Protein biomarkers and alternatively methylated cell-free DNA detect early stage pancreatic cancer

Abstract

Objective

Pancreatic ductal adenocarcinoma (PDAC) is commonly diagnosed at an advanced stage. Liquid biopsy approaches may facilitate detection of early stage PDAC when curative treatments can be employed.

Design

To assess circulating marker discrimination in training, testing and validation patient cohorts (total n=426 patients), plasma markers were measured among PDAC cases and patients with chronic pancreatitis, colorectal cancer (CRC), and healthy controls. Using CA19-9 as an anchor marker, measurements were made of two protein markers (TIMP1, LRG1) and cell-free DNA (cfDNA) pancreas-specific methylation at 9 loci encompassing 61 CpG sites.

Results

Comparative methylome analysis identified nine loci that were differentially methylated in exocrine pancreas DNA. In the training set (n=124 patients), cfDNA methylation markers distinguished PDAC from healthy and CRC controls. In the testing set of 86 early stage PDAC and 86 matched healthy controls, CA19-9 had an area under the receiver operating characteristic curve (AUC) of 0.88 (95% CI 0.83 to 0.94), which was increased by adding TIMP1 (AUC 0.92; 95% CI 0.88 to 0.96; p=0.06), LRG1 (AUC 0.92; 95% CI 0.88 to 0.96; p=0.02) or exocrine pancreas-specific cfDNA methylation markers at nine loci (AUC 0.92; 95% CI 0.88 to 0.96; p=0.02). In the validation set of 40 early stage PDAC and 40 matched healthy controls, a combined panel including CA19-9, TIMP1 and a 9-loci cfDNA methylation panel had greater discrimination (AUC 0.86, 95% CI 0.77 to 0.95) than CA19-9 alone (AUC 0.82; 95% CI 0.72 to 0.92).

Conclusion

A combined panel of circulating markers including proteins and methylated cfDNA increased discrimination compared with CA19-9 alone for early stage PDAC.

CPP and YD are joint senior authors.

WHAT IS ALREADY KNOWN ON THIS SUBJECT

• There is an unmet need for biomarkers that allow non-invasive detection of early-stage pancreatic cancer. Circulating proteins, oncogenic mutations in cfDNA and altered methylation in cfDNA are important candidate markers for PDAC early detection.

WHAT THIS STUDY ADDS

• A combination of plasma proteins and pancreas-specific methylation markers in cfDNA improved detection of pancreatic cancer compared to CA-19-9 and to each marker alone.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• A combination of protein and tissue-specific methylation cfDNA markers may allow for detection of pancreatic cancer at an earlier and curable stage.

Introduction

Pancreatic cancer is a leading cause of cancer death worldwide.¹ This high mortality results in large part from >80% of patients presenting with locally advanced or metastatic disease that is incurable. In contrast, patients who present with earlier stage disease can be treated with multimodality therapy and achieve long-term survival.^{2 3}

Early detection of pancreatic ductal adenocarcinoma (PDAC) remains difficult. The disease causes few early warning symptoms and has few risk factors with high penetrance.⁴ Thus far, patients with a strong family history or genetic predisposition and those with pancreatic cystic lesions have been the primary focus of early detection programmes.^{5 6} These programmes predominantly use abdominal MRI and endoscopic ultrasound to serially evaluate the pancreas for the development of cancer. Blood-based early detection approaches may allow for identification of those patients who would most benefit from imaging or endoscopic procedures.⁷ Circulating carbohydrate antigen 19–9 (CA19-9) is commonly used as a marker of treatment response in PDAC, but it may also have utility as an ‘anchor’ marker on which to add further blood-based technologies for early cancer detection.⁸ Thus, we sought to develop a panel of circulating markers that included CA19-9 and could be used in a screening setting to assess asymptomatic individuals for pancreatic cancer.

Recent studies have evaluated mutations in cell-free DNA (cfDNA) in plasma as a biomarker for the presence of early cancer.⁹ This approach has appeal for detection of pancreatic cancer, as ~90% of patients with this malignancy have point mutations in the oncogene KRAS.^{10 11} Nevertheless, detection of driver gene mutations in cfDNA has thus far had modest sensitivity for early stage malignancies. Furthermore, identification of a mutation in a gene such as KRAS does not identify the tissue origin of malignancy, as mutated genes are shared across tumour types. Oncogenic mutations in cfDNA may also reflect non-malignant conditions, such as clonal haematopoiesis.¹² A potential approach to increase sensitivity for detection of early cancers and assist in determining the malignant site of origin leverages tissue-specific DNA methylation patterns.^13–15 The methylation of cytosines adjacent to guanines (CpG sites) is an essential determinant of gene expression and can serve as a definitive marker of cell identity.¹⁶ Therefore, cfDNA molecules derived from genomic loci with tissue-specific methylation patterns can be used to identify the relative contribution of specific cell types to cfDNA and estimate the rate of cell death in specific tissues.¹³ Since tissue-specific methylation markers are largely maintained on oncogenic transformation, evaluation of cfDNA methylation can provide a powerful tool to detect and determine the tissue of origin for a growing cancer.¹⁷ Recently, several studies have examined altered methylation of circulating cfDNA for detection of single cancer types or for multicancer detection.^17–19

Given the potential utility of adding cfDNA assays to multimarker panels for asymptomatic PDAC detection, we evaluated three protein markers (CA19-9, TIMP1, LRG1),²⁰ KRAS mutation from cfDNA, and exocrine pancreas-specific methylation markers from cfDNA in PDAC cases and controls. In over 400 patients, we demonstrate the utility of combining protein markers with measures of cfDNA tissue-specific methylation in detecting early stage PDAC.

Methods

Study populations

Training, testing and validation case-control sets were identified or enrolled with collection of clinical data and biospecimens (figure 1). For the training set, we identified 125 patients with PDAC and colorectal cancer (CRC), and healthy controls treated at Dana-Farber/Brigham and Women’s Cancer Center (DF/BWCC) between 2010 and 2017. Patients with cancer had no prior cancer-directed treatment, except one patient with localised PDAC who was excluded due to blood collection after surgical resection. The final cohort included 24 with localised PDAC, 25 with metastatic PDAC, 25 with localised CRC, 25 with metastatic CRC and 25 healthy controls. Institutional, Clinical Laboratory Improvement Amendments (CLIA)-certified DNA sequencing of matched tumour DNA was available for all patients, except one with metastatic CRC whose sequencing could not be completed due to low tumour cellularity.

Figure 1. Training, testing and validation sets for characterisation of protein and cell-free DNA markers to detect early stage pancreatic cancer. AJCC, American Joint Committee on Cancer; cfDNA, cell-free DNA; CRC, colorectal cancer; PDAC, pancreatic ductal adenocarcinoma In the bar plot, lighter color shade indicates receipt of neoadjuvant therapy prior to pathologic staging among patients who went to the operating room for surgical resection. All blood samples collected at the time of cancer diagnosis prior to any treatment or surgery. Circulating markers measured: (a) Training set: cfDNA mutation and methylation; (b) Testing set: cfDNA mutation and methylation, CA19-9, TIMP1, and LRG1; (c) Validation set: cfDNA methylation, CA19-9, TIMP1, and LRG1.

For the testing set, we prospectively accrued 86 patients with previously untreated localised PDAC who underwent subsequent surgical resection, 86 healthy controls matched by age and sex, and 50 patients with chronic pancreatitis, enrolled between 2015 and 2019 at three institutions (DF/BWCC, Beth Israel Deaconess Medical Centre and Columbia University-Irving Cancer Centre). Healthy controls had no history of cancer 5 years prior to sample collection and were recruited at the time of a screening colonoscopy or when accompanying a non-blood-related relative to the GI cancer clinic. Patients with chronic pancreatitis were identified in specialty gastroenterology clinics and aetiology was identified by medical record review. For the validation set, we enrolled another prospectively accrued 40 patients with previously untreated, localised PDAC and 40 age-matched and sex-matched healthy controls from the University of Pennsylvania between 2016 and 2021. Healthy controls were recruited at the time of screening colonoscopy. All patients provided informed consent for access to medical records and blood samples. Medical record review identified clinical data and tumour characteristics. Blood samples collected prior to therapy were processed to EDTA plasma within 2–3 hours and aliquots were stored at −80°C. A portion of samples from the validation set was processed in Streck tubes rather than EDTA plasma. Participant identity was blinded to laboratory personnel.

Circulating cfDNA mutation and methylation assays

Droplet digital PCR (ddPCR) and next generation sequencing (NGS) for KRAS and other genes were performed as previously described (online supplemental methods, table 1).²¹ To identify exocrine pancreas-specific cfDNA markers (online supplemental table 2), we performed comparative analysis of human tissue and cell type methylomes from public sources and methylomes generated from freshly isolated, sorted cells from surgical material, using whole genome bisulfite sequencing.²² CpG sites found to be uniquely methylated or unmethylated were selected as potential markers distinguishing cfDNA from the exocrine pancreas. For each candidate CpG, we verified that it retained its methylation pattern in the TCGA collection of methylomes from PDAC and other tumours. To maximise tissue specificity of methylation patterns we leveraged the regional nature of DNA methylation and defined a marker as a genomic locus of <150 bp (typical nucleasome size of cfDNA fragments) that contains at least four CpGs in addition to the identified anchor site. A molecule was assigned pancreas origin when all CpGs within it had a homogenous methylation pattern consistent with that seen in exocrine pancreas. cfDNA was treated with bisulfite, PCR-amplified in multiplex and sequenced as described (online supplemental methods).²³ To correct for the presence of cfDNA derived from other tissues, we multiplied the fraction of pancreas-specific molecules by the total concentration of cfDNA in each sample to provide concentration of exocrine pancreas-specific cfDNA in a sample, expressed as pancreas genome equivalents per mL plasma. Pancreas cfDNA signal was calculated as the average signal obtained from the multiple markers.

Protein biomarkers

Plasma protein concentrations for CA19-9, LRG1 and TIMP1 were measured by bead-based ELISA using Luminex multiplexed assay technology, as previously described.²⁰ This approach was used to minimise required sample volume for research purposes and was not for diagnostic application.

Statistical analysis

Clinical characteristics were described using mean (SD) for continuous variables and number (percent) for categorical variables. Receiver operating characteristic (ROC) curves were generated using logistic regression with calculation of area under the ROC curve (AUC) to provide a measure of model discrimination.²⁴ Confidence limits for AUC were calculated by the Wald method. Comparisons of differences between AUCs were tested using Delong’s non-parametrical approach.²⁵ Sensitivity was reported at ≥90%, ≥95% and≥98% specificity, given the desire to limit false-positive results when detecting a malignancy of low incidence. Assay cut points for positivity were determined from ≥98% specificity in the testing set.

Patient and public involvement

Patients were not directly involved in conduct of the study; however, the public has strongly advocated for advancements in PDAC early detection.

Results

Methylation markers of the exocrine pancreas

Comparative methylome analysis resulted in the identification of two loci that were specifically methylated in pancreatic acinar cells (termed acinar-1 and acinar-2). To validate specificity and examine sensitivity, we amplified these loci from bisulfite-treated genomic DNA derived from a panel of human tissues and sequenced the products. Each marker was fully methylated in 74%–83% of DNA from acinar cells, whereas no fully methylated molecules were identified in leucocytes or other tissues (figure 2A). To assess marker sensitivity, we spiked acinar genomic DNA into leucocyte DNA in known proportions and measured the signal obtained after PCR and sequencing. We detected the presence of as little as 0.1% pancreas DNA in the mixture (figure 2B). Moreover, we verified that these markers retained their altered methylation patterns in PDAC (online supplemental figure 1). Due to modest sensitivity for localised PDAC with two markers in the training set (see below), we expanded our marker set to include seven additional loci differentially methylated in acinar cells (acinar-3 to acinar-7) or ductal epithelial cells (duct-1, duct-2) for the testing and validation sets. All seven new markers showed high organ specificity, spike-in experiments demonstrated identification of exocrine pancreas DNA when comprising as little as 0.1% of DNA in a mixture (figure 2), and these markers retained their methylation patterns in PDAC (online supplemental figure 1). We further verified specificity compared with other cancer types (online supplemental figure 2).

Figure 2. Tissue specificity and spike-in sensitivity of exocrine pancreas methylation markers for pancreas acinar and ductal cells. Methylation status of acinar and duct-derived markers in genomic DNA from multiple human tissues (A). Each color represents a locus that is differentially methylated or unmethylated in a specific cell type. Shown is the methylation score of multiple CpG sites in each block (i.e. the fraction of molecules that are fully methylated or unmethylated in a given sample). Sensitivity of acinar-derived (B) or duct-derived (C) methylation markers. Pancreas-specific DNA was spiked into leukocyte DNA as indicated and the fraction of pancreas DNA was assessed using bisulfite conversion, multiplex PCR amplification of acinar markers and sequencing.

Training set

We investigated cfDNA mutation and methylation in 124 patients with PDAC and CRC, and healthy controls (online supplemental tables 3,4). The cfDNA sequencing for KRAS mutations identified 21% of localised PDAC, 72% of metastatic PDAC, 12% of localised CRC, 36% of metastatic CRC and 4% of healthy controls (online supplemental figure 3A). In an exploratory analysis of amplicon-based NGS, including for KRAS, TP53, GNAS, SMAD4, RNF43, CDKN2A and BRAF, sensitivity was not improved, and more false-positive results were identified (online supplemental table 4). Therefore, this approach was not pursued in the testing set.

We next examined exocrine pancreas-specific cfDNA by bisulfite sequencing of the acinar-1 and acinar-2 loci in the training set (online supplemental figure 3B). Comparing PDAC cases (n=49) to CRC cases and healthy controls (n=75), the AUC by ROC curve analysis was 0.77, with sensitivity for PDAC detection of 57% at 95% specificity, and 55% at 98% specificity. When considered by stage at diagnosis with 98% specificity, sensitivity was 29% for localised PDAC and 80% for metastatic PDAC.

Testing set

Given the ability to detect PDAC with circulating cfDNA approaches in the training set, we next examined the testing set that included 86 patients with localised PDAC, 86 healthy controls and 50 patients with chronic pancreatitis (online supplemental tables 5,6). Although chronic pancreatitis is rare in the general population, this patient group was included to assess the specificity of markers in the context of an inflammatory condition of the pancreas. Given the potential role of CA19-9 as an anchor marker for PDAC detection,⁸ we first measured CA19-9 in the testing set cases and controls (table 1). CA19-9 had an AUC of 0.89 (95% CI 0.84 to 0.94) comparing patients with localised PDAC to healthy controls and AUC of 0.85 (95% CI 0.79 to 0.91) comparing patients with localised PDAC to those with chronic pancreatitis.

Table 1

Discrimination of patients with early stage pancreatic cancer, healthy controls and patients with chronic pancreatitis by protein and cell-free DNA markers in the testing set

Biomarker	No. cases	No. controls	AUC (95% CI)	Sensitivity at designated specificities
				≥90%	≥95%	≥98%
Local PDAC versus healthy controls
CA19-9	86	86	0.89 (0.84 to 0.94)	67%	67%	64%
TIMP1	86	86	0.83 (0.77 to 0.89)	59%	38%	19%
LRG1	86	86	0.79 (0.73 to 0.86)	50%	36%	31%
cfDNA KRAS mutation	86	86	0.54 (0.50 to 0.57)	9%	9%	9%
2-loci cfDNA methylation panel*	84	82	0.55 (0.49 to 0.61)	18%	13%	11%
9-loci cfDNA methylation panel†	84	82	0.69 (0.61 to 0.77)	40%	33%	21%
Local PDAC versus chronic pancreatitis
CA19-9	86	50	0.85 (0.79 to 0.91)	67%	64%	52%
TIMP1	86	50	0.68 (0.59 to 0.78)	38%	17%	13%
LRG1	86	50	0.68 (0.59 to 0.78)	27%	12%	12%
cfDNA KRAS mutation	86	50	0.53 (0.48 to 0.57)	9%	9%	2%
2-loci cfDNA methylation panel*	84	48	0.58 (0.53 to 0.63)	21%	19%	17%
9-loci cfDNA methylation panel†	84	48	0.69 (0.60 to 0.78)	40%	24%	7%

*2-loci cfDNA methylation panel that includes two exocrine pancreas loci encompassing 17 CpG sites.

†9-loci cfDNA methylation panel that includes nine exocrine pancreas loci encompassing 61 CpG sites.

AUC, area under the receiver operating characteristic curve; cfDNA, cell-free DNA; PDAC, pancreatic ductal adenocarcinoma.

Given the potential for protein markers to increase sensitivity when combined with cfDNA detection,^{26 27} we measured plasma TIMP1 and LRG1, which we previously identified as protein biomarkers for early stage PDAC (table 1). TIMP1 had an AUC of 0.83 (95% CI 0.77 to 0.89) comparing patients with early stage PDAC to healthy controls and AUC of 0.68 (95% CI 0.59 to 0.78) when compared with patients with chronic pancreatitis. Plasma LRG1 had an AUC of 0.79 (95% CI 0.73 to 0.86) when patients with early stage PDAC were compared with healthy controls and AUC of 0.68 (95% CI 0.59 to 0.78) when compared with patients with chronic pancreatitis.

We next evaluated cfDNA mutation and methylation detection in the testing set (table 1). Among cfDNA detection approaches, the highest AUC for the comparisons of localised PDAC to both healthy controls and chronic pancreatitis was identified for the cfDNA methylation approach using nine methylation haplotype blocks, with AUC of 0.69 in comparison to both control groups. Since greater sensitivity may be achieved with the combination of several markers, we next examined whether the protein and cfDNA markers provided additional discrimination beyond CA19-9 alone for cases and controls. The AUC for discrimination of early stage PDAC from healthy controls increased with addition of TIMP1, LRG1 or the 9-loci cfDNA methylation panel to CA19-9 (table 2), but not with the addition of cfDNA KRAS mutation or the 2-loci cfDNA methylation panel. Little benefit in AUC was identified for these markers when added to CA19-9 in discriminating early stage PDAC from patients with chronic pancreatitis (table 2).

Table 2

Discrimination of patients with early stage pancreatic cancer, healthy controls and patients with chronic pancreatitis by protein and cfDNA markers when added to CA19-9 in the testing set

Biomarker	No. cases	No. controls	AUC (95% CI)	P value	Sensitivity at designated specificities
					≥90%	≥95%	≥98%
Local PDAC versus healthy controls
CA19-9	84	82	0.88 (0.83 to 0.94)	Reference	67%	67%	63%
+ TIMP1	84	82	0.92 (0.88 to 0.96)	0.06	80%	75%	63%
+ LRG1	84	82	0.92 (0.88 to 0.96)	0.02	77%	77%	68%
+ cfDNA KRAS mutation	84	82	0.88 (0.83 to 0.94)	0.39	67%	67%	63%
+ 2-loci cfDNA methylation panel*	84	82	0.89 (0.83 to 0.94)	0.56	67%	67%	63%
+ 9-loci cfDNA methylation panel†	84	82	0.92 (0.88 to 0.96)	0.02	76%	70%	65%
Local PDAC versus chronic pancreatitis
CA19-9	84	48	0.85 (0.78 to 0.91)	Reference	67%	63%	52%
+ TIMP1	84	48	0.84 (0.78 to 0.91)	0.88	67%	55%	48%
+ LRG1	84	48	0.85 (0.79 to 0.91)	0.51	68%	60%	57%
+ cfDNA KRAS mutation	84	48	0.85 (0.78 to 0.91)	1.00	67%	63%	52%
+ 2-loci cfDNA methylation panel*	84	48	0.86 (0.79 to 0.92)	0.19	68%	62%	55%
+ 9-loci cfDNA methylation panel†	84	48	0.86 (0.80 to 0.92)	0.35	67%	56%	54%

*2-loci cfDNA methylation panel that includes two exocrine pancreas loci encompassing 17 CpG sites.

†9-loci cfDNA methylation panel that includes nine exocrine pancreas loci encompassing 61 CpG sites.

AUC, area under the receiver operating characteristic curve; cfDNA, cell-free DNA; PDAC, pancreatic ductal adenocarcinoma.

We also considered the performance of the protein and cfDNA markers among cases identified as CA19-9-negative (figure 3). Among the 31 patients without CA19-9 elevation, TIMP1 was elevated in 3 cases, LRG1 in 8 cases and the 9-loci cfDNA methylation panel in 7 cases, when considering cut points for positivity that conveyed ≥98% specificity for the individual marker. When considering these markers together, 14 (45%) of the 31 CA19-9-negative cases were positive for one or more of these markers, indicating the presence of an early stage PDAC among cases without elevated CA19-9. We then constructed several multimarker panels with CA19-9 as the anchor marker and including combinations of the two protein markers and the 9-loci cfDNA methylation panel. All panels performed similarly in the testing set with AUCs of 0.94, which were statistically significant improvements over the AUC of 0.88 with CA19-9 alone (p<0.05 for all models; online supplemental table 7).

Figure 3. Cumulative positivity for early stage pancreatic cancer and healthy controls by protein and cell-free DNA markers in the testing set. cfDNA, cell-free DNA; PDAC, pancreatic ductal adenocarcinoma Each column represents one subject and each row represents their value for the designated biomarker, with red bar indicating positive and blue bar indicating negative value by heatmap scale. Cases and controls are each sorted from highest to lowest CA19-9 values for CA19-9 positive cases and then in order by positive values for TIMP1, LRG1, and 9-loci cfDNA methylation panel. Orange track color indicates detected cases by CA19-9. Purple track color indicates detected cases by other markers among CA19-9 negative cases.

Validation set

In an independent validation set including 40 patients with PDAC and 40 healthy controls (online supplemental tables 8,9), we looked to validate the four prediction models combining proteins and the 9-loci cfDNA methylation panel. We first evaluated the discrimination of individual biomarkers in the external validation set (online supplemental table 10). Compared with AUC values in the testing set, LRG1 performed substantially less well in the validation set, while the other three markers had similar AUC, including CA19-9 (AUC, 0.82; 95% CI 0.72 to 0.92), TIMP1 (AUC, 0.76; 95% CI 0.65 to 0.87) and the 9-loci cfDNA methylation panel (AUC, 0.69; 95% CI 0.58 to 0.81). We subsequently examined discrimination of the multimarker panels with fixed model coefficients calculated from the testing set (figure 4, online supplemental table 7). The fixed model containing CA19-9, TIMP1 and the 9-loci cfDNA methylation panel (AUC, 0.86; 95% CI 0.77 to 0.95) had a higher AUC than CA19-9 alone (AUC, 0.82; 95% CI 0.72 to 0.92).

Figure 4. Receiver operating characteristic curves for plasma CA19-9 alone and in combination with TIMP1 and the 9-loci cfDNA methylation panel for distinguishing early stage pancreatic cancer from healthy controls in the testing (A) and validation (B) sets. AUC, area under the receiver-operator characteristic (ROC) curve; cfDNA, cell-free DNA.

Discussion

Diagnosis of early stage disease greatly improves the chance of long-term survival of patients with PDAC. Liquid biopsies for molecular characterisation and disease follow-up during treatment have entered clinical care,²⁸ and large studies are now being pursued to apply liquid biopsies to early cancer detection.^{15 29} However, complementary technologies will likely be needed to identify early cancers with high sensitivity, and optimal marker combinations may differ by cancer type. Here, we demonstrate improved sensitivity when protein and cfDNA markers are added to CA19-9, a potential anchor marker on which to build multimarker detection approaches for PDAC.⁸

We evaluated the performance of circulating markers in the training and testing cohorts and then assessed additive discriminatory capacity for early stage disease in combination with plasma CA19-9 in an independent validation cohort. Although accepted as a prognostic marker for pancreatic cancer, CA19-9 is not routinely used in the screening setting. However, among patients in the testing and validation sets of the current study, the AUC was 0.82–0.89 comparing PDAC cases to healthy controls, and sensitivity was ≥60% at a specificity of ≥98%. These discriminatory statistics suggest that CA19-9 may function satisfactorily as an anchor marker on which to add further early detection technologies. Additionally, genotyping inherited genetic variants may further increase CA19-9 performance, given the ~10% of individuals who do not synthesise CA19-9 due to biallelic inactivating polymorphisms in FUT3.³⁰ Nevertheless, it is important to note that differences between CA19-9 assays can complicate threshold selection and cross-study comparisons,³¹ and studies of prediagnosis plasma suggest that elevations are likely to occur predominantly in the 6–12 months prior to cancer diagnosis,^{8 32} necessitating relatively frequent measurements to capture patients early in the disease process. In addition, the performance of CA19-9 as a potential anchor marker may vary among screening populations, including those at high risk and in the general population.

Although CA19-9 serves as a candidate marker for PDAC early detection, a sizeable number of patients will not be identified by using CA19-9 alone. In the current study, patients with CA19-9 below the positive threshold were identified by assessment of additional protein markers (TIMP1 and LRG1) or exocrine pancreas-specific methylated cfDNA, suggesting the complementarity of different markers for PDAC detection. More than 90% of PDACs have a point mutation in the KRAS oncogene,^{10 11} suggesting this mutation as a cfDNA marker for early detection. Nevertheless, studies thus far have indicated modest sensitivity for KRAS mutation detection in early stage localised PDAC,^{26 27 33 34} and need for large plasma volumes to detect rare tumour DNA fragments. In the current study, multiplexed ddPCR for KRAS codons 12 and 61 identified only 10% of patients with early stage PDAC at 98% specificity in the testing cohort. In contrast, the large majority of patients with metastatic disease were identified in the training cohort. Alternative high-sensitivity detection approaches or the use of larger amounts of plasma may improve on these results, but cfDNA KRAS mutation detection did not provide additive information with plasma CA19-9 in the current study.

Recurrent mutations in PDAC are identified primarily in KRAS and TP53, limiting the areas of the cfDNA genome that are informative for early disease detection. In contrast, many pancreas-specific methylation markers are conserved in the genome, potentially enhancing the detection of rare DNA fragments originating from cancer. Furthermore, given the organ specificity of methylation markers, the tissue of origin may be inferred within the same assay platform, potentially helping to guide clinical evaluation.^{14 15 35} Notably, false-positive test results due to clonal haematopoiesis are also reduced with methylation-based approaches that measure organ-specific cfDNA fragments compared with detection of mutations.^{36 37} In the current study, pancreas-specific methylation markers added discriminatory capacity beyond CA19-9 for early stage disease, but only with our larger panel of methylation blocks. Further increases in the number of pancreas-specific markers may facilitate even greater assay sensitivity.^{38 39} However, the use of a relatively small number of methylation markers harbouring very high pancreas specificity allows for measurement of essentially all DNA molecules containing each marker sequence in a sample (ie, coverage of >1000 ×), potentially providing higher sensitivity for detection of pancreas cfDNA at a lower cost.

Previous studies have identified methylation or hydroxymethylation changes in PDAC and then evaluated their occurrence in cfDNA,^40–42 but these methylation changes were not exclusive to pancreatic cancer compared with other tumour types or well conserved across pancreatic cancers. In the current study, we identified methylation signatures of the normal exocrine pancreas that were preserved in a large cohort of pancreatic cancers. This approach was designed to facilitate high specificity, while also enhancing sensitivity due to evaluation of methylation signatures common across pancreatic tumours and measurement of the joint effect of cell death of tumour cells and adjacent normal cells. Although this approach could theoretically detect non-malignant pancreatic pathologies, we did not identify higher levels of pancreas-derived cfDNA in patients with chronic pancreatitis, possibly due to the natural slow-progressive course of the disease. This finding is seemingly in contrast to our previously published work,⁴³ in which patients with chronic pancreatitis had elevated levels of pancreas-derived cfDNA. In the previous work, plasma samples were obtained from patients with severe chronic pancreatitis who were hospitalised and required surgical intervention. In our current study, samples were obtained in the outpatient clinic in the absence of an acute flare. We believe this is the source of the conflicting findings. Additional studies in patients with acute pancreatitis or benign biliary disease would be informative, although these conditions are readily diagnosed by clinical and laboratory evaluation, so unlikely to reduce test specificity in a screening population.

The addition of TIMP1 and circulating methylated cfDNA to CA19-9 increased PDAC discrimination; however, a combined marker approach can increase the number of false-positive results. Thus, the clinical utility of combining additional markers with CA19-9 will be dependent on the population for evaluation and the false-positive rate deemed tolerable. Notably, multianalyte assays may need to be tuned differently to interrogate the general population versus the high-risk groups, such as those with family history of PDAC, pancreatic cystic lesions or recent-onset diabetes,^{44 45} and future decisions regarding threshold values for a positive test will need to be made with the intended use population in mind.

The current study has a number of important strengths. The subject populations were drawn from multiple institutions using unified sample collection and processing protocols. Multiple patient groups were evaluated in training, testing and validation sets, including subjects with PDAC, CRC, chronic pancreatitis and healthy controls. The testing and validation sets were prospectively collected and included only patients with an initial diagnosis of early stage disease, constituting an important target population for early disease identification. Plasma CA19-9 was used as an anchor marker in the testing and validation sets, with the utility of further markers assessed with respect to their additive discriminatory capacity. Laboratory personnel were blinded to the case-control status of study participants, and analyses were conducted centrally using a prespecified analysis plan.

The study also has limitations that deserve consideration. Sensitivities for our cfDNA approaches in the testing and validation sets were somewhat lower than anticipated.^{15 26 27} Since circulating tumour DNA is thought to be proportional to tumour burden,^{33 46} the fact that our population was heavily skewed towards patients with early stage disease likely resulted in fewer tumour DNA fragments in the blood of these patients. Our ability to detect tumour cfDNA fragments may also have been affected by the 2 mL volume of plasma used for our cfDNA studies. Larger volumes of plasma may be helpful to identify very early stage tumours. The increased sensitivity with our greater multiplexing of methylation sites also suggests that evaluating larger areas of the genome may assist in detecting these rare tumour-derived cfDNA fragments. We evaluated our cfDNA methylation panels in patients with PDAC, CRC, chronic pancreatitis and healthy controls. Additional studies to evaluate the specificity of our protein and cfDNA methylation markers in blood for PDAC compared with other cancer types will be necessary. All blood samples were collected at the time of cancer diagnosis. As we described recently with protein markers,⁸ it is critical to define the timeframe during which circulating markers are detectable before a cancer would be diagnosed clinically, such that screening intervals can be rationally designed. The current study did not evaluate all possible early detection technologies and other promising approaches, such as genome-wide cfDNA fragmentation and circulating exosomes that are currently under evaluation.^{33 47–51}

In summary, the combined detection of protein markers and pancreas-specific methylation in circulating cfDNA may improve discrimination for detection of early stage, localised PDAC compared with plasma CA19-9 alone. Additional studies are needed to determine whether this and other abovementioned approaches can lead to diagnosis of asymptomatic early stage PDAC in the general population or high-risk individuals and reduce mortality from this highly lethal malignancy.

The authors thank members of the Dor, Wolpin-Nowak, Aguirre, Hanash and Carpenter laboratories, along with members of the Translational Research Laboratory of the Belfer Center for Applied Cancer Science for their constructive feedback. The authors also thank the patients who volunteered to provide clinical data, blood samples and tumour specimens for the work described.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study involves human participants, who provided informed consent to participate in the study before taking part. The study was approved by the institutional review boards at Dana-Farber/Harvard Cancer Center and Penn Medicine.

Footnote

Twitter @labrat3

Contributors RS, CPP, YD and BMW conceived and initiated the project and designed the experiments. LB, JN, CY, ML, ABa, TD, AZ, AH, NB, KJ, LL, TEC, DAR, KN, JAC, FK, MK, AJA, PAJ, NB, BZS, MHO, JT, AM, EC, ABu, JMG and SH provided materials and consultation. RB, QW, JZ, JGS, JFF, RL, EI, DN, SP, OG and AH performed experiments. RB, QW, JZ and RL conducted analysis. RB, QW, YD and BMW wrote the manuscript.

Funding Primary support for this work came from NIH/NCI U01 CA210171 and Stand Up To CancerLustgarten Foundation Pancreatic Cancer Interception Translational Cancer Research Grant (Grant Number: SU2C-AACR-DT25-17). Stand Up To Cancer is a programme of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C. This work was also supported by the NCI Early Detection Research Network (NCI U01 CA200468) and the NCI Molecular and Cellular characterization of Screen Detected Lesions (MCL) Consortium (NCI U01 CA196403) to AM, SH and JFF, and the Khalifa Bin Zayed Foundation to AM. CPP acknowledges further support from R01 CA240592. ABa acknowledges support from NIH (grant K07 CA222159) and the Bob Parsons Fellowship. KN acknowledges support from the Broman Family Fund for Pancreatic Cancer Research. YD acknowledges additional research support by grants from the Ernest and Bonnie Beutler Research Program of Excellence in Genomic Medicine, The Alex U Soyka pancreatic cancer fund, The Israel Science Foundation, the Waldholtz/Pakula family, the Robert M and Marilyn Sternberg Family Charitable Foundation, the Helmsley Charitable Trust, Grail and the DON Foundation (to YD). YD holds the Walter and Greta Stiel Chair and Research grant in Heart studies. BMW acknowledges additional research support from the Hale Family Center for Pancreatic Cancer Research, Lustgarten Foundation Dedicated Laboratory programme, NIH grant U01 CA210171, NIH grant P50 CA127003, Stand Up to Cancer, Pancreatic Cancer Action Network, Noble Effort Fund, Wexler Family Fund, Promises for Purple and Bob Parsons Fund.

Competing interests ABu has received consulting fees from Exelixis and Geistlich Pharma. KM declares research funding from Celgene and Trovagene. AJA has consulted for Oncorus, Inc., Arrakis Therapeutics, and Merck & Co., Inc, and has research funding from Mirati Therapeutics, Syros, Deerfield, Inc., and Novo Ventures that is unrelated to this work. AM is listed as an inventor on a patent that has been licensed by Johns Hopkins University to Thrive Earlier Detection and serves as a consultant for Tezcat Biosciences. BG, RS and YD have filed patents on cfDNA analysis technology, and received research funding from GRAIL. BMW declares research funding from Celgene, Eli Lilly Novartis and Revolution Medicines, and consulting for Celgene, GRAIL, Ipsen and Mirati.

Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

References

¹ Siegel RL, Miller KD, Fuchs HE, et al. Cancer statistics, 2021. CA Cancer J Clin 2021; 71: 7–33. doi:10.3322/caac.21654

² Conroy T, Hammel P, Hebbar M, et al. FOLFIRINOX or Gemcitabine as adjuvant therapy for Pancreatic cancer. N Engl J Med 2018; 379: 2395–406. doi:10.1056/NEJMoa1809775

³ Blackford AL, Canto MI, Klein AP, et al. Recent trends in the incidence and survival of stage 1A Pancreatic cancer: A surveillance, epidemiology, and end results analysis. J Natl Cancer Inst 2020; 112: 1162–9. doi:10.1093/jnci/djaa004

⁴ Kamisawa T, Wood LD, Itoi T, et al. Pancreatic cancer. Lancet 2016; 388: 73–85. doi:10.1016/S0140-6736(16)00141-0

⁵ Canto MI, Almario JA, Schulick RD, et al. Risk of neoplastic progression in individuals at high risk for Pancreatic cancer undergoing long-term surveillance. Gastroenterology 2018; 155: 740–751. doi:10.1053/j.gastro.2018.05.035

⁶ Vasen H, Ibrahim I, Ponce CG, et al. Benefit of surveillance for Pancreatic cancer in high-risk individuals: outcome of long-term prospective follow-up studies from three European expert centers. J Clin Oncol 2016; 34: 2010–9. doi:10.1200/JCO.2015.64.0730

⁷ Singhi AD, Koay EJ, Chari ST, et al. Early detection of Pancreatic cancer: opportunities and challenges. Gastroenterology 2019; 156: 2024–40. doi:10.1053/j.gastro.2019.01.259

⁸ Fahrmann JF, Schmidt CM, Mao X, et al. Lead-time trajectory of Ca19-9 as an anchor marker for Pancreatic cancer early detection. Gastroenterology 2021; 160: 1373–83. doi:10.1053/j.gastro.2020.11.052

⁹ Wan JCM, Massie C, Garcia-Corbacho J, et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nat Rev Cancer 2017; 17: 223–38. doi:10.1038/nrc.2017.7

¹⁰ Cancer Genome Atlas Research Network. Integrated Genomic characterization of Pancreatic Ductal adenocarcinoma. Cancer Cell 2017; 32: 185–203. doi:10.1016/j.ccell.2017.07.007

¹¹ Qian ZR, Rubinson DA, Nowak JA, et al. Association of alterations in main driver genes with outcomes of patients with Resected Pancreatic Ductal adenocarcinoma. JAMA Oncol 2018; 4: e173420. doi:10.1001/jamaoncol.2017.3420

¹² Hu Y, Ulrich BC, Supplee J, et al. False-positive plasma Genotyping due to Clonal Hematopoiesis. Clin Cancer Res 2018; 24: 4437–43. doi:10.1158/1078-0432.CCR-18-0143

¹³ Moss J, Magenheim J, Neiman D, et al. Comprehensive human cell-type methylation Atlas reveals origins of circulating cell-free DNA in health and disease. Nat Commun 2018; 9: 5068. doi:10.1038/s41467-018-07466-6

¹⁴ Shen SY, Singhania R, Fehringer G, et al. Sensitive tumour detection and classification using plasma cell-free DNA Methylomes. Nature 2018; 563: 579–83. doi:10.1038/s41586-018-0703-0

¹⁵ Liu MC, Oxnard GR, Klein EA, et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol 2020; 31: 745–59. doi:10.1016/j.annonc.2020.02.011

¹⁶ Dor Y, Cedar H. Principles of DNA methylation and their implications for biology and medicine. Lancet 2018; 392: 777–86. doi:10.1016/S0140-6736(18)31268-6

¹⁷ Schrag D, Beer TM, McDonnell CH, et al. Blood-based tests for Multicancer early detection (PATHFINDER): a prospective cohort study. Lancet 2023; 402: 1251–60. doi:10.1016/S0140-6736(23)01700-2

¹⁸ Nicholson BD, Oke J, Virdee PS, et al. Multi-cancer early detection test in symptomatic patients referred for cancer investigation in England and Wales (SYMPLIFY): a large-scale, observational cohort study. The Lancet Oncology 2023; 24: 733–43. doi:10.1016/S1470-2045(23)00277-2

¹⁹ Majumder S, Taylor WR, Foote PH, et al. High detection rates of Pancreatic cancer across stages by plasma assay of novel Methylated DNA markers and Ca19-9. Clin Cancer Res 2021; 27: 2523–32. doi:10.1158/1078-0432.CCR-20-0235

²⁰ Capello M, Bantis LE, Scelo G, et al. Sequential validation of blood-based protein biomarker candidates for early-stage Pancreatic cancer. J Natl Cancer Inst 2017; 109: djw266. doi:10.1093/jnci/djw266

²¹ Oxnard GR, Paweletz CP, Kuang Y, et al. Noninvasive detection of response and resistance in EGFR-mutant lung cancer using quantitative next-generation Genotyping of cell-free plasma DNA. Clin Cancer Res 2014; 20: 1698–705. doi:10.1158/1078-0432.CCR-13-2482

²² Loyfer N, Magenheim J, Peretz A, et al. A DNA methylation Atlas of normal human cell types. Nature 2023; 613: 355–64. doi:10.1038/s41586-022-05580-6

²³ Neiman D, Gillis D, Piyanzin S, et al. Multiplexing DNA methylation markers to detect circulating cell-free DNA derived from human Pancreatic Β cells. JCI Insight 2020; 5: e136579. doi:10.1172/jci.insight.136579

²⁴ Hajian-Tilaki K. Receiver operating characteristic (ROC) curve analysis for medical diagnostic test evaluation. Caspian J Intern Med 2013; 4: 627–35.

²⁵ DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a Nonparametric approach. Biometrics 1988; 44: 837–45. doi:10.2307/2531595

²⁶ Cohen JD, Li L, Wang Y, et al. Detection and localization of surgically Resectable cancers with a multi-Analyte blood test. Science 2018; 359: 926–30. doi:10.1126/science.aar3247

²⁷ Cohen JD, Javed AA, Thoburn C, et al. Combined circulating tumor DNA and protein biomarker-based liquid biopsy for the earlier detection of Pancreatic cancers. Proc Natl Acad Sci U S A 2017; 114: 10202–7. doi:10.1073/pnas.1704961114

²⁸ Mattox AK, Bettegowda C, Zhou S, et al. Applications of liquid biopsies for cancer. Sci Transl Med 2019; 11: eaay1984. doi:10.1126/scitranslmed.aay1984

²⁹ Lennon AM, Buchanan AH, Kinde I, et al. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science 2020; 369: eabb9601. doi:10.1126/science.abb9601

³⁰ Abe T, Koi C, Kohi S, et al. Gene variants that affect levels of circulating tumor markers increase identification of patients with Pancreatic cancer. Clin Gastroenterol Hepatol 2020; 18: 1161–9. doi:10.1016/j.cgh.2019.10.036

³¹ Haab BB, Huang Y, Balasenthil S, et al. Definitive characterization of CA 19-9 in Resectable Pancreatic cancer using a reference set of serum and plasma specimens. PLoS ONE 2015; 10: e0139049. doi:10.1371/journal.pone.0139049

³² Udgata S, Takenaka N, Bamlet WR, et al. Thbs2/Ca19-9 detecting Pancreatic Ductal adenocarcinoma at diagnosis Underperforms in Prediagnostic detection: implications for biomarker advancement. Cancer Prev Res (Phila) 2021; 14: 223–32. doi:10.1158/1940-6207.CAPR-20-0403

³³ Bettegowda C, Sausen M, Leary RJ, et al. Detection of circulating tumor DNA in Early- and late-stage human malignancies. Sci Transl Med 2014; 6: 224ra24. doi:10.1126/scitranslmed.3007094

³⁴ Le Calvez-Kelm F, Foll M, Wozniak MB, et al. KRAS mutations in blood circulating cell-free DNA: a Pancreatic cancer case-control. Oncotarget 2016; 7: 78827–40. doi:10.18632/oncotarget.12386

³⁵ Shemer R, Magenheim J, Dor Y. Digital Droplet PCR for monitoring tissue-specific cell death using DNA methylation patterns of circulating cell-free DNA. Curr Protoc Mol Biol 2019; 127: e90. doi:10.1002/cpmb.90

³⁶ Boettcher S, Ebert BL. Clonal Hematopoiesis of indeterminate potential. J Clin Oncol 2019; 37: 419–22. doi:10.1200/JCO.2018.79.3588

³⁷ Ptashkin RN, Mandelker DL, Coombs CC, et al. Prevalence of Clonal Hematopoiesis mutations in tumor-only clinical Genomic profiling of solid tumors. JAMA Oncol 2018; 4: 1589–93. doi:10.1001/jamaoncol.2018.2297

³⁸ Chan KCA, Woo JKS, King A, et al. Analysis of plasma Epstein-Barr virus DNA to screen for Nasopharyngeal cancer. N Engl J Med 2017; 377: 513–22. doi:10.1056/NEJMoa1701717

³⁹ Jamshidi A, Liu MC, Klein EA, et al. Evaluation of cell-free DNA approaches for multi-cancer early detection. Cancer Cell 2022; 40: 1537–49. doi:10.1016/j.ccell.2022.10.022

⁴⁰ Eissa MAL, Lerner L, Abdelfatah E, et al. Promoter methylation of Adamts1 and Bnc1 as potential biomarkers for early detection of Pancreatic cancer in blood. Clin Epigenetics 2019; 11: 59. doi:10.1186/s13148-019-0650-0

⁴¹ Henriksen SD, Madsen PH, Larsen AC, et al. Cell-free DNA promoter Hypermethylation in plasma as a diagnostic marker for Pancreatic adenocarcinoma. Clin Epigenetics 2016; 8: 117. doi:10.1186/s13148-016-0286-2

⁴² Guler GD, Ning Y, Ku C-J, et al. Detection of early stage Pancreatic cancer using 5-Hydroxymethylcytosine signatures in circulating cell free DNA. Nat Commun 2020; 11: 5270. doi:10.1038/s41467-020-18965-w

⁴³ Lehmann-Werman R, Neiman D, Zemmour H, et al. Identification of tissue-specific cell death using methylation patterns of circulating DNA. Proc Natl Acad Sci U S A 2016; 113: E1826–34. doi:10.1073/pnas.1519286113

⁴⁴ Abe T, Blackford AL, Tamura K, et al. Deleterious Germline mutations are a risk factor for neoplastic progression among high-risk individuals undergoing Pancreatic surveillance. J Clin Oncol 2019; 37: 1070–80. doi:10.1200/JCO.18.01512

⁴⁵ Yuan C, Babic A, Khalaf N, et al. Diabetes, weight change, and Pancreatic cancer risk. JAMA Oncol 2020; 6: e202948. doi:10.1001/jamaoncol.2020.2948

⁴⁶ Abbosh C, Birkbak NJ, Wilson GA, et al. Phylogenetic ctDNA analysis Depicts early-stage lung cancer evolution. Nature 2018; 554: 446–51.: 264. doi:10.1038/nature25161

⁴⁷ Liu X, Liu L, Ji Y, et al. Enrichment of short mutant cell-free DNA fragments enhanced detection of Pancreatic cancer. EBioMedicine 2019; 41: 345–56. doi:10.1016/j.ebiom.2019.02.010

⁴⁸ Allenson K, Castillo J, San Lucas FA, et al. High prevalence of mutant KRAS in circulating Exosome-derived DNA from early-stage Pancreatic cancer patients. Ann Oncol 2017; 28: 741–7. doi:10.1093/annonc/mdx004

⁴⁹ Yang S, Che SPY, Kurywchak P, et al. Detection of mutant KRAS and Tp53 DNA in circulating Exosomes from healthy individuals and patients with Pancreatic cancer. Cancer Biol Ther 2017; 18: 158–65. doi:10.1080/15384047.2017.1281499

⁵⁰ Cristiano S, Leal A, Phallen J, et al. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature 2019; 570: 385–9. doi:10.1038/s41586-019-1272-6

⁵¹ Yang Z, LaRiviere MJ, Ko J, et al. A Multianalyte panel consisting of extracellular Vesicle miRNAs and mRNAs, cfDNA, and Ca19-9 shows utility for diagnosis and staging of Pancreatic Ductal adenocarcinoma. Clin Cancer Res 2020; 26: 3248–58. doi:10.1158/1078-0432.CCR-19-3313