|Home | About | Journals | Submit | Contact Us | Français|
Bone disease in myeloma occurs as a result of complex interactions between myeloma cells and the bone marrow microenvironment. A custom-built DNA single nucleotide polymorphism (SNP) chip containing 3404 SNPs was used to test genomic DNA from myeloma patients classified by the extent of bone disease. Correlations identified with a Total Therapy 2 (TT2) (Arkansas) data set were validated with Eastern Cooperative Oncology Group (ECOG) and Southwest Oncology Group (SWOG) data sets. Univariate correlates with bone disease included: EPHX1, IGF1R, IL-4 and Gsk3β SNP signatures were linked to the number of bone lesions, log2 DKK-1 myeloma cell expression levels and patient survival. Using stepwise multivariate regression analysis, the following SNPs: EPHX1 (P = 0.0026); log2 DKK-1 expression (P = 0.0046); serum lactic dehydrogenase (LDH) (P = 0.0074); Gsk3β (P = 0.02) and TNFSF8 (P = 0.04) were linked to bone disease. This assessment of genetic polymorphisms identifies SNPs with both potential biological relevance and utility in prognostic models of myeloma bone disease.
Multiple myeloma is a tumor of plasma cells that depends on the bone marrow microenvironment for growth and survival.1,2 Bone disease in myeloma occurs as a result of the complex interactions between myeloma cells and the bone marrow osteoclasts, osteoblasts plus other accessory cells and microenvironmental components.2
Myeloma bone disease is characterized by a unique combination of enhanced osteoclast numbers and function plus reduced osteoblast differentiation and function.1–12 The important elements in osteoclast activation are myeloma cellderived MIP-1α, which activates osteoclast CCRX-5 plus microenvironmental-derived RANK-ligand (RANK-L), which activates osteoclast RANK and competes with stromal-derived osteoprotegrin (OPG).10–12 Recent studies have emphasized the central role of the Wnt (Wingless-type MMTV integration site family (mammalian homologue))-signaling inhibitor DKKK-1 in the pathogenesis of the osteolytic bone lesions in myeloma.6 DKK-1 inhibits both osteoblast differentiation and function and increases osteoclast activity. Attention is focused both on the mechanisms responsible for the upregulation of DKK-1 synthesis in plasma cells and the interactions with the microenvironment. 7–10 Expression of DKK-1 is regulated by a combination of intrinsic genomic factors and interactions with the bone marrow microenvironment.8
To assess the predilection to bone disease, it was elected to study the effect of single nucleotide DNA polymorphisms (SNP) in a well-characterized population of myeloma patients for whom DKK-1 expression and gene expression profile (GEP) gene signature data were also available.13 We focused on several pathways involved in the pathogenesis of myeloma bone disease, including the Wnt pathway, in particular GSK3, as well as insulin growth factor, interleukin 4, bradykinin receptors and β3 adrenergic receptors.
Peripheral blood DNA from 282 patients enrolled in the UARK 2003–33 ‘Total Therapy 2’ (TT2) protocol was studied using the previously reported Affymetrix 3k BOAC custom DNA chip to assess the presence or absence of relevant genetic polymorphisms.14–16 Here, we present evidence that several SNPs significantly correlate with both the clinical extent of the bone disease, as well as DKK-1 expression.
These analyses included 282 patients with previously untreated multiple myeloma enrolled in the TT2 trial between October 1998 and February 2004. Details of patient characteristics plus treatment and clinical outcomes have been reported.14 All participants had provided written informed consent in keeping with institutional and National Cancer Institute (NIH, Bethesda, MD, USA) guidelines. All details of the protocol had been approved by institutional guidelines and the United States Food and Drug Administration, and were monitored by a data safety and monitoring board as required for Phase III trials. The multiple myeloma baseline evaluation included serum and urine protein electrophoresis, quantitative immunoglobulin measurements, total 24-h urine protein excretion, serum β2-microglobulin (Sβ2M), C-reactive protein, and lactic dehydrogenase (LDH) plus bone marrow aspirate and biopsy evaluations.
Imaging included baseline magnetic resonance imaging (MRI) and complete skeletal survey radiological examination (myeloma bone survey (MBS)) in a prospective manner.14 The MRI included the axial skeleton and pelvis plus any additional areas requiring diagnostic evaluation for pain or other medical issues. MRI studies were carried out with a series of sequences to permit identification of focal or diffuse bone marrow involvement, including spin echo (T2-wt), short T, inversion recovery (STIR) and gadolinium-enhanced spin echo sequences with and without fat suppression. Myeloma bone survey encompassed the long bones and were carried out with digital radiographs incorporating two views of the chest; views of ribs, lateral skull, vertebral column; anteroposterior views of the pelvis, shoulders; and the extremities including hands and feet.
Focal lesions on both MRI and myeloma bone survey were identified as areas with an axial diameter of at least 0.5 cm. The MRIs were reviewed independently by four individuals who recorded the size, number and location of all focal lesions compatible with myeloma. Full details have been previously published.14
X-ray was the primary classification system for bone disease. The exception was 12 patients with extensive focal MRI disease, but no focal changes on X-ray. On the basis of detailed previous analyses,14 this 4% subset was added to the ‘extensive bone disease’ category to give 183/282 (65%) within this extensive bone disease group. The remaining 99 patients (35%) all had negative X-rays and no extensive focal disease on MRI.
Using X-ray results only, validation of the TT2 findings was conducted comparing results in separate Eastern Cooperative Oncology Group (ECOG) and Southwest Oncology Group (SWOG) data sets.15,16 For these analyses, patients with completely negative X-rays were compared with those having > 3 focal lesions on X-ray.
Peripheral blood was collected in heparinized green top tubes and centrifuged to recover mononuclear cell pellets. DNA was extracted from the mononuclear cell pellets and genotyped using the Affymetrix (Santa Clara, CA, USA) Genchip scanner 3000 Targeted Genotyping System (GCS 3000 TG System) using molecular inversion probes to simultaneously identify the 3404 pre-selected SNPs in 983 genes.15,16 All genotyping experiments were carried out in strict adherence to the manufacturer’s protocol.
A directed, custom SNP chip design was developed with specific criteria from public and commercial databases. Full details are described elsewhere.15,16 In essence, a custom SNP chip was developed, focusing on functionally relevant polymorphisms known to have a role in normal and abnormal cellular functions related to inflammation, immunity and drug responses.
Several methods were used to assess possible correlations between SNPs and the presence or absence of bone disease.
We used Fisher’s exact test as a univariate screening tool to determine the association of SNPs with bone disease. The top 50 rank-ordered SNPs were selected and a recursive-partitioning algorithm was carried out to determine the combination of SNPs that best distinguished the bone disease subgroups. In recursive partitioning, each genotype was evaluated on its ability to make a correct prediction, creating a decision node.17 Recursive partitioning allowed for interactions of SNPs and also included SNPs further down the rank-ordered list. Univariate association between clinical parameters was assessed using continuous and categorical variables.18 The non-parametric Kruskal–Wallace test was used for continuous variables and the χ2-test was used for categorical variables.19 Multivariate logistic regression was used to test for associations of SNPs and clinical parameters with bone disease.20 Survival curves were constructed according to Kaplan and Meier.21
The 282 patients were divided into 99 patients (35%) with no bone disease (X-rays negative) and 183 patients (65%) with definite/extensive bone disease (X-rays positive and/or extensive focal lesions on MRI (12 patients)). This separation best identified the two sub-populations in detailed analyses of the imaging results for the TT2 data set.14
1. Univariate correlations between bone disease and SNPs in TT2 data set: Fisher’s exact test was used as a univariatescreening tool to determine the association of SNPs with bone disease. Results are shown in Table 1, which displays the top SNPs most highly correlated with bone disease. The top-ranked SNP, EPHX1(P = 0.0003), is rs3766934 (GG), which is an expoxide hydrolase SNP. Several SNPs linked to bone biology were among the top-ranked SNPs, including IGFIR (P=0.003: #6), IL-4 (P=0.009: #16) and Gsk3β (P=0.015: #23).
2. Recursive partitioning: The top 50 SNPs with the lowest P-values were selected for recursive partitioning analysis. The results of recursive partitioning analysis are shown in Figure 1. The 4 SNPs providing the best correlation were: rs3766934, EPHX1, RANK #1; rs3783408, Gsk3β, RANK #23; rs1052637, DDX18, RANK #26; and rs3181366, TNFSF8, RANK #29 in the univariate correlations (Table 1). The 4 SNP combination was then used as a search engine to identify further correlations. The results are shown in Figures 2a and b, respectively. There were excellent correlations with both numbers of individual focal bone lesions (P values=0.001) and the directly measured DKK-1 expression levels for individual patients (P=0.05).
3. Stepwise multivariate regression analyses: Several logistic regression models were used to further assess the correlations with the identified SNPs. Results are displayed in Table 2. Again, the previously identified SNPs prove to be statistically significantly associated with the bone disease status. The individual SNPs (EPHX1, Gsk3β and TNSF8), DKK-1 and lactic dehydrogenase (serum level) are predictive in the displayed multivariate analysis.
4. Correlations with progression-free and overall survival: Figure 3 shows the correlations between SNP pattern and outcomes. The cross correlations between known and predicted survivals are highly significant.
Cross-validation in additional clinical data sets with bone disease defined by X-ray only. These statistical analyses used 207 patients with zero or more than three X-ray focal lesions from the original TT2 data set plus 62 patients from Southwest Oncology Group (S9321) and 69 patients from Eastern Cooperative Oncology Group (E1A00 and E9486). Collectively, there were 163 patients with no X-ray evidence of bone disease and 175 patients with more then three focal lesions evident on X-ray. A majority of the SNPs from the TT2 only analyses were again highly ranked in combined analyses. For example, EPHX1 (previously ranked #1, now #9); IGFIR (previously ranked #6, now #13) and IL-4 (previously ranked #16, now #19) again showed significant correlations. Conversely, the SNPs for BDKRB1, ADRB3 and DDX18 were not highly ranked.
Stepwise multivariate logistic regression analysis was then repeated incorporating top SNPs identified by cross-validation. The results are displayed in Table 3. This further cross-validation assessment shows that the EPHX1 SNP is still the top SNP in both the univariate and multivariate regressions. The TREX1 SNP, previously ranked number 11, acquires greater significance in these univariate and multivariate regressions. Other significant correlations were with DDK-1, lactic dehydrogenase, the 17-gene gene expression profile high risk, plus again the SNPs for Gsk3β and TNFSF8.
In this study, several SNPs are correlated with the likelihood of bone disease. The top SNP is EPHX1 (rs3766934: GG genotype versus GT/TT), an epoxide hydrolase. Although EPHX1 has been evaluated in multiple studies of genetic polymorphisms of biotransformation enzymes related to cancer, the functional significance of this specific GG genotype is currently unclear.22 Nonetheless, it is known that epoxide hydrolase is involved in both the inflammatory response linked to the bioactivation of leukotoxins23 and the activation of the dioxin response element by benzo[a] pyrene compounds.24 Further studies are necessary to investigate the potential significance of this EPHX1 SNP genotype in laboratory, clinical and epidemiological studies.
The Gsk3β SNP (Table 1 and Figure 1) was the second SNP selected as part of the recursive partitioning decision tree. This SNP is especially interesting as binding of GSK3βi with axin and APC forms a critical complex involved in Wnt-activated release or stabilization of β-catenin.25–29 This pathway is central to osteoblast function.30 Increased Wnt signaling through Wnt 3A results in an increase in the bone mineral density and a decrease in the osteoclast/osteoblast ratio.31–34 Gsk3 β is the target of upregulation by thalidomide and is central to reactive oxygen species-mediated thalidomide-induced apoptosis.28
Other identified SNPs linked to bone related pathways (see Table 4) included the following: insulin-like growth factor 1 receptor (ranked number 6: Table 1);35–39 bradykinin receptor B1 (ranked number 10: Table 1);40 adrenergic receptor B3 (ranked number 14: Table 1);41,42 and interleukin-4 (ranked number 16: Table 1).43 Several SNPs linked to drug and/or toxin metabolism and/or DNA metabolism and repair were noted and are summarized in Table 4. As dioxins have been linked to the etiology of myeloma,44 it is noteworthy that EPHX145–47 is important in dioxin and polycyclic aromatic hydrocarbon metabolism. In addition, the DPYD SNP (rs1399291) ranked number 28 (Table 1) is involved with pyrimidine metabolism, and has, in addition, been identified in a separate recent largescale screening.48
Testing with the 3400 SNP custom chip has, thus, revealed several SNPs that are significantly correlated with the likelihood of bone disease in patients with myeloma. Larger studies are currently underway, for example, in collaboration with the National Cancer Institute (NCI) epidemiology branch, to further explore the relationships with identified SNPs.48
This investigation was supported in part by an unrestricted grant from the International Myeloma Foundation (Bank on a Cure project), as well as by the following PHS Cooperative Agreement grant numbers awarded by the National Cancer Institute, DHHS: CA32102 and CA38926 (SWOG); and CA21115 (ECOG); plus CA 97513 (JDS and BB).
Conflict of interest
The authors declare no conflict of interest.