Clonal hematopoiesis in the setting of hematopoietic cell transplantation

Abstract

Clonal hematopoiesis (CH) in autologous transplant recipients and allogeneic transplant donors has genetic features and clinical associations that are distinct from each other and from non-cancer populations. CH in the setting of autologous transplant is enriched for mutations in DNA damage response pathway genes and is associated with adverse outcomes, including an increased risk of therapy-related myeloid neoplasm and inferior overall survival. Studies of CH in allogeneic transplant donors have yielded conflicting results but have generally shown evidence of potentiated alloimmunity in recipients, with some studies showing an association with favorable recipient outcomes.

Introduction

Compared with non-cancer populations, clonal hematopoiesis (CH) has distinct features and clinical impacts in the setting of both autologous and allogeneic transplantation for hematologic malignancies. Moreover, the mutation patterns and patient outcomes associated with CH in autologous stem cell transplantation (ASCT) are distinct from those observed in allogeneic hematopoietic cell transplantation (alloHCT). This review summarizes the available evidence concerning the biology and clinical associations of CH in the context of transplantation.

The first studies characterizing CH in non-cancer populations analyzed whole exome sequencing performed on peripheral blood samples,^1–3 but over 90% of somatic mutations found in these studies were confined to fewer than 20 genes. Therefore, subsequent studies of CH in specialized contexts have largely used targeted sequencing approaches. In studies of CH in both ASCT and alloHCT, researchers have primarily performed targeted DNA sequencing on aliquots of G-CSF mobilized apheresis products or bone marrow aspirates, which are enriched for hematopoietic stem and progenitor cells.

CH in ASCT

ASCT is a common treatment strategy for lymphoid malignancies. For lymphomas, it is typically used as a second-line therapy for relapsed or refractory disease,⁴ whereas it is commonly employed as part of intensive first-line therapy for high-risk multiple myeloma.⁵ The purpose of ASCT is to permit the delivery of high-dose cytotoxic therapy that would otherwise be myeloablative. Autologous hematopoietic cells are collected and preserved prior to the administration of this therapy, then infused afterward. The autologous nature of the transplant means that the infused product has no direct immunologic role in treating the cancer itself.

ASCT can be curative for some forms of aggressive lymphoma,⁶ and it significantly prolongs relapse free survival in myeloma⁵ and some indolent lymphomas. It is also associated with a high rate of therapy-related myeloid neoplasm (TMN) relative to other cytotoxic therapies.^7,8 Following the identification of CH in cross-sectional studies, multiple research groups assessed the frequency of CH in ASCT cohorts and its link to TMN.

CH characteristics in the setting of ASCT

The majority of CH mutations in ASCT patients are concentrated in four genes that can be clustered into two groups.^9–18 DNMT3A and TET2 encode epigenetic modifiers with subtle, pleiotropic effects on global gene expression.^19–22 Mutations in these genes are also common in non-cancer CH cohorts^1–3 and lead to diminished or abolished function of the encoded proteins.^23–25 This in turn confers a replicative advantage to mutated hematopoietic stem cells (HSCs) relative to unmutated counterparts, which over time drives the outgrowth of the mutant clone.

PPM1D and TP53 mutations, on the other hand, lead to decreased activity of the DNA damage response pathway (DDR) and are significantly more frequent in ASCT recipients than in non-cancer cohorts (Figure 1). Mutations in TP53 lead to decreased activity of p53,²⁶ which integrates the cellular response to DNA damage and activates apoptotic gene expression programs.²⁷ Mutations in PPM1D are all truncating events clustered in the terminal exon²⁸ and drive increased expression of the PPM1D (also called WIP1) phosphatase via deletion of a C-terminal degron motif.²⁹ Increased expression augments/enforces the normal physiologic consequences of PPM1D phosphatase activity, leading to attenuation of the DDR.³⁰ Mutations in both genes confer a distinct selective advantage to mutant HSCs in the setting of cytotoxic therapy,³¹ to which ASCT recipients are uniformly exposed.

Figure 1.

Cohort characteristics and mutation frequency in selected CH transplant studies. To normalize comparison, only CHIP (clonal hematopoiesis of indeterminate potential, i.e., CH with variant allele fraction ≥ 0.02) is shown, and only studies for which frequency of CHIP was reported or inferable are included. Frequencies in the heat map reflect the fraction of patients with CHIP who have mutations in each gene, ranging from 0.6 (dark red) to 0 (dark blue). HM, hematologic malignancies.

In ASCT cohorts, the frequency of PPM1D mutations is higher in lymphoma patients than in myeloma patients (Figure 1). This is attributable to differences in disease-specific treatment regimens received before ASCT. Standard lymphoma regimens contain multiple conventional DNA-damaging agents that activate the DDR,³² whereas standard myeloma regimens³³ are built around drug classes like proteasome inhibitors and immunomodulatory agents that act by different mechanisms and may not select for cells with genetic inactivation of the DDR pathway.

Somatic mutations in other components of the DDR are less common than those in PPM1D and TP53 but are still enriched in ASCT cohorts relative to non-cancer populations. Pathogenic missense or truncating mutations in ATM, which encodes a kinase that controls cell cycle pausing after DNA damage, cause Ataxia-Telangiectasia syndrome when mutated in the germline but can also be selected by cytotoxic therapy. Similarly, somatic inactivating mutations in CHEK2, which encodes a kinase downstream from ATM that pauses the cell cycle and stabilizes p53, occur in a small percentage of chemotherapy- or radiation-exposed individuals.³⁴ Truncating mutations in SRCAP, which codes for a chromatin-remodeling protein that also has a role in mediating homologous recombination in response to DNA damage,³⁵ are also selectively enriched in therapy-exposed cohorts relative to unselected population cohorts.³⁶

As in the non-cancer population, the prevalence of CH at the time of ASCT is strongly associated with advancing age, but not with other demographic variables or exposure histories. Studies in model systems have shown distinct advantages for PPM1D and TP53-mutant HSCs in the setting of specific chemotherapeutic agents, such as platinum compounds.³¹ These have been difficult to validate in human cohorts due to the complex treatment histories of patients receiving ASCT.

Post-ASCT outcomes associated with CH

The presence of CH at the time of ASCT has primarily been associated with adverse outcomes (Table 1). First, patients with CH have poorer age-adjusted stem cell mobilization than those without CH and are more likely to require multiple days to collect an adequate number of CD34+ cells.^9,12 There is, however, no apparent difference in hematologic function after transplantation between patients with and without pre-ASCT CH, including time to engraftment of any lineage, blood counts, or erythrocyte indices.

Table 1.

Studies of CH in the setting of autologous transplantation.

	Disease	Design	Findings
			TMN	OS
Gibson9	Lymphoma	Retrospective cohort	Increased	Decreased
Husby10	Lymphoma	Retrospective cohort	Increased *	Decreased *
Lackraj11	Lymphoma	Retrospective cohort	No effect	Decreased
Soerensen12	Lymphoma	Case-control	Increased	-
Slavin13	Lymphoma	Case-control	-	-
Gramegna14	Lymphoma	Cross-sectional	-	-
Ortmann15	Lymphoma	Cross-sectional	-	-
Mouhieddine16	Myeloma	Retrospective cohort	No effect	Decreased
Stelmach17	Myeloma	Retrospective cohort	-	No effect
Chitre18	Myeloma	Cross-sectional	-	-

The red cell background indicates adverse associations.

^*:Husby et al found an adverse OS association with CH defined by mutations in the DNA damage response pathway, but not CH overall.

Patients have been reported to have a 5–10% chance of developing TMN within 5 years of ASCT.^7,8 In lymphoma patients, this risk is 3.5–6.5 times higher for those with CH than those without CH.^9,10,12 In myeloma patients, conversely, there is no difference in TMN risk for those with and without CH.¹⁶ This difference may be explained by the observation that the risk of TMN is higher for DDR pathway mutations than other CH mutations.^9,10,12 Moreover, the majority of TMN risk attributable to DDR pathway mutations appears to be driven by TP53 mutations, rather than PPM1D or other DDR pathway mutations. Analysis of available TMN samples frequently shows the expansion of TP53 mutations that were present at the time of ASCT, whereas PPM1D, ATM, and SRCAP mutations are often either not expanded or no longer detectable.^9,36 This suggests that TP53 mutant HSC clones have actual leukemogenic capacity, whereas PPM1D-mutant clones may not directly drive TMN evolution. In some cases, therefore, CH present at the time of ASCT is clearly the initiator of leukemic transformation, while in other cases it may primarily be a marker of elevated TMN risk.

The adverse outcomes associated with CH in ASCT extend beyond TMN alone. Patients with CH have inferior overall survival after ASCT compared with non-CH patients, which is independent of age and consistent across multiple studies.^9–11 This effect is driven entirely by increased non-relapse mortality for CH patients, as CH status has no impact on the risk of lymphoma or myeloma relapse. Most individual studies have been too small to identify differences in cause-specific mortality other than TMN, but it appears that patients with CH are globally at increased risk of adverse events such as infection, sepsis, and organ failure.^9–11

The risk of death after ASCT has been reported to be higher for CH with DDR mutations than other mutations, particularly in lymphoma patients.¹⁰ As opposed to TMN risk, CH with PPM1D mutations is associated with reduced OS independent of co-occurring TP53 mutation.⁹ It is possible that this reflects a direct cellular effect of mutant PPM1D, which also regulates the NFKB and MAP kinase pathways, in terminally differentiated leukocytes.^37–39 Alternatively, PPM1D-CH may reflect reduced patient fitness for transplant due either to intrinsic constitutional characteristics or to global tissue injury from prior cytotoxic therapy. Consistent with this hypothesis, PPM1D mutations have been associated with reduced relative telomere length in hematopoietic cells, possibly reflecting diminished capacity for regeneration after tissue damage.^40,41

Clinical Implications

Despite consistent results across multiple studies, the recognition of CH as a risk factor for adverse outcomes after ASCT has not resulted in substantial practice changes for lymphoma or myeloma patients. All studies of CH in ASCT published to date sequenced only a single sample, either peripheral blood drawn prior to transplantation or an aliquot of the mobilized autologous stem cell product. Accordingly, while it is clear from these results that the presence of CH predates the transplant itself, it is not known when CH arose in relation to the patient’s original diagnosis and receipt of pre-transplant therapy, nor is it known whether the transplant accelerates the expansion of the aberrant clone.

These limitations leave two key questions unanswered that likely contribute to the lack of clinical impact. First, since all ASCT CH studies have been retrospective single-arm cohorts in which all patients received ASCT, it is not known whether the risk of adverse outcomes is already “locked in” at the time of transplant due to the presence of CH or is compounded by the transplant itself. Second, as above, it is not known whether CH at the time of ASCT is a modifiable risk factor or is a biomarker of poor patient fitness, such as reduced HSC reserve or shortened telomere length. Consequently, it is not known whether these risks could be mitigated by choosing an alternative treatment, such as immune therapies, cellular therapies, or alloHCT.

There is currently insufficient evidence to recommend deferring ASCT in most patients with CH. The alternative therapies listed above are not typically viewed as direct substitutes for ASCT,⁴² and transitioning from ASCT to any of these therapies is not generally recognized as standard of care. Given this lack of broad actionability, there is no basis to recommend routine screening for CH in ASCT candidates, with the possible exception of CH with TP53 mutations due to the excessive risk of TMN. In selected fit patients with TP53-CH, it would be prudent to consider alloHCT. This approach relies on the assumption that alloHCT can effectively eradicate TP53-mutant CH but this hypothesis has not been prospectively tested.

Donor CH in alloHCT

AlloHCT is commonly employed for the treatment of high-risk myeloid malignancies, such as acute myeloid leukemia, myelodysplastic syndrome, and myelofibrosis, as well as treatment-refractory lymphoid malignancies and non-malignant hematologic conditions such as bone marrow failure syndromes or hemoglobinopathies. Successful alloHCT requires (1) engraftment of donor hematopoietic stem cells in the recipient, (2) reconstitution of a functional donor-derived immune system, and (3) favorable balance of on-target alloimmune clearance of malignant cells (graft-versus-leukemia effect, GVL) with off-target toxicity to normal cells (graft-versus-host disease, GVHD).

Donor cell leukemia (DCL) is a rare complication of alloHCT in which a new myeloid malignancy arises post-transplant in cells that can be definitively shown to be of donor origin.^43,44 Donor-engrafted CH was first described in the context of DCL where some mutations present in DCLs could be found in donor samples banked at the time of stem cell collection.⁴⁵ The recognition that CH frequently occurs in the non-cancer population without progressing to hematologic malignancy led to the hypothesis that CH in alloHCT donors might be transferred to recipients with much greater frequency than the rarity of DCL would suggest. This was first shown in alloHCT patients who had persistent cytopenias post-transplant without DCL,⁴⁶ but multiple studies have since examined the frequency and impact of donor CH in cross-sectional studies of alloHCT recipients.^47–54

Donor CH characteristics in the setting of allogeneic transplant

Prior to stem cell collection, alloHCT candidate donors undergo screening to determine their eligibility for donation. Individuals with complete blood count (CBC) abnormalities and those with a prior history of cancer are usually excluded from donation. Therefore, suitable alloHCT donors comprise a healthy subset of the general population with no evidence of hematologic disease. This observation is useful for interpreting the characteristics of donor CH, especially in comparison with that of ASCT CH.

The mutation spectrum of donor CH closely resembles that found in cross-sectional studies of non-cancer cohorts. DNMT3A is by far the most commonly mutated gene, followed by TET2 and, to a much lesser extent, ASXL1 (Figure 1). In contrast to ASCT cohorts, mutations in PPM1D and TP53 are less common. Additionally, the prevalence of CH with JAK2, SF3B1, and SRSF2 is very low in stem cell donors. These mutations have been reported at low frequency in unselected cohorts,^1,2 and have been linked with high risk of developing myeloproliferative neoplasms (JAK2^55,56) and myelodysplastic syndrome (SF3B1^57,58 and SRSF2⁵⁹). Their low prevalence in donor CH cohorts may mean that they do not usually occur without concomitant CBC abnormalities that would trigger exclusion from donor candidacy.

The majority of donors with CH only have a single mutation.^47,49 As in all other studies, advancing donor age is strongly associated with the frequency of CH, and there are no other consistent demographic associations. In some studies, sibling donors appear to have a higher incidence of CH than unrelated donors,⁴⁹ but this association is entirely due to the higher average age of siblings relative to the unrelated donor pool.

Post-alloHCT outcomes associated with donor CH

Clinical outcomes associated with donor CH are neutral to favorable. However, the results from existing cross-sectional studies of alloHCT cohorts are not fully congruent with each other and must be interpreted with caution.

Donor-engrafted CH was first described in conjunction with the evolution of DCL, which develops in about 0.1% of transplant recipients.^44,45,60 Descriptive assessments of DCL in donor CH studies have shown that donor CH with single DNMT3A or TET2 mutations poses a very low risk of evolution to DCL.⁴⁹ Rather, the majority of DCL arises in recipients of donors who have CH with TP53 mutations or splicing factor mutations, or occult germline risk alleles for hematologic malignancy such as mutations in DDX41.⁶¹ Given the rarity of TP53 and splicing factor mutations in donor CH cohorts, their enrichment in DCL cases suggests that they pose an increased risk of leukemic transformation in recipients, even though formal assessments of the magnitude and significance of the association have not been performed.

The impact of donor CH on broader recipient outcomes has now been assessed in five cross-sectional studies of alloHCT cohorts published in the last three years.^47–51 Direct comparisons of these studies, which are summarized in Table 2, are not possible given differences in cohort size and composition, sequencing depth and platform, and thresholds for defining CH. In four of the five studies, however, recipients of donor CH displayed evidence of potentiated alloimmunity. In two studies, this manifested as an increase in the risk of acute graft-versus-host disease (aGVHD) without a concomitant increase in chronic graft-versus-host disease (cGVHD) risk.^48,50 In two other studies, recipients of donor CH had an increased risk of cGVHD and a reciprocal decrease in relapse risk, but no difference in the risk of aGVHD.^47,49

Table 2.

Studies of donor CH.

	SD	Sample	Timepoint	Donor age	VAF	N (CH/total)	Outcomes
							aGVHD	cGVHD	Relapse	PFS	OS
Frick47	RC	D	Pre	≥50	0.02	80/500	No effect	Increased	Decreased	Increased	NS
Gibson49	RC	D	Pre	≥40	0.01	241/1727	No effect	Increased	Decreased	Increased	Increased
Oran48	RC	D	Pre	≥55	0.02	65/363	Increased	No effect	No effect	No effect	No effect
Gillis50	RC	D	Pre	≥55	0.02	41/299	Increased	No effect	No effect	No effect	No effect
Kim51	RC	D	Pre	11–75	0.006	25/372	No effect	No effect	No effect	No effect	No effect
Newell52	CC	R	Pre + Post	27–69	0.005	15/209	No effect	Increased	No effect	No effect	No effect
Heumüller53	CS	R	Post	<40	0.05	1/22	-	-	-	-	-
Boettcher54	CS	D+R	Post	15–65	0.01	13/42	-	-	-	-	-

SD (Study Design): RC, retrospective cohort; CC, case-control; CS, cross-sectional. Sample: D, donor; R, recipient. Timepoint of sample: pre- or post-transplant. VAF, variant allele fraction. For retrospective cohort studies, the VAF shown is that reported for outcome associations. Outcomes: aGVHD, acute graft-versus-host disease; cGVHD, chronic GVHD; PFS: progression-free survival; OS: overall survival. The red background indicates adverse associations, blue background indicates favorable associations.

No study of donor CH has shown an adverse association with recipient overall and relapse free survival (RFS). In the two studies that show an association between donor CH and increased cGVHD/reduced relapse, those with donor CH had improved RFS compared with those without donor CH.^47,49 In the larger of these two, this translated to an improvement in OS as well, though the effect was confined only to those who received traditional calcineurin-inhibitor-based GVHD prophylaxis (that is, those in whom the cGHVD/relapse effect was evident).⁴⁹ The two studies in which donor CH was only associated with aGVHD risk did not show a trend towards a survival effect. Large registry-based studies will be required to resolve these divergent findings.

The mechanistic basis of potentiated alloimmunity in donor CH is not yet understood. Subset analysis in one study showed that the increased risk of cGVHD, and reciprocal decrease in relapse risk, was primarily driven by DNMT3A-CH and did not extend to patients who received post-transplant cyclophosphamide (PTCy),⁴⁹ which potently suppresses the proliferation of alloreactive T-cells.⁶² In this same study, recipients of donor CH had evidence of a reinforced network of proinflammatory cytokines following transplant, and in particular had increased levels of IL-12p70, which is produced by myeloid cells and stimulates Th1 polarization and IFN-gamma production by CD4+ T-cells.^63,64 These findings suggest a model in which myeloid cells with somatic DNMT3A mutations potentiate alloimmunity through enhanced crosstalk with alloreactive T-cells, whose presence is necessary to mediate the effect.

In an alternative model of donor CH, which is not mutually exclusive with enhanced myeloid/T-cell crosstalk, DNMT3A mutations in HSCs propagate into T-cells during lymphoid differentiation and directly augment their function. Evidence supporting this model is, so far, primarily inferential. First, DNMT3A mutations in CH can be found in T-cell lineages with greater frequency than some other mutations, particularly TET2.⁶⁵ Second, de novo DNA methylation has been shown to be important for aspects of both T-cell development and function. In model systems, deletion of DNMT3A accelerates the development of long-lived memory T-cells⁶⁶ and confers resistance to exhaustion in CD8+ T-cells.⁶⁷ Finally, Dnmt3a knockout in CD8+ T-cells promotes the development of severe GVHD and confers superior graft-versus-tumor activity in mice.⁶⁸ However, many missense DNMT3A mutations found in CH are predicted to confer only partial loss of methyltransferase function on a single allele,⁶⁹ and the average clone size is small,⁴⁹ meaning CH is not biologically equivalent to the full DNMT3A deletion utilized in functional studies. The extent to which the potentiated alloimmunity associated with donor CH is a direct effect of DNMT3A mutations in T-cells is therefore uncertain and is probably variable.

Clinical Implications

The results from studies of donor CH have not yet been translated into clinical practice. The barriers to clinical implementation can be divided into those concerning donors and those concerning recipients.

Systematic screening for CH may mandate disclosure of results to candidate donors. For many, a disclosure of confirmed CH will amount to an unwanted diagnosis of an adverse condition for which there are currently no available treatments. In other settings, a diagnosis of CH has been shown to induce significant anxiety and emotional distress.⁷⁰ New risk prediction models for CH that apply to the general population^71,72 have shown that many individuals with CH have a low risk of CH-associated adverse outcomes over 10 years, raising the possibility that not all CH must be disclosed to candidate donors. These questions, however, must be prospectively addressed in practice guidelines to ensure that CH screening programs are standardized and equivalent across all transplant centers.

Practically, implementation of donor CH screening programs will require standardized informed consent and disclosure processes for both local transplant centers, which handle sibling and related candidate donor evaluations, and national donor registries, which handle unrelated donor evaluations. These protocols must include mechanisms for referrals to genetic counselors who can provide validated confirmatory testing that discriminates between somatic and inherited gene mutations since many of the genes affected in CH can also be mutated in germline cancer predisposition syndromes.^73,74 When available, centers may also refer donors to specialized CH clinics, which currently primarily provide surveillance but could in the future offer enrollment in clinical trials aimed at mitigating risks associated with CH.^75,76

For recipients, the immediate practical question is whether the identification of CH should influence the decision to include or exclude a candidate donor.^77,78 Answering this question will require confirmation of the observation that donor CH is associated with potentiated alloimmunity in large registry studies and delineation of the specific genetic characteristics associated with both favorable and adverse recipient outcomes. These assessments must define the magnitude of impact on recipient outcomes in comparison with other components of current donor selection algorithms, including HLA mismatch,⁷⁹ donor age,^80,81 and donor/recipient relatedness.⁸² This will be necessary to resolve the most likely donor choice scenarios, such as whether to select an older related donor with DNMT3A-CH or a younger unrelated donor without CH. Sufficiently large registry studies may also be able to determine the feasibility of a risk-adapted approach to donor selection in which donors with DNMT3A-CH are preferentially selected for recipients at highest risk of relapse and excluded for those with lower risk of relapse or non-malignant indications for transplant, who would not be expected to benefit from augmentation of the GVL.

Absent confirmation that DNMT3A-CH potentiates alloimmunity and benefits some recipients, there is currently insufficient evidence to justify the implementation of broad donor screening for CH. There could, however, be a role for limited donor screening for mutations that confer a high risk of DCL, such as TP53, splicing factor, and rare germline risk alleles, assuming processes for donor consent and disclosure are in place. Although this sequencing would usually be negative, the cost savings associated with preventing even a single case of DCL could be substantial enough to justify it.

It is important to be aware that these considerations only apply to true CH discovered in the process of routine systematic screening of candidate donors, which as above does not currently exist. At present, candidate donors usually only undergo molecular testing in response to CBC abnormalities.⁸³ Clonal hematopoietic mutations discovered in this scenario cannot be classified as CH, which by definition is not associated with cytopenias,⁸⁴ and should be treated with caution, especially in the case of multiple mutations or mutations in genes other than DNMT3A or TET2.

The most important implications of the current evidence for donor CH, and DNMT3A-CH in particular, may extend beyond matters of donor choice to deeper insights about the alloimmunity GVL effect. It is notable that in the two studies in which DNMT3A-CH associates with increased cGVHD and reduced relapse, it does so without increasing the risk of NRM. Confirming this effect in larger cohorts and discerning the mechanism underlying it could open the door to new strategies for augmenting the GVL. These might include pharmacologic therapies post-alloHCT that phenocopy the effect of DNMT3A-CH, and gene editing of cellular products to enhance the efficacy of donor lymphocyte infusions for relapsed disease.

Conclusions

The mutational spectrum and clinical associations of CH have been repeatedly assessed for both autologous and allogeneic transplantation, but distinct outstanding questions and challenges remain for each. For autologous transplantation, the frequency and spectrum of mutations and their clinical implications are clear, but it is not yet known whether the risk associated with CH can be modified by alternative therapeutic choices. For allogeneic transplantation, large registry-based studies are needed to resolve conflicting observations from existing studies of donor CH. Functional assessments to determine the mechanistic basis of donor CH’s association with potentiated alloimmunity will be necessary to extend the impact of these observations from limited questions surrounding donor choice to deeper insights that benefit broad subsets of allogeneic transplant recipients.