Myeloid clonal hematopoiesis of indeterminate potential in patients with chronic lymphocytic leukemia

Chronic lymphocytic leukemia (CLL) is the most common leukemia in adults, and its incidence increases with age. A common age-related phenomenon is clonal hematopoiesis of indeterminate potential (CHIP), characterized by the presence of somatic mutations in hematopoietic stem cells without clear evidence of a hematologic malignancy. Both myeloid CHIP (M-CHIP) and lymphoid CHIP have previously been described, each defined by characteristic somatic mutations that increase the risk of myeloid and lymphoid malignancy, respectively.¹ M-CHIP mutations may also penetrate the lymphoid lineage and impact pathobiology.^2-5 Understanding the intersection of M-CHIP with CLL will help establish its prevalence in this population and its effect on CLL-related outcomes.

Prior work in this area has been limited among both untreated and treated patients with CLL.^6,7 We, thus, sought to describe the prevalence of M-CHIP in patients with CLL, and to describe the association of M-CHIP with CLL-related disease and treatment-related outcomes.

We performed a retrospective cohort study of a prospective CLL tissue bank and database at the Dana-Farber Cancer Institute, on patients who had undergone a targeted 95-gene next-generation sequencing (NGS) panel on whole blood or bone marrow, performed as part of standard of care, which was implemented at the Dana-Farber Cancer Institute in 2015 (supplemental Table 1).⁸ This panel is generally performed at time of initial consultation for new patients, and if treatment is being planned or the patient is being enrolled on a clinical trial. The NGS panel is also generally repeated immediately before starting a new line of treatment. In patients who had >1 NGS panel done, the first panel results were used for this study. All patients provided written informed consent, and research ethics approval for studies involving the prospective tissue bank was granted by the Dana-Farber research ethics board.

We classified patients into 3 cohorts for the analysis. The first cohort was patients who had their first NGS performed before any treatment for CLL (cohort A, untreated). The second cohort was patients who had their first NGS after receiving CLL-directed therapy (cohort B, previously treated). The final cohort combined all patients (cohort C, all patients).

We used published classification criteria for M-CHIP to create a list of 56 candidate genes (supplemental Table 2).¹ Given the overlap between M-CHIP and CLL driver genes, we excluded overlap genes that were more common in CLL than in M-CHIP in order to specifically evaluate the contribution of M-CHIP. To do this, we first established the proportion of patients with CLL who have each driver gene mutation, identified from a molecular map reported from 1148 patients.⁹ We then calculated the proportion of patients with M-CHIP who had each gene mutation from a published study of 3139 patients with M-CHIP.¹ When the genetic mutation was more common among patients with CLL than among patients with M-CHIP, we excluded these genes from potential classification as M-CHIP. This list included SF3B1, TP53, BRAF, KRAS, BCOR, NRAS, CREBBP, and SETD2 (supplemental Table 3; supplemental Figure 1),⁹ leaving 48 candidate genes for M-CHIP classification.

We used the following strategy to adjudicate variants as somatic vs germ line, in the absence of sequencing data from paired normal tissue. If the variant allele frequency (VAF) was <30%, the variant was categorized as somatic. When the VAF was ≥30%, we explored the population allele frequency (PAF) using the Broad Institute Genome Aggregation database.¹⁰ If the PAF was ≥0.1%, the variant was categorized as germ line; if the PAF was <0.1%, the variant was categorized as somatic; if the PAF was unknown, the variant was categorized as unknown. Next, among non–germ line variants, we considered whether the variant could potentially be M-CHIP based on the gene involved. If so, we applied a previously published M-CHIP classification schema¹ to categorize the variant. In scenarios in which the mutation could be M-CHIP based on this schema but the VAF was 40% to 60% (n = 18), the final determination was performed by adjudication with a M-CHIP expert (C.J.G.).

The independent variable in the analyses was M-CHIP, analyzed as a binary (present/absent) and categorical (≥2 vs 1 vs 0 mutations) predictor. In exploratory analyses, we also assessed the impact of VAF of the largest M-CHIP clone as a binary (VAF ≥10% or <10%) and continuous predictor.

Covariates included age at diagnosis; age at time of NGS panel; sex; and CLL-related prognostic factors including immunoglobulin heavy chain variable region mutation status, cytogenetic testing, presence of TP53 mutation, and prior treatment history. If multiple discordant immunoglobulin heavy chain variable region mutation clones were present, patients were characterized as having unmutated disease because this represented the higher risk finding.

For cohort A, we analyzed time to first treatment (TTFT) and overall survival (OS) from time of NGS testing. For cohorts B and C, we analyzed OS from time of NGS testing.

Given the association of M-CHIP with cardiovascular disease (CVD),¹¹ we described the prevalence of CVD (coronary artery disease, heart failure, cerebrovascular disease, and peripheral vascular disease) in patients with M-CHIP compared with those without M-CHIP. Given its association with myeloid malignancies,¹¹ we also described the incidence of myeloid neoplasms including myeloproliferative neoplasms, acute myeloid leukemia, and myelodysplastic syndrome in patients with M-CHIP compared with those without M-CHIP.

Continuous variables are described using median and interquartile range (IQR) or mean and standard deviation, as appropriate. Categorical variables are reported with numbers and frequency. The Kaplan-Meier method was performed to generate survival curves. We performed Cox proportional hazards models to compute unadjusted and adjusted hazard ratios with 95% confidence intervals (CIs) for TTFT and OS. In the multivariable models, we adjusted for covariates that differed between groups. For the exploratory analysis of incidence of myeloid malignancy between groups, we analyzed cumulative incidence with death as a competing event. All analyses were performed using SAS version 9.4.

We included 966 patients who had an NGS test performed between March 2015 and May 2022, including 747 (77%) who had it performed before CLL treatment (cohort A), and 219 (23%) who first had it performed after receiving treatment (cohort B). NGS was primarily performed on peripheral blood (96%; 4% on bone marrow). Median age at time of NGS was 65 years (IQR, 58-72) and 38% of patients were female (Table 1 for patient characteristics). Median follow-up time from diagnosis was 4.4 years (IQR, 2.0-9.1) and from NGS testing was 1.9 years (IQR, 0.5-3.4).

The prevalence of M-CHIP was 12% (95% CI, 10-15; n = 92) in cohort A, and 24% (95% CI, 18-30; n = 52) in cohort B (P < .0001). The difference between groups remained significant even when adjusted for age (adjusted odds ratio, 2.0; 95% CI, 1.4-3.0; P = .0003). The prevalence of M-CHIP increased by age in cohort A but was consistent across age groups in cohort B (Table 1). Patients with M-CHIP in cohort A were older than patients without M-CHIP, whereas patients with M-CHIP in cohort B appeared more likely to have previous chemotherapy exposure. There were no other significant differences in baseline characteristics between groups (Table 1).

Table 2 describes characteristics of M-CHIP among patients. Most patients had only 1 M-CHIP mutation, and the most common mutation was DNMT3A. Patients with prior treatment did not appear to have a significantly larger clone size when analyzed categorically or continuously. Among patients with multiple M-CHIP mutations (n = 28), the most likely co-occurring pair was DNMT3A and TET2 mutations, which co-occurred in 8 patients (29%), followed by ASXL1 and TET1, ASXL1 and EZH1, and DNMT3A and BRCC3, which occurred in 2 patients each (7%).

From time of diagnosis, median TTFT in this cohort was 7.1 years (95% CI, 6.5-8.3). M-CHIP as a binary and categorical predictor was not associated with TTFT, and as a binary predictor was not associated with OS (supplemental Tables 4 and 5). However, M-CHIP as a categorical predictor (≥2 mutations) was significantly associated with OS in both univariable and multivariable analyses controlling for age (supplemental Table 5; Table 3). The 2-year OS from time of NGS testing was 96.2% (95% CI, 94.5-98.1) for patients without any M-CHIP mutations (n = 655) compared with 97.4% (95% CI, 92.5-100) for patients with 1 M-CHIP mutation (n = 77) and 67.5% (95% CI, 33.9-100) for patients with ≥2 M-CHIP mutations (n = 15). In exploratory analyses, VAF as a categorical or continuous predictor was not associated with TTFT or OS (supplemental Tables 4 and 5).

M-CHIP was not associated with OS as a binary or categorical predictor (Table 3; supplemental Table 5). In exploratory analysis, VAF as a categorical (<10% vs ≥10%) and continuous variable was associated with OS (supplemental Table 5).

M-CHIP as a binary predictor was associated with OS in univariable analysis, but this association did not retain significance in multivariable analysis controlling for age and history of prior treatment (supplemental Table 5). However, M-CHIP as a categorical predictor was associated with OS in multivariable analysis controlling for these factors (Table 2; Figure 1); 2-year OS from time of NGS was 92.9% (95% CI, 90.7-95.0) in patients without M-CHIP vs 89.4% (95% CI, 82.4-96.4) with 1 M-CHIP mutation vs 64.7% (95% CI, 42.0-87.5) in patients with ≥2 M-CHIP mutations. In exploratory analysis, VAF as a categorical and continuous variable was also associated with OS (supplemental Table 5).

The cause of death in patients with ≥2 M-CHIP mutations was primarily CLL (n = 6, 75% of deaths), with other deaths from infections or other causes (n = 2, 25% of deaths). No deaths were due to myeloid malignancies.

The prevalence of CVD in patients with M-CHIP (median age, 68 years) was 17.4% compared with 11.7% in patients without M-CHIP (median age, 64 years). Adjusting for age, there was no difference between groups (adjusted odds ratio, 0.85; 95% CI, 0.51-1.41).

Myeloid neoplasms were diagnosed in 1.4% (n = 2) of patients with M-CHIP and 0.4% (n = 3) of patients without M-CHIP. Accounting for follow-up time and considering death as a competing risk, there was no difference in the cumulative incidence between groups (P = .16).

The term CHIP was coined in 2015 to describe somatic mutation(s) in hematopoietic stem cells in the absence of a recognized clonal entity.¹² Those with CHIP have an increased risk of development of a hematologic malignancy. Recently, a scoring system called the clonal hematopoiesis risk score has helped identify those at low, intermediate, and high risk for progression to a hematologic neoplasm.¹³ Although patients with CHIP may be at an increased risk for developing a hematologic neoplasm, CVD is one of the leading causes of mortality in these patients.^12,14 Additionally, other studies have demonstrated that those with CHIP are at higher risk for other medical comorbidities such as autoimmune dysfunction, osteoporosis, gout, and lung disease.^15,16 

As people age, CHIP becomes more prevalent, with ∼10% of people aged >60 years harboring these mutations.^17,18 However, CHIP can also be found in patients who have had prior cytotoxic chemotherapy or radiation. In these situations it is believed that the prior therapy may have allowed for selective expansion of these clones.^19-21 In this study, cohort A demonstrated that the prevalence of M-CHIP increased with age, with rates appearing to match known population prevalence of CHIP mutations. However, in cohort B comprising patients who had received prior therapy, age did not affect the prevalence of CHIP, supporting the theory of a selective expansion of M-CHIP clones with prior cytotoxic therapy. Additionally, in the entire cohort, the incidence of M-CHIP was found to be higher in those who received therapy compared with those who did not (24% vs 12%).

Our study builds upon prior work. In the untreated cohort (cohort A), the most commonly found mutations were DNMT3A and TET2, and the majority of patients had only 1 mutation, matching prior population-based literature.²² The prevalence of M-CHIP in cohort A was also similar to a prior study of 285 untreated patients with CLL with a median age of 71 years, among whom the prevalence of M-CHIP was 12%.⁷ Notably, in that study, the incidence of M-CHIP did not appear to increase significantly with age although caution was noted given the small number of patients aged <50 years or ≥80 years. In cohort B, 25% of patients with M-CHIP had ≥2 M-CHIP mutations, which was similar to the prevalence of 31% noted in a study of patients with lymphoma undergoing autologous stem cell transplant (ASCT).²³ The prevalence of M-CHIP in these pretreated patients was also similar to a prior study of patients with CLL treated with venetoclax with a median 3 prior lines of therapy (81% with prior fludarabine exposure), which found that 10 of 92 patients (11%) had a therapy-related myeloid neoplasm, whereas 27 (29%) had clonal hematopoiesis with a myeloid disorder associated gene.⁶ In a subgroup analysis of 41 patients with low CLL burden (<5% by flow cytometry) or undetectable minimal residual disease, this study did classify overlap mutations with CLL such as TP53 and SF3B1 as myeloid in origin, given that they were assessing a non-CLL compartment, although all patients with such mutations also had other mutations more definitively myeloid in origin (eg, ASXL1 and TET2). In our study, of 165 patients with TP53 mutations (which our schema classified as CLL related), only 7 (4%) had ASXL1 mutations and 2 (1%) had TET2 mutations. Similarly, of 135 patients with pathogenic SF3B1 mutations (which our schema classified as CLL related), only 5 (5%) had ASXL1 mutations, 8 (6%) had TET2 mutations, and 1 patient (1%) had both, supporting our decision to classify TP53 and SF3B1 as CLL-related mutations in our cohort, because we could not be certain which compartment was being analyzed. Future work will involve single-cell sequencing to determine the lineage of origin of mutations found by NGS in order to more accurately classify M-CHIP– vs CLL-related mutations, with an important aspect of this technique being purification of samples to ensure testing is conducted on myeloid cells.

OS significantly differed comparing those with no M-CHIP with those with ≥2 mutations, or higher VAF, although this finding should be confirmed in a larger cohort and/or with longer follow-up time, given that there were only 8 deaths among patients with ≥2 M-CHIP mutations in this study. Notably, in a prior study on CHIP in patients with lymphomas undergoing ASCT, patients with CHIP had inferior OS, although the cohort was enriched with patients who had PPM1D mutations, which was associated with the lowest OS.²³ A prior population-based study on CHIP in patients with lymphoma undergoing ASCT assessed 565 patients and found that CHIP in general was not associated with survival in multivariable analysis, but patients with mutations in DNA repair genes (PPM1D, TP53, RAD21, and BRCC3) had inferior OS.²⁴ In our cohort, only 4 patients had PPM1D mutations, and only 18 patients had mutations in DNA repair genes (PPM1D, RAD21, and BRCC3). Thus, the effect of DNA repair pathway mutations may have been undercaptured in our cohort. In addition, in prior work, patients with higher VAF percentage or those who had a VAF clone of ≥10% had a higher risk of hematologic cancer and/or coronary artery disease.²² 

In exploratory analyses, we found that patients with M-CHIP did not have increased odds of prevalent CVD compared with patients without M-CHIP. This may be because of the fact that patients with CLL have a higher risk of CVD compared with the general population,²⁵ which may mask the effect of M-CHIP in this cohort. We note that the overall prevalence of CVD in our cohort in both the M-CHIP and non–M-CHIP groups appears to be higher than rates reported for similar age groups by the US Centers for Disease Control and Prevention.²⁶ The lack of a significant difference between groups could also relate to the fact that we do not report incident heart disease but rather prevalent disease. We also found no significant difference in risk of myeloid malignancy between groups, likely because of limited follow-up time after NGS testing of 1.9 years.

A strength of our study was the large sample size, as well as the fact that all patients since 2015 seen in consultation at our center receive upfront NGS testing. This allowed capture of the prevalence of M-CHIP in untreated patients who did not necessarily have cytopenias or treatment indications. This is supported by the fact that the median TTFT from time of NGS testing in cohort A was 3.4 years, and the time from diagnosis to NGS testing was similar in patients with M-CHIP and those without M-CHIP.

There are significant limitations to this study. We had limited by follow-up time, which may have limited power in detecting differences in outcomes between groups. Because of limited events, we did not assess associations such as type of M-CHIP mutation with outcome. From a M-CHIP perspective, because of overlap mutations between CLL and M-CHIP, certain somatic mutations were excluded such as TP53 and SF3B1, because we were unable to distinguish myeloid from lymphoid compartments. By excluding these common M-CHIP mutations,¹ we may have underestimated the true prevalence and effect of M-CHIP. However, this approach matches prior work of M-CHIP in CLL in which genes more prevalent in CLL were excluded from classification as M-CHIP (TP53, NOTCH1, SF3B1, MYD88, FBXW7, and BRAF).⁷ 

In conclusion, the prevalence of M-CHIP in a large cohort of untreated (12%) and treated (24%) patients with CLL was similar to that reported in prior literature. M-CHIP prevalence appeared to increase with age in untreated patients but appeared consistent across age groups in treated patients, suggesting that treatment (85% had prior chemotherapy) may have affected M-CHIP emergence or growth even in younger patients. The presence of ≥2 M-CHIP mutations was associated with OS, even accounting for prior treatment and age, but these findings were driven by small numbers of patients and should be viewed as hypothesis generating only. These findings support continued work into characterizing the effects of M-CHIP in patients with CLL.

Contribution: A.V. was responsible for study conceptualization, data curation, methodology, formal analysis, writing the original draft, and review and editing of the manuscript; V.O.V. was responsible for writing the original draft; A.S. was responsible for study conceptualization, and review and editing of manuscript; R.S.A, M.M., C.J.G., S.P.M., and R.F. were responsible for data curation, and review and editing of manuscript; S.T. and Y.R. were responsible for formal analysis, and review and editing of manuscript; S.M.F. was responsible for data curation, and review and editing of manuscript; B.A.K. was responsible for study conceptualization, and review and editing of manuscript; C.K.H. was responsible for review and editing of manuscript; G.G. was responsible for study conceptualization, and review and editing of manuscript; C.J.W. and M.S.D. were responsible for review and editing of the manuscript; and J.R.B. was responsible for study conceptualization, data curation, supervision, and for review and editing of the manuscript.

Conflict-of-interest disclosure: A.S. holds stock in Vertex Pharmaceuticals. B.A.K., C.K.H., G.G., C.J.W. are inventors on a patent “Compositions, panels, and methods for characterizing chronic lymphocytic leukemia” (PCT/US21/45144). G.G. reports research funding from International Business Machines Corporation (IBM), Pharmacyclics/AbbVie, Bayer, Genentech, and Ultima Genomics; is an inventor on patent applications related to MSMuTect, MSMutSig, MSIDetect, POLYmorphic loci reSOLVER (POLYSOLVER), SignatureAnalyzer-GPU, MinimuMM-seq, and DLBclass; is founder, consultant, and holds privately held equity in Scorpion Therapeutics; and is founder and holds privately held equity in PreDICTA Biosciences. C.J.W. reports research funding from Pharmacyclics; holds equity in Biontech; and serves on scientific advisory board for Adventris, Repertoire, and Aethon Therapeutics. M.S.D. reports institutional research funding from AbbVie, AstraZeneca, Ascentage Pharma, Genentech, MEI Pharma, Novartis, Surface Oncology, and TG Therapeutics; and reports personal consulting income from AbbVie, Adaptive Biosciences, Ascentage Pharma, AstraZeneca, BeiGene, Bristol Myers Squibb, Eli Lilly, Genentech, Genmab, Janssen, Merck, Mingsight Pharmaceuticals, Nuvalent, Secura Bio, TG Therapeutics, and Takeda. J.R.B. is a consultant for AbbVie, Acerta/AstraZeneca, Alloplex Biotherapeutics, BeiGene, Bristol Myers Squibb, Galapagos NV, Genentech/Roche, Grifols Worldwide Operations, InnoCare Pharma Inc, iOnctura, Kite Pharma, Loxo/Lilly, Merck, Numab Therapeutics, Pfizer, and Pharmacyclics; reports research funding from BeiGene, Gilead, iOnctura, Loxo/Lilly, MEI Pharma, and TG Therapeutics; and serves on the data safety monitoring board for Grifols Therapeutics. The remaining authors declare no competing financial interests.

Correspondence: Jennifer R. Brown, Department of Medical Oncology, Dana-Farber Cancer Institute, Mayer 226 44 Binney St, Boston, MA 02115; email: jennifer_brown@dfci.harvard.edu.