Inferred clonal hematopoiesis from tumor DNA sequencing among men with prostate cancer: correlation with somatic tumor alterations and outcomes

Abstract

Introduction:

Clonal hematopoiesis of indeterminate potential (CH) is the clonal expansion of hematopoietic stem and progenitor cells harboring somatic mutations in driver genes typically associated with hematologic malignancies, and occurring in patients without known hematologic malignancies.¹ The major risk factor for the development of CH in both sexes is advancing age, with an estimated prevalence up to 20% among people over 70 years of age.^2,3 In the general population, CH is associated with an increased risk of cardiovascular events and all-cause mortality.^4,5 The gold standard method for detecting CH is by using next-generation DNA sequencing (NGS) of peripheral blood leukocytes. Historical estimates showed that CH is detected in about 8-35% of men with prostate cancer.^6-10 Much of what we know is based on this being identified in the setting of research studies, but this is not routinely done in clinical practice. The presence of tumor-infiltrating-leukocyte populations in tumor biopsy and surgical specimens can contaminate NGS results, thereby potentially identifying patients with CH even without sequencing of their peripheral blood.⁶ CH detected in this manner is considered to be “inferred,” since it cannot be proven to arise in the leukocyte fraction. The clinical significance of inferred CH is not well established and it is unclear if this may also be associated with poor prognosis. Additionally, whether or not CH is associated with other concurrent somatic alterations that may arise on an NGS sequencing report is not well understood.

Methods

We conducted a retrospective analysis of 396 consecutive men with prostate cancer who underwent NGS as part of routine clinical care at the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center between 2013 and 2023. Patients with clinical-grade NGS done using targeted panel-based sequencing on the primary prostate tissue using the Foundation Medicine platform, either from biopsy or radical prostatectomy, were included. This was done in order to have a treatment naïve sample for analysis. Patients with tumor mutational burden-high status (>10 muts/Mb) or with microsatellite unstable tumors (one each) were removed because these can result in excess mutations in many genes including those potentially attributable to CH. Demographics including age at diagnosis, self-reported race, Gleason score, stage at diagnosis, and baseline PSA were abstracted from the medical records. NGS reports were reviewed for the presence of a pathogenic or likely pathogenic mutation reported in one or more of the 27 genes commonly associated with CH (Table 1).⁹ Genes that may be found in CH but are more likely to be tumor-related, such as TP53, were excluded because we would be unable to disambiguate the cell of origin based purely on the reports. This study was approved by the Institutional Review Board at Johns Hopkins University.

Patient demographic and disease characteristics were summarized via descriptive statistics by the presence or absence of inferred CH. The group differences between presence or absence of CH in baseline characteristics were examined via 2 sample t-test for continuous variables, and Chi-square test or Fisher exact test as appropriate for categorical variables. Overall survival (OS) was defined from date of prostate cancer diagnosis to date of death, or censored at the date of last follow-up for patients still living. OS was summarized using the Kaplan-Meier method, and the difference between groups (presence versus absence of CH) was compared via log-rank test. Multivariable analysis was performed using the variables found to be positively associated with CH. To assess the pattern of consistency on the OS differences between CH status, subgroup analyses were conducted without multiplicity adjustment for exploration via Cox proportional hazard models. Furthermore, group differences between CH status in prevalence of somatic mutations were compared via Chi-square test or Fisher exact test as appropriate. All P-values were 2-sided with <0.05 considered statistically significant. All analyses were conducted using R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria).

Results:

A total of 396 patients were included in the analysis. Patients had a median follow-up time of 7.8 years from the date of diagnosis (range 0.2-21.4 years). Approximately 12% of patients had inferred CH (n = 46) detected from tumor NGS analysis (Figure 1). ASXL1, TET2 and DNMT3A were the most commonly affected genes, with a prevalence of 2.3% (n = 9), 1.8% (n = 7), and 1.5% (n = 6), respectively (Figure 1, Supplementary Table 1). Of those with CH, 89% of patients (n = 41 of 46) had 1 CH mutation, 2 patients had 2 concurrent mutations, and 3 patients had 3 concurrent CH alterations found. Demographic information and baseline disease characteristics are shown in Table 2. Patients in whom CH was detected were on average 5 years older than those without inferred CH (mean age 68 vs 63 years; P < .001 for the difference). A higher proportion of CH-positive patients self-identified as African American (26%) compared to those who did not have inferred CH (15%; P for difference = .05). A similar proportion of patients had metastatic disease at the time of diagnosis (28% in those without inferred CH compared to 33% in those with inferred CH, P for difference .5). Neither Gleason grade group at diagnosis, percentage with lymph node invasion, extraprostatic extension, nor seminal vesical invasion varied significantly by the presence or absence of inferred CH (Table 2). There was also no difference in the treatments patients received over the course of their disease, with similar rates of chemotherapy received (P = 0.14 for difference) and androgen receptor pathway inhibitors received (P = 0.22 for difference).

Figure 1.

Prevalence of inferred clonal hematopoiesis, overall and by specific gene.

Open in new tabDownload slide

In univariate analysis, those with inferred CH mutations based on clinical NGS testing had an 81% increased risk of death compared to those without inferred CH (hazard ratio 1.8, 95% CI, 1.1, 3.1, P = .03) (Figure 2). However, after adjusting for age at diagnosis and race, the impact of CH on overall survival was attenuated and no longer statistically significant (HR 1.1, 95% CI, 0.6, 2.0, P = .77).

Figure 2.

Kaplan-Meier curve showing overall survival, from time of diagnosis, in those with and without inferred clonal hematopoiesis

Open in new tabDownload slide

Next, we explored the co-occurrence of common tumor mutations with the presence or absence of CH. There was no difference in prevalence between those with and without inferred CH with respect to the identification of mutations in many prostate cancer-associated genes including TP53, PTEN, SPOP, APC, BRCA2, RB1, or KMT2C. However, we did find that those with inferred CH were more likely to also have mutations in PIK3CA (prevalence of 15% vs 4%), CTNNB1 (prevalence of 13% vs 2%), and BRCA1 (4% vs 0.3%); P for difference < .05 for all three comparisons (Figure 3).

Figure 3.

Prevalence of other concurrent tumor driver mutations, by the presence or absence of inferred clonal hematopoiesis. *P-value < .05 for difference

Open in new tabDownload slide

Discussion

In this study, we found that inferred CH was prevalent among men with prostate cancer who underwent somatic tumor DNA sequencing of primary prostatic tissue as part of standard clinical care and was detected in approximately 12% of patients in our cohort. This is perhaps surprising in prostate cancer as there is a paucity of immune cells that typically are found in primary prostate tissue.¹¹ Overall, the prevalence of inferred CH in this cohort is similar, but on the low end of the rates of patients with prostate cancer who have been found to have CH in prior studies, where estimates range from 6% to as high as 57% although the distribution of mutations is different than the general population^6,7,9,10,12 Given that the results in this study are inferred from NGS testing reports of primary tumor tissue and not from the blood, we would anticipate that the rates are lower than those reported from leukocyte sequencing (ie, true CH prevalence). Similar to other studies, there was an association between the identification of CH and older age, supporting that what is being detected clinically may in fact be CH.^6,7,9,10,12 While CH has not been associated with an increased risk overall of developing prostate cancer, the clinical significance of CH among men with coexisting prostate cancer has not been well studied.¹³ In patients with prostate cancer, cardiovascular disease is a major cause of morbidity and mortality (a risk that is further accentuated by use of androgen deprivation therapies) and CH may influence who is more at risk of cardiovascular complications. One study suggested that overall, there was no impact on outcomes in those with advanced metastatic castration-resistant prostate cancer who had CH compared to those who did not have CH.¹² However, that prior study did find that those with TET2-mutated CH were more likely to have a cardiovascular event while on treatment with the oral hormonal agents enzalutamide and abiraterone.¹² Additionally, poly-ADP ribose polymerase (PARP) inhibitors, that are approved for use in men with certain germline or somatic DNA-repair gene mutations, have a rare but serious side effect of treatment-related myelodysplasia, and concurrent CH may be a risk factor for the development of this complication.¹⁴ In addition, there have been some preliminary data, including among men with prostate cancer, that individuals with germline mutations (in particular in CHEK2) who would be eligible for PARP inhibitors, may be at higher risk of CH.¹⁵ Despite these links with complications related to prostate cancer treatment, when inferred based on clinical-grade testing, there was no significant impact of CH on overall survival after adjusting for age in our current study.

Another novel finding of our work is the potential association of CH with other somatic alterations, notably PIK3CA, CTNNB1, and BRCA1. To the best of our knowledge, this is one of the first studies to identify such an association. From this study, we cannot determine if these co-occurring mutations are in the tumors, suggesting that mutated myeloid cells may be impacting the mutational profile of the tumor itself. Alternatively, these somatic mutations may also be occurring in the myeloid cells containing CH mutations and contaminating the tumor specimen being sequenced. Previous research has estimated an approximate 10% rate of interference with CH mutations when interrogating for mutations in DNA repair pathway genes.¹⁶ While BRCA1 is the only actionable gene identified in our study (for PARP inhibitor eligibility) the numbers of patients in this sample are very small. CTNNB1 and PIK3CA are in the WNT and PTEN/PI3K pathways respectively, and treatments targeting those mutations are currently being explored in prostate cancer. If not truly tumor-derived mutations, then these may represent additional mutations for clinicians to be aware of as they could be contributing to interference and may affect the interpretation of clinical NGS reports.

There are a number of limitations to our study. This is a retrospective study and was based on a single-institution experience. This population is also skewed to those with advanced cancer who underwent genomic testing for clinical reasons in a Medical Oncology clinic and may not be representative of everyone with advanced prostate cancer or those who did not have tissue available for analysis. These samples may also not reflect the tumor landscape of primary prostate cancers that are managed with curative-intent therapy and that never recur. These are from primary prostate tissues and thus these patients are generally treatment naïve. We did not include metastatic tumor biopsies in this study to remove the potential for bias based on prior treatments for cancer as it has been established that treatments used for cancer, for example, radiation therapy and PARP inhibitors, can contribute to the progression of CH among patients with prostate cancer as well as other cancers.^9,17,18 We also did not include blood-based (circulating tumor DNA) analysis as this is not commonly done in our practice and typically not done at the same time as sequencing of the primary tumor.¹⁹ Finally, we did not conduct DNA sequencing of patient-matched leukocytes (buffy coat from blood), so we were unable to prove the presence of CH alterations through our analysis of tumor biopsies, and we do not have serial blood samples to assess changes in CH clones over time. The association with age and similar prevalence to prior studies suggests that it is likely that at least some of those identified are true examples of CH.

Conclusion

CH can be commonly detected on the results of somatic NGS testing in men with advanced prostate cancer but did not clearly have a negative prognostic impact on overall survival in this small study. Interestingly, inferred CH may be associated with the identification of other somatic tumor mutations, either in the tumor or also in myeloid cells, and may have further implications for interpretation of clinical NGS reports. These findings require large-scale confirmation and validation by conducting paired analysis of tumor and leukocyte DNA.

Funding

CHM–National Cancer Institute P30 CA006973, V Foundation, Prostate Cancer Foundation, Winn Career Development Award. E.S.A. is partially supported by National Cancer Institute grant P30 CA077598 and Department Of Defense grant W81XWH-22-2-0025.

Conflicts of interest

C.H.M reports personal fees from Tempus, Astellas, and grants to institution from AstraZeneca. (none relevant to this work). E.S.A. reports grants and personal fees from Janssen, Sanofi, Bayer, Bristol Myers Squibb, Curium, MacroGenics, Merck, Pfizer, AstraZeneca, and Clovis; personal fees from Aadi Bioscience, Aikido Pharma, Astellas, Amgen, Blue Earth, Corcept Therapeutics, Exact Sciences, Hookipa Pharma, Invitae, Eli Lilly, Foundation Medicine, Menarini-Silicon Biosystems, Tango Therapeutics, Tempus, and Z-alpha; grants from Novartis, Celgene, and Orion; and has a patent for an AR-V7 biomarker technology that has been licensed to Qiagen

Data availability

The data supporting this research may be made available upon request from the corresponding author (CHM). The data are not publicly available due to privacy restrictions given the clinical nature of the data.

References

Author notes