Purpose: Plasma cell-free DNA (cfDNA) genotyping is increasingly used in cancer care, but assay accuracy has been debated. Because most cfDNA is derived from peripheral blood cells (PBC), we hypothesized that nonmalignant mutations harbored by hematopoietic cells (clonal hematopoiesis, CH) could be a cause of false-positive plasma genotyping.

Experimental Design: We identified patients with advanced non–small cell lung cancer (NSCLC) with KRAS, JAK2, or TP53 mutations identified in cfDNA. With consent, PBC DNA was tested using droplet digital PCR (ddPCR) or next-generation sequencing (NGS) to test for CH-derived mutations.

Results: We first studied plasma ddPCR results from 58 patients with EGFR-mutant NSCLC. Two had KRAS G12X detected in cfDNA, and both were present in PBC, including one where the KRAS mutation was detected serially for 20 months. We then studied 143 plasma NGS results from 122 patients with NSCLC and identified 5 JAK2 V617F mutations derived from PBC. In addition, 108 TP53 mutations were detected in cfDNA; for 33 of the TP53 mutations, PBC and tumor NGS were available for comparison, and 5 were present in PBC but absent in tumor, consistent with CH.

Conclusions: We find that most JAK2 mutations, some TP53 mutations, and rare KRAS mutations detected in cfDNA are derived from CH not tumor. Clinicians ordering plasma genotyping must be prepared for the possibility that mutations detected in plasma, particularly in genes mutated in CH, may not represent true tumor genotype. Efforts to use plasma genotyping for cancer detection may need paired PBC genotyping so that CH-derived mutations are not misdiagnosed as occult malignancy. Clin Cancer Res; 24(18); 4437–43. ©2018 AACR.

See related commentary by Bauml and Levy, p. 4352

This article is featured in Highlights of This Issue, p. 4349

Translational Relevance

Despite increasing use of plasma cell-free DNA (cfDNA) genotyping for cancer care, this biospecimen remains poorly understood. Here, we report recurrent mutations in JAK2, TP53, and KRAS detected in cfDNA on validated plasma genotyping assays that are in fact derived from white blood cells (i.e., clonal hematopoiesis, CH) rather than from tumor. Our approach highlights that clinicians must be aware that mutations detected in cfDNA, especially in CH-related genes, may not always reflect tumor genotype. Furthermore, future efforts to develop plasma cfDNA genotyping as a tool for cancer detection may require paired genotyping of peripheral blood cells so that CH-derived mutations are not misdiagnosed as occult malignancy.

Genomic analysis of plasma cell-free DNA (cfDNA) is increasingly clinically adopted as a method for noninvasive genotyping of advanced cancers. Such “liquid biopsy” approaches potentially allow cancer genotyping in the absence of adequate tumor tissue, resulting in the FDA approval of the cobas EGFR mutation test v2 for plasma genotyping of non–small cell lung cancer (NSCLC) when tumor tissue is unavailable (1). More recent advances in the development of highly sensitive plasma next-generation sequencing (NGS) assays have motivated ambitious plans to expand the clinical use of plasma genotyping (2), to develop assays that can facilitate early detection of residual disease following curative therapy (3), and to eventually develop cancer screening tests.

Unfortunately, discordance between genotyping of tumor tissue and plasma cfDNA has hampered the use and interpretation of these new assays (4). One recognized cause of tumor/plasma discordance is variable shed of tumor DNA into the plasma, causing false-negative plasma genotyping (5). A second well-documented cause of tumor/plasma discordance is tumor heterogeneity, particularly in the setting of drug resistance (6). However, variable DNA shed and tumor heterogeneity fail to fully account for the recurring reports of inaccurate plasma genotyping reported in the literature (7, 8).

Because a large proportion of cfDNA is derived from peripheral blood cells (PBC; ref. 9), we hypothesized that somatic mutations within nonmalignant hematopoietic cells, known as clonal hematopoiesis (CH; ref. 10), might be a recurring source of discordance between tumor genotyping and plasma cfDNA genotyping. Our specific aim was to identify false positives due to CH, that is, instances where plasma genotyping, used with the intent of identifying NSCLC genotype, instead identified mutations derived from CH and not the tumor.

Six large publications studying CH were reviewed to identify genes of interest (11–16). Per patient prevalence estimates of mutations in relevant CH genes were calculated by review of publicly available data (Fig. 1). The genes most commonly mutated (DNMT3A, ASXL1, and TET2; refs. 11, 12) are not covered by most clinically available plasma genotyping assays. However, several other genes with potential importance to cancer biology have been reported as occasionally mutated in CH: JAK2, TP53, GNAS, IDH2, and KRAS. For this analysis, we focused on mutations in three cancer-associated genes that could be genotyped using available highly sensitive assays: JAK2, TP53, and KRAS.

Figure 1.

Genes commonly mutated in CH. Review of six publications (11–16) studying CH reveals three commonly mutated genes: DNMT3A, TET2, and ASXL1 (left). In addition, several cancer-related genes (right) are less commonly mutated in CH but are covered by many commercial plasma genotyping assays. Genes not reported in a given publication are marked with an asterisk.

Figure 1.

Genes commonly mutated in CH. Review of six publications (11–16) studying CH reveals three commonly mutated genes: DNMT3A, TET2, and ASXL1 (left). In addition, several cancer-related genes (right) are less commonly mutated in CH but are covered by many commercial plasma genotyping assays. Genes not reported in a given publication are marked with an asterisk.

Close modal

Patient cohort

Patients with advanced NSCLC who had undergone plasma cfDNA genotyping as part of ongoing research or routine clinical care were identified. Plasma cfDNA genotyping was performed either using a validated droplet digital PCR (ddPCR) assay (performed in a research lab and described previously; ref. 5) or using a commercially available plasma NGS assay (Guardant Health). Under an institutional review board–approved protocol, patients were consented for blood collection so that PBC, which are usually discarded after spinning the plasma, could be saved for further genomic analysis.

ddPCR

DNA was extracted from PBC using the DNeasy blood and tissue kit (QIAGEN) per the manufacturer's recommendation. DNA was tested for recurring point mutations (e.g., KRAS G12X and JAK2 V617F) using ddPCR. Primer/probes for the JAK2 V617F assay were custom made by Life Technologies.

  • – Forward primer sequence: 5′-AAGCTTTCTCACAAGCATTTGGTTT-3′.

  • – Reverse primer sequence: 5′-GAAAGCCTGTAGTTTTACTTACTCTCGT-3′.

  • – Probe sequences: 5′-VIC-CTCCACAGACACATACT-MGB-NFQ-3′, 5′-FAM-CTCCACAGAAACATACT-MGB-NFQ-3′.

  • – Cycling conditions: 10 minutes at 95°C followed by 40 cycles of a two-step thermal profile of 15 seconds at 94°C denaturation and 60 seconds at 58°C.

Highly sensitive NGS

NGS of PBC DNA was performed using a custom QIAseq V3 amplicon-based panel containing PCR amplicons spanning the entire exome of TP53 and KRAS, and sequenced on an Illumina NextSeq at an average mean read depth of 1,200. DNA libraries were prepared using a custom QIAseq V3 targeted DNA kit principally, as per the manufacturer's instructions. Note that 15 ng of isolated DNA was enzymatically fragmented and end repaired in a 25-μL reaction. The reaction conditions were as follows: 4°C for 1 minute, 32°C for 14 minutes, and 65°C for 30 minutes. Immediately after, adapter ligation reagents were directly added to the 25 μL of DNA: 10 μL 5X ligation buffer, 5 μL DNA ligase, 2.8 μL IL-N7## barcoded adapters, and DNAse-free water were added to a total of 50 μL. The reaction continued at 20°C for an additional 15 minutes, after which the reaction was purified using 90 μL of QIAseq beads and eluted into 12 μL of nuclease-free water. 9.4 μL of DNA was mixed with 5 μL of the QIAseq targeted panel, 4 μL of 5X TEPCR buffer, 0.8 μL of IL-forward primer, and 0.8 μL of HotStarTaq DNA polymerase. The PCR enrichment condition was 95°C for 13 minutes, 98°C for 2 minutes, six cycles of 98°C for 15 seconds and 65°C for 15 minutes, and 72°C for 5 minutes. Each reaction was cleaned once using QIAseq beads to remove unused primers. Enriched DNA was combined with 400 nmol/L of IL-universal primer, 400 nmol/L of IL-index primer, 1X of UPCR buffer, and 1 μL of HotStarTaq DNA polymerase in a volume of 20 μL. The universal PCR condition was 95°C for 13 minutes, 98°C for 2 minutes, 20 cycles of 98°C for 15 seconds and 60°C for 2 minutes, and 72°C for 5 minutes. The DNA library was purified and sequenced on Illumina NextSeq (paired-end, 2 × 151 bp) following the manufacturer's user manual (Illumina).

Tumor genotyping

Tumor NGS was performed at the Center for Advanced Molecular Diagnostics of Brigham and Women's Hospital using a hybrid-capture–based NGS platform spanning 275 to 447 genes. An institutional database of NSCLC tumor NGS results was used for comparisons (17).

False-positive plasma ddPCR due to CH

We first queried our multiyear institutional experience performing plasma genotyping of key EGFR and KRAS mutations in patients with advanced NSCLC using a recently validated ddPCR platform with a negligible false-positive rate (5). Reviewing a cohort of 221 patients with advanced NSCLC undergoing plasma ddPCR for EGFR and KRAS mutations, 58 were known to harbor an EGFR mutation based on tumor genotyping (Fig. 2A). Two of these cases (3%) had plasma ddPCR that was positive for a KRAS codon 12 mutation. PBC DNA was collected and tested and determined to be the source of the KRAS mutation in both cases.

Figure 2.

Testing for CH using ddPCR of PBC from patients with advanced NSCLC. A, 221 patients completed plasma ddPCR for EGFR and KRAS driver mutations. Two patients with EGFR mutations on tumor NGS had KRAS G12X mutations detected in plasma. PBC for both were positive for the KRAS mutation. B, One of these patients had advanced EGFR-mutant lung cancer with a durable response to erlotinib therapy (arrow). C, Serial ddPCR of plasma cfDNA reveals clearance of the EGFR exon 19 deletion but sustained low levels of a KRAS G12D mutation over 20 months on therapy. D, Genotyping of DNA from the PBC identifies the presence of the KRAS G12D mutation, and further fluorescence-activated cell sorting (FACS) determined the mutation to be present in lymphocyte subpopulations, but not in monocytes or granulocytes. AF, allelic fraction; N/D, not detected.

Figure 2.

Testing for CH using ddPCR of PBC from patients with advanced NSCLC. A, 221 patients completed plasma ddPCR for EGFR and KRAS driver mutations. Two patients with EGFR mutations on tumor NGS had KRAS G12X mutations detected in plasma. PBC for both were positive for the KRAS mutation. B, One of these patients had advanced EGFR-mutant lung cancer with a durable response to erlotinib therapy (arrow). C, Serial ddPCR of plasma cfDNA reveals clearance of the EGFR exon 19 deletion but sustained low levels of a KRAS G12D mutation over 20 months on therapy. D, Genotyping of DNA from the PBC identifies the presence of the KRAS G12D mutation, and further fluorescence-activated cell sorting (FACS) determined the mutation to be present in lymphocyte subpopulations, but not in monocytes or granulocytes. AF, allelic fraction; N/D, not detected.

Close modal

The first case was a 78-year-old female with newly diagnosed advanced EGFR-mutant NSCLC whose initial plasma ddPCR revealed both an EGFR exon 19 deletion (2.2% allelic fraction, AF) and a KRAS G12D (2.0% AF) mutation. Treatment with erlotinib resulted in a durable response on imaging and clearance of the plasma EGFR mutation; however, the KRAS mutation persisted in cfDNA over the course of 20 months on therapy (Fig. 2B and C). Occult malignancy was suspected, but imaging with abdominal ultrasound and CT colonography revealed no occult malignancy. Noting that KRAS mutations have been reported in rare patients with CH (12, 14), the KRAS G12D mutation was retested in paired cfDNA and PBC and detected in plasma at 1.2% AF and PBC at 0.5% AF, whereas the EGFR exon 19 deletion was undetectable in PBC (Supplementary Fig. S1). To pinpoint the origin of this KRAS mutation, PBC were sorted via fluorescence-activated cell sorting (Fig. 2D); ddPCR detected the KRAS mutation only in lymphocytes (1.7% AF) and across four lymphoid subpopulations. NGS confirmed the presence of the KRAS G12D mutation, as detected with ddPCR, but the absence of any other mutations in 8 cancer-associated genes (Supplementary Fig. S1). These data suggest that a common lymphoid progenitor acquired a KRAS G12D mutation and underwent clonal expansion, consistent with CH (5).

The second case was a 61-year-old female with advanced EGFR-mutant NSCLC being treated with erlotinib. Prior tumor NGS of a resection specimen had detected an EGFR L858R mutation but no KRAS mutation. Upon acquired resistance to erlotinib, commercial plasma NGS did not detect the EGFR L858R mutation but did detect a KRAS G12S mutation; the KRAS G12 mutation was then confirmed at our institution using ddPCR at 0.38% AF. We performed ddPCR of DNA from PBC, which revealed a KRAS G12 mutation (0.41% AF) and no detectable EGFR L858R. NGS of a tumor biopsy performed following erlotinib resistance detected the EGFR L858R mutation and a new T790M mutation but no KRAS mutation.

False-positive plasma NGS due to CH

We then studied 143 commercial plasma NGS results from 122 patients with advanced NSCLC treated at our center (Fig. 3A). Fourteen mutations were detected in three genes associated with CH (JAK2, GNAS, IDH2), including 6 cases (4.9%) positive for a JAK2 V617F mutation, which is rarely seen on tumor genotyping of NSCLC (0.26%, P = 0.0001). PBC were available for analysis in 5 cases, and ddPCR detected the JAK2 V617F mutation in each (AF 0.13%–4.66%, Table 1; Supplementary Fig. S2); 3 had tumor NGS available, and all were JAK2 wild type.

Figure 3.

Testing for CH through NGS of PBC from patients with advanced NSCLC. A, One hundred twenty-two patients had plasma NGS, revealing 14 mutations in CH-related genes and 108 mutations in TP53. PBC were available for analysis for 5 JAK2 mutations and 33 TP53 mutations (see Table 1). B, Comparing the number of coding TP53 mutations across 80 plasma specimens and 1,245 tumor specimens with a TP53 mutation detected using NGS, more TP53 variants were detected per sample in plasma as compared with tumor (P < 0.0001, χ2 test), with 0.5% of tumor samples showing >2 mutations compared with 8.8% plasma samples. C, Three cases had multiple TP53 mutations found on plasma NGS, shown in relationship to the known oncogenic driver mutations (black), where some TP53 mutations were derived from tumor (green) and some were derived from CH (orange).

Figure 3.

Testing for CH through NGS of PBC from patients with advanced NSCLC. A, One hundred twenty-two patients had plasma NGS, revealing 14 mutations in CH-related genes and 108 mutations in TP53. PBC were available for analysis for 5 JAK2 mutations and 33 TP53 mutations (see Table 1). B, Comparing the number of coding TP53 mutations across 80 plasma specimens and 1,245 tumor specimens with a TP53 mutation detected using NGS, more TP53 variants were detected per sample in plasma as compared with tumor (P < 0.0001, χ2 test), with 0.5% of tumor samples showing >2 mutations compared with 8.8% plasma samples. C, Three cases had multiple TP53 mutations found on plasma NGS, shown in relationship to the known oncogenic driver mutations (black), where some TP53 mutations were derived from tumor (green) and some were derived from CH (orange).

Close modal
Table 1.

Genotyping of peripheral blood cells identifies clonal hematopoiesis in patients with plasma genotyping positive for JAK2 mutations (top) or TP53 mutations (bottom)

IDAgePlasma NGS result (AF %)PBC testing methodTumor NGS resultPBC testing result (AF %)Mutation source
JAK2 analysis 
4433 58 JAK2 V617F (19.5) ddPCR Not detected Detected (4.66) CH 
4379 63 JAK2 V617F (2.5) ddPCR Not available Detected (0.19) CH 
4014 59 JAK2 V617F (0.5) ddPCR Not detected Detected (0.68) CH 
349 84 JAK2 V617F (0.3) ddPCR Not detected Detected (0.14) CH 
4447 87 JAK2 V617F (0.2) ddPCR Not available Detected (0.13) CH 
TP53 analysis 
426 64 TP53 Y220C (1.9) NGS Not detected Detected (1.72) CH 
4188 73 TP53 M237I (1.2) NGS Not detected Detected (2.66) CH 
3557 70 TP53 H193P (0.7) NGS Not detected Detected (1.1) CH 
3717 73 TP53 V272M (0.5) NGS Not detected Detected (1.29) CH 
423 67 TP53 R110L (0.3) NGS Not detected Detected (0.56) CH 
427 47 TP53 G244C (11.2) NGS Detected Not detected Tumor 
424 49 TP53 V157F (8.5) NGS Detected Not detected Tumor 
3856 43 TP53 R280I (5.9) NGS Detected Not detected Tumor 
4188 73 TP53 Y220C (5.6) NGS Detected Not detected Tumor 
4250 59 TP53 R248Q (5.1) NGS Detected Not detected Tumor 
3983 47 TP53 R175H (4.6) NGS Detected Not detected Tumor 
4096 59 TP53 H179R (4.4) NGS Detected Not detected Tumor 
4321 50 TP53 Y220C (2.5) NGS Detected Not detected Tumor 
4254 68 TP53 E271* (1.7) NGS Detected Not detected Tumor 
3557 70 TP53 F134L (1.2) NGS Detected Not detected Tumor 
426 64 TP53 S127F (0.9) NGS Detected Not detected Tumor 
3847 50 TP53 V172F (0.8) NGS Detected Not detected Tumor 
3717 73 TP53 E204* (0.3) NGS Detected Not detected Tumor 
3706 59 TP53 R337L (0.2) NGS Detected Not detected Tumor 
4014 59 TP53 H193Y (13.2) NGS Not detected Not detected Uncertain 
1258 81 TP53 Y220C (0.5) NGS Not detected Not detected Uncertain 
4488 68 TP53 R282W (0.3) NGS Not detected Not detected Uncertain 
4096 68 TP53 E258* (0.3) NGS Not detected Not detected Uncertain 
4055 48 TP53 R249S (0.3) NGS Not detected Not detected Uncertain 
4188 73 TP53 C277Y (0.2) NGS Not detected Not detected Uncertain 
3856 43 TP53 R213* (0.2) NGS Not detected Not detected Uncertain 
1258 81 TP53 I162F (0.2) NGS Not detected Not detected Uncertain 
426 64 TP53 V157L (0.2) NGS Not detected Not detected Uncertain 
424 49 TP53 E298* (0.2) NGS Not detected Not detected Uncertain 
3856 43 TP53 G199E (0.1) NGS Not detected Not detected Uncertain 
3557 70 TP53 C238Y (0.1) NGS Not detected Not detected Uncertain 
3553 64 TP53 Y236C (0.1) NGS Not detected Not detected Uncertain 
2966 57 TP53 Y220C (0.1) NGS Not detected Not detected Uncertain 
IDAgePlasma NGS result (AF %)PBC testing methodTumor NGS resultPBC testing result (AF %)Mutation source
JAK2 analysis 
4433 58 JAK2 V617F (19.5) ddPCR Not detected Detected (4.66) CH 
4379 63 JAK2 V617F (2.5) ddPCR Not available Detected (0.19) CH 
4014 59 JAK2 V617F (0.5) ddPCR Not detected Detected (0.68) CH 
349 84 JAK2 V617F (0.3) ddPCR Not detected Detected (0.14) CH 
4447 87 JAK2 V617F (0.2) ddPCR Not available Detected (0.13) CH 
TP53 analysis 
426 64 TP53 Y220C (1.9) NGS Not detected Detected (1.72) CH 
4188 73 TP53 M237I (1.2) NGS Not detected Detected (2.66) CH 
3557 70 TP53 H193P (0.7) NGS Not detected Detected (1.1) CH 
3717 73 TP53 V272M (0.5) NGS Not detected Detected (1.29) CH 
423 67 TP53 R110L (0.3) NGS Not detected Detected (0.56) CH 
427 47 TP53 G244C (11.2) NGS Detected Not detected Tumor 
424 49 TP53 V157F (8.5) NGS Detected Not detected Tumor 
3856 43 TP53 R280I (5.9) NGS Detected Not detected Tumor 
4188 73 TP53 Y220C (5.6) NGS Detected Not detected Tumor 
4250 59 TP53 R248Q (5.1) NGS Detected Not detected Tumor 
3983 47 TP53 R175H (4.6) NGS Detected Not detected Tumor 
4096 59 TP53 H179R (4.4) NGS Detected Not detected Tumor 
4321 50 TP53 Y220C (2.5) NGS Detected Not detected Tumor 
4254 68 TP53 E271* (1.7) NGS Detected Not detected Tumor 
3557 70 TP53 F134L (1.2) NGS Detected Not detected Tumor 
426 64 TP53 S127F (0.9) NGS Detected Not detected Tumor 
3847 50 TP53 V172F (0.8) NGS Detected Not detected Tumor 
3717 73 TP53 E204* (0.3) NGS Detected Not detected Tumor 
3706 59 TP53 R337L (0.2) NGS Detected Not detected Tumor 
4014 59 TP53 H193Y (13.2) NGS Not detected Not detected Uncertain 
1258 81 TP53 Y220C (0.5) NGS Not detected Not detected Uncertain 
4488 68 TP53 R282W (0.3) NGS Not detected Not detected Uncertain 
4096 68 TP53 E258* (0.3) NGS Not detected Not detected Uncertain 
4055 48 TP53 R249S (0.3) NGS Not detected Not detected Uncertain 
4188 73 TP53 C277Y (0.2) NGS Not detected Not detected Uncertain 
3856 43 TP53 R213* (0.2) NGS Not detected Not detected Uncertain 
1258 81 TP53 I162F (0.2) NGS Not detected Not detected Uncertain 
426 64 TP53 V157L (0.2) NGS Not detected Not detected Uncertain 
424 49 TP53 E298* (0.2) NGS Not detected Not detected Uncertain 
3856 43 TP53 G199E (0.1) NGS Not detected Not detected Uncertain 
3557 70 TP53 C238Y (0.1) NGS Not detected Not detected Uncertain 
3553 64 TP53 Y236C (0.1) NGS Not detected Not detected Uncertain 
2966 57 TP53 Y220C (0.1) NGS Not detected Not detected Uncertain 

Finally, we studied TP53 mutations found on plasma NGS, as these are common in NSCLC but also can be seen in CH. Reviewing the 143 plasma NGS results, 108 TP53 mutations were detected. Interestingly, the number of TP53 mutations per specimen was higher than expected, with 7 specimens having >2 different TP53 mutations, which is rare on tumor NGS and suggests a polyclonal process (Fig. 3B). For 33 of the TP53 mutations detected on plasma NGS, PBC and tumor NGS were available for comparison. Fourteen TP53 mutations were detected on tumor NGS, but not in PBC, consistent with tumor-derived variants, whereas the majority of the TP53 mutations were not detected on tumor NGS (Table 1). For five mutations, NGS of PBC detected the TP53 mutation, but tumor NGS did not, consistent with CH; in some of these cases, TP53 mutations from the NSCLC and CH were both found concurrently in plasma (Fig. 3C). Fourteen TP53 mutations were detected on neither tumor NGS nor PBC NGS, such that their source could not be determined; 13 of these had an AF ≤0.5% plasma NGS, near the limit of detection (LOD) of most plasma genotyping assays and making validation using an orthogonal assay extremely difficult.

The enormous enthusiasm behind the potential clinical applications of plasma cfDNA analysis has begun to outpace our biological understanding of this new biospecimen. A range of data indicate that plasma cfDNA is a complex mixture of DNA from many sources, including germline, fetal, infectious, and malignant (2, 18). For example, it is clear that mutations detected on plasma genotyping can be of somatic or germline origin (18), and can be truncal tumor mutations or heterogeneous resistance mutations (6). Here, we demonstrate an added element of complexity—that mutations in the cfDNA can be of hematopoietic origin yet risk being mistaken as tumor-derived mutations. These CH-derived mutations can be seen at a range of AF in cfDNA, with a median AF of 0.9% across the 10 CH-derived mutations we detected.

Our findings affect the use of plasma genotyping for tumor profiling in search of targeted therapy options. We and others have advocated for the use of plasma genotyping in patients with advanced NSCLC where tumor tissue is unavailable for genomic analysis, with the caveat that a positive result can be trusted while a negative result should reflex to tumor genotyping. The finding that KRAS and TP53 mutations detected in plasma may not represent tumor genotype adds complexity to this approach. Importantly, targetable mutations in genes like EGFR, BRAF, and MET have not been described in CH; if these are detected in plasma with a rigorously validated assay, they can likely be trusted and used for initiation of targeted therapy. However, the same may not be true for KRAS or TP53 mutation—a clinician might mistakenly assume that these mutations in cfDNA are tumor derived and indicate the absence of a targetable genotyping. Our finding of CH-derived mutations in KRAS and TP53 supports the established paradigm that treatment can be initiated based upon a targetable mutation detected in cfDNA, whereas the absence of a targetable mutation in cfDNA should be followed by genotyping of a tumor biopsy specimen (6).

Second, our findings affect the development of plasma genomics as a tool for detecting minimal residual disease (MRD) or as a cancer screening assay. For example, in a patient who has completed curative therapy for cancer, detection in plasma of mutations in cancer-associated genes like TP53, KRAS, or JAK2 could be interpreted as indicating the presence of residual cancer (3), whereas these mutations might in fact be derived from CH. It has been reported that CH is seen in 25% of patients with nonhematologic cancers (16); indeed, TP53 mutations may be even more common in PBC of patients previously treated with chemotherapy as part of curative therapy (19), increasing the chance of finding such mutations at a time point when MRD is being evaluated. The clearest way to overcome this challenge would be paired genotyping of both plasma cfDNA and PBC DNA, which could allow an assay to screen out hematopoietic-derived mutations. Such a step would certainly increase the complexity and cost of cfDNA analysis, and would require the saving of PBC, which are routinely discarded when spinning plasma. Mutations detected in cfDNA and not in PBC could then be considered potentially tumor derived and indicative of risk. Of course, it is then possible that the incidental detection of JAK2 mutations in plasma or in PBC could lead to the overdiagnosis of hematologic malignancies that are clinically nonsignificant. This is merely one symptom of a broader challenge, which is that there remains no established management strategy for patients incidentally diagnosed with CH (10).

Further prospective study is needed to better capture the prevalence of CH-related false positives and their potential impact on the positive predictive value (PPV) of these assays. In our analysis, the prevalence of CH-derived KRAS mutations was 3%, which, given KRAS mutations are common in many cancer types, is unlikely to impact the PPV of KRAS plasma genotyping. In contrast, JAK2 mutations are rare in solid tumors, so the PPV of JAK2 plasma genotyping may be low—most of these mutations in cfDNA are likely to be CH derived. We note that others have found rare JAK2 mutations in plasma NGS from a large cohort of patients with colorectal cancer (20); however, this study did not perform PBC and tumor sequencing to confirm the hematopoietic etiology of these mutations. In another case report, an IDH2 mutation found in the cfDNA of a patient with metastatic colon cancer was determined to be CH derived after sequencing of a bone marrow biopsy specimen (21); our data highlight that bone marrow biopsy is not needed for detection of CH-derived mutations, as these can be easily detected in PBC. Our most surprising result is the finding that many TP53 mutations in cfDNA may be CH derived—only 14 of 33 TP53 mutations could be confirmed to be tumor derived, though many low AF TP53 mutations could not be found in PBC as well. Thus, the PPV of TP53 plasma genotyping remains unclear but could be lower than previously believed. Some might disagree with whether mutations in cfDNA due to CH should in fact be considered false positives as these are not technical errors but rather real variants derived from nontumor tissue. However, because plasma cfDNA genotyping is intended to offer a liquid biopsy of the tumor, the clinical impact of such CH-derived mutations is a false-positive result that potentially misdiagnoses a cancer patient with an incorrect genotype. Because CH-derived mutations cause a misleading diagnostic result, they should be acknowledged as false positives, much like a false-positive d-dimer in a patient with cancer undergoing work-up for thromboembolism (22) or a false-positive CT scan in a patient undergoing lung cancer screening who is found to have benign nodules (23).

Our approach was technically limited in its leveraging of ongoing commercial plasma NGS efforts, which increases the clinical relevance of our findings, but means we were unable to perform plasma and PBC sequencing with identical assays. Prospective efforts that include paired NGS of cfDNA and PBC DNA using identical assays will be needed. It should be acknowledged that available estimates of CH prevalence [5%–13% of persons over 70 years old (11, 13) and 25% of cancer patients (16)] are based upon moderate-depth sequencing efforts with an LOD around 2% AF. With the highly sensitive NGS approaches being offered by commercial plasma NGS laboratories (LOD ∼0.5% AF or lower), CH may in fact be much more common, and thus a condition we will need to be ready to recognize as a part of our routine cancer care and a potentially common finding in plasma genotyping results.

P.A. Jänne reports receiving commercial research grants from Astellas, AstraZeneca, Daiichi Sankyo, Eli Lilly, and PUMA, holds ownership interest (including patents) in Gatekeeper Pharmaceuticals, and is a consultant/advisory board member for AstraZeneca, Boehringer Ingelheim, Chugai Pharmaceuticals, Eli Lilly, Genentech/Roche, Ignyta, LOXO Oncology, and Pfizer. C.P. Paweletz reports receiving speakers bureau honoraria from AstraZeneca and BioRad, and is a consultant/advisory board member for DropWorks. G.R. Oxnard is a consultant/advisory board member for AstraZeneca, Boehringer Ingelheim, DropWorks, Genentech, GRAIL, Inivata, LOXO, Sysmex, and Takeda/Ariad. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Hu, B.C. Ulrich, N.B. Feeney, C.P. Paweletz, G.R. Oxnard

Development of methodology: Y. Hu, B.C. Ulrich, Y. Kuang, N.B. Feeney, C.P. Paweletz, G.R. Oxnard

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Hu, B.C. Ulrich, J. Supplee, Y. Kuang, P.H. Lizotte, N.B. Feeney, N.M. Guibert, M.M. Awad, K.-K. Wong, P.A. Jänne, C.P. Paweletz, G.R. Oxnard

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Hu, B.C. Ulrich, J. Supplee, P.H. Lizotte, N.B. Feeney, P.A. Jänne, C.P. Paweletz, G.R. Oxnard

Writing, review, and/or revision of the manuscript: Y. Hu, B.C. Ulrich, J. Supplee, Y. Kuang, N.B. Feeney, N.M. Guibert, M.M. Awad, K.-K. Wong, P.A. Jänne, C.P. Paweletz, G.R. Oxnard

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Hu, J. Supplee, N.B. Feeney

Study supervision: C.P. Paweletz, G.R. Oxnard

This work was supported in part by the Damon Runyon Cancer Research Foundation (G.R. Oxnard); the NIH/NCI (R01 CA135257; to P.A. Jänne and G.R. Oxnard); NCI Cancer Clinical Investigator Team Leadership Award (P30 CA006516 supplement; to G.R. Oxnard); the U.S. Department of Defense (W81XWH-14-1-0128; to G.R. Oxnard); the Anna Fuller Fund (G.R. Oxnard); and the Expect Miracles Foundation (C.P. Paweletz). The authors would like to thank Benjamin Ebert, MD, PhD, for his critical review of this article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
cobas EGFR Mutation Test v2 [about 2 screens]
. [cited 2016 Dec 19]. Available from: https://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm504540.htm.
2.
Wan
JCM
,
Massie
C
,
Garcia-Corbacho
J
,
Mouliere
F
,
Brenton
JD
,
Caldas
C
, et al
Liquid biopsies come of age: towards implementation of circulating tumour DNA
.
Nat Rev Cancer
2017
;
17
:
223
38
.
3.
Chaudhuri
AA
,
Chabon
JJ
,
Lovejoy
AF
,
Newman
AM
,
Stehr
H
,
Azad
TD
, et al
Early detection of molecular residual disease in localized lung cancer by circulating tumor DNA profiling
.
Cancer Discov
2017
;
7
:
1394
403
.
4.
Kuderer
NM
,
Burton
KA
,
Blau
S
,
Rose
AL
,
Parker
S
,
Lyman
GH
, et al
Comparison of 2 commercially available next-generation sequencing platforms in oncology
.
JAMA Oncol
2017
;
3
:
996
8
.
5.
Sacher
AG
,
Paweletz
C
,
Dahlberg
SE
,
Alden
RS
,
O'Connell
A
,
Feeney
N
, et al
Prospective validation of rapid plasma genotyping for the detection of EGFR and KRAS mutations in advanced lung cancer
.
JAMA Oncol
2016
;
2
:
1014
22
.
6.
Oxnard
GR
,
Thress
KS
,
Alden
RS
,
Lawrance
R
,
Paweletz
CP
,
Cantarini
M
, et al
Association between plasma genotyping and outcomes of treatment with osimertinib (AZD9291) in advanced non-small-cell lung cancer
.
J Clin Oncol
2016
;
34
:
3375
82
.
7.
Stetson
D
,
Dougherty
B
,
Ahmed
A
,
Lubinski
T
,
Markovets
A
,
Thress
K
, et al
Examination of analytical factors impacting concordance of plasma-tumor testing by next-generation sequencing (NGS) [abstract]
.
In:
Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1–5
;
Washington, DC. Philadelphia (PA):
AACR; Cancer Res
2017
;
77
(13 Suppl):Abstract nr LB-249
.
8.
Torga
G
,
Pienta
KJ
. 
Patient-paired sample congruence between 2 commercial liquid biopsy tests
.
JAMA Oncol
2017 Dec 14
.
[Epub ahead of print]
.
9.
Xia
L
,
Li
Z
,
Zhou
B
,
Tian
G
,
Zeng
L
,
Dai
H
, et al
Statistical analysis of mutant allele frequency level of circulating cell-free DNA and blood cells in healthy individuals
.
Sci Rep
2017
;
7
:
7526
.
10.
Steensma
DP
,
Bejar
R
,
Jaiswal
S
,
Lindsley
RC
,
Sekeres
MA
,
Hasserjian
RP
, et al
Clonal hematopoiesis of indeterminate potential and its distinction from myelodysplastic syndromes
.
Blood
2015
;
126
:
9
16
.
11.
Genovese
G
,
Kahler
AK
,
Handsaker
RE
,
Lindberg
J
,
Rose
SA
,
Bakhoum
SF
, et al
Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence
.
N Engl J Med
2014
;
371
:
2477
87
.
12.
Jaiswal
S
,
Fontanillas
P
,
Flannick
J
,
Manning
A
,
Grauman
PV
,
Mar
BG
, et al
Age-related clonal hematopoiesis associated with adverse outcomes
.
N Engl J Med
2014
;
371
:
2488
98
.
13.
Xie
M
,
Lu
C
,
Wang
J
,
McLellan
MD
,
Johnson
KJ
,
Wendl
MC
, et al
Age-related mutations associated with clonal hematopoietic expansion and malignancies
.
Nat Med
2014
;
20
:
1472
8
.
14.
McKerrell
T
,
Park
N
,
Moreno
T
,
Grove
CS
,
Ponstingl
H
,
Stephens
J
, et al
Leukemia-associated somatic mutations drive distinct patterns of age-related clonal hemopoiesis
.
Cell Rep
2015
;
10
:
1239
45
.
15.
Zink
F
,
Stacey
SN
,
Norddahl
GL
,
Frigge
ML
,
Magnusson
OT
,
Jonsdottir
I
, et al
Clonal hematopoiesis, with and without candidate driver mutations, is common in the elderly
.
Blood
2017
;
130
:
742
52
.
16.
Coombs
CC
,
Zehir
A
,
Devlin
SM
,
Kishtagari
A
,
Syed
A
,
Jonsson
P
, et al
Therapy-related clonal hematopoiesis in patients with non-hematologic cancers is common and associated with adverse clinical outcomes
.
Cell Stem Cell
2017
;
21
:
374
82
e4
.
17.
Sholl
LM
,
Do
K
,
Shivdasani
P
,
Cerami
E
,
Dubuc
AM
,
Kuo
FC
, et al
Institutional implementation of clinical tumor profiling on an unselected cancer population
.
JCI Insight
2016
;
1
:
e87062
.
18.
Hu
Y
,
Alden
RS
,
Odegaard
JI
,
Fairclough
SR
,
Chen
R
,
Heng
J
, et al
Discrimination of germline EGFR T790M mutations in plasma cell-free DNA allows study of prevalence across 31,414 cancer patients
.
Clin Cancer Res
2017
;
23
:
7351
9
.
19.
Gibson
CJ
,
Lindsley
RC
,
Tchekmedyian
V
,
Mar
BG
,
Shi
J
,
Jaiswal
S
, et al
Clonal hematopoiesis associated with adverse outcomes after autologous stem-cell transplantation for lymphoma
.
J Clin Oncol
2017
;
35
:
1598
605
.
20.
Strickler
JH
,
Loree
JM
,
Ahronian
LG
,
Parikh
AR
,
Niedzwiecki
D
,
Pereira
AAL
, et al
Genomic landscape of cell-free DNA in patients with colorectal cancer
.
Cancer Discov
2018
;
8
:
164
73
.
21.
Zhang
BM
,
Aleshin
A
,
Lin
CY
,
Ford
J
,
Zehnder
JL
,
Suarez
CJ
. 
IDH2 mutation in a patient with metastatic colon cancer
.
N Engl J Med
2017
;
376
:
1991
2
.
22.
Kabrhel
C
,
Mark Courtney
D
,
Camargo
CA
 Jr
,
Plewa
MC
,
Nordenholz
KE
,
Moore
CL
, et al
Factors associated with positive D-dimer results in patients evaluated for pulmonary embolism
.
Acad Emerg Med
2010
;
17
:
589
97
.
23.
Swensen
SJ
,
Jett
JR
,
Hartman
TE
,
Midthun
DE
,
Sloan
JA
,
Sykes
AM
, et al
Lung cancer screening with CT: Mayo Clinic experience
.
Radiology
2003
;
226
:
756
61
.