Alkylating agents

Melphalan is a nitrogen mustard that alkylates guanine bases in DNA, causing interstrand crosslinks (ICLs) that impair DNA replication and transcription, which results in cytotoxicity (reviewed in [51–53]). A derived compound named melflufen (melphalan flufenamide ethyl ester) has been proposed as an alternative to overcome melphalan resistance without differential toxicity [54, 55].

At high doses, cyclophosphamide metabolization by cells with low levels of aldehyde dehydrogenases produces the alkylating agent phosphoramide mustard, which causes irreversible DNA ICLs as melphalan does. Cyclophosphamide also decreases the immune response and it has been proposed that in low doses it can be used as an immunomodulatory drug (IMiD) in MM (reviewed in [56, 57]).

Historically, alkylating agents, mainly melphalan, were the standard-of-care for MM patients. Although their use is still widespread, frequent relapse of melphalan-treated patients has led to its use in combination with other drugs described in the following sections.

Proteasome inhibitors (PIs)

Tumor cells often accumulate misfolded or aberrant proteins, which leads to increase proteasome levels to eliminate these proteins and cope with their potential toxicity. Hence, proteasomes are an attractive therapeutic target to treat some types of cancer [58]. Such is the case of MM, in which malignant PCs produce high amounts of monoclonal Ig and, therefore, are very sensitive to proteasome inhibition [59].

Bortezomib temporarily inhibits the activity of the proteasome subunit beta type-5 (PSMB5) subunit of the 26S proteasome, which impairs degradation of pro-apoptotic proteins and activation of the anti-apoptotic nuclear factor kappa B (NF-κB) pathway, among others [59]. Other PIs with less secondary effects have been developed. Carfilzomib, the second-in-class PI [60], binds to and irreversibly inhibits the proteasome, even in bortezomib-resistant cells, and has shown improved safety compared to bortezomib [61].

The first orally available PI, ixazomib was more recently developed. As bortezomib, it reversibly inhibits the PSMB5 proteasome subunit [62].

IMiDs

The IMiDs induce immune activation and impair angiogenesis and proliferation of tumor cells through several targets. The first-in-class IMiD was thalidomide, that exerts anti-angiogenic and anti-tumoral activity. Thalidomide disrupts MM cells interaction with the BM microenvironment by inhibiting the production of cytokines that are essential for cell growth and survival [63, 64]. Its primary target is the E3 ubiquitin ligase complex cereblon (CRBN), whose inhibition leads to accumulation of proteins causing cytotoxicity [65] and down-regulates the transcription factors Ikaros (IKZF1) and Aiolos (IKZF3) leading to IRF4 and MYC MM on cogene downregulation [37, 66].

Two analogs of thalidomide have been developed, lenalidomide and pomalidomide, which have shown more potent anti-MM, anti-inflammatory, and immunomodulatory effects than thalidomide [67], and in addition, induce cell arrest and apoptosis [68–71].

Corticosteroids

Glucocorticoids (GCs) are a type of corticosteroids used to treat cancer for over 50 years. Dexamethasone, prednisolone, and methylprednisolone are the GCs used in MM. They bind GC receptors in the cytosol and induce their nuclear relocation. Once in the nucleus, they bind the GC response elements to regulate gene expression, either activating (Annexin I, MAPK) or inhibiting it (NF-κB, AP-1) [72, 73]. GC promotes anti-inflammatory and immunosuppressive activities. They induce apoptosis by affecting several pathways, poly (ADP-ribose) polymerase (PARP) cleavage and caspase 3 activation [74–76]. Due to dexamethasone toxicity, the use of prednisone instead has been proposed, especially for older patients [77]. A detailed review on GC use in MM has been recently published elsewhere [76].

Anthracyclines

Anthracyclines are a class of drugs extracted from Streptomyces bacteria. Doxorubicin (also known as adriamycin) is the most relevant anthracycline in the treatment of MM. It acts as a DNA intercalating agent, inhibiting DNA replication and transcription. Doxorubicin forms a stable DNA-topoisomerase II-anthracycline complex that impairs topoisomerase II-mediated DNA religation, leading to double-strand breaks (DSBs) accumulation and cell death [78, 79].

However, doxorubicin high cardiotoxicity has limited its use. In the 90s, it was reformulated in liposomes [pegylated liposomal doxorubicine (PLD)] to improve its delivery to cells and reduce its toxicity [80].

Histone deacetylase inhibitors (HDACis)

Acetylation and deacetylation of lysine residues in the N-ter of histones that form the nucleosome regulate gene expression. These post-translational modifications are epigenetic markers catalyzed by histone acetyltransferases (HAT) and histone deacetylases (HDAC). Histone acetylation relaxes chromatin structure and increases transcription levels while deacetylation has the opposite effect. The expression of critical genes related to cancer, like TP53 or BCL-2, is regulated by acetylation [81]. Several HDACis have been evaluated in clinical trials: vorinostat [82, 83], ricolinostat [84], romidepsin [85], ACY-241 [86], and panobinostat; of them, only panobinostat has been approved for the treatment of MM [87–89].

Panobinostat is a pan-HDACi that reduces MM cell proliferation, arresting cell cycle at G1/S by affecting the p53 pathway, and induces apoptosis at low doses [87]. Panobinostat alone does not have a significant effect on MM; therefore, it is used in combination with other drugs [87, 90–92]. For a recent review on panobinostat use and clinical trials in MM see [89].

Monoclonal antibodies (mAbs)

Immunotherapy uses mAbs that recognize highly specific antigens on the surface of MM cells leading to cell death. mAbs act through different mechanisms, such as direct cytotoxicity or boosting immune response against malignant cells. In the case of MM, mAbs have been developed to target two proteins: CD38 and CD319. On one hand, CD38 is a transmembrane glycoprotein with functions in signal transduction, calcium signaling and cell adhesion; it is normally expressed by plasmablasts and PCs, and overexpressed by MM cells, which makes it a selective target for the treatment of the disease. On the other hand, CD319 [also known as signaling lymphocytic activation molecule family member 7/CD2 subset-1 (SLAMF7/CS1)] is a stable surface marker of hematopoietic cells, in particular of NK cells, and normal and malignant PCs, which promotes cell growth and survival, and is involved in the interaction of MM cells with BM microenvironment [93, 94].

Several anti-CD38 mAbs have been developed: daratumumab (DARA), isatuximab, MOR202 and TAK-079; of them, only DARA has been approved to treat MM [95–97]. DARA induces immune-mediated cytotoxicity of CD38⁺ cells, and has been suggested to exert immunomodulatory activity, improving clinical responses in heavily pretreated patients [49, 98–100]. Currently, DARA is being studied as monotherapy or in combination with other drugs. The first clinical trials for DARA have provided encouraging results about its use alone [101] or together with other drugs [102–104].

Elotuzumab is an anti-CD319 mAb approved for the treatment of relapsed MM patients. It causes myeloma cell death by activating NK cells or through antibody-dependent cellular toxicity [105]. Elotuzumab has no significant activity when used as monotherapy [106], but it has synergistic anti-myeloma activity in combination with other drugs [107, 108].

The use of mAbs in MM has been recently reviewed elsewhere [49, 109].

Selective inhibitors of nuclear export (SINEs)

Multi-drug resistance (MDR) is a common phenomenon in MM patients. MDR is mainly attributed to the overexpression of the ABC superfamily of ATP-dependent efflux transporters, in particular of P-glycoprotein (P-gp) and multidrug resistance-associated protein 1 (MRP1), encoded by the ABCB1 and ABCC1 genes respectively, and also lung resistance protein (LRP). These transporters can export drugs out of the cell, reducing their intracellular concentrations and diminishing their effect [110].

The cell surface transporter P-gp was identified a long time ago as responsible for drug pump out of the cell. It was associated with resistance to anthracyclines or taxanes, and later to alkylating agents or IMiDs. Moreover, it has been shown that many chemotherapeutic drugs, such as carfilzomib or doxorubicin, induce an up-regulation of P-gp expression [42].

In 2019, the first inhibitor of nuclear export, Selinexor, has been approved (reviewed in [111]). Selinexor binds to exportin 1 (XPO1), a transporter that mediates nuclear export of RNAs and proteins involved in tumor suppression, cell cycle, growth and apoptosis, and has been involved in hematologic malignancies [112]. Of note, XPO1 overexpression has been reported in MM associated to poor prognosis [113].

The use of selinexor in MM has been recently reviewed elsewhere [114].

High-dose chemotherapy rescued by ASCT

ASCT was introduced in the 80s as a therapy in MM [115, 116]. Currently, ASCT is a standard-of-care for young and some eligible elderly patients at first diagnosis. Nevertheless, ASCT is not exclusive for newly diagnosed patients and can be indicated at relapse or progression of the disease [117].

Briefly, CD34⁺ stem cells from the patient are collected from blood, an induction treatment with 2–3 drugs is followed by high-dose melphalan or sometimes radiation to deplete the BM where malignant PCs are, and afterwards the CD34⁺ stem cells are infused back into the patient’s blood [118–120]. However, despite significant event-free survival observed post-ASCT, it does not rule out the possibility of relapses, which remain very frequent [121].

For more insight into ASCT protocols and perspectives, see [121] and [122].

Genomics

For a long time, the classification of newly diagnosed MM (NDMM) patients and subsequent treatment decisions were based on conventional use of cytogenetics, FISH, and single nucleotide polymorphism (SNP) arrays [126]. Identification of chromosome TCs, deletions and amplifications allowed the initial stratification as low- or high-risk MM.

There is not a single common genetic driving event of MM, rather many genetic alterations that differ from patient to patient. A commonly accepted classification defines 7 big molecular groups associated to different risks [17] and some often-mutated genes with prognostic value have been identified [13, 127]. Given this genetic complexity, the limited information provided by karyotype, FISH and SNP microarrays is not sensitive enough to identify the genetic signatures required to guide treatment decisions. Highly performant next generation sequencing (NGS) techniques allow the accurate sequencing of a complete genome within a short time. In MM patients, this capacity to fully sequence the genome is particularly interesting, since the natural course of the disease involves the fast accumulation of new mutations that may lead to relapse or refractory MM (RRMM). Custom capture NGS panels have been developed to identify rearrangements at the IGH locus, CNAs, and frequently mutated genes, with high sensitivity and specificity [128, 129]. This kind of approaches can capture somatic point mutations and provide more precise information than conventional techniques, which makes them good candidates to eventually replace them in clinical practice. Several consortium or groups have developed large datasets from multiple MM patient cohorts (Table 2) representing an extremely useful resource widely used in genomic studies to tackle questions about MM physiopathology and response to different treatments. These studies paved the way for the advent of genomics based-precision medicine in MM.

The importance of genomics studies in MM has been recently reviewed elsewhere [130].

Transcriptomics

The level of expression of potential prognostic genes is as relevant as DNA alterations themselves to understand MM, assessing risk at diagnosis, and designing targeted therapies. In the past decades, the revolution in the sensitivity of transcriptomics approaches has revealed new molecular mechanisms and provided a deeper insight into others already not fully understood.

Combination of WES and RNA-seq determined that the majority of mutations found in a group of 10 MM patients occurred within low- or non-expressed genes, which suggests that they may have low or no functional relevance [187]. This idea highlights the added value of combining DNA sequencing and gene expression data to identify the most relevant genes in MM and the sometimes-low correlation between genomics and transcriptomics data.

RNA-sequencing has also been used to create the first map of gene fusions in MM. Data from 71 HMCLs and 255 NDMM patients showed that patients have on average 5.5 expressed fusion genes, most of them affecting kappa/lambda light chains and IGH genes, and that the number of fusions increases with age. Four hundred and ninety five genes were found to be involved in at least one fusion, specially genes from chromosome 19. The most affected pathways were TNFα signaling via NF-κB, p53 and apoptosis pathways. Of the 36 genes most recurrently involved infusions, 2 were associated to lower progression-free survival (CSNK1G2 and CCND1) and 2 to shorter overall survival (MMSET and BCL2L11), and patients with more than 16 fusions had poorer prognosis [188].

MM dependence on NF-κB pathway was described a long time ago [189], but the underlying molecular mechanisms were poorly known. GEP identified increased expression of NF-κB target genes as a common feature of MM primary cells and HMCLs, and several genetic and epigenetic mechanisms responsible for this NF-κB upregulation that could be interesting therapeutic targets [151]. Major TCs involving the IGH locus, such as t(4;14) or t(11;14), are well known and extensively characterized in MM. However, the frequency of less common TCs is poorly studied. Targeted RNA-seq of 21 patients allowed the identification of 2 novel fusions (HGF/CACNA2D1 and SMC3/MXI1) whose biological and clinical relevance for MM physiopathology has not yet been determined [190]. Additionally, 94 genes at 8 regions that were not previously associated to MM risk have been identified by integrating genome-wide and transcriptome-wide association studies [191]. As already mentioned, personalized clinical management of MM requires predictive scores to assess risk at diagnosis and anticipate responses to drugs. A number of prognostic gene expression signatures have been identified in MM [157, 159, 161, 163, 165, 170–172] (Table 3) from direct comparison of patients with survival or from important biological processes of prognostic importance. A high-risk MM score was defined with70 deregulated genes [“University of Arkansas for Medical Sciences-Total Therapy 70 (UAMS70)” score], most of them mapping to chromosome 1. Interestingly, a model including 17 of those genes (KIF14, SLC19A1, CKS1B, YWHAZ, MPHOSPH1, TMPO, NADK, LARS2, TBRG4, AIM2, ASPM, AHCYL1, CTBS, MCLC, LTBP1) had similar predictive capacity as the 70-gene model [157]. The same group reported a GEP study of PC 48 h after thalidomide and dexamethasone, or bortezomib treatment, developing the “UAMS80” score, which includes 80 genes predicting survival [158]. Of note, 3 genes were common to UAMS70 and UAMS80: PMSD4, BIRC5 and KIAA1754. Similarly, GEP analysis found that high-risk MM patients are characterized by overexpression of cell cycle genes (cell cycle progression, mitosis, spindle assembly checkpoint, DNA replication and repair). These data were used to build a survival model including 15 genes belonging to these molecular pathways, such as MAD2L1, PARP1, BUB1B, and ZWINT, to stratify low- and high-risk patients [163]. In agreement with these results, another GEP study determined that proliferation is a central prognostic factor in MM [164]. Similarly, we used Affymetrix microarrays data to develop and validate a score based on highly transcribed genes in MM that demonstrates prognostic value in NDMM [170]. More recently, Chng and collaborators [192] analyzed the interest to combine different already published gene expression risk-signatures to improve prognostic stratification in MM. Interestingly, the comparison of chromosome instability genome event count signature (CINGECS), CI, EMV92, HMCL7, HZCDC, IFM15, PI, UAMS70 and UAMS80 (Table 3) and analysis of the performance of different score combinations determined that EMC9 + HZCDC was the top-performing prognostic signature combination [192]. This conclusion indicates that integrating several predictive scores may be the best approach to optimal risk estimation.

Single-cell RNA-seq of PCs from patients has identified some of the most affected pathways (i.e. cell metabolism and protein homeostasis pathways) during MGUS to MM progression, providing a new signature for prognosis and therapy stratification [193]. Targeted transcriptome single-cell analysis was used to develop a bioinformatic tool to analyze gene expression changes induced by PI at the subclonal level that could contribute to improve therapeutic choices [194]. Of note, we demonstrated that a collection of 40 HMCLs derived from patients recapitulated the molecular diversity of in vivo MM. This analysis provided a genetic signature for stratification of patient risk and proved that this HMCLs collection is an invaluable in vitro model for the study of MM and the screen of potential new therapies [171]. A recent study described single cell dissection of malignant PC transcriptome in symptomatic and asymptomatic MM patients [195]. They described intrasubject transcriptional heterogeneity and revealed that circulating MM cells reflect the molecular heterogeneity in MM cells of BM.

Numerous molecular pathways are disrupted in MM, involving a plethora of affected factors and pathological consequences. For example, DNA damage repair (DDR) pathways are commonly disrupted in cancer and DNA repair inhibitors are intensively studied as chemotherapeutic agents. Unluckily, resistance to DNA damaging agents remains a common and critical problem. Using gene expression data of factors from the DNA repair pathways homologous recombination (HR), non-homologous end joining (NHEJ), and nucleotide excision repair (NER), we developed a DNA repair risk score to predict event-free and overall survival of NDMM patients [27]. This score offers a rationale to exploit DNA damaging agents, such as melphalan, doxorubicin or cyclophosphamide, in combination with DNA repair inhibitors as a therapeutic option. It could also provide further insight into the molecular mechanisms of resistance to DNA damaging drugs [32]. Similarly, a drug response prediction score based on GEP allows to predict sensitivity to bortezomib and melphalan in single treatment in high-risk patients [196]. More recently, we also identified several kinases involved in major signaling pathways that have prognostic value in MM, such as PBK, SRPK, CDC7-DDF4, MELK, CHK1, PLK4, and MPS1/TTK. Specific CHK1, MELK and PBK inhibitors decreased cell viability in HMCLs and primary myeloma cells and re-sensitized lenalidomide and melphalan resistant cell lines, suggesting a potential interest for kinases as therapeutic targets in combination treatments [197]. A recent study of 42 patients refractory to both PIs and IMiDs performed WES (n = 40) and RNA-seq (n = 27) to attempt to identify chemotherapy resistance mechanisms. It found that the number of subclonal mutations increased during the course of the treatment and p53 pathway was the most frequently mutated. The only mutational signature found in refractory patients was linked to exposure to alkylating agents, indicating that resistance to IMiDs and PIs is generally not due to mutation of their targets [198]. More recently, it was shown that PSMB6 and PSMB7 proteasome subunits, but not PSMB5, are essential for MM cell survival. Resistance to PI have been described with activating mutations or upregulation of PSMB5, PSMB6 and PSMB7 [199].

The application of these and other predictive scores (Table 4) at diagnosis and during the follow-up of MM patients would help clinicians to take better educated decisions regarding the best therapy for each patient. Moreover, low level of expression of certain genes related to therapeutic targets could explain low efficacy of the treatment in some cases, which suggests that RNA-seq tests could help to take better therapeutic decisions.

Dormant myeloma cells in specific BM niches are thought to be responsible for MM relapse [206]. A single-cell RNA-seq study with mice found that these cells express a unique transcriptome signature enriched in immune and myeloma cell differentiation genes, that are switched off during reactivation to growing MM [207]. These data provide a rationale to design therapies to prevent relapse by targeting dormant myeloma cells. A small study compared the transcriptome of myeloma cells in the BM from 4 untreated MM patients and 4 healthy donors and evaluated differential transcript expression, mutations, new alternative splicing variants, and fusion genes. This analysis identified possible candidate genes associated with myeloma genesis, such as EEF1G, ITM2C, FTL, CLPTM1L, and CYBA [208]. Mesenchymal stem cells (MSCs) are essential components of the BM microenvironment with a role in MM. Affimetrix gene-expression microarrays allowed the identification of 3 genes (COL4A1, NPR3 and ITGBL1) expressed in MSCs but little or not in PC. These data were used to build a score predictive of MM progression-free survival and SMM to MM progression, which reinforced the notion that the surrounding microenvironment has a critical impact on MM development [209]. Extramedullary progression of MM is associated to drug resistance and high mortality rate. Single-cell RNA-seq of BM and myelomatous pleural effusions or ascites from 15 patients showed activation of several pathways regulating proliferation, protein degradation, antigen presentation, glycolysis, and oxidative phosphorylation. Moreover, data suggested that myeloma cells possess immune evasion mechanisms, like upregulated inhibitory molecules for cytotoxic T and NK cells [210]. Transcriptomic profiling from a murine myeloma model found that bone morphogenetic protein (BMP) pathway, involved in bone formation and resorption, is upregulated in stromal progenitor cells, which makes it a potential new therapeutic target to treat myeloma-induced bone disease [211].

Epigenomics

Epigenomics focuses on the analysis of epigenetic marks on chromatin, both on nucleic acids and histones, that alter gene expression without changing the DNA sequence. DNA methylation and histone modifications (acetylation, ubiquitination, methylation, and phosphorylation) are the epigenetic marks that drive cancer pathogenesis [212]. Consequently, drug epigenetic reprogramming as an anti-cancer strategy is an emerging field [213].

MM arises from malignant transformation within the B-cell lineage, which is tightly regulated by differentiation and proliferation mechanisms that rely on epigenetic modifications. Interestingly, more than half of MM patients have been found to carry mutations or CNAs on epigenetic enzyme genes [214]. An integrative study of gene expression, epigenetic profiling and in situ promoter capture Hi-C data for the 23 known loci associated with MM susceptibility identified disruption of developmental transcriptional regulators and altered B-cell differentiation as key features for MM risk [135]. Moreover, analysis of 2 available GSE datasets found more than 100 differentially expressed genes (51 upregulated and 78 downregulated) in MM patients compared to controls, mainly related to B-cell receptor, hematopoietic cell lineage, and NF-κB pathways [215].

Misregulation of epigenetic regulators is tightly related to cancer [216]. Methylation of DNA and histones is the most studied epigenetic modification. DNA hypermethylation is responsible for the inhibition of genes that control growth inhibition, apoptosis and cell differentiation, and can take place at particular loci. More precisely, in MM hypermethylation occurs outside CpG islands and is associated with intronic enhancer regions [217]. On the contrary, hypomethylation, which is associated with gene activation, is observed genome wide [218, 219]. Gene expression analysis of MM cells has identified obvious differences in methylation between normal cells and different myeloma stages. Indeed, methylation studies showed global DNA hypomethylation from MGUS to MM, hypermethylation of tumor suppressor genes in MM, and global hypermethylation in the transition from MM to plasma cell leukemia (PCL) [220, 221]. Furthermore, methylation analyses indicate that MM cells reacquire a methylation signature of undifferentiated cells that is associated to MM physiopathology [217]. Combination of DNA methylation and gene expression data identified 4 genes with tumor suppressor functions (GPX3, RBP1, SPARC and TGFBI) whose hypermethylation is associated with shorter overall survival [222]. These data suggest that methylation enzymes could be good therapeutic targets. Thus, DNA methyltransferase inhibitors (DNMTi), like decitabine, have been studied as potential myeloma treatments and a DNA methylation score based on 47 genes is available to predict the efficacy of decitabine [200]. Moreover, pre-clinical studies with human cells and murine models indicated that decitabine modulates Wnt/b-catenin signaling pathway and may increase myeloma cells sensitivity to bortezomib [223, 224], and that decitabine combination with the HDACi quisinostat may have anti-myeloma effect [203, 225]. Therefore, the study of DNA methylome in MM vs. normal PC can allow a better understanding of the mechanisms that control the progression of the disease through all its stages as well as the identification of new therapeutic targets/combinations.

Active DNA demethylation occurs mainly at enhancers and gene bodies. The first step of this process is the conversion of 5-methylcytosine (5mC) into 5-hydroxymethylcytosine (5hmC). Genome-wide 5hmC profiling of tumor PC has been recently performed. This analysis showed that proximity to a 5hmC-enriched region correlated with higher gene expression and indicated that 5hmC may participate in a myeloma-specific gene expression program. Moreover, FAM72 was found to be a new prognosis gene involved in MM: 5hmC of FAM72 enhances its expression and FAM72 high expression is associated to resistance to bortezomib and sensitivity to HDACi/DNMTi [226]. 5hmC is associated with clinical aspects in purified MM cells from patients overlapping with active chromatin marks related to major on cogenes in MM [227].

In MM, poor prognosis t(4;14) TC results in upregulation of MMSET methyltransferase, and consequent global Lysine 36 of histone H3 (H3K36) hypermethylation and Lysine 27 (H3K27) low methylation [220, 228]. Mutations that truncate LSD1/KDM1A, a demethylase of Lysine 4 of histone H3 (H3K4) that regulates hematopoietic stem cell renewal, are associated to familial and early-onset MM. Missense KDM1A mutations are also frequent in patients without MM family history. These mutations correlate with an enrichment of Myc transcriptional targets [229]. Misregulation of histone demethylase KDM3A, which demethylates H3K9me1/2, is also involved in MM. It is upregulated in MM cells under hypoxia and its knockdown has anti-MM effect [219, 230, 231]. Another epigenetic regulator, KDM3B, is the most frequently mutated histone demethylase in MM patients [232]. Polycomb repressive complexes (PRC1 and PRC2) are major chromatin modifiers. PRC1 ubiquitinates H2AK119 and PRC2 trimethylates H3K27 to regulate the expression of important genes related to cell proliferation and differentiation [233]. The role of PRC complexes in MM has only been started to be deciphered. For example, we proved that inhibition of enhancer of zeste homolog 2 (EZH2), a PCR2 member, induces cell cycle arrest and apoptosis, and combination with lenalidomide has synergistic effect. Based on these data, we developed an “EZ score” to identify poor prognosis patients that would benefit from EZH2 inhibition, pointing at a new therapeutic target [205].

Methylation is not the only important epigenetic modification in MM. Using gene-expression data, we and others found that HDACi treatment mainly deregulated tumoral immunomodulatory pathways, which suggested an interest on combining HDACi and IMiDs [203]. In addition, we built scores to predict response to DNMTi, HDACi, or combination of both kinds of drugs [200, 202, 234].

Furthermore, recent studies started to decipher the chromatin regulatory network underlying pathophysiology of MM cells. The first genome-wide profiling of H3K27me3 and H3K4me3 found a set of genes unique to primary MM cells, some with H3K27me3 alone and others with both epigenetic marks. Increased silencing of H3K27me3 target genes was found at advanced MM stages and correlated with poor patient survival. Moreover, EZH2 inhibition was shown to have antimyeloma effects and induce cell apoptosis [235]. Another study mapped regulatory elements, open chromatin, and transcription factor footprints in primary MM cells and compared them to normal cells. The authors found that MM cells present changes in enhancer activity connected to deregulation of transcription factor genes, which leads to aberrant expression of genes involved in proliferation, survival, signaling, adhesion, and DNA methylation. This study also reported widespread decompaction of heterochromatin associated with increased activity of regulatory elements in MM cells [236]. The extensive activation of regulatory elements has been recently confirmed, and linked to upregulation of several signaling pathways, including NF-κB, p53, Notch, and mTOR [237].

In light of all these discoveries, it is not surprising that epigenetic drugs are getting increasing attention as potential MM therapy. The use of sequencing and gene expression profiles to study the role of epigenetic mutations and modifications in MM and normal B cell differentiation, and epigenetic drugs as a strategy to treat MM have been reviewed by others [219, 238].

Proteomics

Proteomics methods allow identification and quantification of proteins or peptides from complex mixtures and purified/enriched samples. Large-scale analysis of proteins provides further information about cell composition both in basal conditions and in response to drugs, and is tightly related to the rest of the omics approaches.

MM is commonly characterized by aberrant secretion of a monoclonal Ig or a fragment of it, known as M-protein or paraprotein [239]. Detection of Ig light chains by electrophoresis in serum or urine from patients is an easy and well-established method to monitor aberrant proteins secreted by PCs for diagnostic purposes [240, 241]. Serum/urine monoclonal Ig detection is used in combination with immunofixation electrophoresis and serum free light-chain assay to diagnose monoclonal gammopathies [242, 243]. M-protein levels are usually elevated at diagnosis and drop upon treatment, serving as a therapy response readout. Changes on free light chain secretion can reflect a clonal evolution linked to proliferation of a new dominant clone, which can have important consequences for therapeutic decisions [244–246]. It has been proposed that monitoring clone-specific peptides is more sensitive and specific than other current analytical methods to detect Igs [247].

Despite the broad use of these techniques in MM diagnosis and monitoring, more powerful proteomics approaches have been proposed. For example, quantitative mass spectrometry (MS) has been used to quantify the serum levels of Igs along the progression of MM, showing increased sensitivity compared to standard clinical methods [248]. However, this method has not been incorporated in clinical practice so far. In addition, the same study coupled proteomics and RNA-sequencing as a strategy for personalized detection of myeloma tumor burden with higher performance [248]. Detection of Ig variable regions from patient sera by MS has recently been proposed as a highly efficient, specific and non-invasive way to assess MRD [249].

M-protein monitoring is not the only way to characterize MM progression by proteomics. Targeting particular groups of proteins or enzymes and their targets provides information about MM molecular mechanisms and drug mechanisms of action. Deregulated activity of several kinases in MM cells, such as ERK, JUNK, STAT, MAPK, and AKT, indicates that phosphorylation pathways could be important for the disease physiopathology and clonal expansion [250–253]. Unbiased phosphoproteomics found differential activation of kinases that was linked to RAS mutations in different HMCLs, which allowed to build a predictive score linking pharmacologic and genetic kinase dependencies in MM [254]. Stable isotope labeling with amino acids in cell culture (SILAC) and liquid chromatography mass spectrometry (LC-MS)/MS analyses of primary MM cells treated with bortezomib identified novel bortezomib-induced phosphorylation sites, mostly in nucleic acid binding proteins, like splicing and translation factors [255]. Moreover, increased phosphorylation of stathmin, a microtubule destabilizer, was reported to play a role in mediating apoptosis upon bortezomib treatment [255]. Proteomics has helped dissecting the apoptosis pathway induced by dexamethasone and its resistance mechanisms in MM [256]. In 2018, the first public resource of phosphorylations and histone modifications induced by treatment with 90 drugs in 6 HMCLs was published, providing a tool for new therapeutic opportunities [257].

For a long time, proteomics has been used to elucidate drug resistance mechanisms in MM. For example, resistance to melphalan in HMCLs has been studied by protein fractionation and liquid chromatography coupled to multiple reaction monitoring (LC-MRM) [258], global proteomic coupled to transcriptomics [259], and proteomics coupled to metabolomics [260]. Melphalan resistance seems to depend on alterations on NF-κB pathway, unbalance of pro-apoptotic, anti-apoptotic signals and DDR factors [258], and metabolic and oxidative stress response pathways [259].

Resistance to bortezomib-based treatments in RRMM patients correlates with accumulation of proteasome subunits and/or proteins involved in the response to oxidative stress and cell redox homeostasis, and misregulation of apoptosis and programmed cell death factors [261]. A correlation between serine synthesis and bortezomib resistance has also been reported [262]. Proteomics analysis of PI-resistant MM cells has shown that adaptation to bortezomib and carfilzomib involves quantitative changes in more than 600 proteins, including downregulation of apoptosis factors, transcription and translation factors, as well as upregulation of metabolism, homeostasis and protein folding and destruction proteins, with MDR1 as the most upregulated protein in carfilzomib-resistant cells [263]. Quantitative proteomics of patients treated with bortezomib-liposomal doxorubicine-dexamethasone or lenalidomide-bortezomib-dexamethasone identified candidate predictors of favorable response to bortezomib, DNA damaging agents, and IMiDs [264].

Upon dexamethasone treatment of MM.1S cell line, global protein expression analysis by two-dimensional polyacrylamide gel electrophoresis (2DPAGE) identified upregulation of proteins involved in post-translational modifications, protein folding and trafficking, and downregulation of proteins involved in cell survival and proliferation. It also suggested that inability to induce FKBP5, a member of the steroid receptor complex, may be related to dexamethasone resistance [256]. The proteomic profile of PCs from NDMM and healthy donors was analyzed by two-dimensional gel electrophoresis (2-DE). Several differentially expressed proteins were identified, including annexin A1. Subsequent functional studies showed that knockdown of annexin A1 potentiated dexamethasone effects insensitive and resistant HMCLs [265].

The effect of thalidomide-based therapy on 39 newly diagnosed responder (n = 22) and non-responder (n = 17) MM patients was analyzed by immunodepletion, two-dimensional difference gel electrophoresis (2-D DIGE) analysis and MS. The best candidate misregulated proteins found in non-responders (ZAG, VDB, SAA, B2M, and Hp) were validated by enzyme-linked immunosorbent assay (ELISA) in a new cohort of patients, helping to establish a thalidomide-response predictive score based on protein quantification [266].

Even the effect of ASCT on serum proteome has been studied by differential scanning calorimetry, a biophysical technique to characterize the stability and conformation of biomolecules in solution, showing that changes in serum paraprotein level upon transplantation can be detected by this minimally invasive method [267].

In the search for new biomarkers, analyzing proteins from serum is a less invasive alternative than performing proteomics from purified BM PCs. For example, proteomic profiling of sera from MM patients that have developed resistance to the triple combination cyclophosphamide-thalidomide-dexamethasone has aimed to identify biomarkers that could predict the response to the therapy. In this study, patient sera were collected before treatment with bortezomib-thalidomide-dexamethasone or bortezomib-doxorubicin-dexamethasone. Subsequently, patient responses to therapy were correlated with their initial protein profiling, identifying 54 proteins that could be potential biomarkers [268].

SILAC allows proteins quantification and relative abundance determination, but its application is limited to proliferating cells with active protein synthesis. Super-SILAC is a variant of this method that overcomes this problem (for more details see [269]), which makes it useful for secondary plasma cell leukemia (sPCL) studies, the last and most aggressive form of MM. sPCL is poorly studied from a molecular point of view because it is a very rare condition and patients present a short survival [270]. In 2017, the first super-SILAC quantitative proteomics study of one patient at both MM and sPCL stages showed almost 800 differentially expressed proteins, revealing a shift in cell metabolism towards aerobic glycolysis (known as “Warburg effect”) in the progression of the disease, which suggests that glycolysis enzymes could be potential therapeutic targets in the treatment of sPCL [271]. Further studies with more patients would be beneficial for the understanding of MM to sPCL progression.

The “Warburg effect” is the switch from mitochondrial oxidative phosphorylation to aerobic glycolysis to generate ATP in many cancer cells, even when oxygen is available. This phenomenon is regulated by several oncogenes, like MYC, and implies the upregulation of glycolytic pathway enzymes [272–274]. Interestingly, some of the above-mentioned proteomics studies [259, 271], as well as other metabolic and molecular studies [275–277], point at the “Warburg effect” association with drug resistance mechanisms in MM. These data support a gene expression analysis that showed a positive correlation between high levels of CD147 and glycolytic enzymes involved in the “Warburg effect”, which was associated to poor prognosis in MM [278]. Thus, aerobic glycolysis factors are interesting therapeutic targets in drug resistant patients which need further study.

A primary study of serum peptide profiles by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) and Clinprot bioinformatics analysis allowed the identification of four peptides in NDMM patients that were used to develop a preliminary diagnostic model with potential interest for early diagnosis [279]. Based on these findings, the same researchers aimed to validate their model by studying the soluble components of the BM niche, which change with the evolution of the disease. They confirmed that the identified four circulating peptides (dihydropyrimidinase-like 2, fibrinogen alpha chain, platelet factor 4 and alpha-fetoprotein) may have biomarker value in MM diagnosis, remission and relapse [280].

Currently, it is well established that BM microenvironment has a central role in MM progression and drug resistance. It contributes to the physiopathology of the disease by secreting molecules important for cell proliferation and adhesion, and by releasing extracellular vesicles (EVs) [281]. Regarding this aspect of MM, proteomics approaches have been used to study the remodeling of the extracellular matrix protein composition from MGUS to MM, providing more insight into the mechanisms of development of a permissive BM microenvironment, and identifying LGALS1 and ANXA2 as biomarkers for MM overall survival [282]. Moreover, a LC-MS/MS study of EVs composition extracted from MM cell lines and patient sera found that EVs with high levels of CD44, an important molecule for cell adhesion and BM secretion of interleukin (IL)-6, could be related to reduced overall survival [283].

Due to the difficulty of primary malignant PCs obtention and culture, most of the reported proteomics studies in MM focus on serum profiling or HMCLs models. However, some effort has been done also on analysis of patient cells. For example, ion intensity-based label-free quantitative MS identified a panel of 9 upregulated, and 9 downregulated proteins in PCs isolated from the BM of MM patients when compared to healthy donors [284].

The importance of proteomics for precision medicine in MM has been recently reviewed elsewhere [285].

Metabolomics

Altered metabolism is one of the hallmarks of cancer [286]. Metabolites are the substrate, intermediate or final products of metabolism, such as amino acids, proteins, lipids or sugars, which exert numerous cellular functions, and are a direct readout of the organism physiology at a given moment. It is now well established that metabolism adaptation provides notorious advantages to cancer cells and that metabolites are responsible for phenotypic manifestations of cancer progression. The study of metabolites, known as metabolomics, allows the identification of diagnostic biomarkers to assess patient risk, disease progression, and predict the response to treatment, as well as the discovery of potential new druggable targets.

Metabolomics approaches can be untargeted, when as many metabolites as possible are measured from the samples, or targeted, when a particular set of metabolites is measured to answer a specific question. Therapeutic drugs are metabolized by the organism and, in general, the end molecules can be detected in a non-invasive way from biological fluids, like blood, saliva or urine, providing huge amounts of valuable information for drug safety and toxicologic screens. In general, metabolites are identified by MS-based technologies, but metabolomics sample processing and data analysis are not well standardized yet and their improvement remains a challenge for the near future [287]. Since metabolic differences between patients can be partially dependent on age, gender and body mass index, performance of metabolomics studies should be carefully designed to use controlled cohorts and implement unbiased analysis methods to avoid misleading biased data.