Mosaic Chromosomal Alterations Identify Ultra‐High Risk Clonal Hematopoiesis in Patients With Lymphoma Undergoing Intensive Chemotherapy

To the Editor,

Clonal hematopoiesis of indeterminate potential (CHIP, variant allele frequency [VAF] > = 2%) is an age‐associated condition identified by small somatic genetic aberrations in hematopoietic cells. CHIP has been linked to inferior overall survival (OS) in the general population and to an increased risk of therapy‐related myeloid neoplasms (tMN) in patients with lymphoma [1, 2, 3]. High‐dose chemotherapy followed by autologous stem‐cell transplantation (ASCT) has historically been the standard salvage therapy for refractory or relapsed lymphoma and still remains a therapeutic option for late relapse of diffuse large‐cell lymphoma or follicular lymphoma with early progression after first‐line therapy and following failure of chimeric antigen receptor (CAR) T cell therapy [4, 5, 6].

Prior studies have reported conflicting findings regarding the prognostic impact of CHIP on survival following high‐dose chemotherapy for lymphoma. Meanwhile, inferior overall survival has been observed in carriers of some CHIP mutations, while the importance of other CHIP mutation subgroups remains incompletely defined [2, 7, 8, 9]. Mosaic chromosomal alterations (mCAs) are large acquired somatic chromosomal aberrations, typically spanning kilobases to megabases, that are present in a subset of cells, and their prevalence in hematopoietic cells increases with age. Autosomal mCAs are detectable in 3.6% of healthy individuals of European ancestry aged 40–70 years, and are associated with an increased risk of hematologic cancer and mortality [10, 11, 12, 13, 14].

In healthy individuals and in patients with solid cancer, the co‐occurrence of CHIP with mCAs confers a higher risk of hematologic malignancy than either lesion alone [15, 16]. To our knowledge, the combined impact of CHIP and mCAs in lymphoma has not yet been examined. We hypothesized that concurrent CHIP and mCA may constitute a high‐risk subgroup among patients with lymphoma undergoing intensive chemotherapy.

From a cohort of 440 patients with various lymphoma subtypes treated with high‐dose chemotherapy followed by ASCT and analyzed for CHIP variants [7] (Analyzed CHIP genes are found in Table S1), we selected 188 peripheral blood stem‐cell harvest samples for mCA analysis based on microarray capacity. A total of 106 patients with CHIP mutations (“CHIP” group) were selected by sample availability and 82 patients without CHIP (“No CHIP” group) were selected based on age‐matching with the CHIP group. DNA (525 ng/sample) was genotyped on the Illumina Infinium OmniExpress‐24 single‐nucleotide polymorphism (SNP) array (Illumina, San Diego, CA, USA) at Eurofins Genomics (Ebersberg, Germany).

Clinical and pathological data were updated for the present study. As in the previous study [7] we assessed the development of tMN, which included both pathologically confirmed and clinically suspected events (at least one bone marrow biopsy due to unexplained cytopenia, also described as therapy‐related CCUS [17]).

For the identification of mCAs, SNP array IDAT files were converted to VCF using gtc2vcf and human reference genome GRCh37 (https://github.com/freeseek/gtc2vcf); samples were phased with SHAPEIT4 [18] and mCAs detected using MoChA with recommended filters. Undetermined mCAs were retained, as they represent likely clonal events without sufficient signal resolution for confident copy‐number classification [13, 14]. To weigh the significance of each CHIP‐mCA co‐occurrence, we devised a log enrichment score (Note 1) representing co‐occurrence count in relation to the product of individual CHIP and mCA counts. Fisher's exact test was used for statistical comparisons unless explicitly noted. Overall survival was compared by log‐rank test and cumulative incidence functions were compared using Gray's test with death as a competing event. Cox proportional hazards models were used to estimate cause‐specific hazard ratios for tMN and overall survival.

To investigate differences in types of lymphoma patients from the original cohort, we compared the various histologies in the full cohort and the subset analyzed for mCAs and CHIP. Only two lymphoma subtypes showed a difference in prevalence of more than 2%. As seen, DLBCL was more prevalent in our study cohort compared to Husby et al. [7] (44.1% in this study vs. 36.8% in Husby et al.), while HL was less prevalent (HL, 10.1% vs. 14.5%) (Table S2).

A total of 54 mCAs (42 autosomal, eight loss of Y in males and four loss of X in females) were identified in 35 patients (18.6%) (Table S3, “Autosomal mCA overview”). The mCA landscape varied with regards to type of mCAs (gains, losses, copy‐number neutral loss of heterozygosity [CNLOH]), with different chromosomes and events between the groups (Figure S1, “Chromosomal mCA overview”). The autosomal mCA size ranged from 7.2 kb to 141.2 Mb with a median size of 3.19 Mb (Table S4, size distribution shown in Figure S2). Given the limited ability to detect sex chromosome mCAs in this cohort, subsequent analyses focused exclusively on autosomal events (hereafter “mCA”).

Autosomal mCAs were present in 16 patients in the ‘CHIP’ group (15.1% of CHIP carriers), and eight patients (9.8%) in the ‘No CHIP’ group (Table 1, cohort summary where undetermined mCAs are removed is found in Table S5). In the ‘CHIP’ group, six patients (5.7%) carried a single mCA, while seven patients (8.5%) in the “No CHIP” group had a single mCA (Figure 1A p = 0.565 compared to CHIP). Patients with CHIP were significantly enriched for two or more autosomal mCAs; 10 patients (9.4%) from the “CHIP” group had two or more mCAs, while one patient (1.2%) in the ‘No CHIP’ group had two or more mCAs (Figure 1A p = 0.025). Meanwhile, among patients in the “CHIP” group, DNA repair CHIP conferred a higher risk of having two or more mCAs compared to “other CHIP” (OR = 4.89, p = 0.03, Table S6).

TABLE 1.

Baseline characteristics of study participants stratified by CHIP.

		Full cohort	CHIP		No CHIP
		All patients (n = 188)	No autosomal mCA (n = 90)	Autosomal mCA (n = 16)	No autosomal mCA (n = 74)	Autosomal mCA (n = 8)
Age (years)	Median (IQR)	60 (53–64)	60 (53–64)	62 (56–64)	59 (51–63)	63 (59–65)
Sex	Male	119 (63.3%)	55 (61.1%)	9 (56.2%)	51 (68.9%)	4 (50.0%)
	Female	69 (36.7%)	35 (38.9%)	7 (43.8%)	23 (31.1%)	4 (50.0%)
Lymphoma Subtype	DLBCL	83 (44.1%)	36 (40.0%)	7 (43.8%)	36 (48.6%)	4 (50.0%)
	PTCL	30 (16.0%)	21 (23.3%)	3 (18.8%)	6 (8.1%)	0 (0.0%)
	MCL	34 (18.1%)	14 (15.6%)	0 (0.0%)	18 (24.3%)	2 (25.0%)
	HL	19 (10.1%)	9 (10.0%)	0 (0.0%)	9 (12.2%)	1 (12.5%)
	FL	13 (6.9%)	7 (7.8%)	3 (18.8%)	3 (4.1%)	0 (0.0%)
	Other	9 (4.8%)	3 (3.3%)	3 (18.8%)	2 (2.7%)	1 (12.5%)
ASCT line of therapy	1. Line	74 (39.4%)	35 (38.9%)	4 (25.0%)	32 (43.2%)	3 (37.5%)
	≥ 2. Line	114 (60.6%)	55 (61.1%)	12 (75.0%)	42 (56.8%)	5 (62.5%)
Radiotherapy prior to ASCT	Total body irradiation	3 (1.6%)	2 (2.2%)	1 (6.2%)	0 (0.0%)	0 (0.0%)
	Field radiation	22 (11.7%)	9 (10.0%)	1 (6.2%)	12 (16.2%)	0 (0.0%)
	No radiotherapy	111 (59.0%)	51 (56.7%)	11 (68.8%)	43 (58.1%)	6 (75.0%)
	Unknown	52 (27.7%)	28 (31.1%)	3 (18.8%)	19 (25.7%)	2 (25.0%)

FIGURE 1.

Landscape of mCA and CHIP in patients with lymphoma undergoing ASCT. (A) Number of patients with 0, 1 and 2+ autosomal mCAs in the CHIP and No CHIP groups, p values are shown between groups for each mCA count, significant p values are bold. (B) Co‐occurrence of mCAs and CHIP mutations. Custom log enrichment scores for mCA and CHIP co‐occurrences with missing events masked, only showing the five most mutated CHIP genes. Fisher's p values are shown under the enrichment value, significant p values are marked in red. (C) Overall survival according to CHIP occurrence and autosomal mCA co‐occurrence. Number at risk is placed below the x‐axis. Log‐rank p value is calculated based on the 10‐year overall survival. (D) Cumulative incidence of observed and suspected tMN according to CHIP occurrence and autosomal mCA co‐occurrence. Number at risk is placed below the x‐axis. Grays‐test p value is calculated based on the 10‐year cumulative incidence. A.

Next, we investigated how the CHIP and mCA landscape correlated. TP53 and PPM1D CHIP mutations co‐occurred with multiple mCA subtypes, of which TP53 with chromosome 9 gain (p = 0.019), and PPM1D with chromosome 9q loss were significant (p = 0.038, Figure 1B); TET2 associated with chr20q copy‐neutral loss of heterozygosity (CN‐LOH) and DNMT3A with chr2q loss, while ASXL1‐mutant patients had no mCAs (“Autosomal mCA overview”, Tables S7, S8, Figure 1B). Grouping of patients by different CHIP mutations showed a trend toward more autosomal mCAs in patients with TP53 (26.6% of patients with TP53, p = 0.087) and PPM1D mutations (23.8% of patients with PPM1D, p = 0.091) compared to patients without CHIP (Figure S3A, grouping summary in Table S9). It should be mentioned that co‐occurring variants identified might be found by chance and not necessarily co‐exist in the same clone. We observed a higher frequency of mCAs per individual among 22 individuals with intermediate VAF (5%–10%) compared to 21 individuals with VAF ≥ 20% (Negative binomial p = 0.034, Figure S3B).

Next, we investigated the effect of mCAs on survival after ASCT. CHIP carriers with an autosomal mCA had significantly worse 10‐year OS than patients with neither CHIP nor mCAs (log‐rank p = 0.001) and worse OS than those with CHIP alone (p = 0.019; Figure 1C). Excluding undetermined mCAs did not change the significance of the overall survival results (Figure S4). To contextualize the association between CHIP, mCAs, and inferior overall survival, causes of death were reviewed in 74 patients with available follow‐up. Progressive lymphoma was the predominant cause of death across all groups. However, therapy‐related deaths in patients with disease progression increased across groups, from 0% in No CHIP & no mCA, to 13.5% in CHIP & no mCA, 20% in No CHIP & mCA, and 25% in CHIP & mCA. In contrast, deaths from isolated progressive disease remained similar between groups (Table S10).

Even when including information on DNA repair pathway CHIP mutations, the presence of mCAs was still a negative prognostic factor for overall survival (Figure S5). Among patients with “other CHIP” mutations, those with a co‐occurring mCA had significantly reduced OS compared to “other CHIP” patients without mCAs (p = 0.001) and to patients with neither CHIP nor mCA (p < 0.001) (Figure S5). Notably, the presence of “other CHIP” and mCA showed a trend toward even worse outcomes than with DNA repair pathway CHIP mutations, with all patients deceased at approximately year 5.

tMN developed in 12 patients and was clinically suspected, but not pathologically confirmed, in 22 additional patients (specific counts for patients with CHIP or mCA is shown in Table S11). Incidence of tMN did not differ significantly across groups defined by CHIP or mCA status (Figure 1D). In subgroup analyses stratified by CHIP subtype and mCA status, a higher cumulative incidence was observed in patients with DNA pathway repair mutations without mCAs, when compared to “No CHIP & no mCA” patients (p = 0.021), and to patients with “other CHIP & no mCA” (p = 0.033). The analysis of “other CHIP & mCA” was insufficiently powered (Figure S6, group counts are found in Table S12). In cause‐specific Cox models, concurrent CHIP and mCA were independently associated with inferior overall survival, whereas no independent association with development of tMN was observed after adjustment for age and ASCT line (Figure 2).

FIGURE 2.

Association between CHIP/mCA groups and outcomes in Cox proportional hazards models. CHIP and mCA stratified hazard ratios are shown relative to the reference group (No CHIP & no mCA). For tMN, cause‐specific hazard ratios were estimated by modeling tMN as the event and censoring at death or last follow‐up (time‐to‐event defined as time to tMN, death, or last follow‐up). Adjusted models include age ≥ 60 years and line of ASCT (≥ 2 vs. 1).

In this study, we demonstrate that mCAs are prevalent and prognostically important in lymphoma patients undergoing ASCT. This study focuses on a subset of patients with CHIP mutations and age‐matched patients without CHIP; therefore, some differences in lymphoma subtype prevalence were observed. Patients with co‐occurring CHIP and mCAs had markedly increased mortality compared to patients without mCAs, regardless of CHIP status. These findings indicate that large‐scale acquired hematopoietic chromosomal aberrations confer prognostic risk beyond that associated with canonical CHIP mutations. Although limited by small subgroup sizes, cause‐of‐death data suggest that the inferior survival associated with mCAs is not primarily driven by manifest t‐MN or relapse biology alone. Instead, mCA and CHIP clones may reflect an underlying vulnerability to progression‐associated treatment complications and impaired hematopoietic resilience after ASCT. This is supported by a higher proportion of therapy‐related deaths during lymphoma progression in affected patients. These observations are descriptive and do not establish causality between mCAs, CHIP, and mortality.

The enrichment of multiple autosomal mCAs among CHIP carriers, and specifically those with dna repair mutations suggest an interplay between gene‐level mutations and chromosomal instability, previously described in healthy individuals and solid cancers, though not yet observed in lymphoma [16, 19]. Single‐cell whole‐genome sequencing is needed to fully validate this finding.

Interestingly, in contrast to a prior study [16] demonstrating increasing co‐occurrence of mCAs with larger CHIP clone size (e.g., higher VAF), we observed a higher frequency of mCAs among individuals with intermediate VAF (5%–10%) compared to those with VAF ≥ 20%. This discrepancy may reflect cohort‐specific effects related to lymphoma or chemotherapy exposure. Thus, the sample size of the cohort does not permit firm conclusion regarding the relationship between VAF and mCA co‐occurrence.

Prior national and single‐center cohort studies have shown that CHIP mutations in DNA repair genes predict late mortality after ASCT [2, 7, 9]. By profiling mCAs in a subset of our national cohort, we demonstrate that mCAs co‐exist with CHIP and confer risk exceeding that of CHIP alone, even after exclusion of patients harboring DNA repair mutations.

Notably, several late tMN events (> 4 years) after ASCT occurred in patients without detectable CHIP or mCA. These late tMN events may reflect subsequent lines of chemotherapy and are unlikely to be captured by CHIP and mCA analyses performed at the time of ASCT.

We confirmed increased cumulative incidence of tMN in patients harboring DNA repair pathway CHIP and no mCA, while the lack of significant associations with tMN and CHIP or mCA is likely attributable to small subgroup sizes and limited event numbers, resulting in insufficient statistical power. Our analysis assumes that analyzed samples predominantly contain hematopoietic cells rather than circulating lymphoma cells, as patients are expected to be in complete remission at the time of stem cell collection, and the lymphoma subtypes in our cohort rarely present with leukemic disease. Given detection thresholds of 4% for CHIP and 1% for mCAs, substantial contamination by circulating lymphoma cells is unlikely, although it cannot be entirely excluded. In addition, the sample size limits detection of rare mCA‐CHIP combinations. Furthermore, longitudinal studies with sampling before and after intensive chemotherapy will be essential to delineate clonal dynamics and establish causality.

In conclusion, the combined assessment of mCAs and CHIP in 188 patients undergoing high‐dose chemotherapy and ASCT identifies a distinct ultra‐high risk subgroup of patients with increased mortality. Integration of clonal hematopoiesis analysis with both mCA and CHIP profiling may therefore provide a valuable biomarker for adverse outcomes and support non‐chemotherapy containing treatment strategies.

Author Contributions

M.Æ.H. and S.H. initiated the study. S.H. and K.G. obtained funding. M.Æ.H., K.G., and S.H. designed the study. M.Æ.H., S.H., and F.G.R.‐G. performed the sequencing. M.Æ.H. and F.F. did bioinformatic analysis. M.Æ.H. did the statistical analysis, survival analysis, and prepared all figures and tables with supervision from S.H., M.K.A., and N.J.N. aided in interpretation of mutational patterns. M.Æ.H. and G.Ø.J. catalogued patient samples and prepared them for S.N.P. array analysis. S.H., C.N., L.H.E, E.K.H. gathered patient samples. S.H., N.J.N., P.J., M.T., T.C.E.‐G., P.B., T.S.L., M.K.A. and K.G. collected clinical data and provided clinical interpretation. K.O.L. collected updated patient variables from the national pathology database. All authors participated in interpretation of data, manuscript writing, and review of the manuscript.

Funding

This work was supported by Danish Cancer Society Research Center, R306‐A18107, Holms Mindelegat.

Ethics Statement

This study was approved by all regional ethical committees and the national Danish Data Protection Agency (j.nr. 2012‐58‐0004). The requirement for informed consent was waived by the ethics committees in accordance with Danish regulations for use of anonymized archived samples. Detailed consent procedures have been described previously [7].

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: ajh70392‐sup‐0001‐DataS1.xlsx.

Data S2: ajh70392‐sup‐0002‐Supinfo1.pdf.

Table S1: Next‐Generation sequencing panel used to define CHIP used in this study. The sequencing panel is referenced in [7].

Table S2: Characteristics of lymphoma subtypes in the full national cohort [7], compared to our study cohort.

Table S3: Frequency of mCAs by category.

Table S4: mCAs grouped by size interval.

Table S5: Baseline characteristics of study participants stratified by CHIP. Patients with undetermined mCA type are treated as not having an mCA.

Table S6: Patients with CHIP stratified by ≥ 2 mCAs and CHIP subtype (DNA repair vs. other CHIP), Fisher's exact test for significance.

Table S7: mCA overview of patients with two or more autosomal mCAs.

Table S8: mCA overview in patient with a TP53 and TET2 mutation, harboring five mCAs.

Table S9: Overview of CHIP mutations.

Table S10: Cause of death of patients grouped by CHIP and mCA.

Table S11: tMN status of patients grouped by CHIP and mCA.

Table S12: tMN status of patients grouped by CHIP (DNA repair CHIP, other CHIP or no CHIP) and mCA status.

Figure S1: Frequency (fraction) and type of mCAs in the CHIP and No CHIP groups stratified by chromosome. CN‐LOH, copy number neutral loss of heterozygosity.

Figure S2: Cell fraction and autosomal mCA size. Distribution of autosomal mCA size and mCA clone cell fraction on a log scale, with numerical values for easy interpretation. Some undetermined mCAs had no cell fraction and was removed from the plot.

Figure S3: Additional characteristics of mCA and CHIP co‐occurrence. A Proportion of patients with DNMT3A, TET2, PPM1D, TP53 and ASXL1, split by mCA occurrence. Only showing the five most mutated CHIP genes. For each CHIP gene, p values indicate one‐sided Fisher's exact tests comparing the proportion of patients with autosomal mCAs among carriers of the indicated CHIP mutation with patients without CHIP. B Average number of autosomal mCAs per patient, stratified by maximum VAF of the CHIP mutation found in the patient with the mCA, value shown is average over all patients in the given VAF subgroup. Negative Binomial test was used to calculate p values based on the count of patients in each VAF subgroup. Only p values below 0.2 are shown. Counts of patients in each subgroup is labeled under the point.

Figure S4: Overall survival according to CHIP occurrence and autosomal mCA co‐occurrence, with mCAs of undetermined type removed. Number at risk is placed below the x‐axis. Log‐rank p value is calculated based on the 10‐year overall survival.

Figure S5: Overall survival according to CHIP subtype and autosomal mCA co‐occurrence. Number at risk is placed below the x‐axis. Log‐rank p value is calculated based on the 10‐year overall survival. Significant p values (<= 0.05) are written in bold.

Figure S6: 10‐year cumulative incidence of confirmed or suspected tMN, according to CHIP subtype and autosomal mCA occurrence. Cumulative incidence of observed and suspected tMN, according to CHIP occurrence and type (DNA repair vs. Other) and autosomal mCA co‐occurrence. Number at risk is placed below the x‐axis. Gray's test p value is calculated based on the 10‐year cumulative incidence.

Note S1: Co‐occurrence log enrichment score.

Data Availability Statement

The data supporting the findings of this study are derived from Danish national health registries and are subject to restrictions under the General Data Protection Regulation (GDPR). Therefore, the data are not publicly available. Access may be granted through the relevant Danish authorities upon reasonable request and with the necessary approvals.