Browse the corpus

Tumor-Infiltrating Clonal Hematopoiesis. BACKGROUND: Clonal hematopoiesis of indeterminate potential (CHIP) is an age-related condition associated with increased mortality among patients with cancer. CHIP mutations with high variant-allele frequencies can be detected in tumors, a phenomenon we term tumor-infiltrating clonal hematopoiesis (TI-CH). The frequency of TI-CH and its effect on tumor evolution are unclear. METHODS: We characterized CHIP and TI-CH in 421 patients with early-stage non-small-cell lung cancer (NSCLC) from the TRACERx study and in 49,351 patients from the MSK-IMPACT pan-cancer cohort. We studied the association of TI-CH with survival and disease recurrence and evaluated the functional effect of TET2-mutant CHIP on the biologic features of lung tumors. RESULTS: Among patients with NSCLC, 42% of those with CHIP had TI-CH. TI-CH independently predicted an increased risk of death or recurrence, with an adjusted hazard ratio of 1.80 (95% confidence interval [CI], 1.23 to 2.63) as compared with the absence of CHIP and an adjusted hazard ratio of 1.62 (95% CI, 1.02 to 2.56) as compared with CHIP in the absence of TI-CH. Among patients with solid tumors, 26% of those with CHIP had TI-CH. TI-CH conferred a risk of death from any cause that was 1.17 times (95% CI, 1.06 to 1.29) as high as the risk with CHIP in the absence of TI-CH. TET2 mutations were the strongest genetic predictor of TI-CH; such mutations enhanced monocyte migration to lung tumor cells, fueled a myeloid-rich tumor microenvironment in mice, and resulted in the promotion of tumor organoid growth. CONCLUSIONS: TI-CH increased the risk of disease recurrence or death among patients with NSCLC and the risk of death from any cause among patients with solid tumors. TI-CH remodeled the tumor immune microenvironment and accelerated tumor organoid growth, findings that support a role for an aging-related hematologic clonal proliferation in cancer evolution. (Funded by the Royal Society and others.).

Clonal hematopoiesis of indeterminate potential (CHIP) is a prevalent age-associated condition involving the expansion of blood cells derived from a somatically mutated hematopoietic stem cell without hematologic disorders.1,2 CHIP increases the risk of hematological malignancies,3–6 and chronic inflammatory diseases, such as cardiovascular disease, chronic obstructive pulmonary disease, and chronic liver disease.7–11 CHIP is also associated with increased incidence of lung cancer,12,13 and risk of death in solid tumors.14,15 CHIP-derived immune cells have been detected in solid tumor tissues from patients with CHIP,16–18 with an enrichment in non-small cell lung cancer (NSCLC),18 where they have the potential to alter the local microenvironment and influence tumor evolution.19,20 While more than 20% of patients with cancer have CHIP,14,15 the impact of tumor-infiltrating immune cells with CHIP mutations on tumor progression and patient outcomes remains poorly understood. Here, we define the presence of CHIP mutations with high variant allele frequencies within tumors as tumor-infiltrating clonal hematopoiesis (TI-CH). We determine the prevalence of TI-CH in 421 patients with early-stage NSCLC and 49,351 patients across 75 cancer types from a real-world cohort, study its associations with patient outcomes, and evaluate its functional impact using preclinical lung cancer models.

We studied 421 patients with treatment-naïve stage IA-IIIA NSCLC from the TRACERx study (NCT01888601) (Table 1 and Fig S1), including metastases sampled at autopsy from two patients also enrolled in the CRUK autopsy PEACE program (NCT03004755).21,22 Patients were subject to post-surgical multi-region tumor sampling and to blood collection before surgery. A total of 1,560 tumor regions were analyzed (median 3 regions per patient, range 2-10). We included a pan-cancer MSK-IMPACT cohort of 49,351 patients across 75 cancer types comprising primary (n=31,556) and metastatic (n=17,795) tumors with matched blood.23,24 Findings from TRACERx were validated in 2,602 patients with stage I-III NSCLC from MSK-IMPACT (Table 1). CHIP was detected in blood using a 2% variant allele frequency (VAF) cutoff across 77 myeloid driver genes (Table S1), as per prior studies.5,15,25–27 CHIP mutations were genotyped in the matched tumor samples. Spatial immunophenotyping was performed using multiplex imaging mass cytometry for 163 tumor samples from 70 TRACERx patients.28 Single-cell mutational analysis of immune cells was performed for two TRACERx patients harboring TI-CH.29

prior studies.5,15,25–27 CHIP mutations were genotyped in the matched tumor samples. Spatial immunophenotyping was performed using multiplex imaging mass cytometry for 163 tumor samples from 70 TRACERx patients.28 Single-cell mutational analysis of immune cells was performed for two TRACERx patients harboring TI-CH.29 Tet2-mutant CHIP mice were generated by transplanting 1:1 mix of congenically-marked Tet2-mutant and wild-type bone marrow cells into busulfan-conditioned wild-type recipients.30 Lung tumors were induced via orthotopic transplantation of 3LL lung adenocarcinoma (LUAD) cells.31 Tumor, adjacent lungs and blood were analyzed by flow cytometry. For the organoid experiment, TRACERx patient-derived lung tumor organoid cells were cultured with human myeloid cells isolated from lungs of immunodeficient mice engrafted with human TET2-mutant or wild-type hematopoietic stem and progenitor cells.32 Detailed methods are available in the Supplementary Appendix. CHIP and TI-CH presence was associated with clinical parameters, tumor genomic alterations, tumor microenvironment composition, and patient outcomes. Time-to-event analysis was performed using Cox proportional hazards regression controlled for patient age, sex, ethnicity, smoking status, treatment, tumor stage, and histology. Adjustments for multiplicity were not made to 95% confidence intervals (CI), where relevant. When examining the association between multiple cancer types and CHIP or TI-CH, the Benjamini-Hochberg correction for multiple testing was applied.

We studied 421 patients with treatment-naïve stage IA-IIIA NSCLC from the TRACERx study (NCT01888601) (Table 1 and Fig S1), including metastases sampled at autopsy from two patients also enrolled in the CRUK autopsy PEACE program (NCT03004755).21,22 Patients were subject to post-surgical multi-region tumor sampling and to blood collection before surgery. A total of 1,560 tumor regions were analyzed (median 3 regions per patient, range 2-10). We included a pan-cancer MSK-IMPACT cohort of 49,351 patients across 75 cancer types comprising primary (n=31,556) and metastatic (n=17,795) tumors with matched blood.23,24 Findings from TRACERx were validated in 2,602 patients with stage I-III NSCLC from MSK-IMPACT (Table 1).

CHIP was detected in blood using a 2% variant allele frequency (VAF) cutoff across 77 myeloid driver genes (Table S1), as per prior studies.5,15,25–27 CHIP mutations were genotyped in the matched tumor samples. Spatial immunophenotyping was performed using multiplex imaging mass cytometry for 163 tumor samples from 70 TRACERx patients.28 Single-cell mutational analysis of immune cells was performed for two TRACERx patients harboring TI-CH.29

Tet2-mutant CHIP mice were generated by transplanting 1:1 mix of congenically-marked Tet2-mutant and wild-type bone marrow cells into busulfan-conditioned wild-type recipients.30 Lung tumors were induced via orthotopic transplantation of 3LL lung adenocarcinoma (LUAD) cells.31 Tumor, adjacent lungs and blood were analyzed by flow cytometry. For the organoid experiment, TRACERx patient-derived lung tumor organoid cells were cultured with human myeloid cells isolated from lungs of immunodeficient mice engrafted with human TET2-mutant or wild-type hematopoietic stem and progenitor cells.32 Detailed methods are available in the Supplementary Appendix.

CHIP and TI-CH presence was associated with clinical parameters, tumor genomic alterations, tumor microenvironment composition, and patient outcomes. Time-to-event analysis was performed using Cox proportional hazards regression controlled for patient age, sex, ethnicity, smoking status, treatment, tumor stage, and histology. Adjustments for multiplicity were not made to 95% confidence intervals (CI), where relevant. When examining the association between multiple cancer types and CHIP or TI-CH, the Benjamini-Hochberg correction for multiple testing was applied.

The associations between CHIP and risk of mortality have not been evaluated in early-stage treatment-naïve NSCLC or with lung-cancer specific endpoints. In TRACERx, CHIP mutations were observed in 34% (143/421) of patients with early-stage treatment-naïve NSCLC (Table 1 and Fig S1), with a 5% median VAF (interquartile range [IQR] 2.9-11%), most commonly affecting DNMT3A, TET2, and ASXL1. Age was the only baseline variable associated with CHIP in multivariable analysis (Fig S1). CHIP was equally prevalent in LUAD and lung squamous carcinoma (LUSC) (Fig S2), and NSCLC genetic drivers were similar between CHIP-positive and negative patients (Table 1 and Fig S3).21,22 Patients with CHIP had shorter recurrence-free and overall survival compared to patients without CHIP (hazard ratio [HR]=1.42, 95% CI 1.07-1.88 for recurrence-free survival, and HR=1.59, 95% CI 1.18-2.15 for overall survival). The median recurrence-free survival was 2.7 (95% CI 2.0-5.0) and 6.0 (95% CI 3.8-not reached) years in the presence and absence of CHIP, respectively (Fig 1B). For overall survival, it was 4.0 (95% CI 2.8-not reached) and 6.0 (95% CI 5.5-not reached) years, respectively (Fig S4). We assessed the association in multivariable analysis between CHIP and patient outcomes for the risk of lung-cancer related death, all-cause mortality, and tumor recurrence or new primary lung cancer. The adjusted hazard ratios were 1.75 (95% CI 1.17-2.62), 1.59 (95% CI 1.16-2.17), and 1.39 (95% CI 0.99-1.94), respectively (Fig 1C and Fig S5).

on in multivariable analysis between CHIP and patient outcomes for the risk of lung-cancer related death, all-cause mortality, and tumor recurrence or new primary lung cancer. The adjusted hazard ratios were 1.75 (95% CI 1.17-2.62), 1.59 (95% CI 1.16-2.17), and 1.39 (95% CI 0.99-1.94), respectively (Fig 1C and Fig S5). The associations between CHIP and lung-cancer specific outcomes prompted us to evaluate its relationship with the tumor microenvironment. The 46 primary tumor regions from 21 patients with CHIP had a higher representation of myeloid cells compared to the 117 tumor regions from 49 patients without CHIP (Fig 1D and Fig S6; age-adjusted odds ratio [OR]=1.20, 95% CI 1.01-1.43). Among myeloid cell subtypes, monocytes (odds ratio [OR]=1.37, 95% CI 1.18-1.59) and neutrophils (OR=1.31, 95% CI 1.14-1.51) were specifically enriched (Fig S6). There was no difference in the abundance of T and B cells. The enrichment of myeloid cells within primary lung tumors of CHIP-positive patients suggested these cells may originate from CHIP lineages. Thus, for each patient with CHIP, we genotyped the corresponding CHIP mutation(s) in the matched lung tumor samples (179 mutations across 143 patients with CHIP, median 1 mutation per patient, range 1-4) (Fig S1 and Fig S7). In 96% (137/143) of patients with CHIP, the corresponding CHIP mutation was detected in tumors. The VAF of CHIP mutations in tumors had a median of 1.6% (IQR 0.88-3.4%) (Fig 1E), positively correlated with the VAF in blood (Pearson’s r = 0.67, p<0.001, Fig S8), and was consistently lower than the VAF in blood.

The enrichment of myeloid cells within primary lung tumors of CHIP-positive patients suggested these cells may originate from CHIP lineages. Thus, for each patient with CHIP, we genotyped the corresponding CHIP mutation(s) in the matched lung tumor samples (179 mutations across 143 patients with CHIP, median 1 mutation per patient, range 1-4) (Fig S1 and Fig S7). In 96% (137/143) of patients with CHIP, the corresponding CHIP mutation was detected in tumors. The VAF of CHIP mutations in tumors had a median of 1.6% (IQR 0.88-3.4%) (Fig 1E), positively correlated with the VAF in blood (Pearson’s r = 0.67, p<0.001, Fig S8), and was consistently lower than the VAF in blood. To identify tumors with robust infiltration of CHIP-derived immune cells, we defined tumor-infiltrating clonal hematopoiesis, or TI-CH, as the presence of CHIP mutations within tumors at a VAF of 2% or more in at least one region. Among the 143 patients with CHIP, 42% (60/143) had TI-CH (Fig 1E). In LUAD and LUSC, TI-CH was observed in 34% (27/80) and 50% (25/50) of patients with CHIP, respectively. TI-CH was equally prevalent in patients with stage I (42%, 32/77), II (41%, 18/44), or III (45%, 10/22) disease. We assessed TI-CH intratumor heterogeneity by examining 226 tumor regions from the 60 TRACERx patients with TI-CH (median 4 regions per patient, range 2-8) (Fig S9). Among these patients, 83% had TI-CH in at least half of the tumor regions. In 4 patients with synchronous or metachronous primary lung tumors, the TI-CH clone identified in one primary tumor was also detected in the other primary tumor, and 84% (16/19) of tumor regions qualified for TI-CH (Fig S10).

ange 2-8) (Fig S9). Among these patients, 83% had TI-CH in at least half of the tumor regions. In 4 patients with synchronous or metachronous primary lung tumors, the TI-CH clone identified in one primary tumor was also detected in the other primary tumor, and 84% (16/19) of tumor regions qualified for TI-CH (Fig S10). We examined the prevalence of TI-CH in primary tumors from a validation cohort of 2,602 patients with stage I-III NSCLC (Table 1). CHIP mutations were detected in 35% (917/2,602) of patients, with 6.0% (IQR 3.2-13%) median VAF in blood and 1.3% (IQR 0.56-3.0%) in tumors. Among patients with CHIP, 36% (333/917) had TI-CH, a similar proportion to TRACERx. In LUAD and LUSC, TI-CH was observed in 35% (237/685) and 45% (59/132) of patients with CHIP, respectively. The higher prevalence of TI-CH in LUSC compared to LUAD (OR=1.53 95% CI 1.03-2.27, p=0.03) may reflect histology-specific differences in the composition of the tumor immune microenvironment.28 In patients with low or intermediate/high clonal hematopoiesis risk score,6 TI-CH was observed in 32% (128/405) and 45% (70/156) of cases, respectively. Patients with TET2-mutant CHIP had the highest frequency of TI-CH (44%, 63/144), followed by patients with ASXL1-mutant CHIP (41%, 20/49), DNMT3A-mutant CHIP (34%, 161/467), and PPM1D-mutant CHIP (24%, 14/59) (Fig S11).

atopoiesis risk score,6 TI-CH was observed in 32% (128/405) and 45% (70/156) of cases, respectively. Patients with TET2-mutant CHIP had the highest frequency of TI-CH (44%, 63/144), followed by patients with ASXL1-mutant CHIP (41%, 20/49), DNMT3A-mutant CHIP (34%, 161/467), and PPM1D-mutant CHIP (24%, 14/59) (Fig S11). To assess which immune cells harbored CHIP mutations in TI-CH, we performed single-cell genotyping29 on the tumor immune infiltrate from two TRACERx patients with TI-CH that were not part of the discovery cohort (Fig 1F and Fig S12A–B): CRUK1214, harboring mutations in TET2 (VAF of 43% in blood and 26% in tumor) and SRSF2 (VAF of 43% in blood and 20% in tumor), and CRUK1358, with a single TET2 mutation (VAF of 12% in blood and up to 3% in tumor). Despite the high VAFs of CHIP mutations, CRUK1214 did not exhibit monocytosis (Table S2). In both patients, the CHIP mutations were enriched across myeloid populations, including CD206+ macrophages, other mononuclear phagocytes (macrophages, monocytes, dendritic cells), as well as natural killer cells, present in smaller fractions of B cells, and practically absent in T cells (Fig 1F and S12C). We further validated the enrichment of CHIP mutations in tumor-infiltrating myeloid cells using published data from blood and metastatic brain tumor samples from a NSCLC patient with TET2-mutant CHIP (Fig S12D–E).33

ler cells, present in smaller fractions of B cells, and practically absent in T cells (Fig 1F and S12C). We further validated the enrichment of CHIP mutations in tumor-infiltrating myeloid cells using published data from blood and metastatic brain tumor samples from a NSCLC patient with TET2-mutant CHIP (Fig S12D–E).33 We assessed the associations between TI-CH, recurrence-free and overall survival in TRACERx. We defined three categories of patients: (i) without CHIP; (ii) with CHIP in absence of TI-CH (i.e. blood-only CHIP); and (iii) with TI-CH (Table S3). There was no clinically relevant difference in age of patients with blood-only CHIP compared to TI-CH (median age 71 vs. 74 years, which was not significant). Patients with TI-CH had shorter recurrence-free and overall survival compared to patients without CHIP (HR=1.84, 95% CI 1.28-2.65, for recurrence-free survival, and HR=2.01, 95% CI 1.37-2.95, for overall survival). Recurrence-free and overall survival did not differ between patients with blood-only CHIP and patients without CHIP. Across the three categories of patients, the median recurrence-free survival was 6.0 years (95% CI 3.84-not reached), 4.8 years (95% CI 2.1-not reached), and 2.0 years (95% CI 1.2-3.6), respectively (Fig 2A). The median overall survival was 6.0 years (95% CI 5.5-not reached), 4.9 years (95% CI 3.7-not reached), and 3.1 years (95% CI 1.6-not reached), respectively (Fig S13).

rrence-free survival was 6.0 years (95% CI 3.84-not reached), 4.8 years (95% CI 2.1-not reached), and 2.0 years (95% CI 1.2-3.6), respectively (Fig 2A). The median overall survival was 6.0 years (95% CI 5.5-not reached), 4.9 years (95% CI 3.7-not reached), and 3.1 years (95% CI 1.6-not reached), respectively (Fig S13). In a multivariable analysis, TI-CH was an independent predictor of shorter recurrence-free survival (HR=1.80, 95% CI 1.23-2.63) compared to patients without CHIP, while blood-only CHIP was not (HR=1.26, 95% CI 0.88-1.79) (Fig 2B). In the subset of patients with CHIP, the adjusted hazard ratio comparing TI-CH with blood-only CHIP was 1.62 (95% CI 1.02-2.56) (Fig S14A). These associations were independent of NSCLC genetic drivers and correlates of immune infiltration, including tumor purity, tumor-infiltrating neutrophils, and the proportion of myeloid cells (Fig S14B–D).28 We further recapitulated these associations using an alternative metric of TI-CH that estimates the fraction of non-tumor cells harboring CHIP mutations (Supp Methods, Fig S14E). TI-CH was also an independent predictor of the risk of all cause-mortality, lung-cancer related death, and tumor recurrence or new primary lung cancer in TRACERx (Fig S15).

associations using an alternative metric of TI-CH that estimates the fraction of non-tumor cells harboring CHIP mutations (Supp Methods, Fig S14E). TI-CH was also an independent predictor of the risk of all cause-mortality, lung-cancer related death, and tumor recurrence or new primary lung cancer in TRACERx (Fig S15). Given the known association between CHIP VAF and disease outcomes,6,14,15 we further confirmed the prognostic value of TI-CH relative to blood-only CHIP using the ratio of tumor over blood VAF of CHIP mutations. This ratio was an independent predictor of shorter recurrence-free survival (Fig S16), suggesting that the impact of TI-CH on outcomes is not merely due to clonal expansions in the blood, and the degree of CHIP-derived tumor infiltration might influence disease outcomes. In the MSK-IMPACT stage I-III NSCLC validation cohort, the adjusted hazard ratio for the risk of all-cause mortality for patients with TI-CH was 1.35 (95% CI 1.03-1.77) compared to patients without CHIP and was 1.20 (95% CI 0.94-1.52) for patients with blood-only CHIP (Fig S17). We investigated CHIP and TI-CH in 31,556 patients with matched blood and primary tumor samples across 75 cancer types (Table S4 and Fig S18–S19).23,24 CHIP was observed in 24% (7,450/31,556) of patients with a 5.3% median blood VAF (IQR 3.1-11.9). Adjusting for baseline parameters (see Methods), NSCLC was enriched for CHIP compared to other tumor types. In contrast, colorectal cancer and renal cell carcinoma had reduced likelihood of CHIP (Fig 3A).

Fig S18–S19).23,24 CHIP was observed in 24% (7,450/31,556) of patients with a 5.3% median blood VAF (IQR 3.1-11.9). Adjusting for baseline parameters (see Methods), NSCLC was enriched for CHIP compared to other tumor types. In contrast, colorectal cancer and renal cell carcinoma had reduced likelihood of CHIP (Fig 3A). Among patients with CHIP, 26% had TI-CH (1,974/7,450; range 9-42% across cancer types). Adjusting for baseline parameters, tumor purity and blood VAF of CHIP mutations, NSCLC, head and neck cancer, pancreatic cancer, and mesothelioma were enriched for TI-CH compared to other tumor types. Prostate, endometrial, ovarian, and small-cell lung cancers had reduced likelihood of TI-CH (Fig 3A).

ss cancer types). Adjusting for baseline parameters, tumor purity and blood VAF of CHIP mutations, NSCLC, head and neck cancer, pancreatic cancer, and mesothelioma were enriched for TI-CH compared to other tumor types. Prostate, endometrial, ovarian, and small-cell lung cancers had reduced likelihood of TI-CH (Fig 3A). We assessed the associations between TI-CH and outcome in 22,141 patients with available overall survival data. In a multivariable analysis, the risk of all-cause mortality associated with blood-only CHIP was 1.16 (95% CI 1.09-1.23) and for TI-CH was 1.36 (95% CI 1.24-1.48) relative to patients without CHIP (Fig 3B). TI-CH conferred a 1.17 (95% CI 1.06-1.29) greater risk of all-cause mortality relative to blood-only CHIP. In the subset of 14,694 patients with stage I-III disease, the adjusted hazard ratio for the risk of all-cause mortality for patients with TI-CH was 1.44 (95% CI 1.28-1.62) compared to patients without CHIP (Fig S20), and 1.25 (95% CI 1.10-1.42) compared to patients with blood-only CHIP. As in our discovery cohort (Fig S16), the tumor over blood VAF ratio of CHIP mutations was an independent predictor of shorter overall survival (Fig S21), suggesting that these associations were not solely the result of variations in the burden of CHIP in the blood. TI-CH remained an independent predictor for the risk of death after accounting for cases that subsequently transformed with hematologic neoplasms as censoring events (n=119 cases, treated as censoring events).

uggesting that these associations were not solely the result of variations in the burden of CHIP in the blood. TI-CH remained an independent predictor for the risk of death after accounting for cases that subsequently transformed with hematologic neoplasms as censoring events (n=119 cases, treated as censoring events). TI-CH was prevalent in metastatic samples. CHIP was observed in 26% (4,686/17,795) of patients with matched blood and metastatic tumors sampled from various organ sites, of which 28% (1,304/4,686) had TI-CH in the metastatic samples. For CHIP-positive patients with NSCLC, hepatobiliary cancer, and renal cell carcinoma, more than 35% of metastatic samples analyzed exhibited TI-CH (Fig S22). In 54 primary tumors with TI-CH and their paired progression metastases in MSK-IMPACT (Fig S23A), 81% (44/54) of metastases exhibited TI-CH. The median CHIP mutation VAF was higher in metastases compared to primaries (6.1% vs. 4.3%). We further tracked TI-CH in 49 metastases obtained at progression or autopsy from two TRACERx patients with TI-CH-positive primary tumors enrolled in PEACE (Fig S23B), the first with DNMT3A and TET2 mutations, and the second with a DNMT3A mutation. In both patients, the mutations were detected (without a VAF threshold) in all metastases, except for two located in the brain. In the first patient, 81% (18/22) of metastases qualified for TI-CH. In the second patient, all 9 liver metastases qualified for TI-CH, compared to 17% (3/18) of metastases at other organ sites. Collectively, these findings indicate that TI-CH recurs in metastases.

n all metastases, except for two located in the brain. In the first patient, 81% (18/22) of metastases qualified for TI-CH. In the second patient, all 9 liver metastases qualified for TI-CH, compared to 17% (3/18) of metastases at other organ sites. Collectively, these findings indicate that TI-CH recurs in metastases. Similar to the NSCLC validation cohort (Fig S11), patients with TET2-mutant CHIP from the pan-cancer cohort had the highest frequency of TI-CH (33%, 381/1191), followed by ASXL1-mutant (32%, 124/392), DNMT3A-mutant (25%, 926/3,753), and PPM1D-mutant CHIP (13%, 83/622) (Fig S24). TET2 mutations had the largest relative tumor-infiltrating clone size compared to other CHIP gene mutations (median tumor over blood VAF ratio of 0.18 versus 0.13, p<0.001) (Fig S24). To assess whether distinct CHIP driver genes were differentially associated with TI-CH, we performed a multivariable analysis adjusted for baseline parameters, tumor purity and blood VAF of CHIP mutations (Fig S24D). TET2 mutations were an independent positive predictor of TI-CH compared to other CHIP mutations (adjusted OR=1.78, 95% CI 1.39-2.27), while the opposite was observed for PPM1D mutations (adjusted OR=0.43, 95% CI 0.29-0.63) (Fig 4A). Determining genotype-specific risks associated with TI-CH for clinical outcomes will require larger cohorts.

n independent positive predictor of TI-CH compared to other CHIP mutations (adjusted OR=1.78, 95% CI 1.39-2.27), while the opposite was observed for PPM1D mutations (adjusted OR=0.43, 95% CI 0.29-0.63) (Fig 4A). Determining genotype-specific risks associated with TI-CH for clinical outcomes will require larger cohorts. Given the preponderance of TET2-mutant cells to infiltrate tumors (Fig 4A), we explored the functional link between TET2-mutant CHIP and lung TI-CH using mouse models. Tet2-mutant CHIP mice were generated by transplanting a 1:1 mix of congenically-marked Tet2-mutant and wild-type bone marrow cells into busulfan-conditioned wild-type recipients (Fig S25A).30 Consistent with prior studies,34 Tet2-mutant cells expanded significantly more in blood relative to wild-type counterparts (Fig S25B–C). We induced lung tumors in Tet2-mutant CHIP mice by orthotopically transplanting KrasG12C 3LL LUAD cells, and analyzed blood, tumor and adjacent lung tissues (Fig S26A–C, S27A).31 Similar to our human results (Fig 1E), in mouse tumors Tet2-mutant cells were enriched within the myeloid and natural killer cell compartments relative to other cell populations (Fig S27B).

by orthotopically transplanting KrasG12C 3LL LUAD cells, and analyzed blood, tumor and adjacent lung tissues (Fig S26A–C, S27A).31 Similar to our human results (Fig 1E), in mouse tumors Tet2-mutant cells were enriched within the myeloid and natural killer cell compartments relative to other cell populations (Fig S27B). Consistent with the clinical association between CHIP and a myeloid-rich lung tumor microenvironment (Fig 1D), the percentage of Tet2-mutant cells in blood positively correlated with the overall myeloid infiltration in the tumor (Fig 4B). The correlation was driven by Tet2-mutant tumor-infiltrating myeloid cells (Fig 4C), including mononuclear phagocytes (monocytes, macrophages and dendritic cells) and granulocytes (neutrophils and eosinophils) (Fig S27C–D), although the former were affected more profoundly by blood CHIP levels (Fig S27E). In a functional migration assay,35 blood-derived Tet2-mutant monocytes migrated to 3LL LUAD cells significantly more than their wild-type counterparts (Fig S27F, 4D). In contrast, neutrophil migration was not affected by Tet2 status (Fig S27G). In vivo quantification of Tet2-mutant versus wild-type myeloid cell subsets revealed that Tet2-mutant CD11b+ monocyte-derived macrophages preferentially accumulated in the tumor relative to adjacent normal lung (Fig S27H), an effect that was not observed for their monocyte precursors or neutrophils (Fig S27H–I). These data suggest that Tet2-mutant monocytes preferentially migrate towards lung cancer cells and accumulate as CD11b+ monocyte-derived macrophages in the tumor microenvironment.

umor relative to adjacent normal lung (Fig S27H), an effect that was not observed for their monocyte precursors or neutrophils (Fig S27H–I). These data suggest that Tet2-mutant monocytes preferentially migrate towards lung cancer cells and accumulate as CD11b+ monocyte-derived macrophages in the tumor microenvironment. To evaluate whether TET2-mutant myeloid cells influence lung tumor cell growth, we cultured TRACERx patient-derived LUAD organoids36 with human TET2-mutant or wild-type myeloid cells from lungs of humanized mice engrafted with TET2-mutant or wild-type hematopoietic stem cells (Fig S28A–B).32 Co-cultures of tumor organoid cells with TET2-mutant myeloid cells resulted in larger and more numerous organoids compared to co-cultures with wild-type myeloid cells (Fig 4E–F). Thus, TET2 mutations in myeloid cells can functionally impact tumor cell clonogenicity and growth.

The associations between CHIP and lung-cancer specific outcomes prompted us to evaluate its relationship with the tumor microenvironment. The 46 primary tumor regions from 21 patients with CHIP had a higher representation of myeloid cells compared to the 117 tumor regions from 49 patients without CHIP (Fig 1D and Fig S6; age-adjusted odds ratio [OR]=1.20, 95% CI 1.01-1.43). Among myeloid cell subtypes, monocytes (odds ratio [OR]=1.37, 95% CI 1.18-1.59) and neutrophils (OR=1.31, 95% CI 1.14-1.51) were specifically enriched (Fig S6). There was no difference in the abundance of T and B cells.

The enrichment of myeloid cells within primary lung tumors of CHIP-positive patients suggested these cells may originate from CHIP lineages. Thus, for each patient with CHIP, we genotyped the corresponding CHIP mutation(s) in the matched lung tumor samples (179 mutations across 143 patients with CHIP, median 1 mutation per patient, range 1-4) (Fig S1 and Fig S7). In 96% (137/143) of patients with CHIP, the corresponding CHIP mutation was detected in tumors. The VAF of CHIP mutations in tumors had a median of 1.6% (IQR 0.88-3.4%) (Fig 1E), positively correlated with the VAF in blood (Pearson’s r = 0.67, p<0.001, Fig S8), and was consistently lower than the VAF in blood.

We assessed the associations between TI-CH, recurrence-free and overall survival in TRACERx. We defined three categories of patients: (i) without CHIP; (ii) with CHIP in absence of TI-CH (i.e. blood-only CHIP); and (iii) with TI-CH (Table S3). There was no clinically relevant difference in age of patients with blood-only CHIP compared to TI-CH (median age 71 vs. 74 years, which was not significant). Patients with TI-CH had shorter recurrence-free and overall survival compared to patients without CHIP (HR=1.84, 95% CI 1.28-2.65, for recurrence-free survival, and HR=2.01, 95% CI 1.37-2.95, for overall survival). Recurrence-free and overall survival did not differ between patients with blood-only CHIP and patients without CHIP. Across the three categories of patients, the median recurrence-free survival was 6.0 years (95% CI 3.84-not reached), 4.8 years (95% CI 2.1-not reached), and 2.0 years (95% CI 1.2-3.6), respectively (Fig 2A). The median overall survival was 6.0 years (95% CI 5.5-not reached), 4.9 years (95% CI 3.7-not reached), and 3.1 years (95% CI 1.6-not reached), respectively (Fig S13).

associations using an alternative metric of TI-CH that estimates the fraction of non-tumor cells harboring CHIP mutations (Supp Methods, Fig S14E). TI-CH was also an independent predictor of the risk of all cause-mortality, lung-cancer related death, and tumor recurrence or new primary lung cancer in TRACERx (Fig S15). Given the known association between CHIP VAF and disease outcomes,6,14,15 we further confirmed the prognostic value of TI-CH relative to blood-only CHIP using the ratio of tumor over blood VAF of CHIP mutations. This ratio was an independent predictor of shorter recurrence-free survival (Fig S16), suggesting that the impact of TI-CH on outcomes is not merely due to clonal expansions in the blood, and the degree of CHIP-derived tumor infiltration might influence disease outcomes. In the MSK-IMPACT stage I-III NSCLC validation cohort, the adjusted hazard ratio for the risk of all-cause mortality for patients with TI-CH was 1.35 (95% CI 1.03-1.77) compared to patients without CHIP and was 1.20 (95% CI 0.94-1.52) for patients with blood-only CHIP (Fig S17).

We investigated CHIP and TI-CH in 31,556 patients with matched blood and primary tumor samples across 75 cancer types (Table S4 and Fig S18–S19).23,24 CHIP was observed in 24% (7,450/31,556) of patients with a 5.3% median blood VAF (IQR 3.1-11.9). Adjusting for baseline parameters (see Methods), NSCLC was enriched for CHIP compared to other tumor types. In contrast, colorectal cancer and renal cell carcinoma had reduced likelihood of CHIP (Fig 3A). Among patients with CHIP, 26% had TI-CH (1,974/7,450; range 9-42% across cancer types). Adjusting for baseline parameters, tumor purity and blood VAF of CHIP mutations, NSCLC, head and neck cancer, pancreatic cancer, and mesothelioma were enriched for TI-CH compared to other tumor types. Prostate, endometrial, ovarian, and small-cell lung cancers had reduced likelihood of TI-CH (Fig 3A).

Similar to the NSCLC validation cohort (Fig S11), patients with TET2-mutant CHIP from the pan-cancer cohort had the highest frequency of TI-CH (33%, 381/1191), followed by ASXL1-mutant (32%, 124/392), DNMT3A-mutant (25%, 926/3,753), and PPM1D-mutant CHIP (13%, 83/622) (Fig S24). TET2 mutations had the largest relative tumor-infiltrating clone size compared to other CHIP gene mutations (median tumor over blood VAF ratio of 0.18 versus 0.13, p<0.001) (Fig S24). To assess whether distinct CHIP driver genes were differentially associated with TI-CH, we performed a multivariable analysis adjusted for baseline parameters, tumor purity and blood VAF of CHIP mutations (Fig S24D). TET2 mutations were an independent positive predictor of TI-CH compared to other CHIP mutations (adjusted OR=1.78, 95% CI 1.39-2.27), while the opposite was observed for PPM1D mutations (adjusted OR=0.43, 95% CI 0.29-0.63) (Fig 4A). Determining genotype-specific risks associated with TI-CH for clinical outcomes will require larger cohorts.

Given the preponderance of TET2-mutant cells to infiltrate tumors (Fig 4A), we explored the functional link between TET2-mutant CHIP and lung TI-CH using mouse models. Tet2-mutant CHIP mice were generated by transplanting a 1:1 mix of congenically-marked Tet2-mutant and wild-type bone marrow cells into busulfan-conditioned wild-type recipients (Fig S25A).30 Consistent with prior studies,34 Tet2-mutant cells expanded significantly more in blood relative to wild-type counterparts (Fig S25B–C). We induced lung tumors in Tet2-mutant CHIP mice by orthotopically transplanting KrasG12C 3LL LUAD cells, and analyzed blood, tumor and adjacent lung tissues (Fig S26A–C, S27A).31 Similar to our human results (Fig 1E), in mouse tumors Tet2-mutant cells were enriched within the myeloid and natural killer cell compartments relative to other cell populations (Fig S27B). Consistent with the clinical association between CHIP and a myeloid-rich lung tumor microenvironment (Fig 1D), the percentage of Tet2-mutant cells in blood positively correlated with the overall myeloid infiltration in the tumor (Fig 4B). The correlation was driven by Tet2-mutant tumor-infiltrating myeloid cells (Fig 4C), including mononuclear phagocytes (monocytes, macrophages and dendritic cells) and granulocytes (neutrophils and eosinophils) (Fig S27C–D), although the former were affected more profoundly by blood CHIP levels (Fig S27E).

mor (Fig 4B). The correlation was driven by Tet2-mutant tumor-infiltrating myeloid cells (Fig 4C), including mononuclear phagocytes (monocytes, macrophages and dendritic cells) and granulocytes (neutrophils and eosinophils) (Fig S27C–D), although the former were affected more profoundly by blood CHIP levels (Fig S27E). In a functional migration assay,35 blood-derived Tet2-mutant monocytes migrated to 3LL LUAD cells significantly more than their wild-type counterparts (Fig S27F, 4D). In contrast, neutrophil migration was not affected by Tet2 status (Fig S27G). In vivo quantification of Tet2-mutant versus wild-type myeloid cell subsets revealed that Tet2-mutant CD11b+ monocyte-derived macrophages preferentially accumulated in the tumor relative to adjacent normal lung (Fig S27H), an effect that was not observed for their monocyte precursors or neutrophils (Fig S27H–I). These data suggest that Tet2-mutant monocytes preferentially migrate towards lung cancer cells and accumulate as CD11b+ monocyte-derived macrophages in the tumor microenvironment.

To evaluate whether TET2-mutant myeloid cells influence lung tumor cell growth, we cultured TRACERx patient-derived LUAD organoids36 with human TET2-mutant or wild-type myeloid cells from lungs of humanized mice engrafted with TET2-mutant or wild-type hematopoietic stem cells (Fig S28A–B).32 Co-cultures of tumor organoid cells with TET2-mutant myeloid cells resulted in larger and more numerous organoids compared to co-cultures with wild-type myeloid cells (Fig 4E–F). Thus, TET2 mutations in myeloid cells can functionally impact tumor cell clonogenicity and growth.

Cancer evolution, once regarded as a step-wise accumulation of oncogenic mutations, is now seen as a multifactorial process impacted by the tissue microenvironment, local and systemic immune responses, environmental exposures and aging.37 Recent studies have highlighted the influence of the aging hematopoietic system on lung cancer development.38 Here, age-associated tumor-infiltrating clonal hematopoiesis, or TI-CH, emerges as an important facet influencing cancer progression. TI-CH is a pan-cancer phenomenon, observed in 25% of patients with CHIP, representing over 6% of all solid tumor cases, and associated with adverse patient outcomes. The likelihood of having TI-CH varies across cancer types, with enrichment in NSCLC,18 head and neck cancer, pancreatic cancer, and mesothelioma. In NSCLC, approximately 40% of patients with CHIP had TI-CH, which independently increased their risk of death or recurrence. Across CHIP mutations, TET2 emerged as an independent predictor of TI-CH. In a NSCLC model, Tet2-mutant CHIP caused increased myeloid cell infiltration into tumors, mimicking observations in human disease. Tet2-mutant monocytes preferentially migrated towards tumor cells and accumulated as macrophages within murine tumors. Consistent with published evidence that TET2 mutations can functionally remodel myeloid responses,39 TET2-mutant myeloid cells displayed enhanced protumorigenic capacity in tumor organoid co-cultures compared to wild-type myeloid cells.

y migrated towards tumor cells and accumulated as macrophages within murine tumors. Consistent with published evidence that TET2 mutations can functionally remodel myeloid responses,39 TET2-mutant myeloid cells displayed enhanced protumorigenic capacity in tumor organoid co-cultures compared to wild-type myeloid cells. The associations between TI-CH and adverse outcomes, the increased odds of TI-CH with TET2-mutant CHIP, and the impact of TET2-mutant immune cells on the tumor microenvironment and cancer cell growth, support the role of TI-CH in tumor evolution and its potential utility in cancer diagnostics.18 Future studies should evaluate these findings in larger and more ethnically diverse cancer cohorts and further interrogate the functional impact of TI-CH on tumor progression.

e tumor microenvironment and cancer cell growth, support the role of TI-CH in tumor evolution and its potential utility in cancer diagnostics.18 Future studies should evaluate these findings in larger and more ethnically diverse cancer cohorts and further interrogate the functional impact of TI-CH on tumor progression. CHIP is linked to cardiovascular disease, where mutant myeloid cells produce proinflammatory cytokines, accelerating atherosclerosis.7,40–42 Tet2-mutant macrophages were shown to secrete higher levels of IL-1β, and IL-1β inhibition reduced the risk of cardiovascular complications.7,40–42 Interestingly, anti-IL1β therapy reduces the incidence of lung cancer,43 particularly in patients with TET2-mutant CHIP.44 However, clinical trials have shown that anti-IL-1β therapy is ineffective in established NSCLC.45 Several other cytokines upregulated in Tet2-mutant macrophages, including IL-6, PF4, CXCL1, CXCL2 and OSM,7,40 are implicated in promoting lung tumor progression in model systems.46–49 Understanding the impact of TI-CH on myeloid-cancer cell communication in solid cancer initiation and growth may be crucial for therapeutic interception.

kines upregulated in Tet2-mutant macrophages, including IL-6, PF4, CXCL1, CXCL2 and OSM,7,40 are implicated in promoting lung tumor progression in model systems.46–49 Understanding the impact of TI-CH on myeloid-cancer cell communication in solid cancer initiation and growth may be crucial for therapeutic interception. The ability of age-associated somatic clonal expansions in one tissue, such as the hematopoietic compartment, to regulate malignant progression in another tissue represents a new perspective in cancer biology. Given the pervasiveness of somatic mosaicism in humans,50 more examples of the interplay between non-malignant somatic clones, aging, and human disease may emerge, opening a role for prevention strategies attenuating somatic clonal proliferation and tissue inflammation.