The Light Chain IgLV3-21 Defines a New Poor Prognostic Subgroup in Chronic Lymphocytic Leukemia: Results of a Multicenter Study

The development and progression of chronic lymphocytic leukemia (CLL) depends on stimuli received by the microenvironment mediated by the B-cell receptor (BCR; ref. 1). This immune receptor is composed of two heavy chains and two light chains and has been studied extensively in CLL. In particular, the level of somatic hypermutations within the variable region of the immunoglobulin heavy chain (IgHV) has been used as a prognostic biomarker for over 15 years (2, 3). Patients expressing an IgHV that has undergone somatic hypermutation and shows less than 98% identity with the closest germline are IgHV-mutated (M) and more often have an indolent disease course. Patients with unmutated (UM) IgHV (more or equal 98% identity to germline) experience more rapid progression and shorter survival.

Studying the nature of the BCR in CLL is crucial to understand how CLL clones are selected and how CLL cells receive survival signals. Indeed, BCR engagement rescues CLL cells from apoptosis (4) and the response to IgM stimulation has been linked to CLL prognosis (5, 6). These in vivo survival stimulations could be due to an external antigen or an antigen-independent cell-autonomous signaling whereby a BCR interacts with intrinsic BCR structures (7). The fact that the immunoglobulin gene repertoire in CLL is biased (8) and that 30% of CLL patients express very similar (sometimes even identical) BCRs is in line with reports of antigen-driven selection and clonal expansion (9). Indeed, this phenomenon is unlikely to occur by chance and suggests that similar epitopes or antigens could stimulate and select CLL clones.

So far, studies of the BCR in CLL have focused mainly on the IgHV, but the role of the immunoglobulin light chain in influencing prognosis is largely unknown. As in the case of the heavy chain repertoire, the light chain repertoire in CLL is also biased and some IgKV/IgLV genes are used more often in CLL compared with normal B cells (10). The light chain IgLV3-21 (previously named “VL2-14”) represents the most abundant lambda light chain in CLL (∼21%) and is the third most abundant light chain (∼7%, after IgKV3-20 and IgKV1-39) if both lambda and kappa light chains are considered (10). Interestingly, the majority of CLL patients with the heavy chain IgHV3-21 also express the light chain IgLV3-21. In addition, more than half of these patients display highly similar variable heavy complementarity determining region 3 (VH CDR3) sequences and belong to the largest stereotyped receptor group (∼3% of all CLL) named “subset #2” (11). Patients belonging to subset #2 (IgHV3-21/IgLV3-21) are characterized by a shorter overall survival, similar to UM IgHV patients, despite the fact that two thirds of these patients are IgHV-mutated (12). Interestingly, non-subset #2/IgHV3-21 patients differ from the aggressive subset #2 patient subgroup, and their prognosis is dependent on their IgHV mutational status, as in all other CLL cases (11).

The abnormal frequency of the light chain IgLV3-21 in an initial cohort of 32 patients with high-risk CLL led us to further investigate its prognostic significance. In the present study, we therefore correlated the presence of IgLV3-21 with treatment-free survival (TFS) and/or overall survival (OS) in a total of 813 patients from 4 different CLL cohorts: an initial cohort of 32 high-risk patients (A), 2 cohorts of patients at diagnosis (B. n = 270; C. n = 289) and one cohort (D) of 224 CLL patients from 2 Fludarabine-Cyclophosphamid-Rituximab (FCR)-based UK clinical trials.

This study was conducted in accordance with the declaration of Helsinki, approved by the institutional review board (local Ethics Committee of Jules Bordet Institute and John Radcliffe Hospital) and was performed using samples collected from CLL patients after written informed consent. All CLL patients had a typical CD19⁺CD5⁺CD23⁺ phenotype and a Catovsky score of 4/5 or 5/5. Additional details about patient populations, sample collection, preparation and DNA/RNA extraction can be found in Supplementary Text S1 and Supplementary Tables S1–S5.

IgHV gene mutational analysis was performed by next-generation deep sequencing and/or Sanger sequencing, as previously described (13), and the sequences were aligned with those in the international ImMunoGeneTics (IMGT) information system database (http://www.imgt.org/; ref. 14). Sequences with ≤2% deviation from any germline IgHV sequence were considered unmutated (2). Amino acid sequences were generated from the IMGT website, protein sequences were aligned using Clustal Omega (15) and the consensus sequence was computed with Weblogo (16). Patients were assigned to the major stereotyped subset using the “ARResT/AssignSubsets” online tool (17). Analysis of other prognostic factors, such as Binet stage, zeta-associated protein 70 (ZAP70), lipoprotein lipase (LPL), CD38 molecule (CD38) expression, cytogenetic abnormalities, and lymphocyte doubling time (LDT), as well as target resequencing for the analysis of recurrent mutations and genome-wide SNP analyses, were performed as previously described (18–20). Additional details can be found in Supplementary Text S2.

Whole transcriptome sequencing libraries were prepared using the TruSeq Stranded Total RNA Sample Preparation Kit (Illumina). Libraries underwent 2 × 76 bp paired-end sequencing on a HiSeq 2500 instrument (Illumina). The median number of paired-end reads was 60.5 million (range, 49.7–79.7 million). Data processing and analyses were conducted using bioinformatics pipelines using the Computational Genomics Analysis and Training (CGAT) code collection (21) and CGAT Pipeline repository (https://github.com/CGATOxford/CGATPipelines). Sequence data (Bam files) has been deposited at the European Genome-phenome Archive (EGA), which is hosted by the EBI and the CRG, under accession number EGAS00001002894. Further information about EGA can be found on https://ega-archive.org and "The European Genome-phenome Archive of human data consented for biomedical research" (http://www.nature.com/ng/journal/v47/n7/full/ng.3312.html). Estimates of the number of reads per transcript were obtained using kallisto (22) and all transcripts in Ensembl build 82 (hg38; http://sep2015.archive.ensembl.org/index.html). Gene-level read counts were obtained by aggregating across all isoforms per gene. Differential expression and functional enrichment analyses are detailed in Supplementary Text S3.

The estimated read counts for each variable gene region from the RNA-Seq data were used to determine the light chain usage for each patient. Sequences were confirmed by Sanger sequencing using the primers listed in Supplementary Table S6. To screen additional cohorts of patients for the presence of IgLV3-21, we developed a rapid and simple method by real-time PCR (qPCR; Supplementary Text S4/Supplementary Fig. S1) that was validated by Sanger Sequencing.

TFS, OS, and progression-free survival (PFS) distributions were plotted using Kaplan–Meier estimates and were compared using the log-rank test. TFS and OS were calculated from the time of diagnosis until the date of first treatment and the date of death, respectively. The OS curves presented here include all death events, even if they were not CLL-related. When we considered only CLL-related deaths (CLL-OS), similar results were obtained (Supplementary Fig. S2). PFS was calculated from the end of first-line therapy to disease progression or death of the patient. We used Fisher's exact test to describe associations between clinical markers. Comparisons between 2 groups were assessed using a non-parametric Mann–Whitney test. Multivariate Cox regression stepwise analysis was used to evaluate the effects of the different prognostic variables on TFS and OS. All tests were two-sided. An effect was considered to be statistically significant if P < 0.05. All analyses were performed with GraphPad Prism 5.0 software or IBM SPSS 13.0 software.

Using RNAseq data, we investigated the expression of the heavy and light chains in an initial cohort of 32 patients with high-risk CLL. All sequences were verified by Sanger sequencing (Supplementary Table S2). Surprisingly, 9 of the 32 (28%) patients expressed the light chain IgLV3-21, but only 1 of the 9 (11%) patients expressed the heavy chain IgHV3-21. The expected frequency of IgLV3-21 is approximately 7% (10) in an unbiased CLL population. However, this population was composed of high-risk CLL patients in need of treatment. For 30 of the 32 patients, TFS and OS data were available. IgLV3-21 patients had a poorer prognosis than the other patients in this cohort in terms of both TFS (P = 0.0270) and OS (P = 0.0287; Fig. 1A and B). The prognosis of IgLV3-21 patients was significantly different from that of M IgHV patients (TFS: P = 0.0242; OS P = 0.0098) but not from that of UM patients (TFS: P = 0.0631; OS P = 0.2851; Fig. 1C and D). However, the number of patients in this study was small, thus these results should be confirmed in a larger cohort. It should be noted that, among the 9 IgLV3-21 patients, 4 (44%) were IgHV M.

Differential expression analysis of IgLV3-21 CLL patients (compared with non- IgLV3-21 patients) revealed 1,712 differentially expressed transcripts (41.4% upregulated and 58.6% downregulated; Supplementary Table S7) and 1,555 differentially expressed genes (45.2% upregulated and 54.8% downregulated; Supplementary Table S8). Of the 1,555 differentially expressed genes, 723 were protein coding and had a fold-change greater than 1.5-fold (Supplementary Table S9). Hierarchical clustering based on the 723 genes differentially expressed in IgLV3-21 patients grouped 8/9 IgLV3-21 patients together, demonstrating that the gene expression differences observed between IgLV3-21–positive and negative patients were consistent (Fig. 1E).

We performed gene set enrichment analysis to identify gene ontology (GO) categories and Biocarta, KEGG, and Broad/MIT pathways that were enriched in the genes differentially expressed between our two classes of samples. Interestingly, when all differentially expressed genes (Supplementary Table S10) or only upregulated genes (Supplementary Table S11) were considered, GO categories/pathways relating to translation (ribosome, translational reactome, peptide chain elongation, RNA metabolism—P < 0.0001) and MYC target genes (P = 0.0003) were significantly overrepresented (Fig. 1G–I). However, no pertinent categories/pathways were enriched within the downregulated genes (Supplementary Table S12). To confirm the increased expression of genes within the identified enriched pathways, we selected 7 genes and validated their increased expression by real-time PCR (Fig. 1F).

After obtaining results from the initial 32 patients, we investigated the prognostic significance of IgLV3-21 expression in 2 validation cohorts (populations B and C) composed of CLL samples obtained at diagnosis. In population B, which was composed of 270 patients with a median follow up of 83 months (range, 1–397), we used real-time PCR to identify 30 (11%) IgLV3-21–expressing patients, including 7 (3%) patients from subset #2 (IgHV3-21/IgLV3-21). Patients with an IgLV3-21 light chain had a significantly shorter median TFS (29 months) compared with other patients (82 months, P = 0.0002; Fig. 2A). Similar results were observed for OS (137 vs. 231 months, P = 0.0483; Fig. 2B). Upon subdividing patients into subset #2, other IgHV/IgLV3-21 + non-subset #2, other IgHV UM, and other IgHV M, we found that the first 3 groups had similar median TFS (P < 0.0001) and OS (P < 0.05) rates that were statistically shorter than those in IgHV M patients (Fig. 2C and D). Because the presence of IgHV3-21 did not influence prognosis, we pooled all patients with IgLV3-21. IgLV3-21 patients had a median TFS of 29 months, which was not significantly different from that of IgHV UM patients (36 months; P = 0.6798) but was significantly worse than that of IgHV M patients (144 months, P < 0.0001; Fig. 2E). Similar results were obtained for OS (Fig. 2F). It should be noted that, among the 30 IgLV3-21 patients, 12 (40%) were IgHV M.

To determine whether the presence of IgLV3-21 is an independent prognostic factor, we performed a stepwise multivariate analysis designed to identify the optimal set of independent variables to predict TFS/OS, including the following classical prognostic factors: Binet stages, IgHV mutational status, ZAP70 (Zeta-associated protein), LPL (lipoprotein lipase), CD38 and the use of subset #2 stereotyped receptor. The stepwise analysis identified the presence of IgLV3-21 as a prognostic factor for TFS prediction with a non-significant P value (P = 0.0717; Supplementary Table S13A–S13B). However, when repeating the same analysis for IgHV M patients only (n = 166), the presence of IgLV3-21 was identified as an independent prognostic factor for both TFS and OS prediction with a significant P value (P < 0.001), suggesting that IgLV3-21 may have a significant and independent prognostic impact in patients with mutated IgHV (Supplementary Table S13 C–S13D).

In population C, which was composed of 289 patients with a median follow up of 78 months (range, 0.3–288), we identified 33 (11%) IgLV3-21–positive patients by real-time PCR including 9 (3%) patients from subset #2 (IgHV3-21/IgLV3-21). The prognostic impact of IgLV3-21 light chain presence that was identified in populations A and B was validated in population C (Supplementary Fig. S3). IgLV3-21-expressers had a median TFS of 31 months, which was not significantly different from that of IgHV UM patients (21 months; P = 0.9458) but was significantly worse than that of IgHV M patients (128 months, P < 0.0001; Supplementary Fig. S3E). Similar results were obtained for OS (Supplementary Fig. S3F). It should be noted that, of the 33 IgLV3-21 expressers, 16 (48%) were IgHV M. Similar results were obtained from clinical trial population D (see Supplementary Fig. S4 and Supplementary Text S5).

Recurrent mutations and cytogenetic data were available for 181 patients in population D. Multivariate analysis for TFS prediction was performed, including recurrent mutations in CLL (TP53, ATM, NOTCH1, SF3B1, MYD88, SAMHD1), common cytogenetic abnormalities [del(17p), del(11q), del(6q), del(13q) trisomy 12], the use of subset #2 stereotyped receptor and the occurrence of IgLV3-21. This analysis selected the presence of IgLV3-21 as an independent prognostic factor. The analysis also selected del(13q), TP53 mutation and IgHV status as independent prognostic factors, although only IgHV status had a p-value < 0.05 in the final model (IgLV3-21: P = 0.0603) of TFS prediction (Supplementary Table S14 A). However, when we performed the same analysis in M patients only (n = 69), the presence of IgLV3-21 (P = 0.0415) was selected as an independent predictor of poor prognosis (Supplementary Table S14 B). Interestingly, association analysis showed that the presence of the IgLV3-21 chain was significantly associated with the presence of SF3B1 (P < 0.0001) and ATM (P = 0.0245) mutations, stereotyped receptor from subset #2 (P < 0.0001), and the absence of both NOTCH1 mutation (P = 0.0367) and trisomy 12 (P = 0.0033; Supplementary Table S15).

Although IgLV3-21 is a significant predictor of TFS and OS (Figs. 1 and 2; Supplementary Fig. S3), we did not observe a significant difference in terms of PFS after FCR-based treatment between IgLV3-21 and other patients (Fig. 3A). However, IgLV3-21 patients have a shorter PFS compared with IgHV- mutated patients (Fig. 3B) but we could not exclude that this difference is due to a high proportion of IgHV-unmutated patients carrying IgLV3-21 light chain (Fig. 3D). Indeed, we did not observe a significant difference in term of PFS between IgLV3-21/other M IgHV and the other M IgHV patients (P = 0.1247). A multivariate Cox analysis for PFS prediction including IgLV3-21 usage, identified presence of subset #2 rearrangement, IgHV mutational status and classical cytogenetic abnormalities [del(13q), trisomy 12, del(11q), del(17p)] selected IgHV mutational status (HR, 3.1; P < 0.0001) and deletion 17p (HR, 7.3; P < 0.0001) as independent predictors of PFS, whereas IgLV3-21 was not considered (Supplementary Table S16).

For the following analysis, we pooled all patients from the different populations in order to increase the number of patients for subgroup analysis. To investigate whether the poor prognosis of IgLV3-21-expressing patients could be due to IgHV mutational status, we divided the population into 5 subgroups: M IgHV, UM IgHV, subset #2 (IgHV3-21/IgLV3-21), M IgHV/IgLV3-21, and UM IgHV/IgLV3-21. As presented in Fig. 4, all patients with the IgLV3-21 light chain have a comparable median TFS and OS to that of UM patients (P < 0.0001), independent of their mutational status or the presence of a stereotyped heavy chain IgHV3-21 assigned to subset #2. As presented in Supplementary Table S17 (multivariate Cox analysis including IgLV3-21 usage, subset #2 usage, IgHV mutational status and classical cytogenetic abnormalities), IgLV3-21 was selected as an independent prognostic factor for TFS (HR, 1.5; P = 0.0004, n = 685) and OS (HR, 1.97; P = 0.0031; n = 471) prediction. More interestingly, when only IgHV-mutated patients were considered, patients with IgLV3-21 had a 2.4 times greater risk of undergoing treatment (P < 0.0001, n = 333) and an 8.7 times greater risk of death (n = 250, P < 0.0001) compared with patients with hypermutated IgHV and no IgLV3-21 usage. The additional prognostic information added by IgLV3-21 usage to the CLL international prognostic index (CLL-IPI) is presented in Supplementary Fig. S5. IgLV3-21 is able to identify patients with poor prognosis in the “low” subgroup and in the “intermediate” subgroup (at least for TFS). In the “high” and “very high” subgroups, IgLV3-21 did not add any prognostic information.

To further characterize the BCR of patients carrying an IgLV3-21 rearrangement, we pooled all available IgLV3-21 sequences from the present study (n = 128) with those previously published by Stamatopoulos and colleagues (10; n = 20) and compared these HV/LV sequences to two control groups [patients with κ (n = 174) and other λ (n = 76) light chains; ref. 10]. Data are available in Supplementary Text S6, Supplementary Fig. S6 and Supplementary Table S18.

Over the past decade, clinical and biological evidence has established the BCR as one of the key drivers of clonal selection and survival of CLL cells (23–26). So far, the majority of studies investigating the prognostic significance of the BCR have focused on the heavy chain (2, 3), although there is emerging evidence that the light chains might also play a crucial role in BCR function in CLL (27).

In the current study, we present a comprehensive characterization of gene expression, prognosis, and immunogenic features of patients carrying the IgLV3-21 rearrangement and reveal its association with short TFS and OS in 4 separate cohorts of over 800 patients. Importantly, we showed that IgLV3-21 patients had a poor prognosis irrespective of the presence or absence of a BCR assigned to subset #2. In IgHV-mutated patients, IgLV3-21 usage is independent of other classical prognostic factors, cytogenetic abnormalities and recurrent driver mutations and defines a poor risk group within this good prognosis subgroup. In addition, we highlight that IgLV3-21 usage could indicate poor prognosis in patients classified as low risk by the CLL-IPI.

Based on an initial cohort of 32 patients with high-risk CLL, we found a higher frequency (28% vs. 7%) of IgLV3-21 expressers compared with all CLL patients (10). Interestingly, the frequency of IgLV3-21 was 28% (population A) and 25% (population D) in high risk cohorts with a short median TFS of 24 and 30 months, respectively, whereas it fell to 11% for an unbiased cohort obtained at diagnosis (populations B and C), which is close to what has been reported previously (10).

Gene-expression profiles obtained from total RNA sequencing revealed genes that were differentially expressed between IgLV3-21-expressers and other patients. Specifically, genes involved in translational regulation (including ribosome biogenesis and RNA processing) and MYC target genes were overrepresented in those patients. These data are compatible with the data presented by Yeomans and colleagues (28), who reported that the engagement of BCR in CLL induced a global increase of translation as well as an increase in MYC-specific mRNA translation. In addition, several of the overexpressed genes in patients carrying IgLV3-21 have roles in translation (NPM1, RLP18, RPS6, MRPL44) and were also reported by Pozzo and colleagues (29) to be overexpressed in NOTCH1-mutated patients and to be modulated by NOTCH1 signaling via MYC transcription. In the present study, NOTCH1 mutation was significantly associated with the absence of IgLV3-21, suggesting that the increase in NPM1 and other ribosomal proteins is not linked to NOTCH1 pathway deregulation. Alternatively, the association of mutated SF3B1 with IgLV3-21 could lead to splicing dysregulation affecting multiple cellular functions, including DNA damage response, telomere maintenance, and Notch signaling (30). However, half of IgLV3-21 patients were SF3B1-wildtype (population D), indicating that our results were at least partially independent of SF3B1. The overexpression of NPM1 and other ribosomal proteins is compatible with an enhanced biosynthetic pathway leading to a growth advantage in IgLV3-21–expressing cells and subsequently to a poor prognosis of these patients.

In the initial cohort of 32 patients with high-risk CLL, we observed a shorter median TFS and OS for patients with IgLV3-21. These results, which were obtained using a small cohort, were validated in three other independent cohorts of 270, 289, and 224 patients collected in different centers across Europe. Although the majority of IgHV3-21 cases (B: 80%, C: 83%, D: 86%) were associated with the light chain IgLV3-21, as previously reported (31), the inverse was not necessarily true. In each of these 3 cohorts, IgLV3-21 patients had a median TFS/OS that was not significantly different from that of IgHV UM or subset #2 patients but was shorter than that of M patients. However, IgLV3-21 was not found as a predictive marker of FCR-based treatment response, at least in the present cohort (population D). The median follow-up after the end of the first treatment of this population was 70 months (range, 0.2–95.2). These results should be confirmed in an independent cohort and with a longer follow-up.

It is well established that the heavy and the light chain repertoire is biased in CLL (8–10). When we analyzed the HV repertoire of IgLV3-21, we observed that this light chain is also predominantly associated with particular heavy chains, because the IgHV3-21, IgHV3-48, and IgHV3-23 families represent 50% of all IgHVs. Interestingly, 90% of IgLV3-21 light chains were unmutated and the corresponding heavy chains were less frequently mutated compared with other control groups. This confirms that the BCR, including its light chain, is globally conserved, as illustrated by the consensus sequence. In accordance with previous studies, we show that some amino acid changes are recurrent, such as the S→G replacement in the CDR3 of the light chain, which occurs in 30% of patients (10, 27). It has been suggested that the introduction of a glycine from the germline sequence at this position likely facilitates BCR-BCR interactions (32). However, at least in our patients, this did not have prognostic significance.

Using the ARResT/AssignSubsets tool (17), we observed that, except in subset #2 patients (representing 27% of the IgLV3-21 patients), the other heavy chain was rarely assigned to a major stereotyped subset. In the other 104 IgLV3-21 patients, only 3 patients displayed stereotyped IgHV (subset #5, #5 and #7H). However, recent work shows that immunoglobulin B-cell receptors of unrelated clonotypes may recognize common epitopes that are not predictable from the molecular features of the immunoglobulin B-cell receptor, including CDR3 composition and length (33). It is thus premature to exclude the stereotypy hypothesis based only on amino acid sequences.

The biological reasons underlying the poor prognosis of IgLV3-21 patients are still unclear. Two hypotheses based on BCR stimulation could explain the poor prognosis of these patients: canonical (auto-)antigen binding (34) or cell-autonomous antigen-independent signaling (7). It is now well-accepted that CLL selection is an antigen-driven process and that BCR is a key player in CLL cell survival (35). Regarding the extremely conserved sequence of the IgLV3-21, we could speculate that this particular light chain may recognize a common external antigen (viruses or other pathogens) or an autoantigen (apoptotic bodies).

On the other hand, the BCR stimulation leading to CLL cell survival and thus a poor prognosis for IgLV3-21 patients could be the result of cell-autonomous antigen-independent signaling, as described by Dühren-von Minden and colleagues (7), who were the first to suggest that CLL cells could be activated by the triggering of BCR internal epitopes. In line with this hypothesis, Minici and colleagues (32) demonstrated the importance of the light chain IgLV3-21 in BCR self-recognition and autonomous signaling by studying the crystal structure of homotypic B-cell receptor interactions in subset #2 patients. Together, these results suggest that the IgLV3-21 light chain might play a pivotal role in homotypic interactions and suggest that CLL cells expressing IgLV3-21 might undergo continuous and autonomous constitutive stimulation promoting CLL cell activation, proliferation, and survival.

In line with these two hypotheses, we observed a decrease in CXCR4 cell surface expression in IgLV3-21 patients (Supplementary Fig. S7). CXCR4 is known to be downregulated after IgM stimulation (36) and in cells receiving stimulus from a stromal microenvironment (37). Interestingly, all IgLV3-21 patients (with UM or M IgHV or with IgHV3-21) presented a low level of CXCR4 surface expression, which was significantly different from patients with M IgHV-expressing another light chain. These results suggest that IgLV3-21 patients undergo an in vivo BCR stimulation (through interactions with their microenvironment or through autonomous signaling), which downregulates CXCR4 and provides a survival signal to the cells.

In conclusion, we have demonstrated that the presence of the IgLV3-21 light chain independently defines a new subgroup in CLL characterized by a distinct gene expression profile linked to translational upregulation, increased MYC target gene expression and a prognosis similar to IgHV UM patients, irrespective of whether the IgLV3-21 rearrangement occurs in IgHV-mutated or unmutated patients. Therefore, similar to its heavy chain counterpart, the light chain IgLV3-21 might prove a valuable marker to differentiate high-risk from low-risk IgHV M CLL patients.

P. Hillmen reports receiving commercial research grants from AbbVie, Gilead, Janssen, Pharmacyclics, and Roche Pharmaceuticals, and speakers bureau honoraria from AbbVie, Acerta Pharma, Janssen, and Roche Pharmaceuticals. A. Schuh is a consultant/advisory board member for AbbVie, Gilead, Janssen, and Roche. No potential conflicts of interest were disclosed by the other authors.

The views expressed are those of the authors and not necessarily those of the UK's National Health Service, National Institute of Health Research, or Department of Health.

Conception and design: B. Stamatopoulos, P. Hillmen, A. Schuh

Development of methodology: B. Stamatopoulos, E. Crompot, B. Verhasselt

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Mraz, P. Robbe, A. Burns, A. Timbs, D. Bruce, P. Hillmen, N. Meuleman, P. Mineur, R. Firescu, M. Maerevoet, V. De Wilde, A. Efira, J. Philippé, B. Verhasselt, F. Offner, H. Dreau, A. Schuh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Stamatopoulos, T. Smith, R. Clifford, P. Robbe, A. Burns, F. Offner, D. Sims, A. Heger, A. Schuh

Writing, review, and/or revision of the manuscript: B. Stamatopoulos, T. Smith, D. Bruce, P. Hillmen, R. Firescu, M. Maerevoet, J. Philippé, F. Offner, D. Sims, A. Schuh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Pieters, P. Robbe, R. Firescu, J. Philippé, H. Dreau

Study supervision: P. Hillmen, D. Sims, A. Schuh

This work was supported by a Télévie grant provided by the F.R.S.-FNRS (Fonds National de la Recherche Scientifique) of Belgium, the “David and Alice Van Buuren Fund,” the “Fonds IRIS-Recherche,” “Les Amis de l’Institut Jules Bordet,” “Wallonie-Bruxelles International World” (WBI.World), the Bekales Fundation and the foundation Lambeau-Marteaux, and the National Institute for Health Research (NIHR) Oxford Biomedical Research Center (BRC). M. Mraz was supported by the Ministry of Health of the Czech Republic, grant 16-29622A.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.