Search

Purpose: Molecular properties associated with complete response or acquired resistance to concurrent chemotherapy and radiotherapy (CRT) are incompletely characterized. Experimental Design: We performed integrated whole-exome/transcriptome sequencing and immune infiltrate analysis on rectal adenocarcinoma tumors prior to neoadjuvant CRT (pre-CRT) and at time of resection (post-CRT) in 17 patients [8 complete/partial responders, 9 nonresponders (NR)]. Results: CRT was not associated with increased tumor mutational burden or neoantigen load and did not alter the distribution of established somatic tumor mutations in rectal cancer. Concurrent KRAS/TP53 mutations (KP) associated with NR tumors and were enriched for an epithelial–mesenchymal transition transcriptional program. Furthermore, NR was associated with reduced CD4/CD8 T-cell infiltrates and a post-CRT M2 macrophage phenotype. Absence of any local tumor recurrences, KP/NR status predicted worse progression-free survival, suggesting that local immune escape during or after CRT with specific genomic features contributes to distant progression. Conclusions: Overall, while CRT did not impact genomic profiles, CRT impacted the tumor immune microenvironment, particularly in resistant cases.

ATP-competitive fibroblast growth factor receptor (FGFR) kinase inhibitors, including BGJ398 and Debio 1347, show antitumor activity in patients with intrahepatic cholangiocarcinoma (ICC) harboring activating FGFR2 gene fusions. Unfortunately, acquired resistance develops and is often associated with the emergence of secondary FGFR2 kinase domain mutations. Here, we report that the irreversible pan-FGFR inhibitor TAS-120 demonstrated efficacy in 4 patients with FGFR2 fusion–positive ICC who developed resistance to BGJ398 or Debio 1347. Examination of serial biopsies, circulating tumor DNA (ctDNA), and patient-derived ICC cells revealed that TAS-120 was active against multiple FGFR2 mutations conferring resistance to BGJ398 or Debio 1347. Functional assessment and modeling the clonal outgrowth of individual resistance mutations from polyclonal cell pools mirrored the resistance profiles observed clinically for each inhibitor. Our findings suggest that strategic sequencing of FGFR inhibitors, guided by serial biopsy and ctDNA analysis, may prolong the duration of benefit from FGFR inhibition in patients with FGFR2 fusion–positive ICC.Significance:ATP-competitive FGFR inhibitors (BGJ398, Debio 1347) show efficacy in FGFR2-altered ICC; however, acquired FGFR2 kinase domain mutations cause drug resistance and tumor progression. We demonstrate that the irreversible FGFR inhibitor TAS-120 provides clinical benefit in patients with resistance to BGJ398 or Debio 1347 and overcomes several FGFR2 mutations in ICC models.This article is highlighted in the In This Issue feature, p. 983

During cancer therapy, tumor heterogeneity can drive the evolution of multiple tumor subclones harboring unique resistance mechanisms in an individual patient1–3. Previous case reports and small case series have suggested that liquid biopsy (specifically, cell-free DNA (cfDNA)) may better capture the heterogeneity of acquired resistance4–8. However, the effectiveness of cfDNA versus standard single-lesion tumor biopsies has not been directly compared in larger-scale prospective cohorts of patients following progression on targeted therapy. Here, in a prospective cohort of 42 patients with molecularly defined gastrointestinal cancers and acquired resistance to targeted therapy, direct comparison of postprogression cfDNA versus tumor biopsy revealed that cfDNA more frequently identified clinically relevant resistance alterations and multiple resistance mechanisms, detecting resistance alterations not found in the matched tumor biopsy in 78% of cases. Whole-exome sequencing of serial cfDNA, tumor biopsies and rapid autopsy specimens elucidated substantial geographic and evolutionary differences across lesions. Our data suggest that acquired resistance is frequently characterized by profound tumor heterogeneity, and that the emergence of multiple resistance alterations in an individual patient may represent the ‘rule’ rather than the ‘exception’. These findings have profound therapeutic implications and highlight the potential advantages of cfDNA over tissue biopsy in the setting of acquired resistance. Direct prospective comparison of circulating tumor DNA and tissue biopsy sequencing shows the superiority of liquid biopsies for capturing clinically relevant alterations mediating resistance to targeted therapies in cancer patients.

3565 Background: Over half of patients (pts) with oligometastatic CRC treated with curative intent surgery or radiotherapy experience cancer recurrence with or without adjuvant chemotherapy. ctDNA detection post-definitive treatment could identify high risk pts for additional intervention to eliminate molecular residual disease. Here we report results of a prospective observational study aiming to determine ctDNA detection rates using a sensitive liquid biopsy and to correlate post-procedure ctDNA detection (post-ctDNA (+)) with radiographic mCRC recurrence. Methods: Pts with mCRC intending to undergo a curative intent procedure were prospectively recruited at two US sites. ctDNA was collected pre-procedure, 3 weeks post-procedure, and at multiple structured follow-up timepoints. The presence of ctDNA was evaluated using a plasma-only integrated genomic and epigenomic assay (Guardant Reveal, Guardant Health). A bioinformatic classifier was applied to differentiate tumor derived versus non-tumor derived cell-free DNA. Results: Among 52 enrolled pts, post-ctDNA data is available for 45 pts (87%), with a median of 4 (range 1-10) timepoints per pt. The sample analysis failure rate was 1% (2/217). As of 1/1/2021, the radiographic recurrence rate was 60% with a median follow-up time of 50 (range 4-192) weeks. 23 of 25 pts with post-ctDNA(+) have had recurrence (Positive Predictive Value [PPV], 92%). On average, ctDNA was detected 28 weeks before radiographic recurrence (mean 12 vs. 40 weeks, respectively). The two pts with post-ctDNA(+) but no recurrence have > 3 years follow-up; one pt received adjuvant chemotherapy and cleared ctDNA. With a median event-free follow-up time of 97 (range 4-192) weeks, 4 of 20 pts with no post-ctDNA detection (-) have recurred (Negative Predictive Value, 80%). 3 of 4 pts with recurrence despite post-ctDNA(-) also were pre-ctDNA(-). We observed a sensitivity of 85% and a specificity of 89% for the ctDNA assay. The median time to radiographic recurrence was 36 wks for ctDNA(+) vs. not reached for ctDNA(-) (Hazard Ratio, 7.7; 95% CI, 2.6-22.5; P

6560 Background: PROs assessing quality of life (QOL) and symptoms at a single timepoint frequently correlate with clinical outcomes in patients with cancer, yet efforts to understand how longitudinal changes in PROs can predict for treatment outcomes are lacking. In practice, oncologists often use changes in serum TMs (CEA and CA19-9) to monitor patients with GI cancer, and thus we sought to examine associations of 1-month changes in PROs and TMs with treatment response and survival outcomes among patients with advanced GI cancer. Methods: We prospectively enrolled patients with metastatic GI cancer prior to initiating chemotherapy at Massachusetts General Hospital from 5/2019-12/2020. At baseline (start of treatment) and 1-month later, we collected PROs (QOL [Functional Assessment of Cancer Therapy General {FACT-G}], physical symptoms [Edmonton Symptom Assessment System {ESAS}], and psychological symptoms [Patient Health Questionnaire-4 {PHQ-4}]) and TMs. We used regression models to examine associations of 1-month changes in PROs and TMs with treatment response (clinical benefit [defined as decreased or stable tumor burden] or progressive disease at the time of first scan) and survival outcomes (progression-free survival [PFS] and overall survival [OS]), adjusted for baseline values of each respective variable. Results: We enrolled 159 of 191 patients approached (83.2% enrollment); 134 had 1-month follow-up data (median age = 64 years [range: 28 to 84 years], 64.2% male, 46.3% pancreaticobiliary cancer). For treatment response, 63.4% had clinical benefit and 36.6% had progressive disease at the time of first scan (mean time to first scan = 2.01 months). Changes in PROs (ESAS-Total: OR = 0.97, p = 0.022; ESAS-Physical: OR = 0.96, p = 0.027; PHQ-4 depression: OR = 0.67, p = 0.014; FACT-G: OR = 1.07, p = 0.001), but not TMs (CEA: OR = 1.00, p = 0.836 and CA19-9: OR = 1.00, p = 0.796), were associated with clinical benefit at the time of first scan. Changes in ESAS-Total (HR = 1.03, p = 0.004), ESAS-Physical (HR = 1.03, p = 0.021), PHQ-4 depression (HR = 1.22, p = 0.042), FACT-G (HR = 0.97, p = 0.003), and CEA (HR = 1.00, p = 0.001) were predictors of PFS. Changes in ESAS-Total (HR = 1.03, p = 0.006) and ESAS-Physical (HR = 1.04, p = 0.015) were predictors of OS, but 1-month changes in TMs (CEA: HR = 1.00, p = 0.377 and CA19-9: HR = 1.00, p = 0.367) did not significantly predict for OS. Conclusions: We found that 1-month changes in PROs can predict for treatment response and survival outcomes in patients with advanced GI cancers. Notably, 1-month changes in CEA only correlated with PFS, while changes in CA19-9 did not significantly predict treatment response or survival outcomes. These findings highlight the potential for early changes in PROs to predict treatment outcomes while underscoring the need to monitor and address PROs in patients with advanced cancer.

474 Background: Severe treatment-induced lymphopenia is a poor prognostic factor for multiple solid malignancies. Emerging data suggests that increased TCR repertoire is associated with improved survival. We hypothesized that increased TCR repertoire following severe treatment-induced lymphopenia would be associated with better survival after accounting for known prognostic factors. Methods: Patients within a prospective cohort were retrospectively analyzed for severe (grade 4) lymphopenia during radiotherapy (RT) at a single institution between 2007-2016. Peripheral blood samples were collected weekly. TCR repertoire at 3 time points were evaluated: pre-RT, during or immediately post-RT at absolute lymphocyte count (ALC) nadir, and ~2 months post-RT. Peripheral blood was analyzed via multiparametric flow cytometry. TCR beta sequencing was performed on immunoSEQ platform from Adaptive Biotechnologies. A Gini diversity index reflecting TCR repertoire was calculated as well as percentage change between ALC nadir and ALC 2 months post-RT. Primary endpoint was overall survival (OS). Cox modeling and Kaplan Meier estimates were used for survival analyses. Results: 109 patients were enrolled. 20 patients were evaluable with complete data. Median follow up was 40 months. Primary malignancy was pancreatic adenocarcinoma in 40%, gastric adenocarcinoma in 30%, and cholangiocarcinoma in 30%. Clinical stage was I-II in 29%, III in 50%, and IV in 21%. 80% underwent chemotherapy (CTX) pre-RT and none received CTX post-RT. 80% had concurrent CTX during RT. Median RT dose was 56 Gy. Among those with pathology available (n = 12), 17% had a pathologic complete response (pCR) and 83% had R0 resection. Median ALC at each of the three study timepoints was 1.26 k/ul, 0.32 k/ul, and 1.02 k/ul, respectively. 4 (20%) patients had increased TCR repertoire as defined by the change in Gini diversity index while 16 (80%) had decreased TCR repertoire post-RT. There were no significant differences in baseline characteristics between the two cohorts. On univariable analysis, increased TCR repertoire from ALC nadir to 2 months post-RT significantly correlated with better OS (HR = 0.20 [95% CI 0.04-0.93]; p = 0.041). Other variables including ECOG, stage, RT dose, CTX use, pCR, R0 resection, baseline WBC, and ALC at three study time points did not correlate with OS. Patients with increased vs decreased TCR repertoire post-RT had a 3-year OS of 75% vs 22%, respectively (log rank p = 0.026). Conclusions: Increased TCR repertoire post-RT was an independent prognostic factor for better OS in our cohort of GI malignancies. Our findings provide biologic insight and refine the association between ALC and cancer outcomes. TCR repertoire warrants further study as a predictive and prognostic biomarker in GI malignancies. Therapeutic approaches to limit severe lymphopenia are ongoing and should include TCR assessment.

7006 Background: Loss of skeletal muscle mass (quantity) is common in patients with advanced cancer, but little is known about muscle density (quality). Hospitalized patients with advanced cancer are a highly symptomatic population at risk for the adverse effects of muscle loss. Thus, we sought to describe associations between muscle mass and density, symptom burden, health care use, and survival in these patients. Methods: We prospectively enrolled hospitalized patients with advanced cancer from 9/2014-4/2017. Upon admission, patients reported their physical (Edmonton Symptom Assessment System [ESAS]) and psychological (Patient Health Questionnaire 4 [PHQ4]) symptoms. We used computed tomography (CT) scans performed per routine care ≤45 days prior to enrollment to evaluate muscle mass and density at the level of the third lumbar vertebral body. We categorized patients as sarcopenic using validated sex specific cutoffs. We used regression models to examine associations between muscle mass and density and patients’ symptom burden, health care use, and survival. Results: Of 1,121 patients enrolled, 677 had evaluable CT scan data (mean age = 62.86±12.95 years; 51.1% female). The most common cancer types were gastrointestinal (36.8%) and lung (16.7%) cancer. Most met criteria for sarcopenia (64.0%). Older age and female sex were associated with lower muscle mass (age: B = -0.16, p < .01; female: B = -6.89, p < .01) and density (age: B = -0.33, p < 0.01; female: B = -1.66, p = .01), while higher BMI was associated with higher muscle mass (B = 0.58, p < .01) and lower muscle density (B = -0.61, p < .01). Higher muscle mass was significantly associated with improved survival (HR = 0.97, p < .01), but not with symptom burden or health care use. Higher muscle density was significantly associated with lower ESAS physical (B = -0.17, p = .02), ESAS total (B = -0.29, p < .01), PHQ4 depression (B = -0.03, p < .01) and PHQ4 anxiety (B = -0.03, p < .01) symptoms. Higher muscle density was also associated with decreased hospital length of stay (B = -0.07, p < .01), risk of readmission or death in 90 days (OR = 0.97, p < .01), and improved survival (HR = 0.97, p < .01). Conclusions: Most hospitalized patients with advanced cancer have muscle loss consistent with sarcopenia. We found that muscle mass (quantity) correlated with survival, whereas muscle density (quality) was associated with patients’ symptoms, health care use, and survival. These findings underscore the added importance of assessing muscle quality when seeking to address the adverse effects of muscle loss in oncology.

149 Background: Non-coding repeat RNAs in cancers are pervasive and “mimic” viruses with activation of pattern recognition receptors and the innate immune response. Many repeat RNAs replicate in cancer genomes through a reverse transcriptional intermediate analogous to retroviruses. Nucleoside reverse transcriptase inhibitors (NRTIs) block this retroviral life cycle to increases repeat RNAs in p53 mutant colon cancer cell cancer lines. We initiated a Phase 2 study of lamivudine (3TC) in TP53 mutant mCRC. Methods: Two-stage design with target accrual of 20 patients (pts) in stage 1 and total of 32. Eligibility: pts with p53 mutant refractory mCRC with progression on or intolerance to 5FU, oxaliplatin and irinotecan and anti-EGFR if RAS WT. RNA sequencing was performed on pre-treatment (tx) and on tx biopsy to evaluate for repeat RNA expression and expression of other genes linked to 3TC response/resistance. Radiation was allowed. 9 pts were treated with 3TC 150 mg po bid for 28-day cycles, the maximum FDA approved dose of 3TC in HIV. Subsequent pts were treated at 600 mg po bid, previously tested in P1 trials. Tumor assessments were performed every 8 weeks until documented disease progression by RECIST 1.1 criteria or drug intolerance. Results: 29/32 pts have been treated. Median age: 60 yrs. (27-82) 18 males, 11 females. 2/ 9 (22%) pts on standard 3TC dosing had stable disease (SD) on single agent 3TC with a duration of tx of 169 and 167 days, respectively. Both pts had an initial drop in CEA upon initiation of 3TC. Of the next 20 pts on high dose 3TC, 19 were evaluable. 4 had SD, for 110, 159, 130 and 228+ days. 14 pts had tx-related adverse events (TRAE). 1 pt with a definite Grade 1 TRAE (fatigue). No pts with Grade ≥3 TRAEs. We obtained pre-tx fresh frozen biopsies on 24/29 pts. Of those with SD, 4 had biopsies and differential expression identified significantly higher HSATII repeat RNA in pts with SD compared to PD. There was an association of decreased epigenetic gene expression in HSATII repeat RNA high tumors. Conclusions: This proof-of-concept study demonstrates the safety and activity of single-agent 3TC. Repeat RNA levels appear to correlate with clinical benefit and can be measured in biopsies. Further combination studies and correlatives are planned. Clinical trial information: NCT03144804.

753 Background: The use of comprehensive genomic profiling (CGP) is increasing in pancreatic ductal adenocarcinoma (PDAC) as knowledge improves regarding molecular drivers of tumorigenesis and effective targeted therapies emerge. However, adequate tissue sampling is often limited. Plasma-based CGP offers a non-invasive approach to assess biomarkers that may impact treatment decisions. Methods: We retrospectively evaluated genomic and clinical data from 97 PDAC patients with circulating tumor DNA (ctDNA) testing from 9/2016-8/2019 (Guardant Health, Inc.). ctDNA analysis included single nucleotide variants (SNV), fusions, indels and copy number variations (CNV) of up to 74 genes. ctDNA results were assessed across clinical variables. We evaluated for actionable alterations. Results: A total of 114 samples were obtained from 97 patients for ctDNA testing. ctDNA alterations were detected in 82% (93/114) of all samples, including 90% (18/20) at diagnosis, 88% (59/67) at progression, and 56% (10/18) while on stable therapy. ctDNA alterations were found at each stage of PDAC: in 25% (1/4) of samples with resectable disease, 75% (3/4) with borderline resectable disease, 82% (9/11) with locally advanced disease, and 85% (81/95) with metastatic disease. One or more KRAS alterations were detected in 55% (51/93) of patients with alterations present. The median maximum mutant allele frequency was similar between the cohort of patients with KRAS detected (0.55%) versus not detected (0.70%). 8% (8/97) of patients had potentially actionable alterations (2 activating BRAF SNVs, 1 ERBB2 CNV, 1 ERBB2 activating SNV, 1 KRAS G12C, and 3 indels in Homologous Recombination Deficiency genes). Median turnaround time was 8 days. 51% (49/97) of patients had both plasma-based CGP and tissue-based CGP. Of these patients, tissue-based CGP showed ≥ 1 alterations detected in 82% (40/49), test failure in 14% (7/49), and no alterations detected in 4% (2/49). Conclusions: Plasma-based CGP detected ctDNA alterations in 90% of samples tested at diagnosis and 82% of all samples. Potentially actionable mutations were found in 8% of patients, with prompt processing time allowing for rapid decision making.

515 Background: There is increasing use of ablative radiotherapy (RT) for oligometastatic and/or oligoprogressive cancer, but the population who may benefit from this more aggressive treatment remains poorly defined. We aimed to identify factors associated with improved outcomes following ablative RT for oligometastatic/oligoprogressive liver tumors. Methods: We retrospectively analyzed 106 patients who had tumor genomic profiling and received a 5, 6, or 15-fraction course of ablative RT for liver metastases from 2008-2019. The interval off systemic therapy post-RT was calculated for patients who did not continue treatment through RT. Overall survival (OS) was estimated using the Kaplan-Meier method. The association between clinical and genomic variables and OS were assessed using uni- and multivariable Cox regression. Results: Median follow-up was 12.6 months. Median age was 61.3 years and 57% were male. The most common primary site was colorectum (42%), followed by pancreas (25%) and non-small cell lung cancer (10%). 42% had colorectal adenocarcinoma, 46% had other adenocarcinoma, and 12% had other histology. A BRAF/RAS family mutation (KRAS, NRAS, and/or BRAF) was present in 41%, 69% had > 1 metastasis, and 38% had extra-hepatic disease. Median biological effective dose (α/β = 10) (BED) was 92 Gy. The RT field encompassed all liver metastases in 91%, and 11% received radiosensitizing chemotherapy. Median time off systemic agents was 5 months. Patients with 1 vs > 1 metastasis had a longer interval off systemic therapy (9 vs 4 months, p = 0.026). Median OS was 12.6 months. The table shows the multivariable Cox model for OS. Conclusions: Presence of a BRAF/RAS family mutation, extra-hepatic metastases, exclusion of liver metastases from RT fields, lower BED, and concurrent radiosensitizing chemotherapy were associated with worse OS. This may inform patient selection and RT delivery for aggressive local therapy for liver metastases. [Table: see text]

The acquisition of mismatch repair deficiency is common in many tumor types. A unique subtype of mismatch repair deficient tumors, called tumors with microsatellite instability (MSI), has been recently approved for treatment with PD-1\PD-L1 blockade immunotherapy. Therefore, developing a quick, inexpensive, and easy-to-implement technique for classifying tumors as MSI is of utmost importance. Here, we describe such a method using next-generation sequencing data, and show that it can detect MSI from cell-free DNA (cfDNA), even when there is low fraction of tumor DNA in the cfDNA. We first analyzed the unique mutational features of MSI vs. microsatellite stable (MSS) tumors. Using our SignatureAnalyzer tool, we analyzed the signatures of the different insertions and deletions within microsatellite loci (MS-indels), and found that MSI tumors harbor elevated MS-indels of certain types, which we further validated using whole genome sequencing data from The Cancer Genome Atlas. We next used this information to develop a method (MSIClass) that analyzes the WGS data of a given sample by capturing the features unique to MSI tumors. This method assigns each sample a score based on aggregating information from all MS loci simultaneously; since MSIClass is a sample-based tool, it can call the MSI status of tumor samples that have no patient-matched normal sample. We further demonstrate that MSIClass requires minimal genomic information; indeed, sequencing coverage as low as 0.05x (current sequencing cost of ~$2/sample) is sufficient for accurate classification by MSIClass. Using in silico and in vitro DNA mixing we show that MSIClass can accurately identify MSI cases in very low purity cases, at 0.1% tumor DNA (using 0.5x sequencing coverage). We show that MSIClass, applied on cfDNA, could accurately classify 29 colon cases -- 22 MSS and 7 MSI cases, 3 of which are early stage disease and with low tumor content. We then applied MSIClass to ~2000 cfDNA samples from various tumor types and found MSI cases in tumors that are not associated with MSI such as breast. These results demonstrate that MSIClass is an inexpensive MSI test that can be applied in various cases: bulk sequencing where a matched normal is not available, cases of very low purity tumors, and even using cfDNA from blood in cases a biopsy is not available or possible. Finally, this methodology can be the basis for monitoring minimal residual disease, and potential be used for early detection of MSI tumors. Citation Format: Yosef E. Maruvka, Ruslana Frazer, Jonna Grimsby, Emily E Van Seventer, Carrie Cibulskis, Viktor Adalsteinsson, Ryan Corcoran, Gad Getz. MSIClass - Identification and classification of microsatellite unstable tumors using cell-free DNA from ultra low pass sequencing [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr A094. doi:10.1158/1535-7163.TARG-19-A094

Purpose Third-generation epidermal growth factor receptor (EGFR) inhibitors like nazartinib are active against EGFR mutation–positive lung cancers with T790M-mediated acquired resistance to initial anti-EGFR treatment, but some patients have mixed responses. Methods Multiple serial tumor and liquid biopsies were obtained from two patients before, during, and after treatment with nazartinib. Next-generation sequencing and droplet digital polymerase chain reaction were performed to assess heterogeneity and clonal dynamics. Results We observed the simultaneous emergence of T790M-dependent and -independent clones in both patients. Serial plasma droplet digital polymerase chain reaction illustrated shifts in relative clonal abundance in response to various systemic therapies, confirming a molecular basis for the clinical mixed radiographic responses observed. Conclusion Heterogeneous responses to treatment targeting a solitary resistance mechanism can be explained by coexistent tumor subclones harboring distinct genetic signatures. Serial liquid biopsies offer an opportunity to monitor clonal dynamics and the emergence of resistance and may represent a useful tool to guide therapeutic strategies.

Clonal heterogeneity associated with acquired resistance presents a critical therapeutic challenge. Whole-exome sequencing of paired tumor biopsies and targeted sequencing of cell-free DNA (cfDNA) from BRAF(V600E) colorectal cancer patients receiving BRAF inhibitor combinations identified 14 distinct alterations in MAPK pathway components driving acquired resistance, with as many as eight alterations in a single patient. We developed a novel pooled clone system to study clonal outgrowth during acquired resistance, in vitro and in vivo. In vitro, the dynamics of individual resistant clones could be monitored in real-time in cfDNA isolated from culture media during therapy. Outgrowth of multiple resistant clones was observed during therapy with BRAF, EGFR, and MEK inhibitor combinations. However, ERK inhibition, particularly in combination with BRAF and EGFR inhibition, markedly abrogated clonal outgrowth in vitro and in vivo. Thus, convergent, upfront therapy may suppress outgrowth of heterogeneous clones harboring clinically observed resistance alterations, which may improve clinical outcome.

Treatment with immune checkpoint blockade (CPB) therapies often leads to prolonged responses in patients with metastatic melanoma, but the common mechanisms of primary and acquired resistance to these agents remain incompletely characterized and have yet to be validated in large cohorts. By analyzing longitudinal tumor biopsies from 17 metastatic melanoma patients treated with CPB therapies, we observed point mutations, deletions or loss of heterozygosity (LOH) in beta-2-microglobulin (B2M), an essential component of MHC class I antigen presentation, in 29.4% of patients with progressing disease. In two independent cohorts of melanoma patients treated with anti-CTLA4 and anti-PD1, respectively, we find that B2M LOH is enriched threefold in non-responders (~30%) compared to responders (~10%) and associated with poorer overall survival. Loss of both copies of B2M is found only in non-responders. B2M loss is likely a common mechanism of resistance to therapies targeting CTLA4 or PD1., Resistance to immune-checkpoint blockade often occurs in treated patients. Here, the authors demonstrate that B2M loss is a mechanism of primary and acquired resistance to therapies targeting CTLA4 or PD-1 in melanoma patients.

IL-4 drives expansion of Th2 cells that cause generation of alternatively activated macrophages (AAMs). Filarial infections are established early in life, induce increased IL-4 production are co-endemic with tuberculosis (TB). We sought to understand, therefore, how mycobacteria are handled in the context of IL-4-induced AAM. Comparing IL-4 generated in vitro monocyte derived human AAMs to LPS and IFN-γ generated classically macrophages (CAMs), both infected with mycobacteria (BCG), we demonstrated increased early BCG uptake and increased IL-10 production in AAMs compared to CAMs. We further demonstrated that increased IL-10 production is mediated by upregulation of tumor progression locus 2 (TPL-2), an upstream activator of extracellular signal related kinases (ERKs) in AAMs but not in CAMs, both at the transcript as well as the protein level. Pharmacologic inhibition of TPL-2 significantly diminished IL-10 production only in BCG-infected AAMs. Finally, we validated our findings in an in vivo C57Bl/6 model of filarial infection, where an exaggerated Th2 induced lung-specific alternative activation led to TPL-2 and IL-10 upregulation on subsequent TB infection. These data show that in response to mycobacterial infection, IL-4 generated AAMs in chronic filarial infections have impaired immune responses to TB infection by increasing IL-10 production in a TPL-2 mediated manner.

The inevitable emergence of acquired resistance is a major limitation to the clinical benefit of precision medicine strategies. Single-lesion tumor biopsies have long been the mainstay of understanding acquired resistance, but recent data suggest tumor biopsies may under-represent the molecular heterogeneity of acquired resistance. Alternatively, studies have suggested that liquid biopsy approaches analyzing cell-free DNA (cfDNA) may offer significant advantages, but extensive prospective comparisons of matched liquid vs. tumor biopsies obtained at the time of acquired resistance are lacking. Here, we assess systematic liquid biopsy upon acquired resistance to targeted therapy in 44 patients across seven molecularly defined subtypes of gastrointestinal cancers. Liquid biopsy at disease progression identified at least one functionally validated molecular mechanism of resistance in 75% of patients, wherein 52% exhibited >1 resistance alteration (range 2-9, median 3 per patient). In 23 patients in whom a matched post-progression tumor biopsy could be obtained, tumor biopsy was less effective than liquid biopsy in identifying resistance mechanisms, with resistance alterations detected in only 48% of patients, and multiple resistance mechanisms detected in only 9% of cases. In matched cases, liquid biopsy detected at least one resistance alteration not detected in tumor biopsy in 78% of cases. Targeted analysis and whole-exome sequencing of serial cfDNA, multiple post-progression biopsies, and rapid autopsy specimens from select cases revealed key insights into the geographic and complex characteristics of heterogeneity captured by liquid biopsy in the setting of acquired resistance. These data illustrate that acquired resistance is characterized by frequent and profound tumor heterogeneity, and suggests that liquid biopsy may more effectively identify heterogeneous clinically relevant resistance alterations compared to standard tumor biopsy. Citation Format: Aparna R. Parikh, Ignaty Leshchiner, Liudmila Elagina, Lipika Goyal, Chaya Levovitz, Giulia Siravegna, Dimitri Livitz, Kahn Rhrissorrakrai, Liz Martin, Emily E. Van Seventer, Megan Hanna, Kara Slowik, Filippo Utro, Christopher J. Pinto, Alicia Wong, Brian P. Danysh, Ferran Fece de la Cruz, Isobel J. Fetter, Brandon Nadres, Heather A. Shahzade, Jill N. Allen, Lawrence S. Blaszkowsky, Jeffrey W. Clark, Bruce Giantonio, Janet E. Murphy, Ryan D. Nipp, Eric Roeland, David P. Ryan, Colin D. Weekes, Eunice L. Kwak, Jason E. Faris, Francois Aguet, Ipsita Guha, Mehlika Hazar-Rethinam, Dora Dias-Santagata, David T. Ting, Andrew X. Zhu, Theodore S. Hong, Todd R. Golub, A J. Iafrate, Viktor Adalsteinsson, Alberto Bardelli, Laxmi Parida, Dejan Juric, Gad Getz, Ryan B. Corcoran. Liquid biopsy versus tissue biopsy to assess acquired resistance and tumor heterogeneity in gastrointestinal cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr LB-257.

3563 Background: Colorectal cancer (CRC) incidence in patients younger than 50 years of age is steadily rising by 2% annually. Early-onset CRC usually presents with more aggressive features; however, data on prognosis are widely conflicting. Clinicians may hold an age-related bias in treating younger patients, but this proclivity and its effects have not been quantified. Methods: Patients with a history of metastatic CRC who consented to a departmental chart review protocol were collected between 2014 and 2018 at Massachusetts General Hospital. The cohort was divided into two groups based on age at initial diagnosis: < 50 and ≥50. Data were gathered on treatments and clinicopathological features. A log-rank test compared survival from the diagnosis of metastatic disease between age groups. The distributions of clinicopathological features were compared using Wilcoxon rank sum tests. Results: 464 metastatic CRC patients were identified. 155 patients (33%) were < 50 (median age 43, 49% female) and 309 patients (67%) were ≥50 (median age 61, 45% female). Sex did not significantly differ between the two groups (p = 0.45). Patients < 50 received more lines of therapy after metastatic diagnosis than patients ≥50 (mean 2.7 v. 2.2; p = 0.002). Younger patients also received more resections of distant metastases (mean 0.62 v. 0.48; p = 0.01). A higher rate of enrollment in clinical trials for patients < 50 approached significance (p = 0.06). Even so, patients < 50 did not see a significant survival benefit over older patients (2/5-year survival from metastatic diagnosis 77%/47% v. 73%/38%, p = 0.23). Patients < 50 had a lower proportion of right-sided tumors (p = 0.0002) and BRAF mutations (p = 0.0009). There was no difference in MSI status (p = 0.28), RAS mutational status (p = 0.40), mucinous features (p = 0.53), or signet ring features (p = 0.26). Conclusions: Overall survival in patients < 50 is similar to patients ≥50, despite patients < 50 receiving more aggressive therapy. Further study is warranted to better understand these differences. Potential areas of interest include performance status, age-related treatment bias, and biological factors.

3590 Background: While tissue-based assays have yields above 90% in solid tumors, there is less known about factors that influence the sensitivity of ctDNA for detecting mutations. Methods:We retrospectively evaluated mCRC patients (pts) who had plasma-derived NGS utilizing a highly-sensitive targeted 68-73-gene ctDNA assay. In a case-control design, pts with a known mutation on tissue and radiologic evidence of metastatic disease but no detectable ctDNA mutation were matched 1:3 with randomly selected pts with detectable mutations and compared according to clinical, laboratory, and radiologic characteristics. A prediction score for ctDNA detection was built using a binary logistic backward stepwise regression analysis and tested in two independent data sets from different institutions. Area under the curve (AUC) from receiver operating characteristics curves (ROC) were used for internal and external validation. Results: From 416 pts who met inclusion criteria, plasma-derived NGS did not find tumor mutations in 66 cases (15.9%); 198 pts with detectable alterations were selected as controls. After multivariate analysis, the detection of ctDNA was associated with increasing age (OR 1.05; 95%CI 1.02-1.09; p = .001), presence of liver (OR 5.82; 95%CI 2.55-12.49; p < .001) and lymph node metastases (OR 3.28; 95%CI 1.51-7.60; p = .004), archival TP53 mutations (OR = 2.88; 95%CI 1.37-6.17; p = .006). A key determinant was timing of collection relative to disease status: plasma collected in newly diagnosed metastatic disease or after evidence of progression was substantially more likely to have detectable alterations (OR 9.24; 95%CI 4.11-22.40; p < .001); The simplified prediction model performed well in internal (AUC = 0.88) and external validation (AUC = 0.95; 163 pts). Conclusions: Our validated prediction model provides clinicians and researchers with a tool to screen for patients in whom ctDNA testing can outperform tissue-based testing in detecting genomic alterations.

3602 Background: ctDNA identifies patients (pts) at high risk for disease recurrence post CRC resection (post-op). Current ctDNA residual disease detection approaches assess only genomic alterations (alts) and rely on tissue sequencing to identify tumor-derived alts. We evaluated a plasma-only ctDNA assay to identify high risk pts. Methods: 72 CRC pts (surgery only = 42; adjuvant therapy (adj) = 30) had post-op and/or post-adj plasma samples (3-4mL). Extracted cfDNA (median 27 ng) was analyzed using a single-sample NGS test validated in early stage CRC that integrates assessment of genomic alts with epigenomic cancer signature (Guardant Health, CA). A variant classifier was applied to differentiate tumor-derived from non-tumor derived alts in a tumor tissue-uninformed approach. Results: In the surgery cohort, samples were collected a median of 31 days (d) post-op. 7/8 pts with post-op ctDNA detected (ctDNA+) recurred (PPV 88%; median time to recurrence (mTTR) 248d). The recurrence-free pt has < 180d follow-up. 7/34 pts without ctDNA detected (ctDNA-) recurred (NPV 79%; mTTR 333d). 1/1 Stage 0-II ctDNA+ pt recurred (PPV 100%; TTR 440d) while 1/20 ctDNA- recurred (NPV 95%; TTR 440d). 27 pts in the adj cohort had samples collected a median of 37d post-adj. 6/6 ctDNA+ pts recurred (PPV 100%, mTTR 239d). 4/21 ctDNA- pts recurred (NPV 81%, mTTR 466d). 2/2 ctDNA+ and 0/11 ctDNA- Stage III pts recurred (PPV, NPV 100%, mTTR 420d). All 3 post-op ctDNA+/post-adj ctDNA+ (ctDNA persistence) pts recurred. 1/2 post-op ctDNA+/post-adj ctDNA- (ctDNA clearance) pts is recurrence free (306d). 2 post-op ctDNA-/post-adj ctDNA+ pts recurred. In the entire cohort, ctDNA+ after standard therapy completion had a recurrence PPV 93%, NPV 80%, HR 11.29 (p < 0.0001). Conclusions: In post-op CRC, ctDNA detection utilizing a tumor-uninformed integrated genomic and epigenomic assay has high recurrence PPV and NPV following standard therapy completion. ctDNA identifies pts who may benefit from post-op adj therapy or additional/modified post-adj therapy. These findings demonstrate that ctDNA detection from a single post-op/post-adj plasma sample stratifies high/low risk pts and informs therapy decision making.

3057 Background: ctDNA sequencing has been rapidly adopted for the identification of targetable somatic alterations (alts) in patients with advanced cancers. However, early stage disease detection has been hindered by low levels of ctDNA in circulation and the presence of confounding non-tumor-related somatic alts. We developed and validated a ctDNA assay that combines somatic and epigenomic signals to detect early stage tumors without tumor tissue or white blood cells (WBC). Methods: Using a single input sample, our assay integrates the sensitive detection of genomic alts with quantification of epigenomic signals associated with cancer. Non-tumor alts (e.g., clonal hematopoiesis of indeterminate potential; CHIP) are excluded using a newly developed bioinformatic classifier. To assess analytical sensitivity, specificity, and positive and negative reproducibility, we tested 337 clinical and contrived samples. Results: Clinical specificity was determined using 80 plasma samples from 50-75 year old presumptive cancer-free donors, and resulted in a single false positive (99% specificity). Analytical sensitivity (limit of detection) was established using a dilution series of 4 different late stage CRC pts tested in triplicate at a clinically relevant DNA input (30 ng) across multiple batches. 100% sensitivity was maintained even at the lowest tested level (estimated 0.1% tumor level). Positive/negative reproducibility was assessed by testing triplicates of diluted late stage samples and age-matched healthy donors, respectively, across different cfDNA inputs, and multiple reagent lots. Both positive and negative calls were 100% concordant across all replicates. Independent estimation of tumor levels from epigenomic or genomic signals produced highly concordant results (correlation r-value: 0.82, p-value: 3e-16). Conclusions: We designed and validated a highly specific ctDNA assay that integrates both genomic and epigenomic signals to allow for accurate and quantitative tumor level detection in early stages of the disease without requiring tumor tissue or WBC.

3514 Background: mCRC remains a lethal cancer and immunotherapy in MSS mCRC has yet to show significant activity. In preclinical models, radiation induced cellular damage may increase responsiveness to immunotherapy via the abscopal effect, with evidence for synergy between radiation therapy (RT) and dual checkpoint blockade. In this study, we assessed dual blockade of CTLA-4 and PD-1 combined with RT as a strategy to stimulate an immune response for patients with MSS mCRC. Methods: In this open-label, single arm phase-2 study, we enrolled 40 MSS mCRC patients (pts). Eligible pts had histologically-confirmed MSS mCRC, ECOG PS 0-1, and progression on at least two lines of therapy. Treatment (Tx) consisted of ipilimumab (1mg/kg q6 weeks), nivolumab (240 mg q2 weeks) and 3 fractions of 8 Gy of RT at cycle 2 every other day. Tx continued until progression, discontinuation or withdrawal. The primary endpoint was Disease Control Rate (DCR), with radiological evaluations every 3 months. Exploratory endpoints included ORR, PFS, OS and safety. Response was defined as disease control outside the radiation field. We obtained serial tumor biopsies pre-tx, during checkpoint blockade alone (cycle 1) and 2 weeks after initiation of radiation. Intention-to-treat analysis (ITT) includes all pts receiving at least one dose of study agent. Results: 40 pts (median age 59 years (26-83), 58% male) enrolled and started treatment from 7/2017 to 12/2018. DCR was 17.5% (7/40) with a 7.5% (3/40) ORR by ITT. Median duration of disease control was 77 days in the ITT; 252 days for those with disease control (n=7) based on the first re-staging scans at 3 months or censored (n=3) and 70.5 days for pts with PD (n=17) or who did not receive RT due to clinical progression (n=13). In the modified ITT (all pts receiving RT and restaged), N=24 pts, excluding 1 pt pending 1st scans post-RT, DCR was 29.2% (7/24) and ORR 12.5% (3/24). Median duration of disease control in mITT was 77.5 days: 252 days for those with disease control and 77 days for those with PD. TRAEs were reported in 22/40 (55%). 20/40 (50%) with grade ≥3 toxicities, with fatigue, nausea, vomiting, diarrhea, infusion-related reaction and dyspnea being the most common. 1(2%) pt died of respiratory failure possibly related to tx. Conclusions: Dual blockade of CTLA-4 and PD-1 with RT is feasible and demonstrates durable activity in pts with MSS mCRC. There are 3 pts who have not completed RT or had their post-RT re-staging. We will report updated efficacy data and outcomes from correlative serial tumor biopsies upon trial completion. Clinical trial information: NCT03104439.

562 Background: This study was designed to assess the capability of perioperative ctDNA analysis to predict surgical outcome and recurrence following neoadjuvant CRT for LARC. Methods: Thirty-one patients (pts) with newly diagnosed LARC ( n = 29) or locally recurrent (non-metastatic) rectal cancer ( n= 3) were treated between 7/2013 - 7/2017. Pts received long-course neoadjuvant CRT prior to surgical resection: 50.4 Gy/28 fractions. Serum ctDNA was typically measured at baseline, weekly during CRT, pre-operatively, and post-operatively. Next-generation sequencing was used to identify mutations in the primary tumor, and mutation-specific droplet digital PCR was used to detect mutation fraction in ctDNA. Results: The median age of the cohort was 53 years (IQR 46.5-65.3 y). The overall R0-node negative (R0-NN) resection rate was 66.7%. The rate of R0-NN resection was significantly higher among pts with undetectable preoperative ctDNA ( n = 17, 88%) compared to pts with a detectable preoperative ctDNA ( n = 10, 30%, Fisher’s exact p = 0.036). The overall pathologic complete response (pCR) rate was 18.5%. The pCR rate among pts with undetectable preoperative ctDNA was 23.5% vs 10% among pts with detectable preoperative ctDNA ( ns). For patients with a positive assay at baseline and weekly draws available, those who went on to have an R0NN resection were observed to exhibit ctDNA clearance by the second week of chemoradiation. Recurrence free survival (RFS) was calculated for the subset ( n= 22) who were able to undergo surgery and had post-operative ctDNA available. Pts with detectable postoperative ctDNA experienced poorer RFS than those with undetectable ctDNA, HR 8.0, p < 0.001. Conclusions: Undetectable preoperative ctDNA is associated with R0-NN surgical outcome, and detectable postoperative ctDNA is associated with worse RFS in a cohort of pts treated with neoadjuvant CRT for LARC. Validation is ongoing in a larger cohort of 39 pts.

659 Background: BRAF non-V600 mutations occur in 2-3% of colorectal cancer. These mutants can be classified as RAS independent (class 2) or RAS dependent (class 3). We reported BRAF non-V600 mutations could be a negative predictive factor for anti-EGFR therapy in patients (pts) with pretreated metastatic colorectal cancer (mCRC), while mCRC pts with class 3 BRAF mutations could respond to anti-EGFR therapy because of its dependency on receptors and RAS. Methods: This study evaluates the efficacy to anti-EGFR therapy in a large cohort of pts with BRAF non-V600 mutated mCRC. Pts with mCRC referred to the participating centers from 2010 to 2017 were included. Clinicopathological features, efficacy of anti-EGFR therapy, and survival outcomes were stratified by BRAF mutational class. Results: One hundred seventeen pts with BRAF non-V600 mutated mCRC were identified. Median age was 58 years (range, 27-83), 68 pts (58%) were male, and 38 pts (33%) had right-sided tumors. Mucinous histology was seen in 11 cases (9%); concurrent RAS mutations occurred in 31 cases (27%), and 3 cases (3%) were MSI-H. Also, TP53 mutations were detected in 74 pts among 90 analyzed cases (82%). Regarding BRAF mutation subtype, 25/63/29 pts were classified as class 2/3/not reported (NR), respectively. Median OS in RAS wild-type/mutant were 44.8/34.6 months, respectively (p=0.082). The median OS in RAS wild-type pts with BRAF non-V600 mutations for class 2, 3, and NR were 25.7, 44.2, and 79.1 months, respectively (class 2 vs. 3, p=0.219). Among 40 pts treated with anti-EGFR therapy, response rates were 14%, 44%, and 40% for class 2, 3, and NR, respectively. Median PFS was 4.4, 8.3, 4.0 months for class 2, 3, and NR, respectively. Moreover, in 25 pts receiving anti-EGFR therapy as third or later line, response rate was 0%, 27%, and 50% in class 2, 3, and NR, and median PFS was 2.8, 3.7, and 4.0 months (p=0.762), respectively. Conclusions: Pts with class 2 BRAF mutations tend to have a poor prognosis compared to those with class 3 mutations. While almost half of pts with class 3 BRAF mutations responded to anti-EGFR therapy, response was rare for pts with class 2 BRAF mutations, and none achieved objective response in the third or later line.

391 Background: Pancreatic ductal adenocarcinoma (PDAC) remains a highly lethal cancer. Immunotherapy (IO) has shown minimal activity. In preclinical models, radiation (XRT) increases likelihood of response to IO via an abscopal effect where local tx (treatment) of a tumor leads to an antitumor response distantly, with synergy between XRT and dual checkpoint blockade. In this study, we assessed CTLA-4 and PD-1 blockade with XRT as a strategy to stimulate an immune response for patients (pts) with PDAC. Methods: In this open-label, single arm phase-2 study, we enrolled 25 metastatic PDAC pts in an exploratory cohort. Eligible pts had histologically-confirmed PDAC, ECOG PS 0-1, and progression on at least 1 line of tx. Tx consisted of Ipilimumab (1 mg/kg every 6 wks), Nivolumab (240 mg every 2 wks) and 3 fractions of 8 Gy of XRT at cycle 2. Tx continued until PD, discontinuation or withdrawal. Endpoints include Disease Control Rate (DCR), ORR, PFS, OS and safety. Radiological evaluations were every 3 months. Response was defined as disease control outside of the radiation field. We obtained serial tumor biopsies pre-tx, during checkpoint blockade alone and 2 weeks after XRT. Intention to treat analysis includes all pts receiving at least one dose of study tx. Results: 22 pts were enrolled and evaluable from 6/2017-6/2018, median age 60 years (32-75), 73% male and 100% MSS. DCR was 27% and ORR was 14% with 1 pt having a complete response. All responses were out of the radiation field. Median PFS was 76 days in the entire cohort; 163 days for pts with disease control vs 62.5 days for pts with PD or who came off study prior to initial imaging. 7 pts did not receive XRT due to clinical progression. Treatment-related adverse events (AEs) were reported in 12/22 pts (54.5%). 8/22 pts (36.4%) experienced grade ≥ 3 toxicities. Elevated lipase, lymphopenia, fatigue, hyperglycemia, mucositis and hepatitis were the most common AEs. 1/22 (4.5%) pt had a grade 5 AE possibly related to tx. Conclusions: Dual blockade of CTLA-4 and PD-1 with XRT is feasible and demonstrates promising activity in pts with metastatic PDAC. We will report the updated efficacy and safety data as well as outcomes from the correlative serial biopsies upon completion. Clinical trial information: NCT03104439.

Genetic alterations in the fibroblast growth factor receptor (FGFR) pathway are promising therapeutic targets in many cancers, including intrahepatic cholangiocarcinoma (ICC). The FGFR inhibitor BGJ398 displayed encouraging efficacy in patients with FGFR2 fusion–positive ICC in a phase II trial, but the durability of response was limited in some patients. Here, we report the molecular basis for acquired resistance to BGJ398 in three patients via integrative genomic characterization of cell-free circulating tumor DNA (cfDNA), primary tumors, and metastases. Serial analysis of cfDNA demonstrated multiple recurrent point mutations in the FGFR2 kinase domain at progression. Accordingly, biopsy of post-progression lesions and rapid autopsy revealed marked inter- and intralesional heterogeneity, with different FGFR2 mutations in individual resistant clones. Molecular modeling and in vitro studies indicated that each mutation led to BGJ398 resistance and was surmountable by structurally distinct FGFR inhibitors. Thus, polyclonal secondary FGFR2 mutations represent an important clinical resistance mechanism that may guide the development of future therapeutic strategies. Significance: We report the first genetic mechanisms of clinical acquired resistance to FGFR inhibition in patients with FGFR2 fusion–positive ICC. Our findings can inform future strategies for detecting resistance mechanisms and inducing more durable remissions in ICC and in the wide variety of cancers where the FGFR pathway is being explored as a therapeutic target. Cancer Discov; 7(3); 252–63. ©2016 AACR. See related commentary by Smyth et al., p. 248. This article is highlighted in the In This Issue feature, p. 235

Background: Patients with gastrointestinal (GI) malignancies undergoing treatment are surveyed for disease relapse with radiographic imaging and measurement of standard tumor markers. However, there is a need for more sensitive techniques to detect early cancer progression. Measurement of circulating tumor DNA (ctDNA) is emerging as an important diagnostic tool that can provide real-time information about disease response during treatment. In this study, we evaluated whether dynamic monitoring of ctDNA may predict cancer progression earlier than standard methods in patients with GI malignancies undergoing targeted therapy. Methods: Tumor and blood specimens were obtained from patients treated at the Massachusetts General Hospital (MGH) under IRB-approved studies. Tumor biopsies were performed at the time of diagnosis and were subjected to the MGH standard molecular diagnostics panel for 104 known cancer genes. Blood samples for ctDNA analysis were collected at baseline at the start of targeted therapy, and one more more driver mutations were followed longitudinaly at two week intervals starting one month after treatment. ctDNA was extracted from plasma using QIAGEN-based protocols and amplified by digital droplet PCR (ddPCR) using primers for tumor-specific point mutations. Tumor markers and CT scans were part of routine clinical care and procured using standard procedures at MGH. RECIST measurements, if available, were obtained in a blinded fashion using the Tumor Imaging Metrics Core. Results: We studied 44 patients with GI malignancies undergoing treatment with targeted therapies. Represented cancer types included colorectal (53%), gastroesophageal (21%), biliary (20%), and pancreatic cancers (6%). We correlated the % change in ctDNA mutant allele fraction (MAF) or standard tumor markers 4 weeks after treatment initiation with radiographic response by RECIST. Tumor-specific MAF decreased by 88% in patients who went on to show partial response (PR) or stable disease (SD), compared to an increase by 51% in patients shown to have progressive disease (PD) (p=1.8x10-5). All patients who achieved PR, except one, had a reduction in ctDNA levels by at least 93% or more. In contrast, standard tumor markers did not predict response as there was no significant difference in the % change at 4 weeks in the PR/SD group compared to PD (p=0.64). Conclusions: Real-time monitoring of patient-specific tumor mutations by ctDNA analysis has the potential to predict response to targeted therapy earlier than standard surveillance mechanisms. Integration of liquid biopsies into medical decision-making may allow for earlier identification of cancer progression and consequent modification of therapeutic approach. However, more studies are needed to determine whether incorporation of ctDNA analysis into clinical practice would impact patient outcomes. Citation Format: Jaime L. Schneider, Aparna Parikh, Mehlika Rethinam, Brandon Nadres, Emily Van Seventer, Heather Shahzade, Ryan B. Corcoran. Circulating tumor DNA as a tool for predicting response to targeted therapy in gastrointestinal malignancies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1575.

272 Background: Curative resection is possible following neoadjuvant treatment for LAPC, yet there is no method to identify patients beforehand who will have a favorable surgical outcome. This study was designed to assess the ability of ctDNA measured during neoadjuvant chemoradiation (CRT) to predict surgical outcome for LAPC. Methods: 38 LAPC pts were enrolled at our institution between 10/2015 - 5/2017. Pts received neoadjuvant FOLFIRINOX followed by CRT: either short-course ( n = 9, 25 Gy/5 fractions), or long-course ( n = 29, 50.4 Gy/28 fractions). Serum ctDNA was measured at baseline, weekly during CRT, and preoperatively. After extracting DNA from plasma, a tumor mutation specific droplet digital PCR assay was used to detect the fraction of ctDNA molecules. The following clinical and pathologic outcomes were noted: CA19-9, CEA, RECIST score, tumor grade, T stage, tumor regression grade (TRG), R-resection status, pathologically involved lymph nodes, LVI, and PNI. Results: The median age of the cohort was 67 years (range 42-85 years). Following CRT, 32 (84%) were operable. The overall R0-node negative (R0-NN) resection rate was 66% for the entire cohort. The rate of R0-NN resection was significantly higher among patients with an undetectable preoperative ctDNA ( n= 16) compared to those with a detectable ( n= 22) preoperative ctDNA (88% R0-NN vs 50% R0-NN, respectively, Fisher’s exact p = 0.036). On univariate logistic regression, only ctDNA status was significantly associated with R0-NN resection (p = 0.025), whereas preoperative RECIST score, CA19-9, and CEA were not associated. On multivariable logistic regression, ctDNA remained a borderline significant predictor for R0-NN resection when controlling for CA19-9 ( p = 0.058). For patients who received surgery, the ctDNA allele fraction was significantly correlated with TRG ( Pearson R = 0.399, p = 0.024). Conclusions: Undetectable preoperative ctDNA is associated with R0-NN surgical outcome in a cohort of patients treated with neoadjuvant CRT for LAPC. This approach is worthy of further study to establish guidelines for incorporating ctDNA into clinic with the goal of improving patient selection for surgery.

Clonal heterogeneity associated with the emergence of acquired resistance to therapy presents a critical challenge for therapeutic strategies to overcome resistance. We investigated the molecular landscape of acquired resistance in BRAF-mutant colorectal patients treated with BRAF inhibitor combinations. Through whole-exome sequencing of paired pre-treatment and post-progression tumor biopsies and targeted sequencing of pre-treatment and post-progression plasma circulating tumor DNA (ctDNA), we identified 14 unique alterations in MAPK pathway components driving acquired resistance, affecting KRAS, NRAS, BRAF, MEK1 and MEK2. Analysis of ctDNA at the time of disease progression revealed profound tumor heterogeneity associated with acquired resistance, with multiple concurrent resistance alterations detectable in ctDNA in individual patients, with one patient harboring as many as 8 co-existing resistance alterations. These findings necessitate development of a strategy capable of simultaneously overcoming multiple heterogeneous resistance mechanisms. To evaluate potential strategies for convergent targeting of multiple concurrent resistance alterations, we generated individual resistant models harboring the full spectrum of clinically observed mutations driving acquired resistance. Individually, these alterations drove resistance to BRAF inhibitor combinations currently in clinical trials by maintaining MAPK signaling. However, since acquired resistance is thought to arise from pre-existing clones that emerge during treatment, we developed a novel pooled clone model system to study clonal outgrowth under the selective pressure of therapy. In this system, each resistant clone was pooled at an abundance of 1% in a background of the sensitive parental cells and exposed to an array of potential therapies, both in vitro, and in vivo as xenografts. The change in clonal abundance from baseline to the completion of therapy was assessed for each clone by digital PCR to measure the degree of clonal outgrowth. Moreover, in vitro, we were able to monitor clonal dynamics in real-time by isolating cell-free DNA (cfDNA) from the cell culture media every 3-4 days during therapy. We observed rapid outgrowth of resistant clones during BRAF-EGFR and BRAF-MEK therapy, and delayed, but robust outgrowth during BRAF-MEK-EGFR therapy, all of which are therapies that have been evaluated in recent clinical trials. However, ERK inhibitor alone, and to a greater degree BRAF-ERK and BRAF-ERK-EGFR combinations markedly abrogated the clonal outgrowth of resistant clones. Moreover, in xenograft tumors derived from clonal pools, BRAF-ERK-EGFR triple combination resulted in profound tumor regressions and completely prevented the outgrowth of all resistant clones. In conclusion, we observed the potential for profound heterogeneity of acquired resistance mechanisms in BRAF-mutant colorectal cancer patients, with multiple alterations observed in ctDNA from individual patients. Our data suggest that convergent, upfront therapy with RAF-ERK or RAF-ERK-EGFR inhibitor combinations may suppress outgrowth of clones harboring clinically observed resistant alterations, offering the potential for improved clinical outcome. Citation Format: Mehlika Hazar-Rethinam, Marianna Kleyman, G. Celine Han, David Liu, Leanne G. Ahronian, Heather A. Shahzade, Lifeng Chen, Aparna R. Parikh, Jill N. Allen, Jeffrey W. Clark, Eunice L. Kwak, Jason E. Faris, Janet E. Murphy, Theodore S. Hong, Emily E. Van Seventer, Brandon Nadres, Catriona B. Hong, Joseph M. Gurski Jr., Nicholas A. Jessop, Dora Dias-Santagata, A. John Iafrate, Eli M. Van Allen, Ryan B. Corcoran. Modeling convergent therapeutic strategies to overcome the heterogeneity of acquired resistance in BRAF-mutant colorectal cancer [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2017 Oct 26-30; Philadelphia, PA. Philadelphia (PA): AACR; Mol Cancer Ther 2018;17(1 Suppl):Abstract nr LB-A34.

833 Background: Changes in ctDNA and serum TMs (CEA and CA19-9) can serve as predictors of response to systemic therapy in GI cancer patients (pts). Similarly, PROs correlate with survival and treatment response. We present a preliminary analysis of ctDNA, TMs, and PROs in predicting treatment response. Methods: We are enrolling 200 pts in a prospective study with metastatic pancreatic (PDAC), colorectal (CRC), gastroesophageal (GE), and biliary cancers. We are collecting ctDNA, TMs (CEA for all tumor types; CA19-9 for PDAC, GE, biliary), and PROs (FACT-G for QOL [higher scores indicate better QOL]; ESAS-r and PRO-CTCAE for symptoms; and PHQ-4 [consists of GAD-2 and PHQ-2 for anxiety and depression]; higher ESAS-r, PRO-CTCAE, and PHQ-4 scores reflect greater symptom burden) at baseline and 4 weeks. ctDNA is benchmarked against somatic tissue alterations, and serially assessed by digital droplet PCR. We correlated median percent change from baseline to 4 weeks for ctDNA, TMs, and PROs with treatment response (clinical benefit [CB], progressive disease [PD]). Results: From April to August 2019, we have enrolled 38/45 (84.4%) eligible pts (median age = 64 years; 36.8% female). Among these 38 pts, tumor types are PDAC (36.8%), CRC (31.6%), GE (28.9%), and biliary (2.6%). 18/38 pts were evaluable for ctDNA. Change in ctDNA was -94.5% in pts with CB (n = 10) and -19.5% in pts with PD (n = 8; p = 0.025). No correlation was observed between CEA and treatment response (p = 0.367). Change in CA19-9 was -1.5% for pts with CB and +47% for pts with PD (p = 0.019). Changes in PRO-CTCAE (p = 0.345), GAD-2 (p = 0.697), and ESAS scores (p = 0.743) did not differ between pts with CB and PD. However, changes in PHQ-2 (CB 0% v. PD +22.5%; p < 0.001), PHQ-4 (CB -8.5% v. PD +5%; p = 0.015), and FACT-G (CB +30% v. PD +5%; p = 0.049) were significant. Conclusions: Preliminary analysis suggests that ctDNA and PROs demonstrate promising utility for early prediction of treatment response, with favorable performance relative to standard TMs. Further analyses of larger pt numbers in this ongoing study may clarify the use and integration of these measures to better predict pt outcomes.

393 Background: Nearly half of patients with G/GE cancer do not receive or complete post-operative chemotherapy and/or chemoradiation (CRT). Total neoadjuvant therapy (TNT) is as an emerging alternate treatment strategy. We have previously reported a 28% pCR with FOLFIRINOX followed by CRT. However, TNT outcomes with FLOT or FOLFOX followed by CRT are lacking. Methods: We retrospectively analyzed patients after resection of locally advanced G/GE after receiving TNT. Patient received neoadjuvant FOLFOX or FLOT x 8 cycles, CRT (G 45 Gy, GE 50.4 Gy) with concurrent chemotherapy (5FU, carboplatin/paclitaxel). The primary aim was to explore TNT completion rates. Secondary aims included pCR and toxicity. We performed descriptive statistics, t-test, chi-squared, and Fisher’s exact tests as appropriate. Results: From 12/2015 to 8/2019, 57.1% (40/70) completed TNT and resection (15.7% active treatment, 15.7% progressive disease, 11% treated elsewhere). Median age was 66.0 (range:27-79) and 73% male. Tumor locations included 57.5% G, 30.0% GE, and 12.5% overlapping. Neoadjuvant chemotherapy included FLOT 22.5% (n = 9) or FOLFOX 77.5% (n = 31). Overall we found a 25% pCR without significant differences between type of neoadjuvant chemotherapy. Conclusions: TNT followed by resection is feasible with acceptable rates of treatment completion and toxicity. Notable limitations include the retrospective analysis, small sample size, and heterogenous treatment. The pCR rate is promising and warrants further prospective study to optimize TNT approaches. [Table: see text]

186 Background: PROs assessing quality of life (QOL) and physical symptoms often correlate with clinical outcomes in patients (pts) with cancer. Yet, data are lacking about the use of PROs to predict treatment response. We evaluated associations of baseline PROs with treatment response, healthcare use, and survival among pts with advanced gastrointestinal cancer. Methods: We prospectively enrolled pts with metastatic gastrointestinal cancer prior to initiating chemotherapy at Massachusetts General Hospital. At baseline (start of treatment), pts reported their QOL (Functional Assessment of Cancer Therapy General [FACT-G], subscales assess QOL across 4 domains: functional, physical, emotional, social well-being) and symptom burden (Edmonton Symptom Assessment System [ESAS]). Higher scores on FACT-G indicate better QOL, while higher scores on ESAS represent a greater symptom burden. We used regression models to examine associations of baseline PRO scores with treatment response (clinical benefit [CB] or progressive disease [PD] at the time of first scan based on clinical documentation), healthcare use (unplanned hospital admissions), and survival. Results: From 5/2019-3/2020, we enrolled 112 of 131 (85.5% enrollment) consecutive pts (median age = 62.8, 61.6% male, 45.5% pancreatobiliary cancer). For treatment response, 64.3% had CB and 35.7% had PD. Higher ESAS-physical (B = 1.04, p = .027) and lower FACT-G functional (B = 0.92, p = .038) scores at baseline were significant predictors of PD. On the specific ESAS items, pts who experienced PD were more likely to report moderate/severe poor well-being (57.9% vs 29.7%; p = .001), pain (44.7% vs 25.0%; p < .050), drowsiness (42.1% vs 20.3%; p = .024), and diarrhea (23.7% vs 4.7%; p = .008) at baseline. Lower FACT-G total (HR = 0.96, p = .003), FACT-G physical (HR = 0.89, p < .001), FACT-G functional (HR = 0.87, p < .001), and higher ESAS-physical (HR = 1.03, p = .028) scores at baseline were significantly associated with greater risk of hospital admission. Lower FACT-G total (HR = 0.96, p = .009), FACT-G emotional (HR = 0.87, p = .014), as well as higher ESAS-total (HR = 1.03, p = .018) and ESAS-physical (HR = 1.03, p = .040) scores at baseline were significantly associated with greater risk of death. Conclusions: We found that baseline PROs predict treatment response in pts with advanced cancer, namely physical symptoms and functional QOL, in addition to healthcare use and survival outcomes. These findings further support the use of PROs to predict important clinical outcomes, including the novel finding of treatment response.