Comprehensive molecular profiling of multiple myeloma identifies refined copy number and expression subtypes

Multiple myeloma is characterized by the infiltration of malignant plasma cells in the bone marrow, and is associated with high levels of monoclonal (M) protein in the blood and/or serum. It can occur de novo or evolve from benign monoclonal gammopathy of undetermined significance (MGUS), which is characterized by low amounts of bone marrow plasmacytosis and M protein, with absence of osteolytic lesions, anaemia, hypercalcemia and renal function impairment. Approximately 1% of individuals with MGUS evolve to multiple myeloma per year. Multiple myeloma has a prevalence of 50,000 patients in the United States, occurring in approximately 16,000 new individuals each year. Clinical manifestations include bone disease, hypercalcemia, cytopenia, renal dysfunction, hyperviscosity and peripheral neuropathy. In spite of conventional and high-dose chemotherapies, it remains uniformly fatal owing to intrinsic or acquired drug resistance (Box 1).

Current translational research focusing on the development of molecularly-based combination therapies has the potential to achieve a high frequency of durable responses in most patients with multiple myeloma. Two major advances are making this goal possible. First, recent advances in molecular cytogenetic, genomic and proteomic studies of tumour cells and their normal counterparts have allowed for increased understanding of multiple myeloma pathogenesis, provided the basis for molecular prognostic classification in the context of uniform therapies, identified potential therapeutic targets, and provided the rationale for combining targeted therapies to increase tumour cell cytotoxicity, reduce non-specific cytotoxicity and overcome conventional drug resistance. Second, there is now an increased understanding of how the adhesion of multiple myeloma cells to bone marrow further affects gene expression in multiple myeloma cells and in bone marrow stromal cells (BMSCs), thereby increasing tumour growth, survival, drug resistance and migration in the bone marrow milieu. As a result of these advances, a new treatment paradigm has emerged in multiple myeloma, based on the concurrent targeting of both the tumour cell and its bone marrow microenvironment to overcome drug resistance and improve patient outcome. For example, thalidomide, lenalidomide and bortezomib, three agents recently approved by the US Food and Drug Administration (FDA) for the treatment of multiple myeloma, counteract the protective effect of the bone marrow milieu. Ongoing studies are using oncogenomics to identify, through gene microarray and array-based comparative genomic hybridization (aCGH), new therapeutic targets in multiple myeloma cells and the bone marrow microenvironment. These findings will probably lead to next generation therapies, and will inform the design of rational combination trials for multiple myeloma treatment. This new paradigm has great promise not only to change the natural history of multiple myeloma, but also to serve as a model for identifying and validating new targeted therapeutics directed to improve the outcome of patients with other haematological malignancies and solid tumours.

Advances in oncogenomics of myeloma

Multiple myeloma is characterized by a distinctive combination of gains and losses of whole chromosomes, by non-random chromosomal translocations causing dysregulation of genes at breakpoints, and by point mutations. In addition, epigenetic changes such as gene methylation and a constellation of small, focal lesions in the genome of multiple myeloma cells, have recently been identified by aCGH (see below).

Multiple myeloma can be subdivided into two groups based on the pattern of chromosomal gains and losses. Approximately 55-60% of patients have a hyperdiploid karyotype (number of chromosomes from 48 to 74), with trisomies of chromosomes that happen to be odd-numbered, including 3, 5, 7, 9, 11, 15, 19 and 21. The remaining cases (non-hyperdiploid group) include patients with hypodiploid, near-diploid, pseudodiploid or near-tetraploid chromosome number (less than 48 or more than 74 chromosomes). The mechanisms that underlie this pattern have not been elucidated, but importantly, ploidy status rarely changes during disease progression, and patients with hyperdiploid multiple myeloma tend to have a better prognosis than those with non-hyperdiploid disease. Recent genomic and expression-profiling studies have proposed further subclassifications of the hyperdiploid and non-hyperdiploid groups. Within the hyperdiploid group, aCGH based classification has identified a subset of patients presenting additional gains on 1q and/or losses of chromosome 13. These patients have a worse prognosis than the patients in the non-hyperdiploid group (see below).

In addition to whole chromosome gains and losses, multiple myeloma patients present with non-random, early-onset reciprocal chromosomal translocations involving the immunoglobulin H (IgH) locus (at 14q32.3), and less frequently the IgL locus (2p12, κ or 22q11, λ). In these primary translocations, various genes are juxtaposed to a strong Ig enhancer that dysregulates their mRNA expression. Two translocations directly increase the expression of cyclins: t(11;14)(q13;q32), which occurs in 15-20% of multiple myeloma patients, induces cyclin D1 overexpression; and t(6;14)(p21;q32), present in 5% of multiple myeloma cases, increases expression of cyclin D3 (Ref. 24). Another translocation, t(4;14)(p16.3;q32), is present in approximately 15% of patients and dysregulates the expression of the Wolf-Hirschhorn syndrome candidate 1 gene (WHSC1, also known as MMSET), which encodes a protein with homology to histone methyltransferases, and the receptor tyrosine kinase fibroblast growth factor receptor 3 (FGFR3) gene. Finally, the t(14;16)(q32;q23) dysregulates the oncogene MAF, a basic-leucine zipper transcription factor, in 5-10% of patients, and t(14;20)(q32;q11) affects another member of this family, MAFB, in 5% of cases. These rearrangements seem to be mutually exclusive; although in 5% of MGUS and 25% of advanced multiple myeloma cases two independent translocations can be found in the same patient. The oncogenic consequences of these translocations on multiple myeloma pathogenesis have not yet been extensively explored, with the exceptions of those caused by MAF and FGFR3 deregulation. MAF promotes multiple myeloma cell proliferation and increases multiple myeloma cell adhesion to BMSCs. Inhibition of FGFR3 in t(4;14) multiple myeloma induces plasma cell differentiation and apoptosis. However, FGFR3 is not expressed in 25% of patients with t(4;14) multiple myeloma, questioning its universal relevance as a therapeutic target in this group of patients. Importantly, these chromosomal translocations more often affect non-hyperdiploid patients and are linked to prognosis. Specifically, t(14;16) and t(4;14) translocations are associated with a poor prognosis, whereas patients presenting with t(11;14) have a longer survival relative to all other genetic subtypes.

Expression studies have identified subtypes within the non-hyperdiploid group associated with specific chromosomal translocations. Moreover, it has been shown that translocations affect, directly or indirectly, the expression of cyclins. The fact that cyclin D1 is also consistently dysregulated in the hyperdiploid group, by a yet-to-be-defined mechanism, suggests an important role for D-type cyclin dysregulation in the early pathogenesis of all cases of multiple myeloma.

In addition to the hyperdiploid-non-hyperdiploid dichotomy and the presence of chromosomal translocations, gains or losses of specific chromosomal regions occur in all patients with multiple myeloma and are linked to prognosis. These include chromosome 13 monosomy, loss of the short arm of chromosome 17 (where the tumour-suppressor gene TP53 resides) or the short arm of chromosome 1 (Ref. 13), and gains or amplifications of the long arm of chromosome 1 (Refs 11,17,40). The oncogene MYC is involved in chromosomal translocations and/or amplifications in up to 45% of patients with advanced multiple myeloma. Deletions of 17p and 1p, as well as loss of chromosome 13, are linked to poor prognosis, although the prognostic significance of chromosome 13 remains controversial. Indeed, chromosome 13 loss portends poor prognosis to conventional low-dose and high-dose chemotherapy, but not to the proteasome inhibitor bortezomib. Gains or amplifications of 1q have also recently been proposed as an adverse prognostic factor. In particular, in the most comprehensive expression-profiling survey of patients with multiple myeloma published to date, Shaughnessy et al. identified 70 genes linked to early disease-related death, providing the first validated classifier for prognosis prediction in patients with uniformly-treated multiple myeloma. Strikingly, 30% of these genes were located on chromosome 1, with most of the downregulated genes located on the short arm of chromosome 1 and the upregulated genes mostly on 1q. These recent expression-profiling findings confirm the crucial role of genes located on this chromosome in the survival of patients with multiple myeloma.

aCGH surveys, coupled with expression analyses, have recently identified many focal genetic lesions related to multiple myeloma initiation and progression. Tumour-suppressor genes are found to be mutated, including TP53, the phosphatase and tensin homologue PTEN and the cyclin-dependent kinase inhibitors CDKN2A and CDKN2C, suggesting an important role for several tumour-suppressor genes in multiple myeloma. In addition, in approximately 10% of patients with t(4;14) multiple myeloma, oncogenic, activating mutations of FGFR3 are present. Importantly, two members of the Ras family (NRAS and KRAS) are mutated at codons 12, 13 and 61 in 40-55% of multiple myeloma patients versus only 5% in MGUS, suggesting that the activation of the mitogen-activated protein kinase (MAPK) pathway has an important role in the progression from MGUS to multiple myeloma. Indeed, oncogenomics studies have identified only a few differences distinguishing MGUS from multiple myeloma. Both conditions can present either a hyperdiploid or hypodiploid karyotype and similar chromosomal translocations affecting the IgH or IgL locus. Using expression array profiling, a previous study did not find significant differences between MGUS and multiple myeloma, whereas a more recent analysis has identified a signature of MGUS shared by a subgroup of multiple myeloma patients with a better prognosis. A higher incidence of t(4;14) has been reported in multiple myeloma versus MGUS, but not in all studies. Therefore, at present Ras mutations are the major genetic difference identified between MGUS and multiple myeloma.

These studies over the past decade have highlighted several rearrangements affecting the genome of multiple myeloma cells at different levels. However, one of the peculiarities in the pathogenesis of this disease is the essential role exerted by the bone marrow microenvironment and by the mutual interactions between the cells that reside in this niche and multiple myeloma cells. Studying the bone marrow compartment has been challenging, but recent work has reinforced the relevance of bone marrow accessory cells and the extracellular matrix (ECM) in multiple myeloma pathogenesis. Oncogenomics studies are crucial to understand the genetic heterogeneity of multiple myeloma cells that is likely to influence their interaction with the bone marrow, which in turn induces growth, survival and drug resistance in multiple myeloma cells.

Role of the bone marrow microenvironment

The bone marrow microenvironment consists of various ECM proteins, including fibronectin, collagen, laminin and osteopontin; and cell components including haematopoietic stem cells, progenitor and precursor cells, immune cells, erythrocytes, BMSCs, bone marrow endothelial cells (BMECs), as well as osteoclasts and osteoblasts. The physical interaction of multiple myeloma cells with ECM proteins and accessory cells in the bone marrow milieu, which can be modelled in vitro and in vivo, has a crucial role in multiple myeloma pathogenesis. BMSCs, osteoclasts, osteoblasts and endothelial cells secrete factors including interleukin 6 (IL6), insulin-like growth factor 1 (IGF1), vascular endothelial growth factor (VEGF), B-cell activating factor (BAFF), fibroblast growth factor (FGF), stromal cell-derived factor 1α (SDF1α), and tumour necrosis factor-α (TNFα), which are further upregulated by tumour cell adhesion to ECM proteins and/or BMSCs. The direct interactions of multiple myeloma cells with accessory cells and secreted cytokines activate signalling pathways mediating growth, survival, drug resistance and the migration of multiple myeloma cells, as well as osteoclastogenesis and angiogenesis (Fig. 1).

Homing and adhesion of multiple myeloma cells to the bone marrow. The homing of multiple myeloma cells to the bone marrow is mediated by the chemokine SDF1α, which interacts with its receptor CXCR4 on multiple myeloma cells. SDF1α induces motility, internalization of CXCR4, and cytoskeletal rearrangement in multiple myeloma cells; conversely, specific CXCR4 inhibitors and anti-CXCR4 antibodies inhibit migration of multiple myeloma cells in vitro, suggesting that the SDF1α-CXCR4 interaction is a crucial regulator of multiple myeloma homing. Both homotypic and heterotypic adhesion of multiple myeloma cells to either ECM proteins or BMSCs is then mediated through several adhesion molecules. For example, CD44, very late antigen 4 (VLA4), VLA5, leukocyte function-associated antigen 1 (LFA1), neuronal adhesion molecule (NCAM), intercellular adhesion molecule (ICAM1), syndecan 1, and MPC1 all mediate the adhesion of multiple myeloma cells to the bone marrow. VLA4 expressed on multiple myeloma cells mediates both binding to the ECM and to BMSCs, through fibronectin and VCAM1, respectively (Fig. 1). The binding to fibronectin upregulates p27 (encoded by CDKN1B) and induces nuclear factor κB (NFκB) activation in multiple myeloma cells, which confer cell adhesion-mediated drug resistance (CAMDR) (Box 1) to conventional chemotherapy. Most recently, the upregulation of p27 and p21 (encoded by CDKN1A) following the adhesion of mantle cells and other non-Hodgkin lymphoma cells to BMSCs was associated with the downregulation of SKP2, a subunit of SCF(Skp2) ubiquitin ligase, indicating that cell-cell contact modulates the tumour cell cycle through ubiquitin-proteasome proteolytic pathways. However, it remains to be assessed whether a similar mechanism also occurs in multiple myeloma cells.

Syndecan 1 is a transmembrane heparan sulphate-bearing proteoglycan expressed in most multiple myeloma cells. The adhesion of multiple myeloma cells to the ECM through the binding of syndecan 1 to type I collagen induces the expression of matrix metalloproteinase 1 (MMP1), thereby promoting bone resorption and tumour invasion. Increased serum soluble syndecan 1 correlates with increased tumour cell mass, decreased MMP9 activity and poor prognosis. Importantly, soluble syndecan 1 promotes the growth of multiple myeloma cells in vivo. Therefore, delineating the mechanisms whereby multiple myeloma cells home and localize in the bone marrow both provides new biomarkers and identifies potential therapeutic strategies to overcome CAMDR.

BMSC-multiple myeloma cell interactions. Multiple myeloma cell adhesion to BMSCs not only localizes tumour cells in the bone marrow microenvironment, but also has important functional sequelae. Multiple myeloma cell adhesion to BMSCs triggers the NFκB-dependent transcription and secretion of cytokines such as IL6 in BMSCs, which further stimulate multiple myeloma cell growth, survival, drug resistance and migration (Fig. 1). Conversely, the inhibition of NFκB activity abrogates this response. Moreover, multiple myeloma cells localized in the bone marrow milieu secrete cytokines such as TNFα, transforming growth factor-β (TGFβ) and VEGF, which further upregulate IL6 secretion from BMSCs. NFκB also mediates the expression of many adhesion molecules expressed on both multiple myeloma cells and BMSCs. Moreover, activation of NFκB by cell adhesion and cytokines (such as TNFα) augments the binding of multiple myeloma cells to BMSCs, which in turn induces IL6 transcription and secretion in BMSCs. Within the bone marrow, interaction of the TNF receptor superfamily member 5, CD40, on multiple myeloma cells with its ligand CD40L on bone marrow accessory cells upregulates adhesion molecules (such as LFA1 and VLA4) and augments multiple myeloma cell adhesion, thereby increasing IL6 and VEGF secretion in BMSCs. Conversely, anti-CD40 antibodies (SGN-40 or CHIR-12.12) inhibit multiple myeloma cell adhesion to fibronectin and BMSCs, thereby decreasing IL6 and VEGF secretion from BMSCs. These antibodies are currently under evaluation in clinical trials for the treatment of multiple myeloma, confirming the value of oncogenomics for the identification of molecular and clinical strategies that target the multiple myeloma cell in its bone marrow milieu (Table 1).

Interactions of multiple myeloma cells with the bone marrow microenvironment -- either directly through cell adhesion molecule-mediated interactions between multiple myeloma cells and BMSCs, or indirectly by the effect of growth factors released by both cell types and trapped in the ECM -- activate a pleiotropic proliferative and anti-apoptotic cascade. In addition to NFκB, several signalling pathways are involved in this response: the phosphatidylinositol 3-kinase (PI3K)-Akt pathway; the Ras-Raf-MAPK kinase (MEK)-extracellular signal-regulated kinase (ERK) pathway; and the Janus kinase 2 (JAK2)-signal transducer and activator of transcription 3 (STAT3) pathway. Downstream sequelae include the cytoplasmic sequestration of many transcription factors (such as FKHR); upregulation of cell-cycle regulatory proteins (such as D-type cyclins) and anti-apoptotic proteins (such as BCL-X, MCL1 and caspase inhibitors) (Fig. 1); and increased activity of telomerase. Importantly, all these events promote multiple myeloma cell growth, survival and migration, contributing to multiple myeloma progression and constitutive drug resistance (Box 1). Moreover, many of the growth factors secreted by multiple myeloma and BMSCs stimulate osteoclastogenesis (IL6, IL1, VEGF, SDF1α, macrophage inflammatory protein 1α (MIP1α)) and angiogenesis (VEGF). Therefore, targeting these interactions and related growth factors provides the basis for new treatment strategies. New agents (bortezomib, thalidomide and lenalidomide) that target not only multiple myeloma cells directly, but also tumour-host interactions and the bone marrow milieu (cytokines and angiogenesis), can overcome CAMDR and the growth advantage conferred by the bone marrow, and therefore hold great promise to improve patient outcome.

Moreover, genetic lesions in multiple myeloma may modulate the ability of tumour cells to interact with their bone marrow milieu. For example, multiple myeloma cells with t(14;16) translocation overexpress the transcription factor MAF, which not only transactivates the cyclin D2 promoter, thereby increasing multiple myeloma cell proliferation; but also upregulates β7-integrin expression, thereby promoting tumour cell adhesion to BMSCs. By contrast, multiple trisomies and cyclin D1 overexpression are under-represented or absent in plasma cell leukaemia (PCL), which is independent of the bone marrow microenvironment. These examples highlight the biological significance of the interplay between specific genetic lesions and those signalling pathways that mediate multiple myeloma-microenvironmental interactions.

BMEC-multiple myeloma cell interactions. Following the initial homing of multiple myeloma cells to the bone marrow stromal compartment, adhesion between multiple myeloma cells and BMECs upregulates many cytokines with angiogenic activity. Angiogenesis promotes multiple myeloma cell growth by increasing the delivery of oxygen and nutrients, removing catabolites and through the associated secretion of growth factors from endothelial cells (Fig. 1). Multiple myeloma cells may also constitutively produce, as a result of oncogene activation and/or genetic mutations, factors such as VEGF, basic fibroblast growth factor (bFGF) and MMPs that stimulate bone marrow angiogenesis. Conversely, BMECs secrete growth factors, including VEGF, IL6 and IGF1, which promote multiple myeloma cell growth in the bone marrow milieu. Importantly, growth factors and cytokines such as VEGF and IL8, which are secreted by BMSCs, allow multiple myeloma cells to recruit new blood vessels in the bone marrow. The BMECs in these new multiple myeloma-associated vessels further support tumoral cells through cytokines and direct adhesive interactions. Importantly, these autocrine and/or paracrine loops in the bone marrow milieu may mediate the progression of multiple myeloma. BMECs from multiple myeloma patients not only express VEGF and bFGF, but also several other pro-angiogenic molecules including angiopontin 1, TGFβ, PDGF, hepatocyte growth factor (HGF) and IL1.

The level of bone marrow angiogenesis, as assessed by grading and/or microvessel density (MVD), is consistently increased in patients with progressive multiple myeloma compared with those with inactive disease or MGUS. In active multiple myeloma, MVD correlates with the degree of plasma cell proliferation and infiltration, and is therefore an adverse prognostic marker. Moreover, a recent analysis in newly diagnosed multiple myeloma found a correlation between MVD, syndecan 1 in blood and bone marrow, and level of serum HGF, further highlighting the potential use of MVD as a new biomarker of progression of multiple myeloma. Importantly, bone marrow neovascularization can be targeted by new agents. For example, thalidomide inhibits BMEC-mediated secretion of VEGF, bFGF and HGF, BMEC proliferation and capillarogenesis in patients with multiple myeloma; its empiric use in multiple myeloma was based on its anti-angiogenic effects.

Osteoclast-multiple myeloma cell interactions. Following homing to the bone marrow, multiple myeloma cells trigger osteoclast activity, inducing bone resorption, lytic bone lesions, and/or diffuse osteopaenia. BMSCs and osteoblasts regulate osteoclastogenesis by producing receptor activator of NFκB ligand (RANKL) and osteoprotegerin (OPG). Importantly, RANKL expression is associated with differentiation and is induced preferentially from immature osteoblasts. RANKL binds to RANK on osteoclasts, thereby stimulating their differentiation and activity, whereas OPG functions as a decoy receptor by binding to RANKL and preventing its interaction with RANK, thereby inhibiting its ability to stimulate osteoclastogenesis (Fig. 1). The blockade of RANKL binding to RANK receptor by OPG inhibits osteoclast maturation and bone destruction. By contrast, the binding of multiple myeloma cells through α4β1 integrin to VCAM1 on BMSCs decreases the secretion of OPG and increases the expression of RANKL, thereby promoting osteolysis. Importantly, multiple myeloma cells affect the OPG:RANKL ratio in the bone marrow environment, thereby promoting lytic bone lesions.

Another mechanism that contributes to bone destruction in multiple myeloma is mediated by MIP1α, which is secreted by multiple myeloma cells. MIP1α is a potent inducer of osteoclast formation independently of RANKL, and promotes both RANKL-stimulated and IL6-stimulated osteoclast formation. MIP1α binds to CCR1 on osteoclasts and CCR5 on multiple myeloma cells, and blocking CCR1 and CCR5 inhibits osteoclast formation and multiple myeloma cell adhesion to BMSCs, respectively. Levels of MIP1α in the bone marrow plasma of patients with multiple myeloma are increased, and correlate with the presence of osteolytic lesions. Moreover, an antisense oligonucleotide against MIP1α decreases bone destruction, multiple myeloma adherence to BMSCs, and tumour burden in a mouse model of multiple myeloma.

Importantly, osteoclasts produce several factors that stimulate multiple myeloma cell growth, including IL6. In addition, IL6 secretion from BMSCs, triggered by multiple myeloma-BMSC adhesion, not only induces multiple myeloma cell proliferation and survival, but also stimulates osteoclastogenesis and contributes, along with cytokines such as IL1β, VEGF and HGF, to the increased bone resorption observed in multiple myeloma. Moreover, p38 MAPK in BMSCs upregulates the production of IL11, RANKL and MIP1α, which induce osteoclast formation and activity. Conversely, a p38 MAPK inhibitor can suppress these cytokines, resulting in decreased multiple myeloma cell proliferation and adhesion, as well as decreased osteoclastogenesis. Chondroitin synthase 1, which is upregulated by multiple myeloma cell-osteoclast interactions, induces Notch signalling and survival of multiple myeloma cells, and therefore represents a novel therapeutic target.

Osteoblast-multiple myeloma cell interactions. Decreased activity of osteoblasts also contributes to osteolytic lesions in multiple myeloma. The formation and differentiation of osteoblastic cells from mesenchymal stem cells require the activity and function of the transcription factor RUNX2 (Refs 120-122). The binding of VLA4 on multiple myeloma cells to VCAM1 on osteoblast progenitors downregulates their RUNX2 activity. Soluble factors expressed by multiple myeloma cells, such as Dickkopf 1 (DKK1) and IL3, may also contribute to the inhibitory effects of multiple myeloma cells on osteoblast differentiation and RUNX2 activity (Fig. 1). DKK1 is a Wnt inhibitor and is detected in multiple myeloma cells from bone marrow biopsy specimens. As DKK1 inhibits the canonical Wnt pathway, which mediates the differentiation of osteoblast progenitor cells, DKK1 expression by multiple myeloma cells could inhibit osteoblastogenesis. Importantly, recombinant human DKK1 or bone marrow plasma containing increased levels of DKK1 inhibits the differentiation of osteoblast precursor cells in vitro, and increased DKK1 levels in bone marrow plasma and peripheral blood from patients with multiple myeloma are associated with focal bone lesions. Conversely, a neutralizing anti-DKK1 antibody inhibits bone resorption and tumour cell growth in a mouse model of multiple myeloma. As the expression of OPG and RANKL is also regulated by Wnt signalling, DKK1 may have a key role mediating the osteolytic process in the bone marrow microenvironment in multiple myeloma. In addition, multiple myeloma cell lines and patient multiple myeloma cells constitutively produce another soluble Wnt inhibitor, Frizzled-related protein 2 (FRP2), which also significantly suppresses osteoblast differentiation.

Although IL3 has been shown to stimulate osteoclast activity, a recent study has reported that IL3 also inhibits basal osteoblast formation stimulated by bone morphogenic protein 2 (BMP2) without affecting cell growth, suggesting that IL3 has a role in the bone destructive process in multiple myeloma by inhibiting osteoblast formation. TGFβ from multiple myeloma cells augments IL6 secretion from BMSCs and osteoblasts, thereby further stimulating osteoclast activity. HGF, which can directly inhibit osteoblastogenesis in vitro, is increased and inversely correlated with bone-specific alkaline phosphatase, a marker of osteoblast activity, in sera from multiple myeloma patients. Therefore, multiple cytokines mediate the imbalance of osteoblast and osteoclast activity in multiple myeloma. Importantly, osteoblasts can be activated by proteasome inhibitors, suggesting that these novel agents, which achieve responses even in advanced multiple myeloma, may also induce new bone formation.

Impact of bone marrow interactions on multiple myeloma cell molecular profiles. Constitutive genetic alterations in multiple myeloma cells are not the sole determinants of the biological behaviour of multiple myeloma cells in their local microenvironment. Instead, the molecular profiles of multiple myeloma cells show a wide range of inducible alterations at the transcriptional and proteomic levels that mediate the protective effects of the bone marrow microenvironment on multiple myeloma cells, promote the adhesion of multiple myeloma cells to BMSCs and bone marrow accessory cells, recruit new blood vessels to provide nutrients and growth factor support for multiple myeloma cells, and/or trigger increased osteoclastic bone resorption. For example, the interaction of multiple myeloma cells with BMSCs and bone marrow accessory cells upregulates transcripts for cytokines such as IL6 (Refs 68,137), HGF and insulin-like growth factors (IGFs); the anti-apoptotic protein MCL1 (Ref. 138); and heat shock proteins (HSPs), which regulate the conformation and function of proliferative and anti-apoptotic proteins. This increased autocrine production of cytokines, along with paracrine cytokine production from the BMSCs, stimulates proliferative and anti-apoptotic signalling cascades in multiple myeloma cells. Upregulation of MCL1 or HSPs contributes to increased multiple myeloma cell survival and drug resistance in the bone marrow milieu. Cytokine-driven signalling events also lead to the transcriptional upregulation of adhesion molecules, such as integrins and ECM proteins, and promote the binding of multiple myeloma cells to BMSCs, which in turn promotes the viability of multiple myeloma cells in the bone marrow milieu by activating anti-apoptotic signalling cascades (for example, the focal adhesion kinase (FAK)-PI3K axis) both directly, through adhesion molecules and cell-cell contact, and indirectly by increasing the secretion of anti-apoptotic cytokines. These inducible molecular alterations in multiple myeloma cells interacting with their microenvironment also influence multiple myeloma-associated neo-angiogenesis and bone resorption. Finally, constitutive chromosomal translocations, amplifications, deletions or gene mutations in multiple myeloma cells modulate the transcriptional, proteomic and functional sequelae triggered in multiple myeloma cells interacting with their local bone marrow milieu. For example, IL6 secretion is upregulated by the multiple myeloma-BMSC interaction, and may be of particular importance in multiple myeloma cells that overexpress IL6 receptor (IL6R) owing to amplifications in chromosome 1q21 (Ref. 11) (Fig. 2). These studies suggest that oncogenomics can be used to show multiple myeloma subtype-specific interactions with the bone marrow milieu.

Validation of new therapeutic targets

High resolution genomic profiling has both identified new potential therapeutic targets in tumours and defined distinct new clinical pathogenetic subgroups of multiple myeloma. Specifically, merging aCGH data with gene-expression profiling has identified amplified and overexpressed genes that are associated with the prognosis of patients with multiple myeloma. For example, hyperdiploid multiple myeloma is a genetically heterogeneous disease, and chromosome 1q gain and/or chromosome 13 loss identify a subset of these patients with poor prognosis. To functionally validate the oncogenic potential and therapeutic relevance of such genes, both gain-of-function and loss-of-function (antisense, small interfering RNA or short hairpin RNA) studies can confirm their role in multiple myeloma cell growth, survival, drug resistance and migration using in vitro models of the bone marrow milieu. Monoclonal antibodies or vaccines can then be developed against validated targets on the multiple myeloma cell surface, and prototype small-molecule inhibitors against intracellular targets. It is possible that histone deacetylase (HDAC) inhibitors (such as SAHA and depsipeptide) or inhibitors of DNA methyltransferase (such as 5 aza-2'deoxy-cytidine) might be used to reverse the epigenetic silencing of tumour-suppressor genes in multiple myeloma. Targeting cyclin D dysregulation (as discussed above) could potentially have a major impact on the treatment of all molecular subtypes of multiple myeloma. Several studies have attempted to target cyclin D directly by inhibiting its mRNA expression or translation (that is, desferroxamine and eicosapentaenoic acid), by inducing its degradation through post-translational modifications (ubiquitylation and proteasomal degradation) or by inhibiting its function (selective CDK kinase inhibitors). Additional specificity may be achieved by targeting the genes directly dysregulated by IgH translocations, such as t(4;14)(p16.3;q32) translocation, which occurs in approximately 15-20% of multiple myeloma tumours, and results in the dysregulated expression of two putative oncogenes, MMSET and FGFR3. Preclinical studies have validated FGFR3 as a therapeutic target in t(4;14) multiple myeloma, and FGFR3 small-molecule tyrosine kinase inhibitors are under clinical evaluation; inhibitors of histone methyltransferases are being developed, and studies are underway to validate MMSET as a target in t(4;14) multiple myeloma.

In vivo validation assays can establish the role of target genes or pathways emerging from oncogenomic analyses. In this regard, animal models have been developed to define mechanisms that underlie multiple myeloma cell homing to bone marrow, factors that mediate multiple myeloma cell-bone marrow interactions, and factors within the bone marrow milieu that promote angiogenesis as well as multiple myeloma cell growth, survival, drug resistance or migration (Fig. 3). These models have expanded the range of therapeutic targets in multiple myeloma. Among the models recently introduced, the severe combined immunodeficiency (SCID) mouse models, in particular the SCID/non-obese diabetic (NOD) model of diffuse multiple myeloma lesions, the SCID-human (hu) model and the transgenic Em-directed X-box Binding Protein-1s (XBP-1s) mouse model, permit the evaluation of the cellular, gene transcript and protein alterations in multiple myeloma cells in the context of the bone marrow microenvironment, and can be used to validate promising targeted therapies in vivo. Human multiple myeloma cells engineered to stably express fluorescent proteins such as green fluorescent protein (GFP) can be xenografted in the SCID/NOD and SCID-hu multiple myeloma models. This allows direct real-time imaging of migration, binding and growth of human multiple myeloma cells in bone marrow, as well as the isolation of multiple myeloma cells for the further characterization of transcriptional and protein expression changes induced by treatment of the host with a targeted therapy. Importantly, the SCID-hu model allows for the validation of a targeted therapy in the context of human multiple myeloma cells interacting with human BMSCs and the ECM. Finally, the transgenic Em-directed XBP-1s mouse model of multiple myeloma has recently been generated by Em-directed overexpression of a spliced isoform of XBP-1s, a transcription factor that is required for plasma cell differentiation. Both XBP-1s transcript and protein are expressed at high levels in multiple myeloma cells versus normal plasma cells. In this Em-XBP-1s model, mice develop features of MGUS and progress to multiple myeloma. Interestingly, both premalignant B cells and multiple myeloma cells in this model have dysregulation of genes implicated in human multiple myeloma, including cyclin D1, cyclin D2, MAF, MAFB, IL6R, GP130, BAFF and APRIL. This mouse model therefore mimics human multiple myeloma and represents a valuable platform for studying the biology of human multiple myeloma and assessing new therapies in vivo.

Bench to bedside combination therapies

In spite of conventional and high-dose chemotherapies, which can often achieve high overall frequency of durable responses in multiple myeloma, disease inevitably relapses owing to the acquisition of drug resistance, albeit at dramatically different rates. Salvage therapies can achieve only transient responses and multiple myeloma remains incurable. To overcome resistance to conventional and high-dose therapies, we have delineated gene and protein modulations in multiple myeloma cells and the bone marrow microenvironment associated with sensitivity versus resistance to single conventional and novel agents in vitro and in vivo. Many new agents target the multiple myeloma cell in its bone marrow milieu. These agents specifically target the tumour cell surface such as CD40 and connecting segment-1 (CS1) fibronectin, or are directed at cytokines or receptors such as FGFR3, IGF1, VEGF and BAFFR, whereas small-molecule inhibitors are directed at specific intracellular targets such as MEK, protein kinase C (PKC), cyclin D, NFκB, IκB kinase (IKK), Akt and proteasomes (see Tables 1,2 for a list of the agents that are currently being tested in clinical trials). However, the complexity of constitutive and microenvironmentally induced genetic events in multiple myeloma cells and the bone marrow milieu suggests that combination therapies will be required to increase cytotoxicity and overcome drug resistance. In vivo models allow for preclinical identification and validation of the most promising combination therapies for translation to clinical trials.

One of the first small-molecule inhibitors to be tested in multiple combination strategies was the proteasome inhibitor bortezomib. It has been combined with HSP90 inhibitors, lenalidomide, Akt inhibitors, HDAC inhibitors, BCL2 inhibitors, pro-apoptotic peptides and other proteasome inhibitors in preclinical studies to provide the framework for phase I and II clinical trials (see Table 3 for a list of combination strategies that are currently being tested in clinical trials). For example, gene and protein profiling of bortezomib-treated multiple myeloma cells shows cleavage of DNA repair enzymes, providing the rationale for combining bortezomib with DNA-damaging agents to increase sensitivity or overcome resistance to these conventional therapies. Clinical trials combining bortezomib with akylating agents and anthracyclines have validated the clinical relevance of this strategy. Our microarray profiling shows that bortezomib induces HSP90 gene transcripts in human multiple myeloma cells, and that combining bortezomib with the HSP90 inhibitor 17AAG can block this stress response and increase cytotoxicity. An ongoing clinical trial combining bortezomib and 17AAG demonstrates that the combination can overcome bortezomib resistance. Bortezomib induces apoptosis in vitro primarily through caspase 9, whereas lenalidomide induces apoptosis through caspase 8 (Ref. 167); combined bortezomib and lenalidomide treatment to trigger dual apoptotic signalling can achieve clinical responses even in patients with multiple myeloma resistant to either agent alone. Bortezomib inhibits growth (targeting ERK signalling), survival (targeting Jak and STAT signalling) and migration (targeting PKC signalling), but activates the Akt survival pathway in vitro; combining bortezomib with the Akt inhibitor perifosine abrogates this reponse and induces synergistic multiple myeloma cytotoxicity in vitro. A phase II clinical trial evaluating this combination is ongoing. Importantly, the HDAC inhibitors tubacin or LBH589 (Ref. 148) block protein degradation through the aggresome autophagy pathway and upregulate proteasomal degradation; conversely, blockade of the proteasome with bortezomib upregulates aggresome activity. Preclinical studies have shown that combinations that block both pathways of protein degradation induce synergistic multiple myeloma cell cytotoxicity, and a phase II trial of LBH589 is now ongoing in multiple myeloma, with a combination LBH589 and bortezomib trial to quickly follow.

Lenalidomide, an immunomodulatory drug, has similarly been combined in preclinical studies with several other agents including steroids, mammalian target of rapamycin (mTOR) inhibitors, humanized monoclonal antibodies against CD40 (Ref. 181), a VEGF inhibitor (bevacizumab), and perifosine in order to inform the design of rational combination clinical trials. For example, the addition of dexamethasone, which induces caspase 9-dependent apoptosis, to lenalidomide, which activates caspase 8-dependent apoptosis, triggers dual apoptotic signalling in vitro and improves clinical responses in multiple myeloma. The immunomodulatory effect of lenalidomide markedly upregulates antibody-dependent cell-mediated cytotoxicity (ADCC) induced by humanized monoclonal antibodies in preclinical studies, setting the stage for a combination clinical trial of the humanized anti-CD40 monoclonal antibody and lenalidomide in multiple myeloma. By contrast, multiple combinations found to be antagonistic preclinically (such as perifosine and an HDAC6 inhibitor) are not moving forward to the clinic. These 'bench to bedside' studies confirm the utility of combining therapies to overcome genetic changes that confer tumour cell growth, survival and drug resistance in the bone marrow, and to inform the design of rational combination clinical trials.

Future directions

Oncogenomics studies will allow for the delineation of the molecular sequelae induced when multiple myeloma cells interact with the bone marrow; and will validate the ability of new agents to abrogate these molecular changes and induce tumour cell cytotoxicity in the bone marrow milieu. It will also define those patient subsets that are most likely to respond to individual targeted therapies. For example, microarray profiling studies have identified gene-expression signatures in multiple myeloma cells correlating with response versus resistance to bortezomib. Moreover, correlative studies of patient samples may allow for the identification of biomarkers that are predictive of response. For example, qualitative and quantitative assessment of proteasome inhibition in tumour cells from patients receiving proteasome inhibitor therapies will both determine the extent and type of proteasome inhibition associated with clinical response and allow more selective therapeutics to be developed. Gene and protein profiling before the initiation of therapy and at relapse may provide the basis for combining therapies to overcome resistance (for example, bortezomib and an HSP90 inhibitor), and to improve the outcome of high-dose therapy and stem cell transplantation, as well as for selecting cocktails of targeted therapies for specific patients. Technological advances in high-throughput screening will facilitate the development of more potent and less toxic targeted combination therapies. Ultimately, oncogenomics in the context of clinical trials will allow for both patient selection and rational combination therapeutics to increase cytotoxicity, overcome drug resistance and improve patient outcome in multiple myeloma.