Incipient clonal hematopoiesis is accelerated following CD30.CAR-T therapy

Abstract

Chimeric antigen receptor (CAR) T-cells are an emerging therapy for refractory lymphomas. Clonal hematopoiesis (CH), the preferential outgrowth of mutated bone marrow progenitors, is enriched in lymphoma patients receiving CAR-T cells. CAR-T therapy requires conditioning chemotherapy and often induces systemic inflammatory reactions, both of which have been shown to promote expansion of CH clones. Thus, we hypothesized that pre-existing CH clones could expand during CAR-T cell treatment. We measured CH at 154 timepoints longitudinally sampled from 26 patients receiving CD30.CAR-T therapy for CD30+ lymphomas on an investigational protocol (NCT02917083). Pre-treatment CH was present in 54% of individuals and did not correlate with survival outcomes or inflammatory toxicities. Longitudinal tracking of single clones in individual patients revealed distinct clone growth dynamics. Initially small clones, defined as VAF <1%, expanded following CAR-T administration, compared with relatively muted expansions of larger clones (3.37-fold vs. 1.20-fold, p=0.0014). Matched clones were present at low magnitude in the infused CD30.CAR-T product for all CH cases but did not affect the product’s immunophenotype or transduction efficiency. As cellular immunotherapies expand to become frontline treatments for hematological malignancies, our data indicates CAR-T recipients could be enriched for CH, and further longitudinal studies centered on CH complications in this population are warranted.

Introduction

Clonal hematopoiesis (CH) is a ubiquitous phenomenon of human aging defined by peripheral blood cells that are disproportionately derived from a single hematopoietic progenitor,¹ cells we refer to henceforth as “CH clones.” Large CH clones are an age-independent risk factor for the development of cardiovascular disease, thrombotic events, and hematologic malignancy, as well as all-cause mortality.² CH is enriched among recipients of cytotoxic chemotherapy. Indeed 25% of lymphoma survivors regardless of age display CH,³ and a subset of these patients have shortened survival and more frequent treatment complications.

Chimeric antigen receptor (CAR) T-cell therapy has demonstrated remarkable efficacy in curing heavily pretreated lymphoma patients.⁴ Consequently, the number of individuals entering survivorship clinics continues to grow rapidly. Long-term survivors, however, display higher rates of secondary leukemias and cardiovascular and thrombotic events;^5–8 these sequelae have been shown to be accelerated by large CH clones.² CAR-T therapy may directly influence the growth rates of CH clones because it requires pre-infusion conditioning chemotherapy and often induces systemic inflammation, both of which have been shown in animal models to promote CH.^9,10 Therefore, we hypothesized that CAR-T recipients with CH at baseline would display expanded CH clones following CAR-T administration.

In this report, we leverage longitudinal post-CAR T infusion sampling to understand the frequency and clonal dynamics of CH in a cohort of heavily pretreated lymphoma patients receiving anti-CD30.CAR T-cell therapy on an investigational protocol at our institution (RELY-30, NCT02917083).¹¹

Methods

All patients had CD30-expressing lymphomas (69% Hodgkin lymphoma, 31% non-Hodgkin lymphoma), were heavily pretreated [median of 5 therapies (range 2–12)] and had progressive disease after their last line of therapy. 154 blood samples were available across serial sampling from 26 patients. All patients had pre-infusion blood specimens available for genomic analysis. Post-infusion blood samples ranged from 1 to 48 months after CD30.CAR-T administration. Per patient, 1 to 5 blood specimens were collected post-infusion. In addition, pre-manufacturing autologous cells and the engineered CAR-T product were available for 24 of 26 subjects. Error-corrected sequencing targeting a panel of 215 candidate CH driver genes was performed on genomic DNA purified from blood samples (supplementary Methods, supplementary File 1). Targeted genes include candidate drivers described in population CH studies and included in the WHO definition of CH, additional genes known to be recurrently mutated in myeloid and lymphoid neoplasms, and genes whose impairment confers increased hematopoietic stem cell fitness in laboratory studies.^12–22 Statistical approaches are described in the supplementary Methods.

Results & Discussion

Among 26 patients, we detected CH in 14 patients (54%) at baseline. After adjusting for detection sensitivity differences, our incidence was comparable to recent cohorts of lymphoma patients receiving CD19.CAR-T (37%)²³ and mixed lymphoma and multiple myeloma patients receiving CAR-T therapy (48%),¹⁵ but greater than that observed in patients with solid cancers^17,24 and those receiving autologous hematopoietic stem cell transplantation.^3,25

Subjects with CH were older than those in the non-CH group, reflecting the known impact of age on CH incidence (median age at diagnosis for CH vs. no CH: 48 vs. 25 years, p=0.009). Individuals with or without CH had received a similar number of prior therapies (median 5, range 2–12). NHL and HL cases were comparably represented in both groups. Within the constraints of our sample size, we observed no statistically significant differences in CH incidence when comparing diagnoses, stage, sex, prior transplant status, or number of prior lines of therapy (Figure 1A).

Figure 1. Clonal hematopoiesis in subjects treated with CD30.CAR-T cells.

A) Associations between CH, patient characteristics, and CAR-T toxicities. Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for categorical variables.

B) Overall survival of patients treated with CD30.CAR-T stratified by absence or presence of clonal hematopoiesis. Survival was compared using a log-rank test.

C) Progression-free survival of patients treated with CD30.CAR-T stratified by absence or presence of clonal hematopoiesis. Survival was compared using a log-rank test.

D) Number of clones detected per patient, with pre-infusion clone size indicated by color.

E) Number of mutations per gene among CH in cohort, stratified by clone size at pre-infusion timepoint.

Abbreviations: Auto-SCT = autologous stem cell transplant; allo-SCT = allogenic stem cell transplant; CAR = chimeric antigen receptor; CH = clonal hematopoiesis; CRS = cytokine release syndrome; HL = Hodgkin lymphoma; ICANS = immune effector cell-associated neurotoxicity syndrome.

Non-significant trends towards improved disease-free survival were observed among CH carriers compared with non-CH subjects (3-year DFS of 36% vs. 0%, respectively; p=0.092) but not overall survival (3-year OS 63% vs. 79%, respectively; p=0.209) (Figure 1B–C), aligning with previous cross-sectional studies evaluating CAR-T outcomes in CH carries.^15,23 Previous reports have demonstrated that CH clones are modulated by and influence systemic inflammation, such as observed during cytokine release syndrome (CRS).^15,23 However, we observed no significant difference in CAR-T-related immune-toxicities amongst groups in our study (Figure 1A).

We next examined the degree baseline CH status influenced hematopoietic toxicity or recovery post-CAR T cells as well as influence on CAR-T expansion and persistence. CH mutations are often associated with a myeloid bias during hematopoietic differentiation²⁰. However, preexisting CH status did not affect the recovery of either neutrophils or lymphocytes following CAR-T administration (supplementary Figure 1A–C), possibly because of the heterogeneity in CH mutations and their impact on differentiation, as well as baseline clone sizes (see below) observed in this study.

When adjusted for dose level, we observed a higher level of CAR-T cell persistence among CH carriers at the highest dose level, but no difference at other dose levels. Notably, we did not observe any differences in peak expansion levels of CAR-T cells between CH and non-CH patients (supplementary Figure 1D–E). We note that additional factors, beyond just the dose level infused, might contribute to any differences in CAR-T persistence between CH and non-CH subjects, but performance of a robust regression analysis was limited by sample size.

Multiple CH clones were detected in most patients (median 2, range 1–6) (Figure 1D). DNMT3A was the most frequently mutated gene, followed by PPM1D and CHEK2 (Figure 1E). DNMT3A clones have been associated with cytopenia in HSCT recipients and can cause a differentiation bias towards myeloid cells, but our leukocyte recovery data indicates that in this cohort, myeloid recovery post-CAR T cells was not influenced by DNMT3A mutation status (or any other mutation). The high prevalence of mutations in the DNA damage repair genes PPM1D and CHEK2 is likely due to clonal selection during prior chemotherapy and recapitulates the pattern of CH mutations in solid tumor patients.^17,24

In addition to pre-infusion CH characterization, and distinct from prior cross-sectional CH studies, we had access to peripheral blood at several timepoints post-CD30.CAR T infusion for every patient. This allowed us to analyze and visualize clonal dynamics after CAR-T treatment for every patient (Figure 2A–C, supplementary Figure 2).

Figure 2. Longitudinal dynamics of CH during CAR-T therapy.

A) Oncoplot of all detected CH clones at pre-infusion timepoint and at latest available timepoint within one year. Pre- and post-infusion clone sizes are plotted adjacent. Each vertical line represents clones observed in a single subject; mutated genes are shown by row. Clone size is indicated by dot size and mutation consequence is indicated by color.

B) Change in clone size, normalized to and stratified by pre-infusion clone size. Linked data points indicate matched clones between timepoints.

C) Clone size fold-changes from 12-month timepoint in B, stratified by pre-infusion size. Comparison using t-test.

D) Change in absolute clone size between pre-infusion blood, procured mononuclear cells for manufacture, and the CAR-T product, stratified by initial clone size. Linked data points indicate matched clones between timepoints.

E) Clone size fold-change of greatest cell line VAF from D over pre-infusion size, stratified by pre-infusion size. Comparison using t-test.

An equivalent number of post-infusion samples were queried between CH and non-CH groups (median 3, range 2–9) (Figure 1A). In the CH group, after CD30.CAR-T infusion we observed a median 1.40-fold increase in clone size within six months (supplementary Figure 2A). Elevated clone sizes were sustained at 2.62-fold over pre-infusion levels at least three years post-infusion. However, when considered in aggregate, individual clone dynamics were heterogenous and spanned three orders of magnitude (range 0.56- to 24.3-fold) (supplementary Figure 2A).

Our serial sampling and consensus sequencing approach allowed for the detection of small clones beyond the limits of cross-sectional sampling and standard sequencing approaches. Thus, to observe CH clonal dynamics with greater granularity, we stratified observed clones based on their pre-infusion clone size. Clones with pre-infusion VAF ≤1% were considered low-magnitude (small); clones with pre-infusion VAF >1% were considered high-magnitude (large). Our cutoff of 1% VAF was selected because it is the limit of detection for previous studies using non-error corrected approaches, thus representing the threshold under which CH likely would not have been detected.

Small clones at baseline expanded dramatically (3.37-fold) following CAR-T administration, compared with muted expansions in large clones (1.2-fold) (p=0.0014) (Figure 2B, supplementary Figure 2B). The post-infusion increase in CH magnitude was sustained at three years post-infusion (Figure 2B) and was similarly observed when clone size was stratified by a 2% cutoff (supplementary Figure 2C). In initially small clones, average absolute clone size increased from 0.4% VAF to 1.7% VAF within a year post-infusion, and to 2.2% VAF by three years post-infusion (supplementary Figure 2B). Although all patients had short-term increases in clone sizes, clonal acceleration was only sustained in patients that developed CRS (supplementary Figure 2D). No specific mutation was associated with CRS incidence. We did not have matched lymphoma tissue available for mutational analysis. Conceivably, inadvertent detection of circulating tumor DNA could occur with our assay. However, we did not observe correlation of CH magnitude increases with disease relapse in any patient, and we extracted DNA from plasma-reduced PBMC samples only to minimize the risk of capturing cell-free DNA.

To evaluate the prevalence of CH clones in the infused T-cell product, we queried both non-engineered autologous cells and the corresponding CD30.CAR-transduced T-cell products. Our previous report from this investigational study demonstrated high CAR transduction rates of >90% and products almost exclusively composed of CD3+ T lymphocytes.¹¹ Although all subjects with CH displayed at least one matched clone in their T-cell product (supplementary Figure 3), we observed no difference in the immunophenotype of the final T-cell product between individuals with or without CH (supplementary Figure 4). Detectable matched clones were smaller in magnitude in the T-cell product compared with pre-infusion peripheral blood samples. Indeed, large peripheral blood clones (VAF >1%) demonstrated the greatest reduction in clone size (0.46-fold) in comparison to small clones (1.40-fold; p=0.03) (Figure 2D–E) post-CAR-T manufacturing. Taken together, these data indicate that the clonal acceleration observed following CAR-T administration is not due to infused cells with CH mutations. These data also corroborate the lineage specificity of CH mutations, indicating that T lymphocytes are less likely to harbor any CH mutations, compared with myeloid or B cells.²⁶

In summary, we observed that CH clones are enriched after CAR-T administration, likely because of chemotherapy exposure and systemic inflammation that occur during treatment. Small clones below the VAF limit of detection for standard CH detection assays display a sustained increase in clone size, indicating an influence of the CAR-T process to accelerate their expansion but not induce new CH clones. Indeed, none of the 26 patients developed de novo CH post-CAR T cells. From our data, we cannot uncouple the contribution of conditioning chemotherapy versus the CD30.CAR-T cells themselves to the observed clonal expansions. Less aggressive conditioning regimens may ameliorate clonal expansions observed following cell therapies. Nonetheless, as cellular immunotherapies become more efficacious and expand to become frontline treatments for hematologic malignancies and other cancers, these data suggest that CAR-T recipients could be enriched for CH and its associated complications.

Data Sharing Statement:

Error-corrected sequencing data has been deposited in NCBI BioProject #PRJNA983927. Further information and requests will be fulfilled by Chiraag Kapadia (chiraag.kapadia@bcm.edu).