Clonal hematopoiesis is not significantly associated with COVID-19 disease severity

TO THE EDITOR:

The severity of COVID-19 disease, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is highly variable ranging from asymptomatic to a self-limited flulike illness to severe respiratory failure, often accompanied by cardiovascular events, coagulopathy, thrombosis, and high mortality.^1-3 Although several risk factors for severe disease have been identified, including age, sex, ethnicity, genetic variation, and a range of comorbidities, these only partially predict disease severity and additional determinants remain to be identified.^2-6 

Clonal hematopoiesis (CH) describes the disproportionate expansion of a hematopoietic stem cell and its progeny, in association with leukemia-associated somatic mutations, most commonly affecting the genes for epigenetic regulators DNMT3A, TET2, and ASXL1.^7,8 The prevalence and size of such clones rise with age, in association with changes in the driver gene landscape.⁹ CH is associated with an increased risk of hematologic malignancies, but also of cardiovascular disease (CVD), independently of other known CVD risk factors.^10,11 The basis for this increased CVD risk has been linked to hyperinflammatory positive feedback loops driven by increased cytokine release from clonal myeloid cells, particularly interleukin-6 and interleukin-1β.^12-16 

The close association of CH with advancing age and chronic inflammation led us to hypothesize that it may be another factor associated with increased risk of severe COVID-19 disease, through hyperactivation of abnormal, clonally derived, myeloid cells, including monocytes and macrophages, following SARS-CoV-2 infection.

To investigate a possible association between CH and COVID-19 disease severity, we studied 568 patients aged 50 to 90 years old (median age, 64 years), including 120 nonhospitalized individuals with asymptomatic or mild disease, 241 hospitalized patients not requiring intensive care unit (ICU) support, and 207 critically ill patients who required ICU admission, or mechanical ventilation or went on to die (Table 1). All patients had laboratory-confirmed SARS-CoV-2 infection during the first 6 months of 2020. All participants provided written informed consent as part of ethics committee-approved studies (supplemental Note on the Blood Web site).

To identify individuals with CH, we performed error-corrected targeted sequencing of blood DNA for 56 genes implicated in CH using a custom set of RNA baits (Twist Bioscience design ID TE-99420296; supplemental Table 1). Sequences were mapped to human reference genome GRCh38 and CH somatic driver mutations with a variant allele fraction (VAF) of 1% to 40% were identified using Shearwater (SNV)¹⁷ and Mutect2 (Indels).¹⁸ With median sequencing coverage of 2000×, we detected 266 CH driver mutations within 22 genes (supplemental Tables 2 and 3), with 188/568 (33%) patients having at least 1 mutation (Figure 1A). DNMT3A and TET2 mutations were most common in all 3 groups, with no significant enrichment for particular genes in any group (Figure 1B).

CH mutations in at least 1 gene were identified in 37 (31%) nonhospitalized, 74 (31%) hospitalized, and 77 (37%) critically ill patients (Figure 1C). There was no significant difference in the prevalence of CH between groups (P = .35, χ² test). We next examined CH prevalence by age and found that, although this increased with advancing age in all 3 groups, the groups did not differ significantly when comparing individuals in the same age ranges (Figure 1D). In addition, with most CH carriers harboring 1 or 2 mutations, there were no differences between the 3 groups with regard to the mean number of mutations per patient (P = .80, 1-way analysis of variance test; Figure 1E) or the average clone size as measured by VAF (P = .23, 1-way analysis of variance test; Figure 1F). To investigate whether mutation-bearing myeloid cells were preferentially mobilized or expanded during the acute clinical course of COVID-19, we studied available paired samples, taken 8 days apart, from 54 critically ill patients. CH was identified in 16 (32%). Comparison of VAFs between day 1 and day 9 samples did not differ significantly (P = .27, pairwise t test) (Figure 1G), indicating that there was no preferential expansion of myeloid progeny arising from the CH clone.

To take account for covariates previously implicated in COVID-19 disease severity including age, sex, ethnicity, diabetes, chronic obstructive pulmonary disease/asthma, CVD, cancer/neoplasm, immunodeficiencies, and smoking status, we applied a multivariable proportional odds model to retest for a possible association between COVID-19 disease severity and CH (supplemental Table 4). We found that male sex (adjusted odds ratio [OR] = 2.84; 95% confidence interval [CI], 1.91-4.24; P < .01), diabetes (adjusted OR = 1.56; 95% CI, 1.05-2.44; P = .044), cardiovascular disease (adjusted OR = 1.64; 95% CI, 1.01-2.44; P = .046), and immunodeficiency (adjusted OR = 2.10; 95% CI, 1.11-4.07; P = .024) were all significantly associated with COVID-19 hospitalization and ICU admission, consistent with previous findings.⁵ However, even after adjusting for these factors, the presence of CH was not associated with an increased risk of severe COVID-19 (OR = 1.24; 95% CI, 0.82-1.89; P = .31; Figure 1H). Similarly, neither were the number of mutations per patient (adjusted OR = 1.09; 95% CI, 0.88-1.36; P = .43) nor the CH clone size (adjusted OR = 1.63; 95% CI, 0.05-53.88; P = .78) (Figure 1H).

Given the reported links between TET2 and DNMT3A mutations and hyperinflammation or response to infection,^15,16,19 these 2 gene mutations were interrogated individually for a possible association with COVID-19 severity using a proportional odds model. Mutations in TET2 and DNMT3A were also not associated with COVID-19 disease severity (Figure 1H). Finally, given that large clone size is more strongly associated with CVD,¹⁰ we analyzed the risk associated with large CH clones (VAF ≥ 5%), and again found no association with COVID-19 disease severity (OR = 1.29; 95% CI, 0.69-2.42; P = .42; Figure 1H). Similarly, we found no association between CH with VAF ≥ 10% and COVID-19 disease severity (OR = 1.06; 95% CI, 0.50-2.28; P = .88).

In summary, our study found no evidence that CH is associated with COVID-19 disease severity, even after adjusting for covariates known to affect the risk of severe disease. Previous studies examining the association between COVID-19 disease severity and clonal hematopoiesis CH have produced conflicting results. Three smaller studies concluded that CH is not associated with COVID-19 disease severity.^20-22 However, conclusions were less definitive because of comparisons only to historical non-COVID-19 controls,²⁰ small sample size/power to detect potentially relevant associations,²¹ different sequencing platforms used for cases and controls,²⁰ and/or limited availability of additional comorbidity/risk factor data.²¹ A larger study (n = 413) examined the relationship among patients with solid cancers at various stages during treatment (MSK-IMPACT cohort), and reported that nonputative driver clonal hematopoiesis mutations were associated with COVID-19 disease severity.²³ This finding may have reflected the impact of prior cancer treatment on patients with reduced hematopoietic stem cell numbers/reserve. The same study used an independent noncancer cohort (n = 112) for validation, and found no significant associations within this smaller cohort, although fixed-effects meta-analysis of the combined 2 cohorts remained positive, likely driven by the larger MSC-IMPACT cohort.²³ Also, a recent study reported an association between mosaic chromosomal alterations, a distinct form of CH, and risk of COVID-19 hospitalization.²⁴ Our current study attempted to mitigate these prior limitations by studying the largest number of patients to date, directly comparing relevant patient groups (asymptomatic/mild, hospitalized, critically ill), incorporating covariates from well-characterized additional risk factors, and performing sequencing and mutation calling using the same platforms. Overall, we found no evidence of an association between CH and COVID-19 severity, resolving much of the uncertainty surrounding this question. Although it is never possible to rule out an association with absolute certainty, our study indicates that the clinical impact of any theoretical association is unlikely to be substantial (Figure 1H).