a deletion or monosomy). The most common overlapping cytogenetic abnormalities were translocations with the presence of another IgH abnormality 59 . Monosomy 13/del(13q) was seen in 57% of patients with a concurrent IgH abnormality and rarely without this abnormality or a trisomy. 36% of patients had both a trisomy and IgH abnormality and p53 abnormalities tended to occur with either a translocation or a trisomy and were rarely seen independently. Good risk cytogenetics have generally been considered to be the trisomies, and poor risk cytogenetics included translocation events and p53 mutations. The presence of high risk FISH without a trisomy conveyed a poor prognosis with median OS of 3 years. However, the same high risk FISH with at least one trisomy conferred a standard prognosis 59 . This bene cial effect of trisomy was seen irrespective of the type of high risk cytogenetic defect (translocation or del(17p)).

Proteomics
Whilst numerous individual proteins have been shown to carry prognostic signi cance in MM, unlike genomic analyses, there are comparatively few protein 'signatures' that have been developed for this purpose. Proteomic pro ling of bone marrow plasma cells from relapsed and/or refractory MM patients revealed a protein signature associated with proteasome inhibitor resistance. In particular, increased expression of proteasome activator complex subunit 1 (PSME1) in patients not achieving a very good partial response (VGPR) correlated with the observed clinical resistance to bortezomib-based therapy 75 .
Another study also examined changes in the proteome that predicted response to either bortezomib or lenalidomide containing treatment regimens 76 . Such proteome changes included those relating to endoplasmic reticulum stress and acute phase response signalling, suggesting responsiveness to protesome inhibitors or IMiDs, respectively 76  However, sometimes the target of a drug, though initially an obvious biomarker choice, may not turn out to be useful. For example, IMiDs such as lenalidomide and pomalidomide target cereblon (CRBN), which has been proposed as a predictive biomarker of response. However, there are con icting reports in the literature concerning the clinical utility of CRBN and its downstream targets as predictive biomarkers of IMiD response, since downregulation of CRBN is not able to explain a lack of response 84,85 . On the other hand, immunoglobulin lambda (Igλ) translocations confer a poor outcome to patients receiving IMiDs through a hitherto unknown mechanism 86 . Similarly, mechanisms of resistance to proteasome inhibitors are still not well understood, with mutations in proteasome subunits or expression of multi-drug resistance transporters evident in only a minority of cases and of questionable involvement 4 .

Liquid Biopsies -Circulating Tumour Cells And Dna
Due to the variable nature of bone marrow involvement in MM, bone marrow biopsy is unlikely to represent the spatial or temporal mutational landscape of the disease. Circulating tumour cells (CTCs) are released from primary tumour or metastatic sites into the bloodstream with higher numbers suggested to be an adverse prognostic nding for MM patients at diagnosis, after ASCT and during concordance rate for detecting tumour-derived mutations with allele fraction as low as 0.25% was obtained 96 . Furthermore, the authors reported 3 cases with detectable clonal populations in the ctDNA but not in matched bone marrow samples suggesting determination of sub-clones through the analysis of blood plasma may be more informative than that obtained from single bone marrow aspirates. Despite these encouraging ndings, the utility of ctDNA for monitoring MRD is controversial with con icting reports in the literature. For example, one study demonstrated a high correlation between next generation sequencing (NGS) of IgH gene rearrangements in the ctDNA of diagnostic and post-treatment samples with 8-colour ow cytometry 97 . On the other hand, detectable ctDNA of diagnostic samples could only be tracked in 39% of patients with VGPR or a worse response whilst another study found ctDNA was undetectable in 69% of samples which were clearly MRD positive in the bone marrow 98,99 . However, it should be noted that ctDNA may decline more rapidly than in other plasma cell compartments 100,101 .
Overall, the use of ctDNA as a tool for MRD monitoring is not fully mature and requires integration of molecular techniques and bioinformatic analysis.

Measurable (Minimal) Residual Disease
As therapy for MM improves and more patients achieve a stringent complete response (sCR), new highly sensitive techniques are required to detect residual disease. Two techniques have demonstrated the ability to accurately measure residual clonal plasma cells in the marrow, including next generation ow cytometry (NGF, e.g. EuroFlow) and NGS of immunoglobulin genes (e.g. Clonoseq: Adaptive Technologies) 102,103 . Continually strengthening data suggest that MRD could be used as a biomarker to evaluate treatment e cacy, inform on therapeutic decision making and predict PFS and OS 104  patients with acute lymphoblastic leukaemia and chronic lymphocytic leukaemia, NGS has proved to also be applicable to MM and is more sensitive than traditional MFC and ASO-qPCR. Imaging tools such as PET/CT also play an important part in MRD detection, speci cally for the detection of extramedullary disease and early tumour activity. The IMWG has de ned response criteria categories of MRD negativity to allow uniform and structured reporting (Table 4). Table 4 International Myeloma Working Group MRD Criteria 10 .

Result
Criteria De nition* Sustained MRD negative MRD negativity in the marrow (NGF or NGS, or both) and by imaging as de ned below, con rmed minimum of 1 year apart. Subsequent evaluations can be used to additionally specify the duration of negativity (e.g. MRD negative at 5 years).

Flow MRD negative
Absence of phenotypically aberrant clonal plasma cells by NGF on bone marrow aspirates using the Euro-Flow standard operation procedure for MRD detection in multiple myeloma (or validated equivalent method) with a minimum sensitivity of 1 in 10 5 or greater nucleated cells.

Sequencing MRD negative
Absence of clonal plasma cells by NGS on bone marrow aspirate in which presence of a clone is de ned as fewer than two identical sequencing reads obtained after DNA sequencing of bone marrow aspirates with the LymphoSIGHT platform (or validated equivalent method), with a minimum sensitivity of 1 in 10 5 or greater nucleated cells.
Imaging plus MRD negative MRD negativity as de ned by NGF or NGS plus disappearance of every area of increased tracer uptake found at baseline or a preceding PET/CT or decrease to less than mediastinal blood pool SUV or decrease to less than that of surrounding normal tissue.
* These criteria require achieving complete response on the basis of the standard International Myeloma Working Group response criteria 10 . MRD, minimal residual disease; NGF, next-generation ow; NGS, next-generation sequencing; SUV, standardized uptake value. currently no established guidelines on therapeutic tailoring based on the MRD result. However, it is likely that in future MRD will indeed guide such therapy decisions as is the case in other haematological malignancies [111][112][113] . There a many so called response-adaptive clinical trials underway at this time, for example, the PERSEUS phase III randomized study of daratumumab/bortezomib/lenalidomide/dexamethasone versus bortezomib/lenalidomide/dexamethasone in 690 newly-diagnosed MM patients (NCT03710603).
Patients in the daratumumab arm with sustained MRD negativity (10 − 5 ) for 12 months after a minimum of 24 months of daratumumab/lenalidomide maintenance therapy will stop daratumumab until progressive disease results. Upon recurrence of MRD or loss of CR, patients will re-commence daratumumab.
Until recently, the majority of clinical trials have not included MRD as a primary clinical endpoint. This is largely due to a number of MRD aspects that remain debated, including (1) patient selection (those in sCR and CR or also VGPR), (2) the timing of MRD testing during the treatment course, (3) the optimal cut-off  (Table 5). Recently, a large meta-analysis that included data from 6 randomised trials (3,283 newly-diagnosed MM patients) con rmed that achievement of MRD negativity strongly correlated with prolonged PFS 114 . Thus, MRD met the Prentice criteria for PFS surrogacy 115 .   Traditional clinical trials are not optimal for the rapid evaluation of precision medicine approaches.
Recently, integrated platform trials known as master protocols have been developed to facilitate the simultaneous and rapid testing of multiple treatments in the same trial 129 . An example is the MMRF Myeloma-Developing Regimens Using Genomics (MyDRUG) study (NCT03732703). This is a Phase I/II trial master protocol designed to develop novel precision medicine combinations for patients with high risk MM whose disease has rapidly progressed in spite of standard-of-care therapies. Genomic alterations that can be therapeutically targeted are being identi ed with study arms for individuals with RAS/RAF mutations, CDK activating alterations, FGFR3 and IDH2 mutations, as well as t (11;14) translocations.
Patients without these genetic aberrations receive an immunotherapy (daratumumab) and all patients receive a backbone of ixazomib, pomalidomide and dexamethasone (IPd). However, a major problem for precision medicine approaches is that of tumour clonal heterogeneity 69 . Speci cally, therapy targeting molecular aberrations present in only a subset of MM cells might not achieve clinical bene t and may result in the rapid growth of sub-clones. With regard to the MyDRUG trial, the genetic lesion being targeted must be present in a relatively large proportion of the MM cells, having an allelic fraction > 0.3.
Despite these advances, a number of obstacles remain. Precision medicine MM drugs will require appropriate companion diagnostic tests to detect the targetable lesions, however, assay availability and cost remain barriers. The FDA recently approved the MSK-IMPACT™ and FoundationOne CDx™ panels which will hopefully help to accelerate reimbursement for clinical sequencing. Moreover, assays need to be developed to detect functional changes such as signalling pathway activation rather than only the presence or absence of a genetic or protein lesion. Additionally, bone marrow biopsies in MM are limited in their ability to fully represent the genetic heterogeneity of the disease. As discussed previously, liquid biopsies are emerging as promising tools in MM to satisfy the requirements of precision medicine approaches, however, much work is needed to optimise this methodology for use clinically 91 . Finally, in recent years new immunological treatments including anti-CD38 and anti-SLAMF7 monoclonal antibodies have become available 4,130 . B-cell maturation antigen is currently being investigated clinically through a variety of Chimeric Antigen Receptors (CARs), T-cell engagers and monoclonal antibody therapeutics, however, reliable methods to evaluate a patient's immune system are not available, which is frequently deranged in MM patients, particularly those heavily pre-treated. Immune pro ling will therefore be essential to match patients with the appropriate combinations of targeted and/or immune-based therapies.

Conclusions
There continues to be much interest and indeed much progress in elucidating biomarkers that assist with determining prognosis and treatment selection for patients with MM, with the future moving towards precision medicine and individualised patient management. It may be that an integrated approach which includes clinical, serological, imaging, genetic and protein biomarkers is required to guide both therapy selection as well as prognostication, and ongoing efforts are incorporating new biomarkers such as miRNAs, non-coding RNA and splicing events. Moreover, the presence of functional events such as downstream activating mutations in key signalling pathways, as seen with KRAS and BRAF mutations in the ERK pathway are being considered. Importantly, it will also be necessary to understand the dynamic and changing clinical impact of multiple clones, sub-clones, their evolution and the molecular mechanisms driving these clones, on disease outcome in each MM patient.
The ever-increasing number and complexity of MM drug classes, including immunotherapeutic approaches such as bi-speci c monoclonal antibodies, antibody-drug conjugates and CAR-T cells that not only target the malignant plasma cell but also harness the immune system have markedly improved response rates in MM patients. Ultimately, clinically useful biomarkers will need to be relevant in the era of such novel therapies, which in some instances can overcome the poor prognosis of MM patients harbouring traditionally poor prognostic biology. Moreover, to improve upon the tremendous progress made so far, high-throughput technologies are being incorporated into research with array-based methods giving way to sequencing-based methods. In parallel, new bioinformatics methodologies are being developed for detailed analyses of the large amount of data generated. Therefore, the eld of MM biomarker development is rapidly progressing, spurred on by new technologies and therapeutic approaches underpinned by a deeper understanding of MM biology with individualised patient management the goal.

Declarations
Ethics approval and consent to participate: Nothing to declare Consent for publication: The authors give full consent to publishing this work in the Journal of Hematology & Oncology

Availability of data and materials
There is nothing to declare

Competing interests
There are no competing interests to declare

Funding
The authors received no speci c funding for this work Authors' contributions CWB and RLM reviewed the literature and wrote the manuscript