key: cord-0889737-lxsymt38
authors: Bolton, K. L.; Koh, Y.; Foote, M. B.; Im, H.; Jee, J.; Sun, C. H.; Safonov, A.; Ptashkin, R.; Moon, J. H.; Lee, J. Y.; Jun, J.; Kang, C. K.; Song, K.-H.; Choe, P. G.; Park, W. B.; Oh, M.-d.; Kim, H. B.; Song, H.; Kim, S.; Patel, M.; Derkach, A.; Gedvilaite, E.; Tkachuk, K. A.; Braunstein, L. Z.; Gao, T.; Papaemmanuil, E.; Babady, E.; Pessin, M. S.; Kamoj, M.; Diaz, L. A.; Ladanyi, M.; Rauh, M. J.; Nataranjan, P.; Machiela, M. J.; Awadalla, P.; Joseph, V.; Offit, K.; Norton, L.; Berger, M.; Levine, R. L.; Kim, E. S.; Kim, N. J.; Zehir, A.
title: Clonal hematopoiesis is associated with risk of severe Covid-19
date: 2020-11-27
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2020.11.25.20233163
sha: 4d6232337892586af33f8a91a9d8541743e10783
doc_id: 889737
cord_uid: lxsymt38

Acquired somatic mutations in hematopoietic stem and progenitor cells (clonal hematopoiesis or CH) are associated with advanced age, increased risk of cardiovascular and malignant diseases, and decreased overall survival. These adverse sequelae may be mediated by altered inflammatory profiles observed in patients with CH. A pro-inflammatory immunologic profile is also associated with worse outcomes of certain infections, including SARS-CoV-2 and its associated disease Covid-19. Whether CH predisposes to severe Covid-19 or other infections is unknown. Among 515 individuals with Covid-19 from Memorial Sloan Kettering (MSK) and the Korean Clonal Hematopoiesis (KoCH) consortia, we found that CH was associated with severe Covid-19 outcomes (OR=1.9, 95%=1.2-2.9, p=0.01). We further explored the relationship between CH and risk of other infections in 14,211 solid tumor patients at MSK. CH was significantly associated with risk of Clostridium Difficile (HR=2.0, 95% CI: 1.2-3.3, p=6x10-3) and Streptococcus/Enterococcus infections (HR=1.5, 95% CI=1.1-2.1, p=5x10-3). These findings suggest a relationship between CH and risk of severe infections that warrants further investigation.

Acquired mutations that lead to clonal expansion are common in the normal aging hematopoietic system (clonal hematopoiesis, or CH), yet are known to alter stem/progenitor and lymphoid function and response to environmental stressors, including systemic infections 5, 6, 9, 10 . The mutational events that drive CH overlap with known drivers of hematologic malignancies. However, the majority of mutations in CH appear to occur outside of canonical cancer driver genes 11, 12 . The impact of individual mutational events on hematopoietic stem and progenitor cells differs by the nature of the genomic aberration. For example, chromosomal aneuploidies result in a predisposition for lymphoid fate specification and transformation 13, 14 while point mutations in DNMT3A result in increased myeloid differentiation 6, 15 . Heterogeneity also exists across CH phenotypes by driver gene in regards to its impact on inflammatory signaling 6 . For example, mutations in TET2 result in heightened secretion of several cytokines including IL-1β/IL-6 signaling that may partially explain the increased risk of cardiovascular disease 5, 9, 16 . Moreover, systemic infections and the resultant inflammatory signals can lead to increased clonal fitness of TET2 mutant cells and clonal expansion 10, 17, 18 .

Despite these important insights, the relationship between different CH events, infectious risk and infectious disease severity has not been studied. The severity of Covid-19 is also associated with advanced age, cardiovascular and malignant comorbidities, and elevated circulating IL-6 levels; features which are seen with ageassociated CH [19] [20] [21] [22] [23] . Given the common inflammatory profile of CH and Covid-19 infection, we investigated the relationship between CH and Covid-19 including the potential for an association of CH with increased Covid-19 disease severity.

Our study included patients from two separate cohorts. The first cohort was composed of patients with solid tumors treated at Memorial Sloan Kettering Cancer Center (MSK) with blood previously sequenced using MSK-IMPACT, a previously validated targeted gene panel capturing all commonly mutated CH-associated genes (Supplementary Table 1) 24 . Of these patients, 1,626 were tested for SARS-CoV-2 (the virus that causes All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Covid-19) RNA between March 1st 2020 and July 1st 2020; 403 (24.8%) individuals tested positive for SARS-CoV-2 (Methods and Table 1 ). The second cohort included 112 previously healthy individuals without cancer who were hospitalized for Covid -19 between January and April 2020 at four tertiary hospitals in South Korea (KoCH cohort).

The KoCH cohort was sequenced using a custom targeted NGS panel from Agilent (89 genes) which was designed to include commonly occurring CH genes (Supplementary Figure 1 ). In the KoCH cohort, CH was observed in 25% and 15.9% of patients with severe versus non-severe Covid-19, respectively (adjusted OR 1.85, 95% CI 0.53-6.43, Figure 1 ). In a fixed effects meta-analysis of odds-ratio estimates from the multivariable logistic regression models employed in each separate cohort analysis, the presence of CH was associated with an increased risk of severe Covid-19 (OR=1.85, 95%=1.15-2.99, p=0.01) (Figure 1 ).

Using previously described methods 24 , CH mutations were classified as known or hypothesized cancer putative drivers (PD-CH) or non putative drivers (non-PD CH). The All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Given the evidence of an association between CH and Covid-19 severity, we sought to explore the relationship between CH and other types of infections. We analyzed billing codes from 14,211 solid tumor patients treated at MSK who underwent blood sequencing by MSK-IMPACT. Using a previously established phenome-wideassociation study (Phe-WAS) methodology 25 , we mapped patient ICD-9 and ICD-10 billing codes to categories of infectious disease. Multivariable Cox proportional hazards regression was used to estimate the hazard ratio (HR) for risk of infection among CH positive compared to CH negative individuals. Given the number of model covariates, we limited the analysis to 32 infection subclasses that affected at least 80 individuals (see Methods). Multiple infection types were associated with CH, although many All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint associations were not statistically-significant after multiplicity adjustment ( Figure 2B ).

In summary, we show in cancer and non-cancer patients that CH is associated with increased Covid-19 severity. In a large cancer patient cohort, CH is also associated with other severe infections, namely Streptococcus/Enterococccus and Clostridium difficile infections. Our exploratory analysis suggests that the relationship between CH and Covid-19 and CH and Clostridium difficile infection may be partly driven by non-driver CH. Clonal expansions characterized by non-driver mutational events could be facilitated by multiple mechanisms. Many classes of genetic alterations, such as copy number events (CNVs), structural variants, non-coding, and epigenetic changes, are not detectable using the targeted panels included in this study. As such, the observed events that are highly enriched in CH could be 'passenger' mutations that co-occur with a positively selected, undetected 'driver' mutation such as recurrent CNVs. 12, 13 Alternatively, driver mutations may have been incompletely classified as "non-driver" events using our methodology. However, cancer driver genes tend to recur in multiple patients, and the majority of witnessed non-driver mutated genes in our cohort were non-recurrent suggesting that clonal expansion, and not the specific event driving clonal expansion, may be associated with Covid-19 disease severity. This will need to be further studied in larger cohorts. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint

The hematopoietic system is a key regulator of inflammation and immunity. A substantial body of evidence now links somatic alterations in hematopoietic stem and progenitor cells to a variety of health outcomes, with inflammation emerging as a key mediator. [2] [3] [4] [5] [10] [11] [12] [13] Our data, along with results in the accompanying manuscript by Zekavat et al., demonstrate a similar association between CH and increased infection severity. This association may be due to residual confounding by variables that are unknown and unaccounted for in our models. Alternatively, this could represent a novel pathophysiology that links CH-induced changes in hematopoietic stem, progenitor, and lymphoid cell function with immune regulation and infection response. Future investigation including functional studies will be important to clarify the mechanisms underlying the association between CH and infection risk and to develop potential interventional strategies to attenuate inflammation, clonal expansion, and infectious sequelae in patients with and without cancer.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint

The study population included 9,307 patients with non-hematologic cancers at MSKCC who were alive on May 1st 2020 and underwent matched tumor and blood sequencing before this date using the MSK-IMPACT panel on an institutional prospective tumor sequencing protocol (ClinicalTrials.gov number, NCT01775072). Subjects who had a hematologic malignancy diagnosed after MSK-IMPACT testing or who had an active hematologic malignancy at the time of blood draw were excluded. Demographics, smoking history, exposure to oncologic therapy and primary tumor site were extracted from the electronic health record. Accuracy of populated information was manually checked in the EMR by three independent physicians (K.B, M.F, A.S). The presence of co-existing medical comorbidities known to correlate with Covid-19 severity including diabetes, COPD, asthma, hypertension and cardiovascular disease, were ascertained from ICD-9 and ICD-10 billing codes. SARS-CoV-2 status was determined using RT-PCR. We defined severe Covid-19 as the presence of hypoxia requiring supplemental oxygen (supplemental oxygen device >1 L or hypoxia <94%) resulting from Covid-19

infection. There were seven subjects with Covid-19 for whom there was minimal documentation of clinical course following Covid-19 infection and these individuals were excluded. There were three individuals with metastatic cancer and progression of disease at the time of Covid-19 where it was unclear whether documented hypoxia could be attributed to Covid-19 or disease progression. These subjects were also excluded.

Laboratory-confirmed patients with Covid-19 between January and April 2020 in four tertiary hospitals in Republic of Korea were approached for consent to this study. Blood was drawn following confirmation of Covid-19 positivity. All of four hospitals have been running national-designated isolation units, which are located in Seoul, Gyeonggi, or All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Daegu. These provinces had the highest numbers of Covid-19 cases during the period 14 . Clinical and laboratory characteristics were retrospectively reviewed using the electronic medical record systems of each institution. Hypoxia requiring supplemental oxygen was defined as supplemental oxygen device >1 L with O2 <94% resulting from preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. 

Variants from the MSK and KoCH cohort were uniformly annotated according to evidence for functional relevance in cancer (putative driver or CH-PD). We annotated variants as oncogenic if they fulfilled any of the following criteria: 1) truncating variants in NF1, DNMT3A, TET2, IKZF1, RAD21, WT1, KMT2D, SH2B3, TP53, CEBPA, ASXL1, All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. RUNX1, BCOR, KDM6A, STAG2, PHF6, KMT2C, PPM1D, ATM, ARID1A, ARID2,   ASXL2, CHEK2, CREBBP, ETV6, EZH2, FBXW7, MGA, MPL, RB1, SETD2, 

We used multivariable logistic regression to evaluate for an association between clonal hematopoiesis and Covid-19 severity adjusting for age (measured as a continuous variable), gender, race, smoking history and co-existing medical comorbidities including diabetes, COPD/asthma and cardiovascular disease all classified as per Table 1 . This was done separately for the MSK and KoCH cohorts due to limitations of data sharing.

For solid tumor patients at MSK we also adjusted for primary tumor site (thoracic or non-thoracic cancer) and receipt of cytotoxic chemotherapy before and after IMPACT blood draw . We performed a fixed effects meta-analysis of the MSK and KoCH cohorts to jointly estimate the odds ratio for severe Covid-19 among CH positive compared to CH negative individuals.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint

We analyzed billing codes from 14,211 solid tumor patients at MSKCC who had their blood sequenced using MSK-IMPACT. We applied the phecode nomenclature developed at Vanderbilt 9 to map ICD-9 and ICD-10 billing codes to infectious disease subtypes. Subjects who were billed using a ICD9/10 code within the phecode for the first time following their sequencing blood draw with evidence of CH were considered to have an incident infection. Those who were billed for an ICD9/10 code within the phecode prior to blood draw were removed from the analysis of that phecode. In order to evaluate the accuracy of the billing code data, the presence of a documented Clostridium Difficile or Streptococcus infection in an EMR physician note was manually checked for patients respectively identified by billing codes (N=525 patients) by three independent physicians using shared criteria for infection onset. Billing codes were highly accurate in identifying the presence of the respective infectious disease (concordance >95%).

We used Cox proportional hazards regression to estimate the hazard ratio for risk of infection among those with CH compared to CH negative individuals. The date of blood draw (used for MSK-IMPACT sequencing) served as the onset date for this time-toevent analysis; the end-date was the date of billing code entry for the infectious disease subtype phecode, death or last follow-up, whichever came first. All models were adjusted for age, gender, race, smoking, tumor type, and cumulative exposure to cytotoxic chemotherapy prior to blood draw and after blood draw as previously described 10 . Following the 10:1 rule regarding the number of covariates in a multivariable model in proportion to the number of events 16 , we excluded infection subclasses populated with less than 80 individuals. The analysis utilized multiplicity correction with the Benjamini-Hochberg method to establish adjusted q-values for hazard ratio with a prespecified false-discovery-rate (FDR) <0.10.

All the statistical analyses were performed with the use of the R statistical package (www.r-project.org).

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Table 1 . Characteristics of study participants ............................................................................ 22 preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint

plot of the log(Hazard ratio) of infection with CH using multivariable cox proportional hazards regression. B) Association between CH subtype defined by putative driver status and risk of Clostridium Difficle and Streptococcus/Enterococcus infection using cox proportional hazards regression. All models were adjusted for age, gender, race, smoking, diabetes, cardiovascular disease, COPD/asthma, cancer primary site (if history of malignancy), exposure to cytotoxic cancer therapy. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Table 1 . Characteristics of study participants All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Extended Data Figure 3 . Association between CH and Covid-19 severity stratified by the number of mutations. Shown are the results from logistic regression comparing the odds ratios of severe Covid-19 among those with one mutation and those with two or more mutations. Models were adjusted for age, gender, race, smoking, diabetes, cardiovascular disease, COPD/asthma, cancer primary site (if history of malignancy), exposure to cytotoxic cancer therapy for the MSK and Korea Consortia. Summary statistics for a fixed effects meta-analysis are shown.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Extended Data Figure 4 . Association between maximum VAF of CH-mutation(s) and Covid-19 severity. Shown are the results from logistic regression comparing the odds ratios of severe Covid-19 among those with one or more CH mutations <5% VAF compared to no CH and CH with a VAF >5% and no CH. Models were adjusted for age, gender, race, smoking, diabetes, cardiovascular disease, COPD/asthma, cancer primary site (if history of malignancy), exposure to cytotoxic cancer therapy for the MSK and Korea Consortia. Summary statistics for a fixed effects meta-analysis are shown.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint Extended Data Table 1 . Frequency of clonal hematopoiesis by Covid-19 status.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 27, 2020. ; https://doi.org/10.1101/2020.11.25.20233163 doi: medRxiv preprint

Clonal Hematopoiesis and Blood-Cancer Risk Inferred from Blood DNA Sequence

Age-Related Clonal Hematopoiesis Associated with Adverse Outcomes

Prediction of acute myeloid leukaemia risk in healthy individuals

Somatic mutations precede acute myeloid leukemia years before diagnosis

Genetic Interleukin 6 Signaling Deficiency Attenuates Cardiovascular Risk in Clonal Hematopoiesis

Inherited causes of clonal haematopoiesis in 97,691 whole genomes

An inflammatory cytokine signature predicts COVID-19 severity and survival

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Tet2-Mediated Clonal Hematopoiesis Accelerates Heart Failure Through a Mechanism Involving the IL-1β/NLRP3 Inflammasome

Inhibition of Inflammatory Signaling in Tet2 Mutant Preleukemic Cells Mitigates Stress-Induced Abnormalities and Clonal Hematopoiesis

Clonal hematopoiesis, with and without candidate driver mutations, is common in the elderly

Synonymous mutations reveal genome-wide driver mutation rates in healthy tissues

Insights into clonal haematopoiesis from 8,342 mosaic chromosomal alterations

Chromosomal alterations among age-related haematopoietic clones in Japan

Identification of Clonal Hematopoiesis Mutations in Solid Tumor Patients Undergoing Unpaired Next-Generation Sequencing Assays

Clonal hematopoiesis and risk for atherosclerotic cardiovascular disease

Microbial signals drive pre-leukaemic myeloproliferation in a Tet2-deficient host

Antibiotic treatment ameliorates Ten-eleven translocation 2 (TET2) loss-offunction associated hematological malignancies

Chemotherapy and COVID-19 Outcomes in Patients With Cancer

Clinical Characteristics of Coronavirus Disease 2019 in China

Coronavirus Disease 2019 (COVID-19) in Italy

Factors associated with hospital admission and critical illness among

The authors declare the following competing interests: K.B. has received research funding from GRAIL; Y.K is a co-founder in Genome Opinion. M.F.B is on the advisory board for Roche and recieves research support from Illumina. R.L.L. is on the supervisory board of Qiagen and is a scientific advisor to Loxo, Imago, C4 Therapeutics and Isoplexis which include equity interest. He receives research support from and consulted for Celgene and Roche and has consulted for Lilly, Janssen, Astellas, Morphosys and Novartis. He has received honoraria from Roche, Lilly and Amgen for invited lectures and from Gilead for grant reviews. A.Z. received honoraria from Illumina. E.P receives research funding from Celgene.D.G and has received honoraria for speaking and scientific advisory engagements with Celgene, Prime Oncology, Novartis, Illumina and Kyowa Hakko Kirin and is a co-founder in Isabl Technologies. M. Ladanyi serves on the