SUPPLEMENTARY INFORMATION
Patients
Of the 288 patients with samples available, 54 patients were randomly selected for discovery analyses and tested for serum miRNAs that were consistently expressed across this subset. The remaining 234 patients (117 from each of the two treatment arms of the protocol) constituted the validation subset and were investigated for specific miRNA levels selected through the discovery set analysis. Patients’ baseline clinical and cytogenetic characteristics are reported in Table A1. Overall, no significant differences were observed in the frequency and distribution of the clinical, cytogenetic, and biochemical characteristics between the subsets of patients included in these analyses versus those in the same protocol who did not have material available for analysis (data not shown) and therefore were excluded. Median progression-free survival (PFS) (medians: 2.3 vs 2.5 years; p=0.4) and overall survival (OS) (medians: not reached vs. 5.1 years; p=0.9) were also similar between the included and the excluded patients.
Treatment response and progression was assessed using the International Myeloma Working Group Uniform Response Criteria1. PFS was defined as time from enrollment until the date of progression, relapse or death from any cause. OS was defined as time from enrollment until the date of death from any cause. Any patients who were event-free at their last evaluation were censored at that time.
Fluorescent in Situ Hybridization (FISH) analyses were performed as part of the phase III trial, using purified CD138+ cells obtained from bone marrow at diagnosis following standard procedures, as previously reported2. If 10% or more cells presented a deletion, or 15% or more cells carried a translocation, the sample was considered positive for that chromosomal abnormality. Patients with del17p13, t(4;14), or t(14;16) were classified as high risk3.
Healthy subjects were enrolled as control if they had no monoclonal gammopathy of undetermined significance, nor past or present diagnosis of any neoplastic disease. All MM patients and healthy subjects included in these analyses were enrolled on IRB approved protocols and provided informed consent.
Serum RNA extraction
Briefly, samples stored at -80°C were thawed and the total RNA was extracted from 500ul of serum with mirVana paris kit (Ambion) following the manufacturer’s instructions. To increase the RNA recovery 1µg of RNA carrier (#4382878, Ambion) was added. To normalize the extraction efficiency among different samples, we spiked-in hsa-miR-759 (5pm) in each serum sample, which we previously observed by using Nano-string count and qRT-PCR in a training samples of newly diagnosed MM patients (30) to be absent. Concentration and quality of RNA was assessed using NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA).
miRNA analysis
RNA isolated from 500µl of serum (250 ng) obtained from randomly selected 54 MM patients was analyzed by NanoString assay performed as described by the manufacturer (NanoString Technologies, Inc, Seattle, Washington, USA.)4. NanoString raw data, which were proportional to copy number, were log-transformed and normalized by the quantile method. P-values were used to rank miRNAs of interest, and correction for multiple comparisons was done using the Benjamini-Hochberg method5. We selected for confirmation analysis only those miRNAs consistently expressed over the background threshold in at least 20% of samples. Stem loop real-time PCR (qRT-PCR) was performed using the TaqMan PCR Kit (Applied Biosystems, Foster City, CA) and the Applied Biosystems 7900HT Sequence Detection System (Applied Biosystems) according to manufacturer’s instructions. The miRNAs were analyzed using a log2 transformation, and all analyses described utilized these transformed data.
Approximately 10 ng of RNA was used in each reaction. TaqMan probes for miRNA quantification were purchased from Applied Biosystems (Life Technologies, Grand Island, NY, USA). All qRT-PCR analyses were performed at least in triplicate and included in the analysis. Only markers with SD among replicates p-value <0.001 were considered for further analysis. Synthetic miR-759 spiked into each sample was used as internal control. The comparative cycle threshold (Ct) method for relative quantification was used to determine miRNA expression levels.
Data were filtered to exclude relatively invariant miRNAs using InterQuantileRange (IQR), a measure of statistical dispersion, and miRNAs below the detection threshold (defined for each sample by a cutoff corresponding to twice the standard deviation [SD] of negative control probes) in at least half of the samples were excluded from the analysis. Using R/Bioconductor and the filtered dataset, linear models for multidimensional data analysis6 were employed with a contrast matrix for the comparisons.
Statistical analysis
Variable selection was also assessed using the leaps and bounds method of Furnival and Wilson10, where we examined best subsets based on the score (Cp) statistic to identify the best overall model in terms of prognosis for PFS or OS. In addition, to accommodate the potential for inherent differences in PFS and OS based on treatment received, all multivariable models for these clinical outcomes were stratified by treatment arm in the models. Given the nature of this study, multiple comparison corrections were not utilized when assessing prognostic utility for clinical outcomes.
Spearman rank correlation coefficients were calculated between significant miRNA in the multivariate setting and other continuous clinical variables at diagnosis (age, lactate dehydrogenase (LDH), % of bone marrow plasma cells (BMPC), ẞ2-microglobulin (β2m), C reactive protein (CRP), albumin, hemoglobin, creatinine). Differential expression of significant miRNA in the multivariate setting and categorical variables [treatment arm, gender, ISS stage I-II vs. II, del13, del17, t(4;14), t(11;14), t(14;16)] was evaluated using two-sample t-tests or the non-parametric Wilcoxon rank sum test, dependent on distribution.
In addition, the clinical and molecular factors described above were also evaluated in the univariate setting in relation to PFS and OS, where log-rank tests were used to determine statistical significance.
All analyses were conducted using the statistical program R (version 2.15.2 GUI 1.53 Leopard build 64-bit), and statistical significance was declared if p<0.05.
miRs and del13
We discovered a significant relationship between deletion 13 and OS (HR 1.76, p=0.03). To further define the impact of deletion 13, Cox regression analyses for overall survival were performed dividing up the entire trial population into three groups – one group with no high risk genetic features (as defined above) or deletion 13 by FISH analysis of the patient’s diagnostic bone marrow biopsy, one group with only deletion 13 but no high risk genetic features, and one group with high risk genetic features regardless of the presence or absence of deletion 13 (Figure A2). The presence of deletion 13 alone had borderline significance from the base group (p=0.057), and the presence of high risk genetic features predicted significantly worsened OS (p=0.012). We were surprised to find that Del13 was significant in a multivariable model for overall survival7, but we suspect that the presence of Del13 was significant as it is often a passenger with other higher risk mutations that were not tested in this trial such as 1q amplification, 1p deletion, and karyotype abnormalities.
miRs and Hemoglobin
Several reports have indicated that circulating miRNAs may largely originate from blood cells, thus their expression reflects the effect of cancer on blood cells and not the cancer itself8, 9. When we tried to correlate hemoglobin levels and the level of these miRNAs, including the expression of S-miR-16 and S-miR-451 that have been previously reported to be highly expressed in blood cells8, 10, we found a weak but statistically significant relationship (Figure A4). These observations strongly suggest that additional analyses must be performed to better understand the source of differentially expressed miRNA in MM patients. It is possible that the immune system plays a role in modifying expression levels of miRNAs, and this interplay of mechanisms translates to better prognosis for MM patients. In this trial, collinear circulating miRNAs reflected overall survival much better than PFS, suggesting that they may be a more accurate indicator of patient health than disease aggressiveness.
Other circulating miRNA reports
Several attempts have been made to classify MM patients in distinct risk groups based on miRNA expression profiling, to date these efforts have produced no convincing results11. This is likely due to the relatively small sample sizes in the face of significant clinical and cytogenetic heterogeneity. While small studies have supported the use of a specific set of miRNAs such as S-miR-720 and S-miR-92a as biomarkers in MM pathogenesis12, 13, no large group of newly diagnosed patients uniformly treated has been analyzed for the prognostic role of circulating miRNAs.
Supplementary Information figure legends
Figure A1. CONSORT Diagram. Sample availability from the parent clinical trial and analysis done in this study. Discovery and miRNA profiling was conducted using NanoString nCounter Technology on 54 serum samples randomly selected from the parent clinical trial. Validation of miRNA identified in discovery subset was performed on an expanded set of the remaining 234 patients from the same parent clinical trial.
Figure A2. Deletion 13 and Overall Survival. Using Cox regression analysis, three groups were defined based on FISH results that included no deletion 13 (green), deletion 13 alone (blue), and high risk markers that included only del17, t(4;14), and/or t(14;16) regardless of deletion 13 (red).
Figure A3: Del13 vs log miRNA values. miRNA log values for S-miR-16, 19b, -25, and -30a with and without deletion 13 by FISH using Wilcoxon Rank Sum test.
Figure A4: Hemoglobin vs. log miRNA values. miRNA log values for S-miR-16, -19b, -25, and -30a including p-values and Spearman correlation coefficients.
Supplementary Information Figures
Figure A1
Figure A2
Figure A3
Figure A4
Supplementary Information Tables
Table A1. Patient characteristics
MM Discovery SetNanoString Count / MM Expanded Set
qRT-PCR / Healthy
qRT-PCR
Patient number / 54 / 234 / 67
Age
Median (range) / 71 (56-79) / 70 (55-86) / 73 (55-88)
Gender
Male, % / 24 (44%) / 117 (50%) / 31 (46%)
Female, % / 30 (56%) / 117 (50%) / 36 (54%)
B2M (mg/L)
Median (range) / 5.1 (1.2 - 13.0) / 4.0 (0.2 - 24.8)
≤ 3.5 / 34 / 74
>3.5 / 11 / 120
Data missing / 9 / 40
Albumin (g/L)
Median (range) / 3.8 (2.4 - 5.0) / 2.7 (1.3 - 5.1)
Data missing / 5 / 30
ISS
1 / 9 (20%) / 53 (28%)
2 / 17 (39%) / 77 (41%)
3 / 18 (41%) / 57 (31%)
Data missing / 10 / 47
Creatinine (mg/dl)
Median (range) / 1.0 (0.6 - 2.5) / 1.0 (0.5 -2.5)
Data missing / 0 / 0
Hemoglobin (g/dL)
Median (range) / 10.0 (7.2 - 15.1) / 10.5 (5.9 - 15.7)
Data missing / 3 / 22
LDH (U/L)
Median (range) / 297 (106 – 1283) / 280 (106 – 1283)
Data missing / 5 / 36
FISH abnormalities
del13 / 28 (54%) / 98 (52%)
del17 / 7 (13%) / 29 (15%)
t(4;14) / 11 (21%) / 38 (20%)
t(11;14) / 9 (17%) / 31 (16%)
t(14;16) / 4 (8%) / 9 (5%)
High risk / 18 (34%) / 60 (32%)
Treatment Arm
VMP / 33 (61%) / 117 (50%)
VMP-VT / 21 (39%) / 117 (50%)
Abbreviations: B2M - ẞ2-microglobulin; ISS - International Staging System; LDH - lactate dehydrogenase; FISH - Fluorescent in-situ hybridization; VMP - bortezomib-melphalan-predinisone; VMPT-VT - borteomib-melphalan-predinisone-thalidomide followed by maintenance therapy with bortezomib-thalidomide
Table A2. Median qRT-PCR cycles required to detect S-miRNA expression
miRNA ID / Detected in MM # samples (out of 54) Nano-String Analysis (%) / median Ct valuemiR-92a / 100 / <30
miR-718 / 98 / >33
miR-30a / 98 / <32
miR-451 / 98 / <30
miR-563 / 95 / >33
miR-1283 / 79 / >33
miR-21 / 79 / <30
miR-16 / 73 / <30
miR-126 / 73 / <32
miR-720 / 63 / <32
miR-181a / 62 / >33
miR-423-5p / 56 / High variability
miR-181d / 50 / >33
miR-526 / 47 / >33
miR-25 / 46 / <30
miR-455-3p / 43 / >33
miR-223 / 41 / <32
miR-1323 / 40 / >33
miR-19b / 38 / <32
miR-144 / 35 / >33
miR-1975 / 30 / >33
miR-885-5p / 28 / >33
miR-495 / 25 / >33
miR-1206 / 23 / >33
miR-934 / 22 / >33
Table A3. Relationships between the 10 differentially expressed continuous miRNA markers and clinical factors
Characteristics / miRs-92a / -451 / -19b / -21 / -30a / -223 / -16 / -25 / -126 / -720
Treatment arm / 0.92 / 0.39 / 0.08 / 0.39 / 0.44 / 0.24 / 0.044 / 0.13 / 0.70 / 0.27
Gender / 0.74 / 0.40 / 0.93 / 0.72 / 0.14 / 0.34 / 0.83 / 0.23 / 0.82 / 0.47
Age / 0.035 / 0.16 / 0.036 / 0.35 / 0.06 / 0.08 / 0.17 / 0.01 / 0.013 / 0.34
ISS stage
(1-2 vs. 3) / 0.19 / 0.23 / 0.12 / 0.30 / 0.022 / 0.05 / 0.25 / 0.11 / 0.40 / 0.38
LDH / 0.58 / 0.40 / 0.032 / 0.11 / 0.51 / 0.84 / 0.29 / 0.36 / 0.88 / 0.51
%BMPC / 0.90 / 0.54 / 0.43 / 0.52 / 0.52 / 0.09 / 0.10 / 0.25 / 0.50 / 0.52
B2M / 0.38 / 0.17 / 0.19 / 0.25 / 0.20 / 0.01 / 0.52 / 0.14 / 0.47 / 0.44
CRP / 0.25 / 0.81 / 0.86 / 0.58 / 0.30 / 0.70 / 0.96 / 0.96 / 0.68 / 0.76
Albumin / 0.67 / 0.82 / 0.44 / 0.47 / 0.87 / 0.38 / 0.91 / 0.90 / 0.53 / 0.93
Hemoglobin / 0.22 / 0.01 / 0.001 / 0.11 / 0.15 / 0.004 / 0.02 / 0.03 / 0.10 / 0.40
Creatinine / 0.59 / 0.20 / 0.06 / 0.30 / 0.01 / 0.003 / 0.18 / 0.03 / 0.63 / 0.59
del13 / 0.05 / 0.10 / 0.01 / 0.07 / 0.20 / 0.36 / 0.003 / 0.09 / 0.38 / 0.90
del17 / 0.96 / 0.42 / 0.89 / 0.34 / 0.66 / 0.56 / 0.83 / 0.51 / 0.69 / 0.75
t(4;14) / 0.33 / 0.71 / 0.31 / 0.60 / 0.02 / 0.19 / 0.30 / 0.38 / 0.31 / 0.98
t(11;14) / 0.38 / 0.48 / 0.96 / 0.64 / 0.15 / 0.25 / 0.88 / 0.84 / 0.59 / 0.21
t(14;16) / 0.78 / 0.93 / 0.30 / 0.11 / 0.08 / 0.69 / 0.85 / 0.50 / 0.96 / 0.49
High risk FISH / 0.41 / 0.82 / 0.11 / 0.94 / 0.13 / 0.30 / 0.51 / 0.20 / 0.88 / 0.88
Abbreviations: % BMPC - % of bone marrow plasma cells; CRP - C reactive protein
Table A4. Results from analyses of characteristics and miRNAs in univariate models for PFS and OS, where models were also stratified on treatment arm to adjust for possible inherent differences based on treatment arm.
Charateristic / PFS HR / PFS p-value / OS HR / OS p-valueVMPT (vs. VMP) / 0.67 / 0.013 / 0.83 / 0.41
Del13 / 1.24 / 0.23 / 1.76 / 0.03
Del17* / 2.12 / 0.005 / 2.13 / 0.02
t(4;14) / 1.09 / 0.70 / 1.21 / 0.54
t(11;14)* / 0.67 / 0.10 / 1.22 / 0.54
t(14;16) / 0.49 / 0.13 / 1.02 / 0.97
Creatinine** / 1.30 / 0.14 / 1.89 / 0.03
ISS stage* / 1.25 / 0.06 / 1.42 / 0.039
Male gender / 1.03 / 0.84 / 1.80 / 0.014
miR-92a / 0.99 / 0.80 / 0.93 / 0.19
miR-451 / 0.97 / 0.51 / 0.93 / 0.18
miR-19b / 0.96 / 0.47 / 0.88 / 0.08
miR-21 / 1.03 / 0.66 / 1.01 / 0.87
miR-30a** / 0.97 / 0.49 / 0.86 / 0.016
miR-223 / 0.97 / 0.57 / 0.93 / 0.22
miR-25 / 0.92 / 0.034 / 0.81 / 0.0005
miR-16 / 0.94 / 0.13 / 0.87 / 0.008
miR-126* / 0.96 / 0.45 / 0.91 / 0.17
miR-720** / 0.92 / 0.077 / 0.87 / 0.06
*failed assumptions of proportional hazards and adjusted for non-proportionality of hazards either by hypothesis test or by graphical interpretation of the Shoenfeld residuals