Abstract
Molecular profiling of circulating tumor DNA (ctDNA) is a promising noninvasive tool. Here, next-generation sequencing (NGS) of blood-derived ctDNA was performed in patients with advanced colorectal cancer. We investigated ctDNA-derived genomic alterations, including potential actionability, concordance with tissue NGS, and serial dynamics in 78 patients with colorectal cancer using a clinical-grade NGS assay that detects single nucleotide variants (54–73 genes) and selected copy-number variants, fusions, and indels. Overall, 63 patients [80.8% (63/78)] harbored ctDNA alterations; 59 [75.6% (59/78)], ≥1 characterized alteration (variants of unknown significance excluded). All 59 patients had actionable alterations potentially targetable with FDA-approved drugs [on-label and/or off-label (N = 54) or with experimental drugs in clinical trials (additional five patients); University of California San Diego Molecular Tumor Board assessment]: 45, by OncoKB (http://oncokb.org/#/). The tissue and blood concordance rates for common specific alterations ranged from 62.3% to 86.9% (median = 5 months between tests). In serial samples from patients on anti-EGFR therapy, multiple emerging alterations in genes known to be involved in therapeutic resistance, including KRAS, NRAS, BRAF, EGFR, ERBB2, and MET were detected. In conclusion, over 80% of patients with stage IV colorectal cancer had detectable ctDNA, and the majority had potentially actionable alterations. Concordance between tissue and blood was between 62% and 87%, despite a median of 5 months between tests. Resistance alterations emerged on anti-EGFR therapy. Therefore, biopsy-free, noninvasive ctDNA analysis provides data relevant to the clinical setting. Importantly, sequential ctDNA analysis detects patterns of emerging resistance allowing for precision planning of future therapy.
Introduction
Colorectal cancer is a heterogeneous disease that develops as a consequence of different combinations of epigenetic and genetic alterations, with significant variability observed in individual patient prognosis and therapy response, perhaps due to molecular heterogeneity. Advances in the treatment of patients with metastatic colorectal cancer have led to an improvement in patient survival partly due to the use of biologic agents targeting the EGFR signaling pathway or tumor angiogenesis (1–5). In addition, certain biomarkers with prognostic or predictive value have been identified. For example, activating mutations of KRAS and NRAS can predict a lack of response to anti-EGFR therapy. These mutations frequently coexist with alterations in the PI3K/Akt/mTOR pathway encoding genes (6). Hypermutated colorectal cancer tumors have been found to be excellent targets for programmed death 1 (PD-1) inhibitors (7). Finally, BRAF mutation is a known poor prognostic feature (1, 8, 9).
Further understanding of underlying genomic alterations in colorectal cancer has been made possible by recent improvements in DNA sequencing technology. The Cancer Genome Atlas network conducted a comprehensive genome-wide analysis of somatic mutations in colorectal tumors. The most commonly altered genes were as follows in nonhypermutated colorectal cancers: APC (81%), TP53 (60%), KRAS (43%), and PIK3CA (18%). BRAF mutations were frequently associated with hypermutated colorectal cancers (10). In addition, understanding of heterogenous genetic makeup of colorectal cancer has led to a consensus on molecular subtypes used to classify colorectal cancer (11).
The development of next-generation sequencing (NGS) platforms and their implementation into clinical use allows for rapid clinical-grade genomic analysis to identify actionable genomic alterations, and thus has led to the use of targeted therapies matched to patients' specific alterations (12–15). Until recently, genomic sequencing was typically performed on archival tumor tissue. However, acquisition of tumor tissue is not always feasible, and a tissue biopsy at a single time point may not, due to tumor heterogeneity, provide a recent or complete picture of the molecular background of tumor evolution, response, and resistance (16, 17). One strategy to overcome these challenges is to investigate circulating biomarkers. Circulating tumor DNA (ctDNA) is shed into bloodstream from cancer cells and can be isolated from blood (also known as a “liquid biopsy”; refs. 18–20). More recently, the technology has rapidly advanced, and evaluating multiple genes by performing NGS on ctDNA has proven useful when applied in the clinic (21–26).
Herein, we describe the results of ctDNA testing in 78 patients with colorectal cancer whose blood-derived ctDNA was interrogated by a targeted NGS panel. Type, distribution, and frequency of genomic alterations, potential actionability, concordance with tissue testing, and emerging resistance alterations detected in ctDNA after anti-EGFR–based therapy (including an illustrative case) are described.
Materials and Methods
Patients
We investigated genomic alterations and clinicopathologic data from electronic medical records in 78 consecutive patients with colorectal cancer followed at the University of California San Diego (UCSD) Moores Cancer Center (San Diego, CA), for whom ctDNA testing was performed on their blood samples between December 2014 and November 2016. This was a retrospective analysis of all eligible patients with available data during the time period. A total of 1,005 eligible patients with various cancers had ctDNA analysis during this time period.
NGS
ctDNA analyses.
Digital sequencing was performed by clinical laboratory improvement amendments (CLIA)-licensed and College of American Pathologist (CAP)-accredited laboratory (Guardant Health, Inc., http://www.guardanthealth.com/). ctDNA was extracted from whole blood collected in 10 mL Streck tubes, and 5–30 ng of ctDNA was prepared for sequencing as described previously (23). The fractional concentration or variant allele fraction for a given somatic mutation is calculated as the fraction of ctDNA harboring that mutation in a background of wild-type ctDNA fragments at the same nucleotide position. Germline alterations are filtered out and not reported. This ctDNA assay has high sensitivity [detects 85%+ of the single nucleotide variants (SNV) detected in tissue in patients with advanced cancer] and specificity (> 99.9999%; ref. 23). Throughout the timeframe of this study, the ctDNA assay expanded gene panels from 54 to 73 genes (Supplementary Table S1). There were 104 blood samples collected from a total of 78 patients in this study (17 patients had multiple blood samples assayed): four samples were tested with the 54 gene panel; 44 with the 68 gene panel; 55 with the 70 gene panel; and one with the 73 gene panel. The assay reports SNVs in all genes and selects fusions, copy-number variants (CNV), and indels (Supplementary Table S1). Degree of CNVs were reported as follows: 1+, 2.13–2.40, which is the 10th to 50th percentile; 2+, 2.41–4.00, which is >50th to 90th percentile; and 3+, greater than 4.0 copy numbers, which is >90th percentile. For ctDNA, the following information was evaluated; the number of total alterations and the number of characterized alterations. We counted both variants of unknown significance (VUS) and characterized alterations when we refer to total alterations. Synonymous alterations were not included in any analysis.
Tumor tissue analyses.
Tissue testing was performed on formalin-fixed, paraffin-embedded tissue by Foundation Medicine (FoundationOne, http://www.foundationone.com), which is a clinical-grade NGS test (315 genes; ref. 27). For tissue NGS, we included only characterized alterations in our analyses.
Concordance
We investigated the concordance for the 61 patients who had both tests performed (tissue and blood). Concordance between tissue and blood was calculated using a kappa coefficient. We examined specific concordance rates for the most frequent alterations including TP53, KRAS, APC, PIK3CA, BRAF, and MYC, and all the genes examined were present in the tissue and ctDNA panels.
Definition of actionability
An actionable alteration was defined by UCSD Molecular Tumor Board as a genomic alteration that produces a protein product serving as either the direct target or as part of a signaling pathway that could be impacted by drugs. Drug impact could be impact or via differential expression on tumor versus normal cells. Drugs might be available for on-label (FDA-approved for colorectal cancer) and/or off-label use (FDA-approved for an indication other than colorectal cancer) or in experimental clinical trials. Protein products of genes were considered actionable by small-molecule inhibitors if the compound impacted the protein at low 50% inhibitory concentrations. Genes that produced proteins that were recognized as the main target of an antibody were also considered actionable. We also utilized OncoKB (28) as an additional source to define actionability.
Statistical analysis
The study was performed under an Institutional Review Board–approved registry type study (PREDICT). PREDICT has both prospective and retrospective components but, in this case, data was gathered retrospectively. This study was performed in accordance with the UCSD Moores Cancer Center Institutional Review Board guidelines for NCT02478931 and for any investigational therapies to which the patients consented.
Results
Patient characteristics
The median age at first blood draw was 52 years (range, 27–82 years); 40 patients (51.3%) were men (Table 1). All patients had a diagnosis of colorectal adenocarcinoma. Overall, 53 individuals (67.9%) had low-grade tumor; 11 (14.1%), high-grade tumor; and 6 had mucinous adenocarcinoma. Regarding location, 60 patients had tumors involving the colon (21 right colon; 36 left colon; 3 transverse colon), and 18 involved the rectum. Seventy-seven patients had stage IV disease at the time of first blood draw; one, stage III disease.
Patient characteristics and genomic alterations in patients with colorectal cancer (N = 78)
Genomic alterations among colorectal cancer; ctDNA results and percentage
The total number of genomic alterations were 310 and, of these, 214 (69.0%) were characterized alterations, including substitutions (N = 148), amplifications (N = 64), and indels (N = 2). No fusions were observed. Ninety-six alterations (31.0%) were VUSs (Table 1). Of the 214 characterized alterations, 34 were genomically distinct, and 119 were molecularly distinct alterations (e.g., TP53 R248W and TP53 R248Q were considered genomically identical but molecularly distinct; Supplementary Tables S2 and S3).
Among 78 patients, 63 patients (80.8%) had ctDNA alterations detected with 59 (75.6%) having ≥1 characterized alteration and 48 (61.5%) having ≥2 characterized alterations. The median number of alterations per patient was 3 (range, 0–26); median number of characterized alterations, 2 (range, 0–13; Table 1).
Focusing on characterized alterations, the most frequently altered gene was TP53 [52.6% (41/78)] followed by KRAS [35.9% (28/78)], APC [28.2% (22/78)], EGFR [16.7% (13/78)], BRAF [15.4% (12/78)], and PIK3CA [15.4% (12/78; Fig. 1A; Supplementary Table S2)].
A, Frequency of genomic alterations among patients with colorectal cancer (N = 78). Includes alterations with > 5% frequency. *, Multiple alterations indicate that the patient had > 1 type of alteration in the same gene (substitutions, amplifications, VUS, etc). Abbreviations: VUS, variant of unknown significance. B, Percent ctDNA among frequently altered genes in patients with colorectal cancer. Depicted are seven frequently altered genes (not including amplification). The median of % ctDNA with SE is depicted. Presented here are characterized alterations that were detectable.
When alterations were grouped depending on the oncogenic pathways, genes involved in the MAPK signaling pathway were altered in 59.0% (46/78) of patients; TP53-associated genes were altered in 53.8% (42/78); tyrosine kinase families were altered in 37.2% (29/78); the Wnt signaling was altered in 33.3% (26/78); the PI3K signaling was altered in 17.9% (14/78); and cell-cycle associated genes were altered in 14.1% (11/78; Table 2).
Selected actionable genomic alterations and examples of possible targeted therapies (N = 78)
Among frequently altered genes, APC had the highest median mutant allele ctDNA fraction of 6.6% (range, 0.1–55.5). Most of the other characterized mutations had a median ctDNA fraction of less than 5%. There was no clear association between the frequencies of gene alterations and the fraction of ctDNA detected in the blood (Fig. 1B).
Actionable genomic alterations among patients with colorectal cancer
We determined actionability based on our Molecular Tumor Board assessments (Supplementary Table S4) and by OncoKB (28). Per our Molecular Tumor Board (UCSD) assessment, of the 214 characterized alterations, 70.1% (150/214) were potentially targetable with FDA-approved drugs as on-label and/or off-label use, and an additional 26.2% (56/214) were theoretically targetable with drugs that are currently in clinical trials. Altogether, 96.3% (206/214) were actionable alterations that were targetable either with FDA-approved drugs or with experimental drugs in clinical trials (Supplementary Tables S3 and S4).
Of the 59 patients who harbored ≥1 characterized alteration, all had ≥1 potentially actionable alteration. FDA-approved, on-label therapies were available for 20.5% (16/78) of patients and another 48.7% (38/78) had alterations potentially targetable ith FDA-approved drugs in a different indication (off-label use). An additional 6.4% (5/78) of patients had alterations theoretically targetable with experimental drugs in clinical trials (Supplementary Fig. S1; Supplementary Tables S3 and S4).
Per OncoKB assessment (http://oncokb.org/#/), of the 214 characterized alterations, 5.1% (11/214) were potentially targetable alterations for colorectal cancer and an additional 34.1% (73/214) were theoretically targetable for cancer types other than colorectal cancer. Altogether, 39.2% (84/214) were actionable alterations that were considered to be targetable alterations (Supplementary Tables S3). Of the 59 patients who harbored ≥1 characterized alteration, 45 had ≥1 potentially actionable alterations that were targetable per OncoKB assessment. At least one targetable alteration indicated for colorectal cancers was found in 10.3% (8/78) of patients and another 47.4% (37/78) had alterations that were theoretically targetable for cancer types other than colorectal cancer (Supplementary Tables S3).
Concordance of the ctDNA test with tissue NGS test
Of the 78 patients who had ctDNA test, 61 patients also had a tissue genomic test. The time between biopsy and blood draw ranged from 1 day to 6.6 years, with a median of 5.0 months. We examined concordance for the most frequent alterations. The concordance rates were 70.5% for TP53, 77.0% for KRAS, 62.3% for APC, 80.3% for PIK3CA, 86.9% for BRAF, 83.6% for MYC (Table 3). There was no statistically significant difference in concordance rates in patients for whom the time interval between biopsy and blood draw used for testing was ≤6 months (n = 35) versus those for whom the time interval was >6 months (n = 26).
Concordance for specific alterations in patients with colorectal cancer (N = 61)
Emerging resistance alterations in ctDNA along with therapeutic intervention
We documented that emerging resistance alterations could be detected in ctDNA from patients treated with anti-EGFR antibodies. There were 9 patients treated with anti-EGFR–based therapy who also had tissue NGS prior to or soon after the therapy. Tissue NGS test was performed at various time points (range, 19 days to 26.9 months) before the initiation of anti-EGFR–based therapy; one patient had it performed one month after the initiation of the therapy (#12). Pretreatment ctDNA was also available in two patients, and both showed no detectable alterations (#1, #30; Table 4).
Analysis of resistant markers from anti-EGFR–based therapy with NGS (tissue and/or liquid; N = 9)
In total, we observed 31 alterations that were not detected in the tumor and only appeared in the ctDNA after the therapy was initiated. Of these, we found 15 alterations in genes known to be involved in therapeutic resistance to anti-EGFR therapy (Table 4, #2, #12, #38, #68). The progression-free survival of these 4 patients were 4.6 months, 12.4 months, 6.7 months, and 2.5+ months, respectively; 3 patients (#2, #12, #38) experienced disease progression at the time of blood draw after a prior response/stable disease, and 1 patient (#68) was assessed as having stable disease at the time of blood draw, but lost to follow up thereafter. KRAS alterations were observed in 2 patients (#12, #38); 1 patient had five molecularly distinct KRAS alterations (KRAS G12A, KRAS G12C, KRAS G13D, KRAS Q61H, and KRAS amplification) and the other had KRAS amplification. Two NRAS mutations in 1 patient (#12) were observed; NRAS Q61H, NRAS Q61K. BRAF G469A mutation was observed in 1 patient (#12). EGFR S492R mutation was observed in 2 patients (#12, #68). In addition, amplifications in ERBB2 (n = 3; #2, #12, #38) and MET (n = 1; #68) were observed.
A representative case (#38) with emerging KRAS and ERBB2 alterations upon progression on anti-EGFR–based therapy is presented (Fig. 2).
KRAS wild-type colon cancer patient treated with anti-EGFR–based therapy. ctDNA analysis at progression with multiple emerging resistance alterations. 49-year-old woman with metastatic colon cancer. Tissue NGS at diagnosis showed alterations including TP53 and APC (Table 4, case ID #38). After progressing on FOLFIRI plus bevacizumab, patient was started on FOLFOX plus cetuximab. Patient initially had response in left adrenal and retroperitoneal lymph node metastases. However, after 6.7 months of therapy, left adrenal node was progressing, thus patient was taken off from therapy. Postprogression ctDNA showed persistent TP53 and APC alterations as well as emerging KRAS and ERBB2 amplifications. Left, pretreatment PET/CT. Top figure shows left adrenal metastasis. Bottom figure shows retroperitoneal lymph node metastases. Middle, day 139 from FOLFOX plus cetuximab with response in left adrenal and peritoneal lymph node metastases. Right: day 202 from FOLFOX plus cetuximab with worsening left adrenal metastasis.
Discussion
We investigated genomic alterations in 78 patients with colorectal cancer (77, stage IV) using a targeted NGS assay that analyzed blood-derived ctDNA. Of the total, 63 patients (80.8%) had ≥1 detectable alteration(s), which is consistent with previous reports (23–25, 29). Of 59 patients (75.6%) who harbored at least one characterized alteration, the most frequent alterations were TP53 mutations (52.6%), followed by alterations in KRAS (35.9%), APC (28.2%), EGFR (16.7%), BRAF (15.4%), and PIK3CA (15.4%). The mutation frequencies for KRAS and BRAF are similar to those reported in tissue, which generally show that about 35% to 45% of patients have a KRAS mutation (30, 31) and about 10% of patients with colorectal cancer have a BRAF alteration (32). Examining ctDNA, Thierry and colleagues (33) noted a higher rate of KRAS mutations in ctDNA (∼59%), while our rate of 35.9% was more consistent with the lower range found in tissue studies. Differences between studies can be attributable to different sample sources (tissue vs. ctDNA in blood), relatively small sample size, technical differences such as those in depth of sequencing, and other factors relating to the phenotype of patients tested.
EGFR mutations were detected in three patients (3.8%; two, S492R and one, G465R), and BRAF mutations in 4 patients (5.1%) (three, V600E and one, G469A). Several previous studies have shown that resistant EGFR ectodomain mutations (S492R, G465R) emerge in patients treated with anti-EGFR therapy (34–36). In our study, all three patients who harbored EGFR mutations in ctDNA were previously treated with cetuximab-based regimens, and available pretreatment tumor specimens from 2 patients confirmed that the EGFR S492R mutation was not present before cetuximab treatment (Table 4). Finding this mutation in ctDNA has particular clinical relevance because patients with EGFR S492R mutation appearing under the pressure of cetuximab therapy may still respond to panitumumab (35).
We also examined the concordance in the patients who had both a tissue and ctDNA test. Of interest in this regard, Khan and colleagues, (37) have shown that a significant proportion of patients with colorectal cancers defined as RAS wild-type based on tissue evaluation harbor alterations in the RAS pathway in pretreatment cfDNA and do not benefit from EGFR inhibitors. In our study, the concordance rates for common specific alterations were in the range of 62.3%–86.9% with a median of 5 months between tests (Table 3). Testing of ctDNA therefore demonstrated acceptable concordance with NGS testing of tumor tissue obtained during clinical practice in this study. Previous studies have shown that concordance decreases with the temporal separation between tissue and ctDNA tests (25), but we did not observe such a difference, perhaps because of the limited number of patients in the study. It is also unclear why our results showed a lower concordance for APC compared with other genes (62.3% vs. 70.5%–86.9%); this may reflect a lower rate of APC mutation (28.2%) in our ctDNA versus 67.2% (41/61) in tissue, perhaps suggesting that APC-bearing ctDNA is less likely to shed into the circulation. In one study of mutation analysis in ctDNA using a targeted NGS panel in a variety of metastatic solid tumor patients including colorectal cancer, the concordance rate for all mutations found across 54 genes was 85.9% between tissue and ctDNA (38). In another study of 42 patients with advanced stage non–small cell lung cancer, concordance between tissue and plasma was 76% when concordance for mutations in EGFR, KRAS, PIK3CA, and TP53 was measured using a targeted sequencing method (39). In these studies, tissue and blood sampling were performed concurrently, while in our study, the time between biopsy and blood draw ranged from 1 day to 6.6 years (median = 5.0 months). Some discordance between genomic profiles from tissue and ctDNA is generally observed. On one hand, ctDNA tests can theoretically detect shed DNA from multiple tumor sites, while tissue biopsy demonstrates only the alterations found at the specific site of sampling. Recent reports focusing on discordance of the KRAS mutational status between primary tumor and paired metastasis showed discordance rates of 3.6%–17.5% (40). On the other hand, in cancers with small tumor burden, mutations identified in tissue may not be detected in ctDNA test due to the low content of ctDNA. In addition, emergence of resistance alterations can occur under the pressure of therapy during the elapsed interval between each test (41–43); at the same time, targeted therapies may suppress ctDNA (44). Of note, we observed that ctDNA samples harbor unique CNVs that were absent from tissues. For instance, amplifications in BRAF (N = 6), KRAS (N = 3), PIK3CA (N = 3), MET (N = 2), FGFR1 (N = 2), and FGFR2 (N = 2) were only detected in ctDNA (not in tissue in the 61 patients who had both tested), which, at least theoretically, may be attributed to the inadequate representation of tumor heterogeneity by tissue testing. In contrast, APC mutations were more frequently detected in tissue than in ctDNA (67.2% vs. 28.2%). Thus, complementary assessment of both tissue and ctDNA appears advantageous to assess dynamic tumor profiles.
Regarding actionability to guide therapeutic decisions, per the UCSD Molecular Tumor Board assessment, 69.2% of patients had at least one actionable alteration that could be impacted by an FDA-approved drug, and an additional 6.4% had alterations targetable with experimental drugs in clinical trials. Genes involved in the MAPK pathway, TP53-associated genes, tyrosine kinase families and Wnt signaling were frequently altered (33.3%–59.0%). Although less frequent, alterations in PI3K and cell-cycle machinery were also observed (14.1%–17.9%). Per OncoKB (28), the percent of patients with at least one actionable alteration was 57.7% (45/78). Recent actionability studies performed on ctDNA in various solid tumor patients reported that 71%–81.5% of patients had at least one clinically actionable alteration (24, 25), which is consistent with our results. Accumulating evidence suggests that biomarker-based treatment approach may be able to improve clinical outcome (12–15). Clinical utility of this approach in genomically matched patient populations continues to be investigated in large prospective clinical trials.
Finally, analysis of ctDNA in serially collected plasma samples allowed detection of emerging mutant ctDNA, which can be used to monitor patients for disease progression as well as to find mechanisms of resistance. In our patients treated with anti-EGFR agents, we could detect multiple emerging alterations in genes known to be involved in therapeutic resistance to anti-EGFR therapy, KRAS, NRAS, BRAF, and EGFR mutations, and amplification of KRAS, ERBB2, and MET, consistent with the implication of many of these genes in both primary as well as acquired resistance to anti-EGFR therapy (29, 34–36, 42, 43, 45–49; see illustrative case; Fig. 2). A limitation of our study is that we did not administer therapy that would impact resistant alterations; future studies should address this issue to validate the role of these alterations in resistance.
The study has additional limitations. Importantly, understanding the underlying mechanisms for response and resistance cannot be fully elucidated in the clinic and further bench-side investigations are needed. The median age of patients was 52, which is young for colorectal cancer, and might reflect a selection bias because of the referral pattern to a tertiary care center or because of physician discretion in ordering ctDNA testing. Another limitation relates to whether individual mutations are drivers versus passengers (50). Regarding treatment, some patients who received EGFR inhibitors in this study were also given chemotherapy. This causes some problems in interpretation of genomic findings, although the presence of RAS mutations is highly suggestive of an independent effect of anti-EGFR antibodies and clonal expansion under their selective pressure. Finally, the timing of sample acquisition varied as NGS was performed on the basis of physician discretion. Future studies would benefit from analysis at prospective, predefined time points.
To conclude, over 80% of patients with advanced colorectal cancer had detectable ctDNA alterations and more than half [76% (59/78)] per UCSD Molecular Tumor Board and 57.7% (45/78) by OncoKB (28) had potentially actionable alterations. Concordance between tissue and blood for common specific alterations was acceptable. In addition, we could detect emerging resistance alterations in ctDNA from patients treated with anti-EGFR–based therapy. These data suggest that biopsy-free, noninvasive ctDNA analysis is becoming a valuable option when an invasive tissue biopsy is not feasible, and ctDNA can yield promising candidate biomarkers for the detection and monitoring of colorectal cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: I.S. Choi, S. Kato, P.T. Fanta
Development of methodology: I.S. Choi, S. Kato, R.B. Lanman
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.S. Choi, S. Kato, P.T. Fanta, V.M. Raymond
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I.S. Choi, S. Kato, R. Okamura, V.M. Raymond, R.B. Lanman, S.M. Lippman, R. Kurzrock
Writing, review, and/or revision of the manuscript: I.S. Choi, S. Kato, P.T. Fanta, L. Leichman, R. Okamura, V.M. Raymond, R.B. Lanman, S.M. Lippman, R. Kurzrock
Acknowledgments
This study was funded in part by the Joan and Irwin Jacobs fund and by National Cancer Institute grant P30 CA023100.
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.
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/).
Mol Cancer Ther 2019;18:1852–62
- Received August 25, 2018.
- Revision received December 3, 2018.
- Accepted July 12, 2019.
- Published first July 18, 2019.
- ©2019 American Association for Cancer Research.
References
Volume 18, Issue 10
