A total of 465 diagnostic and selected MRD-positive follow-up samples from 280 patients aged 18-55 years with BCP-ALL were investigated within the frame of a research project associated with the German Multicenter Adult Acute Lymphoblastic Leukemia (GMALL) 08/2013 therapy optimization trial (EudraCT-No.: 2013-003466-13). Initial diagnostic samples were subjected to IGH amplicon-based HTS, multiparameter flow cytometry, and transcriptome sequencing (RNA-Seq) analyses. An overview of patient characteristics is detailed in Supplementary Table 1. From all 280 patients either diagnostic BM (n=217) or PB (n=63) samples were available (Figure 1, cohort #1). Concurrent diagnostic PB and BM samples were available for pairwise comparisons in 46/280 cases (Figure 1, cohort #2). In addition, a total of n=139 MRD-positive early or late follow-up and relapse samples of 63 patients were selected. Clonal IGH evolution was monitored during early therapy phases, in particular during the cyclophosphamide/dexamethasone-containing prephase (Figure 1, cohort #3, 66 paired samples from 33 patients) and after Induction I including one dose of rituximab (Figure 1, cohort #4, 82 paired samples from 41 patients). Ultimately, we studied 149 diagnostic and follow-up samples of 34 patients to track the kinetics of clonal evolution over time in longitudinal case reports (Figure 1, cohort #5). Main results and details on IGH sequences within the cohorts #1-5 are summarized in Supplementary Table 4. All patients provided informed consent. All research described herein was approved by the Frankfurt Research Ethics Board (188/15F) and performed in accordance with the Declaration of Helsinki.

Initial diagnoses were established by standard routine diagnostics. Initial immunophenotype, BCR::ABL1 and KMT2A rearrangement status, initial molecular IG/TR markers, and MRD at follow-up were obtained by GMALL trials central diagnostic reference laboratories (Berlin & Kiel, Germany).

RNA was extracted according to standard procedures recommended by the manufacturer (Trizol, Life Technologies, Carlsbad, CA). Library preps, transcriptome sequencing, and molecular subtype calling were performed as described previously (26). Cases with intermediate or divergent gene expression profiles could not be assigned to any existing molecular profile category and were thus described as “other”.

We employed the EuroClonality-NGS assay and the IGH-VJ-FR1 and IGH-DJ primer sets to sequence diagnostic and follow-up samples. In general, 100 ng of DNA, extracted according to standard procedures recommended by the manufacturer (AllPrep, Qiagen, Hilden, Germany), was used for the analysis of the diagnostic samples. In the longitudinal analysis of diagnostic and follow-up samples, we used 500 ng DNA to ensure adequate sensitivity and direct comparability of results. The EuroClonality-NGS central in-tube quality/quantification control (cIT-QC) was spiked into most diagnostic samples and all follow-up samples as a library-specific quality control and for the normalization of abundance from reads to cells (25, 27, 28). (www.euroclonality.org/protocols). Samples were sequenced on a MiSeq (Illumina, San Diego, CA, USA) with 2x250bp reads (sequencing depth analysis in Supplementary Material S1).

IGH sequences were analyzed with ARResT/Interrogate (25). Usable reads, the denominator for percentage abundance calculations, were defined as sample reads with identified junctions after the exclusion of cIT-QC reads, in other words, reads with patient-only IGH rearrangements. For better readability, especially in figures, we will refer to IGH segment rearrangements using name acronyms omitting ‘IGH’ from gene names, e.g., V_H instead of IGHV or DJ_H instead of IGHD-IGHJ.

For IGH clonal evolution assessment, we isolated the ‘DNJ-stem’ of nucleotide junctions - a subsequence of the junction that consists of the last maximum of 3 D_H nucleotides (or if none are identifiable in a VJ_H rearrangement, of the N region), any N nucleotides between D_H and J_H, and all J_H nucleotides (Figure 2). The DNJ-stem definition is adapted to the understanding that the corresponding junction subsequence is the one remaining generally stable after multiple D_H/V_H to DJ_H (D_H/V_H-DJ_H) recombination and V_H replacement events, and it is therefore used as the link, a shared denominator, between potential family members – whether those clonotypes should indeed be considered related and whether the DNJ-stem should be considered evolving was thoroughly examined by deploying further rules.

We initially removed - as potential noise - clonotypes highly similar to the most abundant DNJ-stem family member, as well as of abundance below 3 reads or below 0.01% of usable reads. We then considered the number of clonotypes removed above (a high number may indicate an overall noisy DNJ-stem), the absolute and relative (to the sample total) number of remaining DNJ-stem family members, the number of their different 5’ genes and junction lengths as a sign of expected junctional variability, as well as the DNJ-stem’s N region complexity to increase confidence in its specificity. Combined consideration of all these data points led to a consensus call on whether a DNJ-stem was evolving or stable. Please see Supplementary Material S2 for the detailed pseudocode of this algorithm.

V_H replacement was identified by considering (i) the general stability of the D_H region’s 5’-site and its preceding N1 region’s 3’-site and (ii) the presence of a sequence remnant (fingerprint) of the replaced V_H in the new junction. Evolving DNJ-stems with the absence of aforementioned signs were categorized as D_H/V_H-DJ_H recombinations.

If a DNJ-stem was detected in both the IGH-VJ and IGH-DJ library, the respective DNJ -stem was termed ‘rooted’, indicating that the underlying root DJ_H rearrangement was also detectable.

A marker DNJ-stem was defined if any of the following conditions were fulfilled: (i) abundance as a sum of all DNJ-stem family members ≥ 5% of usable reads or ≥ 1% of cells (normalized abundance); (ii) presence of clonal evolution (‘evolving’) at any abundance. We recovered the locus IGHV and IGHD gene order from IMGT (29).

Statistical analyses were performed in R version 4.1 using Fisher’s exact test with default settings, meaning the significance threshold was set at 0.05.

We studied DNJ-stems and the respective VDJ_H and DJ_H rearrangements in the diagnostic cohort of 280 BCP-ALL patients (Figure 1, cohort #1). DNJ-stems with only DJ_H rearrangements were further analyzed only if evolving or otherwise noteworthy.

Clonal evolution, i.e. at least one evolving DNJ-stem, was found in 84/280 (30%) patients. Stable (non-evolving) DNJ-stems were found in 160/280 (57.1%) patients and 36/280 (12.9%) patients had no detectable IGH rearrangements to constitute a marker DNJ-stem – the latter being a rate comparable to previous studies (30, 31). In summary, clonal evolution was prevalent in our cohort affecting one-third of the cases (Figure 2).

In total, 511 marker DNJ-stems (i.e., of ≥ 5% usable reads or evolving at any abundance) were identified in 244/280 (87.1%) patients (range 1-11, average 2.1 per patient). In 116 (65.5%) out of the overall detected 177 evolving DNJ-stems, their most abundant DNJ-stem family member stayed below the generally accepted marker threshold of 5% usable reads and therefore may have been ignored on its own. In other words, applying the standard marker screening approach (by following single clonotypes and not DNJ-stems and using the conventional 5% threshold), 65.5% of markers would not have been reported. By screening for DNJ-stem instead of clonotypes, but still applying the 5% threshold, two-thirds (74) of those 65.5% of markers would still not have been reported. A visual breakdown of the DNJ-stems and family member distribution is depicted in Figure 2.

We studied alignments of junction nucleotide sequences of evolving DNJ-stem families from the diagnostic BCP-ALL samples (Figure 1, cohort #1) and observed signs of the two main mechanisms driving IGH clonal evolution, V_H to DJ_H (V_H-DJ_H) recombinations and V_H replacements. Additionally, we identified a variant of somatic recombination, in which D_H to DJ_H (D_H-DJ_H) recombination was evident (10, 11). Both mechanisms, D_H/V_H-DJ_H recombination and V_H replacement, leave distinctive changes in the recombined nucleotide sequences, by which they can be specifically discriminated (Figure 3).

Ongoing D_H/V_H-DJ_H recombination was observed in 78% of evolving DNJ-stems. This mechanism results in sequence trimming during the rearrangement of the D_H/V_H to the existing DJ_H, which in addition to the random editing of N nucleotides results in sequence variability outside the DNJ-stem. This is exemplified in Figure 4A, case #61 and Figure 4B, lower part, case #126. In the illustration, the top row is the ‘mother’ clonotype for the family of clonotypes to be produced by parallel D_H-DJ_H and V_H-DJ_H recombination (case #61) or solely by V_H-DJ_H recombination (case #126).

The vast majority of D_H/V_H-DJ_H cases featured only V_H-DJ_H recombination-driven clonal evolution, but there were notable exceptions. In particular, of 398 DJ_H marker DNJ-stems identified in 144/280 (51.4%) patients, 21/398 (5.2%) DNJ-stems were evolving in 15 patients - in 12/15 (80.0%) of these patients this ongoing D_H-DJ_H recombination accompanied the V_H-DJ_H recombination, both driving clonal evolution (Figures 4A, B). The mechanism of V_H replacement was not active in those cases. Of note, the only two patients with D_H-DJ_H recombination without accompanying V_H-DJ_H recombination were assigned to the rare molecular subtype CDX2/UBTF, a novel BCP-ALL subgroup with the need of intensified treatment that was identified in 7% of our cohort (26) (see below).

Ongoing V_H replacement was observed in 22% of evolving DNJ-stems. T his mechanism is defined by its most distinct signs: (i) the general stability of the 5´-D_H region site and its preceding 3´-N1 region site; (ii) the actual nucleotide remnants of the replaced V_H in the new junction. There were very few exceptions to the requirement of the incoming V_H to be downstream from the mother clonotype segment: 1.6 upstream segments on average (range 1-5) in lowly abundant clonotypes, in 13/39 DNJ-stems evolving by V_H replacement, representing 1.9% of all family members. These could be erroneous gene annotations facilitated by sequence errors, false positive DNJ-stem members, or perhaps true members having originated previously. A case with ongoing V_H replacement is detailed in Figure 4B (upper part, case #126), where the top row is the ‘mother’ clonotype for the family members to be produced by replacement of its V_H segment. The resulting family members underneath show no further deletions in the D_H segment, because any sequence loss was buffered by the N1 region and remnants of the replaced V_H (highlighted in yellow) and now residing inside the new N1 regions.

Overall, and in contrast to V_H replacement that most commonly takes place in the presence of one or more highly abundant mother clonotypes, ongoing V_H-DJ_H recombination results in bursts of numerous lowly abundant family members with the role of the mother clonotype taken by the DJ_H root. This translates to an average abundance of the most abundant family member of 51% for V_H replacement and 7% for V_H-DJ_H recombination in the IGH-VJ library of our diagnostic cohort.

We selected one DNJ-stem from each patient or two DNJ-stems from each of two patients with concomitant V_H-DJ_H recombination and V_H replacement - a total of 53 V_H-DJ_H-evolving stems and 33 V_H replacement-evolving stems - focusing on evolving DNJ-stems with numerous family members, to highlight specific characteristics related to both mechanisms driving clonal evolution. For each DNJ‐stem, abundances and IGH evolution classifications such as evolving or stable, DJ_H root detectable or not, D_H/V_H-DJ_H recombination or V_H replacement were annotated along with its locus-ordered V_H gene profile and basic patient metadata. The results are summarized in an overview tabular heatmap (Figure 5).

Of the 244 diagnostic samples of cohort #1 with identified marker DNJ-stems, 41 (16.8%) were classified as pro-B ALL and 203 (83.2%) as pre-B/c-ALL by multiparameter flow cytometry.

We detected significantly higher rates of clonal evolution in patients with a pro-B ALL compared to those with a pre-B/c-ALL immunophenotype (61.0% vs. 29.1%, respectively, p=0.0008), suggesting that the maturation arrest at earlier B-cell development stages might increase the frequency of ongoing IGH recombination.

Other noteworthy observations are that (i) V_H-DJ_H overwhelmingly featured ‘early’, J-proximal V_H genes, mainly V_H6-1 (42/53 (79.2%) DNJ-stems and most family members) - in strong contrast, V_H6‐1 featured only twice in V_H replacement cases, in which the ‘late’, J-distal V_H3-74 segment was the most frequent; (ii) the DJ_H root was present in the majority of V_H-DJ_H DNJ-stems but in only one V_H replacement DNJ-stem; (iii) D_H-DJ_H-driven clonal evolution was observable in V_H-DJ_H cases but not in V_H replacement cases; (iv) the two mechanisms, V_H-DJ_H recombination and V_H replacement, were almost mutually exclusive (Supplementary Tables 2, 3).

The BCP-ALL molecular subtypes were assigned based on RNA-Seq data available for 167/280 (59.6%) cases and then correlated with the IGH clonal evolution patterns. Overall, 155/167 cases could be assigned to a previously described molecular profile category, while 12 cases showed other gene expression profiles and thus were summarized as “other”. Notably, the aforementioned higher rates of clonal evolution in pro-B ALL were almost exclusively linked to the KMT2A subtype (94.4% evolving), while other molecular subtypes with a pro-B ALL immunophenotype were mostly stable, particularly CDX2/UBTF (cases with UBTF::ATXN7L3 gene fusion, only 22.2% evolving and those only in the IGH-DJ library) and ZNF384 (only 30% evolving) (Figure 6A). Further, the KMT2A subtype showed significantly higher rates of ongoing V_H-DJ_H (16/17 (94.1%) evolving patients, p=3e^‐11), prevalently initiating the recombination with ‘early’ J-proximal V_H genes, mostly V_H6-1 (Figure 5). The preferential usage of ‘early’ J-proximal V_H genes was also seen in the molecular subtypes PAX5 P80R and PAX5alt. Although (and in contrast to the KMT2A subtype) most DNJ‐stems in those subtypes were stable (30/40, 75%), the evolving DNJ-stems (10/40, 25%) clearly showed a prevalent (as in 7/10) usage of ‘early’ J-proximal V_H segment genes including V_H6-1, V_H1‐2, V_H1-3 and V_H4-4. Notably, the three DNJ-stems using more J-distal V_H segments belonged to a single patient with concurrent V_H replacement and V_H-DJ_H recombination activity (Supplementary Table 2). While the KMT2A subtype was strongly associated with V_H-DJ_H recombination as the prevalent clonal evolution mechanism, the molecular subtypes Ph-like and DUX4 featured no V_H-DJ_H but were instead associated with V_H replacement (7/37 and 4/13, p=0.00009 and p=0.067, respectively) (Figure 6A).

The D_H-DJ_H evolution was a less frequent observation and only detectable in 11 of 155 (7.1%) patients with molecular subtype assignment, with the low number of patients impeding significant associations with molecular subtypes. Despite the fact that the KMT2A subtype not only showed the highest rate of V_H-DJ_H evolution, but also significantly higher rates of D_H-DJ_H evolution (27.8% evolving, p=0.0001), we found that the presence of D_H-DJ_H and V_H-DJ_H was not fully concordant (Figure 6B). Strikingly, two patients of the rare CDX2/UBTF subtype showed only D_H-DJ evolution (p=0.04), whereas there were no signs of V_H-DJ_H evolution. Overall, the evolving D_H-DJ_H rearrangements showed a remarkable preference for the usage of D_H2-2 as the incoming D_H (14/15 (93.3%) of DNJ-stems) (Supplementary Table 3), which was also observed by others but in the context of normal rearrangements in blood B cells from healthy donors (32).

We sought to understand the overlap of DNJ-stems and their family members between diagnostic BM and PB of the 46 patients with both diagnostic materials available (Figure 1, cohort #2). We identified 39 patients with paired BM and PB DNA samples and with marker DNJ-stem(s) in at least one of these two compartments. The total number of DNJ-stems was 93, with 66 DNJ-stems in both compartments as marker DNJ-stems (see Methods for definition), 21 DNJ-stems also in both compartments but appearing in one compartment outside the marker DNJ-stem rules (and almost always at very low abundance), and the remaining 6 appearing only in BM compartment (Figure 7). This translated into 37/39 patients with at least one of the 87 DNJ-stems in both materials, or 36/39 when we demanded marker DNJ-stems in both materials, i.e., considering only the 66 DNJ-stems; all other cases had the DNJ-stem in BM only. Of the 66 DNJ-stems in both the BM and PB compartments of the 36 patients, all but two had concordant clonal evolution status between BM and PB, with the two exceptional cases evolving in BM and stable in PB. In 32/36 patients we saw a complete marker DNJ-stem overlap between BM and PB.

When we compared the overlap of ~3300 family members of 26 DNJ-stems that were evolving both in BM and PB, we found the overlap to be only 8.7%, with the rest split almost equally between BM and PB (Figure 7). Evolving DNJ-stem abundance in BM was on average 18.7% reads (range 0.6-89.9%), while the average abundance of family members was 0.16% reads (range 0.0008-71.8%). In PB the evolving DNJ-stem abundance was on average 12.9% reads (0.2-76.8%) and the average abundance of family members 0.11% (range 0.0002-73.8%). The fact that the minimum abundance of family members is below the 0.01% reads threshold used for family members when calling clonal evolution, is due to searching for members across materials at any abundance to maximize the sensitivity and information content of the overlap analysis.

In order to understand the longitudinal dynamics of evolving clonotypes during and after treatment, we compared therapy-naive BCP-ALL samples (day 0) and matched samples from different follow-up time points (Figure 1, cohort #3-5): (i) day +6 samples after the 5-day cyclophosphamide/dexamethasone-containing prephase, no rituximab (cohort #3); (ii) day +22 samples after Induction I including one dose of rituximab (cohort #4); and (iii) later time points including additional doses of rituximab and refractory disease/relapse-related secondary immunotherapies, e.g. blinatumomab (cohort #5).