Medina et al. Blood Cancer Journal (2020)10:108 https://doi.org/10.1038/s41408-020-00377-0 Blood Cancer Journal ARTICLE Open Access Comparison of next-generation sequencing (NGS) and next-generation flow (NGF) for minimal residual disease (MRD) assessment in multiple myeloma Alejandro Medina 1, Noemi Puig1, Juan Flores-Montero 2, Cristina Jimenez1, M.-Eugenia Sarasquete 1, María Garcia-Alvarez1, Isabel Prieto-Conde1, Carmen Chillon1, Miguel Alcoceba1, Norma C. Gutierrez1, Albert Oriol3, Laura Rosinol4, Joan Bladè4, Mercedes Gironella5, Miguel T. Hernandez6, Veronica Gonzalez-Calle1, Maria-Teresa Cedena7, Bruno Paiva 8, Jesus F. San-Miguel 8, Juan-Jose Lahuerta 7, Maria-Victoria Mateos 1, Joaquin Martinez-Lopez 7, Alberto Orfao2, Marcos Gonzalez1 and Ramon Garcia-Sanz 1 Abstract Detecting persistent minimal residual disease (MRD) allows the identification of patients with an increased risk of relapse and death. In this study, we have evaluated MRD 3 months after transplantation in 106 myeloma patients using a commercial next-generation sequencing (NGS) strategy (LymphoTrack®), and compared the results with next- generation flow (NGF, EuroFlow). The use of different marrow pulls and the need of concentrating samples for NGS 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; biased the applicability for MRD evaluation and favored NGF. Despite that, correlation between NGS and NGF was high (R2 = 0.905). The 3-year progression-free survival (PFS) rates by NGS and NGF were longer for undetectable vs. positive patients (NGS: 88.7% vs. 56.6%; NGF: 91.4% vs. 50%; p < 0.001 for both comparisons), which resulted in a 3-year overall survival (OS) advantage (NGS: 96.2% vs. 77.3%; NGF: 96.6% vs. 74.9%, p < 0.01 for both comparisons). In the Cox regression model, NGS and NGF negativity had similar results but favoring the latter in PFS (HR: 0.20, 95% CI: 0.09–0.45, p < 0.001) and OS (HR: 0.21, 95% CI: 0.06–0.75, p = 0.02). All these results reinforce the role of MRD detection by different strategies in patient prognosis and highlight the use of MRD as an endpoint for multiple myeloma treatment. Introduction transplantation (ASCT)4 have increased the proportion of Multiple myeloma (MM) is a B-cell malignancy where patients achieving complete response (CR) in this disease, plasma cells arise from a single clone accumulated in the which has been translated into better progression-free sur- bone marrow1, and usually produce monoclonal immu- vival (PFS) and overall survival (OS) rates during the last 10 noglobulins (M-protein or monoclonal component). New years5–7. This improvement led to the definition of new agents such as proteasome inhibitors and immunomodula- response categories (i.e. stringent complete response, sCR)8, tory drugs2,3, together with autologous stem cell but even myeloma patients with such good responses still relapse due to the presence of residual tumor cells in the bone marrow (i.e. minimal residual disease, MRD)9–11 that Correspondence: M.-Eugenia Sarasquete (mealonsos@saludcastillayleon.es) remain undetectable by conventional serological and mor- 1 Departamento de Hematología, Hospital Universitario de Salamanca (HUSA/ phological methods. New methods are capable of detecting IBSAL), CIBERONC, CIC-IBMCC (USAL-CSIC), Salamanca, Spain 2 MRD with a higher prediction ability, since negative patients Centro de Investigación del Cáncer-IBMCC (USAL-CSIC), CIBERONC, Salamanca, Spain had longer survival rates compared to positive patients12–14, Full list of author information is available at the end of the article © The Author(s) 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 2 of 10 even when only patients in CR are considered15. Accord- Materials and methods ingly, the International Myeloma Working Group (IMWG) Patient and sample selection developed new criteria to define MRD negative responses, Patient selection was based on the following criteria: characterized by the absence of clonal malignant plasma (i) newly diagnosed MM, transplant-candidate patients cells in patients with suspected CR, assessed with a sensi- included in the Spanish GEM2012 clinical trial7, whose tivity of at least 10−5—one malignant cell per hundred clonotypic rearrangements were previously identified, thousand normal cells16; therefore, only very sensitive and (ii) NGF evaluation performed at diagnosis and follow- reliable methods can be applied to detect MRD. up per protocol, and (iii) enough amount of gDNA Nowadays, MRD in MM can be studied using two dif- available for MRD studies. Bone marrow samples from ferent approaches: immunophenotypic (multiparametric eligible MM patients were collected at day 100 follow- flow cytometry, MFC)17 and molecular techniques (next- ing transplantation: two independent pulls were generation sequencing —NGS—, digital PCR)18. Imaging obtained and subsequently processed to perform flow tools (PET-CT, MRI) can add some extra information in and molecular studies. terms of prognostic evaluation19, however, standardized Patients with t(4;14), t(14;16), or del17p by FISH were procedures to perform MRD assessment in myeloma have grouped together in a high-risk cytogenetic subgroup only been developed for MFC20 and NGS21. Despite the according to the IMWG criteria26. The revised interna- added value for MRD assessment, next-generation stra- tional staging system (R-ISS)27 was also applied to stratify tegies are currently used only in a few centers and usually patients according to the ISS stage, serum LDH, and limited to the context of clinical trials. cytogenetics. The study was conducted in accordance MFC allows the identification and quantification of with Declaration of Helsinki principles, was approved by abnormal plasma cells based on an aberrant protein- the Committee of the University Hospital of Salamanca, marker expression profile displayed by myeloma cells. and written informed consent was required and obtained Recently, the EuroFlow consortium has developed a two- from each patient prior to their inclusion. tube, eight-color flow assay that allows the simultaneous analysis of up to 10 million cells (next-generation flow, DNA extraction and quantification NGF)22; NGF has been already approved by the IMWG as gDNA was isolated from bone marrow aspirates of a reference method to detect and define immunopheno- newly diagnosed MM patients using the automated DNA typic CRs in MM16. NGF reaches a sensitivity of 2 × 10−6, Purification kit Maxwell® (Promega, Madison, WI, USA). overcoming previous flow protocols (10−4–10−5), but it is gDNA quality was first assessed using NanoDrop2000 highly dependent on the precise identification of the (ThermoFisher, Waltham, MA, USA). Samples were pathologic immunophenotype, requiring a high level of quantified using Qubit 2.0 and the dsDNA BR assay expertise; in addition, the quality of the sample must be (ThermoFisher, Waltham, MA, USA). high and should be promptly processed23. To improve the quality of follow-up samples, those with Molecular methods use the clonal immunoglobulin gene insufficient DNA concentration for MRD purposes (i.e. rearrangement as target for the detection of MRD levels. those with <100 ng/µl) were ethanol-precipitated. To do Quantitative PCR uses allele-specific oligonucleotides this, 1/10 volume sodium acetate, as well as twice the targeting the junction region of the immunoglobulin sample volume of 100% ethanol (stored at −20 °C) were genes; it has been extensively used as the molecular gold- added to the samples and incubated overnight at −20 °C. standard approach to detect MRD levels in lymphoid Afterwards, samples were centrifuged at 17,900 × g and malignancies24,25. Nonetheless, it is a labor-intensive 4 °C for 10 min. The nucleic acid pellet was then washed technique, requiring the construction of a standard curve with 500 µl ethanol (70%) and centrifuged again at for every single patient, and its applicability accounts only 17,900 × g and 4 °C for 5 min. Finally, the pellet was dried for 40–75% of myeloma patients24,25. By contrast, NGS and rehydrated in ≈12 µl of water, and quantified using strategies have been rapidly introduced in the clinical the Qubit dsDNA BR assay. setting, especially for myeloma. However, the only strategy approved by the IMWG16 and cleared by the FDA — Gene amplification and sequencing Adaptive’s clonoSEQ®— is not commercially available, The LymphoTrack® IGH panel (Invivoscribe Tech- and its cost makes it unaffordable for most centers. nologies, San Diego, CA, USA) was used for the analysis We aimed at validating the potential applicability and of MRD samples to detect previously characterized clo- usefulness of a new NGS methodology for the diagnosis notypic rearrangements of the immunoglobulin heavy and MRD detection in MM patients. To that end, we have chain loci (IGH). Clonal rearrangements were first evaluated a commercial NGS panel, LymphoTrack®, in a determined in matched baseline samples by PCR ampli- cohort of 106 myeloma patients, and compared the results fication and Sanger sequencing using the BIOMED-2 with NGF. primers as previously described28. Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 3 of 10 Briefly, this commercial NGS strategy uses primers developed NGF methodology (Table S1), following the targeting the immunoglobulin framework regions in order EuroFlow guidelines, as described elsewere20,22. Events to amplify V(D)J rearrangements. In one-step PCR (Fig. from two eight-color tubes (per sample) were merged S1) amplicons are generated and one-side indexed, using the merge function of the INFINICYT™ allowing the simultaneous sequencing of up to 24 samples v2.0 software (Cytognos S.L. Salamanca, Spain). A sample in a single run. Whenever possible, ≥650 ng of DNA were was considered positive when at least 20 aberrant plasma used to reach a sensitivity level of at least 10−5 (assuming cells were detected. Hemodilution of bone marrow sam- 6.5 pg of DNA per cell, input for NGS would be at least ples for NGF evaluation was assessed through the iden- 100,000 cells). In addition, DNA from one clonal, well- tification of a significant decrease in non-plasma cell characterized B cell line was added in each reaction as a populations: mast cells (CD117hi), erythroblasts (CD45−/ control spike-in (corresponding to 100 cells), to allow the sideward-scatterlo) and B-cell precursors (CD19+/ absolute quantification of tumor plasma cells. CD38hi/CD45lo). The complete analysis of NGF perfor- After a purification step with Agentcourt AMPure XP mance for the GEM2012 trial has already been microbeads (Beckman Coulter Inc., Brea, CA, USA) and published28. 70% ethanol, and purity and quantity assessment using the TapeStation 4200 (Agilent, Santa Clara, CA, USA) and Statistical analysis KAPA library quantification kit (KAPA Biosystems, Bos- Patients’ characteristics were analyzed with the SPSS ton, MA, USA) or Qubit 2.0, respectively, amplicon 23.0 software (IBM, Armonk, NY, USA) using Fisher’s libraries of 12–20 pM were prepared. These libraries were exact test for discrete variables and the Mann–Whitney later sequenced in a MiSeq platform (Illumina, San Diego, test for continuous variables. Bland–Altman plots were CA, USA) using v3 reagent kits and 2 × 251 sequencing used to test the potential agreement between methods. cycles, aiming at one million reads per sample. The Kaplan–Meier method and the log-rank test were used to plot and compare PFS and OS curves. Cox Sequence analysis and MRD evaluation regression was used to perform univariate and multi- Resulting FastQ files were processed using the Lym- variate analyses. By Landmark analysis, only those patients phoTrackAnalysis® software and the LymphoTrack® who had not progressed or died on day 100 post ASCT MRD Data Analysis tool 1.1.0 (Invivoscribe Technologies) were evaluated. PFS was defined as the time from MRD to retrieve sequences from virtually every clonal B cell of assessment to the last follow-up visit, disease progression the samples. This allowed the identification of residual or death by any cause. OS was defined as the time from tumor cells, if present, by tracking their clonotypic IGH MRD assessment to the last follow-up visit or patient’s complementarity-determining region 3 (CDR3) that had decease by any cause. All reported p values were obtained previously been characterized. The software performs a by a two-sided exact method, at the conventional 5% calculation considering the number of spike-in cells used significance level (p < 0.05). for each reaction, as well as tumor and spike-in read counts. Sequencing results were considered invalid when Results <20,000 total reads were obtained. A sample was con- Patient characteristics sidered positive for MRD assessment when at least two One hundred and six (106) patients met the inclusion identical clonotypic reads were detected; conversely, a criteria. Clinical variables of our cohort are described in sample was considered negative (undetectable) for MRD Table S2. Median age at diagnosis was 59 years. Male to when criteria for positivity were not met. Vidjil (Uni- female proportion was 58/42. Patients with high-risk versity of Lille) and Arrest/Interrogate (EuroNGS) tools cytogenetics represented 22.5% of the present series [n = were also used to analyze FastQ files from MRD 23/102, one of them with t(4;14) plus del17p]. The R-ISS sequencing. stages I, II, and III represented 25.2%, 59.2%, and 15.5% of In order to make comparisons, patients were eventually patients, respectively. After a median follow-up of classified into different groups according to the MRD 39.5 months from MRD assessment (interquartile range: status (MRD-positive and MRD-negative), the MRD level 33.8–46.2 months), 31 patients had relapsed (29.2%) and (MRD-negative, MRD-positive below 10−5, MRD-positive 16 had died (15.1%), two of them without progression, due between 10−5 and 10−4, and MRD-positive higher than to infections. 10−4), the conventional response (VGPR/PR and sCR/ At the corresponding MRD evaluation time point, the CR), and the cytogenetic risk (high- and standard-risk). overall response rate was 96.2%: 32.1% of patients achieved sCR, 23.6% CR, 34.0% very good partial response MFC studies (VGPR) and 6.6% partial response (PR). Only four patients Follow-up samples were processed within 24 h after (3.8%) had stable/progressive disease (SD/PD). Three-year collection. Analysis was carried out using the recently PFS and OS rates were 72.5% and 86.7%, respectively. Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 4 of 10 Table 1 Cell input for NGS- and NGF-based MRD studies. Table 2 MRD stratification for next-generation flow and next-generation sequencing. Input cells NGS NGF MRD level Current study GEM20127 5 a <10 10.4% 0% 105 − 2 × 105 32.1% 0% NGS NGF NGF 5 >2 × 10 57·5% 100% Negative 53 (50.0%) 58 (54.7%) 173 (54.6%) 6 >10 0% 100% <10 −5 16 (15.1%) a 9 (8.5%) 18 (5.7%)a Median cell input 210,672 9,200,276 ≥10 −5 −4 and <10 17 (16.0%) 21 (19.8%) 65 (20.5%) [Interquartile range] [136,382–264,110] [5,702,369–10,305,207] −4 ≥10 20 (18.9%) 18 (17.0%) 61 (19.2%) Median sensitivity 10−5 2 × 10−6 Total 106 106 317 Cell input range 3542–726,264 1,548,175–15,000,000 MRD results, obtained using NGS or NGF, were compared for the 106 patients in Each cell shows the frequency of samples that were analyzed at different cell our cohort. Results obtained in the GEM2012 Spanish trial,7 (intent-to-treat input levels. The number of cell equivalents used for NGS was calculated based population) were used as a reference. Absolute numbers and percentage (in on Qubit2.0 quantification, assuming 6.5 pg DNA/cell. parenthesis) for each MRD category are shown in each cell. a One sample was sequenced with a cell input below 10,000 cells. MRD minimal residual disease, NGF next-generation flow, NGS next-generation sequencing. a Fisher’s exact test showed statistical differences in the proportions of each subgroup (two-sided p = 0.004). Comparison of NGS and NGF performance Sample quality and ethanol precipitation were the main variables affecting NGS applicability for MRD analysis. Because of that, the minimum number of cell equivalents intermediate ISS status and LDH was normal. Of note, to reach a virtual sensitivity of 10−5 could only be used in two of these patients had extramedullary disease (one 95 (89.6%) cases (Table 1). According to the inclusion already present at diagnosis), while another one relapsed criteria, NGF was successfully used for the evaluation of after more than 3 years in sustained MRD negativity, one the entire series, with a significantly higher number of year following their treatment suspension. The other two cells required to perform the studies at the aimed sensi- patients had conventional disease progressions tivity threshold. The analysis of precursor B cells, ery- ~36 months after transplantation. throblasts, and mast cells did not reveal significant hemodilution of any sample evaluated by flow cytometry. Undetectable MRD was clinically significant by NGS and MRD-negativity was achieved by 53 (50.0%) and 58 NGF (54.7%) patients by NGS and NGF, respectively. Further Negative patients showed a significantly better 3-year stratification of patients according to the MRD level is PFS rate than positive patients (p < 0.001), either basing shown in Table 2. on NGS or NGF results (88.7% vs. 56.6%; 91.4% vs. 50%, As it is shown in Fig. 1A, we found a good correlation Fig. 2A and B, respectively, p < 0.001). Three-year OS between methods (r = 0.951; R2 = 0.905), with 15 dis- rates were very high for all subgroups: considering NGS, cordant cases (5 NGF+/NGS−; 10 NGF−/NGS+). It is 3-year OS rates of negative and positive patients were worth mentioning that the majority of NGS-only MRD- 96.2% and 77.3%, respectively (Fig. 3A; p < 0.01). Con- positive cases (7/10) presented values below 10−5, while sidering NGF, 3-year OS rates of negative and positive this situation was found in two out of five NGF-only patients were 96.6% and 74.9%, respectively (Fig. 3B; p < MRD-positive cases. Bland–Altman analysis demon- 0.01). There were no statistical differences in PFS between strated an excellent level of agreement between NGS and double-negative and discordant cases (36-month PFS: NGF (Fig. 1B). 91.7% and 80%, respectively, p = 0.55, Fig. S2), while both subgroups had significantly better PFS rates than double- Characteristics of patients with undetectable MRD who negative cases (median PFS: 34.2 months, p < 0.05 for had disease progression both comparisons). With a median follow-up of 39.5 months, disease pro- Moreover, survival rates progressively increased from gression was confirmed in 31 cases: 23 double-positive patients with high-positive MRD to patients with unde- patients, three discordant cases (two positive cases by tectable residual disease (Fig. S3): by NGS, 36-month PFS NGF, one positive case by NGS), and five patients with rates of ≥10−4, 10−5 − 10−4, <10−5 and negative groups double-negative MRD (by both NGS and NGF). were 55%, 52.9%, 62.5%, and 88.7%, respectively (p < 0.01). Table S3 lists the characteristics of the five patients who By NGF, they were 38.9%, 47.6%, 77.8%, and 91.4%, had undetectable MRD and progressed: all of them respectively (p < 0.001). OS rates of these subsets followed achieved CR at the time of MRD assessment, had low/ the same trend. Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 5 of 10 Fig. 1 Comparison of MRD results. A Linear regression. Ninety-one out of one hundred and six samples (91/106, 85.8%) were concordant between techniques. Only 15/106 cases (14.2%) were discordant. B Bland–Altman plot comparing NGS and NGF performance (n = 43, only double-positive cases were evaluated). Mean MRD values of methods (shown in the x-axis) were calculated. Differences in log10 scale for each case (y-axis) were calculated as follows: log10(higher MRD value/lower MRD value). Then, negative values were assigned to those cases where the MRD level estimated by NGS > NGF, while positive values were assigned to those case where the MRD estimated by NGS < NGF. Normal distribution of the differences was first assessed (Kolmogorov–Smirnov’s p > 0.05, n = 42 degrees of freedom). The Student’s T-test (t = 0.33, SD = 1.15; n = 42 degrees of freedom) was used to calculate the average of differences (bias), where a positive value indicates a general overestimation made by NGF, and a negative value indicates an overestimation made by NGS. Upper and lower limits of agreement were calculated as the bias ± 1.96 multiplied by the standard deviation of the differences. 95% confidence interval limits for mean and agreement limits are represented as gray shades. Overall, the bias was non- significant (mean: 0.06, p > 0.05), which means that the average estimation made by NGF is 100.06 or 1.15 times higher than that made by NGS. Differences between methods were homogenously distributed across the range of MRD levels (x-axis), with the limits of agreement approximately set in ±102 and only 3/43 cases (7%) outside the acceptable range. Fig. 2 Kaplan–Meier curves comparing progression-free survival of MRD-positive and MRD-negative subsets. A Progression-free survival of NGS-based results. B Progression-free survival of NGF-based results. Time was calculated from the time of MRD assessment, 3 months after transplantation. Negative patients are represented in black; positive patients are represented in red. Patients at risk are shown at each time point below plots; events are represented between parentheses. Median PFS of positive patients was 46.7 and 34.2 months for NGS and NGF, respectively. Median PFS was not achieved by negative patients. MRD minimal residual disease, NGF next-generation flow, NGS next-generation sequencing. MRD overcomes response-to-treatment and cytogenetics usage, ISS, R-ISS, cytogenetics, and conventional as a predictive factor response. We only detected a significant association As in prior reports, we correlated the MRD status with between achieving CR/sCR and MRD negativity (Fig. S4). different clinical variables such as heavy and light chains Thus, by NGS we observed that 71.2% of patients in CR/ Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 6 of 10 Fig. 3 Kaplan–Meier curves comparing the overall survival of MRD-positive and MRD-negative subsets. A Overall survival of NGS-based results. B Overall survival of NGF-based results. Time was calculated from the time of MRD assessment, 3 months after transplantation. Negative patients are represented in black; positive patients are represented in red. Patients at risk are shown at each time point below plots; events are represented between parentheses. Median overall survival was not achieved by positive or negative patients (3-year OS rates of NGS-positive vs. NGS- negative patients: 77.3% vs. 96.2%; 3-year OS rates of NGF-positive vs. NGF-negative patients: 74.9% vs. 96.6%). MRD minimal residual disease, NGF next-generation flow, NGS next-generation sequencing. sCR were MRD negative, while only 23.3% of patients in alteration (OS rates >90% for all categories, p > 0.05 for all VGPR/PR achieved MRD negativity (p < 0.0001); when paired comparisons); conversely, survival rates of high- NGF data were used, 81.4% of patients in CR/sCR were risk, MRD-positive patients were significantly lower than MRD-negative versus only 20.9% of patients in VGPR/PR the other subgroups (median OS: 21.7 months for NGS (p < 0.0001). No differences were observed in the pro- and NGF, p < 0.01 for all paired comparisons). Interest- portion of each group when both techniques were com- ingly, 55% (6/11) of t(4;14) patients achieved MRD pared (p > 0.05). negativity, while only 12.5% (1/8) of patients with del17p A significant relationship between negative MRD and were MRD negative. longer PFS was observed irrespective to the response (Fig. S5) when patients achieving at least VGPR were con- Univariate and multivariate analyses sidered (n = 95). Remarkably, MRD-negative patients When the univariate analyses for PFS and OS were achieving VGPR showed comparable 3-year PFS rates to performed (Table 3), the MRD status detected by NGS those achieving CR or sCR; this result was similar either and NGF were amongst the most statistically significant considering NGS (VGPR: 88.9%, CR/sCR: 88·1%; p > 0.05) variables associated with the outcome: hazard ratios (HRs) or NGF data (VGPR: 100%, CR/sCR: 89.6%; p > 0.05). On for PFS according to the MRD status after transplantation the other hand, MRD-positive patients had inferior out- were 0.20 for NGF (95% CI: 0.09–0.44, p < 0.001) and 0.29 comes, with no differences observed between responses. for NGS (95% CI: 0.14–0·63, p = 0.002); HRs for OS were Then, we stratified patients (n = 102) into four groups 0.18 for NGF (95% CI: 0.05–0.62, p = 0.007) and 0.21 for considering the MRD status (positive or negative) and NGS (95% CI: 0.06–0.75, p = 0.02). cytogenetics (high-risk or standard-risk). MRD-negative In the multivariate Cox regression models (Table 4), patients showed excellent 3-year PFS rates independently MRD negativity achieved by NGF (HR: 0.20, 95% CI: of their cytogenetic status (Fig. S6), while the outcome of 0.09–0.45, p < 0.001) and the R-ISS I or II (HR: 0.37, 95% MRD-positive patients was significantly determined by CI: 0.17–0.78, p = 0.01) were associated with the lowest their cytogenetic profile: 3-year PFS rates were prolonged risk of progression or death with independent value. for standard-risk patients (67.1% by NGS, 61.8% by NGF), Similarly, MRD negativity achieved by NGF was sig- while for high-risk patients it was 28.6% by NGS and nificantly associated with a decreased risk of death (HR: 15.4% by NGF (median PFS: 14.5 months by NGS and 0.21, 95% CI: 0.06–0.75, p = 0.02), along with the R-ISS 12.8 months by NGF, p < 0.01 compared to the other (HR: 0.13, 95% CI: 0.05–0.38, p < 0.001). MRD negativity groups). For OS (Fig. S7), all MRD-negative patients by NGS had similar results but it was removed from the showed comparable 3-year survival rates to those patients stepwise analysis because of its close association with showing MRD positivity but no high-risk cytogenetic MRD by NGF. Nonetheless, when the analysis was done Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 7 of 10 Table 3 Univariate analysis for progression-free and overall survival. Variable (N) Univariate analysis for PFS Univariate analysis for OS Median survival HR [95% CI] p Median survival HR [95% CI] p Sex Male (61) NA 1.36 [0.68–2.78] >0.05 NA 1.01 [0.38–2.71] >0.05 Female (45) NA NA Cytogenetic risk High risk (23) 40.02 2.99 [1.45–6.16] 0.003 NA 6.33 [2.24–17.86] <0.001 Standard risk (79) NA NA LDH High (20) NA 1.73 [0.80–3.72] >0.05 NA 2.00 [0.69–5.75] >0.05 Low (86) NA NA ISS stage I or II (70) NA 0.63 [0.31–1.31] >0.05 NA 0.22 [0.07–0.67] <0.01 III (33) NA NA R-ISS stage I or II (87) NA 0.34 [0.16–0.74] 0.008 NA 0.12 [0.04–0.44] <0.001 III (16) 34.20 36.1 Conventional response sCR/CR (59) NA 0.54 [0.27–1.08] >0.05 NA 0.36 [0.12–1.04] >0.05 VGPR/PR (43) NA NA MRD status (NGF) Negative (58) NA 0.20 [0.09–0.44] <0.001 NA 0.18 [0.05–0.62] 0.007 Positive (48) 34.20 NA MRD status (NGS) Negative (53) NA 0.29 [0.14–0.63] 0.002 NA 0.21 [0.06–0.75] 0.02 Positive (53) 46.72 NA Cox’s proportional hazards model was used to test individual variables. The number of patients corresponding to each variable’s category is shown in parenthesis. Hazard ratios were calculated comparing first vs. second subgroup of each variable. CI confidence interval, CR complete response, HR hazard ratio, ISS international staging system, LDH lactate dehydrogenase, MRD minimal residual disease, NA not achieved, NGF next-generation flow, NGS next-generation sequencing, PR partial response, R-ISS revised international staging system, sCR stringent complete response, VGPR very good partial response. using the 10-5 cutoff, considering only the 95 patients not evaluate the deepest responses to treatment, measured affected by preanalytical pitfalls, MRD negativity by NGF either by flow or sequencing16. Moreover, the FDA has (HR: 0.16, 95% CI: 0.07–0.37, p < 0.001) and the R-ISS recently cleared the use of NGS to perform MRD analyses. (HR: 0.30, 95% CI: 0.12–0.72, p = 0.007) were still selected However, to date only one strategy has been approved for PFS, while MRD negativity by NGS (HR: 0.15, 95% CI: (Adaptive’s clonoSEQ), which makes access difficult for 0.04–0.57, p = 0.005) and the R-ISS (HR: 0.13, 95% CI: many centers, and raises the need to find other suitable 0.04–0.39, p < 0.001) were selected for OS, indicating a strategies. Here, we have evaluated and compared two similar potential for prediction. different strategies: NGF, developed by the EuroFlow group and recommended by the IMWG, and the NGS- Discussion based LymphoTrack® IGH panel, a commercial strategy MRD evaluation has arisen as a very promising tool to designed by Invivoscribe Technologies. predict the outcome of hematologic patients. On this MRD results are highly dependent on the quality and behalf, the IMWG validated the use of this parameter to concentration of the samples, which is directly related to Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 8 of 10 Table 4 Multivariate analysis for progression-free and However, all studies were done on an intention-to-study overall survival. analysis, mimicking the intention-to-treat analysis that we usually see in clinical trials. Variable Multivariate analysis Multivariate Despite being limited by the concentration method and for PFS analysis for OS the poor quality of certain samples, NGS was found to be fairly concordant with NGF in terms of MRD detection HR [95% CI] p HR [95% CI] p and quantification ability. Although we detected 15 dis- R-ISS stage cordant cases, only three of those patients relapsed (two I or II vs. III 0.37 [0.17–0.78] 0.01 0.13 [0.05–0.38] <0.001 positive cases by NGF, one positive case by NGS). In contrast, highly concordant results were observed in the MRD status (NGF) MRD evaluation of the remaining 28 relapsing patients: Negative vs. 0.20 [0.09–0.45] <0.001 0.21 [0.06–0.74] 0.02 23 were double-positive and only five were negative by positive both methods, which could be explained by risk factors MRD status (NGS) already present at diagnosis, extramedullary relapses24, or Negative vs. >0.05 >0.05 infiltration levels below the limit of detection. positive Most discordant cases had MRD levels below 10−5, that may be explained either by a higher sensitivity for one Those variables with a significant impact in the univariate analysis were method over the other, or in the case of NGS, if technical introduced in a multivariate Cox regression model to determine which ones were predictive for survival. Hazard ratios were calculated comparing first vs. inaccuracies due to the use of the spike in for quantifi- second subgroup of each variable. Cytogenetics were included only as part of cation were present (i.e. underestimation of the MRD the R-ISS score. CI confidence interval, HR hazard ratio, MRD minimal residual disease, NGF next- level). The only six discordant cases (NGS = 3; NGF = 3) generation flow, NGS next-generation sequencing, R-ISS revised international with MRD levels over 10−5 could be explained based on staging system. differences in the sampling procedure (different tubes from different bone marrow pulls), since sensitivity was preanalytical conditions9–11,20. Since flow cytometry set in the range of 10−5−10−6 for both techniques. Only requires an immediate experimental procedure, this is three discordant cases relapsed, and accordingly no dif- usually taken as a disadvantage compared to NGS, whose ferences were seen with double-negative patients in terms samples can be frozen and stored after gDNA extraction of PFS, although an intermediate survival rate could be for later analyses. On the other hand, NGS relies on the expected in a broader population with longer follow up. identification of clonotypic rearrangements at diagnosis, According to the standard design of clinical trials, MRD making that up to 10% of cases cannot be followed due to is evaluated at several timepoints irrespective of the somatic hypermutation in primer-annealing regions, response. In our study, eight patients who had achieved though this was not the case in this study, and once it is VGPR as confirmed by immunofixation and electro- characterized, CDR3 regions represent a stable and reli- phoresis had undetectable residual disease by next- able tool for MRD evaluation29. Other major flaws of NGS generation techniques. This finding is concordant with a are the turnaround time, which is longer than for NGF situation where tumor plasma cells had been eradicated in (5–7 days compared to 24–48 h), and the need to run a the bone marrow, while the monoclonal component was high number of samples together. In contrast, inter- still present in the blood stream. Noteworthy, five of them pretation of results is usually more difficult for NGF, achieved ≥CR after consolidation, supporting the notion requiring high expertise, while LymphoTrack’s solution is that, in the clinical practice, MRD should be assessed more user-friendly and semi-automated. whenever patients achieve CR16,30. In our study, the use of ethanol precipitation resulted in Cytogenetic status still maintained its potential to suboptimal sample quantity, given that, in our experience, stratify patients, but only when applied to the MRD- up to 40% of gDNA may be lost during its processing positive subset, a feature that has already been described workflow, and the use of ethanol and sodium acetate for patients treated with the combination of bortezomib, usually turns to an increment of contaminating sub- lenalidomide, and dexamethasone31. As far as this study is stances (salts and alcohol) that could interfere with PCR concerned, it seems that this treatment scheme could amplification. Accordingly, all the samples were optimal overcome the negative prognostic impact of t(4;14) and be for NGF studies (criteria for inclusion), while 10% of the able to provide MRD-negative responses. By contrast, 17p samples were suboptimal for NGS. If we had considered deletion still retained its association with poor survival only optimal samples for both technologies, results would rates, and was strongly correlated with MRD positivity. have favored NGS based on its slightly higher sensitivity Regarding the multivariate analysis, NGF retained a using fewer cells; this represents a major advantage over more independent informative capacity for survival than NGF, that requires the acquisition of 10,000,000 events. NGS because it was selected before in the multivariate Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 9 of 10 analysis and the results of the two methods were closely Author contributions associated. This result has to be interpreted with caution A.M., R.G.-S., N.P., J.F.-M. and M. González conceived and designed the study. N.P., N.C.G., A.O., L.R., J.B., M. Gironella, M.T.H., V.G.-C., J.F.S.M., J.-J.L., M.-V.M., J.M.-L., M. because the preanalytical conditions strongly affected González, and R.G.-S. recruited patients for the Spanish Multiple Myeloma Working NGS and could not be considered by the multivariate Group (GEM-PETHEMA) clinical trials. A.M., C.J., M.E.S., I.P.-C., and M.G.-A. produced analysis. Despite that, NGS provided consistent and and analyzed the molecular results. N.P., J.F.-M., M.-T.C., B.P., and A.O. produced and analyzed the immunophenotypic results. A.M., M.E.S., V.G.-C., and R.G.-S. reviewed comparable results to NGF when only non-biased cases the data, performed the statistical analyses, and wrote the first draft of the were evaluated. manuscript. N.P., J.-F.-M., I.P.-C., M.G.-A., C.C., M.A., V.G.-C., and R.G.-S. provided As Perrot et al. have recently stated14, both NGF and support for statistical analyses and clinical data interpretation. All the authors reviewed the final manuscript and agreed on its publication. R.G.-S. and M. NGS can reach a sensitivity of 10−6 when more cells are González obtained funding for the study. R.G.-S., the head of the group, produced used and, in this context, the most important factor is not the final revision of the manuscript and gave final approval for its publication. the technique but the ability to detect deeper levels of Conflict of interest MRD, avoiding false-negative results due to lower detec- The authors report grants PI15/01956, CIBERONC-CB16/12/00233, CEI-2010-1-0010, tion limits. Direct comparison of MRD quantification by FEHH, FI19/00320, IMS Young Investigator Award 2019, CPII18/00028 and CB16/ NGS and NGF in the same conditions (i.e. reaching the 12/00400 during the conduct of the study. All Spanish funding is co-sponsored by the European Union FEDER program. N.P. has consulted for and has received same sensitivity level and using the same marrow pull) honoraria from Janssen, Takeda, Amgen and Celgene. A.O. has consulted or should be performed to shed light into this issue, although served in an advisory role for Amgen, Celgene, Takeda and Janssen. L.R. has an intrinsic false-negativity risk will always exist for both received honoraria from Celgene and Janssen. J.B. has received honoraria from Janssen. B.P. has received honoraria, consulted or served in an advisory role for methodologies due to the patchy nature of the disease. As Amgen, Bristol-Myers Squibb, Celgene, EngMab, Janssen, Merck, Novartis, Roche, it was mentioned before, samples undergoing laboratory Sanofi and Takeda. J.F.S.M. has consulted or served in an advisory role for Janssen, analysis are not the same, and a great variation can be Celgene, Takeda, Amgen, BMS, MSD, Novartis, Sanofi, and Roche. J.-J.L. has received honoraria from Amgen, Celgen, and Janssen. M.-V.M. is on the board of detected from the first marrow pull (which is usually sent directors or an advisory committee and has received honoraria from Janssen, to the morphology laboratory) to the last one (which is Celgene, Takeda, Amgen, and BMS. J.M.-L. has consulted or served in an advisory distributed for molecular, cytogenetic, and immunophe- role for BMS, Janssen, and Novartis. R.G.-S. has consulted for and received honoraria from Janssen, Takeda, BMS, Hospira, Pharmacyclics, Gilead, and Incyte. notypic studies)23,32. This could be a feasible explanation The remaining authors declare no conflict of interest. for the 15 discordant cases in our cohort. Nonetheless, our study demonstrates the similarity between NGS and NGF, with analogous results to those reported in the Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in GEM2012 overall analysis of MRD by NGF30 but using published maps and institutional affiliations. ~10 times less cells. In summary, our results support the use of the Lym- Supplementary Information accompanies this paper at (https://doi.org/ 10.1038/s41408-020-00377-0). phoTrack® strategy to detect and evaluate MRD in MM patients, with excellent applicability and comparable Received: 23 July 2020 Revised: 14 September 2020 Accepted: 2 October results to NGF. Altogether, these findings reinforce the 2020 use of MRD assessment as an endpoint in MM clinical trials and underline the need of standardization and quality assessment in future studies for all MRD approaches in MM. References 1. Rajkumar, S. V. et al. International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. Lancet Oncol. 15, e538–e548 (2014). Acknowledgements 2. Ocio, E. M. et al. New drugs and novel mechanisms of action in multiple The authors wish to thank José Pérez, Alicia Antón, Rebeca Maldonado, myeloma in 2013: a report from the International Myeloma Working Group Montserrat Hernández-Ruano, Estrella Arnés, Mercedes Jiménez, Alejandra (IMWG). Leukemia 28, 525–542 (2014). Martín, Isabel Sánchez, Ana Balanzategui, Rocío Corral, and Francisco Boix 3. Goldschmidt, H., Ashcroft, J., Szabo, Z. & Garderet, L. Navigating the treatment (University Hospital of Salamanca, Spain) for their technical support, as well as landscape in multiple myeloma: which combinations to use and when? Ann. the Spanish Multiple Myeloma GEM-PETHEMA group for providing clinical Hematol. 98, 1–18 (2019). data. The authors are also very grateful to the patients who participated in 4. Mikhael, J. et al. Treatment of multiple myeloma: ASCO and CCO Joint Clinical this study. Practice Guideline. J. Clin. Oncol. 37, 1228–1263 (2019). 5. Lahuerta, J. J. et al. Busulfan 12 mg/kg plus melphalan 140 mg/m2 versus melphalan 200 mg/m2 as conditioning regimens for autologous transplan- Author details tation in newly diagnosed multiple myeloma patients included in the 1 Departamento de Hematología, Hospital Universitario de Salamanca (HUSA/ PETHEMA/GEM2000 study. Haematologica 95, 1913–1920 (2010). IBSAL), CIBERONC, CIC-IBMCC (USAL-CSIC), Salamanca, Spain. 2Centro de 6. Mateos, M. V. et al. Bortezomib, melphalan, and prednisone versus bortezo- Investigación del Cáncer-IBMCC (USAL-CSIC), CIBERONC, Salamanca, Spain. mib, thalidomide, and prednisone as induction therapy followed by main- 3 Hospital Germans Trias i Pujol, Institut Català d’Oncología (ICO), Institut Josep tenance treatment with bortezomib and thalidomide versus bortezomib and Carreras, Badalona, Spain. 4Hospital Clínic i Provincial, Institut de Investicacions prednisone in elderly patients with untreated multiple myeloma: a rando- Biomediques August Pi i Sunyer (IDIBAPS), Barcelona, Spain. 5Hospital Vall mised trial. Lancet Oncol. 11, 934–941 (2010). d’Hebrón, Barcelona, Spain. 6Hospital Universitario de Canarias, Santa Cruz de 7. Rosinol, L. et al. Bortezomib, lenalidomide, and dexamethasone as induction Tenerife, Spain. 7Hospital 12 de Octubre, i + 12, CNIO, Universidad therapy prior to autologous transplant in multiple myeloma. Blood 134, Complutense, Madrid, Spain. 8Clínica Universidad de Navarra (CUN), Centro de 1337–1345 (2019). Investigación Médica Aplicada, IDISNA, CIBERONC, Pamplona, Spain Blood Cancer Journal Medina et al. Blood Cancer Journal (2020)10:108 Page 10 of 10 8. Durie, B. G. et al. International uniform response criteria for multiple myeloma. 21. Mateos, M. V. et al. Daratumumab plus bortezomib, melphalan, and pre- Leukemia 20, 1467–1473 (2006). dnisone for untreated myeloma. N. Engl. J. Med. 378, 518–528 (2018). 9. Mailankody, S. et al. Minimal residual disease in multiple myeloma: bringing 22. Flores-Montero, J. et al. Next generation flow for highly sensitive and stan- the bench to the bedside. Nat. Rev. Clin. Oncol. 12, 286–295 (2015). dardized detection of minimal residual disease in multiple myeloma. Leukemia 10. Landgren, O. & Owen, R. G. Better therapy requires better response evaluation: 31, 2094–2103 (2017). paving the way for minimal residual disease testing for every myeloma 23. Bai, Y., Orfao, A. & Chim, C. S. Molecular detection of minimal residual disease patient. Cytometry B 90, 14–20 (2016). in multiple myeloma. Br. J. Haematol. 181, 11–26 (2018). 11. Paiva, B., García-Sanz, R. & San Miguel, J. F. Multiple myeloma minimal residual 24. Sarasquete, M. E. et al. Minimal residual disease monitoring in multiple mye- disease. Cancer Treat Res. 169, 103–122 (2016). loma: a comparison between allelic-specific oligonucleotide real-time quan- 12. Munshi, N. C. et al. Association of minimal residual disease with superior titative polymerase chain reaction and flow cytometry. Haematologica. 90, survival outcomes in patients with multiple myeloma: a meta-analysis. JAMA 1365–1372 (2005). Oncol. 3, 28–35 (2017). 25. Puig, N. et al. Critical evaluation of ASO RQ-PCR for minimal residual disease 13. Lahuerta, J. J. et al. Depth of response in multiple myeloma: a pooled analysis evaluation in multiple myeloma. A comparative analysis with flow cytometry. of three PETHEMA/GEM clinical trials. J. Clin. Oncol. 35, 2900–10. (2017). Leukemia 28, 391–397 (2014). 14. Perrot, A. et al. Minimal residual disease negativity using deep sequencing is a 26. Chng, W. J. et al. IMWG consensus on risk stratification in multiple myeloma. major prognostic factor in multiple myeloma. Blood 132, 2456–2464 (2018). Leukemia 28, 269–277 (2014). 15. Martinez-Lopez, J. et al. Prognostic value of deep sequencing method for 27. Palumbo, A. et al. Revised International Staging System for Multiple Myeloma: minimal residual disease detection in multiple myeloma. Blood 123, a report from International Myeloma Working Group. J. Clin. Oncol. 33, 3073–3079 (2014). 2863–2869 (2015). 16. Kumar, S. et al. International Myeloma Working Group consensus criteria for 28. van Dongen, J. J. M. et al. Design and standardization of PCR primers and response and minimal residual disease assessment in multiple myeloma. protocols for detection of clonal immunoglobulin and T-cell receptor gene Lancet Oncol 17, e328–e346 (2016). recombinations in suspect lymphoproliferations: report of the BIOMED-2 17. Paiva, B., Merino, J. & San Miguel, J. F. Utility of flow cytometry studies in the Concerted Action BMH4CT98-3936. Leukemia 17, 2257–2317 (2003). management of patients with multiple myeloma. Curr. Opin. Oncol. 28, 29. Puig, N. et al. The predominant myeloma clone at diagnosis, CDR3 defined, is 511–517 (2016). constantly detectable across all stages of disease evolution. Leukemia 29, 18. Ladetto, M. et al. Next-generation sequencing and real-time quantitative PCR 1435–1437 (2015). for minimal residual disease detection in B-cell disorders. Leukemia 28, 30. Paiva, B. et al. Measurable residual disease by next-generation flow cytometry 1299–1307 (2014). in multiple myeloma. J. Clin. Oncol. 38, 784–792 (2020). 19. Cavo, M. et al. Role of 18F-FDG PET/CT in the diagnosis and management of 31. Goicoechea, I. et al. Deep MRD profiling defines outcome and unveils different multiple myeloma and other plasma cell disorders: a consensus statement by modes of treatment resistance in standard and high risk myeloma. Blood. 2020 the International Myeloma Working Group. Lancet Oncol. 18, e206–e217 Jul 21:blood.2020006731. https://doi.org/10.1182/blood.2020006731 [Online (2017). ahead of print]. 20. Stetler-Stevenson, M. et al. Consensus guidelines for myeloma minimal resi- 32. Rawstron, A. C. et al. Report of the European Myeloma Network on multi- dual disease sample staining and data acquisition. Cytometry B 90, 26–30 parametric flow cytometry in multiple myeloma and related disorders. Hae- (2016). matologica 93, 431–438 (2008). Blood Cancer Journal